Liquids, Solutions, and Interfaces: From Classical Macroscopic Descriptions To Modern Microscopic Details (Topics in Analytical Chemistry) (Fawcett 2004)

Liquids, Solutions, and Interfaces
This page intentionally left blank

From Classical Macroscopic Descriptions
to Modern Microscopic Details
W. RONALD FAWCETT
University of California, Davis
1
2004
Oxford New York

Auckland Bangkok Buenos Aires Cape Town Chennai
Dar es Salaam Delhi Hong Kong Istanbul Karachi Kolkata
Kuala Lumpur Madrid Melbourne Mexico City Mumbai Nairobi
Sao Paulo Shanghai Taipei Tokyo Toronto
Copyright # 2004 by Oxford University Press, Inc.

Published by Oxford University Press, Inc.
198 Madison Avenue, New York, New York, 10016
www.oup.com
Oxford is a registered trademark of Oxford University Press
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,
electronic, mechanical, photocopying, recording, or otherwise,
without the prior permission of Oxford University Press.
Library of Congress Cataloging-in-Publication Data
Fawcett, W. Ronald.
Liquids, solutions, and interfaces / W. Ronald Fawcett.
p. cm.
Includes bibliographical references and index.
ISBN 0-19-509432-8
1. Solution (Chemistry) 2. Interfaces (Physical sciences). I. Title.
QD541.F39 2004
541.3 0 4dc21
2003013539
Cover illustration: A Li+cation solvated by four tetrahedrally disposed

acetonitrile molecules. The Li+cation is the principal component of the
non-aqueous electrolyte solution used in rechargeable lithium ion
batteries found in laptop computers and cell phones. Acetonitrile is often
used as one of the solvents in these batteries.
7 6
2 1
Printed in the United States of America

on acid-free paper
To my three muses,
Zuzana, Natalka, and Tetiana
Time is a never ending Fugue

An Interplay of Slow and Fast
Of Silence . . .
And of Light and Dark
That Flickers with Hypnotic Rhythm
Preface
This book developed from a series of lectures given to freshmen graduate students
specializing in analytical and physical chemistry and chemical engineering at the
University of California, Davis. The purpose of these lectures is to introduce the
students to modern topics in solution chemistry. Solutions are involved in every
practical chemistry laboratory, in chemical analysis, in biochemistry, in clinical
chemistry, and in chemical synthesis. When I was a student, solution chemistry
occupied a major fraction of physical chemistry textbooks. At that time it dealt
mainly with classical thermodynamics, phase equilibria, and non-equilibrium phenomena, especially those related to electrochemistry. Much has happened in the
intervening period with the development of important new experimental techniques. At the present time, solutions are examined experimentally at the molecular
level. In X-ray and nuclear diraction experiments, the structure of liquids and
solutions is described in atomic detail to give the time average of the distribution
and orientation of the component molecules and ions. Laser spectroscopy provides a route to the time resolution of molecular events occurring in the femtosecond time range. Non-linear spectroscopic techniques are being used to study
the molecular composition and structure of interfaces. Both theory and experiment have seen tremendous advances since the 1950s. The purpose of this book is
to bring the student through these developments from the classical macroscopic
descriptions to the modern microscopic details.
The subject matter in this monograph falls into three general areas. The rst of
these involves liquids and solutions at equilibrium. These subjects are discussed in
chapters 15, and include the thermodynamics of solutions, the structure of
liquids, electrolyte solutions, polar solvents, and the spectroscopy of solvation.
viii
PREFACE
An attempt is made to familiarize the student with the fundamental background

material together with important aspects of current research in each of these
areas. Chapters 6 and 7 deal with non-equilibrium properties of solutions, and
the kinetics of reactions in solutions. In the latter chapter, emphasis is placed on
fast reactions in solution and femtochemistry. Chapters 810 involve important
aspects of solutions at interfaces. These include liquids and solutions at interfaces,
electrochemical equilibria, and the electrical double layer. This subject matter is of
interest to analytical chemists, physical chemists, biochemists, and chemical and
environmental engineers.
The material in this book should be suitable for senior undergraduates and
graduate students who have completed two semesters or three quarters of chemical thermodynamics and statistical mechanics. A solid background in mathematics and physics is also necessary to understand the subject matter. The
instructor will note that detailed derivations of the thermodynamic equations
have been given. In the case of quantum mechanics and statistical mechanics,
some derivations are not given but instead, the background for the theory is
carefully described. In addition, the student is referred to the original literature.
The material is arranged in such a way that it may be presented at dierent levels.
Thus, if one wishes to present only the fundamentals of describing the structure of
liquids from chapter 2, the material dealing with the statistical mechanical
description of liquid systems can be largely avoided. The same comment applies
to most of the material in this monograph. Introductions to thermodynamics,
statistical thermodynamics, and chemical kinetics are included in chapters 1, 2,
and 7, respectively. This material is given to help students review the subject
matter from earlier courses in physical chemistry and to introduce the symbols
used in this book. General references giving other sources for the material covered
in a given chapter, and to compilations of experimental data are listed at the end
of each chapter.
As far as possible only SI units have been used in writing equations and presenting experimental data. Angstroms and calories, which still appear in the
scientic literature, are avoided. Instead, nanometers and picometers are used
for atomic and molecular dimensions, and joules for units of energy. Pressure is
discussed in terms of pascals and bars rather than torrs and atmospheres.
Equations involving the molecular dipolar properties, namely the dipole moment
and polarizability, assume units of coulomb meters and farad square meters,
respectively, for these quantities. However, tabulated data are given in the more
familiar cgs system with debyes for the dipole moment and cubic nanometers for
the polarizability. This follows the usage in most data tabulations at the present
time. The connection between the SI and cgs units is explained in chapter 2. The
symbols recommended by the International Union of Pure and Applied
Chemistry [1] are used as much as possible.
This preface would not be complete without an expression of gratitude to those
who made this project possible. First, I am grateful to those who introduced me to
the topics discussed in this book during my student years at the University of
Toronto. There are many teachers in this group but special mention goes to
Professors Frank Wetmore and Mike Dignam in the Department of Chemistry.
PREFACE
ix
Second, I would like to thank my students, especially those at UC Davis who

listened to and discussed the lectures on which this book is based. Third, I thank
my typists Elizabeth Bogren and Zuzana Kovacova, who patiently and faithfully
prepared the manuscript including all the complicated equations. Special thanks
are also due to Alex Tikanen and Dmytro Verbovy for their work with the
diagrams. Finally, I would like to thank those who reviewed individual chapters
including Rafael Andreu, Imre Bako, Bob de Levie, Dennis Evans, Manuel
Galan, Douglas Henderson, Andrzej Lasia, Shiraz Markarian, Roger Parsons,
Oleg Petrii, Dino Tinti, Nancy True, and Galina Tsirlina. I am especially indebted
to Oldr ich and Eva Fischer, who have gone through the text with painstaking care
and helped to eliminate numerous small errors and to establish a consistent
system of symbols. The helpful comments of all those involved in the review
process resulted in a signicant improvement in this monograph during the
long gestation period.
Reference
1. Mills, I.; Cvitas , T.; Homann, K.; Kallay, N.; Kuchitsu, K. Quantities, Units, and
Symbols in Physical Chemistry; Blackwell Scientic Publications: Oxford, U.K., 1986.
Contents
Fundamental Constants xvi

1. The Thermodynamics of Liquid Solutions
1.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
1.9
1.10
1.11
1.12
1.13
1.14
Most Liquid Solutions Are Not Ideal, 3

Concentration Units, 4
Thermodynamic Quantities, 6
Partial Molar Quantities, 9
Ideal SolutionsRaoults Law, 15
Thermodynamics of Ideal Solutions, 16
Non-Ideal Solutions, 18
Thermodynamics of Non-Ideal Solutions, 21
Regular Solutions, 24
An Empirical Approach to Non-Ideal Solutions, 30
Ideally Dilute Solutions, 33
Thermodynamics of Ideally Dilute Solutions, 34
Experimental Determination of Solution Activities, 38
Concluding Remarks, 40
References, 41
Problems, 41
2. The Structure of Liquids

2.1
2.2
2.3
2.4
2.5
45
What Is a Liquid?, 45
The Statistical Thermodynamics of Liquids, 47
Intermolecular Forces, 52
Distribution and Correlation Functions, 61
The Experimental Study of Liquid Structure, 65
xi
xii
CONTENTS
2.6 The Direct Correlation Function and the Mean Spherical

Approximation, 70
2.7 Computer Simulations of Simple Liquids, 73
2.8 Estimation of Thermodynamic Properties from the Pair
Correlation Function, 75
2.9 The Properties of a Hard-Sphere Fluid, 79
2.10 The Structure of Water, 84
2.11 Distribution Functions for Liquid Solutions, 88
2.12 Concluding Remarks, 90
References, 91
Problems, 92
3. Electrolyte Solutions
3.1
3.2
3.3
3.4
3.5
3.6
3.7
3.8
3.9
3.10
3.11
3.12
95
Electrolyte Solutions Are Always Non-Ideal, 95

Ionic Size in Solutions, 97
The Thermodynamics of IonSolvent Interactions, 100
IonSolvent Interactions According to the Born Model, 102
IonSolvent Interactions According to the Mean Spherical
Approximation, 106
The Thermodynamics of Electrolyte Solutions, 111
The Experimental Determination of Activity Coecients for
Electrolytes, 116
IonIon Interactions According to the DebyeHuckel Model,
121
IonIon Interactions According to the MSA, 130
The Thermodynamics of Ion Association, 135
Ion Association According to the MSA, 140
References, 144
Problems, 145
4. Polar Solvents
148
4.1 What Constitutes a Polar Liquid?, 148

4.2 Some Important Properties of Polar Solvents, 149
4.3 The Static Solvent Permittivity on the Basis of Continuum
Models, 153
4.4 The Static Solvent Permittivity According to the MSA, 162
4.5 Dielectric Relaxation Phenomena, 169
4.6 The Permittivity of Electrolyte Solutions, 176
4.7 The Dielectric Relaxation Parameters, 180
4.8 Ion Solvation in Polar Solvents, 184
CONTENTS
xiii
4.9 Polar Solvents as Lewis Acids and Bases, 191

References, 200
Problems, 201
5. Spectroscopic Studies of Liquid Structure and Solvation
204
5.1
5.2
5.3
5.4
5.5
5.6
5.7
5.8
5.9
5.10
What Spectroscopic Techniques Are Available?, 205

X-Ray and Neutron Diraction Experiments, 206
Nuclear Magnetic Resonance Spectroscopy in Solutions, 213
NMR Studies of Ion Solvation in Water, 219
NMR Studies of Ion Solvation in Non-Aqueous Solvents, 223
Vibrational Spectroscopy in Solutions, 226
Infrared Spectroscopy of Polar Solvents, 232
Infrared Spectroscopy of Non-Electrolyte Solutions, 239
Infrared Spectroscopy of Electrolyte Solutions, 242
UltravioletVisible Spectroscopy and Solvatochromic Eects,
245
References, 251
Problems, 252
6. Non-Equilibrium Phenomena in Liquids and Solutions
6.1
6.2
6.3
6.4
6.5
6.6
6.7
6.8
6.9
6.10
6.11
6.12
254
Non-Equilibrium Processes Are Usually Complex, 254

The Thermodynamics of Irreversible Processes, 255
The Viscosity of Liquids, 259
Isothermal Diusion in Solutions, 264
Linear Diusion from a Wall, 266
The Electrochemical Potential, 271
The Conductivity of Electrolyte Solutions, 274
Experimental Studies of Conductivity, 283
The DebyeOnsager Model for Conductivity, 288
Transport Phenomena in Non-Aqueous Solutions, 294
Proton Transport Phenomena, 298
References, 301
Problems, 301
7. Chemical Reaction Kinetics in Solution
304
7.1 What Time Scales Are Involved for Chemical Reactions in

Solution?, 304
7.2 Fundamental Concepts, 305
xiv
CONTENTS
7.3
7.4
7.5
7.6
7.7
7.8
7.9
7.10
7.11
7.12
General Types of Solution Reactions, 312

Temperature Eects and Transition State Theory, 323
Diusion-Controlled Rapid Reactions, 329
Relaxation Techniques for Rapid Reactions, 332
Laser Spectroscopy and Femtochemistry in Solutions, 338
The Theory of Homogeneous Electron Transfer, 346
NMR Spectroscopy and Chemical Exchange Reactions, 358
Medium Eects in Solution Reactions, 366
Linear Gibbs Energy Relationships, 375
References, 378
Problems, 380
8. Liquids and Solutions at Interfaces
383
8.1 The Molecular Environment at the Interface Is Dierent than

in the Bulk, 383
8.2 The Interfacial Tension of Liquids, 385
8.3 The Thermodynamics of Fluid Interfaces, 390
8.4 The Electrical Aspects of Interfaces, 395
8.5 The Work Function for Electrons in Metals, 398
8.6 The Liquid|Gas Interface and the Adsorption Isotherm, 401
8.7 Experimental Measurement of the Volta Potential Dierence
at Interfaces, 408
8.8 The Metal|Solution Interface, 422
8.9 The Liquid|Liquid Interface, 426
8.10 Surface Films on Liquids, 433
8.11 Spectroscopy at Liquid Interfaces, 437
References, 443
Problems, 444
9. Charge Transfer Equilibria at Interfaces
447
9.1 Electrochemical Equilibria Occur at a Wide Variety of

Interfaces, 447
9.2 Electrochemical Cells, 448
9.3 The Thermodynamic Basis of the Nernst Equation, 456
9.4 The Absolute Electrode Potential, 461
9.5 Experimental Studies of Electrochemical Cells, 464
9.6 Electrochemical Cells for Electroanalysis, 474
9.7 The Liquid Junction Potential, 477
9.8 Membrane Potentials and the Donnan Eect, 484
CONTENTS
xv
9.9 Ion-Selective Electrodes, 494

9.10 p-Functions and the Denition of pH, 502
References, 505
Problems, 505
10.
The Electrical Double Layer
508
10.1 The Electrical Double Layer Is an Example of Electrostatic

Equilibrium, 508
10.2 The Thermodynamics of the Ideally Polarizable Interface, 510
10.3 The Experimental Study of the Double Layer, 516
10.4 The Structure of the Double Layer, 530
10.5 The Potential of Zero Charge and the Role of the Metal, 535
10.6 The GouyChapman Model of the Diuse Double Layer, 542
10.7 The Structure of the Inner Layer in the Absence of
Adsorption, 552
10.8 The Specic Adsorption of Ions, 558
10.9 The Adsorption of Molecules at Electrodes, 569
References, 577
Problems, 579
Appendix A. Mathematical Background
A.1
A.2
A.3
A.4
582
Laplace Transforms, 582

Fourier Transforms, 584
Complex Numbers and Functions, 585
Power Series, 586
Appendix B. The Laws of Electricity and Magnetism

Appendix C. Numerical Methods of Data Analysis
C.1
C.2
C.3
C.4
C.5
C.6
C.7
The Principle of Least Squares, 595

Linear Regression, 599
Multiple Linear Regression, 605
Numerical Methods, 608
Numerical Interpolation, 610
Numerical Integration, 612
Numerical Dierentiation, 614
Index 617
589
595
Fundamental Constants
Fundamental Constants
Velocity of light
Fundamental charge
Electron rest mass
Proton rest mass
Neutron rest mass
Avogadro constant
Planck constant
Boltzmanns constant
Gas constant
Faraday constant
Permittivity of vacuum
c
e0
me
mp
mn
NL
h
kB
R
F
"0
2:997925 108 m s1

1:602177 1019 C
9:109390 1031 kg
1:672623 1027 kg
1:67493 1027 kg
6:022137 1023 mol1
6:626076 1034 Js
1:380658 1023 J K1
8:314510 J mol1 K1
96485.3 C mol1
8:854188 1012 C2 N1 m2
T
g
273.15 K
9.80665 m s2
atm
cal
A
101,325 Pa
4.184 J
100 pm
107 J
101.325 J
1:6022 1019 J
2:9979 109 esu
3:3356 103 esu
8:9875 1011 esu
3:3356 1030 C m
104 Tesla
Dened Constants
Temperature of H2O freezing
Standard acceleration of free fall
Conversion Factors
Standard atmosphere
Thermochemical calorie
Angstrom
1 erg
1 litre atm
1 eV
1C
1V
1F
1 Debye
1 Gauss
From Cohen and Taylor, Rev. Mod. Phys., 59 (1987) 1121

xvi
The Thermodynamics of Liquid

Solutions
Joel Hildebrand was born in Camden, New

Jersey, in 1881. His interest in science
developed in high school and he went on to
study physics and chemistry at the University
of Pennsylvania, obtaining a B.S. degree in
1903. He stayed at Pennsylvania for a doctoral
degree in chemistry, which he obtained in 1906.
Like several young American scientists of that
period, he went to Germany to study the
emerging eld of physical chemistry.
Hildebrand spent a year at the University of
Joel Henry Hildebrand
Berlin in the laboratory of Walther Nernst and
attended lectures given by Nernst and vant
Ho. He then returned to the University of Pennsylvania, where he held the
position of instructor in chemistry until 1913. Then, at the invitation of G. N.
Lewis, he joined the faculty of the University of California at Berkeley, where
he remained for the rest of his scientic career. Hildebrands main scientic
research was in the area of the physical chemistry of liquids and non-electrolyte
solutions. He was a major contributor to the theory of regular solutions. Much
of his work in this eld is summarized in his monograph with Robert Scott,
The Solubility of Non-electrolytes [1]. He was famous at Berkeley for his
lectures, especially to freshman classes. His scientic work was recognized by
many awards during his long career, including the Priestley Medal of the
American Chemical Society in 1962. Hildebrand was also an avid sportsman
and was particularly fond of skiing in the Sierra Nevada mountains. One of his
last papers was published in the Annual Review of Physical Chemistry in 1981
at the age of 100. He died in 1983.
1.1 Most Liquid Solutions Are Not Ideal

Chemistry in the laboratory very often involves the use of liquid solutions. This is
especially true in chemical analysis, where the amount of analyte is easily manipulated when it is dissolved in a solution. Solutions are often the medium for
chemical reactions which form the basis of titrations. Other simple analytical
procedures are based on absorption spectroscopy, which is used to determine
the concentrations of an analyte in solution.
3
LIQUIDS, SOLUTIONS, AND INTERFACES
Most liquid solutions, also called liquid mixtures, are non-ideal. This follows
from the fact that the components are in intimate contact with one another, and
that the forces between the various species are usually not the same. As a result,
the physical properties of the solution, for example, the vapor pressure of a given
component, are usually not simply related to its concentration. This non-ideality
leads to the concept of the activity of a solution component. As far as the analytical chemist is concerned, only concentration is ultimately of interest. Thus, if an
analysis is based on the measurement of a physical property which in turn
depends on the activity of a component, it is very important that the relationship
between activity and concentration be understood for the system in question.
Activity and its relationship to concentration is dened within the context of
chemical thermodynamics. Using the laws which govern phase equilibria and the
laboratory observations relating to these processes one can develop a detailed
understanding of this relationship. In this chapter the macroscopic concepts of
chemical thermodynamics which are relevant to solutions are reviewed. In addition, some simple models based on molecular concepts are discussed. The examples chosen are mainly limited to non-electrolyte solutions, especially those
involving polar molecules.
1.2 Concentration Units

Concentration of one component in a two-component system can be expressed in
several ways: as a weight/weight ratio, as a volume/volume ratio, or as a weight/
volume ratio. Physical chemists clearly prefer to express concentration as a
weight/weight ratio because then one has the possibility of estimating the number
of moles of both components in the solution. In this case, solution composition is
independent of temperature and pressure. On the other hand, the analytical chemist prefers to use a weight/volume ratio. This is usually because one component,
namely, the analyte, is present at low concentration. Then, one refers to this
component as the solute, and the majority component as the solvent. However,
in this case the concentration changes when temperature or pressure is changed.
Consider the simple example of a solution of acetonitrile in water formed by
mixing 10 g of acetonitrile with 90 g of water. The concentration of acetonitrile
can be simply stated as 10% by weight. Another way of expressing the concentration is in terms of the relative number of moles of these molecules. Given that
the molecular mass of acetonitrile is 41.04 g, the number of moles of acetonitrile
nB used to form the solution is 10.000/41.04 0.2437. The corresponding number
of moles of water nA is 90.000/18.02 4.9945 where 18.02 is its molecular mass.
Thus, one may express the concentration as the mole fraction, xB , of acetonitrile
where
xB
nB
0:2437
0:0465
nA nB 4:9945 0:2437
1:2:1
Another commonly used concentration unit in physical chemistry is molality.

It is dened as the number of moles of component B per 1000 g of pure component A, which is regarded here as the solvent. In the present case, the molality is
THE THERMODYNAMICS OF LIQUID SOLUTIONS
0:2437 1000=90 2:708 m. This is still a weight/weight ratio but has units of
mol kg1 . The relationship between mole fraction and molality can be written
xB
mB
1000=MA mB
1:2:2
where MA is the molecular weight of the solvent, that is, component A. In dilute
solutions for which mB 1000/MA this relationship becomes
xB
m B MA
1000
(dilute solutions)
1:2:3
For example, a 0.02 m solution of acetonitrile in water corresponds to a mole

fraction xB equal to 3:6 104 .
The concentration unit used for analysis is molarity, that is, the number of
moles of solute per liter of solution. It should be noted that the molarity involves a
weight/volume ratio, and that the volume involved is that of the total solution. In
order to determine the molarity of the system being considered here, one must
know the density of the solution. In general, this property cannot be determined
from the densities of the individual components but must be found in an independent experiment. The density of acetonitrilewater solutions as a function of
the weight fraction of acetonitrile is shown in g. 1.1. From these data one nds
that the density of the solution made of 10 g acetonitrile and 90 g water is 0.979 g
mL1 at 25 C. Thus, the volume of the same solution is 102.15 mL and the
corresponding molarity, 0.2437/0.10215 2.386 M. The relationship between
the molarity cB and mole fraction xB is
xB
cB
1000 r cB MB =MA cB
1:2:4
Fig. 1.1 Plots of the density of aqueous solutions of acetonitrile (AcN,*) and sodium
hydroxide (^) against their weight fraction in the solution.
Table 1.1 Concentration of Acetonitrile in Aqueous Solutions

Using Different Expressions for the Relative Amounts of
Acetonitrile and Water at 25 C
Weight
Percent
Mole
Fraction/xB
Molality,
mB /mol kg1
Molarity,
cB /mol L1
Solution Density,
r=g L1
0.01
0.1
1
10
4:39 105
4:39 104
4:42 103
0.0465
2:44 103
2:44 102
0.246
2.708
2:44 103
2:44 102
0.245
2.386
0.997
0.997
0.995
0.979
where r is the density of the solution in g mL1 . When the solution is dilute, one
may neglect the terms in the denominator involving cB so that the expression for
xB becomes
xB
cB MA
1000 rA
(dilute solutions)
1:2:5
where rA is the density of the pure solvent A. On comparing equations (1.2.3) and
(1.2.5), one sees that for aqueous solutions where rA 1:0, the molarity is equal
to the molality when the solution is dilute. These calculations are illustrated in
table 1.1 for a change in acetonitrile concentration by a factor of 1000 in the range
of dilute solutions. It is clear that molality and molarity are equal for dilute
aqueous solutions. However, if water is not the solvent, the density of the solution
is probably suciently dierent from unity that these quantities are no longer
equal. Notice also that the denition of molarity is temperature dependent
because the volume of the system depends on temperature. Thus, the analyst
should always cite the temperature at which solutions were prepared.
A plot of the density of sodium hydroxide solutions of varying composition is
also shown in g. 1.1. Electrolyte solutions are considerably more dense than
water when they are concentrated. The important point to remember about
these systems is that their compositions cannot be varied over the whole range
because the solute is normally a solid at room temperature. Thus, the range of the
weight fraction scale is determined by the solubility of the solid solute. In order to
convert from molality to molarity, the solution density must be determined.
Density data for common solutions can be found in data compilations such as
the LandoltBornstein tables.
1.3 Thermodynamic Quantities

The composition of a solution is obviously one of its important properties. In the
preceding section various ways of describing the composition of a two-component
system were described. Other properties include its volume, V, internal energy, U,
and entropy, S. In order to specify any one of these, one must specify not only the
amounts of each of the components but also the temperature, T, and pressure, P.
These quantities are known as the independent variables of the system. They are
to some extent arbitrary, but nevertheless convenient. If the system possesses

more than two components, the number of variables required to be specied to
determine any quantity such as V is N 2, where N is the number of components.
Consider rst of all the volume of the solution. The volume of any solution
may be estimated from the mass of each component and its density. Volume is an
extensive property, since its value depends on the total amount of solution. A
quantity of more fundamental interest is the specic volume, Vs , that is, the
volume per gram. It is simply the reciprocal of the density. This is an intensive
quantity, since its value does not depend on the size of the solution, only on its
composition, temperature, and pressure. From the point of view of chemists, an
even better way to describe this property is in terms of the molar volume, that is,
the volume per mole of solution. For a two-component solution, the molar
volume Vm is related to the density as follows:
Vm
M A xA M B xB
r
1:3:1
Notice that the units of this quantity are L mol1 if the density is expressed in
g L1 . To calculate the volume from the molar volume one must know the number
of moles of each component, nA and nB . Thus,
V nA nB Vm
1:3:2
where the relative amounts of nA and nB are those required to give the mole
fractions xA and xB .
It was pointed out above that the volume is a function of the number of moles
of each component, temperature, and pressure. Thus, one may write for a twocomponent system
V VnA ; nB ; T; P
1:3:3
It follows that the total derivative of the volume dV is given by

@V
@V
@V
@V
dnA
dnB
dT
dP
dV
@nA nB ;T;P
@nB nA ;T;P
@T nA ;nB ;P
@P nA ;nB ;T
1:3:4
where each partial derivative species the change in the volume with a given
independent variable, holding the other independent variables constant. It is
very important that this specication be made; otherwise one does not know
exactly what change is being measured.
The internal energy of the system, U, is dened on the basis of the rst law of
thermodynamics, which, in simple terms, states that energy cannot be created or
destroyed. For a closed system, the gain in internal energy during a process
involving a change in the values of its independent variables is equal to the
heat gained by the system, q, plus the work done on the system, w.
Mathematically, this is expressed as
U q w
1:3:5
If the changes in q and w are innitesimally small, this relationship may be written
dU dq dw
1:3:6
Furthermore, if the work is limited to pressurevolume changes, then

dU dq PdV
1:3:7
The relationship dw PdV reects the fact that when mechanical work is done
on the system, its volume decreases. Otherwise stated, if the volume of the system
increases during a change in state, the system must do work against the surrounding pressure, which leads to a net loss in its internal energy.
The second law of thermodynamics states that all spontaneous processes lead
to an increase in disorder, which is quantitatively measured by means of the
systems entropy, S. For an innitesimally small process involving a ow of
heat into the system, dq, the entropy change, dS, is given by
dS
dq
T
1:3:8
Combining equations (1.3.7) and (1.3.8), one obtains an important result summarizing the rst and second laws in dierential form:
dU TdS PdV
1:3:9
One must remember that this expression applies to a closed system, that is, one in
which no matter enters or leaves (dni 0). If one relaxes this condition for a twocomponent system, then the general expression for the change in internal energy
becomes

@U
@U
dnA
dn
1:3:10
dU TdS PdV
@nA nB ;S;V
@nB nA ;S;V B
where the last two derivatives describe the change in internal energy with the
number of moles of each component. Since equation (1.3.10) gives the total
dierential of U, one arrives at the following denitions of temperature and
pressure:

@U
T
1:3:11
@S nA ;nB ;V

@U
1:3:12
P
@V nA ;nB ;S
Furthermore, equation (1.3.10) suggests that the internal energy is best described
as a function of entropy, volume, and the number of moles of each component,
that is
U US; V; nA ; nB
1:3:13
As a result, chemists have introduced other thermodynamic functions so that the

properties of a system may be considered with respect to more convenient independent variables, especially temperature and pressure.
The three remaining thermodynamic variables and their denitions are as
follows: the enthalpy,
H U PV
the Helmholtz energy,
1:3:14
A U TS
1:3:15
G U TS PV
1:3:16
and the Gibbs energy,

The dierential form of the enthalpy for a closed system may be found by writing
the total derivative:
dH dU PdV VdP
1:3:17
Combining this result with equation (1.3.9), one obtains

dH TdS VdP
1:3:18
For the Helmholtz energy, the corresponding result is

dA SdT PdV
1:3:19
dG SdT VdP
1:3:20
and for the Gibbs energy

It follows that the Gibbs energy for a closed system is conveniently described as
a function of temperature and pressure. Relaxing the condition that the system be
closed, the total derivative of G in an open two-component system becomes

@G
@G
dn
dn
1:3:21
dG SdT VdP
@nA nB ;T;P A
@nB nA ;T;P B
This is a very important equation in discussing chemical equilibria, and is the
starting point for deriving other important results. One sees that the entropy can
be dened from the temperature dependence of the Gibbs energy,

@G
S
1:3:22
@T P;nA ;nB
and the volume from the pressure dependence

@G
V
@P T;nA ;nB
1:3:23
The thermodynamic variables, U; S; H; A; and G introduced above are extensive quantities like the volume V. Thus, the amount of internal energy in a sulfuric
acid solution depends on whether one has 250 mL beaker, a 4 L bottle, or a full
railway tank car. Just as for volume, it is necessary to dene intensive variables
giving the internal energy per gram, Us , or the internal energy per mole, Um . Since
the present discussion is concerned with chemistry, we will use only the molar
quantities Um , Sm , Hm , Am , and Gm which are dened from the corresponding
extensive quantity by equations like equation (1.3.2).
1.4 Partial Molar Quantities

Partial molar quantities are used to describe the change in properties of a multicomponent system when one component is added at constant temperature, pres-
10
sure, and amounts of all other components. In the present section, partial molar
quantities will be considered with respect to the volume of a two-component
system. For example, the partial molar volume of component A is dened as

@V
A
1:4:1
@nA nB ;T;P
and that of B as

B
@V
@nB

1:4:2
nA ;T;P
It follows from equation (1.3.4) that, at constant temperature and pressure, a

change in the volume of a two-component solution is given by
dV A dnA B dnB
1:4:3
This expression may be integrated under conditions that the relative amounts of
nA and nB , that is, the composition of the solution, do not change:
V A n A B n B
1:4:4
The resulting equation states that the volume of the solution may be calculated
given the number of moles of each component and their partial molar volumes. In
terms of the molar volume, this equation becomes
Vm A xA B xB
1:4:5
It can be shown that there is a relationship between the partial molar volumes for
a given solution composition. Taking the total derivative of the volume on the
basis of equation (1.4.4), one obtains at constant temperature and pressure
dV A dnA nA dA B dnB nB dB
1:4:6
Comparing equations (1.4.3) and (1.4.6), one nds that

nA dA nB dB 0
1:4:7
or dividing by the total number of moles, nA nB ,

xA dA xB dB 0
1:4:8
This equation, which is one example of the GibbsDuhem equation, shows that
changes in the partial molar volume of one component may be related to changes
in the same quantity for the other component. Experimentally, it means that one
only has to measure one partial molar volume as a function of composition
provided one has a value of the second partial molar volume at a reference
point. In order to illustrate this point, equation (1.4.8) is written in a form suitable
for calculating A from B :
xB
d
1:4:9
dA
xA B
If the reference point for the integration is a solution consisting of pure component A (xB 0), then the integration constant required is the molar volume of
pure component A, and one may write
11
Fig. 1.2 Plots of the molar volume of aqueous solutions of acetonitrile (AcN) and
methanol (MeOH) against their mole fraction in solution.
xB
A VmA
xB 0
xB
d
xA B
1:4:10
Since the data tabulated in the literature for thermodynamic quantities are
intensive, one needs a method of determining partial molar quantities from
these data. However, the partial molar quantity involves the rst derivative of
an extensive quantity such as the volume with respect to the number of moles of a
particular component. Although the resulting derivative is intensive in nature, it
itself implies that it comes from extensive quantities. The raw data that are used to
estimate @V=@ni are normally the molar volume Vm as a function of the mole
fraction of a given component xi . Thus, it is reasonable to examine the relationship between @V=@ni and @Vm =@xi . A plot of Vm against xB , the mole fraction of
acetonitrile, is shown in g. 1.2 using the data given in table 1.2 for the acetonitrilewater system. Considering the relationship between V and Vm (equation
(1.3.2)),

@V m
1
@V
V

1:4:11
@nA nB nA nB @nA nB nA nB 2
Furthermore,
@xB
@nA
nB
nB
nA nB 2
1:4:12
Now dividing equation (1.4.11) by (1.4.12), one obtains the result that
@V m

V
A
@xB
x B nB
This equation is rearranged to give
1:4:13
12
Table 1.2 Molar Volume Data for the AcetonitrileWater System at

25 C
Partial Molar Volumes
Mole Fraction Molar Volume
Excess
of Acetonitrile of the Solution Molar Volume
Water
Acetonitrile
1
xAcN
Vm /mL mol1 V ex
AcN / mL mol1 w /mL mol1
m /mL mol
0
0.2
0.4
0.6
0.8
1.0
18.07
24.56
31.48
38.60
45.85
53.01
0
0.50
0.56
0.44
0.17
0
Vm A xB
51.83
52.51
53.03
53.08
53.01
@V m
@xB
18.07
17.75
17.46
16.95
16.91
1:4:14
One may interpret this equation by relating it to the equation for a simple straight
line. Referring to g. 1.2, it follows that a line drawn with a slope equal to @Vm
=@xB at the point (Vm , xB ) will intersect the Vm axis at A , which is the y intercept.
Such a line is shown on the molar volume plot in g. 1.2 at the point ^ (xB 0.5)
for the methanolwater system. In a similar way, it is easily shown that
Vm B xA
@V m
@xA
1:4:15
Since a plot of the molar volume against the mole fraction of B is easily converted
to one against the mole fraction of A (xA 1 xB ), equation (1.4.15) shows that
the intercept on the right-hand ordinate of g. 1.2 (xB 1, xA 0) gives the
partial molar volume of component B.
On the basis of the above analysis it has been shown the partial molar quantities are easily obtained from intensive quantities like the molar volume Vm when
this quantity is plotted as a function of an intensive composition variable like the
mole fraction. The plots in g. 1.2 show that the molar volume is almost a linear
function of the mole fraction of solute. If the curves in g. 1.2 were actually
perfect straight lines, the partial molar volumes would be constant independent
of solution composition. Such a situation would arise if the solution were perfectly
ideal. In reality, very few solutions are ideal, as will be seen from the discussion in
the following section. In order to see more clearly the departure from ideality, one
denes and calculates a quantity called the excess molar volume. This quantity is
equal to the actual molar volume less the molar volume for the solution if it were
ideal. The latter can be considered as the volume of the solution that would be
found if the molecules of the two components form a solution without expansion
or contraction. Thus, the ideal molar volume can be dened as
V id
m xA VmA xB VmB
1:4:16
and is calculated directly from the molar volumes of the pure components A and
B. It follows that the excess molar volume is given by

id
V ex
m Vm V m Vm xA VmA xB VmB
13
1:4:17
The excess molar volume is also called the molar volume of mixing.
Plots of V ex
m against the mole fraction of solute are shown for the acetonitrile
water and methanolwater systems in g. 1.3. The V ex
m function focuses attention
on the non-ideality of the solution. The plots demonstrate clearly that the behavior of these solutions is not simple. Both systems have negative excess volumes.
This indicates that the molecules occupy a smaller volume in solution than they
do as pure liquids. Such an observation is undoubtedly due to attractive forces
between the two components, which are stronger than the forces between molecules in the pure liquids.
The partial molar volumes may be calculated from the data for the excess
molar volume in a manner similar to that used with the molar volume data. On
the basis of equation (1.4.17), one may write
@V ex
@V m
m
VmA VmB
@xB
@xB
1:4:18
Then substituting equation (1.4.13), one obtains after rearrangement

A Vm xB
@V ex
m
xB VmA VmB
@xB
1:4:19
Using the denition of V ex

m given by equation (1.4.17), this simplies to
A V ex
m xB
@V ex
m
VmA
@xB
1:4:20
In a similar way, one may show that
Fig. 1.3 Plots of the excess molar volume for the acetonitrile (AcN)water and methanol
(MeOH)water systems against the mole fraction of these solutes.
14
B V ex
m xA
@V ex
m
VmB
@xA
1:4:21
EXAMPLE
The application of equations (1.4.20) and (1.4.21) to determine partial molar

volumes in the methanolwater system is now illustrated. The excess molar
volume for this system is fairly symmetrical with respect to the mole fraction of
methanol, and can be tted with reasonable accuracy by a cubic equation using
least squares. The result is
V ex
m 0:0286 4:163xB 4:341xB2 0:1977xB3
1:4:22
The resulting curve is shown in g. 1.3, drawn through the experimental points.
On the basis of this analytical expression, one may now write an analytical
expression for the rst derivative at any point on the curve. This equation is
@V ex
m
4:163 8:682xB 0:5931xB2
@xB
1:4:23
Now, suppose one wants the values of the partial molar volumes at xB 0.4.
1
Substituting into equation (1.4.22), one nds that V ex
m 0:955 mL mol ;
ex
1
similarly, from equation (1.4.23), @V m =@xB 0:785 mL mol . The molar
volume of pure water, VmA , is 18.07 mL mol1 , and that of pure methanol,
VmB , 40.72 mL mol1 . Calculating the partial molar volume of water from
equation (1.4.20), one obtains
A 0:955 0:4 0:785 18:07
17:43 mL mol1
1:4:24
Similarly, from equation (1.4.21) for the partial molar volume of methanol,
B 0:955 0:6 0:785 40:72
39:29 mL mol1
1:4:25
These calculations can be carried out for any value of xB .

Partial molar quantities can be dened for any of the remaining thermodynamic functions including the internal energy U, the enthalpy H, the Helmholtz
energy A, and the Gibbs energy G. Those most used in chemistry are the chemical
potentials, which are dened from the Gibbs energy for the system. Thus, the
chemical potential of component A is dened as

@G
1:4:26
mA
@nA nB ;T;P
and that for component B as
@G
mB
@nB

1:4:27
nA ;T;P
15
These quantities are connected to the molar Gibbs energy for the solution by the
equation
Gm xA mA xB mB
1:4:28
The application of these quantities to understanding physical and chemical equilibria in solutions is investigated in the sections which follow.
1.5 Ideal SolutionsRaoults Law

The concept of an ideal solution is important in the development of an understanding of the properties of real solutions. In a liquid solution, molecules are in
intimate contact with one another so that the question of ideality is determined by
the nature of the intermolecular forces. Suppose a solution is formed by mixing
two liquids, A and B. Then, the solution is ideal if the intermolecular forces
between A and B molecules are no dierent from those between A and A, or B
and B molecules.
An indication of whether or not the above condition for ideality is met is
obtained from the vapor pressure of the solution. At a given temperature, the
vapor pressure of a pure liquid is a measure of the ability of molecules to escape
from the liquid to the gas phase. By studying the vapor pressure of a solution as a
function of its composition at constant temperature one may assess the solutions
ideality or its degree of departure from ideality. For an ideal solution, the tendency of molecule A to escape is proportional to its mole fraction, that is, to its
concentration expressed in terms of the fraction of molecules which are of type A.
The proportionality constant must be the vapor pressure of pure component A
because this vapor pressure is reached when the mole fraction is unity. This result
is Raoults law, which is expressed mathematically as
PA PA xA
1:5:1
Similarly, for the other component B,

PB PB xB
1:5:2
A system which is close to ideal in its behavior is a solution of benzene and

toluene. These molecules are very similar in structure so that the intermolecular
forces between benzene and toluene do not dier greatly from those between
benzene molecules or toluene molecules in each pure liquid. The vapor pressure
diagram for this system at 25 C is shown in g. 1.4. Since toluene has a higher
molecular weight, its vapor pressure as a pure liquid is lower (3.25 kPa) than that
of benzene (12.69 kPa). The total vapor pressure PT is obtained by adding those
of components A and B, so that
PT PA PB PA PB PA xB
1:5:3
The total vapor pressure also a linear function of the mole fractions xA and xB .
It is clear from the above that Raoults law denes ideality by relating the
properties of the liquid solution to the vapor with which it is in equilibrium.
Since one knows how to deal with the thermodynamic properties of ideal gaseous
16
Fig. 1.4 Vapor pressure of benzene, PBZ , and toluene, PTL , and total vapor pressure, PT ,
plotted against the mole fraction of toluene xTL for benzenetoluene solutions at 25 C.
solutions, one now has a route to develop the thermodynamics of ideal liquid
solutions.
1.6 Thermodynamics of Ideal Solutions

The thermodynamic equations for ideal solutions are derived by considering the
equilibrium between a given component in the vapor phase and liquid solution.
Thus, for component A in a solution containing two components,
Aliquid Avapor
1:6:1
The thermodynamic condition for equilibrium is that the chemical potential of A

in the liquid phase be equal to that of A in the vapor, that is,
msA mvA
1:6:2
where msA is the chemical potential of A in the liquid solution, and mvA , that for A
in the vapor phase. If one now assumes that the vapor phase behaves ideally, one
may write
mvA mv;
A RT ln PA
1:63
where mv;
A is the standard chemical potential of A in the vapor phase measured
when the partial pressure of A is 1 bar, and PA , the actual partial pressure of A.
Since Raoults law applies when the liquid solution is ideal, one may also write
v;
msA mvA mA
RT ln xA PA
1:6:4
17
where xA is the mole fraction of A and PA the partial pressure of the pure liquid
v;
and PA are constants for any given temperature and pressure,
A. Since mA
equation (1.6.4) may be rewritten as
msA ms;
A RT ln xA
1:6:5
s;

mv;
mA
A RT ln PA
1:6:6
where
s;
mA
Thus,
is the standard chemical potential of component A in the liquid solution, which can be measured when the mole fraction of A is one (pure A). A
similar analysis for the other component B leads to the equation
msB ms;
B RT ln xB
1:6:7
v;

ms;
B mB RT ln PB
1:6:8
where
The thermodynamic quantities associated with the mixing of pure liquids to

form a solution are important in assessing solution properties. Suppose nA moles
of component A are combined with nB moles of component B to form a solution.
The Gibbs energy change associated with this process is given by
s;
nB ms;
mix G nA msA nB msB nA mA
B
1:6:9
On the basis of equations (1.6.5) and (1.6.7), this may be rewritten as

mix G nA RT ln xA nB RT ln xB
1:6:10
Dividing both sides by the total number of moles, nA nB , one obtains

mix Gm xA RT ln xA xB RT ln xB
1:6:11
where the subscript m indicates that the Gibbs energy change is given on a
molar (intensive) basis. It is easy to see that this result can be generalized to a
multicomponent system with n components by writing
mix Gm
n
X
xi RT ln xi
1:6:12
i1
Since the mole fraction xi is less than one, its logarithm is negative. Thus, mix Gm
is a negative quantity, indicating that the mixing process is spontaneous, as one
would expect.
The entropy change associated with mixing can be obtained by taking the
temperature derivative of mix Gm . Accordingly,
mix Sm
n
X
@mix Gm

xi R ln xi
@T
i1
1:6:13
This quantity is clearly positive, since the mixing process results in an increase in
entropy.
On the basis of the denitions of Gibbs energy and enthalpy (equations
(1.3.14) and (1.3.16), the enthalpy of mixing is given by
18
mix Hm mix Gm Tmix Sm
1:6:14
Substituting in equations (1.6.12) and (1.6.13), one obtains

mix Hm 0
1:6:15
This result gives one of the important properties of ideal solutions, namely, that
the mixing process does not involve any heat. Since the components of the solution interact with each other in exactly the same manner that they interact with
themselves in the pure liquid, mixing is neither exothermic nor endothermic.
By taking the pressure derivative of mix Gm , one may determine the volume
change associated with mixing:
mix Vm
@mix Gm
0
@P
1:6:16
The fact that the volume change associated with mixing the components is zero
gives another important property of an ideal solution. On the other hand, a
volume change does accompany the formation of most solutions. One example
was analyzed above in section 1.4. This change is another reection of the fact
that the energy due to the intermolecular forces between the components changes
with solution composition.
In summary, there are three important characteristics of ideal solutions that
one should remember in assessing the properties of any non-ideal system: (i) the
vapor pressure of each component is proportional to its mole fraction in solution
over the whole composition range (Raoults law); (ii) the enthalpy of mixing is
zero; (iii) the volume change associated with mixing is zero. The sections which
follow deal with non-ideal solutions.
1.7 Non-Ideal Solutions

Most solutions are non-ideal. This is simply a result of dierences in the chemical
nature of the molecular components in the solution, and in the way in which they
interact with each other. A convenient way of examining intermolecular forces in
a pure liquid is in terms of its internal pressure, Pi , which is dened as

@U
1:7:1
Pi
@V T
This quantity was investigated extensively by Hildebrand [1], who showed that the
internal pressure is approximately equal to the enthalpy of vaporization divided
by the molar volume. Thus,
Pi
vap H m
Vm
1:7:2
Values of the internal pressure for some commonly used solvents are given in
table 1.3. It is apparent that the internal pressure varies considerably from one
solvent to another, water having the highest value among those considered.
19
Table 1.3 Values of Internal Pressure for Some Common Solvents at

25 C
Solvent
Acetonitrile
Acetone
Carbon tetrachloride
Chloroform
Dimethylsulfoxide
Hexane
Nitrobenzene
Methanol
Propylene carbonate
Water
Molar Volume
Vm =cm3 mol1
Enthalpy of Vaporization
vap Hm / kJ mol1
Internal
Pressure
Pi /kJ dm3
52.9
74.0
97.1
80.7
71.3
131.6
102.7
40.7
85.2
18.07
33.2
30.8
32.4
32.2
52.9
31.55
52.5
37.43
42.8
43.99
628
416
334
399
742
240
511
917
502
2434
Thus, one has a clear indication that intermolecular forces are signicantly
dierent in these liquids. Under these circumstances, a solution formed from
two of them would not be ideal and would generally exhibit positive deviations
from Raoults law. It is also obvious that not every pair of liquids formed from
those shown in the table are miscible. Thus, a very non-polar solvent such as
hexane is immiscible with a very polar one like water. For those which are
miscible, the dierence in internal pressure gives a good indication of the extent
of departure from ideality. A very few systems show a negative deviation from
Raoults law behavior. This occurs when there is a strong attraction between the
two molecules forming the solution, a well-known example being the chloroformacetone system.
An example of a system exhibiting a small positive deviation from Raoults
law is a methanolwater solution (see g. 1.5). It should be noted that when the
concentration of methanol is small (xMeOH < 0.1), the vapor pressure of water is
close to the value expected on the basis of Raoults law. Similarly, for a dilute
solution of water in methanol (xMeOH > 0.9, xw < 0.1), the vapor pressure of
methanol is approximately equal to that in an ideal solution. These conditions
are often observed in dilute solutions and have important consequences with
respect to their thermodynamic properties, as will be seen in the following
section.
A much more complex behavior is demonstrated by the acetonitrilewater
system [2] (g. 1.6). The vapor pressure curves show an interesting change in
slope at a mole fraction of acetonitrile close to 0.8. However, when the mole
fraction of acetonitrile is less than 0.06, the vapor pressure of water is close to
the value predicted by Raoults law.
As mentioned above, a few systems show negative deviations from Raoults
law, a well-known example being the acetonechloroform system (g. 1.7). In this
case there is attractive interaction between the two components, specically,
between the electron-rich oxygen in acetone, and the hydrogen atom in chloroform. As a result, the escaping tendency of either molecule from the solution is
20
Fig. 1.5 Vapor pressure of methanol and water for methanolwater solutions plotted
against the mole fraction of methanol at 40 C. The left-hand ordinate scale gives the
vapor pressure of water and the right-hand scale that of methanol.
less than it would be if the chemical interaction did not occur. These systems are
similar to those showing positive deviations from Raoults law in that the vapor
pressure of the predominant component approaches ideal behavior when the mole
fraction of the minority component is very small, that is, with a mole fraction less
than 0.1.
Now we proceed to examine the thermodynamic properties of non-ideal
solutions.
Fig. 1.6 Vapor pressure of acetonitrile and water for acetonitrilewater solutions at 25 C
plotted against the mole fraction of acetonitrile. The left-hand ordinate scale gives the
vapor pressure of water and the right-hand scale, that of acetonitrile.
21
Fig. 1.7 Vapor pressure of acetone and chloroform for acetonechloroform solutions at
35 C plotted against the mole fraction of acetone.
1.8 Thermodynamics of Non-Ideal Solutions

The starting point for developing the thermodynamics of non-ideal solutions is
the same as that for ideal solutions. Thus, one considers the equilibrium between
each component in the liquid solution and its vapor. It follows from section 1.6
that one may write in general for component A,
msA mvA mv;
A RT ln PA
1:8:1
This equation states that chemical potentials of component A in the liquid solution and vapor are equal and that each relates to the vapor pressure of A.
However, one would like to have a way of relating the chemical potential of A
to its mole fraction in solution. This is achieved by relating the vapor pressure of
A to its mole fraction in the liquid solution using a correction factor to make the
value of PA predicted by Raoults law equal to the true value. Thus, one writes
PA gA xA PA
1:8:2
where gA is the correction factor, known as the activity coecient of A. It is clear

that gA is greater than unity when the system exhibits positive deviations from
Raoults law, and less than unity when the deviations are negative. Furthermore,
the value of gA depends on xA , since the extent of departure varies with solution
composition (see gs. 1.51.7). Combining equations (1.8.1) and (1.8.2), one
obtains
msA ms;
A RT ln gA xA
1:8:3
s;
is given by equation (1.6.6). It is emphasized that the standard state for
where mA
this denition of the activity coecient is based on the properties of pure component A. As the mole fraction of A approaches unity, the activity coecient gA
also approaches unity, as can be seen from the vapor pressure plots presented
earlier. It is also emphasized that the standard chemical potential has the same
22
value as it would have in an ideal solution because it only depends on the properties of pure component A. The product gA xA is known as the activity of component A in solution. Thus, one may also write
s;
RT ln aA
msA mA
1:8:4
a A g A xA
1:8:5
where
In a similar way, one may write for component B,

s;
msB ms;
B RT ln aB mB RT ln gB xB
1:8:6
where mBs; is given by equation (1.6.8) and

gB PB =xB PB
1:8:7
It is interesting to evaluate the thermodynamic functions of mixing for the nonideal solution. On the basis of equations (1.8.3) and (1.8.6),
mix G nA RT ln gA xA nB RT ln gB xB
1:8:8
mix Gm xA RT ln gA xA xB RT ln gB xB
1:8:9
or on a molar basis
In the case of non-ideal solutions the mixing functions are often referred to the
value they would have in an ideal solution, mix Gid
m , thereby dening the excess
Gibbs energy of solution formation:
id
mix Gex
m mix Gm mix Gm xA RT ln gA xB RT ln gB
1:8:10
In order to calculate the other excess functions one must know the temperature
and pressure derivatives of the activity coecients gA and gB . The excess entropy
of mixing is given by
mix Sex
m xA R ln gA xA RT
@ ln gA
@ ln gB
xB R ln gB xB RT
@T
@T
1:8:11
and the excess enthalpy by

2
mix H ex
m xA RT
@ ln gA
@ ln gB
xB RT 2
@T
@T
1:8:12
Finally, the excess volume of mixing is obtained from the pressure derivatives of
the activity coecients:
mix Vmex xA RT
@ ln gA
@ ln gB
xB RT
@P
@P
1:8:13
EXAMPLE
The vapor pressure of acetonitrile above an acetonitrilewater solution with a

mole fraction of 0.395 in acetonitrile is 9.727 kPa at 25 C. The corresponding
vapor pressure of water is 2.874 kPa. At the same temperature the vapor
23
pressures of the pure liquids are 11.983 kPa and 3.166 kPa for acetonitrile and
water, respectively. Estimate the Raoult law activity coecients for each component, the molar Gibbs energy of mixing, and the excess value of this function. The enthalpy of mixing for this solution is 876.1 J mol1 . Estimate the
entropy of mixing and its excess value.
The Raoult law activity coecient for acetonitrile is
gAcN
9:727
2:055
0:395 11:98
1:8:14
The corresponding quantity for water is

gw
2:874
1:500
0:605 3:166
1:8:15
The molar Gibbs energy of mixing is

mix Gm 0:395 2479:4 ln2:055 0:395
0:605 2479:4 ln1:500 0:605
1:8:16
349:9 J mol1
The excess molar Gibbs energy of mixing is
mix Gex
m 0:395 2479:4 ln2:055 0:605 2479:4 ln1:500
1313:6 J mol1
1:8:17
The enthalpy of mixing is also the excess enthalpy of mixing because an ideal
solution has zero enthalpy of mixing. Now, the entropy of mixing can be
calculated:
mix S m
mix H m mix Gm 876:1 349:9

4:111 J K1 mol1
298:2
T
1:8:18
and
mix Sex
m
ex
mix H ex
876:1 1313:6
m mix Gm
1:467 J K1 mol1
298:2
T
1:8:19
Values of the excess Gibbs energy, enthalpy, and entropy for the acetonitrile
water system which show signicant departures from ideality are shown as a
function of solution composition in g. 1.8. The excess enthalpy is positive over
the whole composition range, reaching a maximum value of 1067 J mol1 in the
vicinity of an acetonitrile mole fraction equal to 0.7. These data give a direct
measure of the endothermic nature of the mixing process. The excess entropy
displays a rather complex behavior, being negative at lower concentrations of
acetonitrile and positive for values of xAcN greater than 0.7. Excess thermodynamic data such as those shown in g. 1.8 provide a convenient way of recording
the properties of non-ideal solutions and are often found in tables for liquidvapour equilibria. However, they provide information about the solution as a
24
Fig. 1.8 Excess Gibbs energy, enthalpy, and entropy for acetonitrilewater solutions at
25 C plotted against the mole fraction of acetonitrile.
whole, not about the individual components. If one wants the activity coecients
for each component, one must also have vapor pressure data for each component
as a function of solution composition.
In assessing the above treatment of non-ideal solutions, it must be kept in mind
that it is applicable to a limited number of systems. This follows from the fact that
we are often dealing with solutions of solids in liquids, and also because not all
liquids are miscible over the whole composition range. Under these circumstances
it is not convenient to dene the standard state for one component in terms of the
pure substance. Thus, for the majority of solutions, the majority component is
treated as the solvent and its thermodynamics discussed with respect to its pure
state within the context of Raoults law. The other minority component, which is
the solute, is discussed using a standard state based on the properties of an ideally
dilute solution. These systems are considered in more detail later in this chapter.
1.9 Regular Solutions

In examining the properties of non-ideal solutions, it became clear that some
systems dier from ideality in a manner which could be treated by quite simple
statistical mechanical models. The solution non-ideality is often reected in the
experimental observation that the enthalpy of mixing is not zero, so that AB
interactions are dierent from AA and BB interactions. On the basis of his
study of the properties of a large number of liquid solutions, Hildebrand [1, 3]
introduced the concept of a regular solution. This is a system for which the
enthalpy of mixing is non-zero and the entropy of mixing has its ideal value so
that Sex is zero. This is equivalent to assuming that the molecules are randomly
distributed in the mixture so that the dierences in the intermolecular forces
which lead to the non-zero value of H ex cannot be large. Guggenheim [4, 5]
25
discussed strictly regular solutions which have the additional restriction that
mix V ex
m be zero. These types of systems are considered in this section.
Consider a mixture of two molecules A and B which are both approximately
spherical in shape. As perfect spheres they can pack together in a face-centered
cubic lattice to form a liquid with a coordination number c of 12 in each pure
liquid. The mixture packs in the same way provided the molecular sizes are not
too dierent, more specically, provided the molecular volumes do not dier by
more than a factor of two [4]. If the free volume between molecules in the mixture
does not dier from the sum of those in the two pure liquids used to form the
mixture, then the volume of mixing is eectively zero, and the interaction energy
experienced by a given molecule in the mixture may be calculated by summing the
contributions from nearest neighbors. Mixtures with these properties are strictly
regular.
Let us now consider how one can estimate the enthalpy of mixing given the
enthalpies associated with AA, AB, and BB interactions at the molecular level.
If each molecule has c nearest neighbors, then the number of interactions experienced by type A molecules is cnA /2, and the number of B molecules, cnB /2, where
the factor of two appears in order to avoid counting the interactions twice. If one
denes the number of AA interactions as nAA , the number of BB interactions as
nBB , and the number of AB interactions as nAB , it follows that
cnA 2nAA nAB
1:9:1
cnB 2nBB nAB
1:9:2
and
By adding these equations, one obtains an expression for twice the total number
of interactions in the solution.
Suppose that the enthalpy associated with an AA interaction is hAA , that with
BB, hBB , and that with an AB interaction, hAB . Then the enthalpy of nA molecules in pure A associated with intermolecular interactions is
cn
1:9:3
HA A hAA
2
Similarly, for pure B,
HB
cnB
h
2 BB
1:9:4
The enthalpy of the solution formed from these pure liquids, associated with
intermolecular interactions is
Hsl nAA hAA nBB hBB nAB hAB
1:9:5
As a result, the enthalpy of mixing becomes

mix H Hsl HA HB

hAA hBB
nAB hAB
2
1:9:6
In order to develop this model further one has to obtain an expression for nAB
in terms of nA and nB . The assumption used is that the molecular composition
26
around a given molecule is completely random and therefore reects the overall
solution composition. This is precisely the assumption used by Hildebrand to
dene a regular solution. Thus, the number nAB can be calculated from the
mole fractions dening solution composition, and is given by
nAB cnxA xB
1:9:7
where n is the total number of molecules (n nA nB ). The enthalpy of mixing

can now be expressed as
mix H cnxA xB h
1:9:8
where
h hAB
hAA hBB
2
1:9:9
Since the mixing process is completely random, one may use the value of Smix
for an ideal solution, which for the present system is given by
mix S nA R ln xA nB R ln xB
1:9:10
Thus, the expression for the Gibbs energy of mixing becomes

mix G nA RT ln xA nB RT ln xB cnA nB xA xB h
1:9:11
One can now derive expressions for the chemical potentials of the individual
components. Since the Gibbs energy of the solution is
s;
nB mBs; mix G
Gsl nA mA
1:9:12
the chemical potential of A is obtained by dierentiating Gsl with respect to nA

(equation (1.4.26)):
2
mA ms;
A RT ln xA cxB h
1:9:13
Similarly, one nds for B,

mB mBs; RT ln xB cx2A h
1:9:14
Thus, according to the model for a regular solution the activity coecients gA and
gB are given by
ln gA
cx2B h
RT
1:9:15
ln gB
cx2A h
RT
1:9:16
and
These quantities, in turn, may be related to the vapor pressure of each component
over the solution using equations. (1.8.2) and (1.8.7):
!
cx2B h

PA xA PA exp
1:9:17
RT
and
PB xB PB exp
cx2A h
RT
27
1:9:18
A plot of vapor pressure data for a hypothetical system, assuming ch=RT is

unity, is shown in g. 1.9. The positive deviations from ideality indicate that the
mixing process is endothermic. As the parameter h is increased, the deviations
increase in the positive direction. Obviously, the theory also predicts negative
deviations from Raoults law when h is negative, that is, when the mixing
process is exothermic. Under these conditions the intermolecular forces between
the two species A and B are attractive, and the escaping tendency of each is less
than it would be if the solution were ideal (h 0).
EXAMPLE
Estimate the vapor pressure of the two components in a regular solution for
which ch=RT 1 and xA 0.4 given that the vapor pressure of pure
component A is 15.0 kPa and that of pure B, 20.0 kPa. Also calculate the
Raoult law activity coecients. Repeat the calculation for the case that
ch=RT 1.
From equation (1.9.17) at xA 0.4, the vapor pressure of A is
PA 0:4 15:0 exp 1 0:36 8:60 kPa
1:9:19
The vapor pressure of B is

PB 0:6 20:0 exp 1 0:16 14:08 kPa
1:9:20
The activity coecients are

gA
PA
8:60
1:433

xA pA 0:4 15:0
1:9:21
Fig. 1.9 Vapor pressure for a hypothetical regular solution for which chRT 1 plotted
against the mole fraction of component B. The vapor pressure of pure component B is 26.7
kPa, and that of component A, 20.0 kPa. The broken lines show Raoult law behavior.
28
and
gB
PB
14:08
1:173

xB PB 0:6 20:0
1:9:22
In the case that the mixing process is exothermic and ch=RT 1, the
vapor pressures are
PA 0:4 15:0 exp1 0:36 4:19 kPa
1:9:23
PB 0:6 20:0 exp1 0:16 10:23 kPa
1:9:24
and
The activity coecients are

gA
PA
4:19
0:698

xA PA 0:4 15:0
1:9:25
gA
PB
10:23
0:853

xB PB 0:6 20:0
1:9:26
and
It is clear that there must be a limit to the endothermicity associated with the
mixing process, if a stable solution is to be formed. As the enthalpy of mixing
increases at a given composition, eventually a value is reached where the Gibbs
energy of mixing is zero. This follows from the fact that the entropy of mixing for
a regular solution is constant for xed solution composition and equal to the ideal
value (equation (1.9.10)). Values of mix Gm for a regular solution as a function of
composition for increasing values of ch=RT are shown in g. 1.10. When this
parameter reaches a value of 3, mix Gm increases over part of the composition
Fig. 1.10 The molar Gibbs energy of mixing in units of RT plotted against the mole
fraction of component B for regular solutions with increasing values of the ratio
w chRT.
29
range; this result indicates that a solution does not form. One may regard the
diagrams in this gure as referring to a unique system at dierent temperatures.
Thus, as temperature decreases, and ch=RT increases, one eventually reaches a
temperature at which the solution separates into its component liquids. The temperature at which phase separation begins to take place is called the critical
temperature, and for regular solutions corresponds to ch=RT equal to 2. It is
easily apparent that the value of mix Gm is approximately constant over most of
the composition range for this value of ch=RT.
The critical temperature at which phase separation begins is dened by the
conditions [4]
@2 mix Gm RT
0
@x2B
1:9:27
@3 mix Gm =RT
0
@x3B
1:9:28
and
In other words, the slope of a plot mix Gm against xB must be constant and equal
to zero over the composition region where the critical phenomenon is observed.
From equation (1.9.11), it follows that
@2 mix Gm =RT
1
1
2ch
0

xA xB
RT
@x2B
1:9:29
@3 mix Gm =RT
1
1
2 2 0
3
@xB
xA xB
1:9:30
Thus, at the critical temperature, when xA xB ,

ch
2
RT
1:9:31
At lower temperatures, this ratio is larger, and phase separation occurs.

The important feature of the above treatment, which is also known in statistical mechanics as the BraggWilliams approximation [6], is that the molecular
composition around a given molecule reects the bulk composition. This cannot
be the case in general because of the dierences in intermolecular forces between
the solution components. Thus, if molecule A interacts more strongly with molecule B than with itself, the local composition of B around A is higher than the
average value. Recognition of this fact leads to more complex descriptions of
mixing phenomena such as that based on the quasi-chemical or Bethe approximation [5]. However, as soon as one accepts that the local molecular composition is
not the same as the average bulk composition, it follows that the entropy of
mixing is not given by the ideal value (equation (1.9.10)) and that the solution
is no longer regular. The quasi-chemical model and other models for non-ideal
molecular solutions have been considered in some detail in the development of
theories for molecular liquid solutions but are not considered further here.
30
1.10 An Empirical Approach to Non-Ideal Solutions

Most real solutions are neither ideal nor regular. As a result a realistic description
of their thermodynamic properties must consider the fact that both the excess
enthalpy of mixing, mix H ex , and excess entropy, mix S ex , are non- zero. Wilson
[7] has proposed an empirical description of the excess thermodynamic properties
of non-ideal systems which provides an excellent description on the basis of two
adjustable parameters. His approach includes systems in which the component
molecules have dierent sizes, and estimates the Gibbs energy of mixing on the
basis of the local volume fractions of each component. It is presented here for the
case of binary mixtures but can easily be extended to systems with more components.
Consider the solution composed of two molecules A and B. The number of
AA interactions, nAA , with respect to the number of AB interactions, nAB , is
given by the overall ratio of A to B in the solution weighted by factors which
account for the enthalpy associated with these interactions, namely, hAA and hAB .
Thus, one writes
nAA nA exphAA =RT
nAB nB exphAB =RT
1:10:1
Similarly, estimating the ratio of the number of BB interactions to AB interactions, one obtains
nBB nB exphBB =RT
nAB nA exphAB =RT
1:10:2
The volume fractions of each molecule, A and B , are then dened using these
ratios and the molar volumes of the two pure components, VmA and VmB :
A
nA V mA exp hAA =RT

nA V mA exp hAA =RT nB V mB exp hAB =RT
1:10:3
B
nB V mB exp hBB =RT

nA V mA exp hAB =RT nB V mB exp hBB =RT
1:10:4
and
The Gibbs energy of mixing is then assumed to be

mix Gm RTxA ln A RTxB ln B
1:10:5
Subtracting o the Gibbs energy of mixing for the ideal solution (equation
(1.6.11)), one obtains for mix Gex
m :
mix Gex
m RTxA lnA =xA RTxB lnB =xB
1:10:6
This may be rewritten as

mix Gex
m RTxA ln1 rBA xB RTxB ln1 rAB xA
1:10:7
where
rBA 1
V mB exp hBB =RT

V mA exp hAB =RT
1:10:8
31
and
rAB 1
V mA exp hAA =RT

V mB exp hAB =RT
1:10:9
The parameters rAB and rBA are treated as adjustable and are chosen to obtain a
good t with the experimental data.
One may now derive expressions for the activity coecients of the two components. On the basis of equation (1.8.10)
mix Gex nA RT ln gA nB RT ln gB
1:10:10
so that

@ mix Gex =RT
@ ln gA
@ ln gB
ln gA nA
nB
@nA
@nA
@nA
1:10:11

@ mix Gex =RT
@ ln gA
@ ln gB
ln gB nA
nB
@nB
@nB
@nB
1:10:12
and
From the GibbsDuhem relationship (see equation (1.4.8) and associated discussion), the last two terms in equations (1.10.11) and (1.10.12) are equal to zero. It
follows that the derivatives of mix Gex =RT with respect to nA and nB give
directly the activity coecients of components A and B. Thus, dierentiating
equation (1.10.7) with respect to nA and simplifying, one obtains
ln gA ln1 rBA xB
xA xB rBA
x2B rAB
1 rBA xB 1 rAB xA
1:10:13
In a similar way, the expression for ln gB is

ln gB ln1 rAB xA
xA xB rAB
x2A rBA
1 rAB xA 1 rBA xB
1:10:14
Wilson [7] demonstrated the eectiveness of this model using data for both
binary and ternary systems, one example being the carbon tetrachlorideacetonitrile system. This system shows positive deviations from Raoults law, as shown
from the data for mix Gex
m presented in g. 1.11. Wilson [7] found that the best
values of the parameters rBA and rAB are 0.6118 and 0.8287 at 25 C, respectively,
where B refers to carbon tetrachloride and A to acetonitrile. The tted curve is
also shown in the gure, from which it is clear that there is excellent agreement
between the model and experiment.
EXAMPLE
Using Wilsons parameters for the carbon tetrachlorideacetonitrile system,

estimate the Raoult law activity coecients for each component in a equimolar
solution. Then estimate the molar Gibbs energy of mixing.
If acetonitrile is component A, then its activity coecient is given by equation (1.10.13) so that
32
Fig. 1.11 Plot of the excess Gibbs energy of mixing for the carbon tetrachlorideacetonitrile system against the mole fraction of acetonitrile at 45 C. The points show the
experimental results and the solid curve was calculated using equation (1.10.7) with
the parameters given by Wilson [7] (see text).
ln gA ln1 0:6118 0:5
0:52 0:6118
0:52 0:8287
1 0:6118 0:5 1 0:8287 0:5
0:3651 0:2204 0:3538 0:4985
1:10:15
The value of gA is 1.646.

Similarly, for carbon tetrachloride which is component B
ln B ln1 0:8287 0:5
0:52 0:8287
0:52 0:6118
1 0:8287 0:5 1 0:6118 0:5
0:5350 0:3538 0:2204 0:4016
1:10:16
The value of gB is 1.494.

The molar Gibbs energy of mixing is given by equation (1.8.9) so that
mix Gm 0:5 8:3145 298:2 ln1:646 0:5 0:5 8:3145
298:2 ln1:494 0:5
603:1 J mol1
1:10:17
Equation (1.10.7) can also be used to derive expressions for the excess entropy
and enthalpy functions. Dierentiating this equation with respect to temperature,
the expression for mix Sex
m is
mix Sex
m RxA ln1 rBA xB RxB ln1 rAB xA
RTxA xB @rBA RT xA xB @rAB

1 rBA xB @T
1 rAB xA @T
1:10:18
33
Combining equations (1.10.7) and (1.10.18), the expression for mix H ex

m is
mix H ex
m
RT 2 xA xB @rBA RT 2 xA xB @rAB

1 rBA xB @T
1 rAB xA @T
1:10:19
The two derivatives @rBA =@T and @rAB =@T constitute two additional parameters
which are obtained by tting mix H ex
m data to the model. Obviously, description
of the entropy requires four parameters.
The eectiveness of Wilsons model lies in the fact that only two parameters are
required to describe the Gibbs energy at a given temperature. Its weakness lies in
the fact that there is no clear molecular interpretation of these parameters.
Wilsons approach works for a great variety of systems but when the departures
from ideality are complex, more detailed models are required. Some extensions of
Wilsons work have been discussed by Renon and Prausnitz [8] but they require
introduction of more adjustable parameters.
1.11 Ideally Dilute Solutions

For many solutions, it is not possible to vary the composition of the components
over the whole range of mole fractions. This is obviously true of solutions made
up of a solid and a liquid. For these systems it is better to choose a standard state
which is based on the properties of a dilute solution. This leads to the denition of
an ideally dilute solution. Such a system is easily dened on a molecular basis as
one in which the solute molecule only comes in contact with solvent molecules,
and never with another solute molecule. In the previous discussion of regular
solutions it was concluded that, when the two components are of equal size, the
coordination number for the other molecules around a central one is twelve. This
suggests that an ideally dilute solution must have a solute mole fraction which is
less than 1/13, that is, 0.08.
The above approximate guideline for an ideally dilute solution can only be
made more exact by examining vapor pressure data for a specic system. The case
of the methanolwater system discussed earlier is used as an illustration. For very
dilute solutions, that is, when xMeOH is less than 0.04, the vapor pressure of
methanol is linear in its mole fraction. This is the region where Henrys law is
obeyed. As the mole fraction increases, the actual vapor pressure falls below that
predicted by Henrys law, quite signicant deviations being found when xMeOH
reaches 0.1. Henrys law for component B in a two-component system of A and B
may be expressed as
PB kH xB
1:11:1
where kH is the slope of the Henry law line. Raoult law behavior is also shown in
g. 1.12. For this non-ideal system, which exhibits positive deviations from
Raoults law, the slope of the Raoult law line, which is equal to the vapor pressure
of pure methanol, is much less than that for the Henry law line. It should be
remembered that in the concentration range over which Henrys law holds for the
solute, Raoults law is valid for the solvent (see g. 1.5). This fact gives one
34
Fig. 1.12 Vapor pressure of methanol for dilute solutions of methanol in water plotted
against the mole fraction of methanol. The straight line shows the vapor pressure according
to Henrys law, and the broken line, that according to Raoults law.
another convenient guideline for judging the concentration range for ideally dilute
behavior.
The value of the Henry law constant and concentration range over which this
law is valid depends very much on the system. This is easily seen by comparing the
behavior of the methanolwater and acetonitrilewater systems (gs. 1.5 and 1.6).
In the latter case ideally dilute solution behavior is observed for a lower range of
mole fractions of the solute, that is, when xAcN is less than 0.03.
The most important application of low concentration behavior of solutes is for
solid solutes, especially electrolytes. Electrolyte solutions are examined in detail in
chapter 3. Some general thermodynamic methods for describing the properties of
very dilute solutions are considered in the following section.
1.12 Thermodynamics of Ideally Dilute Solutions

Just as was done previously, one develops the thermodynamic description of an
ideally dilute solution by considering the equilibrium between the dilute solute
component in the liquid solution and in the vapor phase. If the minority component is designated B, then one may write its chemical potential as
msB mvB mBv; RT ln PB
1:12:1
Since Henrys law holds when the solution is ideally dilute, this can be rewritten as
msB ms;
B RT ln xB
1:12:2
mBs; mBv; RT ln kH
1:12:3
where
35
It should be noted that ms;

B gives the standard chemical potential for the ideally
dilute solute in a hypothetical system in which the mole fraction of B is unity. This
is obviously a ctitious state which is impossible in reality but whose properties
are obtained by extrapolating the Henrys law line to xB 1 (see g. 1.12). When
Henrys law is not obeyed, an activity coecient gH
B is introduced so that the
product gH
k
x
is
equal
to
the
vapor
pressure
P
B . The activity of the dilute
B H B
H
is
dened
to
be
g
x
.
Thus,
the
general
expression for the concomponent aH
B
B B
centration dependence of msB becomes
H
s;
H
msB ms;
B RT ln aB mB RT ln gB xB
1:12:4
where gH
B is the Henrys law activity coecient on the mole fraction scale.
Because of the inconvenient nature of the standard state dened above, the
concentration units used to describe the concentration dependence of the chemical
potential are usually dierent. More convenient choices for concentration are
molality and molarity. When the solution is dilute the relationship between
mole fraction and molality is quite simple (see equation (1.2.3)). In terms of
molality, the expression for the concentration dependence of the chemical potential of component B becomes
msB mBs; RT ln mB
where

v;
ms;
B mB RT ln
kH MA
1000
1:12:5

1:12:6
and MA is the molecular weight of the solvent. Now, the standard chemical
potential is that for a hypothetical system which obeys Henrys law for a solute
concentration of 1 m. For the methanolwater system, xB is equal to 0.018 at this
concentration. Not only is this a dilute solution by the criteria that have been
discussed here, but it is also a system which approximately obeys Henrys law
on the basis of the data shown in g. 1.12. It should be emphasized that this will
not always be the case, signicant departures from Henrys law being observed
for very low concentrations for some systems, for example, acetonitrilewater
mixtures.
When the dilute system does not obey Henrys law, one introduces an activity
coecient as above to correct to the experimentally observed value. Thus, in
general, one may write
PB
gH
B kH MA
mB
1000
1:12:7
where gH
B is the Henrys law activity coecient for component B on the molality
scale. Then, on the basis of equation (1.12.1), the chemical potential of component B may be written
s;
H
msB mBs; RT ln aH
B mB RT ln gB mB
1:12:8
where aH
B is the activity of B on the molality scale with the standard state dened
according to Henrys law (equation (1.12.6)).
36
Alternatively, one may use molarity as the concentration unit. In this case, one
must know the density of the solution to order to relate mole fraction to molarity.
For very dilute solutions, using equation (1.2.5), the concentration dependence of
the chemical potential is given by
msB mBs; RT ln cB
1:12:9
where
mBs; mv;
B RT ln
k H MA
1000rA
1:12:10
rA being the density of the solvent. When the system does not obey Henrys law,
one introduces an activity coecient, gH
B which is the multiplicative correction
factor required to make the vapor pressure of B predicted by Henrys law equal to
that observed experimentally. Thus, gH
B is dened by the equation
PB
gH k H M A
c
1000rA B
1:12:11
and the expression for the concentration dependence of msB becomes

H
s;
H
msB ms;
B RT ln aB mB RT ln gB cB
1:12:12
EXAMPLE
Carbontetrachloridebenzene solutions can be regarded as regular with an

enthalpy parameter ch equal to 324 J mol1 at 25 C. Given that the vapor
pressure of pure carbon tetrachloride is 14.13 kPa, determine the Henrys law
constant for this component by examining its vapor pressure at mole fractions
in the range 0.001 to 0.1. Then estimate the Henrys law activity coecient at a
mole fraction of 0.1 on the mole fraction and molality scales.
From the theory for regular solutions, the Raoult law activity coecient for
component B in a solution of A and B is
ln gR
B
chx2A
RT
1:12:13
Values of gR
B were estimated for values of xB in the range 0.001 to 0.005 and are
recorded in the following table. Then, the partial pressure of B (carbon tetrachloride) was estimated using the relationship

PB gR
B PB xB
1:12:14
Finally, the ratio PB =xB was also calculated.

xB
gR
B
PB / Pa
PB =100xB
mB
0.001
0.002
0.003
0.004
0.005
1.139
1.139
1.139
1.138
1.138
16.1
32.1
48.3
64.4
80.4
161
161
161
161
161
0.013
0.026
0.039
0.051
0.064
It is clear from these results that the Henrys law constant is 16,100 Pa.
37
At xB 0.1, PB is equal to 1571 Pa. Henrys law predicts that PA is 1610 Pa.
Therefore, the Henrys law activity coecient is 0.976.
The molality is easily calculated from the mole fraction using the relationship
mB
1000xB
MA xA
1:12:15
where MA is the molecular mass of benzene (78.1). Values of the molality for
the dilute solutions are also recorded in the table. The Henrys law constant on
this scale is 1253 Pa kg mol1 . A mole fraction of 0.1 in carbon tetrachloride
corresponds to a molality of 1.42 m. Thus, the predicted vapor pressure by
Henrys law is 1780 Pa. As a result the Henrys law activity coecient is 0.883.
On the basis of the above, the denitions of the standard state (equations
(1.12.3), (1.12.6), and (1.12.10)) and of the activity coecient gH
B (equations
(1.12.4), (1.12.7), and (1.12.11)) depend on the choice of concentration units
used for the dilute solution component. It is emphasized that one must always
state the concentration units in dening these two quantities. This point is further
illustrated in table 1.4, where activity coecients for the components in dilute
aqueous solutions of methanol are tabulated. Notice rst of all that the molality
of methanol has reached 6.2 m when the mole fraction is 0.1. This is due to the
dierence between the molecular mass of water (18 g) and that of methanol (32 g).
On the basis of the Raoult law activity coecients the behavior of water is ideal
over most of this concentration range, the activity coecient being 1.000 to four
Table 1.4 Raoult Law and Henry Law Activity
Coefcients for Dilute Solutions of Methanol in
Water at 40 C
Activity Coecients
Methanol Concentration Raoults Law
xMeOH
mMeOH
/mol kg1
gR
MeOH
gR
w
0.0001
0.0002
0.0005
0.001
0.002
0.005
0.01
0.02
0.05
0.1
0.006
0.011
0.028
0.056
0.111
0.279
0.561
1.133
2.921
6.167
1.7054
1.704
1.704
1.704
1.702
1.696
1.687
1.669
1.618
1.539
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.001
1.005
Henrys Law*
gH
gH
MeOH
MeOH
x s:s (m s:s
1.000
1.000
1.000
1.000
0.999
0.996
0.990
0.979
0.949
0.903
1.000
1.000
1.000
1.000
1.000
0.997
0.987
0.967
0.908
0.819
*The rst column gives the Henry law activity coefcient for
methanol on the basis of a standard state of xMeOH 1 and the
second on the basis of a standard state of mMeOH 1.
38
signicant gures up to a mole fraction of 0.02. At the same time, the behavior of
methanol with respect to Raoults law is non-ideal over the range considered. Since
deviations from ideal behavior are positive (see g. 1.5), the Raoult law activity
coecient for methanol is greater than one; in addition, it changes signicantly in
this concentration range, decreasing steadily in magnitude as the mole fraction
increases. On the other hand, the values of gH
MeOH are unity at the lowest concentrations in the region where Henrys law holds. As concentration increases, these
activity coecients fall below unity and dier according to the choice of standard
state. By multiplying the concentration in particular units by the appropriate
activity coecient, one obtains the Henry law activity. For instance when
xMeOH 0.1, the activity of methanol on the mole fraction scale is 9:90 103 .
On the molality scale, the activity is 0.554 mol kg1 of water. By multiplying the
latter quantity by MA =1000, that is, 0.018 kg mol1 for water, one obtains approximately the activity on the mole fraction scale (see equation (1.2.3)).
The results recorded in table 1.4 emphasize the necessity of clearly stating the
standard conditions for dening activity coecients in liquid solutions. As stated
above, the molality scale is preferred to the mole fraction scale for most systems,
especially those which are not miscible over the whole composition range. The
molality scale is preferred over the molarity scale because the denition of solution
concentration is independent of temperature and pressure. On the other hand, the
molarity concentration scale is so popular in chemistry that one often nds activity
coecients also recorded in the literature using a Henrys law scale on the basis of
molarity. This is especially true of electrolyte solutions, which are always nonideal. When one recognizes that the majority of solutions involve solid solutes,
the importance of the Henrys law denition of the standard state becomes clear.
These solutes usually have a negligible vapor pressure for the conditions that the
solutions are used. Thus, a question arises regarding determination of the activity
of non-volatile solutes. This is dealt with in the following section.
1.13 Experimental Determination of Solution Activities

In order to determine the activity of a component in solution, one must measure
its vapor pressure. In the case of volatile liquids such as those discussed in most of
this chapter, vapor pressure measurement is not a problem so that very accurate
determination of activity is possible over the whole composition range for which a
solution is formed. However, many solutes, for example, most solids, have negligible vapor pressures. Under these circumstances, one makes use of the Gibbs
Duhem relationship between the activities of the two-components in solution.
Since the vapor pressure of the solvent can be measured, its activity can be
determined, and then used to estimate the activity of the solute.
On the basis of equation (1.4.28), for any innitesimal change in a twocomponent system, one may write
dGm xA dmA mA dxA xB dmB mB dxB
1:13:1
From the rst and second laws of thermodynamics for this system (equation
(1.3.21)), one has
dGm Sm dT Vm dP mA dxA mB dxB
39
1:13:2
Therefore, in general, one may write

Sm dT Vm dP xA dmA xB dmB 0
1:13:3
and at constant temperature and pressure

xA dmA xB dmB 0
1:13:4
This is the form of the GibbsDuhem equation needed to relate the activity of
component B in solution to that of component A. Choosing the Raoult law
activity for the solvent A, and the Henry law activity for the solute B, equation
(1.13.4) may be rewritten as

xA
d ln aH
dln aR
B
A
xB
1:13:5
In the limit of very dilute solutions, both activity coecients approach unity so
that the activity of A can be replaced by xA , and the activity of B by its molality
(assuming that one has chosen molality as the concentration unit). However, as xB
becomes very small the ratio xA /xB becomes very large. Thus, in practice, one may
not choose the innitely dilute solution as a reference point but instead a very
dilute solution for which Henrys law is valid. Then, integrating equation (1.13.5)
between this very dilute concentration designated m1 , to any other concentration
m2 , one obtains
2
H
ln aH
B ln gB m2 ln m1
1
xA R
d ln aA
xB
1:13:6
An example of application of the GibbsDuhem relationship to determination

of the activity coecients of sucrose in aqueous solutions is shown in table 1.5.
Sucrose (C12 H22 O11 ) has a very high molecular mass (342 g) compared to water so
that solutions with very low solute mole fractions have relatively high molalities.
By measuring the water vapor pressure as a function of sucrose concentration
down to very low concentrations, one is able to determine its activity coecient
and thus its activity on the Raoult law scale. It should be noted that the vapor
pressure of water begins to depart from ideal behavior at a quite low sucrose mole
fraction (0.005). By carrying out the integration dened by equation (1.13.6) one
is able to calculate the Henrys law activity for sucrose and, thus, the activity
coecients given in table 1.5.
It should be noted that the values of gH quickly become non-unity and are
greater than one. This is indicative of strong attractive solutesolvent interactions
and negative deviations from Raoult law behavior. In the case of the methanol
water system for which positive deviations from Raoults law is observed (table
1.4), the Henry law activity coecients are less than one.
The above procedure for determining activity coecients for solid solutes is
often applied to electrolytes. This important class of solutes always behaves nonideally. The properties of electrolyte solutions are considered in detail in chapter 3.
40
Table 1.5 Activity Coefcient Data for Sucrose and

Water as Determined from Water Vapor Pressure
Measurements for Their Solutions at 25 C
Concentration of Sucrose
Activity Coecients
Water
Sucrose
Molality (Raoults Law) (Henrys Law)
gR
gH
Mole Fraction /mol kg1
w
s
1 104
1 103
5 103
0:02
0:04
0:07
0:1
5:6 103
0:056
0:279
1:134
2:315
4:182
6:173
1:000
1:000
0:9999
0:998
0:990
0:968
0:939
1:000
1:001
1:042
1:206
1:52
2:15
2:91
1.14 Concluding Remarks

The material in this chapter explains the relationship between the concentration of
a solution component and its activity. The activity is monitored through the vapor
pressure of the components, which are volatile for most of the examples considered. Thus, it is very easy to understand why the activity of a given component
can also be dened as its escaping tendency. It is obvious from the fact that most
solutions are non-ideal that the relationship between activity and concentration is
not simple. When the solution is very dilute, Henrys law holds for the solute and
Raoults law for the solvent. Then the activity is proportional to the concentration
over a nite concentration range which must be determined for each system.
A review of chemical thermodynamics, especially as it relates to the properties
of liquid solutions, has also been presented. Partial molar quantities such as the
chemical potential are an important feature of the treatment of this subject. It is
often the case that the activity and chemical potential of one quantity is relatively
easy to determine directly by experiment, whereas that of another component is
not. Under these circumstances, the change in chemical potential of one component can be related to that of another through the GibbsDuhem equation. This
relationship and its use in estimating thermodynamic properties are extremely
important in solution chemistry.
Two simple models for non-ideal solutions have been discussed. These are the
HildebrandGuggenheim model for regular solutions, and Wilsons empirical
approach to non-ideal solutions. They give the avor of the subject but represent
only a very small fraction of the theoretical work done in this area. The extension
of the model for regular solutions using the quasi-chemical approximation was
mentioned earlier. In this way, the approximation that the distribution of molecules is random in solution is relaxed. Another subject which has been examined
in detail in the literature is solutions with strong association. This association is
often due to hydrogen bonding and therefore is important in understanding the
properties of solutions formed with water and the alcohols. Polymer solutions
41
have also been considered and their thermodynamic properties derived. More
details regarding this subject and an introduction to the original literature in
this eld can be found in the monographs by Prigogine [9] and Marcus [10].
General References
G1. Hirata, M.; Ohe, S.; Nagahama, K. Computer Aided Data Book of Liquid-Vapor
Equilibria; Elsevier: Amsterdam, 1975.
G2. Levine, I. N. Physical Chemistry, 4th ed.; McGraw-Hill: New York, 1995.
G3. Landolt-Bornstein. Densities of Binary Aqueous Systems and Heat Capacities of Liquid
Systems; New Series, Group IV; Springer-Verlag: Berlin, 1977; Vol. 1b.
References
1. Hildebrand, J. H.; Scott, R. L. The Solubility of Non-Electrolytes, 3rd ed.; Dover: New
York, 1964.
2. Treiner, C.; Tzias, P.; Chemla, M.; Poltoratsky, G. M. J. Chem. Soc., Faraday Trans. I
1976, 72, 2007.
3. Hildebrand, J. H. J. Am. Chem. Soc. 1929, 51, 66.
4. Fowler, R. H.; Guggenheim, E. A. Statistical Thermodynamics; Cambridge Press,
1939; Chapter 8.
5. Guggenheim, E. A. Mixtures; Oxford University Press 1952; Chapter 4.
6. Rushbrooke, G. A. Introduction to Statistical Mechanics; Oxford University Press, 1949.
7. Wilson, G. M. J. Am. Chem. Soc., 1964, 86, 127.
8. Renon, H.; Prausnitz, J. M. AIChE J. 1968, 14, 135.
9. Prigogine, I.; Bellemans, A.; Mathot, V. The Molecular Theory of Solutions; North
Holland: Amsterdam, 1957.
10. Marcus, Y. Introduction to Liquid State Chemistry; Wiley-Interscience: New York, 1977.
Problems
1. The density of LiClO4 solutions in dimethylsulfoxide at 25 C is given by
r 1.0965 6.204 102 c
where r is the density in g mL1 and c, the concentration of LiC1O4 in M.
Calculate the molality and mole fraction of the solute for solutions with
concentrations of 0.1, 0.2, and 0.5 M.
2. The following density data are reported for the carbon tetrachlorideacetonitrile system at 25 C.
Mole Fraction
Acetonitrile
Density
/g mL1
0
0.1657
0.3145
0.4406
0.5498
0.6473
0.7348
0.8113
1.5844
1.5066
1.4248
1.3441
1.2640
1.1827
1.1003
1.0194
42
0.8797
0.9430
1.0000
0.9390
0.8575
0.7766
Calculate the excess molar volume for the solution and partial molar volumes
of the two components at mole fractions equal to 0, 0.2, 0.4, 0.6, 0.8, and 1.0.
Use the appropriate numerical interpolation and dierentiation techniques
(see appendix C).
3 The following results are reported for the carbon tetrachlorideacetonitrile
system at 45 C.
Mole Fraction CCl4
Total Vapor Presure
Liquid
Vapor
/kPa
0
0.0347
0.1914
0.3752
0.4790
0.6049
0.8069
0.9609
1.000
0
0.1801
0.4603
0.5429
0.5684
0.5936
0.6470
0.8001
1.000
27.783
33.062
44.796
48.604
49.274
49.466
48.362
41.908
34.501
Use an interpolation method to obtain the values of the mole fraction of CCl4
in the vapor and the total vapor pressure for values of the mole fraction in the
liquid phase equal to 0.2, 0.4, 0.6, and 0.8. Then calculate the vapor pressure
and the activity coecient of each component for the same values. Finally
estimate the molar Gibbs energy of mixing on the Raoult law scale at these
four points.
4. The following data are reported for the molar enthalpy of mixing for the
system discussed in question 3.
Mole Fraction of CCl4
Liquid
mix Hm
/J mol1
0
0.128
0.317
0.407
0.419
0.631
0.821
1.0
0
414
745
862
858
930
736
0
Determine the value of mix Hm at xCCl4 equal to 0.2, 0.4, 0.6, and 0.8.
Combine these data with those obtained in question 3 to prepare a plot of
mix Gm , mix Hm , and Tmix Sm against xCCl4 .
43
5. Calculate the excess Gibbs energy, entropy, and enthalpy of mixing for the
carbon tetrachlorideacetonitrile system discussed in questions 3 and 4.
Prepare a plot of these data and compare the results with those obtained in
the previous question.
6. The following data are available for solutions of acetone and chloroform at
50 C.
Mole Fraction of Acetone
Total Pressure
Liquid
Vapor
/kPa
0
0.10
0.20
0.30
0.38
0.40
0.50
0.60
0.70
0.80
0.90
1.00
0
0.071
1.165
0.279
0.380
0.408
0.550
0.684
0.789
0.890
0.955
1.000
69.5
66.0
63.2
61.7
61.1
61.3
62.5
65.2
68.1
72.0
76.8
81.6
Calculate the Raoult law activity coecients of both components and plot
them as a function of the mole fraction of acetone. Determine the range of
composition with respect to acetone that the solution can be regarded as
regular. Calculate the enthalpy parameter for acetonechloroform interactions on the basis of a one-parameter least-squares t of the data in this
range using an appropriate plot.
7. J. J. Van Laar gave a useful semiempirical equation for the molar excess
Gibbs energy of solutions,
mix Gex
m
b12 x1 x2
b1 x1 b2 x2
where b12 , b1 , and b2 are characteristic constants. Show that the Van Laar
relation implies that
p

A12 x1
1 1=2
1
p
ln g1
B12 x2
A12
p

1 1=2
1
B12 x2
p
ln g2
A12 x1
B12
where A12 b12 =b2 RT and B12 b12 =b; RT
44
8. The following vapor pressure data are found for methanolwater solutions at
40 C.
xMeOH
PMeOH =kPa
0.04
0.08
0.12
0.2
0.4
0.6
0.8
1.0
2.1
4.3
6.4
9.9
17.2
23.3
29.3
36.0
Calculate the Raoult law and Henry law activity coecients for methanol on
the mole fraction scale.
The Structure of Liquids
Douglas (Doug) Henderson was born in

Calgary, Alberta, Canada in 1934. He grew up
in Vancouver, British Columbia and attended
the University of British Columbia where he
obtained a bachelors degree in mathematics in
1956. Henderson then went to the University of
Utah in Salt Lake City to work with Henry
Eyring on the theory of liquids. He obtained
his Ph.D. in physics under Eyrings direction in
1961. After leaving graduate school he held
faculty positions at universities in both the
Douglas James Henderson
United States and Canada. During 196667,
Henderson was an invited scientist at the Commonwealth Scientic and
Industrial Research Organization (CSIRO) in Melbourne, Australia. There he
began a long and very successful collaboration with John Barker. One of the
major results of their work was the perturbation theory of liquids, some of
which was outlined in Reviews of Modern Physics in 1976 in their well-known
paper What is liquid? Understanding the states of matter [1]. Barker and
Henderson moved together to the IBM Almaden Research Center in San Jose,
California as research scientists in 1969. Hendersons research interests in the
statistical mechanics of condensed phases involve many aspects, an important
one in recent years being the electrical double layer. He was instrumental in
bringing our understanding of interfacial phenomena at polarizable interfaces
beyond the primitive level. He is currently professor of chemistry at Brigham
Young University in Provo, Utah. His career is noteworthy for the
international collaborations which he has established in many countries where
research in statistical mechanics is carried out. He has published over 400
papers in this area. In addition, he has won numerous awards for his scientic
work, the most recent being the Hildebrand Award of the American Chemical
Society in 1999 for his work on the structure of liquids.
2.1 What Is a Liquid?

It is well known from studies of the properties of matter that the liquid state is
much more complex than either the gaseous or solid states. Studies of the properties of gases quickly lead to the ideal gas law, which describes the properties of
real gases at low pressures and high temperatures. This success is clearly due to
45
46
the fact that the molecules in a dilute gas are far from one another so that the
eects of intermolecular forces and of the nite volume occupied by the gas
molecules are negligible. As the pressure of a gas is increased and its temperature
lowered, the eects of non-ideality become apparent, and the equation of state
becomes more complex. These changes are those required to convert the gas to a
liquid. As the molecules come closer together, the inuence of intermolecular
forces becomes greater and the free volume available for the gas molecules is
signicantly reduced because of the space occupied by the molecules themselves.
The statistical mechanical description of a gas relies upon the concept that the
molecules are in constant movement with trajectories determined by collisions
with the walls of the container and with other molecules. The probability of
nding another molecule in the immediate vicinity of a given molecule is extremely low and does not vary signicantly with distance from the reference
molecule.
On the other hand, solids are characterized by a very ordered structure in
which each ion or molecule is surrounded by a xed number of neighbors
whose nature and orientation are determined by the interparticle forces in the
crystal. These may be chiey ionion interactions, as in an ionic crystal, or intermolecular forces, as in a molecular crystal. Because of the high state of order in
crystals it is a reasonably straightforward problem to calculate their thermodynamic properties on the basis of quite simple statistical mechanical models.
One way of conceptualizing a liquid is as a very disordered solid. If one disrupts the structure of the nearest neighbors around a reference molecule in a
molecular crystal, the eect of the disruption extends quite far. As a result,
there is some local order around the reference point but the extent of order
falls o rapidly with distance so that at distances equivalent to four or ve
molecular diameters the system does not possess order with respect to the reference point. Theories of the liquid state based on an approach involving disordered
solids were pursued from the 1930s to the 1960s but did not meet with much
success. On the other hand, the liquid may be regarded as an extremely imperfect
gas. In this approach, which has been quite successful, the statistical mechanical
techniques used to describe the properties of non-ideal gases are extended to
liquids.
Considerable advances have been made since the 1950s in developing the
theory of liquids [G1, G2]. The purpose of this chapter is to give an introduction
to this subject, outlining the main theoretical and experimental topics. Rigorous
development of the theory is not possible without appealing to more sophisticated
mathematics. However, an understanding of the basic concepts involved in this
subject is helpful, not only in reading further in the area of liquid structure, but
also in developing the other topics considered in this monograph.
Liquids may be classied according to the intermolecular (or interatomic)
forces existing between the components. In this way, seven dierent kinds of
liquids may be identied:
1. The simplest liquids are those formed by the inert gases, He, Ne, Ar, etc.
These atoms interact via van der Waals forces and strongly repel each other
at short interatomic distances.
THE STRUCTURE OF LIQUIDS
47
2. Homonuclear diatomic molecules such as H2, N2, and Cl2 form the second
group. They are similar to the rst in that they do not possess a dipole
moment but have electrical quadrupole moments and are not spherical.
3. Liquid metals such as Hg make up the third group. Because of the mobility of
electrons in metals, these systems have long-range coulombic forces, and are
softer with respect to short-range repulsive forces. In addition, electrical
screening eects are important in liquid metals.
4. Molten salts are systems in which the components are ions but which are
electrically neutral on a local scale. Coulombic forces are long range in molten salts and electrical screening is important. The complexity of these systems depends on the nature of the ions, that is, whether they are monoatomic
or polyatomic. For polyatomic ions, other electrostatic forces may be
involved.
5. Aprotic polar liquids such as dimethyl sulfoxide and acetonitrile make up
another group. These molecules have high dipole moments, so that dipole
dipole interactions are an important part of the description of intermolecular
forces.
6. Another group is composed of protic polar liquids such as water and the
alcohols. In these systems, hydrogen bonding adds a further complication to
the description of intermolecular forces.
7. Finally, non-polar molecular liquids, such as carbon tetrachloride and the
hydrocarbons, form a group. Many of these systems possess no permanent
dipole moment so that the intermolecular forces are similar to those existing
in simple atomic liquids such as liquid argon. However, internal modes of
motion are important in describing the properties of the molecular liquid.
Chemists are obviously concerned mainly with liquids in the last three groups.
However, they are the most dicult to model from the point of view of theory.
Much of the theoretical eort has been directed to understanding the properties of
the simplest liquids, namely, the inert gases. In the following sections, the statistical mechanical approach developed to understand liquid properties is outlined.
The purpose of this subject is to establish a connection between the properties of
the individual atoms or molecules in the liquid and the bulk properties of the
system. An important part of this development is the experimental study of liquid
structure which is also outlined in the following discussion.
2.2 The Statistical Thermodynamics of Liquids

Statistical thermodynamics uses statistical arguments to develop a connection
between the properties of individual molecules in a system and its bulk thermodynamic properties. For instance, consider a mole of water molecules at 25 C and
standard pressure (1 bar). The thermodynamic state of the system has been
dened on the basis of the number of molecules, the temperature, and the pressure. In order to relate the macroscopic thermodynamic properties such as U, G,
H and A to the properties of the individual molecules, one would have to solve the
Schrodinger wave equation (SWE) for a system composed of 6 1023 interacting
water molecules. This is an impossible task at present but if it were possible, one
would obtain a wave function, j , and an energy, j , for the system. Moreover,
48
the values of j and j would uctuate with time because of the changing positions of the molecules and their interactions. The task of statistical thermodynamics is to relate the time average properties of the many microstates which arise
in the uctuating system to its macrostate. In the following section the basic
principles of statistical thermodynamics are outlined, rst for non-interacting
systems, such as ideal gases, and then for interacting systems, that is, non-ideal
gases and liquids.
The present discussion concerns the canonical ensemble. An ensemble is a
hypothetical collection of an large number of non-interacting systems, each of
which is in the same thermodynamic state (macrostate) but in a dierent microstate. A fundamental postulate of statistical thermodynamics is that the measured
time average of a macroscopic property of the system is equal to the average value of
that property in the ensemble. Thus, time averaging can be replaced by averaging
over the systems in the ensemble. The fact that the ensemble is canonical means
that each system in the ensemble has the same volume, temperature, and composition (number of molecules). This means that each system is rigid (constant
volume), and has impermeable walls (constant composition), but walls which
are thermally conductive. All of the systems can be imagined as being immersed
in constant temperature bath.
The internal energy of the macrosystem, U, is found by averaging the energy j
for each microsystem in the ensemble. The averaging is carried out over all
possible quantum states of the microsystem, taking into account the probability
that the total energy is equal to j . A second postulate used in this process is that
for a thermodynamic system of xed volume, composition, and temperature, all
quantum states that have equal energy have equal probability of occurring.
Estimation of the probability for a given quantum state leads to the denition
of the canonical partition function of the macrosystem:
X
ebj
2:2:1
Q
j
where b 1=kB T and kB is the Boltzmann constant. The partition function is of

central importance in determining the thermodynamic properties of the macrosystem given its microscopic properties. Sometimes the summation is dened in terms
of available energy levels rather than quantum states. In this case, each term in the
summation must be multiplied by the degeneracy of the corresponding energy level.
On the basis of the denition of the partition function, the probability that the
microsystem is in quantum state j is given by
fj
ebj
Q
Since the internal energy, U, is given by

X
U
fj j
it is easily shown that
U

@ ln Q
@ ln Q
kB T2
@b V;N
@T V;N
2:2:2
2:2:3
2:2:4
49
In the same way, one may estimate the systems pressure P by noting that
X
P
fj Pj
2:2:5
where Pj is the pressure for the microsystem in microstate j. It follows that

1 @ ln Q
P
2:2:6
b @V T;N
An expression for the entropy S of the ensemble can be found by relating the
temperature derivative of the entropy to the heat capacity at constant volume
using equation (1.3.9). Accordingly,

@U
@S
CV
T
2:2:7
@T V;N
@T V;N
Since
CV
it follows that

@
@ ln Q
kB T2
@T
@T V;N
2:2:8
!

@S
@ ln Q
@2 ln Q
2kB
k T
@T V
@T V;N B
@T2
2:2:9
V;N
The corresponding dierential equation to obtain the entropy is

!

@ ln Q
@2 ln Q
dS 2kB
dT kB T
dT
@T V;N
@T2
V;N

@ ln Q
@ ln Q
kB
dT kB d T
@T V;N
@T V;N
2:2:10
Integrating with respect to temperature from 0 K, one obtains the expression

S S0 kB ln Q kB ln Q0 kB T
@ ln Q
@T
2:2:11
where S0 and Q0 are the entropy and partition function at 0 K. The partition
function Q0 is simply the degeneracy of the lowest available energy state of the
system and is a temperature-independent quantity. Equating S0 and kB ln Q0 , the
resulting expression for the entropy is
S
U
kB ln Q
T
2:2:12
On the basis of these denitions, one easily obtains expressions for the Helmholtz
energy
A U TS kB T ln Q
and for the Gibbs energy
2:2:13
50

@ ln Q
G U TS PV kB T ln Q kB TV
@V T;N
2:2:14
The thermodynamic functions are introduced here for a one-component system,

but are easily extended to multicomponent systems [G3, G4].
Two other ensembles are commonly dened in statistical thermodynamics,
namely, the microcanonical ensemble and the grand canonical ensemble. In the
microcanonical ensemble, the number of moles, volume, and internal energy, U,
are xed in each system. This means that each system is totally isolated, being
unable to exchange either energy or matter with the surroundings. In the case of
the grand canonical ensemble, the volume, temperature, and chemical potential of
the single component in each system are held constant. Thus, for such an ensemble, both energy and matter can be exchanged with the surroundings. Further
details regarding these ensembles and the relationships between the three types
introduced can be found elsewhere [G3, G4].
The treatment of the canonical ensemble for the special case of an ideal gas is
now outlined. First of all the partition function for the macrosystem is related to
the properties of the individual molecules. At the same time, the energy of a given
state in the ensemble j can be related to the energy of an individual molecule i .
In the case of a gas or a liquid the particles in the system are indistinguishable
because of random motion. Furthermore, if the particles do not interact with one
another as in an ideal gas, it is easily shown that
qN
2:2:15
N!
where q is the partition function for a given molecule, which is given by
X b
q
e i
2:2:16
Q
The summation for q is made over the available quantum states for the molecule
as determined by solution of the SWE.
In order to proceed further one must write down the Hamiltonian for the
molecule, that is, the expression for its kinetic and potential energy. The former
contains up to ve contributions, namely, for its translational, rotational, vibrational, electronic, and nuclear energies. Since the molecules do not interact with
one another the potential energy is zero. Thus, one may write
H Htrans Hrot Hvib Helec Hnuc
2:2:17
For present purpose this is more conveniently written as

H Htrans Hint
2:2:18
where Hint represents all of the internal modes of possessing energy. With the
Hamiltonian operator H one obtains the SWE whose solution gives the energy of
an individual molecule, :
trans int
2:2:19
where trans and int are the contributions of the translational and internal modes
to the energy, respectively. These may then be used to estimate the molecular
partition function, q:
q qtrans qint
51
2:2:20
Expressions for the individual partition functions have also been developed.
From the solution of the SWE for a particle in a box it is well known that

2pmkB T 3=2
V
qtrans
V 3
2:2:21

h2
where m is the mass of the particle, V, the volume of the box and h, the Planck
constant and

h
2pmkB T1=2
2:2:22
The expression for the canonical partition function of an ideal gas thus becomes
Q
N
qN
qN V N
trans Qint
int 3N
N!
N!
2:2:23
Another way of writing the molecular partition function is to use the classical
expression for Htrans in an integral form. By classical mechanics, Htrans for each
molecule is given by
1
P2 P2y P2z
2:2:24
Htrans
2m x
where Pi is the linear momentum in the ith direction. Then,
1
qtrans 3
expbHtrans dPdr
2:2:25
h
where dP dPx dPy dPz and dr dxdydz. The factor 1/h3 appears in this expression in order to keep the classical result the same as that from the quantummechanical summation. Integration with respect to the volume element dr yields
the volume V of the container. Substituting equation (2.2.24) into equation
(2.2.25), one may write
21
33
V4
2pmkB T3=2
qtrans 3
expbP2 =2mdp5
V
2:2:26
h
h3
1
Thus, the classical expression for qtrans yields the same result as that obtained
from the SWE (equation (2.2.21)).
It is useful to keep the classical way of expressing the partition function and
extend its application to more complex situations. First of all, one may write an
expression for the canonical partition function of an ideal gas in terms of the
partition function for each molecule. On the basis of the total Hamiltonian for
each molecule (equation (2.2.18)), Hj , the canonical partition function is

N
1
expbHj dPj drj
Q
2:2:27
N!h3N
Each integral within the brackets is sixfold with three directional and three
momentum coordinates involved in the integration process. Alternatively, the
partition function may be expressed as
52
1
.
.
.
expbHdP1 . . . dPN dr1 . . . drN
N!h3N
where
H
Hj
2:2:28
2:2:29
is the Hamiltonian of the entire system. In the notation used, particle 1 at position
r1 (a vector) is located in volume element dr1, and so on. Now the partition
function is expressed in terms of one 6N-fold integral.
The above treatment of an ideal gas is easily extended to a non-ideal system, at
least, in a conceptual fashion. In the non-ideal system the molecules interact with
each other so that the system has a potential energy, U, which depends on the
coordinates of each molecule in the system. In the real system in which molecular
positions change with time, the potential energy also uctuates. This is reected in
the canonical ensemble by variation in the potential energy with the molecular
conguration in a given microstate. The Hamiltonian for such a system containing N molecules is
H
N
N
X
1 X
P2xj P2yj P2zj
Hintj Ur1 ; r2 ; . . . ; rN
2m j1
j1
2:2:30
Performing the integration in equation (2.2.27), one obtains the following expression for the canonical partition function Q:
Q
qN
int
Z
N!3N
where Z is the congurational partition function, that is,
Z . . . expbUdr1 dr2 . . . drN
2:2:31
2:2:32
It should be noted that it is assumed that the intermolecular forces do not aect
the internal degrees of freedom so that qint is independent of whether these forces
are present or not. When they are absent (U 0), the integral Z collapses to VN
and equation (2.2.31) becomes the same as equation (2.2.23). The important task
of the statistical thermodynamics of imperfect gases and liquids is to evaluate Z.
This subject is discussed in detail later in this chapter. However, the nature of the
intermolecular forces which give rise to the potential energy U is considered next.
2.3 Intermolecular Forces

Molecules exhibit relatively long-range attractive forces between themselves which
give rise to the cohesive forces in liquids. These forces arise because the electronic
distribution in the molecule or atom making up the liquid is not uniform either on
a time-averaged basis or with respect to its instantaneous value. Non-uniformity
in the time-averaged electronic distribution in a molecule is a well-known
phenomenon, and is discussed in terms of the experimentally measured dipole
53
moment, and higher-order moments such as the quadrupole and octapole

moments. The major attractive component of intermolecular forces is due to
dipoledipole interactions, of which there are three principle contributions,
namely, that due to permanent dipole interactions, that due to induced dipole
permanent dipole interactions, and the dispersion contribution arising from interactions between the instantaneous dipole moments. These are collectively called
van der Waals forces and give rise to an attractive potential between two molecules which is inversely proportional to the distance between them to the sixth
power. The nature of van der Waals forces is now examined in more detail.
van der Waals forces are usually treated using the physics of ideal dipoles. A
real dipole consists of equal and opposite charges separated by a nite distance d.
This concept is easily applied to molecules where dierences in atomic electronegativities leads to polar molecules in which the equal and opposite charges are
separated by distances the order of a typical bond length. A simple example is
carbon monoxide, which has a dipole moment of 0.12 debye because of the
polarity associated with the CO bond. An ideal dipole is one in which the nite
distance separating the charges shrinks to zero but the charge separation is maintained. Thus, the ideal dipole is a point with a vector character associated with the
charge separation. The dipole moment is measured in coulomb meters in SI units
and in debyes in traditional units.* The dipole vector points along the axis of the
dipole from the negative to the positive charge. Because of the vector character of
the dipole moment, the electrostatic equations describing its properties are much
more complex than those for a point charge.
The potential due to an ideal dipole p at a distance r from its center is
p
pr
p cos y
3
4pe0 r
4pe0 r2
2:3:1
where y is the angle between the dipole vector and the line from the dipole to the
point where the potential is measured, and "0, the permittivity of free space. The
maximum potential is obtained in the direction of the dipole vector where cos y is
equal to one. On the other hand the potential perpendicular to the dipole is zero
(cos 90 0). The corresponding equation for a point charge q is
q
2:3:2
q
4pe0 r
In this case the potential is independent of the direction from the charge at which
it is measured. The dipole potential in a given direction falls o as 1=r2 , whereas
that due to a point charge, as 1=r.
The eld due to the ideal dipole is found by taking the gradient of the potential.
Thus,
*All electrical equations in this monograph are written in SI units. As a result, the dipole moment
must be given in C m, not in its usual units of debyes (1 debye 3.336 1030 C m). In addition, the
molecular polarizability, which is usually tabulated in m3, has units of C2 m N1 in the SI system
a 0 cgs aSI=4pe0 . Thus, a polarizability of 1030 m3 in traditional units is equivalent to 1.113
1040 C m2 V1 in SI units. Although the equations are written for SI units, dipole moments and
polarizabilities are recorded in the tables in this monograph in the traditional units of debyes and m3.
54
Ep gradq
3p r
p
r
4pe0 r5
4pe0 r3
2:3:3
The rst term on the right-hand side gives the component of the eld in the
direction r dened by the vector from the ideal dipole to the point of measurement. The second term gives the component in the direction dened by the dipole.
The net eld, Ep , which is also a vector, points in a direction between p and r
dened by their vector sum. When the eld is measured along the direction of the
ideal dipole, the expression simplies to
Ep y 0
p
2pe0 r3
2:3:4
Now one may estimate the energy of two dipoles on the basis of the work done to
bring dipole p2 from innity to a point located at distance r from dipole p1. This is
u12 p2 E
3p1 r
p p
p r 1 2 3
4pe0 r5 2
4pe0 r
2:3:5
In polar coordinates, equation (2.3.5) becomes

u12 r
p1 p2
2 cos y1 cos y2 sin y1 sin y2 cosf1 f2
4pe0 r3
2:3:6
where yi is the angle that dipole i makes with the z-axis, chosen as the direction
dened by the line between the dipoles, and fi is the angle that dipole i makes
with the x-axis (see g. 2.1). There are several important congurations that are
among all possible congurations of two dipoles. The rst corresponds to the case
that the two dipoles point along the z-axis with the head of one dipole pointing to
the tail of the other (g. 2.1). In this case
y1 y 2 0
2:3:7
and
u12 r
2p1 p2
4pe0 r3
2:3:8
Fig. 2.1 Schematic diagram showing two

dipoles pa and pb separated by a distance r
along the z-axis with orientations dened by
the angles ya, fa and yb, fb. The four simple
congurations for two dipoles and their
interaction energies are also illustrated with
ur pa pb =4pe0 r3 :
55
On the basis of equation (2.3.6), this is the conguration of minimum energy. The
conguration of maximum energy occurs when the two dipoles are oriented with
their heads or tails pointing to one another. This occurs when
y1 0
y2 1808
and
2:3:9
and gives an energy

u12 r
2p1 p2
4pe0 r3
2:3:10
Two other congurations are important in assessing dipoledipole interactions.

The rst occurs when the two dipoles are parallel to each other but point in
opposite directions. This is obtained when
y1 908;
y2 2708;
f1 f2
and
2:3:11
with the result that

u12 r
p1 p2
4pe0 r3
2:3:12
Finally, when the two dipoles are parallel and point in the same direction
y1 908;
y2 908;
and
f1 f2
2:3:13
so that
u12 r
p1 p2
4pe0 r3
2:3:14
The important result from this analysis is that the dipoledipole interaction
energy depends on the product of the two dipole moments and decreases with
distance as 1=r3 .
In order to estimate the contribution of interactions between permanent
dipoles to intermolecular forces, one must consider all possible relative orientations at a given separation distance r. The interaction energy is then averaged,
taking into account the probability of a given orientation on the basis of a
Boltzmann factor. In this way, orientations with attractive interaction are
weighted in favor of those resulting in repulsive interaction. The expression giving
the net average interaction energy between two dipoles is
u r expu12 r=kB Tdo1 do2

12
2:3:15
udd r
expu12 r=kB Tdo1 do2
where
doi sin yi dydf
2:3:16
Integration over all possible values of y and f for the case that u12 r is small with
respect to kB T gives the following result for the average dipoledipole interaction
energy:
udd r
2p21 p22
34pe0 2 kB Tr6
2:3:17
56
In the averaging process, the factor p1 p2 =4pe0 r3 ; which appears in the expression
for the energy of a specic orientation, is squared, so that the average attractive
interaction depends on 1/r6. Thus it is important to distinguish between the interaction energy of two isolated dipoles, xed with respect to their relative orientations, which depends on the distance separating them as 1=r3 , and the interaction
energy between the same two dipoles, when it is averaged over all possible mutual
orientations, which depends on 1=r6 .
As an example, consider the case of two water molecules in contact, each with a
dipole moment equal to 1.85 debyes (6.17 1030 C m) and represented as
spheres with diameters of 274 pm. The attractive energy between these molecules
when they are oriented in the conguration of maximum attraction is 3:25
1020 J, that is, 7:9kB T at room temperature. On the other hand, when the
attractive energy is estimated using Boltzmann averaging, it is 4:30 1020 J
or 10:4kB T. The Boltzmann-averaged value is more than 50% of the maximum
attractive energy. It follows that dipoledipole interactions represent a major
component of the cohesive energy which leads to formation of the liquid state
in water.
A second component of van der Waals forces is that resulting from the induced
dipole moment formed in one molecule in the presence of another molecule
possessing a permanent dipole moment. If a1 and a2 are the polarizabilities of
the two molecules, the potential energy due to dipoleinduced dipole interactions
is
uid r
p21 a2 p22 a1
4pe0 2 r6
2:3:18
This is derived by considering the interaction of two polarizable dipoles in a

general orientation and then averaging over all possible orientations. Assuming
a molecular polarizability equal to 1.46 1030 m3 (1.62 1040 C m2 V1) for
water, the value of uid r for water molecules in contact is 2:3 1021 J, or
0:6kB T at room temperature. As expected, this contribution is much smaller
than that due to the interaction of the permanent dipole moments.
The third contribution to the attractive forces comes from London dispersion
forces. These arise from the instantaneous dipoles on the molecules formed by the
moving electrons. Thus, inert gas atoms such as helium and argon possess instantaneous dipole moments formed by the electron cloud, which is constantly in
motion. Dispersion forces were analyzed using quantum mechanics by London
[2] who derived the following expression for the attractive potential:
udis
3I1 I2 a1 a2
2I1 I2 4pe0 2 r6
2:3:19
where I1 is the ionization potential of molecule 1, and I2 , that of molecule 2. For

two water molecules in contact, the dispersion energy is 1:85kB T at room temperature given that the ionization potential of a water molecule is 12.6 eV or 2.02
1018 J molec1 .
57
EXAMPLE
Estimate the contributions to the van der Waals energy in dimethylsulfoxide at

25 C by considering two molecules in contact as hard spheres with diameters of
491 pm. The dipole moment of dimethyl sulfoxide is 3.96 debyes, its polarizability, 7.99 103 nm3, and its ionization potential, 9.01 eV.
The dipoledipole attractive potential energy for two identical spherical
molecules in contact is
udd
2p4
34pe0 2 kB Ts6
2:3:20
where s is the molecular diameter. For dimethyl sulfoxide,

udd
23:96 3:336 1930 4

31:113 1010 2 1:38 1023 298:2 4:91 1010 6
2:3:21
2:84 1020 J molec1 ; or 6:91kB T

Similarly, for the dipoleinduced dipole interactions,
uid
2p2 a
4pe0 2 s6
23:96 3:336 1030 2 7:99 1030 1:113 1010
1:113 1010 2 4:91 1010 6
2:3:22
1:79 1021 J molec1 ; or 0:43kB T

For the London dispersion energy,
udis
3Ia2
44pe0 2 s6
39:01 1:602 1019 7:99 1030 2
4 4:91 1010 6
2:3:23
4:93 1021 J molec1 ; or 1:20kB T

Note that the calculations of uid and udis in which the polarizability is involved
are simpler when the SI polarizability a is replaced by a 0 =4pe0 , which has
units of m3. Then equations (2.3.22) and (2.3.23) may be written as
uid
2p2 a 0
4pe0 s6
2:3:24
udis
3Ia 02
4s6
2:3:25
and
Values of the dipoledipole, dipoleinduced dipole, and London dispersion

contributions for two molecules in contact are summarized for a range of liquids
58
Table 2.1 Estimation of van der Waals Interaction Energies for Two Molecules at
Contact Using a Spherical Representation of the Molecules
Atom or
Molecule
Argon
Nitrogen
Hexane
Benzene
Water
Acetone
Acetonitrile
Dimethyl sulfoxide
Molecular Dipole
Molecular Ionization van der Waals Energy*
Diameter Moment Polarizability Potential
s=pm
p=debye
a=nm3
I=eV
udd =kB T uid kB T udis =kB T
370
400
592
563
274
476
427
491
0
0
0.09
0
1.82
2.88
3.92
3.96
1:63 103
1:76 103
0:0118
0:0103
1:46 103
6:41 103
4:41 103
7:99 103
15.8
15.6
10.1
9.25
12.6
9.71
12.2
9.01
0.0
0.0
0.0
0.0
10.38
2.33
15.34
6.91
0.0
0.0
0.0
0.0
0.55
0.22
0.54
0.43
0.48
0.34
0.95
0.90
1.85
1.00
1.14
1.20
Estimated at 25 C.
in table 2.1. It should be kept in mind that equations (2.3.17)(2.3.19) are generally not applicable at such short distances, since they were derived on the basis
of point dipole models. However, they give a good idea of the relative magnitude
of the three components and how they change when one goes from an inert gas
like argon to very polar molecules like water and dimethyl sulfoxide. It is also
easily seen from these equations that the intermolecular potential energy falls o
rapidly with increase in distance between the molecules. Thus, if the separation is
two molecular diameters, the value of the potential energy drops by a factor of
1/26, that is, 1/64; for three molecular diameters this decreases to 1/36, or 1/729,
and so on.
The above simple picture of the attractive component of intermolecular forces
is certainly not complete. Since a complete electrostatic description of a molecule
usually requires the introduction of higher-order moments, one must also consider the potential energy due to dipolequadrupole interactions, quadrupole
quadrupole interactions, and interactions of moments of higher order. These
lead to terms proportional to r8 , r10 , and so on. Obviously, the potential energy
due to these components falls o much more rapidly than those due to dipole
dipole interactions, and therefore, is less important. Another factor not considered in the above discussion is the role of hydrogen bonding, which is very
important for protic molecules such as water and the alcohols. It is well recognized that hydrogen atoms bonded to more electronegative atoms such as oxygen
and nitrogen can form a bond with another electron-rich atom in an adjacent
molecule. The best-known example of hydrogen bonding is that in water, where
the hydrogen atom on one molecule bonds with the oxygen atom on an adjacent
molecule. Other well-known examples are the alcohols, certain amides like formamide and acetamide, and carboxylic acids. In order to estimate the strength of
hydrogen bonding, one must develop a quantum-mechanical model of the system.
This is a complex topic and is not considered further here. It is sucient to point
out that hydrogen bonding accounts for an important fraction of the cohesive
energy in liquids when it is present.
59
Hydrogen bonding is an example of a relatively short-range attractive force

between specic types of molecules. In general, one must consider another shortrange force which is experienced by all atoms and molecules and which is repulsive. This repulsive interaction arises when the electronic clouds in two molecules
are so close that they repel one another. It may be treated quantum mechanically
in the adiabatic approximation in which the nuclei in the molecules are assumed
to be motionless, and the wave functions for the electrons calculated. An approximate form for the repulsive energy between two molecules is
urep AR ear
2:3:26
where AR and a are constants depending on the properties of the individual

molecules. This is more conveniently approximated by a function of the twelfth
power of the intermolecular distance, r, so that equation (2.3.26) becomes
urep
AR
r12
2:3:27
When one compares this term with the attractive van der Waals forces considered
above it is clear that the repulsive energy falls o much more rapidly with distance
than the attractive one.
When one adds the attractive van der Waals potential terms to the repulsive
term, one obtains the LennardJones expression for the intermolecular potential
energy for a simple uid such as an inert gas like argon. On the basis of the above,
the LennardJones potential function may be written
ur
ALJ BLJ
6
r12
r
2:3:28
where the second term represents the attractive van der Waals component.
Obviously, there is a minimum potential energy, LJ , corresponding to the optimum distance between any two molecules. In addition, one may identify the
molecular diameter s as corresponding to the distance at which the intermolecular potential energy is zero. As a result, the constants ALJ and BLJ are
connected by the equation
ALJ =BLJ s6
2:3:29
dur
ALJ
B
12 13
6 LJ
0
dr
r7e
re
2:3:30
At the optimum separation re
so that
r6e
2ALJ
BLJ
2:3:31
It follows that at the minimum

LJ
ALJ BLJ
B2
B2
B2
6 LJ LJ LJ
12
4ALJ 2ALJ
4ALJ
re
re
2:3:32
Solving equations (2.3.29) and (2.3.32) for the constants ALJ and BLJ , one obtains
60
ALJ 4s12 LJ
2:3:33
BLJ 4s6 LJ
2:3:34
and
It follows that the LennardJones equation is

s 12
s6
ur 4LJ

r
r
2:3:35
It is easily seen that the function r12 was chosen to replace ear for the repulsive
component in order that a simple mathematical result could be obtained.
EXAMPLE
Estimate the LennardJones energy of molecular nitrogen at the optimum

separation re and at twice this value. Use the parameters given in table 2.2.
From the table,
LJ 95:2 1:381 1023 1:245 1021 J molec1
2:3:36
When r is equal to re , ur is equal to LJ . The relationship between re and s is

r6e 2s6
2:3:37
so that re is 420.4 pm, given that s is 374.5 pm.

When r is equal to 2re
"
#

s12
s6
1
1
ur 4LJ

LJ 12 5
2
2
221=6 s12 221=6 s6
2:3:38
0:031 1:245 1021 3:86 1023 J molec1

Thus, the LennardJones energy at 2re is 3:86 1023 J molec1 :
A plot of the LennardJones potential against distance between two molecules
is shown in g. 2.2. It is clear that the repulsive component dominates for values
of r less than the molecular diameter, where it rises to values over 5LJ for
r 0:9s. The stable minimum occurs at r 1:12s. By the time the separation between the two molecules is equal to three times their diameter, the intermolecular
Table 2.2 LennardJones

Parameters for Simple Molecules
Atom or Molecule
Ar
Xe
N2
CO2
s=pm
LJ k1
B =K
350
410
374.5
433
117.7
222.3
95.2
198.2
*kB is the Boltzmann constant, so that LJ k1

B
has units of temperature.
61
Fig. 2.2 Plot of the

LennardJones potential
ur in units of the attractive
potential energy LJ against
intermolecular distance, r, in
units of the molecular
diameter s. The vertical and
horizontal straight lines at r
s show the potential
energy for a hard-sphere
representation of the system.
potential is negligible. This potential energy function has often been used in the
development of the theory of simple liquids such as the noble gases at low temperature. For many liquids that chemists use, a much more complex expression for the
intermolecular potential is required. However, the properties of a Lennard
Jones liquid are reasonably easily derived and form a convenient point of reference with respect to which the properties of more complex systems can be
considered.
2.4 Distribution and Correlation Functions

The experimental and theoretical descriptions of liquid structure are most conveniently achieved in terms of distribution functions. This is because there is
short-range structure in the liquid, but at large distances from the point of reference, the distribution of molecules is random. In this section, the fundamental
aspects of distribution and correlation functions, especially, the pair correlation
function, are outlined.
Using the congurational partition function derived earlier (equation (2.2.32)),
one may write an expression for the probability of any given conguration in
which the particle located at r1 is in volume element dr1, that at r2 in dr2, and so on
(g. 2.3). This probability is given by
f N r1 ; . . . rN dr1 . . . drN
expbU
dr1 . . . dN
Z
2:4:1
The ratio exp(bU=Z is the important part of this expression, giving the fraction
of the conguration integral associated with a given potential energy U.
Otherwise, the expression is of little practical use other than as a starting point
for deriving simpler distribution functions. The most elementary of these is the
single-particle distribution function f 1 r1 , which gives the probability of nding
a particle or molecule at position r1 . It is found by integrating the N-particle
density function f N over the remaining coordinates r2 ; . . . rN and multiplying
by the number of ways of choosing particle 1, namely N. The result is
62
Fig. 2.3 Coordinate system in

which the location of each
particle is designated by a vector
ri with its associated volume
element dri. Three particles are
shown at r1, r2, and r3. The scalar
distances between them are r12
jr1 r2 j and r13 jr1 r3 j.
f 1 r1
N
. . . expbUdr2 . . . drN
Z
2:4:2
Since the system is uniform, the potential energy is a function of relative coordinates only so that Z can be written as the product of the integral in the numerator
times the volume of the system V. It follows that the singlet distribution function
is simply the number of particles or molecules in the uid N per unit volume V,
that is,
f 1 r1
N
V
2:4:3
One may also dene the pair distribution function f 2 r1 ; r2 as the probability of
nding a particle in the volume element dr2 at r2 given that there already is a
particle in the volume element dr1 at position r1 . By analogy with equation (2.4.2),
the pair distribution function is given by
NN 1
2
. . . expbUdr3 . . . drN
f r1 ; r2
2:4:4
Z
By integrating f 2 r1 ; r2 over dr2 , that is, over the volume of the system, one
obtains the singlet distribution function:
NN 1
2:4:5
f 2 r1 ; r2 dr2 N 1f 1 r1
V
In the same way, one may dene higher-order distribution functions. For
instance, f 3 r1 ; r2 ; r3 is the probability of nding a particle at r3 given that
there are already particles at r1 and r2 . A dening relationship for this function is
f 3 r1 ; r2 ; r3 dr3 N 2f 2 r1 ; r2
2:4:6
The pair distribution function leads to the pair correlation function, which
illustrates how the local order found near a given molecule is lost as distance
from the molecule increases. This quantity is of fundamental theoretical interest
and may be determined in X-ray and neutron scattering experiments, as discussed
below. The denition of the pair correlation function gr12 is
gr12
N2 f 2 r1 ; r2
V2
63
2:4:7
This function gives the normalized probability for nding a molecule at r2 when
another molecule is at r1 . Usually, the reference molecule is assumed to be at the
origin of the coordinate system and the subscript 12 is dropped so that the pair
correlation function becomes simply gr. In the limit that r goes to innity, gr
goes to unity, demonstrating that there is no correlation between the molecule at
the origin and one innitely far away. Another way of thinking about this function is to consider gr as a factor which multiplies the average bulk density, N=V,
to give the local density, Ngr=V, about a central molecule. The correlation
function for a uid obeying the LennardJones potential as calculated by
Verlet [3] is shown in g. 2.4. At the optimum distance from the central molecule
or particle, which is slightly larger than the molecular diameter s, gr is close to
three because of attractive intermolecular forces. As distance increases, the correlation function oscillates, reecting the nite size of the particles in the system.
A second maximum is found at twice the optimum distance but its height is
signicantly reduced with respect to that of the rst. Eventually, the oscillations
die away so that, at a distance equal to four or ve particle diameters, the eect of
the central molecule is scarcely felt.
Several important quantities may be obtained from gr. One is the potential of
mean force Wr. This is dened by the equation
Wr kB T ln gr
2:4:8
The problem of obtaining a relationship between Wr and the intermolecular

potential ur is central to the statistical mechanics of non-ideal gases and liquids.
Various methods have been described for making this connection as discussed in
the statistical mechanical literature [G1G4]. As with any potential energy, one
may obtain the force acting on the central particle in the direction of any other
particle by dierentiation with respect to the distance vector r1 . Thus, the mean
force acting on the central particle due to a molecule at r1 is
Fig. 2.4 The correlation

function gr for a uid
obeying the LennardJones
potential with LJ 1:69kB T
and a molecular density N=V
0:88s3 [3].
64
Fr1
@Wr kB T @gr
@r1
gr @r1
2:4:9
Another function commonly used in discussing the structure of liquids is the

total correlation function hr. It is simply related to gr by the equation
hr gr 1
2:4:10
Clearly, when the separation between two molecules is suciently large, hr is

zero (see g. 2.4).
From the total correlation function, one obtains the structure factor SkD by
Fourier transformation. Thus,
SkD 1 N=V hreikD r dr

2:4:11
where kD is the reciprocal distance. The structure factor may be determined
experimentally from X-ray and neutron diraction experiments, and thus provides a direct route to information about the pair correlation function in liquids.
A plot of SkD against reciprocal distance kD based on the data obtained by
Yarnell et al. [4] for liquid argon near its triple point is shown in g. 2.5. Damped
oscillations of the same kind are found for the structure factor as are seen for gr.
More information about Fourier transforms is found in appendix A.
One still has the problem of relating the potential energy of the system U, which
appears in the congurational integral (equation (2.2.32)) and in the pair correlation function (equations (2.4.4) and (2.4.7)) to the intermolecular potential energy
between any two molecules, uij . The potential energy can be expanded in terms of
Fig. 2.5 An example of an experimentally determined structure factor, namely, that for
liquid argon near its triple point [4]. (Reproduced from Physical Review A, by permission.)
65
a sum of interactions between pairs of molecules, triplets of molecules, and so on.

It follows that
Ur1 ; . . . ; rN
N
X
i<j1
urij
N
X
vrij ; rik ; rjk . . .
2:4:12
i<j<k1
where urij is the pair interaction energy already discussed, and vrij ; rik ; rjk the
triplet interaction energy. The contributions of the triplet and higher-order terms
given by the second and higher sums in equation (2.4.12) constitute about 5% of
the total potential energy. Estimation of eects beyond the pair potential contribution represents an important problem in the statistical thermodynamics of
liquids and several strategies have been developed to solve this problem. More
detailed discussion of this topic, including the use of cluster diagrams, can be
found elsewhere [G1, G2, G4]. In this chapter, only the integral equation
approach is considered and is presented later.
2.5 The Experimental Study of Liquid Structure

The structure of pure liquids and liquid solutions is conveniently studied using
diraction techniques [5]. The most common of these is X-ray diraction. Two
other useful techniques are neutron and electron diraction. In these experiments,
radiation, which is usually monoenergetic, penetrates the liquid sample and is
scattered through an angle y (see g. 2.6). The analytical information is obtained
by studying the intensity of the scattered radiation as a function of this angle. In
Fig. 2.6 Schematic representation of a scattering experiment in which the intensity of

scattered radiation is measured as a function of the scattering angle y (From P. A.
Egelsta, An Introduction to the Liquid State, 2nd edition, Oxford University Press, New
York (1992), by permission.)
66
X-ray diraction, the X-rays are scattered by the electron clouds around individual atoms. Since the atoms and molecules of the liquid sample are not xed in
space, the information resulting from the diraction experiment must be interpreted in terms of statistical averages. The neutrons used in a neutron diraction
experiment are scattered by the nuclei of the atoms in the liquid sample so that the
scattering pattern is quite dierent from that for X-rays. In electron diraction,
the electrical potential, which depends on the spatial conguration of the nuclei
and electronic density distribution, determines the diraction pattern. Early
experiments involved simple monoatomic liquids such as the inert gases and liquid
metals. However, many molecular liquids have also been studied, including polar
liquids such as water, the alcohols, and amides [5]. In this section, attention is
focused on two of these techniques, namely, X-ray and neutron diraction.
The incoming radiation in the experiment is characterized by its wavelength, l,
or its frequency n, and by its intensity I0 . Special eorts must be made to achieve
monochromatic radiation. The wavelength is chosen so that it is commensurate
with the average distance between atoms, but much smaller than the sample
dimensions. A typical wavelength for neutrons is the order of 100 pm, which
corresponds to a kinetic energy of 0.08 eV; the wavelength and the linear momentum are connected through the de Broglie equation:
l
h
P
2:5:1
where P is the momentum of the neutrons. The scattering results are presented as
a function of the variable kD , which has the dimensions of reciprocal distance:
kD
4p siny=2
l
2:5:2
Consider rst of all the scattering of neutrons by the nuclei in a monoatomic

liquid. This process is characterized by a scattering cross-section, sD , which, in
turn, is related to bD , the bound atom scattering length. For slow neutrons,
sD 4pbD 2
2:5:3
Two types of scattering are observed in these experiments, namely, coherent and
incoherent scattering. The former is directly related to the structure of the sample
and can be used to determine the structure factor SkD of a simple monoatomic
liquid. Incoherent scattering arises due to isotope or spin eects. The presence of
dierent isotopes in a sample leads to dierent values of scattering length bD ,
which depends in a complex way on the number of nucleons in the nucleus from
which it is scattered. In addition, a given nucleus may possess a spin which is
randomly oriented in the sample. This characteristic of the system contributes to
incoherent scattering. The incoherence only gives information about an individual
atom and no information about the collective motion of atoms.
The relationship between the scattered intensity and the structure factor for a
monoatomic liquid is
Iy aD yhbD i2 SkD D
2:5:4
67
where aD y is a factor which depends on the properties of the instrument used for
the experiments and inelastic scattering from the sample, and D gives the contribution from incoherent scattering. hbD i denotes the value of the atomic scattering length averaged with respect to isotopic composition and spin state. The two
unknown quantities in equation (2.5.4), namely, aD y and D are determined by
measuring the intensity at y 0 and y 1. The structure factor of undeected
neutrons is related to the sample atomic density NA and isothermal compressibility, T by the following equation:
S0 NA T kB T
2:5:5
At very large angles, the structure factor goes to unity (see g. 2.6)
S1 1
2:5:6
As a result, one may show for monoatomic liquids that

SkD
IkD I0
I1 I0
2:5:7
In the case of X-rays, scattering occurs from electrons in the atoms of the liquid
sample, and depends on the electron density of the isolated atom, Ne r=V, where
Ne r is the number of electrons at distance r. The quantity corresponding to hbD i2
in the neutron diraction experiment is the atomic scattering intensity for isolated
atoms, fD 2 kD , where
ra
Ne r
sin kD r
4pr2
dr
2:5:8
fD kD
V
kD r
0
In contrast to hbD i
depends on the angle of scattering y as expressed
through the parameter kD . The structure factor SkD is simply the ratio of the
scattered intensity IkD to the incident intensity I0 divided by the intensity ratio
that would be observed for the same number of atoms of the same atomic mass as
individuals. The resulting expression for SkD is
2
; f 2D kD
SkD
IkD
Z2
I0 NZ f 2D kD
2:5:9
where NZ is the number of atoms of atomic number Z.

Both the X-ray and neutron diraction techniques are subject to experimental
diculties. It is important to be able to carry out the experiments to high values of
kD so that Fourier transformation can be performed accurately. A practical limitation for the value of kD in X-ray and classical neutron diraction experiments is
180 nm1 . It is also dicult to measure SkD for very small values of kD .
Problems due to incoherent scattering with neutrons and with absorption of
radiation also have to be dealt with. However, it is apparent from g. 2.5 that
data of high precision can be obtained for simple systems such as the rare gases in
the liquid state.
Initially, diraction experiments were performed on monoatomic liquids such
as argon and neon, or on liquid metals such as mercury. The analysis of scattering
data from molecular liquids is more dicult. Nuclei with dierent atomic masses
are present, and the incident radiation is scattered by dierent amounts by each
68
type of nucleus. In the case of symmetrical molecules such as methane or carbon

tetrachloride, the scattering data can be interpreted in a reasonably straightforward fashion. However, in the case of molecules like water, the resulting spectra
are much more complex.
For liquids with molecules containing dierent atoms, the change in intensity
with angle y is given by
"
#
X
dI
2
nA
ci f i kD FT kD
2:5:10
dy
i
where nA is the number of dierent atoms (nuclei) in the sample, ci , the molar
concentration of atom i; fi kD , the inherent scattering amplitude of the same
atom, and FT kD , the structure function for the liquid being studied. For example, in the case of water there are two atoms, the H atom and the O atom. In
neutron diraction, the scattering amplitude fi kD is an inherent property of the
atom and is tabulated for each atom on the basis of its nuclear composition. That
means that it is dierent for normal hydrogen, 1 H, and heavy hydrogen (deuterium), 2 H. Since ordinary water contains some heavy water, an ordinary water
sample actually contains three dierent atoms. In the above discussion of scattering from monoatomic liquids, fi kD was referred to as bD ; the simpler notation
was used because the scattering fraction is independent of kD in neutron experiments.
The structural information from the scattering experiment is contained in the
structure function FT kD which is dened as follows for both X-ray and neutron
experiments:
!2
XX
X
FT kD
2 dij ci cj fi kD fj kD S ij kD 1=
ci fi kD
2:5:11
ij
where S ij kD is the partial structure factor for atoms i and j, and dij is the
Kronecker delta. In the case of a monoatomic uid such as liquid argon, there
is only one kind of atom i j and the structure factor S Ar is easily extracted
from the structure function. In the case of liquid water, ignoring minor isotopic
components, there are three partial structure factors, namely, S OO ; S OH , and S HH .
The individual structure factors must be separated in the structure function in
order to obtain the corresponding pair correlation functions. This separation is
usually achieved using isotopic substitution techniques as described below. It
should be noted that the partial structure factors S OH and S HO are the same
because of the nature of the experiment. It follows that there are six independent
partial structure factors for a triatomic molecule such as HOD, namely,
S OO ; S OD ; S OH ; S DD ; S DH , and S HH .
If the individual partial structure factors S ij kD can be determined, one can
calculate the individual partial pair correlation function gij r on the basis of a
Fourier transformation:

1
V
gij r 1
S ij kD 1kD sinkD r dkD
2:5:12
2p2 Nr
0
69
However, if this is not possible, one is forced to Fourier transform the whole
structure function FT kD to obtain the overall correlation function Gr, which is
dened as
1

V
Gr
FT kD sinkD r dkD
2:5:13
2p2 Nr
0
Because the fractions fi kD and fj kD depend on the nature of the atom, Gr is

not a linear combination of the individual gij r functions, so that their separate
determination is quite dicult in general.
As pointed out above, X-radiation is scattered by the electrons around each
atom, and the scattering amplitude increases monotonically with the atomic number Z of the atom [G2]. The value of fi kD for the hydrogen atom is so small that
the contribution of partial structure factors involving this atom to FT kD can be
neglected. As a result, in an X-ray scattering experiment in pure water, the only
signicant contribution to FT kD is from S OO kD . The overall correlation function Gr obtained from these experiments shows a single sharp maximum at 285
pm [6, 7]. More detailed analysis of the data indicates that each water molecule is
surrounded on the average by four nearest neighbors.
In the case of neutron diraction, the radiation is scattered by the atomic
nuclei, not by the electrons. It turns out that nucleons such as 1 H and 2 H have
very dierent scattering amplitudes. This means that isotope eects are very
important in developing experimental strategies. Soper and Phillips [8] used
data for the structure function obtained in mixtures of normal and heavy water
to extract values of the partial structure factors for water. In this way they were
able to determine all of the pair distribution functions for water from their diffraction data. These are gHH r; gOH r, and gOO r. More details of their experimental results are given in section 2.10.
X-ray and neutron diraction studies have also been carried out in other
liquids with hydrogen bonding. Results with three or more dierent atoms
become quite complex. A good example is methanol, which has three dierent
atoms, namely, C, O, and H. Moreover, there are two dierent types of H atom,
that on oxygen, which is involved in hydrogen bonding, and those in the methyl
group, which are not. On this basis there are ten distinct contributions to the
structure function, that is, S CC ; S CO ; S CH ; S CM ; S OO ; S OH ; S OM ; S HH ; S HM ; and
S MM , where M designates a hydrogen atom in the methyl group, and H, the
hydrogen atom of the hydroxyl group. In an X-ray experiment, only three of
these are signicant, namely, S CC ; S CO , and S OO . Fourier transformation of the
structure function FT kD for methanol gives a correlation function Gr with one
sharp peak at 280 pm [9]. This corresponds to the distance between two oxygen
atoms in adjacent methanol molecules. The peak is sharp because hydrogen
bonding results in a strong correlation between oxygen atoms in liquid methanol.
Further analysis of the data demonstrates that there are, on the average, two
other molecules associated with the central one. This result is interesting because
it demonstrates the importance of hydrogen bonding in the structure of protic
solvents. Water, which has two sites for hydrogen bonding per molecule, has
approximately four nearest neighbors associated by hydrogen bonding, whereas
70
methanol, which can form one hydrogen bond per molecule, has two. More
recently, a complete description of the ten pair correlation functions in methanol
has appeared [10]. The experimental data were obtained using neutron diraction
and involved the use of isotope techniques to separate the component structure
factors. These were then Fourier transformed to give the individual pair correlation functions.
Other polar liquids which have been studied using X-ray and neutron diraction are acetonitrile [11, 12], formamide [13, 14], and dimethyl sulfoxide [15]. The
analysis of X-ray data for molecules containing the methyl group is simpler
because this group can be treated as a simple atom. On this basis, acetonitrile
contains only three distinguishable atoms, namely, CH3, C, and N. As a result
there are only six partial structure factors to be considered in analyzing the X-ray
diraction data. Studies of protic solvents such as formamide show that there are
hydrogen-bonded chains in this liquid. On the other hand, aprotic solvents like
acetonitrile and dimethyl sulfoxide show evidence of dimerization due to strong
dipoledipole interactions.
Diraction studies of liquid structure constitute an active area of contemporary research. When suitable isotopes are available the data may be resolved to
give the individual structure factors and pair correlation functions. These studies
help elucidate the role of intermolecular forces in determining liquid structure,
and hence bulk properties.
2.6 The Direct Correlation Function and the Mean

Spherical Approximation
One of the important methods of developing a model for the potential energy of
the liquid system is that based on integral equations. These include the Kirkwood
integral equation, the BornGreenYvon equation, and the OrnsteinZernike
equation. The last approach leads to the denition of the direct correlation function cr12 . It is the approach which is most frequently used and is the one which is
considered here.
In earlier discussion, the total correlation function hr12 was dened to be
gr12 1. This quantity is a measure of the total inuence molecule 1 has on
molecule 2 at a distance r12 . Ornstein and Zernike [16] proposed that this inuence
could be considered as composed of two parts, a direct part and an indirect part.
The direct part, which measures the direct inuence of molecule 1 on molecule 2,
is given by cr12 . The indirect part is the inuence propagated by molecule 1 on
molecule 3, which then aects molecule 2 either directly or indirectly through
other molecules. The indirect eect is weighted by the density and averaged
over all positions of molecule 3. As a result one may write
N
hr12 cr12
cr13 hr23 dr3
2:6:1
V
v
This equation is the OrnsteinZernike (OZ) equation and gives the mathematical
denition of cr12 with the indirect eect being expressed as a convolution
integral of h and c. By Fourier transformation, one obtains
Nc~kD h~kD
h~kD c~kD
V
71
2:6:2
where c~ and h~ are the Fourier transforms of cr and hr, respectively. Using the
denition of the structure factor, one may also write
Nc~kD SkD 1
V
SkD
2:6:3
It follows that the direct correlation function may be calculated from the experimental structure factor after Fourier transformation of the expression on the
right-hand side of this equation.
Three other important equations in the statistical thermodynamics of liquids
involve the direct correlation function and provide a connection between the
intermolecular potential ur for two molecules and the potential of mean force
Wr. One way of deriving these equations is by writing an approximate expression for the indirect contribution to the pair correlation function gr. Keeping in
mind that Wr is dened as ln gr=b, and dening gind r as the contribution to
gr from indirect interactions, an approximate expression for cr is
cr gr gind r gr expfbWr urg
2:6:4
The approximation consists of assuming that the potential of mean force Wr less
the direct interaction energy ur gives a correct estimate of the indirect eect of
the other molecules in the system. Equation (2.6.4) may also be written as
cr grf1 expburg
2:6:5
This result is known as the PercusYevick (PY) approximation [17]. It has been
applied with considerable success to the evaluation of the properties of a liquid
composed of hard spheres. In this case, the interaction energy ur is zero so that
cr 0;
r>s
2:6:6
Another approximation for cr is obtained by expanding the exponential in

equation (2.6.4). Assuming that the rst two terms in the expansion suce, one
obtains
cr gr 1 bWr bur hr ln gr bur
2:6:7
This expression for cr is called the hypernetted chain (HNC) approximation [18].
On the basis of cluster diagrams [G4] it has been shown that the HNC equation
provides a better description of most systems than the PY equation. It has been
applied extensively in the development of the theory of liquids and electrolyte
solutions. Its drawback is that the OZ equation can only be solved numerically
using this approximation.
Values of the direct correlation function estimated for the LennardJones uid
described earlier (g. 2.4) using the PY equation are shown in g. 2.7. It is
apparent that cr rises to a maximum close to the rst maximum in the total
correlation function hr and then rapidly dies away, reaching a value of zero close
to r 2:5s. When cr is estimated by the HNC equation, qualitatively similar
results are obtained but the exact values of cr dier by a few percent.
72
Fig. 2.7 Total correlation function hr () and the direct correlation function cr () for a
liquid obeying the LennardJones potential with the same parameters as in g. 2.4.
Since cr is short ranged, an even simpler approximation can often be applied

to estimate it in the region outside of a hard sphere. The logarithmic term in
equation (2.6.7) can be written as ln[1 hr]. Expanding this using only the rst
two terms, the expression for cr becomes
cr bur;
r>s
2:6:8
This approximation is known as the mean spherical approximation (MSA). For the
case of a hard-sphere uid for which ur 0, the MSA is equivalent to the PY
approximation. For the case that the hard spheres have embedded point charges,
the function ur is simply Coulombs law. Although the MSA provides the least
detailed expression for cr, it is popular because the OZ equation can often be
solved using this approximation to yield an analytical expression for gr. The
equation for gr within a hard sphere is
gr 0;
r<s
2:6:9
Equations (2.6.8) and (2.6.9) give what is known in statistical mechanics as the
closure conditions. To solve a particular problem given ur, one must rst determine cr for r < s. After Fourier transformation, one can calculate the total
correlation function in Fourier space, namely, h~k, using equation (2.6.2).
Inverse Fourier transformation then gives hr and gr.
Wertheim [19] showed that cr within a non-interacting hard sphere on the
basis of the MSA is given by
cr l1
where
62 zr zl1 r3

;
s
2s3
r<s
2:6:10
73
Fig. 2.8 The pair correlation function gr for a uid composed of hard spheres at a
packing fraction of z 0:49 calculated as a function of distance, r, using the Ornstein
Zernike equation with the direct correlation function given by equations (2.6.6) and
(2.6.10). The data shown as () are from a Monte Carlo calculation.
l1
1 2z2
1 z4
2:6:11
1 z=22
1 z4
2:6:12
and
l2
The parameter z is the packing fraction given by

z
Nps3
6V
2:6:13
The radial distribution function is now obtained using the OZ equation. The
resulting expressions for gr are rather lengthy and are not written out here.
However, it is clear that the calculated gr is very close to the value that one
obtains on the basis of a computer simulation (g. 2.8). A brief description of
computer calculations is given in the following section.
As will be seen in later chapters, the integral equation approach has been
applied to other important problems relating to liquids and solutions. The
MSA used to dene cr in the region outside of a given sphere has proven to
be especially useful because of its simplicity.
2.7 Computer Simulations of Simple Liquids

Two types of reference uids have been studied extensively by computer experiments. One system is composed of the non-interacting hard spheres discussed
74
above and is often referred to as PercusYevick (PY) uid. The other system is
made up of spheres which interact by van der Waals forces according to a
LennardJones potential. Such a uid corresponds to a good approximation
to a noble gas in the liquid state, for example, liquid argon. In a computer
experiment, the uid is imagined to consist of a cell with a known number of
particles N and volume V. The value of N is usually chosen to be between 100
and 1000, its exact magnitude depending on the speed and memory capabilities
of the computer. The real liquid is imagined to be made up of many such cells,
all being identical (see g. 2.9). Surface eects in the experiment are minimized
by assuming that, if a particle leaves the cell on one side, it re-enters from an
image cell on the opposite side. In this way, surface eects are minimized and
calculations based on a relatively small number of particles are capable of
estimating accurately the properties of a macroscopic system containing the
order of 1023 molecules.
The two simulation techniques commonly used are the Monte Carlo method
[20, 21] and the method of molecular dynamics [22, 23]. In Monte Carlo experiments, an initial conguration for the N particles is chosen and the corresponding
potential energy Ur1 ; r2 ; . . . ; rN is calculated. This conguration, chosen by the
programmer, is usually something simple, for instance, a face-centered cubic
arrangement of the particles. Then, the position of one particle is changed at
random, generating a new conguration. The potential energy U is calculated
and compared with that of the previous conguration. If it is smaller, the new
conguration is accepted as a contribution to the estimation of the congurational partition function for the ensemble, Z. However, if U is greater than that
for the previous conguration, a choice is made on the basis of the increment in
the potential energy, U. The fraction expbU is calculated and compared
with a random number r in the interval 0 to 1 generated by the computer. If the
fraction is greater than r, the new conguration is accepted and added to the chain
of congurations describing the system. If the fraction is less than the random
number r, the new conguration is rejected and the previous conguration is
added again into the chain of congurations describing the system. This process
is repeated many times, thereby generating a chain of congurations. The overall
properties of the system then converge to those of a canonical ensemble for which
Fig. 2.9 Schematic diagram

illustrating the periodic boundary
conditions used in computer
simulations. The lines represent
cell boundaries. The positions of
the particles in each cell, assumed
to be ve in this example, are
identical so that surrounding cells
may be considered images of a
central one. Each particle is given
a dierent symbol to clarify the
imaging eects.
75
V, N, and T are held constant. In this way, the congurational integral Z is

calculated. This immediately allows one to calculate the partition function Q
for the ensemble (equation (2.2.31)) and thus the thermodynamic functions for
the system. In addition, the pair correlation function for the uid is obtained in
this calculation.
The acceptance or rejection of congurations with high potential energy is a
special feature of modern Monte Carlo techniques [20, 21]. In the case of hardsphere systems, this choice is particularly simple because U is equal to zero when
spheres do not overlap, and innity when they do. Details about application of
this technique to more complex systems such as molecular uids involve some
variations which are described in the original literature.
In a molecular dynamics calculation, both the position and velocity of the
particles in the cell are specied. The movement of the particles is governed by
Newtons laws of motion, and the initial velocities are chosen from a Maxwell
distribution. This computer experiment is particularly easy for non-interacting
hard spheres because the molecules move in straight lines and undergo elastic
collisions. The velocity after the collision is found through the laws of conservation of momentum and energy. Since the total energy of the system is constant, it
corresponds to a microcanonical ensemble with constant U, V, and N. The temperature is found from the time average of the molecular kinetic energy. Since
velocity is included in the description of the system both equilibrium and transport properties can be obtained from this type of computer calculation. When
these calculations are performed for systems with continuous functions for the
potential energy U, nite dierence techniques are used. This type of calculation is
more complex and takes more computer time. In the end one obtains the same
information as in the Monte Carlo experiment, namely, the pair correlation function and the thermodynamic properties of the uid. The additional information
from molecular dynamics allows one to calculate velocity correlation functions,
and ow properties such as viscosity and diusion coecients.
The results of computer experiments are often used to test approximate
theories of liquids such as those based on the OZ equation discussed above.
For this reason they are an important part of understanding the structure of
liquids and how structure aects liquid properties.
2.8 Estimation of Thermodynamic Properties from the

Pair Correlation Function
In section 2.2 it was shown that the thermodynamic properties of a system may be
estimated from its partition function, Q. In a liquid, estimation of Q requires
knowledge of the congurational partition function, Z. In the following section
the relationship between Z and the pair correlation function gr is examined. In
principle, gr may be determined from experiment. Therefore, it is important to
understand the connection between gr and the thermodynamic properties of the
system.
The starting point in the derivation is the partition function Q (equation
(2.2.31)), which is written in terms of logarithms:
76
ln Q N ln qint ln Z ln N! 3N ln l
2:8:1
Now the internal energy U can be found by taking the derivative with respect to b,

@ ln Q
@ ln qint
@ ln Z
@ ln l
U
N

3N
2:8:2
@b V;N
@b V;N
@b V;N
@b
V;N
On the basis of the denition of l (equation (2.2.22)) and qint , this may be
rewritten as

3NkB T
@ ln Z
Neint
2:8:3
U
2
@b V;N
where the rst term on the right-hand side gives the average kinetic energy of the
system, the second, the contribution from the internal energy of the molecules,
and the last, the average potential energy of the system due to intermolecular
interactions. Dierentiating equation (2.2.32) with respect to b, the following
expression is obtained for the average potential energy:
@ ln Z 1
. . . U expbUdr1 dr2 . . . drN

hUi
2:8:4
@b
Z
The multiple integral in this equation expresses the fact that the averaging process
is carried out by multiplying the potential energy of each conguration U by its
probability given by the factor expbU=Z. In order to proceed further, the
important assumption is made that U is made up of the sum of potentials between
pairs of molecules:
X
NN 1
ur12
U
urij
2:8:5
2
1 i j N
where ur12 is the pair potential energy. After substitution of equation (2.8.5) into
NN 1
hUi
. . . ur12 expbUdr1 dr2 . . . drN
2Z

2:8:6
NN 1
ur12 . . . expbU dr3 . . . drN dr1 dr2

2Z
Comparing the inner integral with the denition of the pair distribution function
given by equations (2.4.4) and (2.4.7), one may now write
1
hUi
ur12 f 2 r1 r2 dr1 dr2
2
2:8:7
N2
ur12 gr12 dr1 dr2

2V 2
Integrating over the coordinates of particle 1, the following result is obtained for
the average potential energy:
N2
hUi
2V
urgr4pr2 dr
0
2:8:8
77
The integral in this equation gives the total intermolecular potential energy
between a central molecule and other molecules in the uid located between r
and r dr. Thus, evaluation of hUi involves adding up contributions for all values
of r and multiplying by the factor N=2, since any one of the N molecules can be
considered as central. The factor of two arises so that each pair interaction is not
counted twice. The nal expression for the internal energy is
3NkB T
N2
Neint
U
2
2V
urgr4pr2 dr
2:8:9
It should be remembered that the approximation involved in equation (2.8.5)

may not always be valid. When three-body interactions are important they must
be included in the estimation of U. Under most circumstances higher-order interactions account for approximately 5% of the potential energy so that equation
(2.8.5) represents a reasonable approximation.
The next quantity calculated from the radial distribution function is the
pressure. On the basis of equations (2.2.6) and (2.2.31), one may write

1 @ ln Z
2:8:10
P
b @V T;N
where it is assumed that the partition function due to internal degrees of freedom
is independent of volume. Derivation of the relationship between the volume
derivative of ln Z and the radial distribution function gr is presented elsewhere
[G3, G4] and is not reproduced here. The result expressed as the equation of state
is
PV
N
1
NkB T
6VkB T
gr
dur
4pr3 dr
dr
2:8:11
Estimation of the pressure requires that the derivative of the pairwise interaction
energy with respect to distance be known. This is certainly not a problem for
simple systems. Equation (2.8.11) was also derived on the basis of the assumption
that the potential energy U can be written as a sum of two-body interactions. To
this extent the result is approximate.
In order to complete the thermodynamic description, the Helmholtz energy A
is now calculated. On the basis of equation (1.3.15), one may write
so that
A U
S
T T
2:8:12

A
dU
1
Ud
d
dS
T
T
T
2:8:13
dU
P dV
dS
T
T
2:8:14
Since
78
it follows that

@A=T
U
@1=T V
2:8:15
Thus, integration of equation (2.8.9) with respect to 1=T yields the Helmholtz
energy A. This in turn requires that the radial distribution function gr be known
as a function of temperature. Since gr is generally not available as a function of
temperature, equation (2.8.15) is not a convenient route to obtain A. On the basis
of equation (1.3.19), one obtains the relationship

@A
P
2:8:16
@V T
Thus, another route to estimating A is by integration of the pressure equation
(equation (2.8.11)) with respect to volume. This calculation requires that the
dependence of gr on volume or density be known. Since this is usually not
known, this method of estimating A is also not convenient in most cases.
Another method used to estimate the Helmholtz energy in a uid with intermolecular forces is to introduce a coupling parameter, x. This quantity, which
varies between 0 and 1, measures the extent to which a central reference molecule
is coupled to all other molecules in the system. Thus, when x 0, one imagines
that the reference molecule experiences no intermolecular interactions, whereas all
other molecules interact with each other in the normal fashion. In the case of van
der Waals interactions, this is equivalent to setting the central molecules dipole
moment equal to zero at all moments of time. As the parameter x increases from 0
the dipolar characteristics of the central molecule are gradually turned on until
when x 1, the system returns to its normal state. In order to evaluate the eects
of this imaginary process, one needs to evaluate the pair correlation function for
dierent values of x. This quantity is designated gr; x where gr; 1 is equal to
gr. Of course, the value of the pair correlation function during this imaginary
process cannot be determined experimentally, but it can be estimated using
various theoretical models, for instance, the MSA model discussed earlier.
When the number of molecules is very large, the chemical potential m can be
calculated from the increase in Helmholtz energy obtained by adding molecules,
one by one, to the system. Thus,

@A
AN; V; T AN 1; V; T
2:8:17
m
@N V;T
From equations (2.2.13) and (2.2.31), one may write
AN; V; T NkB T ln qint kB T ln ZN kB T ln N! 3NkB T ln
2:8:18
where ZN is the conguration integral for the system with N molecules. It follows
that

ZN
m kB T ln qint kB T ln
2:8:19
kB T ln N 3kB T ln l
ZN1
On the basis of the denition of the conguration integral (equation (2.2.32)) it
can be shown that

1
ZN
@ ln ZN
dx
ln
ln V
ZN1
@x
79
2:8:20
Now making use of the relationship between the average potential energy hUi and
the pair correlation function (equation 2.8.8), one obtains
1 1
ZN
N
ln
urgr; x4pr2 drdx
ln V
ZN1
VkB T
2:8:21
0 0
The resulting expression for the chemical potential is

!
1 1
Nl3
N
urgr; x4pr2 drdx
m kB T ln
Vqint
VkB T
2:8:22
0 0
If one has a model for gr; x) one may estimate the chemical potential m, and
hence, all of the other thermodynamic functions.
The above analysis demonstrates the importance of the pair correlation function in estimation of the thermodynamic properties of simple liquids. In the
following section, the properties of the simplest uid, namely, one based on
non-interacting hard spheres, are developed on the basis of the relationships
presented in this section.
2.9 The Properties of a Hard-Sphere Fluid

It is useful to examine the properties of a uid made up of non-interacting hard
spheres. Such a system may be regarded as an important reference liquid, albeit
ctitious, with respect to which the properties of real systems can be compared. Its
properties are most easily obtained on the basis of the PercusYevick (PY)
approximation. Since the spheres do not interact, the interaction energy ur is
zero outside any sphere:
ur 0;
r>s
2:9:1
Since the spheres are hard, that is, they do not penetrate one another,
ur 1;
r<s
2:9:2
It is apparent that the interaction energy is discontinuous at the boundary of a

sphere. From the discussion in section 2.6, discontinuities are also possessed by
the pair correlation function gr and the direct correlation function cr at the
boundary of a sphere. In order to remove problems associated with these discontinuities, one introduces a new function, yr dened as follows:
yr gr expbur
2:9:3
Outside of a sphere where ur is zero, one nds that

yr gr;
rs
2:9:4
80
Referring to the PY approximation (equation (2.6.5)), inside a sphere where gr is

zero,
yr cr;
r s
2:9:5
Although cr and gr are discontinuous at r s; yr is continuous.

The rst step in obtaining the properties of the hard-sphere system is to calculate its equation of state on the basis of equation (2.8.11). From equations (2.9.1)
and (2.9.2), the derivative dur=dr is zero for all values of r except r s where it
is a negative-going delta function. This follows from the fact that ur drops from
1 to 0 at the boundary of a hard sphere. It can be shown that the integral in the
equation of state (equation (2.8.11)) simplies to the value of the integral at r s
multiplied by kB T so that one obtains
PV
2pNs3
1
gs 1 4zys
NkB T
3V
2:9:6
where the packing fraction z is introduced to simplify the result (see equation
(2.6.13)). On the basis of equations (2.9.4) and (2.9.5), one may write
gs ys cs
2:9:7
where s and s indicate that the function is evaluated at the positive and
negative side of the discontinuity at s, respectively. Thus, gs may be calculated
on the basis of equations (2.6.10)(2.6.12), giving the result that
gs
1 z=2
1 z2
2:9:8
Substituting equation (2.9.8) into equation (2.9.6) one obtains the following equation of state:
PV
1 2z 3z2
NkB T
1 z3
2:9:9
In the limit of very low densities, the parameter z goes to zero and the ratio
PV=NkB T approaches one. This is the expected ideal gas limit. At higher densities, the ratio PV=NkB T is greater than unity and goes to innity when z 1.
In fact, for a real system, the value
p of 3z is limited by the density of hard spheres in
a close-packed
system,
which
is
2=s . The corresponding value of the parameter
p
z is 2p=6, or 0.74, and that of PV=NkB T, 61.
The equation of state may also be calculated from an equation relating the
compressibility of the system to the pair correlation function gr. The result is
PV
1 z z2
NkB T
1 z3
2:9:10
For low densities, this equation gives the same value of PV=NkB T as that
estimated from the pressure equation (equation (2.9.9)). However, at high densities, the estimate from the compressibility equation is much higher (see g. 2.10).
The disagreement between the equations of state obtained by the two dierent
methods clearly is a result of the approximation made in deriving the PY
equation.
81
Values of PV=NkB T have been calculated for the hard-sphere system by

computer simulation. At high densities, the results obtained in this way fall
between those estimated by the pressure and compressibility equations. This led
Carnahan and Starling [24] to propose an equation of state obtained by combining one-third of the result from the pressure equation (equation (2.9.9)) together
with two-thirds of the compressibility result (equation (2.9.10)). The result is
PV
1 z z2 z3
NkB T
1 z3
2:9:11
The CarnahanStarling equation of state agrees well with the result of computer
simulations over the range shown in g. 2.10 and is used in all further calculations
presented here.
EXAMPLE
Estimate the value of PV=NkB T for a hard-sphere uid using the Carnahan
and Starling equation of state assuming a concentration of 10 M and a hardsphere diameter of 300 pm.
The concentration 10 M corresponds to
10 6:023 1023
6:023 1027 molec m3
103
2:9:12
The packing fraction z is equal to

z
3:1416 6:023 1027 3 1010 3

8:514 102
6
2:9:13
Thus,
PV
1 8:514102 8:514102 2 8:514 102 3
1:427
NkB T
1 8:514102 3
2:9:14
At this concentration the hard-sphere uid has properties which are signicantly dierent from those of an ideal uid (PV=NkB T 1).
Fig. 2.10 Plot of PV=NkB T

for a hard sphere uid
against the ratio of its
density N=V to the
p closepacked density 2=s3 .
The data designated () were
obtained using the pressure
equation (2.9.9), those
designated (^) using the
compressibility equation
(2.9.10), and the smooth
curve, using the Carnahan
Starling equation (2.9.11).
82
The Helmholtz energy of the system is now calculated using equations (2.8.16)
and (2.9.11).
Accordingly,
A
P
1 z z2 z3
dV
2:9:15

dV
NkB T
NkB T
V1 z3
Recalling that
z
pNs3
6V
2:9:16
it follows that
dV
pNs3
dz
6z2
2:9:17
so that the integral in equation (2.9.15) can be written in terms of z. The result is
A
1 z z2 z3
dz
2:9:18
NkB T
z1 z3
This integral can be broken up into four integrals, all of which can be found in
standard tables. The integration is carried out from a nite but very low density
where the parameter z is z0 , and the properties of the system are those of an ideal
gas, to any higher value of z. The result is

A
A0
z
3 2z 3z2 3
2:9:19
ln

NkB T NkB T
z0
2
21 z2
where A0 is the Helmholtz energy of an ideal gas. On the basis of equations
(2.2.13) and (2.2.23) this contribution is
A0
ln V0 ln N 1 3 ln l
NkB T
2:9:20
where V0 is the volume of the ideal gas. Noting that

z
V
0
V
z0
the expression for the Helmholtz energy becomes

A
N
3 2z 3z2 3
3 ln l ln

NkB T
V
2
21 z2
2:9:21
2:9:22
One may now derive the remaining thermodynamic functions. First of all, the
internal energy is obtained by dierentiation of A=T with respect to the reciprocal
temperature:

@A=T
@ ln l
3NkB T
U
2:9:23
3NkB
@1=T V
@1=T
2
83
The internal energy is simply the kinetic energy of the system and therefore does
not dier from the result for an ideal gas containing the same number of molecules. One may now write an expression for the entropy of the system:

S
UA
N
3 2z 3z2
3 ln l ln
3
2:9:24

NkB NkB T
V
21 z2
Finally, the Gibbs energy is obtained by adding PV to the Helmholtz energy:

G
A PV
N
3 2z 3z2 1 z z2 z3 3
3 ln l ln

NkB T
NkB T
V
2
21 z2
1 z3
2:9:25
For an ideal monoatomic gas, the Gibbs energy at constant temperature and
pressure varies with the logarithm of the molecular density. The additional
terms in z in equation (2.9.25) give the contribution to G due to the non-ideality
of the hard-sphere system with respect to an ideal gas.
EXAMPLE
Estimate the internal energy, Helmholtz energy, and entropy of a hard-sphere

uid in units of NkB T assuming a concentration of 10 M, a diameter of 300 pm,
a molecular mass of 30 g, and a temperature of 258C.
The concentration of hard spheres is 6:023 1027 m3:
The mass of each sphere is
30 103
4:98 1026 kg
6:022 1023
The parameter l (equation (2.2.22)) is
l
6:626 1034
2 3:1416 4:98 1026 1:38 1023 298:21=2
1:847 1011 m
2:9:26
3 ln l 74:14

N
ln
63:97
V
2:9:27
2:9:28
From the previous example, z is equal to 8:514 102 :

Thus,
3 2z 3z2
7:52
21 z2
2:9:29
A
74:14 63:97 7:52 1:5 4:15
NkB T
2:9:30
and
The internal energy in units of NkB T is 1.5.

The entropy is given by
84
Fig. 2.11 Plots of the Helmholtz energy for a hard-sphere uid (s 0:37 nm) against the
logarithm of the pressure at constant temperature as indicated.
S
UA
1:5 4:15 2:65
NkB NkB T
2:9:31
Values of the Helmholtz energy estimated as a function of pressure at constant

temperature on the basis of equation (2.9.22) are shown in g. 2.11. These plots
are reasonably linear in the logarithm of the pressure at low pressures. This is to
be expected, since the density is proportional to the pressure under these conditions, and the eects of non-ideality are relatively unimportant. However, at
higher pressures the value of A, starts to rise sharply due to non-ideality.
Eventually, one reaches positive values of A, indicating that the uid is not stable.
It has been shown that the hard-sphere system undergoes a phase transition from
uid to solid when Ns3 =V 0:943. For the system considered in g. 2.11, this
corresponds to a pressure of 760 bars at 300 K. Many interesting calculations
relating to phase transitions and critical phenomena in hard-sphere and
LennardJones uids have been carried out.
Just as the ideal gas forms a convenient point of reference in discussing the
properties of real gases, so does the hard-sphere uid in discussing the properties
of liquids. This is especially true at low densities, where the role of intermolecular
forces in real systems is not so important. In this limit, the hard-sphere model is
useful in developing the theory of solutions, as will be seen in chapter 3.
2.10 The Structure of Water

The discussion in this chapter has largely concerned very simple liquids such as a
hypothetical uid composed of non-interacting hard spheres, or spheres interacting via the LennardJones potential function. The most common liquid, namely,
water, is much more complex. First, it is a molecule with three atoms, and has a
85
large permanent dipole moment. Second, it does not possess a center of symmetry,
unlike methane, or carbon tetrachloride. Finally, it is more complex than other
polar solvents because of the strong hydrogen bonding.
In spite of its complexity, the structure of liquid water has been determined
with precision using neutron diraction techniques [8]. The structure factor may
be analyzed to obtain correlation functions for each type of atomatom interaction in the system. For water, the three atomatom distribution functions are
gOO r; measuring the distribution of oxygenoxygen separations, gOH r; measuring the oxygenhydrogen distribution, and gHH r, measuring the hydrogen
hydrogen distribution.
Since the scattering amplitude for oxygen is much smaller than that for hydrogen, the gOO r distribution is the most dicult to determine. The experimental
results are illustrated in g. 2.12. This function rises from a value of zero at
200 pm to a sharp maximum of 3.09 at a distance of 288 pm. This means that
the local density of oxygen atoms is 3.09 times greater than the average value.
However, in order to estimate of the oxygenoxygen coordination number one
must consider the distribution density over the whole region of the maximum in
gOO r. One way of achieving this is to integrate the distribution function up to the
rst minimum, considering shells of volume 4pr2 dr. The coordination number is
then dened as
rmin
ro gOO r 4pr 2 dr
nc
2:10:1
where ro is the density of oxygen atoms in the liquid. When this integration is
carried out to a distance of 330 pm from the central oxygen atom, the result is that
nc 4:3. This suggests that the central water molecule is approximately in a
tetrahedral environment surrounded on the average by four nearest neighbors.
Fig. 2.12 Oxygenoxygen distribution function in liquid water determined at 258C by

neutron diraction [8].
86
Beyond the rst minimum, gOO r oscillates with a second maximum at 450 pm. In
order to understand these results in more detail, one needs to carry out a molecular dynamics calculation. However, it is clear that the behavior of water is
roughly similar to that of much more simple liquids if one considers the oxygen
atom alone.
A much dierent pattern is seen for the oxygenhydrogen distribution function
(g. 2.13). These results have been corrected for the intramolecular contribution
to gOH r, and show only the eects of the intermolecular correlation. A fairly
sharp maximum occurs at 185 pm, and a broader but higher one at 330 pm. The
rst maximum is clearly associated with the hydrogen atoms in surrounding
molecules which are hydrogen bonded to the oxygen atom in a central molecule.
Integration out to the rst minimum gives a coordination number of 1.7. The
second maximum occurs at the rst minimum on the gOO r distribution curve. At
this point, the total number of hydrogen atoms around the central oxygen is close
to eight, that is, two times the number of oxygen atoms in the same volume of
solution. These observations all follow what is expected on the basis of the molecular composition of water.
The distribution function for hydrogen atoms with respect to a central hydrogen is shown in g. 2.14. The function gHH is the easiest to determine experimentally because of the large scattering amplitude for hydrogen with respect to that
for oxygen. These results have also been corrected for the intramolecular contribution to gHH r. The rst maximum at 240 pm is followed by one which is less
high at 390 pm. The coordination number estimated by integrating out to the rst
minimum at 310 pm is 5.8. These hydrogen atoms are associated with a central
molecule via hydrogen bonding. Comparing gOH r and gHH r, one sees that the
maxima on the HH distribution function are located further from the central
atom by 60 pm with respect to those on the OH distribution. This is perfectly
reasonable because of the OH bond length in water.
Another method of studying water structure is Raman spectroscopy [25]. Using
this technique, one is able to distinguish spectral features which arise from intra-
Fig. 2.13 Oxygenhydrogen distribution function in liquid water determined at 258C by

87
Fig. 2.14 Hydrogenhydrogen distribution function in liquid water determined at 258C by

molecular and intermolecular vibrational modes of the water molecules.

Interpretation of the data is assisted by comparing spectra obtained in H2O
and D2O, and by changing the temperature. A picture which is consistent with
both the Raman spectra and the neutron diraction data shows considerable
order around a given water molecule. Because of hydrogen bonding, one molecule
is surrounded on the average by four hydrogen-bonded nearest neighbors (see g.
2.15). The orientation of these neighbors can be such that the system has a high
degree of symmetry. However, keeping in mind the dynamic character of liquid
structure, rotation of the molecules hydrogen bonded to the central molecule can
reduce the symmetry. The dynamic character of the system also results in a very
Fig. 2.15 Diagram of a water molecule

surrounded by four nearest neighbors
whose orientation is determined by
hydrogen bonding. Hydrogen bonds are
denoted by broken lines and OH
bonds by solid lines. (Reproduced from
reference 25, with permission.)
88
weak correlation between the central molecule and those located further away
than the four nearest neighbors. The vibrational spectra of water are discussed in
more detail in section 5.7A.
It is clear that the physical properties of water are very much inuenced by the
important role played by hydrogen bonding in determining its structure. These
properties include the high dielectric permittivity, which cannot be explained on
the basis of dipoledipole interactions alone. It is also clear that electrolytes have
a very disruptive eect on water structure. Cations are solvated by the lone
electron pairs on the oxygen atom of the water molecule and thus cause considerable disruption in the local water structure. This leads to changes in the bulk
physical properties of water, such as its permittivity.
Many aspects of the structure and properties of aqueous solutions can be
understood in terms of the qualitative structural changes that occur because of
solutesolvent interactions. This subject is discussed in more detail in the following chapters.
2.11 Distribution Functions for Liquid Solutions

The techniques used to describe the properties of pure liquids can be extended in a
fairly straightforward fashion to liquid solutions [26, 27]. This treatment is normally restricted to liquids in which the molecules behave as non-interacting hard
spheres or as dipolar species interacting via a LennardJones potential. The discussion here is limited to two-component mixtures but it is easily extended to
more complex systems.
The energy equation for a liquid mixture or solution is
3NkB T
N2 X
2p
xx
U
2
V ij i j
uij rgij rr2 dr
2:11:1
where x1 is the mole fraction of component 1, x2 , that of component 2, and

N N1 N2 . The summation involves three terms and requires specication of
the interaction energies u11 ; u12 and u22 as well as the pair correlation functions
g11 ; g12 and g22 . In equation (2.11.1), the contribution to U from the internal
degrees of freedom of the two molecules in the solution has not been included.
Several approaches to estimation of the internal energy have been described.
These involve assuming specic relationships between the pair correlation functions gij and the form of the interaction energy uij . The simplest of these is based
on the van der Waals treatment of uids and its application of the law of corresponding states. Examination of typical radial distribution functions for mixtures
such as those shown in g. 2.16 reveals that the maximum in each distribution
function gr occurs close to the diameter s describing the distance of closest
approach for the two molecules involved. Thus, it is better to describe the radial
distribution function in terms of the reduced distance r=s instead of the distance r.
This conclusion leads to the assumption that
89
Fig. 2.16 Typical radial distribution functions of a two-component mixture plotted against
the reduced distance for the 22 distribution. (From reference 26, with permission.)

g11

r
r
r
g22
g12
s11
s22
s12
2:11:2
When the interaction energy is given by the LennardJones potential, one may
write
uij r 4eij
h
s n
s m i
ij
ij

r
r
2:11:3
where n is a coecient describing repulsive interactions and m, one describing

attractive interactions. As discussed earlier, n is often assumed to be 12, and m is 6
for dipoledipole interactions (see equation (2.3.35)). This expression includes the
case of a hard-sphere uid when n 1 and the term in m is neglected. As far as
liquids are concerned the more important contribution comes from the repulsive
component which predominates for molecules in close contact. The Lennard
Jones function is conformal because it has the same form independent of the
molecule, that is, of the values of eij and sij .
Changing the variable in equation (2.11.1) from r to y r=sij and substituting
the expression for uij r one obtains
3NkB T
N2 X
8p
U
xxe s 3
2
V ij i j ij ij
If one denes

1
1
g yy2 dy

yn ym ij
2:11:4
90
s3mix
xi xj sij 3
2:11:5
ij
and
emix
xi xj eij sij 3 =
ij
xi xj sij 3
2:11:6
ij
the energy equation simplies to

3NkB T
N2
8p
e s3
U
2
V mix mix

1
1
gyy2 dy

yn y m
2:11:7
where gy is any one of the distribution functions dened in equation (2.11.2).

This equation shows that the composition dependence of the internal energy U is
specied only through the quantities emix and smix For this reason, the properties
of the system at a given composition are those of a hypothetical pure uid with
properties dened by emix and smix . Thus, the model is called a one-uid model.
In the case of mixtures of simple molecules, the Gibbs energy for a one-uid
system may be written as
G emix kB T=emix 3NkB T ln smix
2:11:8
where the function kB T=emix is found by curve-tting the experimentally

observed data for G for the pure liquids as a function of temperature. The
other thermodynamic functions are determined by dierentiation.
The van der Waals one-uid theory is quite successful in predicting the properties of mixtures of simple molecules. Unfortunately, the systems usually considered by chemists are considerably more complex, and often involve hydrogen
bonding and other chemical interactions. Nevertheless, the material presented
here outlines how one could proceed to develop models for more complex systems
on the basis of the integral equation approach.

The fundamental quantity used to describe liquid structure is the pair correlation
function gr. It has been estimated both experimentally and on the basis of
computer calculations for a great variety of liquid systems. Simple monoatomic
liquids such as the inert gases were of great interest to physicists because they
allowed them to test theories based on fairly simple models. Intermolecular forces
can be described using the LennardJones potential and the thermodynamic
properties of these liquids can be estimated reasonably well. On the other hand,
chemists are interested in much more complex polyatomic systems. When the
constituent molecule is polar, description of the intermolecular forces must at
least include dipoledipole interactions and often higher-order multipole interactions. In the case of protic liquids such as water, hydrogen bonding is a very
important factor in determining its structure as a liquid. This was seen directly
in the case of water from the correlation functions derived from neutron dirac-
91
tion data and from the structural information obtained from Raman spectroscopy.
In discussing the statistical thermodynamics of liquids, emphasis has been
placed on the OrnsteinZernike equation and the integral equation approach to
obtaining thermodynamic properties from molecular ones. This is by no means
the only approach, an alternative one being based on the perturbation theory of
liquids. Perturbation theory has been developed in some detail and is quite successful in describing the properties of simple liquids [28, 29]. In addition, in
discussing the integral equation approach, simple approximations such as the
mean spherical approximation (MSA) have been emphasized, since they often
give analytical expressions for the thermodynamic functions. A good example
of such an outcome is the PercusYevick results for a hard-sphere uid discussed
in section 2.9. The nature of the approximations used in these methods can be
understood in detail using cluster diagrams. More about this method of describing
the integrals which arise in a detailed description of intermolecular interactions
can be found in monographs dealing with liquid structure [G1, G2].
In conclusion, the material in this chapter is meant to give only an introduction
to the subject of liquid structure. Much of what has been presented has dealt with
systems which can be represented as point dipoles embedded in hard spheres. Very
few liquid systems of chemical interest can be described in such simple terms.
However, the simple models can often be modied to make them more realistic.
For example, the eects of chemical interactions can be introduced by assuming
that the hard sphere experiences sticky interactions in a given direction with
respect to the central dipole. Other methods are available for dealing with the
eects of non-sphericity. Thus, the simple models can often be made relevant to
chemical systems after suitable modication.
In the following chapters, some of the theories introduced here are used to
discuss other systems, including polar solvents and electrolyte solutions. The
statistical mechanical tools introduced here are important because they help
one to develop an understanding of the way that molecular properties of a
given system inuence its macroscopic properties.
General References
G1. Egelsta, P. A. An Introduction to the Liquid State, 2nd ed.; Oxford University Press:
New York, 1992.
G2. Watts, R. O. and McGee, I. J. Liquid State Chemical Physics; John Wiley: New York,
1976.
G3. Eyring, H.; Henderson, D.; Stover, B. J.; Eyring, E. M. Statistical Mechanics and
Dynamics, 2nd ed.; John Wiley: New York, 1982.
G4. McQuarrie, D. A. Statistical Mechanics; Harper and Row: New York, 1976.
References
1. Barker, J. A.; Henderson, D. Rev. Mod. Phys. 1976, 48, 587.
2. London, F. Trans. Faraday Soc. 1937, 33, 8.
3. Verlet, L. Phys. Rev. 1968, 165, 201.
4. Yarnell, J. L.; Katz, M. J.; Wenzel, R. G.; Koenig, S. H. Phys. Rev. 1973, A7, 2130.
92
5. Kalman, E.; Palinkas, G. In The Chemical Physics of Solvation, Part B, Spectroscopy

of Solvation; Dogonadze, R. R., Kalman, E., Kornyshev A. A., Ulstrup, J., eds.;
Elsevier: Amsterdam, 1986, Chapter10.
6. Narten, A. H.; Levy, H. A. J. Chem. Phys. 1971, 55, 2263.
7. Hajdu, F.; Lengyel, S.; Palinkas, G. J. Appl. Crystallogr. 1976, 9, 194 .
8. Soper, A. K.; Phillips, M. G. Chem. Phys. 1986, 107, 47; Soper, A. K. Chem. Phys.
2000, 258, 121.
9. Narten, A. H.; Habenschuss, A. J. Chem. Phys. 1984, 80, 3387.
10. Yamaguchi, T.; Hidaka, K.; Soper, A. K. Mol. Phys. 1999, 97, 603.
11. Bertagnolli, H.; Zeidler, M. D. Mol. Phys. 1978 35, 177.
12. Radnai, T.; Itoh, S.; Ohtaki, H. Bull. Chem. Soc. Jpn. 1988, 61, 3845.
13. Kalman, E.; Serke, I.; Palinkas, G.; Zeidler, M. D.; Weisman, F. J.; Bertagnolli, H.;
Chieux, P. Z. Naturforsch. 1983, 38a, 231.
14. Bellisent Funel, M.C.; Nasr, S.; Bosio, L. J. Chem. Phys. 1997, 106, 7913.
15. Luzar, A.; Soper, A. K.; Chandler, D. J. Chem. Phys. 1992, 96, 8460.
16. Ornstein, L. S.; Zernike, F. Proc. Acad. Sci. Amsterdam 1914, 17, 793.
17. Percus, J. K.; Yevick, G. J. Phys. Rev. 1958, 110, 1.
18. van Leeuwen, J. M. J.; de Boer, J. Physica (Utrecht) 1959, 25, 792; Green, M. S.
J. Chem. Phys. 1960, 33, 1403.
19. Wertheim, M. S. J. Math. Phys. 1964, 5, 643.
20. Metropolis, N.; Rosenbluth, A. W.; Rosenbluth, M. N.; Teller, E.; Teller, A. H.
J. Chem. Phys. 1953, 21, 1087.
21. McDonald, I. R.; Singer, K. Quart. Rev. 1970, 24, 238.
22. Rahman, A. Phys. Rev. 1964, 136, A405.
23. Alder, B. J.; Wainwright, T. E. J. Chem. Phys. 1957, 27, 1208; 1959, 31, 459; 1960, 33,
1439.
24. Carnahan, N. F.; Starling, K. E. J. Chem. Phys. 1970, 53, 600.
25. Walrafen, G. E. J. Chem. Phys. 1964, 40, 3249.
26. Henderson, D.; Leonard, P. J. In Physical Chemistry, an Advanced Treatise; Eyring,
H., Henderson, D., Jost, W., eds.; Academic Press: New York, 1971; Vol. VIII B,
Chapter 7.
27. McDonald, I. R. In Statistical Mechanics, Specialist Periodical Reports; Chemical
Society: London, 1973; Vol. 1, p 134.
28. Henderson, D.; Barker, J. A. In Physical Chemistry, an Advanced Treatise Eyring, H.,
Henderson, D., Jost, W., eds.; Academic Press: New York, 1971; Vol. VIII B.
29. Smith, W. R. In Statistical Mechanics, Specialist Periodical Reports; Chemical Society:
London, 1973; Vol. 1, p 71.
Problems
1. The following molecules form liquids at room temperature. Calculate the
energy involved between two similar molecules at contact (r s) due to
dipoledipole, induced dipoledipole and London dispersion forces. The
molecules are assumed to be spherical with a diameter s. Express the results
in units of kB T at 258C.
Molecule
Diameter
s=pm
Dipole Moment
p/debye
Polarizability
1030 a=m3
Ionization Potential
I=MJ mol1
CCl4
CH2Cl2
CH3OH
538
450
371
0
1.60
2.87
10.49
6.48
3.26
1.107
1.090
1.047
93
2. Calculate the LennardJones interaction potential for the following molecules and plot them on the same graph. The calculations should be carried
out for increments of r by 0.02s in the range 0.91.2s, and by 0.1s in the
range 1.22.4s.
Molecule
LJ k1
B =K
s=pm
Tb =K
423
333
298
592
491
431
342
462
374
Hexane
Dimethylsulfoxide
Nitromethane
Does the interaction potential give a good measure of cohesive forces in the
corresponding liquids? Discuss in terms of the boiling point.
3. The following table gives the distribution function gr for liquid argon at
85 K. The LennardJones parameters for argon are s 350 pm and
LJ k1
B 118 K. Estimate the LennardJones potential at each value of r,
and then use the PercusYevick equation to calculate the direct correlation
function cr (equation (2.6.5)). Plot gr and cr against r.
r=pm
gr
r=pm
gr
r=pm
gr
r=pm
gr
0
100
200
300
320
327
334
341
347
354
361
368
375
381
388
395
402
409
415
422
429
436
443
449
456
0
0
0
0
0.072
0.321
0.736
1.283
1.886
2.441
2.851
3.054
3.040
2.850
2.555
2.234
1.947
1.722
1.555
1.424
1.306
1.186
1.062
0.946
0.847
463
477
504
518
552
579
592
613
627
647
667
688
715
729
756
769
797
804
810
817
824
831
844
872
878
0.773
0.686
0.580
0.560
0.609
0.729
0.804
0.914
1.004
1.140
1.229
1.269
1.254
1.230
1.125
1.056
0.947
0.921
0.894
0.870
0.850
0.836
0.825
0.836
0.845
855
892
899
906
913
919
926
933
940
947
953
960
967
981
994
1008
1022
1042
1062
1075
1090
1103
1124
1144
1165
0.859
0.876
0.894
0.911
0.928
0.943
0.959
0.976
0.994
1.014
1.033
1.049
1.063
1.083
1.099
1.108
1.104
1.084
1.065
1.045
1.020
0.999
0.973
0.944
0.934
1219
1253
1273
1293
1345
1376
1410
1444
1485
1566
1655
1805
1900
1975
2057
2118
2227
2295
2377
2479
2547
2620
2683
1
0.954
0.984
1.014
1.034
1.046
1.026
1.007
0.989
0.971
0.993
1.020
0.988
1.000
1.009
1.000
0.994
1.000
1.004
1.000
0.998
1.000
1.002
1.000
1.000
94
4. Use the above data to calculate the internal energy per mole of argon at 85 K
given that its density is 21.25 atoms nm3 . The contribution to the internal
energy from internal degrees of freedom (equation (2.8.9)) should be
neglected. Develop a condition in your computer program so that numerical
integration is terminated when the integral becomes constant.
5. Estimate the direct correlation function for liquid argon at 85 K using the
hypernetted chain approximation with the data given in problem 3. Compare
the result with that found using the PercusYevick approximation.
6. Estimate the packing fraction for a hard-sphere liquid with a density of 21.25
atoms nm3 and a hard-sphere diameter of 350 pm. Use this result to calculate the PercusYevick product for the system at 85 K using the Carnahan
Starling equation of state (equation (2.9.11)).
7. Calculate the internal energy and entropy of the hard-sphere system
described in problem 6. Assume that the atomic weight of the hard sphere
is 39.95 g mol1 .
8. The following table gives the distribution function gOO r in liquid water at
258C. Use these data together with the density of water (0.997 g cm3 ) to
calculate the coordination number for a central oxygen atom assuming that
the pair correlation function should be integrated up to the rst minimum.
r=pm
gOO r
r=pm
gOO r
r=pm
gOO r
200
213
223
233
243
248
253
258
263
268
273
0
0.008
0.040
0.116
0.233
0.306
0.399
0.541
0.782
1.179
1.746
278
283
288
293
298
303
308
313
318
323
333
2.388
2.907
3.092
2.869
2.351
1.758
1.273
0.966
0.813
0.752
0.734
343
353
363
373
383
393
403
413
423
433
443
0.749
0.773
0.807
0.850
0.898
0.950
1.003
1.050
1.089
1.116
1.333
Electrolyte Solutions
Wilhelm Ostwald was born to a German family

in 1853 in Riga, Latvia, where he grew up and
attended school. In 1872 he entered Dorpat
University (now Tartu University in Estonia)
where he studied chemistry, obtaining the
diploma in 1875. At that time the Baltic States
were part of the Russian Empire with some
institutions of higher learning operating in
German for the German minority who lived in
that region of Europe. Ostwald stayed at
Friedrich Wilhelm Ostwald
Dorpat to study physics under Arthur von
Oettingen and chemistry under Carl Schmidt.
In 1877 he was appointed lecturer at Dorpat, and he obtained his doctoral
degree in 1878. He moved back to Riga in 1881 as professor of chemistry at
the Polytechnic Institute. In 1887 he accepted the Chair of Physical Chemistry
at Leipzig University and remained there until 1906. Ostwald is generally
considered to be the founder of modern physical chemistry. He began the rst
journal dealing with this subject, namely, Zeitschrift fur physikalische Chemie in
1887, and edited it until 1922. The main area of research interest for Ostwald
was electrolyte solutions, but he also contributed to other areas of physical
chemistry, especially reaction kinetics and catalysis. He had several famous
associates including S. A. Arrhenius, J. H. vant Ho, and W. Nernst, all of
whom won the Nobel Prize in Chemistry. Many young North American
chemists studied in his laboratory, and then went on to develop the new area
of physical chemistry at home. These included G. N. Lewis, A. A. Noyes, and
W. Lash Miller. Ostwald spent much of his time during his long retirement
studying the scientic basis for color. He won the Nobel Prize in Chemistry in
1909 for his work on catalysis, chemical equilibria, and reaction kinetics. He
died in 1932.
3.1 Electrolyte Solutions Are Always Non-Ideal

Electrolyte solutions are important in all branches of chemistry, but especially in
analytical chemistry, and biochemistry. These systems by their nature are always
non-ideal, and represented an early challenge to theoreticians interested in
describing their thermodynamic properties. The solute components are ions,
cations, and anions, which carry opposite charges and thus interact very dier95
96
ently with one another. The existence of electrolyte solutions depends on the polar
properties of the solvent through which the individual ions are stabilized. When
one recognizes the molecular nature of the solvent, one must also consider the
interactions between solvent dipoles and the ion. This results in changes in solvent
structure in the immediate vicinity of the ions. It follows that a complete description of an electrolyte solution at the molecular level requires the consideration of
iondipole, ionion, and dipoledipole interactions. In addition to these simple
electrostatic interactions, one must also consider the role of hydrogen bonding in
protic solvents like water.
In very dilute electrolyte solutions, the most important consideration is ion
dipole interactions. One expects these interactions to be dierent for cations and
anions. This follows from the fact that the solvent molecule is not a simple dipole
in the electrostatic sense but instead it has a chemical structure which is dierent
at each end of the molecular dipole. Each ion interacts locally with four to six
solvent molecules in its immediate surroundings. In the case of water, the concentration of water molecules in the pure liquid is 55.5 M; it follows that the
number of water molecules experiencing direct interaction with ions in dilute
solutions represents a small fraction of the total number.
As the electrolyte concentration increases, ionion interactions become more
important in determining the thermodynamic properties of the solution. The
electrostatic eld of an ion is long ranged, decreasing with the reciprocal of the
distance from the charge center of the ion. As a result a given ion has an ionic
atmosphere in which the concentration of oppositely charged ions in its vicinity is
slightly greater on the average than that of ions of the same charge. The properties
of the ionic atmosphere depend on temperature, that is, on the randomizing
eects of thermal motion. Thus, the composition of the ionic atmosphere uctuates with time, and the relative dierence between the local concentration of
cations and anions in the vicinity of a given ion varies with both temperature and
distance. As the electrolyte concentration increases, the average distance between
ions and the thickness of the ionic atmosphere both decrease.
Under some circumstances ionion interactions can be more important than
iondipole interactions. This is especially true when the valence of the ion is
greater than one, and the electrolyte concentration is high. Then, the formation
of ion pairs and higher aggregates is possible. Two types of ion pairs have been
recognized, namely, contact ion pairs in which the cation and anion are in physical contact, and solvent-separated ion pairs in which one or two solvent molecules are situated between the cation and anion. Ion pairing must be considered in
developing a complete picture of an electrolyte solution.
As was seen in chapter 2, both dipoledipole interactions and hydrogen bonding are important in determining the structure and thermodynamic properties of
pure water. In the immediate vicinity of an ion, the solvent structure is disrupted
so that local dipoledipole interactions and hydrogen bonding are dierent than
they are in pure water. These changes are also important because they aect the
local permittivity and the strength of ionion interactions.
In the present chapter, the properties of electrolyte solutions in water are
discussed in detail. Initially the solvation of ions in innitely dilute solutions is
considered on the basis of the Born theory. Then, the DebyeHuckel model for
ELECTROLYTE SOLUTIONS
97
ionion interactions in dilute solutions is described. Finally, the role of ion pairing
is outlined. Methods of improving the description of electrolyte solutions on the
basis of the MSA are also presented.
3.2 Ionic Size in Solutions

In order to apply the models which are introduced in this chapter, one needs an
estimate of the size of the ions which make up the electrolyte. This is most easily
done for monoatomic ions, since they are spherical, so that ionic size is known by
determining the ionic radius. Furthermore, accurate interparticle distances are
available from X-ray diraction studies of ionic crystals, and more importantly,
from X-ray and neutron diraction studies of aqueous electrolyte solutions [1].
One of the most widely known and used set of ionic radii are those estimated
by Pauling [2] on the basis of interionic distances in ionic crystals. He noted that
repulsive eects between ions of the same charge depend on the relative size of the
cation and anion in the crystal, and also took into consideration the coordination
number of the ion with oppositely charged neighbors in the crystal lattice. The
results obtained for the alkali metal and halide ions for the case that the coordination number is six (rock salt structure) are summarized in table 3.1.
As the collection of X-ray diraction data became more extensive, it was
possible to describe the electron density distribution in ionic crystals in more
detail. Using these data one can divide up the internuclear distance in the crystal
on the basis of the minimum in the electron density between the two oppositely
charged ions [3, 4]. For example, in the case of NaCl for which the internuclear
distance is 281 pm, the minimum in the electron density leads to radii of 117 pm
for Na and 164 pm for Cl. Radii derived on this basis are larger for cations and
smaller for anions than those of Pauling.
Table 3.1 Radii for the Alkali Metal and Halide Ions for a
Coordination Number of Six Estimated from Crystallographic
Data Together with IonOxygen Atom Distances from
Diffraction Studies of Aqueous Solutions
Ionic Radius / pm
Ion
Ion-Oxygen Distance in
Pauling [2] Shannon and Prewitt [5] Aqueous Solution [1] / pm
Li
Na
K
Rb
Cs
61
96
133
148
166
88
116
152
163
184
210
242
280
315
F
Cl
Br
I
134
181
195
217
119
167
182
206
262
310
334
363
98
Shannon and Prewitt [5] considered a very large collection of crystallographic

data for metal oxides and metal uorides. Included in these data were transition
metal compounds in which the bonding between cation and anion is both electrostatic and covalent in character. They noted that the eective radius for a given
ion depends on its coordination number in the crystal. In addition, for transition
metal ions the radius depends on whether the d electrons in the ion are in a high or
low spin state. By assigning a radius of 126 pm to the O2 ion and 119 pm to the
F ion when they are surrounded by six counter ions, they found that the eective
radii of the alkali metal ions and halide ions are very close to those obtained from
electron density maps for the corresponding crystals. Their results for these ions
are also summarized in table 3.1.
Another source of information directly relevant to ionic size in solution is
X-ray and neutron diraction studies of interatomic distances in aqueous electrolyte solutions [1]. Values of the ionoxygen distance found in these experiments
are given in table 3.1. In the case of the alkali metal ions, the dierence between
the ionoxygen distance and the ShannonPrewitt radius increases from 122 pm
in the case of Li to 131 pm in the case of Cs, the average dierence being
127 pm. The average value is very close to the eective radius of oxygen on the
ShannonPrewitt scale. In the case of Li, the average number of water molecules
around this cation is probably closer to four in concentrated solutions [6]. Thus, a
more appropriate value for the Li ion radius in water is 82 pm. As the atomic
number of the alkali metal cation increases, the average number of associated
water molecules increases and the eective radius in solution changes with respect
to the Shannon and Prewitt estimate accordingly.
In the case of the halide ions, the dierence between the ionoxygen distance
and the ShannonPrewitt radius increases from 143 pm for F, to 157 pm for I,
the average dierence being 149 pm. Assessment of this result is more dicult
because water molecules are oriented around anions via hydrogen bonding (see
g. 3.1). The fact that the average distance between an ion and the oxygen atom in
a water molecule is larger for an anion than a cation of the same size is easily
explained on the basis of the expected dierence in water orientation. Some
increase in the average number of water molecules associated with a given
anion is expected with increase in the anions atomic number.
Studies of pure water have resulted in the conclusion that water has a diameter
of 284 pm in a spherical representation [G3]. Thus, if the electrolyte solution is
represented as a collection of hard spheres, the distance between the center of a
Fig. 3.1 The orientation of a deuterated water molecule at a Li ion and Cl ion as
determined by neutron diraction. (From reference G3, with permission.)
99
K ion and that of an adjacent water molecule is 294 pm. The corresponding
distance for a Cl ion is 309 pm. Examination of these estimates with respect to
the experimental diraction data reveals that they are reasonable. However, if the
hard-sphere model relies on a point dipole description of the water molecule, the
chemical nature of the ionsolvent interactions is clearly neglected.
ShannonPrewitt radii for other monoatomic ions at various coordination
numbers are recorded in table 3.2. The choice of radius is important when estimating the solvation parameters of highly charged ions such as Al3. It is also
important to note that the estimated radii for transition metal ions such as Mn2
and Fe3 depend on the spin state of the d electrons in the ion. This also leads to
important dierences in solvation energies.
Radii have also been tabulated for polyatomic ions [7]. Since these species are
not truly spherical, such radii must be regarded as eective. However, it is useful
to have estimates of eective radii when comparing ionic solvation parameters. In
the following section, methods of determining the solvation parameters of single
ions in innitely dilute electrolyte solutions are considered. This is followed by a
discussion of simple models of ionic solvation in which ionic size is an important
factor.
Table 3.2 Shannon and Prewitt Radii for Monoatomic Ions

Determined from Crystallographic Data for Various
Coordination Numbers* [5]
Ion
Ag
Al3
Ba2
Ca2
Cd2
Co2
Co3
Cr2
Cr3
Cu2
FFe2
Fe3
Ga3
Coordination
Number
6
4
6
6
6
4
6
6
6
6
6
6
6
6
6
4
6
6
6
6
4
6
(ls)
(hs)
(ls)
(hs)
(ls)
(hs)
(ls)
(hs)
(ls)
(hs)
Radius
/pm
129
53
67
150
114
98
109
79
88
67
75
87
96
76
87
117
75
91
69
79
61
76
Ion
Hg2
In3
Li
Mg2
Mn2
Mn3
Na
Ni2
O2Pb2
S2
Se2
Sr2
Tl
Tl3
V2
V3
Zn2
Coordination
Number
6
6
4
4
6
6
6
6
6
4
6
6
6
6
6
6
6
6
6
6
4
6
(ls)
(hs)
(ls)
(hs)
Radius
/pm
116
93
73
63
86
81
96
72
79
113
84
126
132
170
184
130
164
102
93
78
74
89
*hs refers to the high-spin conguration of the d electrons in a transition

metal cation, and ls, to the low-spin conguration.
100
3.3 The Thermodynamics of IonSolvent Interactions

A very simple experiment that has been carried out for many electrolytes in water
is the measurement of the enthalpy associated with the dissolution of the electrolyte, which is often a solid, in water. This process can be either exothermic or
endothermic, and has an enthalpy change which depends on the relative amounts
of electrolyte and water. By studying the enthalpy of solution for one mole of
electrolyte as a function of the number of moles of water, which increase from one
experiment to the next, one can determine the enthalpy of solution associated with
the formation of an innitely dilute solution. In the case of NaCl, the relevant
process is
NaClcry 1H2 Ol ! NaClsl ; 1H2 O
3:3:1
with an enthalpy change equal to 3.9 kJ mol1, where the subscript cry represents the crystalline state, l, the liquid state, and sl, the solution. Under these
experimental conditions, and assuming complete dissociation of the electrolyte,
the value of the enthalpy change reects ionsolvent interactions in solution and
the lattice energy of the NaCl crystal. Since the solution is innitely dilute, ionion
interactions do not play a role in determining the magnitude of the enthalpy
change.
A much more suitable point of reference for assessing ionsolvent interactions
is the unsolvated ion in the gas phase. Thus, one would prefer to know the
enthalpy change associated with the reaction

Na
g Clg 1H2 Ol ! NaClsl ; 1H2 O
3:3:2

where Na
g and Clg are gas phase ions. In order to determine this quantity, the
enthalpy change associated with the reaction

NaClcry ! Na
g Clg
3:3:3
must be found. On the basis of tabulated thermodynamic data [G4], the enthalpy
change for reaction (3.3.3) is 787.4 kJ mol1. It follows that the enthalpy of reaction (3.3.2) is 783.4 kJ mol1. This quantity is called the enthalpy of solvation of
NaCl, sH. On the basis of other thermodynamic experiments, one can determine
the entropy change, sS, and the Gibbs energy change, sG, associated with the
same process.
Thermodynamic data for electrolyte solvation have been found experimentally
for many dierent electrolytes. Ultimately, one would like to be able to analyze
these results further to obtain separate contributions from the cation and anion.
However, that is not possible without making an extrathermodynamic assumption. As a result, a scale of single ion solvation parameters has been dened
relative to those for the H ion. For example, the enthalpy associated with the
process

H
g Clg / H2 O ! HCl; 1H2 O
3:3:4
is 1469.9 kJ mol1. If the enthalpy of solvation of H is dened to be zero, then

the enthalpy of solvation of Cl is 1469.9 kJ mol1. Using this result and the
101
value of sH for NaCl, the enthalpy of solvation of the Na ion is 686.5 kJ mol1.
The resulting scale of solvation enthalpies is called the conventional scale.
Another way of describing solvation parameters on the conventional scale is
by means of an exchange reaction between the ion in question and the proton.
Thus, the solvation parameters for the Na ion are dened by the reaction
Na
g H ; / H2 O ! Na ; 1H2 O Hg
3:3:5
From thermodynamic tables, the enthalpy of formation of H in an innitely

dilute aqueous solution is zero by denition; the same quantity for Na from
tabulated data is 240.1 kJ mol1. On the basis of these results, the enthalpy
associated with reaction (3.3.5) is 686.7 kJ mol1. This result agrees within experimental error with that obtained by comparing the heats of formation of innitely
dilute aqueous solutions of NaCl and HCl.
Estimation of the entropy of solvation requires calculation of the entropy of
the ion in the gas phase. For a monoatomic ion, the main contribution to the
entropy comes from its translational energy. Simple ions formed from the main
group elements have the electronic structure of an inert gas and therefore do not
have an electronic contribution to the entropy. On the other hand, ions formed
from transition metals may have an electronic contribution to the gas phase
entropy, which depends on the electronic conguration of the ions ground
state and of any other electronic states which are close in energy to the ground
state. The translational entropy is given by the SackurTetrode equation, which is
obtained from the solution of the SWE for a particle in a box (see section 2.2)
"
#
5R
2pmkB T 3=2 kB T
R ln
Strans
3:3:6
2
P
h3
Here m is the mass of the ion and the other symbols have their usual meaning. For
standard conditions of 25 C and 1 bar pressure, this equation becomes
Strans 108:856 12:472 ln M
3:3:7
where M is the atomic mass for the ion expressed in grams. The entropies of the
ions in solution have been estimated from a variety of thermodynamic data and
are given in standard tables [G4].
EXAMPLE
Estimate the enthalpy, entropy, and Gibbs energy of solvation of the S2 ion at
25 C and 1 bar using tabulated thermodynamic data together with the electron
anities of sulfur in the gas phase, which are
Sg e ! S
g
H 200:4 kJ mol1
3:3:8
H 456 kJ mol1
3:3:9
and
2
S
g e ! Sg
The enthalpy of vaporization of solid sulfur is 278.8 kJ mol1. The enthalpy

of formation of H in the gas phase is 1536.2 kJ mol1. The enthalpy of formation of S2 in an innitely dilute aqueous solution is 33.1 kJ mol1. The
enthalpy of solvation of S2 is dened by the process
102

2
2H
g Sg ! H2 S; / H2 O
3:3:10
On the basis of the above data, sH for S2 is 3573.7 kJ mol1.
Using the SackurTetrode equation the entropies for H and S2 in the gas
phase are 108.95 and 152.10 J K1 mol1, respectively. The entropy of S2 in
an innitely dilute aqueous solution is 14.60 J K1 mol1 and that for H,
zero by denition. It follows that sS for S2 is 384.61 J K1 mol1.
The Gibbs energy of solvation is obtained using the relationship
s G s H Ts S
3:3:11
On the basis of the above results, s G for S2 is 3459.0 kJ mol1

Values of sG and sS on the conventional scale are summarized in table 3.3.
Large variations are found in these quantities with ionic size and charge but they
are dicult to interpret without conversion to an absolute scale. Various methods
have been described to determine the absolute values of the solvation parameters
but they all require an extrathermodynamic assumption. On the other hand, the
Gibbs energy and enthalpy of formation of small water clusters containing a
single ion may be determined using mass spectrometry [8]. The number of
water molecules in a cluster is small, varying from one to six. By extrapolating
these results to obtain the values of G and H for an innite number of water
molecules in the cluster, one obtains the absolute solvation parameters, sG(abs)
and sH(abs). A method of carrying out this extrapolation has been developed [9]
which leads to the conclusion that sG(abs) is 1104.5 kJ mol1 and sH(abs),
1150 kJ mol1 for H ions at 25 C. The corresponding value of sS(abs) is
153 J K1 mol1. The value of sG(abs) obtained by the mass spectrometric
method is quite close to that estimated on the basis of measurements of the
work function for the H ion in an innitely dilute solution. The latter estimate,
which is 1096 kJ mol1 at 25 C is discussed in more detail in chapter 8.
On the basis of the values of sG(abs) and sS(abs) obtained in the mass
spectrometric experiments one may convert the results obtained on the conventional scale to the absolute scale. These results are also recorded in table 3.3. Of
course, it is an easy matter to obtain the value of sH on either scale using
relationship (3.3.11).
Having obtained absolute values of the thermodynamic properties describing
ion solvation it is interesting to examine various theories for ion solvation. The
best known of these is the Born model, which is described in the following section.
3.4 IonSolvent Interactions According to the

Born Model
The Born model [11] provides a means of estimating the Gibbs energy of solvation
for an ion in an innitely dilute solution. It is based on a continuum description of
the solvent as a uniform dielectric with a relative permittivity of es. The work of
transferring the ion from vacuum to the dielectric medium is estimated on the
basis of the following three-step process: (a) the ion is reversibly discharged in
vacuum; (b) the discharged ion, which is assumed to be a sphere of radius, ri, is
Table 3.3 Solvation Parameters for Monoatomic Ions in Water at

25 C and 1 bar [10]
Conventional Values
Ion
Absolute Values
sG / kJ mol1 sS / J K1 mol1 sG / kJ mol1 sS / J K1 mol1
Cations from the Main Group Elements

H
Li
Na
K
Rb
Cs
Be2
Mg2
Ca2
Sr2
Ba2
Al3
Ga3
In3
Tl
Tl3
Sn2
Pb2
0.0
575.1
680.8
752.6
775.2
798.4
289.3
277.6
600.7
729.9
857.3
1362.4
1370.2
820.7
746.1
822.0
619.1
681.6
0.0
10.7
20.0
56.9
66.1
72.2
48.1
68.8
9.9
20.7
57.3
144.8
165.9
7.9
59.3
40.3
32.5
53.0
1104
529
424
352
329
306
2498
1931
1608
1479
1352
4676
4684
4134
358
4135
1590
1527
153
164
133
96
87
81
354
375
296
285
249
604
625
451
94
499
273
253
84.5
9.5
174.4
29.4
64.0
125.5
74.6
89.0
11.1
57.2
14.4
55.2
23.1
10.7
3946
1933
4544
1868
1957
4429
2017
2095
583
2109
489
2057
1829
1859
543
315
633
335
370
584
380
395
164
363
139
361
329
295
429
304
278
243
1238
115
53
37
14
79
Transition Metal Cations

Sc3
Cr2
Cr3
Mn2
Fe2
Fe3
Co2
Ni2
Cu
Cu2
Ag
Zn2
Cd2
Hg2
632.7
276.0
1230.7
340.7
251.5
1115.8
192.2
113.5
521.1
100.2
615.8
152.2
379.8
349.8
Anions from the Main Group Elements

F
Cl
Br
I
S2
1533.6
1408.5
1382.2
1347.1
3446.6
268.3
205.8
190.1
166.9
384.6
104
transferred from vacuum to the dielectric liquid; and (c) the sphere is reversibly
charged up to become the original ion in the dielectric medium. In order to
estimate the associated work, one must be able to calculate the potential on the
surface of the ionic sphere due to its charge. According to Coulombs law, the
potential due to a charge zie0 at a distance ri is given by
zi e0

3:4:1
4pemed e0 ri
where zi is the ion charge number, e0, the electronic charge, emed, the relative
permittivity of the medium, and e0, the permittivity of free space. The work
done in charging up the sphere from zero charge to a nal charge of zie0 is then
z
i e0
w
0
q
dq
4pemed e0 ri
3:4:2
where q is any intermediate value of the charge. Carrying out the integration one
obtains
w
zi e0 2
8pemed e0 ri
3:4:3
The total work associated with the above three-step process for solvation can be
written
ws wa wb wc
3:4:4
Assuming that wb, the work associated with introducing the uncharged sphere
into the dielectric liquid, is zero, then
ws
zi e0 2
z e 2
i 0
8pe0 ri 8pes e0 ri
3:4:5
where the rst term corresponds to the work associated with discharging the
sphere in vacuum (emed 1) and the second, to charging up the sphere in the
liquid (emed es). The work done per mole of ions is equated to the Gibbs
solvation energy of the ion, which may be expressed as

N z e 2
1
s Gi L i 0
1
3:4:6
es
8pe0 ri
where NL is Avogadros constant. This is the basic result of the Born model.
Having obtained the Gibbs energy of solvation, one may now determine the
other thermodynamic parameters. The entropy is obtained from the temperature
derivative of s Gi :
s S i
ds Gi NL zi e0 2 1 des
dT
8pe0 ri e2s dT
3:4:7
where des/dT is the temperature derivative of the relative solvent permittivity. It

should be noted that the radius of the ion has been assumed independent of
temperature, a reasonable assumption for monatomic ions. Since the solvent
permittivity decreases with increase in temperature, the Born model predicts
105
that the entropy of solvation is negative. This is an expected result, since the
solvation process is imagined to involve reorganization of local solvent molecules
around the ion to stabilize its presence in the medium.
The Gibbs energy and entropy may now be combined to obtain an expression
for the enthalpy of solvation:

N z e 2
1 T de
3:4:8
s Hi s Gi Ts Si L i 0
1 2 s
8pe0 ri
es es dT
The enthalpy of solvation is clearly larger in magnitude (more negative) than the
Gibbs energy because the sign of the entropy and Gibbs energy contributions are
the same.
Estimates of sG and sS according to the Born model using the Shannon and
Prewitt radii are given in table 3.4. From the results, it is clear that the Born
estimates of sGi for the alkali metal cations are all too large in magnitude, the
dierence for the smallest ion, Li, being the largest. In the case of the halide
anions, the dierence between the Born estimate and experiment is not as great,
the Born estimate always being larger in magnitude. On the other hand, the Born
estimates of the entropy change accompanying solvation are usually too small.
The generally accepted explanation for the failure of the Born model is that the
structure of the solvent is signicantly altered near the ion by the eect its eld has
on the local water molecules. This eect is especially important for cations, which
are stabilized by the electron density on the oxygen atoms of water. One can
describe the local disruption of water structure as resulting in an eective decrease
in the dielectric constant of the solvent near the ion. As a result, the work associated with the charging the ionic system in solution is less negative, and the Born
estimates are too high in magnitude. In the case of anions, hydrogen bonds from
surrounding water molecules are important in stabilizing the ion. Disruption of
the local solvent structure also occurs but it is not as severe. As a result the Born
Table 3.4 Experimental Value and Estimates According to
the Born Model and Mean Spherical Approximation for
the Gibbs Energy and Entropy of Solvation of the Alkali
Metal Cations and Halide Anions at 25 C
Gibbs Energy s Gi / kJ mol1
Entropy s Si / J K1 mol1
Ion
Expt*
Born
MSA
Expt*
Born
MSA
Li
Na
K
Rb
Cs
F
Cl
Br
I
529
424
352
329
306
429
304
278
243
779
591
451
421
373
576
411
377
333
483
403
333
316
288
396
310
291
264
164
133
96
87
81
115
53
37
14
46
35
27
25
22
34
24
22
20
199
143
101
92
78
138
89
79
66
*Absolute value of the Gibbs energy of solvation from table 3.3.
106
estimates for anions are closer to the experimental quantities than for cations. It is
also apparent from table 3.4 that the Born model underestimates the magnitude of
the entropy change resulting from ion solvation especially for the smaller ions.
This is specically because it does not take into account the local restructuring of
water, which increases the orderliness of the system near the ion.
Many attempts have been made to improve the Born description of ion solvation. Most of these rely upon continuum descriptions of the solvent in which the
permittivity varies from a low value near the ion to the bulk value farther away.
This variation is described mathematically either as two or three regions with a
constant permittivity in each, or as a solvent with a continuously varying permittivity over a region of a few molecular diameters thick. Macroscopic concepts
such as the permittivity are not really valid at molecular dimensions. For this
reason these models are not considered further here. Instead, in the next section, a
model based on a discrete description of the electrolyte solution as a collection of
hard spheres is discussed.
3.5 Ion-Solvent Interactions According to the Mean

Spherical Approximation
In the Born model, the solvent was considered to be a continuum with uniform
dielectric properties, right up to an ion, which carries a discrete charge. As soon as
one recognizes the molecular nature of the solvent, one can imagine how the local
solvent structure near an ion is disrupted by the high electrical eld near the ion.
Another way of representing a polar solvent is as a collection of close-packed hard
spheres with centrally located point dipoles. Such a description ignores the chemical composition of the solvent molecule but takes into account its most important electrostatic feature, namely, that it has a dipole moment. No consideration is
given to the presence of higher-order moments, or to the possibility that the
solvent may also be involved in hydrogen bonding.
The simplest non-primitive model for representing a dilute electrolyte solution
is the mean spherical approximation (MSA). In its present application, the system
consists of hard spheres of two sizes, one size for the solvent molecules, and
another for the ions, both cations and anions. Since this model is only applied
here in the limit of innite dilution the restriction that cations and anions have
equal radii is of no consequence. On the basis of the discussion in chapter 2, the
MSA requires that the pair correlation function for any two spheres goes to zero
within the reference sphere because they are hard. Outside of any sphere, the
direct correlation function is set equal to the interaction energy between the
spheres divided by kBT. Solution of the statistical mechanical problem involves
applying the OrnsteinZernike (OZ) equation to obtain the pair correlation function outside of a given sphere. Since the system consists of a mixture of ions and
dipoles, the OZ equations appropriate for a mixture must be used with consideration of the ionion, iondipole, and dipoledipole interactions that are present.
Once the pair correlation functions have been found, the thermodynamic properties of the system may be calculated. In the following presentation, the equations
dening the problem are given as well as the solution, no attempt being made to
107
present all of the mathematics required to obtain the solution. The interested
reader is referred to the original literature [1214] for the mathematical details.
In the following, the radius of a solvent molecule represented as a sphere is
denoted by rs, and that of either a cation or anion by ri. The conditions which
dene the direct correlation function are the following.
For ionion interactions,
gij r 0;
r < 2ri
3:5:1
and
cij r buij r
zi zj be20
;
4pe0 r
r > 2ri
3:5:2
where zi and zj are the charge numbers on ions i and j, and b 1=kB T.
For iondipole interactions,
gid r; o 0;
r < ri rs
3:5:3
and
cid r; o
zi e0 pbso
sr
;
4pe0 r2
r > ri rs
3:5:4
where p is the dipole moment of the solvent molecule, so , a unit vector in the
direction of the dipole, sr , one in the direction r, and o, an angular variable
dening the dipole orientation.
Finally, for dipoledipole interactions,
gdd o1 ; r; o2 0;
r < 2rs
3:5:5
and
cdd o1 ; r; o2
bp2
s1
sr s2
sr s1 s2 ;
4pe0 r3
r > rs
3:5:6
where the dipole directions are dened by unit vectors s1 and s2 . Using the
OrnsteinZernike equation (2.6.1), one can now determine the total correlation
function for each of these interactions.
Only one result from this model is presented here, namely, the Gibbs energy of
iondipole interactions in the limit of innite dilution, Gid . Further, if one identies this with the Gibbs solvation energy, the result is

N L zi e0 2
1
1
3:5:7
s G i
1
8pe0 ri
es
1 rs =ls ri
where ls is the polarization parameter for the solvent. This parameter is obtained
from the dielectric properties of the solvent, and on the basis of the work of
Wertheim [15] is given by
l2s 1 ls 4 16es
3:5:8
By using the bulk dielectric permittivity, the value of ls includes consideration not
only of the dipole moment of the solvent but also its polarizability and other
features which are necessary to describe iondipole and dipoledipole interac-
108
tions. In the case of water at 298 K, the value of ls estimated from es (78.46) is
2.65. It should be noted that the role of dipoledipole interactions in determining
s Gi is ignored in equation (3.5.7).
A more convenient way of writing the Gibbs solvation energy is

N z e 2
1
1
s G i L i 0
3:5:9
1
8pe0
es
r i ds
where ds is equal to rs/ls and depends only on the nature of the solvent. If ds is
zero, the expression for s Gi becomes equal to that given by the Born model. It is
clear that ds is a correction to the ionic radius and results in estimates of s Gi
which are considerably smaller than the Born estimates. A reasonable value of rs
for water is 142 pm, so that the value of ds is 53.6 pm. This is not insignicant with
respect to the radii of simple monoatomic ions (see tables 3.1 and 3.2).
EXAMPLE
Acetonitrile is a polar solvent with a relative permittivity of 35.9. It may be

represented as a hard sphere with a diameter of 427 pm. Estimate the Gibbs
energy of solvation of Na in acetonitrile according to the Born and MSA
models. Compare the theoretical estimates with the experimental estimate
given that the Gibbs energy of transfer for Na from water to acetonitrile is
15.1 kJ mol1.
The Gibbs energy of solvation according to the Born model is given by
equation (3.4.6). The constant NLe02/8pe0 is equal to 6:945 105 J m mol1.
The radius of Na according to Shannon and Prewitt is 116 pm (table 3.1).
The factor (1 1/es) is equal to 0.972. The resulting value of s Gi is
581.9 kJ mol1.
In order to estimate s Gi according to the MSA, one must rst nd the
value of ls using equation (3.5.8). Solving this cubic equation, one obtains the
result that ls is equal to 2.26. Using a radius of 213.5 pm for acetonitrile, the
corresponding value of ds is 94.5 pm. Thus, the MSA estimate of s Gi for Na
is 320.7 kJ mol1 (equation (3.5.9)).
The absolute value of the Gibbs energy of solvation of Na in water according to experiment is 424 kJ mol1 (table 3.4). After transfer to acetonitrile,
s Gi is 409 kJ mol1.
The MSA estimate of s Gi is closer to the experimental value than the Born
estimate.
The entropy of solvation is obtained from the temperature derivative of s Gi
[16]. Recognizing that ls depends on temperature, the resulting expression is
"
#

ds Gi N L zi e0 2 1
1
des
1
1 dds
s S i

1
3:5:10
es dT
dT
8pe0
e2s ri ds dT
ri ds 2
where the temperature coecient of ds is given by
dds
r dl
s2 s
dT
ls dT
3:5:11
109
dls/dT can be determined by dierentiation of the Wertheim equation with

respect to temperature, which gives

dls
ls 1 ls 1 des
3:5:12
3ls 1 2es dT
dT
Since the dielectric permittivity decreases with temperature, the coecient dls/dT
is negative, and dds/dT is positive. As a result, both terms in the square brackets in
equation (3.5.10) give negative contributions to s S i . Given that des/dT for water
is 0.360 K1, the value of dls/dT is 2:48 103 K1 and dds/dT is 0.050 pm
K1.
Values of s Gi and s S i calculated according to the MSA for the alkali metal
cations and halide anions are recorded in table 3.4. The estimates of s Gi for the
cations are much better than those by the Born model and give acceptable results.
On the other hand, the results for the anions are signicantly lower than the
experimental values. The MSA estimates of s S i for most ions are too high in
magnitude. In general, the MSA provides a simple extension of the Born continuum model and gives better values for the fundamental thermodynamic quantities relating to ionic solvation. The major reason the simple MSA model fails is
that it has ignored dipoledipole interactions in the estimation of s Gi and s S i .
Dipoledipole interactions in the immediate vicinity of an ion are expected to be
repulsive in character and reduce the estimates of s Gi ands S i from those given
by equations (3.5.9) and (3.5.10).
An expression for the dipoledipole contribution to the solvation energy has
been derived on the basis of the MSA [12, 13] but it is generally felt to overestimate the repulsive eect. When one analyzes the functional form of the dipole
dipole term, it can be shown that it is similar to that for the iondipole contribution. As a result, a simple way of writing an improved expression for the Gibbs
solvation energy is

N L zi e0 2
1
1
1 f dd
s Gi
3:5:13
8pe0 ri ds
es
where fdd is a term depending on dipoledipole interactions which is much less
than one.
Another problem with the MSA is that it does not distinguish between the
solvation of cations and anions of the same size. Thus, although the K and F
ions have approximately the same radius, the F anion is more strongly solvated
than the K cation (table 3.4). This can be understood in terms of the eect that
each ion has on local water structure. The K ion disrupts this structure more so
that the stabilizing eect of the local iondipole interactions is oset by the work
done to break up the water structure, that is, to disrupt attractive dipoledipole
interactions and hydrogen bonding between local water molecules. This means
that the parameter ds should be dierent for cations and anions in the same
solvent (table 3.5).
It is interesting to examine the data for ions from the main group elements on
the basis of equation (3.5.13) using the absolute values of s Gi given in table 3.3.
In order to do this, the equation is rearranged as follows:
110
Table 3.5 MSA Parameters Relating to the

Solvation of Monoatomic Ions of the Main Group
Elements in Water on the Basis of Equation (3.5.14)
Polarization
Parameter, ds / pm
DipoleDipole Interaction
Parameter, fdd
49
0
0
0.26
Cations
Anions
z2i
ks ri ds
s Gi
3:5:14
where
ks
N L e20 1
8pe0 es
f dd es 1
3:5:15
It follows that a plot of z2i s Gi 1 should be linear against the ionic radius ri,
and that fdd and ds can be calculated from the slope and radius, respectively.
The quality of the t of equation (3.5.14) to the experimental data is illustrated
in g. 3.2. It is clear from this plot that the behavior of cations and anions is
signicantly dierent. On the other hand, the linear plots obtained for each species are very good. The radius correction parameter ds for cations is slightly
smaller than the value predicted by the Wertheim equation but that for anions
is considerably smaller. This is clearly a result of the fact that cations disrupt the
structure of water much more than anions. On the other hand, dipoledipole
Fig. 3.2 Plots of z2i s Gi 1 against ionic radius ri for the alkali metal cations (*),
alkaline earth metal cations (&), halide anions (~) and the sulde anion (!). The parameters ds and f dd are determined from the intercept and slope of these plots, respectively,
according to equation (3.5.14). The plot for cations has been shifted vertically by 1 mol
MJ1 for the sake of clarity.
111
interactions are negligible for water molecules arranged around a cation, whereas
they are signicant in the solvation of anions. The latter result reects the role of
hydrogen bonding in anion solvation. In assessing the results one must keep in
mind the fact that it has been assumed that the appropriate radii for all ions are
those estimated for a coordination number of six. This is probably not correct for
smaller ions such as Li and Mg2. For these ions, the radii estimated for a
coordination number of four may be more appropriate. If smaller radii are chosen
for the smaller ions, the characteristics of the plots in g. 3.2 change. This would
aect both the polarization parameter ds and the dipoledipole interaction parameter fdd.
The above empirical model can be extended to estimation of the entropy of
solvation. When equation (3.5.13) is dierentiated with respect to temperature,
one must consider the temperature variation of ds and fdd as well as the temperature dependence of es. However, the number of data points is too few to determine
all the temperature coecients which arise.
In summary, the empirical approach to ionic solvation based on the MSA is
quite successful for monoatomic ions of the main group elements. It helps one to
understand the important dierences between the way cations and anions are
solvated in water. It can also be applied to other ions, including polyatomic
ions, provided the solvation is essentially electrostatic in character. Thus, one
may estimate eective radii for anions such as nitrate and perchlorate from the
Gibbs solvation energy using the value of ds calculated for the halide ions.
Considering the simplicity of the model, it provides an useful means of understanding the thermodynamics of solvation.
3.6 The Thermodynamics of Electrolyte Solutions

Electrolyte solutions are non-ideal by their nature. This follows from the fact that
the interactions among the three species in solution, namely, cations, anions, and
solvent molecules, are all dierent. Ions with the same charge strongly repel one
another, whereas oppositely charged ions attract one another. Interactions of ions
with the solvent molecules depend on ionic charge. From a fundamental point of
view one would like to understand each interaction in detail at the molecular level.
However, thermodynamically one may study the properties of the electrolyte as a
whole, or of the solvent. In this section, the thermodynamic conventions for
dealing with the properties of electrolytes are presented. As will be seen, these
conventions depend largely on the fact that cations and anions are always present
at concentrations such that the net charge in the solution is zero.
In general one distinguishes two types of electrolyte solutions. The rst is
formed from strong electrolytes, also called ionophores. Strong electrolytes exist
as ions in the solid state, a well-known example being NaCl. When they dissolve
in a polar solvent such as water, the individual ions are stabilized by iondipole
interactions with solvent molecules. Otherwise the properties of the system reect
long-range electrostatic ionion interactions. On the other hand, a weak electrolyte or ionogene is a compound which is molecular in nature as a pure substance.
It forms ions by interaction with water molecules. A familiar weak electrolyte is
112
acetic acid, which interacts with water to form solvated protons and acetate ions.
Usually, dissociation of the molecule into ions is not complete, so that a dissociation constant is required to determine the ionic concentrations in solution.
It is important to note that ionophores are not always completely dissociated.
For example, when NaCl is dissolved in a solvent of lower relative permittivity,
such as methanol, it is ion paired to some extent. The thermodynamics of systems
with ion pairing is considered separately in section 3.10. Under these circumstances the ionophore behaves in the same way as a weak electrolyte. On the
other hand, all ionogenes are not weak electrolytes. For example, HCl, which is
a molecule in the gas phase, is completely dissociated in water and therefore is a
strong electrolyte. Acetic acid is completely dissociated in liquid ammonia, which
is a much stronger base than water. Thus, the solvent plays an important role in
determining the extent of electrolyte dissociation in solution. In the following
discussion the traditional terms, strong and weak electrolytes, are used.
A. Strong Electrolytes
Consider a simple 11 electrolyte MX which is completely dissociated in dilute
solutions. The chemical potential of MX can be written as
mMX mMX RT ln aMX
3:6:1
where aMX is the activity of the electrolyte. Activity is related to the concentration
of the electrolyte on the appropriate scale. Although one cannot directly measure
separately the cationic activity aM and anionic activity aX, one can consider these
quantities conceptually. They are dened by the relationship
aM aX aMX
3:6:2
and are not necessarily equal, in spite of the fact that the concentrations of M and
X must be equal. The electrolyte concentration is commonly expressed either in
terms of molality or molarity. Using the former, the activity of ion i is given by
ai gi m e
3:6:3
where gi is the activity coecient on the molal scale and me, the electrolyte
molality. The mean ionic activity coecient is dened by the equation
g2 gM gX
3:6:4
Since individual ionic activity coecients cannot be measured experimentally,

only the mean quantity is tabulated. It follows that the activity of a 11 electrolyte
is given by
3:6:5
aMX g2 m2e
and its chemical potential by

2RT ln g me
mMX mMX
3:6:6
The concentration dependence of g is an important feature involved in the

experimental and theoretical evaluation of electrolyte behavior. The chemical

potential of the standard state, mMX
, is that for a hypothetical one-molal solution
in which all real interactions are imagined to be absent (g 1.00). Thus, the
113
standard state for the electrolyte solution is chosen on the basis of the Henry law
convention (see section 1.11).
One may also write expressions for the chemical potential of the electrolyte on
the molarity and mole fraction scales. In the former case, the expression is
mMX mMX 2RT ln y ce
3:6:7
mMX
is the standard chemical potential and y, the corresponding activity

where
coecient. On the mole fraction scale, one writes
mMX mMX 2RT ln f xe
D
3:6:8
where mMX is the standard chemical potential, f, the corresponding activity

coecient, and xe, the mole fraction of the electrolyte. As with the molal scale,
the standard states are dened with respect to hypothetical ideal solutions of unit
concentration. The mole fraction scale is not often used for electrolytes for the
obvious reason that the electrolyte concentration cannot be varied over the whole
range of mole fractions because of its limited solubility. On the other hand, the
molar scale is popular. In order to convert values of activity coecients which are
usually tabulated on the molal scale [G1] to the molar scale, one needs to know
the concentrations on the two scales, ce and me, and the density of the pure
solvent, rs . The relationship between the activity coecients is
c
g e y
3:6:9
me rs
The above equations for a 11 electrolyte are easily extended to a more complex electrolyte MaXb. Dissociation of this electrolyte give cations Mb and
anions Xa-:
Ma Xb ! n Mb n Xa
3:6:10
The stoichiometric numbers n and n give the number of moles of cations and
anions, respectively, resulting from the dissolution of one formula of electrolyte.
Their sum is designated as n:
n n n
3:6:11
On the basis of equation (3.6.2) the electrolyte activity is

ae a n a n
3:6:12
Each of these activities is the product of an activity coecient and a concentration. On the molality scale, one may write
ae g m n g m n
3:6:13
m n m e
3:6:14
m n m e
3:6:15
Now,
and
The mean molality m is dened on the basis of a geometric average so that
114

n
m n m n m n n nn mne
3:6:16
Now one denes a mean activity coecient using the equation

g n g n g n
3:6:17
It follows that the electrolyte activity is

ae n n n n g n mne
3:6:18
One may also dene a mean electrolyte activity, a , which is given by

n
a n nn 1=n g me
3:6:19
It is easily seen that this expression yields the correct result for a 11 electrolyte
when n and n are set equal to unity.
EXAMPLE
A 0.1 m solution of MgCl2 in water has a density of 1.0047 g mL1 at 258C. The
mean ionic activity coecient on the molal scale is 0.528. Calculate the mean
activity and electrolyte activity on this scale. Repeat the calculations for the
molarity scale.
For this electrolyte, the concentration of Mg2 ions is 0.1 m n 1 and
that of Cl ions 0.2 m n 2. Thus the mean ionic molality on the basis of
equation (3.6.16) is
m 1 22 1=3 me 41=3 0:1 0:1587
3:6:20
The mean activity is

a g m 0:528 0:1587 0:0838
3:6:21
The electrolyte activity on the molality scale is

ae a3 0:08383 5:89 104
3:6:22
The molecular mass of MgCl2 is 95.22 g. Thus, 9.522 g of MgCl2 are dissolved in 1000 g of water to form a 0.1 m solution. Given that the density of the
solution is 1.0047, its volume is
V
9:522 1000
1004:8 mL
1:0047
3:6:23
0:1
0:0995 M
1:0048
3:6:24
Its molar concentration is

ce
The mean ionic concentration is

c 41=3 0:0995 0:1579 M
3:6:25
1
Given that the density of pure water at 258C is 0.997 g mL , the activity
coecient on the molar scale (equation (3.6.9)) is
y
me rso g 0:1 0:997 0:528

0:529
0:0995
ce
Thus, the mean electrolyte activity is
3:6:26
a y c 0:529 0:1579 0:0835
115
3:6:27
The electrolyte activity on the molarity scale is

ae a3 5:83 104
3:6:28
It is important to remember that the mean ionic concentration and activity

are based on the geometrical mean of the concentrations of the two ions in
non-symmetrical electrolytes, not on their arithmetic mean.
B. Weak Electrolytes
In order to develop the thermodynamics of a weak electrolyte, one extra parameter is required, namely the dissociation constant Kdiss . Using the same terminology as above, the dissociation process for a weak 11 electrolyte is
MX (
+ M X
The equilibrium constant for this process is
a a
Kdiss M X
aMX
3:6:29
3:6:30
where aM and aX are the activities of the ions, and aMX , that of the undissociated
molecule. On the molality scale, this may be written as
Kdiss
g2 m2
gMX mMX
3:6:31
where g is the mean ionic activity coecient on the molality scale, m , the mean
ionic molality, gMX , the activity coecient of the molecular species, and mMX , its
molality. This relationship is often written in terms of the fraction of weak electrolyte which is dissociated, namely, ai . If me is the stoichiometric molality of the
weak electrolyte, that is, the molality of the undissociated molecule plus that of
the dissociated form, then equation (3.6.31) bcomes
Kdiss
g2 a2i me
gMX 1 ai
3:6:32
An expression can now be written for the chemical potential of the weak
electrolyte system. On the basis of the composition of the solution

mMX ai m
M ai mX 1 ai mMX
m
M
3:6:33
m
X
and
are the chemical potentials of the free ions and mMX , that of the
where
undissociated molecular. Writing each chemical potential in terms of its standard
value and the corresponding activity, one obtains
mMX mMX ai RT ln aM aX 1 ai RT ln aMX
3:6:34
mMX ai mM ai mX 1 ai mMX
3:6:35
where
In these equations, mi is the standard potential of species i and ai , its activity.
Finally, from equilibrium (3.6.29), one may write that
116

mMX m
M mX
3:6:36
RT ln Kdiss mMX mM mX
3:6:37
so that
By substituting equations (3.6.30) and (3.6.37) into equation (3.6.34), a much
simpler expression for mMX is obtained:
mMX mM mX RT ln aM aX
3:6:38
This result shows that the chemical potential of the weak electrolyte system may
be expressed in terms of the activities of the ions only, without explicitly including
the activity of the undissociated molecule. Equation (3.6.38) is no dierent in
form from those for a strong electrolyte (equations (3.6.1) and (3.6.2)). Of course,
the activities of the ions are much less for the weak electrolyte than those for the
strong electrolyte for a given molality. Thus, on the basis of the present analysis
for a weak electrolyte
aM aX a2i g2 m2e
3:6:39
When these equations are written in terms of the concentration (molarity) the
results are
mMX mM X RT ln aM aX
3:6:40
aM aX a2i y2 c2e
3:6:41
where
mM
mX
and
are the standard potentials of the two ions on the molarity scale, y ,
the mean ionic activity coecient, and ce , the molarity of the weak electrolyte
system.
3.7 The Experimental Determination of Activity

Coefcients for Electrolytes
In general, one may distinguish two methods of determining the activity of an
electrolyte in solution. One type is based on the direct measurement of electrolyte
activity. The most popular technique in this category involves electromotive force
measurements with electrochemical cells. This subject is not discussed here, but
later in the chapter dealing with electrochemical equilibria (chapter 9). The second
type of measurement involves determination of the activity of the solvent, for
instance, by measuring its vapor pressure. These activity data are used to calculate
the activity of the electrolyte on the basis of the GibbsDuhem relationship.
Methods related to determination of the solvents vapor pressure include the
determination of the lowering of the melting point, and the elevation of the
boiling point. The latter two techniques have the inconvenience that the activity
coecients are determined at the freezing point or boiling point of water, and not
at room temperature or the usual experimental temperature of 258C. Thus, the
most often used technique is the direct determination of the vapor pressure of the
solvent as a function of electrolyte concentration at 258C. The experimental techniques have been described in detail by Robinson and Stokes [G1].
117
Having obtained the solvent vapor pressure, one may immediately relate it to
the activity of the electrolyte using the GibbsDuhem relationship. Thus,

P
3:7:1
ne d ln ae ns d ln as ns d ln s
Ps
where the subscript e denotes electrolyte, and the subscript s, solvent.
Furthermore, if one works with molalities, such that ne is equal to me and ns to
1000 M 1
s where Ms is the molecular mass of the solvent, then
1000
nme d ln g me
d ln as
3:7:2
Ms
At this point, it is convenient to introduce a quantity called the osmotic coecient, j. It is dened according to the equation
1000
ln as
3:7:3
nme j
Ms
One often nds values of j tabulated for electrolyte solutions as well as the mean
activity coecient g [G1, G5]. Taking the total derivative of equation (3.7.3), one
obtains
1000
d ln as
3:7:4
nme dj nj dme
Ms
Combining this with equation (3.7.2), one may write
d ln g dj j 1 d ln me
3:7:5
Remembering that both g and j go to unity as me goes to zero, one obtains

m
e
3:7:6
ln g j 1 j 1 d ln me
0
This relationship is used to calculate mean ionic activity coecients using values
of the osmotic coecient, j. Alternatively, one may calculate the osmotic coecient given the mean ionic activity coecient. From equations (3.7.2) and (3.7.3)
dnme j
1000
d ln as nme d ln g me
Ms
3:7:7
Integrating from zero molality, one obtains

m
e
nme j
nme d ln g me
3:7:8
or
1
j1
me
EXAMPLE
m
e
me d ln g
3:7:9
The activity of water (Pw/Pw ) in a 6:4 103 m NaCl solution is 0.9997754.

Estimate the osmotic coecient and mean ionic activity coecients for the
118
electrolyte. In more dilute solutions, it may be assumed [G1] that the relationship between the osmotic coecient and molality is
j 1 0:330 m1=2
e
3:7:10
From equation (3.7.3), the osmotic coecient is

j
1000
1

ln 0:9997754 0:9738
18:02 2 6:4 103
3:7:11
Note that the value of the osmotic coecient agrees with that estimated from
equation (3.7.10).
Combining equations (3.7.6) and (3.7.10), one obtains
m
e
ln g 0:0262
0:330 m1=2
e d ln me
0
m
e
0:0262
0:330 me1=2 dme
3:7:12
0:0262 2 0:330 me1=2 0:0790

The value of g is 0.924.
Values of the water activity and osmotic coecient for NaCl solutions in water
for concentrations up to 6 m are shown in g. 3.3. First, it is apparent that the
vapor pressure of water decreases continuously with increase in salt concentration
as one would expect. The value of aw at the most dilute concentration considered
(0.1 m) is 0.99665. Measurements of vapor pressure lowering for lower concentrations are extremely dicult. The concentration dependence of the osmotic coecient j reported in the same gure is quite dierent from that of the water activity.
This quantity at rst decreases, and then rises to values which are signicantly
Fig. 3.3 Plots of the water activity, aw, and corresponding osmotic coecient, j, for
aqueous solutions of sodium chloride at 25 C.
119
greater than unity. Tables of the osmotic coecients for a large number of electrolytes in water at 25 C are available in the monograph by Robinson and Stokes
[G1], and in the more recent compilation by Lobo [G5].
In order to calculate the mean ionic activity coecient, g, from the osmotic
coecient, j, one must perform a numerical integration to evaluate the integral in
equation (3.7.6). The value of the integrand, chosen to be 1 j=me , is shown as
a function of electrolyte molality using the NaCl data in g. 3.4. This function is
clearly nite in the limit that me goes to zero, and is equal to one. As a result, there
is no problem in evaluating the integral by numerical techniques, and thereby
estimating ln g.
In many cases, it is more convenient to use activity coecients on the molarity
scale. Not only is molarity more commonly used as a concentration unit in
chemistry but values of y are more directly related to the results of statistical
mechanical theories of electrolyte solutions discussed later in this chapter. For a
given molality, me, one must calculate the corresponding molarity, ce, using the
relationship
ce
1000me rs
me Me 1000
3:7:13
where rs is the density of the solution and M e, the molecular mass of the electrolyte. The solution density is not related in a simple way to either the molality or
molarity. It has been tabulated [G5] for many electrolyte solutions but often not
at the specic concentrations of interest. In this regard, there is a useful empirical
relationship which can be used to describe the variation in solution density with
molality [G6]:
ln rs ln rs
Ae me Me
me Me 1000
3:7:14
Fig. 3.4 Plots of the function (1j)/me, where j is the osmotic coecient against solute
molality, me, for solutions of sodium chloride in water at 25 C.
120
Table 3.6 Values of the Constant

Ae in Equation (3.7.14) which
Relates the Solution Density to Its
Molality for Aqueous Solutions of
the Alkali Metal Halides
Anion
F
Cl
Br
I
1.0260
0.8471
0.9710
1.2595
0.5667
0.6938
0.6323
0.7748
0.8308
0.8248
0.8388
0.7624
0.9077
0.8536
0.8697
0.9146
0.8310
0.848
0.6420
Cation
Li
Na
K
Rb
Cs
Values of the constant Ae determined via least-squares ts of density data for

aqueous solutions of the alkali metal halides are given in table 3.6. Similar ts
may be made for other electrolyte solutions so that the conversion of molality to
molarity is easily carried out. Once the concentration units are known, values of
y are easily calculated from g using equation (3.6.9). These quantities are given
in table 3.7 for the NaCl system for concentrations up to 2.4 m.
The above results illustrate the importance of non-ideality for electrolyte solutions and also of the use of the GibbsDuhem relationship in obtaining electrolyte
Table 3.7 Mean Activity Coefcients

for the Aqueous NaCl System on the
Molality and Molarity Scales
Molality/me
Molarity/ce
0.010
0.020
0.050
0.100
0.200
0.300
0.500
0.600
0.700
0.800
0.900
1.000
1.200
1.400
1.600
1.800
2.000
2.200
2.400
0.903
0.872
0.822
0.779
0.735
0.710
0.681
0.673
0.667
0.662
0.659
0.657
0.654
0.654
0.657
0.661
0.668
0.675
0.683
0.010
0.020
0.050
0.100
0.199
0.298
0.494
0.592
0.689
0.786
0.883
0.979
1.170
1.360
1.548
1.734
1.919
2.101
2.282
0.900
0.869
0.820
0.780
0.738
0.714
0.687
0.680
0.675
0.672
0.670
0.669
0.669
0.671
0.677
0.684
0.694
0.705
0.716
121
activity coecients from vapor pressure measurements for the solvent. Although
activity coecients can be measured directly for some electrolytes, they can
always be obtained indirectly from the solvent activity.
In the next section a model developed to account for non-ideality is considered
and its ability to predict experimental results assessed.
3.8 IonIon Interactions According to the

ckel Theory
DebyeHu
The rst successful theory of electrolyte solutions was proposed by Debye and
Huckel. It is based on a model in which the ions are represented as point charges
embedded in a dielectric continuum. The model uses fundamental electrostatics to
calculate the average local potential at any given ion due to its ionic atmosphere
[17]. The main problem is to determine a distribution function for the ions in the
atmosphere. This is achieved by assuming that the probability of nding an ion at
a distance, r, from the central ion is given by the Boltzmann distribution law
which accounts for the randomizing eects of thermal motion in the presence
of the electrical eld. The potential at the reference ion may then be used to
calculate the excess Gibbs energy due to the ionion interactions present in the
electrolyte solution. In the DebyeHuckel model, solution non-ideality is attributed to the ionion interactions and related to the excess Gibbs energy through
the activity coecient, yi for ion i.
It is important to recognize that the average potential in a uniform phase, such
as an electrolyte solution, is constant in the absence of an external eld, such as an
electrical eld. However, the problem solved by Debye and Huckel is the estimation of the average local potential which is dierent from the average potential.
The local potential is sometimes called the micropotential, whereas the average
potential is called the macropotential. It will be seen that the micropotential
uctuates in the solution, being more positive at a cation and more negative at
an anion. The theory presented below gives an estimate of how the ionic atmosphere aects these uctuations.
Two equations are required to solve the problem in question. The rst is the
Poisson equation, which describes how the micropotential, , varies in space. As
a partial dierential equation it is written as
r2 div grad
rz
es e0
3:8:1
where rz is the average charge density at a distance, r, from the central ion. It is
derived by combining two fundamental laws of electrostatics, namely, Coulombs
law and Gauss law. In vector notation, these are
E grad
3:8:2
div D rz
3:8:3
and
122
where E is the electrical eld and D, the electric displacement (D ee0 E) (see
Appendix B). The Laplacian of , r2 , describes the spatial variation of the
potential and in Cartesian coordinates is given by
r2
@2 @2 @2
2 2
@x2
@y
@z
3:8:4
However, the present problem is more conveniently solved in spherical coordinates because an estimate of the average charge density at a distance r is sought.
In this case, depends only on r and not on the Eulerian angles, which dene an
exact location in spherical coordinates. As a result, equation (3.8.1) can be written
as

1 d 2 d
r
2
3:8:5
r 2
r
z
dr
dr
e
r
s e0
The second equation is obtained by applying Boltzmann statistics to estimate
the charge density, rz. The local concentration of ion i at a distance r from the
central reference ion, cri , diers from its average concentration, ci , because the
potential, , is dierent from the average potential. Using Boltzmann statistics
these quantities are connected by the relationship
cri ci expzi f
3:8:6
where ci the concentration in moles per cubic meter and f e0/(kBT) F/(RT). If
the central ion is a cation, and is positive, then the local concentration of other
cations is lower and that of anions higher. The extent to which the local concentration diers from the average concentration depends on the relative values of
the electrostatic energy, zie0, and the thermal kinetic energy, kBT. For a simple
electrolyte solution containing one cation with charge number z and one anion
with charge number z, the charge density at distance r is
rz z Fc expz f z Fc expz f
3:8:7
where F is the Faraday constant (F NLe0). In a more general situation where the
solution contains more than two types of ions, one may write
rz
n
X
zi Fci expzi f
3:8:8
i1
Combination of equations (3.8.5) and (3.8.8) gives the PoissonBoltzmann

equation for the system:

n
X
1 d 2 d
zi Fci
r

expzi f
3:8:9
dr
ee
r2 dr
i1 s 0
There are two important problems with this equation. First, it is a non-linear
dierential equation because of the exponential terms in and therefore cannot
be solved in a simple way. The second problem is that the solution of the statistical problem is not consistent with the fundamental laws of electrostatics. These
problems were overcome by Debye and Huckel [17] in a fairly simple way. In the
case that the electrostatic energy is small in comparison to the thermal kinetic
123
energy, the exponential term in equation (3.8.9) can be expressed by the rst two
terms in its Taylor series expansion. Thus, the expression for the charge density
becomes
rz
n
X
zi Fci
i1
n
X
z2i Ffci
3:8:10
i1
The rst term on the right-hand side of equation (3.8.10) gives the sum of the ionic
charges in the solution, which must add to zero. From the second term, one
denes an important quantity used in assessing the properties of electrolyte solutions, namely, the ionic strength, I. The denition is
I
n
1X
z2 c
2 i1 i i
3:8:11
For a 11 electrolyte, the concentrations of cations and anions are equal to the
overall bulk concentration ce and the charge numbers zi are 1 and 1. Then, the
ionic strength is also equal to the bulk concentration ce . For a 21 electrolyte,
such as ZnCl2, the cationic concentration is equal to ce , whereas the anionic
concentration is 2ce . In this case, the ionic strength is 3ce and recognizes the
higher charge on the cations. For a 22 electrolyte, such as ZnSO4, the ionic
strength is four times the bulk concentration. Of course, the ionic strength may
be expressed in other concentration units, including the molality me. It is an
important property of an electrolyte solution, especially in evaluating the extent
of departure of its properties from ideality.
When the DebyeHuckel approximation for the charge density rz is used, the
PoissonBoltzmann equation can be written as

1 d 2 d
2FfI
r
k2
3:8:12
2 dr
dr
es e0
r
where the constant k is given by
k

2FfI 1=2
es e0
3:8:13
Dimensional analysis shows that k has units of reciprocal length, and it is called
the DebyeHuckel reciprocal distance. It depends on the ionic strength of the
solution, the dielectric properties of the solvent, and temperature. For an aqueous
solution containing a 11 electrolyte at a concentration of 1 M (1000 mol m3) at
25 C, k is equal to 3.288 nm1. As will be seen below, 1/k corresponds to the
eective thickness of the ionic atmosphere, which would be 304 pm for a 1 M
solution.
The solution to the PoissonBoltzmann equation is now rather easily obtained.
First of all, one makes the substitution
y r
Then,
3:8:14
124
dy
d
r

dr
dr
3:8:15
d2 y
d
d2
r 2
2
2
dr
dr
dr
3:8:16
and
The left-hand side of equation (3.8.12) gives

1 d 2 d
2 d d2
r
2
dr
r dr
r2 dr
dr
3:8:17
so that the PoissonBoltzmann equation becomes

d2 y
k2 y
dr2
3:8:18
This is a well-known dierential equation in mathematical physics, whose general

solution is
y r k1 ekr k2 ekr
3:8:19
The constants k1 and k2 are determined by the boundary conditions for the
system. One of these states that the potential, , must go to zero as the distance,
r, goes to innity. This leads to the conclusion that term in ekr cannot contribute
to the potential ; as a result the constant k2 is zero. The solution to the Poisson
Boltzmann equation with one undetermined constant k1 is then
k1 kr
e
r
3:8:20
In order to determine the constant k1, one makes use of the relationship
between the charge density rz and local potential . From equations (3.8.10)
and (3.8.11),
n
X
rz
z2i Ffci 2FfI es e0 k2
3:8:21
i1
The total charge in the solution outside a given reference ion must be equal and
opposite to the charge on the ion in order to maintain electroneutrality. Each ion
has an excluded volume around it into which the center of another ion may not
enter. For spherical ions this volume may be considered also to be spherical and
to have a radius a related to the sum of the radii of the cation and anion in a
simple electrolyte but not necessarily equal to that sum (see g. 3.5).
The cuto distance a denes the lower limit for integrating the charge density
outside of a reference ion in a calculation which allows one to determine the
constant k1. If the charge on the reference ion is zie0, one may write
1
4pr2 rz dr zi e0
3:8:22
where the integration involves adding the charge density in spherical shells of
volume 4pr2dr. Combining this with equations (3.8.20) and (3.8.21), one obtains
125
Fig. 3.5 Models for the excluded

volume around (a) a contact ion pair
and (b) a solvent-separated ion pair
with consideration of interaction of
their solvation atmospheres. The
radius of the excluded volume is
designated as a.
1
4prk1 es e0 k2 ekr dr zi e0
3:8:23
Integrating by parts, the result is

4pk1 es e0 kaeka eka zi e0
3:8:24
so that
k1
zi e0 eka
1
4pe0 es 1 ka
The solution of the PoissonBoltzmann equation is thus

ka
zi e0 ekr
e

4pe0 es r 1 ka
3:8:25
3:8:26
To proceed further, one must separate the contribution to due to the ionic
atmosphere from the contribution that the ion makes itself in the absence of other
ions, that is, the so-called self-atmosphere potential. The latter quantity is given
by
ze
self i 0
3:8:27
4pe0 es r
It follows that atm estimated at r a, the distance to which other ions are absent,
is

zi e0
1
zi e0
zi e0
k

atm

3:8:28
4pe0 es a 1 ka
4pe0 es a
4pe0 es 1 ka
The remaining part of the problem involves relating the work done in forming
the ionic atmosphere to the activity coecient yi. If all departure from ideality is
attributed to the eects of ionion interactions, and the work done is estimated in
an imaginary charging process in which the reference ion acquires its charge in the
presence of all of the other ions, then
z
i e0

z2i e20
k
atm dzi e0
8pe0 es 1 ka
3:8:29
For a mole of these ions, the relationship between the contribution to the Gibbs
energy due to the non-ideality and the work done is
126
RT ln yi

NL z2i e20
k
8pe0 es 1 ka
3:8:30
Note that the activity coecient estimated is designated yi because the concentration of the ions is expressed in molarity units. Since k 3.288 I1/2 nm1 at 25 C in
water, the expression for ln yi for these conditions is
ln yi
1:174 z2i I 1=2

1 3:288 a I 1=2
3:8:31
ADH z2i I 1=2

1 BDH aI 1=2
3:8:32
In general one may write

ln yi
where ADH and BDH are the well-known DebyeHuckel constants. If the ionic
strength is expressed in moles per liter, then these constants are given by the
expressions

1=2
NL e20 F
2000
3:8:33
ADH
8p
e0 es RT3
and

2000 1=2
3:8:34
BDH F
e0 es RT
where the factor of 1000 is introduced to convert moles per liter to moles per cubic
meter. Values of ADH and BDH for water in the temperature range 0 to 50 C are
given in table 3.8.
The DebyeHuckel theory is also used to estimate activity coecients for dilute
solutions on the molality scale. In this case, equation (3.8.32) becomes
ln gi
0
ADH
z2i I 1=2
0
1 BDH aI 1=2
3:8:35
where I 1/2, A 0 , and B 0 are now the ionic strength and the DebyeHuckel constants
on the molality scale. Thus,
Table 3.8 Values of the DebyeHuckel Constants on the
Molarity Scale Together with the Density and Relative
Permittivity of Water in the Temperature Range 050 C
T/ C
rs/kg L1
es
ADH / L1/2 mol1/2
BDH/nm1 L1/2 mol1/2
0
10
20
25
30
40
50
0.99984
0.99970
0.99820
0.99705
0.99565
0.99222
0.98804
87.81
83.99
80.27
78.46
76.67
73.22
69.90
1.1309
1.1455
1.1638
1.1741
1.1856
1.2100
1.2374
3.2473
3.2611
3.2784
3.2881
3.2987
3.3212
3.3462
127
1X 2
z i mi
3:8:36
2
0
0
where mi is the molality of ion i. The constants ADH
, and BDH
for dilute solutions
(see equation (3.7.13)) are given by
1=2
0
ADH
ADH rs
3:8:37
I
and
0
BDH
BDH rs 1=2

0
ADH
1=2
0
BDH
3:8:38
1:1724 kg mol
and
3:2832 nm kg mol1=2
Thus, at 25 C,
on the molality scale.
The DebyeHuckel equation as presented above is often called the extended
DebyeHuckel law (EDHL) because a simpler expression is used for very dilute
solutions. When the ionic strength is less than 0.001 M, the term in the denominator of equation (3.8.32) goes to one, and one may write
1=2
ln yi ADH z2i I 1=2
1
1=2
3:8:39
This result is known as the limiting DebyeHuckel law (LDHL) and gives a
remarkably simple way of estimating the eects of non-ideality in very dilute
electrolyte solutions. Although the model does not recognize the nite size of
the ions in solution or the structure of the solvent in a molecular sense, it points
out the importance of the ionic strength in determining the degree of departure
from ideality.
In order to examine the predictions of the model with respect to experimental
data, one must estimate the mean molar activity coecient. On the basis of
equation (3.6.17), this can be obtained from the relationship
n ln y n ln y n ln y
n z2 ADH I 1=2 n z2 ADH I 1=2
3:8:40
Electroneutrality requires that

jn z j jn z j
3:8:41
jn z2 j jn z z j
3:8:42
jn z2 j jn z z j
3:8:43
so that
and
Using equation (3.6.11), the expression for the LDHL becomes

ln y ADH jz z jI 1=2
3:8:44
In the same way, the corresponding expression for the EDHL is

ln y
ADH jz z jI 1=2
1 BDH aI 1=2
3:8:45
128
EXAMPLE
Estimate the mean ionic activity coecient for 0.04 M CaCl2 using the LDHL
and the EDHL at 258C assuming that the ion size parameter a is 0.47 nm.
Compare with the experimental value, which is 0.597.
The ionic strength of the solution is
I 1=20:04 4 0:08 1 0:12 M
3:8:46
According to the LDHL,

ln y Ajz z jI 1=2 1:174 2 0:121=2 0:8134
or
y 0:443
3:8:47
According to the EDHL,

ln y
0:8134
0:5298
1 3:288 0:47 0:121=2
or
y 0:589
3:8:48
The result from the EDHL is much closer to the experimental value, as one
would expect.
Experimental values of ln y for NaCl are plotted against the square root of the
ionic strength in g. 3.6. Also shown are the values of ln y predicted by the
LDHL. It is clear that the limiting law agrees with the experimental data only
for the most dilute solutions (I < 0:001). At higher concentrations, the experimental results fall above the limiting law predictions, mainly because one must
Fig. 3.6 Plot of y on a logarithmic scale against the square root of the ionic strength, I 1=2
for aqueous NaCl at 258C. The straight line shows the prediction of the limiting Debye
Huckel law (equation (3.8.39)).
129
consider the eect of ionic size as expressed through the extended DebyeHuckel
equation. For electrolytes containing ions of higher charge the t between theory
and experiment is limited to even more dilute solutions.
In order to examine the eect of ionic size on the value of ln y , the experimental data are plotted in another way. Equation (3.8.45) may be rearranged to
give
Ajz z jI 1=2
1 BaI 1=2
ln y
3:8:49
Accordingly, a plot of the function on the left-hand side should give a straight line
when plotted against I 1=2 with a slope equal to Ba and an intercept of unity on the
y-axis. Such a plot with data for NaCl is shown in g. 3.7. It is apparent that these
conditions are met for ionic strengths less than 0.1 M I 1=2 < 0:3. At higher
concentrations there are positive deviations from the value of the ordinate predicted by the EDHL. The slope of the initial linear portion is 1.39, from which the
estimate of the ion size parameter is 420 pm.
Several reasons can be cited for the failure of the DebyeHuckel theory at
higher concentrations. One important reason is the fact that this model only
considers the fact that the central ion has size, and ignores the nite size of the
other ions in the atmosphere. As a result the thickness of the ionic atmosphere is
underestimated in concentrated solutions. In addition, the extra work involved in
introducing additional electrolyte ions into a concentrated solution while maintaining constant volume is neglected. This fact is mainly responsible for the
experimentally observed increase in ln y with increase in ionic strength (see g.
3.6). Finally, the structure of the solvent, water, is strongly aected by ionsolvent
interactions. As the concentration of the electrolyte increases, the fraction of
Fig. 3.7 Plot of the activity coecient data for aqueous NaCl at 258C according to equation (3.8.49). The value of the ion size parameter a from the slope in the limit of low
concentrations is 420 pm.
130
water molecules associated with ions increases, and the dielectric permittivity of
the solution decreases. This results in an increase in the strength of ionion
interactions and eventually can lead to ion pairing. In the following section a
method for estimating the activity coecients of electrolytes in more concentrated
solutions is considered, using a version of the MSA which ignores the molecular
nature of the solvent.
3.9 Ion-Ion Interactions According to the MSA

Although the DebyeHuckel theory of electrolyte solutions gives a good description of their physical properties when they are very dilute, it fails in a serious way
at higher concentrations, especially for electrolytes involving ions of higher
charge. In these systems the neglect of the nite size of all constituent ions is an
important defect and leads to overestimation of the eects of the ionion interactions. One way of overcoming this defect is to represent the electrolyte solution
as a collection of hard spheres of varying sizes, corresponding to the constituent
cations, anions, and solvent molecules. This approach was described earlier in the
discussion of ionsolvent interactions (section 3.5), and has also been applied to
the assessment of ionion interactions, especially in very dilute solutions [12].
Such an approach is called non-primitive because it considers the discrete nature
of the solvent molecules and ions which make up the electrolyte solution. A
method of treating these systems which falls between the DebyeHuckel model
and a non-primitive model represents the ions as hard spheres and the solvent as a
dielectric continuum. Such an approach is denitely primitive because it neglects
the molecular nature of the solvent, but it allows one to consider the eects of
ionic size more carefully than in the DebyeHuckel model, in which only the size
of the central ion was taken into account. In the MSA, all of the ions have a nite
size. As a result the distribution of ions around the central one is determined not
only by the local electrostatic potential but also by how well they can be packed in
space. It follows that the ionic distribution functions are not always smooth but
can have an oscillatory character, especially near the point of reference at the
central ion. When the cation and anion are assumed to have equal sizes the
present MSA treatment is called a restricted primitive model. Although presentations of the unrestricted version of the MSA are available [1820] the discussion
here is limited to the restricted version.
In the following model, the electrolyte is assumed to consist of ions having
radius ri and charge number zi . Since the solution is electrically neutral, the sum
of the concentration of each ion, ci times its valence zi must equal zero, that is
X
zi ci 0
3:9:1
i
The ions are embedded in a dielectric continuum of relative permittivity es . The

MSA boundary conditions for this system are
gij r 0;
and
r < 2ri
3:9:2
cij r buij r
zi zi be20
;
4pe0 es r
r > 2ri
131
3:9:3
On the basis of these conditions and the OrnsteinZernike equation, one obtains
the distribution functions gij r and cij r for all values of r.
An important parameter obtained in the solution of this problem is , which is
related to the thickness of the ionic atmosphere:
!1=2
NL e20 b X
z2i ci
3:9:4

4e0 es i 1 2ri 2
The concentration ci is expressed in moles per cubic meter. For the case that the
ions all have the same radius ri , 2ri is equal to the distance of closest approach, a,
dened in g. 3.5. Then, the summation in equation (3.9.4) can be related to the
ionic strength and the expression for can be written as
k
3:9:5

21 a
where k is the DebyeHuckel reciprocal distance dened earlier (equation
(3.8.13)). It is obvious that in the limit that the ion size parameter a goes to
zero, 2 is equal to the reciprocal distance k. Solving equation (3.9.5) for ,
one obtains

1 2ak1=2 1
2a
3:9:6
In dilute solutions, the pair distribution function is

gij r
bzi zj e20 2r

e
4pe0 es r
3:9:7
This is exactly the form expected for gij r on the basis of the DebyeHuckel
model if 2 replaces k as the screening parameter.
The thermodynamic properties of the electrolyte solution are derived by rst
calculating the excess internal energy due to electrostatic interactions, Ues . This
quantity is given by
N e2 X z2i ci
3:9:8
Ues L 0
4pe0 es
1 2ri
i
where Ues has units of J m3 molec1 . For the case that all ions have the same
size, this reduces to
Ues
NL e20
I
2pe0 es 1 a
3:9:9
The corresponding excess Helmholtz energy is estimated using the thermodynamic relationship
@bAes
Ues
@b
3:9:10
132
After some algebra, one obtains the result

Aes Ues
3
3pb
3:9:11
Since b 1=kB T, the term kB 3 =3p can be identied as the excess entropy
per unit volume associated with electrostatic interactions. One may now extract
the electrostatic contribution to the single ion activity coecient, which is

@bAes
be20 z2i
bz2 Ues
ln yi es
i
3:9:12

NL @ci
4pe0 es 1 a
2NL I

After application of equation (3.8.42), the expression obtained for the mean ionic
activity coecient is
ln y es
bjz z jUes
2NL I
3:9:13
In order to complete the MSA estimate of ln yi one must add the hard-sphere
contribution, which accounts for the fact that work must be done to introduce the
ions as hard spheres into the solution. It is obtained from the PercusYevick
model for non-interacting hard spheres. For the case that all ions (spheres)
have the same radius, the result is (see equation (3.9.22))
ln yi hs
6z 3z2 2z
2 3
3:9:14
where
1z
3:9:15
and
NL pa3
z
ci
3:9:16
Since ln yi is equal to ln yi hs when all the ions have the same size, the resulting
expression for the mean ionic activity coecient is
ln y ln y es ln y hs
3:9:17
Estimation of this quantity requires that only one adjustable parameter be specied, namely the mean ionic diameter a.
EXAMPLE
Estimate the mean activity coecient for 0.25 M NaCl according to the MSA
assuming an ion size parameter equal to 360 pm and a temperature of 258C.
For an ionic strength of 0.25 M, the DebyeHuckel parameter k is equal to
1.644 nm1 . The MSA is now estimated using equation (3.9.6):

1 2 0:36 1:6441=2 1
0:6635 nm1
2 0:36
3:9:18
133
Now, Ues is calculated using equation (3.9.9). The constant NL e20 =2pe0 es is
equal to 3:541 106 J m mol1 so that
Ues
3:541 106 6:635 108 0:25 1000

1 0:6635 0:36
3:9:19
4:741 105 J m3 mol1

From equation (3.9.13)
ln y es
4:741 105
0:382
2 8:3144 298:2 250
3:9:20
In order to estimate the hard-sphere contribution the packing fraction z is

calculated rst:
z
6:022 1023 p 3:6 1010 3 250 2

7:36 103
6
3:9:21
Using equation (3.9.14),

ln y hs
6 7:36 103 3 7:36 103 2 2 7:36 103
0:060
0:9926
0:99262
0:99263
3:9:22
Thus,
ln y 0:382 0:060 0:322
3:9:23
The resulting estimate of y (0.725) agrees very well with the experimental
value (0.726).
The MSA model was tted to the activity coecient data for aqueous NaCl
solutions on the molarity scale following the procedure used in the above example. The results are shown in g. 3.8 for the case that the mean ionic diameter is
assumed to be 360 pm. An excellent t was found for concentrations up to 0.3 M.
At higher concentrations, the estimated value of ln y is higher than that found
experimentally. This is clearly a result of the hard-sphere contribution, which is
positive. Results obtained using the extended DebyeHuckel model are shown in
the same gure. This model requires that a higher average ionic diameter be
assumed and produces results which agree with experiment up to 0.1 M. The
DebyeHuckel model overestimates the eects of non-ideality at higher concentrations because it ignores the eects due to the size of all the ions in the system,
considering only the eect of the size of the central ion.
The t of the MSA to activity coecient data for aqueous electrolyte solutions
can be considerably improved if one takes into consideration the decrease in
solvent permittivity which accompanies the increase in electrolyte concentration.
This phenomenon is clearly related to the eect that ions have on solvent structure
and was studied originally in aqueous solutions by Hasted et al. [21, 22]. More
recently, data have been collected for a large number of electrolytes by Barthel
and coworkers [23]. In the case of NaCl solutions, the change in dielectric permittivity with electrolyte concentrations up to 2 M is given by
134
Fig. 3.8 Plot of the mean ionic activity coecient for NaCl at 258C on a logarithmic scale
against the square root of the molar ionic strength. The points give the experimental data;
the curves give the theoretical ts for the MSA and extended DebyeHuckel models with
ion size parameters as indicated.
es 78:46 19ce 5c3=2

e
3:9:24
At a concentration of 0.1 M, the value of es is 76.7 but when the concentration is

1 M, es falls to 64.5. The extent to which the permittivity changes with concentration also depends on the nature of the electrolyte. This aspect of solvent properties
is discussed in more detail in section 4.6.
When the variation in es with electrolyte concentration is considered the
expression for ln yi es becomes
ln yi es
bz2i Ues
be20 I
@es
2NL I
2e0 e2s 1 a @ci
bz2 Ues bUes @es

i
2NL I
NL es @ci
3:9:25
Making use of equation (3.9.13), the mean ionic activity coecient is

ln y es
jz z jbUes bUes @es

2NL I
nNL es @ce
3:9:26
where
@es
@e
@e
n s n s
@ce
@c
@c
3:9:27
@es =@ce can be measured experimentally and in the present example is calculated
from the concentration derivative of equation (3.9.24).
The t of the MSA model with varying solvent permittivity is shown in g. 3.9.
These results demonstrate the importance of considering the true solvent permit-
135
Fig. 3.9 As in g. 3.8, but with a varying dielectric permittivity for the solvent in the MSA
model (see equation (3.9.26)).
tivity in estimating ionion interactions. The value of the size parameter a which
yields the best t (436 pm) is somewhat larger than that found for the simpler
MSA t with a constant permittivity equal to that of the pure solvent. At concentrations higher than 1 M, the MSA curve lies above the experimental results.
This undoubtedly reects a small extent of ion pairing in concentrated solutions
of 11 electrolytes so that the true ionic strength is somewhat less than that
estimated on the basis of the stoichiometric concentration.
It should be noted that redening the DebyeHuckel constants so that they
change with electrolyte concentration to reect the corresponding change in es
does not extend the concentration range over which this model ts experimental
data. This result emphasizes that it is important to include the nite size of all ions
in a model which is applied in a concentration range greater than 0.1 M.
Inclusion of the change in solvent permittivity in the MSA description is an
eective way of dealing with the change of solvent properties which accompany
the addition of an electrolyte to a polar solvent. Since permittivity data are now
available for a large number of electrolyte solutions in water [23], the MSA model
can be applied to a wide variety of systems. However, there is one feature of
electrolyte solutions which has been neglected in the treatments presented up to
this point, namely, the existence of ion aggregates. This feature of electrolyte
solutions is discussed in the following sections of this chapter.
3.10 The Thermodynamics of Ion Association

Two important factors which determine whether ionic association is present in an
electrolyte solution are the charge on the ions and the dielectric permittivity of the
solvent. Their roles are clearly seen on the basis of Coulombs law, which shows
136
that the force between two ions increases with increase in the product of their
charges and with decrease in the solvents permittivity, that is
Fc
zi zj e20
4pe0 es r2ij
3:10:1
where zi and zj are the valences of the two ions, and rij , the distance between them.
Another factor to be considered is chemical bonding between two ions to form a
molecule. This phenomenon falls under the general classication of weak electrolytes, which were discussed in section 3.6.
Derivation of the thermodynamic equations for an electrolyte system with ion
pairing follows the same procedure given for a weak electrolyte. However, in the
following the ion pairing equilibrium is dened in terms of an association process.
For a 11 electrolyte, ion pairing is described as
M X (
+ MXip
3:10:2
where MXip represents the neutral ion pair. The association equilibrium constant
Kas is given by
aip
3:10:3
Kas
aM aX
where aM and aX are the activities of the cation and anion, respectively, and aip ,
that of the ion pair. The association constant may be written in terms of the
standard potentials of each of these species as follows:
RT ln Kas mM mX mip
3:10:4
Here mM and mX are the standard chemical potentials for the two ions on the
molarity scale, and mip , that for the ion pair.
As a result of ion pairing, a fraction ai of the dissolved electrolyte remains in
solution as ions and a fraction 1 ai exists as ion pairs. Thus, the chemical
potential of the electrolyte can be expressed as
mMX ai mM ai mX 1 ai mip
3:10:5
where mM and mX are the chemical potentials of the ions and mip , that of the ion
pair. Substituting equations (3.10.3) and (3.10.4) into equation (3.10.5), one
obtains
mMX mM mX RT ln aM aX
3:10:6
This result shows that the chemical potential of the electrolyte can be expressed in
terms of the activities of the two ions without considering that of the ion pair. It
was obtained earlier for weak electrolytes as equations (3.6.38) and (3.6.40).
Because of ion pairing the ionic activities aM and aX are less than they would be
on the basis of the stoichiometric concentration of the electrolyte, ce . Expanding
the activities in equation (3.10.3) in terms of activity coecients and concentrations, one can write
Kas
yip cip
1 ai yip
yM c M yX c X
y2 a2i ce
3:10:7
137
where yip is the activity coecient of the ion pair, cip , its concentration, yM and
yX , the activity coecients for the individual ions, y , the mean ionic activity
coecient, and cM and cX , the concentration of the two free ions, which are equal
to each other. One may also write the activity product aM aX as
aM aX a2i y2 c2e
3:10:8
It is clear that the value of Kas must be known in order to assess the thermodynamic properties of an electrolyte system in which ion pairing takes place. The
above equations may also be written on the molality scale (see section 3.6).
It should be noted that experimental activity coecient data for systems
known to undergo ion pairing are tabulated ignoring ion pairing. In this case
the tabulated ionic activity coecient on the molality scale must be regarded as
giving ai g , not . For example, moderate ion pairing occurs in NaF solutions.
Given that the tabulated mean ionic activity coecient for 0.1 m NaF is 0.765 and
assuming Kas 1 and gip 1, the value of ai in this solution is 0.941 on the basis
of equation (3.10.7). It follows that the true mean ionic activity coecient is
0.765/0.941, that is, 0.813. This calculation requires a value of Kas which must
be obtained on the basis of additional experimental information. Small changes in
the assumed value of Kas have a large eect on the result of the calculation,
especially at high electrolyte concentrations.
EXAMPLE
A reasonable estimate of the thermodynamic association constant for MgSO4

is 164. The experimental value of g in 0.4 m MgSO4 is 0.0756. Estimate the
true value of g assuming that the activity coecient of the ion pair is unity.
On the basis of equation (3.10.7), one may write on the molal scale with
gip 1,
1a
Kas 164
g2 a2 me
3:10:9
Substituting for ai g , 0.0756, and for me , 0.4, one obtains

1 ai 0:375
3:10:10
Thus, the fraction dissociated ai is 0.625 and the true activity coecient g is
0.121.
The above example shows that the same analysis can be applied to a symmetrical electrolyte containing ions of higher charge, for example MgSO4. In the case
of non-symmetrical electrolytes such as MgCl2, the ion association process can be
more complex. For the 21 electrolyte MX2, two association steps are possible:
M2 X (
+ MX
Kas1
3:10:11
+ MX2
MX X (
Kas2
3:10:12
and
+
Formation of the ion pair MX is expected to take place under most circumstances. However, formation of the ion triplet MX2 may not be signicant, especially in aqueous solutions. Important factors determining the magnitude of the
138
association constants are the charges on the ions and the solvent permittivity, as
can be seen from equation (3.10.1). An analysis of the properties of a system
containing more than one association step to obtain the concentrations and activities of the component ions is obviously a complicated problem. However, if the
association constants are known, methods exist for estimating the activity coecients of the system over a wide concentration range [24].
Considerable eort has been made to develop a model for the parameter Kas on
the basis of statistical theories using simple electrostatic concepts. The rst of
these was proposed by Bjerrum [25]. It contains important ideas which are
worth reviewing. He assumed that all oppositely charge ions within a certain
distance of a central ion are paired. The major concept in this model is that
there is a critical distance from the central ion over which ion association occurs.
Obviously, it must be suciently small that the attractive Coulombic forces are
stronger than thermal randomizing eects. Bjerrum assumed that at such short
distances there is no ionic atmosphere between the central ion and a counter ion
so that the electrostatic potential due to the central ion may be calculated directly
from Coulombs law. The value of this potential at a distance r is
zi e0
4pe0 es r
3:10:13
where zi e0 is the charge on the central ion. The concentration of counter ions with
charge zj e0 at the same distance is estimated using Boltzmann statistics so that
crj cj expzj f
3:10:14
where cj is the bulk concentration in ions per cubic meter and f e0 b. Combining
these equations one obtains an expression for the probability Pr of nding two
oppositely charged ions at a distance r from each other:
!
zi zj e20 b
2
Pr 4pr cj exp
3:10:15
4pe0 es r
This function has two opposing trends: rstly, as r increases the exponential term
decreases due to the decrease in Coulombic attraction; secondly, the volume
associated with the sphere containing possible counter ions increases as 4pr2 .
Thus, the probability has a minimum whose location can be found by dierentiating Pr with respect to r. The condition for a minimum is
q
q
dPr
q
B
B
8prcj exp
4pr2 cj B2 exp
0
3:10:16
dr
r
r
r
where
zi zj e20 b
3:10:17
4pe0 es
Dening rB as the distance corresponding to the minimum, it follows that
qB
rB
zi zj e20 b
qB

8pe0 es
2
3:10:18
139
The value of the Bjerrum distance rB depends on the ionic charges, the nature of
the solvent and the temperature but not on electrolyte concentration. For aqueous
solutions of 11 electrolytes at 258C, it is equal to 357 pm. Bjerrum proposed that
all ions contained in a sphere with this radius are paired. This is a reasonable
proposal for aqueous systems, since ions of typical size would be close to contact
at such small separations. However, if the solvent has a lower dielectric permittivity, the distance over which ions are considered to be paired increases and the
assumption that they are in contact is more dicult to accept. For example, when
the solvents relative permittivity is 10, the Bjerrum distance rB increases to
2800 pm.
A plot of the Bjerrum probability function against distance r is shown for an
aqueous solution in g. 3.10. The probability of nding an oppositely charged ion
rises rapidly from the minimum at 357 pm as the distance between ions gets
smaller. An increase in probability is also seen as distance increases from the
minimum. The function Pr is limited for small values of r by the fact that
oppositely charged ions can approach the central ion up to the distance of closest
approach a. However, there is no limit in Pr for large values of r. Bjerrum
obtained an estimate of the number of paired ions by integrating Pr from the
distance of closest approach a to the minimum on the probability curve rc . In this
way, he was able to obtain an estimate of the association constant Kas . Further
details may be found in the monograph by Robinson and Stokes [G1].
As suggested by the above discussion, there are serious problems with the
Bjerrum model. One of these relates to the fact that unreasonably large critical
distances are involved in dening an ion pair in solutions of low permittivity. The
second relates to the fact that the probability distribution is not normalized and
continues to increase with increase in distance r. The latter problem is eectively
avoided by considering only those values of Pr up to the minimum in the curve.
Fig. 3.10 Bjerrum probability function Pr for oppositely charged ions plotted against r,
the distance from a central ion for a 11 electrolyte at 1 M in water at 258C with es 78:46:
140
Fuoss [26] revised the distribution function taking into account the fact that the
probability should go to zero for large values of r. However, the modied function
describing Pr is essentially the same as that given by Bjerrum for small values of
r. Therefore, the simpler theory is that usually used.
Fuoss developed a new theory of ion association in 1958 [27] which overcame
some of the diculties associated with the Bjerrum approach. The cations in the
solution were assumed to be conducting spheres of radius a and the anions to be
point charges. The ions are assumed to be immersed in a dielectric continuum of
permittivity es . Only oppositely charged ions separated by the distance a are
assumed to form ion pairs. The resulting expression for the association constant
is

4000pa3 NL
jqB j
Kas
exp
3
a
3:10:19
The most important quantity determining the magnitude of Kas is the ion size
parameter a, which appears to the third power in the numerator and also in the
exponential term. The role of the solvent and the charges on the two ions is
expressed through the parameter qB (equation (3.10.17)). This quantity increases
with the charge product jzi zj j and with decrease in the solvent permittivity. In the
case of MgSO4, assuming a is equal to 450 pm, the value of jqB j=a is 6.35. The
estimate of Kas according to equation (3.10.19) is 132. This result is the correct
order of magnitude on the basis of other experimental evidence.
The Fuoss estimate of Kas is based on a more reasonable model than that of
Bjerrum and therefore is preferred. However, there are also problems with the
Fuoss treatment in so far as it considers the solvent to be a dielectric continuum.
Dielectric saturation eects are expected to be important, especially near the ions
involved in ion pair formation. The second problem relates to the choice of the
eective size for the ions. In the calculation made here the value of a for MgSO4
was chosen to be much bigger than the crystallographic radius of Mg2 . This
presumably is because the cation is strongly hydrated in aqueous solution. One is
then faced with the question whether the ion pair involves contact of the two ions
or whether it is better considered to be a species in which the two ions are
separated by at least one water molecule. These questions can only be properly
resolved using other experimental methods.
In summary, both models are useful from a qualitative point of view in assessing the strength of ion pairing in a given system. They demonstrate the importance of the ionic charge and the dielectric permittivity of the medium in
determining Kas . In order to obtain a good quantitative estimate of Kas , as, the
size of the ions involved must be carefully determined.
3.11 Ion Association According to the MSA

In the treatment of ionic activity coecients according to the MSA it was emphasized that the nite size of the ions is an important factor in estimating their Gibbs
energy. Accordingly, the work done to introduce an additional ion to the solution
141
increases at constant solution volume with increase in electrolyte concentration.

This is the hard-sphere contribution to the activity coecient which depends on
the size of the ions and on their concentration.
When ion pairing is present there are three solute species, the additional one
being the ion pair. This species is expected to be larger than either the cation or
the anion, and to be a factor in determining the hard-sphere contribution to the
Gibbs energy. In more traditional treatments of ion pairing, the activity coecient
of the ion pair was assumed to be unity. On the basis of the MSA it is greater than
one, and depends on electrolyte concentration.
For the case of symmetrical electrolytes with ion pairing, the electrolyte activity
is given by
aM aX a2i y2 c2e
3:11:1
where ai is the fraction of the ions remaining free, y , the mean activity coecient
on the molarity scale, and ce , the electrolyte molarity. On the basis of the ion
pairing equilibrium
Kas
1 ai yip
y2 a2i ce
3:11:2
where Kas is the ion pairing equilibrium constant and yip , the activity coecient of
the ion pair. Estimation of the activity coecients and the fraction ai requires
knowledge of two parameters, namely Kas , and a, the mean diameter of the three
solute species. As outlined earlier (section 3.9), y consists of an electrostatic
contribution and a hard-sphere contribution. The rst is
ln y es
jz z jbUes bUes @es

2NL I
nNL es @ce
3:11:3
where Ues is dened by equation (3.9.9) and ce is the concentration in moles per
cubic meter. The hard sphere contribution is given by equations (3.9.14)(3.9.16)
in the restricted approximation. It is then equal for all solute species so that
ln yip ln y hs ln yi hs
3:11:4
Finally, the experimentally available activity coecient is (y exp where

y exp ai y ai y es y hs
3:11:5
Given values of a and Kas , the activity coecient y is calculated in the following way. Initially assuming ai 1, the value of (y es is estimated using equations
(3.11.3) and (3.11.4) with the values of ce and a. Now equation (3.11.2) may be
solved for ai given Kas . The process is repeated using the new estimate of ai until
successive estimates of y agree to the required precision.
Results of such an analysis for the MgSO4 system are shown in g. 3.11. The
best values of a and Kas for this system are 610 pm and 185 L mol1 , respectively.
On the basis of dielectric relaxation experiments, the permittivity of MgSO4
solutions as a function of electrolyte concentration is given by
es 78:5 31:8ce 12:8c3=2
e
3:11:6
An excellent t between the MSA model and experimental data is possible for
concentrations up to 1.5 M (see g. 3.11). At higher concentrations, the theory
142
Fig. 3.11 Plot of the eective mean ionic activity coecient for MgSO4 at 258C on a
logarithmic scale against the square root of the molar ionic strength. The curve gives the
t with the modied MSA assuming a 610 pm and Kas 185 L mol1 ; the straight line
shows the prediction of the limiting DebyeHuckel law.
predicts values of y which are higher than those found experimentally. This is
probably because the eective size of the ions, especially the strongly hydrated
Mg2 ion, decreases at high concentrations, where there are fewer water molecules
available per ion.
It is very interesting to examine how the mean activity coecient for the ions,
y , the ion pair, yip , and the fraction of electrolyte dissociated, ai , vary with
electrolyte concentration. These quantities are plotted against ce for the range
over which the theory is successful in g. 3.12. The mean ionic activity coecient
Fig. 3.12 Plots of yip , y , and ai on a logarithmic scale for the MgSO4 system according to
the modied MSA model at 258C against electrolyte concentration ce in water.
143
y falls to very low values as one would expect for a 22 electrolyte at high ionic
strength. In this regard it should be remembered that a concentration of 1.5 M
corresponds to an ionic strength of 6 M, so that the theory has been very successful in describing the electrostatic interactions. At the same time the activity coecient of the ion pair increases reaching values greater than 2. This has an
important inuence on the estimate of ai and the nal estimate of the overall
mean electrolyte activity coecient ai y . The fraction of electrolyte dissociated is
initially close to 0.6 and then slowly increases. This reects changes in the apparent association constant with ionic strength and is an important factor in determining the ionic strength. On the basis of equation (3.11.2) the apparent
association constant may be dened as
Kap
Kas y2
yip
3:11:7
It decreases with increasing ionic strength due to the corresponding decrease in

y and increase in yip . As a result, there is a slow increase in ai , the fraction
dissociated, at higher concentrations.
In conclusion, the MSA provides an excellent description of the properties of
electrolyte solutions up to quite high concentrations. In dilute solutions, the most
important feature of these systems is the inuence of ionion interactions, which
account for almost all of the departure from ideality. In this concentration region,
the MSA theory does not dier signicantly from the DebyeHuckel model. As
the ionic strength increases beyond 0.1 M, the nite size of all of the ions must be
considered. This is done in the MSA on the basis of the hard-sphere contribution.
Further improvement in the model comes from considering the presence of ion
pairing and by using the actual dielectric permittivity of the solution rather than
that of the pure solvent.
3.12. Concluding Remarks

The theory of electrolyte solutions developed in this chapter relies heavily on the
classical laws of electrostatics within the context of modern statistical mechanical
methods. On the basis of DebyeHuckel theory one understands how ionion
interactions lead to the non-ideality of electrolyte solutions. Moreover, one is able
to account quantitatively for the non-ideality when the solution is suciently
dilute. This is precisely because ionion interactions are long range, and the
ions can be treated as classical point charges when they are far apart. As the
concentration of ions increases, their nite size becomes important and they are
then described as point charges within hard spheres. It is only when ions come
into contact that the problems with this picture become apparent. At this point
one needs to add quantum-mechanical details to the description of the solution so
that phenomena such as ion pairing can be understood in detail.
The above comments also apply to the description of ionsolvent interactions. The traditional treatment of a solvent molecule is to represent it as a hard
sphere with a central point dipole. Such a model ignores the details of the
144
chemical nature of the interactions between the ion and the surrounding solvent
molecules.
In summary, the models discussed in this chapter focus on the physical
aspects of electrolyte solutions but they ignore the chemical aspects. This is
especially apparent in the treatment of ion solvation where an empirical correction to the MSA model was applied to treat the dierences in behavior seen for
cations and anions in water. The same problem arises in using classical electrostatics to describe ion pairing. In spite of the fact that the Bjerrum and Fuoss
models give a good qualitative description of an ion association, this phenomenon can only be understood in detail by using quantum-mechanical methods.
Needless to say, such calculations in condensed media are much more dicult to
carry out.
The problems outlined here can be greatly elucidated using spectroscopic methods. With the appropriate technique one can probe the strength of ionsolvent
interactions, and measure the extent of contact ion pairing. Spectroscopic studies
of electrolyte solutions have certainly greatly improved the understanding of these
important systems. Major spectroscopic methods and results of their application
to these systems are considered in detail in chapter 5.
General References
G1. Robinson R. A.; Stokes, R. H. Electrolyte Solutions, 2nd ed.; Butterworths: London,
1968.
G2. Petrucci, S. Ionic Interactions; Academic Press: New York, 1971; Vols. I and II.
G3. Marcus, Y. Ion Solvation; Wiley-Interscience: New York, 1985.
G4. Wagman, D. D.; Evans, W. H.; Parker, V. B.; Schumm, R. H.; Halow, I.; Bailey, S.
M.; Churney, K.L.; Natall, R. L. J. Phys. Chem. Ref. Data 1982, 11, Supp. 2.
G5. Lobo, V. M. M. Handbook of Electrolyte Solutions; Elsevier: New York, 1989.
G6. Horvath, A. L. Handbook of Aqueous Electrolyte Solutions; Ellis Horwood:
Chichester, 1985.
References
1. Ohtaki, H.; Radnai, T. Chem. Rev. 1993, 93, 1157.
2. Pauling, L. The Nature of the Chemical Bond, 3rd ed.; Cornell University Press: Ithaca,
NY, 1960; Chap. 13.
3. Gourary, B. S.; Adrian, F. J., Solid State Phys. 1960, 10, 125.
4. Morris, D. F. C. Struct. Bonding 1968, 4, 63.
5. Shannon, R. D.; Prewitt, C. T. Acta Crystallogr. 1969, B25, 925.
6. Szasz, G. I.; Heinzinger, K.; Palinkas, G. Chem. Phys. Lett. 1981, 78, 194.
7. Jenkins, H. D. B.; Thakur, K. P. J. Chem. Educ. 1979, 56, 576.
8. Bowen, K. H.; Castleman, A. W. J. Phys. Chem. 1996, 100, 12911.
9. Tissandier, M. D.; Cowen, K. A.; Feng, W. Y.; Gundlach, E.; Cohen, M. H.; Earhart,
A. D.; Coe, J. V.; Tuttle, Jr., T. R. J. Phys. Chem. A 1988, 102, 7787.
10. Fawcett, W. R. J. Phys. Chem. B 1999, 103, 11181.
11. Born, M. Z. Phys. 1920, 1, 45.
12. Chan, D. Y. C.; Mitchell, D. J.; Ninham, B. W. J. Chem. Phys. 1979, 70, 2946.
13. Garisto, F.; Kusalik, P. G.; Patey, G. N. J. Chem. Phys. 1983, 79, 6294.
14. Blum, L.; Fawcett, W. R. J. Phys. Chem. 1992, 96, 408.
15. Wertheim, M. S. J. Chem. Phys. 1971, 55, 4291.
16. Fawcett, W. R.; Blum, L. J. Chem. Soc. Faraday Trans. 1992, 88, 3339.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
145
Debye, P.; Huckel, E. Phys. Z. 1923, 24, 185.

Blum, L. Mol. Phys. 1975, 30, 1529.
Blum, L.; Hoye, J. S. J. Phys. Chem. 1977, 81, 1311.
Humray, A. A. J. Phys. Chem. 1983, 87, 5521.
Hasted, J. B.; Ritson, D. M.; Collie, C. H. J. Chem. Phys. 1948, 16, 1.
Haggis, G. H.; Hasted, J. B.; Buchanan, T. J. J. Chem. Phys. 1952, 20, 1452.
Barthel, J.; Buchner, R.; Munsterer, M. Electrolyte Data Collection, Part 2: Dielectric
Properties of Water and Aqueous Electrolyte Solutions; Dechema: Frankfurt/Main,
1995.
Pitzer, K. S. Activity Coecients in Electrolyte Solutions, 2nd ed.; CRC Press: Boca
Raton, FL, 1991.
Bjerrum, N. Kg. Danske Videnskab. Selskab. 1926, 7, 9.
Fuoss, R. M. Trans. Faraday Soc. 1934, 30, 967.
Fuoss, R. M. J. Am. Chem. Soc. 1958, 80, 5059.
Problems
1. Nitromethane (NM) is a polar solvent with a relative permittivity of 35.8
at 258C. It has a diameter of 431 pm when represented as a sphere.
Estimate the Gibbs energy of solvation of K+, whose diameter is
276 pm, in NM according to the Born model and MSA at 258C.
Compare these estimates with the experimental estimate, which is
300 kJ mol1 .
2. Estimate the entropy of solvation of K+ in NM given that the temperature coecient of the permittivity is 0:161 K1 . Use both the Born model
and the MSA. Compare these with the experimental estimate of
181 J K1 mol1 at 258C. What are the contributions to long-range interactions which depend on des =dT, and local interactions which depend on
dds=dT, in this estimate?
3. Estimates of the Gibbs energy of solvation of the alkali metal ions in
acetonitrile are 464 (Li+), 374 (Na+), 307 (K+), 288 (Rb+), and
257 kJ mol1 (Cs+) at 258C. Estimate the MSA parameter ds for acetonitrile using a plot of s Gm 1 against cation radius. Compare this
result with the value estimated from the permittivitry which is equal to
35.9.
4. The activity coecient on the molality scale for NaI at 258C is 0.938 at
0.1 m and 0.991 at 1 m. Estimate the corresponding concentrations in mol
L1 using the density data summarized in table 3.6, and hence, the activity
coecients on the molarity scale.
5. Use the extended DebyeHuckel theory to estimate the mean ionic activity
coecient for Na2SO4 at concentrations of 0.01 and 0.1 M and 258C
assuming an ion size parameter of 400 pm. Also calculate the mean electrolyte activity and the electrolyte activity.
146
6. Given the following activity coecient data determine the average ionic
radius for each electrolyte according to the extended DebyeHuckel equation. All data were obtained in water at 258C.
y
Conc/M
NaCl
CaCl2
0.001
0.002
0.005
0.01
0.02
0.05
0.1
0.2
0.5
0.965
0.953
0.927
0.900
0.869
0.820
0.780
0.738
0.687
0.888
0.851
0.787
0.729
0.664
0.577
0.517
0.469
0.444
Rearrange the extended DebyeHuckel equation so that the experimental

data may be plotted as a straight line depending on the square root of the
ionic strength whose slope is proportional to the average ionic radius.
7. On the basis of the equation
ln y
ADH zA zB I 1=2
CI
1 BDH aI 1=2
derive an expression for the ionic strength at which ln y is a minimum.

Given that the minimum value of y for CaCl2 is 0.444, and using the
results obtained in question 6, determine the appropriate value of C to t
the equation to these data. Calculate y for ce 1 and 3 M. Then plot ln
y versus I 1=2 showing the prediction of the LDHL on the graph.
8. The true solubility product (I 0) for calcium hydroxide is 6:5 106 at
258C. Estimate its solubility in (a) water, and (b) 0.1 M NaNO3 using the
EDHL BDH a 1 to estimate activity coecients. Calculate the pH of
the solutions in each case.
9. Estimate the activity coecient for KBr in water at 258C using the MSA
with an ion size parameter of 470 pm at concentrations of 0.1 M and
0.5 M. Estimate the solution permittivity using equation (3.9.18).
Compare with the experimental values, which are 0.818 (0.1 M) and
0.666 (0.5 M).
10(a). Use the Fuoss equation (3.10.19) to estimate an association constant for
ZnSO4 assuming an ion size parameter of 650 pm.
(b). Estimate the fraction a of free ions using the formula
Kas
147
1a
g2 a2 ce
which assumes that the activity coecient of the ion pair is unity. It is
necessary to use an iterative technique to solve this problem.
11(a). Estimate the DebyeHuckel constants ADH and BDH in dimethylsulfoxide
at 258C given that the density of DMSO is 1.096 g cm3 and its permittivity 46.7.
(b). Use the following activity coecient data for LiCl in DMSO to estimate
an ion size parameter for the electrolyte on the basis of the EDHL.
m
0.005
0.01
0.02
0.03
0.05
0.08
0.10
0.12
0.859
0.803
0.736
0.698
0.644
0.600
0.578
0.559
Polar Solvents
Peter Debye was born in 1884 at Maastricht,

the Netherlands, where he also grew up. After
secondary school he studied electrical
technology at the Institute of Technology in
Aachen and graduated in 1905. After two years
as an assistant in the same institute, Debye
moved to Munich University to study physics
and obtained a Ph.D. in physics in 1908. In
1911, he was appointed Professor of
Theoretical Physics at Zurich University. He
Peter Joseph Wilhelm Debye held chairs at Utrecht, and Gottingen, before
moving to ETH Zurich in 1920 as Professor of
Physics and Principal. In 1927, he held the same post at Leipzig University and
in 1934 he was appointed Director of the Max Planck Institute at the Kaiser
Wilhelm Institute of Physics in Berlin. With the outbreak of war in Europe,
Debye moved to Cornell University in the United States as Professor of
Chemistry and Chair of the Chemistry Department. He held that position until
his retirement in 1952. Debyes work is remembered in several, quite dierent,
areas of physical chemistry. One of his rst important contributions was made
in 1912 to the theory of specic heat of solids, in which he extended the earlier
work of Einstein. In 1923, while at Zurich, he published the famous theory of
strong electrolytes, together with Erich Huckel. He worked extensively on the
dielectric properties of matter and published his famous monograph Polar
Molecules [1] in 1929. Another area in which he contributed signicantly was
X-ray crystallography. Debye was awarded the Nobel Prize for Chemistry in
1936 for his contribution to explaining the structure of molecules, his research
into dipole moments, and his work on the electronic diraction of gases. His
devotion to science gained him many distinctions and international awards. He
died in Ithaca, New York, in 1966.
4.1 What Constitutes a Polar Liquid?

Polar solvents are those liquids whose relative permittivity is suciently high that
electrolytes can be dissolved in them. The best-known example of such a liquid is
water. The oxygen end of this simple molecule is electron-rich and can stabilize
cations. The hydrogen atoms are electron-poor and thus are involved in the
solvation of anions. The structure of pure water is very much inuenced by the
148
POLAR SOLVENTS
149
hydrogen bonding between the negative end of the molecular dipole at oxygen
and a hydrogen atom on an adjacent molecule. The special properties of water as
a solvent for electrolytes are the central reason for its importance in living systems.
There are many other solvents which can be classied as polar. Some of them,
such as the alcohols, have the same polar group as the water molecule, namely,
the hydroxyl group OH. These solvents are also involved in hydrogen bonding,
and are generally classied as protic. Other examples of protic solvents are simple
amides such as formamide and acetamide. In these systems, the protic group is
NH2, the hydrogen atom being involved in hydrogen bonding with the oxygen
atom in the carbonyl group on an adjacent molecule.
There are other polar solvents which are not protic. These involve liquids with
large dipole moments. Some examples are acetonitrile, propylene carbonate, and
dimethylsulfoxide. In each case, the solvent molecule possesses an electronegative
group which is rich in electrons. The opposite end of the molecule is electron
decient but does not have acidic hydrogen atoms which can participate in hydrogen bonding. This class of solvents is called aprotic.
In this chapter, the properties of polar solvents are discussed, especially as they
relate to the formation of electrolyte solutions. Polar solvents are arbitrarily
dened here as those liquids with a relative permittivity greater than 15.
Solvents with zero dipole moment and a relative permittivity close to unity are
non-polar. These include benzene, carbon tetrachloride, and cyclohexane.
Solvents with relative permittivities between 3 and 5 are weakly polar, and
those with values between 5 and 15 are moderately polar. The latter systems
are not considered in the discussion in this chapter.
Initially, some relevant thermodynamic and molecular properties of polar solvents are considered. Then, their dielectric properties are considered in detail. Ion
solvation in these solvents is also discussed with emphasis on some non-thermodynamic methods of dividing experimentally measured data for electrolytes into
contributions for the cation and anion. Finally, the important characteristics of
the solvent in its direct interaction with the solute, namely, its acidity and basicity,
are also described.
4.2 Some Important Properties of Polar Solvents

A number of properties, including dielectric properties, should be considered in
assessing a given polar solvent. Important bulk properties are density, vapor
pressure, thermodynamic properties related to vaporization, heat capacity, viscosity, compressibility, and surface tension. Some of these are summarized in table
4.1 and are discussed briey in this section.
Water is unique among polar solvents in that it is a small molecule with a low
molar volume. As a result, the concentration of water in pure water is 55.5 M.
This means that the mole fraction of water in dilute aqueous solutions is close to
one, and the partial molar quantities in these solutions are close to the corresponding quantities for the pure solvent. Other solvents have considerably higher
molar volumes, and therefore, lower concentrations in the pure solvent. The
Table 4.1 Thermodynamic Properties of Selected Polar Solvents at 25 Ca
Solvent
Molecular
Mass
/g mol1
Molar
Volume
/cm3 mol1
Enthalpy of
Vaporization
/kJ mol1
Entropy of
Vaporizationb
/K1 mol1
Molar Heat
Capacity
/J K1 mol1
18.02
32.04
46.07
60.10
74.12
45.04
59.07
18.07
40.73
58.69
75.14
92.0
39.89
59.14
44.0
37.4
42.3
47.5
52.4
65.0
59.6
108.9
104.8
110.1
111.3
110.4
98.1
99.8
75.3
81.5
112.3
143.8
177.1
107.6
123.8
58.08
41.05
103.12
69.11
87.12
73.09
78.13
179.20
99.13
123.11
61.04
55.08
102.08
120.17
116.16
74.04
52.86
103.06
87.87
93.05
77.4
71.3
175.7
96.4
102.7
54.0
70.9
85.2
95.3
120.3
31.3
32.9
55.5
39.3
49.2
47.5
52.9
61.1
54.0
55.0
38.3
36.1
65.3
79.5
45.5
88.4
84.0
98.9
88.1
98.7
90.0
93.3
112.2
94.0
84.2
91.9
83.6
93.1
107.3
101.4
124.9
91.5
190.3
155.8
176
148.4
153.2
321.3
166.4
177.1
105.8
119.7
160.2
180.0
229.6
Protic
1.
2.
3.
4.
5.
6.
7.
Water (W)
Methanol (MeOH)
Ethanol (EtOH)
1-Propanol (PrOH)
1-Butanol (BuOH)
Formamide (F)
N-methylformamide (NMF)
Aprotic
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
a
Acetone (AC)
Acetonitrile (AcN)
Benzonitrile (BzN)
Butyronitrile (BuN)
Dimethylacetamide (DMA)
Dimethylformamide (DMF)
Dimethylsulfoxide (DMSO)
Hexamethylphosphoramide (HMPA)
N-methylpyrrolidinone (NMP)
Nitrobenzene (NB)
Nitromethane (NM)
Propionitrile (PrN)
Propylene carbonate (PC)
Tetramethylenesulphone (TMS)
Tetramethylurea (TMU)
Obtained from compilations by Marcus [2] and Riddick et al. [G4].

Estimated at the normal boiling point.
POLAR SOLVENTS
151
concentration of methanol in pure methanol is 24.6 M; the corresponding concentration of 1-butanol is 10.9 M. The largest solvent molecule considered here is
hexamethylphosphoramide, which has a concentration of only 5.7 M in the pure
liquid. The molar volume of the solvent is obviously an important parameter in
assessing the degree of departure from ideality of its solutions.
The enthalpy of vaporization is an important parameter which can be used
to assess intermolecular forces in a given liquid. As discussed earlier, the ratio
vap Hm =Vm gives the internal pressure of the liquid (section 1.7). The entropy of
vaporization is also useful. When it is measured at the normal boiling point,
this quantity is close to 89 kJ mol1 in an unstructured solvent according to
Troutons rule. The data in table 4.1 show that vap S is signicantly greater
than 89 J K1 mol1 for the protic solvents, which are hydrogen-bonded liquids.
However, this quantity is also larger in some aprotic solvents for which dipole
dipole interactions are strong.
The dielectric properties of the solvents considered here are summarized in
table 4.2. These properties are important in evaluating the solvation of ions in
polar solvents under both static and dynamic conditions. The relative permittivity
of a solvent at high frequencies, eop, can be calculated from the refractive index,
nop, the relationship being
eop n2op
4:2:1
Relaxation parameters obtained by studying the frequency dependence of the

relative permittivity are also summarized in this table. In the case of the aprotic
solvents the summarized parameters were usually obtained by analyzing the
experimental data on the basis of the Debye model, which is discussed in detail
later in this chapter. Protic solvents follow a more complex behavior because of
hydrogen bonding and are generally considered to exhibit multiple relaxations.
For these systems the relaxation time listed in table 4.2 is that for the slowest
solvent relaxation, whereas the value of the high-frequency permittivity, e1, is
that for the fastest relaxation process. More details about the behavior of these
systems are given later. It should be noted that e1, which is usually measured in
the far infrared region, is slightly higher then eop, which is measured at visible
optical frequencies. The distinction between these quantities results from the fact
that the Debye treatment is based on a continuum model which does not recognize the molecular nature of the solvent and the features of its vibrational spectrum in the infrared region.
Some molecular properties of polar solvents are summarized in table 4.3. The
dipole moment and molecular polarizability are the molecular parameters which
lead to the solvent permittivity. The other parameters listed are the molecular
diameter and the LennardJones interaction energy, eLJ. These are of interest in
assessing the role of van der Waals forces in determining the properties of a polar
liquid.
In assembling the properties of polar solvents, attention has been focused on
their dipolar properties, that is, their ability to stabilize electrolytes in solution. As
will be seen in the following chapter, it is not sucient to consider only the
electrical properties of these molecules as dened in classical electrostatics. The
Table 4.2 Dielectric Properties of Selected Polar Solvents at 25 Ca
Solvent
Relative
Static
Permittivity
es
Temperature
Coecient
des =dT
Refractive
Index
nop
Temperature
Coecient
dnop =dT 104
Debye
Relaxation Timec
tD =ps
Temperature
Coecientb
HD =kJ mol1
High-Frequency
Permittivity
e1
78.46
32.70
24.55
20.33
17.51
111.0
182.4
0.360
0.197
0.147
0.142
0.132
0.72
1.6
1.3325
1.3265
1.3594
1.3837
1.3973
1.4468
1.4300
1.05
3.83
4.0
3.72
3.9
1.44
3.8
8.32
51.5
163
329
517
37.3
128
18.2
15.8
18.0
20.0
23.8
15.2
19.3
4.49
2.79
2.69
2.44
2.22
4.48
3.20
20.7
35.9
25.2
24.5
37.78
36.71
46.68
30.0
32.0
34.82
35.8
28.2
66.1
43.7
23.1
0.0977
0.160
0.091
0.108
0.213
0.178
0.106
0.175
0.140
0.180
0.161
0.119
0.240
0.115
1.3560
1.3416
1.5259
1.3820
1.4356
1.4282
1.4773
1.45
1.4680
1.5500
1.3796
1.3636
1.4209
1.4820
1.4493
5.0
4.5
4.8
4.3
5.4
4.6
3.58
3.6
5.0
4.6
4.5
4.5
3.75
3.4
3.2
3.2
37.9
6.6
16.0
10.4
18.9
80.0
21.0
45.0
4.0
5.0
43.1
31
6.5
8.3
(10.6)
8.8
11.6
10.0
11.7
(14.6)
(12.1)
13.1
6.6
8.65
15.8
1.89
2.26
3.80
(2.5)
3.04
2.94
5.3
3.3
4.06
4.10
2.0
(2.4)
4.14
4.5
Proticc
1.
2.
3.
4.
5.
6.
7.
W [5, 6]
MeOH [5, 6]
EtOH [5, 6]
PrOH [5, 7]
BuOH [7, 8]
F [6, 9]
NMF [6, 9]
Aprotic
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
a
AC [10]
AcN [9, 11]
BzN [12]
BuN [11]
DMA [9, 13]
DMF [6, 9]
DMSO [14]
HMPA [15]
NMP [16]
NB [12]
NM [6]
PrN [11]
PC [6, 9]
TMS
TMU [17]
The majority of the data were obtained from compilations by Marcus [2], Riddick et al. [G4], and Karapetyan and Eychis [3]. Specic references to relaxation data are given with
each solvent. Numbers in parentheses are estimates.
b
Temperature coefcient of D dened as R @ln D/@(1/T).
c
In the case of the protic solvents which have several relaxation processes, the value of D is for the rst relaxation.
POLAR SOLVENTS
153
Table 4.3 Molecular Properties of Selected Polar Solvents
Solvent
Dipole Momenta
p/debye
Polarizabilityb
103 ap /nm3
Diameterc
s/nm
LennardJones
Energyd
eLJ k1
B =K
1.85
1.70
1.69
1.58
1.66
3.73
3.83
1.47
3.26
5.13
6.96
8.79
4.23
6.05
0.274
0.371
0.436
0.498
0.540
0.382
0.452
79
234
324
320
2.88
3.92
4.18
4.07
3.80
3.82
3.96
5.54
4.09
4.22
3.46
4.05
4.98
4.81
3.47
6.41
4.42
12.54
8.11
9.63
7.90
7.99
18.97
10.62
12.97
4.95
6.26
8.56
10.77
12.80
0.476
0.427
0.574
0.532
0.548
0.517
0.491
0.698
0.569
0.574
0.431
0.477
0.536
0.581
0.544
362
275
520
450
380
333
670
609
298
400
Protic
1.
2.
3.
4.
5.
6.
7.
W
MeOH
EtOH
PrOH
BuOH
F
NMF
Aprotic
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
AC
AcN
BzN
BuN
DMA
DMF
DMSO
HMPA
NMP
NB
NM
PrN
PC
TMS
TMU
From gas phase data when possible [4] or from the compilations of Marcus [2] (1 debye is equal to 3:3356
1030 C m).
b
Estimated using the LorentzLorenz equation (4.3.21).
c
Estimated from gas phase solubilities [18]; when these data were not available, the solvents diameter was
estimated on the basis of a linear correlation between existing experimental data and the cube root of the molar
volume.
d
From the compilations of Marcus [2].
a
specic chemical interactions between an ion and polar solvent molecule are also
important. Methods of including consideration of both electrical and chemical
interactions are outlined later in this chapter.
4.3 The Static Solvent Permittivity on the Basis

of Continuum Models
The task of statistical thermodynamics in describing the dielectric properties of a
liquid is to develop a model for the liquid permittivity in terms of the dipole
moment and polarizability of the component molecules. The usual approach has
been to assume that the liquid molecule can be represented as a polarizable hard
154
sphere with a point dipole which is embedded in a dielectric continuum. In the

presence of an applied external eld E, the molecule experiences a local eld Ee . An
important feature of the model is the method used to estimate the local eld.
A. The Debye Model
The rst solution to this problem as given by Debye [1] for the case of dipolar
molecules in the gas phase. The potential energy of a molecule with dipole
moment p in a eld E is equal to
u pE cos y
4:3:1
where y is the angle between the eld vector E and the dipole. The degree of
orientation of the molecules in the eld varies with the temperature upon which
randomizing thermal motion depends. The average value of the angle y is now
determined on the basis of the Boltzmann distribution law. The number of molecules lying within a solid angle, d, in the direction y, is given by
dNV AN ebu d
4:3:2
where AN is a constant related to the total number of molecules per unit volume,
NV and b 1/(kB T). It follows that
2p
NV
AN expbp E cos yd
4:3:3
where the integration is carried out over all possible values of the solid angle, .
The increment d can be estimated by considering cones of semi-angle y and y
dy dened with respect to the eld vector, E. On this basis it follows that
d 2p sin y dy
4:3:4
The component of the dipole moment in the direction of the eld is p cos y. To
obtain the average value hpi, one adds up values of p cos y over all possible angles.
Thus, the average is
p
hpi
p cos y expbp E cos y sin y dy

p
4:3:5
expbp E cos y sin y dy
After integration by parts one obtains

hpi p cothbpE
1
bE
4:3:6
The function coth x 1=x is called the Langevin function so that the average
dipole moment is equal to its magnitude times the Langevin of bpE. The value of
bpE is usually much smaller than unity. For instance, the value of bp for a
molecule with a dipole moment of 3 debyes is 2.4 109 m V1 at 25 C. The
applied eld would have to be 4.1 108 V m1 in order to make bpE unity. The
POLAR SOLVENTS
155
applied eld is normally much smaller, so that the average value of the dipole
moment can be estimated using the value of coth(bpE) in the limit that bpE is
small. Expanding the exponential terms in coth(bpE) retaining the rst four terms
in the expansion, one obtains
b2 p2 E2
1
bpE
2!
cothbpE

3
b3 p3 E3 bpE
2bpE 2
3!
22
4:3:7
Thus, the approximate expression for hpi is

hpi
bp2 E
3
4:3:8
The total polarization of the medium per unit volume considering also the molecular polarizability becomes
!
bp2
P NE p
4:3:9
3
The units of the polarization P are coulombs per square meter; this quantity gives
the eective dipole of the medium per unit volume containing NV molecules.
EXAMPLE
A parallel-plate capacitor is set up at 25 C with a voltage of 105 V between the

plates, which are separated by 1 mm. Dimethylsulfoxide (DMSO) vapor at low
pressure is introduced into the gap between the plates. Estimate the average
dipole moment of each DMSO molecule in the direction of the eld.
The eld in the capacitor gap is 108 V m1 . Since a dilute gas is present in the
gap the local eld can be assumed to be equal to the applied eld.
The dipole moment of DMSO is 3.96 debyes (table 4.3):
p 3:96 3:336 1030 1:32 1029 C m
4:3:10
The average dipole moment in the direction of the eld is

hpi
1:32 1029 2 108

1:41 1030 C m
3 1:38 1023 298:2
4:3:11
This is equivalent to 0.423 debyes.

Even though the eld is quite high, the eects of thermal motion result in the
average dipole moment of the molecule in the direction of the eld being much
less than the permanent dipole moment.
In applying this relationship to dipolar molecules in dense gases and liquids the
eld E is understood to be the local or eective eld Ee. In order to estimate Ee
one must solve the statistical mechanical problem which relates the local quantity
to the macroscopic applied eld E. The problem is solved by estimating the local
eld in a spherical cavity within the dielectric (g. 4.1). The cavity is assumed to
have molecular dimensions with diameter a. The material within the sphere is
considered in terms of individual molecules, whereas that outside the sphere is
156
Fig. 4.1 Schematic diagram of a

spherical cavity within a
homogenous dielectric between
the plates of a parallel-plate
capacitor.
assumed to have homogeneous properties with an isotropic relative permittivity

es. One also imagines the dielectric to be between the plates of a parallel-plate
capacitor so that the eld and polarization vectors have a dened direction.
In order to estimate Ee one must consider three contributions, namely, the eld
E1 imposed on the medium via the charge on the plates of the capacitor, the eld
E2 due to the polarization charge on the surface of the spherical cavity, and
nally, the eld E3 due to the molecules within the sphere. E1 is obviously identied with the experimentally controlled eld E. The eld E2 is calculated by integrating the polarization in the direction of the eld over the surface of the
spherical cavity. This gives
p
E2
0
P cos y2pr2 sin y dy

4pe0 r2
4:3:12
where 2pr2 sin y dy is the area of an element on the sphere dened by the angles y
and y dy. The result is
E2
P
3e0
4:3:13
In order to estimate E3, one must know how the individual molecules are
arranged within the cavity. For example, if they are located on a cubic lattice
or when the molecules form an ideal gas, E3 is zero. In the solution to the problem
given here, E3 is assumed to be zero. Thus,
Ee E
P
3e0
4:3:14
This relationship was originally derived by Lorentz [19].

In order to proceed further one must make use of the fundamental electrostatic
relationship between the electric displacement D, the electric eld E, and the
polarization P, which is
D e0 E P e0 es E
4:3:15
P e0 Ees 1
4:3:16
It follows that
POLAR SOLVENTS
157
It is clear from these relationships that the electric displacement D has the same
units as the polarization P, namely, C m2 . Now, from equation (4.3.14)
Ee
es 2
E
3
4:3:17
Identifying the eld in equation (4.3.9) as the local eld Ee when this expression is
applied to liquids, and substituting in equation (4.3.16) for P, one obtains the
result
!
es 1
N
bp2

4:3:18
es 2 3e0 p
3
for N molecules. For one mole of molecules, one may write
Nv
NL
N
Lr
Vm
M
4:3:19
where NL is the Avogadro constant, Vm , the molar volume of the liquid, r, its
mass density, and M, its molecular mass. Equation (4.3.18) is then rewritten in the
form
!

es 1 M N L
bp2
Pm

4:3:20
es 2 r
3e0 p
3
where the quantity on the left-hand side is called the molar polarization. It is
familiar from the analysis of dielectric permittivity data for liquids and solids, and
is usually expressed in units of cubic centimeters per mole. Equation (4.3.20) is the
well-known Debye equation.
For pure liquids, the Debye equation suggests that the molar polarization
should be a linear function of the reciprocal temperature. Furthermore, one
should be able to analyze relative permittivity data for a polar liquid like water
as a function of temperature to obtain the dipole moment and polarizability from
the slope and intercept, respectively. In fact, if one constructs such a plot using
data for a polar solvent, one obtains results which are unreasonable on the basis
of known values of p and ap from gas phase measurements. The reason for the
failure of the Debye model in liquids is the fact that it neglects the eld E3 due to
dipoles in the immediate vicinity of a given molecule. However, it provides a
reasonable description of the dielectric properties of dilute polar gases. In liquids,
relatively strong forces, both electrostatic and chemical, determine the relative
orientation of the molecules in the system, and lead to an error in the estimation
of the orientational component of the molar polarization.
There are two circumstances under which equation (4.3.20) proves useful. The
rst of these is at frequencies suciently high that the orientational component of
Pm is eectively zero. Then, one can write

eop 1 M N L ap
Rm
4:3:21
eop 2 r
3e0
This is called the LorentzLorenz equation, and is used to estimate the molecular
refraction Rm from the refractive index or eop. Since the polarizability ap is often
158
not known from gas phase data, it has been customary to estimate it for many
molecules using equation (4.3.21).
EXAMPLE
Given that the refractive index of water at 25 C is 1.3325 and its molar volume
18.07 cm3, estimate the molecular polarizability of the water molecule. Then
estimate the same quantity at 50 C where the value of n is 1.3291, and the
molar volume 18.234 cm3.
At 25 C
eop n2 1:7756
4:3:22
eop 1
0:7756
18:07 3:712 cm3 mol1
Vm
3:7756
eop 2
4:3:23
Thus,
The corresponding value of ap is

ap
3:712 106 3e0

1:637 1040 C m2 V1
NL
4:3:24
This result may be expressed in the more customary units of cubic meters or
cubic nanometers by dividing by 4pe0 . Thus, ap is equal to 1:472 103 nm3 at
258C. When the calculation is repeated at 508C the result is ap 1:471 103
nm3 . One expects the polarizability to be independent of temperature in a
range where the electrons in the molecule remain in the same molecular orbitals. The small change in the polarizability reects the weakness of the
LorentzLorenz model, which is based on continuum concepts. However,
the estimated change is small, so that one may assume that the model is reasonably good.
The second situation where equation (4.3.20) is often applied is to the analysis
of data for the dielectric permittivity of dilute solutions of a polar compound in a
non-polar solvent such as benzene or carbon tetrachloride. Under these circumstances local structure due to dipoledipole interactions can be neglected, and the
dipole moment of the polar solute can be calculated from the change in molar
polarization with concentration of the polar solute. Then, the molar polarization
can be written

es 1 M1 x1 M2 x2
4:3:25
Pm x1 Pm1 x2 Pm2
es 2
r
where Pm1 is the molar polarization of the non-polar solvent, Pm2 , that due to the
polar solute, and r, the density of the solution. Pm1 is assumed to be constant,
independent of solution composition, and Pm2 is determined as a function of the
mole fraction x2 . From the experimental data, one extrapolates to x2 0 to
obtain the limiting value, P0m2 . The dipole moment is then calculated using the
Debye equation with the polarization component being estimated from the corresponding data for the refractive index. Although this method has been often
POLAR SOLVENTS
159
applied, it can be criticized on fundamental grounds [G1]. Therefore, molecular

data obtained using this approach should be regarded as approximate.
EXAMPLE
Using the data given below for dilute solutions of nitromethane in benzene,
calculate the molar polarization and refraction of the solute in the limit of
innite dilution. Then, estimate the dipole moment of nitromethane on the
basis of the Debye model.
The following data were reported by Balakier [20] at 293.2 K.
Mole Fraction
of Nitromethane
x2
Solution
Density
r/g cm3
Relative
Permittivity
es
Refractive
Index
nop
0.0000
0.0197
0.0392
0.0593
0.0798
0.8789
0.8830
0.8860
0.8893
0.8926
2.284
2.587
2.863
3.188
3.457
1.5011
1.4992
1.4976
1.4958
1.4943
Given that the molecular weight of nitromethane is 61.04 and that of benzene, 78.12, one rst calculates the molar volume of the solution. For example,
when x2 0:0197,
Vm
61:04 0:0197 78:12 0:9803

88:09 cm3 mol1
0:8830
4:3:26
The molar polarization is estimated using the dielectric constant following

equation (4.3.20):
Pm 88:09
1:587
30:48 cm3 mol1
4:587
4:3:27
The molar refraction is calculated in the same way using the optical dielectric
constant, that is, the square of the refractive index:
Rm 88:09
1:50112 1
25:87 cm3 mol1
1:50112 2
4:3:28
The results obtained as a function of mole fraction are tabulated below.

Mole Fraction
of Nitromethane
x2
Molar
Volume
Vm =cm3 mol1
Molar
Polarization
Pm =cm3 mol1
Molar
Refraction
Rm /cm3 mol1
0.0000
0.0197
0.0392
0.0593
0.0798
88.88
88.09
87.42
86.71
85.99
26.64
30.48
33.49
36.57
38.72
26.19
25.87
25.61
25.32
25.05
160
Now one calculates the molar polarization due to the solute. For example,
when x2 0:0197,
Pm2
30:48 26:64 0:9803

221:4 cm3 mol1
0:0197
4:3:29
Similarly, for the molar refraction due to the solute:

Rm2
25:87 26:19 0:9803

10:1 cm3 mol1
0:0197
4:3:30
Values of Pm2 and Rm2 are plotted against the mole fraction x2 in g. 4.2.
Extrapolation to x2 0 gives the values of these parameters in the limit of
innite dilution. They are Pm20 232:7 cm3 mol1 and Rm20 9:8 cm3 mol1 .
From equation (4.3.20), one may write
Pm20 Rm20
NL bp2
9e0
4:3:31
Thus,
9 8:854 1012 1:381 1023 293:2 232:7 9:8 106
p
6:022 1023
1:09 1029 C m 3:27 debyes
!1=2
4:3:32
B. Other Models for the Static Permittivity

Another approach to a continuum model for the electric permittivity which considers the eld E3 was given by Onsager [21]. This model also makes use of a
Fig. 4.2 Plot of the molar polarization Pm2 and molar refraction Rm2 for nitromethane
against its mole fraction in benzene.
POLAR SOLVENTS
161
spherical cavity but limits it to one molecule. As a result it is applicable only to

liquids without strong directional forces. The equation obtained by Onsager is
es eop 2es eop 2
es eop 22
Vm
NL bp2
9e0
4:3:33
where Vm is the molar volume of the liquid. It is clear that the distortional
contribution to the polarization is not explicitly available in this expression but
is included in a complex way in the quantity on the left-hand side. The model
corrects some of the problems of the Debye model but in the end it still fails to
account for the correlation of a given dipole to its neighbors.
Further developments in the theory of the structure of polar liquids included
estimates of the correlation of a given dipole to its neighbors. Important contributions were made in this direction by Kirkwood [22] and Frohlich [23]. In
Kirkwoods model, the eld E3 is calculated by considering all possible orientations of surrounding dipoles in a spherical cavity for a xed orientation of
the central dipole. By averaging over these orientations, Kirkwood obtained
an improved estimate of the polarization of the medium. For the case of nonpolarizable dipoles the result is
es 12es 1
N g bp2
Vm L k
3e0
3es
4:3:34
where gK is a correlation parameter describing the degree of local order. This

parameter is unity when the average dipole moment of the molecules immediately
surrounding a central molecule which is held xed is the same as the dipole
moment of an isolated molecule. In other words, xing the position of one molecule does not aect the positions of the surrounding dipoles except through longrange electrostatic forces. On the other hand, if the neighboring dipoles tend to
line up in a parallel fashion as a result of the xing of a central dipole, the
parameter gK rises to values greater than unity. Similarly, if xing one dipole
tends to line up the surrounding dipoles in an antiparallel direction, the parameter
gK will be less than one. Equation (4.3.34) diers from the Onsager equation
(equation (4.3.33)) in two ways. If the distortional contribution in equation
(4.3.33) is neglected by setting eop equal to one, and the correlation between
dipoles is neglected by setting gK equal to one in equation (4.3.34), these expressions are the same. The distortional contribution is not an important factor for
polar molecules so that the version of the Kirkwood equation given here may be
applied to the systems considered in this chapter. Inclusion of the correlation
parameter gK in this equation gives an important improvement in the description
of the interactions between molecules in these systems. This is especially important when the interactions are chemical in nature, for example, when hydrogen
bonding is present. In principle, it is possible to calculate gK if the structure of the
liquid is known, but in general, it is considered to be an adjustable parameter.
Dielectric permittivity data for water and acetonitrile in the temperature range
from 0 to 508C are plotted according to the Kirkwood equation in g. 4.3. The
straight lines shown are based on one-parameter least-squares ts, the slopes
giving the value of NL gK p2 =3kB e0 . On the basis of this analysis for water, assum-
162
Fig. 4.3 Plot of the molar polarization according to the Kirkwood model against 1000/T,
where T is the absolute temperature for water (*) and acetonitrile (^) in the temperature
range 0508C.
ing that its dipole moment is 1.85 D, the value of the correlation parameter gK is
4.5. This result is higher than one would estimate on the basis of the structure of
liquid water determined by neutron diraction and Raman spectroscopic studies
(see section 2.10). The high value reects the eects of hydrogen bonding in this
very structured liquid. In the case of acetonitrile, assuming p 3:92 D, the value
of gK is 1.3. In this solvent the degree of correlation between adjacent molecules is
much weaker. Since hydrogen bonding is absent in acetonitrile, the only source of
correlation would be the formation of dimers resulting from the antiparallel
association of the dipole vectors due to the CN groups in this molecule.
The above examples illustrate that continuum models such as the Kirkwood
model are reasonably successful in describing the static permittivity, provided one
has an independent means of estimating the correlation parameter gK .
Unfortunately, these estimates are available for only a few polar solvents, so
that gK must be considered an independent parameter. The version of
Kirkwoods theory presented here only considers orientational polarization.
When distortional polarization, that is, the eect of molecular polarizability, is
included, interpretation of experimental results is less clear. Since the approach
taken here involves continuum concepts, it is necessarily limited. In the following
section, a simple model based on a molecular description of a polar liquid is
presented.
4.4 The Static Solvent Permittivity According to the MSA

The approach taken in non-primitive models for the dielectric properties of polar
solvents is quite dierent from that described above. The rst step in a molecular
model is to calculate the pair correlation function for molecules in the liquid.
POLAR SOLVENTS
163
Then, when one knows the dipole moment and polarizability, one should be able
to estimate the dielectric constant of the medium using the position and average
orientation of each molecule in the system. Since dipoledipole interactions are
long range, the problem is not simple and approximations are made in order to
nd a solution. The usual approximation is to represent each molecule as a hard
sphere with a central point dipole p and polarizability ap . Such a representation of
the system should give a good representation of aprotic solvents but does not have
any feature which is able to describe hydrogen bonding.
A treatment for polar solvents on the basis of the mean spherical approximation was rst given by Wertheim [24, 25]. The closure conditions are based simply
on the dipoledipole interaction energy between the polar molecules in the system.
Neglecting molecular polarizability, these conditions are
go1 ; r; o2 0
for
r > ss
4:4:1
and
co1 ; r; o2
bp2
s1 sr s2 sr s1 s2
4pe0 r3
for
r > ss
4:4:2
where ss is the diameter of a hard sphere, sr , a unit vector in the direction r, and s1
and s2 , unit vectors in the directions of the two dipoles which give rise to the
interaction energy. By solving the OZ equation (see section 2.6), using this simple
expression for the direct correlation function, cr, one can nd the form of the
pair correlation function gr as a function of distance from any given hard
sphere, averaging out the angular dependence. One may then estimate the thermodynamic properties of the system using the methods described earlier (see
section 2.8).
On the basis of Wertheims solution of the MSA for dipolar hard spheres it is
convenient to dene a polarization parameter ls which is obtained directly from
the relative static permittivity, that is, by solving the equation
l2s 1 ls 4 16es
4:4:3
The value of ls for a typical polar solvent is approximately two. This equation
was introduced earlier in the development of the MSA for ionsolvent interactions (section 3.5). It was seen that the MSA gives an improved description of ion
solvation parameters with respect to the Born model. However, it fails to distinguish between the solvation of cations and anions of the same size. In other
words, it fails to distinguish between the short-range chemical interactions
which stabilize ions of diering charge.
It is obvious that a model for the dielectric properties of a polar solvent must
consider the eects of hydrogen bonding and other chemical interactions which
are not included in a point dipole model for solvent molecules. A simple way of
doing this is to introduce a directional sticky parameter [26]. Thus, each solvent
molecule is represented by a hard sphere with a central point dipole which is
polarizable. The surface of the hard sphere is modied so that it can interact
attractively or repulsively with the surface of an adjacent sphere. Stickiness is
an eective method of introducing short-range chemical forces. It amounts to
assuming that interactions along the line of the dipole vector may be stronger
164
or weaker than one estimates on the basis of the simple point dipole model.
Equation (4.4.3) then becomes

t0 2
2
4
16es
4:4:4
ls 1 ls 1
ls
where t0 is the longitudinal stickiness parameter.
Further development of the theory leads to an expression for the excess
internal energy in terms of molecular solvent parameters and ls . The result is
ex

Um
dp2 b2 bVm
2ps3s
4:4:5
where
b2
6ls 1
ls 2
4:4:6
Vm is the molar volume of the solvent, and dp , a parameter related to the strength
of dipoledipole interactions. With consideration of dipole polarizability, dp is
given by
"
#
NL a2 bp2 a ap
2
dp
4:47
Vm 3e0
e0
where
a
3s2s
3s3s b2 ap pe0
4:4:8
and NL is the Avogadro constant. The parameter dp may also be estimated on the
basis of the polarization parameter ls and the stickiness parameter t0 dened by
equation (4.4.4). This relationship is
"
#

l2s ls 22
t0 2
16
2
1
dp
2
4:4:9
9
ls
ls 1 ls 4
Equations (4.4.4)(4.4.9) give a relationship between the molecular parameters p
and ap and the macroscopic relative permittivity es via the two MSA parameters
ls and t0 . Equations (4.4.4) and (4.4.9) are non-linear equations from which one
obtains ls and t0 once the parameter dp has been estimated.
The above equations can be solved using iterative techniques to obtain values
of ls , t0 , and dp which are self-consistent. First of all, equation (4.4.4) is solved for
ls , assuming that t0 is zero. This value of ls is used to obtain an initial estimate for
dp using equations (4.4.6), (4.4.7), and (4.4.8). Equation (4.4.9) is now solved for
t0 . The entire calculation is repeated using this estimate of t0 to obtain a second
estimate of ls from equation (4.4.4). By going through an iterative scheme several
times, the solution converges to give constant values of ls and t0 . In the case of
water, the revised estimate of ls is 2.50 and the stickiness parameter is 0:40. The
negative value of t0 indicates that dipoledipole interactions are enhanced by the
sticky chemical interactions. In the case of water, dp2 is 30% larger with these
interactions than when estimated without them.
POLAR SOLVENTS
165
EXAMPLE
Estimate the MSA parameters for DMSO using the dielectric properties given
in table 4.2 and the molecular properties in table 4.3.
The rst step is to estimate ls on the basis of the Wertheim equation (4.4.3)
and the value of es (46.68). The equation is
ls 1 ls 2 16 46:681=2 27:33
4:4:10
The solution of this cubic equation gives ls 2:385.

The molecular properties of DMSO are
ss 4:91 1010 m
p 3:96 3:336 1030 1:32 1029 C m
ap 4pe0 7:99 1030 C m2 V1
The second step is to estimate dp from equation (4.4.9). The estimate of b2
(equation (4.4.6)) is
b2
6 1:385
1:895
4:385
4:4:11
The estimate of a (equation (4.4.8)) is

a
3 4:91 1010 3
1:206
3 4:91 1010 3 4 7:99 1030 1:895
4:4:12
The number density for the solvent molecules is

NL 6:022 1023
8:45 1027 molec m3

Vm
7:13 105
4:4:13
It follows that
1:2062 8:45 1027 1:32 1029 2
4p 1:206 8:45
3 8:854 1012 4:115 1021
1027 7:99 1030
19:60 1:02 20:62
4:4:14
dp2

9dp2
t0 2
16
1
2
2
2
ls
ls ls 2
ls 1 ls 4
9 20:61
16
1:718
2
2
2:385 4:385
2:3852 3:3854
and t0 0:74:
Now a new value of s is estimated using equation (4.4.4):

16 46:68 1=2
ls 1 ls 2
20:85
1:718
4:4:16
166
The result is ls 2:129. dp2 is now estimated again, the result being 19.56. The
iteration process is repeated until successive values are the same to a reasonable
level of precision. The results for DMSO are ls 1:85 and t0 1:53.
Values of ls and t0 have been estimated for the polar solvents considered in this
chapter and are summarized in table 4.4. Also included in this table are values of
the Kirkwood correlation parameter gK . It is dened in the MSA model as
gK
es 12es 1
3es dp2
4:4:17
As can be seen from the results the values of ls are usually close to two. In
general, they are higher for the protic solvents than the aprotic ones. The stickiness parameter is negative for aprotic solvents and positive for most protic solvents. A negative value of t0 indicates that the degree of order in the solvent is
higher than one predicts on the basis of the simple point dipoledipole model. It
also reects the fact that the dipole moment is not centrally located in the solvent
molecules considered.
Table 4.4 MSA Parameters for Selected Polar Solvents
Solvent
Polarization
Parameter
ls
Stickiness
Parameter
t0
Molecular
Polarizability
103 ap /nm3
Kirkwood
Correlation
Parameter
gK
2.50
2.58
2.62
2.68
2.69
2.08
3.28
0.40
0.80
1.11
1.36
1.47
2.36
0.33
1.48
3.29
5.21
7.10
8.99
4.26
6.20
2.79
2.99
3.08
3.23
3.26
2.04
4.52
1.70
1.47
1.52
1.39
1.99
1.80
1.85
1.71
1.74
1.77
1.55
1.45
1.98
1.74
1.81
0.79
2.53
1.65
1.93
0.77
1.30
1.53
1.28
1.29
1.32
2.14
2.17
1.70
1.77
0.62
6.51
4.48
13.05
8.29
9.91
8.12
8.24
19.61
10.99
13.54
5.04
6.37
8.80
11.19
13.16
1.49
1.18
1.23
1.09
1.89
1.60
1.67
1.44
1.52
1.56
1.38
1.15
1.86
1.53
1.64
Protic
1.
2.
3.
4.
5.
6.
7.
W
MeOH
EtOH
PrOH
BuOH
F
NMF
Aprotic
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
AC
AcN
BzN
BuN
DMA
DMF
DMSO
HMPA
NMP
NB
NM
PrN
PC
TMS
TMU
POLAR SOLVENTS
167
A trend in the value of t0 can be seen for a series of solvents with the same polar
groups and alkyl groups of increasing size. For example, for water and the alcohols, t0 increases from 0:40 for water to 1.47 for butanol. A similar trend is seen
for the nitriles where t0 increases from 2:54 for acetonitrile to 1:67 for benzonitrile. These values of t0 are plotted in g. 4.4 against b2p =d 3 e0 , the quantity which
characterizes the strength of dipoledipole interactions in the medium. There is a
clear linear relationship between these parameters. As the size of the alkyl groups
increases, the dipolar group moves further away from the center of the molecule.
This results in a decrease in dipole moment, an increase in molecular size, and a
corresponding decrease in the strength of dipoledipole interactions. These
changes require that the stickiness parameter change as well to account for the
fact that the dipoledipole interactions cannot be treated by the simple point
dipole model. Examination of the data for aprotic solvents in table 4.4 also reveals
a general trend according to which t0 becomes more negative as the strength of
dipoledipole interactions increases [26].
The estimates of the correlation parameter gK give an idea of the degree of
local structure in the polar solvent. In the case of water and the alcohols, this
parameter is relatively high and increases with the length of the hydrocarbon
chain. The result for water, namely, 2.79, is remarkably close to the value estimated by Kirkwood [22] on the basis of the well-known tetrahedral structure for
the surrounding hydrogen-bonded water dipoles (2.65) (see also g. 2.14) [G2].
The increase in gK in the alcohols suggests a tendency to form hydrogen-bonded
chains in these systems. The value of gK is also large for N-methyl formamide, a
strongly structured solvent [27]. In the case of the aprotic nitriles, the average
value of gK is 1.2. This result indicates that local ordering is minimal in these
systems. Dimethylsulfoxide, which has approximately the same dipole moment as
acetonitrile, is known to be strongly dimerized in the liquid state [28]. As a result it
Fig. 4.4 Plot of the stickiness parameter t0 against the values of bp2 =e0 d3 ), which characterizes the strength of dipoledipole interactions for two selected solvent groups.
168
has a higher value of gK . An even higher value is found for propylene carbonate
(1.86), which is also associated in the liquid state. In summary, the values of gK
give a useful indication of local order in polar solvents and may be related to more
detailed structural information when this is available.
One may also use the MSA model to describe the permittivity of the system at
optical frequencies. Under these circumstances the system responds to the electrical eld only through electronic polarization, the orientational component
being frozen. The directional stickiness of the dipoles is then unimportant so
that t0 is eectively zero at very high frequencies. Under these circumstances, the
polarization parameter is given by
l2op 1 lop 4 16eop
4:4:18
where lop is the value of the polarization parameter at optical frequencies. The
expression for dp , the parameter which characterizes the excess energy of the
system, becomes
dp2
NL aap
e0 V m
4:4:19
where a is given by equation (4.4.8) and

b2
6lop 1
lop 2
Alternatively, dp may be estimated from the following equation:

l2op lop 22
1
1
dp2
eop
9
4:4:20
4:4:21
Since values of the molecular polarizability are not available from gas phase
measurements for most of the polar solvents considered here, they may be estimated using the MSA in the following way. First, a value of lop is calculated by
solving equation (4.4.18) using the experimental value of eop . Then, the parameter
dp is estimated using equation (4.4.21). Finally, the polarizability is found from
equation (4.4.19). The results obtained are summarized in table 4.4. They are very
close to estimates obtained on the basis of the LorentzLorenz equation (see
equation (4.3.21) and table 4.3), usually being a few percent higher.
In summary, the MSA provides a remarkably simple and eective way of
describing the dielectric properties of polar solvents. By including stickiness in
the model one has a very easy method of dealing with the eects of non-sphericity,
the non-central location of the polar group, and hydrogen bonding. These eects
collectively make up the chemical part of the problem of estimating the permittivity of a polar liquid from its dipole moment and polarizability. One should note
that the polarization parameter used here is quite dierent from that introduced
in section 3.4 in the discussion of ionsolvent interactions. Although it arises in
the model in the same way, it is dierent because the chemical interaction between
a solvent dipole and an ion depends on the chemical nature of the polar group in
the solvent and whether it is able to participate in hydrogen bonding. This leads to
the necessity of introducing dierent values of this parameter for cations and
POLAR SOLVENTS
169
anions. On the other hand, the polarization parameter used here describes the
interaction of one dipole with neighboring dipoles. These examples show that
simple modication of the MSA leads to an excellent description of the properties
of real systems without recourse to complex mathematical treatments.
4.5 Dielectric Relaxation Phenomena

When a polar solvent is placed in a changing electrical eld, the molecules must
realign so that their dipole vectors maintain the orientation corresponding to
minimum energy. Because of intermolecular forces, this process does not occur
innitely fast but on a time scale which depends on the properties of the medium
and which is usually on the order of 1100 ps. Dielectric relaxation experiments
provide very useful information about molecular motion in polar liquids and the
ability of the solvent molecules to respond to changing electrical conditions.
The theory of dielectric relaxation is based on a macroscopic model which
considers how the polarization of the medium changes with time. As seen earlier,
the polarization is made up of an orientational and a distortional component so
that one may write
P Po Pd
4:5:1
When an electrical eld is applied to the system, the distortional component

relaxes very quickly on a time scale corresponding to electronic motion. On the
other hand, the orientational component relaxes on a time scale consistent with
molecular motion. In deriving an expression for the kinetics of the relaxation
process, it is assumed that the rate of relaxation is proportional to the departure
of the orientational polarization from its equilibrium value, that is,
dPo Poe Po
dt
t
4:5:2
where Poe is the equilibrium value of the orientational polarization and t is the
relaxation time. The reciprocal of t is the equivalent of the rst-order rate constant for the relaxation process.
From the fundamental laws of electrostatics, the relation between the polarization and the electrical eld is
P e0 Ee 1
4:5:3
where e represents the value of the permittivity at time t. If E is changing, then the
polarization depends on the time dependence of both E and e. At equilibrium the
total polarization is given by
Pe e0 E es 1
4:5:4
This equation shows that the static permittivity is the appropriate value when the
system is equilibrated. At very high frequencies only the distortional component
of the polarization remains so that
Pd e0 E e1 1
4:5:5
170
The permittivity e1 is the value reached at the highest frequency used in the
relaxation experiment. Normally this corresponds to microwave frequencies of
the order of 1 THz but it can extend into the far infrared.
Now, one may estimate the orientational polarization as
Po P Pd e0 E e e1
4:5:6
At equilibrium, this contribution is

Poe Pe Pd e0 E es e1
4:5:7
It is also important to be able to relate the orientational polarization to the

electric displacement D. On the basis of equation (4.3.15), one may write
D e0 E Po Pd
4:5:8
P o D e0 e1 E
4:5:9
so that
Under the usual circumstances, dielectric relaxation is studied at an electrical

eld which is controlled, for example, by controlling the voltage drop across a
parallel-plate condenser. On the basis of equation (4.5.2), substituting for Po and
Poe using equations (4.5.7) and (4.5.9), one obtains the following dierential
equation in D:
t
dD e0 e1 E
D e0 e1 E e0 Ees e1
dt
4:5:10
Suppose that the dielectric system relaxes due to the application of a constant eld
E at time t 0. Then equation (4.5.10) simplies to
t
dD
D e0 es E D e
dt
4:5:11
where D0 is the equilibrium value of the electric displacement. The dierential

equation may also be written as
dD De
dt

D De
t
4:5:12
D De D0 De et=t
4:5:13
for which the solution is
where Do is the initial value of D. This equation describes an exponential decay in

D which is characterized by a relaxation time t, from the initial value D0 to the
equilibrium value De .
Most experimental studies involve measurements of the electrical permittivity
as a function of frequency using an alternating voltage or eld E. Then, the eld
varies with time according to the equation
E Em cos ot
4:5:14
where Em is the amplitude of the a.c. eld and o is its angular frequency. These
experiments are commonly discussed in terms of the frequency f , where
POLAR SOLVENTS
o
2p
171
4:5:15
The electrical displacement also varies in a sinusoidal fashion but is not necessarily in phase with the electrical eld. The relationship between these quantities is
obtained by solving equation (4.5.10). When the alternating eld is initially
applied to the system, there are some transients which describe the initial relaxation. The solution to the dierential equation presented here ignores the initial
eects, and considers only the properties of the system at a steady state. Under
such circumstances any uctuation in the system occurs at the same frequency as
that of the perturbing eld E.
In order to make the mathematics easier, the dierential equation is solved
using a complex number representation of the sinusoidally varying quantities (see
appendix A). Thus, the eld E is written as
E Em ejot
4:5:16
where
ejot cos ot j sin ot
4:5:17
p
and j is equal to 1. By comparing the magnitudes to the quantities associated
with cos ot and sin ot, one is able to estimate the phase angle of the alternating
quantity. The steady-state solution to the dierential equation for D is given by
D AD ejot
4:5:18
Substituting this expression for D into equation (4.5.10) one obtains

jotAD ejot e0 e1 Em ejot AD ejot e0 e1 Em ejot e0 Em ejot es e1
4:5:19
Simplifying and solving for AD , the result is

AD
e0 Em es j!te1 es o2 t2 e1
1 o2 t2
4:5:20
Under non-equilibrium conditions

D e e0 E
4:5:21
A D e e0 E m
4:5:22
It follows that
Now one can write the following equation for the varying permittivity:
e e
e e
4:5:23
e e1 s 212 jot s 212
1o t
1o t
This result obtained by Debye [1] describes the frequency dependence of the
permittivity e. In the following discussion the relaxation time measured with
the electrical eld E as the controlled variable is called the Debye relaxation
time and is given the symbol tD .
From equation (4.5.23) the in-phase component of the permittivity is
e e
ein e1 s 212
4:5:24
1 o tD
172
At very low or static frequencies (o 0, the in-phase component is equal to es as

one would expect. In the limit of very high frequencies, the in-phase component is
equal to e1 . For angular frequencies which are close to 1=tD , ein drops rapidly
from es to e1 , reaching the value (es e1 =2 when ot is equal to one.
The out-of-phase component is
eout
otD es e1
1 o2 t2D
4:5:25
It is zero in the limits of very low and very high frequencies, and reaches a
maximum value equal to (es e1 when ot is one. The fact that this term is
multiplied by j in the overall expression shows that the out-of-phase component
lags the in-phase component by 908.
Plots of the in- and out-of-phase components of the permittivity of a hypothetical solvent with typical parameters (es 50, e1 2, and tD 20 ps) are shown
as a function of the logarithm of the frequency in g. 4.5. The frequency range
over which most of the change in these quantities occurs is from 100 MHz to 1
THz. The upper limit is beyond the range of most microwave experiments, which
is about 300 GHz. The out-of-phase component reaches a maximum value when
otD 1, which occurs at a frequency of 8 GHz in this example. This is also the
frequency at which the rate of change in ein with frequency is a maximum.
Obviously, the frequency range shown in this plot could not be covered in a
normal microwave experiment. Thus, extrapolation techniques are often used to
estimate e1 and tD from real experimental data.
An alternative experiment is one in which permittivity changes are studied at
constant electrical displacement D, for example, for constant charge on the plates
of a capacitor. The resulting relaxation phenomena are then characterized by a
Fig. 4.5 Plots of the in-phase, ein , and out-of-phase, eout , components of the dielectric
permittivity of a hypothetical Debye solvent against the logarithm of frequency. The parameters assumed are ein =50, eout =2, and tD 20 ps.
POLAR SOLVENTS
173
dierent relaxation time and it is useful to examine this situation further. On the
basis of equation (4.5.10) when D is constant,
tD e0 e1
dE
D e0 es E
dt
4:5:26
Rearranging, one may write

tD e1 dE D e0 es E
e
es
es 0 dt
4:5:27
de0 es E D
D e0 es E
dt
4:5:28
or
tL
where
tL
e1 tD
es
4:5:29
tL is known as the longitudinal relaxation time. Integration of equation (4.5.28)

leads to the result that
D e0 es E D e0 es E0 et=tL
4:5:30
where E0 is the initial value of E. Since the electric displacement is constant and
equal to the equilibrium value De , equation (4.5.30) can be rewritten as
E Ee E0 Ee et=tL
4:5:31
where De is related to Ee through equation (4.5.21). Equation (4.5.31) shows that

the decay of E back to its equilibrium value Ee is described by a rate constant
1=L . It is analogous to equation (4.5.13), which was obtained for the case of
constant electrical eld E.
The relaxation properties of a polar solvent may be studied by carrying out
experiments in which the electrical eld E or displacement D are changed suddenly in a step fashion. In the case that the eld is the independently controlled
quantity, one follows the change in D with time. An analysis of the transient
observed experimentally allows one to determine the Debye relaxation time tD
on the basis of equation (4.5.13). Similarly, if one studies the relaxation transient
under conditions that D is suddenly changed, for instance, by suddenly changing
the charge on the plates of a capacitor, one observes a transient involving the
electrical eld E. Analysis of these data allows one to determine the longitudinal
relaxation time tL . If both experiments are conducted, comparison of tD and tL
allows one to estimate the high-frequency permittivity e1 assuming that the static
permittivity is already known.
However, dielectric relaxation experiments are usually conducted in the presence of sinusoidally varying electrical elds as a function of frequency. Many
such experiments have been carried out in various pure liquids and solutions [G5,
416, 29], and values of the in-phase and out-of-phase permittivity reported as a
function of frequency over a wide range. Analysis of these data to obtain es , e1 ,
and tD can be carried out in a variety of ways. On the basis of equations (4.5.24)
174
and (4.5.25), eliminating the common factor es e1 =1 o2 t2D , one obtains the
equation
ein e1
eout
otD
4:5:32
This shows that a plot of ein against eout =o should be linear, with e1 being
obtained from the intercept and 1=tD from the slope. This method of data analysis
normally involves extrapolation because the experimental frequency range does
not include values of ein close to e1 . An alternative way of analyzing the data
results in estimates of es and tD from a linear plot. Multiplying equation (4.5.25)
by otD and adding (es e1 =1 o2 t2D to both sides, one obtains
otD eout
es e1
es e1
1 o2 t2D
4:5:33
Combining this with equation (4.5.24), the result is

ein es otD eout
4:5:34
Accordingly, a plot of ein against oeout should be linear with a slope of tD . Such
a plot is shown in g. 4.6 using data obtained for water at 258C [G5]. A linear
relationship is obtained with the data for frequencies less than 40 GHz. On the
basis of a straight line tted in this region, the value of es from the intercept is
78.20 and the value of tD from the slope, 8.38 ps r 0:9998.
A plot of the same data according to equation (4.5.32) is shown in g. 4.7. A
straight line can be drawn through the points obtained at frequencies below 40
GHz. The slope of this line gives 1=tD from which the value of tD is 8.37 ps. This
is in excellent agreement with the result from the previous plot. The intercept gives
the value of e1 , which is 6.02 for this system.
Fig. 4.6 Plot of ein against oeout using dielectric relaxation data for water at 258C [G5]. The
straight line is drawn considering only the data obtained at frequencies below 40 GHz.
POLAR SOLVENTS
175
Fig. 4.7 Plot of ein against eout =o using dielectric relaxation data for water at 258C [G5].
The straight line was tted using the data points shown as lled circles.
An important question is why the data obtained at frequencies above 40 GHz

do not follow the simple Debye model. A plot of eout against ein for data in the
high-frequency region is shown in g. 4.8. It is clear that the experimental results
lie above the values predicted by the Debye model using the best values of the
parameters found from data at lower frequencies. This suggests that another
relaxation process involving much higher frequencies is present in water. In the
alcohols [47], two high-frequency relaxation processes are found. The complex
nature of dielectric relaxation in this group of solvents is attributed to hydrogen
bonding.
Fig. 4.8 Plot of eout against ein using dielectric relaxation data for water in the frequency
range 60410 GHz [G5]. The solid line shows the contribution from the low-frequency
relaxation process.
176
It is often possible to describe the dielectric data for complex liquids in terms of
multiple relaxation processes, each of which follows a Debye relaxation based on
equation (4.5.2). The resulting equations for ein and eout are
ein en1
n
X
ei0 ei1
i1
1 o2 t2i
4:5:35
and
eout
n
X
ei0 ei1 oti
i1
1 o2 t2i
4:5:36
ei0 and ei1 are the low- and high-frequency values of the permittivity for the
relaxation process with relaxation time ti . The static permittivity es is equal to
the low- frequency permittivity for the rst relaxation process e10 . In addition, the
high-frequency permittivity for the ith relaxation ei1 is equal to the low-frequency
permittivity for the (i 1)th relaxation ei10 . Thus, for a liquid having n relaxation processes, there are n values of ei0 which must be specied together with n
dierent relaxation times, ti , and nally the high-frequency permittivity for the
last relaxation process, en1 .
In the case of water, it has been shown that the dielectric data can be described
in terms of two relaxation processes. Further analysis of the data shown in g. 4.8
leads to the following parameters: e20 6:02, e21 4:57, and t2 0:97 ps [G5].
The second relaxation process is much faster than the rst and is characterized by
frequencies in the terahertz range. The slow relaxation process which is characterized by a large amplitude (e10 e11 72 is attributed to a cooperative process
involving clusters of water molecules connected through hydrogen bonding. The
fast process is attributed to hydrogen bond formation and breakup. More will be
said about the nature of these processes in section 4.7.
Other relationships which have been used to describe dielectric relaxation data
include the ColeCole and ColeDavidson equations [29]. These are preferred
when a distribution of relaxation times rather than a single relaxation time is
more appropriate to describe the data in a given frequency range. Nevertheless,
the Debye model in its simple version or multiple relaxation versions works quite
well for most of the solvents considered here.
The principal problem with existing data in the literature is that they often are
not available over a suciently wide frequency range so that a reasonable analysis
at the high-frequency end cannot be made. Only recently have experiments been
carried out above 100 GHz. For solvents with rapid relaxation processes experiments above 1 THz in the far infrared region are required.
4.6 The Permittivity of Electrolyte Solutions

When an electrolyte is added to a polar solvent, the resulting solution has very
dierent electrical properties than those of the pure solvent. The most obvious
change is the fact that the solution has an easily measured conductivity due to the
presence of the ions. The solvent is a dielectric which behaves as a capacitor but its
POLAR SOLVENTS
177
dielectric properties change in the presence of the electrolyte. This is partially due
to the fact that the structure of the solvent is disrupted in the immediate vicinity of
the ions. Thus, in terms of an equivalent circuit, an electrolyte solution consists of
a resistor and capacitor in parallel.
By studying the frequency dependence of the impedance of an electrolyte solution one can learn something about both ionsolvent and ionion interactions and
how these aect solution structure. Dielectric relaxation data have been collected
for a number of electrolyte solutions, both aqueous and non-aqueous [G5, 29].
These data are obtained by measuring the impedance of the solution and then
correcting for solution conductivity. When simple electrolytes which do not ion
pair are involved, the permittivity data often can be treated by the Debye model.
A typical example is KCl in water. Values of ein plotted as a function of oeout for
this system at two dierent concentrations are shown in g. 4.9. On the basis of
equation (4.5.34), these plots give the value of es from the intercept and tD from
the slope. In this case, es is 76.58 at a concentration of 0.1 M, and 67.05 at 1 M.
The static permittivity decreases noticeably with increase in electrolyte concentration. This result is attributed to two factors. When the electrolyte ions do not
possess a dipole moment, as is the case for monoatomic ions, they do not respond
to the electrical eld applied to the system in the relaxation experiment. However,
they occupy volume in the system, so that the overall permittivity of the solution
decreases with increase in the mole fraction of the electrolyte. The second eect
related to the addition of electrolyte is connected to the breakup of solvent structure due to solvation of the ions. This is especially important in water solutions,
where hydrogen bonding plays an important role in determining solvent structure.
The value of the relaxation time also decreases with increase in electrolyte concentration. In the present case, tD varies from 8.26 ps for 0.1 M KCl to 8.05 ps for
1 M KCl.
Fig. 4.9 Plot of ein against oeout using dielectric relaxation data for aqueous KCl solutions:
(*) 0.1 and (^) 1 M.
178
The extent to which the dielectric permittivity of the solution departs from that
of the pure solvent depends not only on the electrolyte concentration but also on
the nature of the electrolyte. Considering the alkali metal halides, the largest drop
in es at a given concentration is found for Li electrolytes [G5, 29]. This reects
the disruptive eect of the small cation which causes reorientation of the solvent
molecules around it and a subsequent breaking of hydrogen bonds. The dielectric
decrement decreases with increase in cation atomic number for a given anion. On
the other hand, in the case of anions the dielectric decrement increases with
increase in anion atomic number for a given cation. Anions are less disruptive
of water structure, so that the main eect of changing the anion is a change in the
volume fraction of solution occupied by ions.
Representative plots of es as a function of concentration are shown in g. 4.10.
The principal feature depending on electrolyte concentration is the initial slope
which is called the dielectric decrement. In addition all plots are curved, suggesting that a limiting permittivity is reached at high electrolyte concentration. The
dependence of es on electrolyte concentration ce has been examined by Friedman
[30], who derived the following relationship:
es es0 de ce be c3=2
e
4:6:1
where es0 is the dielectric permittivity of the pure solvent, de , the dielectric decrement, and be , a curvature parameter. This equation describes the concentration
dependence of es quite well for most 11 electrolytes over a wide concentration
range. The curvature parameter is small for 11 electrolytes, usually falling in the
range 25 L3=2 mol3=2 . When be is set equal to 5 L3=2 mol3=2 , the values of de for
the alkali metal halides fall in a range from 14 to 22 L mol1 [31].
Fig. 4.10 Plots of the static permittivity of three 11 electrolyte solutions against electrolyte concentration. The ordinate scale is correct for the NaCl data; for KF, the data have
been shifted up by 10 units, and for KBr, down by 10 units for the sake of clarity.
POLAR SOLVENTS
179
EXAMPLE
Use the following permittivity data for aqueous NaClO4 solutions at 258C to
determine the best values of the parameters de and be for this electrolyte:
ce =M
es
0.5
1.0
1.5
2.0
3.0
4.0
5.0
70.41
63.56
58.17
53.44
47.28
40.29
35.21
The permittivity of pure water at 258C is 78.46. Thus, equation (4.6.1) can
be rearranged as follows:
78:46 es
de be c1=2
e
ce
4:6:2
The parameters de and be are determined from a plot of (78:46 es =ce against
ce1=2 . The required data are given in the following table:
1=2
c1=2
e =M
1
78:46 es c1
e =M
0.707
1.000
1.225
1.414
1.732
2.000
2.236
16.1
14.9
13.5
12.5
10.4
9.5
8.7
These data are tted to a straight line by least squares. From the intercept,
de 19:4 M1 and be 5:1 L3=2 mol3=2 .
When ion pairing is present an additional relaxation is observed at low frequencies. A typical example is the MgSO4 system in water. The ion pair has a
dipole moment and therefore this species reorients in the alternating electrical
eld. The relaxation time associated with the reorientation is much longer than
that associated with the reorientation of water molecules. It depends not only on
reorientation of the ionpair dipole but also on the kinetics of its formation and
decomposition. For this reason, the parameter for the low-frequency relaxation
process is strongly concentration dependent.
Dielectric relaxation data for a 0.08 M Mg2 SO4 solution are shown in g. 4.11.
On the basis of an analysis of these data by Barthel and coworkers [29, 32], three
relaxation processes may be discerned. The rst one, involving the ion pair, occurs
between permittivity values of 82.9 and 75.2 and involves a relaxation time of 181
ps. The second process, which is attributed to the slow reorientation of water
clusters, takes place between the permittivity values of 75.2 and 8.4 with a relaxation time of 8.4 ps. Finally, the high-frequency process, which occurs between 8.4
180
Fig. 4.11 Plot of ein against eout for dielectric relaxation data obtained with 0.08 M MgSO4
in water. The semicircles show the resolved Debye relaxation processes for (a) the ion pair;
(b) the slow reorientation of water clusters; (c) the fast process in water.
and 4.6 with a relaxation time of 1.5 ps, is attributed to the formation and breaking of hydrogen bonds.
Obviously, the dielectric behavior of systems with ion pairing is much more
complex than those without it. The presence of the ion pair gives the solution a
higher permittivity at lower frequencies than it otherwise would have. This feature
is important in understanding the equilibrium properties of these solutions. The
permittivity data for the low-frequency process may be used to determine the ion
pairing equilibrium constant and the rate constants for formation and breakup of
this species. Thus, dielectric relaxation experiments in electrolytes provide valuable information about ion association equilibria.
A variety of dielectric relaxation data are now available for both aqueous and
non-aqueous solutions. These results help one understand the properties of these
solutions in more detail. They are complementary to information obtained from
thermodynamic, spectroscopic, and conductivity experiments, and provide an
important basis for understanding solution structure.
4.7 The Dielectric Relaxation Parameters

According to the Debye model there are three parameters associated with dielectric relaxation in a simple solvent, namely, the static permittivity es , the Debye
relaxation time tD , and the high-frequency permittivity e1 . The static permittivity
has already been discussed in detail in sections 4.3 and 4.4 . In this section attention is especially focused on the Debye relaxation time tD and the related quantity, the longitudinal relaxation time tL . The signicance of these parameters for
solvents with multiple relaxation processes is considered. The high-frequency
permittivity e1 and its relationship to the optical permittivity eop is also discussed.
POLAR SOLVENTS
181
As dened, the Debye relaxation time is the reciprocal of a rst-order rate

constant. Thus, it is expected to depend on temperature according to the usual
Arrhenius relationship

HD
tD AD exp
4:7:1
RT
where AD is a pre-exponential factor and HD , the activation enthalpy associated
with the relaxation process. Values of HD are available for many polar solvents
and fall in the range from 5 to 25 kJ mol1 (see table 4.2). For unassociated polar
solvents the relaxation process is assumed to involve rotation of a molecular
dipole in the alternating electrical eld and is called rotational diusion. The
major barrier to this process is dipoledipole interactions. Thus, it was shown
that HD is a linear function of the enthalpy associated with dipoledipole interactions in aprotic solvents [6]. Although tD is not a molecular parameter, it is
related in a complex way to the molecular events which take place in a polar liquid
subjected to an alternating electrical eld [G2].
The temperature dependence of the longitudinal relaxation time L is also an
important quantity. For a Debye solvent, tL is given by the relationship
e
4:7:2
tL 1 tD
es
The enthalpy associated with the temperature dependence is dened as
HL R
@ ln tL
@ ln e1
@ ln es
HD R
R
@1=T
@1=T
@1=T
4:7:3
The temperature derivatives of the high- and low-frequency contributions to the

permittivity are easily estimated from experimental data (see table 4.2). In some
cases the temperature derivative of e1 is not available but one may assume that it
is approximately equal to the temperature derivative of eop .
Unfortunately, the Debye model provides only an approximate description of
aprotic solvents. It has been applied extensively to determine their relaxation
properties quite successfully, mainly because permittivity data are available
over a limited frequency range. As a result, the high-frequency parameter is
usually obtained by a long extrapolation. As experimental methods have become
available at frequencies above 50 GHz, it has been found that the behavior of
aprotic solvents is more complex [9].
Acetonitrile (AcN) can be described by the Debye model using data up to 40
GHz [9]. When the frequency range is extended to 90 GHz, the data can only be
tted if a distribution of relaxation times is assumed. Dimethylsulfoxide (DMSO)
and propylene carbonate (PC) have been studied up to 89 GHz [9]. These data can
be tted to the Debye model with two relaxation processes. DMSO is known to
form dimers in the bulk [28]. This association is undoubtedly the reason for its
complex relaxation behavior. The same is true for PC, a molecule with a very high
dipole moment.
The dielectric relaxation data for dimethylformamide (DMF) and dimethylacetamide (DMA) can be described by two Debye processes [9]. The lowfrequency, high-amplitude process is attributed to rotational diusion. For
182
these systems, the high-frequency relaxation is ascribed to hindered intramolecular rotation about the CN bond [9].
All protic solvents undergo multiple relaxation processes due to the presence of
hydrogen bonding. In the case of water and formamide (F), the data can be
described in terms of two Debye relaxations. For the alcohols and N-methylformamide (NMF), three Debye relaxations are required for the description. In
all of these solvents, the low-frequency process involves the cooperative motion of
hydrogen-bonded clusters. In the case of water and the alcohols the highfrequency process involves the formation and breaking of hydrogen bonds. The
intermediate process in the alcohols is ascribed to rotational diusion of monomers. Studies of dielectric relaxation in these systems have been carried out for the
n-alkyl alcohols up to dodecanol [8]. Values of the relaxation parameters for water
and the lower alcohols are summarized in table 4.5.
F and NMF are also highly structured solvents as a result of hydrogen bonding. The low-frequency relaxation process in these systems can be attributed to the
cooperative motion of hydrogen-bonded clusters. The process at the highest frequencies has a similar relaxation time to those observed in DMF and DMA.
Thus, it is probably due to intramolecular rotation about the CN bond in the
monomer. The intermediate relaxation observed in NMF is attributed to rotational diusion of a monomer. Relaxation parameters for F and NMF are also
summarized in table 4.5.
Estimation of the longitudinal relaxation time in solvents with multiple Debye
relaxation processes is not straightforward. In fact, tL is a function of time in
these systems [33, 34], and varies between two limiting values. For a solvent with
two relaxation processes, the low-frequency limit for tL is
e
tL0 1 f1 tD1 f2 tD2
4:7:4
es
where
fi
ei0 ei1
e s e1
4:7:5
At high frequencies, the limiting value of tL is

e1 f 1
f2 1
tL1
es tD1 tD2
4:7:6
Table 4.5 Dielectric Relaxation Parameters for Some Protic

Solvents at 258C [5, 9]
Solvent
es e10
t1 =ps
e20
t2 =ps
e30
t3 =ps
e1
W
MeOH
EtOH
PrOH
F
NMF
78.45
32.63
24.35
20.44
109.5
186.0
8.32
51.5
163
329
37.3
128
6.18
5.91
4.49
3.74
7.08
6.13
1.02
7.09
8.97
15.1
1.16
7.93
4.90
3.82
3.20
4.60
1.12
1.81
2.40
0.78
4.49
2.79
2.69
2.44
4.48
3.20
POLAR SOLVENTS
183
The time characterizing transition between tL1 and tL0 is tm , which is given by
tm f1 tD1 f2 tD2
4:7:7
It is interesting to calculate these parameters for water on the basis of the data
given in table 4.5. For this solvent, the fractions f1 and f2 are 0.977 and 0.023,
respectively. As a result, the low-frequency value of tL , tL0 , is equal to 0.47 ps.
Because f1 is close to one, the inuence of the high-frequency process on tL0 is
minimal. At high frequencies, the limiting value of tL , tL1 , is 0.41 ps. For this
system the dierence between the limiting values of tL is very small and beyond
the limit of experimental detection under most circumstances. The time characterizing transition between tL1 and tL0 is 1.19 ps. Thus, for processes occurring at
frequencies greater than 10 THz, the appropriate value of tL is tL1 . For frequencies lower than 100 GHz, the corresponding value is tL0 .
When the solvent has three relaxation processes, the above relationships for tL
become
tL0
e1
f t f2 tD2 f3 tD3
es 1 D1
4:7:8

e1 f1
f
f 1
2 3
es tD1 tD2 tD3
4:7:9
and
tL1
Variation of tL between the high- and low-frequency limits is characterized by

two exponentially decaying functions with dierent transition times [34]. In the
case of the alcohols the dierence between these limits is signicant and can lead
to complex relaxation behavior in these solvents. For methanol, the highfrequency value of tL is 1.0 ps and the low-frequency value, 4.0 ps. For 1-propanol, the dierence is greater, the corresponding values being 5.3 and 36.5 ps,
respectively.
The high-frequency permittivity e1 reects the value of this property when
reorientation of the solvent dipoles no longer contributes to polarization of the
medium. As the frequency of the a.c. electrical eld increases, the ability of the
molecules to respond rapidly decreases. Eventually, the solvent dipoles become
frozen when the a.c. frequency is very high. This occurs typically for frequencies close to 1 THz, corresponding to relaxation times in the order of 1 ps.
Experimentally, these processes occur in the far infrared at frequencies of 30
cm1 or greater. The molecules now respond to the electromagnetic eld through
distortional and electronic polarization. Distortional polarization corresponds to
movement of the molecular nuclei with respect to one another and is equivalent to
molecular vibration. As the frequency of the electromagnetic eld increases into
the infrared, discontinuities are found in the permittivity of the liquid corresponding to the frequency at which a given molecular vibration occurs. The permittivity
after each of these vibrations is activated decreases until the distortional contribution to the permittivity is lost. Eventually in the visible region, the molecules
only respond via electronic polarization, that is, response of the electronic cloud
in the molecules to the alternating electrical eld. This response is normally
184
determined by measuring the refractive index of the liquid at the sodium D line
(589 nm) and thereby estimating eop . Because of the distortional contribution to
the permittivity, e1 is always larger than eop . For example, in the case of water,
e1 is 4.49 and eop , 1.78.
Relaxation parameters such as tD and tL are used in the analysis of the kinetics
of very fast processes in solution. As one might expect, movement of solvent
molecules can be inuential in determining the rates of these processes. Thus,
the study of dielectric relaxation not only provides valuable information about
solvent structure but also relaxation parameters relevant to fast solution kinetics.
This subject is discussed in more detail in chapter 7.
4.8 Ion Solvation in Polar Solvents

The variation in ion solvation with solvent nature is an interesting subject because
it is relevant to basic ideas about the structure of electrolyte solutions. According
to the classical Born model, the Gibbs energy of ion solvation depends on solvent
nature only through its static dielectric permittivity es . When this subject was
examined for aqueous solutions in chapter 3, it was concluded that local ion
solvent interactions are very important in determining the magnitude of the Gibbs
energy of solvation and that these interactions result in disruption of local solvent
structure. Therefore, it is interesting to examine ion solvation in a wide variety of
polar solvents in order to assess the relative importance of long-range electrostatic
and short-range chemical forces.
The most direct way of determining s G is by measuring the work associated
with transferring the electrolyte from the gas phase, where the ions are innitely
far apart from one another, to the liquid solution. However, from a practical
point of view, it is much simpler to solve this problem starting with the solid
electrolyte. Consider a simple 11 electrolyte MX which dissolves in solvent S:
MXcry !MXsl S
4:8:1
The Gibbs energy change associated with formation of the solution may be
written as
sl G sl G RT ln aM aX
4:8:2
where sl G is the value of sl G at unit activity on the molarity scale. When the
dissolution process (4.8.1) is at equilibrium, the electrolyte solution is saturated
and one may write
sl G RT lnaM aX sat RT ln Ksp
4:8:3
where Ksp is the thermodynamic solubility product. Following the usual thermodynamic conventions for electrolytes, sl G is the standard Gibbs energy change
associated with dissolution of solid electrolyte in a hypothetical 1 M solution in
which all real interactions are absent (see section 3.6). In practice, solubility
products are reported in the literature in terms of concentrations. Thus, determi-
POLAR SOLVENTS
185
nation of sl G usually requires that one estimate the mean activity coecient of
the electrolyte in the saturated solution. One way of avoiding this problem is to
use an electrochemical cell to determine Ksp with electrodes reversible to the
cation and anion of the electrolyte (see chapter 9).
In order to remove the contribution to sl G from the solid phase, values of
this quantity are compared in two dierent solvents. If one imagines that the same
salt is dissolved in two dierent solvents S1 and S2 , then one may estimate the
Gibbs energy of transfer of MX according to the process
MXsl S1 !MXsl S2
4:8:4
The corresponding standard Gibbs energy change is called the Gibbs energy of
transfer:
tr G sl G S2 sl G S1
4:8:5
tr G depends only on the dierences in the interaction of the ions with the two
solvents, the contribution from the solid phase having been lost in subtraction.
Values of sl G have been determined for many electrolytes in most of the
polar solvents considered in this chapter. Various reference solvents have been
considered for reporting tr G [3538] but that used in most recent work is water.
Thus, tr G gives the Gibbs energy associated with transferring one mole of the
electrolyte in the hypothetical standard state from water to the given non-aqueous
solvent. This quantity is positive for most simple 11 electrolytes because the
electrolyte is less soluble in the non-aqueous medium [36, 37].
Obviously, it is more interesting to have single ion quantities for the Gibbs
energy of transfer. This requires that an extrathermodynamic assumption be
made to achieve separate cation and anion contributions. Many assumptions
have been considered and evaluated in the literature [38, 39]. Only one is described
here and its use may be considered a convenient convention which leads to reasonable separation of the cationic and anionic contributions to tr G . By choosing an electrolyte with very large ions whose charge is buried in the center of the
ion, it can be assumed that the cation and anion have equal contributions to
tr G . The electrolyte used is tetraphenylarsonium tetraphenylborate (TATB)
which is sparingly soluble in water, but more soluble in non-aqueous media. As
a result, the values of tr G for this electrolyte are negative. The TATB assumption is expressed quantitatively through the equation
tr G TA tr G TB
tr G TATB
2
4:8:6
By choosing a series of electrolytes which includes salts containing the tetraphenylarsonium cation and the tetraphenylborate anion, the individual contributions
to tr G from the more common ions may be extracted.
EXAMPLE
The following data are for the Gibbs energy of transfer from water to propylene carbonate for 11 electrolytes at 258C [37]. Use these data to estimate the
corresponding quantities for the individual ions on the basis of the TATB
assumption.
186
Electrolyte
LiI
NaCl
KCl
KI
TAI
TATB
tr G =kJ mol1

43.7
53.1
44.4
23.4
19.2
72.0
From the value of tr G for TATB,

tr G TA tr G TB 72:0=2 36:0 kJ mol1
4:8:7
From the value of tr G for TAI,

tr G I 19:2 36:0 16:8 kJ mol1
4:8:8
From the data for the other iodides,

tr G Li 43:7 16:8 26:9 kJ mol1

tr G K 23:4 16:8 6:6 kJ mol
1
4:8:9
4:8:10
From the data for the chlorides,

tr G Cl 44:4 6:6 37:8 kJ mol1
4:8:11
tr G Na 53:1 37:8 15:3 kJ mol1
4:8:12
and
These results are slightly dierent than the values recorded in tables 4.6 and
4.7. The tabulated data are based on a larger collection of results for 11
electrolytes and therefore are better estimates of the single ion values.
Values of tr G for the alkali metal cations together with the value for the
tetraphenylarsonium ion are given in table 4.6 for the solvents considered in this
chapter. For the alkali metal ions, tr G is both positive and negative, the latter
values indicating that the ion is more stable in the non-aqueous environment. In
the case of the Na ion, tr G varies from 17 kJ mol1 in hexamethylphosphoramide (HMPA) to 36 kJ mol1 in nitrobenzene. It is clear that there is no
correlation between tr G and the static permittivity es or a related function of es ,
such as that found in the Born equation (3.4.6). The solvent with the highest
permittivity among those considered is NMF. Its value of tr G for Na is
negative but not as large as that seen for solvents like HMPA or DMSO.
Similarly, BuOH has a very low permittivity but its value of tr G is not as
large as that of NB. This point is further demonstrated by considering three
aprotic solvents with approximately the same values of es , namely, AcN, DMF,
and NM. The values of tr G for the Na ion in these solvents are 15.1, 9:6, and
32 kJ mol1 , respectively. This shows emphatically that chemical interactions are
POLAR SOLVENTS
187
Table 4.6 Gibbs Energy of Transfer for Monovalent Cations

from Water to a Non-Aqueous Solvent at 258C [39]
tr G /kJ mol1
Solvent
Li
Na
Rb
Cs
Ph4 As
(529)
4.4
11
11
11
10
20
(424)
8.2
14
17
19
8
7
(352)
9.6
16.4
17
20
4.3
6
(329)
9.6
16
19
23
5
8
(306)
8.9
15
17
19
6
7
24.1
21.2
25
20
23.9
33
10
25
36
22
10
15
35
38
48
27
23.8
6
23
10
15.1
21
16
12.1
9.6
13.4
17
15
36
31.6
15
14.6
3
14
4
8.1
19
11.7
10
13.0
16
11
21
15.4
11
5.3
4
13
4
6.3
15
8
9.7
10.4
10
8
19
11.0
1
9
10
4
6.0
13
7
10.8
13.0
7
10
18
5.6
7
10
32
32.8
35
40
38.5
37.4
39
40
36
32.6
36
36
Protic
1.
2.
3.
4.
5.
6.
7.
Wa
MeOH
EtOH
PrOH
BuOH
F
NMF
Aprotic
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
AC
AcN
BzN
BuNb
DMA
DMF
DMSO
HMPA
NMP
NB
NM
PrN
PC
TMS
TMUb
Values in parentheses give the absolute Gibbs energy of solvation in water

estimated from experiment.
b
Estimated from polarographic halfwave potential data [40].
much more important in determining solvation in these systems than long-range

electrostatic eects.
Values of tr G for some halide ions, the perchlorate ion, and the tetraphenylborate ion are given in table 4.7. This quantity is always positive for the halide
ions, and is much larger in aprotic solvents than in protic systems. Thus, small
monoatomic anions are especially dicult to solvate when hydrogen bonding is
not available in the system. The same is not true for the perchlorate ion, which
actually is more stable in some aprotic solvents than in water. Because the negative charge in the ion is distributed over ve atoms, it dissolves more easily than
the Cl anion in all non-aqueous media. For this reason perchlorate salts are
often used to introduce cations into a non-aqueous solvent. Values of tr G for
the tetraphenylborate anion are always large and negative, and by denition equal
to the values for the tetraphenylarsonium cation.
188
Table 4.7 Gibbs Energy of Transfer for Monovalent

Anions from Water to a Non-Aqueous Solvent at 258C
[39]
tr G /kJ mol1
Solvent
Cl
Br
I
ClO
4
Ph4 B
(304)
13.2
20.2
26
29
13.7
(278)
11.1
18.2
22
24
10.7
(243)
7.3
12.9
19
22
7.3
(165)
6.1
10
17
22
12
24.1
21.2
25
20
23.9
33
57
42.1
47
54.9
48.3
40.3
58
51
35
37.7
39.8
47
42
31.3
37
44.0
36.2
27.4
46
37
29
29
30
35
25
16.8
25
21
20.4
10.4
30
19
18
18.9
13.7
21
6
2
13
4
1
7
12
10
4.7
3
32
32.8
35
40
38.5
37.4
39
40
36
32.6
36.0
36
Protic
1.
2.
3.
4.
5.
6.
7.
Wa
MeOH
EtOH
PrOH
BuOH
F
NMF
Aprotic
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
AC
AcN
BzN
BuN
DMA
DMF
DMSO
HMPA
NMP
NB
NM
PrN
PC
TMS
TMU
a
Values in parentheses give the absolute Gibbs energy of solvation in
water estimated from experiment.
Extensive data also exist for the enthalpy of transfer of 11 electrolytes from
water to non-aqueous solvents. These data are obtained in calorimetric experiments in which the enthalpy of solution is measured as a function of electrolyte
concentration. The value of the enthalpy of solution per mole at innite dilution is
obtained by extrapolation. The enthalpy of transfer is dened in the same way as
the Gibbs energy of transfer according to process (4.8.4). Thus, the standard
enthalpy of transfer is given by
tr H sl H S2 sl H S1
4:8:13
The reference solvent in common use is water [39].

Several extrathermodynamic assumptions have been considered to separate
experimental values of tr H into cationic and anionic contributions [35]. The
one commonly used is the TATB assumption, namely, that the contributions to
POLAR SOLVENTS
189
tr H for the large tetraphenylarsonium and tetraphenylborate ions are equal.
This leads to the values shown in tables 4.8 and 4.9 for the alkali metal and halide
ions.
Values for tr H for the alkali metal cations are negative in almost all cases.
Exceptions are those for Li and Na+ in NM, and Li in PC and TMS. It
follows that the interaction of these monoatomic monovalent cations is more
exothermic with most polar solvents than with water. On the other hand, the
enthalpy of transfer of the smaller anions, Cl and Br , is almost always positive.
This is a direct indication that these species are dicult to dissolve, especially in
aprotic solvents. In the case of I and ClO
4 , the enthalpy of transfer is usually
negative. Comparison of the enthalpy data with the Gibbs energy data recorded in
tables 4.6 and 4.7 shows that these quantities are very dierent. When both Gibbs
energies and enthalpies of transfer are available, one can calculate the entropy of
transfer [39].
Table 4.8 Enthalpy of Transfer for Monovalent Cations from Water to a

Non-Aqueous Solvent 258C [39]
tr H =kJ mol1
Solvent
Li
Na
Rb
Cs
Ph4 As
(578)
21.7
20.2
18.4
6.0
(464)
20.7
19.4
18.8
16.5
22.5
(381)
19.0
19.6
18.3
17.9
(355)
16.5
17.8
(330)
14.1
11.8
12.7
17.7
2.2
0
1.5
0.5
7.6
8.0
25.4
27.1
57.4
26
25.8
2.8
12
13.3
41.2
32.4
29.2
50.6
41
11.5
10.5
16
22.9
35.7
35.4
46.6
47
18.5
22.5
26
24.6
36.1
38.1
43.8
55
17.7
24.9
28
26.1
34.6
33.0
45.0
50
16.6
27.5
26
11.1
13.7
17.2
10.6
25.8
17
6.9
13.1
11
Protic
1.
2.
3.
4.
5.
6.
7.
Wa
MeOH
EtOH
PrOH
BuOH
F
NMF
Aprotic
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
AC
AcN
BzN
BuN
DMA
DMF
DMSO
HMPA
NMP
NB
NM
PrN
PC
TMS
TMU
a
Values in parentheses give the absolute enthalpy of solvation in water estimated from
experiment.
190
Table 4.9 Enthalpy of Transfer for Monovalent Anions from

Water to a Non-Aqueous Solvent 258C [39]
tr H /kJ mol1
Solvent
Cl1
Br
I
ClO
4
Ph4 B
(320)
8.4
10.4
8.4
3.5
14.2
(289)
4.5
5.5
2.5
1.5
(247)
1.0
0.7
1.5
6.8
(176)
3.1
2.7
0
20
11.8
0.9
0
1.5
0.5
7.6
19.3
35.6
17.9
20.0
38.2
27
21.2
26.2
27
8.0
17.2
0.6
4.6
17.7
13
7.1
15.2
13
7.6
0.7
15.0
11.5
5.8
2
7.8
1.6
8
16.0
15.3
23.4
18.2
20.7
11
19.2
16.3
20
11.1
13.7
17.2
10.6
25.8
17
6.9
13.1
11
Protic
1.
2.
3.
4.
5.
6.
7.
Wa
MeOH
EtOH
PrOH
BuOH
F
NMF
Aprotic
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
AC
AcN
BzN
BuN
DMA
DMF
DMSO
HMPA
NMP
NB
NM
PrN
PC
TMS
TMU
a
Values in parentheses give the absolute enthalpy of solvation in water estimated
from experiment.
Using the absolute values of the Gibbs energy of solvation for the individual
ions in water one can calculate the absolute value of this quantity in any solvent
using the relationship
s GS s GW tr G
4:8:14
where s GS is the Gibbs energy of solvation in solvent S and s G(W) the

corresponding quantity in water. According to the Born model for ion solvation
presented in section 3.4, the Gibbs solvation energy for a given ion depends on
solvent nature only through the dielectric permittivity es . Specically, this model
predicts that s G(S) should be linear in (1 1=es ) provided the ionic radius ri
does not depend on solvent nature. However, plots based on the data in tables 4.6
and 4.7 for simple cations and anions show no correlation between s G(S) and
(1 1=es ). These results emphasize that a simple electrostatic model cannot
account for the variation in the Gibbs energy of solvation with solvent nature.
POLAR SOLVENTS
191
As will be seen in the following section, the chemical nature of the interaction
between the ion and the solvent must be considered in order to understand this
variation.
The data in tables 4.64.9 are very helpful in predicting the solubility of a given
electrolyte in a solvent. In addition, they give an indication of the extent to which
ion pairing is important. If the solvent interacts strongly with the ion, then its
tendency to form ion pairs is reduced. Data are available for more ions than those
considered here, and these can be consulted to assess the properties of other
electrolytes in a given solvent [39, 40].
4.9 Polar Solvents as Lewis Acids and Bases

Examination of the variation in thermodynamic properties of electrolytes in polar
solvents shows conclusively that continuum electrostatic models are not able to
account for the experimental data. As a result, chemists have looked for other
ways of explaining the variation in these quantities with solvent nature. Since ions
are stabilized by the dipolar character of the solvent molecule, a method of
measuring the ability of a given solvent to solvate cations and anions was sought.
Cations are stabilized by the electronegative end of the solvent dipole. Thus, the
important property of the solvent molecule is its ability to donate a pair of
electrons, that is, to act as a Lewis base. On the other hand, anions are stabilized
by the positive end of the solvent dipole. The relevant molecular property is
solvent acidity or the ability to accept a pair of electrons. In the case of Lewis
acidity, the ability to form hydrogen bonds is also a contributing factor. The fact
that the chemical nature of the ends of the molecular dipole is quite dierent
means that a given solvent can have a very dierent Lewis acidity and basicity.
Early work by Dimroth and Reichardt [41, 42] led to the development of a
polarity scale ET based on a solvatochromic dye. This dye, namely, 4-(2,4,6triphenyl-pyridinium)-2,6-diphenylphenoxide undergoes a pp transition in the
visible or ultraviolet which is accompanied by a large decrease in the molecular
dipole moment (see g. 4.12). In the ground state, the dye interacts strongly with
the solvent through the electronegative oxygen atom. The wavelength associated
with the pp transition depends strongly on the nature of the solvent and is
clearly related to solvent acidity, that is, to the solvents ability to solvate the
electronegative oxygen atom in the ground state. Thus, the pp transition is
observed at 810 nm in a non-polar solvent such as diphenyl ether, whereas it
occurs at 453 nm in water. ET is dened on the basis of the energy associated
with this transition in kJ mol1 . It provides an important empirical parameter for
measuring solvent acidity (see table 4.10).
Another scale for measuring solvent acidity was formulated by Mayer et al.
[43]. It is called the solvent acceptor number (AN) and is based on the relative
values of the 31P NMR chemical shifts produced by a given solvent with a strong
Lewis base, triethylphosphine oxide (g. 4.13). The data were normalized so that
the acceptor number of hexane is zero and that for the 1:1 adduct with the strong
Lewis acid, SbCl5, 100 when dissolved in 1-2 dichloroethane. The attractive feature of this scale is that it varies over a wide range for the polar solvents con-
192
Fig. 4.12 Chemical structure

of the solvatochromic dye
used to dene the ET
polarity scale: left, ground
state; right, excited p state.
sidered here, namely, from 10.6 for hexamethylphosphoramide to 54.8 for water
(table 4.10).
As one would expect, there is a strong linear correlation between the parameters ET and AN, the relationship being
Table 4.10 Empirical Parameters for Solvent

Acidity and Basicity
Acidity
Solvent
Basicity
AN
ET
DN
Bsc
54.8
41.5
37.9
37.3
36.8
39.8
32.1
264.0
231.8
217.1
212.1
210.0
236.8
226.4
18.0
19.0
19.2
19.8
19.5
24
27
591
589
589
589b
589
598
604
12.5
18.9
15.5
15.2a
13.6
16.0
19.3
10.6
13.3
14.8
20.5
22.7a
18.3
19.2
10.7a
176.6
190.8
173.6
180.3
182.8
183.3
188.7
171.1
176.6
172.4
193.7
195.0
184.1
171.5
171.5
17.0
14.1
11.9
16.6
27.8
26.6
29.8
38.8
27.3
4.4
2.7
16.1
15.1
14.8
31
569
573
572
573b
608
602
613
633
604b
522
530
572b
554
562
596
Protic
1.
2.
3.
4.
5.
6.
7.
W
MeOH
EtOH
PrOH
BuOH
F
NMF
Aprotic
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
a
AC
AcN
BzN
BuN
DMA
DMF
DMSO
HMPA
NMP
NB
NM
PrN
PC
TMS
TMU
Estimated on the basis of equation (4.9.1).

Estimated on the basis of equation (4.9.2).
POLAR SOLVENTS
193
Fig. 4.13 Lewis base, triethylphosphine oxide, used to

dene the AN scale.
ET 150:6 1:954 AN
4:9:1
with a correlation coecient of 0.964. The slope of this relationship demonstrates

that the variation in ET is only 47% of that in AN, so that AN is the preferred
parameter to assess solvent acidity. Equation (4.9.1) can be used to estimate
values of AN which were missing from the original compilation [43]. These are
also given in table 4.10.
A popular method of estimating solvent basicity is based on the donor number
scale, DN, introduced by Gutmann [44, 45]. The donor number is dened as the
value of the heat of reaction of the polar solvent with the strong Lewis acid SbCl5
when these reactants are dissolved in 1,2-dichloromethane (g. 4.14). The scale
suers from the problem that DN cannot be measured directly for protic solvents
because of the instability of SbCl5 in these systems. However, values of DN for
these solvents have been estimated by a variety of other techniques [46].
Another method of estimating solvent basicity makes use of the solvatochromism of a Cu2 complex, namely Cu(II) N;N;N 0 ;N 0 -tetramethylethylenediamine
acetoacetonate [47]. This parameter which is designated Bsc is especially convenient because it may be measured directly for all the solvents considered here,
both protic and aprotic. The relationship between Bsc and DN is
Bsc 525:2 2:89DN
4:9:2
with a correlation coecient of 0.942. Unfortunately, experimental values of Bsc

are not available for all the solvents considered here but they may be estimated
using equation (4.9.2).
It was pointed out earlier that the Gibbs energy of transfer of simple monatomic ions such as Na and Cl1 is not related in any simple way to the dielectric
permittivity of the solvent. On the other hand, tr G for the alkali metal cations is
correlated with solvent basicity. A plot of this quantity for the Na ion against the
solvents donor number DN is shown in g. 4.15. Although the correlation is not
excellent, it shows a clear trend in which tr G becomes more negative as DN
increases. This is exactly what one expects because the ability of the solvent to
stabilize the cation in solution increases with increase in solvent basicity.
On the basis of the MSA, the molar Gibbs solvation energy for a monovalent
ion i in solvent S is
Fig. 4.14 Lewis acid, antimony pentachloride, used to

dene the DN scale.
194
Fig. 4.15 Plot of the Gibbs energy of transfer of Na from water to various non-aqueous
solvents against the Gutmann donor number DN.
s Gi S

NL e20
1
1
1
e s r i ds
8pe0
4:9:3
where NL is the Avogadro constant, e0 , the fundamental charge, e0 , the permittivity of free space, es , the solvents dielectric permittivity, ri , the ions radius, and
ds , the MSA parameter correcting ri . The latter quantity depends not only on the
nature of the solvent but also on whether the ion is a cation or an anion. Dening
1/ds as Bs or solvent basicity, for the case of a monovalent cation, equation (4.9.3)
may be rewritten as

NL e20
1
Bs
4:9:4
s GM S
1
es 1 Bs ri
8pe0
This shows that the Gibbs solvation energy changes with solvent nature due to
changes in es and Bs . Since the factor (1 1=es ) is close to unity for the polar
solvents discussed here, and assuming that Bs ri does not change signicantly with
respect to one, then the main reason for the correlation seen in g. 4.15 is the
change in Bs . Thus, the reciprocal of the MSA parameter for cations is directly
related to solvent basicity [48, 49].
Values of Bs estimated using the data for Na ion given in table 4.6 are plotted
against the Gutmann DN in g. 4.16. A good linear correlation is found with a
correlation coecient of 0.921. The equation relating Bs to DN is
Bs 16:30 0:314DN
4:9:5
Thus, Bs provides an alternate empirical scale for solvent basicity.

In the case of the halide anions, the Gibbs energy of transfer is correlated with
solvent acidity. A plot of tr G for the chloride anion is shown as a function of
the Gutmann AN in g. 4.17. A very good correlation is found based on the data
POLAR SOLVENTS
195
Fig. 4.16 Plot of the MSA parameter Bs estimated in various solvents for the Na ion
(equation (4.9.4)) against the Gutmann donor number DN.
which are available for 18 solvents (r 0:954). According to the MSA, the Gibbs
solvation energy is given by equation (4.9.3) with the provision that the parameter
ds depends on the nature of the solvent but is not the same as that found for the
alkali metal cations [48, 49]. Dening 1/ds as As or solvent acidity for the case of a
halide anion, equation (4.9.3) can be rewritten as
s GX S

NL e20
1
As
1
es 1 As ri
8pe0
4:9:6
Fig. 4.17 Plot of the Gibbs energy of transfer of Cl from water to various non-aqueous
solvents against the Gutmann acceptor number AN.
196
As lls the role of an acidity parameter because the factors (1 1=es ) and As ri do
not vary markedly for the polar solvents considered in the present analysis. As a
result, s GX S correlates well with the acidity parameter AN.
Values of As estimated using the data for Cl1 ion given in table 4.7 are plotted
against the Gutmann acceptor number in g. 4.18. A very good linear correlation
with a correlation coecient of 0.978 is obtained. The relationship between As
and AN is
As 7:826 0:172AN
4:9:7
As provides another empirical scale for solvent acidity.

The above treatment may be extended to the analysis of the Gibbs energy of
transfer of simple electrolytes and related thermodynamic quantities. Because the
electrolyte contains both cations and anions, the thermodynamic property is
expected to depend on both the acidity and basicity of the solvent. A simple
way of expressing this dependence is
Q Q0 as AN bs DN
4:9:8
where Q is the solvent-dependent thermodynamic quantity, Q0 , its value in a

solvent with zero acceptor number and zero donor number, as , the response to
solvent acidity and bs , the response to solvent basicity [50]. Values of Q0 , as , and
bs may be determined using a three-parameter least-squares t to express Q as a
linear function of both AN and DN. Graphical presentation of the results of this
t requires using a three-dimensional plot.
Results of analyzing Gibbs energy of transfer data on the basis of equation
(4.9.8) for ve alkali metal halides and TATB from water to various non-aqueous
solvents are summarized in table 4.11. Acceptable ts to equation (4.9.6) are
obtained for these systems, the correlation coecient r decreasing with increase
in the size of the alkali metal cation and halide anion. However, there is a problem
in assessing the properties of the t on the basis of the response factors as and bs .
Fig. 4.18 Plot of the MSA parameter As estimated in various solvents for the Cl ion
(equation (4.9.5)) against the Gutmann acceptor number AN.
POLAR SOLVENTS
197
Table 4.11 Analysis of the Dependence of the Gibbs Energy of Transfer of 11

Electrolytes [36, 37] on Solvent Acidity and Basicity (Equation (4.9.8))
Salt
LiCl
NaCl
KCl
KBr
CsBr
TATB*
No. of
Solvents
Q0
/kJ mol1
as
bs
as
9
10
8
10
10
12
133.1
106.2
87.9
74.2
67.4
94.6
1:48
1:31
1.13
0.87
0.82
1.60
2:82
1:86
1.33
1.38
1.15
0.35
0.55
0.62
0.62
0.50
0.53
0.88
bs
Stand.
Dev.
/kJ mol1
Corr.
Coe.
r
0.45
0.38
0.38
0.50
0.47
0.12
2.6
2.8
4.2
6.1
9.7
6.9
0.996
0.993
0.987
0.963
0.894
0.959
*TATB refers to (C6 H5 )4 As (C6 H5 )4 B.
Since the range over which AN and DN vary depends on the solvents chosen in
the study, the relative importance of solvent acidity and basicity in determining
the Gibbs energy of transfer is not easily assessed.
By calculating relative partial regression coecients, the role of solvent acidity
and basicity in determining the thermodynamic quantity can be clearly seen [50].
In order to do this, one must estimate the variance for the independent and
dependent variable involved in the multiparameter analysis. For the parameter
Q, the variance is dened as
!2
X
X
2
nQ
Qi
Qi =n
4:9:9
i
where the Qi s are the individual values of Q in a data set containing n dierent
values. The variances of AN and DN, nAN and nDN , respectively, are calculated in
a similar way. The partial regression coecients are dened as
a0s jas jnAN =nQ 1=2
4:9:10
b0s jbs jnDN =nQ 1=2
4:9:11
and
The parameters a0s and b0s are now normalized to the same scale so that their
relative values indicate the relative importance of solvent acidity and basicity in
the given analysis. A more direct way of assessing this importance is in terms of
relative partial regression coecients as and bs , which are dened as follows:
a s
a0s
a0s
b0s
4:9:12
a0s
r0s
b0s
4:9:13
and
bs
198
The parameter as gives the fraction of the explained variation in the parameter Q,
which is due to solvent acidity. In the same way, b s gives the corresponding
fraction due to solvent basicity.
The values of as and bs recorded in table 4.11 show that solvent acidity is
usually more important than solvent basicity in determining solvation of the
electrolytes considered. The result for TATB for which as 0:88 and bs 0:12
is especially surprising. According to the TATB assumption used to estimate
single ion Gibbs energies of transfer, these quantities should each be equal to
0.50. If results for the three protic solvents W, MeOH, and F are removed from
the data set, and the analysis performed with the remaining nine aprotic solvents,
the values of as and bs are then equal to 0.50. This result suggests that the TATB
assumption is valid for aprotic solvents but not very good for protic solvents with
strong hydrogen bonding [48].
A more general relationship for analyzing solvent eects for thermodynamic
quantities was given by Koppel and Palm [51]. They extended equation (4.9.9) to
include non-specic solvent eects related to solvent permittivity and polarizability. Their equation can be written as
Q Q0 as AN bs DN gs X ds Y
4:9:14
where X is the solvents polarity, Y, its polarizability, and gs and ds , the corresponding response factors. Solvent polarity is dened to be
X
es 1
es 2
4:9:15
and is directly related to the molar polarization dened in Debyes theory of the
dielectric permittivity of polar uids (see equation (4.3.20)). The polarizability Y
is given by
Y
eop 1
eop 2
4:9:16
According to the LorentzLorenz equation (4.3.21) for the molar refraction at

optical frequencies, Y is directly proportional to the molecular polarizability ap .
The KoppelPalm equation has also been applied to the analysis of solvent eects
on thermodynamic quantities related to the solvation of electrolytes [48, 49]. In
the case of the systems considered in table 4.11, addition of the parameter X to the
linear equation describing the solvent eect improves the quality of the t to the
experimental data, especially in the case of alkali metal halide electrolytes involving larger ions. The parameter Y is not important for these systems but does
assist in the interpretation of other thermodynamic quantities which are solvent
dependent [48, 49]. Addition of these parameters to the analysis is only possible
when the solvent-dependent phenomenon has been studied in a large number of
solvents.
In conclusion, solvent acidity and basicity are useful parameters for the interpretation of solvent eects related to the solvation of electrolytes in polar media.
This role demonstrates the importance of the chemical nature of the interaction
between the ion and solvent molecule. Chemical interactions cannot be rationalized simply in terms of concepts from electrostatics based on point charges and
POLAR SOLVENTS
199
point dipoles. Although the parameters used to assess solvent acidity and basicity
are empirical they are related in a direct way to the molecular description of ion
solvation according to the MSA.

Attention in this chapter has been focused on the dielectric properties of polar
solvents and how they relate to the solvation of electrolytes. The important
problem discussed is how one uses statistical mechanics to derive the solution
permittivity from the molecular properties of polar liquids, namely, the dipole
moment and polarizability. In the classical or continuum approach developed
by Debye, Onsager, Frohlich, and Kirkwood, it became apparent that description of the molecular properties within a spherical cavity in the liquid required
the introduction of a correlation parameter gK . This parameter takes into
consideration the fact that the orientation of molecular dipoles with respect
to one another is not just determined by dipoledipole interactions but also by
chemical considerations such as hydrogen bonding. This fact is especially
apparent for liquid water in which the presence of strong hydrogen bonding
results in a rather high solvent permittivity. In the molecular model based on
the MSA, an additional parameter is required namely the molecular diameter in
a spherical representation of the individual molecule. This model is only successful when the molecular sphere with its embedded point dipole also is
sticky in a given direction. The stickiness factor plays the same role as the
correlation parameter gK in the classical approach. It accounts for chemical
interactions such as hydrogen bonding and the eects of non-sphericity in the
real system. The success of the MSA model is that it predicts values of gK
which are consistent with structural information about the liquid obtained
spectroscopically.
Dielectric relaxation experiments provide valuable information about the properties of polar solvents in an alternating electrical eld. The important parameter
characterizing the ability of the solvent dipole to respond to the changing eld is
the Debye relaxation time. Solvents with hydrogen bonding have several relaxation times, one of which corresponds to the rotational diusion of clusters of
molecules. The information provided by these experiments helps one to understand the role of the solvent in fast reactions occurring in solution. More about
this aspect of solvent eects is discussed in chapter 7. When these experiments are
conducted in electrolyte solutions one is able to assess ion pairing by examining
relaxation data at lower frequencies.
Ion solvation in polar solvents is also an important aspect of the subject matter
considered here. This is traditionally studied by measuring the Gibbs energy and
enthalpy of transfer of a variety of electrolytes from water to another polar
solvent. Single ion quantities are then derived on the basis of the TATB assumption. Study of these quantities for simple monoatomic ions like the alkali metal
cations and the halide anions leads to the conclusion that specic molecular
properties, namely, Lewis acidity and basicity, are important in ion solvation.
On the other hand, the dielectric permittivity, a non-specic bulk property,
200
does not play an important role. The analysis presented points out the importance
of the chemical structure of the solvent in stabilizing ions in the electrolyte solution. Thus, the conclusions reached here are in accord with those from chapter 3.
Physical laws such as Coulombs law and bulk parameters such as the permittivity
help one to understand long-range ionion interactions. However, to understand
close-range interactions, one must understand the chemical properties of the
solvent. In order to assess Lewis acidity and basicity, empirical parameters such
as the Gutmann acceptor number and donor numbers have proven to be quite
useful.
Ionsolvent interactions are conveniently studied by several spectroscopic techniques. These provide information about the structure of the solvent around the
ion including the solvation number, and the strength of the ionsolvent bond.
Information about these techniques is presented in the following chapter.
General References
G1. Bottcher, C. J. F.; Bordwijk, P. Theory of Electric Polarization, 2nd ed.; Elsevier:
Amsterdam, 1978.
G2. Hill, N. E.; Vaughn, W. E.; Prince, A. H.; Davies, M. Dielectric Properties and
Molecular Behaviour; Van Nostrand: London, 1969.
G3. Politzer, P., Murray, J. S. Quantitative Treatments of Solute-Solvent Interactions;
Elsevier: Amsterdam, 1994.
G4. Riddick, J. A.; Bunger, W. B.; Sakano, T. K. Organic Solvents, 4th ed.; Wiley: New
York, 1986.
G5. Barthel, J.; Buchner, R.; Munsterer, M. Electrolyte Data Collection, Part 2: Dielectric
Properties of Water and Aqueous Electrolyte Solutions; Dechema: Frankfurt-amMain, 1995.
References
1. Debye, P. Polar Molecules; Chemical Catalog Co.: New York, 1929.
2. Marcus, Y. Properties of Solvents; Wiley: New York, 1998.
3. Karapetyan, Yu, A.; Eychis, V. N. Physico-chemical Properties of Non-Aqueous
Electrolyte Solutions (in Russian); Khimia: Moscow, 1989; Chapter 1.
4. Lide, D. R., Ed. Handbook of Physics and Chemistry, 84th ed.; CRC Press: Boca
Raton, FL, 20032004.
5. Barthel, J.; Bachhuber, K; Buchner, R.; Hetzenauer, H. Chem. Phys. Lett. 1990, 165,
369.
6. Fawcett, W. R. Chem. Phys. Lett. 1990, 174, 167.
7. Castner, E. W., Jr.; Bagchi, B.; Maroncelli, M.; Webb, S. P.; Ruggiero, A. J.; Fleming,
G. R. Ber. Bunsen-Ges. Phys. Chem. 1988, 92, 363.
8. Garg, S. K.; Smyth, C. P. J. Phys. Chem. 1965, 69, 1294.
9. Barthel, J.; Bachhuber, K.; Buchner, R.; Gill, J. B.; Kleebauer, M. Chem. Phys. Lett.
1990, 167, 62.
10. Calderwood, J. H.; Smyth, C. P. J. Am. Chem. Soc. 1956, 78, 1295.
11. Krishnaji A.; Mansingh, A. J. Chem. Phys. 1964, 41, 827.
12. Davies, G. J.; Evans, G. J.; Evans, M. W. J. Chem. Soc., Faraday Trans. 2, 1979, 75,
1428.
13. Brownsell, V. L.; Price, A. H. J. Phys. Chem. 1970, 74, 4004.
14. Elie, V. Bull. Chem. Soc. Belges 1984, 93, 839.
15. Behret, H.; Schmidthals, F.; Barthel, J. Z. Phys. Chem. NF 1975, 96, 73.
POLAR SOLVENTS
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
42.
43.
44.
45.
46.
47.
48.
49.
50.
51.
201
Dachwitz, E.; Stockhausen, M. Ber. Bunsen-Ges. Phys. Chem. 1985, 89, 959.
Gauman, T. Helv. Chim Acta 1958, 41, 1933.
Wilhelm, E.; Battino, R. J. Chem. Phys. 1989, 55, 4012.
Lorentz, H. The Theory of Electrons; Dover: New York, 1952.
Balakier, G. Pol. J. Chem. 1980, 54, 2297.
Onsager, L. J. Am. Chem. Soc. 1936, 58, 1486.
Kirkwood, J. G. J. Chem. Phys. 1939, 7, 911.
Frohlich, H. Theory of Dielectrics; Oxford University Press: London, 1949.
Wertheim, M. S. J. Chem. Phys. 1971, 55, 4291.
Wertheim, M. S. Mol. Phys. 1973, 25, 211; 1973, 26, 1425.
Blum, L.; Fawcett, W. R. J. Phys. Chem. 1993, 97, 7185; 1996, 100, 10423.
Reid, D. S.; Vincent, C. A. J. Electroanal. Chem. 1968, 18, 427.
Fawcett, W. R.; Kloss, A. A. J. Chem. Soc., Faraday Trans. 1996, 92, 3333.
Barthel, J.; Buchner, R. Pure Appl. Chem. 1991, 63, 1473.
Friedman, H. J. Chem. Phys. 1982, 76, 1092.
Fawcett, W. R.; Tikanen, A. C. J. Phys. Chem. 1996, 100, 4251.
Barthel, J.; Hetzenauer, H.; Buckner, R. Ber. Bunsen-Ges. Phys. Chem. 1992, 96, 988.
Hynes, J. T. J. Chem. Phys. 1986, 90, 3701.
Fawcett, W. R. Chem. Phys. Lett. 1992, 199, 153.
Krishnan, C. V.; Friedman, H. L. In Solute-Solvent Interactions; Coetzee, J. F.,
Ritchie, C. D., eds.; Marcel Dekker: New York, 1976; Vol. 2, Chapter 9.
Cox, B. G.; Hedwig, G. R.; Parker, A. J.; Watts, D. W. Aust. J. Chem. 1974, 27, 477.
Cox, B. G.; Waghorne, W. E. Chem. Soc. Rev. 1980, 9, 381.
Alexander, R.; Parker, A. J.; Sharp, J. H.; Waghorne, W. E. J. Am. Chem. Soc. 1972,
94, 1148.
Marcus, Y.; Kamlet, M. J.; Taft, R. W. J. Phys. Chem. 1988, 92, 3613.
Gritzner, G. Pure Appl. Chem. 1990, 62, 1839.
Dimroth, K.; Reichardt, C. Liebigs Ann. Chem. 1969, 727, 93.
Reichardt, C. Solvents and Solvent Eects in Organic Chemistry, 2nd ed.; VCH
Publishers: New York, 1988.
Mayer, U.; Gutmann, V.; Gerger, W. Monatsh. Chem. 1975, 106, 1235.
Gutmann, V.; Wychera, E. Inorg. Nucl. Chem. Lett. 1966, 2, 257.
Gutmann, V. Coord. Chem. Rev. 1976, 19, 225.
Marcus, Y. J. Solution Chem. 1984, 13, 599.
Sandstrom, M.; Persson, I.; Persson, P. Acta Chem. Scand. 1990, 44, 653.
Fawcett, W. R. J. Phys. Chem. 1993, 97, 9540.
Fawcett, W. R. In Quantitative Treatments of Solute/Solvent Interactions; Politzer, P.
Murray, J. S., eds.; Elsevier: Amsterdam, 1994; Chapter 6.
Krygowski, T. M.; Fawcett, W. R. J. Am. Chem. Soc. 1975, 97, 2143.
Koppel, I. A.; Palm, V. A. In Advances in Linear Free Energy Relationships; Chapman,
N. B., Shorter, J., eds.; Plenum: London, 1972; Chapter 5.
Problems
1. Show that there is a relationship between the molar volume (table 4.1) and
molecular diameter (table 4.3) for the solvents considered in these compilations. Check to see whether the slope of the least-squares line has the expected
value.
2. The following data were obtained for dilute solutions of nitromethane in
carbon tetrachloride at 208C.
202
Mole fraction
NM
Density
r/kg m3
Relative
Permittivity
Refractive
Index
0
0.0343
0.0486
0.0693
0.1202
1594.0
1584.1
1580.6
1573.8
1557.1
2.238
2.710
2.958
3.257
4.038
1.4605
1.4582
1.4569
1.4559
1.4529
Using the method outlined in section 4.3A, estimate the molar polarization
and refraction in the limit of innite dilution. Use these results to estimate the
dipole moment and polarizability of NM and compare with the results in
table 4.3.
3. Estimate the Kirkwood polarization parameter for dimethylsulfoxide using
the MSA model described in section 4.4. Your result should agree with that
given in table 4.4.
4. Given the following dielectric relaxation data for methanol, nd the values of
the es , e1 , and tD , assuming there is only one relaxation process.
Frequency
/MHz
In-Phase
e0
Out-of-Phase
e00
5
1520.5
2016
4183
5216
7122
32.30
27.45
25.04
15.42
13.45
9.57
0
10.30
11.55
13.10
12.12
10.40
5. The following data were obtained for the permittivity of aqueous solutions of
LiBr.
Conc./M
es
0.5
1.0
1.5
2.0
3.0
4.0
5.0
69.29
62.75
56.53
51.85
42.56
35.66
30.44
Determine the best value of de and be in equation (4.6.1) using a least-squares

procedure, and assuming that es for pure water is 78.46. Compare these
values with the results for NaCl (equation (3.9.24)) and comment.
6. The following data are available for methanol assuming one relaxation
process:
POLAR SOLVENTS
Temperature/8C
es
e1
tD /ps
43.4
37
25
0
30
50.6
47.9
44.4
37.7
31.9
6.9
6.7
6.9
6.4
6.0
333
263
186
95
47
203
Estimate the longitudinal relaxation time at 25, 0, and 258C. Estimate the
activation enthalpy associated with both tD and tL .
7. The following data have been obtained for the free energy of transfer of Li
and Br ions from water to the given non-aqueous solvent. Analyze the
dependence of tr G on solvent using the equation Q Q0 as AN bs DN.
tr G =kJ mol1
Solvent
Li
Br
AN
DN
H2 O
MeOH
PC
DMF
AcN
NM
NB
DMSO
0
4.4
23.8
10
25
48
38
15
0
11.1
30.0
36.2
31.3
30
29
27.4
54.8
41.5
18.3
16.0
18.9
20.5
14.8
19.3
18.0
19.0
15.1
26.6
14.1
2.7
4.4
29.8
Determine the relative contributions of solvent acidity and basicity to the

observed variation in tr G . Compare the t with the above equation to a
simple linear t involving the more important parameter. Is the addition of
the second parameter justied?
8. The following data are reported for the Gibbs energy of transfer of the salts
LiCl and KBr. Perform the analysis used in question 7 with these data. What
do you conclude about the suitability of equation (4.9.8)?
tr G =kJ mol1
Solvent
LiCl
KBr
H2 O
MeOH
F
NMF
DMF
AN
DMSO
PC
NB
NM
0
16.7
5.86
13.8
31.0
66.1
20.1
61.9
0
20.5
2.93
22.6
38.5
13.4
35.6
57.2
44.1
Spectroscopic Studies of Liquid

Structure and Solvation
Gabor Palinkas
Erika Kalman
Gabor Palinkas was born in Budapest, Hungary, in 1942. He entered Eotvos

Lorand University in Budapest, graduating with a Ph.D. in physics in 1968. He
then joined the Laboratory of Structural Chemistry of the Hungarian Academy
of Sciences, where he undertook X-ray diraction studies of water and aqueous
electrolyte solutions. In 1973 he began a collaboration with Erika Kalman to
develop electron diraction methods for studying the structure of liquids. Their
work resulted in the rst experimental derivation of the pair correlation
functions in liquid water using existing neutron, X-ray, and electron diraction
data. He expanded his interests in 1975 to include molecular dynamics
simulations of these systems in a collaboration with Professor Karl Heinzinger
at the Max Planck Institute in Mainz, Germany. He has published extensively
in the area of solution structure, both aqueous and non-aqueous. He is
currently Director of the Chemical Research Center of the Hungarian Academy
of Sciences, and holds the position of Professor at the Technical University in
Budapest.
Erika Kalman was born in Budapest in 1942, where she grew up and went
to school. She attended university in Dresden, Germany and obtained a Ph.D.
in chemistry in 1970 after studying in the laboratory of Professor Kurt
Schwabe. She then returned to Budapest to begin work on the study of liquid
structure at the Hungarian Academy of Sciences. In 1973 she began a very
fruitful collaboration with Gabor Palinkas. She spent a sabbatical year in the
United States in 1975, where she worked in the laboratories of Harold
Friedman at Stony Brook and Alan Narten at Oak Ridge National
Laboratories. In the mid 1980s she helped edit an important series of books on
204
SPECTROSCOPIC STUDIES OF LIQUID STRUCTURE AND SOLVATION
205
the chemical physics of solvation. She has expanded her interests in recent
years to include studies of corrosion and nanochemistry. She currently directs
the Surface Chemistry and Corrosion Research group at the Chemical
Research Center of the Hungarian Academy of Sciences and holds the position
of Professor at the Technical University in Budapest.
Erika Kalman and Gabor Palinkas were married in 1974 and have two
children.
5.1 What Spectroscopic Techniques Are Available?

Spectroscopy involves the study of the interactions of electromagnetic radiation
with matter. In the case of liquids, radiation of a wide range of frequencies, and
thus energies, has been used, all the way from radio-frequency waves to X-rays.
Experiments involving neutrons, which are associated with very short wavelengths, are also important. In the spectroscopic experiment the incident radiation
may be either absorbed or scattered and the experimental information is obtained
by examining the intensity and direction of the radiation after it has passed
through the sample.
Several spectroscopic techniques will be considered in this chapter. X-ray and
neutron diraction techniques are powerful tools for studying the structure of
liquids and have been introduced in chapter 2. They may also be used to study the
structure of solutions and determine distribution functions for both the solute and
solvent. The feasibility of these experiments depends on the number of dierent
nuclei involved in the system. UV-visible spectroscopy is mainly used to study
electronic transitions in polyatomic species. These species are often complex ions
formed between the electrolyte and the solvent, or between the cation and one or
more anions. Vibrational spectroscopy involves electromagnetic radiation of
lower energy, usually in the infrared region. It is used to study intramolecular
vibrational modes and how they are altered by the environment in solution. It can
also be used to study the bonds formed between solute and solvent in the solvation process. Finally, nuclear magnetic resonance spectroscopy and its application
to the study of solvation will be discussed. This is a particularly powerful technique because it provides information about the environment of a given nucleus,
and experiments specic to a given nucleus can be carried out provided the
nucleus has a non-zero magnetic moment. Several other spectroscopic techniques
are commonly used [G1] but those considered here provide a representative
picture of what can be learnt from those experiments.
One should remember that the atoms and molecules in liquids are not motionless but in a state of ux determined by the intermolecular interactions and
temperature. From the study of microwave spectroscopy discussed in chapter 4,
it was found that rotational diusion processes in liquids are characterized by
relaxation times the order of a few picoseconds. When a liquid is irradiated with
UV or visible radiation which involves frequencies greater than 1015 Hz, the atoms
and molecules appear to be motionless because the frequency of the electromagnetic radiation is much greater than that associated with molecular motion
(1012 Hz and lower). The same is true for infrared spectroscopy except in the
206
far infrared where the frequency of the probing radiation is sometimes close to
that of the dynamic processes in the liquid. In the case of NMR spectroscopy,
radio-frequency radiation is used to study the relaxation processes of a given
nucleus in a magnetic eld. Since most relaxation processes in solution take
place at higher frequencies than the resonant frequencies in NMR, only the
average environment of a given nucleus in solution is involved in the experimental
observations.
In this chapter much of the discussion involves electrolyte solutions in both
aqueous and non-aqueous media. The role of spectroscopy in elucidating the
chemical interactions which cannot be understood on the basis of classical
electrostatics is emphasized. More details about each of the four spectroscopic
techniques and their application are now given.
5.2 X-Ray and Neutron Diffraction Studies

The important experimental features of X-ray and neutron diraction were discussed earlier in chapter 2. The incident radiation is characterized by an intensity
I0 and a wavelength l. In the case of neutrons the wavelength is calculated from
the momentum of the particles using the de Broglie equation. The neutron diffraction experiments discussed earlier involved monochromatic particles, that is,
neutrons with the same velocity and thus the same kinetic energy. Another type of
neutron diraction experiment involves neutrons with a range of energies. This is
called a time-of-ight experiment, and it requires detectors which respond to
particles of dierent energies at dierent scattering angles.
The quantity of interest in a diraction experiment is the intensity I(y)
observed at a scattering angle y or at wave number kD dened as
kD
4p siny=2
l
5:2:1
where kD has the dimension of reciprocal length. The scattered intensity at angle y
arises from three sources, namely, coherent, incoherent, and multiple scattering.
As was seen earlier in section 2.5, the structure function FT(kD) for the liquid
sample is determined from the coherent scattering. The relationship between
FT(kD) and the individual partial structure factors S ij(kD) for any two dierent
atoms i and j in the liquid is
!2
XX

X
FT kD
2 dij ci cj fi kD fj kD S ij kD 1 =
ci fi kD
5:2:2
ij
The ultimate goal of the diraction experiment is to determine each contribution

S ij (kD) to the structure function FT(kD). Then one can determine the individual
pair correlation functions gij(r) by Fourier transformation.
As pointed out in section 2.5, there are signicant dierences between results
from X-ray and neutron diraction experiments because of the way these species
are scattered by the atoms of the liquid. In the case of X-rays the scattering is due
to the electrons around each nucleus, and the scattering amplitude increases with
207
the atomic number of the atom. This means that when X-rays interact with liquid
water, they are signicantly scattered only by the oxygen atoms. As a result one
partial structure factor, namely, S OO(kD), makes a contribution to FT(kD). On the
other hand, in a neutron diraction experiment neutrons are scattered by both the
proton 1 H, and the oxygen atom 18 O.
X-ray and neutron diraction experiments dier signicantly in another
important way. In an X-ray experiment the scattering amplitude for atom i,
fi (kD), depends strongly on the reciprocal distance or wave number kD. As kD
increases fi (kD) decreases monotonically [1]. This means that the disentanglement
of the individual partial structure factors S ij(kD) requires knowledge of fi (kD). On
the other hand, in a neutron diraction experiment fi (kD) is independent of kD
and depends only on the nature of the atom and its nuclear components. As a
result, fi (kD) is given dierent names in these experiments. In an X-ray experiment, fi (kD) is usually referred to as the form factor, whereas fi (kD) in a neutron
diraction experiment is called the coherent scattering length [1].
Because of the important dierences between the two diraction experiments,
the strategies used in carrying out these studies are not the same. X-ray experiments are certainly more common since the equipment used is easily obtained.
Neutron experiments are carried out at a nuclear reactor site or at an accelerator
with the appropriate facilities. In the following sections, some results from diffraction experiments are presented with emphasis on the structural information
which has been obtained regarding ion solvation in electrolyte solutions.
A. X-Ray Diffraction Studies of Electrolyte Solutions
X-ray diraction has been used extensively to study the structure of aqueous
electrolyte solutions [G1]. When an electrolyte with monoatomic ions is dissolved
in water the number of dierent atoms in the system is four, namely, M and X,
from the electrolyte, and H and O from water. As a result there are ten independent partial structure factors. They are S OO, S OM, S OX, S OH, S HH, S HM, S HX,
S MM, S MX, and S XX. In a more complex system with n dierent atoms, the
number of independent partial structure factors is nn 1=2. In an X-ray
experiment, all partial structure factors involving the H atom are not important
so that the number is reduced from ten to six. Of these, two more may be
neglected, namely, those involving only cations or anions, that is, S MM and
S XX. Obviously, ions of the same sign tend to be as far as possible away from
each other. Finally, if ion pairing is negligible, S MX may also be neglected. Thus,
three signicant partial structure factors, namely, S OO, S OM, and S OX usually
contribute to the observed structure function FT(kD).
Because the scattering amplitudes fi (kD) vary with reciprocal distance, a normalized structure function H(kD) is usually dened in reporting X-ray diraction
data. Thus,
HkD FT kD =MkD
5:2:3
where M(kD) is chosen to change the scale to that characteristic of one molecule
and remove some of the breadth due to the electron distribution in the molecule
[2]. A convenient choice in the case of electrolyte solutions is
208
MkD
!2
xi fi kD
5:2:4
where xi is the mole fraction of atom i [3]. The value of kDH(kD) oscillates in an
irregular fashion and with decreasing amplitude. These data have been analyzed
on the basis of the rst-neighbors model [3] according to which
HkD xM HOM kD xX HOX kD xW HOO kD
5:2:5
where xM, xX, and xW are the mole fractions of cationwater aggregates, anion
water aggregates, and free water, respectively, and Hij is the normalized structure factor for atoms i and j. Accordingly, there are two states for water molecules, those bound in hydration shells and those in the normal (free) water
structure. In addition, the hydration shells only involve nearest neighbors which
are arranged in a very symmetrical fashion. Using this model, Palinkas et al. [3]
resolved the data for FT(kD) to obtain a model for alkali and alkaline earth metal
halide solutions. In this way they developed a model for ion hydration in these
systems. Their results are shown for 2 m solutions of the alkali metal chlorides in
gs 5.1 and 5.2. It is clear from these data that gOM(r) has a well-dened maximum at a value of r which is greater than 200 pm. In addition, the position of the
maximum shifts to higher values as the atomic mass of the alkali metal cation
increases. On the other hand, the position of the maximum for Cl on the gOX(r)
function is approximately independent of cation nature and occurs at 314 pm.
Fig. 5.1 Values of the normalized

structure factor kDH(kD) for 2 m
solutions of the alkali metal chlorides in
water as a function of reciprocal distance
kD [3]. The data for KCl, NaCl, and
LiCl have been shifted vertically by 20,
40, and 60 nm1, respectively, for the
sake of clarity.
209
Fig. 5.2 Pair correlation functions for

the cationwater, gOM(r), and anion
water interactions, gOX(r), in 2 m
aqueous solutions of the alkali metal
halides [3]. The data for NaCl, KCl, and
CsCl have been shifted vertically by 5,
10, and 15 units, respectively, for the
sake of clarity.
Many X-ray diraction studies of electrolyte solutions have been carried out in
aqueous solutions [G1, 4, 5]. Values of the most probable distance, roi, between
the oxygen atom in water and a number of monoatomic ions are summarized in
table 5.1. In the case of the cations, this distance reects the radius of the cation
plus the eective radius of the water molecule measured in the direction of the
lone pairs on oxygen. In the case of alkali metals, the eective radius of water
increases from 122 pm for Li to 131 pm for Cs when the Shannon and Prewitt
radii are assumed for the cations (see section 3.2), the average value being 127 pm.
This result can be attributed to the observation that the coordination number for
water molecules around an alkali metal or alkaline metal earth cation changes
with cation size and electrolyte concentration. In the case of the Li ion, this
number decreases from six in very dilute solutions to four in concentrated solutions [5]. Because of the electrostatic character of the interaction between the
cation and water molecules, these molecules exchange rapidly with other water
molecules in their vicinity. For this reason, the solvation coordination number
should be considered as an average.
In the case of transition metal cations, the coordinated water molecules are
covalently bonded to the cation, and the coordination number can be interpreted
as xed in the absence of other ligands. If the cations Mn2, Fe2, Co2, and Ni2
are assumed to be in the high spin state, then the eective radius of coordinated
water molecules in these systems is 123 pm on the basis of the X-ray data.
Interpretation of the roi data for monoatomic anions is much more dicult
because of the fact that anions are solvated in water via hydrogen bonding (see
210
Table 5.1 Most Probable Distance roi Between the Oxygen

Atom in Water and Monoatomic Ion i in Aqueous Electrolyte
Solutions [5]
Ion
Distance roi / pm
Coordination Number
Solution
Li
Na
K
Cs
Ag
210
242
280
315
242
6
6
6
6
4
2.2 M LiI
2 M NaCl
2 M KCl
2 M CsCl
3.3 m AgNO3
Mg2
Ca2
Sr2
212
244
264
6
6
8
1.1 M MgCl2
2 m CaBr2
1.6 m SrCl2
Al3
In3
187
215
6
6
3.8 m Al(NO3)3
5.0 m In(ClO4)3
Mn2
Fe2
Co2
Ni2
Cu2
220
212
210
207
194
6
6
6
6
4
2 m MnSO4
2.2 m Fe(ClO4)2
3.7 m Co(ClO4)2
3 m NiCl2
2.8 m Cu(ClO4)2
F
Cl
Br
I-
262
310
334
363
4.5
6
6
6.9
4.2 m KF
3 m CoCl2
2 m CaBr2
2.2 m LiI
g. 3.1). Thus, the ionoxygen atom distance cannot be interpreted in a simple

way as it is for cations [5].
X-ray diraction studies have also been carried out in non-aqueous electrolyte
solutions. In the case of methanol, there are two atoms which scatter X-rays,
namely, carbon and oxygen. When a monoatomic electrolyte is added the number
of scattering atoms increases to four. As a result, such a system has ten partial
structure factors. If the ionion correlations are neglected this reduces to seven. A
system which has been analyzed in some detail is MgCl2 in methanol [6, 7].
Analysis of the data gives a MgO distance of 207 pm and a ClO distance of
318 pm for coordination numbers of six [6].
As the number of atoms in the solvent molecule increases disentanglement of
the partial structure factors becomes more dicult. One way of making the analysis easier is to use dierent isotopes of the atoms involved. This technique has
been applied with success in neutron diraction experiments which are described
in the following section.
B. Neutron Diffraction Studies of Electrolyte Solutions
In the case of neutron diraction, the radiation is scattered by the atomic nuclei,
not by the electrons [1]. As a result, the scattering amplitude fi (kD) is determined
by the nucleons and depends in a complex way on their total number. This means
211
that isotope eects are very important in developing experimental strategies, and
that very light atoms such as hydrogen and deuterium have signicant values of
fi (kD) [8, 9]. As discussed earlier there are ten partial structure factors which can
contribute to the structure function Fi (kD) for a monoatomic electrolyte MX
dissolved in water. By using isotope substitution techniques, six of these can be
factored out, and only two of the remaining contributions are normally signicant.
Consider as an example the studies of LiCl in D2O by neutron diraction [8].
The isotopes 6Li and 7Li have very dierent scattering amplitudes, so that the
structure functions Fi (kD) are not the same for solutions of 6LiCl and 7LiCl.
Moreover, the scattering amplitude fi (kD) is independent of kD, that is, of scattering angle y, and depends only on the nature of atom i. By subtracting the
overall structure functions obtained in the presence of dierent Li isotopes, the
contributions from partial structure factors not depending on Li are removed.
Thus,
FT kD ; Li FT kD ; 7 Li FT kD ; 6 Li
5:2:6
depends only on the partial structure factors S LiO, S LiD, S LiLi, and S LiCl.
Furthermore, it is assumed that solution structure is independent of isotopic
nature, so that the partial structure factors for 7Li are exactly the same as those
for 6Li. Two of the above partial structure factors are negligible, namely those for
LiLi interactions, and LiCl interactions assuming that ion pairing is not signicant. Fourier transformation of FT(kD, Li) gives the overall correlation function GLi(r) which depends on the pair correlation functions gLiO and gLiD. A plot
of GLi(r) against distance r determined for LiCl in D2O is shown in g. 5.3. Two
clearly resolved peaks are seen. The rst corresponds to the most probable distance between the Li ion and the oxygen on the nearest water molecules (195 pm)
and the second to the corresponding distance between the Li ions and D atoms
(255 pm). If the same dierence is determined for 6LiCl and 7LiCl solutions in
light water, the dierence pair correlation function depends on gLiO and gLiH.
Fig. 5.3 The rst-order

dierence function GLi(r) for
LiCl in D2O as a function of
interatomic distance r. (From
212
Using the data obtained in light and heavy water, the individual pair correlation
functions can be separated. For the data in D2O, one has
GLi 2cLi cO fLi fO gLiO 1 2cLi cD fLi fD gLiD 1
5:2:7
where the independence of fi from kD has been noted. Similarly in H2O, one has
GLi 2cLi cO fLi fO gLiO 1 2cLi cH fLi fH gLiD 1
5:2:8
These are two equations in the unknowns gLiO and gLiD, assuming that gLiD and
gLiH are exactly the same [8, 9]. Thus, the two pair correlation functions may be
obtained separately. This method requires that careful and precise experiments be
carried out.
Another type of isotope experiment can be used to determine the eect of the
electrolyte on water structure. If one compares the overall pair correlation functions for a given LiCl concentration in H2O and D2O one obtains the dierence
function GLi(r). This function depends on the values of gij(r) which are due to
the two H isotopes. They are gOH, gLiH, gClH, and gHH, where the subscript H
designates both 1H and 2H. To a good approximation, one expects to nd changes
only in gHH with change in electrolyte concentration. Experiments have shown
[10] that the pair correlation function gHH in 1 m LiCl is almost the same as that in
pure water, thereby demonstrating that the hydrogen bonding in the solution is
very little perturbed by the added electrolyte. However, when the electrolyte
concentration is increased to 10 m, there is a signicant change in gHH, indicating
that the number of hydrogen bonds in the very concentrated solution is about
70% lower than in pure water.
Using the second dierence technique the partial pair correlation functions
gClH and gClO have been determined for the Cl ion in water [11]. The experiments
involve rst determining the dierence function GCl(r) using chloride salts rich
in 35Cl or 37Cl. Then each of these experiments is done in either H2O or D2O to
achieve separation of the gClH and gClO partial pair correlation functions. The
results conrm that the orientation of the water molecule with respect to the Clion is that shown in g. 3.1.
It is useful to consider the general characteristics of the partial pair correlation
function for a metal ion coordinated to water molecules. Given suciently precise
data and metal isotopes with dierent scattering fractions fi(kD), the pair correlation function gMO is the ultimate goal of experiments designed to study ion
solvation. The shape of gMO for the cases of strongly and weakly coordinated
water molecules is shown in g. 5.4. When the water is tightly bound to the metal
ion, the rst peak in gMO(r) is high and sharp, reaching values greater than 2. The
position of the peak gives the MO distance. By integrating gMO(r) up to the rst
minimum at rs, one obtains an estimate of the coordination number for the ion
(see equation (2.10.1)). For Ni2 in water, this number is 6 as one would expect [8,
9]. In addition, the average residence time for such a water molecule is quite long,
specically, the order of 10 ms. When the water molecules are weakly bound to the
central metal ion the appearance of the partial pair correlation function gMO(r)
changes signicantly (see g. 5.4). The rst peak on this curve is lower and
broader and the subsequent minimum much less deep. Cations with this type of
213
Fig. 5.4 Generic radial distribution functions gMO for the metal ion Mn in water for cases
of (a) strong coordination and (b) weak coordination. rMO is the optimum metal ion
oxygen distance in the solvated species. The coordination number is estimated by integrating the distribution function out to the minimum at rs. (From reference 8, with permission.)
gMO are less strongly hydrated and have water residence times in the rst solvation layer which are of the order of 10 ps.
Not only does neutron diraction allow one to determine ionic size and hydration numbers in solution but it can also be used to assess changes in hydration
with concentration. In the case of Li, the hydration number is 6 in dilute solutions but it drops to values below 4 in very concentrated solutions. Similar conclusions have been reached regarding divalent cations such as Ca2 for which the
ionsolvent interactions are mainly electrostatic in nature. For this system the
hydration number decreases from 10 in 1 M CaCl2 to 6 in a 4.5 M solution of the
same salt.
The important limitation of neutron diraction experiments is that the necessary isotopes are often not available so that the dierence technique described
here cannot be applied. Thus, neutron diraction studies have been carried out to
study solvation of Li and K cations but not Na. In the latter case, X-ray
techniques, which do not provide information about hydrogen bonding, are used.
In summary, diraction techniques provide a powerful means of investigating
the structure of electrolyte solutions. They give information about the pair correlation functions which can be directly related to modern theoretical techniques
such as molecular dynamics calculations. This information can also be used to
improve the statistical thermodynamic models of electrolyte solutions discussed in
chapter 3.
5.3 Nuclear Magnetic Resonance Spectroscopy in

Solutions
Nuclear magnetic resonance (NMR) spectroscopy is based on the magnetic properties of the nuclei which make up the molecules and ions in solution. Each
nucleus has a spin quantum number I which depends on the number of protons
214
and neutrons in the nucleus and the way in which the spins of these nucleons
interact. The spin quantum number can have integral or half-integral values so
that I may equal 0, 1/2, 1, 3/2, . . . NMR spectroscopy was rst applied to protons
(1H) for which I is equal to 1/2. In the case of deuterium, which has one proton
and one neutron, the nuclear spin quantum number is 1. Atoms with an even
number of protons and an even number of neutrons have I equal to zero; examples are 12C, 16O, and 32S. The magnetic properties of some nuclei which have
been studied experimentally are summarized in table 5.2.
The magnetic energy level is degenerate with a total of 2I 1 quantum states.
In the presence of an external magnetic eld these quantum states have dierent
energies, each with a quantum number mM. For example, the proton for which
I 1/2 has two energy levels with mM 1/2 or 1/2, and the deuteron for which
I 1 has three energy levels with mM 1, 0, or 1.
Another important property of a magnetic nucleus is its magnetogyric ratio
gM , which is also given in table 5.2. This gives a measure of how strongly or
weakly the nucleus interacts with an external magnetic eld, and can be considered as the magnetic analog of the extinction coecient in optical spectroscopy.
Using this quantity and the spin quantum number one can calculate the magnetic
moment for the nucleus, which is given by
lM
gM Ih
2p
5:3:1
In the presence of a magnetic eld B0 the energy levels split according to the
individual values of lM. The dierence in energy or energy gap between these
levels is given by
hn
lM B0 gM hB0
I
2p
5:3:2
and depends on the strength of the magnetic eld. This is illustrated for the proton
in g. 5.5.
Table 5.2 Magnetic Properties of Some Nuclei
Nucleus
1
H
H
7
Li
13
C
14
N
17
O
19
F
23
Na
35
Cl
33
S
2
Spin Quantum Number I
Magnetogyric Ratio gM 107 =rad T1 s1
1/2
1
3/2
1/2
1
5/2
1/2
3/2
3/2
3/2
26.75
4.11
10.40
6.73
1.93
3.63
25.18
7.08
2.64
2.05
215
Fig. 5.5 Energy level splitting for a proton

(I 1=2) in an applied magnetic eld B0.
EXAMPLE
Calculate the magnetic moment for a proton from its magnetogyric ratio.
What is the frequency corresponding to the energy gap in a magnetic eld of
4.69 T?
The magnetic moment lM is
lM
26:75 107 1=2 6:626 1034

1:41 1026 J T1
2p
5:3:3
The energy gap is

hn
1:413 1026 4:69

1:325 1025 J
1=2
5:3:4
This corresponds to a frequency of 200 MHz.

From a classical point of view a nucleus which possesses a magnetic moment
(I 6 0) can be considered as a spinning charged particle. If I 1=2, the nucleus
behaves as a spinning charged sphere; if I is larger, then the nucleus is equivalent
to a non-spherical spinning charge, and, in addition, possesses a quadrupole
moment.
When a magnetic nucleus is placed in a magnetic eld B0 it tends to adopt the
direction of the eld. However, because of its spin, it is not aligned exactly with
the external eld but instead at an angle to it, as shown in g. 5.6. The net result is
that the magnetic moment vector lM precesses about the magnetic eld vector
B0. The angular velocity for this precession is called the Larmor precessional
frequency and is given by
oL gM B0
5:3:5
If oL is expressed as a normal frequency n, it is clear that n is given by equation

(5.3.2).
Now consider the eect of a small additional magnetic eld B1 applied perpendicular to the main eld B0 . If the eld B1 rotates around B0 at the Larmor
precessional frequency oL , a torque is produced on vector lM which acts to
216
Fig. 5.6 Magnetic moment vector lM

precessing about the applied eld vector B0 (a);
the same system with addition of a rotating
magnetic eld B1 perpendicular to B0 (b).
change the angle y (g. 5.6). On the other hand, for other frequencies, the interaction between lM and B1 varies depending on the relative phases of the two
motions and there is no sustained eect. When B1 rotates at the Larmor frequency
the nuclear magnetic moment can absorb energy and assume a higher energy level
in which lM points in the opposite direction. The absorption of energy by the
nucleus under these circumstances is called nuclear magnetic resonance.
In the NMR experiment the sample containing the magnetic nuclei is exposed
to a xed magnetic eld B0 and a rotating eld B1 , as described above. The
frequency of the rotating eld is swept through a range of values including the
Larmor frequency for the nucleus in question. The absorption of energy at this
frequency is observed when a plot of energy absorbed against frequency is made.
This is the NMR spectrum. The rotating magnetic eld is generated using a
linearly oscillating eld of radio frequency. The linearly oscillating eld can be
resolved into two rotating magnetic elds of equal magnitude but opposite rotational directions. Only the component rotating in the same direction as lM interacts with the nuclei.
The above experiment is called a frequency sweep experiment. Equivalent
results are obtained if the radio frequency is held constant and the magnetic
eld B0 is changed. This is called the eld sweep technique.
The relative populations of the separate energy levels for a nucleus in a magnetic eld and the relaxation processes from a higher energy level to a lower one
are also of fundamental interest. On the basis of Boltzmanns law and equation
(5.3.2) the ratio of the number of nuclei in the higher energy level to that in the
lower level for a proton (I 1=2) is

Nu
gM hB0
5:3:6
exp
N1
2pkB T
Since the energy gap between these levels is small with respect to kB T for typical
values of B0 , the exponential can be replaced by the rst two terms in its expansion as an innite series. Thus,
Nu
g hB
1 M 0
N1
2pkB T
217
5:3:7
EXAMPLE
Estimate the ratio of Nu =N1 for protons in a magnetic eld of 4.69 T at 258C.
The energy gap between the two levels is
hn
26:75 107 6:626 1034 4:69

1:325 1025 J
2p
5:3:8
The value of kB T at 258C is

kB T 1:381 1023 298:2 4:118 1021 J
5:3:9
Therefore the ratio Nu =N1 is

Nu
1:325 1025
1
1 3:22 105
N1
4:118 1021
5:3:10
Thus the dierence in population for these two levels corresponds to 32.2 ppm.
There are two kinds of relaxation processes in NMR, namely, spinlattice and
spinspin relaxation. The lattice is the general collection of atoms and molecules
surrounding the nuclei which interact with the magnetic eld. The lattice system is
involved in vibrational and rotational motion reecting the thermal energy of the
system. Interaction between a nucleus which has absorbed energy from the magnetic eld and the lattice causes this nucleus to return to the ground state, thereby
transferring energy which increases the vibrational and rotational motion of the
lattice. This is spinlattice relaxation which is a rst-order process characterized
by a relaxation time T1 . It is also called longitudinal relaxation because it is
associated with change in the component of the magnetization vector of the
nucleus Mz in the direction of the applied magnetic eld B0 . The relaxation
time T1 depends on the magnetogyric ratio of the nucleus gM and thus on the
Larmor frequency at which it absorbs energy. It also depends on the characteristics of the lattice. The fact that T1 is nite results in broadening of the lines in the
NMR spectrum. In liquids, values of T1 fall in the range from 102 to 102 s. The
relationship between the width of the NMR line and T1 is
n
1
2pT1
5:3:11
The other mechanism by which the system returns to equilibrium after absorption of energy is spinspin relaxation. This is attributed to interaction of the spin
vectors on adjacent nuclei. The precessing moment on nucleus 1 sets up an oscillating eld at nucleus 2; this eld momentarily comes into phase with that of the
other nucleus which absorbs energy from nucleus 1. The resulting transfer of
energy between the nuclei is called spinspin relaxation. The other mechanism
of line broadening associated with spinspin relaxation is due to the local inhomogeneity of the magnetic eld B0 . Because the environment of each nucleus
contains other magnetic nuclei, the local value of B0 , Blocal , is slightly dierent.
As a result there are a range of Larmor frequencies for the responding nuclei
218
depending on the range of values of Blocal . Spinspin relaxation can also be

described in terms of the magnetization vectors in the x; y-plane perpendicular
to the main magnetic eld B0 . These vectors Mx and My achieve nite values after
the absorption of energy by the magnetic nuclei. They decay to zero, their equilibrium values, with a rate constant 1=T2 . Thus, spinspin relaxation is also called
transverse relaxation. Both longitudinal and transverse relaxation of the magnetization vectors Mx , My , and Mz are illustrated in g. 5.7.
An understanding of relaxation processes in the resonance process allows one
to explain the line shapes observed experimentally. As will be seen in chapter 7,
chemical reactions, for example, proton transfer reactions, can lead to band
broadening in the NMR spectrum. Thus, this feature of the spectra has very
important applications in the study of fast chemical reactions in solution.
In the discussion of the Larmor frequency above, the eect of the magnetic
environment on the absorbing nucleus was ignored. The eectiveness of NMR
spectroscopy in chemistry is denitely a result of the sensitivity of the absorbing
nucleus to the nature of the nuclei in its immediate environment. Consider for
example the case of ethanol, C2H5OH. One may distinguish three dierent protons with respect to environment, namely, the proton in the OH group, the
protons in the methyl group, and the protons in the CH2 or methylene group.
Each of these species experiences a dierent magnetic environment, so that the
NMR spectrum of ethanol has three absorption peaks at low resolution. The
shielding eect of the environment is expressed quantitatively in terms of the
shielding constant s. Thus, if oL is the Larmor angular frequency in the absence
of a magnetic environment, the angular velocity in the presence of this environment is
o gM B0 1 s
5:3:12
Fig. 5.7 Illustration of the magnetization vectors Mx , My , and My for a system (a) in the
presence of an external magnetic eld B0 in the z-direction but no transverse eld B1 ; (b)
after application of B1 in the (x; y)-plane and absorption of energy; and (c) after partial
relaxation back to the equilibrium conguration in (a).
219
The observed peaks in the NMR spectrum are usually cited in terms of chemical
shifts measured with respect to the peak for a reference compound. In the case of
1
H spectra this compound is usually tetramethylsilane, (CH3)4Si, which is added
to the sample placed in the spectrometer. The protons in this reference are all
equivalent and therefore give one NMR peak. The chemical shifts for the other
peaks observed are dened as

o or
d 10
or
6
5:3:13
where or is the angular frequency at which the reference peak is observed. Values
of d are given in parts per million through the factor 106. Positive contributions to
s are termed diamagnetic, whereas negative contributions are paramagnetic.
The description of NMR spectroscopy given here is applicable to a continuous
wave instrument in which the radio-frequency is swept continuously over the
appropriate range. Most modern instruments are pulsed or Fourier transform
spectrometers (FT-NMR). In the pulsed instrument, the sample is irradiated
with periodic pulses of radio-frequency energy that are directed through the
sample at right angles to the magnetic eld. The short pulse, which lasts the
order of 10 ms, results in a change in the net magnetization of the sample system
away from the z-direction of the permanent magnetic eld. When the pulse ends,
the magnetization relaxes back to its equilibrium position in a process called free
induction decay (FID). The FID signal is detected by the spectrometer and stored
digitally. By repeating this process many times, the signal-to-noise ratio of the
FID is improved. Finally, the average FID which gives the NMR spectrum in the
time domain is Fourier transformed to give the frequency domain spectrum.
In the early stages of NMR spectroscopy, most work involved studies of protons in organic compounds. However, instruments are now available for studying
a wide variety of other nuclei. In this chapter, attention is focused on NMR
studies of ion solvation. Work carried out in both aqueous and non-aqueous
solutions is considered in the following sections.
5.4. NMR Studies of Ion Solvation in Water

One of the most important applications of NMR spectroscopy is the determination of the solvation number of ions in aqueous solutions. This can be accomplished in some cases using 1H NMR or using 17O NMR in isotopically enriched
water. An example involving the highly charged Al3+ ion is shown in g. 5.8. The
proton spectrum of an aqueous solution of Al(ClO4)3 at a temperature of 46:38C
shows two peaks, a large one corresponding to the protons in bulk water, and a
lower one shifted downeld corresponding to the protons associated with the
Al3+ ion. By integrating the area under these peaks one can determine the solvation number of Al3+ which is 6. As the temperature is raised, both lines broaden
and move toward coalescence due to proton exchange processes, so that at
26:48C the absorption due to coordinated water molecules is scarcely visible
[G1, G3].
220
Fig. 5.8 Proton NMR spectra of

aqueous 2.1 m Al(ClO4)3 at (a) 46:38
and (b) 26:48C. (From reference
G1, with permission.)
The above experiment only works when the rate of exchange of the solvent is
slow with respect to the frequency separation between the coordinated and bulk
water molecules. If this frequency is 1000 Hz, then the relaxation time characterizing solvent exchange should be signicantly greater than 1 ms. This is the case
for highly charged main group cations or for most transition metal ions for which
water is a ligand. Data for some diamagnetic cations are summarized in table 5.3.
It can be seen that d increases roughly with the ratio zi =ri , that is, with the ratio of
ionic charge to ionic radius. However, this is not a complete explanation, as can
be seen from the Al3+, Ga3+, In3+ series. The downeld shift also increases with
the polarizability of the ion.
Paramagnetic cations such as Mn2+, Fe2+, Co2+, and Ni2+ cause a much
larger downeld chemical shift of the protons in coordinated water molecules. In
some cases 1H NMR data can be used to study hydration for these systems in 17Oenriched water. Since the gyromagnetic ratio for 17O is much less than that for 1H
the downeld chemical shift in hertz is not as large. 17O NMR spectroscopy has
the advantage that 17O exchange processes between adjacent water molecules are
much slower than the corresponding 1H exchange processes. As a result, one
source of line width broadening is greatly reduced.
When the rate of solvent exchange around the ion is fast, only one line is seen
in the NMR spectrum. In this case, the 1H nucleus in the solvent may experience
all possible environments in the solution within the time corresponding to the
Table 5.3 1H NMR Data for Water in the Primary
Solvation Shell of Diamagnetic Cations
Cations
Cation Chemical Shifta / ppm
Solvation Number
Be2+
Mg2+
Al3+
Ga3+
In3+
(Water)
8.04
5.55
8.74
8.98
7.22
4.13
4
6
6
6
6
Measured at low temperatures with respect to ethane gas as reference.
221
radio frequency of the NMR experiment. Such a situation holds in a solution of

NaCl. To a good approximation, the proton in the water molecule experiences
three environments, that near the cation, that near an anion, and that in the bulk
solvent far away from either ionic component. The experimentally observed chemical shift is an average for the protons in each of these environments. As a result
the observed chemical shift changes with electrolyte concentration [12, 13].
Proton chemical shifts observed on addition of simple electrolytes to water are
given in table 5.4. They depend on the nature of the electrolyte and can be either
positive or negative. In the limit of dilute solutions the shifts are linear in concentration, reecting the fact that solventsolvent interactions are being replaced
by ionsolvent interactions. The non-linearity at higher concentrations is attributed to ion pair formation. Two phenomena contribute to the observed chemical
shift. One involves the breaking of hydrogen bonds, which leads to a positive
shift. The other involves polarization of the water molecule in the eld of the ion,
which gives a negative shift. The former eect predominates for most electrolytes.
Obviously, the chemical shifts given in table 5.4 depend on both cation and
anion, and it is interesting to try to separate their contributions. The eect of the
cation was studied by Davies et al. [13], who measured the limiting slope of the
chemical shift against concentration plots for a series of alkali metal and alkaline
earth metal chlorides (see g. 5.9). When the molal chemical shift is plotted
against the polarizing strength of the cation estimated as the charge/radius
ratio, z=r, one nds that the most positive shifts in these series are for K+ and
Ba2+. As z=r increases the molal shift becomes more negative due to the increase
in polarizing strength of the cation. In the alkali metal series, this trend changes
direction at K+, the molal shifts for Cs+ and Rb+ being less positive than one
would expect on the basis of a simple linear correlation. This probably reects the
polarizability of the largest cations which should also be considered in estimating
cationic polarizing strength.
The trends for the alkali metals and alkaline earth metals are separated by
0.033 ppm. As argued by Davies et al. [13], this dierence gives the eect of the
extra Cl ion which is in the alkaline earth metal solution (MCl2). By shifting the
dependence for the alkaline earth metal chloride solutions down by two times
0.033 ppm, and that for the alkali metal chlorides down by 0.033 pm, one trend is
Table 5.4 Molal Chemical Shifts for the Proton Resonance of Water in Electrolyte
Solutions at 258C [13]
Electrolyte
NaF
KF
LiCl
NaCl
KCl
RbCl
CsCl
a
Chemical Shifta d / ppm kg mol1

0.021
0.008
0.047
0.096
0.111
0.092
0.080
With respect to that for pure water.
Electrolyte
NaBr
NaI
NaCN
NaNO3
NaBF4
NaSCN
NaClO4
Chemical Shifta d / ppm kg mol1

0.128
0.175
0.113
0.081
0.123
0.143
0.140
222
Fig. 5.9 Molal chemical shifts for the alkali metal and alkaline earth metal chlorides in
aqueous solution plotted against the polarizing strength of the cation estimated as z=r
(Pauling scale). The data shown as () have been corrected for the contribution of Cl ,
and represent estimates of the cation contribution (see text).
found for both series. This analysis gives an approximate method of separating
cation and anion eects. In assessing it, one should remember that the anion
interacts directly with the proton whose magnetic resonance is being measured,
whereas the cation interacts indirectly through the oxygen atom in the water
molecule. The eect of the halide ions for a given alkali metal ion is such that
the molal chemical shift becomes increasingly more negative with increase in
polarizing strength. Anions are known to have very little disruptive eect on
water structure, so that the trend observed mainly reects the polarization of
the proton by the anion.
Another way of studying solvation eects for simple diamagnetic ions is to
measure chemical shifts for the solute ion. In the case of aqueous solutions of
Na+ electrolytes, experiments have shown that the 23Na+ chemical shift is proportional to the mole fraction of the electrolyte over a wide concentration range
[14, 15], so that it may be expressed as
d d e xe
5:4:1
where de is the value of the chemical shift characterizing the given electrolyte at
unit mole fraction and xe , the mole fraction of the electrolyte. Similar results are
available for 39K [16] and other alkali metal cations [14]. These results shed
considerable light on the nature of the interactions of these simple cations with
their environment.
Values of de for seven Na+ electrolytes are summarized in table 5.5. A positive
shift occurs when the interaction of the Na+ cation with the anion is less than that
with a water molecule. This situation occurs in NaClO4 solutions over a very wide
range of concentrations. At a mole fraction of 0.25, every Na+ ion is accompa
nied by one ClO
4 and three molecules of water in solution, so that ClO4 ions are
223
Table 5.5 Value

of de for various
11 electrolytes
containing 23Na+
[15]
Electrolyte
NaClO4
NaNO3
NaBPh4
NaCl
NaBr
NaI
NaOH
de /ppm
24
19
0
7
10
13
22
expected to be in the immediate environment of the cation. The NMR results

show that there is no signicant interaction between the electron clouds of these
two ions. As the basicity of the donor ion or molecule increases, the chemical shift

becomes more negative according to the following series: ClO
4 < NO3

< H2 O < Cl < Br < I < OH . This series also follows an increasing tendency to form ion pairs. Thus, in the case of Na+ and I , signicant overlap
can occur between the electron clouds of these ions, so that I shields the 23Na+
nucleus more eectively than a water molecule. As a result there is a moderate
tendency to form ion pairs in concentrated NaI solutions.
Studies of the nuclear resonances of 35Cl, 81Br, and 127I have also been carried
out in various alkali metal halide solutions [17]. The magnitude of the chemical
shift increases with electrolyte concentration and also with atomic number of the
anion. In the case of the alkali metal ions the chemical shift becomes more positive
in the series Na < K < Li < Rb < Cs . The results were attributed to direct
interaction between the cation and anion in solutions containing K+, Rb+, and
Cs+. With the smaller cations, interactions between the halide ion and the water
molecules solvating the cation are more important.
The above discussion provides a brief introduction to the NMR spectroscopy
of solvation in water. Much more has been learnt about the kinetic aspects of
solvation by studying relaxation times associated with the magnetic resonance
lines. Detailed information is available in several reviews [G1, 18, 19]. Clearly,
NMR spectroscopy is a powerful tool in the study of ionsolvent and ionion
interactions in aqueous solutions and has helped greatly to improve the understanding of electrolyte solutions at the microscopic level.
5.5 NMR Studies of Ion Solvation in Non-Aqueous

Solvents
It is not surprising that 1H NMR was also applied to the study of cation solvation
in non-aqueous media. Most solvent molecules contain protons, but in the case of
224
aprotic solvents these protons often do not interact directly with the ions of the
electrolyte. Proton NMR has been used to determine solvation numbers of highly
charged cations such as Al3+ in solvents such as methanol, dimethylformamide,
and dimethylsulfoxide [G3]. Since the solvent molecules are tightly bound to the
cation, their exchange with molecules in the bulk is slow and two peaks are seen in
the NMR spectrum, one corresponding to the solvent molecules coordinated to
the cation and the other to free solvent molecules. By comparing peak areas, the
number of coordinating solvent molecules per cation can be determined. This is
typically six for cations such as Mg2+, Zn2+, Al3+, Ga3+, Co2+, and Ni2+ [G3].
Studies have also been carried out with systems where the exchange of solvent
molecules between the solvation shells of the ions and the bulk is rapid. In this
case the chemical shift of the nucleus being probed, for example protons, depends
on the nature of the electrolyte and its concentration. A typical example is the 1H
chemical shifts observed for some 11 electrolytes in acetonitrile [20], which are
summarized in table 5.6. In all cases the shift is to a lower frequency, indicating
that the protons at the positive end of the molecular dipole are more strongly
shielded in the presence of the electrolyte than in pure acetonitrile. Distinct trends
are dicult to discern due to complications resulting from dierent degrees of ion
pairing, but the perchlorate anion has the least eect among the anions considered. Among the halide anions, the trend follows the order Cl Br > I . The
eect of the large tetraalkylammonium ions is small because of their size and the
fact that they interact with the negative end of the molecular dipole. The inuence
of the alkali metal cations increases approximately in the order K < Na < Li ,
that is, in the order of their polarizing power.
Another method of examining ionsolvent interactions is to study the chemical
shifts of the solute ions. This type of investigation has been carried out for the
7 + 23
Li , Na+, and 205Tl+ cations and for the 35Cl, 81Br, and 127I anions.
Table 5.6 Molar Chemical Shifts

for the Proton Resonance of
Acetonitrile in Electrolyte Solutions
at 258C [20]
Electrolytea
LiI
NaI
KI
LiClO4
NaClO4
Et4 NCl
Et4 NBr
Bu4 NCl
Bu4 NBr
Bu4 NI
Chemical Shiftb d/Hz M1

6.5
7.5
6.0
4.0
3.0
7.5
7.4
6.3
6.9
5.0
Et4 N is tetraethylammonium and Bu4 N ,

tetrabutylammonium.
b
With respect to that for pure acetonitrile.
a
225
Variation in the 23Na+ chemical shift with solvent nature has been very helpful in
elucidating the nature of the ionsolvent interactions, especially for aprotic solvents. Popov [21] showed that there is a linear correlation between the chemical
shift observed for 23Na+ in an aprotic solvent and the Gutmann donor number
DN when an electrolyte with minimal ion pairing is used. Values of d obtained in
NaPh4B are plotted against DN for 11 aprotic solvents in g. 5.10. A very good
correlation is found with a correlation coecient of 0.976. Data points for formamide, water, and methanol are also shown in this plot. The result for formamide falls on the regression line but those for water and methanol are clearly
below it. This led to the suggestion [21] that the eective donor numbers of water
and the alcohols are much higher than the values given by Gutmann [22] when
these solvents have their bulk structure with the accompanying hydrogen bonding. However, the Gutmann values of DN for these solvents are conrmed by
independent estimates of solvent basicity (see section 4.9).
Chemical shifts for the 7Li nucleus have also been studied as a function of
solvent nature for a variety of electrolytes including lithium perchlorate and
lithium tetraphenylborate [21]. Although they change with solvent nature, there
is no correlation with solvent donicity. Popov attributed the failure of the 7Li
system to monitor basic properties of the solvent to the fact that the paramagnetic
and diamagnetic screening constants for this atom are approximately equal, so
that they cancel one another. In the case of 23Na, the paramagnetic screening
constant is much larger than the diamagnetic one, so that this system can be used
to monitor the basicity of aprotic solvents.
Solvation of the halide ions in various solvents has been studied by measuring
the chemical shifts for 35Cl, 79Br, and 127I in water and several other polar
solvents [23]. The chemical shifts are strongly solvent dependent and correlate to
the UV absorption band energy corresponding to charge transfer to the solvent.
Fig. 5.10 Chemical shift for 23Na using sodium tetraphenylborate dissolved in various
solvents [21]. The regression line was drawn considering only aprotic solvents (lled
circles). Abbreviations for solvents given in chapter 4: PYR pyridine.
226
Polar molecular solutes have been used to probe the donoracceptor properties
of polar solvents. 19F chemical shifts have been measured for interaction between
triuoroiodomethane and the solvent molecule as electron pair donor [24]. As
interaction between the donor molecule and the iodine atom in this molecule
increases, electron density at the uorine atoms increases with a resulting positive
chemical shift in the 19F NMR signal. An excellent correlation between these
shifts and the Gutmann donor number was reported [24].
The 31P chemical shifts in triethylphosphine oxide measured in polar solvents is
used to dene the acceptor number scale for solvent acidity [25]. In this case, the
oxygen atom in the PO bond acts as an electron pair donor to the solvent as a
Lewis acid. The resulting inductive eect lowers the electron density at the phosphorus atom and results in a chemical shift which depends on solvent acidity.
In summary, NMR techniques based on the chemical shifts of a solute atom
provide a useful way of probing solutesolvent interactions. They demonstrate
clearly that the chemical nature of these interactions is important, and must be
considered in developing an overall understanding of the structure of both electrolyte solutions and solutions of polar molecules.
5.6 Vibrational Spectroscopy in Solutions

Vibrational spectroscopy is a powerful tool for studying solvation in polar solvents. These solvent molecules have permanent dipole moments and therefore
absorb radiation in the infrared region of the electromagnetic spectrum due to
the vibrational modes of motion. In the present section, the principal experimental techniques used in vibrational spectroscopy are outlined with emphasis on
methods for analyzing solvent structure and solvation.
A non-linear polyatomic molecule with N atoms has 3N 6 degrees of vibrational freedom. Each vibrational mode is characterized by a frequency which is
determined by the force constant characterizing the vibrational motion, and the
reduced mass of the system in a harmonic oscillator representation. For example,
acetonitrile (CH3CN) has six atoms and 12 degrees of vibrational freedom. A very
important vibrational mode in this molecule is the CN stretch, which is associated with the polar cyano group. If this group interacts with an ionic or molecular solute in the solvation process, the electron density in the bond is aected,
and a change in the vibrational frequency is observed. Other vibrations in this
molecule are associated with the methyl groups and the CC bond.
EXAMPLE
Estimate the force constant for the CC bond in acetonitrile using the harmonic oscillator approximation, given that the frequency of the CC stretching
mode expressed in wave numbers is 918 cm1 .
In the harmonic oscillator approximation the vibrational frequency is given
by
n

1 k 1=2
2p m
5:6:1
227
where k is the force constant and m, the reduced mass. The reduced mass is
estimated assuming that the vibrating groups are methyl (15.04 g) and cyano
(26.02 g):
m
0:01504 0:02602
1:583 1026 kg molec1
6:022 1023 0:01504 0:02602
5:6:2
In SI units, the frequency is

n 918 2:998 1010 2:752 1013 s1
5:6:3
On the basis of equation (5.6.1), the force constant is

k 4p2 1:583 1026 2:752 1013 2 473 N m1 molec1
5:6:4
There are two kinds of vibrational spectroscopy commonly used to study

solvent structure and solvation, namely, infrared and Raman spectroscopy.
These methods dier signicantly with respect to the nature of the vibrational
features which may be observed. Therefore, some fundamental concepts and
experimental methods are introduced before specic results are presented.
A. Infrared Spectroscopy
Infrared spectroscopy is normally carried out in the mid-infrared region, that is,
from 4000 to 200 cm1 . The near infrared region is at higher frequencies (12,500
to 4000 cm1 ) and the far infrared, at lower frequencies (200 to 10 cm1 ).
Experiments in these regions require special techniques and are not considered
further here.
Infrared radiation is only absorbed by the irradiated molecule at the appropriate frequency if the corresponding vibration results in a change in molecular
dipole moment. This means that not all vibrational modes are infrared active. An
analysis of which vibrational modes in a polyatomic molecule are active is based
on group theory and the symmetry properties of the molecule. More details about
this subject may be found in monographs devoted to spectroscopy [G4].
Modern infrared spectroscopy is normally carried out in the Fourier transform
mode (FTIR). Thus, the experimental information is gathered as an interferogram
in the time domain and then Fourier transformed to obtain absorbance as a
function of wavelength. A number of interferograms may be recorded in a few
seconds. These are averaged before Fourier transformation. In this way, the
signal-to-noise ratio in the resulting spectrum is signicantly enhanced with
respect to a spectrum obtained in a traditional dispersive spectrometer
In a simple transmission experiment the liquid sample is examined in a cell
made of a suitable infrared transparent medium. These include sodium chloride,
potassium bromide, zinc selenide, cadmium telluride, and germanium. Materials
like sodium chloride should not be used to study solutions in protic solvents like
methanol and water.
In the case of strongly absorbing media, infrared experiments are more conveniently carried out using internal reectance techniques. An example of an
228
Fig. 5.11 Schematic diagrams illustrating (A) an attenuated total reection experiment at
the interface between a hemispherical window and a test solution and (B) the evanescent
wave, which decays rapidly with distance into the test medium.
experimental setup to carry out such an experiment is shown in g. 5.11. The

infrared radiation enters a hemispherical window at an angle y to the normal
drawn from the window j solution interface. When the refractive index of the
solution is less than that of the window, the radiation is totally reected at the
interface if the angle y is greater than the critical angle yc . On the basis of the wave
nature of the radiation the beam actually penetrates the optically less dense medium a very short distance. This penetrating wave is called the evanescent wave
because it decays exponentially with distance. From the theory of physical optics
the depth of penetration is the order of the wavelength.
EXAMPLE
Estimate the depth of penetration of infrared light with a frequency of

2000 cm1 into an aqueous solution given that it is transmitted through a
ZnSe hemisphere (nop 2:93) at an angle 28 greater than the critical angle.
The critical angle is given by
s
nop
yc sin
nw
op
1
5:6:5
where nsop is the refractive index of the solution and nw

op , that of the window.
Assuming that nsop 1:333,

1:333
yc sin1
278
5:6:6
2:93
The depth of penetration is given by
dp
lw
2 1=2
2psin y nsop =nw
op
2
5:6:7
where lw is the wavelength of the radiation in the window. The required

wavelength is
lw
2:93 101
1:465 105 m
2000
229
5:6:8
Thus, for an angle of incidence equal to 298

dp
1:465 105
1:39 105 m
2p0:235 0:2071=2
5:6:9
For this system the depth of penetration is 14 mm. This distance is comparable
to the wavelength of the infrared radiation which is 5 mm.
The internal reectance technique is usually called attenuated total reection
(ATR) spectroscopy. It is especially useful for studying strongly absorbing media,
for example, aqueous solutions. When the infrared radiation is absorbed in the
test medium, one obtains a spectrum similar to that from a transmission experiment. However, there are distortions in the ATR spectrum, especially in the
region of intense bands. One reason for distortion is the fact that the depth of
penetration varies with wavelength. The other eect is due to the change of the
refractive index of the solution in the region of the intense band. ATR spectra
should be corrected for these eects so that they may be compared to normal
transmission spectra.
The combination of ATR and FTIR spectroscopies gives a powerful tool for
studying the eects of solvation in solutions made from polar solvents. Some
recent data obtained in experiments using these methods are discussed later in
this chapter.
B. Raman Spectroscopy
In Raman spectroscopy, the vibrational modes of a molecule are activated but in
an entirely dierent way. In such an experiment the system is irradiated with a
strong source of light, usually from a laser which emits in the visible region. A
molecule whose major dimension is less than the wavelength of the light will
scatter the light if it is not absorbed. The electrons in the irradiated molecule
oscillate as a result of this interaction and thereby produce radiation of the same
frequency but scattered in all directions with respect to the incident beam. This
phenomenon is known as Rayleigh scattering. However, a small fraction of the
incident radiation can interact with the molecule via its vibrational degrees of
freedom. If the vibrational motion results in a change in molecular polarizability,
then the quantum of scattered light is slightly dierent in energy from that of the
stimulating radiation. When this occurs, the phenomenon is known as Raman
scattering.
EXAMPLE
Given that carbon dioxide is a linear molecule, which of its vibrational modes
are infrared active, and which, Raman active?
As a linear triatomic molecule, carbon dioxide has four degrees of vibrational motion. These are the symmetrical stretch (n1 ), the asymmetrical stretch
(n3 ), and the bending mode (n2 ). The later vibration is doubly degenerate and
can be described in two directions perpendicular to the interatomic axis.
230
Because of its symmetry with respect to the central carbon atom, carbon
dioxide has no net dipole moment. In a symmetrical stretching vibration, the
dipole moment of the molecule remains zero. Therefore the n1 mode is not
infrared active. However, the electron density along the interatomic axis is
alternately elongated and condensed. Thus, the molecular polarizability
changes with symmetrical stretching and the n1 mode is Raman active.
Exactly the opposite conclusions are reached for the asymmetric stretching
mode (n3 ). Since one CO bond contracts while the other stretches, the molecule develops a dipole moment during vibration. Thus, the n3 mode is infrared
active. At the same time the increase in polarizability due to the stretching
bond is oset by the decrease due to the contracting bond. As a result there
is no net change in polarizability and the n3 is Raman inactive. The same kind
of arguments apply to the bending modes (n2 ). Thus, n2 is infrared active and
Raman inactive.
This example demonstrates that infrared and Raman spectroscopies are
often complementary. By applying a normal mode analysis using group theory,
one may determine which vibrational modes in more complex molecules are
infrared or Raman active.
The shift in frequency of the Raman scattered radiation from the stimulating
radiation may result in quanta of lower or higher frequency. The former process is
known as Stokes scattering and the latter as anti-Stokes scattering. The quantummechanical theory of the Raman eect involves the formation of a virtual
excited quantum state at an energy level higher than the ground state by an
amount hn0 , where n0 is the frequency of the stimulating radiation. The vast
majority of molecules which reach the virtual excited state decay back to the
ground state with the emission of Rayleigh scattered light of exactly the same
frequency. However, a very few molecules in the virtual excited state decay down
to the rst excited vibrational level of the ground state, thereby giving rise to
Stokes scattered light. The energy of the quantum of scattered light will be less
than hn0 by an amount corresponding to the energy dierence between the ground
vibrational state and the rst excited vibrational state. This is hnvib . It follows that
the frequency of the Stokes line is
nst n0 nvib
5:6:10
A very few molecules in the sample are in the rst excited vibrational level when
they interact with the stimulating radiation. They reach a virtual excited state
which is correspondingly higher in energy than most excited molecules. The
majority of these fall back to the same vibrational level, thereby emitting scattered
Rayleigh light of frequency n0 . However, some also reach the ground vibrational
level and emit a quantum of anti-Stokes scattered light. Because of the total
energy change involved, the frequency of the anti-Stokes line is
naSt n0 nvib
5:6:11
Because the population of the rst excited vibrational level is much less than that
of the ground state, the intensity of the anti-Stokes lines is less than that of the
231
Stokes lines. In most Raman experiments, only the Stokes part of the spectrum is
examined.
The Raman spectrum of carbon tetrachloride is shown in g. 5.12. The stimulating radiation, which has a wavelength of 488.0 nm, was obtained from an
argon-ion laser. On the wave number scale this corresponds to 20,492 cm1 ,
which is blue light. Three lines are resolved on each side of the central peak,
which is itself due to Rayleigh scattering. These lines correspond to ve vibrational modes in the CCl4 molecule which are Raman active. The intensities of the
Stokes lines which occur on the low-frequency side are higher than those of the
anti-Stokes lines. In addition, the shift of each Stokes line from the central
Rayleigh line is exactly equal to that for the corresponding anti-Stokes line.
Thus, the features of the spectrum conform to the quantum-mechanical analysis
discussed earlier.
Raman spectroscopy is an important technique used in the study of solvent
structure and solvation. Because of its complementary nature it is often used in
connection with infrared spectroscopy. Although most of the examples discussed
in the following sections involve infrared experiments, many examples of Raman
experiments can be found in the literature.
Fig. 5.12 Raman spectrum of carbon tetrachloride obtained using an argon-ion laser
operating at 488.0 nm. Each line is labeled in terms of its shift from the central Rayleigh
line. (From reference G5, with permission.)
232
5.7 Infrared Spectroscopy of Polar Solvents

All polar solvents considered in chapter 4 are polyatomic species and therefore
have vibrational spectra. Some vibrational modes are infrared active and others
are Raman active. Especially interesting are those modes associated with the
molecular dipole because they give information about intermolecular interactions.
In this section, attention is focused on the infrared spectra of pure polar solvents.
A. Protic Solvents
The most common protic solvent is water. It is also one of the most complex from
the point of view of vibrational spectroscopy because of its highly structured
nature. Since water is a triatomic, non-linear molecule it has three vibrational
modes, which are illustrated in g. 5.13. The n1 mode is the symmetrical stretch; n2
is the bending mode; and n3 is the asymmetrical stretch. All three vibrational
modes for water are active in the infrared because they involve changes in the
dipole moment. Activity in the Raman spectrum requires that the polarizability of
the molecule changes during vibration. Analysis of this aspect of molecular properties is more dicult but it shows that all three modes are also Raman active. A
summary of the frequencies of these vibrations for H2O, and the isotopes D2O,
and HOD determined from gas phase spectra are given in table 5.7.
The infrared and Raman spectra of liquid water are broad and complicated by
several types of vibrational interactions. Three regions can be distinguished in the
infrared spectrum (see g. 5.14). The OH stretching region due to the n1 and n3
modes is observed between 2800 and 3800 cm1 . The broad character of this band
is a result of the large, uctuating distribution of hydrogen bonds. In fact, one can
distinguish two classes of OH stretching vibrations which depend on whether the
OH group is involved in a weak or strong hydrogen bond. The second region
involves the n2 bending mode observed between 1500 and 1800 cm1 . A librational or restricted rotational region is observed between 250 and 800 cm1 .
Finally, there is a translational or hydrogen bond stretching region in the far
infrared from 20 to 250 cm1 which is not shown in g. 5.14.
The diculties in interpreting the vibrational spectrum of water are obvious.
Because of the broad nature of the bands, the two stretching modes n1 and n3
overlap. In addition, overtones complicate the assignment of bands. Thus, the
rst overtone of the n2 bending mode, 2n2 , lies close in frequency to the stretching
modes n1 and n3 . For these reasons, there are advantages in studying HOD, a
molecule for which the vibrational frequencies are quite dierent. As a result,
Fig. 5.13 Normal modes of vibration of a water molecule. The arrows indicate the relative
motion of each atom.
233
Table 5.7 Vibrational

Frequencies for H2O, D2O,
and HOD from Gas Phase
Measurements
Frequency/cm1
Molecule Mode
H2O
HOD
D2O
n1
n2
n3
3657 1595 3756

2727 1402 3707
2671 1178 2788
overtone and combination bands are more widely separated. Thus, the vibrational
properties of water are conveniently studied using dilute solutions of HOD in
H2O or D2O [18].
Protic solvents always have more complex infrared spectra because of the
presence of hydrogen bonding in the liquid state. In methanol, this involves
interaction of the acidic proton on the OH group in one molecule with the oxygen
atom in an adjacent molecule (g. 5.15). The infrared spectrum shows a wide band
centered at 3346 cm1 which is due to the OH stretch. When methanol is dissolved as a dilute solute in carbon tetrachloride, this band is sharp and appears at
3644 cm1 . An OH bending mode appears at 1449 cm1 . Another broad band
due to OH out-of-plane deformation is centered at 663 cm1 . The other features
of the methanol spectrum are due to the vibrational modes of the CH3 group or
to skeletal vibrations [27].
Another example of a protic solvent is formamide in which hydrogen bonding
involves the proton in the NH2 group and the carbonyl oxygen in an adjacent
molecule (see g. 5.16). A broad band centered at 3320 cm1 is due to the asym-
Fig. 5.14 Infrared spectrum of puried water in the frequency range 4000400 cm1 .
(From reference 26, with permission.)
234
Fig. 5.15 Ball-and-stick diagram of a methanol molecule together with its infrared
spectrum as a liquid. (From reference 26, with permission.)
metric and symmetric stretching vibrations of the NH2 group. Other bands due
to this group occur at 1605, 1050, and 602 cm1 . An important band for this
molecule is the CO stretching vibration at 1685 cm1 . However, this band is
not sharp, other spectral features being very close in frequency. The band at
1392 cm1 is assigned to the symmetric OC N stretching mode and that at
1308 cm1 , to the CH deformation [27, 28]. As one would expect, the frequencies
of the bands due to vibrational modes associated with hydrogen bonding change
when formamide is a dilute solute in an aprotic solvent. Then, the NH2 band at
3320 cm1 and the CO band at 1683 cm1 shift to higher frequencies.
B. Aprotic Solvents
As one would expect the vibrational spectra of aprotic liquids are usually much
simpler than those for protic liquids. Acetonitrile is an example of an aprotic
solvent whose polar properties are due to the large dipole moment associated
1
with the C
N stretch at 2254 cm is a prominent feature
N bond. The C
235
Fig. 5.16 Ball-and-stick diagram of a formamide molecule together with its infrared
of the IR spectrum of liquid acetonitrile (g. 5.17). Other important bands include
1
1
the C C
N bending mode at 748 cm , the C C stretch at 918 cm , the CH3
1
rocking mode at 1040 cm and CH3 deformation modes at 1376 and 1445 cm1 .
The CH3 stretching modes are seen at 2945 and 3003 cm1 . The spectrum contains
two Fermi resonance bands in the region shown, an important one arising at
2293 cm1 . This band is due to a combination of the bands at 918 cm1 and
1
1375 cm1 , and interacts somewhat with the C
N band at 2254 cm because it
1
is only separated by 39 cm . Because of the electron density on the nitrogen
atom, this end of the molecule interacts strongly with cations and with molecular
solutes which are good Lewis acids. For this reason, the C
N stretching band is
often used to assess interactions of acetonitrile as a solvent with ionic and molecular solutes. More information about this aspect of the IR spectroscopy of the
acetonitrile molecule is given in the following section.
Acetone is an aprotic liquid with a very dierent structure. Its infrared spec stretch.
trum possesses a strong band at 1716 cm1 due to the asymmetric
CO
Other prominent features of its spectrum include the CH3 asymmetric and sym-
236
Fig. 5.17 Ball-and-stick diagram of an acetonitrile molecule together with its infrared
metric deformation modes at 1420 and 1364 cm1 , respectively, the C C C

asymmetric stretch at 1222 cm1 , and the CO deformation mode at 530 cm1
(g. 5.18). The CH3 stretching vibrations appear around 3000 cm1 . The CO
stretching band can be used to monitor solutesolvent interactions in acetone
solutions but it is not as sharp as the C
N stretching band in acetonitrile.
This shows that several overlapping bands occur in this region of the acetone
spectrum. The two principal components are due to the CO stretch of the
acetone monomer, and to the same stretching mode in an acetone dimer which
occurs at slightly smaller wave numbers. Because of the high dipole moment
associated with the double bond in this molecule, dimers are formed in two
dierent congurations. Dimerization in acetone has been studied by both
Raman [29] and infrared spectroscopy [30].
Dimethylformamide (DMF) has a much more complex structure than either
acetonitrile or acetone, but it is interesting to compare its infrared spectrum with
that of formamide. By replacing the two acidic protons on the amino group by
methyl groups, one obtains an aprotic liquid. As can be seen by comparing gs
5.16 and 5.19, the infrared spectrum of DMF is much simpler than that of for-
237
Fig. 5.18 Ball-and-stick diagram of acetone together with its infrared spectrum as a liquid.
mamide. However, the features of the spectrum due to the CO group and the
nitrogen atom are similar. The strongest band in the spectrum occurs at
1675 cm1 and is due to the CO stretching mode. Another prominent feature
N stretching mode. The CH3 rocking mode
at 1389 cm1 is assigned to the OC
1

is seen at 1095 cm and the O C N deformation at 659 cm1 .
The infrared spectra of other aprotic solvents are also characterized by a strong
band or bands associated with the polar group in the molecule. For example,
dimethylsulfoxide has a strong absorption at 1057 cm1 due to the SO stretching mode. However, absorption in this region is complicated by several overlapping bands, one of which is due to DMSO dimers [31]. Propylene carbonate,
238
Fig. 5.19 Ball-and-stick diagram of dimethylformamide together with its infrared spectrum in the liquid state. (From reference 26, with permission.)
which has a CO bond attached to a ve-member heterocyclic ring, has a

strong CO band at 1797 cm1 . In the case of nitrobenzene and nitromethane,
the polar group is triatomic. In nitromethane, the strongest band in the infrared
spectrum is at 1563 cm1 due to the asymmetric stretching mode of the NO2
group. The symmetric stretch mode at 1378 cm1 is considerably weaker. The
NO2 group has other vibrational modes at lower frequencies.
239
In summary, the infrared spectra of aprotic solvents are much simpler than
those of protic solvents because of the absence of complications due to hydrogen
bonding. On the other hand, the aprotic liquids are characterized by very high
molecular dipole moments. As a result, dimers can be formed in which some of
the molecules line up with their dipole vectors in an antiparallel fashion [30]. This
can lead to the spectra being complicated by additional spectral features due to
solvent aggregates. Nevertheless, the prominent band due to the polar group in
these solvents can be used to probe solutesolvent interactions in non-aqueous
media. This subject is discussed in more detail in the next section.
5.8 Infrared Spectroscopy of Non-Electrolyte Solutions

When a polar solute such as acetonitrile is dissolved in another organic solvent,
the shift in the frequency of the polar group may be used to monitor solute
solvent interactions. This type of study is especially interesting because it demonstrates clearly the importance of short-range chemical interactions in determining
the properties of the solution.
A number of studies of the IR spectra of polar molecules dissolved in a variety
of solvents both polar and non-polar have been carried out [32]. In the case of
acetonitrile, the C
N stretching frequency (n2 band) is shifted to higher values
in solvents which are stronger Lewis acids. Many of these solvents are protic and
interact with the electronegative C
N group via hydrogen bonding. The electron density in this bond shifts toward the Lewis acid. Since the electron density is
associated with an antibonding orbital, the net eect is to make the C
N bond
stronger, and a blue shift, that is, a shift to higher frequencies is observed.
A plot of the shift of the n2 band for both acetonitrile and deuterated acetonitrile
(CD3CN) as a function of the solvents acceptor number AN is shown in g. 5.20
for 12 solvents which are more acidic than acetonitrile. An excellent correlation
between n2 and the acceptor number AN is found except for formamide (F) and
N-methyl formamide (NMF). In the latter cases, the solvents are both strong Lewis
acids and strong Lewis bases. Corresponding results are obtained when acetonitrile
is dissolved in solvents which are stronger Lewis bases. In this case, there is a good
correlation between n2 and the solvents donor number DN, with the solvent
shifts being negative (g. 5.21). These experiments point out the importance of
Lewis acidbase properties in determining the nature and strength of short-range
interactions in solution. The strong Lewis bases interact with the CH3 group, which
is at the positive end of the molecular dipole. This interaction aects the C
N
group, so that more electron density is associated with this part of the molecule. As
a result, a red shift occurs, that is, negative values of n2 are observed.
Solvent-induced frequency shifts have also been studied for dimethylsulfoxide
(DMSO), which is a strong Lewis base. Most solvents behave as Lewis acids and
interact with DMSO via the electron density on the SO group. This interaction leads to a red shift for this band which is as large as 56 cm1 in the presence of
acetic acid [34]. The magnitude of the shift correlates well with the solvents
acceptor number giving the relationship
nSO 1080 1:375AN
5:8:1
240
Fig. 5.20 Plot of n2 for CH3CN () and CD3CN () in more acidic solvents against the
solvents acceptor number AN. The best straight lines were estimated without the data for
F and NMF. The data for CD3CN have been shifted vertically by 20 cm1 for clarity.
Fig. 5.21 Plot of n2 for CH3CN () and CD3CN () in more basic solvents against the
solvents donor number DN. The best straight lines were estimated without the data for F
and NMF. The data for CD3CN have been shifted vertically by 10 cm1 for clarity. (From
241
For this system, removal of electron density from the SO bond results in its
being weakened as one would expect without knowing the detailed molecular
orbital structure.
Acetone provides another interesting system for studying intermolecular interactions. The basicity of this solvent, whose DN is 17.0, is close to that of water.
On the other hand, it is a very weak Lewis acid with an AN equal to 12.5. Thus,
one expects the CO stretching frequency n3 to shift with change in the acidity of
the solvent in which acetone is dissolved. A good correlation is found between the
frequency of this band in 22 solvents and the solvents acceptor number AN, with
the frequency shifting in the red direction as AN increases. The equation describing the correlation is
n3 1718 0:286AN
5:8:2
with a correlation coecient equal to 0.876. Interaction of the acetone molecule
with a Lewis acid results in a weakening of the CO bond due to withdrawal of
electron density. The correlation is weakened by the eects of acetoneacetone
interactions which are stronger in solvents which are weak Lewis acids [29].
The eects of intermolecular interactions are seen much more clearly for the n17
mode which is the skeletal asymmetric stretch. Values of the frequency of this
vibration are plotted against solvent AN in g. 5.22. The equation describing the
linear correlation is
5:8:3
n17 1216 0:339AN
with a correlation coecient of 0.984 [30]. This mode is relatively unaected by
acetone dimerization, and shifts in the blue direction with increase in solvent
acidity.
Other systems in which solvent-induced frequency shifts have been studied
include benzonitrile, cyclopentanone, and tetramethylurea [32]. These studies all
Fig. 5.22 Plot of the frequency of the asymmetric stretching band of the molecular skeleton (n17 ) of acetone as a dilute solute in various solvents against the acceptor number of the
solvent, AN.
242
demonstrate the importance of specic solvation eects in determining intermolecular interactions in these systems. Vibrational spectroscopy provides a valuable
tool for probing these interactions in liquid solutions.
5.9 Infrared Spectroscopy of Electrolyte Solutions

Ions are stabilized in electrolyte solutions by the dipolar character of the solvent
molecules. Thus, one expects the iondipole interaction to result in a shift in
electron density along the direction of the dipole vector. This, in turn, should
result in a change in frequency of the vibrational modes associated with the
molecular dipole. As has been seen earlier, there is a strong band in the spectrum
of aprotic molecules which is associated with the electronegative group. Changes
in the frequency of this group allow one to monitor interactions of the solvent
molecules with cations. Interactions with anions are more dicult to see in the
vibrational spectrum of these systems but can be assessed under some circumstances. In the case of protic solvents, hydrogen bonding is important in the
solvation of anions. Thus, features of the IR spectrum associated with this aspect
of solvent structure are aected by the anion in the system.
Another way of assessing cationsolvent interactions is to examine the vibrational characteristics of the bond formed between the cation and the negative end
of the solvent molecules dipole. This bond is mainly electrostatic in character and
therefore is much weaker than intramolecular covalent bonds. As a result the
infrared band corresponding to vibration of the metalsolvent bond is normally
observed in the far infrared at frequencies less than 500 cm1 . Some results
obtained for the alkali metal cations in four aprotic solvents are summarized in
table 5.8.
It is clear that the frequency of the metalsolvent band decreases with increase
in the size of the alkali metal, that is, with decrease in the eld due to the charge on
the ion. When ion pairing is absent, the value of the frequency is independent of the
nature of the anion. This is denitely the case in DMSO, which strongly solvates
cations because of its high donor number. In addition, the relatively high permittivity of this solvent reduces the inuence of long-range ionion interactions.
Because instrumentation capable of working in the far infrared is not commonly available, studies of interactions between polar solvent molecules and ions
have more commonly involved intramolecular vibrational modes in the solvent
Table 5.8 Frequency of CationSolvent

Band, n/cm1 [G1]
Solvent Cation
AC
DMSO
NMP
PC
Li
Na
K
Rb
Cs
425
195
140

429
199
153
124
109
398
204
140
106
397
186
144
115
112
243
molecule. In the case of cations, vibrations associated with the polar group at the
negative end of the molecular dipole can be studied. Thus, cationsolvent interactions can be assessed by following changes in the C
N stretching frequency
in acetonitrile, in the SO stretching frequency in dimethylsulfoxide, or in the
CO stretching frequency in acetone.
The eect of dissolving LiClO4 in acetonitrile (AcN) is illustrated in g. 5.23 on
the basis of a dierence spectrum. The lower spectrum in this gure shows the
absorbance features of pure AcN in the region from 600 to 3400 cm1 . The upper
spectrum was obtained by subtracting the spectrum of pure AcN from the spectrum of an AcN solution containing 0.7 M LiClO4. As a result the dierence
spectrum has both positive-going features corresponding to new bands in the
electrolyte solution, and negative-going features corresponding to bands whose
intensity decreased in the electrolyte solution. The most important changes are
1
those associated with the C
N stretching mode at 2253 cm . In the dierence
spectrum a negative-going band is observed at this frequency, indicating that the
concentration of free acetonitrile has decreased in the electrolyte solution. These
AcN molecules are now coordinated to Li+ ions through the negative charge
density on the N atom and have their C
N stretching frequency shifted in
the blue direction to 2280 cm1 . Negative-going bands are also observed at
918 cm1 (C C stretching mode) and 748 cm1 (C C
N bending overtone).
Spectral properties associated with the CH3 deformation modes at 1375 and
1453 cm1 are also aected by the addition of electrolyte. Perchlorate ion has
its own vibrational spectral features which appear at 625 and 1100 cm1 .
By following the concentration dependence of the negative-going band at
2253 cm1 , one can determine the average number of AcN molecules which are
coordinated to the metal ion in solution. Experiments show that the integrated
intensity of this band is linear in the electrolyte concentration over reasonable
concentration ranges. Interpretation of the negative slope of these plots requires
that the extinction coecient of free acetonitrile molecules be determined in a
Fig. 5.23 Dierence spectrum for an

acetonitrile solution containing
0.7 M LiClO4 obtained by subtracting the
spectrum of pure acetonitrile from that of
the solution, together with the spectrum of
pure acetonitrile.
244
suitable inert medium. This was done by observing the intensity of the C
N
stretching mode for AcN dissolved in carbon tetrachloride as solvent [35, 36]. In
this way, the average number of solvent molecules coordinated to a single cation
was determined. Results for several monovalent and divalent cations in acetonitrile are summarized in table 5.9.
It is clear from the results in table 5.9 that the average coordination number nc
of solvent molecules to a cation varies signicantly with cation nature. This
variation is attributed to a corresponding variation in the extent of ion pairing.
If the coordination number is less than the maximum value for a given cation,
then the value of nc can be used to estimate the average number of anions associated with a cation, and thus, the average charge on the cationic species. The
maximum number of solvent molecules which can coordinate to a cation is
usually considered to be six, but for small cations it can be four.
The observed spectral shift for the C
N stretching mode n2 varies considerably with cation nature. It is always in the blue direction, indicating that the C
N
bond is stronger in the presence of the positive electron-withdrawing eld of the
cation. This observation suggests that the electron density withdrawn from the
C
N system involves antibonding orbitals. The variation in n2 with cation is
usually attributed to variation in the cations electrical eld. The latter quantity is
usually estimated from the ratio of the cationic charge zi e0 to its radius ri .
However, a reasonable correlation between n2 and zi e0 =r is not obtained without considering the eects of ion pairing. On the other hand, if the eld is estimated using the eective average cationic charge, zi0 , estimated from the average
coordination number nc an acceptable correlation is obtained (see g. 5.24).
Cations also aect other solvent bands for AcN including the combination
band at 2293 cm1 (n3 n4 ), the C C stretching mode at 918 cm1 (n4 ), and
1
the C C
N deformation mode at 746 cm (2n8 ). Changes in the intensity of
these bands with electrolyte concentration may also be used to assess ion pairing
in acetonitrile solutions [35, 36].
Anions aect the features of the AcN spectrum to a much smaller extent [37].
Any change in the spectrum is usually associated with CH3 symmetrical stretch-
Table 5.9 Average Coordination Number for Acetonitrile

to the Metal Cation and Frequency Shift of the n2 Band
( C
N stretch) from IR Spectra of Metal Perchlorate
Solutions in Acetonitrile [36]
Cation
Average Coordination Number nc
Frequency Shift n2
Li+
Na+
Mg2+
Ca2+
Sr2+
Ba2+
Zn2+
Cd2+
Pb2+
3.7
3.8
3.4
4.9
5.9
4.8
5.7
5.4
5.0
23
13
42
24
42
12
41
32
11
245
Fig. 5.24 Plot of the shift of the C

N stretching frequency of acetonitrile, n2 observed in
solutions containing various metal perchlorates against the eective eld of the metal
cation, z0i =ri [36].
ing mode at 2944 cm1 (n1 ) and the corresponding antisymmetrical mode at
3002 cm1 (n5 ). Polyatomic anions such as ClO4 and NO3 have very little eect
on this region of the spectrum. On the other hand, monoatomic ions such as the
halides, and linear polyatomic ions such as NCS aect this region of the spectrum strongly. This suggests that the anion is interacting with the methyl group at
the positive end of the molecular dipole, and that the extent of interaction
depends on the electrical eld associated with the anion [37].
Electrolyte eects on the vibrational spectra of polar solvents have been studied in a number of other solvents including acetone, pyridine, dimethylsulfoxide,
dimethylformamide, tetramethylsulfone, and nitromethane [37]. These studies
provide valuable information on the nature of solvation of ions in these systems.
In addition, they can be used to assess ion pairing and its variation with electrolyte nature and concentration.
5.10 UltravioletVisible Spectroscopy and

Solvatochromic Effects
Ultraviolet-visible spectroscopy usually involves electronic transitions in which
valence electrons in molecules and polyatomic molecules move to higher unlled
orbitals. This type of spectroscopy is well known in organic chemistry and often
involves the p electrons in groups such as CC, CO, NN, and NO. The
electronic transitions taking place are p ! p transitions, and in molecules with
heteroatoms, n ! p transitions, that is, from nonbonding orbitals for electrons
on atoms such as oxygen to an antibonding p orbital. Electronic transitions also
occur between s bonding orbitals and s antibonding orbitals but the associated
246
energy is usually quite high, so that the absorption is observed in the vacuum
ultraviolet at a wavelength less than 150 nm.
The discussion in this section is mainly concerned with species containing a
transition metal ion and the involvement of its d electrons in electronic transitions.
The simplest description of electronic energy levels in a transition metal complex
is based on crystal eld theory, but a more detailed description requires the
application of molecular orbital theory. When the metal ion is surrounded by
six ligands it is in an octahedral eld, and the d orbitals are no longer energetically
degenerate. The three orbitals with electron density directed away from the axes
of the reference coordinate system, namely, dxy , dyz , and dxz have lower energy
than those with electron density directed along these axes, namely, dx2 y2 and dz2 .
The energy dierence between the three lower levels (t2g orbitals) and the two
higher ones (eg orbitals) is called the crystal eld splitting energy and is designated
as 0 (see g. 5.25). The value of 0 depends both on the nature of the metal ion,
on the number of d electrons, and on the nature of the ligands. On the basis of a
large number of spectroscopic studies the splitting energy has been determined for
a wide variety of complexes. The variation in 0 with metal ion nature is illustrated with ions formed from the elements in the rst transition series of the
periodic table in table 5.10 for the case that the ligands are six water molecules.
The energy associated with the electronic transition varies from a minimum of
9.3 meV for Mn2+ to a maximum of 26.0 meV for Mn3+. Considerable variation
in 0 is also found with the nature of the ligand. In this case 0 increases in the

series I < Br < SCN < Cl < N
3 < OH < H2 O < NCS < NH3 < NO2

< CN < CO. The sequence is known as the spectrochemical series.
Another common coordination number observed in transition metal complexes
is four. When the ligands are situated as far away from each other as possible, one
obtains a tetrahedral conguration. In this case the ligands interact repulsively
with the dxy , dyz and dxz orbitals, which have their electron density directed
Fig. 5.25 Crystal eld

splitting of the d orbitals in
a transition metal ion in an
octahedral eld and a
tetrahedral eld due to
surrounding ligands (Lewis
bases).
247
Table 5.10 Crystal Field Data for the Complex Ions M(H2O)n
6 from the First
Transition Series
Ion
d Electrons
Ti3+
Ground State for Free Ion
Octahedral Field Conguration
0 =cm1
t12g
20,100
3+
t22g
19,950
V2+
t32g
12,100
Cr3+
t32g
17,400
Cr2+
Mn3+
4
4
D
5
D
t32g e1g
t32g e1g
14,000
21,000
Mn2+
t32g e2g
7,500
3+
t32g e2g
14,000
Fe2+
Fe
t42g e2g
9,350
3+
Co
t62g
20,760
Co2+
9,200
t52g e2g
2+
Ni
t62g e2g
8,500
Cu2+
t62g e3g
12,000
between the axes of the x, y, z coordinate system. As a result, the d orbitals are
split into two degenerate sets with the dx2 y2 and dz2 orbitals (e orbitals) having
lower energy than the dxy , dyz , and dxz orbitals (t2 orbitals). The energy dierence
is designated as t (see g. 5.25). Another conguration with four ligands is the
square planar geometry. As the name implies, the ligands lie in a plane centered
on the transition metal ion. This system is favored by metal ions with a d 8 conguration and has a more complicated distribution of d orbital energy levels than
either the octahedral or tetrahedral systems. Other geometries observed for transition metal complexes include a linear one with two ligands, and a trigonal one
with three ligands.
One normally thinks of the electronic spectroscopy of transition metal complexes as involving d ! d transitions. In fact, many of the spectra observed for
these systems do involve the promotion of d electrons from a lower lying level to
one which is higher because of the crystal eld splitting [38]. However, there are
also charge transfer transitions from an orbital associated with a ligand to an
empty one associated with the metal (LMCT) or from a d orbital on the metal to
an empty orbital associated with the ligand (MLCT). These are more frequent
when there are p bonds in the ligands of the transition metal complex.
Solvatochromic eects occur then the color of a solute depends on the nature
of the solvent in which it is dissolved. Two such systems were discussed in section
4.9. The rst is the betaine dye, 4-(2,4,6-triphenylpyridinium)-2,6-diphenylphenoxide, which is used to dene the DimrothReichardt parameter ET . This molecule undergoes a p ! p transition in the visible region whose wavelength is very
solvent dependent. In the strongly acidic solvent water it is equal to 453 nm which
corresponds to an energy gap of 264 kJ mol1 . In the very weakly acidic solvent
diphenylether, the wavelength for the same transition is 810 nm. This corresponds
248
to an energy gap and value of ET equal to 148 kJ mol1 . The change in the color of
the dye with solvent nature is clearly related to the charge transfer which accompanies the p ! p transition as illustrated in g. 4.12. Since the negatively charged
O atom in the ground state is stabilized to an extent which depends on solvent
acidity, the ET parameter gives an excellent way of estimating this molecular
property of the solvent.
Solvatochromic eects related to solvent acidity have been found for several
transition metal complexes. Burgess [39] studied the visible spectra of bis-(2,2 0 bipyridyl)biscyanoiron(II), bis-(1,10-phenanthroline)biscyanoiron(II), and related
iron complexes in a wide range of solvents, both polar and non-polar. In most
cases the wavelength of maximum absorption depends on solvent acidity and is
linear in the solvents acceptor number. The structure of the phenanthroline
complex is shown in g. 5.26. It is clear that the solvent interacts with the negative
charge density on the two cyano groups so that the complex probes the ability of
the solvent to act as a Lewis acid. Iron(II) is a d 6 system and the ligand eld
splitting of the d orbitals is large for these systems. Thus, all six d electrons are in
the low-lying t2g orbitals. Burgess [38] reported the frequency for an MLCT band,
that is, for charge transfer from a t2g orbital on the metal to a p orbital on the
bipyridyl or phenanthroline ligand. As the acidity of the solvent increases and its
interaction with the cyano ligands becomes stronger, the splitting between the t2g
and p orbitals increases. At the same time the frequency observed at maximum
absorption shifts in the blue direction. For example, the phenanthroline complex
absorbs at 634 nm in hexamethylphosphoramide, an extremely weak Lewis acid,
so that the solution is blue in color. In ethanol, a moderate Lewis acid, maximum
absorption occurs at 557, and in triuoroacetic acid, a very strong Lewis acid, at
395 nm. The relationship between the frequency (cm1 ) at maximum adsorption
and the Gutmann acceptor number AN is
nmax 107:18AN 14,340
5:10:1
on the basis of data in 12 polar solvents r 0:980 [40]. A similar study with bis(2,20 bipyridyl)biscyanoruthenium(II) and related complexes has also been carried
Fig. 5.26 Chemical structure of bis-(1,10phenanthroline)biscyanoiron(II). Solvent

molecules can interact with the electron
density on the cyano groups.
249
out [41]. These results show the same sensitivity to solvent acidity seen in the
earlier work with iron complexes.
Another group of transition metal complexes behave as solvatochromic indicators of solvent basicity. The best known of these is Cu(II) N; N; N 0 ; N 0 -tetramethylethylene diamine acetoacetonate or Cu(II)(tmen)(acac) (see g. 5.27). In
this complex the two chelates are located in a plane so that the resulting geometry
is approximately square planar. The transition of maximum frequency corresponds to a d electron associated with the o-planar orbitals dxz and dyz being
promoted to the dx2 y2 orbital which has its electron density in the plane containing the ligands. As the strength of interaction of the solvent molecules in the axial
positions with the metal ion increases, the eective geometry of the complex
changes from square plane through tetragonal to octahedral. At the same time
the splitting between the lowest occupied and the highest empty d orbital
decreases so that the frequency associated with the d ! d transition shifts in
the red direction. Values of the wavelength associated with the transition for
the polar solvents considered in chapter 4 are given in table 4.10 where they are
designated as Bsc , the solvatochromic basicity. Considerable variation in wavelength is observed, from a minimum of 522 nm in nitrobenzene, a weakly basic
solvent to 633 nm in hexamethylphosphoramide which is a strong Lewis base. For
this system, the wavelength correlates strongly with solvent basicity as assessed by
the Gutmann donor number (see equation (4.9.2)).
Another group of complexes exhibiting solvatochromic properties are Ru(II)
and Ru(III) ammine complexes containing one or more other ligands with p
bonds [40, 41]. A simple example is the Ru(II) pentammine pyridine cation,
which has an MLCT band at 407 nm in acetonitrile. This transition corresponds
to a d electron in a t2g orbital on the metal being transferred to a p orbital on
pyridine. When the solvent is changed to a more basic one such as dimethylsulfoxide, the wavelength increases. As a result there is a correlation between the
wavelength for the MLCT band and the Gutmann donor number DN, namely,
lSC 390:9 1:698DN
5:10:2
The correlation is very strong but the experimental study was limited to aprotic
solvents.
One can argue that correlations such as equations (4.9.2) and (5.10.2) are not
physically reasonable in that they are based on the wavelength of the electronic
transition related to the solvatochromic eect, not its frequency. The frequency is
Fig. 5.27 Chemical structure of Cu(II)N,N,N0 ,N0 -tetramethylethylenediamine

acetoacetonate. Solvent molecules can
interact as Lewis bases with the axial sites
in the complex as shown.
250
directly proportional to the associated energy dierence, which in turn reects the
strength of the interaction of the solvent as a Lewis base with the transition metal
complex. Thus, the frequency is the more relevant parameter to consider in assessing the relation between the solvatochromic eect and the Gutmann donor
number. In fact, equally strong correlations are found between the frequency
and the DN for both of these systems. The fact that linear correlations are
found with both frequency and wavelength, which is related to the reciprocal of
the frequency, reects the fact that the net change in either parameter with respect
to its average value is small. In this case the reciprocal of one quantity is approximately linear in that quantity over a limited range on the basis of a Taylors series
expansion. Equations (4.9.2) and (5.10.2) are used here because the parameters
BSC and lSC can then be used as empirical measures of solvent basicity which
increase with increase in basicity.
The examples of electronic transitions in transition metal complexes discussed
here have emphasized solvatochromic eects and often involve intramolecular
electron transfer. The role of the solvent in electron transfer is an important
aspect of these processes and is discussed in more detail in chapter 7. Many
electronic transitions do not display the type of solvent eects considered here
but depend instead on bulk solvent properties. The systems discussed above
interact in a specic way with the solvent usually as a result of a signicant change
in the dipole moment of the molecule or the complex ion as a result of the
electronic transition. There are many other interesting aspects of this area of
spectroscopy which have been discussed in more detailed treatments [38, 43].

The discussion in this chapter has only considered representative examples of the
application of spectroscopy to the elucidation of the structure of liquids and
solutions. Raman spectroscopy is complementary to infrared spectroscopy and
has been used just as frequently to study solutesolvent, solutesolute, and solventsolvent interactions in solution. Electron spin resonance spectroscopy
involves the spins of unpaired electrons in molecules and ions and their interaction with a magnetic eld. It has been used to study hydrogen bonding and
hydrophobic interactions in aqueous solutions. Mossbauer spectroscopy involves
the nuclear transitions that result from the absorption of g-rays by certain nuclei.
It is used to study the coordination spheres of the corresponding ions in solution.
Dielectric spectroscopy is used to determine the dielectric properties of liquids and
solutions at microwave frequencies. As shown in chapter 4, this technique gives
information about molecular reorientation in an alternating electrical eld.
Ultrasonic absorption spectrometry is a study of the density changes which
occur when sound energy is passed through a liquid or solution and provides
information about structural relaxation times. Electron diraction studies are
related to neutron and X-ray diraction. Electrons are diracted by the local
electrical potential in the sample, which depends on the spatial conguration of
the nuclei and density distribution of the associated electrons. It is complementary
to X-ray diraction because it is more sensitive to the distribution of light atoms
251
in the liquid sample. More information about all these techniques, including those
discussed in more detail in this chapter, are available in the reviews in reference
[G1].
The important contribution of spectroscopic studies is structural information
about liquids and solutions at the atomic and molecular level. Most of the information discussed in earlier chapters is thermodynamic in nature and does not
provide a microscopic description of the system. When the spectroscopic details of
intermolecular interactions are available, one is able to understand thermodynamic results in much greater detail. Spectroscopic experiments were developed
mainly in the latter half of the twentieth century. As a result of the detailed
information they provide, understanding of the structure of condensed phases
is now at the same level as our understanding of the much simpler gas phase.
General References
G1. Dogonadze, R. R.; Kalman, E.; Kornyshev, A. A.; Ulstrup, J. The Chemical Physics
of Solvation, Part B: Spectroscopy of Solvation; Elsevier: Amsterdam, 1986.
G2. Marcus, Y. Ion Solvation; Wiley-Interscience: New York, 1986.
G3. Burgess, J. Metal Ions in Solution; Ellis Horwood: Chichester, 1978.
G4. Drago, R. S. Physical Methods in Chemistry, 2nd ed.; Saunders, Fort Worth, TX,
1992.
G5. Nakamoto, K. Infrared and Raman Spectra of Inorganic and Coordination Compounds,
5th ed.; Wiley Interscience, New York, 1997.
References
1. Enderby, J. E.; Neilson, G. W. Rep. Prog. Phys. 1981, 44, 593.
2. Blum, L.; Narten, A. H. Adv. Chem. Phys. 1976, 34, 203.
3. Palinkas, G.; Radnai, T.; Hajdu, F. Z. Naturforsch. 1979, 35a, 107.
4. Magini, M.; Licheri, G.; Paschina, G.; Piccaluga, G.; Pinna, G. X-Ray Diraction of
Ions in Aqueous Solutions: Hydration and Complex Formation; CRC Press: Boca
Raton, FL, 1988.
5. Ohtaki, H.; Radnai, T. Chem. Rev. 1993, 93, 1157.
6. Radnai, T.; Kalman, E.; Pollmer, K. Z. Naturforsch. 1984, 39a, 464.
7. Tamura, Y.; Spohr, E.; Heinziger, K.; Palinkas, G.; Bako, I. Ber. Bunsen-Ges Phys.
Chem. 1992, 96, 147.
8. Enderby, J. E. Chem. Soc. Rev. 1995 24, 159.
9. Neilson, G. W.; Enderby, J. E. J. Phys. Chem. 1996, 100, 1317.
10. Tromp, R. H.; Neilson, G. W.; Soper, A. K. J. Chem. Phys. 1992, 96, 8460.
11. Powell, D. H.; Neilson, G. W.; Enderby, J. E. J. Phys. Condens. Mater 1993, 5, 5723.
12. Hindman, J. C. J. Chem. Phys. 1962, 36, 1000.
13. Davies, J.; Ormonroyd, S.; Symons, M. C. R. Trans. Faraday Soc. 1971, 67, 3465.
14. Deverell, C.; Richards, R. E. Mol. Phys. 1996, 10, 551.
15. Templeman, G. J.; van Geet, A. L. J. Am. Chem. Soc. 1972, 94, 5578.
16. Bloor, E. G.; Kidd, R. G. Can. J. Chem. 1972, 50, 3926.
17. Deverell, C.; Richards, R. E.; Mol. Phys. 1969, 16, 421.
18. Desnoyers, J. E.; Jolicoeur, C. In Comprehensive Treatise of Electrochemistry;
Conway, B. E., Bockris, J. OM., Yeager, E., eds.; Plenum: New York, 1983; Vol. 5.
19. Covington, A. K.; Newman, K. E. In Modern Aspects of Electrochemistry; Bockris, J.
OM., Convway, B. E., eds.; Plenum: New York, 1977; Vol. 12, Chapter 2.
20. Coetzee, J. F.; Sharpe, W. R. J. Solution Chem. 1972, 1, 77.
21. Popov, A. I. Pure Appl. Chem. 1975, 41, 275.
252
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
42.
43.
Gutmann, E.; Wychera, E. Inorg. Nucl. Chem. Lett. 1966, 2, 257.

Stengle, T. R.; Pan, Y.-C. E.; Langford, C. H. J. Am. Chem. Soc. 1972, 94, 9037.
Spaziante, P. M.; Gutmann, V. Inorg. Chim. Acta 1971, 5, 273.
Mayer, U.; Gutmann, V.; Gerger, W. Monatsh. Chem. 1975, 106, 1235.
Hansen, D. L., Ed. Sprouse Collection of Infrared Spectra, Book IV, Common Solvents:
Condensed Phase, Vapor Phase and Mass Spectra; Sprouse Scientic Systems: Paoli,
PA.
Lin-Vien, D.; Colthrup, N. B.; Fateley, W. G.; Grasselli, J. G. The Handbook of
Infrared and Raman Characterstici Frequencies of Organic Molecules; Academic
Press: New York, 1991.
Smith, C. H.; Thompson, R. H. J. Mol. Spectrosc. 1972, 42, 227.
Perelygin, I. S. Fluid Phase Equilib. 1996, 116, 395.
Cha, D. K.; Kloss, A. A.; Tikanen, A. C.; Fawcett, W. R. Phys. Chem. Chem. Phys.
1999, 1, 4785.
Fawcett, W. R.; Koss, A. A. J. Chem. Soc., Faraday Trans. 1996, 92, 3333.
Fawcett, W. R. In Quantitative Treatments of Solute/Solvent Interactions; Politzer, P.,
Murray, J. S., eds.; Elsevier: Amsterdam, 1994; Chapter 6.
Fawcett, W. R.; Liu, G.; Kessler, T. E. J. Phys. Chem. 1993, 97, 9293.
Fawcett, W. R.; Kloss, A. A. J. Phys. Chem. 1996, 100, 2019.
Fawcett, W. R.; Liu, G. J. Phys. Chem. 1992, 96, 4231.
Fawcett, W. R.; Liu, G.; Kloss, A. A. J. Chem. Soc., Faraday Trans. 1994, 90, 2697.
Perelygin, I. S. In Ionic Solvation; Krestov, G. A., Ed.; Nauka: Moscow, 1987; Chapter
3.
Lever, A. B. P. Inorganic Electronic Spectroscopy, 2nd ed.; Elsevier: Amsterdam, 1984.
Burgess, J. Spectrochim. Acta 26A, 1970, 1369, 1957.
Soukup, R. W.; Schmid, R. J. Chem. Educ. 1985, 62, 459.
Timpson, C. J.; Bignozzi, C. A.; Sulliman, B. P.; Kober, E. M.; Meyer, T. J. J. Phys.
Chem. 1996, 100, 2915.
Curtis, J. C.; Sullivan, B. P.; Meyer, T. J. Inorg. Chem. 1983, 22, 224.
Murrell, J. N. The Theory of Electronic Spectra of Organic Compounds; Wiley: New
York, 1963.
Problems
1. The overall pair correlation function for acetonitrile is obtained by X-ray and
neutron diraction experiments. List the component pair correlation functions which make up Gr in each experiment. Devise a strategy for separating
these based on isotopic substitution.
2. Write down the pair correlation functions which are involved in a neutron
diraction study of aqueous LiNO3 solutions. Devise a strategy for separating these using isotopes.
3. Estimate the fraction of 7 Li atoms in each of the four quantum states when
this atom is in a magnetic eld of 5 T at 258C.
4. The equation relating the chemical shift for
23
Na to the DN is
d23 Na 16:78 0:510DN

Estimate the chemical shift for 23 Na in nitrobenzene, benzonitrile, and butyronitrile using the DN data in table 4.10.
253
5. Examine the importance of isotope eects on the IR spectrum of acetonitrile.

Estimate the frequencies of the symmetric and asymmetric CH3 stretching
modes in deuterated acetonitrile given that they are 2945 and 3003 cm1 in
13 15
normal acetonitrile. What is the frequency of the C
N stretch in CH3 C N?
6. The following frequencies have been observed for the C
N stretch in various
nitriles:
2253 cm1
2247 cm1
2228 cm1
Acetonitrile
Propionitrile
Benzonitrile
Can these results be accounted for quantitatively using the simple harmonic
oscillator model?
7. In the NMR spectroscopy of liquids, the spin lattice relaxation time falls in
the range 102 to 102 s. This inuences the width of the NMR adsorption line.
Estimate the corresponding minimum width of the line using the relationship
n
1
2pT1
What is the source of this relationship?

8. The correlation between the wavelength of the solvatochromic absorption by
the Cu complex shown in g. 5.27 is
lSC 525:2 2:89DN
where DN is the Gutmann donor number. Re-express this equation in terms
of the energy hn associated with the transition. Show that the energy is also
approximately a linear function of DN for the energy range in question.
Non-Equilibrium Phenomena in
Liquids and Solutions
Lars Onsager was born in 1903 in Oslo,

Norway, where he grew up and attended
school. In 1920 he entered the Norwegian
Institute of Technology in Trondheim to study
chemical engineering, graduating in 1925. After
spending several months with Debye in Zurich,
he became interested in the theory of
electrolytes and developed the model for
electrolyte conductivity which he later
published with Debye. In 1928 he moved to the
United States and held positions at Johns
Lars Onsager
Hopkins and Brown Universities. He moved to
Yale University in 1933, where he was
appointed Gibbs Professor of Theoretical Chemistry in 1945, and remained
there until he retired. Onsager is best known for his work in the area of
irreversible thermodynamics where he developed the reciprocal relations which
are now known after him. He submitted this work as a doctoral thesis to his
alma mater in Trondheim but it was not accepted. Clearly his work in this area
of physical chemistry was very advanced for its time. Onsager also worked in
other areas of physical chemistry including colloids, dielectrics, orderdisorder
transitions, metals and superuids, hydrodynamics, and fractionation theory.
He received the Nobel Prize in chemistry in 1968 for his discovery of the
reciprocal relations. His work was also recognized by many other awards both
in the United States and internationally. He died in 1976.
6.1 Non-Equilibrium Processes Are Usually Complex

The topics considered up to this point have involved liquids and solutions at
equilibrium. Attention is now turned to systems which are not at equilibrium,
and the processes which occur spontaneously in such systems. The physical phenomena involved can be quite complex, so that the task faced in early experiments
was to separate the various processes and understand the physical properties of
the system which govern them.
Consider what happens when a beaker of pure isothermal water is placed on a
hot plate. The water near the bottom rises in temperature and a temperature
gradient is set up. As a result heat ows from the bottom of the beaker, producing
254
NON-EQUILIBRIUM PHENOMENA IN LIQUIDS AND SOLUTIONS
255
a gradual increase in temperature in the water at a given height above the bottom.
In addition, the temperature varies with distance, being highest at the bottom and
lowest at the top. Eventually, the temperature of the water in the beaker is uniform and equal to that of the hot plate, assuming that the water does not boil.
However, the ow of heat is not the only process resulting from the heat source.
The density of the hot water is less than that of the cold water, so that a convection process is set up in order to achieve uniform density. Convection results in
cold water moving down into the hot region so that the ow of water molecules
assists the ow of heat. The changes which occur in this system cannot be understood without considering both processes.
A system undergoing an irreversible change involving an electrolyte is electrolysis in an electrochemical cell. When current ows between two copper electrodes
in an aqueous solution of CuSO4, the charge in solution is carried by migration of
Cu2 ions moving in one direction and SO2
4 ions moving in the opposite. At the
cathode, the incoming Cu2 ions are reduced to metallic copper, thereby lowering
the concentration of these ions in the electrodes vicinity. At the anode, Cu metal
is oxidized to produce Cu2 ions in the solution, so that the local concentration of
cations is increased. The changes in ionic concentrations near each electrode result
in the establishment of concentration proles. As a result, diusion of ions occurs
as well as migration in the electrical eld. If the density of the solution is signicantly changed near the electrode, convection is also present so that three
irreversible processes contribute to the mass transfer, and thus, to the total current observed in the electrolysis experiment.
In general, non-equilibrium processes are a function of both position in the
system and time. A convenient way of discussing these processes is in terms of
uxes, that is, the ow of matter or energy through unit area per second in a given
direction. The ux is a vector quantity which depends on the specic location in
the system, and it is also usually a function of time. When the ux is independent
of time, the non-equilibrium process is called a steady-state process. Irreversible
thermodynamics is used to relate the uxes in the non-equilibrium system to the
forces which give rise to them. As will be seen in the next section in this chapter,
this subject gives a very useful way of relating these quantities when there is more
than one of each, as illustrated in the examples discussed above.
In this chapter, attention is focused on mass transfer processes in isothermal
systems, especially in electrolyte solutions. Heat transfer is not discussed, but the
methods used to treat this problem are very similar to those used to analyze
diusion problems. Chemical reactions in solutions are another example of
non-equilibrium processes. These are discussed in detail in chapter 7.
6.2 The Thermodynamics of Irreversible Processes

Onsager [1] developed an approach to the study of non-equilibrium processes
which is based on elementary thermodynamic concepts. This provides a useful
background to the subjects discussed in this chapter and helps show the connection between various processes occurring in a system undergoing an irreversible
change.
256
The fundamental quantity describing the irreversible process is the ux vector

Ji for species i. In the case of mass transfer, it describes a ow in a given direction
in space in units of moles or grams crossing unit area per second. It can also be
described as the product of the local concentration ci times the velocity vi at which
molecules or ions are moving. Thus,
Ji c i v i
6:2:1
The latter denition is conceptually helpful when the concentration is changing

with position in the system. One may write a similar relationship for an energy
ux, Je , that is
J e re v e
6:2:2
where re is the energy density in joules per cubic meter and ve , its velocity in a
given direction.
The change in the ux with position in the system may be related to the time
derivative of the local concentration by applying Gauss theorem. Consider a
system with volume V and surface area A. Suppose that substance i is owing
out of this volume. Then the rate of substance i leaving in moles per second can be
found by integrating the ux Ji over the surface area A, so that
dn
i Ji
dA
6:2:3
dt
A
where dni =dt is the rate of loss of the substance ni in moles per second. This can
be related to the divergence of the ow, which is dened in vector notation as
divJi
@Ji @Ji @Ji
@x @y @z
6:2:4
Physically, a positive divergence of the ow at a given point x; y; z in the system

means that substance i is owing out of that point, that is, the point is a source.
On the other hand, if the divergence is negative, substance i is owing into the
point and it is a sink. By integrating div Ji over the volume of the system, one can
determine the net change in mass with respect to time for the system. Thus,
divJi dV Ji
dA
6:2:5
V
This is Gauss theorem stated with respect to mass transfer.

The nal part of the analysis involves relating the change in the number of
moles of component i to the changes in its concentration within the system.
Assuming conservation of mass and no chemical reactions, this means that
dni
@ci
dV
6:2:6
dt
@t
V
Combining equations (6.2.3)(6.2.6), one obtains
dni
@ci
dV Ji
dA divJi dV
dt
@t
V
6:2:7
257
It follows that
@ci
div Ji
@t
6:2:8
This is an important result connecting the variation in concentration with time at

a given point in the system to the divergence of the ux at the same point. It is
called the equation of continuity with respect to mass.
The cause of the ux in the irreversible process is called a force. This term is not
used in the strict Newtonian sense but instead it can refer to any source of
perturbation. Forces resulting in uxes include a temperature gradient, a concentration gradient, a density gradient, and a gradient in electrical potential. These
forces are vectors and are represented by the general symbol Xi . If concentration
is varying in the x direction in a system, the gradient is @ci =@x. In order to keep the
discussion general, the gradient is written in vector notation so that variations in
ci in all possible directions can be considered. Thus,
Xc rci
@ci
@c
@c
i ij ik
@x
@y
@z
6:2:9
where Xc is the force due to the concentration gradient and i, j, and k are unit
vectors in the x; y, and z directions respectively. The force due to a temperature
gradient is
XT rT
6:2:10
that due to a gradient in electrical potential

X r
6:2:11
and so on. It is obvious that forces dened in this way have dierent units.
The general equation relating forces and uxes from Onsagers theory is
Ji
n
X
Lik Xk
6:2:12
k1
where Lik is the phenomenological coecient relating force Xk to ux Ji . This

linear relationship only holds when the departure from equilibrium is not large.
The exact limitations for equation (6.2.12) can only be determined experimentally.
It is usually found that each force leads to a dierent ux. Thus, if there are two
forces, there are two uxes and four phenomenological coecients. A system with
three forces has three uxes and nine phenomenological coecients. Fortunately,
these coecients are not all unique, as will be seen below.
The example of the beaker of water on a hot plate is now considered with
respect to equation (6.2.12). There are two forces in this system, namely, a temperature gradient XT , and a density gradient, Xr . Thus, one may write,
JT LTT XT LTr Xr
6:2:13
Jr LrT XT Lrr Xr
6:2:14
and
258
where LTT , LTr , LrT and Lrr are the phenomenological coecients. JT and Jr are
the uxes for the temperature and density, respectively. LTT is directly related to
the coecient of thermal conductivity for water, and Lrr , to the viscosity of
water. The coecients LrT and LTr are related to the eect one process has on
the other. According to one of Onsagers theorems, the cross-coecients are equal
to one another if the forces are dened appropriately. This is an example of an
Onsager reciprocal relation. In this case, only three phenomenological coecients
are required to describe the irreversible process in the system. The easiest way to
achieve an appropriate denition is to dene the force as the gradient of the
energy associated with the process. This conclusion is illustrated in more detail
in this chapter.
Some of the phenomenological coecients relating forces and uxes are
already familiar from less general treatments of the subject. For example, the
phenomenological coecient relating a concentration gradient and a mass transfer ux is the diusion coecient. Other phenomenological coecients are related
to the ionic mobility, the coecient of thermal conductivity, and the solvent
viscosity. These are discussed in more detail later in this chapter.
EXAMPLE
Compare the phenomenological coecient for mass transfer dened in terms

of the concentration gradient with that based on a Newtonian force dened for
the gradient of the chemical potential.
The traditional relationship between the ux of species i and its concentration gradient is
Ji Di rci
6:2:15
where Di is the diusion coecient, that is, the traditional phenomenological

coecient. The negative sign expresses the fact that the ux vector is pointed in
the opposite direction to that in which the concentration gradient increases. If
rci has units of mol m4 and Ji , units of mol m2 s1 , the units of the diusion
coecient are m2 s1 .
In order to dene a Newtonian force, one must relate the concentration
gradient to an energy gradient. The easiest way of achieving this is to use
the chemical potential of species i, which is the thermodynamic measure of
the species ability to do useful work. The chemical potential of species i may
be written as
mi mii0 RT ln ai
6:2:16
where mi0 is the standard chemical potential and ai , the activity of species i. The
gradient of the chemical potential is
rmi RT r ln ai
6:2:17
Ji Li RT r ln ai
6:2:18
Now one may write
where Li is the Newtonian phenomenological coecient for mass transfer. The

units of rmi are J m1 mol1 or N mol1 . Thus the units of Li are mol2 m2
N1 s1 .
259
Another important aspect of Onsagers theory is the principle of microscopic

reversibility. This is usually discussed with respect to the detailed balancing of
chemical reactions. When it is considered with respect to the phenomenological
coecients, it leads to the reciprocal relations discussed above, namely that
Lik Lki
6:2:19
In other words, the matrix giving the phenomenological coecients for the system
is symmetric. Another important part of this theory is that one can estimate the
entropy generated in an irreversible process. Onsager showed that the local rate of
entropy production per unit volume is
y
1 @S 1 X
J
X
V @t T i i i
6:2:20
This relationship can only be applied when the forces are appropriately dened.
In the case of mass transfer, if the force is measured in J m1 mol1 and the ux in
mol m2 s1 , the rate of entropy production is obtained in J K1 m3 s1 .
Onsagers theory provides an elegant background for the consideration of nonequilibrium processes. It is especially helpful when more than one mechanism of
irreversibility is present. This subject has only been briey introduced here. More
details can be found in monographs devoted to this area of physical chemistry
[G1, 2].
6.3 The Viscosity of Liquids

Viscosity is the property of a uid which characterizes its resistance to ow. It is
often measured by timing the ow of a liquid through a cylindrical tube under the
inuence of gravity. In order to understand the denition of the viscosity, consider a uid owing between two large plane parallel plates (g. 6.1). The velocity
of the uid in the direction of the ow, vx , varies with position. It is at its maximum midway between the plates and decreases to zero between each plate on the
basis of experimental observation. Now imagine that the uid is made up of
horizontal layers which are parallel to the plates. The movement of one layer
with respect to another is retarded by a frictional force which is related to the
uids viscosity. The origin of this friction is clearly intermolecular forces.
In order to dene the viscosity, one imagines a cylinder of thickness dx and
radius r located on the central axis of the tube. The retarding force is proportional
to the surface area of the cylinder involved in the ow and to the gradient of the

liquid owing between two parallel
plates. The arrows indicate the
magnitude of the velocity and the
direction of the ow.
260
velocity in the r-direction (see g. 6.1). This leads to Newtons law of viscosity,
which is
Fx ZA
dvx
dr
6:3:1
where Fx is the force in the x-direction, dvx =dr, the velocity gradient in the
r-direction, Z, the viscosity, and A, the area of contact. The outer area of the
cylinder for the example considered is 2pr dx. In the case of laminar ow,
the velocity on the outer surface of the cylinder is constant. This means that
the retarding force due to the viscous drag is exactly counterbalanced by the
force driving the liquid forward, which usually is a pressure dierential in the
tube. In the classical experiments used to measure liquid viscosity the pressure
dierential is established by gravity. If the pressure gradient is dP=dx, then the
force in the x-direction on the cylinder is
Fx pr2
dP
dx
6:3:2
where the minus sign takes into account the fact that the pressure drops as x
increases. Equating these forces with the appropriate value for the contact area,
one obtains
dvx
r dP
dr
2Z dx
6:3:3
If the radius of the tube is r0 , integration of equation (6.3.3) gives the result that
vx
1 2
dP
r r20
4Z
dx
6:3:4
This shows that the forward velocity is a parabolic function of position in the tube
varying from zero at the wall where r is equal to r0 , to the maximum value at the
center of the cylindrical tube, where vx is equal to r20 dP=dx=4Z. It is emphasized that equation (6.3.4) is only applicable to laminar or non-turbulent ow in a
cylindrical tube. When a liquid ows under dierent geometrical boundary conditions, the relationship between the ux and the force is not the same.
From equation (6.3.1), the SI units of viscosity are N s m2 or Pa s. In the older
literature, the viscosity is cited in cgs units, that is, in dyn s cm2 . 1 dyn s cm2 is
called a poise. Since one Newton is equal to 105 dynes, it follows that one poise is
equal to 0.1 Pa s. At room temperature, liquids typically have viscosities of 1 cP,
which corresponds to 1 mPa s.
Another important relationship used to describe the movement of a heavier
solid object through a uid was derived by Sir George Stokes. In the case that the
object is a sphere, the viscous drag or force on it in the direction opposite to its
movement is
F 6pZrv
6:3:5
where r is the radius of the moving sphere. When a heavier solid sphere, for
instance, a marble, is dropped into a viscous liquid, its velocity, which should
continuously increase under the force of gravity alone, is observed to decrease and
261
eventually become constant. At this point, the gravitational force corrected for
buoyancy is exactly balanced by the viscous drag. This leads to the following
relationship:
vmax
2r2 g
r r1
9Z s
6:3:6
Here vmax is the maximum velocity, g, the acceleration due to gravity, rs , the
density of the solid, and r1 , the density of the liquid. Stokes law is important
in relating the movement of ions and molecules dissolved in liquids to other
transport properties.
In order to connect the above expressions to Onsagers theory, it is necessary to
extend equations such as (6.3.1) to three dimensions. This equation shows that a
force in one direction leads to velocity changes in the other two spatial directions.
A general treatment of uid ow requires that one identify the components of the
pressure tensor associated with the force which leads to uid ow. The force
vector F has, in general, a component in each of the three Cartesian directions.
The component in the x-direction, Fx , gives rise to three pressure components,
one in the same direction, Pxx , and two shear components, Pxy and Pxz . Six more
pressure components are obtained from the force components in the y- and zdirections, Fy and Fz . As a result, there is a second-rank pressure tensor Pij with
nine components. Analysis on the basis of classical dynamics leads to the conclusion that the o-diagonal elements, Pij and Pji , are equal. As a result, six distinct
elements of this tensor must be determined to dene it. The second important step
is to write an equation of continuity in terms of momentum. Dening the momentum density as
rv
6:3:7
where r is the local density and v, the local velocity, the momentum equation of
continuity is
@
r
J
@t
6:3:8
where J is the ux of momentum. J is also a second-rank tensor so that the

relationship between and F can only be obtained by considering the tensorial
nature of the problem. As solution of the problem was obtained many years ago
and resulted in the NavierStokes equation [3]. This equation is the basis of the
subject of hydrodynamics. An excellent introduction to this topic can be found in
the monograph by McQuarrie [3].
Values of the viscosity for the polar liquids discussed in chapter 4 are collected
in table 6.1. A large variation in viscosity is seen, varying from 0.3 mPa s for
acetone to 10 mPa s for tetramethylenesulfone. The viscosity is a direct reection
of intermolecular forces, which obviously change considerably in strength in this
group of liquids. Debye [5] considered the viscosity in more detail in this regard
and pointed out that there should be a connection with the relaxation time tD
observed when a polar liquid is exposed to a high frequency alternating electrical
eld (see chapter 4). This parameter also reects the eect of intermolecular forces
when a molecule or group of molecules undergo rotational diusion. Debye
262
Table 6.1 Values of the Viscosity

and the Enthalpy Associated with Its
Temperature Dependence for
Selected Polar Solvents at 258C [4]
Solvent
Viscosity
Z=mPa s
Enthalpy
HZ =kJ mol1
0.890
0.545
1.083
1.956
2.593
3.302
1.65
2.09
9.3
12.5
17.2
18.6
18.6
12.1
0.303
0.341
1.237
0.549
0.927
0.802
1.996
3.245
1.666
1.795
0.610
0.410
2.513
10.29
1.395
5.9
6.1
12.0
8.5
9.4
8.5
13.6
15.0
12.7
7.5
6.6
14.9
20.0
Protic
1.
2.
3.
4.
5.
6.
7.
W
MeOH
EtOH
PrOH
BuOH
F
NMF
Aprotic
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
AC
AcN
BzN
BuN
DMA
DMF
DMSO
HMPA
NMP
NB
NM
PrN
PC
TMS
TMU
developed a model for tD in which it was imagined that one polar molecule which
is assumed to be a sphere undergoes rotational diusion in a liquid continuum
whose intermolecular forces are evaluated by the bulk viscosity Z. On this basis,
he derived the relationship
tD
4pZr3m
kB T
6:3:9
in which rm is the radius of the molecule. If this is applied to estimate the value of
tD for the solvents listed in table 6.1, the result obtained for aprotic solvents is of
the correct magnitude, but always too high.
EXAMPLE
Estimate the Debye relaxation time for nitromethane at 258C given that its
viscosity is 0.61 mPa s and assuming that its molecular diameter is 431 pm.
263
On the basis of equation (6.3.8),

tD
4p6:1 104 2:155 1010 3

1:86 1011 s
1:38 1023 298:2
6:3:10
This estimate of 19 ps is much larger than the experimental estimate of 4 ps,

which is obtained by applying the Debye model to dielectric relaxation data for
the pure solvent (see table 4.2). However, an approximate relationship between
tD and Zr3m is found when data for more solvents are considered, as shown
below.
The fact that estimates of tD are the correct order of magnitude suggests that
there is an approximate relationship between tD and Z, at least for liquids which
obey the Debye relaxation model (see section 4.5). This point is illustrated in g.
6.2 where tD is plotted against Zr3m with logarithmic scales for the aprotic solvents
listed in table 6.1. A very good correlation is found, conrming that equation
(6.3.9) is approximately correct. On the basis of a least-squares t for 14 solvents,
the relationship is
ln tD 5:4 103 0:91 lnZr3m
6:3:11
with a correlation coecient of 0.961. Addition of the data for the protic solvents
results in a much poorer t. This is not surprising because the value of tD for these
liquids reects the rotational diusion of hydrogen-bonded clusters, not that of
single molecules as assumed in the simple Debye model.
The simplest model for the temperature dependence of the viscosity is that
proposed by Arrhenius for a reaction rate constant which leads to the equation
Fig. 6.2 Plot on logarithmic scales of the Debye relaxation time tD for aprotic solvents
against the product Zr3m , where Z is the viscosity and rm , the molecular solvent radius.
264
HZ
Z A exp
RT

6:3:12
where HZ is the activation enthalpy associated with viscous ow and A, a preexponential factor. Values of HZ are also summarized for the polar solvents
considered here in table 6.1. Since the temperature dependence of tD is assumed to
have the same form (equation (4.7.1)) the analysis presented in g. 6.2 allows one
to relate HD and HZ for aprotic solvents. Assuming that rm is independent of
temperature, the relationship is
HD 0:91HZ
6:3:13
Comparison of the data in tables 4.2 and 6.1 shows that this is approximately
correct.
The discussion here has been directed to the viscosity of pure liquids. When an
electrolyte or other solute is added, the viscosity changes. Many data exist in the
literature for the viscosity of solutions. This subject is of special importance to
engineers interested in the ow properties of solutions. The discussion here is
meant to be only a brief introduction to the basic concepts. In the following
sections, attention is turned to the transport of solute species in solutions.
6.4 Isothermal Diffusion in Solutions

As seen in the earlier discussion of Onsagers theory (section 6.2), in the classical
approach diusion is a response to a concentration gradient. If species i is diusing then the ux of i is given by
Ji Di rci
6:4:1
where Di is the diusion coecient of i. This relationship is known as Ficks rst

law. It is most readily applied to describe a steady-state situation in which the ux
Ji is independent of time. Very often, the mass transfer problem involves conditions in which the local concentration is also a function of time. By applying the
equation of continuity with respect to mass one obtains the result
@ci
rJi Di r2 ci
@t
6:4:2
This is known as Ficks second law; it states that the time derivative of the
concentration at a given point in space is equal to the diusion coecient times
the Laplacian of the concentration. In deriving this equation it has been assumed
that the diusion coecient is independent of concentration.
In a simple situation the variation of concentration only occurs in one direction
in space. Then Ficks rst law becomes
Jxi Di
@ci
@x
6:4:3
where Jxi is the component of the ux in the x-direction. In one dimension, Ficks
second law is
@ci
@2 c
Di 2i
@t
@x
265
6:4:4
This is a second-order partial dierential equation whose solution gives the concentration of species i as a function of position and time, that is, of x and t.
As seen earlier, the dimensions of the diusion coecient in SI units are m2 s1.
Typical diusion coecients in aqueous solutions are the order of 1 109
m2 s1 . On the basis of Onsagers theory the quantity leading to the ux, that
is, the force should have the dimensions of Newtons. On the basis of classical
thermodynamics this is most directly achieved by writing Ficks laws in terms of
the chemical potential rather than the concentration. This means that Ficks rst
law becomes
Ji Li rmi
6:4:5
where rmi is the gradient of the chemical potential of species i, and Li , the
phenomenological coecient relating the ux to the force for the diusion process. In general, the chemical potential can be related to the activity of species i,
and thus to its concentration using the equation
mi mi0 RT ln ai mi0 RT ln yi ci
6:4:6
where ai is the activity on the concentration scale, yi , the activity coecient and
mi0 , the standard chemical potential. Combining equations (6.4.5) and (6.4.6), one
obtains
Ji Li RT r ln ai
Li RT
rai
ai
6:4:7
If the solution is ideal (yi 1) or if the activity coecient is constant for the range
of concentrations involved in the experiment, then this equation becomes
Ji
Li RT
rci
ci
6:4:8
It follows that the relationship between the Onsager phenomenological coecient

and the traditional diusion coecient is
Li
Di ci
RT
6:4:9
In many cases, for example, electrolyte solutions, the eects of solution nonideality must be considered. Then equation (6.4.7) can be written as
Ji Li RTr ln ci Li RT r ln yi
Li RT
L RT r ln yi
rci i
rc
ci
ci r ln ci i
6:4:10
It follows that the diusion coecient depends on the concentration of the diusing species according to the relationship
Di Di0 Di0
r ln yi
r ln ci
6:4:11
266
where Di0 is the diusion coecient in the ideal solution. In one dimension this
equation has the form
Di Di0 Di0
d ln yi
d ln ci
6:4:12
EXAMPLE
The diusion coecient of Na is 1:34 109 m2 s1 at innite dilution in

water at 258C. Estimate its value for a Na concentration of 0.1 M in a
solution of the same ionic strength. Use the extended DebyeHuckel theory
to estimate the concentration dependence of the activity coecient assuming
that the ion size parameter is 400 pm.
The activity coecient for the Na+ ion is given by
ln yi
ADH z2i I 1=2

ADH ci1=2

1 BDH aI 1=2
1 BDH ac1=2
i
6:4:13
where I is the ionic strength, and the DebyeHuckel constants are ADH
1:174 M1=2 and BDH 3:288 M1=2 nm1 . The derivative of this quantity
with respect to ci is
d ln yi
0:5ADH ci1=2
0:5ADH BDH a
1=2
dci
1 BDH aci1=2 2
1 BDH aci
6:4:14
d ln yi
0:5ADH c1=2
0:5ADH BDH aci
i
1=2
d ln ci
1 BDH aci
1 BDH aci1=2 2
6:4:15
Thus,
Substituting into equation (6.4.15) for the case that ci is 0.1 M, the value of
d ln yi =d ln ci is 0:093.
Now, on the basis of equation (6.4.12), the diusion coecient for Na+ in
the 0.1 M solution is 1.22 109 m2 s1 .
In practice, conditions in a diusion experiment can be maintained such that
the activity coecient does not change signicantly with electrolyte concentration. For electrolyte solutions, this is done by adding an inert electrolyte whose
ions are not involved in the mass transfer process. Then, it can be assumed that
the ionic strength is constant and independent of position in the system.
Otherwise, the mass transfer problem becomes extremely dicult to solve and
requires information about the diusion coecient as a function of concentration.
6.5 Linear Diffusion from a Wall

In this section, a simple example of the solution of a mass transfer problem is
presented. When one solves dierential equations, it is very important to state the
boundary conditions clearly. In the present example, diusion takes place in the
267
positive x-direction into a solution, but half of the space is lled with a wall
extending in the minus x-direction for all values of y and z. The diusing material
or solute dissolves into the solution and its concentration at the wall is maintained
constant and equal to c0 . As a result the value of the concentration varies with
distance from the wall, x and with time, t. Since the system does not reach a
steady state, the partial dierential equation to be solved is
@c
@2 c
D 2
@t
@x
6:5:1
c cx; t
6:5:2
where
It is also assumed that the diusion coecient D is independent of location in the
solution (x > 0).
The boundary conditions with respect to x are
c0; t c0
6:5:3
c1; t 0
6:5:4
and
The second condition implies that diusion does not occur for a very long time, so
that the concentration far from the wall is zero. The boundary condition with
respect to time is
cx; 0 0
6:5:5
which states that the solute is not present anywhere in the solution initially. The
solution to equation (6.5.1) is obtained using the method of Laplace transforms.
Since there are two independent variables in the present problem, this means that
Laplace transformation must be performed twice, once with respect to the variable t, and then again with respect to the variable x. The Laplace transform of the
concentration with respect to time is given by (see appendix A)
1
cx; s
cx; test dt
6:5:6
The variable s now contains the information about the time dependence of the
problem. Applying Laplace transformation to the derivative on the right-hand
side of equation (6.5.1), one obtains
d 2c
dx2
@2 c st
e dt
@x2
6:5:7
Laplace transformation of the left-hand side is more dicult and requires use of
the product rule. In the present case, this gives
dc est est
Thus, one obtains
@c
s est c
@t
6:5:8
268

1
1
1

@c st
st
e dt ce s c est dt sc
@t
0
6:5:9
Now the partial dierential equation has been transformed into an ordinary
dierential equation in c, namely
D
d2 c
sc
dx2
6:5:10
At this point, Laplace transformation is carried out with respect to x using the
relationship
1
c~u; s
cx; seux dx
6:5:11
The most dicult part of the solution involves transformation of the derivative on
the left-hand side of equation (6.5.10) which requires that the product rule be used
twice:
1
1
1
d2 c ux
dc ux
dc ux
e
e dx
e
dx
u

dx
dx
dx2
0
0
1
1

dc0; s
uc eux u2 c eux dx
dx
0
6:5:12
dc0; s
uc0; s u2 c~

dx
Laplace transformation of the right-hand side gives
1
sc eux dx s c~
6:5:13
The result can be expressed as

Du2 c~ DAu DB s c~
6:5:14
A c0; s
6:5:15
where
and
B
d c; 0; s
dx
6:5:16
Equation (6.5.14) is rearranged to obtain an expression for c~:

c~
where
DAu DB Au B
2
Du2 s
u v2
6:5:17
v2
s
D
269
6:5:18
By using the method of partial fractions this result may be expressed as

c~
a
b
uv uv
6:5:19
where the constants a and b may be found by equating the numerators of equations (6.5.17) and (6.5.19) when equation (6.5.19) is in complex form. In other
words
au v bu v Au B
6:5:20
From the coecients of u,

abA
6:5:21
vb a B
6:5:22
and from the constant term,
At this point, the inverse Laplace transformation of equation (6.5.19) is carried

out. In order to do this, one must consult a table of Laplace transforms such as
the one given in appendix A. Operating on each term individually, one obtains

a
L1
a evx
6:5:23
uv
and
L
1

b
b evx
uv
6:5:24
As a result, the general expression for cx; s is

cx; s aevx bevx
6:5:25
Now, the boundary conditions with respect to x are applied so that the solution
specic to the problem addressed can be found. When x is innitely large, the
concentration goes to zero (equation (6.5.4)). The only way that this result is
obtained is by setting the constant b equal to zero. When x is equal to zero,
the concentration is constant and equal to c0 (equation (6.5.3)). Laplace transformation of this boundary condition with respect to time gives
1
c0 est dt
1
c0 st
c
e 0 c0; s
s
s
0
6:5:26
By comparing this result with equations (6.5.15) and (6.5.21), it is clear that a selfconsistent solution has been obtained. Thus, one may now write
cx; s
c0 vx
e
s
6:5:27
270
The nal step is to perform inverse Laplace transformation on cx; s to obtain

cx; t. Using the table of Laplace transformations, inverse transformation of
equation (6.5.27) results in

x
cx; t c0 erfc
6:5:28
4Dt1=2
where erfc is the error function complement.
The error function of the parameter z is dened as
z
2
2
erfz 1=2 ey dy
p
6:5:29
This function corresponds to the area under the normal error distribution curve
from its maximum value to a value z. The error function has a value of 0 when z is
zero, and a value of one when z is innity. The error function complement is
simply dened as
erfcz 1 erfz
6:5:30
This function goes from one to zero.

The nal result in equation (6.5.28) shows that the concentration falls from its
initial value of c0 at the wall in a manner determined by the error function and the
diusion coecient D. Typical concentration proles estimated at various values
of t are shown in g. 6.3. Obviously, the concentration increases at a given
distance from the wall as time increases.
It is also useful to calculate the ux of the diusing species using Ficks rst
law:
Jx; t D
@cx; t
@x
6:5:31
Fig. 6.3 Typical concentration proles estimated for various times for diusion from a
semi-innite wall with D 1 109 m2 s1 .
271
This can be done by dierentiating equation (6.5.28) with respect to x, and

requires that the rst derivative of the error function be estimated. On the basis
of equation (6.5.29),
2 dz
d erfz
2
1=2 ez
dx
dx
p
6:5:32
d erfcx=2Dt1=2 expx2 =4Dt
dx
pDt1=2
6:5:33
!
1=2
D
x2
Jx; t c0
exp
pt
4Dt
6:5:34
1=2
D
J0; t c0
pt
6:5:35
It follows that
Thus,
At the wall, the ux is
This is an important result used in electroanalytical chemistry.

The importance of the boundary conditions in obtaining a solution to Ficks
equation is apparent from the example presented here. This equation has been
solved for many dierent situations, including cylindrical and spherical geometries [G2, G3]. It has often been applied to mass transfer problems in electroanalysis and important examples can be found in monographs devoted to this
subject [G4].
6.6 The Electrochemical Potential

When mass transfer occurs in electrolyte solutions, the component ions can move
independently of one another. Because of the ionic charge, the Gibbs energy of an
ion depends not only on its local concentration but also on the local electrostatic
potential. It is also necessary to consider circumstances in which changes in the
local potential are brought about by application of an external eld. Thus, it is
important to be able to deal with the chemical and electrostatic contributions to
the Gibbs energy separately. The thermodynamic tool for achieving this is the
electrochemical potential.
The electrochemical potential of charged species i, m~ ai is dened as the work
done when this species is moved from charge-free innity to the interior of a
homogeneous phase a which carries no net charge [6]. The opposite process is
known as the work function and is familiar for the case of removal of an electron
from a metal. In fact, the work function for single ions in electrolyte solutions can
also be measured experimentally, as described in detail in chapter 8. This means
that the electrochemical potential is an experimentally determinable quantity.
However, separation of the electrochemical potential into chemical and electrostatic contributions is arbitrary, even though it is conceptually very useful.
272
A very helpful way of understanding the contributions to m~ ai has been described

by Parsons [6]. It is imagined that phase a, for example, an electrolyte solution, is
a sphere isolated in free space (g. 6.4). This sphere is divided into two regions, a
central spherical region where the distribution of ions and solvent molecules
reects the homogeneous nature of the bulk, and a thin spherical shell at the
surface which contains the interfacial double layer and any excess surface charge
associated with phase a. The double layer at the aqueous electrolyte solution|air
interface is due to any net orientation of water molecules at the surface, and to the
fact that hydrated cations and anions may have dierent distances of closest
approach due to diering degrees of hydration. Electrolyte solutions do not normally possess excess charge, but this could easily be the case if phase a were a
metal.
Now, the two parts of phase a are separated (g. 6.4) and the process of
introducing a charged species to each part repeated. The work done in the case
of the homogeneous phase a is the chemical potential mai . On the other hand, the
work done in moving the charged particle through the surface shell into the empty
interior is associated with the electrostatic characteristics of phase a and is written
as zi Ffa . fa is called the inner potential of phase a, and cannot be measured for
an isolated phase. This fact can be understood from the point of view that one is
only able to measure a potential dierence in electrostatics. One is also not able to
measure the chemical potential of a single charged species, for example, a single
ion. On the other hand, one is able to determine the chemical potential of a
neutral electrolyte in which the charges on the individual ions add to zero.
As a result of the thought experiment described above, the expression for the
electrochemical potential of species i in phase a is
m~ ai mai zi Ffa
6:6:1
For a simple 11 electrolyte MX, the electrochemical potential of the cation is

m~ aM maM Ffa
6:6:2
m~ aX maX Ffa
6:6:3
and that of the anion
Fig. 6.4 Schematic representation of the

division of the electrochemical potential of
species i in phase a into chemical and electrical
contributions.
273
The electrochemical potential of the electrolyte is

m~ aMX m~ aM m~ aX maM maX maMX
6:6:4
Thus, the distinction between the electrochemical potential and the chemical
potential is lost when the properties of the electrolyte as a whole are considered.
The development of the thermodynamic properties of the electrolyte are those
discussed earlier in section 3.6.
The inner potential is an important property of individual phases. Much more
will be said about this property when the interface between two phases is discussed in chapter 8. For the moment, fa is regarded as a property of phase a
which is the same throughout the phase with a value dened with respect to
charge-free innity. On the other hand, in the case of an electrolyte solution,
the local electrostatic potential varies from point to point due to the presence
of discrete charges on the ions. Thus the electrostatic potential is more positive at
a cation and more negative at an anion. These uctuations occur about the
average value, fa . This can be seen more clearly by writing out the chemical
potential of ion i in terms of its concentration ci and activity coecient yi .
Thus, from equation (6.6.1)
m~ ai mai0 RT ln yi ci zi Ffa
6:6:5
The activity coecient is directly related to the work done to create the ionic
atmosphere around ion i, so that this equation can also be written as
m~ ai mai0 RT ln ci zi Fai
6:6:6
ai
is the micropotential or local potential experienced by this ion (see

where
section 3.8). The relationship between fa and ai is
ai fa ia
6:6:7
where ia is the discreteness-of-charge potential. On the basis of this denition, ia

is the dierence between the local potential and its average value which arises
because the charge in the solution is localized on ions. The discreteness-of-charge
potential may be expressed in terms of two contributions as follows:
a
a
zi F ia zi F self
zi F atm
6:6:8
a
self
is the self-potential of the ion which depends on its radius and the dielectric
a
, is the self atmosphere potential due to the
properties of the medium, and atm
ionic cloud formed around ion i. The latter quantity is directly related to the
activity coecient yi as follows:
a
RT ln yi zi F ;atm
6:6:9
Since the self-potential of ion is not included in the denition of yi , the standard
states in equations (6.6.5) and (6.6.6) are not the same (see section 3.8).
The concept of the electrochemical potential is also helpful in analyzing
changes in the Gibbs energy of dipoles and multipoles when the electrical environment is changed. In the case of a dipole, the electrostatic energy is determined
by the local eld Ee . The local eld can change due to a change in environment or
274
the imposition of the eld from an external electrical source. A well-known example of an environmental change occurs when a dipolar molecule is adsorbed at an
interface. Under these circumstances the local eld can be quite dierent, so that
the electrostatic contribution to the Gibbs energy changes signicantly.
The electrochemical potential of a dipolar molecule may be written as
m~ ai mai0 RT ln ai pi Ee
6:6:10
where ai is the activity of molecule i, pi , its dipole moment, and Ee , the local eld.
The electrostatic energy is expressed as a vector dot product because the angle
between the vectors must be known in order to estimate it. An example of the
application of dipolar electrochemical potentials is given in chapter 10.
Discussion of non-equilibrium processes involving ions in terms of the micropotential is especially helpful because it focuses attention on the fact that major
source of non-ideality in these systems is electrical in character. The arbitrary
nature of the separation of the electrochemical potential into chemical and electrical contributions has often been pointed out in the literature. In fact, chemical
interactions are fundamentally electrical in nature. However, the formal separation discussed here is conceptually important. Its usefulness becomes clear when
one tackles problems related to the movement of ions in electrolyte solutions
under the inuence of concentration and electrostatic potential gradients. These
problems are discussed in the following section.
6.7 The Conductivity of Electrolyte Solutions

Electrolyte solutions contain ions which can move in response to a gradient in
electrical potential. The transport properties of these systems are important in
devices such as batteries and in living systems. The movement of ions in solution
is very dierent from the movement of electrons in metallic conductors, and it is
important to understand the fundamental laws which govern the conductivity of
electrolyte solutions. Ions move according to the classical laws of physics, whereas
the movement of electrons is quantal.
The force governing the movement of ions is the local gradient of the electrostatic potential, ri . If eects due to the presence of an ion at a given point in the
solution and other ions surrounding it are ignored, and only the macroscopic
potential gradient is considered, then ri can be replaced by rf. Local electrostatic eects are generally considered to be the major source of non-ideality in
electrolyte solutions at equilibrium. In the classical treatment of conductivity all
eects due to non-ideality are included in the phenomenological coecients,
namely, the mobility and the molar conductance discussed below. The ux of ion
i is normally written in terms of the current density, ii . Recalling that ux has units
of moles per square meter per second, the corresponding current density is given by
ii zi FJi
6:7:1
where zi is the valence of ion i. The total current density which is observed
experimentally is the sum of the contributions from each ion:
zi FJi
275
6:7:2
Since cations and anions move in opposite directions under the inuence of an
electrical eld, the vector JM for a cation is in the opposite direction to that for an
anion JX . This means that the ux Ji multiplied by the ionic valence zi has the
same sign for all ions; as a result each ion contributes to the solution conductivity
in such a way that it becomes larger.
The phenomenological coecient relating the current density and the macroscopic potential gradient is the conductivity due to the ion, ki . Thus, one has the
general relationship
ii ki rf
6:7:3
The minus sign reects the fact that a positive ion moves in the direction opposite
to the increase in electrical potential. If rf is positive, the electrical potential
increases in a specied direction, for instance the x-direction, but a cation
moves in the opposite direction. This equation is a special form of Ohms law
written in terms of the current density and the potential gradient, rather than the
current and the potential dierence. The experimentally measured quantity is the
conductivity ke , which is the sum of the contributions from each ion in solution:
X
ke
ki
6:7:4
i
In order to measure the conductivity of an electrolyte solution, a special cell is

constructed in which the two electrodes of accurately known dimensions are
placed at a known distance from one another (g. 6.5). The electrodes are usually
made of an inert material such as platinum. If a d.c. voltage is applied to such a
cell, no current ows until a certain minimum value is reached. Below this minimum, no reactions occur at the electrodes, and the electrode|solution interface
behaves as a capacitor. When d.c. current ows, reactions occur at each electrode
and ions move in the electrolyte solution. The nature of the reactions depends on
the nature of the electrolyte but certainly the component ions are involved in these
reactions. For this reason, d.c. experiments are not used in the precise determination of solution conductivity. On the other hand, if an a.c. voltage of low amplitude is applied to the cell, its conductivity may be determined in the absence of
any net change in the composition of the solution. Traditionally, this experiment
was carried out by making the conductance cell one arm of an a.c. Wheatstone
bridge designed to measure resistance.
Fig. 6.5 Schematic diagram of

a conductivity cell (a) and its
impedance representation as an
equivalent circuit (b).
Capacitances C1 and C2 are at
the electrode|solution
276
The relationship between the solution resistance Rs and the conductivity ke is

ke
d
Rs A
6:7:5
where d is the separation between the plates of the conductivity cell and A is the
area of the electrodes. The units of ke are 1 m1 . In order to estimate ke from
the experimental value of the resistance, the cell parameters d and A must be
carefully determined. In practice, the ratio d=A is found by calibrating the conductivity cell with a solution of known conductivity.
Another important parameter related to electrolyte solution conductivity is the
transport number ti . This is dened as the fraction of the current carried by
species i. On the basis of equation (6.7.4), it is clear that ti is given by
ti
ki
ke
6:7:6
The classical way of measuring transport numbers is in a Hittorf cell, which is

illustrated schematically in g. 6.6. The cell consists of three compartments designated L (left), M (middle), and R (right). In the present example the two electrodes are composed of pure copper, and the electrolyte is an aqueous solution of
Cu(NO3)2. Direct current ows through the cell, and the amount of charge which
passes through the system, q, is measured by the coulometer in the circuit.
For the example chosen, the reaction at electrode L, which is the cathode, is
Cu2 2e !Cu
6:7:7
This is accompanied by the transport of Cu2 ions into L and transport of NO
3
ions out of L. The net change in number of moles of Cu2 ion in compartment L
is
Fig. 6.6 Hittorf cell for determining transport

numbers with associated circuitry. Circuit element C
is a coulometer. The three compartments are labeled
L (left), M (middle), and R (right).
nL Cu2
q
t q
t q

2F 2F
2F
277
6:7:8
Since there are only two kinds of ions, the fractions t and t must add to unity.
The number of moles of NO
3 ion moving out of L is
nL NO
3
t q
F
6:7:9
Since each nitrate ion has a charge of 1, the number of moles is obtained by
dividing the charge moved by the Faraday constant F. Thus, the total change in
the number of moles of Cu(NO3 2 is t q=2F with the requirement of electroneutrality being maintained.
The reaction at the anode R is the opposite, namely
Cu!Cu2 2e
6:7:10
In this compartment, Cu2 ion is transported out and NO

3 ion moves in.
Analysis of the next changes leads to the result
nR Cu2
nR NO
t q
3

2
2F
6:7:11
The net change in the middle compartment is zero. By analyzing the contents
of each compartment after passing a known charge q through the system, the
transport numbers t and t can be calculated.
Two other quantities are commonly used in discussing the transport properties
of electrolytes. The rst is the molar conductivity of ion i, which is dened as
li
ki
ci
6:7:12
where ci is the concentration of the ion. In order to obtain reasonable dimensions

for li , the concentration is expressed in mol m3 so that li has units of 1 m2
mol1 . As the concentration of ion i increases one expects the conductivity of the
solution to increase as well. To a rst approximation, this increase should be
proportional to the change in concentration, so that li is a constant. In fact, li
decreases with increase in electrolyte concentration as a result of the eects of
non-ideality. This observation is discussed in detail in the following section.
Another way of presenting information about the concentration dependence of
ki is in terms of the equivalent conductance dened as
li
ki
jzi jci
6:7:13
One equivalent of ion i is the amount which contains one mole of univalent
charge. This quantity is often found in the older literature.
Another important parameter used to discuss the transport properties of electrolytes is the ionic mobility, ui . This is the velocity of the ion in a unit electrical
eld. It can be found from the molar conductivity by dividing it by the charge on a
mole of ions:
278
ui
li
jzi jF
6:7:14
Dimensional analysis shows that ui has units of m2 V1 s1 in the SI system. It
is also useful to relate ui to the conductivity of the solution. Combining equations
(6.7.4), (6.7.12), and (6.7.14), one obtains
X
ke
jzi jF ui ci
6:7:15
i
EXAMPLE
The molar conductance of 0.5 M MgCl2 in water is 165.52 cm2 mol1 1 at
258C. The cationic transport number is 0.308 at this concentration. Estimate
the molar conductances of the individual ions. Also estimate the ionic mobilities.
The molar conductance for Mg2 is given by
lMg2 0:308 165:52 50:98 cm2 mol1 1
For the Cl ion,
lCl
1 0:308 165:62
57:27 cm2 mol1 1
2
Using equation (6.7.14), the mobility of Mg2 is

uMg2
50:98
2:64 104 cm2 s1 V1
2 96;485
and that of Cl is

uCl
57:27
5:94 104 cm2 s1 V1
96;485
The Mg2 ion is much more strongly solvated by water molecules than Cl .
Thus, it moves with more water molecules and its mobility is much less.
The relationship between ionic conductivity and Onsagers theory can now be
presented in terms of the electrochemical potential. By expressing the force leading to the transport of ions in terms of the gradient of m~ i , one nds important
relationships between the diusion coecients of the ions, and the molar conductivity and mobility. Furthermore, when the force has the correct Newtonian
units, one is also in a position to calculate the rate of entropy production. On the
basis of the thermodynamics of irreversible processes, the relationship between
the ux of ion i and the force rm~ i is
Ji Li rm~ i
6:7:16
where Li is the appropriate phenomenological coecient. Using equation (6.6.5),

this can be written as

L RT
r ln yi
1
6:7:17
rci Li zi Frf
Ji i
ci
r ln ci
279
The factor in front of rci is simply the diusion coecient Di (equation (6.4.10)).
Now, using equations (6.7.1) and (6.7.3), the contribution to the ux from the
potential gradient may be expressed in terms of the molar conductivity. The result
is
k
6:7:18
Ji Di rci i rf
zi F
Since
ki jzi jF ui ci
6:7:19
the ux of ion i can also be written as

Ji Di rci
z i ui c i
rf
jzi j
6:7:20
Comparing equations (6.7.17) and (6.7.20), one nds that

Li
Di ci
uc
i i
RT jzi jF
6:7:21
In the limit of innite dilution, where the eects of non-ideality are absent, one
obtains the Einstein relationship between these parameters for a single ion:
Di0
RT
u
jzi jF i0
6:7:22
The subscript 0 designates the values of Di and ui in the innitely dilute limit.
A typical mobility for an ion in water is 5 108 m2 s1 V1 . For a monovalent
ion, the corresponding diusion coecient at 258C is 1:3 109 m2 s1 .
The diusion coecient of an individual ion may also be related to its molar
conductance. Making use of equation (6.7.14), one may write
li0 jzi jui0 F
z2i F2 Di0
RT
6:7:23
For the whole electrolyte, the limiting molar conductivity is

i0 n l0 n l0
F2
n z2 D n z2 D0
RT 0
6:7:24
where n and n are the stiochiometric numbers of the cation and anion, respectively. This is known as the Nernst-Einstein equation.
Another important relationship relevant to the mobility of ions may be derived
from Stokes law. If it is imagined that the ion is a sphere moving through a
solvent of viscosity Z at a constant velocity vi , then the force due to the electrical
eld is exactly counterbalanced by the force due to the viscous drag. The electrical
force is the ionic charge zi e0 times the electrical eld E. Thus, using equation
(6.3.5), one may write
jzi je0 E 6pZri vi
6:7:25
280
The ratio of vi to E is simply the ionic mobility ui so that

jz je
ui i 0
6pZri
In terms of the diusion coecient, one obtains
kT
Di
6pZri
6:7:26
6:7:27
This relationship is known as the StokesEinstein equation. Strictly speaking it

should only be applied at innite dilution to monoatomic ions. However, in
practice it is applied to more complex ions and at nite ionic strengths. If the
diusion coecient for the ion is measured experimentally, an eective radius for
the ion can be estimated using the viscosity of the pure solvent.
EXAMPLE
Using the data given in the previous example and in table 6.1, estimate the
diusion coecients for Mg2 + and Cl in 0.5 M MgCl2 . Estimate also their
eective Stokes radii.
The mobility of Mg2 is 2:64 104 cm2 s1 V1 . The diusion coecient is
equal to
0:0257
3:39 106 cm2 s1
DMg2 2:64 104
2
Using equation (6.7.27), the Stokes radius for this ion at 258C is
kT
4:11 1021
6pZDi 6p 8:9 104 3:39 106 104

7:24 1010 m 724 pm
ri
For Cl , the diusion coecient is

DCl 5:94 104 0:0257 1:527 105 cm2 s1
The Stokes radius is
ri
4:11 1021
1:60 1010 m 160 pm
6p 8:9 104 1:527 109
The results for the Stokes radii are reasonable but they should have been
obtained using the viscosity of the solution, not that of the pure solvent.
Unfortunately, the necessary data are usually not available.
At nite concentrations, the uxes of the ions of the electrolyte are not completely independent of one another because of ionion interactions. This can be
seen clearly when this phenomenon is examined within the context of Onsagers
treatment of ion transport. It is assumed in the following that the electrolyte MX
contains two univalent ions. The ux of the cation is
JM LMM rm~ M LMX rm~ X
6:7:28
where rm~ M and rm~ X are the gradients of the electrochemical potentials of the
cation and anion, respectively, LMM , the phenomenological coecient describing
281
the ux of the cation resulting from rm~ M and LMX , that describing the ux of the
cation resulting from rm~ X . The cross-coecient LMX is expected to be much
smaller than the principal coecient LMM , but both must be considered in a
general treatment. For the anionic ux, the corresponding expression is
JX LXM rm~ M LXX rm~ X
6:7:29
where LXX is the principal phenomenological coecient for the anion, and LXM ,
the cross-coecient. Because of the principle of microscopic reversibility, the
coecients LMX and LXM are equal. This means that three phenomenological
coecients must be determined in order to obtain the transport properties of
the system. In terms of traditional treatments, the three quantities determined
experimentally at a given electrolyte concentration are the diusion coecient
of the electrolyte and the mobilities of each ion. By examining the experiments
used to determine each of these quantities on the basis of the Onsager treatment,
the relationships among LMM , LMX , LXX , and the traditional parameters can be
obtained.
In a conductivity experiment, the conductance is determined in the absence of a
concentration gradient by using an a.c. potential gradient rfac so that the force
on ion i is zi F rfac . Thus, the ionic uxes in this experiment are
JM zM LMM F rfac zX LMX F rfac
6:7:30
Jx zM LMX F rfac zX LXX F rfac
6:7:31
and
Using equation (6.7.2) the a.c. current density for a 11 electrolyte is

iac FJM FJX F2 LMM 2LMX LXX rfac
6:7:32
Thus, according to equations (6.7.3) and (6.7.4), the conductivity of the solution is
k F2 LMM 2LMX LXX
6:7:33
Expressions may also be written for the transport numbers tM and tX using the
above analysis:
tM
FJM
LMM LMX
iac
LMM 2LMX LXX
6:7:34
and
tX
FJX
LXX LMX
iac
LMM 2LMX LXX
6:7:35
As required by the denition of the transport numbers, tM and tX add to unity.

When diusion occurs in an electrolyte solution in the absence of an external
electrical eld, local electrical neutrality must be maintained. When rf is zero, the
ionic uxes are
JM LMM rmM LMX rmX
and
6:7:36
282
JX LMX rmM LXX rmX
6:7:37
The net electrical current in the solution must be zero so that

FJM FJX 0
6:7:38
LMM LMX rmM LXX LMX rmX
6:7:39
It follows that
The gradient of the chemical potential of the electrolyte is given by

rme rm~ M rm~ X rmM rmX
6:7:40
Combining equations (6.7.39) and (6.7.40), one nds after some straightforward algebra that
rmM
LXX LMX
rm
LMM 2LMX LXX e
6:7:41
rmX
LMM LMX
rm
LMM 2LMX LXX e
6:7:42
and
Now, using equations (6.7.36) and (6.7.37), one nds that

JM JX
LMM LXX L2MX

rm
LMM 2LMX LXX e
6:7:43
The ux of each ion is also equal to the overall ux of the electrolyte Je when
electroneutrality is maintained. The relationship between this quantity and rce is
Je De rce
6:7:44
where De is the overall diusion coecient of the electrolyte. The relationship

between rme and rce is
rme RTr ln ae
RT
rc
ce e e
6:7:45
where
e 1
r ln ye
r ln ce
6:7:46
This leads to the nal result that

De
RT
LMM LXX L2MX
e
ce
LMM 2LMX LXX
6:7:47:
Equations (6.7.33), (6.7.45), and (6.7.34) or (6.7.35) form a system of three

equations in three unknowns which may be solved for the individual phenomenological coecients LMM , LMX , and LXX . The results are
LMM
De ce ke t2M
2
RT
e
F
6:7:48
LMX
283
De ce ke tM tX

RT
e
F2
6:7:49
De ce ke t2X
2
RT
e
F
6:7:50
and
LXX
The non-ideality of the electrolyte solution leads to complex relationships

between the phenomenological coecients and the transport parameters measured experimentally. These only become simple in the limit of innite dilution
where LMX goes to zero and
e to unity. For example, under these conditions
equation (6.7.47) reduces to
De0
RT LMM LXX
ce LMM LXX
6:7:51
Since the diusion coecients for the individual ions can be dened from LMM
and LXX using equation (6.7.21), one may write
De0
DM0 DX0
DM0 DX0
6:7:52
where DM0 and DX0 are the diusion coecients for M and X , respectively, at
innite dilution. This equation was rst derived by Nernst.
As can be seen from the above development, study of the transport properties
of electrolyte solutions led scientists in the late nineteenth and early twentieth
centuries to think about these systems on a microscopic scale. Important connections between the movement of ions under the inuence of thermal and electrical
eects were made by Einstein. This was all brought together in an elegant way by
Onsager. An important question faced by those involved with these studies is
whether the electrolyte is completely dissociated or not. The answer to this question can be found be examining both the equilibrium and non-equilibrium properties of electrolyte solutions. The latter aspect turns out to be more revealing and is
discussed in more detail in the following section.
6.8 Experimental Studies of Conductivity

The study of electrolyte conductivity dates from the latter half of the nineteenth
century when it was realized that simple salts dissociate into ions in aqueous
solution. Furthermore, when a current ows in such a solution, the cations
move to a large extent independently of the anions, resulting in two separate
vectorial components with positive charge moving in one direction and negative
charge in the opposite. These studies showed that the molar conductivity varies
with electrolyte concentration. In addition, it was possible to distinguish two types
of behavior, namely, that characteristic of strong electrolytes, and that characteristic of weak electrolytes.
Many of the careful early experiments were carried out in the laboratory of
Kohlrausch who showed that the conductivity of solutions of dilute strong elec-
284
trolytes was linear in the square root of the electrolyte concentration. The
Kohlrausch result may be expressed as
0 ke c1=2
e
6:8:1
where 0 is the molar conductance at innite dilution and ke , the slope of the
relationship between and c1=2
e . Examples of the conductivity behavior of several
electrolytes are shown in g. 6.7. In all cases, the conductivity decreases as the
concentration increases. This can be understood as an eect of the ionic environment on the motion of individual ions. Three of the electrolytes shown in g. 6.7,
namely, NaCl, KCl, and sodium acetate (NaC2 H3 O2 ) show the behavior predicted by equation (6.8.1). Thus, the conductivity is linear in the square root of
the concentration in the limit of low concentrations so that these systems can be
classied as strong electrolytes. Extrapolation to zero concentration is easily carried out, so that values of 0 can be estimated. In the case of acetic acid, a weak
electrolyte, the behavior is quite dierent and the value of 0 is impossible to
obtain on the basis of the Kohlrausch plot.
The second important parameter obtained directly from experiment is the
transport number for one of the ions of the electrolyte. On the basis of the
analysis in section 6.7, the cationic transport number for a 11 electrolyte can
be written as
t
l
u
l l u u
6:8:2
where li is the molar conductivity of ion i, and ui , its corresponding mobility. If it

is assumed that the individual ionic molar conductivities are also given by an
equation similar to equation (6.8.1), then one may write
Fig. 6.7 Plots of the molar conductivity against the square root of the electrolyte concentration for three strong electrolytes (*) and one weak electrolyte (^). The data for the
strong electrolytes were tted to a straight line at lower concentrations in order to obtain
the limiting molar conductivity 0 .
li li0 1
ki c1=2
e
li0
285
6:8:3
where ki is the constant describing the concentration dependence of li for ion i.

Substituting this expression for each molar conductivity into (6.8.2), the following
expression is obtained for the concentration dependence of t at low concentrations:
t t0 kt c1=2
e
6:8:4
t0 is the transport number in the limit of innite dilution, and kt , the constant
describing the concentration dependence of t . The sign of kt can be positive or
negative depending on the relative values of the corresponding coecients for the
constituent ions, k and k . Since the transport numbers t and t add to unity,
the corresponding equation for t is
t t0 kt c1=2
e
6:8:5
Typical transport number data for 11 electrolytes are shown as a function of c1=2
e
in g. 6.8. It is clear that these data can easily be extrapolated to obtain the value
of the transport number at innite dilution.
Having obtained the limiting conductance for an electrolyte and the limiting
transport number for one of its ions, one may calculate limiting molar conductivities for the individual ions. Examination of experimental data for strong electrolytes shows that l0 for a single cation, say Na , is independent of the nature of
the anion. Similarly, for a series of electrolytes with a common anion, say Cl , the
value of l0 is independent of the nature of the cation. Thus, by selecting contributions from the appropriate ions one can calculate the limiting ionic conductance for any electrolyte whose constituent ions are listed in table 6.2. For
Fig. 6.8 Plots of the cationic transport number 258C against the square root of the electrolyte concentration for three strong electrolytes in water.
286
Table 6.2 Limiting Molar Conductivities of Single Ions

in Water at 258C [7]*
Cation
li0 =cm2 mol1 1
H
Li
Na
K
Rb
Cs
Ag
NH
4
Mg2
Ca2
Sr2
Ba2
Cd2
Zn2
Ni2
Pb2
349.8
38.68
50.10
73.50
77.81
77.26
61.90
73.55
106.1
119.0
118.9
127.3
108.0
107.0
108.0
140.0
Anions
OH
F
Cl
Br
I
NO
3
ClO
4
C2 H3 O
2

N3
CN
SCN
SO2
4
CO2
3
C2 O24
Fe(CN)3
6
Fe(CN)4
6
li0 =cm2 mol1 1

198.3
55.4
76.35
78.14
76.84
71.46
67.36
40.90
69.5
78.0
66.5
160.04
138.6
148.3
302.7
442.0
*In the older literature, the limiting conductance is usually given per
equivalent of the ion rather than per mole as cited here. The equivalent
conductance is equal to the molar conductance divided by the number
of charges on the ion. Thus, the equivalent conductance for the divalent
cations listed above is equal to the molar conductance divided by two.
example, in the case of Zn(ClO4 )2 , there is a contribution of 107.0 cm2 mol1 2
from the Zn2 ion, and a contribution of 134.7 cm2 mol1 1 from the two
perchlorate ions. These add to a value of 0 equal to 241.7 cm2 mol1 1 for the
electrolyte. The same data can be used to estimate the limiting ionic conductivity
for a weak electrolyte. In the case of acetic acid the result is 390.7 cm2 mol1 1 .
Examination of the data for this system presented in g. 6.7 shows that the value
of changes markedly at low concentrations, so that 0 clearly could not be
estimated on the basis of the Kohlrausch plot.
The limiting molar conductance is a fundamental property of ions in electrolyte
solutions. It can be used to estimate the ionic mobility (equation (6.7.14)) and the
diusion coecient of the ion (equation (6.7.22)) in the limit of innite dilution.
When data are available for ions in a given series, trends can be seen in the values
of li0 . In the case of the alkali metal cations, l0 increases with increase in cation
atomic number up to Rb , the values for Rb and Cs being approximately the
same. A similar trend is seen for the halide ions for which l0 increases in the
series F < Cl < Br , and then decreases slightly for I . These results are
usually interpreted in terms of the extent to which the ion is solvated in water.
Since Li is the smallest crystallographic radius in the alkali metal series, it is the
most strongly hydrated and carries some surrounding water molecules when it
moves. On the other hand, the large Cs ion is only weakly hydrated, and therefore, it can move more rapidly through the electrolyte solution.
The concepts that ions move independently and that the individual ionic limiting conductivities can be used to estimate the limiting conductivity of an electro-
287
lyte were major achievements in developing an understanding of the properties of

electrolyte solutions. However, many studies were also made with weak electrolytes which are only partially dissociated. The concentration behavior of these
systems, for example, acetic acid (see g. 6.7), is clearly dierent. When a simple
11 electrolyte is only partially dissociated one has the equilibrium
MX M X
6:8:6
whose equilibrium constant is

Ka
aM aX
y2 a2 ce
aMX
yMX 1 a
6:8:7
ai is the activity of species i, yi , the corresponding activity coecient, ce , the

concentration of the weak electrolyte and a, the fraction of the electrolyte
dissociated. Dening Kc as the equilibrium constant in terms of concentration,
one may write
Kc
yMX Ka
a 2 ce
1a
y2
6:8:8
Kc will be constant only in the limit of very low concentrations where the activity
coecients are unity. The second assumption regarding weak electrolytes is that the
fraction dissociated can be directly obtained from the ionic conductivity, that is,
a
6:8:9
Combining equations (6.8.8) and (6.8.9), and writing the result in a linear form,
one obtains
1
1
c
e
0 Kc 20
6:8:10
This is the Ostwald dilution law for weak electrolytes.

Plots of data for two weak electrolytes, namely acetic acid and ammonium
hydroxide, according to Ostwalds law are shown in g. 6.9. These plots give
reasonable straight lines with somewhat more scatter in the case of NH4OH.
The value of 0 for acetic acid from the intercept is 388.5 cm2 mol1 1 ,
which agrees quite well with the value calculated from the limiting ionic conductivities in table 6.2 (390.7 cm2 mol1 1 ). The value of Kc estimated from the
slope is 7:8 105 M, which is close to the value found in tabulations of acidity
constants. In the case of NH4OH, the estimate of 0 from the intercept is 115.9
cm2 mol1 1 , which is very dierent from the value obtained from limiting
conductances (271.4 cm2 mol1 1 ). The estimate of Kc from the slope is 7:8
105 M compares with 1:8 105 M from basicity constant tabulations. These
calculations show the weakness of the Ostwald analysis. Since 0 is much larger
than the measured conductances, the error associated with estimating it from its
reciprocal in the plots shown is very large. Any error in 0 is then reected in the
estimate of the Kc which comes from the reciprocal of the slope, equal to Kc 20 .
An additional problem comes from the neglect of the activity coecients involved
288
Fig. 6.9 Plot of 1/ at 258C against ce for two weak electrolytes, namely, acetic acid (^)
and NH4 OH (*).
in the thermodynamic association constant Ka . Robinson and Stokes [7] have

described an iterative technique for extracting values of 0 and Ka , which includes
consideration of activity coecients for the constituent ions on the basis of the
DebyeHuckel law. This technique is useful when highly precise conductivity data
are available.
Many studies of electrolyte conductivity have been carried out [7]. This work
certainly helped to conrm modern ideas about electrolyte solutions. One aspect
of the behavior of strong electrolytes which was initially not well understood is the
fact that their molar conductance decreases with increase in concentration.
Although this is now attributed to ionion interactions, early work by
Arrhenius [8] ascribed the decrease in all electrolytes to partial dissociation.
However, it is clear from the vast body of experimental data that one can distinguish two types of behavior for these systems, namely, that for strong electrolytes
and that for weak electrolytes, as has been illustrated here. The theory of the
concentration dependence of the molar conductance of strong electrolytes was
developed earlier this century and is discussed in detail in the following section.
6.9 The DebyeOnsager Model for Conductivity

In chapter 3, it was shown that the DebyeHuckel theory for ionion interactions
is able to account for solution non-ideality in very dilute systems. The same model
forms the basis for understanding the concentration dependence of the conductance observed for strong electrolytes. Thus, Onsager [9] showed in 1927 that the
limiting conductance law for 11 electrolytes has the form
0 B1 0 B2 c1=2
e
6:9:1
289
where B1 and B2 are constants whose signicance is discussed below. The success
of the DebyeOnsager theory is that it is able to account quantitatively for the
behavior observed empirically by Kohlrausch at low concentrations (equation
(6.8.1)).
The inuence of the interionic forces is due to two phenomena, namely, the
electrophoretic eect and the time-of-relaxation eect. The net ionic atmosphere
around a given ion carries the opposite charge and therefore moves in a direction
opposite to the central ion. The nal result is an increase in the local viscosity, and
retardation of the central ion. This is called the electrophoretic eect. The time-ofrelaxation eect is also related to the fact that the ionic atmosphere around a
given ion is moving and therefore disrupted from its equilibrium conguration. It
follows that the ionic atmosphere must constantly be re-formed from new counter
ions as the ion under observation moves through the solution. The net eect is
that the electrical force on each ion is reduced so that the net forward velocity is
smaller.
The derivation of an expression for the electrophoretic eect is based on the
DebyeHuckel theory described earlier in section 3.8. On the basis of the
Boltzmann distribution law, the local concentration of ion i at a distance r
from a central reference ion j is
cri ci expzi f
6:9:2
ci
where
is the average concentration of ion i, zi , its valence, and , the local
potential. Expanding the exponential as an innite series in zi f, the following
expression is obtained for the deviation of the local concentration from its average
value:
cri ci ci
1
X
1n
n1
n!
zi fn
6:9:3
In keeping with the DebyeHuckel theory, it is assumed that the electrostatic

energy of an individual ion is small with respect to kB T so that only the rst
term in the expansion need be considered; as a result
cri ci zi ci f
6:9:4
If fi is the force acting per mole on ion i in a spherical shell which is at a distance r
from the reference ion, and the shell has a thickness dr , then the incremental force
on the ions of type i in the shell is
dFi 4pzi ci fr2 fi dr
6:9:5
For a binary electrolyte with one type of cation and anion, the incremental force is
dF 4pfr2 z c f z c f dr
6:9:6
Using Stokes law (equation (6.3.5)) this can be related to an incremental change
in the velocity of ion j. Thus,
dvj
2
f rz c f z c f dr
3Z j
6:9:7
290
where j is the local potential due to ion j and Z is the viscosity of the solution.
The net change in velocity is found by integrating this expression from the boundary of ion j at r a to innity. Thus,
1
vj
2fj
z c f z c f r dr
3Z
6:9:8
The expression obtained for j in DebyeHuckel theory (equation (3.8.26)) is

ka
zj e0 ekr
e
j
6:9:9
4p"0 "s r 1 ka
where k is the DebyeHuckel reciprocal distance (equation (3.8.13)).
For simplicity the derivation is restricted here to symmetrical zz electrolytes.
Substituting equation (6.9.9) into equation (6.9.8) for the case that the central ion
is a cation,
1
ka
z2 fe0 ce
e
f f ekr dr
v
6pe0 es Z 1 ka
6:9:10
or, after integration,

v
z2 fe0 ce f f
6pe0 es Z k1 ka
6:9:11
For a symmetrical zz electrolyte the ionic strength is equal to z2 ce . Recalling that
the DebyeHuckel reciprocal distance k is given by

2NL e0 f I 1=2
k
6:9:12
e0 es
the equation for the electrophoretic eect on the velocity of a cation may be
further simplied to obtain
v
kf f
12pNL Z1 ka
6:9:13
The corresponding equation for the case that the central ion is an anion is
v
kf f
12pNL Z1 ka
6:9:14
The forces f and f are given by the product of the charge per mole of ions and
the local eld. The latter is equal to the external applied eld E plus the relaxation
eld E. The relaxation eld arises because of the asymmetry in the ionic atmosphere caused by the motion of the ion with respect to its atmosphere. Thus, for
the cation
f NL z e0 E E z FE E
6:9:15
f z FE E
6:9:16
and for the anion
291
Combining these results with equations (6.9.13) and (6.9.14), one obtains
v v
ze0 kE E
6pZ1 ka
6:9:17
This result may be interpreted as showing that the ionic atmosphere has a charge
z, opposite to that of the ion on which attention is focused. Furthermore, the
atmosphere has an eective radius equal to a 1=k. The motion of the atmosphere is in the opposite direction to that of the ion whose velocity is given by
Stokes law.
The theory of the relaxation eect makes use of the equations of hydrodynamics applied with consideration of the eects of interionic forces. It is the most
dicult part of the description of non-equilibrium processes in electrolytes, and
the details of the derivation are not given here. The result obtained for the relative
value of the relaxation eld is
E
jz z je20
q
k

E
12pe0 es kB T 1 q1=2 1 k0
6:9:18
where
q
jz z j
l0 l0
z jz j z l0 jz jl0
6:9:19
In the case of symmetrical electrolytes for which z jz j, the parameter q is

equal to 1/2.
The limiting ionic mobility for ion i, ui 0 is reduced by an amount ui (relx) as a
result of the eld relaxation eect. Thus, ignoring the electrophoretic eect one
may write

E
ui ui0 1
6:9:20
E
where the term in E=E gives the decrement in ui due to the eld relaxation eect.
From equation (6.9.17), the decrement in ui due to the electrophoretic eect is

vi
zi e0 k
E
1
ui elph
6:9:21
6pZ1 ka
E
E
Combining equations (6.9.20) and (6.9.21), one obtains

zi e0 k
E
1
ui ui0
6pZ1 ka
E
6:9:22
The expression for E=E (equation (6.9.18)) is now substituted into equation
(6.9.22) with the result that
!
zi e0
jz z je20 ui0
q
k
6:9:23
ui ui0
6pZ 12p"0 "s kB T 1 q1=2 1 ka
where the second-order term in E=E has been neglected.
292
Table 6.3. Values of the Debye-Onsager Constants B1

and B2 for 11 Electrolytes Together with the Viscosity
of Water in the Temperature Range 0508C
T/8C
Z=mPa s
B1 =L1=2 mol1=2
B2 =cm2 L1=2 1 mol3=2
0
10
20
25
30
40
50
1.792
1.307
1.002
0.890
0.798
0.653
0.547
0.2208
0.2236
0.2272
0.2292
0.2315
0.2362
0.2416
29.71
40.91
53.65
60.57
67.78
83.39
100.3
The nal result is obtained by multiplying each mobility ui by the Faraday to

obtain the individual ionic equivalent conductances li . The values of li for the
ions in the binary electrolyte are added so that
I1=2
Fu u 0 B1 0 B2
6:9:24
1 BDH aI1=2
where
!1=2
jz z je20
q
2NL e20
B1
6:9:25
12p"0 "s kB T 1 q1=2 "0 "s kB T
and
!1=2
NL z jz je20 2NL e20
B2
6pZ
"0 "s k B T
6:9:26
The DebyeOnsager constants B1 and B2 depend explicitly on temperature, and

also on the solvent parameters es and Z, which themselves are temperature dependent. If a factor of 1000 is included in the terms under the square root sign so that
the ionic strength may be expressed in moles per liter, then B1 and B2 may be
written as
B1
jz z je20
q
B
12p"0 "s kB T 1 q1=2 DH
6:9:27
NL z jz je20
BDH
6pZ
6:9:28
and
B2
where BDH is the DebyeHuckel constant dened in equation (3.8.34). Values of

B1 and B2 for water together with its viscosity in the temperature range from 0 to
508C are given in table 6.3.
EXAMPLE
Estimate the molar conductivity of 0.1 M HCl at 258C given that its limiting
molar conductivity is 426.2 cm2 1 mol1 and assuming an ionic size parameter a of 0.45 nm.
293
The calculation is based on equation (6.9.24) and requires estimation of the

DebyeHuckel factor I 1=2 =1 BDH a I 1=2 ), where the value of BDH is given in
table 3.8; the ionic strength for this system is equal to its concentration so that
I1=2
0:316
0:2153 mol 1=2 L1=2
1 BDH aI1=2 1 3:288 0:45 0:316
6:9:29
The estimate of is
426:2 0:2292 426:2 60:57 0:2153 392:1
6:9:30
Thus, the molar conductance is 392.1 cm2 1 mol1 .

Application of the DebyeOnsager model to conductivity data for NaCl in water
is illustrated in g. 6.10. When the size parameter a is set equal to 0.4 nm, an
excellent t of theory to the experimental data is obtained for concentrations up
to 0.5 M (I1=2 0:7 M1=2 ). At higher concentrations the experimental data fall
below the theoretical curve indicating that the model does not describe the
decrease in with concentration properly in this range. A variety of reasons
can be cited for this failure and many of them have been discussed with respect
to the DebyeHuckel theory for the equilibrium properties of electrolyte solutions. They include the failure to describe properly the thickness of the ionic
atmosphere and the strength of ionion interactions at high ionic strength.
Attempts have been made to extend the concentration range over which equation
(6.9.24) is valid by adding extra terms in the concentration. The interested reader
is referred to more extensive discussions of this subject [7, 10] for further details.
In the limit of very dilute solutions where ion size eects may be neglected,
equation (6.9.24) becomes
Fig. 6.10 Plot of the equivalent conductance of NaCl in water according to equation
(6.9.1) against the square root of the ionic strength at 258C. The solid curve shows the
prediction of the DebyeOnsager equation with a 0:4 nm, and the straight line, the
prediction of this model in the limit that ion size eects may be neglected.
294
0 B1 0 B2 I1=2
6:9:31
This is identical with equation (6.9.1) in the case of 11 electrolytes. The predictions of the limiting law for the NaCl system are also shown in g. 6.10. It is valid
for concentrations up to 0.01 M. The success of the theory is clear from this result.
First of all, it conrms that plots of against the square root of ionic strength
provide a valid route for determining 0 , the equivalent conductance in the limit
of innite dilution. In addition, it explains why the slope of the plot in the dilute
solutions regime depends on the nature of the electrolyte.
The above theory can also be applied to account for the concentration dependence of transport numbers, especially in dilute solutions. Since the transport
number can be dened as a ratio of the equivalent conductance of the given
ion to the total ionic conductance (equation (6.7.6)), it is clear that a non-linear
relationship can be derived describing the concentration dependence using equations (6.9.23) and (6.9.24).
In summary, Onsagers extension of the DebyeHuckel theory to the nonequilibrium properties of electrolyte solutions provides a valuable tool for deriving single ion properties in electrolyte solutions. Examination of the large body of
experimental data for aqueous electrolyte solutions helped conrm the model for
a strong electrolyte. In more recent years, these studies have been extended to
non-aqueous solutions. Results in these systems are discussed in the following
section.
6.10 Transport Phenomena in Non-Aqueous Solutions

In the development of the theory of ionic conductance it has been shown that the
viscosity of the solvent is an important parameter determining ionic mobility.
Initially, conductivity data were only available in water so that attention was
focused on the eects of ionic size, structure, and charge in determining mobility
and its concentration dependence. More recently, data have become available in a
wide variety of non-aqueous solvents [11, 12], that is, in media with a wide range
of permittivities and viscosities. On the basis of these data one may examine in
more detail the role of solvent viscosity in determining the transport properties
of single ions. Values of the limiting ionic molar conductance for selected monovalent cations and anions are summarized in tables 6.4 and 6.5, respectively.
By combining the StokesEinstein equation (equation (6.7.27)) with equation
(6.7.23), the following expression for the limiting ionic molar conductance is
obtained:
li0
z2i F2
6pNL ZrST
6:10:1
where Z is the viscosity of the pure solvent and rST , the eective Stokes radius of
the ion. Rearranging, one obtains the following result for the Stokes radius:
rST
z2i F2
6pNL Zli0
6:10:2
295
Table 6.4 Limiting Ionic Conductances for

Monovalent Cations in Various Solvents at
258C
li0 =cm2 1 mol1
Solvent
Li Na K
Cs TEA TBA
38.7
39.6
17.1
8.8
8.3
14.7
50.1
45.2
20.5
10.5
9.9
18.3
73.5
52.4
23.3
12.1
12.4
19.0
77.3
69.0
26.4
13.8
13.5
21.2
32.7
60.5
29.2
15.1
10.4
21.2
19.5
38.9
19.6
10.9
6.6
72.1
69.5
20.3
26.1
11.5
5.3
13.9
14.5
53.9
8.3
4.3
14.5
74.8
76.8
19.6
25.8
30.0
14.1
5.5
14.6
15.9
56.7
9.3
3.6
16.1
78.9
85.1
23.5
25.2
31.6
14.7
5.7
15.2
17.4
59.7
11.2
4.0
15.7
84.4
87.3
27.9
35.4
15.6
6.2
12.3
4.3
16.9
91.4
85.9
24.7
32.9
36.2
17.3
9.1
18.3
16.1
48.1
13.2
3.9
22.1
67.3
62.4
17.2
22.8
26.6
11.4
6.1
11.5
34.7
9.0
2.8
15.0
Protic
1.
2.
3.
4.
6.
7.
W
MeOH
EtOH
PrOH
F
NMF
Aprotic
8.
9.
10.
12.
13.
14.
15.
16.
17.
18.
20.
21.
22.
AC
AcN
BzN
DMA
DMF
DMSO
HMPA
NMP
NB
NM
PC
TMS
TMU
The product li0 Z is known as the Walden product and is the focus of the discussion in the present section. To a rst approximation, li0 Z should be independent
of solvent nature if the ion moves in the solution without any accompanying
solvent molecules. However, some variation in li0 Z is seen on the basis of an
analysis of the data in tables 6.4 and 6.5 using the viscosities reported in table
6.1[13].
Plots of the Stokes radius for the Na and TEA cations estimated using
equation (6.10.2) in 18 dierent solvents are shown as a function of the eective
radius of the solvent molecule, rs , in g. 6.11. Quite a large change in solvent size
is involved for the solvents considered, from the very small water molecule with a
radius of 137 pm to the very large hexamethylphosphoramide molecule with a
radius of 394 pm. The results are quite striking: the Stokes radius of the TEA
cation is approximately constant and equal to 270 pm, whereas the Stokes radius
of the Na cation clearly increases with solvent size. The line shown for the Na
data in this gure was drawn with unit slope and with an intercept on the y-axis of
102 pm, that is, the crystallographic radius of the Na ion. Although there is
considerable scatter about the line, it is clear that the most important factor
determining the size of the moving ion is the size of the solvent molecules
Table 6.5 Limiting Ionic Conductances

for Monovalent Anions in Various
Solvents at 258C
li0 =cm2 1 mol1
Solvent
Cl
Br
I
76.4
52.1
22.0
10.4
17.3
53.8
78.1
56.4
24.1
12.2
17.5
53.4
76.8
62.6
27.2
13.4
16.8
51.1

ClO
4 BPh4
Protic
1.
2.
3.
4.
6.
7.
W
MeOH
EtOH
PrOH
F
NMF
67.4
70.6
31.8
16.3
16.7
51.6
36.5
20.0
10.7
6.3
AC
105.1 115.9 113.0 115.5
AcN
99.2 100.7 102.5 103.7
DMA
45.9 43.2 41.8 42.8
DMF
53.8 53.4 51.1 51.6
DMSO 24.2 23.8 23.6 24.0
HMPA 19.7 18.9 17.2 16.0
NMP
26.5 28.2 27.0 27.2
NB
22.4 22.0 21.3 21.5
NM
61.9 62.3 63.1 65.3
PC
20.2 18.9 18.5 18.6
TMS
9.3
8.9
7.2
6.7
TMU
30.1 28.8 28.4
66.0
58.0
10.7
6.1
12.2
10.8
32.5
8.4
Aprotic
8.
9.
12.
13.
14.
15.
16.
17.
18.
20.
21.
22.
Fig. 6.11 Plots of the Stokes radii rST for Na and TEA estimated using equation (6.10.2)
against the eective solvent radius rs . The data for TEA have been shifted vertically by
300 pm for the sake of clarity.
297
which accompany it. The scatter can be attributed to several factors, one of them
being the obvious fact that not all solvent molecules are perfectly spherical.
However, other factors also play a role. For example, the three solvents EtOH,
AcN, and NM have radii which are close to 216 pm. The Stokes radius for Na in
NM is smallest (237 pm), that in AcN intermediate (313 pm), and that in EtOH
largest (371 pm). This follows the increase in solvent donicity. Since NM is a very
weak Lewis base, the Na ion holds fewer solvent molecules on the average when
it moves, and therefore has a smaller radius. On the other hand, EtOH is a
stronger Lewis base so that the moving ion has a larger radius reecting the
fact that more solvent molecules move with it. TEA is a much larger ion than
Na with a tetrahedral structure and an approximate radius of 337 pm. For this
cation, the Stokes radius is independent of solvent nature with an average value
equal to 270 pm. This result indicates the eective size of the ion when it moves as
an unsolvated species.
The results for the Stokes radii of anions are more complex and are considered
separately for protic and aprotic solvents. In the case of aprotic solvents, the value
of rST decreases in the aprotic solvents with increase in solvent radius (g. 6.12).
Since the interaction between the anion and solvent is weak, the number of
solvent molecules which move with the ion decreases with increase in solvent

size. This trend is weaker for the larger ClO
4 ion than for the smaller Cl ion,

demonstrating that the Cl anion is more strongly solvated in solvents of the
smallest molecular size. In the case of protic solvents, hydrogen bonding is
involved in anion solvation. Data are available in only a few solvents so that a
detailed analysis is not possible at present.
The results presented here demonstrate clearly that the solvent radius is the
important parameter to be considered in discussing the solvent dependence of the
Fig. 6.12 Plots of the Stokes radii for Cl and ClO
4 ions in aprotic solvents estimated
using equation (6.10.2) against the eective solvent radius rs . The data for ClO
4 have
shifted vertically by 300 pm for the sake of clarity.
298
Stokes radius or of the Walden product. Ion size and structure both determine the
extent to which the ion moves with associated solvent molecules.
6.11 Proton Transport Phenomena

The proton is formed from the lightest element in the periodic table and has the
smallest ionic size. It also has a special relationship to the most common polar
solvent, namely, water. These two facts combine to give this ion the highest
mobility in water, approximately 10 times that of the Li ion, which is the next
larger monoatomic cation. In order to understand these observations one must
develop a picture of how the proton is incorporated in the structure of protic
solvents.
Using the analysis of the thermodynamic properties of monoatomic ions in
water presented earlier (section 3.5), the Gibbs energy of solvation of the proton
at innite dilution and 258C is 1104 kJ mol1 . On the basis of the MSA, using
the parameters for monoatomic cations given in table 3.5, the corresponding
radius of the H ion is 13 pm. Thus, the proton radius is close to one-sixth of
the radius of the Li ion (74 pm). This in turn means that the electrical eld due to
the H ion is approximately six times larger than that due to the Li ion. As a
result, the interaction between the proton and surrounding solvent dipoles is
extremely large compared to ions of more usual dimensions, that is, with radii
greater than 100 pm. In the case of water, a covalent bond is formed by interaction of the proton with the lone pair electron density on an adjacent water molecule. The existence of the hydronium ion, H3 O , was demonstrated using infrared
spectroscopy by Falk and Gigue`re [14]. It has also been observed in the solid state
in compounds such as HClO4
H2 O using X-ray crystallography. In dilute aqueous
solutions at equilibrium, the hydronium ion is solvated by other water molecules.
Considering the structure of water, the most probable local species corresponds to
H9 O
4.
The mechanism by which the hydronium ion moves in water is very dierent
than that by which most other ions move. As shown in g. 6.13, the conductance
process involves the transfer of a proton between two adjacent water molecules
with the proton in a hydrogen bond between these molecules. Because of the
relative orientation of the water molecules with respect to one another, the net
motion of the proton is not exactly in the direction of the electrical eld but rather
in zig-zag motion about this direction. This feature of the net forward motion can
also be described in terms of rotation of the water molecule to which the proton is
transferred so that the orbitals on this molecule are favorably oriented with
respect to the direction of the eld. The net eect is that the value of l0 at
258C is 349.8 cm2 1 mol1 , that is, much higher than any other monovalent ion.
The hydroxyl ion moves by a similar mechanism, which involves proton transfer in the opposite direction of the electrical eld. The molar conductance of the
OH ion is smaller (197.6 cm2 1 mol1 ) simply because this ion occupies more
space in the water structure. In addition, the eld due to OH is much less than
that due to H because of the much larger size of the former ion. Anomalous
proton transfer also occurs in the lower alcohols and their mixtures with water.
299
Fig. 6.13 Schematic diagram illustrating

the mechanism for proton and hydroxyl
ion conductance in water.
This also reects the fact that proton transfer in a solvent such as methanol
involves the species CH3 OH
2 . However, in the alcohols there is only one site
for hydrogen bond formation, whereas the water molecule has two. Thus orientational eects are more important in acidied alcohol systems, so that the enhancement of proton mobility with respect to other monovalent cations is not as great
as it is in water.
An important question regarding proton transfer in water is whether it occurs
by a classical mechanism or by quantum-mechanical tunneling. This problem can
be elucidated by comparing the rates of proton and deuteron transfer. In this way
it was concluded that transfer occurs by both mechanisms [15].
Another way of examining the mechanism of proton transfer is by comparing
its temperature dependence with that of other monoatomic cations. Plots of the
molar conductance of three cations, namely, H , Li , and K , on a logarithmic
scale are shown as a function of reciprocal temperature in the range 5558C in g.
6.14. Excellent linear plots are found from which an energy barrier associated
with the process may be calculated. Using a simple Arrhenius expression, the
temperature dependence of li for a small temperature interval is given by
li Al exp

Ul
RT
6:10:3
where Al is the pre-exponential factor, and Ul , the activation energy associated

with the conductance process. From the data presented in g. 6.14, the value of
Ul is 10.0 kJ mol1 for the proton; on the other hand, it is 16.3 kJ mol1 for Li ,
and 14.2 kJ mol1 for K . This type of analysis certainly oversimplies the
physical picture of proton transport because several mechanisms are involved.
However, it clearly shows that the net eect for the proton is that the average
activation barrier is much less than that for other cations.
300
Fig. 6.14 Plot of the molar conductance for H , Li , and K ions against the reciprocal
temperature in the range 5558C.
The study of proton transfer in solutions is a rich area of research [16] which
has only been briey introduced by the present discussion. It is obviously important not only in conduction processes but also in acidbase processes in solution.
More information about these reactions is given in chapter 7.

The discussion in this chapter has emphasized the Onsager approach to non-equilibrium phenomena. This provides an excellent background for understanding the
way in which the uxes observed experimentally often depend on several forces in
the system. Early experimental work dealt mainly with electrolyte solutions.
Conductivity and diusion measurements in these systems were very important
in developing a microscopic picture of the structure of electrolyte solutions.
Electrolyte solutions are non-ideal, and the non-ideality is reected in non-equilibrium processes by the concentration dependence of the phenomenological coecients. The work by Debye and Onsager has shown that the concepts developed to
understand non-ideality in electrolyte solutions at equilibrium are applicable to the
same systems, when they are not at equilibrium. The DebyeOnsager theory
described here is currently being extended to include a more careful description
of ionic atmosphere eects on the basis of the MSA which was introduced in
chapter 3.
Mass transfer and heat transfer are important subjects in engineering. Whole
monographs have been devoted to solutions of the pertinent dierential equations
for a variety of boundary conditions [G2, G3]. Only one example is worked out in
this chapter to give an idea of what is involved mathematically. Mass transfer
problems make up an important part of electroanalytical chemistry. In a typical
experiment the current which ows in the electrochemical cell which is not at
301
equilibrium is used to determine the amount of analyte. This species reacts at the
working electrode and the observed current depends not only on the concentration of the analyte but also on the mass transfer conditions by which it reaches the
reaction site at the electrode. Much more can be found about this subject and the
associated mass transfer problems in monographs devoted to electroanalysis [G4].
General References
G1. Katchalsky, A.; Curran, P. F. Non-Equilibrium Thermodynamics in Biophysics;
Harvard University Press: Cambridge, MA, 1967.
G2. Jost, W. Diusion in Solids, Liquids, Gases; Academic Press: New York, 1960.
G3. Bird, R. B.; Stewart, W. E.; Lightfoot, E. N. Transport Phenomena; Wiley: New York,
1960.
G4. Galus, Z. Fundamentals of Electrochemical Analysis; Ellis Horwood: New York, 1994.
References
1. Onsager, L. Phys. Rev. 1931, 37, 405; 1931, 38, 2265.
2. de Groot, S. R. Thermodynamics of Irreversible Processes; North-Holland:
Amsterdam, 1952.
3. McQuarrie, D. A. Statistical Mechanics; Harper and Row; New York, 1976.
4. Riddick, J. A.; Bunger, W. B.; Sakano, T. K. Organic Solvents, 4th ed.; Wiley: New
York, 1986.
5. Debye, P. Polar Molecules; Chemical Catalogue Co.: New York, 1929.
6. Parsons, R. Modern Aspects of Electrochemistry; Bockris, J. OM., Conway, B. E.,
Eds.; Butterworths: London, 1954; Chapter 3.
7. Robinson, R. A.; Stokes, R. H. Electrolyte Solutions, 2nd ed; Butterworths: London,
1959.
8. Arrhenius, S. Z. Phys. Chem. 1887, 1, 631.
9. Onsager, L. Phys. Z. 1927, 28, 277.
10. Justice, J. C. In Comprehensive Treatise of Electrochemistry; Conway, B. E., Bockris,
J. OM., Yeager, E., eds.; Plenum Press: New York, 1983; Vol. 5, Chapter 3.
11. Karapetyan, Yu. A.; Eychis, V. N. Physico-chemical Properties of Non-Aqueous
Electrolyte Solutions (in Russian); Khimia: Moscow, 1989.
12. Barthel, J.; Neueder, R.; Schroder, P. Electrolyte Data Collection; Dechema:
Frankfurt, 1992; Vol. XII, Part 1; 1993; Vol. XII, Part 1a; 1994; Vol. XII, Part 1b;
1996; Vol. XII, Part 1c.
13. Fawcett, W. R. Mol. Phys. 1998, 95, 507.
14. Falk, M.; Gigue`re, P. Can. J. Chem. 1957, 35, 1195.
15. Conway, B. E.; Bockris, J. OM., Linton, H. J. Chem. Phys. 1956, 24, 834.
16. Lengyel, S.; Conway, B. E. In Comprehensive Treatise of Electrochemistry; Conway, B.
E., Bockris, J. OM., Yeager, E., Eds.; Plenum Press: New York, 1983; Vol. 5,
Chapter 5.
Problems
1. The viscosity of DMSO is 1.996 mPa s at 258C. Estimate its Debye relaxation
time assuming a molecular diameter equal to 491 pm. Compare your answer
with the experimental value, which is 18.9 ps.
2. Ficks law for diusion to a sphere of radius r0 is written
302
"
#
@cr; t
@2 cr; t 2 @cr; t
D
@t
r @r
@r2
Solve this expression for cr; t with the conditions

cr; 0 c ; cr0 ; t 0t > 0; lim cr; t c r ! 1
Show that the ux J at r r0 follows the expression

1
1
J Dc
r0 pDt1=2
(Hint: By making the substitution vr; t r cr; t in Ficks equation and in
the boundary conditions, the problem becomes essentially the same as that
for linear diusion.)
3. Using the limiting ionic conductances recorded in table 6.4 for the alkali
metal ions in acetonitrile, estimate their diusion coecients and Stokes
radii.
4. The following parameters apply to mass transfer in 0.01 M NaCl:
D 1:545 105 cm2 s1
118:53 cm2 1 mol1
t 0:3918
Assume that the coecient
e given in equation (6.7.46) can be estimated
using the DebyeHuckel theory with an ion size parameter equal to 0.42 nm.
Calculate values for the phenomenological coecients LMM , LXX , and LMX
using equations (6.7.48)(6.7.50).
5. The following data are reported for the conductivity of solutions of chloroacetic acid in water at 258C.
Conc./M
/cm2 1 mol1
1.101104
3.027104
5.899104
1.323103
2.821103
3.812103
7.462103
1.4043102
362.10
328.92
295.58
246.15
197.14
177.98
139.85
109.00
Determine the acidity constant for this acid and its limiting ionic conductivity.
6. Determine the best value of the ion size parameter for LiNO3 given the
following data obtained at 188C. At this temperature the DebyeHuckel
303
BDH is 3.274 M1=2 nm1 , and the DebyeOnsager constants are B1

0:2264 M1=2 and B2 50:99 cm2 M1=2 1 mol1 .
ce /M
=cm2 1 mol1
0.001
0.005
0.01
0.05
0.1
0.5
1.0
92.9
90.3
88.6
82.7
79.2
68.0
60.8
7. Estimate the missing data points for Cs in table 6.4 using a linear correlation
between the Stokes radius and the solvent radius as illustrated in g. 6.11.
8. The following data were obtained for ZnSO4 in water at 258C.
ce /M
=cm2 1 mol1
2104
5104
1103
2103
5103
0.01
0.02
0.05
0.1
242.0
234.0
225.0
212.0
190.0
169.0
146.4
121.4
105.0
(a) Determine the limiting ionic conductance assuming that ZnSO4 is a

strong electrolyte
(b) Determine this parameter assuming that ZnSO4 can be considered a
weak electrolyte due to strong ion pairing.
(c) Comment on these results using the parameters given in table 6.2.
Chemical Reaction Kinetics

in Solution
Rudolph (Rudy) Marcus was born in 1923 at

Montreal, Quebec, Canada, where he also grew
up and went to school. He entered McGill
University to study chemistry, obtaining a BSc in
1943 and a Ph.D. in 1946. His doctoral research
under Professor Carl Winkler was in the area of
chemical kinetics. He then spent two years as a
post-doctoral fellow at the National Research
Council in Ottawa, Canada where he worked in
the photochemistry laboratory of E.W.R.
Steacie. As a young student, Marcus had strong
Rudolph Arthur Marcus
interests in both mathematics and chemistry. In
order to develop these interests further he joined
the research group of Professor Oscar Rice at the University of North Carolina
in 1949. There he worked on the theory of unimolecular reactions becoming
one of the contributors to what is now known as RRKM (RiceRamsperger
KasselMarcus) theory. In 1951, Marcus joined the faculty of the Polytechnic
Institute of Brooklyn as Assistant Professor. It was during his years in
Brooklyn that he developed the now famous theory of electron transfer. He has
subsequently held faculty positions at the University of Illinois and, at
California Institute of Technology where he presently holds the Arthur Amos
Noyes Chair of Chemistry. His work on the theory of chemical kinetics has
also involved proton and atom transfer reactions as well as reactions at
interfaces. In 1992, Marcus was awarded the Nobel Prize in Chemistry for his
contributions to the theory of electron transfer reactions in chemical systems.
He has received numerous other honors including the Peter Debye award of
the American Chemical Society and foreign membership in the Royal Society
(London).
7.1 What Time Scales Are Involved for Chemical

Reactions in Solution?
The kinetics of chemical reactions were rst studied in liquid solutions. These
experiments involved mixing two liquids and following the change in the concentration of a reactant or product with time. The concentration was monitored by
removing a small sample of the solution and stopping the reaction, for example,
304
CHEMICAL REACTION KINETICS IN SOLUTION
305
by rapidly lowering the temperature, or by following a physical property of the

system in situ, for example, its color. Although the experiments were initially
limited to slow reactions, they established the basic laws governing the rate at
which chemical changes occur. The variables considered included the concentrations of the reactants and of the products, the temperature, and the pressure.
Thus, the reacting system was examined using the variables normally considered
for a system at equilibrium. Most reactions were found to be complex, that is, to
be made up of several elementary steps which involved one or two reactants.
As the fundamental concepts of chemical kinetics developed, there was a strong
interest in studying chemical reactions in the gas phase. At low pressures the
reacting molecules in a gaseous solution are far from one another, and the theoretical description of equilibrium thermodynamic properties was well developed.
Thus, the kinetic theory of gases and collision processes was applied rst to
construct a model for chemical reaction kinetics. This was followed by transition
state theory and a more detailed understanding of elementary reactions on the
basis of quantum mechanics. Eventually, these concepts were applied to reactions
in liquid solutions with consideration of the role of the non-reacting medium, that
is, the solvent.
An important turning point in reaction kinetics was the development of experimental techniques for studying fast reactions in solution. The rst of these was
based on ow techniques and extended the time range over which chemical
changes could be observed from a few seconds down to a few milliseconds.
This was followed by the development of a variety of relaxation techniques,
including the temperature jump, pressure jump, and electrical eld jump methods.
In this way, the time for experimental observation was extended below the nanosecond range. Thus, relaxation techniques can be used to study processes whose
half lives fall between the range available to classical experiments and that characteristic of spectroscopic techniques.
The experimental techniques for studying fast reactions provided a means of
studying fundamental processes in solution that were previously considered to be
instantaneous. These include electron and proton transfer reactions. Proton
transfer is the elementary step involved in acidbase reactions, which are so
important in classical analytical chemistry. On the other hand, electron transfer
is the elementary step involved in redox reactions. The theory of electron transfer
is especially well developed and is discussed in detail below.
This chapter is mainly concerned with fast reactions and the experimental
methods used to study them. Other than the relaxation techniques already mentioned, spectroscopic methods applied to the study of elementary reactions are
outlined. Especially important in this regard are laser methods which are able to
probe fundamental processes in solution in the femtosecond time range.
7.2 Fundamental Concepts

The experimental study of chemical kinetics traditionally involves the measurement of the concentration of a reactant or a product of the reaction as a function
of time in a homogeneous system. When the experimental data are plotted, one
306
usually nds that the slope of the plot, that is, the rate of reaction, decreases as
time increases. On the basis of a number of dierent experiments, the conclusion
was reached that the rate of reaction could be related to the concentration of one
reactant in a fairly simple way. Thus, for the irreversible reaction
A ! products
7:2:1
the rate of reaction is often given by the general expression

dcA
kr cnA
dt
7:2:2
where dcA =dt is the rate of reaction with respect to the reactant A, cA, its
concentration, n, the order of the reaction, and kr, the rate constant. Typical
values of n are 1 corresponding to a rst-order reaction and 2, corresponding
to a second-order reaction. However, other values are found experimentally
including fractional values such as 1/2 and 3/2. In these cases, a complex mechanism is involved.
Experimental determination of the reaction order can be made in several ways
[G1, G2]. Given values of cA as a function of time, one can estimate dcA/dt by
numerical dierentiation. Then, the reaction order can be estimated on the basis
of a plot of the logarithm of the reaction rate against the logarithm of the reactant
concentration. Thus, from equation (7.2.2)

dcA
ln
ln kr n ln cA
7:2:3
dt
The order of the reaction is estimated from the slope of the plot and the rate
constant from the intercept.
Since estimation of the rate of reaction from the experimental data involves
amplication of experimental error, an improved estimate of the rate constant is
usually sought using an integrated version of the general expression for the rate of
reaction once it is known. For a rst-order reaction (n 1), integration of equation (7.2.2) gives the result
ln cA ln cA0 kr t
7:2:4
where cA0 is the initial concentration of reactant A. It follows that a plot of ln cA

against time t should be linear, and that the rate constant kr can be determined
from the slope of the plot. If the reaction is second order (n 2), the relationship
resulting from integration of equation (7.2.2) is
1
1
kr t
cA cA0
7:2:5
Thus, a plot of 1/cA should be linear in time t with a slope equal to the rate
constant kr. Once the order of reaction is known, equation (7.2.2) can be integrated to obtain a relationship between cA and t. Using this relationship the data
may be analyzed further to obtain the rate constant kr .
Most chemical reactions involve more than one reactant. For example, for an
irreversible reaction involving two reactants such as
aA bB ! products
7:2:6
307
the rate of reaction can be written as

1 dcA
1 dcB
n

kr cm
A cB
a dt
b dt
7:2:7
where a and b are the stoichiometric coecients for reactants A and B, respectively, and m and n, the corresponding orders with respect to each of these
reactants. The overall order is m n. The simplest case for this general system
occurs when the reaction is rst order in A, and rst order in B with an overall
order of two. When the stoichiometric coecients are also unity one can write
dcA
dc
B k r cA cB
dt
dt
7:2:8
Integration of this equation to give an expression for cA or cB as a function of time

is very simple in two limiting cases. If the initial concentration of A is equal to that
of B, then cA is equal to cB for all values of time and equation (7.2.8) reduces to
equation (7.2.2) for the case that n 2. The integrated result is then given by
equation (7.2.5). The other limiting case obtains when the initial concentration of
one reactant is much greater than that of the other, so that the change in concentration of the dominant reactant is negligible. If reactant B is in excess, then
integration of equation (7.2.8) gives the result
ln cA ln cA0 kr cB t
7:2:9
By performing experiments with the concentration of one reactant much smaller

than the others, the order with respect to the minority component is easily found.
This is called the method of isolation.
Integration of equation (7.2.8) for the general case of non-equal concentrations
of A and B is best carried out using a variable x which is equal to the amount of A
or B which has been consumed. Thus, the dierential equation is rewritten as
dcA
dc
dx
kr cA0 xcB0 x
B
dt
dt
dt
7:2:10
where cA0 and cB0 are the initial concentrations of A and B, respectively. Using
the method of partial fractions this becomes

dx
1
dx
dx

kr dt
7:2:11
cA0 xcB0 x cB0 cA0 cA0 x cB0 x
After integration, one obtains

c x
c
ln B0
ln B0 cB0 cA0 kr t
cA0 x
cA0
or

ln

cB
c
ln B0 cB0 cA0 kr t
cA
cA0
7:2:12
7:2:13
This equation describes the time dependence of the two concentrations cA and
cB for all cases except the case that cA0 is equal to cB0.
308
Many elementary reactions are reversible so that an analysis of the kinetics

must consider both the forward and reverse processes. A simple example is a
reversible reaction consisting of two rst-order reactions, which is described as
(B
A+
7:2:14
If the rate constant for the forward step is k1 and that for the backward step, k1,
then the rate of consumption of A is
dcA
k1 cA k1 cB
dt
7:2:15
On the basis of mass balance,

cA cB cA0
7:2:16
where cA0 is the initial concentration of A. Thus, equation (7.2.15) can be rewritten as
dcA
k1 cA k1 cA0 k1 cA
dt
7:2:17
This can be rearranged to give the dierential equation

k1 k1 dcA
k1 k1 dt
k1 k1 cA k1 cA0
7:2:18
After integration and further rearrangement, the relationship giving cA as a

function of time is
cA
k1 cA0 k1 cA0 expk1 k1 t

k1 k1
7:2:19
When the system is at equilibrium (t 1), the concentration of A is

cAe
k1 cA0
k1 k1
7:2:20
cBe
k1 cA0
k1 k1
7:2:21
and that of B,
The equilibrium constant in terms of concentration is

Ke
cBe
k
1
cAe k1
7:2:22
This is an important result linking chemical kinetics and thermodynamics.

The concentration change occurring for a rst-order reaction with a rate constant of 1 s1 is illustrated in g. 7.1. At 7 s, the concentration of the reactant falls
below one-thousandth of its original value. It continues to fall, reaching concentrations which are presumably too small to be detected experimentally. When the
reaction is reversible with a reverse rate constant one hundred times slower than
the forward reaction (k1 0:01 s1), the concentration of the reactant reaches its
equilibrium value in approximately 10 s. In this case, the behavior of the concentration against time is noticeably dierent.
309
Fig. 7.1 Plots of the logarithm of the concentration of reactant A against time for a rstorder reaction with a rate constant equal to 1 s1 and a reversible rst-order reaction with
the same forward rate constant and a reverse rate constant equal to 0.01 s1. The initial
concentration of A is 1 M.
Another important feature of many mechanisms is sequential steps. This is

illustrated here with the simple example of two rst-order reactions, one following
another, that is
k1
k2
A ! B ! C
7:2:23
If the rate constant for the rst reaction is k1, then the concentration of A is given
by
cA cA0 ek1 t
7:2:24
where cA0 is its initial concentration. The dierential equation giving the concentration of B is
dcB
k1 cA k2 cB
dt
7:2:25
where k2 is the rate constant for the second step. Substituting equation (7.2.24)
into equation (7.2.25) and multiplying through by the factor ek2t, one obtains
ek2 t
dcB
k1 cA0 ek2 k1 t k2 cB ek2 t
dt
7:2:26
d
c ek2 t k1 cA0 ek2 k1 t
dt B
7:2:27
or
This is now easily integrated with the result

cB
k1 cA0 k1 t
e
ek2 t
k2 k1
7:2:28
310
where the initial concentration of B is assumed to be zero. The concentration of C

is now found by subtraction:
cC cA0 cA cB cA0 cA0 ek1 t
k1 cA0 k1 t
e
ek2 t
k2 k1
7:2:29
A plot of the concentrations of A, B, and C against time is shown in g. 7.2 for

the case that the rate constant for the second step is much greater than that of the
rst (k1 k2 ). Under these conditions the concentration of B does not reach very
high values during the course of the reaction. For smaller values of k2, the intermediate B reaches higher concentrations which represent a signicant fraction of
the number of moles of the reactant A at the beginning of the experiment (cA0).
When the concentration of B is much smaller than that of A and C, so that any
change in its concentration may be neglected, it can be assumed to be in a steady
state. The steady-state assumption is an important approximation used in the
analysis of the kinetics of complex reactions which results in much more simple
rate expressions. In the present case, the steady-state approximation for the
reactive intermediate B is
dcB
0
dt
7:2:30
It follows from equation (7.2.25) that

cBss
k 1 cA k 1
cA0 ek1 t
k2
k2
7:2:31
For the product C
Fig. 7.2 Plots of the concentration of A, B, and C against time for a consecutive reaction
scheme with irreversible rst-order processes. The solid lines show the results based on the
exact solution assuming k1 1 s1 and k2 100 s1; the points were calculated on the basis
of the steady-state approximation (equations (7.2.31) and (7.2.32)).
cCss cA0 cA0 ek1 t
k1
c ek1 t
k2 A0
311
7:2:32
When the steady-state estimates of cB and cC (equations (7.2.31) and (7.2.32)) are
compared with the exact values (equations (7.2.28) and (7.2.29)) it is apparent that
two conditions are involved, namely, k2 k1 and k2 t 1. Values of cB and cC
obtained on the basis of the steady-state approximation are also shown in g. 7.2.
It is clear that the steady-state values provide excellent estimates of the concentrations of these species at longer times. However, when k2 t < 1, the steady-state
estimates of cB and cC are incorrect.
The examples of reversible and consecutive reactions presented here give a very
modest introduction to the subject of reaction mechanisms. Most reactions are
complex, consisting of more than one elementary step. An elementary step is a
unimolecular or bimolecular process which is assumed to describe what happens
in the reaction on a molecular level. In the gas phase there are some examples of
termolecular processes in which three particles meet simultaneously to undergo a
reaction but the probability of such an event in a liquid solution is virtually zero.
A detailed list of the elementary steps involved in a reaction is called the reaction
mechanism.
In concluding this section, the reaction mechanism involved in a very important solution reaction, namely, enzyme catalysis, is considered. In this reaction,
the enzyme E coordinates with a substrate species S to form a reactive intermediate ES. A very small fraction of the intermediate reacts to form product P releasing the enzyme for further catalysis. The majority of the reactive intermediate
reverts to the substrate and enzyme. Thus, the mechanism can be described as
k1
E S ! ES
7:2:33
k1
7:2:34
k2
7:2:35
ES ! E S
and
ES ! P E
Assuming that the concentration of the reactive intermediate is given the steadystate approximation, one has
dcES
0 k1 cE cs k1 cES k2 cES
dt
7:2:36
Solving for the concentration of ES at steady state, one obtains

cESss
k 1 cE cS
k1 k2
7:3:37
Since the enzyme is a catalyst its total concentration does not change so that
dcE cES dcE
0
dt
dt
7:2:38
The rate of the overall reaction can be expressed in terms of the disappearance of
substrate S,
312
dcS
k1 cE cS k1 cES
dt
7:2:39
or in terms of the formation of product P,

dcP
k2 cES
dt
7:2:40
Combining equations (7.2.37) and (7.2.40), one obtains for the steady state
dcP k1 k2 cE cS
dt
k1 k2
7:2:41
The steady-state concentration of enzyme may be related to its initial concentration as follows:
cEss cE0 cESss cE0
k1 cE cS
k1 k2
7:2:42
where cE0 is the initial enzyme concentration. Simplifying, one nds that
cEss
k1 k2 cE0
k1 k2 k1 cS
7:2:43
Thus, the expression for the rate of reaction becomes

dcP
k1 k2 cS cE0
dt
k1 k2 k1 cS
7:2:44
Since the steady-state approximation requires that k2 k1 , this expression

simplies to
dcS
dc
k k c c
P 1 2 S E0
dt
dt
k1 k1 cS
7:2:45
This result was rst derived by Michaelis and Menten [1].

This section has provided only a brief introduction to the principal concepts in
chemical kinetics. Most of the discussion which follows in this chapter relates to
fast reactions in solution. Before describing the experimental techniques used to
study fast reactions, the important types of solution reactions are outlined and
their properties discussed.
7.3 General Types of Solution Reactions

Important solution reactions are well known from the study of quantitative analysis in analytical chemistry. These include acidbase reactions, redox reactions,
and complex formation reactions. Often the reaction is so fast that the system is
considered to come instantaneously to equilibrium, for example, in the acid
base reaction involved in a titration. In fact, any of these reactions has a nite rate
whose kinetics can be determined using modern experimental techniques.
The above reactions may be described in a more fundamental way in terms of
the elementary act involved in the reaction. For example, the reaction between an
acid and a base involves proton transfer so that it is better described as a proton
313
transfer reaction. Redox reactions involve electron transfer between at least two
components in the system and are better described as electron transfer processes.
Complex formation reactions are really ligand exchange reactions. Fundamental
processes such as electron and proton transfer are of great interest to theoreticians, and have been studied in great detail experimentally. Some of the important
properties of fundamental solution reactions are now outlined.
Bimolecular reactions in the gas phase and in liquid solutions dier in important ways. In the gas phase the progress of the reaction is described in terms of the
collisions which occur between the two reactants A and B. In the liquid solution
the same reactants are surrounded by solvent molecules, and the approach of
these species to each other is controlled by the laws of diusion. If A and B collide
in solution, it is highly probable that they undergo numerous successive collisions
before reaction occurs. Such a set of successive collisions is referred to as an
encounter. The fact that the surrounding solvent molecules can transfer energy
to or from A and B during the course of a solution reaction is important in
developing the theory of reactions in condensed phases. A consequence is that
the couse of the reaction is described in terms of the Gibbs energy of the system
rather than by the potential energy, as is the case for gas phase reactions. More is
said about this aspect of solution reactions in the presentation of transition state
theory in section 7.4
In a bimolecular solution reaction, the reactants A and B diuse to a point
close to one another at which reaction is possible. This process is called formation
of the precursor complex. At this point, rearrangement of bond lengths and bond
angles in the two reactants, and of the surrounding solvent molecules, can occur
to form an activated complex or transition state between the reactants and products. As one would expect, the nature of this process depends on the specic
reaction involved. It is the focus of the development of the theory of the elementary step in the reaction and the associated energy requirements. In some cases it
has been studied experimentally using very fast laser spectroscopic techniques
which provide time-resolved information about the elementary step in the femtosecond range.
In the case of very fast bimolecular reactions, the contributions of both the
diusion process and the rate of decomposition of the precursor complex must be
considered in analyzing the experimental observations. The overall reaction
mechanism can be described as
A B ! AB
AB ! C D
kdiff
7:3:1
kr
7:3:2
where A and B are the reactants, AB, the precursor complex, and C D, the
products. The rate constant for formation of the precursor complex, kdi, depends
on the diusion coecients of A and B. Noyes [2] showed that the overall rate
constant, taking into account the diusion process, is
knet
kdiff kr
kdiff kr
7:3:3
314
In the case of slow reactions for which kr kdiff , the overall rate constant knet is
equal to the rate constant for transformation of the precursor complex, kr.
However, for very fast reactions, kr is greater than kdi, so that the overall rate
constant is close to kdi. Therefore, it is important to understand that there is an
upper limit on reaction rates that can be measured experimentally for diusion
controlled processes. In section 7.5 it is shown that kdi for molecular reactants is
given by
kdiff 4pNL DA DB a
7:3:4
where DA and DB are the diusion coecients for A and B in the given medium,
and a, the distance between their centers in the precursor complex. Assuming that
DA and DB are each equal to 109 m2 s1 and a, 500 pm, the estimate of kdi is
8106 m3 mol1 s1 or 8109 M1 s1. For reactions involving ions, the estimate
is larger or smaller depending on whether the ions attract one another, or repel
one another, respectively. The rate of the diusion process is often comparable to
the rate of electron and proton transfer reactions in solution. Therefore, it must be
considered in analyzing the experimental data.
The distinctive features of the elementary step involved in the important classes
of solution reactions are now considered in more detail.
A. Electron Transfer Reactions
Electron transfer reactions are classied as simple or complex on the basis of the
number of electrons transferred and whether or not chemical bonds are broken or
formed in the overall reaction. An example of a complex electron transfer reaction
is
2
2
3
S2 O2
8 2FeH2 O6 ! 2SO4 2FeH2 O6
7:3:5
In this reaction two electrons are transferred and the peroxide bond in the peroxydisulfate anion is broken. It is possible that this reaction actually involves
three elementary steps. In the rst step an electron is transferred from
Fe(H2O)62 to S2O82 to form a radical anion S2O83 . The radical then breaks

. The nal step involves the
up into the anion SO2
4 and the radical anion SO4
transfer of the second electron to the radical anion SO4 .
An example of a simple electron transfer reaction is
2
3
2
CrH2 O3
6 FeH2 O6 !FeH2 O6 CrH2 O6
7:3:6
In this case, one electron is transferred from Fe(II) to Cr(III), and no bonds are
broken or formed. The reaction is also classied as heteronuclear, since the reactants belong to dierent redox couples. This means that the overall reaction
involves a net change in the standard Gibbs energy, rG . There is considerable
interest also in homonuclear reactions such as

3
3
2
FeH2 O2
6 FeH2 O6 ! FeH2 O6 FeH2 O6
7:3:7
These systems are studied using radioactive isotopes designated by the asterisk (*)
and provide an example of an electron transfer reaction for which rGo is zero.
315
Much of the early work on electron transfer reactions focused on systems

involving transition metal ions. Many redox couples based on the Fe(III)/
Fe(II), Co(III)/Co(II), and Cr(III)/Cr(II) can be prepared in the laboratory by
changing the ligands coordinated to the central metal ion. From these studies it is
possible to distinguish between two types of reactions which would otherwise be
classied as simple. A classic example from the work of Taube [3] is the reaction
2
CoNH3 5 Cl2 CrH2 O2
6 ! CoNH3 5 H2 O CrH2 O5 Cl
7:3:8
In this reaction the Cl ligand is transferred from the substitution-inert coordination sphere of the Co(III) complex to the substitution-inert coordination sphere of
the Cr(III) product. On the other hand, the hexaquo Cr(II) complex is substitution-labile. Thus, it was postulated that the Cl ligand is shared between Cr(II) and
Co(III) in the transition state for the reaction in which one water ligand is displaced from the substitution-labile Cr(II) reactant. Accordingly, the activated
complex is [(NH3)5Co . . . Cl . . . Cr(H2O)5]4+. After electron transfer, the Co(II)
product is substitution-labile so that the Cl ligand remains with the substitutioninert Cr(III) product. This type of electron transfer reaction follows the inner
sphere mechanism. The bridging Cl ion reduces the repulsion between the cations
involved in electron transfer and thus lowers the energy required to reach the
transition state. It may also provide orbitals of the appropriate symmetry to delocalize metalion electrons and thereby provide a continuous path of good orbital
overlap between the two metal ions. Much more about the inner sphere mechanism
of electron transfer can be found in the work of Henry Taube [3].
In the outer sphere mechanism of electron transfer the ligands around each
reactant remain intact in the passage through the transition state. This mechanism
is expected when both reactants are substitution-inert. Formation of the activated
complex may involve a change in bond lengths and bond angles within the reactants but it does not involve the sharing of ligands. The theory of electron transfer
by the outer sphere mechanism has been developed in some detail and is discussed
later in section 7.8. An important aspect of this theory concerns the relative
velocities of the electron being transferred and the nuclei in the reactants. The
transfer of the electron from an orbital on one reactant to an orbital on the other
takes place in a time which is the order of femtoseconds ( 1015 s). Nuclear
motion is much slower and is characterized by the time associated with intramolecular vibration, that is, 100 fs ( 1013 s). As a result, the nuclei in the reactants
are virtually motionless during the electron jump. This results in the Franck
Condon restriction for electron transfer which was originally formulated in the
study photochemically induced electronic transitions in molecules. It follows that
the Gibbs energy of the system remains constant during electron transfer. In
addition, the spatial conguration of the nuclei is the same in the product system
as it was in the reactant system. This conguration is not the same as the low
energy equilibrium conguration of the nuclei in either system and is usually
achieved as a result of random thermal motion. This description of the formation
of the transition state for electron transfer is the basis for the Marcus model of
electron transfer [4].
The Gibbs energy barrier involved in an electron transfer reaction is shown in
g. 7.3. As a result of random thermal processes the reactants achieve the neces-
316
Fig. 7.3 Gibbs energy prole for the formation of an activated complex 6 in a bimolecular
heteronuclear reaction (A) and a bimolecular homonuclear reaction (B).
sary nuclear conguration for electron transfer. This is the transition state for the
reaction, designated as 6. As stated above, the electron is transferred very rapidly
with respect to the time taken to achieve the required nuclear conguration. After
electron transfer the system relaxes to the product state. For a heteronuclear
reaction the Gibbs energy of the nal state P is lower than that of the initial
state R (g. 7.3A). If the reaction is homonuclear (see equation 7.3.7) then the
Gibbs energy of the nal state P is equal to that of the initial state R (g. 7.3B).
Rate constant data for several homonuclear electron transfer reactions involving transition metal complex ions in water are summarized in table 7.1. The
striking feature of the results is that the rate constants vary over a very wide
range from a low 2 107 M1 s1 to a high of 4 108 M1 s1 . Since these
reactions are all assumed to occur by an outer sphere mechanism, the observations show that the potential energy barrier shown in g. 7.3 varies signicantly
with the nature of the reactants. According to the theory of electron transfer
(section 7.8), the magnitude of the barrier depends on the energy required to
Table 7.1 Kinetic Data for Homonuclear Electron

Transfer Reactions Involving Transition Metal Complex
Ions in Water at 258C [5]*
Reaction
[Cr(H2O)6]3=2
[V(H2O)6]3=2
[Fe(H2O)6]3=2
[Co(H2O)6]3=2
[Ru(H2O)6]3=2
[Ru(NH3)6]3=2
[Ru(en)3]3=2
[Ru(bpy)3]3=2
[Co(en)3]3=2
[Co(bpy)3]3=2
Rate Constant ket =M1 s1
Ionic Strength M
2 107
1:5 102
4.2
3.3
60
3:2 103
2:8 104
4:2 108
7:7 105
18
1.0
2.0
0.55
0.5
1.0
0.1
0.1
0.1
1.0
0.1
*The chelates involved in the last four reactions are ethylenediamine (en)
and bipyridine (bpy).
317
rearrange the ligands and solvation atmosphere around the reactants that must
occur prior to the actual transfer of the electron. However, there is another
important feature of these reactions which leads to variation in the rate constants.
Since the reactions involve highly charged ions the observed rate constants are
very dependent on ionic strength. Thus, a comparison of kinetic data for dierent
reactions should be made at the same ionic strength. An even better way of
examining these data is to use estimates of the rate constants obtained in the
limit of zero ionic strength. More details of all of the important features of
electron transfer reactions are given in section 7.8.
The other bimolecular processes which are considered here involve the movement of atoms and ions. The lightest ion is the proton, and its reactions are
discussed in the next section.
B. Proton Transfer Reactions
According to the Brnsted denition an acid is a proton donor, and a base is a
proton acceptor. It follows that acidbase reactions can also be called proton
transfer reactions. Thus, the reaction between acid HA and base B is written as
HA B(
+A HB
7:3:9
where A is the conjugate base formed from the acid HA and HB , the conjugate
acid formed from the base B. A well-known example is the reaction between HCl
and NH3:
HCl NH3 (
+Cl NH
4
7:3:10
Usually these reactions have been studied in water or in other protic solvents such
as the alcohols. Thus, the acidbase properties of the solvent are important in
determining the relative strength of acids and bases which are solutes in water.
This leads to the denition of two other types of proton transfer reaction, namely,
the protolysis reaction,
HA H2 O(
+H3 O A
7:3:11
B H2 O(
+BH OH
7:3:12
and the hydrolysis reaction,
Finally, there is the neutralization reaction,

H3 O OH (
+2H2 O
7:3:13
The thermodynamics of these reactions is well understood and is documented in

terms of the acidity and basicity constants of the corresponding acids and bases.
On the other hand, their kinetics were only determined after relaxation techniques
for studying very fast reactions became available. The neutralization reaction is
one of fastest reactions known and its rate may be estimated, assuming that its
kinetics are only controlled by diusion of the two reactants to the reaction site
(see section 7.5).
When proton transfer reactions are studied in protic solvents such as water the
role of the solvent can be very important. For example, in a neutral solution at
318
pH 7, the reaction between HCl and NH3 is described by equation (7.3.10). If the
pH is low, the predominant reactant is H3O+ which can be formed by protolysis,
so that the corresponding reaction is
H3 O NH3 (
+NH
4 H2 O
On the other hand, if the pH is high, OH
dominant reaction is
7:3:14
is formed by hydrolysis and the
OH HCl(
+Cl H2 O
7:3:15
This means that a complete analysis of this system involves determination of ten
rate constants: two for direct proton transfer, two for the protolysis reaction, two
for the hydrolysis reaction, and four for proton transfer reactions such as (7.3.14)
and (7.3.15). The relationship among these reactions is shown in g. 7.4 for a
general reaction involving the acid HA and the base B in water.
The theory for the elementary act of proton transfer is well developed [6]. It
involves three steps, the rst of which is formation of the precursor complex
[AH

B]:
HA B ! AH

B
7:3:16
This step involves work which can have an electrostatic component if the reactants are charged, and includes an orientational component resulting from the
requirement that HA have a specic orientation with respect to B. The second
step is the actual proton transfer which involves intramolecular and solvent
reorganization to form the transition state for proton transfer:
AH

B ! A

H

B
7:3:17
Finally, there is the disorientation step in which the complex formed after proton
transfer resolvates as the products:
A

HB ! A HB
7:3:18
The energetics of each of these steps have been described in the Marcus model for
proton transfer [6]. Proton transfer is a bimolecular process so that formation of
the transition state involves an energy barrier of the type shown in g. 7.3.
Kinetic data [7] for some acidbase reactions involving the solvated proton
H3O+ are summarized in table 7.2. The forward reaction is always very fast and
close to the diusion-controlled limit. On the other hand, the reverse reaction is
much slower as required by the relationship between the acidity constant for the
Fig. 7.4 Reaction scheme for proton

transfer between the acid HA and base
B in water illustrating the role of
protolysis (P) and hydrolysis (H)
reactions.
319
Table 7.2 Kinetic Data for Some Simple AcidBase Reactions in Water
at 258C [7]
Rate Constants
Forward kf =M1 s1 Backward kb =s1
Reaction
H3 O OH 2H2 O
H3 O F HF H2 O
H3 O HS H2 S H2 O
H3 O HCO
3 H2 CO3 H2 O
H3 O HCOO HCOOH H2 O
H3 O CH3 COO CH3 COOH H2 O
H3 O NH3 NH
4 H2 O
1:4 1011
1:0 1011
7:5 1010
4:7 1010
5 1010
4:5 1010
4:3 1010
2:5 105
7 107
4:3 103
8 106
8:6 106
7:8 106
25
equilibrium Ka and the two rate constants kf and kb . Similar results are found
when data for a collection of reactions between OH1 and a weak acid are
considered [8]. The forward rate constant is very fast, whereas the backward
reaction reects the basicity constant for the weak base in the equilibrium.
Since ions are involved in these reactions, medium eects are important and
must be considered in a complete analysis of the experimental data.
Many more details can be found about the kinetics of proton transfer reactions
[8]. These often involve specic chemical features of the reactants. Since much of
the focus in this chapter is on experimental methods, these reactions are not
discussed further. However, the use of NMR spectroscopy to study proton transfer kinetics is considered in section 7.9.
C. Ligand Exchange Reactions
The kinetics of formation of complex ions is the subject of considerable interest in
inorganic chemistry [9]. If the discussion is limited to octahedral complexes, then
the type of reaction being considered is
z (
nz
MH2 On
H2 O
6 L +MH2 O5 L
7:3:19
The ligand exchanging with the water molecule can be a neutral molecule such as
NH3 (z 0) or monovalent anions such as the halides (z 1). The interesting
observation is that the rate of the reaction is practically independent of the nature
of the ligand but very much dependent on the nature of the metal ion. This has led
to the hypothesis that the mechanism of ligand exchange involves a reactive
intermediate with less than six ligands:
n
(
MH2 On
6 +MH2 O5 H2 O
7:3:20
The reactive intermediate then recombines with a water molecule or with the
or to give the product
ligand Lz to reform the reactant MH2 On
6
MH2 O5Lnz . This type of mechanism in which a reactant intermediate involving fewer than the normal number of ligands or bonds is formed is called the SN1
320
mechanism in organic chemistry. The rate of the reaction is a reection of the

strength of the MOH2 bond.
In order to characterize the ligand exchange process more quantitatively,
studies have been made of the kinetics of water exchange for a wide variety of
metal ions
17
n
(
H2 O
MH2 O6 n H17
2 O+MH2 O5 H2 O
7:3:21
By using water molecules labelled with 17O in the solvent one is able to follow the
kinetics of the reaction. One popular technique for measuring the kinetics of
reaction (7.3.21) is 17O NMR line-broadening experiments. The exchange reaction
is usually quite fast but has a wide range of rate constants. In terms of the rstorder reaction (7.3.20), the rate constant is small and close to 1 s1 when the metal
ion is Al3 , but very fast and greater than 109 s1 when the metal ion is Cs . In the
case of ions from the main group elements, the interaction between the metal ion
and the water ligands is mainly electrostatic in character. In this case, plots of the
logarithm of the rate constant against z2i r1
i , where zi is the cationic valence and ri
its radius, are linear [10]. For the alkali metal ions, the exchange rate is very fast,
with the rate constant falling in the range 108109 s1 . For the divalent alkaline
earth metal ions, the range of observed rate constants is much wider, varying from
103 s1 for Be2 to 109 s1 for Ba2 . For the transition metal ions other
factors play an important role in determining the rate of ligand exchange, especially the ligand eld stabilization energy. Thus, a cation such as Cr3 which has a
high charge and three unpaired d electrons in the t2g orbitals (see g. 5.25) has a
very low ligand exchange rate ( 106 s1 ). For this reason Cr3 is substituently
inert. When ligand eld stabilization energy is lost by adding an electron to the eg
orbitals in chromium to form Cr2 , the rate of ligand exchange increases dramatically to 109 s1 . Thus, Cr2 is a substitutionally labile cation. Values of the rate
constants for water exchange for the common metal ions of the rst transition
series are given in table 7.3. The radii of these ions do not vary greatly so that the
Table 7.3 Rate Constants for Water

Exchange for Common Ions of the First
Row of the Transition Metals [9]
Ion
V2
Cr3
Cr2
Mn2
Fe3
Fe2
Co2
Ni2
Cu2
Zn2
Number of d Electrons
Rate Constant
kex =s1
3
3
4
5
5
6
7
8
9
10
90
3 106
8 109
3:1 107
3 103
3:2 106
2:4 106
3:5 104
8:0 109
3 107
321
water exchange rate depends mainly on the number of d electrons and the associated ligand eld stabilization energy.
According to the above mechanism (equation (7.3.20)) the reactive intermediate in the ligand exchange process is formed in a unimolecular step. The height of
the Gibbs energy barrier in this step decreases as the rate of reaction (7.3.20)
increases in the forward direction. On the other hand, there is no barrier for
the reverse reaction in which the octahedral complex is reformed. The Gibbs
energy prole associated with a unimolecular reaction is illustrated in g. 7.5.
The stable complex has an average energy equal to that at the minimum on the
curve. As a result of thermal uctuations the energy can reach a value suciently
high to form the activated complex designated by the 6 sign. The diagram also
shows that there is no energy barrier associated with the products in the activated
state returning to form the reactant.
The water exchange process described here gives the essential details about the
ligand exchange process for octahedral complexes. There are many other interesting details which become apparent when one considers a wider variety of
ligands and complexes with dierent coordination numbers. Details can be
found in monographs devoted to this subject [9, 11].
D. Atom Transfer Reactions
The atom transfer reaction can be written in general as
A BC ! AB C
7:3:22
A is a base or nucleophile which attacks the species BC, forming a new molecule
AB and the species C. A well-known example from organic chemistry is the
reaction
Cl CH3 I ! CH3 Cl I
7:3:23
This reaction is also called a displacement reaction and it follows the SN2 mechanism. The attacking nucleophile can be an ion or a neutral molecule. Obviously the
solvent plays an important role in the kinetics of the reaction when ions are
involved.
The formation of the transition state for reaction (7.3.23) is shown in g. 7.6.
The electrophilic reactant is the simple molecule methyl iodide, which has tetrahedral geometry with respect to the carbon atom. As the attacking chloride ion
Fig. 7.5 Gibbs energy diagram for the formation of

an activated complex in a unimolecular reaction step.
322
Fig. 7.6 Formation of the transition state in an atom transfer reaction between Cl and
CH3I.
approaches the methyl group, the CH bonds are distorted. The transition state
consists of a ve-atom species with the CH bonds all perpendicular to the linear
ClCI portion of the molecule. As the reaction proceeds to completion the
distance between the carbon and iodine atoms increases until the product state
is reached.
Extensive studies have been made of solvent eects on atom transfer reactions
involving ions [12]. In the case of reaction (7.3.23), the rate constant decreases
from 250 M1 s1 in N-methylpyrrolidinone to 3 106 M1 s1 in methanol.
This eect can be attributed to solvation of the anionic reactant Cl and the
anionic transition state [12]. Since the reactant is monoatomic, its solvation is
much more important. It increases signicantly with solvent acidity leading to
considerable stabilization of the reactants. As a result the potential energy barrier
increases and the rate decreases with increase in solvent acidity. As shown in g.
7.7, this leads to an approximate linear relationship between the logarithm of the
rate constant and the solvents acceptor number AN, an empirical measure of
solvent acidity (see section 4.9). Most of the results were obtained in aprotic
solvents which have lower values of AN. The three data points at higher values
of AN are for protic solvents.
Fig. 7.7 Plot of the logarithm of the rate constant for reaction (7.3.23) in various solvents
against the solvents acceptor number AN (table 4.10).
323
The distinction between atom transfer reactions involving neutral and ionic
nucleophiles is important in developing the theory of these processes [G3].
Actually, proton transfer is also an ionic atom transfer reaction. However, it is
treated separately because of the special relationship that the proton has to the
most common solvent, namely, water. In addition, because of its very low mass,
proton tunnelling must be considered in proton transfer processes.
The Gibbs energy prole involved in formation of the transition state for atom
transfer is that shown in g. 7.3. When the attacking nucleophile is the same as the
leaving group, then the prole is that for a homonuclear bimolecular reaction.
This type of system can be studied using radioactive isotopes.
The above examples illustrate the important types of solution reactions, especially from the point of view of analytical chemistry. But it is not an exhaustive
list. For example, another type of solution reaction which is of interest to theorists
is the isomerization reaction [G3]. In this chapter, the theory of the electron
transfer process is considered in detail. Some of the results from this treatment
can also be applied to other bimolecular processes. More details follow in the later
sections of this chapter.
7.4 Temperature Effects and Transition State Theory

The eect of temperature on reaction kinetics was an important aspect of experimental studies dating from the earliest work. The common observation is that the
rate constant for the reaction increases with increase in temperature. This was
considered on a quantitative basis by Arrhenius, who derived the following
empirical expression:

kr Ar exp
Ea
RT

7:4:1
In this equation, Ar is the pre-exponential factor, which was assumed to be

temperature independent, and Ea , the activation energy. Further development
of the theory of the temperature dependence of the rate constant was based on
reactions in the gas phase. The simple picture is that two molecules which can
react, collide with a relative kinetic energy given by their individual velocities. If
this relative kinetic energy is greater than a certain minimum value called the
activation energy, then reaction can occur. If it is less than the minimum
value, then reaction does not occur and the two molecules behave as if they
had undergone an elastic collision. In the Arrhenius model, the number of successful collisions is estimated using the Boltzmann factor, expEa =RT), which
gives the fraction of collisions with relative kinetic energy greater than Ea . On the
basis of this discussion, the preexponential factor Ar should really be identied
with a collision frequency. Methods of calculating this quantity are well known on
the basis of the kinetic theory of gases. Using this theory it is clear that the
preexponential factor does depend on temperature. However, the temperature
dependence is much less than that expressed explicitly in the Boltzmann factor.
For this reason, it is more dicult to determine experimentally.
324
Transition state theory was also developed as a means of rationalizing rate

constants for gas phase reactions and their temperature dependence. It is most
directly applied to bimolecular reactions and is based on three fundamental
postulates for reactions in solution:
1. One can conceive of a Gibbs energy surface which gives the Gibbs energy of
the reacting system as a function of the distances between the atoms in both
the reactants and products. There exists a most probable path describing the
conguration of the atoms and surrounding solvent molecules in the system
and its Gibbs energy as it changes from reactants to products. This path is
called the reaction coordinate. The point where the Gibbs energy on this path
is a maximum is called the transition state (see g. 7.3).
2. The chemical species which exists at the transition state X6 is in equilibrium
with the reactant state.
3. The rate of reaction is equal to the product of the concentration of X6 and
frequency n6 with which it decomposes to the product state.
On the basis of these postulates, a bimolecular reaction may be described as

+X6
A B(
7:4:2
X ! products
7:4:3
dcA
dc
B k6 cX6
dt
dt
7:4:4
The rate of reaction is

where k6 is the rate constant for reaction (7.4.3). Since this reaction is unimolecular, the rate constant has the units of s1 and is equivalent to the frequency of
passage through the transition state n6 . On the basis of the equilibrium process in
which X6 is formed, its concentration may be determined from the equilibrium
constant for reaction (7.4.2). For solution processes, the equilibrium constant is
a 6
y 6 cX6
K6 X X
7:4:5
aA aB yA yB c A c B
where ai is the activity of species i, and yi , its activity coecient on the concentration scale. Combining equations (7.4.4) and (7.4.5), the rate of reaction
becomes
dcA
dc
y y
y y
B k6 A B K6 cA cB n6 A B K6 cA cB
dt
dt
yX6
yX6
It follows that the rate constant is given by

y y
k6 n6 A B
yX6
7:4:6
7:4:7
The parameters n6 , K6 , and yX6 are dened within the context of the transition
state model, whereas yA and yB are quantities which can be obtained from equilibrium thermodynamics.
In order to estimate the rate constant, the equilibrium constant is written down
in terms of the partition functions of the components of equilibrium (7.4.2):
K6
q6
qA qB
325
7:4:8
where qi is the partition function of component i. Now, one component of q6 , the

partition function for the activated complex or transition state, corresponding to
a weak or loose vibration state, is factored out. This vibration occurs at a
frequency n6 and is the process which leads to decomposition of the transition
state to form the products of the reaction. The expression for the corresponding
vibrational partition function is

h6 1
7:4:9
q6v 1 exp
kB T
Since the vibration is weak, n6 is small and the ratio hn6 =kB T is much less than
unity. As a result the exponential term in equation (7.4.9) can be replaced by the
rst two terms in its series expansion, so that

h6 1 kB T
q6v 1 1
7:4:10
hn6
kB T
Thus the partition function for the activated complex becomes
q6
kB T
q
hn6 6r
7:4:11
where q6r is the remainder of the partition function for the activated complex
accounting for all other degrees of freedom. The equilibrium constant can now be
written as
K6
kB T q6r
k T
B Ke6
hn6 qA qB hn6
7:4:12
where Ke6 is the eective equilibrium constant obtained after factoring out the
contribution from one degree of vibrational motion. The expression for the rate
constant in the solution is
k T yA yB e
K
7:4:13
kr B
h yX6 6
The frequency associated with decomposition of the activated complex is kB T=h.
At room temperature this corresponds to 6:2 1012 s1 .
In the thermodynamic approach to the transition state model the equilibrium
constant Ke6 is related to the standard Gibbs energy change for process (7.4.2).
Thus, one writes
RT ln Ke6 6 G 6 H T6 S
7:4:14
where 6 G is the standard Gibbs energy change, 6 H , the corresponding

enthalpy change, and 6 S , the standard entropy change associated with reaction
(7.4.2). One should remember that these quantities are obtained for an equilibrium constant dened in terms of concentration with the standard state being
1 mol L1 .
326
By combining equations (7.4.7), (7.4.13), and (7.4.14), the expression for the
rate constant for a reaction in solution is obtained:

kB T yA yB
6 G
exp
kr
7:4:15
h yX6
RT
Alternatively, one may write
6

kB T yA yB
S
6 H
exp
kr
exp
h yX6
R
RT
7:4:16
Comparing this result with the Arrhenius equation (equation (7.4.1)) it is clear
that the major part of the temperature dependence of the rate constant is associated with the exponential term containing the standard enthalpy of activation,
6 H .
An approximate relationship between the Arrhenius activation energy Ea and
the standard enthalpy of activation 6 H can be found by taking the derivative
of ln kr with respect to 1=T at constant pressure using equation (7.4.15).
Neglecting the temperature derivatives of the activity coecients, one obtains

@ ln kr
h kB T2 6 S 6 G
6 H RT
7:4:17

kB T h
R
@1=T p
RT
R
Comparing this result with the corresponding derivative obtained from equation
(7.4.1) it follows that
Ea 6 H RT
7:4:18
for reactions in solution.

When the rate of reaction is studied as a function of pressure, the standard
volume of activation can be found. Neglecting the pressure dependence of the
activity coecients, on the basis of equation (7.4.15),

@ ln kr
@ ln6 G =RT
6 V
7:4:19
@P
@P T
RT
T
EXAMPLE
The following rate constants are reported for the reaction between p-bromodimethylaniline and methyliodide in nitrobenzene:
Temperature/8C
Rate Constant/M1 s1
15.0
24.8
40.1
60.0
8:24 106
1:80 105
5:55 105
2:13 104
Estimate the Arrhenius activation energy, the pre-exponential factor, and

the enthalpy and entropy of activation at 258C. Ignore any eects due to
solution non-ideality.
The data are tted using least squares to the Arrhenius equation in the form
327
Ea
RT
7:4:20
6:969 103
T
7:4:21
ln kr ln Ar
The result is
ln kr 12:46
The value of the activation energy is given by

Ea 6:97 103 8:314 58:0 kJ mol1
7:4:22
The pre-exponential factor is

Ar exp12:456 2:57 105 M1 s1
7:4:23
From equation (7.4.18), the enthalpy of activation is

6 H Ea RT 57:9 2:5 55:4 kJ mol1
7:4:24
The Gibbs activation energy is calculated from equation (7.4.15) ignoring the
activity coecients. From equation (7.4.20) the rate constant at 258C is
1:82 105 M1 s1 . Thus,
!
5
1:82

10
7:4:25
6 G RT ln
100:1 kJ mol1
6:2 1012
Now,
6 S
6 H 6 G 55:4 100:1 1000

150 J K1 mol1
298:2
T
7:4:26
The entropy of activation is negative because formation of the activated

complex involves the combination of two molecules.
Estimation of the activation parameters on the basis of fundamental theoretical considerations is a formidable task for solution reactions. Because of the
surrounding solvent molecules which solvate the reactants A and B in a bimolecular process, there are many degrees of freedom which must be considered in
evaluating the partition functions qA , qB , and q6r . Moreover, the number of
degrees of freedom often changes when the transition state is formed. For example, consider the case that A and B are two highly charged cations (reaction
(7.3.6)). Each cation is strongly solvated so that the internal degrees of freedom
associated with both the cation and surrounding solvent molecules must be considered in evaluating qA and qB . The charge on the transition state, zA zB , is
much higher. As a result, more solvent molecules are probably solvating the
activated complex than the total number solvating A and B separately. Thus,
formation of the transition state is accompanied not only by a change in the
energy of the system, but also by a change in entropy. This example illustrates
the importance of considering the Gibbs energy in discussing the course of the
reaction through the transition state.
There is considerable interest in comparing the reaction rate for a given reaction in solution with that in the gas phase [G3]. Assuming that the rate of decomposition of X6 is the same in both phases, the ratio of the rate constants is
328
"
#
s
6 Gs 6 Gg
ksr yA yB K6 yA yB
exp
yX6 Kg6
yX6
RT
kgr
7:4:27
where kir is the rate constant in phase i, Ki6 , the equilibrium constant to form the
activated complex in phase i and 6 Gi , the corresponding standard Gibbs activation energy. If consideration is limited to reactions not involving ions in nonpolar solvents, the activity coecients can be neglected. Furthermore the Gibbs
activation energy in solution can be expressed as follows:
6 Gs 6 Gg 6 Gsolv
6
Gsolv
Gs .
7:4:28
6
Gsolv
where
is the solvent contribution to
In other words,
is the
contribution to the work done due to the presence of the solvent in bringing the
reactants A and B from sites in the solution which are very far apart up to the
transition state site where they are in the same solvent cage. The expression for the
rate constant in solution is then

6 Gsolv
ksr kgr exp
7:4:29
RT
An eective way of estimating 6 Gs and 6 Gg is to carry out molecular
dynamics calculations for the system being studied. In this way, cumbersome
quantum-mechanical calculations involving the reacting system in the solvent
are avoided. From a qualitative point of view, an atom exchange reaction involving neutral reactants in a non-polar solvent is expected to be faster in the solvent
than in the gas phase. This is due to the fact that two neutral species form one
species in the transition state. Very few experimental studies have been carried
out; one example of a reaction which has been studied is
H D2 ! HD D
7:4:30
This reaction is faster in solution than in the gas phase by approximately a factor
of 40. Theoretical calculations agree with the experimental result [G3].
Transition state theory has also been applied in quite another way to reactions
in solution. Reaction (7.4.2) can be described as a reaction precursor equilibrium
which is characterized by the diusion of A and B to a position close enough so
that reaction can take place. This pre-equilibrium has an equilibrium constant
which can also be thought of as a collision frequency. The expression for the rate
constant is then

G
7:4:31
kr ZB exp
RT
where ZB is the collision frequency for the bimolecular process in solution and
G is the standard Gibbs activation energy for reaction (7.4.3). G diers
from 6 G in that it is calculated using the total partition functions for A, B, and
X6 ; on the other hand, 6 G has the contribution from one degree of vibrational
motion in the activated complex factored out. Electron transfer kinetics in solution are usually analyzed on the basis of equation (7.4.31). A more complete
description of this subject as well as methods of estimating ZB are given later
in sections 7.8 and 7.10.
329
Medium eects are very important for solution reactions. They are dealt with
quantitatively through the activity coecient ratio which appears in equation
(7.4.15). Thus, if the nature of the non-reacting components in the solution is
changed, one can expect to observe a change in the rate constant. The easiest way
of dealing with these eects is to keep the nature of the medium constant when
other aspects of the system being studied are changed. On the other hand, there is
an interest in studying the eects of the medium on solution reactions. This
subject is considered in more detail in section 7.10.
7.5 Diffusion-Controlled Rapid Reactions

The fastest reactions in solution are limited in their reaction rates by mass transfer. For most bimolecular reactions this limitation arises because the two reactants must diuse together in order to form a reactant pair. Therefore it is
interesting to estimate the magnitude of the rate constant for a diusion-controlled reaction and to compare it with experimentally determined rate constants
for very fast reactions. The treatment presented here applies to bimolecular reactions involving either molecules or ions.
Consider a bimolecular reaction
A B ! products
7:5:1
which takes place every time these species come into contact. In order to solve the
mass transfer problem, it is imagined that only A moves and B is stationary. On
the basis of the analysis in section 6.7, the ux of species A is
z u c
JA DA ;cA A A A ;(
7:5:2
jzA j
where ;cA is the gradient of the concentration of A, ;( the gradient of the
electrostatic potential, DA , the diusion coecient of A, uA , its mobility, and
zA , its charge. The second term obviously accounts for electrostatic eects
which are only important in the case of reactions involving ions. In the case of
reactions involving molecules, the second term disappears. Using the relationship
between the diusion coecient and mobility of an ion (equation (6.7.21)), the
above equation may be written as

z Fc
7:5:3
JA DA ;cA A A ;(
RT
It is convenient to re-express the electrostatic term in terms of the electrostatic
energy UeA , where
UeA zA F
Thus, the equation for the ux of A may be written as
c
JA DA ;cA A ;UeA
RT
7:5:4
7:5:5
The problem is now solved, assuming that diusion is a spherically symmetrical process and that it occurs at a steady state. Furthermore, the reactants
330
are assumed to be spherical with radii rA and rB . The ux which is a constant

independent of angular direction becomes

dcA cA dUeA
JA DA
7:5:6
dr
RT dr
Introducing the function
U
Cr cA exp eA
RT
one may write

7:5:7

UeA dCr
JA DA exp
dr
RT
7:5:8
The fact that both A and B move in reality is now dealt with by using DA DB
as the mutual diusion coecient; this eectively accounts for the motion of the
coordinate system which is now dened with respect to B. Thus,

UeA dCr
7:5:9
JA DA DB exp
RT
dr
The number of molecules moving through the surface of a sphere of radius r can
now be calculated:

dnA
UeA dCr
2
2
4pr JA 4pr DA DB exp
7:5:10
dt
RT
dr
At a steady state, dnA =dt is a constant so that solution of the dierential equation
is relatively easy.
The rate of reaction is found by integrating equation (7.5.10) in the form
dnA
dt
expUeA =RT
dr
r2
cA
4pDA DB dCr
7:5:11
In the limit that r equals a the molecules A and B are in contact, that is,
a rA rB
7:5:12
In this case, species A reacts with species B, and the concentration of A is zero. At
very large distances, the concentration of A is its bulk value, cA . After integration,
one obtains
dnA
4pDA DB a y cA
7:5:13
dt
where
1
expUeA =RT
dr
7:5:14
y1 a
r2
a
The factor a is introduced into the fraction y, so that it is dened as a dimensionless quantity. Equation (7.5.13) applies to one molecule of B. By multiplying by
the concentration of B, cB , one obtains an expression for the reaction rate:
dnA
4pDA DB a y cA cB
dt
331
7:5:15
The second-order rate constant is

kr 4pNL DA DB a y
7:5:16
where Avogadros constant NL is introduced so that this quantity has units of

m3 mol1 s1 when DA and DB have units of m2 s1 and a, units of meters.
When the two reactants are molecules, the electrostatic energy UeA is zero and
the fraction y is equal to one. Then, the expression for kr is
kr 4pNL DA DB a
7:5:17
When the two reactants are ions, the value of y is most easily estimated for the
case that the solution is very dilute and the electrostatic energy is given by
UeA
zA zB e20 NL a
r
4pe0 es r
7:5:18
where e0 is the electronic charge, es , the relative permittivity of the solvent, and e0 ,
the permittivity of free space. Integration of equation (7.5.14) gives the nal result
1
a
y ea=a 1
7:5:19
a
Finally, if the bimolecular reaction involves the same species, namely,
A A ! products
7:5:20
the rate of reaction is given by

dcA
2kr c2A
dt
7:5:21
In this case the expression for the diusion-limited bimolecular rate constant is
kr 4pNL DA a y
7:5:22
EXAMPLE
Estimate the diusion-limited rate constant for the recombination of I

radicals
in carbon tetrachloride at 258C given that the radius of the radical is 200 pm,
and its diusion coecient, 4:12 109 m2 s1 .
The reaction is
7:5:23
I
I
! I2
Since the reactants are neutral, electrostatic eects are negligible and y is equal
to one. The collision distance a is equal to 400 pm. Thus,
kr 4p 6:022 1023 4:12 109 4:0 1010
1:27 107 m3 mol1 s1
7:5:24
The experimentally observed value for kr is 8:2 106 m3 mol1 s1 . The fact
that the experimental value is smaller than that predicted theoretically suggests
that there may be a small activation energy associated with this process.
332
EXAMPLE
Estimate the diusion-limited rate constant for recombination of H3O+ and

OH to form water at 258C given that the diusion coecients for these species
are 9:34 109 m2 s1 and 5:23 109 m2 s1 , respectively, and assuming that
the eective radius for each species is 250 pm.
The reaction is
H3 O OH ! 2H2 O
7:5:25
This reaction is accelerated by the attractive Coulombic force between the ions.
The electrostatic parameter is
zA zB e20 NL
1 1:602 1019 2 6:022 1023
4pe0 es RTa 4p 8:854 1012 78:5 8:3145 298:2 5 1010

7:5:26
1:429
The electrostatic factor y is
y
1:429
1:879
exp1:429 1
7:5:27
The estimate of the rate constant is

kr 4p 6:022 1023 14:57 109 5 1010 1:879
1:04 108 m3 mol1 s1
7:5:28
The experimental value of kr is 1:4 108 m3 mol1 s1 . The theoretical estimate is low probably because the model of the reactants as hard spheres is
inappropriate for this system. This follows from the fact that the reacting ions
have a special relationship to the solvent, water.
Although the treatment described here is approximate it certainly gives the
correct order of magnitude for the rate constants of very fast reactions in solution.
Methods of improving these estimates have been considered by Noyes [2]. In the
following section experimental methods of studying very fast reactions based on
relaxation techniques are considered.
7.6 Relaxation Techniques for Rapid Reactions

Relaxation techniques were developed by Eigen and coworkers [13, 14] as a means
of studying very fast reactions in solution. The experiment involves applying a
stress rapidly to a system at equilibrium and following the time lag required to
establish new equilibrium concentrations. For example, in the temperature jump
method, the temperature of the system can be changed by several degrees in a time
as short as 106 s by dissipating the energy stored in a high-voltage capacitor in a
solution which contains a conducting electrolyte. Spectrophotometric or conductometric methods are then used to follow the relaxation of the components of the
equilibrium to new concentrations which reect the value of the equilibrium
constant at the higher temperature. Then, the prole of concentration against
333
time is analyzed to nd the rate constants characterizing the reactions taking

place. In the following section some simple examples are considered to illustrate
how the relaxation data are analyzed. Some important experimental techniques
used are also outlined.
A. Analysis of the Experimental Data
As a simple example, consider the equilibrium in a homogeneous reaction
kf
AB C
kb
7:6:1
which is characterized by a forward rate constant kf and a backward rate constant

kb . At equilibrium
cCe
k
Ke f
cAe cBe
kb
7:6:2
The initial concentrations of these components which are present before the stress
is applied to the system are cA0, cB0, and cC0. Then at any moment during the
relaxation one may write
x cA cAe cB cBe cCe cC
7:6:3
where x is the extent that the instantaneous concentrations dier from their new
equilibrium values. The rate of reaction is
dcA dx
kf cA cB kb cC
dt
dt
7:6:4
This equation is now written in terms of the variable x:

dx
kf cAe xcBe x kb cCe x
dt
kf cAe cBe kf xcAe cBe kf x2 kb cCe kb x
7:6:5

kf cAe cBe kb cCe 0
7:6:6
Furthermore, the term in x may be neglected because the departure from equilibrium is not large. The nal expression for the rate of reaction is
dx
kf cAe cBe kb x
dt
7:6:7
This has the same form as the expression derived for a simple rst-order reaction
with the quantity in the square brackets as the rst-order rate constant. The usual
practice in relaxation kinetics is to dene the reciprocal of this quantity as the
relaxation time t for the reaction. Thus,
t kf cAe cBe kb 1
7:6:8
By varying the equilibrium concentration of one reactant while holding the concentration of the other high and approximately constant, the two rate constants
334
Table 7.4 Relaxation Times for Equilibrium

Reactions Involving One Step
Reaction
Relaxation Time/t
kf
kf kb 1
A B
kb
kf
4kf cAe kb 1
2A B
kb
kf
kf cAe cBe kb 1
AB C
kb
kf
AB CD
kb
kf cAe cBe kb cCe cDe 1
which contribute to t can be separated. Relaxation times for several other

mechanisms involving one step are summarized in table 7.4. If these reactions
do not correspond to equilibria but actually are irreversible, then the relaxation
time is found by setting kb equal to zero.
Most chemical reactions involve more than one elementary step. As a result
there is more than one relaxation time associated with the reacting system. In the
case of fast reactions it is necessary to extract information about each step.
Although the mathematical analysis of complex systems can be quite involved,
it is important to understand how such an analysis is made. The following example involves a two-step mechanism in which both steps are reversible.
Consider a bimolecular process
k1f
A B AB
k1b
7:6:9
in which the reactive intermediate AB is formed. The forward rate constant for
this reaction is k1f and the backward rate constant, k1b. The species AB can
undergo an intramolecular rearrangement
k2f
AB C
k2b
7:6:10
to form the nal product C. The second step is characterized by a forward rate
constant k2f and a backward rate constant k2b.
Now one denes variables which relate to the departure of each concentration
from its equilibrium value. They are
xA cA cAe cB cBe
xAB cAB cABe
xC cC cCe
7:6:11
7:6:12
7:6:13
Mass conservation requires that

xA xAB xC 0
The rate of reaction of A is
7:6:14
dcA
k1f cA cB k1b cAB
dt
335
7:6:15
In terms of the variables xA and xAB this becomes

dxA
k1f xA cAe xA cBe k1b xAB cABe
dt
7:6:16
When this expression is simplied by neglecting terms in x2 and using the equilibrium condition
k1f cAe cBe k1b cABe
7:6:17
one obtains
dxA
k1f cAe cBe xA k1b xAB
dt
7:6:18
The expression for the rate of formation of the intermediate AB is

dcAB
k1f cA cB k1b cAB k2f cAB k2b cC
dt
7:6:19
On the basis of equilibrium (7.6.10)

k2f cABe k2b cCe
7:6:20
Making use of equations (7.6.11)(7.6.14), the expression for the rate of formation
xAB is
dxAB
k1f cAe cBe k2b xA k1b k2f k2b xAB
dt
7:6:21
Equations (7.6.18) and (7.6.21) comprise linear rst-order dierential equations in the variables xA and xAB. For simplicity they may be rewritten as
dxA
a11 xA a12 xAB
dt
7:6:22
dxAB
a21 xA a22 xAB
dt
7:6:23

where
a11 k1f cAe cBe
7:6:24
a12 k1b
7:6:25
a21 k1f cAe cBe k2b

a22 k1b k2f k2b
7:6:26
7:6:27
This system of equations may be solved by several methods including Laplace

transformation and matrix algebra. The mathematical details have been given by
Connors [G2].
The solutions to these equations may be expressed in terms of the relaxation
times t1 and t2 for step 1 and step 2, respectively. The sum of their reciprocals is
given by
1
t1
7:6:28
1 t2 a11 a22
and the product by
336
11 t1
2 a11 a22 a12 a21
7:6:29
Using the denitions of the coecients given above these become

1
t1
1 t2 k1f cAe cBe k1b k2f k2b
1
t1
1 t2 k1f cAe cBe k2f k2b k1b k2b
7:6:30
7:6:31
Thus, each relaxation time depends on all four rate constants.

The analysis of the experimental data to obtain the two relaxation times is
equivalent to the analysis of two consecutive rst-order reactions. If the values of
t1 and t2 are quite dierent, this is not a problem and can be accomplished using
standard methods of data analysis [G2]. If t1 and t2 are close to one another,
separate determination of these quantities is very dicult. Typically, the ratio of
the relaxation times must be greater than three to estimate them separately. In this
regard, some variation in the relative values of t1 and t2 can be achieved by
changing the values of cAe and cBe.
Further analysis of the data involves determining t1 and t2 for a range of values
of the sum cAe cBe. On the basis of equations (7.6.30) and (7.6.31), the individual rate constants can then be found using linear plots of t11 t21 against
cAe cBe and (t1t2)1 against the same quantity.
One limiting case of interest occurs when t1t2. This corresponds to reaction
(7.6.9) coming rapidly to equilibrium with the second step (7.6.10) being much
slower. Under these circumstances, the terms in k2f and k2b in the expressions for
a21 and a22 disappear. The reaction is then termed decoupled because the kinetic
processes determining the rst relaxation can be completely separated from those
determining the second.
The above system provides one example of the treatment of a relaxation process for a complex reaction involving a two-step process. Other systems involving
two or more steps can be treated by this method, and details can be found elsewhere [G2, 14].
B. Some Experimental Methods
The temperature jump (T-jump) method is the most widely used relaxation technique. It is based on the temperature dependence of the equilibrium constant for
the reaction. For a one-step process such as reaction (7.6.1), this is given by the
vant Ho equation, that is,
@ ln Ke
H

@1=T
R
7:6:32
where H is the standard enthalpy change for the reaction. The experiment
usually involves a specially designed cell with a small volume of solution. The
temperature jump is accomplished via the joule heating which occurs when a
capacitor is discharged through the solution, which must be electrically conducting. Typical temperature changes are between 1 and 10 C, and these can be
achieved in times as short as 10 ns. Although this technique was originally used
with aqueous electrolyte solutions, it can also be applied to systems with much
lower conductivity using microwave and laser heating techniques. A typical T-
337
jump cell designed for use with an electrical discharge is shown in g. 7.8.
Monochromatic light passes through the cell and out to a detector in which the
absorption transient is recorded. The relaxation time is determined by analyzing
the absorption data as a function of time.
EXAMPLE
A system with an equilibrium constant of 5 105 L mol1 at 25 C is subjected

to a temperature jump of 5 C. Calculate the new equilibrium constant given
that the standard enthalpy change for the reaction is 10 kJ mol1 .
On the basis of equation (7.6.32)

H 1
1
ln Ke T2 ln Ke T1

R T2 T1
7:6:33

10,000
1
1
2

6:65 10
8:314 303:2 298:2

The equilibrium constant at 30 C is 5:3 105 L mol1 corresponding to a 7%
percent change.
The pressure jump (P-jump) relaxation technique employs a sudden change in
pressure for a reversible reaction which has associated with it a signicant volume
change. In this case the change in the equilibrium constant with pressure is given
by
@ ln Ke
V

@P
RT
7:6:34
The sample is mounted in a exible cell which is placed in a pressurized vessel at a

high pressure, say, 60 atm. The pressure is suddenly reduced by rupturing a diaphragm so that it drops to the atmospheric value. Unfortunately, the value of
V associated with most reactions is rather small, so that the equilibrium is only
slightly perturbed by the pressure change.
Fig. 7.8 Experimental setup with a temperature-jump cell designed for Joule heating by
means of an electrical discharge.
338
Fig. 7.9 Illustration of the time

dependence of the forcing function,
for example, pressure, and the
concentration of one reactant in (A) a
step relaxation experiment and (B) an
experiment with a periodic
perturbation.
The electrical eld-jump method is applied to reactions involving ions and

dipoles. The eld induces a shift in the solution equilibrium in the direction of
producing more ions, or in the orientation of dipoles. Thus, when a high electrical
eld is applied to a solution containing a weak electrolyte, the extent of dissociation increases. The method has been used to study metal ion complex formation
reactions and acidbase reactions.
Ultrasound experiments rely on the periodic oscillations which result in a solution when a high-frequency sound wave is passed through it. Sound waves produce periodic oscillations in pressure and temperature which result in a periodical
perturbation in a chemical equilibrium (see g. 7.9). In aqueous solutions, the
pressure change is more important, whereas in non-aqueous solutions the temperature change is predominant. When the frequency of the sound wave is high
enough, the chemical equilibrium cannot respond quickly enough and a phase
shift is observed between the oscillations in the concentration of one component
and the oscillations in pressure and/or temperature. Using transducers, these
oscillations can be converted to electrical signals and precisely followed using
lock-in ampliers or similar electronic instrumentation.
7.7 Laser Spectroscopy and Femtochemistry in

Solutions
The use of short light pulses to study rapid chemical reactions in solution dates
from the work of Norrish and Porter [15]. In a ash photolysis experiment a light
339
pulse of short duration is used to irradiate the solution, where it is absorbed by

molecules or ions. The resulting excited species loses its excess energy either by
returning to the ground state via uorescence or some other process, or by a
chemical reaction which produces radicals and other intermediates of high energy.
In the latter case, further reactions of these species are monitored using a less
energetic spectroash red a short time after the initial photolysis ash. This
procedure is repeated several times with dierent delay times between the photolyzing ash and the spectroash. In this way the kinetics of the reactive intermediate can be elucidated.
In the original ash photolysis experiment a high-voltage capacitor was discharged through a ash photolysis lamp. The typical energy involved about 1000 J
which was dissipated within a few microseconds. This kind of experiment is
carried out at present using lasers. Using these devices one is able to resolve
chemical events at times less than one picosecond. This allows the experimenter
to monitor changes which occur during times for which the vibrational motion in
molecules is essentially frozen. These exciting developments have led to the coining of the new word femtochemistry, which refers to chemistry occurring in the
femtosecond time scale. In this section the important features of an experiment
involving ultrafast pulsed lasers are briey outlined. Then applications of these
experiments to reactions in solution are described.
A. Experimental Aspects
In order to discuss this topic a brief introduction to the properties of common
lasers is given. The word laser is an acronym for light amplication by stimulated
emission of radiation. A molecule can absorb a quantum of light when the associated energy corresponds exactly to a transition from the ground state in the
molecule to an available excited state. Under normal circumstances the excited
molecule decays back to the ground state by some process which may be radiative,
involving the emission of light, or non-radiative, involving the generation of heat.
Photon emission occurring this way is called spontaneous emission. On the other
hand, if the system with molecules in the excited state is irradiated with an
external light source corresponding in energy to that which would be emitted,
each molecule can relax by emitting another photon of the same frequency. Under
these circumstances the probability of emission is enhanced by the presence of
similar photons and the emission occurs preferentially in the same direction as the
external source. This leads to a population inversion with more molecules in the
excited state than in the ground state. This type of emission is called stimulated
emission and is the principal feature of laser operation. Moreover, since the emission occurs preferentially in the direction of the applied beam, this beam is
thereby amplied in intensity. The stimulated photons, have the same phase
and polarization as the incident photons, so that the laser radiation is termed
coherent.
The above discussion implies that a laser may be designed with a absorbing
system involving two quantum states, namely, the ground state and the excited
state. However, an analysis of the rate at which molecules reach the excited state
from the ground state with respect to the rate at which molecules in the excited
340
state return to the ground state by emission shows that a population inversion is
not possible with a two-state system. In order to realize a population inversion a
multilevel system is required. An example of a three-level system which is capable
of lasing action is shown in g. 7.10. The ground state system at energy E1 is
pumped to the excited state E2 . It rapidly undergoes a non-radiative transition to
a lower energy level E3 The lasing action involves the emission of a quantum of
light corresponding to the energy dierence E3 E1 . Other multiple level lasing
schemes have been used to design functioning lasers [15].
The three essential elements of a laser are the resonator containing the gain
medium, a pumping source that excites the gain medium, and mirrors which direct
the light beam back and forth through the gain medium (see g. 7.11). One of
these mirrors is not completely reective, so that part of the laser beam escapes
giving the output light from the laser system.
The gain medium can be a solid, a liquid solution, or a gaseous mixture. One
the rst lasers made use of a solid ruby rod which consists of Al2O3 doped with a
minor Cr3 impurity. This system outputs light at a wavelength of 694.3 nm.
Other solid-state systems make use of Nd3 or Ti3 as the gain medium.
Popular lasers with a gas as the gain medium include the HeNe laser, and the
CO2 laser. Of course, the output frequency depends on the characteristics of the
gain medium and its associated spectroscopic properties.
The energy source used to initiate lasing action is often a high-intensity light
source which is used to excite the gain medium. In the case of gas phase lasers, an
electrical discharge can be used to provide the energy to pump the gas molecules
into the excited state involved in laser action. The properties of the laser cavity are
important in achieving the necessary population inversion with more molecules in
Fig. 7.10 Transitions and energy levels in a three-level laser.
341
Fig. 7.11 Schematic diagram of the essential components of a laser.
the excited state than in the ground state. Thus, the design of the resonating
system by which most of the light travels back and forth between the two mirrors
shown in g. 7.11 is an important feature of the system.
Lasers are operated in two modes, on a continuous-wave basis, or in a pulsed
mode. In the continuous mode the laser is pumped steadily at a rate sucient to
supply the emergent laser energy and the heat losses due to the non-radiative
processes. The pulsed laser is of interest for kinetic studies because it allows a
pulse of optical energy to be applied to the experimental system over a known
period of time. In this mode, the laser is pumped with a pulsed energy source, and
a shutter is used to control the release of energy in the laser. The shutters may be
mechanical or electrooptical, and a variety of pulsing techniques have been developed. A self-mode-locked Ti:sapphire laser is a popular system for ultrafast spectroscopy. It can be tuned to operate over a wide range of frequencies in the visible,
it has high average power output, and it can deliver ultrashort pulses down to 10
fs in duration. More details about the generation of ultrashort light pulses can be
found elsewhere [1618].
Having delivered optical energy to the experimental system using an ultrafast
pulse, the next part of the experiment involves following the changes which occur
over a very short period of time. This is called ultrafast time-resolved spectroscopy. One of these is pump-probe spectroscopy. After the system has been
pumped to an excited optical state it is probed by a series of light pulses which
are used to follow the response of the system as a function of time. For example,
the pump-pulse system may be used to excite a simple molecule from its ground
electronic level to an excited electronic level where a bond in the molecule is
ruptured. The probe pulses can follow the absorption of light by one of the
fragments of the molecule. On the other hand, the probe pulses can be used to
induce uorescence in one of the fragments, and the uorescence is followed as a
function of time. The probing pulse may be the same frequency as the pumping
pulse, or it may have a dierent frequency depending on the design of the experiment. The time delay between probe and pulse is obtained by altering the path
length for the two light signals in the optical system.
When the experimental system emits light after the initial pumping pulse, quite
dierent techniques can be used to obtain a time-resolved spectrum of the sample
emission. The simplest of these is time-correlated single photon counting. The
time resolution of this technique is limited by the design of the photon detectors.
Two other methods used in emission spectroscopy are the streak camera and
342
uorescence up-conversion. Details of these experiments which are beyond the

scope of this monograph can be found elsewhere [1618].
Experimental studies of solution reactions using pulse lasers have blossomed
since the 1980s. Important examples are considered in the following discussion.
B. Time-Resolved Studies of Solvation
Very interesting studies have been made of the eects of solvation on the timeresolved uorescence of certain dye molecules dissolved in polar solvents. The
solute molecules are often characterized by a non-polar ground state and a dipolar excited electronic state. As a result, when the excited state is formed by the
absorption of a quantum of light, the molecule nds itself in a solvation environment with which it is not at equilibrium. With increasing time, the solvent restructures responding to the new charge distribution in the excited state. This results in
a lowering of the energy of the excited state with respect to the ground state. As a
result the uorescence spectrum changes with time with its peak frequency shifting
to lower values. This shift in the emission spectrum in the red direction is called a
Stokes shift (SS). Analysis of the spectral features as a function of time is used to
obtain molecular details of the solvent restructuring process.
The uorescence spectra are analyzed to obtain the correlation function C(t)
dened as
Ct
nm t nm 1
nm 0 nm 1
7:7:1
where nm(t), nm(0), and nm(1) are the frequency of the maximum on the uorescence spectra observed at time t, zero, and innity, respectively. The time
dependence of C(t) is related to the relaxation process of the solvent environment
around the dipolar excited state. A theoretical model for this process has been
derived in which the dye molecule in its excited state is considered to be a point
dipole embedded in a dielectric continuum [19]. In this case the correlation function C(t) is an exponentially decaying function of the form

t
7:7:2
Ct exp
tL
where tL is the longitudinal relaxation time of the solvent (see section 4.7). Studies
carried out in a wide variety of solvents show that this simple model is approximately correct. Results obtained in a variety of solvents with the coumarin dye
C152 (g. 7.12) are summarized in table 7.5.
It is clear from these results that there is an approximate correlation between
the value of the average relaxation time tS and the longitudinal relaxation time tL
(see section 4.7). However, careful examination of the time-dependent Stokes shift
(TDSS) data reveals that Ct is described by more than one relaxation process.
This is not dicult to understand, considering that many aprotic solvents are
found to have more than one relaxation process. This can be attributed to the
formation of dimers and other aggregates in these liquids. This is especially the
case in aprotic solvents such as DMSO and PC, which have very high dipole
moments. Alcohols also have multiple relaxation processes due to the presence
343
Fig. 7.12 Structure of the coumarin dye C152.
of hydrogen bonding (see section 4.7). The most important point is that the Debye
model which is applied to the analysis of the dielectric relaxation data is a macroscopic model. What is really needed to interpret the spectroscopic data is a molecular model for the dielectric properties of the solvents.
The example given is only one of several which have been reported for solvation phenomena studied at very short times after a molecule absorbs a quantum of
light [17, 18]. These studies have only become possible with the advent of highspeed laser technology.
C. Ultrafast Photoisomerization Reactions
Stilbene is a molecule which exists in two isomeric states because of the large
energy barrier associated with rotation around the central double bond (see g.
7.13). It has been known for some time [22] that when stilbene is excited to its rst
singlet state by absorption of UV light, it rapidly uoresces to the ground state
but with some molecules in the other isomeric form. Thus, if cis-stilbene is irradiated with ultraviolet light, one obtains approximately 50% trans-stilbene and
Table 7.5 Average Relaxation Times from Time-Dependent Stokes Shifts

(TDSS) for Coumarin C152 in Various Solvents Together with the
Longitudinal Relaxation Time Determined by Dielectric Relaxation
Spectroscopy [20]
Solvent
MeOH
AC
AcN
BzN
DMF
DMSO
PC
PrN
Longitudinal Relaxation Time*

tL /ps
Average Relaxation Time from TDSS

ts =ps
1.0
4.0
0.3
0.2
5.7
0.7
1.5
2.4
0.4
6.2
0.68
0.52
5.0
1.0
1.2
2.4
0.85
*The values of tL for aprotic solvents with more than one relaxation process are averages; two
values are given for methanol, which has three relaxation processes. (See section 4.7 and [21].)
344
Fig. 7.13 Isomerization of stilbene.
50% cis-stilbene. The kinetics of the photoisomerization reaction is so rapid that

it can only be resolved using a pump-pulse laser technique. These experiments
allow one to study the role that the solvent plays in the kinetics of the isomerization process.
Detailed studies of this photo-initiated reaction have led to a picture of the
energetics associated with the isomerization reaction (g. 7.14). When cis-stilbene
reaches the excited singlet state, it rapidly loses energy by twisting about the
central double bond. It then returns to the ground state from the energy minimum
by uorescence and gives either the trans or cis isomer, depending on how the
molecule leaves the twisted form to return to a low-energy conguration in the
ground state. For example, when cis-stilbene is irradiated at 316 nm emitted light
can be observed at 475 nm when the experiment is carried out at 20 C in decane.
The uorescence decay occurs over a few picoseconds and in this case is characterized by relaxation time t of 1.27 ps [23]. Studies have been made of the
Fig. 7.14 Energy proles for

twisting of the stilbene molecule in
the ground state S0 and rst excited
singlet state S1 .
345
dependence of t on solvent nature, temperature, and the nature of substituents on

the stilbene molecule [22].
Fluorescence decay times measured in several polar solvents are summarized in
table 7.6. It is apparent there is a correlation between the viscosity and t, longer
decay times being observed in more viscous solvents. This eect is related to the
friction of the solvent molecules surrounding stilbene when it undergoes rotation
in the excited state. Solvent friction in solution kinetics is of considerable theoretical interest (see section 7.10). In order to determine the activation parameters
associated with photoisomerization, decay times have been measured in a series of
solvents with the same structure. The temperature of the experiment is varied so
that the solvents all have the same viscosity. A plot of the logarithm of t against
the reciprocal of the absolute temperature using data obtained in this way is
known as an isoviscosity plot. Some data for the isomerism of cis-stilbene in the
alkanes are given in table 7.6. The decay times observed in these solvents are
longer than those obtained in polar solvents indicating that a larger energy barrier
is involved. The isoviscosity plot of these data gives an activation energy of 4.6 kJ
mol1. The corresponding result for trans-stilbene is 14.6 kJ mol1 [22]. These
results are consistent with the energy prole for the excited state shown in g.
7.14. When the same reaction is studied in polar solvents, namely, a series of nalkanenitriles, the barrier is signicantly lower (10.8 kJ mol1). This suggests that
the excited state has some polar character. Much more has been learnt about this
simple unimolecular reaction on the basis of the ultrafast laser studies. It is clear
from the data presented in table 7.6 that the study could not be carried out
without the ability to resolve the spectral data on a time scale less than 1 ps.
This kind of experiment has been applied to several important solution reactions
and represents the forefront of modern research in this area.
Table 7.6 Cis-stilbene Fluorescence Decay Times in

Organic Solvents for Emission at 475 nm [23]
Solvent
Temperature
T/8C
Viscosity
Z/mPA s
Decay Time
t/ps
20
20
20
20
20
0.38
2.20
0.59
2.38
4.18
0.38
0.63
0.50
0.80
0.94
7
31
64
87
0.78
0.78
0.78
0.78
1.58
1.23
1.02
0.92
Polar
Acetonitrile
Dimethyl sulfoxide
Methanol*
2-Propanol
n-Pentanol
Non-polar
Octane
Decane**
Dodecane
Tetradecane
*Measured at 465 nm.
**Measured at 450 nm.
346
7.8 The Theory of Homogeneous Electron Transfer

The theory of electron transfer is the most developed of that for any reaction in
solution. In addition this type of reaction has been studied extensively by experiment. As a result many of the predictions made by theory have been addressed in
the laboratory. As was seen from earlier discussion, simple electron transfer may
proceed by two mechanisms, inner sphere and outer sphere. The inner sphere
mechanism involves both electron and atom transfer. The discussion here is limited to the outer sphere mechanism, which involves the transfer of a single electron without the formation or breaking of any chemical bonds.
Within the context of transition state theory there are two problems to be
addressed in developing a model to estimate the rate constant for electron transfer. One is the estimation of the Gibbs energy of activation for the process; the
second is estimation of the pre-exponential factor. The model used to describe the
activation barrier is considered rst.
Consider the simple electron transfer process
A B ! A B
7:8:1
+A
A e(
B e(
+B
7:8:2
involving the redox couples

7:8:3
When the reactants are at the site where electron transfer occurs, rearrangement
of the bond lengths and angles within the reactants must take place to achieve a
conguration intermediate between that of the reactants and that of the products.
In addition, there must be some rearrangement of the solvation atmosphere
2
around charged reactants. For example, if CrH2 O3
6 and FeH2 O6 are the
reactants, the length of the metal ionligand bonds must change. In the case of
CrH2 O3
6 this means that the bonds are lengthened to achieve a structure inter2
2
mediate between that of CrH2 O3
6 and CrH2 O6 . In the case of FeH2 O6 , the
length of the metal ionligand bonds decreases to achieve a conguration closer to
that of FeH2 O3
6 . This process is called inner sphere reorganization. At the same
time there must be a readjustment of the solvation atmosphere around these
highly charged ions. According to the Born model (section 3.4), the Gibbs energy
of ion solvation depends on the square of the ionic charge. The solvation atmosphere must readjust to achieve a solvation conguration intermediate between
that of reactants and product. This process is called outer sphere reorganization.
Both inner sphere and outer sphere reorganization contribute to the energy barrier associated with electron transfer. These processes occur randomly due to the
thermal uctuations in the system. According to the FrankCondon restriction,
electron transfer is only possible when this, yet unspecied, intermediate conguration is achieved.
In order to visualize the energy barrier between reactants and products, it is
assumed that each system can be represented as a classical harmonic oscillator
along the reaction coordinate. This is illustrated in g. 7.15. The left-hand parabola gives the Gibbs energy of the reactants and the right-hand parabola, that of
347
Fig. 7.15 Gibbs energy barrier for an

outer sphere electron transfer process with
a standard Gibbs energy change of
G(A). Interaction of the surfaces for
reactants and products leads to electronic
coupling and creates an energy gap 2Jr at
the point where they intersect (B).
the products. There are three important characteristics of this diagram. The rst is
that the curvature of the two parabolas is assumed to be the same. This is equivalent to assuming that the work done to achieve a given non-equilibrium conguration of the reactant system is the same as that in the product system when
measured from the parabolas minimum. This is clearly an approximation which
may not be valid for some systems. A second important feature of the diagram is
the distance between the minima of the two parabolas. This is related to the
relative positions of the two reactants before reaction and determines the height
of the energy barrier along the reaction coordinate. As the distance between the
minima increases the value of the Gibbs energy at the point where they intersect
increases. This corresponds to an increase in the Gibbs energy of activation for
the reaction.
The third feature of the energy barrier diagram shown in g. 7.15 is related to
the point where the Gibbs energy surfaces meet. Electron transfer is not possible if
the individual parabolas remain intact. Electronic coupling of the energy surfaces
is necessary if the system is to pass from the reactant state to the product state.
This is illustrated as a fusing of the lower portions of the two parabolas at the
intersection point to create a gap of height 2Jr between the lower and upper
portions of the curve. The magnitude of Jr determines whether the reaction is
adiabatic or diabatic (non-adiabatic). An adiabatic reaction is one for which the
electronic coupling is strong. This means that every reactant system which reaches
the activated state, that is, the intersection point of the two parabolas, passes
through to the product state if solvent friction is absent. If the electronic coupling
is weak for some reason, so that Jr kT, then the reaction is diabatic and the
passage from the reactant state to products is less probable. An important factor
involved in determining the degree of reaction adiabacity is the distance to which
the species A and B, which exchange an electron, can approach one another. If
this distance is large, then the degree of electronic coupling is small and the
reaction is diabatic. An important example of this type of electron transfer occurs
in biochemistry when one of the redox centers is embedded in a large biomolecule,
348
for example, in a protein. On the other hand, electron transfer reactions involving
2
simple transition metal systems such as CrH2 O3
6 =CrH2 O6 are expected to be
adiabatic because the reactants can approach one another to distances the order
of 500 pm.
The estimation of a rate constant involves determination of the Gibbs activation energy and the pre-exponential factor. An important contribution to the
latter is the Coulombic work which is involved in bringing charged reactants to
the reaction site where the precursor complex is formed. The Gibbs activation
energy and pre-exponential factor are the central components of the transition
state model. Methods of estimating the Gibbs activation energy are discussed rst
of all.
A. Estimation of the Gibbs Activation Energy
Using the properties of the parabolas shown in g. 7.15, one can determine a
relationship between the Gibbs energy of activation and the shape and position of
the parabolas on the reaction coordinate. The equation giving the Gibbs energy of
the reactant system is
GR
kf q qR 2
2
7:8:4
where kf is the force constant describing the curvature of the parabola, q, the
position on the reaction coordinate, and qR, the position of the minimum for the
reactant system. The corresponding equation for the product system is
GP
kf q qP 2
G
2
7:8:5
where qP is the position of the minimum for the parabola corresponding to the
products and G , the standard Gibbs energy change for the reaction. In the case
of homonuclear reactions, G is zero. At the intersection point where q is equal
to qC, the values of GR and GP are equal so that
kf qC qR 2 kf qC qP 2
G
2
2
7:8:6
Solving this equation for qC, one obtains

qC
qP qR
G
2
kf qP qR
The Gibbs energy at the intersection point can now be calculated:

2
k q qR 2 kf qP qR
G
GR qC f C
2
2
2
kf qP qR
7:8:7
7:8:8
In order to simplify the last equation the reorganization energy, El, is introduced,
where
El
kf qP qR 2
2
7:8:9
349
El is the energy that the reacting system must be given to reach the position qP
where there is a minimum on the Gibbs energy surface for the products.
Substituting equation (7.8.9) into equation (7.8.8) and simplifying, one obtains
G GR qC GR qR
El G 2
4El
7:8:10
This is the fundamental relationship in electron transfer theory relating the Gibbs
activation energy to the reorganization energy El and the standard Gibbs energy
change for the reaction G .
There are several points which should be noted about equation (7.8.10). First,
although the theory is developed here within the context of the transition state
model, the estimate of the Gibbs energy of activation is made without factoring
out the degree of freedom associated with passage through the transition state.
Thus, G is similar to the quantity dened in equation (7.4.31). Second, the
rounding o of the barrier due to electronic coupling was neglected in deriving
equation (7.8.10). If this is not a negligible eect, it is easily accounted for by
subtracting JR from the right-hand side of equation (7.8.10). Finally, when the
reaction is homonuclear, so that G is zero,
G
El
4
7:8:11
As was discussed above, the reorganization of the reacting system required to

reach the transition state is made up of two components, namely, an inner sphere
contribution and an outer sphere contribution. It follows that
G Gis Gos
7:8:12
where Gis is the inner sphere reorganization energy, and Gos , the corresponding quantity for outer sphere reorganization. Details regarding estimation
of these quantities are given later in this section.
According to equation (7.8.10) the Gibbs activation energy is a quadratic
function of the driving force for the reaction, namely, G . This has important
consequences when the driving force is large in magnitude. As G increases,
one eventually reaches a value at which it is equal to El. Under these circumstances the reaction proceeds without a Gibbs energy barrier. Such a reaction is
called activationless. Thus, if one moves the right-hand parabola down with
respect to the one on the left, one eventually reaches a position where the righthand parabola crosses the left-hand parabola at its minimum (g. 7.16). If the
value of G is even greater then the right-hand parabola intersects the lefthand one on its left side. Clearly the Gibbs activation energy is again positive in
spite of the large driving force for the reaction. The region in which the rate of
reaction decreases with further increase in G is called the Marcus inverted
region. The variation in G with the standard Gibbs energy change G is
shown in g. 7.17 for typical values of the parameters involved.
An interesting demonstration of the inverted Marcus region has been made
with a reaction involving photo-induced back electron transfer within a radical
ion pair. Thiophene (diaminophenothiazin-5-ium chloride) is a dye molecule
which may be excited to the triplet state using UV light. In the triplet state the
350
Fig. 7.16 Conguration of the Gibbs energy surfaces for reactants and products for (A) an
activation less process and (B) a process in the Marcus inverted region.
thiophene cation is strongly electrophilic and forms an ion pair with an donor
molecule in solution as follows:
3
TH D ! 3 TH

D

7:8:13
where 3TH is the triplet state of the thiophene cation and D, the donor molecule.
When this reaction is carried out in buered methanolic solution, the ion pair
disappears in two reactions [24]. One is a dissociation step which breaks up the ion
pair and gives thiophene in the ground state:
Fig. 7.17 Plot of the Gibbs activation energy G against the Gibbs energy barrier G
for an electron transfer reaction with El equal to 100 kJ mol1.
351

3
TH

D
! TH
D
7:8:14
The other reaction involves electron transfer within the ion pair with subsequent
dissociation:
3
TH

D
! TH D
7:8:15
The rate constant for the electron transfer step has been measured in the presence
of at least eight dierent donors with increasingly positive redox potentials. These
data are summarized in table 7.7.
As the driving force for the reaction increases the rate constant increases,
reaches a maximum, and then decreases again. When the rate of reaction is a
maximum, there is no barrier to electron transfer and the process is activationless.
This condition is reached for a rate constant of 1 109 s1 . For higher driving
forces, the electron transfer reaction occurs in the Marcus inverted region. Other
examples of this behavior have been described in the literature. They give strong
conrmation of the model for electron transfer presented here.
An important task in the theory of electron transfer is estimation of the
reorganization energy El. As discussed earlier, this consists of inner sphere and
outer sphere contributions. These are now considered in more detail.
B. The Inner Sphere Reorganization Energy
The model for the inner sphere reorganization was originally based on simple
electron transfer reactions involving octahedral inorganic complexes. For a
homonuclear electron exchange reaction such as
MH2 Oz1
MH2 Ozn (
+MH2 Ozn MH2 Oz1
n
n
7:8:16
the work required to achieve a common value l* of the MOH2 distance in each
reactant is given by [5]
nk
nk
7:8:17
w A l lA 2 B l lB 2
2
2
Table 7.7 Kinetic Data for Reaction (7.8.15) with Various Donor
Molecules Together with the Standard Potentials for Oxidation of the
Donor Molecule to its Cation Radical [24]
Donor Molecule
N; N; N0 ; N0 -Tetramethyl-P-phenyldiamine
p-Phenylenediamine
p-Aminodiphenylamine
1-Naphthylamine
Diphenylamine
9,10-Dimethylanthracene
1,2,4-Trimethoxybenzene
9-Methylanthracene
Electron Transfer
Rate Constant
107 ket =s1
Standard Redox
Potential
E /V
3.3
3.3
15.0
100
95
42
28
17
0.16
0.18
0.27
0.54
0.83
1.05
1.12
1.16
352
where ki and li are the force constant and equilibrium bond length, respectively, in
reactant i, and n, the number of ligands. When this work is minimized with respect
to l*, one nds that
l
kA lA kB lB
kA kB
7:8:18
Substitution of this expression for l * into equation (7.8.17) leads to the result
w
nkA kB lB lA 2
2kA kB
7:8:19
For this type of system, the entropy associated with inner sphere reorganization is negligible [5]. Therefore, the work can be equated to the corresponding
change in Gibbs energy, so that
Gis NL
nkA kB lB lA 2
2kA kB
7:8:20
where NL is the Avogadro constant.

For a more general reaction, the following formulae have been given by
Marcus [4]:
Gis NL m2 E li
7:8:21
1 G c GP c GR
m
2
2El
7:8:22
where
and
E li
X kjR kjP
lj 2
k
k
jR
jP
j
7:8:23
c GP and c GR are the Coulombic work terms for the reactants and products,
respectively. They are discussed in more detail below in the section dealing with
formation of the precursor complex. kjR and kjP are the force constants of the jth
vibrational coordinate in a species when it participates as a reactant and product,
respectively, and lj is the associated change in bond length. The latter quantities
allow one to calculate the work associated with inner sphere reorganization.
EXAMPLE
Calculate the inner sphere reorganization energy for the homonuclear electron
3
transfer between Fe(H2O)2
6 and Fe(H2O)6 given that the force constant for
1
the Fe(II)OH2 bond is 149 N m and its equilibrium length 221 pm, and that
the corresponding quantities for the Fe(III) complex are 416 N m1 and
205 pm, respectively.
The bond length required in these species to achieve electron transfer is
l
149 221 416 205

209 pm
149 416
7:8:24
353
This result shows that the required bond length is closer to that in the Fe(III)
complex than in the Fe(II) complex.
The inner sphere reorganization energy is
6 149 416205 2212 1024
50:7 kJ mol1
2149 416
7:8:25

The above example illustrates the estimation of Gis for a homonuclear reaction involving octahedral transition metal complexes. Most systems are more
complicated. However, if the details of the changes in bond lengths and bond
angles which accompany electron transfer are known, estimates of Gis can be
made. Quantum-mechanical calculations are often required to obtain the necessary information.
Gis 6:023 1023
C. The Outer Sphere Reorganization Energy

Estimation of the outer sphere reorganization energy is based on a continuum
description of the solvent around the reacting species and therefore is based on the
Born model of ionic solvation. Marcus [25] has described a two-step charging
process in which the non-equilibrium polarization of the dielectric medium is
produced in a reversible manner. The two reactants A and B with radii rA and
rB and charges zA and zB are imagined to be conducting spheres whose centers are
separated by a distance a at the moment of electron transfer. In step 1, the charge
on each reactant is changed from zi to zi0 , a value intermediate between what it has
as reactant and what it has as product. At the same time the surrounding solvent
molecules readjust their orientation to accommodate the change in charge. Since
the reversible work is estimated on the basis of a continuum model, it depends on
the static permittivity of the solvent es. In step 2, the charge on each reactant is
changed back from the intermediate value zi0 to the original value zi , holding the
solvent molecules xed. The medium can only respond to this change through its
polarizability. It follows that the reversible work associated with this step depends
on the optical permittivity of the solvent medium, eop. The Gibbs energy required
to produce the non-equilibrium polarization of the solvent in the transition state
for transfer of one electron was shown by Marcus [4, 25] to be

e2
1
1
1
1
1
7:8:26

Gl0 0
4pe0 2rA 2rB a eop es
where e0 is the permittivity of free space, and e0, the fundamental electronic
charge. The term involving the reciprocal relative permittivities is called the
Pekar factor. Since eop is so much smaller than es, the contribution of the
rapid polarization of the medium at optical frequencies dominates, especially in
solvents with high static permittivities such as water. The outer sphere contribution to the Gibbs energy of activation is equal to
Gos NL m2 Gl0
where the fraction m is given by equation (7.8.22).
7:8:27
354
EXAMPLE
Estimate the outer sphere reorganization energy for the homonuclear electron
3
transfer between FeH2 O2
6 and FeH2 O6 assuming that the radii of the
reactants are 353 and 337 pm, respectively, and that the Coulombic work
terms CGR and CGP are zero. In the transition state the reactants are
assumed to be in contact, so that a is equal to rA rB.
The values of the radii of the aquo complexes are assumed to be 132 pm
larger than the MO bond lengths to account for the contribution of the water
molecules to the overall radius. The distance factor in equation (7.8.26) is
1
1
1
1
1
1

1:45 103 pm1

2rA 2rB a 706 674 690
7:8:28
The Pekar factor is

1
1
1
1

0:550

eop es 1:776 78:46
7:8:29
Thus, the outer sphere contribution to the Gibbs reorganization energy is

Gl0
1:602 1019 2
1:451 109 0:550 1:840 1019 J molec1
4p 8:854 1012
7:8:30
For a homonuclear reaction G is zero. Since the Coulombic work terms
are neglected in this calculation, the factor m is equal to 1/2. It follows from
equation (7.8.27) that the outer sphere contribution to the Gibbs activation
energy is equal to
Gos
6:022 1023 1:840 1019

27:7 kJ mol1
4
7:8:31
This is considerably less than the corresponding inner sphere contribution but
certainly not negligible.
On the basis of the analysis described by Marcus [25], the Gibbs energy Gl0
may be written as the dierence between a Born solvation term at static frequencies and one at optical frequencies. Thus, from equation (7.8.26)

e20
1
e20
1
Gl0

7:8:32
1
1
es
eop
4pe0 Ra
4pe0 Ra
where
1
1
1
1

Ra 2rA 2rB a
7:8:33
The Born model certainly overestimates the solvation term at static frequencies.
Using the MSA to account for the eects of dielectric saturation, equation
(7.8.32) can be rewritten as

e20
1
e20
1
Gl0
1
1

7:8:34
es
eop
4pe0 Ra ds
4pe0 Ra
355
where ds is the MSA parameter which corrects the ionic radii to give a more
realistic estimate of solvation energies (see section 3.5). The parameter ds depends
mainly on the nature of the solvent but also on whether the ion is positive or
negative. This modication of the estimates of Gl0 has been used to discuss
experimentally observed dierences in the rates of one-electron oxidation and
reduction of the same molecule [26].
Other modications of the original Marcus model have been suggested [27].
Many reactants are not spherical in shape and are better approximated as ellipsoids. In this case a much more complex expression for the eective distance R is
obtained which depends on the length of the two axes which describe the shape of
the ellipsoid. Another improvement in the model is to describe each reactant as a
dielectric cavity with xed charges located within it. In this case, the calculation of
Gl0 requires a description of the charge distribution within the reactants and an
estimate of the local permittivity in the dielectric cavity.
Attempts have been made to test the Marcus estimate of Gl0 experimentally.
One way of doing this is to study the variation in electron transfer rate constant
with solvent nature. This results in a change in the Pekar factor and thus in Gl0 .
More will be said about these experiments in section 7.10.
D. The Pre-Exponential Factor
The pre-exponential factor is the most dicult part of the rate constant to estimate. Although fairly good estimates can be made of the Gibbs activation energy
using concepts based on transition state theory, this theory is not applicable to
estimation of the pre-exponential factor in a condensed medium. Factors that
need to be considered in a successful model include the electronic coupling parameter J and the role of solvent friction in determining the rate of crossing the
Gibbs energy barrier. These features were included in a model developed by
Zusman [28].
The pre-exponential factor may be expressed as
A Kp net
7:8:35
where Kp is the equilibrium constant for formation of the encounter complex

between the two reactants, and net is a frequency associated with electron transfer.
The parameter Kp can be estimated from the EigenFuoss model, which was
originally derived to estimate the ion pair formation constant (see section 3.10).
One of the ions is assumed to have a radius a which is equal to the distance
between the reactants at the reaction site. Other ions of the redox couple which
are within or on a sphere of radius a can react. The estimate of Kp also takes into
consideration any electrostatic work done to bring the reactants to the reaction
site. The equation for Kp in units of M1 is
Kp
w
4000pNL a3
el
exp
3
RT
7:8:36
where wel is the electrostatic work and NL , the Avogadro constant, the distance a
being expressed in meters. In innitely dilute solutions the electrostatic work is
356
wel
zA zB NL e20
8pe0 es a
7:8:37
where zA and zB are the valences associated with the reactants A and B, e0 , the
electronic charge, es , the relative permittivity of the solvent, and e0 , the permittivity of free space. At nite solution concentrations, the shielding eect of the
ionic atmosphere should be considered in estimating wel . This can be done using
the extended DebyeHuckel theory or the MSA.
EXAMPLE
Estimate the equilibrium constant for formation of the encounter complex

between the Fe3 and Fe2 ions in water at innite dilution and 25 C, assuming that the distance of closest approach is 690 pm.
The electrostatic work is equal to
wel
3 2 6:022 1023 1:602 1019 2

7:697 kJ mol1
8p 8:854 1012 78:46 6:9 1010
The value of Kp is
4000p 6:022 1023 6:9 1010 3
7:697 103
exp
Kp
3
8:3145 298:2
0:0372 M1
7:8:38
7:8:39
The value of Kp for reactions in which wel is zero is close to unity. When the
reactants are charged with the same sign the repulsion results in a lower value of
Kp. In the estimate of wel made at zero ionic strength, the repulsive eect is
overestimated, and the estimate of Kp is too low.
The frequency associated with electron transfer, net, is estimated by quantum
mechanics and depends on the degree of reaction adiabacity as measured by the
coupling parameters, Jr and the reorganization energy, El. In the case of adiabatic
reactions it also depends on the longitudinal relaxation time of the solvent, tL. A
general expression for net is
nd
7:8:40
net
1 ga
where nd is the frequency associated with a diabatic electron transfer process, and
ga, the adiabacity factor. When ga 1, the value of net is nd/ga, which is the
frequency associated with an adiabatic electron transfer. From quantum
mechanics, the expression for nd is
nd
4p2 J2r
NL h4pEl RT1=2
7:8:41
8p2 J2r tL
NL hEl
7:8:42
and that for ga,

ga
When Jr is large and the reaction adiabatic, the expression for net is
357
net

1=2
nd
1
El
ga tL 16pRT
7:8:43
EXAMPLE
Estimate the electron transfer frequency for a reaction occurring in water at

25 C for which El is 314 kJ mol1, assuming that Jr is 0.1 kJ mol1. Repeat the
calculation for the case that Jr is 5 kJ mol1. Compare these estimates with the
value obtained for an adiabatic electron transfer.
The estimate of the diabatic frequency when Jr is 100 J mol1 .
nd
4p2 1002
6:022 1023 6:626 1034 4p 314 103 8:3145 298:21=2
1:00 1010 s1
(7.8.44)
The value of tL in water at 25 C is 0.4 ps (see section 4.7). Thus the adiabacity parameter is equal to
ga
8p2 1002 4 1013

2:5 103
6:022 1023 6:626 1034 314 103
7:8:45
Since ga is very small, the electron transfer frequency net is equal to the diabatic
frequency, that is, to 1:0 1010 s1 .
When Jr is 5000 J mol1 , the corresponding value of nd is 2:50 1013 s1 .
The adiabacity parameter also increases to 6.3. According to equation (7.8.40)
the electron transfer frequency is now
net
2:50 1013
3:42 1012 s1
1 6:3
7:8:46
For an adiabatic reaction

1
314 103
net
4 1013 16p 8:3145 298:2
!1=2
3:97 1012 s1
Thus, when Jr 5000 J mol1, the value of net

(7.8.40) is very close to the adiabatic value.
7:8:47
estimated by equation
Now, the rate constant for electron transfer may be calculated. All of the
examples considered in this section have related to the reaction between Fe3
and Fe2 , so that the rate constant for this process is estimated now.
EXAMPLE
Estimate the rate constant for electron transfer between Fe3 and Fe2 in
water at 25 C, assuming that Jr is 5 kJ mol1 and using the results obtained
above.
The inner and outer sphere reorganization energies for this reaction are 50.7
and 27.7 kJ mol1, respectively. This gives a value of El equal to 314 kJ mol1.
The lowering of the activation barrier due to interaction of the energy proles
358
for the reactants and products is considered in estimating the Boltzmann

factor. Thus, the Gibbs energy barrier is
G 50:7 27:7 5:0 73:4 kJ mol1
and the Boltzmann factor is
!

G
73:4 103
exp
exp
1:39 1013
RT
8:3145 298:2
7:8:48
7:8:49
The pre-exponential factor is

Kp net 0:0372 3:42 1012 1:27 1011 M1 s1
7:8:50
Thus, the electron transfer rate constant is

ket 1:27 1011 1:39 1013 0:018 M1 s1
7:8:51
This is two orders of magnitude smaller than the experimental value of

4.2 M1 s1 (table 7.1). If the eects of ionic strength on Kp were considered
the theoretical estimate would be higher and closer to the experimental value.
The calculations presented here show that many dierent factors must be
considered in estimating the rate constant. Nevertheless, electron transfer theory
is remarkably successful in describing this elementary solution reaction. Theory
has gone much further than described here, especially in developing the quantummechanical description of electron transfer. More details can be found in recent
reviews [29, 30]. There are other related topics which have not been discussed in
this section. They include, for example, photo-induced electron transfer [30], and
the Marcus cross-relation [5].
7.9 NMR Spectroscopy and Chemical Exchange

Reactions
When a proton is transferred from one molecule to another in solutions, it usually
nds itself in a dierent magnetic environment. As a result, if nite amounts of
both reactant and product are present in solution at equilibrium, the proton can
produce two lines in the NMR spectrum, one corresponding to the proton donor
and the other to the proton acceptor. A second feature of this system relates to the
rate with which the proton is exchanged between the donor and acceptor. If the
frequency with which the proton is transferred is comparable to the radio frequency associated with the NMR spectrometer, then the lines corresponding to
the two species are broadened. The extent of line broadening can be used to
determine the rate constants associated with the exchange process [31].
The most direct explanation of line broadening due to exchange processes is in
terms of the Heisenberg uncertainty principle. If t is the average lifetime associated with the proton in one location, then the uncertainty in the frequency of the
corresponding absorption line n is given by (2pt)1. Of course, the principle
involved here is general and applies to all spectroscopies, including UV-visible
359
and infrared spectroscopies. However, the frequencies associated with the latter
experiments are the order of 1013 to 1015 s1. This is much higher than the frequency associated with most chemical exchange processes. As a result, any line
broadening due to chemical exchange is too small to be seen experimentally. On
the other hand, an NMR spectrometer operating at 100 MHz has a frequency
which is the order of 106 times smaller. Under these conditions the relaxation time
associated with chemical exchange is often shorter than that of relaxation processes due to interaction of the radio frequency eld with the magnetic nuclei in
the system.
In the NMR experiment, the sample is exposed to a magnetic eld B0 which is
imagined to act along the z-axis (see section 5.3). As a result, some of the magnetic
nuclei are in energy levels higher than the lowest level. In the case of protons,
there are two levels (m 12), and the relative population at equilibrium is given
by the Boltzmann fraction, which depends on the magnitude of B0 (equation
(5.35)). The system is now exposed to a rotating magnetic eld B1 in the (x; y)plane that is produced by a radio-frequency wave. The frequency of this eld is
swept over a small range which encompases the Larmor frequencies which characterize the nuclei being studied. When the radio frequency corresponds exactly to
the Larmor frequency of a given nucleus, transitions occur and the equilibrium
Boltzmann distribution is altered. Interactions with uctuating local magnetic
elds in the sample cause relaxation. The relaxation in which the system returns
to equilibrium is characterized by two processes, namely, spinlattice relaxation
which has a relaxation time T1 and spinspin relaxation with a characteristic time
T2. T1 corresponds to relaxation of Mz, the component of the magnetization
along the z-axis in the direction of the eld B0. On the other hand, T2 corresponds
to relaxation of Mx and My, the components of the magnetization in the (x; y)plane (see g. 5.7). For small molecules in solution, local eld uctuations, which
are caused by molecular motions, are fast and occur in the time range 1100 ps;
thus, they are considerably faster than the time characterizing the rotating magnetic eld B1 (10 ns). Under these circumstances T1 and T2 are equal. In viscous
media or in solids where T1 and T2 are longer, the relationship between them is
more complex. In order to obtain quantitative relationships characterizing the
time dependence of Mx, My, and Mz, one must solve the Block equations, which
are described in more detail elsewhere [G2, G4]. A more molecular description of
spinlattice and spinspin relaxation is given in section 5.3.
When a chemical exchange reaction occurs in solution, it can lead to a broadening of the related NMR peaks, as described in the following discussion.
Consider rst of all a proton exchange reaction between protons in dierent
environments denoted as HA and HB:
HA(
+HB
7:9:1
Since the protons are in dierent environments they are expected to produce two
distinct lines in the NMR spectrum. The intensities of these lines reect the
relative concentrations of the protons at equilibrium. On the basis of the theory
of chemical relaxation (section 7.6), the relaxation time associated with proton
exchange is
360
t kf kb 1
7:9:2
Here, kf and kb are the rate constants associated with the forward and backward
reactions in equilibrium (7.9.1). The experiment is usually designed so that the
total proton concentration is very small with respect to the concentration of other
solution components, for example, the solvent. Then the expression for the relaxation time can be rewritten as
1 1
1

t tf tb
7:9:3
where
tf
1
kf
7:9:4
tb
1
kb
7:9:5
and
are the relaxation times associated with the forward and backward processes,
respectively. Equation (7.9.3) can be rearranged as
tt
7:9:6
t f b
tf tb
The fraction of molecules present in the HA environment at any time is
tf
xA
7:9:7
tf tb
and that in the HB environment is
xB
tb
tf tb
7:9:8
At this point it is necessary to relate the relaxation time for the chemical
process to the spinspin relaxation process which is always present when the
system absorbs energy from the radio frequency wave. Both processes lead to
broadening of the NMR absorption peak. If Ti is the total relaxation time for
process i, it may be estimated from T2 and ti using the relationship
1
1
1
Ti T2 ti
7:9:9
The relaxation time is now related to the peak width at half height, (n1/2)i as
follows:
n1=2 i
1
pTi
7:9:10
Thus, from equation (7.9.9)

n1=2 i n1=2 i
1
pti
7:9:11
361
where (1=2 i is the peak width at half height for process i in the absence of
chemical exchange. If (1=2 i can be determined in an experiment without the
system undergoing chemical exchange, then the value of ti can be calculated from
the increase in peak width.
Further analysis requires consideration of the frequencies of the two lines
corresponding to HA and HB, and how the system behaves under conditions
of slow and fast chemical exchange. If these frequencies are designated as nA
and nB then one may dene a frequency n0 which is a weighted average depending
on the fractions of protons present as HA and HB:
n 0 xA n A x B n B
7:9:12
n0 falls in between nA and nB and can be used as a point of reference to discuss

precession of the magnetic vectors of the protons at the two sites. At site A, the
magnetic vector is imagined to precess at a frequency n0 nA ; at B, it precesses in
the opposite direction with a frequency nB n0 . Now the eect of the chemical
exchange on the appearance of the NMR spectrum can be considered in greater
detail (see g. 7.18). Four representative situations are considered from a qualitative point of view, namely, very slow, moderately slow, moderately fast, and
very fast exchange processes.
When the chemical exchange process is very slow, the relaxation times tf and tb
are long. This means that protons at sites A and B precess many times before they
exchange. Thus, there is time for the absorption of energy from the radio frequency eld B1 and sharp lines are seen in the spectrum at nA and nB (g. 7.18
(A)). When the exchange process is moderately slow, two eects are seen (g. 7.18
(B)). The lines for the protons at sites A and B are broader, and the positions of
Fig. 7.18 NMR spectra for chemical

exchange between two symmetrical sites
for the cases of (A) very slow, (B)
moderately slow, (C) moderately fast, and
(D) very fast exchange processes.
362
the maxima shift. The band broadening can be attributed to the uncertainty
principle, which plays a more important role as the lifetime of the proton in a
given state decreases. At the same time, the position of the band changes due to a
corresponding change in the average environment of each proton. The value of
the relaxation time t can be found from the peak separation nA nB . If nA nB 0
is the peak separation in the absence of exchange broadening (t 1), then
"
#1=2
A nB
1
1 2 2
nA nB 0
2p t nA nB 20
7:9:13
If the process is symmetrical (tf tb), then the relaxation time in each direction is
equal to t/2. In writing equation (7.9.13) it is assumed that the absorption peaks
have a Lorentzian shape. In addition, the line-broadening eects illustrated in g.
7.18 are those for a symmetrical process.
Eventually, as the exchange process becomes faster, the two bands coalesce to
form one peak at frequency n0. The relaxation time for this condition, tc is given
by
tc
1
21=2 pn
nB 0
7:9:14
Further increase in the exchange rate results in sharpening of the band at n0.
When this process is very fast a sharp line is observed at frequency n0 whose
width is determined by spinspin relaxation in the absence of chemical relaxation
(t 0). Under these circumstances a nucleus in site A cannot precess to a signicant extent before it leaves A to enter site B. From the point of view of the
rotating frame of reference dened with respect to n0, the proton is essentially
stationary. An important method of observing the eects of chemical exchange
illustrated in g. 7.18 is to change the temperature of the system. At low temperatures, the exchange is very slow. As the temperature increases, the exchange
speeds up and the various stages from very slow to very fast are seen.
It is important to understand what time range for kinetic experiments is accessible by NMR. The frequency dierence n which can be resolved in 1H NMR is
of the order of 101000 Hz. When the exchange process is very slow, the relaxation time t is very much greater than n1 (1030.1 s); for a very fast process t is
much less than n1. Thus, values of t in the range 1041 s can be measured
using line-broadening experiments. Other methods are used to determine rates
which cannot be measured by line broadening experiments. One of these is
based on observation of the absorption signal intensity at one of the sites in an
exchanging system, while absorption at the other site is saturated by using a
second radio frequency eld tuned at the required frequency. During the irradiation the absorption intensity falls o to a new steady-state value due to transfer
from the irradiated site. The rate constant for the decay process can be used to
calculate the lifetime of the observed state. More information about this technique, which is called spin relaxation transfer, and other methods used to determine
the kinetics of exchange processes using NMR, can be found in monographs
devoted to dynamic NMR spectroscopy [32, 33].
363
H NMR has been used extensively to study the kinetics of proton transfer
reactions in water and other solvents. In the case of the protolysis reaction
HA S(
+A HS
7:9:15
where S represents the solvent, and HS , its protonated form (lyonium ion), the
relaxation time is given by
t kf cHAe cSe kb cAe cHSe 1
7:9:16
Normally, the equilibrium solvent concentration cSe is much higher than the
concentrations of the other species involved in reaction (7.9.15). Furthermore,
the experiment is designed with the concentrations cHAe and cAe suciently small
that the relaxation times associated with proton exchange can be measured in the
experiment. In addition, the equilibrium concentration of the lyonium ion cHSe is
much less than that of the conjugate base A. As a result, equation (7.9.16)
becomes
t kf cSe kb cAe 1
7:9:17
Since the solvent concentration is normally constant, the term kf cSe is treated as a
rst-order rate constant with units of s1. Also the concentration of A is a
known parameter in the experiment, so that the value of kb as a second-order
rate constant can be extracted from the relaxation time tb associated with the
back reaction. The same approach is used to determine the kinetics of hydrolysis
reactions. The general hydrolysis reaction is written as
B SH(
+BH S
7:9:18
where B is the Brnsted base and SH, the solvent which can donate a proton.
Under conditions that the equilibrium concentration of B is very small with
respect to that of the solvent, and that of the lyanate ion S is very small
with respect to that of BH, the expression for the relaxation time is
t kf cSHe kb cBHe 1
7:9:19
Data for some protolysis and hydrolysis reactions in water and methanol are
summarized in table 7.8. The rate constants for the backward reactions are all
very fast, whereas those for the forward reactions depend very much on the nature
of the reaction. When these data are combined with those in table 7.2 a more
complete description of proton transfer reactions in protic solvents is available. It
is clear that the solvent plays an important role in these processes.
A very eective way of studying a chemical exchange process is to determine
line widths as a function of temperature. This has been applied with success to the
determination of the kinetics of ligand exchange processes [31]. For example, the
kinetics of exchange of acetonitrile (AcN) as a ligand in [Ni(AcN)6]2 with acetonitrile as solvent has been determined by examining the temperature dependence
of the 1H NMR spectra of the ligand [34]. The quantity determined experimentally is

1 1
1
Y

7:9:20
xX T2 T2;0
364
Table 7.8 Kinetic Data Obtained by 1H NMR for Proton Transfer

Reactions in Water and Methanol at 25 C [32]
Reaction
kf cse =s1
kb =M1 s1
6 1011
24.5
4.7
2:1 1011
1:1 106
4:3 1010
3:0 1010
1:7 1010
17:6 1010
0.7
1:85 1010
6 109
Water
H3 O H2 O H2 O H3 O
NH
4 H2 O NH3 H3 O
CH3 3 NH H2 O CH3 3 N H3 O
OH H2 O H2 O OH
CH3 3 N H2 O CH3 3 NH OH
Methanol
CH3 OH
2 CH3 OH CH3 OH CH3 OH2
CH3 3 NH CH3 OH CH3 3 N CH3 OH
2
CH3 O CH3 OH CH3 OH CH3 O
*For totally symmetrical reactions, kf kb :
where xX is the mole fraction of AcN bound to the transition metal cation, in this
case, Ni2, T2, the relaxation time determined from the width at half-height for
the 1H peaks in the NMR spectrum, and T2,0, the same quantity determined in
pure AcN. The results of this experiment are shown in g. 7.19 as a plot of Y
against the reciprocal of the absolute temperature in the range 230420 K. Four
regions are clearly distinguishable in which the variation in y with 1/T is approximately linear. The two regions with steep slopes are connected with the ligand
exchange process, and the rst and last region with variation in the spinspin
relaxation time with temperature. Each of these parts of the plot in g. 7.19 is now
discussed in more detail.
In region I where the temperature is very low, the ligand exchange process is so
slow that it does not contribute to the line broadening. The mechanism of line
broadening involves T2,0, that is, spinspin relaxation related to the protons in the
acetonitrile molecule. As the temperature increases, this process speeds up and the
line width decreases. By extrapolating the line drawn through the data obtained in
this region, estimates of T2,0 at higher temperature are obtained.
Fig. 7.19 Plot of the parameter Y as a

function of 1=T for an exchange process
in regions of (I) negligible exchange, (II)
moderately slow exchange, (III)
moderately fast exchange, and (IV) very
fast exchange.
365
As the temperature increases, the ligand exchange process begins to contribute

to the observed line width. This results in the increase in y observed in region II.
In fact, the parameter y in this region is simply given by
Y
1
tx
7:9:21
where tX is the relaxation time for ligand exchange. The activation parameters for
this process may be determined from the slope of the linear plot in region II.
EXAMPLE
Given that the value of tX is 0.76 ms at 288.2 K and 0.19 ms at 308.2 K,

estimate the enthalpy and entropy of activation for the ligand exchange
process.
Neglecting medium eects, the relationship between tX and the activation
parameters is
6

1
kB T
S
6 H
exp
7:9:22
exp
R
RT
tx
h
The ratio of the values of tX measured at the two temperatures is

0:76 308:2
6 H
6 H
exp
0:19 288:2
308:2R 288:2R
7:9:23
Solving this equation, one obtains a value of the enthalpy of activation equal to
49.0 kJ mol1. When this result is substituted into equation (7.9.17), the result
for the entropy of activation is 15.1 J K1 mol1.
The maximum line width for the system being considered occurs at 313 K. At
this temperature the two lines being followed in the NMR spectrum coalesce.
Further increase in temperature results in a decrease in line width as illustrated
experimentally in region III. The parameter y in this region is given by
Y
1
o2 tx
T2X
7:9:24
where T2X is the relaxation time of the proton in the ligand and
o 2pnX nS
7:9:25
nX being the chemical shift for the proton in the ligand, and nS, the value in the
bulk solvent. Over most of the region the main contribution to Y comes from the
second term, which is proportional to (o)2. In region IV, the exchange is so fast
that the broadening eects related to exchange disappear. This means that o is
zero and the parameter Y is given by the reciprocal of the spinspin relaxation
time T2. This type of experiment has been used succesfully to determine the kinetic
parameters associated with ligand exchange reactions.
Another application of NMR relaxation experiments is the study of intramolecular processes. The classic example is rotation about the CN bond in
amides. Since this bond is partially double in character, the rotation requires
energy and occurs at a nite rate. In addition, it results in exchange of protons
in dierent chemical environments. For the case of NN dimethylacetamide
366
Fig. 7.20 Equivalent structures for N,N-dimethylacetamide illustrating the partial double
bond character of the CN bond.
(DMA), the protons on the methyl groups attached to the nitrogen atom are
either cis- or trans to the carbonyl group (g. 7.20). As a result, the room
temperature 1H NMR spectrum has two lines corresponding to the dierent
environments. At higher temperatures, rapid internal rotation makes the protons of the two methyl groups equivalent and only one peak is observed.
Analysis of the spectral data leads to a value of 50 kJ mol1 for the energy
barrier associated with the exchange process.
The examples chosen here have emphasized the use of 1H NMR. However,
other nuclei can be examined, including 13C, 14N, and 17O. Thus, in the case of
ligand exchange processes involving acetonitrile, one can use 13C NMR with an
isotope-enriched sample [G4]. In addition, the discussion here has been limited to
a few examples of kinetic processes which have been studied using NMR spectroscopy. Closely related experimental techniques are electron paramagnetic resonance and Raman spectroscopies, which are discussed more fully by Strehlow
[G4].
7.10 Medium Effects in Solution Reactions

Since the components of a reaction in solution are surrounded by other species,
most of which do not react, the nature of the intermolecular interactions experienced by the reactants and products can have an important inuence on the Gibbs
energy surfaces, and thus, on the kinetics of the reaction. The most important
nearest neighbor in most cases is the solvent. Thus, the study of solvent eects on
the rate and mechanism of solution reactions has been an important area of
experimental research. Interactions with other solute components may also be
important. This is especially true for reactions involving ions.
Since the equilibrium properties of the medium are involved in determining
these eects they are called static medium eects. However, there is another type
of relevant phenomenon which is related to the dynamic properties of a condensed phase. Since the movement of molecules with respect to one another is
required for the reaction to take place, the local viscosity of the system can also
inuence the rate of reaction. This property is also related to local intermolecular
forces. Eects which depend on local viscosity have also been studied experimentally and are known as dynamic medium eects.
In the following discussion, important examples of both static and dynamic
medium eects are described and discussed with respect to data in the literature.
367
A. Static Medium Effects

As pointed out in section 7.4, the medium eect is described within transition state
theory in terms of the activity coecient ratio in the expression for the rate
constant. Thus, on the basis of equation (7.4.13), the rate constant for a simple
bimolecular reaction may be written as
kr kr 0
y A yB
yx6
7:10:1
where kr 0 is the rate constant in the absence of medium eects, yA and yB, the
activity coecients of the reactants A and B on the molarity scale, and yx6 , the
activity coecient of the activated complex.
The most direct method of changing the medium is to change the solvent in
which the reaction is carried out. The reactions studied are usually limited to
organic and organometallic systems for reasons of solubility. Solvent eects
have been studied for a wide variety of reactions [35] including electron transfer
[36, 37], proton transfer [8, 32], and atom transfer reactions [12]. In the case of
electron transfer reactions, the role of the solvent is expressed directly in the outer
sphere Gibbs activation energy Gos . This depends on the Pekar factor and thus
on the static and optical values of the solvents relative permittivity (section 7.8C),
and has been investigated in detail for these reactions [36]. However, dynamic
medium eects may also be important for electron transfer reactions and are
expressed through the longitudinal relaxation time tL in the pre-exponential factor (section 7.8D). For this reason, solvent eects in electron transfer reactions are
discussed later in this section. Proton transfer reactions have not only been studied in water but also in the alcohols and other protic solvents [8, 32]. As one
would expect, the Gibbs energy barrier and the rate constants for these processes
depend on strength of the intramolecular forces which aect the proton in each
environment. Solvent eects for atom transfer reactions were illustrated earlier in
this chapter (section 7.3D). In the example chosen, an anion is involved in a SN2
reaction. As a result the solvents acidity plays a role in determining the Gibbs
energy barrier for the process.
Another well-known example of a medium eect in solution kinetics is the
eect of ionic strength on reactions between ions [38, 39]. This is normally treated
using the extended DebyeHuckel theory to estimate the activity coecients in
equation (7.10.1). The activity coecient for species i is given by
ln yi
ADH z2i I1=2

1 BDH ai I1=2
7:10:2
ADH and BDH are the DebyeHuckel constants (see section 3.8) and I is the ionic
strength. zi is the charge on species i and ai, the ion size parameter. If it is assumed
that the individual values of ai can be replaced by an average value, a, and that z6
is equal to zA zB, then
ln
yA yB 2ADH zA zB I1=2
yx6
1 BDH aI1=2
7:10:3
368
It follows that
ln kr ln kr 0
2ADH zA zB I1=2
1 BDH aI1=2
7:10:4
so that ln kr should be a linear function of I1/2 in the limit of very low ionic
strengths.
Plots of ln(kr/kr 0) against I1/2 for a series of reactions with dierent values of
zAzB are shown in g. 7.21. In the case that the ions have the same sign (zAzB> 0),
considerable reaction acceleration is observed when the ionic strength increases.
This is clearly a result of the screening eect of the inert electrolyte on the repulsion between the reactants. When the ions have opposite signs, the eect is
reversed and the observed rate constant decreases as ionic strength increases.
The slopes of the lines drawn in g. 7.21 reect the change in the value of zAzB.
The range over which the extended DebyeHuckel theory is applicable is limited
to ionic strengths equal to 0:1 M. This range can be extended signicantly when
the MSA model is used to estimate the activity coecients [40].
The study of static medium eects is obviously important because it provides
direct information regarding the factors which inuence the Gibbs energy barrier
for the reaction. These factors are important in deciding the mechanism of complex reactions. As a result, medium eects are often used by the synthetic chemist
to determine the nal product of a reaction in solution.
As already pointed out, the equilibrium environments are not the only factor
that must be considered in assessing the overall medium eect. During the course
of the reaction these environments change, and molecules and ions move with
respect to one another. The friction involved in this movement can also play a role
in determining the rate constant, as is discussed now.
Fig. 7.21 Plots of log10 (kr/kr0)

against the square root of the ionic
strength for ionic reactions of
various types in water: (A)
CoNH3 5 Br2 Hg2 zA zB 4;

(B) S2 O2
8 I zA zB 2;
(C) COOC2 H5 N : NO2

OH zA zB 1; (D) [Cr(urea)6]3
H2O (), CH3COOC2H5 OH (),
(zAzB 0); (E) H3O Br
H2O2 (zAzB 1); (F)
CoNH3 5 Br2 OH zA zB 2;
and (G) Fe2 CoC2 3
4 (zAzB
6). The lines are drawn with slopes
equal to 2zAzB. (From reference G1,
with permission.)
369
B. Dynamic Medium Effects

Dynamic medium eects in solution kinetics were rst recognized by Kramers
[41]. He treated the problem on the basis of the Langevin equation [42] according
to which the velocity of the reactants along the reaction coordinate and the
friction of the surrounding medium play a role. Details of Kramers theory are
not given here but an introduction to this subject can be found elsewhere [G3].
The parameters involved in quantitatively assessing the dynamic solvent eect are
the frequency associated with the shape of the barrier of the transition state n6
and a friction parameter z6 which is related to solvent viscosity.
From the solution of the Langevin equation, Kramers recognized three
important regimes. For moderate friction, the system behaves as if dynamic
solvent eects were absent. This means that the equilibrium conditions assumed
in the standard formulation of transition state theory are satised and the
reactants move freely over the potential energy barrier to the product state. As
solvent friction increases, the second regime is reached. As a result of collision
with the viscous solvent molecules, the crossing of the Gibbs energy barrier is not
always successful. Thus, the barrier may be crossed several times before the
system succesfully reaches the product state. This means that the eective barrier
crossing frequency is less than n6 with a corresponding decrease in the rate
constant. The magnitude of the eect depends not only on the magnitude of
the friction but also on the shape of the barrier. The third regime corresponds
to systems with very low solvent friction. Since the reactants are now in an
environment in which the solvent appears to be absent, an equilibrium distribution of reactant states is not maintained, and transition state theory is no longer
valid.
The results of Kramers theory [41] have been elaborated in detail by Hynes
[G3]. A brief introduction to this subject is given here. As a result of dynamic
friction the observed rate constant kr can fall bellow the transition state value
kTST, that is,
kr kT kTST
7:10:5
where kT is the Kramers transmission coecient. The rate constant according to

transition state theory is given by

6 G
kTST nR exp
7:10:6
RT
where nR is a frequency associated with the Gibbs energy surface for the reactants
and6 G , the standard Gibbs energy of activation. According to equation
(7.4.13), the expression for nR is
nR
kB T yA yB
h yX6
7:10:7
Static medium eects appear in equation (7.10.6) through nR and 6G , both of
which depend on solvent properties. After solving the Langevin equation, the
following expression is obtained for kT:
370
"
z6
kT 1
4pn6
2 #1=2

z6

4pn6
7:10:8
z6 is the friction parameter related to the viscosity of the medium, and n6, a
frequency associated with the Gibbs energy barrier. If one imagines that the
Gibbs energy barrier has a parabolic shape (see g. 7.3), then n6 can be calculated
from the equation describing the parabola. The exact expression for z6 depends
on the nature of the reaction but it is clear that it must have units of s1 because
kT is a dimensionless fraction.
In the case of moderate friction, z6/4pn6 1 and there is no dynamic solvent
eect. On the basis of equation (7.10.8), kT is equal to unity and the rate constant
kr is given by its transition state theory value, kTST. However, if solvent friction is
large, then z6/4pn6 1, and the expression for the transmission coecient is
!1=2
z6
z6
16p2 n26
2pn6
kT
1
7:10:9
2
4pn6
4pn6
z6
z6
In this limit, the dynamic solvent eect can also be treated as a diusion process
using the Smoluchowski equation. This emphasizes the fact that the friction leads
to multiple crossings of the Gibbs energy barrier. The diusion coecient associated with this process is given by
D6
kB T
n6 z6
7:10:10
where m6 is the reduced mass of the reacting system. Expressing the transmission
coecient in terms of D6, one obtains
kT
2pn6 m6 D6
kB T
7:10:11
Hynes [G3] has discussed the application of this theory to a wide variety of
solution reactions, including electron transfer reactions, atom transfer reactions
involving both neutral and charged species, and isomerization reactions. The
following discussion is limited to electron transfer processes.
An important parameter in the Hynes treatment of the dynamic solvent eect is
the frequency associated with motion of the reactant system in the corresponding
potential energy well. On the basis of the dielectric continuum model, this frequency is given by

2es e1 kB T 1=2
nR
7:10:12
62 gK e1 Irot
where es and e1 are the static and high-frequency values of the solvents relative
permittivity, respectively, gK, the Kirkwood correlation parameter (see section
4.4), and Irot, the eective moment of inertia of an individual solvent molecule
[29]. Values of nR for selected solvents which approximately follow the Debye
equation with a single relaxation process are given in table 7.9 together with the
solvent molecules moment of inertia and the reciprocal of the longitudinal relaxa-
371
Table 7.9 Reactant System Characteristic

Frequencies in Selected Debye Solvents Estimated
Using Equation (7.10.12) Together with the Solvent
Molecules Effective Moment of Inertia and the
Reciprocal of the Longitudinal Relaxation Time
Solvent
Acetone
Acetonitrile
Benzonitrile
Dimethyl sulfoxide
Nitromethan
1047 Irot =kg m2
1
t1
L =ps
nR =ps1
62.51
9.1
303.9
93.13
9.0
3.4
5.0
0.17
0.47
4.5
1.3
4.6
0.52
0.91
4.5
1
tion time t1
L . It is apparent that nR is close to tL in magnitude and follows the
1
same trend, varying from 0.1 ps in slow solvents to 5 ps1 in fast solvents such
as acetonitrile. These results are all smaller than the gas phase transition state
frequency kB T=h, which is equal to 6.2 ps1 at room temperature.
EXAMPLE
Estimate the characteristic reaction frequency nR in acetonitrile given that the

eective moment of inertia of this molecule is 9:1 1047 kg m2.
The ratio (kB T=Irot 1=2 at room temperature is
!1=2

kB T 1=2
4:115 1021
6:72 1012
7:10:13
Irot
9:1 1047
The dielectric parameters for acetonitrile are es 35.9 and e1 2.26 (table
4.2). The Kirkwood correlation parameter is gK 1.18 (table 4.4). The dielectric factor in equation (7.10.12) may now be estimated:

1=2

2es e1 1=2
2 35:9 2:26
0:685
7:10:14
6p2 gK e1
63:14162 1:18 2:26
The value of nR is
nR 0:685 6:72 1012 4:60 ps1
7:10:15
The longitudinal relaxation time for acetonitrile given a Debye relaxation time
of 3.2 ps is
tL
2:26
3:2 0:20 ps
35:9
7:10:16
Its reciprocal value is 5.0 ps1. This value is very close to that of nR.
Hynes [43] has discussed dynamic solvent eects for electron transfer reactions
and described the role of solvent friction for both diabatic and adiabatic reactions. In the case of diabatic reactions the rate is strongly dependent on the
coupling between the energy surfaces for the reactants and products as expressed
through the parameter Jr (see section 7.8D). When Jr is very small, dynamic
372
solvent eects are normally not important and the pre-exponential factor is given
by equations (7.8.35) and (7.8.40). The latter equation is derived from quantum
mechanics for the case that movement from the reactant energy surface to the
product surface at the intersection point is the rate-determining step [44].
As interaction between the two energy surfaces increases, the character of the
reaction changes from diabatic to adiabatic. This interaction aects the shape of
the cusp-shaped barrier associated with the transition state and thus the value of
n6. In the case of a simple Debye solvent the friction parameter is given by
z6 4p2 n2R tL
7:10:17
The expression for the Kramers transmission coecient for adiabatic electron
transfer is therefore
n6
7:10:18
kT
2pn2R tL
Hynes [43] also derived a relationship between n6 and nR for this type of
reaction, which is

n6
pEl 1=2
7:10:19
nR
4RT
where El is the reorganization energy for electron transfer (equation (7.8.9)).
Substituting this into equation (7.10.18) the expression for kT becomes

1=2
1
El
kT
7:10:20
nR tL 16pRT
The corresponding expression for the adiabatic pre-exponential frequency for
electron transfer is

1=2
1
El
net kT nR
7:10:21
tL 16pRT
This is identical with the expression derived by Zusman [28] which was discussed
in section 7.8D.
EXAMPLE
Estimate the frequency associated with the transition state barrier for an adiabatic electron transfer in acetonitrile assuming that El is 100 kJ mol1.
The ratio n6/nR is estimated using equation (7.10.19):
!1=2
n6
p 100 103
5:63
7:10:22
nR
4 8:3145 298:2
Using the value of nR for acetonitrile given in table 7.9, the frequency
associated with the transition state is
n6 5:63 4:60 26:0 ps1
7:10:23
Hynes [43] has also discussed the dynamic solvent eect for solvents with more
complex dielectric relaxation behavior. In the following it is assumed that the
373
simple Debye model is sucient to describe the dynamic medium eect. When one
compares the nal expression for diabatic and adiabatic reactions, one sees that
the solvent eect enters in quite dierent ways. For a diabatic electron transfer
reaction the rate constant can be expressed as

Ad
G
exp
kd
7:10:24
RT
E RT1=2
where the standard Gibbs energy of activation G depends on solvent nature
through the outer sphere contribution Gos , and Ad accounts for the solventindependent terms in the pre-exponential factor (see equation (7.8.41)). The outer
sphere contribution Gos depends on the Pekar factor, 1/eop 1/es (see equation
(7.8.26)). Because of the dominating eect of the exponential term, it is reasonable
to expect ln ket to be linear in the Pekar factor for diabatic reactions with small
values of Gis .
On the other hand, if the electron transfer reaction is adiabatic, the dependence
of the rate constant on solvent nature is quite dierent:
1=2

E
G
1
ka Aa
tL exp
7:10:25
RT
RT
In this equation, Aa gives the solvent-independent portion of the pre-exponential
factor. Separation of the solvent contributions to the exponential and pre-exponential terms in the rate constant is much more dicult for adiabatic reactions.
However, if the outer sphere contribution to G is small, then the contribution
from t1
L in the pre-exponential factor dominates. Under these circumstances a
plot of the logarithm of the electron transfer rate constant ln ket should be linear
in ln tL.
Grampp and Jaenicke [36] have carried out detailed studies of solvent eects on
the kinetics of electron transfer for a variety of organic redox couples. The kinetic
data were obtained by electron spin resonance (ESR) line-broadening experiments
for very fast reactions with rates close to the diusion-limited value. Some of the
reactions were clearly diabatic, and others adiabatic on the basis of the solvent
eect.
An example of a diabatic reaction is the oxidation of tetramethyl-p-diaminobenzene (TMDAB) to its cation radical. Kinetic data were obtained in seven
solvents with Pekar factors ranging from 0.27 (chloroform) to 0.53 (acetonitrile).
The Pekar factor gP is dened as
1
1
gP

7:10:26
eop es
For the TMDAB, the logarithm of the electron transfer rate constant ln ket is
clearly a linear function of the Pekar factor gP (g. 7.22). On the other hand, a
plot of ln ket against ln tL for the same data yields a weak linear correlation with a
positive slope (g. 7.23). This correlation is the opposite to that expected; in other
words, the rate constant should decrease with increase in tL if the reactions were
adiabatic. The estimate of Gis for this reaction is 6.4 kJ mol1; the value of
Gos depends on the nature of the solvent but it is close to Gis in most solvents
considered. The reason that the reaction is diabatic is clearly due to the weak
374
Fig. 7.22 Plots of the logarithm of the electron transfer rate constant for oxidation of
TMDAB () and TTF () against gP , the Pekar factor of the solvent [36].
interaction between the Gibbs energy surfaces for reactants and products which
results in a small value of the interaction parameter Jr .
Quite dierent behavior is found for the electron transfer reaction between
tetrathiofulvalene (TTF) and its cation radical. For this system, there is no correlation between ln ket and the Pekar factor g (g. 7.22). However, there is a
strong correlation between ln ket and ln tL. These results are consistant with an
adiabatic reaction with a dynamic solvent eect. The estimate of Gis for this
Fig. 7.23 Plots of the logarithm of the electron transfer rate constant for oxidation of TTF
(^) and TMDAB () against the logarithm of the solvents longitudinal relaxation
time [36].
375
reaction is 4.9 kJ mol1. The corresponding value of Gos is 9g. Since g is close to
0.5 for many solvents, the outer sphere reorganization energy is close to the inner
sphere value. For this system, interaction between the Gibbs energy surfaces is
stronger, so that the reaction follows an adiabatic path. When the reaction parameters are such that it is neither strongly adiabetic or strongly diabetic, analysis of
the solvent eect is more dicult.
Dynamic solvent eects have been studied for other solution reactions, including proton transfer reactions, isomerizations, and time-resolved uorescence
studies. This is an important area of modern research involving ultrafast kinetic
experiments in solution.
7.11 Linear Gibbs Energy Relationships

As the study of solution kinetics advanced, data were accumulated for specic
types of reactions in which the nature of one of the reactants was changed in a
systematic way. An example is a reaction involving an acid which is studied for a
series of acids of varying acidity constant. As the body of data increased in size it
was observed that the rate constant for the reaction was often a linear function of
an equilibrium constant associated with the overall process, for example, the
acidity constant. This correlation is called the Brnsted relationship and can be
written as
ln kr ln kr 0 bB ln Ke
7:11:1
where kr is the rate constant, Ke, the equilibrium constant, bB, the Brnsted slope,
and kr 0, the rate constant when the equilibrium constant is unity.
An early description of such a relationship was given by Brnsted and
Pederson [45] who presented data for the base-catalyzed decomposition of nitramide (H2N2O2). They found that the logarithm of the rate constant for this
reaction is linear in the acidity constant for the conjugate acid HB of the base
B involved. Their data, which are shown in g. 7.24, demonstrate that the linear
relationship between the two experimental parameters holds over a very wide
range. The slope of this line, 0.78, gives the Brnsted coecient bB. For this
system the rate constant increases with increase in basicity of the anion involved
in the reaction.
Although the Brnsted relation was rst discussed with respect to proton
transfer reactions, it has been found to apply to other types of reactions including
electron and atom transfer reactions. Equation (7.11.1) implies that there is also a
linear relationship between the Gibbs activation energy for the reaction and the
standard Gibbs energy change for the associated equilibrium. Thus, one may also
write
6 G 6 G0 bB G
7:11:2
where 6 G is the standard Gibbs energy of activation, G , the standard Gibbs
energy change for the equilibrium, and 6 G0 , the standard Gibbs energy of
activation when G is zero. The relationship of the Brnsted coecient to the
fundamental parameters of an electron transfer reaction is easily derived on the
376
Fig. 7.24 Plot of the logarithm of the rate constant for the decomposition of nitramide
against the logarithm of the basicity constant for the base used as catalyst [45].
basis of equation (7.8.10). For a series of electron transfer reactions in which the
reorganization energy El is independent of G, one obtains
bB
d6 G 1 G

dG
2 2El
7:11:3
When the reaction is homonuclear (G 0), the Brnsted coecient is exactly
equal to 1/2. Equations similar to equation (7.8.10) have also been used to estimate the Brnsted slope for proton and atom transfer reactions [46].
Another linear Gibbs energy relationship which is used extensively in organic
chemistry is the Hammett equation [47]. This equation is used to relate the rate
constants and equlibrium constants for reactions of meta- and para-substituted
benzene derivatives. The Hammett equation for the rate constants of a given
reaction series is written as
log10 kX log10 k0 sX rH
7:11:4
where kX is the rate constant for the benzene derivative with substituent X, k0, the
rate constant for the unsubstituted compound, sX, the substituent parameter, and
rH, the reaction parameter. The values of sX are based on the dissociation constants of substituted benzoic acids, so that
sX log10 KaX log10 Ka0
7:11:5
where KaX is the acidity constant for the benzoic acid with substituent X and Ka0,
that for unsubstituted benzoic acid. Some values of sX for typical substituents in
the meta and para positions are given in table 7.10. When sX is negative, the
substituents are electron donating and push electron density into the benzene ring.
As a result, the COOH group becomes more electronegative and is able to hold
on to the proton more easily.
377
Table 7.10 Hammett

Substituent Parameters [48]
Position
Substituent
Group
OH
OCH3
CH3
H
F
Cl
Br
I
CN
NO2
Meta
Para
0.12
0.115
0.07
0
0.34
0.37
0.39
0.35
0.56
0.71
0.37
0.27
0.17
0
0.06
0.23
0.23
0.18
0.66
0.78
Substituents with positive values of sX are electron withdrawing and have the
opposite eect. The benzoate anion is stabilized with respect to the acid and the
acidity constant is higher.
An example of application of the Hammett relation is shown in g. 7.25. Data
for the rate constant for bromination of substituted benzeneboronic acids in 20%
aqueous acetic acid show an excellent linear correlation with the Hammett sX
parameter for both para and meta substituents. The reaction parameter rH for
this system is 4.52. The Hammett equation has been shown to successfully
describe kinetic and equilibrium data for over 300 aromatic reactions [49].
Following the original work of Hammett [47] several other linear Gibbs energy
relationships have been derived. The most important of these is the Taft relationship [50] which allows consideration of aliphatic as well as aromatic systems and
also of steric eects. The Taft equation has the same form as equation (7.11.4) but
the substituent constants are dened in another way. More information about this
relationship and other linear Gibbs energy relationships can be found in the
monograph by Hammett [48].
Obviously, linear Gibbs energy relationships are quite useful because of their
predictive power. However, it must be remembered that they are based on fundamental concepts which originate in the theory of elementary reactions. Thus, the
experimental study of these relationships has helped elucidate the energy proles
involved in the basic processes which occur in solutions.

The study of reaction kinetics in solution has seen tremendous advances since the
1960s. This progress has been recognized by four Nobel Prizes. In 1967, Manfred
Eigen, Ronald Norrish, and Sir George Porter received the Nobel Prize for their
studies of extremely fast reactions aected by disturbing a solution equilibrium by
378
Fig. 7.25 Plot of the logarithm of the rate constant for bromination of a substituted
benzeneboronic acid relative to that of the unsubstituted acid, log10(kr/kr0) against the
Hammett substituent parameter, sX.
means of very short pulses of energy. Their work was carried out before lasers
were used in this type of study. Henry Taube received the Nobel Prize in 1983 for
his studies of the mechanism of electron transfer reactions, especially those involving metal complexes. In 1992, Rudy Marcus was given the Nobel Prize for his
work on the theory of electron transfer reactions. The most recent Nobel Prize
was awarded in 1999 to Ahmed Zewail for his studies of the transition states of
chemical reactions using femtosecond spectroscopy. Resolution of chemical
events on such short time scales emphasizes the great progress made in this
area of chemistry.
The discussion in this chapter has focused on fundamental types of solution
reactions including electron transfer, proton transfer, atom transfer, and ligand
transfer reactions. The theory of these processes is of great fundamental interest
and has been discussed in detail for electron transfer processes. Of course, not all
processes studied in solution could be covered in this chapter. A notable example
is oscillation reactions, which have also attracted considerable attention.
The remarkable progress made in this eld clearly has been the result of both
experimental and theoretical advances. As theory asked new questions about
fundamental processes, new and more powerful experiments were designed to
answer them. As a result, this eld now has developed a molecular-level understanding of the dynamics of reactions in condensed phases.
General References
G1. Laidler, K. J. Chemical Kinetics, 3rd ed.; Harper and Row: New York, 1987.
G2. Connors, K. A. Chemical Kinetics, The Study of Reaction Rates in Solution; VCH
Publishers: New York, 1990.
379
G3. Hynes, J. T. In Theory of Chemical Reaction Dynamics; Baer, M., Ed.; CRC Press:
Boca Raton, FL, 1985; Vol. 4, Chapter 4.
G4. Strehlow, H. Rapid Reactions in Solution; VCH: Weinheim, 1992.
References
1. Michaelis, L.; Menten, M. L. Biochem. Z. 1913, 49, 333.
2. Noyes, R. M. Progress in Reaction Kinetics; Pergamon Press: Oxford, 1961; Vol. 1,
p 129.
3. Taube, H. Angew. Chem., Int. Ed. Engl. 1984, 23, 329.
4. Marcus, R. A. J. Chem. Phys. 1965, 43, 679.
5. Sutin, N. Prog. Inorg. Chem. 1983, 30, 441.
6. Marcus, R. A. J. Phys. Chem. 1968, 72, 891; in Proton Transfer, Symposia of the
Faraday Society, No. 10, 1975, p 60.
7. Hague, D. N. Fast Reactions; Wiley-Interscience: New York, 1971.
8. Bell, R. P. The Proton in Chemistry; Cornell University Press: Ithaca, NY, 1973.
9. Basolo, F.; Pearson, R. G. Mechanisms of Inorganic Reactions, A Study of Metal
Complexes in Solution, 2nd ed.; Wiley: New York, 1967.
10. Kruger, H. Chem. Soc. Rev. 1982, 11, 227.
11. Sykes, A. G. Kinetics of Inorganic Reactions; Pergamon: Oxford, 1996.
12. Parker, A. J. Chem. Rev. 1969, 69, 1.
13. Eigen, M.; De Maeyer, L. In Investigation of Rates and Mechanisms of Reaction, 3rd
ed.; Hammes, G. G., Ed.; Wiley-Interscience: New York, 1974; Vol. VI, Part II,
Chapter 3.
14. Bernasconi, C. F. Relaxation Kinetics; Academic Press: New York, 1976.
15. Norrish, R. G. W.; Porter, G. Nature 1949, 164, 658.
16. Andrews, D. L. Lasers in Chemistry, 3rd ed.; Springer: Berlin, 1997.
17. Fleming, G. R. Chemical Applications of Ultrafast Spectroscopy; Oxford University
Press: New York, 1986.
18. Simon, J. D. Ultrafast Dynamics of Chemical Systems; Kluwer Academic Publisher:
Dordrecht, 1994; Acc. Chem. Res. 1988, 21, 128.
19. Van der Zwan, G.; Hynes, J. J. J. Phys. Chem. 1985, 89, 4181.
20. Maroncelli, M. J. Mol. Liq. 1993, 57, 1.
21. Fawcett, W. R. Chem. Phys. Lett. 1992, 199, 153.
22. Waldeck, D. H. Chem. Rev. 1991, 91, 415.
23. Todd, D. C.; Fleming, G. R. J. Chem. Phys. 1993, 98, 269.
24. Grampp, G.; Hetz, G. Ber. Bunsen-Ges. Phys. Chem. 1992, 96, 198.
25. Marcus, R. A. In Special Topics in Electrochemistry; Rock, P. A., ed.; Elsevier:
Amsterdam, 1997; p 161.
26. Fawcett, W. R.; Blum, L. Chem. Phys. Lett. 1991, 187, 173.
27. German, E. D.; Kuznetsov, A. M. Electrochim. Acta 1981, 26, 1595.
28. Zusman, L. D. Chem. Phys. 1980, 49, 295.
29. Heitele, H. Angew. Chem., Int. Ed. Eng. 1993, 32, 359.
30. Kapturkiewicz, A. Adv. Electrochem. Sci. Eng. 1997, 5, 1.
31. Drago, R. S. Physical Methods in Chemistry; Saunders: Philadelphia, 1997; Chapter 8.
32. Jackman, L. M.; Cotton, F. A. Dynamic Nuclear Magnetic Resonance Spectroscopy;
Academic Press: New York, 1975; Chapter 15.
33. Sandstrom, J. Dynamic NMR Spectroscopy; Academic Press: New York, 1982.
34. Ravage, D. K.; Stengle, T. R.; Langford, C. H. Inorg. Chem. 1967, 6, 1252.
35. Burgess, J.; Pelizetti, E. Prog. React. Kinet. 1992, 17, 1.
36. Grampp, G.; Jaenicke, W. Ber. Bunsen-Ges. Phys. Chem. 1991, 95, 904.
37. Sanchez-Burgos, F.; Moya, M. L.; Galan, M. Prog. React. Kinet. 1994, 19, 1.
38. Brnsted, J. N. Z. Phys. Chem. 1922, 102, 169.
39. Bjerrum, N. Z. Phys. Chem. 1924, 108, 82.
380
40.
41.
42.
43.
44.
45.
46.
47.
48.
Fawcett, W. R.; Tikanen, A. C.; Henderson, D. J. Can. J. Chem. 1997, 75, 1649.
Kramers, H. A. Physica 1940, 7, 284.
McQuarrie, D. A. Statistical Mechanics; Harper and Row: New York, 1976.
Hynes, J. T. J. Phys. Chem. 1986, 90, 3701.
Levich, V. G. Adv. Electrochem. Eng. 1965, 4, 249.
Brnsted, J. N.; Pederson, K. Z. Phys. Chem. 1923, 108, 185.
Marcus, R. A. J. Phys. Chem. 1968, 72, 891.
Hammett, L. P. J. Am. Chem. Soc. 1937, 59, 96.
Hammett, L. P. Physical Organic Chemistry, 2nd ed.; McGraw-Hill: New York,
1970.
49. Jae, H. H. Chem. Rev. 1953, 53, 191.
50. Taft Jr., R. W. J. Phys. Chem. 1960, 64, 1805.
Problems
1. Consider the mechanism
k1
A P B
k1
k2
B ! C
where all reactions are rst order. Derive an exact expression for the concentration of B as a function of time. What is expression for cB if it is assumed
to be in a steady state? Calculate cB as a function of time using each expression assuming k1 10 s1, k1 1 s1, and k2 0.01 s1.
2. The relaxation time for the reaction
H OH (
+ H2 O
is 3.5105 s at pH 7 and 25 C. What are the rate constants for the
forward and reverse reactions?
3. The following rate constants were measured for the reaction between
dimethylaniline and methyliodide in nitrobenzene.
Temperature T=8C
Rate Constant kr =L mol1 s1
25
40
60
80
8.39
21.0
77.2
238
Determine the Arrhenius activation energy and the pre-exponential factor.

Use these results to estimate the enthalpy and entropy of activation at 25 C.
4. The following data were obtained for the rate constant of a second-order
reaction as a function of pressure:
Pressure P/atm
Rate Constant kr =L mol1 s1
1
272
544
816
1088
9:30 104
11:13 104
13:1 104
15:3 104
17:9 104
381
Estimate the volume of activation for the reaction.

5. For the rate of proton transfer between MeOH
2 and methanol, and between
MeO and methanol, Grunwald gives the result
10

rate L mol1 s1 8:8 1010 MeOH
2 1:85 10 MeO
What are the second-order rate constants for these processes?

6. The dehydration of acetaldehyde has been studied in the presence of various
weak acids in aqueous acetone at 25 C. The following table gives the name of
the weak acid, its acidity constant, KA, and the rate constant for the catalyzed
reaction, kr. Analyze these data to obtain the parameters of the Brnsted
relation.
Acid
Diethylketoxine
Hydroquinone
Resorcinol
Nitromethane
o-Chlorophenol
Benzoylacetone(enol)
p-Nitrophenol
2,4,6-Trichlorophenol
Propionic
Acetic
Phenylacetic
Diphenylacetic
Glycolic
Dichloroacetic
kr/M1min1
KA/M
0.104
0.013
0.026
0.00084
0.112
0.0088
0.52
1.53
18.0
19.2
33.0
44.2
56.6
773
2:5 1013
4:5 1011
1:55 1010
5:8 1010
3:2 109
5:9 109
6:75 108
3:9 107
1:35 105
1:76 105
4:88 105
1:15 104
1:54 104
5:0 102
7. A common type of enzyme inhibition occurs when an inhibitor combines

with a reaction intermediate to form an inactive complex. This mechanism is
E S ! ES
ES ! E S
ES I ! ESI
ESI ! ES I
ES ! E P
382
Assuming ES and ESI are at a steady state, where I is the inhibitor, derive an
expression for the rate of formation of product P.
8. Consider the following data for the pKas of anilinium ions:
pKa
Substituent
meta
para
H
Br
Cl
CN
F
I
CH3 O
CH3
NO2
4.60
3.58
3.52
2.75
3.57
3.60
4.23
4.72
2.47
4.60
3.86
3.98
1.74
4.65
3.78
5.34
5.10
1.00
Determine the Hammett parameter rH for this equilibrium process using the
Hammett substituent parameters given in table 7.10.
Liquids and Solutions at Interfaces
Alexander Naumovich Frumkin was born in

1895 in Kishinev, Moldova, and grew up and
went to school in Odessa, Ukraine. After
nishing middle school in 1912, he studied in
Western Europe at Strasbourg and Bern for
two years. In 1915, he passed the Diploma
exam at the Physics and Mathematics Faculty
of Novorossijski University in Odessa and
began his study of electrocapillary phenomena.
His doctoral dissertation entitled
Electrocapillary Phenomena and Electrode
Alexander Naumovich Frumkin Potentials was published in 1919. He
continued to teach and carry out research in Odessa until 1922 when he was
invited to join the Karpov Institute of Physical Chemistry in Moscow. In 1928
29, he spent a sabbatical year at the University of Wisconsin, and on return to
Moscow became vice-director of the Karpov Institute. In 1930, he was
appointed Professor of Chemistry at Moscow State University where he
founded the Department of Electrochemistry. He founded the Institute of
Electrochemistry at Moscow in 1958; this became during his lifetime the most
important center for electrochemical research in the world. In his early work,
Frumkin resolved the problem arising from the earlier work of Nernst
regarding the source of the potential drops within an electrochemical cell. He
demonstrated the importance of the point of zero charge of a metal electrode
in determining adsorption processes and reaction kinetics at the electrode
solution interface. His experimental work in interfacial electrochemistry helped
consolidate fundamental ideas in the eld and demonstrate their relationship to
other important areas of physical chemistry. Frumkin received many honors
for his scientic work both in the Soviet Union and internationally. He died
while attending a scientic conference at Tula, Russia, in 1976.
8.1 The Molecular Environment at the Interface Is

Different Than in the Bulk
When the properties of liquids and solutions are considered, attention is normally
focused on the bulk of the phase, and the properties of the system at its boundaries are ignored. Signicant eects are associated with the region near the surface
383
384
of a liquid phase and an understanding of these is an important part of solution

chemistry.
As a simple example, consider a beaker of pure water at room temperature in a
closed inert environment. As has been seen in the consideration of liquid structure, the properties of water are strongly inuenced by hydrogen bonding between
neighboring molecules, and to a lesser extent by dipoledipole interactions. As an
observer at the molecular level, one would nd that the molecules near the boundaries of the water phase have dierent properties. There are two boundaries in this
system, the water | air interface and the water | glass interface. At the water | air
interface, the important feature is the termination of intermolecular interactions,
so that molecules must adjust to an environment where the number of nearest
neighbors is reduced. At the water | glass interface, water molecules meet the
components of glass, a supercooled liquid with silicon dioxide as the major component. Interaction between water and silicon dioxide is dierent from interaction
among water molecules. It is clear that the molecular environment at these interfaces is very dierent than it is in the bulk. As a result, local properties are
dierent.
Now imagine that the water in the beaker is dispersed as a fog, that is to say, as
many very small droplets for which the ratio of surface area to volume is much
larger than for the water in the beaker. It is obvious that the thermodynamic
properties of the fog, a colloidal system, are very dierent from those of the water
as a macrosystem in a beaker. In order to create the fog considerable work must
be done to form a system with a much larger surface area. This means that the
Gibbs energy of a fog containing the same number of water molecules as the
beaker of water is much higher. It follows that the fog is a non-equilibrium system
in which the water droplets grow in size as a result of collisions.
An understanding of the properties of liquids and solutions at interfaces is very
important for many practical reasons. Some reactions only take place at an interface, for example, at membranes, and at the electrodes of an electrochemical cell.
The structural description of these systems at a molecular level can be used to
control reactions at interfaces. This subject entails the important eld of heterogeneous catalysis. In the discussion which follows in this chapter the terms surface
and interface are used interchangeably. There is a tendency to use the term surface
more often when one phase is in contact with a gas, for example, in the case of
solid | gas and liquid | gas systems. On the other hand, the term interface is used
more often when condensed phases are involved, for example, for liquid | liquid
and solid | liquid systems. The term interphase is used to describe the region near
the interface where the structure and composition of the two phases can be
dierent from that in the bulk. The thickness of the interphase is generally not
known without microscopic information but it certainly extends distances corresponding to a few molecular diameters into each phase.
In the present chapter, the properties of interfaces involving liquids and solutions are described from the point of view of thermodynamics, and also at a
molecular level. Examples are given for the liquid | gas, liquid | liquid and liquid
| solid interfaces. The electrical properties of interfaces are also considered and
their relevance to processes such as the extraction of an ion from an electrolyte
solution is discussed. As will be seen, a knowledge of interfacial properties and
LIQUIDS AND SOLUTIONS AT INTERFACES
385
reactions is extremely important to the understanding of many processes in living

systems and in the environment.
8.2 The Interfacial Tension of Liquids

When one considers that a molecule at an interface or surface has fewer nearest
neighbors than a molecule in the bulk of the solution, it is easy to understand that
the interfacial molecule has dierent thermodynamic properties. If one wishes to
increase the area of a liquid, one must bring molecules from the bulk to the
surface, thereby breaking intermolecular bonds and doing work. The work
done to increase the surface area by unit amount is the interfacial or surface
tension. Mathematically, the reversible work dw is given by
dw g dA
8:2:1
where g is the interfacial tension and dA the change in area. The units of interfacial tension are J m2 . This is often expressed as N m1 .
The interfacial tension is obviously an important property of a liquid because it
gives a direct indication of the magnitude of intermolecular forces. As a result of
interfacial tension a liquid which is not in contact with another condensed phase,
such as a water droplet in air, assumes the shape which has minimum area. It
turns out that this shape is a sphere. As a result, there are no elliptical or square
water droplets! By maintaining a spherical shape, the area-to-volume ratio, and
the number of molecules at the surface are their lowest possible values. One is not
surprised by this fact on the basis of experience.
Some experimental methods for measuring interfacial tension involve the contact of three phases, usually glass and air, together with the liquid being studied.
This means that one must understand the properties of three interfaces, namely,
the liquid | air, the liquid | glass, and the glass | air interface, in order to interpret
the experimental results. Since the forces of cohesion in the liquid, that is, the
intermolecular forces, are not necessarily the same as the forces of adhesion
between the glass and the liquid, the interface is usually curved. Furthermore,
the properties of curved interfaces are dierent than those of at interfaces. As a
result, the contact angle formed between the glass and the liquid must be examined carefully. Other methods, for example, the measurement of drop prole, and
the maximum bubble pressure method, depend only on the interfacial tension
between the two phases in contact.
The formation of a contact angle is illustrated in g. 8.1. The forces on a
molecule near the interface between the glass wall and the water are shown as
vectors with a longer vector fa for the adhesive force and a shorter vector fc for the
cohesive force, indicating that adhesive forces are stronger than cohesive forces.
When these two vectors are added, the resulting vector points away from the air |
water interface at an angle determined by the relative magnitudes of fa and fc . In
order for the system to be at equilibrium, this interface must be perpendicular to
the resultant vector fT . Thus, the interface is curved and makes a contact angle y
with the wall, as shown in g. 8.1.
386
Fig. 8.1 Schematic diagram of the contact angle formed

between a glass wall and a liquid where the adhesive force
on a molecule near the air | water interface and wall is fa
and the cohesive force with the bulk, fc . The vector sum of
these forces is fT .
When the contact angle is less than 90 , the liquid is said to wet the wall. In the
case of glass and water this angle is close to 0 , indicating very strong adhesive
forces at this interface. In the case of mercury and glass, the contact angle is 140 ,
demonstrating that adhesive forces are much weaker than cohesive forces. As a
result, the meniscus for a mercury column in a glass tube is convex rather than
concave.
Now, consider again a spherical liquid drop. Because of the curvature of the
interface, there is a pressure dierence between the inside and outside of the drop.
This dierence exists because of the interfacial tension, which tends to reduce the
area of the liquid system, so that equilibrium is maintained with a higher pressure
inside the drop than the atmospheric pressure outside. If the radius of the drop is
r, its surface area is 4pr2 . The incremental work dws done in increasing the radius
by dr is
dws 8pgrdr
8:2:2
This is counterbalanced by mechanical work dwm done on the basis of the volume
change which occurs under a pressure dierence, Pi Patm , where Pi is the internal pressure in the drop and Patm is the external atmospheric pressure. Thus,
dwm Pi Patm dV 4pr2 Pi Patm dr
8:2:3
At equilibrium, dws is equal to dwm so that

Pi Patm
2g
r
8:2:4
This is the YoungLaplace equation applied to a spherical surface. A more general form of this equation is used when the curvature of the interface is not
spherical [G1].
EXAMPLE
Calculate the increase in pressure inside a water droplet at 25 C given that its
radius is 0.1 mm.
The surface tension of water at 25 C is 71.8 mJ m2. From equation (8.2.4),
P
2 71:8
1:44 106 m Pa 1440 Pa
1 104
8:2:5
387
Thus, the pressure drop is 1440 Pa or 0.014 bar. This is not a negligible pressure
dierence.
Two dierent methods for measuring surface tension are now considered. The
rst is based on the fact that a liquid which wets glass rises in a glass capillary
when this is immersed in a beaker of the liquid. The height of the liquid in the
capillary can be related to the interfacial tension at the liquid | air interface. The
meniscus is concave on the liquid side, indicating that the pressure inside the
liquid in the capillary is less than that in the atmosphere. This experiment is
discussed here for the more general case of two uids a and b, as illustrated in
g. 8.2. In the specic example a corresponds to air and b to water, but the
analysis could be applied to any two uids. The meniscus is assumed to be part
of a sphere of radius R whose center is also indicated in the gure. Geometrical
analysis shows that
r
8:2:6
R
cos y
where r is the radius of the capillary tube and y, the contact angle. The dierence
in pressure across this interface is given by
2g
P
8:2:7
R
Now consider the pressure exerted through distance h by phase a outside of the
capillary, where h is height of the rise of uid b in the capillary. The pressure is
given by
Pa ra gh
8:2:8
where ra is the density of phase a and g, the acceleration due to gravity. For
equilibrium the same pressure drop must occur inside of the capillary. This pressure is
Pb rb gh
2g
R
Combining equations (8.2.6)(8.2.9), one obtains the result that

r
g
r ra gh
2 cos y b
Fig. 8.2 Schematic diagram of the rise of uid b

with respect to uid a in a narrow glass capillary.
The distance h is measured from the bottom of the
meniscus in the capillary.
8:2:9
8:2:10
388
This result is general and can be applied to any two uids, provided the
meniscus has a spherical shape. The diculty in using this technique arises
when the contact angle is not known or when the capillary is not a perfect
cylinder. This means that the experimental system is usually calibrated using a
system of known interfacial tension. Other problems arise when the meniscus is
not perfectly spherical. These have been discussed in detail by Adamson [G1].
EXAMPLE
Estimate the height of benzene in a glass capillary with a diameter of 0.01 cm at

20 C given that g is 28.96 mJ m2 and the contact angle is zero. The densities of
benzene and air at this temperature are 0.8785 and 0.0014 g cm3, respectively.
The acceleration due to gravity is 9.807 m s2. Applying equation (8.2.9),
one obtains
h
2 28:86 203
6:7 102 m
104 0:8785 0:0014 103 9:807
8:2:11
Thus the benzene rises a distance of 6.7 cm in the capillary.

Another method for measuring g is the maximum bubble pressure technique,
which has the advantage that its results are independent of the contact angle. This
technique as applied to studies of the mercury | liquid solution interface is
illustrated schematically in g. 8.3. The capillary is bent into a U-shape so that
the mercury emerges in an upward direction. As the pressure on the mercury is
increased, a bubble emerges from the end of the capillary. The sequence of shapes
assumed by the mercury bubble is such that it is always a section of a sphere, its
radius going to a minimum when it is exactly a hemisphere. At this point the
pressure is a maximum, further increase in pressure resulting in dislodgment of a
mercury drop. According to the YoungLaplace equation,

maximum bubble pressure apparatus
used to study the mercury | electrolyte
solution interface. The pressure applied
to the mercury in reservoir R is
measured as a function of the shape of
the emerging mercury bubble at the
capillary tip T. The pressure is a
maximum when this shape is
hemispherical. The experiment is
carried out for dierent voltages V
applied between the reference electrode
and the mercury reservoir.
Pmax
2g
r
389
8:2:12
where r is the radius of the capillary at its orice. Estimation of the pressure
involves consideration of the height of solution above the end of the capillary.
In practice, the apparatus is calibrated with a system of known interfacial tension.
This technique is conveniently adapted to computer-controlled experiments and
has been used to study the interfacial tension of the mercury | solution interface as
a function of electrical potential [1]. The electrical aspects of these experiments are
discussed in detail in chapter 10.
Values of the interfacial tension at the liquid | air interface for the polar
solvents considered in chapter 4 are summarized in table 8.1. There is a very
large variation in g from a high of 71.81 mJ m2 for water to a low of 21.90
mJ m2 for ethanol. The high value for water is a clear indication of the importance of hydrogen bonding in determining its physical properties. Formamide,
which can form two hydrogen bonds per solvent monomer, also has a high value
Table 8.1 Values of the Interfacial Tension and Its

Temperature Coefcient at the LiquidAir Interface
for Selected Polar Solvents at 25 C [2]
Solvent
Interfacial Tension
g=mJ m2
Temperature Coecient
@g=@T=mJ K1 m2
71.81
22.30
21.90
23.30
24.16
58.15
38.7
0.155
0.093
0.088
0.069
0.079
0.793
0.59
22.68
27.56
38.46
26.79
33.15
36.42
42.98
33.8*
40.7
42.43
36.48
25.82
41.39
35.5**
0.123
0.122
0.106
0.100
0.147
0.122
0.121
0.118
0.168
0.110
0.110
Protic
1.
2.
3.
4.
5.
6.
7.
W
MeOH
EtOH
PrOH
BuOH
F
NMF
Aprotic
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
AC
AcN
BzN
BuN
DMA
DMF
DMSO
HMPA
NMP
NB
NM
PrN
PC
TMS
*Measured at 30 C.
**Measured at 20 C.
390
of g. In the case of the alcohols, which can form only one hydrogen bond per
monomer, the values of g are much lower. Considerable variation in g is also seen
among the aprotic solvents. The highest value is found for dimethylsulfoxide,
which is highly dimerized in the bulk. On the other hand, a rather low value is
found for acetone. A proper ranking of these data would require knowledge of the
area occupied by a single molecule at the interface on the basis of its molecular
size and orientation. Then, the interfacial tension data could be re-evaluated as
work done per molecule to take it from the bulk to the interface.
If g is measured at constant temperature and pressure in a one component
system, then it is equal to the Gibbs energy of the interface, Gs . A more precise
denition of Gs is given in the following section, but the idea that work done can
be equated to Gibbs energy is familiar from thermodynamics. It follows that the
temperature derivative of g can be related to the entropy of the interface as
follows:
@Gs
@g
8:2:13

Ss
@T
@T
Thus, the values of @g/@T recorded in table 8.1 give the entropy change when the
area of the interface is increased by 1 m2. In all cases, Ss is positive, indicating that
order is lost when molecules are moved from the bulk of the solution to the
interface. This is a result of the fact that intermolecular bonds are broken in
the change of environment. The largest entropy change is found for formamide
due to its strongly hydrogen-bonded structure. Much smaller entropies are found
for water and the alcohols. In the case of the aprotic solvents, the entropies fall in
the range from 0.1 to 0.2 mJ K1 m2. In order to assess these parameters in more
detail, the entropy per molecule must be estimated on the basis of the area
occupied by each molecule.
In assessing the meaning of the parameters g and @g=@T, one eventually asks
the question What is the interface and how thick is it? The interface certainly
has the thickness of the layer of molecules at the termination of the liquid phase
but the disruption of normal liquid structure may extend somewhat further into
the bulk. Since the thickness of the interface is not known, it is dicult to give a
molecular interpretation of the thermodynamic properties of this region.
However, eective thermodynamic conventions have been developed for discussing interfacial properties. These are outlined in detail in the following section.
8.3 The Thermodynamics of Fluid Interfaces

Fluid interfaces are those systems in which the two phases forming the interface
are mobile. Thus, they include liquid|liquid and liquid|gas interfaces. The example
chosen here to illustrate the application of thermodynamics to these interfaces is a
closed system containing both gas and liquid phases consisting of two components, propanol and water. The liquid phase is designated b and the gas phase a.
Now imagine a microprobe moving up through b to the interface (see g. 8.4).
The concentration of the alcohol is constant in this phase and equal to cbA . In the
gas phase, its concentration is much lower. Near the interface the alcohol concentration rises before it falls to the lower gas phase value caA . This observation is
391
Fig. 8.4 (a) A two-component system

with liquid (b) and gas (a) phases; (b)
concentration of the alcohol component
as a function of vertical position near
the interface.
attributed to the hydrophobic nature of the alkyl group in propanol which likes to
be out of the aqueous environment as much as possible. Under normal circumstances where the surface area/volume ratio is small, any depletion of propanol
from the bulk liquid is negligible. Thus, a change in bulk concentration due to the
preference of propanol for the interface would not be observed experimentally.
However, if the area/volume ratio is increased, for example, by converting phase b
to an aerosol, the alcohol could be suciently depleted from the bulk that its
concentration would change.
Two conventions exist in the treatment of interfacial thermodynamics. The rst,
due to Gibbs [G1], treats the interface as a region with no volume. Since the
interphase is actually a region of nite width extending over several molecular
diameters, this convention is equivalent to associating all interfacial inhomogeneities with a plane placed arbitrarily in the interphase. The choice of position of the
dividing surface is often made so that the surface excess of one component is
arbitrarily dened to be zero. Suppose that the dividing surface is placed at position xs as shown in g. 8.4, and that positions xb and xa are suciently far within
the bulk phases that the interface does not inuence local properties. In the Gibbs
model, the properties of b are assumed to be uniform right up to the plane at xs .
Similarly, the properties of a are assumed to be uniform up to this plane. This
means that inhomogeneities associated with the interphase are now attributed to a
plane. As a result, one may dene the surface excess for propanol, A as follows:
xa
xs
b
x
A cA cA dx cxA caA dx
8:3:1
xb
xs
where cxA is the concentration of propanol at position x. This denition is arbitrary because it depends on where one places the dividing plane xs . The surface
excess has units of mol m2 when the concentration is expressed in mol m3. A
similar expression can be written for the surface excess of water, namely,
xa
xs
b
x
W cW cW dx cxW caW dx
8:3:2
xb
xs
cbW ,
where cxW is the local water concentration,

the average concentration in phase
b, and caW , the corresponding value in phase a. If the dividing plane xs is moved,
392
both A and W change but a simple expression relates the two values. As a result,
one may only determine relative surface excesses and not their absolute values, as
shown in more detail below.
The second convention, due to Guggenheim [G1], denes a surface phase s
whose width is sucient to include the inhomogeneities associated with the surface. Placing the boundaries at xa and xb , the surface excesses may be dened as
follows:
xa
ns
8:3:3
W cxW dx W
A
xb
and
xa
A
cxA dx
nsA
A
8:3:4
xb
nsW
nsA
and
are the number of moles of water and alcohol, respectively, in the
surface phase, and A is the area of the interface. The volume of the surface phase
Axa xb is nite but arbitrary, its value depending on the width of the inhomogeneous region. Just as with the Gibbs convention, one can only measure
relative surface excesses, not absolute values. This aspect of the surface excess
is claried further below.
The fundamental thermodynamic expression for the innitesimal increase in
Gibbs energy for a multicomponent system which can exchange matter with its
surroundings was derived earlier on the basis of the rst and second laws of
thermodynamics (equation (1.3.21)):
X
dG SdT VdP
mi dni
8:3:5
i
mi is the chemical potential of species i and the other symbols have their usual
meanings. This expression was derived assuming that the only work done on the
system is mechanical. If work is also done to change its area A, then the expression becomes
X
mi dni g dA
8:3:6
dG SdT VdP
i
It follows that the interfacial tension g is given by

@G
g
@A T;P;ni
8:3:7
Two points should be noted about equation (8.3.6). First, the number of
components considered in the summation is that which can be added to the
system in an independent way. Thus, if NaCl is a component, Na is not considered separately from Cl, since the concentrations of the two ions are exactly
equal. However, local concentrations near the interface may not be equal. If one
considers Na and Cl as independent components the total number of variables
in equation (8.3.6) increases. In the case of charged components, the chemical
potential mi must be replaced by the electrochemical potential m~ i . It also follows
393
that the number of degrees of freedom in the system is exactly the same because of
the condition of electroneutrality. Second, one is normally interested in the Gibbs
energy change associated with the surface phenomena. These are dicult to detect
when the surface to volume ratio is small. Following the Gibbs convention, one
denes a surface phase s of zero volume, and contributions to dG due to perfectly
homogeneous a and b phases and a s phase with which all inhomogeneity is
associated. Thus,
dG dGa dGb dGs
8:3:8
Now, the right-hand side of equation (8.3.5) is divided up to obtain the contributions from phases a, b, and s under the condition that temperature and pressure
are constant throughout the system. In addition, at equilibrium the chemical
potential of each component is the same in each phase. The total entropy of
the system is
S Sa Sb Ss
8:3:9
Remembering that the surface phase has no volume in the Gibbs convention,
V Va Vb
8:3:10
Finally, any incremental change in the number of moles of component i can be

divided into changes in each phase so that
dni dnai dnbi dnsi
8:3:11
The following equations can now be written:

dGa Sa dT Va dP
mi dnai
8:3:12
dGb Sb dT Vb dP
dGs Ss dT
mi dnbi
8:3:13
mi dnsi g dA
8:3:14
The nal equation, which describes the thermodynamic properties of the surface
phase, is the one which is important for the present analysis.
Now, an integration is performed on equation (8.3.14) at constant temperature
to nd the Gibbs energy of the surface phase. The result is
X
Gs
mi nsi gA
8:3:15
i
Taking the total dierential and comparing with equation (8.3.14), one obtains
X
nsi dmi A dg 0
8:3:16
i
394
This result is known as the Gibbs adsorption isotherm and is the starting point for
analyzing the properties of the surface phase at constant temperature and pressure. It is more commonly written as
X
dg
i dmi
8:3:17
i
where i is the surface excess equal to the number of moles of component i in the
surface phase divided by the area of the surface.
In the case of the propanolwater system discussed above, the Gibbs adsorption isotherm is
dg A dmA W dmW
8:3:18
Since it is impossible to change the chemical potential of A without also changing

that of W, one cannot measure either A or W independently. The practical
signicance of this can be seen using the GibbsDuhem relationship, according
to which
xA dmA xW dmW 0
8:3:19
where the xi s are mole fractions in the bulk of the solution. Accordingly, one may
write

x
8:3:20
dg A W A dmA W
A dmA
xW
W
A is the surface excess of alcohol relative to that of water. The relative value is
approximately equal to the absolute one when W xA xW .
When the solution is suciently dilute such that Henrys law holds for the
solute, then
dmA RT d ln cA
8:3:21
where cA is the concentration of the solute, namely, propanol in the present

example. Its surface excess is then given by

1
@g
A
8:3:22
RT @ ln cA
where it has been assumed that the relative nature of A may be neglected in the
dilute solution.
EXAMPLE
In the concentration range 0.010.03 M, the interfacial tension of solutions of

3-phenylpropanoic acid (PBA) at 21 C is given by the equation
g 67:87 567:3cA
8:3:23
where g is given in mJ m2 and cA is the concentration of PBA. Calculate its

surface excess at cA 0:02 M.
From the relationship given
@g
@g
cA
0:02 567:3 11:35 mJ m2
@ ln cA
@cA
8:3:24
395
The surface excess is

A
1 @g
11:35
4:64 103 mmol m2
RT @ ln cA 8:3145 294:2
8:3:25
It is clear that determination of the change of interfacial tension with solute

concentration in dilute solutions provides a way to nd the surface excess of the
solute. This method has been successfully applied to many two-component systems involving molecules. Application to electrolyte solutions is more complicated
but is discussed further below. Before doing so, more details about the electrical
aspects of interfaces are outlined.
8.4 The Electrical Aspects of Interfaces

When the thermodynamic properties of charged components in condensed phases
are considered it is convenient to evaluate the chemical contributions to the Gibbs
energy separately from the electrostatic contributions. This leads to the denition
of the electrochemical potential of charged species i and the inner potential fa of
the phase a in which species i is located (see section 6.6). It was emphasized in the
earlier discussion that the separation of the electrochemical potential into chemical and electrostatic contributions is arbitrary, but conceptually very useful. The
inner potential arises as a result of the hypothetical work done to take a charged
species from innity across the double layer at the surface of phase a under
circumstances where this species experiences no chemical interactions in the
bulk. The double layer is described as that region at the surface of phase a in
which there may be a potential drop due to free charge on the surface, unequal
adsorption of ions of opposite charge, or net orientation of dipoles [G4]. The
inner potential is a quantity which cannot be measured experimentally, but only
estimated on the basis of a suitable model.
The inner potential is further subdivided into the outer potential ca and the
surface potential wa , that is,
fa ca wa
8:4:1
The outer potential is due to the free or excess charge on the surface of phase a
and can be measured experimentally. The surface potential is due to the dipolar
distribution of charge at the interface due to the unequal adsorption of ions and
orientation of molecular dipoles. It cannot be measured experimentally. Since
these quantities are dened with respect to the process of bringing a charged
species from innity into the phase, the surface potential is positive when the
positive end of the dipolar charge points toward the center of the solution and
the negative end toward charge-free innity.
The method of measuring the outer potential of phase a can be described on
the basis of a conceptual experiment in which a test charge is brought up to the
phase, which bears a net charge q. Assuming that the phase is an isolated sphere
of radius r, the potential experienced by the test charge is
396
q
q
4pe0 d 4pe0 r x
8:4:2
where d is its distance from the center of the sphere, and x, its distance from the
surface of the sphere. When the distance is very small with respect to the radius r,
c is constant and equal to q=4pe0 r. For example, if r is 1 cm, c is constant to
more than 1% for values of x less than 105 m. For very small values of x, other
factors need to be taken into consideration, for example, redistribution of the
charge on the surface of the sphere due to the so-called image eect. As a result
the ideal distance from the surface at which one should measure c is 1 mm
[G2, G4].
When the outer potential of phase a is zero, it is convenient to dene another
quantity, namely the real potential. Recalling that the electrochemical potential
for species i in phase a is given by
m~ ai mai zi Ffa mai zi Fca wa
8:4:3
the real potential of species i in phase a is dened to be the electrochemical

potential when the phase is not charged:
aai m~ ai ca 0 mai zi Fwa
8:4:4
This quantity can also be measured. For example, the electronic work function for
metals discussed in section 8.5 is the negative value of the real potential for an
electron in the metal.
Since the discussion here concerns interfaces, it is also important to consider
the potential dierences which arise between the boundaries of two phases a and
b. The inner potential dierence, which is known as the Galvani potential dierence, can be dened as
a b f
fb fa
8:4:5
The direction in which the potential dierence is measured must be dened so that
the i operator gives the initial phase on the left-hand side and the nal phase on
the right-hand side. Of course, one could choose to dene the potential dierence
in the opposite direction and write
b a f
fa fb
8:4:6
In general, the Galvani potential dierence cannot be measured because the

individual inner potentials cannot be measured. However, there is one important
exception to this rule which is of great practical importance. When the phases a
and b have the same chemical composition, a b f can be found from the dierence in electrochemical potentials for a common species i in the phases. Thus,
m~ bi m~ ai mbi zi Ffb mai zi Ffa zi Fa b f
8:4:7
This principle is used every time one measures the electromotive force of an
electrochemical cell. In this case the potential measuring device determines the
dierence between the electrochemical potentials of electrons in two pieces of the
same metal, for example, in two copper wires. The classical device for doing this
measurement is a Poggendorf compensation potentiometer but a modern in-
397
strument is a very high impedance electronic voltmeter based on operational

ampliers.
The dierence in outer potentials between two phases a and b is known as the
Volta potential dierence and is dened as
a b c
cb ca
8:4:8
Since each outer potential may be determined as described above, the Volta
potential dierence may also be found experimentally. A method for measuring
aib c is discussed later in this chapter.
It is important to keep in mind that, when two condensed phases a and b
come into contact, the potentials associated with the isolated phases can change.
When a and b contain a common charged species i, that is, an electron or an
ion, contact of these phases results in the ow of charge in order to establish
equilibrium with respect to species i. This is accompanied by a movement of free
charge on the surfaces of a and b which are exposed to air or vacuum. The rst
process leads to a change in the Galvani potential dierence a b f on contact of
the two condensed phases. The second process, which involves a change in the
free charge at the interfaces of a and b with air or vacuum, results in a change
in the Volta potential dierence a b c. Concrete examples of these eects
are discussed in more detail later with respect to the metal | solution and the
liquid | liquid interfaces.
The relationship between the Galvani potential dierence, the Volta potential
dierence, and the surface potentials of two condensed phases in contact is illustrated in g. 8.5. Both of the condensed phases a and b are in contact with an
inert gas or vacuum. The outer potential ca is the potential dierence between
charge-free innity and a point just outside of but close to phase a; cb is the
corresponding quantity for phase b. The Volta potential dierence is simply the
dierence between these quantities. wa and wb give the surface potentials of phases
a and b, respectively. They are measured from vacuum into the condensed phase.
The Galvani potential dierence a b f is dened within the condensed phases.
Since the sum of the potential drops around a loop must be zero, one obtains
a b f
wb cb ca wa 0
Fig. 8.5 Illustration of the denitions of the outer

potentials, ca and cb, the surface potentials, wa
and wb, and the Galvani potential dierence
a b for two condensed phases a and b in
contact.
8:4:9
398
or
a b f
a b c wb wa
8:4:10
This is the fundamental relationship between the quantities which describe the
electrical properties of the interface between two condensed phases and the gas
phase.
The electrical properties of systems containing charged species are very important in achieving an understanding of how they behave at interfaces. In this
chapter, attention is focused on two such systems, namely, metals and electrolyte
solutions, that is, the components of electrochemical cells. In the following
section, the properties of electrons in metals and the work required to extract
an electron are examined in more detail.
8.5 The Work Function for Electrons in Metals

A well-known property of metals is their ability to carry an electrical current by
means of the movement of the electronic cloud associated with the metal atoms.
Most metals are solids at room temperature and only a few, such as mercury, are
liquid. Because of their importance in the design of electrochemical cells, their
electronic properties are considered here in some detail.
The work function is dened as the minimum work required to withdraw an
electron from a metal under the conditions that the metallic phase bears no net
charge. On the basis of the earlier discussion, one may write
m
m
m
Wm
el ael mel Fw
8:5:1
where Wm
el is the work function for an electron in metal m. Thus the electronic
work function is the negative value of the real potential of the electron am
el . The
chemical potential of the electron depends on the bulk structure of the metal,
whereas the surface potential wm depends on the arrangement of the metal atoms
at the metal | vacuum interface. In the case of single crystals, the variation in Wm
el
observed when dierent crystal faces are exposed at this interface is due to a
corresponding variation in wm. Because the electronic cloud within the metal
tends to spill out over the edge of the crystal lattice, wm is a positive quantity.
Thus, a test charge moving from vacuum into the metal experiences rst the
negative end of a surface dipole, due to the overspilt electron cloud, and then
the positive end, due to the metal atoms in the lattice at the interface.
The energy of electrons in metals is made up of two contributions, namely,
potential energy and kinetic energy. The potential energy contains contributions
from the bulk and from the surface. The bulk contribution arises from the coulombic interaction of the electronic cloud with the positive charge of the metal ion
cores. The surface contribution arises from the dipole at the surface due to electron overspill, as described above. The kinetic energy of the electrons can be
estimated by the Sommerfeld model [G3] using a quantum-mechanical description
of the translational motion of electrons in the lattice system. On the basis of the
Pauli exclusion principle, two electrons of opposite spin occupy each energy level
399
corresponding to increasing kinetic energy. The highest lled level at 0 K is called

the Fermi level, and its energy is
eF
h2
3p2 rel 2=3
8p2 mel
8:5:2
where rel is the electron density, mel , the electron mass, and h, the Planck constant. The kinetic energy is a positive destabilizing contribution to the overall
energy of the system, whereas the potential energy U is a stabilizing negative
energy. On the basis of this description, the work function is given by
Wm
el U eF UB Us eF
8:5:3
where U B is the bulk contribution to the potential energy and U s, the surface
contribution. These energies are shown schematically in g. 8.6.
When the electrochemical approach to describing the work function (equation
(8.5.1)) is compared to the physical approach (equation (8.5.3)), the following
relationships are obtained. The chemical potential of the electron is related to
bulk properties so that
mm
el UB eF
8:5:4
In the same way, the surface potential energy is related to the surface potential
so that
U s Fwm
8:5:5
When the metal phase is charged by connecting it to a voltage source, the

potential energy U is changed but the kinetic energy eF is not changed. When
the charge is positive, U shifts down, indicating that more work must be done to
extract an electron. On the other hand, a negative charge results in an upward
shift in U, corresponding to easier removal of an electron.
In order to determine the electronic work function the metal must be ultrapure
with no contamination on the surface. The biggest problem in this regard is the
fact that most metals have a surface oxide layer under ambient conditions. A
clean metallic surface can be formed by sputtering the metal in a vacuum chamber, and techniques are available for growing thin metal lms of known crystallographic orientation. Otherwise, the metal surface may be cleaned by positive ion
bombardment techniques at very low pressures. Values of the electronic work
function for polycrystalline metal surfaces are summarized in table 8.2. The lowest
Fig. 8.6 Energy prole at the

metal/vacuum boundary
showing the separate
contributions of the potential
and kinetic energy of
electrons to the work
function.
400
Table 8.2 Values of the Electronic Work

Function for Polycrystalline Metal Surfaces*
Metal
Ag
Al
Au
Ba
Be
Bi
Ca
Cd
Ce
Co
Cr
Cs
Cu
Fe
Ga
Hf
Hg
K
Wm
el =eV
4.30
4.25
4.30
2.49
3.92
4.40
2.80
4.10
3.20
4.41
4.58
1.81
4.40
4.31
3.96
3.53
4.52
2.22
Metal
In
Ir
La
Li
Mg
Mn
Mo
Na
Nb
Nd
Ni
Os
Pb
Pd
Pr
Pt
Rb
Re
Wm
el =eV
3.80
4.70
3.30
2.38
3.64
3.83
4.30
2.35
3.99
3.07
4.50
4.70
4.00
4.80
2.54
5.32
2.16
5.00
Metal
Rh
Ru
Sb
Sc
Sm
Sn
Sr
Ta
Te
Th
Ti
Tl
U
V
W
Zn
Zr
Wm
el =eV
4.75
4.60
4.08
3.30
2.70
4.38
2.35
4.12
4.73
3.41
3.95
4.00
3.74
4.12
4.54
4.24
3.90
*As given by Trassatti [G3] and Fomenko [3].
work function is found for cesium, the heaviest commonly available alkali metal.
On the other hand, high values of Wm
el are observed for the precious metals gold,
palladium, and platinum. As stated earlier, the exact value depends on the potential and kinetic energy of the electron in the metal (equation (8.5.3)).
Quantum-mechanical models have been developed for the electrons in sp
metals, the best known of these being the jellium model [4]. Sp metals are systems
such as the alkali and alkaline earth metals in which d electrons are not involved
as valence electrons. For these systems, reasonable estimates of mm
el can be made;
these are summarized in table 8.3 for the alkali metals. It is interesting that this
Table 8.3 Experimental Value of the Work

Function (Welm ) and Theoretical Estimates of
the Chemical Potential of an Electron (mm
el ), the
Surface Potential (wm), the Fermi Kinetic
Energy (eF), and the Potential Energy of
Electrons (U) in the Alkali Metals [G3]
Metal
Li
Na
K
Rb
Cs
Wm
el =eV
mm
el =eV
wm =V
eF =eV
U=eV
2.38
2.35
2.22
2.16
1.81
1:65
1:95
2:00
1:94
1:80
0.73
0.40
0.22
0.22
0.01
4.72
3.23
2.12
1.85
1.58
7:10
5:58
4:34
4:01
3:39
401
quantity does not change in a monotonic fashion but reaches its minimum value
at potassium. By subtracting mm
el from the value of the work function, one
obtains an estimate of the surface potential wm. This quantity is always positive,
reecting the fact that electrons spill out of the metal at the metal | gas interface.
The surface potential is largest for Li and smallest for Cs. The values of the Fermi
energy eF and the depth of the potential energy, U also increase monotonically
with increase in the number of electrons in the alkali metal. These results reect
the fact that loss of one electron from Cs is much less important than the loss of
one electron from Li.
The fact that electrons tend to spill out from the metal reects the mobility of
the electron gas within these systems. The distance involved in the formation of
the surface dipole is approximately 100 pm. The exact value of wm depends on the
crystallographic structure of the metal at the surface and reects the density of
metal atoms associated with this structure. For metals with a low melting point,
the data for the polycrystalline surface recorded in table 8.2 give the best indication of surface properties. Since the metal atoms are relatively mobile at room
temperature, the system tends to maintain the properties described in the jellium
model at the surface.
For metals with high melting points, it is possible to cut single crystals so that
surfaces with dierent packing densities of metal atoms are exposed. In the case of
face-centered cubic systems such as metallic gold, the packing density increases
for the low Miller indices in the order (110) < (100) < (111). The electronic work
function increases in the same direction, being lowest for Au(110) and highest for
Au(111). Since the bulk structure of gold is independent of its surface structure,
the value of mm
el is constant. Thus, the change in work function reects a change in
surface potential wm so that wm has its largest value for Au(111), which also has
the highest density of gold atoms at the surface.
Polycrystalline surfaces result from a mixing of all possible crystal orientations.
From an energetic point of view, the low index faces discussed above predominate. The work function of the polycrystalline surface reects a weighted average
of the work functions for each crystallographic orientation. In the case of facecentered cubic systems, it falls between that of the (110) and (100) single crystal
surfaces. Since polycrystalline metals are involved in most practical applications,
the eect of surface structure on the work function is not discussed further here.
More information on this topic is available in reviews by Trasatti [G3, 5].
The concept of the electronic work function is very familiar to physicists and
chemists and is discussed in elementary textbooks. A much less familiar concept is
that one may measure experimentally the work required to remove an ion from an
electrolyte solution. This is also discussed in this chapter but before doing so, the
properties of liquids and solutions at gas interfaces are considered.
8.6 The Liquid | Gas Interface and the

Adsorption Isotherm
Much of the early work in interfacial chemistry involved studies of the liquid | gas
interface, especially for aqueous solutions. The techniques described earlier,
402
namely the capillary rise experiment and the maximum bubble pressure experiment are easily applied to measure the interfacial tension of liquid solutions.
Aqueous systems have been studied extensively for both electrolytes and organic
compounds as solutes.
Before discussing the interfacial properties of aqueous solutions it is useful to
consider the structure of pure water at its interface with air. Although the surface
potential of water cannot be measured experimentally, indirect estimates of wW on
the basis of a variety of experiments lead to the conclusion that it has a small
positive value, of the order of 0.1 V. Of course, all of these estimates involve at
least one extrathermodynamic assumption. The positive value of wW suggests that
there is a net small orientation of water dipoles with their oxygen atoms pointing
to the gas phase. As a result, when a positive test charge approaches the surface of
pure water it experiences a small positive dipole with its positive end pointing into
the solution. This result probably reects the strong eect of hydrogen bonding,
which prevails right up to the interface, so that a few water molecules in the last
monolayer are attached to other molecules below by two hydrogen bonds,
whereas most are attached by only one. The orientation of these few would
be such that the molecular dipole vector is approximately perpendicular to the
interface.
When an electrolyte is added to water, a small increase in surface tension is
observed [6]. This is illustrated for several 11 electrolytes in g. 8.7. The Gibbs
adsorption isotherm for these systems can be written as
dg e dme W dmW
8:6:1
Fig. 8.7 Plots of the surface tension of electrolyte solutions g relative to that of pure water
g0 against the molar concentration of the electrolyte [6]. The ordinate scale is correct for the
NaCl system. It has been shifted vertically for the KCl, CsI, and KCNS systems by 0.01,
0.02, and 0.03 units, respectively.
403
where e is the surface excess of the electrolyte, W , that of water, and me and mW,
the corresponding chemical potentials. As described earlier the relative surface
excess of the electrolyte is given by

@g
x
W
e
e e W
8:6:2
@me T;P
xW
where xe and xW are the mole fractions of the electrolyte and water, respectively.
Since the slope of a plot of g against RT ln ae is positive, the surface excess W
is
e
negative, indicating that ions are excluded from the region of the interface. It
should also be noted that over most of the concentration range considered xe is
much less than xW , so that the relative surface excess is approximately equal to its
absolute value e . Since it is not possible to vary the activities of the cation and
anion separately, the individual surface excesses of the two ions cannot be determined. It follows that these experiments give no information about the distribution of charge at the interface.
In order to understand the above results, one must remember that the electrolyte is eectively not present in the gas phase. This is due to the fact that the
relative permittivity of the gas is much smaller than that of water. On the basis of
simple electrostatic concepts using the method of images, an ion of charge zi e0
located at a distance x from the interface is repelled by an image charge located in
the gas phase at a distance 2x with a charge equal to
q zi e0
eW ea
eW ea
8:6:3
where ew is the relative permittivity of the aqueous phase and ea, that of the gas
phase. Since ea is very much less than ew, the image charge is approximately equal
to the ions charge. The force between the ion and its image is
Fim
z2i e20
eW ea
2e e
16pe0 eW x W
a
8:6:4
This is a repulsive force which increases as 1=x2 as the ion gets closer to the
interface. A model based on these concepts was developed by Onsager and
Samaras [6] to estimate values of e for electrolyte solutions. This model is
reasonably successful for dilute electrolyte solutions.
Most studies of the solution | gas interface have involved organic compounds,
for example, organic alcohols and acids. These dissolve in water to some extent as
a result of the polar group. However, the alkyl group in the molecule is hydrophobic, so that the solute molecules are always found in excess at the solution | air
interface. The accumulation of the organic molecule means that its surface excess
is positive so that the interfacial tension of the solution decreases as the concentration of the organic molecule increases. An example of data for the straight
chain alcohols in the series n-butanol to n-octanol is shown in g. 8.8. The results
are presented in terms of the surface pressure , which is dened as the interfacial
tension, with respect to that for the pure solvent, that is,
g0 g
8:6:5
404
The experimental results [7] show that the surface pressure is positive and
increases with alcohol concentration. This indicates that the alcohol accumulates
at the water | air interface. Furthermore, the surface pressure increases with
increase in the alkyl chain length at constant bulk concentration, demonstrating
that the extent to which the alcohol accumulates at the interface depends on the
alkyl chain length, or molecular hydrophobicity.
The Gibbs adsorption isotherm for the alcoholwater system can be written as
dg d A dmA W dmW
The relative surface excess of the alcohol is given by

@g
x

A A W
W
A
xW
@mA T;P
8:6:6
8:6:7
Visual examination of the data in g. 8.8 shows that the slope of the plot of
against the logarithm of the alcohol concentration approaches a constant at the
highest concentrations considered. This demonstrates that the relative interfacial
excess approaches a maximum value, AM W . The maximum interfacial excess
corresponds to a situation in which the alcohol molecules are arranged at the
interface in such a way that further increase in the interfacial excess is not possible. Analysis of the concentration dependence of A W allows one to determine
the form of the adsorption isotherm for the alcohol.
The process of adsorption of the alcohol at the aqueous interface can be
described by the equation
Ab rWad Aad rWb
8:6:8
Fig. 8.8 Plots of the surface pressure for the water | air interface against alcohol concentration for aqueous solutions of butanol (C4), pentanol (C5), hexanol (C6), heptanol
(C7), and octanol (C8) [7].
405
where the subscript ad designates molecules at the interface, and the subscript
b, molecules in the bulk of the solution; r is the number of water molecules
displaced in the adsorption of one alcohol molecule. In order to derive the adsorption isotherm, one writes down the thermodynamic conditions for equilibrium in
terms of the chemical and electrochemical potentials of the individual species. For
the molecules in the solution, the chemical potentials are
mA mA RT ln aA
8:6:9
mW mW RT ln aW
8:6:10
and
where mA and mW are the standard chemical potentials, and aA and aW , the
activities of the alcohol and water, respectively. Since the electrostatic environment at the interface is very dierent at the interface than in the solution, the
electrochemical potentials of the adsorbed molecules must be considered. This
requires knowing the average dipole moment of the interfacial molecule in a
direction perpendicular to the interface, and also the local electrical eld in the
same direction. For the alcohol molecules, the electrochemical potential at the
interface can be written
ad;
ad;
m~ A
mA
RT ln A hPA iEe
8:6:11
is the standard chemical potential at the interface, A , the absolute

where
surface excess of alcohol, hPA i, the average dipole moment perpendicular to the
interface, and Ee, the local eld experienced by the alcohol dipoles in the same
direction. The corresponding expression for interfacial water molecules is
ad;o
mA
ad;
m~ ad
A mW RT ln W hpW iEe
8:6:12
On the basis of equilibrium (8.6.8), one obtains the expression

RT ln A rRT ln W Gad RT ln aA rRT ln aW hpA iEA rhpW iEe
8:6:13
where

ad;
Gad mad;
A rmw mA rmW
8:6:14
If it is assumed that alcohol can completely cover the interface replacing all of the
water molecules originally present, then one may dene a maximum surface excess
for alcohol, Am , which is given by
rAm rA W
8:6:15
As a result the condition for equilibrium becomes

ln A r lnAm A ln Bad aA hpA i rhpW i
Ee
RT
8:6:16
where
ln Bad
Gad
r ln r r ln aw
RT
8:6:17
406
RT ln Bad is the eective standard Gibbs adsorption energy and includes the
activity of water, which is approximately constant, provided the alcohol solutions
are dilute.
A thorough analysis of the electrostatic contribution in equation (8.6.16) is a
dicult problem in statistical thermodynamics. An approximate solution to this
problem is to assume that this contribution is proportional to the surface excess of
the alcohol. In the present case the assumption is reasonable because the hydrophobicity of the alkyl groups causes the alcohol molecules to adopt a similar
orientation with respect to one another at the interface. Then equation (8.6.16)
can be rearranged as
ln A r lnAm A ln Bad a Aad
A
Am
8:6:18
where Aad is the interaction coecient. This equation is known as the Flory
Huggins isotherm and has often been used to analyze the adsorption of molecules
at interfaces when the adsorbate diers in size from the solvent [8]. When Aad is
positive, interaction between adjacent adsorbed molecules is repulsive and larger
bulk concentrations are required to reach maximum coverage. The interaction
term can also be attractive (Aad < 0). Under these conditions, phase transformation is possible at the interface. When interactions between the adsorbed species
are absent (Aad 0), the isotherm becomes
ln A r lnAm A ln Bad aA
8:6:19
In the early literature the adsorption isotherm was usually written with a simpler
coverage term, namely, with r equal to unity. This practice reected work at the
solid | gas interface, where the adsorption process does not involve replacement of
one molecule by another. When size eects are ignored, equation (8.6.18) has the
form

A
Aad A
Bad aA
exp
8:6:20
Am A
Am
This equation is known as the Frumkin isotherm. Isotherms with more complex
versions of the coverage term have also been derived [8].
Plots of the fraction A =Am against the logarithm of the activity of the
adsorbate in the bulk of the solution are shown in g. 8.9. The data were calculated assuming that r is one (equation (8.6.20)), and Bad is 1000, with various
values of Aad =Am . The curves all have the sigmoidal shape expected on the basis
of the coverage factor Am A . When interactions between the adsorbed molecules are absent (Aad 0), the major part of the change in A takes place for a
change in bulk adsorbate activity by a factor of 400. When repulsive interactions
are present a larger change in bulk adsorbate concentration is required to achieve
the same coverage at the interface.
Application of the general adsorption isotherm to experimental data requires
the determination of four parameters, namely, r, Am ; Aad , and Bad . The value of
Am is often apparent from the experimental data. If the size factor r is also
known, equation (8.6.18) is rearranged so that Aad and Bad can be determined
from a linear plot.
407
Fig. 8.9 Plot of the fraction A =Am against the logarithm of adsorbate activity aA according to the Frumkin isotherm (equation (8.6.20)) with the interaction parameter Aad =Am
equal to 0, 2, and 5.
ln

A

r lnAm A ln Bad Aad A
aA
Am
8:6:21
A plot of the function on the left hand side of equation (8.6.21) against A is
linear with a slope equal to Aad =Am and intercept on the ordinate axis equal to
ln Bad.
Most often the parameter r is not known a priori on the basis of molecular size.
Information about molecular orientation is also required for the adsorbate. In the
case of aqueous solutions, water molecules may act in clusters because of hydrogen bonding, so that the eective size of water at interfaces may be larger than
that estimated on the basis of one water molecule.
Plots of data generated on the basis of equation (8.6.18) but plotted according
to equation (8.6.21), assuming that r 1, are shown in g. 8.10. When the value
of r is indeed unity, the plot is linear with a slope equal to Aad and an intercept
on the ordinate axis equal to ln Bad . When r is not unity, the plots are curved. The
best way to determine r involves calculating the function on the left-hand side of
equation (8.6.21) for a reasonable range of r values, and in this way searching for
the value of r which gives the best straight line for a plot of the left-hand side
against A =Am .
Many dierent isotherms have been derived for molecular adsorption at interfaces, including the liquid | gas, liquid | liquid, and liquid | solid interfaces. Some
of these isotherms can be shown to be approximate versions of the general
isotherm derived here.
Although the changes in interfacial composition can be followed by measuring
the change in interfacial tension, these experiments provide no direct information
about the electrical changes at the interface. These are especially important for
408
Fig. 8.10 Plot of the function lnA =aA lnAm A against A =Am using data generated with equation (8.6.18) assuming Bad 1000; Aad 2, and Am 5 molec. nm2.
The value of r is 1, 1.5, or 2.0 for each curve as indicated.
electrolyte solutions in which the microscopic description at the interface can be

dierent for the cation and anion. More details about these systems are given in
the next section.
8.7 Experimental Measurement of the Volta Potential

Difference at Interfaces
A Volta potential dierence usually arises between two condensed phases when
they come into contact. Because these phases have dierent chemical properties,
charge may be redistributed at their interface with air, even though this interface
carried no net charge before contact. Measurement of the Volta potential dierence is possible in a high-impedance electrochemical cell which contains an air
gap. The potential drop across this cell is determined under conditions for which
the potential drop across the air gap is zero, and is known as the compensation
potential dierence.
A schematic diagram illustrating the experimental determination of the Volta
potential dierence between two metals, a and b, is given in g. 8.11. For this
system, electrons are the species common to each phase. The experiment involves
an arrangement in which there is an air gap between a and b. On the other side of
b, there is a second portion of phase a which is designated a 0 , so that the system
can be described as
a j air j b j a 0
8:7:1
Here, the vertical bars represent phase boundaries according to the usual conventions for describing electrochemical cells (see chapter 9). A variable dc potential
409
Fig. 8.11 Experimental setup for determining the

Volta potential dierence between two metals a and
b. The compensation potential is determined with
potentiometer P by moving a with respect to b until
there is zero current in galvanometer G.
may be applied between the metals a and a 0 at the ends of the cell. The eld
strength in the space between a and b is zero if no current ows in the galvanometer when phase a is moved with respect to phase b. Under these conditions,
the Volta potential dierence between the adjacent surfaces of a and b must be
zero. Recalling that the Galvani potential dierence between two phases can be
measured when they have the same chemical composition, this experiment gives
0
Fa a 0 f Ffa Ffa m~ ael m~ ael m~ ael m~ bel
8:7:2
Because of the electronic equilibrium between

phases a and b, the electrochemi0
cal potentials of electrons in these phases, mael and mbel , are equal. Furthermore, the
measurement of a a 0 f is made under conditions that a b c is zero, so that
Fa a 0 f mael Fwa mbel Fwb Wbel Wael
8:7:3
Thus, under these circumstances, the measured Galvani potential dierence is

equal to the dierence in the electronic work functions of the metallic phases a
and b.
At the interface between a 0 and b where there is electronic equilibrium,
0
m~ ael m~ bel
8:7:4
so that one may write

0
mael Ffa mbel Ffb
8:7:5
Rearranging this equation to obtain an expression for b a 0 c, one nds that

0
Fb a 0 c Fca cb mael Fwa mbel Fwb W bel W el
8:7:6
Combining equations (8.7.3) and (8.7.6) the nal result is

c f a a 0 f b a 0 c
W bel W ael
F
8:7:7
where c f refers to the compensation potential dierence. Thus, this experiment

gives the Volta potential dierence between two dierent metals in contact, which
is related in a very simple way to the dierence in work functions for the two
metals.
410
Fig. 8.12 Schematic diagram of

the Kenrick apparatus used to
determine the Volta potential
dierence between two liquids a
and b. Liquid a (Hg in this
example) ows down the center
of tube T in a stream of droplets,
whereas b (0.1M HCl) ows
down the walls.
Another important experiment used to measure the Volta potential dierence

between two liquids was described by Kenrick* [9] and is illustrated here for the
mercury | electrolyte solution interface (see g. 8.12). An air gap between the two
liquids is established in a cylindrical tube T. The mercury emerges from a central
reservoir in a stream of small droplets which ow down the center of the tube. On
the other hand, the solution, namely 0.1 M HCl, ows down its walls. The
experiment is designed so that the two liquids ow suciently rapidly that no
charge can be built up on their surfaces. As a result the Volta potential dierence
across the air gap is zero. The HCl solution is part of an electrochemical half-cell
connected to a calomel electrode. The total cell may be described as
Cu j Hg j Hg2 Cl2 j 0:1 MHCl j Hg 00 j air j 0.1 M HCl 0 j Hg2 Cl2 j Hg 0 j Cu 0
8:7:8
Identical calomel electrodes are on each side of the cell, but that on the left-hand
side is also in contact with the pure mercury which ows down the center of the
cylindrical tube. The purpose of the calomel electrodes is to establish an equilibrium between electrons in mercury and chloride ions in solution through the
process
Hg2 C12 2e 2Hg 2C1
8:7:9
Since these are the only charged species in the equilibrium, any change in the
electrochemical potential of the electron in mercury will be reected in a change in
the electrochemical potential of the Cl in solution. The Galvani potential dierence between the two copper leads, which is measured experimentally, is given by
*Frank Kenrick was a professor in the Department of Chemistry, University of Toronto from 1900
until 1944.
411

0
~ Cu
~ Hg
~ Hg
Fc f FCu Cu 0 f m~ Cu
el m
el m
el m
el
8:7:10
The electrochemical potentials of the electrons in the Cu and Hg in contact are

equal because of their electronic equilibrium. Furthermore, because of the calomel
equilibrium on the left-hand side of the cell,
~ sC1
mHg2 C12 2m~ Hg
el 2mHg 2m
8:7:11
A similar equation can be written for the calomel system on the right-hand side,
which involves Hg 0 and the HCl 0 solution, which has the same concentration as
the solution on the left-hand side. It follows from equation (8.7.11) that the
compensation potential dierence also measures the dierence between the electrochemical potentials of the Cl ions in the two HCl solutions:
0
~ Hg
~ sCl m~ sCl Ffs Ffs
Fc f m~ Hg
el m
el m
8:7:12
The object of the experiment is to measure the Volta potential dierence between
Hg 00 and the HCl solution with which it is in contact. As described above, any
Volta potential dierence between the surfaces of Hg 00 and the HCl 0 solution on
the other side of the air gap is eliminated by streaming these liquids in the central
cylinder of the apparatus. In addition, nitrogen gas is passed up the gap in the
cylinder to prevent condensation of water vapor on the mercury droplets. Since
the resistance of this system is very high it is important that the device used to
measure the potential dierence have an even higher resistance, so that charge
does not ow from the experimental cell to the measuring device.
In order to understand the connection between the observed cell potential
dierence and s Hg 00 c, equation (8.7.12) is rewritten as
0
00
00
c f fs fHg fHg fs
8:7:13
Hg 00
where the term f

has been added and subtracted on the right-hand side. Since
the Volta potential dierence at the air gap between Hg 00 and solution s 0 has been
eliminated:
00
cHg cs 0
8:7:14
Thus, equation (8.7.13) may also be written as

0
00
00
00
c f ws wHg fHg fs cHg cs Hg 00 s c

s
8:7:15
0
where it has been assumed that the surface potential w is equal to ws . This
assumption is perfectly reasonable because the solutions have exactly the same
composition. It follows that the Galvani potential dierence between the copper leads to cell (8.7.8) is equal to the Volta potential dierence from the
mercury streaming in the center of the system to the solution with which it
is in contact.
In a slightly dierent conguration, the Kenrick apparatus can be used to
measure the Volta potential dierence between two miscible liquids. Such a cell is
.
Cu j Hg j Hg2 Cl2 j 0:1M HCl .. S j air j 0:1M HCl 0 j Hg2 Cl2 j Hg 0 j Cu 0
8:7:16
412

.
The three vertical dots .. represent a liquid junction. The exact nature of the
construction of this junction is not given here, but it is used to prevent mixing
of solution S with the 0.1 M HCl solution. More details about liquid junctions and
methods of forming them are given in chapter 9. The other aspect about the liquid
junction that must be considered is that, in general, there is a Galvani potential
drop across it. In the Kenrick experiment, solution S streams down the center of
the vertical tube shown in g. 8.12 and the reference solution, 0.1 M HCl, ows
down the walls of the tube, thereby establishing zero potential drop across the air
gap. On the basis of the analysis given above, the potential drop across the cell
under these circumstances is given by
0
c f Cu Cu 0 f fr fr
0
fr fs fs fr wr ws fs fr s r c
8:7:17
where the superscript r refers to the reference solution, 0.1 M HCl, r 0 , to the
reference solution, 0.1 M HCl 0 , and s, to the solution S.
This system provides a convenient means of studying changes in the surface
potential of a solution associated with adsorption of one of its components at the
solution | air interface. Rewriting equation (8.7.17) as
c f s r f wr w s
8:7:18
where s r f is the liquid junction potential between the two solutions, the compensation potential dierence provides a means of following changes in the surface potential of solution S; ws , provided that s r f is known or is constant.
A. The Surface Potential at the Solution | Gas Interface
The change in surface potential at the air interface has been studied for a variety
of electrolyte solutions in water [6, 10]. The results are reported as w where
w ws wr
8:7:19
wr is the surface potential of a reference solution which is usually the innitely

dilute system. The experimentally observed values of c f were corrected for the
liquid junction potential on the basis of theories discussed in chapter 9. Typical
values obtained by Randles are shown in g. 8.13. For most simple electrolytes,
w is positive and increases in magnitude with increase in electrolyte concentration. The surface potential depends strongly on the nature of the anion and only
weakly on the nature of the cation when this is an alkali metal ion.
One can identify three physical phenomena which lead to the observed values
of w. First, an ionic double layer can be established if the distance of closest
approach for cations and anions to the interface is not the same. Second, if one of
the components of the solution has a dipole moment, it may assume a preferred
orientation at the interface, thereby giving rise to dipolar potential drop. Finally,
the presence of the solute can change the orientation of water molecules at the
interface from that present in the pure solvent. The fact that w is usually positive
is evidence that the anion approaches the surface more closely than the cation.
This is not dicult to understand given that anions are more weakly solvated than
413
Fig. 8.13 Change in the surface potential at the aqueous solution | air interface with
respect to that for the innitely dilute solution plotted against electrolyte concentration.
cations in water. As a test charge approaches the interface it rst experiences an

anionic excess. This leads to a cationic excess further in the solution, with the
result that the net charge in the double layer at the interface is zero (see g. 8.14).
Since the surface potential is measured from the gas phase into the liquid solution
phase, it is positive. The change in the surface potential going from 2 M KCl to
2 M KI is 50 mV. This reects the fact that the iodide ion is more weakly solvated
than the chloride ion, so that the former establishes a greater potential dierence
at the interface. KF is the only electrolyte for which w is negative. This demonstrates that the uoride anion is more strongly solvated by water molecules than
the chloride ion which is in the reference solution.
Values of w have also been determined for solutions of simple inorganic
acids. The surface potential change for the acid is always signicantly larger
Fig. 8.14 Schematic diagram

showing the distribution of K
and I ions at the solution |
gas interface for an aqueous
solution of KI.
414
than that for the potassium salt. For example, w is 2 mV for KCl at a concentration of 1 M, whereas it is 23 mV for 1 M HCl. The results suggest that the
hydronium ion is oriented at the interface with the hydrogen atoms directed into
the solution and the negative electron density on the oxygen atom directed toward
the gas phase. The hydronium ion contributes in two ways to the surface potential, both as a cation and as a dipole, so that the net electrostatic eect is larger.
The eect of organic solutes on the surface potential has been studied in some
detail [G2]. In the cases of short-chain alcohols and acids, and simple ethers, the
value of w is negative and much larger than that found for inorganic systems.
This result is interpreted as evidence that water-soluble organic molecules are
oriented at the interface with the hydrophobic part, for example, the alkyl
group, pointing into the gas phase. The polar group interacts with the water
molecules, which can form hydrogen bonds with electronegative atoms like oxygen and nitrogen. As a result, a test charge entering the solution from the gas
phase experiences the electrostatic eect of a dipolar layer with its positive end
pointing to the gas phase and its negative end into the solution. Some typical
values of w are given in table 8.4.
The value of w increases with increase in the concentration of the organic
solute and reaches a limiting value wm in the same way as the surface excess.
This limit is independent of the length of the alkyl chain but it does depend on the
nature of the functional group. For example, in the case of alcohols wm is
approximately 350 mV, whereas it is about 550 mV for dialkylethers.
The shape of the dependence of w on solute concentration suggests that there
may be a simple relationship between w and the surface excess . This is illustrated for the 1-propanolwater system by the plot in g. 8.15. The simple linear
relationship demonstrates that the alcohol molecule is adsorbed in a monolayer,
and that the orientation of individual molecules at the interface does not change
with coverage. The relationship between w and is not always simple and
depends on the solubility of the organic solute. When this is sparingly soluble
or insoluble, a linear relationship is not observed. In this case, the experimental
observations are attributed to change in the orientation of the organic molecules
in a monolayer at the interface with the extent of lling of the monolayer.
Table 8.4 Typical Values for the Change in Surface

Potential at the Aqueous Solution | Air Interface After
Addition of an Organic Solute
Organic Solute
Electrolyte
1 M CH3OH
1 M C2H5OH
1 M C3H7OH
1 M HCOOH
1M CH3COOH
1M (CH3)2CO
0.5M (C2H5)2O
1M C2H5NH2
0.01 M KCl
0.01 M KCl
0.005 M H2SO4
0.005 M H2SO4
0.005 M H2SO4
Surface Potential Change

w=mV
100
230
220
40
285
380
480
310
415
Fig. 8.15 Plot of the change in surface potential w against surface excess for 1-propanol
in water at the solution | air interface [11].
When the orientation of the organic molecule at the interface does not change
with coverage, a simple relationship exists between w and , namely
w
NL
p cos y
es e0
8:7:20
NL is the Avogadro constant, p, the dipole moment of the adsorbate molecule, y,

the angle that the dipole vector makes with the direction perpendicular to the
interface, es the relative permittivity of the solvent, and e0 , the permittivity of free
space. Assuming that es is 6 in the surface region, the data in g. 8.15 yield an
estimate for p cos y of 1.4 debyes. This result is reasonable when compared to the
dipole moment of 1-propanol (1.7 debyes). It is clear that equation (8.7.20), which
shows that w is proportional to , can only be applied if the orientation of the
organic molecules at the interface does not change with coverage.
Studies of w have been used to make estimates of the absolute value of wW ,
the surface potential at the interface between pure water and air. Since this quantity cannot be determined experimentally, such estimates necessarily involve an
extrathermodynamic assumption. The most convincing evidence about the sign of
wW and thus, the orientation of water at the interface, comes from experiments
designed to study the temperature coecient dwW =dT. The cell used was
Cu j Hg j Hg2 Cl2 j 1:0 M KCl; H2 O T0 j 0:01 M KCl T0 j air
j 0:01 M KCl T1 j 0:01 M KCl T0 j 1:0 M KCl T0 jHg2 Cl2 j Hg0 j Cu0
8:7:21
where the temperature of the KCl solution on the right-hand side of the air gap,
T1 , was varied with respect to the temperature of the other components of the cell
(T0 ). All of the liquid junction potentials in this cell can be estimated except that
416
between the two KCl solutions with dierent temperatures. Assuming that this is
negligible, analysis of the thermodynamics of this cell leads to the result that
dc f dwW
dT
dT
8:7:22
Experimental data were collected for several values of T1 in the vicinity of room
temperature and gave the result that dwW =dT is 2:7 104 V K1 . The fact that
the temperature coecient is negative shows that the surface potential wW is
positive [10]. As temperature increases, thermal disorder results in a decrease in
wW which eventually goes to zero at the critical temperature (647 K).
Various methods for estimating wW were critically reviewed by Randles [10].
Although it is dicult to make a precise estimate, it is clear from the experimental
work that the surface potential of water is positive and approximately equal to 80
mV at 25 C.
Surface potentials have also been estimated for a variety of non-aqueous solvents [12, 13]. Some of these estimates are given in table 8.5. The results are
mainly based on compensation potential measurements in cells with an aqueous
solution on one side of the air gap and the non-aqueous solution on the other.
Details about these measurements and the extrathermodynamic assumption
involved are given in the original literature [12, 13]. It is especially interesting
that wS for the polar non-aqueous systems is always negative, so that water is
unique with a positive value for its surface potential. The negative surface potential for non-aqueous solvents indicates that the hydrocarbon part of the molecule
points into the gas phase, whereas the dipolar functional group points into the
solution. Thus, a test charge passing from the gas phase into solution experiences
a dipole potential dierence which is negative. In the case of an alcohol such as
methanol, the polarOH group prefers to remain in solution where it is involved
in hydrogen bonding. For aprotic systems like acetonitrile, the polar C
N group
is involved in strong dipoledipole interactions with other solvent molecules.
Although one can be quite condent that the sign given to the values of wS is
correct, there is a large uncertainty in the magnitude. Estimation of surface
potentials is closely related to the experimental measurement of the real potentials
Table 8.5 Estimates of the Surface
Potential for Some Polar Solvents
[12, 13]
Solvent
Water
Methanol
Ethanol
Acetone
Acetonitrile
Dimethylformamide
Dimethylsulfoxide
Nitrobenzene
Surface Potential
wS =mV
80
230
300
390
100
260
250
160
417
for individual ions in solution. This is an important part of the study of the
properties of electrolyte solutions. It is discussed in detail in the following section.
B. The Work Function for Ions in Solution
The work function for ions is dened with respect to the process in which one
mole of ions in the ideal standard state of unit concentration in a given solvent are
transferred to charge-free innity in vacuum as unsolvated ions under conditions
that the solution bears no net charge. This process diers considerably from that
for electrons in metals, which involves removal of electrons from the Fermi level
in the metal to charge-free innity. An important aspect of the latter process is
that it involves the denition of zero on the physical potential scale. Thus, an
electron at rest in vacuum is dened to have zero potential energy on that scale.
Using the above denition, the work function for an ion is based on the process
ionsoln!iongas
8:7:23
~ is;
W si m~ g;
i m
8:7:24
and is given by
where m~ xi is the electrochemical potential of ion i in phase x, and the superscript

refers to the standard state. Since the solution phase carries no net charge
(cs 0), and the inner potential of the gas phase can be set equal to zero, one can
write
W si mig; mis; zi Fws0 mig; as;
i
8:7:25
where ws0 is the surface potential of the solution, and as;

i , the standard real
potential of ion i in solution. In order to determine ais; one must be able to
measure the Volta potential dierence between the solution and a metal electrode
which is in electrochemical equilibrium with ion i. This point is illustrated in detail
in the following discussion.
The standard real potential for the H+ ion can be determined using the following cell:
Cu j Hg j air j x m HCl j Pt, H2 j Cu 0
8:7:26
When the compensation potential is determined using the Kenrick apparatus, the
mercury streams down the center of the vertical tube and the HCl solution down
its walls so that the Volta potential dierence across the air gap is eliminated. In
this system both the concentration of HCl and the pressure of hydrogen gas can
be varied. It diers from cells (8.7.8) and (8.7.16) in that no attempt is made to
balance the charge carriers on opposite sides of the cell. The Galvani potential
dierence measured between the two copper leads can be related to the electrochemical potentials of electrons in the mercury and platinum as follows:
0
~ Cu
~ Hg
~ Pt
Fc f FCu Cu 0 f m~ Cu
el m
el
el m
el m
Pt
Pt
Hg
mHg
el mel F f f
8:7:27
418
In order to proceed further, one must relate the electronic equilibria in the metallic terminals of the cell to the electrochemical equilibrium in the hydrogen halfcell.
This equilibrium is
( 1 H2 gas
H soln e Pt +
2
8:7:28
so that the condition for equilibrium is

m~ sH m~ Pt
el
mgH2
8:7:29
where the superscript s designates the solution phase. Expanding each of the
electrochemical and chemical potentials and rearranging, one can show that
Ffs fPt G RT ln aH
RT ln pH2
2
8:7:30
where

G
g;
mH
2
s;
Pt;
mH
mel
8:7:31
Combining equations (8.7.27) and (8.7.30) for the case that the pressure of the
hydrogen gas is 1 bar, one may write
Fc f Ffs fHg RT ln aH G
8:7:32
where
0
G
mg;
H2
2
Hg;
s;
mH
mel
8:7:33
Since the Volta potential dierence across the air gap between the mercury and
the aqueous solution is maintained at zero, this result may also be written as
Fc f Fws wHg RT ln aH G
8:7:34
Equation (8.7.34) shows that the compensation potential depends on the hydrogen ion activity in the electrolyte solution. Strictly speaking, single ion activities
are not subject to experimental determination. However, if the solution is suciently dilute, the single ion activity aH may be estimated by the mean activity of
the electrolyte a to a good approximation.
Cell (8.7.26) was studied in detail by Farrell and McTigue [14] using 13 dierent molalities of HCl in the range 0.0010.02 m. In this range, it is legitimate
to assume that the hydrogen ion activity aH is equal to the mean ionic activity
a , which is easily determined experimentally. Thus, it is possible to determine
c f RT=F ln a from the experimentally measured compensation potential. A
plot of this quantity against the square root of the ionic strength is shown in g.
8.16. It is clear that this plot is linear at lower concentrations, so that the intercept,
which gives the value of c f RT=F ln a in the limit of zero ionic strength, can
easily be found. From g. 8.16 the result is 56:3 mV. Keeping in mind that ws is
equal to ww in the innitely dilute solution, equation (8.7.34) can be rewritten as
419
Fig. 8.16 Plot of c RT=F ln a for cell (8.7.26) against the square root of the ionic
strength due to HCl, I1/2.
s;
W
Fc f0 RT ln a mH
mHg;
FwHg
Fw
el
s;
Hg;

aH
ae
mg;
H2
mg;
H2
2
8:7:35
s;
aH
is the standard real potential for the hydrogen ion in an ideal innitely
where
dilute solution, and aHg;
, the standard real potential for an electron in mercury.
e
Thus, the Kenrick experiment carried out with cell (8.7.26) gives the Gibbs energy
change for the reaction:
1
H gas!H soln e Hg
2 2
8:7:36
In order to estimate the work function for the H+ ion, two other pieces of
information are required. The rst is the work function for an electron in mercury
from table 8.2, which is equal to 4.50 eV (434.2 kJ mol1). The second piece of
information is the work associated with the dissociation and ionization of molecular hydrogen in the gas phase, corresponding to the reaction
1
H gas!H gas e gas
2 2
8:7:37
The energy required for this process is 1517.1 kJ mol1 [15]. Using these data, the
Gibbs energy associated with the reaction
H soln!H gas
8:7:38
is 1088.3 kJ mol1. This is precisely the work function for the H+ ion in aqueous
solution.
Having obtained the work function for one ion it is now a simple matter to
extract work functions for other ions provided the necessary thermodynamic data
420
Table 8.6 Work Functions for Simple Monovalent

Ions at Innite Dilution in Water at 25 C [14]
Ion
Wi =kJ mol1
H+
Li+
Na+
K+
Rb+
Cs+
F
Cl
Br
I
1088
509
409
336
314
283
436
319
305
258
for the appropriate electrolytes are available. For example, in the case of HCl, the
Gibbs energy for the process
HClsoln!H gas Cl gas
8:7:39
1
is equal to 1405.2 kJ mol . Comparing equations (8.7.38) and (8.7.39), it is clear

that the work function for the Cl ion is 316.9 kJ mol1. By choosing electrolytes
with one ion for which the work function is known and the other for which it is
not known, the list of values of W si may be extended. A summary of results for
monovalent monoatomic ions is given in table 8.6.
It is interesting to compare values of the standard real potential for a given ion
in table 8.6 with values of the absolute Gibbs energy of solvation given in table
3.3. On the basis of equation (8.4.4),
s
aiS; W si mS;
i zi Fw0
8:7:40
miS;
where
is the standard Gibbs energy for ion i in solvent S, which is equal to the
standard Gibbs energy of solvation, s G. This comparison gives a method of
obtaining an independent value of s G for the hydrogen ion using the estimate
of the surface potential of water discussed earlier. Thus,
s GH 1088 96:5 0:08 1096 kJ mol1
8:7:41
1
This result is 8 kJ mol more positive than the estimate discussed in section 3.5
(1104 kJ mol1). Considering that both estimates involve extrathermodynamic
assumptions the agreement is quite good.
Values of the work function for simple ions have also been measured in nonaqueous solutions [16]. These experiments involve measuring the compensation
potential in a cell such as
Ag j AgCl j NaCl, W j air j NaCl, S j AgCl j Ag 0
8:7:42
The compensation potential for this system is

0
c f fAg fAg
8:7:43
421
Because of the equilibrium at the Ag | AgCl electrodes, namely,

+ Ag Cl
AgCl e (
8:7:44
the inner potentials of the silver electrodes at each end of the cell can be related to
the electrochemical potentials of the chloride ion in each solution. Thus,
0
~ SC1
Fc f FfAg FfAg m~ W
C1 m
8:7:45
m~ W
C1
where
is the electrochemical potential of the chloride ion in the aqueous
solution, and m~ SC1 , that in the non-aqueous solution involving solvent S. The
Volta potential dierence across the air gap is eliminated, for the example, by
using the Kenrick experiment (g. 8.12). Then, equation (8.8.26) can be rewritten
as
W
S
s
W
S
Fc f mW
C1 Fw mC1 Fw aC1 aC1
8:7:46
where miC1 is the chemical potential of the chloride ion in phase i, and aiC1 its real
potential in the same medium. Since aW
C1 is known, this experiment gives the real
potential of the Cl ion in the non-aqueous solvent S. Now, thermodynamic data
for solvation of whole electrolytes, for example, HCl, can be used to extract real
potentials for other ions. Values of the work function for the hydrogen ion in
several non-aqueous media are summarized in table 8.7 (W SH aSH :
In assessing the change in W si with solvent nature, one should keep in mind
that both the Gibbs energy of solvation and the surface potential contribution
change with solvent nature. The change in the rst contribution can be assessed
using data for the Gibbs energy of transfer which are given in chapter 4. On the
other hand, one may estimate the Gibbs solvation energy using an estimate of the
surface potential term which gives a small contribution to the real potential.
EXAMPLE
The real potential of the Na+ ion in methanol is 435.6 kJ mol1. Estimate the
Gibbs energy of solvation of Na+ ion in methanol using the surface potential
for this solvent given in table 8.5. Compare this result with that obtained from
the Gibbs energy of solvation of Na+ ion in water (424 kJ mol1) and the
Gibbs energy of transfer of Na+ ion from water to methanol (8.2 kJ mol1).
Table 8.7 Values of the

Work Function for the
Proton, WHS , in Different
Polar Solvents
Solvent
Water
Methanol
Ethanol
Formamide
Acetone
Acetonitrile
W SH =kJ mol1
1088
1113
1111
1102
1118
1077
422
The Gibbs solvation energy for ion i in solvent S is given by

s Gi mSi aSi zi FwS
8:7:47
The surface potential for methanol is 0.23 V. Therefore

s GNa 435:6 96:5 0:23 413:4 kJ mol1
8:7:48
The value of the Gibbs energy of solvation estimated from the data in water
together with the Gibbs energy of transfer is
s GNa 424 8:2 416 kJ mol1
8:7:49
The estimates dier by 3 kJ mol1. Considering the fact that extrathermodynamic assumptions are involved in both routes used to obtain s GNa , the
agreement is quite good.
Values of aSi are available for a number of ions in several non-aqueous solvents
[16]. These data provide a route to estimating the Gibbs solvation energy for
individual ions when the estimates of wS given in table 8.5 are used to calculate
the work done in transporting the ion across the double layer at the solutionair
interface.
The Kenrick experiment and other methods for determining the work function
for individual ions in solution provide valuable information about ion solvation.
This subject is of great interest in the development of realistic models for electrolyte solutions and in understanding their properties at the microscopic level.
8.8 The Metal | Solution Interface

The metal | solution interface has special signicance in electrochemistry and
electroanalysis. The usual electrochemical cell contains two such interfaces, and
an understanding of its properties is important in describing the physical chemistry of these systems.
When a metal comes into contact with a polar liquid, a redistribution of charge
occurs at the newly formed interface. At the metal | air interface, the electrons spill
out of the metal atom lattice, giving rise to a positive surface potential, wm . At the
metal | solution interface, this spillover is not the same because the same electrons
encounter the electron density associated with the solvent molecules. Similar
considerations apply to the liquid phase where the molecules usually adopt a
special orientation at the liquid | air interface giving rise to the surface potential
of the solvent or the solution, ws . This orientation may change upon contact with
the metal phase. Finally, there is usually a redistribution of charge between the
two phases. The exact nature of this process depends on the composition of the
two phases, as will be seen from the following discussion.
The electrostatic properties of the metal | solution interface are expressed in
terms of the Galvani potential dierence, s m f, where
s m f
fm fs
8:8:1
423
fm is the inner potential of the metal, and fs that of the solution. The Galvani
potential dierence cannot be measured experimentally. On the basis of the above
discussion, it is made up of three contributions, so that it can also be written as
s m f
s f g m g s
8:8:2
where s f is the potential drop due to the net free charge at the interface, gm , the
dipolar potential in the metal phase, and gs , that in the solution. gm at the interface between the condensed phases is the analog of wm at the metal | gas interface;
in general, these quantities are not equal. The same comments apply to gs and ws .
The value of the Galvani potential dierence may be changed by changing the
composition of the metal phase, for example, by using amalgams of varying
composition, or by changing the composition of the solution. In this way one
can achieve conditions for which there is no net charge at the interface, so that
s f is zero and
s m f
s
gm
0 g0
8:8:3
This is the potential of zero charge, an important point of reference for this type of
interface. In equation (8.8.3), the subscript 0 designates the specic conditions
which are present at this reference potential.
On the basis of g. 8.5, the relationship between the Galvani potential dierence between the metal and the solution, and the Volta potential dierence just
outside of these phases where they form an interface with the gas phase is
s m f
s m c wm ws
8:8:4
Only the Volta potential dierence s m c can be measured experimentally. For

example, using the Kenrick experiment illustrated in g. 8.12, one can measure the
value of s m c between liquid metals like mercury and electrolyte solutions. In the
case of solid metals, the cell used to measure the compensation potential can have
the same conguration as that for the Kenrick experiment (cell (8.7.9)). However,
the air gap is formed between a metal disk placed a small distance above the
solution phase. The potential dierence between these phases is eliminated using
ionizing alpha radiation from a radioactive substance placed in the gap. The end
result is that the Volta potential dierence is obtained for the cell.
Typical values of s m c0 for the metal | solution interface at the PZC are given
in table 8.8. The Volta potential dierence is negative, indicating that the metal |
air interface is negative with respect to the solution | air interface. Combining
equations (8.8.3) and (8.8.4), one obtains
s m c0
m
s
s
gm
0 w g0 w
8:8:5
Thus the Volta potential dierence at the PZC is related to the dipole potential
dierences at the metal | solution, metal | air, and solution | air interfaces.
EXAMPLE
Estimate the Galvani potential dierence at the Hg | aqueous solution interface

at the PZC given that the surface potential at the Hg | air interface is 2.2 V.
The surface potential at the metal | air interface may be estimated by a
variety of techniques, including the jellium model for sp metals. In the case
424
Table 8.8 Values of the Volta

Potential Difference at the
Metal | Water Interface for
Different Metals at the Point
of Zero Charge (PZC)
Metal
Volta Potential Dierence

W m c0 =V
0.295
0.425
0.495
0.248
0.365
0.335
0.365
0.485
Bi
Cd
Ga
Hg
In
Sn
Tl
Zn
of mercury, it is estimated from thermodynamic data for the Gibbs energy of

solvation of metals in mercury [G2].
The estimate of the surface potential of water is 0.08 V. Thus from equation
(8.8.4)
s m f0
0:25 2:2 0:08 1:9 V
8:8:6
This potential dierence is such that electrons tend to remain in the metal
phase and do not easily go over to the solution phase. The value of s m f0
depends greatly on the value of wm , which in turn depends on the nature of the
metal.
The Volta potential dierence has also been estimated at the mercury | solution
interface for dierent polar solvents. The results obtained (table 8.9) demonstrate
that interaction between mercury and the solvent varies greatly with solvent
Table 8.9 Values of the

Volta Potential Difference at
the Mercury | Solution
Interface for Different
Solvents at the PZC
Solvent
Water
Methanol
Ethanol
Acetone
Dimethylformamide
Dimethylsulfoxide
S Hg c0 =V
0.248
0.52
0.45
0.62
0.68
0.69
425
nature. It is clear that the dipolar contribution to interfacial properties is an

important part of its electrostatic description.
Further study of the metal | solution interface reveals two limiting cases in
terms of their thermodynamic properties. One case occurs when the solution
contains cations formed from the metal and corresponds to electrochemical equilibrium. In general, this process can be described as
+M
Mz ze (
8:8:7
where z is equal to the oxidation state of the metal cation. The thermodynamic
condition for equilibrium is
m~ si zm~ m
el mm
8:8:8
m~ m
el ,
m~ si
that for the

where is the electrochemical potential of the cation in solution,
electron in the metal, and mm , the chemical potential of the metal. The electrochemical potential of the cation is given by
s
m~ si ms;
i RT ln ai zFf
8:8:9
mis;
is the standard chemical potential of this species and, ai , its activity.

where
The corresponding equation for electrons is
m;
m
m~ m
el mel Ff
8:8:10
where it is assumed that the electron concentration in the metal cannot be changed signicantly. Therefore, the condition for equilibrium leads to the result that
s m f
G RT
ln ai
zF
zF
8:8:11
where
m;
G mm ms;
i zmel
8:8:12
Equation (8.8.11) is a form of the Nernst equation for equilibrium (8.8.7).

However, it is expressed in terms of the Galvani potential dierence which cannot
be measured experimentally. As will be seen in the discussion in chapter 9, this
problem is overcome in practice by dening a convenient reference half reaction
whose value of G is arbitrarily set equal to zero.
If reaction (8.8.7) is driven in either the forward or reverse direction by passing
a current, the potential s m f does not change provided there is no signicant
change in the ionic activity ai . For this reason, the electrode | solution interface is
termed non-polarizable. In practice, the forward and reverse reactions of the
equilibrium (8.8.7) occur at nite rates.
The other limiting case for the metal | solution interface occurs when there is no
ow of charge across the interface, so that it behaves as a capacitor. A simple
example is the Hg | aqueous KC l interface. By connecting the mercury to an
external d.c. source, the charge density on its surface can be changed without
current owing. However, the potential drop s f changes in response to the
change in charge density at the interface. This leads to a corresponding change
in s m f (equation (8.8.2)), This limiting form of the metal | solution interface is
called polarizable and is characterized by electrostatic equilibrium. In practice,
426
there are limits over which the value of s m f can be changed. For the example
given, if the mercury electrode is made too negative, the K+ ion is reduced by the
reaction
K Hg e !K; Hg
8:8:13
to form potassium amalgam. On the other hand, if the mercury potential is driven
too far in the positive direction, mercury is oxidized in the reaction
2Hg 2Cl !Hg2 Cl2 2e
8:8:14
in which insoluble mercury(I) chloride is formed. If either of these reactions

occurs, the polarizable nature of the interface is lost. However, there is an interval
of approximately two volts over which the system remains polarizable.
The metal | solution interface both in its polarizable and non-polarizable forms
is extremely important in electroanalysis and in practical electrochemical devices.
The discussion in this section has focused on its fundamental electrical properties.
These systems are considered in much more detail in chapter 9, which deals with
electrochemical equilibria, and chapter 10, which is concerned with polarizable
interfaces and the electrical double layer.
8.9 The Liquid | Liquid Interface

Many examples exist of interfaces formed between two immiscible liquids. A wellknown one is the interface between a long-chain hydrocarbon, for example,
dodecane, and water, which is commonly known as the oil | water interface.
Dodecane and water are immiscible because the hydrocarbon phase is nonpolar. Liquid |liquid interfaces are also formed between water and organic liquids
with polar groups such as octanol and heptanoic acid, which also have rather long
hydrocarbon chains. The polar liquid nitrobenzene, which has a relative permittivity of 35, is also immiscible with water. Another well-known system is the
mercury | polar liquid interface. This has been studied extensively, especially
for aqueous electrolyte solutions. However, the mercury | polar liquid interface
is also an example of a metal | solution interface which was considered in the
previous section. The discussion here is limited to liquids with relative dielectric
permittivities falling in the range 1200, and systems which have poor conductivities as pure liquids.
In dening the interfacial tension of liquids in section 8.2, the discussion was
kept general in terms of two uids a and b. When one of these uids is air, then
the interfacial tension has the values which are given in table 8.1 for a collection of
polar liquids. A relationship between the interfacial tension between two immiscible liquids a and b, and that at the liquid | air interface for each pure liquid can
also be dened. These quantities are connected by the work of adhesion wab,
which is the work required to separate the two phases so that an air interface is
formed at both a and b. Thus, if gab is the interfacial tension at the a | b interface,
that is, the work required to increase the area of this interface by unit amount,
then
427
Table 8.10 Interfacial Tension Between

Water and Immiscible Organic Liquids
at Room Temperature [G1]
gab / mJ m2
Interface
Water
Water
Water
Water
Water
Water
Water
|
|
|
|
|
|
|
benzene
chloroform
carbon tetrachloride
nitrobenzene
octane
heptanoic acid
octanol
gab wab ga gb
34.4
33.3
43.8
24.7
49.8
5.3
7.3
8:9:1
where ga and gb are the interfacial tensions at the a|air and b|air interfaces,
respectively. Some typical values of gab for interfaces in which water is one component are given in table 8.10. A wide range of values are found reecting the way
in which the organic system interacts with the water phase at the interface.
There has been considerable interest in recent years in electrochemical processes at liquid|liquid interfaces involving water and an immiscible organic liquid
[17, 18]. When the organic liquid is suciently polar, as is nitrobenzene (NB), an
electrolyte may be dissolved in both liquids, so that they become electrically
conductive. This system is often referred to as an interface involving two immiscible electrolyte solutions, or ITIES. Electrolytes commonly used in NB include
tetraphenylarsonium tetraphenylborate (TATB), and tetra-n-butylammonium tetraphenylborate (TBATB). LiCl is often used as a background electrolyte in water.
Electrolytes such as TATB and TBATB are virtually insoluble in water, whereas
LiCl does not dissolve in NB. Thus, when the NB|water interface is formed the
electrolytes remain almost entirely in their original phases. Such a system behaves
as an ideally polarizable interface over a limited potential range. Thus, the application of a potential dierence between two electrodes, one placed in the aqueous
phase and the other in the NB, results in the accumulation of charge at the
interface.
When the electrolyte is soluble in both phases, an equilibrium is established in
which it is partitioned between the two phases. At equilibrium, the chemical
potential of the electrolyte is equal in the two liquids, that is,
NB
mW
e me
8:9:2
where mie is the chemical potential of the electrolyte in phase i. The electrolyte
activities in the two phases are related by the expression
!
aNB
e
RT ln
meNB; tr Ge
8:9:3
mW;
e
aW
e
428
X;
where aX
the corresponding stane is the electrolyte activity in phase X, and me

dard chemical potential, and tr Ge the standard Gibbs energy of transfer from
water to the non-aqueous phase. The factors controlling the partition equilibrium
can be seen more clearly when the equilibrium condition for the individual ions is
considered. For the sake of simplicity, a simple 11 electrolyte MX is considered.
Equating the electrochemical potentials of the cation in the two phases, one
obtains
W
W
NB;
NB
mW;
RT ln aNB
8:9:4
M Ff
M RT ln aM Ff mM
i
where mW;
M is the standard Gibbs energy of cation M in phase i, aM , its activity,
i
and f , the inner potential of phase i. The Galvani potential dierence at the
water nitrobenzene interface is
!
tr GM RT
aW
M
ln NB
8:9:5
W NB f
F
F
aM
The Galvani potential dierence may also be expressed in terms of the equilibrium
for the anion, which leads to
!
tr GX RT
aNB
X
ln W
8:9:6
W NB f
F
F
aX
The Galvani potential dierence may not be found experimentally because single
ion quantities are not subject to experimental determination. However, because of
the requirement of electroneutrality, the concentration of cations and anions in
each phase is equal. Furthermore, in the limit of dilute solutions, the activities of
these ions are equal in each phase. Thus, by equating (8.9.5) and (8.9.6), one
obtains for dilute electrolyte concentrations
!
aNB
G tr GX
G

tr e
8:9:7
ln W tr M
2RT
2RT
a
On the basis of the thermodynamics of electrolytes, equation (8.9.7) is the equivalent of equation (8.9.3).
Because the partition equilibria observed at the liquid | liquid interface are
relevant to interfacial phenomena in specic ion electrodes and biological membranes, there is an interest in determining single ion quantities associated with
transfer of an ion from water to the non-aqueous phase. This quantity can only be
estimated from experimental data after making an extrathermodynamic assumption. One common assumption discussed earlier in section 4.8 is the so-called
TATB assumption, according to which
tr GTA tr GTB
tr GTATB
2
8:9:8
where tr GTATB is the standard Gibbs energy of transfer of the whole electrolyte,
that is, tetraphenylammonium tetraphenylborate. By determining the partition
ratio in dilute solutions for a series of electrolytes, values of the single ion standard Gibbs energy of transfer may be estimated. Some typical values for the
water | nitrobenzene interface are given in table 8.11. In the case of the mono-
429
Table 8.11 Standard Gibbs Energy of Transfer

for Monovalent Ions from Water to Nitrobenzene
at 25 C (TATB Assumption) [18]
Cation
Li
Na
K
Rb
Cs
TMA
TEA
TBA
TA
tr G =kJ mol1

38.2
34.2
23.4
19.4
15.4
3.4
5.7
24.0
35.9
Ion
tr G =kJ mol1
F
Cl
Br
I
ClO4
BF4
NO3
IO4_
TB
44.0
31.4
28.4
18.8
8.0
11.0
24.4
6.9
35.9
TMA tetramethylammonium;
TEA tetraethylammonium; TBA tetrabutylammonium.
atomic ions, tr G is large and positive, demonstrating that the ion is much less
soluble in NB than in water. However, there is a trend for tr G to decrease as the
radius of the ion increases, as one would expect. Negative values of tr G are
found for most of the tetraalkylammonium cations and the tetraphenyl ions in the
reference electrolyte TATB.
In the study of the interface with two immiscible electrolyte solutions (ITIES),
considerable attention has been focused on the estimation of the Galvani potential dierence at the water | oil interface on the basis of a reasonable extrathermodynamic assumptions. The discussion of these estimates is often made
in terms of the ionic distribution coecient, which is dened on the basis of
equations (8.9.5) and (8.9.6). Generalizing this equation for the a | b interface
at which ion i with charge zi is transferred, one may write
mia; mib; RT
aai
ln

f
a b
zi F
zi F
abi
!
8:9:9
The limiting ionic distribution coecient is dened as
Kab
i
ma; mib;
exp i
RT
!
8:9:10
so that the expression for the Galvani potential dierence may also be written as
!
a
RT
ab ai
ln Ki b
8:9:11
a b f
zi F
ai
For a simple 11 electrolyte one obtains
430
RT
aaM
ln Kab
a b f
M b
F
aM
or
RT
Kab
M
ln

f
a b
2F
Kab
X
RT
aaX
ln Kab

X b
F
aX
RT
aa abX
ln M
2F
abM aaX
!
8:9:12
!
8:9:13
where the subscripts M and X refer to contributions from the cation and anion,
respectively. In dilute solutions where the activity of the cation is equal to that of
the anion, this expression becomes
!
RT
Kab
M
ln ab
8:9:14
a b f
2F
KX
One method of estimating a b f is based on the measurement of cell potential
dierences [19]. Consider the cell
. .
.
. .
. .
Hg j Hg2 Cl2 j sat0 d KCl .. .. MX in W .. MX in b .. .. TEAPic in b .. .. TEAPic in
1
2
3
4
W j Hg2 Pic2 j Hg0
8:9:15
where Pic stands for picrate (2,4,6-trinitrophenolate). The measured potential

drop across the cell is Hg Hg 0 f1 and the important Galvani potential drops at
each liquid | liquid junction have been numbered. The Galvani potential drop of
interest is that at position 2. Those at positions 1, 3, and 4 are designed to be very
small. The second cell considered is
..
..
10
20
30
Hg j Hg2 Cl2 j sat0 d KCl .. .. TEAPic in W .. TEAPic in b .. .. TEAPic in W jHg2 Pic2 j Hg0
8:9:16
In this cell all liquid | liquid junction potentials are negligible, so that the measured potential Hg Hg 0 f2 reects the dierence in potentials of the two reference
electrodes which are common to each cell. As a result
W b f
Hg
Hg0 f1 Hg Hg0 f2
8:9:17
The Galvani potential dierence or distribution potential measured in this way

agrees well with that calculated from values of the standard Gibbs energies of
transfer of the component ions, which have been estimated on the basis of solubility data.
EXAMPLE
Estimate the Galvani potential dierence for KI at the water | nitrobenzene

interface using the data for the standard Gibbs energy of transfer given in table
8.11.
For the K cation, the value of tr G is 23.4 kJ mol1. The corresponding
limiting ionic distribution coecient is

Kab
exp
K
23:4
7:96 105
8:3144 0:2982
For the I anion, tr G is 18.8 kJ mol1 and

18:8
Kab
5:09 104
exp
I
8:3144 0:2982
431
8:9:18
8:9:19
On the basis of equation (8.9.14), the Galvani potential dierence is

!
0:0257
7:96 105
ln
0:048V
8:9:20
a b f
2
5:09 104
The negative result reects the fact that the NB phase is negative with respect
to the water phase. Since the K cation is more dicult to solvate in nitrobenzene than the I anion, the value of a b f shows that electroneutrality is
maintained in the system by attracting K cations into NB and at the same
time repelling I anions from it.
The above calculations are only valid in the absence of ion pairing. Since ion
pairing is strong in NB, the observed values of a b f are expected to change
signicantly with concentration of the electrolyte.
The above discussion applies to a non-polarizable interface with a common
electrolyte MX distributed in two immiscible phases a and b:
.
MX; a .. MX; b
8:9:21
Non-polarizability implies that the Galvani potential dierence a b f does not

change when the activity of MX is changed on one side of the interface. Addition
of MX to phase a results in the transfer of the ions M+ and X through the
interface into phase b in order to maintain the distribution equilibrium as dened
in equation (8.9.13), and constancy of a b f. The rate of the ion transfer processes
is nite so that re-establishment of equilibrium is a kinetic process which is of
fundamental interest in itself.
One may also design interfaces which behave as ideally polarizable systems. In
this case, one or more of the component ions is essentially present in only one
solvent. In order to illustrate this situation, a system with a common anion but
two dierent cations is considered rst:
.
M1 X; a .. M2 X; b
8:9:22
Anion X dissolves easily in both phases, and its concentration in each is determined by the total cation concentration. The distribution coecient for M1, Kab
M1 ,
is a very small number, indicating that this cation is predominantly in the a phase.
On the other hand, the distribution coecinet for M2, Kab
M2 is a very large number
because this cation is predominantly in the b phase. The Galvani potential
dierence is determined by the anion X so that
!
a
RT
ab aX
ln KX b
8:9:23
a b f
F
aX
432
In this system the anion X is called the potential determining ion. Assuming that
the anion can be added to or removed from phase a by adding the electrolyte
M1X, the activity of X remains approximately constant because the electrolyte
concentration in phase b does not change. Thus, changes in aaX are directly
reected in changes in a b f according to equation (8.9.12). As a result, the
Galvani potential dierence is linear in ln aaX . Under these circumstances the
potential changes can be used to detect changes in anion activity as is done
with a specic ion electrode. If aaX is very low or very high, changes occur in
abX , so that a b f is no longer linear in ln aaX .
EXAMPLE
An ITIES is formed between water and nitrobenzene (NB) using electrolytes

with the picrate (Pic) anion at the same activity (0.1 M). The electrolyte in
water is Li Pic and that in NB, tetraphenylarsonium (TA) picrate. Given that
the standard Gibbs energy of transfer of Pic is 4.6 kJmol1 [18], estimate the
Galvani potential dierence at the interface. Use the data in table 8.11 to
estimate the activity of Li in NB.
On the basis of equations (8.9.5) and (8.9.6), one may write for ion i
W NB f
tr Gi RT aw
ln i
zi F aNB
zi F
i
8:9:24
Applying this equation to the Pic ion

W NB f
4:6
0:048 V
96:485
8:9:25
Thus, the NB phase is negative with respect to the W phase. From table 8.11,
the Gibbs energy of transfer for Li is 38.2 kJ mol1. From equation (8.9.24)
W
ln aNB
Li ln aLi
tr GLi
F
f

RT W NB
RT
8:9:26
38:2
ln0:1
38:92 0:048 15:84
8:314 0:2982
The activity of Li in the NB is 1:3 107 M. Although Li is attracted to the
NB phase by the fact that it is negative with respect to water, the activity is very
small because of the large positive value of tr GLi
Another important system is one with a common electrolyte in both phases,
and a second electrolyte in which one ion is soluble in only one phase. Such a
system is
.
MX; MY; a .. MX; b
8:9:27
where the anion Y is present in phase a only. The Galvani potential dierence in
this system is called the Donnan potential dierence since its characteristics are
similar to those observed at membranes which are permeable to some ions but not
to others. The Galvani potential dierence is determined by both the distribution
of the cation M and the anion X, and is given by equation (8.9.13). The
condition of electroneutrality in phase a gives
433
caM caX caY
8:9:28
cbM cbX
8:9:29
whereas in phase b,
Thus, for this system equation (8.9.13) becomes

"
#
a

ab a b
RT
KM
yM yX
RT
cX caY
ln
ln

f
a b
2F
2F
caX
KXab ybM yaX
8:9:30
where yxi is the activity coecient of ion i in solvent x. Under conditions that
caY >> caX , the Donnan or Galvani potential dierence is linear in the logarithm of
the concentration of the anion X in the a phase. This system also has analytical
applications in which a b f is used to determine the concentration of X in the a
phase.
The liquid | liquid interface is ideally polarizable when the electrolyte in phase
a is insoluble in b, and that in b, insoluble in a. Such a system may be designated
as
.
M1 X1 ; a .. M2 X2 ; b
8:9:31
Under these circumstances the distribution coecient for M1X1 is zero, and that
for M2X2, innity. It follows that the Galvani potential dierence is not dened
for this interface. However, the thermodynamics of the interface can be derived on
the basis of the Gibbs adsorption isotherm.
The discussion in this section has focused on liquid | liquid interfaces for
systems in which electrolytes are soluble in both liquid phases. Of course, there
are many examples of systems in which electrolytes are not present in at least one
phase. Examples include some of the systems listed in table 8.10 in which the
organic phase has such a low permittivity that electrolytes are essentially insoluble. One interesting subclassication of such systems includes those in which an
organic phase may be spread on a polar liquid such as water to form a surface lm
with the thickness of a monolayer. These systems are considered in more detail in
the following section.
8.10 Surface Films on Liquids

When an insoluble liquid a is placed at the liquid | air interface of another liquid
b, the rst liquid can spread out as a thin lm which becomes monomolecular if
the area of liquid b is suciently large. Everyone has experienced the colors which
are seen when oil is spread on water in a muddy puddle on the road. The colors
result from the interference of the light rays reected from the oil | air interface
with those reected at the water | oil interface. The interference occurs because the
oil lm is very thin. The properties of thin lms are especially interesting when
their thickness corresponds to one molecule.
The rst question regarding thin lms is how they spread on the surface of the
substrate liquid b (see g. 8.17). In order for this to occur, the Gibbs energy
434
Fig. 8.17 Schematic diagram illustrating

the interface present when a small amount
of liquid a is present on substrate liquid b
as a lens at the air interface.
change associated with the spreading process must be negative. As liquid a

spreads on substrate b, the area of a exposed to air increases and that of b
decreases. At the same time, the area of the a | b interface increases. Since the
interfacial tension is the incremental work done to increase the area of a given
interface, the net work associated with spreading is
Ws gab ga gb
8:10:1
where gab is the interfacial tension between the two liquids. If Ws is negative then
spreading is a spontaneous process.
EXAMPLE
Examine the spreading of benzene on water, given that the interfacial tension
of pure benzene is 28.18 mJ m2, using the data in tables 8.1 and 8.10.
From table 8.1, the interfacial tension of pure water is 71.81 mJ m2, and
from table 8.10, the interfacial tension between water and benzene is
34.4 mJ m2. Thus, the work of spreading is
Ws 34:4 28:2 62:2 0:4 mJ m2
8:10:2
As a result, a small amount of benzene placed on the surface of water rapidly

spreads. However, this simple calculation is na ve in that it ignores the eect of
benzene on the interfacial tension of pure water. When the benzene and water
are in contact, a very small amount of benzene dissolves in the water to yield a
saturated solution. Since the benzene is hydrophobic it accumulates at the
water | air interface and causes a signicant lowering of the interfacial tension
of the aqueous phase. After equilibrium is reached, gb is 62.2 mJ m2. The
work of spreading of benzene on water saturated with benzene is thus
Ws 34:4 28:2 62:2 0:4 mJ m2
8:10:3
Since Ws is positive, benzene does not spread on the saturated solution.

As a result, when benzene is introduced at the water | air interface it
initially spreads on the water. However, the water soon becomes saturated
with benzene. When the interfacial tension of pure water falls to its equilibrium value for the solution, the benzene layer on the surface retracts to a
lens as shown in g. 8.17.
The phenomenon described with benzene on water occurs for most organic
liquids of lower molecular weight. However, most experimental work with
unimolecular lms has involved molecules of higher molecular weight involving
at least ten carbon atoms. These substances are often solids at room temperature
and are introduced to the water surface as a solution in a volatile organic solvent.
For example, a unimolecular lm of octadecanoic acid can be spread on water by
435
introducing it as a solution in pentane. The pentane evaporates rapidly, leaving

the C18 acid on the water surface as a unimolecular lm.
Methods for studying unimolecular lms usually involve the Langmuir lm
balance (g. 8.18), which allows one to study the surface pressure due to the lm,
, as a function of the area available, A. In this experiment the trough is lled
with water and the surfactant is introduced into the section of the trough with the
larger area. The area available to the surfactant lm is varied by moving the sweep
barrier. At the same time, the surface pressure is determined by the torsion wire
connected to the oating barrier on the right-hand side of the trough. The torsion
wire incorporates a pressure transducer which allows very precise determination
of the surface pressure . In order to obtain data of high quality the purity of the
water and surfactant must be carefully controlled. In addition, temperature must
be kept constant, since the relationship between and A is sensitive to this
parameter.
Typical surface pressure against area data obtained for monolayers of a longchain carboxylic acid, namely, pentadecanoic acid, spread on acidied water, are
shown in g. 8.19. Considerable change in the characteristics of the A isotherms is seen with change in temperature. Extensive study of monomolecular
lms has led to the denition of several states for the two-dimensional system [G1,
20]. A gaseous lm is one for which the A product is constant for xed temperature and corresponds to a situation where the area occupied per molecule in
the lm is very large, so that the molecules do not interact with one another. It is
clear that none of the isotherms in g. 8.19 corresponds to a gaseous state.
However, such a state is possible at a temperature higher than those studied. A
liquid expanded state, designated L1, is one in which the molecules can interact
but are not well organized. This state is present for the pentadecanoic acid lm at
35.2 C for areas greater than 0.30 nm2 molec1 (see g. 8.19). The A product is
not constant in this region but instead exhibits the properties of a two-dimen-
Fig. 8.18 Sketch of a Langmuir

lm balance. The trough (1) is
completely lled with water or an
aqueous solution; the lm is
conned between a movable
barrier (2) and a xed oating
barrier (3) which is attached to
the posts of the pressure
transducer system by gold foil
barriers (4); surface pressure is
detected by movement of mirror
(5) attached between a torsion
wire (6) and the oating barrier
(3); the calibration arm (7) and
torsion wire control (8) are used
to relate mirror position to
surface pressure.
436
Fig. 8.19 Surface pressure against area isotherms for pentadecanoic acid monolayers on
acidied water (pH 2). (From reference G1, with permission.)
sional non-ideal gas. In some cases the isothermal data can be tted to a twodimensional version of the van der Waals equation:

a
8:10:4
vw2 A bvw RT
A
Here aVW and bVW are two-dimensional van der Waals constants describing the
isotherm. Below 0.3 nm2 molec1, there is a sharp change in the slope of the A
plot corresponding to transition to an intermediate lm. In this state the monomers begin to assemble with the polar carboxylic acid group in the aqueous phase
and the hydrophobic hydrocarbon tail pointing into the air phase. Further compression of the lm leads to expulsion of water molecules between the monomers
and their alignment as a liquid condensed (L2) lm (see g. 8.20). The A
isotherm shown at 17.9 C is characteristic of an L2 lm showing a discontinuity
Fig. 8.20 Schematic diagram showing

the structural changes in a
unimolecular lm going from the
liquid expanded state (L1) through the
intermediate state (I) to the liquid
condensed state (L2). The hydrophilic
head group is designated by the circle
at the end of a long hydrophobic
hydrocarbon chain.
437
in its two-dimensional compressibility for the range of areas studied. At high

surface pressures the liquid condensed state becomes a solid-state lm for
which the compressibility is negligibly small (dA=d ! 0).
Studies of monolayer lms have included alcohols and carboxylic acids with
hydrocarbon chains of varying length in the range C10C26 [20, 21]. In the liquid
condensed and solid states the A isotherms have similar shapes at high surface
pressures. These data show that the monolayer in the self-assembled state occupies an area of 0.205 nm2 per monomer. This result is independent of chain length
and provides evidence that the monolayer consists of a close-packed structure
with all molecular units oriented with their hydrocarbon chains perpendicular to
the interface. Under these circumstances, strong attractive van der Waals forces
are present between the hydrocarbon tails. As a result, formation of the solid-state
lm can be irreversible, so that the lm does not break up when the surface
pressure is decreased.
Unimolecular lms provide a rich area for study in interfacial chemistry which
has only been briey introduced here. Other investigations include the kinetics of
lm formation, the viscosity of the monolayer, and measurements of surface
potential changes [G1]. Unimolecular lms in condensed format may be removed
from the surface of the liquid on which they are formed using a glass plate or
other suitable solid substrate. The hydrophilic ends of the molecules in the lm
remain in contact with the solid surface, and the hydrophobic ends, with the air.
This type of unimolecular lm, which is called a LangmuirBlodgett lm, is also
the subject of considerable research interest.
8.11 Spectroscopy at Liquid Interfaces

Spectroscopic studies of liquid interfaces provide important information about
the composition and structure of the interfacial region. Early work was mainly
carried out at the solid | liquid interface and involved techniques such as neutron
and X-ray diraction, and reection FTIR spectroscopy. More recently, powerful
techniques have been developed to study the liquid | liquid and liquid | gas interfaces. These studies are especially important because of their relevance to biological systems such as cell membranes. The techniques described here are secondharmonic generation (SHG) and vibrational sum frequency spectroscopy (VSFS).
They are both second-order non-linear optical techniques which are specic to the
interfacial region. Since the second-order eects involve signals of low intensity,
they rely on high-power lasers.
When electromagnetic radiation interacts with a medium it induces a polarization P which is proportional to the electrical eld E due to the radiation:
P w1 E
8:11:1
(1)
The proportionality constant w is called the linear electrical susceptibility. It is

related in a simple way to the permittivity of the medium (see equation (4.5.3)).
Thus,
w1 e0 e 1
8:11:2
438
where e is the relative permittivity of the medium and e0, the permittivity of free
space. In general, both w(1) and e are functions of frequency. In the gas phase, the
susceptibility is related to the polarizability aP of the molecules in the gas.
Equation (8.11.1) is a basic relationship in linear optics and the starting point
for deriving the fundamental equations of spectroscopy.
When a molecule in the gas phase experiences a very large electromagnetic
eld, the resulting polarization in no longer a linear function of the eld. In
this case, the polarization is given by
P aP E bP E2 gP E3 . . .
8:11:3
The coecient bP is called the second-order non-linear polarizability or the hyperpolarizability. The corresponding equation in a condensed medium is
P w1 E w2 E2 w3 E3 . . .
(2)
8:11:4
(3)
w and w are the non-linear contributions to the susceptibility. Non-linear

optical eects become possible with high-powered lasers. SHG and VSFS depend
on w(2), the second-order non-linear susceptibility of the medium at an interface.
More details about these two methods are given below.
A. Second-Harmonic Generation
Optical second-harmonic generation involves the non-linear conversion of two
photons of frequency to a single photon of frequency 2. This process requires
a non-centrosymmetric medium, one example being the interface between
two centrosymmetric media. Since only a few molecular layers are involved in
the symmetry breaking at the interface, the SHG process is a highly
interface-selective optical probe.
SHG arises from polarization of the molecules at the interface and depends on
their second-order non-linear susceptibility. Thus, if the amplitude of the electrical eld due to the incident light is E(), the polarization responsible for the
SHG is given by
P2 2n w2 E2 n
8:11:5
The value of w(2) is related to the molecular hyperpolarizability and depends on

the electronic distribution in a given molecule. Only those molecules which have
an inherent asymmetric electron density distribution, or an asymmetric distribution induced by adsorption, are capable of yielding an interfacial second-harmonic response. The value of w(2) may be calculated for both molecules and metals
given their electronic properties [22]. In the case of molecules it depends on the
number of electrons in the system, and therefore, is proportional to their surface
concentration. For metals, w(2) is directly related to the surface concentration of
free electrons. The intensity of the second-harmonic signal is proportional to the
square of w(2).
SHG experiments may also be used to determine molecular orientation at
interfaces. By determining the polarity of the SH signal with respect to that of
the incident light one may determine the independent components of w(2). In
general, w(2) is a third-rank tensor with 27 elements. When the composition of
439
the monolayer is the same in all directions away from the normal to the interface
2
2
(z-direction), only three of these are unique, namely, w2
xzx , wzxx , and
zzz These
may be related to the corresponding elements of the hyperpolarizability tensor,
and the orientation of the molecule determined [22, 23].
A typical experimental conguration for SHG studies at a liquid | liquid interface is shown schematically in g. 8.21. A nanosecond or picosecond pulsed laser is
the source of the input signal. It is polarized and ltered to ensure that any extraneous second-harmonic light is removed. Typical power densities at the interfaces
are in the range 105108 W cm2. Conversion eciencies are normally very small,
and on the order of 1 photon in 1014. The light emitted at 2 is analyzed with
respect to polarization in order to obtain information about the orientation of the
molecules whose non-linear susceptibility give rise to the SHG signal.
Early studies were carried out at the liquid | gas interface [22, 23]. Castro et al.
[24] studied the adsorption of p-propyl-phenol from aqueous solutions at the air
interface as a function of phenol concentration in the bulk. They showed that the
square root of the second-harmonic intensity plotted against bulk phenol concentration followed a Langmuir isotherm with a standard Gibbs energy of adsorption
equal to 24.3 kJ mol1. Similar results were obtained for other alkylphenols and
alkylanilines. In other work with phenols, the orientation of phenol at the water |
2
air interface was determined by studying the phase of the
xzx
component of the
susceptibility. As expected, the OH was oriented toward the water phase [25] so
that it could participate in the hydrogen-bonded structure of water. The same
conclusion was reached for p-bromophenol and p-nitrophenol.
SHG has also been used to study adsorption at the liquid | liquid interface
[22, 23]. 2(-N-octadecylamino)naphthalene-6-sulfonate (ONS) is an anionic surfacant which is adsorbed at the water | 1,2-dichloroethane (DCE) interface [26]. The
long alkyl chain prefers to be in the relatively non-polar DCE medium with the
charged sulfonate group in the water at the interface. If the interface is polarized
in an electrochemical cell, the change in the ONS surface excess with charge at the
polarizable DCE | water interface can be followed using the SHG signal due to
the surfactant.
The other important area where SHG has been applied is to the study of
adsorption at the metal | gas and metal | electrolyte solution interfaces [22, 23].
Fig. 8.21 Schematic diagram of the experimental apparatus used to study secondharmonic generation at a liquid | liquid interface. (From reference 23, with permission.)
440
The SHG at metal surfaces is almost always dominated by the nonlinear polarizability of the metal electron cloud at the interface. This technique is especially
important in electrochemistry and has been used to study the adsorption of ions,
molecules, and metal atoms at electrodes.
B. Vibrational Sum Frequency Spectroscopy
Sum frequency spectroscopy (SFS) provides a powerful tool for obtaining the
vibrational spectrum of the molecules at the interface between two condensed
phases or one condensed phase and a gas. It is a non-linear optical technique in
which a photon of light with a frequency 1 in the visible region interacts with a
photon of light with a frequency 2 in the infrared region to give a photon with
frequency 1 2 [27, 28]. By scanning the infrared light source over a range of
frequencies i2, the vibrational spectrum is obtained as 1 i2. The relationships between the intensity of the incoming and outgoing optical signals are similar
to those for second-harmonic generation. Thus, if P(2)(1 2) is the second-order
non-linear polarization due to the elds of the incoming light, one may write
P1 n1 n2 w2 n1 n2 En1 En2
8:11:6
where w (1 2) is the non-linear susceptibility at the frequency 1 2, and

E(i), the eld due to radiation at frequency i. The square root of the VSFS
1=2
intensity, ISF
, is proportional to the number density of molecules at the interface.
When the frequency of the infrared beam is equal to the vibrational frequency of a
molecule in the interface, a very large resonance enhancement of the SF signal is
possible. Only vibrational modes that are both IR and Raman active are SF
active. On the other hand, molecules or vibrational modes that possess an inversion center are not SF active. The resulting infrared spectrum of the interface
provides information about what is at the interface. It can also be used to determine molecular orientation when the incoming and outgoing optical signals are
polarized.
In VSF spectroscopy, two pulsed laser beams, one of xed frequency in the
visible and another of variable frequency in the IR region, overlap at the interface.
Since the fraction of incident photons undergoing the SF process is very small,
pulsed lasers are used in order to obtain detectable SF signals. Short pulses in the
picosecond or femtosecond range are optimum. Details regarding operation of the
lasers for dierent optical geometries have been given by Richmond [28]. Typical
geometries for VSFS in the reection mode at the liquid | liquid and liquid | air
interfaces are shown in g. 8.22.
VSFS has been used to carry out studies of water structure at the aqueous
solution | air and water | organic liquid interfaces [27, 28]. Hydrogen bonding
plays an important role in determining water structure at the air interface and is
responsible for the positive dipole potential found at this interface (see section
8.7.A). The VSF spectrum of water at the air interface shows two features, a fairly
sharp band centered at 3700 cm1 and a broad band appearing between 3000 and
3500 cm1 (g. 8.23). The latter feature is also seen in bulk water and reects the
eects of hydrogen bonding on the symmetrical and asymmetrical stretching
modes of water. The high-energy band is due to free OH bonds which do not
(2)
441
Fig. 8.22 Experimental

geometries used in the reection
mode for VSFS at the liquid |
liquid and liquid | air interfaces.
The critical angle for the SF
signal is yc.
participate in the hydrogen-bonded structure. These results conrm that the water
molecules at the surface are strongly hydrogen-bonded so that there is a net excess
of water dipoles with their negative ends pointing to the gas phase. SHG studies of
the water | air interface [23] also led to the conclusion that the net orientation of
water is with the positive end of the molecular dipole pointing to the interior of
the solution at an angle of 70 from a line perpendicular to the interface.
Signicant changes in the VSF spectra of water are seen at the water | carbon
tetrachloride interface (g. 8.23). The band due to freeOH bonds is shifted to
lower energies by 40 cm1. This suggests attractive interaction between the free
OH oscillators and the organic phase. In addition, the spectral characteristics due
to the hydrogen-bonded water molecules are signicantly modied. The intensity
seen at the low-frequency region at the air interface has largely disappeared at the
organic interface, suggesting that hydrogen bonding at the interface is much
weaker in the presence of the organic phase. A detailed analysis of these spectra
with resolution into component bands has been given by Richmond [27, 28].
Stanners et al. [29] used VSFS to study the structure of the alcohols at the
liquid | air interface. For low-molecular-mass alcohols, namely, methanol to
butanol, the alcohol molecule is oriented with the hydroxy group in the liquid
and the alkyl group pointing into the gas phase. This orientation is attributed to
the eects of hydrogen bonding involving the hydroxy group and other alcohol
molecules in the bulk. The infrared data involved both the OH and CH stretch-
442
Fig. 8.23 Vibrational sum frequency spectra of water at the air | water and carbon tetrachloride | water interfaces [28]. The spectrum for the air | water interface has been shifted
vertically by 0.6 units for the sake of clarity.
ing modes. These results are consistent with the negative surface potential estimated for the lower alcohols (see table 8.5).
Richmond [27, 28] has also used VSFS to study surfactants at the water | air
and water | organic interfaces. These studies have provided information not only
about the extent of adsorption but also about changes in the orientation of the
adsorbate molecules with coverage. Other experimental work has included studies
of solvent molecule orientation at the electrode | solution interface [30].
VSFS is obviously a powerful technique for obtaining specic information
about the composition and structure of molecules at interfaces. Because it provides the vibrational spectrum, it is more advantageous than SHG.

The discussion in this chapter has focused on the properties of liquids at interfaces. A related area of contemporary research is the study of solid | gas interface.
The solid surface is quite dierent in that atomic or molecular components of a
solid are relatively motionless compared to those of liquid. For this reason it is
easier to dene a plane associated with a well-dened solid surface. The approach
to studying adsorption on solids has been more molecular with the development
of sophisticated statistical mechanical models. On the other hand, the study of
liquid | gas and liquid | liquid interfaces has been much more macroscopic in
approach with a rm connection to classical thermodynamics. As the understanding of liquids has improved at the molecular level using contemporary statistical
mechanical tools, these methods are being applied now to uids at interfaces.
Emphasis has been placed on the electrical aspects of interfaces in this chapter.
This is especially important in examining the properties of electrolyte solutions at
443
interfaces and leads to the experimental determination of ionic work functions.

The concepts of the inner potential and surface potential are important in dening
a single electrode potential in an electrochemical cell. Although these quantities
cannot be measured experimentally they are important in developing a connection
with fundamental theoretical treatments of the transfer of charge from and into a
condensed phase.
A great variety of diraction and spectroscopic techniques have been developed to study the interfacial region. Most of these have been applied at solid
interfaces with the probing electromagnetic radiation approaching the interface
from the gas or liquid phase. Now SHG and VSFS provide powerful techniques
for obtaining molecular level information at the liquid | liquid and liquid | gas
interface.
The concepts developed in this chapter are very important in understanding the
properties of interfaces in electrochemical cells and in living organisms. In the
following chapters these subjects are developed in more detail, rst for the case of
electrochemical equilibria, and then for the case of electrostatic equilibria.
General References
G1. Adamson, A. W. Physical Chemistry of Surfaces, 5th ed.; Wiley-Interscience: New
York, 1990.
G2. Damaskin, B. B.; Petrii, O. A. Introduction to Electrochemical Kinetics, 2nd ed.;
Vysshaja Shkola: Moscow, 1983; Part I (in Russian).
G3. Trasatti, S. In Comprehensive Treatise of Electrochemistry; Bockris, J. O M.; Conway,
B. E., Yeager, E., eds.; Plenum Press: New York, 1980; Volume 1, Chapter 2.
G4. Parsons, R. In Modern Aspects of Electrochemistry; Bockris, J. OM., ed.;
Butterworths: London, 1954; Vol. 1, Chapter 3.
G5. Somorjai, G. A. Introduction to Surface Chemistry and Catalysis; Wiley-Interscience:
New York, 1994.
References
1. Schirin, D. J. J. Electroanal. Chem. 1969, 23, 168.
2. Riddick, J. A.; Bunger, W. B.; Sakano, T. K. Organic Solvents, 4th ed.; Wiley: New
York, 1986.
3. Fomenko, V. S. The Emission Properties of Materials (in Russian); Naukova Dumka:
Kiev, 1981.
4. Appelbaum, J. A.; Hamann, D. R. Rev. Mod. Phys. 1976, 48, 479.
5. Trasatti, S. In Advances in Electrochemistry and Electrochemical Engineering;
Gerischer, H., Tobias, C. W., eds.; Wiley: New York, 1967; Vol. X, Chapter 4.
6. Randles, J. E. B. Adv. Electrochem. Electrochem. Eng. 1963, 3, 1.
7. Posner, A. M.; Anderson, J. R.; Alexander, A. E. J. Colloid Sci. 1952, 7, 623.
8. Levine, S. J. Colloid Interface Sci. 1971, 37, 619.
9. Kenrick, F. B. Z. Phys. Chem. 1896, 19, 625.
10. Randles, J. E. B. Phys. Chem. Liquids 1977, 7, 107.
11. Frumkin, A. N.; Damaskin, B. B. Pure Appl. Chem. 1967, 15, 263.
12. Parsons, R.; Rubin, B. T. J. Chem. Soc., Faraday Trans. 1, 1974, 70, 1636.
13. Koczorowski, Z.; Zagorska, I.; Kalinska, A. Electrochim. Acta. 1989, 34, 1857.
14. Farrell, J. R.; McTigue, P. J. Electroanal. Chem. 1982, 139, 37.
15. Bard, A. J.; Parsons, R.; Jordan, J. Standard Potentials in Aqueous Solution; Marcel
Dekker: New York, 1985.
444
16. Case, B.; Parsons, R. Trans. Faraday Soc. 1967, 63, 1224.
17. Girault H. H.; Schirin, D. J. In Electroanalytical Chemistry; Bard, A. J., ed., 15, 1,
1989.
18. Vanysek, P. Lect. Notes Chem. 1985, 39, 1.
19. Koryta, J.; Stulik, K. Ion Selective Electrodes, 2nd ed.; Cambridge University Press:
Cambridge, 1983; Chapter 2.
20. Harkins, W. D. The Physical Chemistry of Surface Films; Reinhold Publishing: New
York, 1952.
21. Boyd, G. E. J. Phys. Chem. 1958, 62, 536.
22. Corn R. M.; Higgins, D. A. Chem. Rev. 1994, 94, 107.
23. Eisenthal, K. B. Chem. Rev. 1996, 96, 1343.
24. Castro, A.; Bhattacharyya, K.; Eisenthal, K. B. J. Chem. Phys. 1991, 95, 1310.
25. Kemnitz, K.; Bhattacharyya, K.; Hicks, J. M.; Pinto, G. R.; Eisenthal, K. B.; Heinz,
T. F. Chem. Phys. Lett. 1986, 131, 285.
26. Higgins, D. A.; Corn, R. M. J. Phys. Chem. 1993, 97, 489.
27. Watry, M. R.; Brown, M. G.; Richmond, G. L. Appl. Spectrosc. 2001, 55, 321A.
28. Richmond, G. L. Annu. Rev. Phys. Chem. 2001, 52, 357.
29. Stanners, C. D.; Du, Q.; Chin, R. P.; Cremer, P.; Somorjai, G. A.; Shen, Y.-R. Chem.
Phys. Lett. 1995, 232, 407.
30. Baldelli, S.; Mailhot, G.; Ross, P. N.; Somorjai, G. A. J. Am. Chem. Soc. 2001, 123,
7697.
Problems
1. Given the following data for the interfacial tension at the water | air interface,
determine the entropy of formation of this interface at 20 C.
T / C
g / mJ m2
10
15
20
25
30
74.22
73.49
72.75
71.97
71.18
2. The following data have been obtained for the interfacial tension of benzene
trimethylpentane (TMP) mixtures at 30 C:
Mole Fraction
of TMP
xTMP
Interfacial
Tension
g=mJ m2
0.000
0.186
0.274
0.378
0.483
0.583
0.645
0.794
1.0
27.53
23.40
22.47
21.21
20.29
19.70
19.32
18.74
17.89
445
A simple equation for the surface tension of mixtures is

g g1;0 x1 g2;0 x2 bx1 x2
where g1,0 is the surface tension of pure component 1, g2,0, that of component
2, and x1 and x2 , the corresponding mole fractions. Determine the best value
of b for the methanolwater interfacial tension data. How well does the
model t the data?
3. Using the data given in problem 2 and assuming an ideal solution, estimate
the relative surface excesses of benzene and TMP for mole fractions of 0.2,
0.4, 0.6, and 0.8. Plot the results against the mole fraction of benzene.
4. The following data for the surface tension of aqueous n-hexyl alcohol solutions were obtained at 12 C.
Alcohol Concentration
103 M
Surface Pressure
=mJ m2
0.62
0.81
1.25
1.72
2.50
3.43
4.90
6.86
9.80
2.3
2.5
3.9
5.7
7.9
9.4
13.4
16.3
19.4
Use a numerical method from appendix C to determine the relative surface

excess of n-hexyl alcohol at c 1:25; 2:5, and 5 103 M. Assume that
Henrys law holds for the solute in the given concentration range. Express
the results in units of mol m2 and molec nm2.
5. Consider a mercury sphere, radius r 0:01 cm, carrying a charge density of
10 pC cm2 in vacuo. The potential experienced by unit test charge e0 at a
distance x from the metal surface is

1
q
e0 r

c
4e0 r x 2x2 2xr
where q is the charge on the sphere and e0 1:6 1019 C. What is the origin
of the second term on the right-hand side of the above equation? Calculate c
for decadic increments in x in the range 1071 cm. Over what range of x is c
equal to the outer potential of the mercury drop?
6. On the basis of the MSA the chemical potential of a monovalent cation is

N e2
1
1
mi 0 0 1
es ri s
8e0
446
where ri is the ions radius and s, the MSA parameter for cation in water.
Plot the work functions for the alkali metal cations from table 8.6 against ri
s assuming the s is 49 pm and using the ShannonPrewitt radii from table
3.1. Estimate the surface potential of water from the intercept of best straight
line through the data. Does the slope of the plot have the value expected from
the MSA?
7. Estimate the Galvani potential dierence for NaCl at the water | 1,1- dichloroethane interface given that the standard Gibbs energy of transfer of Na is
29 kJ mol1 and that for Cl, 58 kJ mol1.
8. An insoluble lipopolysaccharide forms an insoluble monolayer on water. Use
the following data to estimate the molecular mass of this compound assuming
ideal behavior:
c=mg m2
=mN m1
0.06
0.09
0.10
0.14
0.17
0.23
10.6
16.4
20.4
25.9
34.3
50.0
9. Two capillary tubes with inside radii of 0.600 and 0.400 mm are inserted in a
liquid with a density of 0.901 g mL1. The dierence in the capillary rises in
the two tubes is 1.00 cm. Estimate the surface tension of the liquid assuming
zero contact angle and an air density of 0.001 g mL1.
Charge Transfer Equilibria

at Interfaces
Walther Nernst was born in Briesen, Prussia

(now in Poland) in 1864. He grew up in
Graudentz and went on to study physics and
mathematics at the Universities of Zurich,
Berlin, and Graz. He completed his doctoral
thesis in 1887 at the University of Wurzburg.
He then joined Professor Ostwald at the
University of Leipzig, where he started his
work in physical chemistry. At this time
Arrhenius and vant Ho were also working in
the Ostwald group. In 1894, he moved to
Walther Hermann Nernst
Gottingen where he founded the Institute of
Physical Chemistry and Electrochemistry. His nal appointment was as
Professor of Chemistry at the University of Berlin from 1905. Nernst
contributed in a signicant way to the development of the theory of electrolyte
solutions and electrochemical cells. His name is connected to the well-known
equation relating the electrode potential to the activities of the components
involved in the electrode reaction. His scientic work was not limited to
electrochemistry and in the latter part of his career he focused on the
properties of matter at very low temperatures. This resulted in an early version
of the Third Law of Thermodynamics. He was awarded the Nobel Prize in
Chemistry for his work in this area in 1920. After retirement in the 1930s he
lived in his country home near Berlin, where he died in 1941.
9.1 Electrochemical Equilibria Occur at a Wide Variety

of Interfaces
Processes in which charge is transferred from one phase to another at an interface
make up an important class of interfacial reactions. Well-known examples are the
reactions which occur at the electrodes of an electrochemical cell. These are
electron transfer reactions, oxidation taking place at one electrode and reduction
at the other. The early study of electrochemical cells provided valuable thermodynamic information about the redox processes occurring in them. When an
electrochemical cell is a source of energy, for example, a battery, chemical energy
is converted to electrical energy. When electrical energy is driven into an electro447
448
chemical cell from an external source, electrode reactions producing products of

commercial interest are possible. Thus the general subject discussed here is of
considerable practical importance.
Another important class of interfacial charge transfer processes occurs at the
membrane | solution interface. Some solute species can move into the membrane
phase, whereas others cannot. When ions are involved in membrane selectivity, a
potential drop is established at the interface. Ion transfer processes at membranes
are extremely important in living organisms and form the basis for the functioning
of the nervous system. Membranes are also involved in ion selective electrodes
such as the ubiquitous pH electrode. These electrodes are often used in modern
analytical techniques based on potentiometry.
In the present chapter, the relationship between the electrode potential and the
activity of the solution components in the cell is examined in detail. The connection between the Galvani potential dierence at the electrode solution interface
and the electrode potential on the standard redox scale is discussed. This leads to
an examination of the extrathermodynamic assumption which allows one to
dene an absolute electrode potential. Ion transfer processes at the membrane |
solution interface are then examined. Diusion potentials within the membrane
and the Donnan potentials at the interface are illustrated for both liquid and solid
state membranes. Specic ion electrodes are described, and their various modes of
sensing ion activities in an analyte solution discussed. The structure and type of
membrane used are considered with respect to its selectivity to a particular ion
over other ions. At the end of the chapter, emphasis is placed on the denition of
pH and its measurement using the glass electrode.
9.2 Electrochemical Cells

The classical description of an electrochemical cell is that it consists of two electrodes immersed in an electrolyte solution. The simplest type of electrode is a
metal, so that a simple cell involves two metal | solution interfaces. However,
experimental study of the electrochemical cell always involves other interfaces
which are introduced when the electrodes are connected to the leads of the potential measuring device. When the cell is a source of energy, an electrical current
ows in an external circuit as a result of a chemical reaction at each electrode.
There are important conventions regarding the presentation of information about
these systems. These are outlined in detail below.
Consider the following cell as an example:
Cu j Pt j H2 1 bar j HCl; H2 O j AgCl j Ag j Cu0
9:2:1
This is called the diagram of the cell and includes all of the components required
to make experimental measurements of its properties. The single vertical bar
represents a phase boundary. The two electrodes in the cell are the hydrogen
electrode and the silver | silver chloride, which are illustrated in g. 9.1. The
electrolyte solution is 1 M HCl in water. Each electrode is connected to a copper
wire which is used as a lead to connect to the measuring device, for example, a
voltmeter with a very high input impedance. Since the wires at the ends of the cell
CHARGE TRANSFER EQUILIBRIA AT INTERFACES
449
Fig. 9.1 Sketches of half-cells corresponding to (a) a metal | metal ion system, (b) the
hydrogen electrode, and (c) the silver | silver chloride electrode.
have the same chemical composition, the Galvani potential dierence between
them can be measured.
.
Two other symbols are used in cell diagrams. They are a dotted vertical bar (..)
.. ..
and a double dotted vertical bar (. .). The rst represents a liquid junction between
miscible liquids, and the second, a double liquid junction in which the liquid
junction potential has been eliminated to a good approximation, for example,
by using a salt bridge. Liquid junctions are formed in practice using fritted
glass disks and gels such as agar. They permit ion transport but greatly reduce
mechanical mixing of the two solutions. The properties of liquid junctions are
examined later in this chapter.
The Galvani potential dierence across the cell is called the EMF (electromotive force) and is given the symbol E. Furthermore, it is always measured as the
potential of the right-hand electrode with respect to the left-hand one. Thus,
0
E Cu Cu0 f fCu fCu
9:2:2
The term electromotive force is no longer used but the abbreviation EMF is still
considered a valid way of referring to E. If the terminals of the cell are connected
by a wire a spontaneous current ows with electrons moving from the left-hand
terminal to the right. This is accompanied by a chemical reaction in the cell. For
cell (9.2.1), the reaction is
H2 AgCl ! 2 HCl 2Ag
9:2:3
When a spontaneous current ows the potential dierence between the terminals
is no longer equal to E.
When the potential dierence is measured under conditions of zero current in
the external circuit, it has important thermodynamic signicance. Specically, the
reactions at each electrode are at equilibrium. For cell (9.2.1), the equilibria are
( 2H2
2H 2e +
(
AgCl e + Ag Cl
9:2:4
9:2:5
450
By convention, these equilibria are always written with the electrons on the lefthand side. In order to obtain the cell reaction (9.2.3), the reaction at the left-hand
electrode is subtracted from that at the right-hand electrode using multiplying
factors such that the net number of electrons in the overall reaction is zero. It
should be noted that the Gibbs energies of the electrons in each electrode are not
equal when the cell has an EMF. This means that the cell is capable of doing
electrical work when current from the cell ows in an external circuit. By measuring E under conditions of zero current, one can calculate the reversible electrical
work which is available from the cell. In addition, the reversible work is equal to
the Gibbs energy change associated with reaction (9.2.3). Thus, one may write
E
G
nF
9:2:6
where n is the number of electrons involved in the cell reaction. For reaction
(9.2.3), n is equal to two.
The classical experimental method for measuring E involves a Poggendorf
compensation potentiometer. In this experiment the voltage due to the electrochemical cell is exactly balanced by an opposing voltage source in the potentiometer (g. 9.2). Under conditions of balance, no current ows from the cell, and
the voltage can be related to the reversible electrical work available from the
system. If the opposing voltage is slightly smaller, a current ows from the electrochemical cell, which then acts as an energy source. On the other hand, if the
opposing voltage is slightly greater, current is forced into the cell so that the
electrode reactions take place in a direction opposite to the spontaneous direction.
In a modern measurement of E, an electronic voltmeter with a very high input
impedance is used. Nevertheless, the classical technique with the Poggendorf
method is very helpful in illustrating the thermodynamic signicance of the
EMF because it involves an exact balance between the electrochemical source
and a variable d.c. source in the potentiometer device, thereby demonstrating
that the cell properties are being measured under reversible conditions.
The value of G and thus E for reaction (9.2.3) is determined by subtracting
the chemical potentials of the reactants from those for the products:

Poggendorf compensation
potentiometer. The contact position C
on slidewire AB is moved until no
current ows in galvanometer G.
Calibration of AB, for which position
is related to voltage, permits
determination of the EMF of cell X.
G 2mHCl 2mAg mH2 2mAgCl
451
9:2:7
Each chemical potential for a component in solution can be expressed in terms of

its standard value plus a term in its activity; for gas phase components which are
not at high pressure, the activity is equal to the partial pressure of the gas. Thus,
expanding these terms in equation (9.2.7) one obtains
G 2mHCl 2RT ln aHCl 2mAg mH2 RT ln PH2 2mAgCl
9:2:8
This may also be written as

G G 2RT ln
aHCl
!
9:2:9
P1=2
H2
where G , the standard Gibbs energy change, is dened as

G 2mHCl 2mAg mH2 2mAgCl
9:2:10
In the case of pure solids such as Ag and AgCl the chemical potential is identical
to the standard chemical potential at 25 C and 1 bar pressure. For solutions, the
standard state of the solute is unit activity at the same temperature and pressure.
In the case of electrolytes as solutes, the activity ae is dened on the concentration
(molarity) scale, and the standard state is the hypothetical ideal state of unit
molarity for which the activity coecient ye is unity. Under these circumstances,
the activity of the solvent, which does not appear explicitly in equation (9.2.9), is
also unity to a good approximation when the solvent is water. For gases the
standard state is a pressure of 1 bar (105 Pa) at 25 C. In the older literature the
standard pressure was 1 atm (101,325 Pa). In data compilations appearing after
1982, the standard state of 1 bar and 25 C is always used for gases [G3].
By combining equations (9.2.6) and (9.2.9), one obtains the Nernst equation
for cell (9.2.1), that is,
!
!
RT
aHCl
RT
aH aCl

ln 1=2 E
ln
9:2:11
EE
F
F
pH
P1=2
H
2
where
E
G
2F
9:2:12
In writing these equations it is recognized that reaction (9.2.3) involves two

electrons. The value of G from the relevant thermodynamic data is 42.84 kJ
mol1. The corresponding standard EMF E is 0.222 V.
Nernst equations may also be written for the equilibria (9.2.4) and (9.2.5).
These equations are written in terms of the EMF of a cell containing the electrode
in question and a standard hydrogen electode (SHE). The latter is a dened point
of reference which is obtained when a hydrogen electrode is operated with a H2
pressure of 1 bar and unit H activity in the electrolyte solution. On this basis the
Nernst equation for the hydrogen electrode is
452
EH =H2
RT
ln
EH =H2
2F
PH2
a2H
!
9:2:13
and that for the silver | silver chloride electrode

EAgCl=Ag EAgCl=Ag
RT
ln aCl
F
9:2:14
Equation (9.2.11) can be obtained by subtracting equation (9.2.13) from equation

(9.2.14). In other words,
Ecell EAgCl=Ag EH =H2
9:2:15
It is emphasized that experiments only provide values of the EMF of the whole
cell, and the polarity of one electrode with respect to the other. The arbitrary
reference system, that is, the SHE, is introduced in order to establish a table of
standard potentials for common electrodes. Since the standard potential for this
system is zero by denition,
EH =H2 0
9:2:16
It follows that the standard EMF of cell (9.2.1) can be used to determine the
standard potential of the silver | silver chloride electrode. On the basis of the
analysis presented for cell (9.2.1),
EAgCl=Ag 0:222 V
9:2:17
It is important to recognize that a SHE cannot be constructed in the laboratory

on the basis of experimental data. This follows from the fact that single ion
activities cannot be measured experimentally. Referring to cell (9.2.1), the activity
of the H ion is, in general, dierent from that of the Cl ion for a given concentration. One is only able to measure the mean activity a of the HCl electrolyte
by experiment. Thus, although there exists a solution in which the H ion activity
is unity, the corresponding concentration of HCl is not known. However, there is
no problem in determining the standard EMF of cell (9.2.1) experimentally as
shown in section 9.5A. The standard potentials for the individual reactions are
dened from this experimental result.
The properties of a cell with a zinc electrode and a silver | silver chloride
electrode are now analyzed:
Cu j Zn j ZnCl2 ; H2 O j AgCl j Ag j Cu0
9:2:18
Since electrons ow spontaneously from the left-hand electrode to the right-hand

one when these terminals are connected externally, the spontaneous cell reaction
is
Zn 2AgCl ! ZnCl2 2Ag
9:2:19
The Gibbs energy change is

G G RT ln aZnCl2
9:2:20
where aZnCl2 is the activity of the ZnCl2 electrolyte in the aqueous solution. The
corresponding equation for the EMF is
E E
RT
ln aZnCl2
2F
453
9:2:21
where it is recognized that reaction (9.2.19) involves two electrons. The standard
Gibbs energy change for this reaction is 190.08 kJ mol-1, and the corresponding
standard EMF, 0.985 V. The individual equilibria at each electrode when no
current is drawn from the cell are
Zn2 2e (
+ Zn
AgCl e (
+ Ag Cl
9:2:22
9:2:23
The Nernst equation for the zinc electrode is

EZn2 =Zn
EZn2 =Zn

RT
1
ln

2F
aZn2
9:2:24
and that for the silver | silver chloride electrode is equation (9.2.14). Equation
(9.2.21) can be obtained by subtracting equation (9.2.24) from equation (9.2.14).
Since the standard potential for the silver | silver chloride half-cell is 0.222 V, it
follows that the standard potential of the zinc half-cell is 0.222 V 0.985 V, that
is, 0:763 V.
Because the cell diagram describes a system from which thermodynamic data
may be obtained, cells can be added or subtracted to obtain new cells. For
example, subtracting cell (9.2.1) from cell (9.2.18), one obtains the cell
..
Cu j Zn j ZnCl2 ; H2 O .. .. HCl; H2 O j H2 1 bar) j Pt j Cu0
9:2:25
The cell reaction for this system is

Zn 2H ! Zn2 H2
9:2:26
and the standard EMF is 0.763 V. This is also the negative value of the standard
potential for the zinc half-cell. Cell (9.2.25) contains a special liquid junction
which prevents mixing of the two electrolyte solutions containing HCl and
ZnCl2 . It is important to note that the liquid junction has been designed to
eliminate any associated potential drop. In general, liquid junctions do have an
associated potential drop which must be considered in assessing the properties of
the system. Liquid junction potentials are considered in detail in section 9.7.
Cells like (9.2.25) may be used to determine the standard potentials for a series
of metal|metal ion half cells with the general reaction
Mn ne (
+M
9:2:27
In these experiments the SHE may be regarded as a reference electrode, and the
standard EMF is equal to the negative value of the standard potential for the
metal | metal ion equilibrium when the electrolyte concentration corresponds to
unit activity. Thus, for the cell
..
Cu j M j Mn ; H2 O .. .. H ; H2 O j H2 (1 bar j Pt j Cu0
E ESHE EMn =M EMn =M
9:2:28
9:2:29
because ESHE is zero by denition. The Nernst equation for equilibrium (9.2.27) is
454
EMn =M
EMn =M

RT
1
ln

nF
aMn
9:2:30
Values of the standard potentials for some metal | metal ion half-reactions are
given in table 9.1. These cover a very wide range from 3.05 V for the strongly
reducing metal Li to 1.4 V for gold, which is a very stable noble metal. Three
factors determine the value of E . One is the Gibbs energy of solvation of the
metal ion in water. As has been seen from the discussion in chapter 3, this depends
on the radius of the ion and the properties of water as a solvating medium. The
second factor is the Gibbs energy of the metal, which depends on its structure as a
solid. Finally, the ionization potential of the metal plays a role as well. Metal |
metal ion reactions are examples of type I electrode reactions. This name derives
Table 9.1 Standard Potentials for

Some Metal-Metal Ion Half-Reactions
in Aqueous Solution at 25 C*
Reaction
Standard Potential
E =V
( Li
Li e +
( Cs
Cs e +
3.040
( Rb
Rb e +
(K
K e +
2.943
( Ba
Ba2 2e +
2
( Sr
Sr 2e +
2.906
( Ca
Ca 2e +
(
Na e + Na
2.869
( Mg
Mg2 2e +
( Al
Al3 3e +
2.357
( Mn
Mn2 2e +
2
( Cr
Cr 2e +
1.182
( Zn
Zn2 2e +
2
(
Fe 23 + Fe
0.762
( Cd
Cd2 2e +
( Co
Co2 2e +
0.402
( Ni
Ni 2e +
2
( Sn
Sn 2e +
0.236
( Pb
Pb2 2e +
2
( Cu
Cu 2e +
0.127
( Ag
Ag e +
( Hg
Hg2 2e +
3
Au
( Au
3e +
3.027
2.936
2.899
2.715
1.676
0.902
0.409
0.282
0.141
+0.339
+0.799
+0.852
+1.52
*Based on data in references [G3] and [1].
455
from the fact that only one equilibrium is involved in the electrode process, and
that the electrode itself takes part in this equilibrium.
Another example of a type I reaction is an amalgam electrode. In such a system
the metal which is in equilibrium with its ion in solution is dissolved in mercury.
The metal activity in the liquid amalgam may be varied by varying its concentration in the amalgam. An example of such an electrode is the system
Zn; Hg j Zn2 ; H2 O
9:2:31
Zn2 2e (
+ ZnHg
9:2:32
in which the half-reaction is
The Nernst equation for this electrode is

EZnHg EZnHg
RT
a
ln Zn
2F aZn2
9:2:33
where aZn is the activity of Zn metal in the amalgam. In this system both aZn and
aZn2 may be varied to change the electrode potential.
Redox processes are examples of interfacial equilibria in which the electrode is
not chemically involved, but acts only as a source or sink for electrons. A simple
example of such an electrode is
Pt j Fe3 ; Fe2 ; H2 O
9:2:34
for which the half-reaction is
Fe3 e (
+ Fe2
9:2:35
In this system, platinum is the inert source or sink for electrons. The Nernst
equation for the electrode process is
EFe3 =Fe2 EFe3 =Fe2
RT aFe2
ln
F
aFe3
9:2:36
There are many examples of redox processes in which all components are in the
electrolyte solution. An example of a complex reaction which is used in analytical
2
chemistry is the reduction of MnO
in acidic solution. The half-reaction
4 to Mn
is
2
(
MnO
4H2 O
4 8H 5e + Mn
for which the Nernst equation is

EMnO4 =Mn2
EMnO =Mn2
4
RT
aMn2
ln

5F
aMnO4 a8H
9:2:37
!
9:2:38
Another example of a redox electrode is a gas electrode such as the hydrogen

electrode discussed above. In this system a constant concentration of molecular
hydrogen is maintained in the electrolyte solution, which also contains the hydrogen ion H. In order to achieve these conditions, hydrogen gas at constant pressure is bubbled over the inert platinum electrode at the point where it contacts the
solution (g. 9.1). Other gas electrodes involve molecular oxygen and chlorine.
456
A type II electrode is one in which two equilibria are involved in determining

the electrode process. These systems often involve a metal electrode coated with
an insoluble salt. A well-known example is the silver | silver chloride electrode
discussed above. The solid silver chloride ensures that the surrounding solution is
saturated with sparingly soluble AgCl. The Ag metal is in equilibrium with Ag
ion in solution:
Ag e (
+ Ag
9:2:39
The activity of the silver ion is controlled by the solubility equilibrium

+ Ag Cl
AgClcry (
9:2:40
These two processes add to give the overall electrode equilibrium expressed by
equation (9.2.23). As a result the silver | silver chloride responds to the activity of
the chloride ion in solution.
Another well-known example of a type II electrode is the calomel system:
Hg j Hg2 Cl2 j Cl ; H2 O
9:2:41
for which the half-reaction is

Hg2 Cl2 2e (
+ 2Hg 2Cl
9:2:42
The Nernst equation for the calomel electrode is

Ecal Ecal
RT
ln a2Cl
2F
9:2:43
This electrode also responds to the chloride ion activity in solution. The corresponding electrode responding to the sulfate ion activity in solution is
Hg j Hg2 SO4 j SO2
4 ; H2 O
9:2:44
These half-cells are often used to make very stable and reliable reference electrodes.
In the same way one may dene a type III electrode process. Such a system
involves three equilibria, one involving the metal | metal ion system and two
equilibria in solution. Examples of type III electrodes are systems involving complex ions undergoing two complex formation steps.
9.3 The Thermodynamic Basis of the Nernst Equation

The Nernst equation for single electrodes was introduced above on the basis of a
cell in which the second electrode is the SHE. Since each electrode reaction in the
cell is at equilibrium the Gibbs energy change for each process must be zero. It
follows that the Nernst equation for a single electrode process cannot be derived
in the same way as that for the overall cell reaction. Moreover, the overall cell
reaction can be written in terms of uncharged species so that the thermodynamics
for the process can be developed on the basis of the chemical potential of each
species involved. However, a single electrode reaction necessarily involves charged
457
species. Therefore, the thermodynamics must be derived using electrochemical

potentials (section 6.6).
In order to illustrate the thermodynamics of the Nernst equation for a single
electrode, the following system is considered:
Cu j Zn j ZnCl2 ; H2 O j AgCl j Ag j Cu0
9:3:1
By choosing electrodes which are reversible to the cation and anion in the electrolyte, a liquid junction in the cell is avoided. The equilibrium process at the lefthand electrode is
Zn2 2eZn (
+ Zn
9:3:2
The condition for equilibrium is

~Zn2 2~
m
mZn
el mZn
9:3:3
~Zn
~Zn2 is the electrochemical potential of the Zn2 ions in solution, m
where m
el , that
of the electrons in the Zn electrode, and mZn , the chemical potential of metallic
zinc. Now each term is written out in the usual way: for the Zn2 ion, one obtains
~Zn2 mZn2 RT ln aZn2 2Ffs
m
9:3:4
where mZn2 is the standard chemical potential of the Zn2 ion, aZn2 , its activity,
and fs , the inner potential of the solution; the expression for the electron is
Zn;
~Zn
m
FfZn
el mel
9:3:5
where mZn;
is its standard chemical potential, and fZn , the inner potential of the
el
Zn electrode; nally, for the zinc metal
mZn mZn
9:3:6
where mZn is the standard chemical potential of Zn. The last result follows from
the fact that the Zn is a pure substance. Substituting equations (9.3.4)(9.3.6) into
the equilibrium condition (equation (9.3.3)), and rearranging, one obtains the
result
s Zn f
fZn fs
GZn2 =Zn
2F
RT
ln aZn2
2F
9:3:7
GZn2 =Zn is the standard chemical potential change associated with equilibrium
(9.3.2), that is,
GZn2 =Zn mZn mZn2 2mZn;
el
9:3:8
Equation (9.3.7) relates the Galvani potential dierence at the Zn | solution interface to the activity of the Zn2 ion in solution. However, since neither s Zn f nor
aZn2 can be measured experimentally, and GZn2 =Zn contains contributions
which are also not available from experiment, equation (9.3.7) is only of theoretical interest. On the other hand, it contains the essential elements of the
Nernst equation for this single electrode system.
The equilibrium process at the right-hand electrode in this cell involves
458
AgCl eAg (
+ Ag Cl
9:3:9
The condition for equilibrium is

~Ag
~Cl
mAgCl m
el mAg m
9:3:10
~Ag
~Cl , the elecwhere m
el is the electrochemical potential for the electrons in Ag, m
trochemical potential for chloride ions in solution, and mAgCl and mAg , the chemical potentials for the pure substances AgCl and Ag metal, respectively. The
electrochemical potential of the electron is given by
Ag;
~Ag
FfAg
m
el mel
9:3:11
where mAg;
is the standard chemical potential for this species, and fAg , the inner
el
potential of the silver electrode. For the Cl ion in solution, the corresponding
expression is
~Cl mCl RT ln aCl Ffs
m
9:3:12
where mCl is the standard chemical potential, and aCl , its activity. The equation
for the Galvani potential dierence associated with equilibrium (9.3.9) is obtained
after substituting equations (9.3.11) and (9.3.12) into equation (9.3.10), which
leads to
s Ag f
fAg fs
GAg=AgCl RT
ln aCl

F
F
9:3:13
where
GAg=AgCl mAg mCl mAgCl mAg;
el
9:3:14
The results obtained for the Galvani potential dierences s Zn f and s Ag f
dene the fundamental electrical properties of the electrode | solution interfaces in
the electrochemical cell (9.3.1). However, as pointed out earlier, they cannot be
measured experimentally. Further analysis of the properties of this system
requires that the potential drops at all phase boundaries be considered. In fact,
there are four phase boundaries in cell (9.3.1) and therefore four Galvani potential
dierences. If the Galvani potential dierence obtained by experiment is Cu Cu0 f ,
then the following relationship holds:
Cu Cu0 f
Cu Zn f Zn s f s Ag f Ag Cu0 f
9:3:15
The contributions involving the Cu metal contacts at each end of the cell can be
determined by recognizing the electronic equilibria which exist between the electrons in each metal. On the left-hand side,
( eZn
eCu +
9:3:16
so that
~Cu
~Zn
m
el m
el
9:3:17
~Cu
where m
el is the electrochemical potential of the electrons in Cu. Writing each
electrochemical potential as a standard chemical potential together with the elec-
459
trical contribution depending on the inner potential of the given phase, equation
(9.3.17) leads to the result
Cu Zn f
fZn fCu
mZn;
mCu;
el
el
F
F
9:3:18
Similarly for the equilibrium

eAg (
+ eCu0
9:3:19
on the right-hand side of the cell,

0
Ag
Cu0
Cu0
f f
f
Ag
mCu ; mAg;
el
el
F
F
9:3:20
Substituting equations (9.3.18) and (9.3.20) into equation (9.3.15), one obtains
mZn;
mCu;
el
el
0

f
Cu Cu
F
F
melCu ; mAg;
el

f

f
Zn s
s Ag
F
F
9:3:21
Since the composition of the Cu wires at0 the ends of the cell is identical, the
standard chemical potentials mCu;
and melCu ; are equal. Thus, with consideration
el
of equation (9.2.2) relating the cell EMF and Cu Cu0 f, equation (9.3.21) simplies
to
!
!
Zn;
mAg;
m
E Cu Cu0 f s Ag f el
s Zn f el
9:3:22
F
F
where the rst term in the square brackets only contains contributions from the
right-hand electrode in the cell and the second, from the left-hand electrode. This
is an important result from a theoretical point of view, but in practice it does not
permit separation of the EMF into contributions from each electrode. It is important to note that the fact that copper contacts are used to measure the Galvani
potential dierence for the cell is not reected in the terms involved in equation
(9.3.22). This means that the cell EMF does not depend on the nature of the metal
used to measure the Galvani potential dierence.
EXAMPLE
Consider the cell

Ag0 j Zn j ZnCl2 ; H2 O j AgCl j Ag
9:3:23
in which the Galvani potential dierence is measured between two pieces of

silver. Show that the value of Ag0 Ag f is equal to that obtained for cell (9.3.1)
which is given by equation (9.3.22).
The expressions for s Zn f and s Ag f are those derived for cell (9.3.1)
(equations (9.3.7) and (9.3.13)).
At the Ag0 | Zn interface,
0
Ag0 Zn f
Zn
f
Ag0
mZn; mAg ;
el el
F
F
9:3:24
The experimentally measured Galvani potential dierence is given by
460

Ag0 Ag f
Ag0 Zn f Zn s f s Ag f
Substituting in equation (9.3.24), this simplies to

!
0 !
;
mAg
mZn;
el
el
s Zn f
Ag0 Ag f s Ag f
F
F
9:3:25
9:3:26
Since the silver wires at each end of the cell are identical, melAg; is equal to
and equation (9.3.26) is the same as equation (9.3.22).
0
;
,
mAg
el
On the basis of the above analysis one may identify an absolute potential
associated with each electrode. In the case of the silver | silver chloride electrode
the absolute potential is
a
EAg=AgCl
s Ag f
mAg;
el
Kabs
F
9:3:27
where Kabs is a constant used to dene the absolute potential scale. Kabs cannot be
determined by thermodynamic means. But other methods have been found to
estimate Kabs so that the electrode potential can be related to other quantities
measured on the absolute scale, for example, the Fermi level of the electrons in the
metal.
For the zinc electrode,
a
EZn
2
=Zn s Zn f
mZn;
el
Kabs
F
9:3:28
The EMF is the dierence between the absolute potentials

a
a
EZn
E EAg=AgCl
2
=Zn
9:3:29
As usual, the EMF is obtained by measuring the potential of the right-hand

electrode with respect to that of the left-hand electrode.
Although the thermodynamic analysis has given results with clear contributions from each of the electrodes, the observed EMF cannot be separated into
these contributions by experiment. As was seen in the previous section, the solution to this problem has been to choose the SHE as a reference and to quote the
standard potentials of all other half-reactions with respect to this point on the
redox potential scale. In order to illustrate the application of this concept using
absolute electrode potentials, the following cell is considered:
Cu j Pt; H2 j HCl; H2 O j AgCl j Ag j Cu0
9:3:30
The equilibrium process for the hydrogen electrode is

2H 2ePt (
+ H2
9:3:31
Analysis of the thermodynamics of cell (9.3.30) using the methods developed

above gives the following expression for the absolute potential of the hydrogen
electrode:
EHa =H2 s Pt f
melPt;
Kabs
F
9:3:32
where
GH =H2 RT
aH

ln 2 2

f

s Pt
2F
2F
aH
461
!
9:3:33
and
GH =H2 mH2 2mH 2melPt;o
9:3:34
Under standard conditions at which the activities of the hydrogen gas and hydrogen ion in solution are both unity, the absolute potential of this system is
EHa; =H s Pt f0
2
mH
melPt;
m
Kabs 2 H Kabs
F
2F
F
9:3:35
s Pt f0
is the value of the Galvani potential dierence dened in equation (9.3.33)

for unit H2 gas pressure and unit H ion activity.
As emphasized above, the standard potential of the hydrogen electrode is
dened to be zero on the conventional scale of standard electrode potentials.
On the basis of equation (9.3.35) there are two factors which cannot be evaluated
thermodynamically in the expression for EHa; =H . First, the value of Kabs is not
2
known. Second, the standard chemical potential of a single ion such as the H ion
cannot be determined experimentally. Nevertheless, attempts have been made to
estimate EHa; =H on the basis of extrathermodynamic assumptions. One approach
2
to determining this parameter is examined in the following section.
9.4 The Absolute Electrode Potential

Although estimation of absolute electrode potentials falls outside of the realm of
thermodynamics, it is of interest to theoreticians, especially those involved with
the quantum-mechanical description of solids and liquids. These quantities allow
one to dene an absolute potential scale which is referenced to charge-free innity.
For example, when semiconductors are used as electrodes, it is helpful to compare
the absolute potential for a redox couple in solution with that of the Fermi level in
the electrode.
Trasatti [2] has described the methods used to estimate the absolute electrode
potential on the basis of suitable extrathermodynamic assumptions. The method
presented here is the one which gives an estimate which can be related to the
potential scale used by physicists. Moreover, the resulting estimates of the absolute values of the standard electrochemical potential are based on experimentally
measured quantities. The analysis is illustrated here for cell (9.3.30), which contains a hydrogen electrode. An air gap is introduced into the cell, so that the
solutions surrounding each electrode are separated. The resulting cell is
Cu j Pt; H2 j HCl; H2 O j air j HCl; H2 O j AgCl j Ag j Cu0
9:4:1
If the Volta potential gap across the air gap is maintained at zero, then the
potential drop across the cell is a compensation potential as described in section
8.7. Thus, one may write
462
c f
Cu Cu0 f
Cu Pt f Pt s f s a f a s f s Ag f Ag Cu0 f 9:4:2
where s represents solution and a, the air. Since the Volta potential dierences at
the gap are zero,
s a c
a s c 0
9:4:3
a s f ws
9:4:4
Thus, it is anticipated that

s a f
s
where w is the surface potential of the solution. Applying the analysis presented in
section 9.3 for the remaining terms in equation (9.4.2), one obtains
!
!
mAg;
melPt;
s
el
S
w s Pt f
w
9:4:5
c f s Ag f
F
F
The two terms in square brackets may be identied with the absolute potentials of
the silver | silver chloride and hydrogen electrode, respectively. For the standard
hydrogen electrode

EHa; =H s Pt fo
2
mH
mPt;p
m
el
w s 2 H ws
F
2F
F
9:4:6
Kabs ws
9:4:7
It is clear from this analysis that
Furthermore, the standard real potential of the hydrogen ion is

H mH Fws
9:4:8
Thus, the standard potential of the hydrogen electrode is

EHa; =H
2
mH2
H

F
2F
9:4:9
Both quantities needed to estimate EHa; =H may be obtained from experiment.

2
In order to estimate EHa; =H the value of mH2 must be obtained with respect to
2
the same reference point as that used for H . The real potential is dened as the
work done to move the H ion from innity into the bulk of the solution phase.
Thus, estimation of mH2 involves calculation of the Gibbs energy change associated with formation of molecular hydrogen from unsolvated protons in the gas
phase according to the reaction

2H
g 2eg ! H2
9:4:10
This process involves recombination of the proton and electron to given a

H atom,

H
g eg ! H g
9:4:11
followed by the reaction of hydrogen atoms to give molecular hydrogen in the gas
phase,
2H g ! H2
9:4:12
463
From the ionization potential for the H atom, the Gibbs energy change for
reaction (9.4.11) is 1312.0 kJ mol1 [G3, 1]. The Gibbs energy change associated
with recombination of H atoms is 406:5 kJ mol1 . Thus, the Gibbs energy
change associated with reaction (9.4.10) is 3030:5 kJ mol1 . It follows from
equation (9.4.9) that the absolute value of the standard potential for the hydrogen
electrode is
1088
3030:5
4:42 V
96:485 2 96:485
EHa; =H
2
9:4:13
The absolute potential of the hydrogen electrode may be estimated by another

route using the well dened properties of a polarizable mercury electrode [2].
Since the metal only acts as a source or a sink for electrons for the hydrogen
electrode reaction, one could use the half-cell
Hg; H2 j H
9:4:14
to set up a standard hydrogen electrode. Although this system is thermodynamically the same as that using a Pt electrode, it is not used in practice because
equilibrium (9.3.31) is established slowly on mercury. Ignoring the kinetic
problem, thermodynamic analysis for this system gives the result that
EHa; =H s Hg f0
2
melHg;
ws
F
9:4:15
where s Hg f0 is the standard Galvani potential dierence at the Hg | solution

interface and mHg;
, the standard chemical potential for an electron in mercury.
el
The Galvani potential dierence may be divided into a Volta potential dierence,
which can be measured, and a dierence in surface potentials, which cannot be
measured. Thus,
EHa; =H s Hg c0 wHg
2
mHg;
el
F
9:4:16
where s Hg c0 is the Volta potential dierence at the standard potential, and wHg,
the surface potential of the mercury. Recalling that the work function for an
electron in mercury is
Hg;
WHg
FwHg
el mel
9:4:17
equation (9.4.16) becomes

EHa; =H s Hg c0
2
WHg
el
F
9:4:18
From Volta potential dierence measurements at the metal | solution interface

discussed in section 8.8, the value of s Hg c at the potential of zero charge (PZC)
of mercury is 0:248 V. However, the PZC is 0.192 V negative of the standard
potential of the SHE. Thus, the value of the Volta potential dierence at the
standard potential of the hydrogen electrode is 0:056 V. Using the work function for an electron in Hg (4.50 eV), the estimate of EHa; =H is
2
EHa; =H 0:056 4:50 4:44 V

2
9:4:19
464
The nal estimate of EHa; =H obtained by averaging the values from the two
2
results is
EHa; =H 4:43 0:01 V
9:4:20
The uncertainty in this estimate is quite realistic in terms of what is known about
the parameters involved.
The above analysis is easily extended to other half-cell reactions. Much of the
data required to do the necessary calculations has been collected for cells involving aqueous solutions [1] and can also be found in thermodynamic tables published by the National Bureau of Standards in Washington [G3]. In practice,
standard potentials are always used on the conventional scale because no extrathermodynamic assumptions are involved in their calculation. Any of these quantities can be converted to the absolute scale by adding the estimate of the absolute
potential of the SHE, that is, 4.43 V, to the conventional value of the standard
potential.
9.5 Experimental Studies of Electrochemical Cells

The experimental study of electrochemical cells contributed in an important way
to the acquisition of thermodynamic data for electrolyte solutions. Experimental
methods for precisely measuring the EMF of these systems were well established
in the rst half of the twentieth century [3]. As a result, many data were obtained
for electrolyte activities and for important solution equilibria. Some examples of
the experiments involved in these studies are given in this section.
A. Measurement of Electrolyte Activity Coefcients
When an electrochemical cell is set up without liquid junctions it can be used to
study the thermodynamic properties of the electrolyte involved. This requires that
one electrode be reversible to the cation of the electrolyte and the other to the
anion. A simple example is the cell
Pt; H2 j HCl; H2 O j Hg2 Cl2 j Hg
9:5:1
which was studied in detail by Hills and Ives [4]. The cell reaction for this system is
H2 Hg2 Cl2 ! 2HCl 2Hg
9:5:2
When the hydrogen pressure is 1 bar, the EMF is given by

E E
RT
RT
ln aHCl E
ln aH aCl
F
F
9:5:3
where E is the standard EMF, and aH and aCl are the activities of the individual ions. A summary of the results obtained by Hills and Ives as a function of
HCl concentration is given in table 9.2. As the concentration of HCl increases, the
activity term on the right-hand side of equation (9.5.3) becomes more negative
and the EMF of the cell decreases. This is clearly seen from the experimental data
recorded in table 9.2.
465
Table 9.2 Experimental Results for Cell (9.5.1)

Molality
of HCL
103 mc
EMF
E=V
Activity
Coecient
g
Molality
of HCl
102 me
EMF
E=V
Activity
Coecient
g
1.6077
3.0769
5.0403
7.6938
10.947
13.968
18.871
0.60080
0.56825
0.543665
0.522675
0.50532
0.49339
0.47870
0.9568
0.9416
0.9276
0.9140
0.9002
0.8903
0.8772
2.5067
3.7690
5.1645
6.4718
7.5081
9.4276
11.9304
0.46490
0.44516
0.42994
0.41906
0.41187
0.40088
0.38948
0.8639
0.8436
0.8278
0.8163
0.8095
0.7984
0.7876
The experimental results for cell (9.5.1) can be used to determine the mean
activity coecient of the electrolyte HCl. Expressing the individual ionic activities
in terms of the mean activity coecient and the electrolyte molality, equation
E E
2RT
ln g me
F
9:5:4
In order to extract the activity coecient g at each concentration one must

rst determine the standard EMF E . Taking the term ln me , which is known, to
the left-hand side, one may write
E
2RT
2RT
ln me E
ln g
F
F
9:5:5
In the limit of very dilute solutions one may assume that ln g is given by the
limiting DebyeHuckel law and write
E
2RT
2RT 0
ln me E
ADH m1=2
e
F
F
9:5:6
where ADH is the DebyeHuckel constant, and the ionic strength has been
expressed in terms of the molality. This expression shows that the quantity on
in the limit of very dilute solutions.
the left-hand side should be linear in m1=2
e
A plot of the data according to equation (9.5.6) is shown in g. 9.3. It is
apparent that it would be very dicult to determine E on the basis of this
plot, which shows curvature even in the lowest concentration range. Under
these circumstances it is better to use the extended DebyeHuckel model to perform the necessary extrapolation. Then equation (9.5.5) becomes
0
E
0
2RT
2RT ADH m1=2
e
ln me Eo
F
F 1 B0DH ame1=2
9:5:7
where BDH is the second DebyeHuckel constant and a, the ion size parameter.
In order to use this equation the ion size parameter must be chosen so that a
linear plot is obtained for data obtained up to 0.05 M. When values of a in the
range 0.20.9 nm are tested, the best linearity is obtained for a equal to 0.6 nm.
The corresponding results are also shown in g. 9.3. Extrapolation to zero
466
Fig. 9.3 Plot of E 2RT=F ln me against ADH me1=2 =1 BDH ame1=2 for cell (9.5.1) using
the data recorded in table 9.2. The value of a is zero for the results designated by (^) and
0.6 nm for those designated by (*).
molality gives a value of E equal to 0.2680 V. The relevant values of g are

then easily calculated at each molality using equation (9.5.5). These are recorded
in table 9.2.
Activity coecients may be determined using electrochemical cells in a
straightforward manner, provided there are electrodes available which are reversible to the cation and anion of the electrolyte. For example, one may determine
the activity coecients for aqueous solutions of ZnSO4 using the cell
Zn j ZnSO4 ; H2 O j Hg2 SO4 j Hg
9:5:8
Electrodes of the second kind are available for many anions. In the case of
cations, one may use the pure metal or the metal amalgam. The latter choice is
convenient for reactive metals such as the alkali and alkaline earth metals.
Of course, amalgams containing very reactive metals such as sodium also react
with water. However, special techniques have been worked out so that activity
coecient measurements can also be carried out with amalgams involving these
metals [5].
B. Determination of Equilibrium Constants
Electrochemical cells can also be used to determine other thermodynamic parameters such as equilibrium constants. For example, the solubility product for the
sparingly soluble salt AgCl may be determined by comparing the properties of the
silver | silver chloride electrode (9.2.23) with those of the silver | silver ion electrode (9.2.39). The potentials of these electrodes are equal when they are in a
saturated solution of AgCl, that is, when the activities of these ions are those given
by equilibrium (9.2.40). Therefore, under these conditions
E EAg =Ag
RT
1
ln
F
aAg
EAg=AgCl
RT
ln aCl
F
467
9:5:9
Rearranging, one obtains

ln Ksp lnaAg aCl

F

EAg=AgCl EAg =Ag
RT
9:5:10
The tabulated standard potentials are 0.799 V for the Ag | Ag electrode and
0.222 V for the silver | silver chloride electrode. It follows that Ksp is equal to 1.76
10-10. Since a saturated aqueous solution of AgCl is very dilute, this result may
be used on either the concentration or the molality scale, and the activity coecients set equal to one. Thus,
cAg cCl Ksp 1:76 1010
9:5:11
Electrochemical cells can also be used to determine the dissociation constants

of weak electrolytes. This application is illustrated with the cell
Pt; H2 j HAm1 ; Na Am2 ; NaClm3 ; H2 O j Hg2 Cl2 j Hg
9:5:12
which can be used to determine the dissociation constant of the weak acid HA.
Since the left-hand electrode is reversible to the H ion and the right-hand one to
the Cl ion, the EMF of the cell is given by
E E
RT
ln aH aCl
F
9:5:13
where E is equal to 0.268 V for this system (see cell (9.5.1)).

The dissociation of the weak acid is described by the equilibrium
HA (
+ H A
9:5:14
with an equilibrium constant given by

Ka
aH aA
aHA
9:5:15
Substituting equation (9.5.15) into equation (9.5.13) and rearranging, one obtains
E E
RT mHA mCl
RT
RT gHA gCl
ln
ln Ka
ln

m A
gA
F
F
F
9:5:16
All of the quantities on the left-hand side are known if one assumes that the
molality of the undissociated acid HA is equal to its stoichiometric molality.
This means that any dissociation of the weak acid is neglected in estimating the
left-hand side. On the other hand, the quantities on the right-hand side are
unknown. The activity coecient of the Cl ion and the anion A from the
acid are functions of ionic strength and are approximately equal to one another.
Thus, one expects the term involving the activity coecients to depend weakly on
the ionic strength. Analysis of the data to extract Ka involves plotting the function
on the left-hand side against ionic strength and extrapolating to zero ionic
strength where all activity coecients are unity. Thus, the intercept of such a
plot yields the value of the acidity constant Ka .
468
Data obtained in a study of the dissociation of propanoic acid are given in

table 9.3. In the experiment, the ionic strength was varied by approximately an
order of magnitude with the concentration of the three solute components being
kept equal. A plot of the function on the left-hand side of equation (9.5.16)
against ionic strength demonstrates that the variation in the ordinate variable is
not large, but it shows a slight tendency to increase with increase in ionic strength
(see g. 9.4). Extrapolation to zero ionic strength gives a value of 0.2884 V from
which the estimate of Ka is 1:33 105 .
The above technique was applied by Harned and his colleagues to determine
acidity constants for a variety of weak acids in both water and in waternonaqueous solvent mixtures [3]. It may also be used to determine the self-dissociation
constant of water. In the case of moderately weak acids the extrapolation procedure requires a more careful consideration of the contribution of H to the ionic
strength. More details can be found in the monograph by Harned and Owen [3].
C. Concentration Cells
Another important area of experimental investigation involves concentration
cells. These devices derive their EMF from the fact that a given cell component
has a dierent concentration in two parts of the cell. A simple example of a
concentration cell is one involving the electrodes. For example, one may set up
a cell with two amalgam electrodes containing the same metal at dierent concentrations:
Zna1 ; Hg j ZnSO4 ; H2 O j Zna2 ; Hg
Table 9.3 Experimental Results

for Cell (9.5.12) for the Case of
Propanoic Acid at 25 C
(mHA mNaA mNaCl )
Molality of
Propanoic Acid
mHA
EMF
E=V
0.004899
0.005918
0.01281
0.01605
0.01643
0.01715
0.01811
0.02220
0.02555
0.02648
0.03113
0.03179
0.03179
0.03569
0.69318
0.68809
0.66846
0.66252
0.66211
0.66098
0.65961
0.65461
0.65082
0.64983
0.64561
0.64561
0.64518
0.64224
9:5:17
469
Fig. 9.4 Plot of Ecor , the function on the left-hand side of equation (9.5.16) against ionic
strength I using the data in table 9.3. Extrapolation of the straight line tted by least
squares to I 0 allows one to estimate the acidity constant Ka for propanoic acid.
The equilibrium at each electrode is

+ Zn; Hg
Zn2 Hg 2 (
9:5:18
The Nernst equation for this process is

EZnHg EZnHg
RT aZni
ln
2F aZn2
9:5:19
where aZni is the activity of zinc in electrode i and aZn2 the activity of Zn2 ion in
solution. Subtraction of the Nernst equation for the left-hand electrode from that
for the right-hand electrode gives
E
RT aZn1
ln
2F aZn2
9:5:20
E is positive when the zinc activity in the left electrode is higher than that in the
right electrode. Study of these systems shows that the amalgam phase behaves as a
non-ideal solution. Thus, activity coecients for amalgam solutions can be determined using cells such as (9.5.17). In these experiments, the amalgam concentration is changed from very dilute, for which the activity coecient can be assumed
to be unity, to more concentrated.
Another type of concentration cell involves gas electrodes with the same gas at
dierent pressures. An example involving the hydrogen electrode is
Pt; H2 P1 j HCl; H2 O j Pt; H2 P2
9:5:21
On the basis of the Nernst equations for this system (equation (9.2.13)) the
expression for the EMF of the cell is
470
RT aH2 P1
ln
2F aH2 P2
9:5:22
where aH2 Pi is the activity of hydrogen gas at pressure Pi . E is positive when P1 is

greater than P2 .
The third type of concentration cell involves two electrolyte solutions of different concentrations. Such a cell can be set up with or without transference. The
simpler system is one without transference and it is considered rst. It can be
constructed by placing two cells of identical format but dierent electrolyte concentrations in opposition to one another. An example based on cell (9.5.8) is
Zn j ZnSO4 m1 ; H2 O j Hg2 SO4 j Hg j Hg2 SO4 j ZnSO4 m2 ; H2 O j Zn0
9:5:23
This cell has potential drops at three locations, namely, at each electrode | solution interface, and across the central electrode. For this reason it is easier to
determine the cell EMF on the basis of the individual Galvani potential drops
which make up the observed Galvani potential dierence Zn Zn0 f. The equilibrium at each electrode is
Zn2 2e (
+ Zn
9:5:24
The Galvani potential dierence at the electrode | solution interface is

s Zn f
GZn2 =Zn
2F
RT
ln aZn2
2F
9:5:25
The function of the central electrode is to facilitate the exchange of sulfate ions
between the two solutions. The equilibrium associated with a single mercury |
mercury sulfate electrode is
Hg2 SO4 2e (
+ 2Hg SO2
4
9:5:26
When two of these systems are put back-to-back, the net equilibrium process is
( 2
SO2
4 m1 + SO4 m2
9:5:27
that is, sulfate ions in solution 1 are in equilibrium with sulfate ions in solution 2.
The Galvani potential dierence between the two solutions is
s1 s2 f
fs2 fs1
RT aSO2
4 2
ln
2F aSO2 1
9:5:28
Now, one can calculate the EMF of the system using the relationship
E Zn Zn0 f Zn s1 f s1 s2 f s2 Zn0 f
9:5:29
Combining equation (9.5.25) written for each electrode | solution interface with
E
RT aZn2 2 aSO2
RT aZnSO4 2
4 2
ln
ln
2F aZn2 1 aSO2 1
2F aZnSO4 1
4
9:5:30
471
On the basis of this result, it is clear that the EMF is positive when the activity of
ZnSO4 in the left-hand compartment is less than that in the right-hand compartment.
It is helpful to analyze the functioning of cell (9.5.23) when it performs electrical work under reversible conditions. This means that zero current ows in the
external circuit and that no net changes in concentration occur in the two electrolyte solutions. The reaction which occurs at the left-hand electrode which is in
contact with the dilute solution is
Zn ! Zn2 m1 2e
9:5:31
The opposite reaction occurs on the right-hand electrode

Zn2 m2 2e ! Zn
9:5:32
The net process at the central electrode is transfer of sulfate ions from solution 2
to solution 1:
2
SO2
4 m2 ! SO4 m1
9:5:33
By adding these reactions, one obtains the overall process

ZnSO4 m2 ! ZnSO4 m1
9:5:34
This result shows clearly how the EMF derives from the dierence in concentration between the two electrolyte solutions.
An electrolyte concentration cell may also be set up with a liquid junction
between two electrolyte solutions with dierent concentrations. An example
based on the Zn | Zn2 system is
.
Zn j ZnSO4 m1 ; H2 O .. ZnSO4 m2 ; H2 O j Zn0
9:5:35
The liquid junction between the two electrolyte solutions is formed physically in a
device which allows electrical contact between the two solutions but prevents their
mixing in a signicant way. One popular device is a fritted glass disk. When this
type of barrier is used there is negligible transport of the components of one
solution to the other. However, since the electrolyte lls the pores of the frit
from both sides, electrical contact is established, and one is able to measure the
ability of the electrochemical cell to do electrical work under conditions of zero
external current.
The electrode equilibria in cell (9.5.35) are exactly the same as those in cell
(9.5.23). However, the process at the liquid junction is very dierent from that at
the central Hg | Hg2 SO4 electrode in cell (9.5.23). In order to maintain electroneutrality, both the cation and anion must be involved in the equilibrium at the
liquid junction. The process at this junction is complicated by the fact that the
individual ions move with dierent mobilities. An easy way of analyzing the
operation of this cell is to consider the cell reactions under reversible conditions.
At the left-hand electrode Zn2 ions enter solution 1 according to reaction
(9.5.31). In order to maintain electroneutrality in solution 1, Zn2 ions move
out to solution 2, and SO2
4 ions move in the opposite direction from solution
2 to solution 1. If the fraction of the current across the liquid junction carried by
the Zn2 ions is t , then transport of Zn2 is described as
472
t Zn2 m1 ! t Zn2 m2
9:5:36
In a similar way, since t is the fraction of the current carried by SO2

4 ions
2
t SO2
4 m2 ! t SO4 m1
9:5:37
In both equations it is assumed that the transport numbers t and t are independent of concentration. Finally, the reaction occurring at the right-hand electrode is given by equation (9.5.32). The net cell process is obtained by adding
equations (9.5.31), (9.5.32), (9.5.36), and (9.5.37):
2
2
1 t Zn2 2 t SO2
4 2 ! 1 t Zn 1 t SO4 1
9:5:38
Since there are only two ions in solution, the fractions t and t add to give unity,
and equation (9.5.38) simplies to
t ZnSO4 m2 ! t ZnSO4 m1
9:5:39
Since the net charge associated with this process on a molar scale is two moles of
electrons, the expression for the EMF is
E
aZnSO4 m2
RT
t ln
2F aZnSO4 m1
9:5:40
Comparison of the results for cells (9.5.23) and (9.5.35) reveals that the EMF
of the latter is smaller by the fraction t . This is a result of the potential drop
which is present across the liquid junction separating the two solutions. Since cell
(9.5.35) has the same metal at either end, the observed EMF can be written as a
sum of Galvani potential dierences. The result is given by equation (9.5.29),
where s1 s2 f now represents the Galvani potential dierence across the liquid
junction. From equation (9.5.25), the Galvani potential drops at the two electrodes are
Zn s1 f
GZn2 =Zn
2F
RT
ln aZn2 m1
2F
9:5:41
RT
ln aZn2 m2
2F
9:5:42
and
s2 Zn f
GZn2 =Zn
2F
Subtracting equations (9.5.41) and (9.5.42) from equation (9.5.40), the following
expression is obtained for the liquid junction potential:
aZn2 m2 aSO2 m2 RT aZn2 m2
RT
s2
s1
4
t ln
ln

s1 s2 f f f
2F
aZn2 m1 aSO2 m1 2F aZn2 m1
4
9:5:43
aSO2 m2
aZn2 m2 RT
RT
4
t ln
t ln
2F aZn2 m1 2F aSO2 m1
4
Because single ion activities cannot be measured experimentally, this result shows
that the liquid junction potential cannot be determined exactly by thermodynamic
measurements. However, if the solutions are both suciently dilute so that the
single ion activities can be replaced by their mean values, then one obtains
473

s 1 s 2 f
am2
RT
t t ln
2F
am1
9:5:44
The liquid junction potential may also be expressed in terms of the EMF of the
total cell. Dividing equation (9.5.44) by equation (9.5.40) one easily obtains the
result
t t
E
9:5:45
s1 s2 f
2t
Another example of a concentration cell with transference is the system
.
Ag j AgCl j KClm1 ; H2 O .. KClm2 ; H2 O j AgCl; Ag0
9:5:46
The equilibrium at each electrode is given by equation (9.3.9) for the silver | silver
chloride system. Reversible operation of this cell leads to a decrease in the KCl
activity in the left-hand compartment and an increase in the right-hand one. At
the same time there is transport of Cl ions across the liquid junction from left to
right, and transport of K ions in the opposite direction. The net cell reaction for
exchange of one electron at each electrode is
t KCl m1 ! t KCl m2
9:5:47
If the transport number is independent of electrolyte concentration, the expression for the EMF is
aKClm1
RT
t ln
E
9:5:48
F aKClm2
Because of the direction of the spontaneous cell reactions it is clear that the
solution in the left-hand compartment is more concentrated than that in the
right-hand one.
Further analysis of cell (9.5.46) leads to the following expression for the liquid
junction potential:
am1
RT
t t ln
9:5:49
s1 s2 f
F
am2
This diers from that derived for cell (9.5.35) because the dierence in the nature
of the electrode reactions. In cell (9.5.35) the reactions involve cations, whereas in
cell (9.5.46) anions are involved.
The above expressions were derived for systems in which the transport numbers are independent of electrolyte concentration. In reality, these parameters
depend on ionic strength, as was discussed earlier in section 6.6. Under these
circumstances the EMF can be calculated by assuming that the liquid junction
is made up of an innite number of junctions of innitesimally small width. In
addition, it is assumed that the activity change across each minijunction is innitesimally small. Under these circumstances, the cell potential can be calculated
on the basis of an integral. In the case of cell (9.5.46), the expression for the liquid
junction potential becomes
1
RT
t t d ln a
s1 s2 f
F
2
9:5:50
474
Obviously, estimation of s1 s 2 f requires that the variation of t with electrolyte

activity be known. A similar expression can be written for cell (9.5.35).
The results obtained for the liquid junction potential show that its magnitude
for a system with the same electrolyte on either side depends on the relative values
of the transport numbers for the cation and anion. For an electrolyte like KCl for
which t and t are approximately equal, the value of s1 s2 f is very small. On the
other hand, if the liquid junction is formed between two HCl solutions, the value
of s1 s2 f is considerably larger because the transport number for the H cation is
larger than that for the Cl anion.
The general subject of liquid junction potentials is important in electrochemistry. These junctions often involve dierent electrolytes on either side of the
boundary in which case the potential dierence is more dicult to estimate as
described later in this chapter. The experimental cells considered in this section
demonstrate that only those systems which do not have a liquid junction can be
treated exactly by thermodynamics. However, most cells used in electrochemistry
have a liquid junction, and therefore can only be treated approximately. This is
especially true for the cells used to perform electroanalysis.
9.6 Electrochemical Cells for Electroanalysis

In the simple cells without liquid junctions considered in the previous section, one
electrode was reversible to a cation in the electrolyte solution and the other to an
anion in the same solution. In a cell designed for electroanalysis, the system must
respond to only one ion in the electrolyte. A well-known example of such a system
is the pH meter, which is designed to respond to the activity of the hydrogen ion
in a test solution. The electrode that gives the response is called the indicator
electrode. The other electrode, which is designed to give a constant response
independent of the composition of the solution, is called the reference electrode.
A constant response is achieved by isolating the reference electrode solution from
the test solution via a liquid junction. Thus, the general construction of a cell
used for potentiometry in electroanalysis is given by
reference electrode j reference electrolyte solution
test electrolyte solution j indicator electrode
..
.
9:6:1
The reference electrode system is kept constant so that the composition of the cell
up to the liquid junction does not change in the experiment. When the composition of the test electrolyte solution is changed, the cell EMF changes. These
changes can be related to the concentration of a potential determining ion in
the test solution via a calibration procedure. The system is also designed to
keep the liquid junction potential constant and as small as possible. Thus, the
observed potential drop across the cell can be written as
Ecell Eind Eref flj
9:6:2
where Eind is the potential of the indicator electrode, Eref , that of the reference
electrode, and flj , the liquid junction potential.
475
Many dierent indicator electrodes may be constructed on the basis of the

systems discussed in section 9.2. For example, a pH electrode could be constructed using the hydrogen electrode with hydrogen gas bubbling over a platinum electrode immersed in the test solution. However, this system is very
inconvenient because it involves handling a gas which is explosive when ignited.
Indicator electrodes responding to metal ions can be constructed using the corresponding metal. For example, a zinc electrode responds to the zinc ion activity in
aqueous solution. However, such a system can only be used over a limited pH
range. If the test solution is too acidic, the zinc metal dissolves due to a chemical
reaction with the acid and the analytical result is too high. If the test solution is
too basic, zinc hydroxide is formed. This may precipitate on the electrode and
change the nature of the electrode process. In modern electroanalysis the classical
electrode systems are usually avoided in favor of specic ion electrodes. These
devices make use of membrane selectivity to obtain a response which depends on
the activity of a given ion. Specic ion electrodes are discussed in detail later in
this chapter.
As stated earlier, the reference electrode in a cell used for electroanalysis is
designed so that its potential is independent of the composition of the test solution. There are several general properties that reference electrodes should have in
order to be useful in analysis: (1) they should be reversible with an electrode
potential which is independent of time and reproducible; (2) they should have a
small temperature coecient; (3) they should be ideally non-polarizable with
negligible eects from the ow a small current through the system; and (4) they
should be easily constructed. The most commonly used reference electrodes are
those based on on the mercury calomel system and the silver | silver chloride
system. The electrolyte most commonly used in these systems is KCl. Relevant
parameters for commonly used reference electrodes are given in table 9.4.
The calomel electrode is based on reaction (9.2.42). Insoluble mercury(I)
chloride (calomel) lies on top of the liquid mercury phase and ensures that the
solution is saturated with Hg2
2 . The potential of the electrode depends on the
chloride ion activity as follows:
Ecal Ecal
RT
RT
ln a2Cl Ecal
ln aCl
2F
F
9:6:3
The most common electrolyte for this reference electrode is saturated KCl. This
choice ensures that the liquid junction potential between the reference electrode
Table 9.4 Potentials of Common Reference
Electrodes at 25 C on the SHE Scale
Electrode
Ag | AgCl | satd KCl
Hg | Hg2 Cl2 | satd NaCl (SSCE)
Hg | Hg2 Cl2 | satd KCl (SCE)
Hg | Hg2 Cl2 | 0.1 M KCl
Hg | Hg2 SO4 | satd K2 SO4
Potential/V
0.197
0.2360
0.2412
0.3337
0.64
476
compartment and the indicator electrode compartment is small. In some cases the
KCl is replaced by NaCl. The latter choice is made when the test solution contains
the ClO
4 anion or another anion which forms sparingly soluble salts with the K
cation. The calomel electrode is also used with lower concentrations of KCl when
contamination of the test solution by Cl ion is a problem. The calomel electrode
is easily assembled and provides excellent reproducibility. Since mercury is a
liquid, this electrode must be kept in an orientation which maintains contact
between the mercury and the metallic lead to the external connection.
The other common reference electrode is the silver | silver chloride system
which is based on reaction (9.2.23). This electrode may be made by anodizing a
silver wire in an electrolyte containing Cl ions. This results in a deposit of AgCl
on the metal surface so that a solution saturated in AgCl is maintained when the
electrode is brought into contact with an aqueous solution. This electrode suers
from two disadvantages with respect to the calomel system. The rst is that the
chemical potential of the silver depends somewhat on its metallurgical history.
This in turn has a small eect on the actual standard potential of the system. The
second problem with this system is that silver chloride is photoreactive and degenerates with continued exposure to light. Thus, commercially available Ag | AgCl
electrodes are encased in bakelite or other suitable material to prevent exposure of
the system to light. On the other hand, this reference electrode may be used in any
orientation in the experimental system. It is incorporated in most commercial
specic ion electrodes as an internal reference.
As already discussed, the standard hydrogen electrode (SHE) is the chosen
reference half-cell upon which tables of standard electrode potentials are based.
The potential of this system is zero by denition at all temperatures. Although this
reference electrode was often used in early work in electrochemistry, it is almost
never seen in chemical laboratories at the present time. It is simply too awkward
to use because of the requirement for H2 gas at 1 bar pressure and safety considerations.
The remaining feature of the electroanalytical cell is the liquid junction potential. As an example, consider the junction which arises in a cell used to determine
the pH of 0.01 M HNO3 solution with a saturated calomel electrode (SCE) as
reference:
.
Sat0 d KCl; H2 O .. 0:01 M HNO3 ; H2 O
9:6:4
Since the electrolytes on either side of the junction are dierent, there is a tendency for the KCl on the left side of the junction to diuse to the right-hand
compartment. At the same time, the HNO3 on the right side diuses to the lefthand compartment. The diusional pressure of each ion is proportional to the
product of the ionic concentration times its mobility. Thus the mass transfer
process is dominated by movement out of the concentrated KCl solution. Since
the mobilities of the K cation and Cl anion are almost the same, the mass
transfer process can proceed without any tendency to develop a potential dierence. On the other hand, the mobility of the H cation is signicantly greater than
that of the NO
3 anion in the solution on the right-hand side. As a result a positive
potential is established on the left-hand side of the junction with respect to the
477
right. This potential drop acts to keep the solution electrically neutral by retarding
mass transfer of the H. For the conditions chosen for junction (9.6.4) mass
transfer of KCl dominates so that the total potential drop at the junction is
small, that is, the order of a few millivolts. It must be remembered that the models
used to estimate flj are approximate. The only test of their approximate validity
is the fact that dierent models give estimates which are not too dierent from one
another.
The above example illustrates the reason why saturated KCl is usually used in
the calomel and silver | silver chloride electrodes. As a result, the liquid junction
potential is kept small. Nevertheless, there is an interest in estimating the value of
the liquid junction potential for the general case. This subject is considered in
more detail in the following section.
9.7 The Liquid Junction Potential

Liquid junctions are found in almost all electrochemical cells used in electroanalysis. In general, there is a potential drop across the liquid junction and it is
important to be able to evaluate it. Because of the dierent electrolyte compositions and concentrations involved, the liquid junction is associated with an irreversible mass transfer process. In this section, methods of estimating the potential
drop due to the liquid junction are outlined.
The case of a liquid junction between electrolyte solutions of the same composition was examined earlier for electrochemical cells with transport (section 9.5).
This situation is now re-examined using Onsagers method for dealing with mass
transfer. The system considered is
.
c1 MX .. c2 MX
9:7:1
where MX is a 11 electrolyte at two concentrations c1 and c2 separated by a

suitable liquid junction barrier. According to the Onsager treatment, the ux of
either ion, Ji is given by equation (6.7.16), that is,
Ji Li rm~ i Li RT r ln ai Li zi Frf
9:7:2
where m~ i is the electrochemical potential of the ion, ai its activity, zi , its valence, f,
the inner potential, and Li , the phenomenological coecient connecting the ux
to the gradient of the electrochemical potential. In order to keep matters simple, it
is assumed that the ux takes place in one direction only, namely, along the x
direction. This means that the gradients of the ionic activity and the electrical
potential can be replaced by their derivatives with respect to x.
Both ions move in the same direction from the solution of higher concentration
to that of lower concentration. If the concentration c1 is higher than c2 then the
ions on the left-hand side move across the liquid junction from the left to the
right. The movement is governed by the law of electroneutrality which requires
that the local concentrations of cation and anion be the same throughout the
system. This condition is met by requiring that the local ux of the two ions be
equal, that is,
478
JM JX
9:7:3
If the ions move in the x direction only, this means that

LM RT
d ln aM
df
d ln aX
df
LX RT
LM F
LX F
dx
dx
dx
dx
9:7:4
where LM and LX are the phenomenological coecients for the cation and anion,
respectively. This equation may be rearranged to obtain an expression for the
potential gradient df=dx:

df
RT
d ln aM
d ln aX

LM
LX
9:7:5
dx
FLM Lx
dx
dx
The phenomenological coecients may be related to the diusion coecients or
to the mobility of the ions (equation (6.7.21). In the present case, they can be
related to the ionic transport numbers because
tM
LM
LM LX
9:7:6
tX
LX
LM LX
9:7:7
and
Another simplication is possible if the individual ionic activities aM and aX are

replaced by their mean value a . Strictly speaking, this simplication is only valid
in dilute solutions.
Equation (9.7.5) now becomes
df RT
d ln a
t t
dx
F
dx
9:7:8
Integrating from solution 1 to solution 2, one obtains

2
RT
d ln a
t t
1 2 f f2 f1
F
dx
9:7:9
If the transport numbers are independent of solution composition, the nal result
is
1 2 f
RT
a c
t t ln 2
F
a c1
9:7:10
This result was obtained earlier on the basis of an analysis of the potential drops
within concentration cells with transport (equations (9.5.44) and (9.5.49)).
In practice, liquid junctions are usually much more complex than the example
just considered. However, it is clear that the potential drop arises because of the
requirement that the solution be locally electroneutral. For a liquid junction
involving dierent ions of varying charges this requirement may be expressed as
X
z i Ji 0
9:7:11
i
479
Applying equation (9.7.2) for the case that mass transfer occurs only in the xdirection, the expression for the potential gradient is
P
df
RT

dx
F
d ln afi
zi Li
dx
P 2
zi Li
9:7:12
The transport number of species i can be found from the fraction of the current
carried by that species when the concentration gradient is zero. The current
carried by ion i is
ii zi FJi
9:7:13
so that the transport number for this ion is

z FJ
ti P i 2 i
zi FJi
9:7:14
Using equation (9.7.2) with d ln ai =dx set equal to zero, one obtains
z2 L
t i Pi 2 i
zi Li
9:7:15
It follows that the potential gradient can now be written as

df
RT X ti d ln ai

dx
F i zi dx
9:7:16
By integrating this equation, one obtains an expression for the liquid junction
potential:
2
RT X ti
d ln ai
1 2 f f2 f1
F
zi
i
9:7:17
Before the value of 1 2 f can be estimated the concentration proles associated

with each ion must be determined.
When dierent electrolytes are present on either side of the boundary, the
electrolyte distribution is time dependent. This means that an exact thermodynamic solution to the problem is not possible. The solution to the problem given
here is a steady-state solution, that is, the solution appropriate to a system in
which mass transfer is occurring but under conditions that the liquid junction
potential is independent of time. The porous diaphragms described earlier are
examples of junctions which meet this condition. There are two well-known solutions to equation (9.7.17), one by Planck [6] and the other by P. Henderson [7].
The latter solution is more often used in practice and therefore is presented here.
There are three important assumptions made in obtaining the Henderson equation for the liquid junction potential. First of all, it is assumed that the concentration of each ion changes linearly from the value that it has in the solution on
480
the left of the junction, ci (1) to the value that it has on the right-hand side, ci 2.
For a junction with a total thickness d, this means that
x
ci ci 1 ci 2 ci 1 ci 1 si x
9:7:18
d
where si is the slope of the linear dependance. The second assumption is that the
activity coecients do not change in the region of the liquid junction. This means
that the derivative in ln ai may be replaced by the corresponding derivative in ln ci .
Finally, the mobility of each ion ui is assumed to be independent of solution
composition in the junction. The latter two assumptions lead to considerable
simplication of the problem, but are clearly in disagreement with existing experimental data.
Writing equation (9.7.16) in terms of ionic concentration, one obtains
df
RT X zi ui ci d ln ci
P 2

9:7:19
dx
F i
zi ui ci dx
i
The sum in the denominator can be written as

X

z2i ui ci 1 z2i ui si x ALJ BLJ
9:7:20
where
ALJ
z2i ui ci 1
9:7:21
and
BLJ
9:7:22
z2i ui si
Thus, equation (9.7.19) becomes

df
RT X
z i ui
dci

dx
F i ALJ BLJ dx
As a dierential equation, one may write

RT X zi ui si
dx
df
F i ALJ BLJ
Integrating from solution 1 to solution 2, one obtains

RT X zi ui si
ALJ BLJ d

f

ln
1 2
F i BLJ
ALJ
Substituting the expressions for si , ALJ , and BLJ , the nal result is
1
0P 2
P
zi ui ci 2 ci 1
zi ui ci 2
RT i
C
B i
P 2
ln@P
A
1 2 f
F
zi ui ci 2 ci 1
z2i ui ci 1
i
This is the Henderson equation.
9:7:23
9:7:24
9:7:25
9:7:26
481
The equation for the liquid junction potential is much simpler when 11 electrolytes with one common ion are present on either side of the junction. For
example, if the same anion is present in both solutions, the liquid junction can
be designated as
.
c1 M1 .. Xc2 M2 X
The sums in the Henderson equation can now be written as
X
zi ui ci 2 ci 1 uM2 cM2 2 uM1 cM1 1 uX cX 2 cX 1
9:7:27
9:7:28
X
i
z2i ui ci 2 ci 1 uM2 cM2 2 uM1 cM1 1 uX cX 2 cX 1

X
9:7:29
z2i ui ci 2 uM2 cM2 2 uX cX 2
9:7:30
z2i ui ci 1 uM1 cM1 1 uX cX 1
9:7:31
X
i
The expression for the liquid junction potential is

1 2 f
RT uM2 cM2 2 uM1 cM1 1 uX cX 2 uX cX 1

F uM2 cM2 2 uM1 cM1 1 uX cX 2 uX cX 1
ln
uM2 cM2 2 uX cX 2
uM1 cM1 1 uX cX 1
9:7:32
When the concentrations c1 and c2 are the same, the expression for the liquid
junction potential is even more simple:
1 2 f
RT M2 X 2
ln
F
M1 X 1
9:7:33
where ij is the equivalent conductance of electrolyte ij. This equation is known as

the LewisSargent equation. If the liquid junction involves a common cation with
equal electrolyte concentrations such as
.
c1 MX1 ; H2 O .. c1 MX2 ; H2 O
9:7:34
the corresponding equation is

1 2 f:
RT MX2 2
ln
F
MX1 1
9:7:35
EXAMPLE
Estimate the liquid junction for the boundary

.
0:1 M NaCl .. 0:01 M HCl
9:7:36
using the data in table 6.2. Repeat the calculation for the case that the concentration of HCl is 0.1 M.
482
Since ratios are involved in the Henderson equation, the molar conductivities may be used directly instead of the mobility
P
zi ui ci 2 ci 1
0:01 349:8 0:1 50:1 0:09 76:35
i
P
0:639
2
zi ui ci 2 ci 1 0:01 349:8 0:1 50:1 0:09 76:35
i
9:7:37
P
i
P
i
z2i ui ci 2
z2i ui ci 1
1 2 f
0:01 349:8 0:01 76:35

0:337
0:1 50:1 0:1 76:35
0:0257 0:639 ln0:337 0:018 V
9:7:38
9:7:39
The potential in the right-hand solution is more negative than that in the lefthand solution. This slows down movement of the most mobile ion, namely H,
in moving from right to left; in addition, the second most mobile ion, namely,
Cl is also retarded in its motion from left to right.
When the concentration of the two electrolytes is the same the more simple
LewisSargent equation may be used:

349:8 76:35
0:031 V
9:7:40

f
0:0257
ln
1 2
50:1 76:35
The increase in HCl concentration results in an increase in the liquid junction
potential.
These estimates could be improved by using a model in which the actual
conductivities at each concentration are used rather than the values at innite
dilution.
The Henderson equation may also be used to illustrate the principle involved in
keeping the liquid junction potential small. Consider the junction between a very
concentrated or saturated electrolyte solution and a dilute electrolyte solution:
.
saturated M1 X1 .. dilute M2 X2
9:7:41
Neglecting the terms in the sums which are due to M2 X2 , the Henderson equation
becomes

uM2 cM2 2 uX2 cX2 2
RT uX1 cX1 1 uM1 cM1 1
ln
9:7:42

f
1 2
F uM1 cM1 1 uX1 cX1 11
uM1 cM1 1 uX1 cX1 1
The ratio in the logarithm term gets larger as the concentration dierence
increases. However, when the mobilities of the cation and anion in the concentrated solution are the same, the sum in front of the logarithm is zero so that 1 2 f
is zero. As stated earlier, the electrolyte most commonly used for which this
condition is approximately met is KCl. Another suitable system is concentrated
sodium formate.
Values of the liquid junction potential between 4.2 M KCl and more dilute
solutions of the same electrolyte estimated by equation (9.7.32) are shown in table
483
Table 9.5 Liquid Junction Potentials Between 4.2

M KCl (Solution 1) and More Dilute Solutions of
HCl and KCl (Solution 2) Estimated by the
Henderson Equation (9.7.32)
Concentration of the
More Dilute Solution
Liquid Junction Potential

1 2 f=mV
0.01 M KCl
0.1 M KCl
1.0 M KCl
0.01 M HCl
0.1 M HCl
1.0 M HCl
3.0
1.8
0.7
3.0
4.6
14.1
9.5. 4.2 M KCl corresponds to a saturated solution. The liquid junction potential
varies from approximately 3 mV for 0.01 M KCl to 1 mV for 1 M KCl on the
basis of the Henderson estimate. Because the mobility of the K ion is slightly less
than that of the Cl ion, small positive values for 1 2 f are obtained. HCl is an
example of an electrolyte in which the cation has a much higher mobility than that
of the anion. For this electrolyte the estimate of 1 2 f for the dilute 0.01 M HCl
solution is the same as that for 0.01 M KCl. As the HCl concentration increases,
the estimate of the liquid junction potential also increases, reaching a value of 14
mV for 1.0 M HCl. These results demonstrate that the liquid junction potential is
an important feature of the electroanalytical system. The most eective way of
dealing with this feature is to calibrate the indicator electrode by making measurements of E over the concentration range where analysis is being carried out.
In some cases one wishes to reduce the liquid junction between two dilute
electrolytes as much as possible. In this case, a salt bridge is used to connect
the two solutions. This can be quite simply a U-shaped tube tted with fritted
glass disks at each end and lled with saturated KCl through a central tap. Thus,
the salt bridge involves two liquid junctions each of which is small. The junction
between the two solutions is described as
..
dilute M1 X1 .. .. dilute M2 X2
9:7:43
where the double dotted vertical bars represent the salt bridge.
Some eort has been made to estimate liquid junction potentials experimentally, especially for systems containing the chloride anion. A suitable cell for these
measurements is
.
Ag j AgCl j c1 M1 Cl .. c2 M2 Cl j AgCl0 j Ag0
9:7:44
The EMF of this cell is

E

aCl 1
RT
ln
1 2 f
F
aCl 2
9:7:45
484
Table 9.6 Liquid Junction Potentials Estimated from Cell

(9.7.44) and from the LewisSargent Equation (9.7.33)
1 2 f=mV
Solution 1
0.1 M KCl
0.1 M NaCl
0.1 M LiCl
0.1 M NaCl
0.01 M KCl
0.01 M NH4 Cl
Solution 2
Experiment
LewisSargent
Equation
0.1 M HCl
0.1 M HCl
0.1 M HCl
0.1 M KCl
0.01 M HCl
0.01 M HCl
26.8
33.1
34.9
6.4
25.7
27.0
28.5
33.4
36.1
4.9
27.5
27.5
The rst term on the right-hand side of equation (9.7.45) may be estimated,
assuming that the single ion activity is equal to the mean ionic activity, which
is known from studies of electrolyte solutions. Values of the liquid junction
potential estimated using the LewisSargent equation are compared with experimental estimates in table 9.6. For most of the systems studied the two quantities
agree to within 1 mV. Thus, this equation, although it is approximate, gives
acceptable results.
It should be stressed that the Henderson model recognizes that the system is
not at equilibrium, and instead assumes that it is in a steady state. In addition, it is
not the only model which was developed to investigate liquid junctions. The
design of the liquid junction is an important aspect of obtaining reproducible
experimental results. More information about this aspect can be found in the
monograph by Koryta and Stulik [8].
9.8 Membrane Potentials and the Donnan Effect

The electrical aspects of membrane phenomena have long been of interest in
science, and attracted the attention in the nineteenth century of famous physical
chemists including Gibbs, Nernst, Planck, and Ostwald. Originally, this interest
was connected to electrical phenomena in biological systems. In the present discussion attention is focused on membranes used in specic ion electrodes.
A membrane can be either a liquid or a solid. Its electrical properties arise
when it allows transport of an ion of one charge but not that of another.
Membranes are usually suciently thick that one can distinguish an inside region
and two outer boundary regions which are in contact with electrolyte solutions.
Two types of membranes are considered here: (1) membranes of solid and glassy
materials; (2) liquid membranes with dissolved ion-exchanging ions or neutral ion
carriers (ionophores). In fact all of these membranes are involved in ion exchange.
It is important to understand how this process aects the potentials which develop
in the system at both sides of the membrane.
The membrane system may be represented as
Electrolyte1 j L
Membrane
R j Electrolyte 2
485
9:8:1
where L and R represent the regions just inside the membrane on the left and
right, respectively. In solid membranes these regions are usually charged because
they contain a local excess of positive or negative charges. For this reason they are
called space charge regions. They correspond to the electrical double layer which
is formed in liquid electrolyte solutions at interfaces such as those involved in
system (9.8.1). The total potential drop across the membrane may now be divided
up as follows:
1 2 f
1 L f L R f R 2 f
9:8:2
where 1 L f is the Donnan potential associated with ion exchange on the lefthand side of the membrane, R 2 f, the Donnan potential associated with the same
process on the right-hand side, and L R f, the diusion potential due to the
concentration gradient within the membrane. Each of these contributions must
be considered in order to understand how the membrane is used in electroanalysis.
The functioning of a membrane is illustrated for the case that it can exchange
cations with solutions 1 and 2. On the other hand, anions cannot enter the
membrane. Furthermore it is assumed that the solvent in the solutions, usually
water, also does not enter the membrane. This situation is not always valid.
Estimation of the membrane potential is more complex when the solvent enters
the membrane due to osmotic eects. The cation involved in membrane transport
is designated M and is assumed to be present in both solutions, so that the
system can now be described as
c1 MX; H2 O j L
Membrane
R j c2 MY; H2 O
9:8:3
In general, the chemical potentials of the cation are not equal in a given solution
and in the membrane. At equilibrium the electrochemical potentials of the cation
must be equal in the two phases. As a result, a potential dierence called the
Donnan potential is established at each interface. Moreover, the concentration of
the cation on the left-hand side of the membrane is not always the same as that on
the right and the cation diuses from the location of high concentration to the one
where it is lower. The non-equilibrium diusion process gives rises to a diusion
potential.
Of course, there are many membranes into which anions may enter and cations
are excluded. Some important examples will be discussed in the following presentation. It turns out that basic principles of the functioning of membranes in
electroanalysis can be illustrated very well using the example of the glass membrane which is involved in the pH electrode. This is discussed in detail in the
following section. Then, the functioning of liquid membranes is considered in the
next section.
A. Membranes of Solid and Glassy Materials
These membranes function as ion-selective media when the cations and anions in
the solid material have very dierent conductivity. For example, in a crystal of
lanthanium uoride (LaF3 ), the cation and anion have approximately the same
conductivity because they have approximately the same size and are packed
486
together compactly in the crystal lattice. However, the mobility of F with respect
to that of La3 can be increased by doping LaF3 with a small amount of EuF3 . In
the case of a glass membrane, the structure is determined essentially by a supercooled solution of silicon dioxide (SiO2 ). By doping SiO2 with sodium oxide
(Na2 O), a small cation, namely, Na, is introduced into the disordered structure,
which is made up of covalently bonded silicon and oxygen atoms. The mobile
species is this type of glass is the Na cation.
In the following discussion, a monovalent cation M is assumed to be the
mobile species. Initially, the simplest situation is considered in which the solutions
on either side of the membrane contain only this cation, without any interfering
cations. An equilibrium is established between cation M in the solution and in
the membrane, which can be described as
+ M membrane
M solution (
9:8:4
Equating the electrochemical potentials of the cation in the solution and the
membrane, one can write for the left-hand side of the membrane (see diagram
9.8.3),
1;
L
RT ln aM1 Ff1 mL;
mM
M RT ln cML Ff
9:8:5
1;
where mM
and mL;
M are the standard chemical potentials of the cation M in
solution 1 and the left-hand side of the membrane, respectively, aM1 , the activity
of M in the solution, cML , its concentration in the membrane on the left side,
and f1 and fL , the inner potentials in solution 1 and the left side of the membrane, respectively. Thus, the potential dierence on the left side due to ion
exchange is
1 L f
RT
RT aM1
ln Kex
ln
M
F
F
cML
9:8:6
L;
m1;
M mM
RT
9:8:7
where
ln Kex
M
The activity coecient of the cation within the membrane is assumed to be constant independent of location, so that it is not explicitly written in the equilibrium
expression. Kex
M is the equilibrium constant for the ion exchange process. This
potential dierence is called the Donnan potential. It arises essentially because
the activities of the ion M are not the same in the solution and in the membrane.
A similar equation can be written for the ion exchange equilibrium on the
right-hand side, namely
2;
R;
mM
RT ln aM2 Ff2 mM
RT ln cMR FfR
9:8:8
The corresponding Donnan potential is

2 R f
where
RT
RT aM2
ln Kex
ln
M
F
F
cMR
9:8:9
ln Kex
M
2;
R;
mM
mM
RT
487
9:8:10
The ion exchange equilibrium constant is the same as that for the process on the
left because the standard potentials in the solution and in the membrane are
1;
L;
R;
independent of location (mM
m2;
M and mM mM :
The diusion potential can be estimated on the basis of the Henderson equation (9.7.26). Since only one ion is moving in the membrane, the sums in this
equation have only one term and the diusion potential is

uM cMR
cMR
RT uM cMR cML
RT
ln
ln

9:8:11
L R f
F uM cMR cML
F
uM cML
cML
where uM is the mobility of ion M in the membrane.
Now using equation (9.8.2) the total potential drop across the membrane is

aM1
aM2
cMR
RT
RT
RT
RT
RT
ex
ex
ln
K
ln
ln
K
ln
ln

f

1 2
M
M
F
F
F
F
F
cML
cMR
cML

aM2
RT
ln

F
aM1
9:8:12
This result shows that the membrane potential is related to the ratio of the
activities of cation M in the two solutions in a very simple way. It illustrates
how a properly chosen membrane can be used to relate the activity of the ion,
which can be transported, on one side of the membrane to that on the other.
The derivation presented here gives a greatly oversimplied picture of the
actual situation in a membrane with xed ion exchange sites. Use of the
Henderson equation implies that the diusion process in the membrane has
reached a steady state with a linear concentration distribution. This is certainly
not the case for solid membranes such as glass which are thick with respect to
the diusion length of the ion. Moreover, it is probably not valid to assume that
the ionic mobility is independent of position in the membrane. More complex
models for the membrane potential have been developed but they lead to essentially the same result. More details can be found in monographs devoted to this
subject [9].
The above analysis ignores the possibility that more than one cation in the test
solution can exchange with cations in the membrane. This is usually not the case
so that operation of an ion-selective electrode requires that the operator be aware
of interference from other species. For example, the pH electrode also responds to
Na ions so that a non-Nernstian response is obtained when the H ion concentration is very low, that is, at high pH, and the Na ion concentration is high.
The eects of an interfering cation are now assessed for the case that there are
two monovalent cations M and N in the test solution which can interact with
the membrane and which have the same charge. The membrane system is now
described as
c1 MX; H2 O j L
Membrane
R j c2 MY; c3 NY; H2 O
9:8:13
488
where the test solution (solution 2) now also contains the electrolyte NY at concentration c3 . The nature of the anions on this side of the membrane is not
important except in as much as they aect the activity of the cations M and
N. The principal eect of the interfering cation N is that it may replace the ion
to be determined, M, from the membrane. The corresponding exchange selectivity equilibrium is
M(membrane) N(solution) (
+ M(solution) N(membrane)
with an equilibrium constant Ksel equal to
aM2 cNR
Ksel
cMR aN2
(9.8.14)
9:8:15
where ai2 is the activity of ion i in the solution and ciR , its concentration on the
right-hand side of the membrane. In writing the equilibrium constant this way, the
eects of non-ideality within the membrane have been ignored. Alternatively, one
could include the activity coecients, which are assumed to be constant because
the membrane concentrations are small. Thus, in writing equation (9.8.15) the
activity coecients for the ions in the membranes are included in the denition of
Ksel .
Now it is assumed that there are a xed number of cation sites within the
membrane on the right-hand side so that one can dene a total concentration
cTR such that
cMR cNR cTR
Substituting for cNR from equation (9.8.15), one obtains
cMR aN2
cTR
cMR Ksel
aM2
9:8:16
9:8:17
so that
cMR
aM2 cTR
aM2 Ksel aN2
9:8:18
The corresponding expression for cNR is

cNR
Ksel aN2 cTR

aM2 Ksel aN2
9:8:19
The expressions for the Donnan potentials at each side of the membrane may
now be written. The Donnan potential on the left-hand side, 1 L f, is given by
equation (9.8.6) because the concentration of the interfering ion N is zero at this
side of the membrane. On the other hand, the Donnan potential on the right-hand
side is given by

aM2 Ksel aN2
RT
RT
ex
ln
K
ln
9:8:20

f
2 R
M
F
F
cTR
The diusion potential must be estimated considering transport of both ions
through the membrane. The interfering ion N diuses from the right-hand side
where its concentration is cNR , to the left-hand side where it is zero. The situation
489
for the ion being determined, M, is the same as it was in the absence of the
interferant. From the Henderson equation one obtains

uM cMR uN cNR
RT uM cMR uM cML uN cNR
9:8:21

f

ln
L R
F uM cMR uM cML uN cNR
uM cML
Only ions of the same charge are involved in diusion so this simplies to

cMR uN cNR =uM
RT
ln
9:8:22
L R f
F
cML
Now equations (9.8.18) and (9.8.19) are used to relate cMR and cNR to the total
concentration of cation sites on the right-hand side of the membrane. After substituting these expressions and simplifying, one obtains
!
aM2 cTR uN Ksel aN2 cTR =uM
RT
ln
9:8:23
L R f
F
cML aM2 Ksel aN2
The total potential drop across the membrane is now estimated by adding the
two Donnan potentials and the diusion potential. The result is

aM2 Ksel aN2
RT
RT aM1 RT
RT
ex
ex
ln KM
ln
ln KM
ln

1 2 f
cML
cTR
F
F
F
F
!
aM2 cTR uN Ksel aN2 cTR =uM
RT
ln

9:8:24
F
cML aM2 Ksel aN2
This simplies to

aM2 uN Ksel aN2 =uM
RT
ln
1 2 f
aM1
F
9:8:25
The parameters determining the response of the membrane to the interfering ion
are the equilibrium constant Ksel for the ion exchange process (9.8.14) and the
mobilities of the two ions uM and uN . Since these quantities are all assumed to be
constants, equation (9.8.25) can be further simplied to
!
0
aM2 Ksel aN2
RT
ln
9:8:26
1 2 f
F
aM1
where
0
Ksel
uM Ksel
uN
9:8:27
Equation (9.8.26) is known as the Nikolsky equation [10]. More complex versions
of this equation have been derived for the case that the interfering ion has a
dierent charge than the ion being determined, or there is more than one interfering ion.
The eect of the interfering ion on the membrane potential is illustrated in g.
9.5. The membrane potential is estimated for a constant activity of M on the lefthand side aM1 0:1 M and constant activity of the interfering ion N on the
490
Fig. 9.5 Plot of the membrane potential 1 2 for a system responding to monovalent
cation M against the logarithm of the cations activity
log10 aM for various activities of an
0
interfering cation N with a selectivity constant Ksel of 10-7: (*) aN 0.001; (&) 0.01; and
(~) 0.1 M.
right-hand
side aN2 0:001; 0:01, and 0.1 M) with a typical selectivity con0
stant (Ksel 10-7 ). The activity of M on the right-hand side is varied over a
to the ratio of
wide range. When aM2 is high, the membrane potential responds
0
M activities on its two sides in a Nernstian fashion because Ksel aN2 << aM 2.
The membrane potential is approximately given by equation (9.8.12) and it can be
used in an analytical application to determine
the activity of M. However, when
0
aM2 is very low so that aM 2 << Ksel aN2 , the membrane potential is constant
and given by
!
0
Ksel aN2
RT
ln
9:8:28
1 2 f
F
aM1
Provided aN2 and aM1 are known, the0 value of 1 2 f in this region can be used
to determine the selectivity constant Ksel for the interfering ion N.
When the mobile ion in the solid membrane is an anion, the basic equations
giving the membrane potential dier with respect to the sign of the contributing
terms. The equilibrium giving rise to the Donnan potential is now
+ X (membrane)
X (solution) (
9:8:29
The equations derived above apply to this system with the charge now being 1
for a monovalent anion. This aects the potential terms in equation (9.8.5) and
those which follow. The nal result for the potential drop across the membrane
when a monovalent anion is the mobile species is
1 2 f
RT aX2
ln
F
aX1
9:8:30
491
There are other types of solid membranes which are used in ion-selective
electrodes. An important group are those based on silver halides and suldes.
In these systems the ionic conductor is Ag. These membranes function in the
same manner as other ionic solids in which one ion has a very dierent conductivity than the other. Thus, the above analysis is also applicable to these systems.
B. Liquid Membranes
There are two important types of liquid membranes used in analytical applications. One type involves an ion-exchanging system dissolved in a hydrophobic
solvent, usually of low permittivity. The other type makes use of an ionophore or
neutral complexing agent dissolved in a similar solvent. In both systems, Donnan
potentials are established on either side of the membrane but a diusion potential
is absent because of the mobility of the solute in the liquid phase. More details
about the functioning of these membranes are given in this section.
Examples of ions and ionophores which can be used in liquid membranes are
shown in table 9.7. These species all have a strong hydrophobic character and can
be used to design membranes which respond to either cations or anions. By using
Table 9.7 Ions and Ionophores Used in Liquid
Junction Membranes
Formula
Primary Determinand
(a) Ion-exchanging ions

Ca2
ClO
4
NO
3
(b) Ionophores
Mg2
Ba2
492
liquid membranes, ion-specic electrodes responding to a wide variety of ions

become possible.
The ion-exchanging system consists of an electrolyte MA dissolved in a hydrophobic solvent S. In the example considered here, the cation M is also present in
the aqueous solutions on either side of the membrane. On the other hand, the
anion A is very hydrophobic, so that its concentration in the aqueous phases is
very small. Thus the membrane system responding to cation M can be described
as
c1 MX; H2 O j cM MA; S j c2 MY; H2 O
9:8:31
where c1 and c2 are the electrolyte concentrations in the aqueous phases, and cM ,
the electrolyte concentration in the membrane. Donnan potentials are established
at each interface as a result of the exchange of cation M between the aqueous
and membrane phases.
The distribution equilibrium involving the electrolyte MA in the membrane
phase is
MA (in membrane) (
+ MA (in aqueous solution)
(9.8.32)
From the condition of equilibrium

m m
w;
w w
mm;
MA RT ln aM aA mMA RT ln aM aA
9:8:33
where aim is the activity of ion i in the membrane, and ai (w), its activity in water.
As a result, one may dene the transfer coecient from the membrane to water as
aMw aAw
Ktr
9:8:34
MA
aMm aAm
where

Ktr
MA exp
w;
mm;
MA mMA
RT

9:8:35
Because of the hydrophobic character of the anion A , Ktr

MA is very small, and
w;
<< mMA
.
The Donnan potentials are established as a result of exchange of monovalent
cations M between the membrane and the aqueous phases. If the activity of M
is constant and fairly high on the left-hand side of the membrane then the corresponding Donnan potential can be written as
m;
mMA
1 m f
RT
RT aM1
ln Kex
ln
M
F
F
aMm
9:8:36
where
1;
m;
RT ln Kex
M mM m M
9:8:37
Kex
M is the exchange equilibrium constant for the cation M and the other symbols
have their usual meanings. In writing the equation this way it is assumed that
there is no diusion potential, the activity of M being uniform throughout the
membrane. The corresponding equation on the right-hand side is

2 m f
RT
RT aM2
ln Kex
ln
M
F
F
aMm
493
9:8:38
where
2;
m;
RT ln Kex
M mM mM
9:8:39
2;
However, the standard potentials for the cation, m1;
M and mM , are equal since both
solutions are aqueous. As a result the exchange constant Kex
M is the same on both
sides of the membrane. Now equation (9.8.38) is subtracted from (9.8.36) to
obtain the potential drop across the membrane,
1 2 f
RT aM1
ln
F
aM2
9:8:40
This simple result shows that the membrane potential is directly related to the
logarithm of the ratio of the activities of M in the two aqueous solutions.
Liquid membranes with an ionophore L function via complex formation
between the ionophore and an ion in the aqueous phase. When this ion is a
monovalent cation M the membrane phase contains the cation in the form
ML, that is, completely complexed with the ionophore. The other components
of the membrane are a hydrophobic solvent S which constitutes the majority of
the liquid phase and a hydrophobic anion A . The concentration of free cation
M within the membrane is very small but must be considered in assessing the
Donnan equilibria on each side. The description of the membrane system is
c1 MX; H2 O j cM ML ; A ; S j c2 MY; H2 O
9:8:41
The complexation equilibrium is

M membrane Lmembrane (
+ ML membrane
9:8:42
with a stability constant KML given by

KML
aMLm
aMm aLm
9:8:43
KML is very high by design so that the activity of free M within the membrane is
extremely small. The system is often designed so that the activity of free ionophore L is also very small.
The equilibrium determining the Donnan potential is given by equation (9.8.4).
Assuming that there is no diusion potential in the membrane, the Donnan
potentials on either side of the membrane are given by the same equations
obtained for the ion-exchanging system, namely, equations (9.8.36) and
(9.8.38). In addition, the expression for the membrane potential is the simple
result given by equation (9.8.40).
Interfering ions can also aect the membrane potential for liquid membranes.
If it is assumed that solution 2 contains cation N which can exchange with the
cation M in the liquid membrane, the system becomes
c1 MX; H2 O j cM MA; S j c2 MY; c3 NY; H2 O
9:8:44
Ion exchange equilibrium (9.8.14) is established at the second boundary. An

expression for the membrane potential is easily derived using the procedure set
494
out above for a solid membrane. However, in the case of a liquid membrane the
diusion potential within the membrane can be neglected because the mobilities
of M and N are the same to a good approximation. The expression for the
Donnan potential between solution 1 and the membrane is still given by equation
(9.8.36). On the basis of the analysis given for solid-state membranes, the Donnan
potential between solution 2 and the membrane is

aM2 Ksel aN2
RT
RT
ex
ln KM
ln
9:8:45
2 m f
F
F
aMm
Thus, the membrane potential is
1 2 f

aM1
RT
ln
F
aM2 Ksel aN2
9:8:46
This has the same form as the Nikolsky equation (9.8.26) derived for solid-state
membranes.
The simplied analysis presented here fails for ion-exchange systems when the
activity of the exchanging ion, M in the test solution, falls to very low values.
The nature of the problem is easily understood on the basis of distribution equilibrium (9.8.32). The activity of the electrolyte MA which is in the aqueous phase is
very small and can be neglected under most circumstances. However, when the
activity of M originally present in the aqueous solution is low and comparable to
that which comes from the membrane, then both sources of M aect the Donnan
potential at that interface. In other words, the ability of the membrane to respond
in a linear way to the logarithm of the activity of M in a given solution is limited
tr
by the value of Ktr
MA . The smaller the value of KMA , the lower the range of linear
response observed experimentally. When the activity of M originally present in
the aqueous solution falls below that which comes from the membrane, the
Donnan potential is approximately constant reecting the contribution from
the distribution equilibrium. More details about this limitation of ion-exchanging
liquid membranes and other problems can be found in the monograph by Koryta
and Stulik [8].
9.9 Ion-Selective Electrodes

An ion-selective electrode (ISE) is a sensor with a membrane which is designed so
that its potential indicates the activity of a specic ion in an electrolyte solution.
The membrane may be a solid, either crystalline or glassy, or a liquid. The best
known ISE is the glass electrode used to determine pH. The membrane not only
develops a potential dierence which responds to the unknown ionic activity in
the test solution, but it also separates completely the test solution from the internal reference solution of the ISE. Three dierent ISEs involving glass, crystalline,
and liquid membranes are shown in g. 9.6. All of these systems involve an
internal reference electrode. In each case, the membrane is at the bottom of the
ISE, which is dipped into the test solution. More details about each of these
electrodes are given in the following discussion.
495
Fig. 9.6 Illustration of the structures of three ion selective electrodes (left to right): a glass
electrode; an electrode with a crystalline membrane; and a liquid membrane electrode.
An ISE usually is constructed with an internal silver | silver chloride reference

electrode. This means that the internal reference solution must contain the Cl ion
together with the ion for which the membrane is selective, for example, M. Thus,
a typical format for the ISE is
Ag j AgCl j a1 MCl; H2 O j membrane j test solution; a2 M
9:9:1
The potential of the silver | silver chloride electrode is

EAg;AgCl EAg;AgCl
RT
ln aCl 1
F
9:9:2
Since the internal reference solution is sealed within the ISE and cannot be
changed, the potential of the silver | silver chloride electrode is constant, depending only on the activity of Cl in the internal reference solution (solution 1) and
on temperature. The potential drop across the membrane is
1 2 f
RT aM 1
ln
aM 2
F
9:9:3
Since aM 1 is constant, the membrane potential depends only on the activity of

the specic ion M in the test solution. The potential of the ISE is
EISE EAg;AgCl 1 2 f EAg;AgCl
RT aCl 1 aM 1
ln
F
aM 2
9:9:4
This can also be written as

EISE EISE
RT
ln aM 2
F
9:9:5
where
EISE EAg;AgCl
RT
ln aCl 1 aM 1
F
and aM 2 is the activity of M in the test solution (solution 2).
9:9:6
496
In the case that the membrane responds to an anion X , the internal reference
solution must contain that anion. As discussed in section 9.8, the sign of the
membrane potential is then reversed. Extending consideration to both cations
and anions of all charges, equation (9.9.5) can be written as
EISE EISE
RT
ln ai2
zi F
9:9:7
where ai2 is the activity of the ion to which the membrane is selective in the test
solution, and zi , its charge.
The cell is completed with a reference electrode, for example, a saturated
calomel electrode (SCE). The observed cell potential is then
E EISE ESCE flj EISE
RT
ln ai2 ESCE flj
zi F
9:9:8
The choice of the SCE ensures that the liquid junction potential between the ISE
and the reference electrode is small, so that variation in E approximately follows
variation in the logarithm of the activity of the selected ion in the test solution.
Interfering ions cause problems with all ISEs. Usually these arise when the
concentration of the analyte ion is very low and that of the interfering ion, very
high.
The expression for the cell potential in the presence of an interfering ion j is

RT
z =z
ln ai Ksel aj i j ESCE flj
E EISE
9:9:9
zi F
where Ksel is the selectivity coecient, aj , the activity of the interfering ion, and zi
and zj , the valences of the analyte and interfering ions, respectively. This equation
is clearly a generalized form of the result obtained earlier in section 9.8 (equation
(9.8.20)). In many cases the valence of the interfering ion is the same as that of the
analyte ion so that the exponent associated with the activity aj reduces to unity.
It is important to characterize a given ISE with respect to interfering ions by
determining the value of Ksel . A common method of doing this is to determine the
value of the cell potential for a xed activity of the interfering ion but varying
activity of the analyte ion. When the activity ai is the dominant term in the
logarithm in equation (9.9.9), the cell potential is given by
E EISE
RT
ln ai ESCE flj
zi F
9:9:10
Thus a plot of E against ln ai is linear with a slope equal to RT=zi F, provided

that the liquid junction potential is constant. On the other hand, when ai falls to
very low values so that the term in aj dominates, the cell potential is

RT
z =z
ln Ksel aj i j ESCE flj
E EISE
9:9:11
zi F
Under these conditions, E is a constant independent of ln ai . If one extrapolates
the two linear sections to the point where they meet, one can estimate Ksel (see g.
9.7). If the value of ai at the intersection point is aix , then
497
Fig. 9.7 Plot of E E against log10 ai for an ISE with an interfering ion activity of 0.1 and
Ksel of 105. The straight lines through the linear sections of the curve meet at ai 106.
Ksel
aix
zi =zj
aj
9:9:12
ISEs can function with interfering ions provided that the value of Ksel is less than
103.
Successful operation of an ISE requires that the operator be aware of the
properties of the membrane used in the electrode and of its limitations. Some
important membranes used in these systems are now outlined.
A. Glass Electrodes
The glass electrode for determining pH is the best known and most widely used
ISE. It was discovered in the late nineteenth century that a thin layer of glass
which contains a relatively high Na content develops a membrane potential in
response to dierences in H activity on either side. Further study of the properties of sodium glass revealed that an optimum composition consisted of 22 wt%
Na2 O, 6 wt % CaO, and 72 wt % SiO2 . The structure of this system is determined
to a large extent by the structure of a supercooled solution of SiO2 . By doping the
SiO2 with sodium oxide a small cation is introduced into the disordered system
which is made up predominately by covalently bonded silicon and oxygen atoms.
The small conductivity of this membrane is due to movement of the Na ion (see
g. 9.8). When the glass is immersed in water, a very thin layer (up to 100 nm) at
the surface becomes hydrated. The hydration process involves the exchange of
Na in the glass with protons in solution:

H
soln Na Gl Nasoln H Gl
9:9:13
498
Fig. 9.8 Cross-sectional view of a glass membrane containing two dierent cations (shaded
circles). Each Si atom (black) is bonded to four oxygen atoms (clear), three of which are in
the same plane, with an additional oxygen atom above or below it.
The equilibrium constant is extremely large so that the mobile ions in the
hydrated layer are essentially protons except at very high pH. Formation of
this layer is essential for most pH electrodes if they are to function well. For
this reason these electrodes are kept in distilled water when not in use.
The membrane potential of a glass electrode is a result of an ion exchange
process between protons in the solution and those at cation sites in the hydrated
layer. Although protons are the mobile species in the hydrated layer, charge is
carried by Na ions in the dry interior. The response to the pH of the test solution
is linear for most systems between 2 and 12. In strongly acid solutions a positive
deviation is observed and in strongly alkaline solutions the observed pH is too
low. The exact extent of the deviations depend on the composition of the glass
and to some extent on its history. The alkaline error is due to the fact that the
system responds to alkali metal ions, each of which has a dierent Ksel . When the
H ion activity is very low, the contribution from the alkali metal cations becomes
signicant and the membrane response suggests that aH is higher than it really is.
The response of the glass at high pH can be improved by modifying its composition, for example, by adding Li2 O. The electrode error at very low pH is much less
well understood than that at very high pH.
By altering the composition of the glass one can fabricate ISEs which are
specic to dierent alkali metal cations. The best known of these is the Na
499
ISE. The glass membrane for this electrode has a composition of 11 wt % Na2 O,
18 wt % Al2 O3 and 71 wt% SiO2 . This system gives a linear response for Na
activities down to 105 M. More information about these systems can be found
elsewhere [8].
B. Crystalline Electrodes
Crystalline membranes function by virtue of the fact that one ion in an ionic
crystal may have signicant conductivity with respect to the others. These can
be systems in which this ion is smaller than the others and can occupy interstitial
positions in the crystal lattice. The best known ISE in this category is the F
electrode in which the membrane is a single crystal of LaF3 . The pure crystal has a
very low conductivity due to its regular structure. However, the conductivity of
the crystals used in ISEs is increased by doping the crystal with a small amount of
EuF3 . With 0.5 wt% EuF3 , the crystal structure is suciently distorted that the
doped crystal has a conductivity of 2 10-6 -1 cm-1. This system has a Nernstian
response for F ion concentrations down to 10-6 M. It is widely used in the
analysis of natural waters for uoride. The most important interferant is the
OH ion, so the electrode cannot be used in basic solutions. The O and OH
ions are only slightly larger than F , so that they can enter the LaF3 crystal and
eventually destroy the ISE. The OH ion is the only serious interferant for the F
electrode.
The construction of an ISE with a crystalline membrane is shown in g. 9.6(b).
The membrane is located at the bottom of the electrode, where it comes into
contact with the test solution. The solution in the reference compartment inside
the electrode contains both the Cl ion, which establishes the potential of the Ag |
AgCl reference electrode, and the ion to which the membrane is responding. For
example, in the case of the F ISE, it also contains the uoride anion.
Crystalline membranes for ISEs are also constructed with silver halides and
sulde. The mobile species in these materials is the Ag cation. The ISE is usually
designed to detect the anion in the silver salt. Thus, an ISE for the Cl anion
makes use of AgCl in the membrane. The membrane is polycrystalline and is
formed from precipitated AgCl. The functioning of these systems as anion indicators is more complicated than the F electrode because of the fact that the Ag
ion is the mobile species. In essence, the Donnan potential at the test solution
interface is controlled by the activity of Ag in the membrane and the solubility
product of the silver halide salt. These systems are widely used for determining
Cl , Br , I , and S anions. ISEs for determining Pb2 and Cd2 are also made
by mixing PbS or CdS, respectively with Ag2 S [8].
C. Electrodes with Liquid Membranes
Commercial ISEs are made with liquid membranes involving both ion-exchanging
systems and ionophores (see table 9.7). The solvent used is a hydrophobic liquid
with a low relative permittivity. Examples include decane-1-ol, 5-phenyl pentan-2ol, octyl phtalate, tri-n-phenyl phosphate, and 2-nitro-p-cymene. Another important property of the solvent is that it have a low vapor pressure so that it is not
500
lost by evaporation. The electrode is constructed with a porous plastic membrane

at the bottom (g. 9.6(c)). This becomes lled with the solution of the ion exchanger or ionophore in the hydrophobic solvent. The porous plastic is positioned
such that it contacts the internal reference solution inside the ISE on one side and
the external test solution on the other side. The internal reference solution contains an electrolyte in which there is both Cl for the Ag | AgCl reference electrode and the ion for which the ISE is selective.
The most popular ISEs using liquid membranes are those for Ca2, K, NO
3,
2
.
Ca
ISEs
are
made
with
both
ion-exchangers
and
with
ionophores.
and ClO
4
The alkyl phosphate anion shown in table 9.7 is an example of an ion-exchanging
system. By using a long alkyl chain on the phosphate its solubility in water
becomes extremely low. Two phosphate anions complex with one Ca2 cation
to form calcium dialkyl phosphate, which is soluble in the liquid membrane. This
system gives a linear response to Ca2 ion activity down to 10-6 M. The principal
interfering ions are Pb2 and other divalent cations. This ISE is often used to
determine Ca2 ion in blood serum and other biomedical applications.

ISEs designed for polyatomic anions such as NO
3 and ClO4 make use of
strongly hydrophobic cations as ion exchangers. Typical examples of suitable
cations are shown in table 9.7. The nitrate electrode responds to other anions,
especially perchlorate. It is widely used to test for nitrates in soil samples and
agricultural products.
A wide variety of ISEs for alkali metals and alkaline earth metals are available
which make use of ionophores. The best known of these is the K ISE. These
systems have varying degrees of selectivity with respect to cations of the same
group. Thus, the K ISE is usually very selective with respect to Na but is less
selective toward Cs. Since this ISE is most often used in clinical applications, its
poor selectivity toward alkali metal cations of higher atomic mass is not important because these cations are not typical components of body uids.
Because the membranes used in these systems contain a liquid, greater care
must be exercised in storing the ISE between experiments to prevent loss of the
membrane system from the porous plastic in which it contacts the test solution.
D. Composite Electrode Sensors
Composite electrode sensors are based on an ISE which is separated from the test
solution by another membrane. The second membrane either selectively separates
a certain component of the analyte or modies it by a suitable reaction. The bestknown composite electrode sensors are those designed for specic gases, and with
biomolecules, especially enzymes. The operation of a gas-sensing system and an
enzyme electrode are now described briey.
An H ISE may be designed to detect gaseous carbon dioxide when it is
isolated from the test solution by a second membrane. Operation of the sensor
is based on the reaction
+ H HCO
CO2 H2 O (
3
9:9:13
which involves the formation of both H and the hydrogen carbonate ion HCO
3.
The gas-permeable membrane at the outer end of the sensor system is a thin
501
microporous lm fabricated from a hydrophobic plastic. The pores of this lm

allow passage of gas molecules but are repellant to water molecules and electrolytes. The solution inside the sensor contains a high concentration of the HCO
3
ion. As a result of equilibrium (9.9.13), there is a change in H activity inside the
sensor but a negligible change in the activity of the hydrogen carbonate ion. Thus,
using equation (9.9.13) one may write
aH Keq
aCO2
aHCO3
9:9:14
where Keq is the equilibrium constant, and ai , the activity of species i. On the basis
of equation (9.9.10) for an H ISE
E const
Keq
RT
RT
RT
ln aH const
ln
ln aCO2
F
F
F
aHCO3
9:9:15
Since the term involving the HCO

3 ion activity is approximately constant, the
expression for the cell potential becomes
E const0
RT
ln aCO2
F
9:9:16
where const0 includes all constant terms in equation (9.9.15). This result shows
that the system responds to the CO2 activity when designed with the appropriate
gas-permeable membrane.
A variety of other gas-detecting sensors are also available. Sensors for NH3
and SO2 are similar to the CO2 system in that they make use of an H ISE which
responds to the acidbase reaction between the gas molecule and water. Other
gas-sensing systems use other ISEs. For example, a sensor for H2 S can be based
on the S ISE. The sulde anion is produced when H2 S dissolves in water by an
acid-base reaction. The S ISE then responds to the resulting activity of the
sulde anion.
Another important category of composite electrode sensors is that of enzyme
electrodes. In these systems, the analyte is brought into contact with an enzyme
immobilized on the surface of the sensor. The analyte then undergoes a catalytic
reaction to yield a species for which an ISE is sensitive, for example, ammonia,
carbon dioxide, or H ions.
The best-known enzyme electrode is that used to analyze for urea in blood. The
enzyme urease is immobilized in a polyacrylamide hydrophilic gel and xed at the
bottom of a glass electrode whose characteristics make it an NH
4 ISE.
Alternatively, the ISE can be a composite system designed to detect NH3 . In
the presence of the enzyme, urea is hydrolyzed according to the reaction

NH3 2 CO 2H2 O H ! 2NH
4 HCO3
9:9:17
The NH
4 produced is then detected by either of the ISE systems mentioned
above. Some problems exist for the NH

4 ISE because of interference by Na
and K ions which are present in blood.

Composite electrode sensors are an active area of research at the present time.
The examples discussed give a general idea of the modications required to
502
develop a composite ISE. Obviously, this type of development extends the range
of substances which can be detected by ISEs in a signicant way.
9.10 p-Functions and the Denition of pH

p-Functions are familiar ways of expressing very small quantities in chemistry.
The best known p-function is the pH. The denition of pH is
pH log10 aH
9:10:1
where aH is the activity of the H ion. Although pH is often thought of in terms

of H ion concentration, the exact denition becomes important in systems of
higher ionic strength. Thus, the pH of a 10-4 M HCl solution is very close to 4.00.
Since the ionic strength of the solution is also 10-4 M, the activity coecients of
the ions are very close to unity. Thus, the H ion activity is very close to its
concentration. However, if the ionic strength of the system is increased by adding
an electrolyte such as NaCl, the H ion activity falls below its concentration. As a
result, the pH of such a solution is greater than 4.0.
p-Functions are used to express other ion activities. pCl refers to log10 aCl
and pNa, to log10 aNa . They are also used to express equilibrium constants.
Thus
pKa log10 Ka
9:10:2
Since logarithms are given to the base 10 it is easy to convert the p-function
mentally to the actual activity to obtain an order of magnitude estimate.
Since p-functions refer to the activity of one ion in an electrolyte solution and
this solution contains at least two ions, there is a fundamental problem with its
experimental measurement. Consider for example the measurement of the pH of
an HCl solution using the cell
glass electrode | x HCl, H2 O || SCE
(9.10.3)
The potential of this cell can be written as

E Eglass ESCE
RT
ln aH flj
F
9:10:4
where flj is the liquid junction potential between the test solution containing
HCl and the SCE. The latter quantity cannot be dened in an exact thermodynamic way, even though the liquid junction was chosen to keep flj as small as
possible. Another problem with this system is the asymmetry potential which
sometimes arises with the glass membrane. For example, if the bulb-shaped membrane was formed by glassblowing, the properties of the hydrated layer inside the
electrode could be dierent than those on the outside. This dierence leads to a
small potential drop across the membrane even when the activities of the H ion
are the same on both sides. Thus, the cell dened by (9.10.3) cannot be used to
determine pH without calibration using solutions of known pH.
The problem can be examined in another way using a cell whose potential is
exactly dened by thermodynamics, namely the system
Pt; H2 1 atm j xHCl; H2 O j AgCl j Ag
503
9:10:5
The potential of this cell is given by

E ESHE EAg=AgCl
RT
ln aH aCl
F
9:10:6
Although the concentrations of the H and Cl ions are exactly equal, their
activities are not necessarily equal. Furthermore, there is no way to determine
the individual ion activities separately. As discussed earlier in this chapter, thermodynamic data for this cell are analyzed in terms of the mean ionic activity a
and mean ionic activity coecient y .
In order to deal with the practical problem of pH measurement, standard
buer solutions are used to calibrate the pH meter. These solutions have been
carefully studied by the National Institute of Standards and Technology (NIST)
in the USA using electrochemical cells designed without liquid junctions as much
as possible. In this way, a very good estimate of the true value of the pH could
be made. Selected results for three buer solutions in the temperature range
545 C are given in table 9.8. In order to make pH measurements using cell
(9.10.3), the operator rst calibrates the instrument using standard buers
which bracket the pH range where experimental results are expected. In this
way, the pH measurements obtained using dierent glass electrodes and dierent
reference electrodes are comparable.
The experimental problem in dening the pH of an aqueous solution exists
for all p-functions related to ionic activity. The fundamental reason for the
problem is that individual ionic activity coecients cannot be measured experimentally. A variety of extra thermodynamic assumptions are used to circumvent
this problem. As might be expected, the exact nature of the assumption depends
on the nature of the ion and the way it interacts with the solvent, which in most
cases is water.
Finally, the distinction between pH dened in terms of H ion activity and H
ion concentration is often ignored in practice. This is especially true in acidbase
titrations. If a calibrated pH meter is used to carry out an acidbase titration, the
Table 9.8 Selected Values of NIST Standard Buffer

Solutions in the Temperature Range 545 C
pH
Temperature
0.05 m
KH2 citrate
0.025 m
KH2 PO4 +
0.025 m
Na2 HPO4
5
15
25
35
45
3.840
3.802
3.776
3.759
3.750
6.951
6.900
6.865
6.844
6.834
0.025 m
NaHCO3 +
0.025 m
Na2 CO3
10.245
10.118
10.012
9.925
9.856
504
observed pH is the exact value. However, this distinction is usually ignored in

calculating a titration curve. Keeping in mind the fact that the pH is varying over
a wide range in a titration, neglect of the dierence between the pH dened in
terms of activity and concentration has a negligible eect on determination of the
equivalence point. Thus, the features of the titration curve are much more easily
determined, ignoring any small variation in activity coecients during the titration. This aspect of titrations and other problems associated with pH determination have been discussed in detail by de Levie [11].

The material presented in this chapter has focused on two types of electrochemical
equilibria, namely, those at the electrodes in an electrochemical cell, and those at
membrane interfaces. When the EMF of an electrochemical cell is exactly
balanced by an opposing voltage source, one can measure the Gibbs energy
change associated with the cell reaction. By studying the EMF as a function of
temperature and pressure, one can measure the corresponding entropy and
volume changes. Thus, one can determine all of the basic thermodynamic properties of the cell reaction on the basis of fairly simple experiments which can be
carried out in a highly precise fashion. This type of study provided a signicant
fraction of the activity coecient data for aqueous electrolyte solutions. Studies
of the properties of electrochemical cells have also provided data for a variety of
solution equilibria, including the dissociation constants of weak acids and bases,
and solubility product data for sparingly soluble electrolytes. Much of the work in
this area was carried out in the rst half of the twentieth century.
The other important subject discussed in this chapter is ionic equilibria at
membranes. The focus of this discussion was the membranes involved in ionselective electrodes. A variety of systems were considered, including both solidstate and liquid membranes. The study of ionic equilibria at membranes is very
important for biological systems. Although biological membranes are often more
complex structurally, the basic phenomena described here apply there as well.
The thermodynamic analyses used in this chapter make use of the electrochemical potential. In this way the electrical aspects of the interfacial equilibria
are clearly dened. Earlier work on this problem, especially that by Volta and
Nernst, had led to dierent conclusions regarding the source of the EMF in an
electrochemical cell [12]. This problem was resolved by Frumkin, essentially, by
writing out the interfacial equilibria using electrochemical potentials. In this
regard, all interfaces in the cell must be considered including those between different metals at the terminals of the cell. This was shown in the discussion of the
thermodynamic basis of the Nernst equation.
Finally, the systems discussed in this chapter have provided an excellent way of
demonstrating how important the electrical aspects of equilibria involving
charged species are. Modern research in this area involves the study of both the
thermodynamic and kinetic aspects of electrochemical reactions. This is especially
true for fuel cells, a very important area of contemporary research. In these
systems the electrode reactions are complex and the kinetics of the individual
505
steps are extremely important. Study of these systems requires an understanding

of electrocatalysis and the mechanistic aspects of complex electrode reactions.
General References
G1. Newman, J. S. Electrochemical Systems; Prentice-Hall: Englewood Clis, NJ, 1973.
G2. Laitinen, H. A.; Harris, W. E. Chemical Analysis, An Advanced Text and Reference,
2nd ed.; McGraw-Hill: New York, 1975.
G3. NBS Tables of Chemical Thermodynamic Properties; J. Phys. Chem. Ref. Data 1982,
11 (Suppl. 2).
References
1. Parsons, R. In Standard Potentials in Aqueous Solutions; Bard, A. J., Parsons, R.,
Jordan, J., eds.; Marcel Dekker: New York, 1985; Chapter 2.
2. Trasatti, S. Pure Appl. Chem. 1986, 58, 955.
3. Harned, H. S.; Owen, B. B. The Physical Chemistry of Electrolytic Solutions, 2nd ed.;
Reinhold: New York, 1950; Chapter X.
4. Hills, G. J.; Ives, D. J. G. J. Chem. Soc. 1951, 318.
5. MacInnes, D. A. The Principles of Electrochemistry; Dover Publications: New York,
1939; Chapter 8.
6. Planck, M. Ann. Phys. Chem. N.F. 1890, 39, 196; 1890, 40, 561.
7. Henderson, P. Z. Phys. Chem. 1907, 59, 118; 1908, 63, 325.
8. Koryta, J.; Stulik, K. Ion Selective Electrodes, 2nd ed.; Cambridge University Press:
Cambridge, 1983.
9. Eisenman, G. Glass Electrodes for Hydrogen and Other Cations, Principles and
Practice; Marcel Dekker: New York, 1967.
10. Nikolsky, B. P. Zh. Fiz. Khim. 1937, 10, 495.
11. de Levie, R. Aqueous Acid-Base Equilibria and Titrations; Oxford Chemistry Primer;
Oxford University Press: New York, 1999.
12. Damaskin, B. B.; Petrii, O. A. An Introduction to Electrochemical Kinetics, 2nd ed.;
Vysokaya Shkola: Moscow, 1986; p 98.
Problems
1. Use the following data at 298.15 K for a cell whose cell diagram is Zn(s) |
ZnCl2 (aq) | AgCl(s) | Ag(s) to determine the value of E .
m/mol kg1
E=V
0.002941
0.007814
0.01236
0.02144
0.04242
0.09048
0.04242
0.09048
0.02211
0.04499
1.1983
1.1650
1.1495
1.1310
1.1090
1.0844
1.1090
1.0844
1.0556
1.0328
Then determine values of g at each molality.
506
2. The value of the acid-dissociation constant of acetic acid can be determined

using the cell Pts j H2 gjHAcaq; NaAcaq; NaClaq j AgCls j Ags,
where Ac represents the acetate ion. Use the following data at 0 C to
determine the value of Ka for acetic acid at 0 C. Take E for the cell to be
0.23655 V at 0 C.
mHAc =mol kg1
mNaAc =mol kg1
mNaCl =mol kg1
E/V
0.004779
0.012035
0.021006
0.049220
0.081010
0.090560
0.004599
0.011582
0.020216
0.047370
0.077960
0.087160
0.004896
0.012426
0.021516
0.050420
0.082970
0.092760
0.61995
0.59826
0.58528
0.56545
0.55388
0.55128
3. Use the Henderson equation to estimate the liquid junction potentials for the
following systems; assume that the limiting molar conductivities given in
table 6.2 can be used to calculate the ionic mobility.
(a) 0.01 M HCl/0.01 M NaCl
(b) 0.1 M HCl/0.1 M NaCl
(c) 0.01 M KCl/1 M KCl
(d) 0.1 M MgCl2 /0.1 M NH4 Cl
4. A calcium ISE is calibrated in a 0.5 M buer solution at 25 C. This solution
contains a constant activity of magnesium ion (0.2 M), the principal interferant. Use the following data to determine the standard potential for the cell,
containing the ISE and the selectivity coecient for Mg2
aCa2 =M
E=V
0.1
102
103
104
105
106
107
94.3
65.1
36.8
15.3
8.7
7.7
7.5
5. EMF data have been obtained at 50 C for the cell

Zn sat0 d; Hg j 0:1 m ZnSO4 ; H2 O jZnx; Hg
where x is the atomic fraction of Zn in Hg. The saturated amalgam contains
0.09608 atomic fraction of Zn and has a dened zinc activity of 1.00. In other
words the saturated solution of zinc in mercury is dened to be the standard
state. Use the following data to estimate the Zn activity and activity coecient on the atomic fraction scale as a function of its concentration.
xZn
E=mV
0.09608
0.08389
0.07283
0.05609
0.04464
0.02918
0.01750
0.520
0.0
1.296
2.652
5.702
8.020
13.111
19.420
35.520
507
6. The standard potential of the Ag | AgCl | Cl electrode has been measured
very carefully over a range of temperatures. The result may be summarized as
Eo 0.23659 4.8564 10
T 3.4205 105 T
5.869 109 T3
where T is the Celsius temperature. Calculate the standard Gibbs energy and
enthalpy of formation of Cl in water and its entropy at 25 C.
7. (a) Estimate the EMF of the cell
Ag j AgCl j NaCl a 0:01 m; H2 O j Na; Hg j NaCl a 0:025 m;
H2 O j AgCl j Ag
(b) Estimate the EMF of the cell
Ag | AgCl | NaCl (a 0:01 m), H2 O | NaCl (a 0:025 m), H2 O | AgCl | Ag
given that the transference number of Na is 0.390. Use the LewisSargent
equation to estimate the liquid junction potential.
8. Consider the eect of the distribution equilibrium (9.8.32) for an ion-exchanging membrane system designed to detect cation M. The total activity of M
in the test solution is
W
W
aW
M;T aM;X aM;IE
W
where aW
M;X is the activity due to the unknown analyte and aM;IE , the activity
W
W
due to the ion exchange process. When aM;X >> aM;IE , the membrane
behaves ideally and gives a Nernstian response to the unknown analyte
W
activity. However, when aW
M;X falls to very low values so that aM;X <<
W
aM;IE , the response is no longer Nernstian. Derive an expression for the
Donnan potential at the test solution interface on the basis of the total
activity of M, aW
M;T which is valid over a wide range of activities. Use equation
(9.8.34) to describe aW
M;IE as a function of other activities in the test solution.
10
The Electrical Double Layer
Roger Parsons was born in 1926 in London,

England, where he grew up and attended school
except for his high school years, which he spent
in Edmonton, Canada. He entered Imperial
College, London in 1944 where he obtained a
B.Sc. in 1947. At the same time he carried out
doctoral research in the laboratory of Dr. John
Bockris, and received a Ph.D. in 1948. After
several more years at Imperial College, he moved
to St. Andrews University in Scotland, and then
Roger Parsons
in 1954 to the University of Bristol, where he
spent almost 25 years. In 1977, he became
Director of the Laboratoire dElectrochimie Interfaciale at the Centre
Nationale de Recherche Scientique in Meudon, France. He returned to
England in 1985 as Professor of Chemistry at the University of Southampton
where he remained until his retirement in 1992. For 37 years, Parsons was
editor of the Journal of Electroanalytical Chemistry and Interfacial
Electrochemistry. Under his stewardship, this became the most important
periodical in the eld. He contributed in signicant ways to several areas in
interfacial electrochemistry, most importantly, to studies of the electrical
double layer and electrode kinetics. He was elected a Fellow of the Royal
Society, London in 1980. Parsons traveled widely during his career and was
known in electrochemical laboratories all over the world. He received many
international honors including the Palladium Medal of the Electrochemical
Society, and the Prix Paul Pascal of the French Academy. With his rare
combination of expertise, diplomacy, and dedication, he played an important
role in bringing electrochemistry from its thermodynamic roots of the ninteenth
century to the molecular science it is today.
10.1 The Electrical Double Layer Is an Example of

Electrostatic Equilibrium
In examining the properties of the metal | solution interface, two limiting types of
behavior are found, namely, the ideal polarizable interface and the ideally nonpolarizable interface. In the former case, the interface behaves as a capacitor so
that charge can be placed on the metal using an external voltage source. This leads
to the establishment of an equal and opposite charge on the solution side. The
508
THE ELECTRICAL DOUBLE LAYER
509
total system in which charge is separated in space is called the electrical double
layer and its properties are characterized by electrostatic equilibrium. An electrical double layer exists in general at any interface at which there is a change in
dielectric properties. It has an important inuence on the structure of the interface
and on the kinetics of processes occurring there.
The classical example of an ideally polarizable interface is a mercury electrode
in an electrolyte solution which does not contain mercury ions, for example,
aqueous KCl. The charge on the mercury surface is altered using an external
voltage source placed between the polarizable electrode and non-polarizable electrode, for example, a silver | silver chloride electrode in contact with the same
solution. Within well-dened limits, the charge can be changed in both the negative and positive directions. When the mercury electrode is positively charged,
there is an excess of anions in the solution close to the electrode. The opposite
situation occurs when the electrode is negatively charged. An important point of
reference is the point of zero charge (PZC), which occurs when the charge on the
electrode is exactly zero. The properties of the electrical double layer in solution
depend on the nature of the electrolyte and its concentration.
In many electrolytes, one or more of the constituent ions are specically
adsorbed at the interface. Specic adsorption implies that the local ionic concentration is determined not just by electrostatic forces but also by specic chemical
forces. For example, the larger halide ions are chemisorbed on mercury due to the
covalent nature of the interaction between a mercury atom and the anion. Specic
adsorption can also result from the hydrophobic nature of an ion. Thus, tetraalkylammonium ions, which are soluble in water, are specically adsorbed at the
mercury | water interface because of the hydrophobic nature of the alkyl groups.
Specic adsorption of molecular solutes, such as the alcohols, occurs for the same
reason.
Double layers can also be formed at non-polarizable electrodes. If a metal
electrode is placed in a solution containing its cation, the metal is either oxidized
or metal ions in the solution are reduced, depending on the concentration of the
electrolyte. For example, when a zinc electrode is placed in a highly concentrated
solution of ZnSO4, some Zn2 ions are chemisorbed at the metal surface, so that it
acquires a net positive charge. The double layer established in solution has an
excess of SO2
4 ions whose charge compensates the positive charge on the electrode. When the ZnSO4 concentration is very low, Zn2 ions dissolve from the
electrode leaving a net negative charge on the metal. The double layer in solution
then contains an excess of Zn2 ions. A null solution is one in which the concentration of metal ions is such that no reaction occurs when the metal electrode is
introduced.
Double layers are established at most other interfaces. An important example
is the semiconductor | solution interface. In this case, a space charge region or
double layer exists both in the semiconductor phase and in the liquid solution.
Double layers are also formed at the liquid | liquid interface when both phases
contain an electrolyte.
In the present chapter, attention is focused on the double layer formed at the
metal | solution interface. The roles of the solvent and the electrolyte in establishing double layer structure are discussed in detail. The relationship of the PZC to
510
fundamental properties of the metal is also considered. The adsorption of ions

and molecules at the interface is described with special attention given to the role
of the electrical variable used to establish interfacial properties.
10.2 The Thermodynamics of the Ideally Polarizable

Interface
The thermodynamic properties of an ideally polarizable interface are most easily
examined by considering an electrochemical cell with one polarizable electrode
and one non-polarizable electrode. An example of such a system is
Hg j KCl in H2 O j Hg2 Cl2 j Hg 0
10:2:1
The potential of the electrode on the right-hand side is determined by the equilibrium
+ 2Hg 2Cl
Hg2 Cl2 2e (
10:2:2
and remains constant, provided the cell is not subjected to large sudden changes in
the potential dierence between the mercury phases at each end. The mercury |
aqueous solution interface is polarizable within certain potential limits, so that
any changes in the potential dierence across the cell are reected in changes in
the potential of the polarizable mercury electrode.
The basic thermodynamic equation describing the properties of the polarizable
interface is the Gibbs adsorption isotherm (GAI, equation (8.3.17)), that is,
X
dg
i dm~ i
10:2:3
i
where g is the interfacial tension, i, the surface excess of component i, and m~ i , its
electrochemical potential. The GAI is applied here to a system with charged
components and therefore is written in terms of electrochemical potentials rather
than chemical potentials. Of course, the electrochemical potential reduces to a
chemical potential when species involved in the electrostatic equilibrium is not
charged. Strictly speaking, equation (10.2.3) is only valid when both phases are
liquids.
The components of the metal (mercury) phase are mercury and excess electrons. In solution, there are water molecules, K ions, and Cl ions. The ions are
chosen as separate components, so that their individual surface excesses may be
assessed. However, it is understood that their bulk concentrations cannot be
dierent, so that other constraints appear in the thermodynamic description.
On the basis of the above, the GAI for the polarizable interface in cell (10.2.1)
becomes
mel K d~
mK Cl d~
mCl W dmW
dg Hg dmHg el d~
10:2:4
where the subscript el stands for electron and W for water. Since mercury is a
component whose chemical composition does not change in this system, the
chemical potential of mercury is constant, and dmHg is zero. The electrochemical
511
potential of the electrons in mercury depends on the inner potential of this phase,
so that
~el mel FfHg
m
10:2:5
d~
mel FdfHg
10:2:6
As a result, one obtains
The change in the electrochemical potential of the electrons may be related to the
surface charge density on the polarizable mercury electrode, sm, as follows:
mel Fel dfHg sm dfHg
el d~
10:2:7
On the solution side, the excess charge in the double layer, ss, is due to the
relative values of the surface excesses of cations and anions, namely, K and
Cl . Thus, one may write
ss sm FK FCl
10:2:8
On the basis of the equilibrium present in the non-polarizable electrode (equation

(10.2.2)), one may obtain a relationship between the electrochemical potential of
the chloride ion and the inner potential of the mercury on the right-hand side of
cell (10.2.1). Thus,
0
mHg2 Cl2 2~
mHg
msCl
el 2mHg0 2~
10:2:9
It follows that
0
Hg
2d~
mHg
2d~
msCl
el 2Fdf
10:2:10
Now the terms in the GAI relating to the ions are re-expressed in terms of the
chemical potential of the electrolyte
me and the inner potential of the mercury in
0
the non-polarizable electrode fHg . First of all, because of electroneutrality, the
chemical potential of the electrolyte in the solution is
~sK m
~sCl
mse m
10:2:11
Thus, the terms due to the ions may be rewritten as follows:

K d~
msK Cl d~
msCl K dmse K d~
msCl Cl d~
msCl
s
s
ms K dme m d~
msCl
K dme s d~
F Cl
F
10:2:12
Combining equations (10.2.4), (10.2.7), (10.2.10), and (10.2.12), one obtains

0
10:2:13
dg sm dfHg dfHg K dme W dmW
0
The potential dierence fHg fHg can be determined experimentally because it

corresponds to a Galvani potential dierence between two phases of identical
composition. It is designated here as E, where the minus sign indicates that
the non-polarizable electrode or reference electrode responds to the activity of
the anion in solution. Furthermore, on the basis of the GibbsDuhem relationship
xe dme xw dmw 0
10:2:14
512
where xi is the mole fraction of component i. Thus, the nal two terms in equation

x
K dme W dmW K e W dme W
dme
10:2:15
K
xW
where W
is the relative surface excess of K ions. In aqueous solutions for
K
which the mole fraction of water is usually much greater than that of the electrolyte, the relative nature of the surface excess can usually be neglected. Thus, the
nal form of the GAI becomes
dme
dg sm dE W
K
10:2:16
It is interesting to derive the GAI for the case that the non-polarizable electrode responds to the concentration of cations in solution. Such a cell is
Cu j Hg j KCl in H2 O j K in Hg j Cu 0
10:2:17
The equilibrium at the non-polarizable electrode is

+ KHg
K Hg e (
10:2:18
If the concentration of K in the amalgam is not too high and there is no signicant
reaction with water during the time frame of the experiment, the electrode on the
right-hand side provides a non-polarizable system which responds to the K ion
activity in solution. The condition for the electrochemical equilibrium is
~Hg
~sK mHg m
m
el mKHg
10:2:19
where the subscript el again denotes the electron. It follows that

0
KHg
d~
msK d~
mHg
F dfCu
el F df
10:2:20
On the basis of equation (10.2.11), one obtains

msK Cl d~
msCl K d~
msK Cl dmsel Cl d~
msK
K d~
s
msK Cl dmsel
m d~
F
Thus, the GAI becomes

0
dg sm dfHg dfCu Cl dmse W dmW

0
W
s
s
sm dfCu dfCu W
Cl dme sm dE Cl dme
10:2:21
10:2:22
where E is the potential of the polarizable Hg electrode with respect to the nonpolarizable electrode reversible to the cation in solution.
When equations (10.2.16) and (10.2.22) are compared, it is clear that the surface excess of the anion appears in the GAI when the reference electrode is
reversible to the cation. Thus, the result valid for any 11 electrolyte can be
summarized as
s
dg sm dE W
dme
10:2:23
513
where the choice of a reference electrode reversible to one of the ions results in the
surface excess of the other.
The above result can be used to examine the dependence of interfacial tension
on cell potential for a given electrolyte at xed concentration. Thus,

@g

sm
10:2:24
@E mse
As the potential drop across the cell increases in the positive direction, the charge
on the polarizable electrode changes from negative values through zero to positive
values. The interfacial tension is a maximum at the PZC and falls to lower values
for both positive and negative charge densities. Typical data obtained for the Hg |
aqueous solution interface are shown in g. 10.1. These data are often referred to
as electrocapillary curves in reference to the fact that they were obtained experimentally using a capillary electrometer. The experiment is based on the capillary
rise eect discussed in section 8.3 and is considered further in the following section
of this chapter.
When interfacial tension data are obtained both as a function of cell potential
and electrolyte concentration, one may extract information about the relative
ionic surface excesses. Using cell (10.2.1), the relative cationic surface excess is
given by

@g
@g

W
10:2:25
@mse E
RT@ ln ase E
where ase is the activity of the electrolyte in solution. The surface excess can also be
expressed in terms of charge density on the solution side of the double layer. Thus,
for a 11 electrolyte,
Fig. 10.1 Experimental data for the interfacial tension of the mercury | aqueous electrolyte
solution (1 M) interface as a function of the potential drop across the cell at 188C. The
reference electrode was a calomel electrode with 1 M KCl.
514
s FW
10:2:26
where s is the charge density due to the cations in the double layer. The charge
due to anions is obtained by using the condition for electroneutrality:
s ss s sm s
This also gives the surface excess of anions, that is,
s
W

F
10:2:27
10:2:28
Thus, having obtained values of sm and one surface excess for given values of E
and ase , the other surface excess is easily estimated.
The above analysis shows how the GAI is applied to the simplest polarizable
interface in contact with a 11 electrolyte. Other more complicated situations have
been analyzed for systems with more complex electrolytes and molecular solutes.
More details can be found in reviews by Mohilner [1] and Parsons [G4]. The
essential feature of these analyses is that an equation is derived which relates
the change in interfacial tension to the change in the potential of the polarizable
electrode with respect to that of a non-polarizable electrode, and to the chemical
potentials of the components of the solution.
The discussion so far has been limited to liquid metal electrodes. At solid metal
electrodes other factors come into play. The interfacial tension of a liquid is the
work done to increase the area of the liquid interface by unit amount. The surface
area of a solid may be increased in two ways. One way is by plastic deformation in
which the number of surface atoms is increased, for example, by cleavage of the
crystal. The other way is by deforming the crystal lattice in a dened direction
such that the interatomic lattice spacing changes but the number of surface atoms
remains constant. This is called elastic deformation. The change in the internal
energy of the surface, dUs , due to a general change in area is
dUs g dAp AYmn demn
10:2:29
where g is the interfacial tension, dAp , the change in area due to plastic deformation, Ymn , the surface stress in a direction dened by the mn component of the
surface stress tensor, and emn , the corresponding surface strain due to the elastic
deformation. The analysis of the thermodynamics of the polarizable interface for
solid metals is obviously much more complex than that for liquids. Nevertheless,
it can be shown that the equation equivalent to (10.2.23) for the solid electrode
system at constant temperature and pressure is
"
"

#

#
@emn
@emn
W
d sm g Ymn
dE g Ymn
dmse
@E mse
@mse E
10:2:30
where @emn =@E mse is the change in surface strain with potential drop across the
system at constant electrolyte composition or electrostriction, and @emn =@mse E ,
the change in surface strain with electrolyte concentration at constant potential
drop. Under the usual experimental conditions involved in the study of the polarizable interface at solid metals, the two derivatives in the surface strain are neg-
515
ligible [2]. It follows that equation (10.2.30) reduces to equation (10.2.23) for these
conditions.
Sometimes it is convenient to analyze interfacial thermodynamic data at constant electrode charge density sm rather than at constant cell potential dierence
E . Once the interfacial tension data at a given concentration have been dierentiated with respect to E to obtain sm , one may calculate the function x which
is given by
x g sm E
10:2:31
This is known as the Parsons function and is often used in the analysis of interfacial thermodynamic data. Taking the total derivative of x , one obtains
dx dg sm dE E dsm
10:2:32
After substitution of equation (10.2.23), the result is

W
dx E dsm
dmse
10:2:33
Thus, the GAI is now expressed as a dierential equation with the electrode
charge density and the chemical potential of the electrolyte as independent variables.
By cross-dierentiating equations like (10.2.23) and (10.2.33), one obtains useful relations between partial derivatives for the thermodynamic properties of the
polarizable interface. From equation (10.2.23), cross-dierentiation leads to the
result
!

@W
@g
@sm

10:2:34
@E @mse
@mse E
@E s
me
This relationship states that the change in charge density on the electrode with
electrolyte concentration measured at constant cell potential dierence is equal to
the change in ionic surface excess with cell potential dierence measured at constant electrolyte concentration. Cross-dierentiation of equation (10.2.33) gives
!

@W
@E

10:2:35

@mse sm
@sm s
me
By extending this procedure one may obtain two more useful relationships.
W
When the independent variables are E and
, it is easily shown that
!
s
@sm
@me
10:2:36
W
@E
@
E
On the other hand, when they are sm and W

, it follows that
!
s
@E
@me
@s
@W
m W

10:2:37
sm
The procedures used to derive these equations are entirely analogous to those
followed in the thermodynamics of bulk solutions (see section 1.3).
516
The above discussion illustrates that the interfacial tension, cell potential difference, electrode charge density, the ionic surface excesses, and the electrolyte
activity are the important properties of the simple system considered. Another
property which is studied experimentally is the dierential capacity. Experimental
methods for measuring the properties of a polarizable electrode are now considered and the analysis of the resulting data outlined.
10.3 The Experimental Study of the Double Layer

The rst experiments to study polarizable interfaces were carried out in the nineteenth century with mercury electrodes. After invention of the dropping mercury
electrode by Heyrovsky [3] it was realized that very precise data could be obtained
for the Hg | solution interface when the components of the system were carefully
puried. Experiments to measure precisely the interfacial capacity were soon
developed. Now these techniques and related ones are applied to study other
interfaces, including liquid j liquid and solid metal j liquid systems.
Experimental study of the double layer is not limited to thermodynamics. A
variety of spectroscopic methods have been applied to determine the structure and
composition of the double layer. Two of these, namely, second-harmonic generation and vibrational sum frequency spectroscopy, have already been described in
section 8.11. Other important techniques are based on the absorption of electromagnetic radiation when it is transmitted through or reected at the interface.
Finally, the scattering of X-rays and neutrons at interfaces has proven to be a
valuable tool for obtaining atomic level information about the interface. In the
following section some of these methods are outlined in more detail.
A. Thermodynamic Methods
On the basis of the GAI, it is clear that the interfacial tension g is the most
important experimental quantity. Three methods are commonly used to determine g at liquid j liquid interfaces, namely, the capillary electrometer method, the
maximum bubble pressure method, and the drop weight or drop time method.
The capillary electrometer method rst used by Lippmann [4] is based on the
capillary rise principle. The interface between the liquid Hg electrode and the
solution is established in a ne capillary, the position of the interface being
determined by the interfacial tension (see g. 8.2). More specically, g is given by
g hm rm hs rs gr cosy=2
10:3:1
where hm is the height of the mercury column, hs , the height of the solution above
the meniscus, rm and rs , the densities of mercury and the solution, respectively, g,
the acceleration due to gravity, r, the radius of the capillary and y, the contact
angle. Providing y and r are constant this equation can be rewritten as
g khc
where the corrected height is given by
10:3:2
hc hm
hs rs
rm
517
10:3:3
and the constant k by

10:3:4
k grm r cosy=2
2
The interfacial tension is normally in the range 200400 mJ m ; it follows that

a mercury column of about 0.6 m can be supported with a capillary radius of
10 mm. In principle the method is absolute, but in practice the device is calibrated
with a solution of known interfacial tension because of the diculty in determining r. The meniscus is viewed with a microscope positioned at an optical at in the
cell wall near the capillary. Further details regarding this method have been given
by Payne [5].
Under certain circumstances, the contact angle between Hg and the solution
can change with applied potential. For this reason, the maximum bubble pressure
method which is independent of contact angle is preferred (see section 8.2). This
technique is easily adapted to computer-controlled experiments.
The nal method discussed here is the drop time technique. At the critical size
when a mercury drop breaks from the end of a capillary placed in an electrolyte
solution, the weight of the drop is balanced by the interfacial tension forces so
that
2prg mg
10:3:5
where r is the radius of the capillary and m, the mass of the drop. If the ow rate
of the mercury, fl , is constant then the mass may be related to the drop time t:
m fl t
10:3:6
It is generally recognized that equation (10.3.5) is an approximate description of

the condition for drop break away. However, this method can be used to obtain
high-quality interfacial tension data provided one considers the more detailed
description of the related physics. More details can be found in the work of
Barradas and Kimmerle [6].
Interfacial tension against electrode potential curves have a parabolic shape
with a maximum value which depends on the nature and concentration of the
electrolyte (see g. 10.1). Detailed results for the mercury j aqueous solution
interface were initially reported by Gouy [7, G5]. Examination of these data for
the alkali metal halides shows that the interfacial tension depends markedly on
the nature of the electrolyte at positive potentials. On the other hand, the variation with electrolyte at negative potentials is rather small. It follows that the
anions in the electrolyte strongly aect the interfacial tension when they predominate in the double layer.
By dierentiating the interfacial tension data, one obtains the surface charge
density on the metal sm (equation (10.2.24)). This quantity is positive at the most
negative potentials; then it decreases in magnitude, goes through zero, and
becomes negative in the positive potential range. The maximum value of g is
the point of reference for a polarizable electrode, namely, the point of zero charge
(PZC, see section 8.8). In the case of mercury electrodes, it is also called the
518
electrocapillary maximum (ECM). Some typical results for the mercury j solution
interface are shown in g. 10.2.
A very sensitive method of studying the properties of a polarizable interface is
the measurement of its dierential capacity, C. This is given by the second derivative of the interfacial tension with respect to electrode potential:
!

@2 g
@sm
10:3:7
C
@E mse
@E2 s
me
Early experiments with mercury were carried out by Grahame [G1] using a precise
capacitance bridge. The experimental setup is shown schematically in g. 10.3. An
a.c. signal is applied to the cell as the fourth arm of a Wien bridge. The cell
consists of two polarizable electrodes and an electrolyte solution. One polarizable
electrode is a dropping Hg electrode of small area and the other, a large-area
electrode, for example, a Hg pool. Thus, the impedance of the cell consists of two
capacitors, corresponding to each electrode j solution interface, and a resistor,
corresponding to the electrolyte solution, in series, that is,
Zcell
1
1
Rs
joC1
joC2
10:3:8
C1 is the capacitance at one interface, C2 , that at the other, Rs , the solution

resistance, and o the angular frequency of the applied a.c. signal. Since the area
of the second electrode is at least 1000 times that of the rst, its contribution to
the impedance may be neglected so that
Zcell
1
Rs
joC1
10:3:9
Fig. 10.2 Plots of the charge on the polarizable Hg electrode for various 0.1 M electrolytes
against electrode potential with respect to a calomel reference electrode with 1 M KCl at
258C [G5].
519
Fig. 10.3 Schematic diagram of the capacitance bridge used to measure the dierential
capacity of the Hg electrode | solution interface.
Since the area of the dropping Hg electrode changes with time, the experimental
system must include a timing device which allows one to balance the bridge at a
precise interval after the birth of a new drop. Using the measured ow rate of the
Hg electrode and the time at which the measurement is made, one may calculate
the area of the Hg drop. The value of C1 obtained from the balanced bridge and
the electrode area are then used to calculate the specic dierential capacitance of
the interface. The d.c. potential of the polarizable interface is established in a
second circuit using a non-polarizable reference electrode (see g. 10.3). The
experiment is repeated at xed intervals of potential over the polarizable range
of the dropping Hg electrode by changing the d.c. potential drop in the polarizing
circuit by the appropriate amount.
Although they are very precise, experiments carried out using the Grahame
bridge are rather tedious because the bridge is balanced manually. In a modern
version of this experiment the in- and out-of-phase components of the cell impedance are recorded as a function of time during the life of the Hg drop using a
phase-sensitive detector. The specic capacity is then easily estimated from the
impedance data using the ow properties of the Hg electrode.
Typical dierential capacity data obtained at the mercury j aqueous solution
interface using various electrolytes at a concentration of 0.1 M at 258C are shown
in g. 10.4. The capacity of the interface depends markedly on the nature of the
anion at positive potentials. These data demonstrate that the anion strongly
520
Fig. 10.4 Plots of the dierential capacity of the Hg | aqueous solution interface for a
variety of 0.1 M electrolytes using a calomel reference electrode (1 M KCl) at 258C [G5].
inuences double layer properties when it predominates in the double layer. On

the other hand, the capacity of the interface depends very little on the nature of
the cation when these are simple monoatomic species [8] (g. 10.5). The presence
of polyatomic hydrophobic cations, such as the tetraalkylammonium cations in
solution, results in much higher capacities in the negative potential range.
Fig. 10.5 Plots of the dierential capacity of the Hg | aqueous solution interface at negative potentials for a variety of 0.1 M chloride electrolytes using a calomel reference electrode (1 M KCl) at 258C [G5].
521
The interfacial capacitance may also be measured at solid polarizable electrodes in an impedance experiment using phase-sensitive detection. Most experiments are carried out with single crystal electrodes at which the structure of the
solid electrode remains constant from experiment to experiment. Nevertheless,
capacity experiments with solid electrodes suer from the problem of frequency
dispersion. This means that the experimentally observed interfacial capacity
depends to some extent on the frequency used in the a.c. impedance experiment.
This observation is attributed to the fact that even a single crystal electrode is not
smooth on the atomic scale but has on its surface atomic level steps and other
imperfections. Using the theory of fractals, one can rationalize the frequency
dependence of the interfacial properties [9]. The capacitance that one would
observe at a perfect single crystal without imperfections is that obtained at innite
frequency. Details regarding the analysis of impedance data obtained at solid
electrodes are given in [10].
An important point of reference for interfacial thermodynamic data is the
PZC. If only dierential capacity data are available, this must be determined in
an independent experiment. For liquid electrodes such as mercury, the PZC is
conveniently measured using the streaming electrode technique [5]. The polarizable electrode is designed so that the liquid metal, for example, mercury, emerges
from a specially designed capillary in small droplets with a short lifetime. The
capillary is connected to a column of mercury on which the pressure may be
increased using an inert gas such as nitrogen. The capillary is connected to a
d.c. potential measuring device and its potential measured against a suitable
reference electrode. The pressure on the mercury, and thus, the rate at which it
streams from the capillary, is increased until the measured d.c. potential is constant and independent of further pressure increase. This value gives the PZC
against the chosen reference electrode. The principle of the technique is to
cause the area of the polarizable electrode to increase so rapidly that no signicant
charge can accumulate at the interface.
Values of the PZC at the Hg j solution interface are shown as a function of
electrolyte concentration in g. 10.6. In the case of NaF, the PZC with respect
to a constant reference electrode is independent of electrolyte concentration.
However, in the cases of the other halides, the PZC moves to more negative
potentials as the electrolyte concentration increases. The latter observation is
considered to be evidence that the anion in the electrolyte is specically
adsorbed at the interface. Specic adsorption occurs when the local ionic concentration is greater than one would anticipate on the basis of simple electrostatic arguments. Anions such as Cl , Br , and I can form covalent bonds
with mercury so that their interfacial concentration is higher than the bulk
concentration at the PZC. Furthermore, the extent of specic adsorption
increases with the atomic number of the halide ion, as can be seen from the
increase in the negative potential shift. A more complete description of specic
adsorption will be given later in this chapter.
The PZC may also be seen on dierential capacity curves when specic adsorption is absent and the electrolyte concentration is low 0:01 M). At this point the
capacity of the diuse part of the double layer is a minimum and can fall below
that of the compact or inner layer. As a result the total double layer capacity may
522
Fig. 10.6 Point of zero charge (PZC) for the Hg | aqueous solution interface for several
alkali metal halides as electrolytes at 258C using a calomel reference electrode (1 M KCl)
plotted against electrolyte concentration.
also reach a minimum. This is the principal method of determining the PZC at
solid electrodes.
Once the PZC is known, the capacity data may be integrated to obtain the
surface charge density:
E
sm
C dE
10:3:10
Ez
where Ez is the PZC on the chosen potential scale E. A complete analysis of

interfacial thermodynamic data also requires that the value of the interfacial
tension at the PZC be known. Then, the capacity data may be integrated twice
to obtain the interfacial tension as a function of cell potential:
E
g gm
C dE
10:3:11
Ez
gm is the value of g at the PZC or electrocapillary maximum. This means that gm

must be measured in an independent interfacial tension experiment at each electrolyte concentration at which the dierential capacity data were obtained.
Unfortunately, these data are often not available.
In the case of the study of the adsorption of molecular solutes at the polarizable interface, dierent tactics are often employed in data analysis. Suppose a
study of the adsorption of pyridine is carried out at the mercury j solution interface from an aqueous solution containing KCl. The Gibbs adsorption isotherm
for this system using a reference electrode reversible to the Cl anion is
dg sm dE W
dmKCl py dmpy W dmW
10:3:12
523
where py is the surface excess of pyridine, mpy , its chemical potential, and the
other symbols have their usual meaning. If capacitypotential data are obtained
at varying pyridine concentration (varying mpy ) holding the KCl activity constant,
it is observed that the capacity is independent of mpy at suciently negative or
positive potentials. This is a result of desorption of the pyridine molecules from
the interface when the electrical eld due to the charge on the electrode is suciently high. Choosing a potential in this region as a point of reference, one may
integrate the capacity data to obtain the change in interfacial tension g as
follows:
E
g g g r
C dE
10:3:13
Er
where gr is the interfacial tension at the point of reference and Er , the potential at
that point. More will be said about the analysis of data for molecular adsorption
in section 10.10.
At solid polarizable electrodes, the charge density on the electrode can be
measured directly using chronocoulometry [11]. This has been applied successfully
to the study of molecular adsorption at single crystal gold electrodes. As discussed
above, a reference potential Er is chosen at which molecular adsorption is absent.
The potential of the polarizable electrode is then stepped to a new value Ei and the
current transient which results is recorded. By integrating this transient, one
obtains the change in charge density between the two potentials:
sm sm Ei sm Er
10:3:14
where sm Ei is the charge density at the new potential and sm Er , that at the
reference potential. The potential of zero charge Ez is determined in an independent capacity experiment. By stepping the potential to the PZC, one can determine the absolute value of the charge density on the electrode at the reference
potential:
smz sm Ez sm Er sm Er
10:3:15
where smz is the change in charge density required to reach the PZC. These
experiments are carried out as a function of potential over the polarizable range of
the electrode, and as a function of the concentration of the molecular adsorbate,
for example, pyridine. The data may then be integrated to obtain the change in
interfacial tension with respect to the reference potential:
E
g g gr
sm dE
10:3:16
Er
More details about these experiments can be found in the work of Lipkowski [11].
Once a complete set of interfacial tension data have been assembled as a
function of electrode potential and solution composition, they may be dierentiated to obtain the relative surface excess. For example, if the solution contains a
single electrolyte and the reference electrode is reversible to the anion in the
524
electrolyte, the relative surface excess of the cation is obtained by estimating the
derivative in equation (10.2.25). In the case of molecular adsorption such as
pyridine, the surface excess is found from the relative change in the interfacial
tension at constant electrode potential and electrolyte activity (see equation
(10.3.12)):
!
@g
10:3:17
py
@mpy
E ;mKCl
Many studies of ionic and molecular adsorption have been carried out using the
techniques described here. They provided quantitative information regarding the
extent of adsorption. Detailed analysis of these data led to an improved picture of
the forces which control the extent of adsorption at polarizable interfaces. This
subject is considered in more detail in the discussion of ionic adsorption in section
10.9 and molecular adsorption in section 10.10.
B. Spectroscopic Methods
In recent years a rich variety of spectroscopic techniques has been developed to
study the properties of the metal j solution interface [1214]. These techniques fall
into all areas of traditional spectroscopy used to study the properties of bulk
phases, and involve absorption, emission, and scattering phenomena. The methods which are particularly useful for investigating the composition and structure
of the electrical double layer are considered in more detail here.
Vibrational spectroscopy provides an excellent tool for examining interfacial
properties. Experiments have been carried out using both the infrared and Raman
techniques [1214]. Discussion is limited here to Fourier transform infrared spectroscopy (FTIR) in the reection mode. It is important to understand how the
infrared radiation interacts with dipolar adsorbates at the interface. Consider an
electromagnetic wave travelling in the (x; z)-plane, which strikes the interface
located in the (x; y)-plane at an angle y with respect to the interface (see g.
10.7). The electrical eld vector associated with the wave can be resolved into
two components, one oscillating in the x; z-plane (the parallel or p-component)
Fig. 10.7 Sketch of the vector

dening the reection of
electromagnetic radiation in the
(x, z)-plane at a solution | metal
interface in the x, y)-plane.
Dipolar adsorbates in the
perpendicular (a) and parallel
orientations (b) are shown with
their corresponding electrostatic
images in the metal phase.
525
and the other, in the (y; z-plane (the perpendicular or s-component). Because the
substrate phase is a metal, electrostatic images are formed from the adsorbed
dipoles as a result of rearrangement of the mobile electron density within the
metal. When the dipole vector is perpendicular to the interface, its image points
in the same direction so that the eective dipole moment is larger. Dipoles in this
orientation interact with the parallel or p-component of the incident radiation.
When the frequency of the incident radiation corresponds to a vibration associated with the dipole, energy is absorbed which can be detected by examining the
reected radiation. On the other hand, if the dipole vector is parallel to the interface, the image dipole in the electrode acts in the opposite direction to that of the
molecular dipole, so that the net eect is zero dipole moment in the (x; y)-plane.
Thus, there is no interaction with the s-component of the electrical eld vector,
which oscillates perpendicular to the plane of incidence. It is clear that polarization of the infrared radiation incident on the interface is useful in determining the
orientation of adsorbate molecules. The fact that dipolar components perpendicular to the interface are infrared active and parallel components are not is known
as the surface selection rule [12, 13].
Because most solutions absorb infrared radiation in their bulk, the design of
the electrochemical cell is an important consideration in interfacial reection
experiments. The optimum conguration involves a very thin solution layer
which is a few microns thick and is sandwiched between the optical window
and the reective electrode (g. 10.8). In order to achieve maximum sensitivity,
the window has a hemispherical or triangular (prism) shape. Nevertheless, most
radiation is absorbed in the bulk of the solution and the eect of interfacially
adsorbed molecules cannot be seen unless special steps are taken. One procedure
is to polarize the light in a cyclical fashion between s- and p-polarized light. If
adsorbate molecules interact with the p-polarized light, the intensity of the
Fig. 10.8 Thin-layer electrochemical cell

used in SNIFTIR studies and reection
optics: (a) Teon cap; (b) N2 inlet; (c)
glass tube; (d) Teon cell body; (e) port
for reference electrode; (f) ceramic tube;
(g) Pt wire counter electrode; (h) single
crystal working electrode; (i)
hemispherical ZnSe window; (k) focal
plane for reection optics; (m) instrument
focal plane; and (n) folding mirrors
526
reected beam also oscillates, being lower when p-polarized light predominates.
The dierence between the signal observed in the presence of p-polarization with
respect to that observed in the presence of s-polarization gives the spectrum of the
interfacial layer. This experiment is known as infrared reection absorption spectroscopy, or IRRAS [15]. It is carried out with the polarizable electrode at constant potential. Another experimental strategy is to compare spectra observed at
two dierent potentials. One potential is chosen as a reference point at which
interfacial conditions are fairly simple, for example, at which adsorption is absent.
By subtracting the spectrum at the reference potential from the spectrum at
another potential, one determines the changes which have occurred in the interfacial region. When this experiment is carried out with an FTIR spectrometer it is
known as subtractively normalized FTIR spectroscopy, or SNIFTIRS [15].
SNIFTIR spectra obtained in a study of double layer structure on an Au
electrode in a non-aqueous electrolyte solution are shown in g. 10.9 [17]. The
electrolyte involved is Cr(DMSO)6 (ClO4 )3 in dimethylacetamide (DMA) as solvent. The cation in which DMSO is a ligand has a band at 924 cm1 due to the
SO vibration in the DMSO molecule. The perchlorate anion absorbs at
1100 cm1 . The reference spectrum was recorded at 0.16 V against the ferrocene
ferrocinium internal reference. Each of the dierence spectra shown in g. 10.9
was obtained by subtracting the reference spectrum from the spectrum observed
at the more negative potential. As the potential moves in the negative direction,
the band at 924 cm1 increases in intensity, demonstrating that the
[Cr(DMSO)6]3+ cation is attracted into the double layer. At the same time a
negative-going band is observed at 1100 cm1 ; this demonstrates that the ClO
4
anion is repelled from the double layer as the electrode potential moves negative
of the PZC. The most interesting feature of these spectra is the bipolar band
found in the region 16001640 cm1 . This is attributed to interfacial solvent
Fig. 10.9 SNIFTIR spectra obtained in the 1700700 cm1 region for a 2 mM [Cr(DMSO)6 ]
(ClO4 )3 solution in dimethylacetamide at an Au electrode. The reference potential was 0.16
V against an internal ferroceneferrocinium couple; spectra were recorded at more negative
potentials at intervals of 100 mV (From reference 17, with permission.)
527
molecules which undergo reorientation as the electrical eld in the double layer
changes sign. When the electrode is positively charged the DMA molecules are
adsorbed on the gold with the negative end of the molecular dipole pointing
toward the metal (g. 10.10(A)). Because of the interaction of the CO group
with the metal, the stretching frequency is red shifted and appears at 1605 cm1 .
As the charge on the electrode is made more negative the DMA molecules reorient
so that the molecular dipole is oriented with its negative end in the solution (g.
10.10(B)). Under these circumstances, the CO stretching frequency is closer
to 1640 cm1 . The reorientation of the interfacial DMA molecules results in a
bipolar band with a positive-going lobe at 1640 cm1 and a negative-going lobe
at 1605 cm1 .
SNIFTIR spectroscopy can also be used to obtain quantitative information
about double layer composition [18]. The surface excess of ClO
4 anion was
estimated for 0.1 M aqueous HClO4 solutions at single crystal Au electrodes as
a function of potential in the region around the PZC. The reference spectrum was
obtained at 0.69 V with respect to a Ag j Ag+ reference electrode, that is, well
positive of the PZC. The negative-going band observed at 1109 cm1 , which is
attributed to ClO
4 , was integrated to obtain an estimate of the change in surface
,
ClO4 , with electrode potential. Data were collected for
excess of ClO
4
Au(111), Au(110), and Au(100) electrodes. Since the PZC depends on the crystallographic orientation of the Au single crystal, the value of ClO4 depends on the
nature of the single crystal at constant electrode potential. However, when plotted
against electrode charge density ClO4 is independent of the nature of the Au
surface in the region 10 mC cm2 sm 10 mC cm2 (see g. 10.11). The integrated intensity may be related to the surface excess estimated by the Gouy
Chapman theory once the extinction coecient is determined. Excellent agreement between these quantities is obtained in the region around the PZC (g.
10.11). However, for charge densities greater than 10 mC cm2 , the integrated
intensity is considerably greater than the GouyChapman estimate. This discre-
Fig. 10.10 Schematic diagram indicating the orientation of dimethylacetamide at a positively charged polarizable electrode (A) and at a negatively charged electrode (B).
528
1
Fig. 10.11 Integrated intensity AR=R of the negative-going ClO
in
4 band at 1109 cm
aqueous 0.1 M HClO4 at Au(hkl) plotted against electrode charge density. The solid line
gives the GouyChapman estimate of the relative surface excess of ClO
4 . The data are for
Au(100) (*), Au (111) (!), and Au(110) (~). (From reference 18, with permission.)
pancy is attributed to specic adsorption of ClO

4 which replaces water molecules
on the electrode surface in this region.
X-ray methods provide an important route to determining interfacial structure
when the metal electrode is a single crystal [13, 14]. One technique, namely,
extended X-ray absorption ne structure, or EXAFS, depends on the absorption
of X-rays by the atoms in the interfacial region. The other method is surface Xray scattering or surface X-ray diraction. Both methods require intense X-ray
sources and therefore can only be carried out at a particle accelerator or synchrotron. The discussion here is limited to scattering experiments.
The power of the scattering technique in determining the details of adsorption at
a single crystal electrode is readily apparent for the case of monoatomic adsorbates,
for example, the adsorption of a monoatomic anion, or the underpotential deposition of one metal on another. Since dierent atoms have dierent sizes and interact
with one other in dierent ways, the adsorption phenomenon does not necessarily
involve one adsorbate atom A sitting on top of each metal atom M. However, if the
surface structure does involve atom A depositing on each atom M (see g.
10.12(A)), the adlayer is termed commensurate with a p(1 1) surface structure.
Another possibility is that the adsorbate atoms only occupy a fraction of the
available sites but they do so with a regular pattern. Figure 10.12(B) shows a
commensurate adlayer with a p(22) surface structure. Finally, the adlayer may
be deposited in a regular fashion but without following a pattern dictated by the
substrate metal; this is called an incommensurate adlayer (g. 10.12(C)). The structure of the adlayer with respect to the substrate is found experimentally by comparing the diraction pattern for the adsorption layer with that for the substrate.
529
Fig. 10.12 Three surface structures for an adlayer of atoms A (shaded) on a substrate of
metal atoms M (*): (a) a commensurate adlayer with a p1 1 structure; (b) a commensurate adlayer with a p2 2 structure; and (c) an incommensurate adlayer.
Surface scattering experiments must be carried out with the X-ray beam reaching the interface at grazing incidence (see g. 10.13). Typically the X-ray beam
makes an angle of less than 18 to the plane dening the electrode surface. If the Xray beam travels in the (x; z)-plane, scattering is chiey in the (x; y)-plane. The
solution layer must be very thin, because the path length through the solution is
large at small incident angles. Finally, the window material must be carefully
chosen in order to avoid signicant absorption by this medium.
X-ray surface diraction has been applied in situ to study several processes at
the electrode j solution interface [13, 14]. An important phenomenon in electrochemistry at Au is surface reconstruction in which a monolayer of atoms on the
surface of a single crystal acquires a dierent arrangement from that of the
Fig. 10.13 Thin layer electrochemical cell for X-ray scattering experiments.
530
atoms in the crystal. Surface reconstruction is best known on Au(100) and is

easily observed in ex-situ experiments. Using X-ray surface diraction, reconstruction of the Au(111) surface was observed at potentials suciently negative
of the PZC in the presence of p
specically
adsorbed Cs+ ions. In this region a
monolayer with a rectangular ( 3 22) structure was detected on the Au(111)

surface [13].
The specic adsorption of halide anions has been studied on Au and Ag single
crystals [14]. On Au(111), these ions form incommensurate hexagonal monolayers
that compress as the electrode potential is changed in the positive direction [19].
However, on Ag(100), Br adsorption occurs at the hollow site formed by four
metal atoms in a square pattern. This type of commensurate monolayer has a
c(2 2) surface structure. These studies demonstrate the role of atomic surface
structure in determining the extent of adsorption. Dierences between adsorption
on Ag(100) and Au(100) are explained in terms of dierences in the strengths of
the metalhalide bonds [14].
Another important area in which X-ray surface scattering is applied is the
underpotential deposition of metals. Underpotential deposition is the phenomenon by which one metal deposits on another at a potential positive of its normal
reduction potential. For example, Pb deposits at underpotentials on Ag. This is
due to the fact that the Gibbs energy for formation of a PbAg bond is less than
that for formation of a PbPb bond. Other metals which undergo underpotential
deposition on Ag, Au, and Pt are Tl and Bi. On the basis of the electrochemistry
observed in formation of the metal monolayers, there is good reason to expect
that they are well ordered. Tl, Pb, and Bi all form an incommensurate monolayers
on Au(111). On Au(100), Tl and Bi form an incommensurate monolayer with a
c(2 2) surface structure [14]. On the other hand, underpotential deposition of Pb
on Ag(111) leads to an incommensurate monolayer [13]. These studies demonstrate clearly that the nature of the monolayer formed depends on both the nature
and structure of the substrate metal.
As stated earlier, there are other spectroscopic tools which have been used to
study the structure and composition of the electrical double layer. However, it is
clear that detailed information can be obtained both at the atomic and molecular
level on the basis of the two techniques discussed here.
10.4 The Structure of the Double Layer

The development of microscopic models of the double layer began over 100 years
ago with work of Helmholtz [20]. He assumed that the charge on the polarizable
metal electrode is exactly compensated by a layer of ionic charge in solution
located at a constant distance from the geometrical electrode j solution interface.
The separation distance zd was assumed to have molecular dimensions. This
simple model which gave rise to the term double layer is the equivalent of a
parallel-plate capacitor with a capacitance given by
C
eH e0
zd
10:4:1
531
where eH is the relative permittivity of the Helmholtz layer. Considering the

diameter of a water molecule and adding to that the radius of a typical monoatomic ion, a reasonable estimate of zd is 400 pm. A typical capacity observed at
the Hg j aqueous solution interface is 20 mF cm2 . The resulting value of eH is 9.
This simple calculation shows that the properties of the solvent at the interface are
very dierent from those in the bulk of the solution where most polar solvents
have considerably higher relative permittivities.
The above model is na ve in so far as it assumes that all the charge on the
solution side is located on one plane. Although the Helmholtz model satises the
electrostatic requirement of charge neutrality at the interphase, it ignores the
randomizing eects of thermal motion. This was recognized independently by
Gouy [21] and Chapman [22] who proposed that the double layer involves a
diuse layer of charge extending a nite distance into the solution. As one
would expect, the average thickness of the diuse layer depends on electrode
potential and electrolyte concentration; as potential and concentration increase,
the diuse character of this region is reduced and the average thickness decreases.
The GouyChapman theory does not consider the nite size of the ionic constituents of the diuse layer. It follows that this model predicts that the capacity of
the double layer goes to innity as the eld due to the electrode charge increases in
the positive or negative direction. As pointed out by Stern [23], this problem can
be resolved by combining the Helmholtz model with the GouyChapman model
of the diuse layer. In this case one recognizes that the ionic constituents of the
diuse layer are limited in their approach to the interface by their size. The plane
of closest approach of the counter ion is located at distance xd from the geometrical boundary of the interface, and is called the outer Helmholtz plane (oHp). It
follows that the capacity of the double layer is composed of a series combination
of the capacity due to the Helmholtz or inner layer, Ci , and that due to the diuse
layer, Cd :
1
1
1

C Ci Cd
10:4:2
In the absence of ionic specic adsorption, the properties of the inner layer are
determined by those of the polarizable metal and of the solvent. Some small
inuence is also seen from the counter ion at the oHp. The latter eect is attributed to changes in the size of the counter ion, which results in a change in the total
thickness of the inner layer. At potentials well removed from the PZC, the diuse
layer capacity is quite large, so that it gives only a minor contribution to the
observed experimental capacity.
As noted above, ionic specic adsorption occurs when the total surface excess
of a given ion is greater than that which can be accounted for on the basis of
electrostatics. It arises because of the formation of a covalent bond between the
given ion and the atoms of the metal electrode, or because the ion is hydrophobic
and therefore accumulates at the interface. For these reasons it is assumed that a
specically adsorbed ion displaces solvent molecules from the monolayer of solvent solvating the electrode (see g. 10.14). Stern [23] treated specic adsorption
using the Langmuir adsorption isotherm, thereby emphasizing that the adsorbed
ions interact chemically with the polarizable metal. Grahame [G1] emphasized
532
that the specically adsorbed ions are often closer to the electrode than the
counter ions at the oHp. He introduced the term inner Helmholtz plane (iHp) to
refer to the location of the charge centers of these ions. This picture of the solution
side of the double layer is called the GouyChapmanSternGrahame model and
is shown schematically in g. 10.14.
From the time of Grahames famous review [G1], published in 1947, until the
early 1980s, the model of the double layer was based on the above picture.
However, during this period of time, experiments were carried out on polarizable
metals other than mercury which showed that the nature of the metal plays an
important role in determining double layer properties [G3]. Experiments on soft
polarizable metals such as bismuth, tin, cadmium, and zinc, on other liquid metals
such as gallium and various amalgams, and on single crystal electrodes such as
silver and gold showed that inner layer properties depend markedly on metal
nature. Moreover, in the case of single crystals, it depends on the crystallographic
orientation of the metal which is exposed to the solution. In addition, the PZC
depends on the nature and crystal character of the metal. This was investigated in
Fig. 10.14 Schematic diagram of the

double layer according to the Gouy
ChapmanSternGrahame model.
The metal electrode has a net
negative charge and solvated
monatomic cations dene the inner
boundary of the diuse layer at the
outer Helmholtz plane (oHp).
Monatomic anions are contact
adsorbed on the electrode at the
inner Helmholtz plane (iHp). Their
presence is stabilized by the
formation of images in the
conducting electode so that each
adsorbed ion establishes a dipole at
the interface.
533
detail by Trasatti [24], who emphasized the relationship between the PZC and the
work function of the metal.
On the basis of these studies, it was clear that the metal could not be treated
simply as a structureless perfect conductor. Both the spatial arrangement of the
surface atoms in the metal and their associated electron cloud play a role in
determining double layer properties. In 1983, the jellium model was introduced
to account for the eects of the metal at the polarizable interface [25, 26]. In its
simplest version, this model treats the metal lattice of positively charged ions as a
constant positive background charge. The electrons are modeled as an inhomogeneous electron gas which interacts with the positive background. At the surface
of the metal crystal, which is called the jellium edge, the positive background
drops abruptly to zero. The properties of the metal are determined by the electronic density in this model. This aects the electronic density prole at the surface,
where the electrons spill out of the atomic lattice over a distance of 100200 pm.
This leads to the surface potential wm at the metal surface, as discussed in section
8.5. The surface potential is an important contribution to the electronic work
function of the metal which depends on its electronic density in a complex way
[26].
When the metal comes in contact with an electrolyte solution, the extent of
electron spillover changes. Water molecules interact to some extent with the
electrophilic metal atom lattice through the electron density on the oxygen
atom. As a result, the electron overspill at the PZC is less than it is at the
metal j vacuum interface. Furthermore, the extent of this overspill changes
when the electrode is polarized to potentials away from the PZC. When the
electrode carries a positive charge, the electron overspill is reduced somewhat;
when the electrode is negatively charged, it is extended. If the surface potential of
the metal at the metal j solution interface is denoted as gm (see section 8.8), then it
is clear that this quantity varies with electrode charge density sm . As a result there
is a contribution to the interfacial capacity from the metal which is dened as
1
@gm
Cm @sm
10:4:3
Recalling that wm is dened for the movement of a test charge from vacuum into
the metal, and dening gm in the same way for movement of a test charge from
solution into the metal, gm is a positive quantity at the PZC. As sm increases from
negative values to positive ones, gm becomes smaller. As a result, Cm is a negative
quantity. Its role in the overall picture of the electrical double layer is to introduce
the polarizability of the metal s electronic cloud into the interfacial model. The net
capacity of the inner layer with consideration of the metal is now
1
1
1
Ci Cm Csolv
10:4:4
where Csolv is the contribution to the inner layer capacity from the solvent molecules. Since physically relevant capacities are always positive, the rst term on the
right-hand side of equation (10.4.4) must be smaller in magnitude than the second
one. As will be seen in the following sections of this chapter, the inner layer
capacity Ci is the dominant contribution to the experimental capacity C over
534
most of the polarizable range of the electrode. The metal contribution is such that
the observed capacity is higher than it would be at a structureless perfect conductor. The simple jellium model, in which the metal atom lattice is treated as a
constant positive background charge, does not account for eects observed at
single crystal surfaces. The jellium model may be extended to take into account
crystal structure by replacing the positive background charge with a lattice of a
pseudopotentials. At each lattice site the electrons are subjected to a potential
which is governed by the electronic repulsion from the atomic cores at short
distances, and by Coulombs law at large distances. The electronic density prole
within the crystal and its surface using a simple form of the pseudopotential
model is shown in g. 10.15. Within the crystal the electron density oscillates,
reaching maxima in the vicinity of a metal atom (ion). The electron overspill at the
surface depends on the surface plane which is exposed. Thus, the revised jellium
model is able to account for the dependence of the work function and PZC of
single crystal electrodes on the crystallographic orientation at the interface. The
introduction of pseudopotentials represents an important extension of the jellium
model.
It is clear from the above discussion that three aspects of the electrical double
layer must be considered in order to understand experimental observations and
double layer phenomena. The rst of these is the role of the metal and its inuence
on double layer properties. The second aspect concerns the inner layer or region
immediately next to the metal. In the simplest case, this region is occupied only by
solvent molecules. If adsorption is present, then some of these molecules are
replaced by ions or solute molecules. In many cases the inner layer plays a dominant role in determining interfacial capacity. Thus, considerable eort has been
expended to develop models for solvent structure in this region and adsorption.
Fig. 10.15 Distribution of electron density within a Ag lattice at the (111) surface on the
basis of the jellium model with pseudopotentials located at the metal atom cores as indicated by the arrows [26]. The broken vertical line shows the position of the metal surface.
535
Finally, the third aspect is the diuse layer. This has traditionally been treated by
the GouyChapman theory. Recently, attempts have been made to develop an
improved model of the diuse layer which is analytical and considers the eects of
nite ion size and change in dielectric properties. These aspects of the double layer
are now considered in more detail.
10.5 The Potential of Zero Charge and the Role of

the Metal
The PZC is an important point of reference in discussing the properties of a
polarizable interface. Its location depends on the nature of all of the components
which are at the interface, that is, on the metal, on the solvent used for the
electrolyte solution, and on the nature and concentration of the solute components in this solution. Its importance was rst pointed out by Frumkin [G3] who
was able to carry out the rst experiments at polarizable electrodes other than
mercury. He showed that there is a fairly simple relationship between the PZC
and the work function of the metal for a given solution composition and reference
electrode. In this section the relationship is derived and its consequences illustrated with experimental data. Then a model which describes the role of the
metal in interfacial properties, namely, the jellium model, is presented.
A. The Potential of Zero Charge and the Metallic Work
Function
In order to derive the relationship between the PZC and the work function, a cell
with the conguration
Cu j M j 0:1 M KCl; H2 O j AgCl j Ag j Cu 0
10:5:1
is considered, where M is a polarizable metal. The most studied metal in this

group is mercury but it also includes metals such as gold, tin, lead, and bismuth
within a dened potential range. Since the silver j silver chloride reference electrode is reversible to the anion of the electrolyte the potential of the cell is designated as E and is given by the Galvani potential dierence between the copper
wires at the terminals. Thus,
0
E Cu0 Cu f fCu fCu
10:5:2
This can be further decomposed into four Galvani potential dierences as follows:
0
E fCu fm fm fs fs fAg fAg fCu
10:5:3
The rst and the last of these may be re-expressed in terms of the standard
potentials of the electrons in the pure metal because of the electronic equilibria
which exist at the metal j metal interfaces. As a result,
!

mAg
mm
el
E s m f
s Ag f el
10:5:4
F
F
536
The rst term on the right-hand side relates to the properties of the polarizable
electrode and the second to those of the reference electrode. By adding the surface
potential of the solution ws to each term within the brackets, these terms may be
related to the absolute potentials for the electrodes (see section 9.4). Thus,
!

mAg
mm
el
s
s
el
w s Ag f
w
E s m f
F
F
10:5:5

mm
el
s
a
s m f
w EAg=AgCl
F
where E aAg=AgCl is the potential of the Ag j AgCl electrode on the absolute scale.
The focus of interest in this section is the Galvani potential drop s m f. As
discussed in section 8.8, this can be resolved into three contributions, so that
s m f
fm fs s f gm gs
10:5:6
s f is the potential drop due to the net free charge at the interface; g is the
dipolar potential due to the metal phase, more specically, to the electron overspill that occurs at the surface of the metal; nally, gs is the dipolar potential due
to the solution phase which arises because of the orientation of solvent molecules
at the interface due to their proximity to the metal, and because of the unequal
distances of closest approach of the cations and anions to the interface. gs is
dened in the opposite direction to gm because the concept of the dipolar potential
originates at the condensed phase j vacuum interface where the denition of the
potential drop is always from vacuum to the condensed phase. The dipolar potential gm arises for the same reasons as the surface potential wm at the metal j
vacuum interface. However, it is not the same because of the eect that the
proximity of the molecules and ions of the solution phase have on the electron
overspill.
At the PZC, the contribution s f is zero and equation (10.5.6) becomes
m
s m f0
s
gm
0 g0
10:5:7
where the subscript 0 indicates the values which apply at this point of reference.
Combining equations (10.5.5) with equation (10.5.7), one obtains the result
s
E0 gm
0 g0
mm
el
ws E aAg=AgCl
F
10:5:8
E0 is the PZC measured with respect to the silver j silver chloride reference
electrode.
The next step is to relate the PZC to the work function for metal M, namely,
Wm
el . The work function is given by
m
m
Wm
el mel Fw
10:5:9
where wm is the surface potential of the metal. Now combining equations (10.5.8)
and (10.5.9), one obtains
E0
Wm
el
m
s
s
a
gm
0 w g0 w E Ag=AgCl
F
10:5:10
537
This result shows that the relationship between the PZC and the work function is
simple only when the dipole contributions are independent of metal nature. The
contribution from the reference electrode is constant, provided the electrolyte
solution remains unchanged when the polarizable metal is changed. As will be
shown below, the dipole potential terms denitely depend on the nature of the
metal.
The PZC has been determined for a wide range of metals in dierent aqueous
electrolyte solutions. Examination of these data shows that it varies considerably
with both the nature of the metal and of the electrolyte. Values of this quantity
determined at the metal j water interface in the absence of ionic adsorption are
summarized in table 10.1. Considerable interest has been expressed in the relationship between the PZC and the work function of the metal, W m
el . When the SHE is
used as reference, equation (10.5.10) becomes
E0
Wm
el
m
s
s
a
gm
0 w g0 w E SHE
F
10:5:11
where E aSHE is the absolute potential of the SHE. The SHE is the most common
reference point used to cite experimental values of the PZC but in fact any convenient reference electrode may be used. In any case, the point of reference must
be cited.
A plot of Ez , that is, E0 against W m
el using data for the sp metals is shown in
g. 10.16. A non-linear correlation between these quantities is obtained, the value
of Ez becoming more positive with increase in W m
el . These results show clearly that
the terms describing the dipole potentials vary with the nature of the metal.
The sp metals have low melting points, some of them being liquids at room
temperature. Thus, the fact that the data are for the polycrystalline form is not
terribly important. If they all interacted with water to the same degree, the plot of
Ez against W m
el should be linear with unit slope. In other words, one would expect
the dipole potential terms to be constant. In fact, the slope of the plot in g. 10.16
Table 10.1 Point of Zero Charge with Respect to the
SHE for Various Metals in Water in the Absence of
Ionic Adsorption [27]
Metal
Ez / V
Metal
Ez / V
Hg
Other sp metals
Bi
Cd
Ga
In
Pb
Sb
Sn
Tl
Zn
0.192
Single crystal sd metals

Ag (100)
Ag (110)
Ag (111)
0.62
0.73
0.46
0.38
0.74
0.67
0.65
0.60
0.17
0.39
0.71
0.92
Au (100)
Au (110)
Au (111)
0.33
0.19
0.56
Cu (100)
Cu (110)
Cu (111)
0.54
0.69
0.20
538
Fig. 10.16 Plot of Ez , the point of zero charge for sp metals in an aqueous solutions with
respect to the SHE with a non-adsorbing electrolyte against W m
el , the vacuum work function of the metal. The dotted line indicates unit slope.
is close to unity for the metals with the largest work functions, namely, mercury
and tin. However, as the work function decreases the slope of the plot also
decreases. This has been discussed in terms of the hydrophilicity of the metal.
Thus, examination of double layer data has led to the conclusion that gallium is
more hydrophilic than mercury [24, 28]. This means that more water molecules
are adsorbed on the metal with the negative electron density on the oxygen atom
interacting with it. As a result, the dipole term gs becomes more negative as the
metal is changed in the series Hg < Ga < Zn. As a result, Ez becomes more
positive in the same series than it would be if gs were constant. The magnitude
of the eect can be judged by drawing a straight line with unit slope through the
point for Hg on the plot. Examination of the data shows that the eect is quite
large.
Other relevant data are available for single crystals of the sd metals, namely,
copper, gold, and silver. The data for the low-index faces of these systems show
interesting trends with crystallographic orientation and with metal nature. As the
density of atoms on the metal face increases both the work function and PZC
move in the positive direction. In addition, for a given crystallographic orientation both parameters increase in the order Cu < Ag < Au. These trends demonstrate a corresponding decrease in the hydrophilicity of the metal with a given
crystallographic orientation at the surface, and a corresponding decrease in the
magnitude of the dipole potential gs .
Some data are also available for the variation in the PZC with metal nature in
non-aqueous solvents [27]. These results are much more limited in terms of the
number of metals studied but they show the same trends that are found in the case
of water.
539
B. The Jellium Model for the Metal

The role of the metal in double layer properties can be understood in greater detail
when the system is examined on the basis of the jellium model. This model was
developed to describe the electron gas within sp metals. It can be used to estimate
several properties of interest, including the chemical potential of an electron in the
metal, the extent of electron overspill, and the work function of the metal. More
recently, it has been extended to describe metal surfaces in contact with polar
solvents [26]. In its simplest form, the metal atoms in the metal are modeled as a
uniform positive background for the electron gas, no consideration being given to
their discrete nature and position in the metal lattice. The most important property
of the system is the average electron density, hNel i, which depends on the number
of metal atoms per unit volume and the number of valence electrons per atom, nv .
Thus, if rm is the mass density of the metal, and Mm , its atomic mass
hNel i
N L nv r m
Mm
10:5:12
where NL is the Avogadro constant. Values of hNel i for some typical sp metals are
given in table 10.2. The electronic work function depends on both the electron
density and on the extent of electron overspill, which itself is related to the
electron density. This suggests that there should be a correlation between the
work function and the electron density. The correlation for the sp metals listed
in table 10.2, which is approximately linear, is shown in g. 10.17. The deviations
Table 10.2 Average Electron Density hNel i in Typical sp
Metals at 258C Together with the Number of Valence Electrons
nv , the Metallic Density rm , and the Pseudopotential Radius rc
Metal
nv
rm =g cm3
hNel i/electrons nm3
rc /pm
Li
K
Cs
1
1
1
0.534
0.869
1.879
46.3
13.7
8.5
70
117
146
Mg
Ca
Ba
2
2
2
1.74
1.54
3.51
86.2
46.3
30.8
69
92
107
Zn
Cd
Hg
2
2
2
7.14
8.64
13.59
131.5
92.6
81.6
57
67
69
Al
Ga
In
3
3
3
2.70
5.90
7.30
133.9
152.9
114.9
59
63
72
Sn
Pb
4
4
6.54
11.34
132.7
131.8
59
57
*Values of rc were taken mainly from S. Elliot, The Physics and Chemistry of
Solids, Wiley, New York (1998).
540
Fig. 10.17 Plot of the metallic work function W m

el against the average electron density hNel i
for sp metals.
from perfect linearity can be attributed to the fact that the simple jellium model
ignores the structure of the metal lattice.
The description of the metal is improved considerably if metallic structure is
introduced by accounting for the local attractive force of the metal atoms on the
free electron gas. This corresponds to the jellium model with pseudopotentials.
Each metal atom in the lattice is pictured as being surrounded by a spherical
volume Vc in which electrostatic eects may be ignored. Outside of the sphere
the metal atom behaves as a point charge of charge number nv . Thus, it has a
pseudopotential fc , where
fc
nv e0
4pe0 r
r rc
10:5:13
and rc is the radius corresponding to spherical volume Vc . The spacing of the

metal atoms in a given direction in the crystal is dm , and depends on both the
nature of the crystal lattice and on the direction. It follows that, when jellium with
pseudopotential is adopted, two additional parameters are introduced, namely, rc
and dm . The parameter dm may be calculated from the metallic density rm , provided that lattice structure and direction have been specied. The radius rc is
estimated on the basis of a quantum-mechanical calculation and is given in
table 10.2.
In the present discussion, attention is focused on the properties of the electron
gas at the surface of the metallic crystal. As far as the metal atoms are concerned
the lattice terminates sharply at the surface, which is called the jellium edge in this
model. The important feature of the electron cloud is that it spills out over the
edge of the lattice to an extent which is determined by lattice properties such as rc
and dm , and by electron gas properties such as hNel i. When the jellium model is
541
applied, one can determine Nel x, that is, the electron density as a function of
position in the crystal and at its edge (x 0). Typical results with pseudopotentials for the individual metal atoms included in the model are shown in g. 10.15.
The electron overspill at the jellium edge determines the surface potential wm of
the metal, and has an inuence on the metallic work function.
In order to determine the electronic density prole Nel x the surface energy of
the electronic plasma Us must be determined. This is written as a functional of the
density prole as follows:
Us Nel x Ubulk Uion Udip Upsp Urep
10:5:14
where Ubulk gives the bulk contributions to the energy, Uion , that due to electrostatic interaction of the plasma with ions in solution, Udip , the corresponding term
for solvent dipoles, Upsp , that portion of the surface energy which arises from the
introduction of pseudopotentials in the model, and Urep , the repulsion of the
electrons from the solution due to short-range forces [26]. Ubulk contains six
dierent contributions which may be estimated from the quantum-mechanical
description of metals [26, 29]. They are the kinetic energy, the exchange energy,
and the correlation energy of the plasma, its electrostatic self-interaction, its
interaction with the positive background charge, and the inhomogeneity contribution. Expressions for each of these terms have been given by Smith [29]. Uion ,
Udip and Urep are described by expressions given by Schmickler and Henderson
[30].
At this point one must choose a functional form for Nel x. This choice is based
on what is known about the physical description of the system. Several appropriate functions have been described in the literature, a choice which has met with
considerable success being the following [30]:

ajel sm ajel x
1
Nel x hNel i
e
hNel i
x < 0
e
2

0
10:5:15
ajel sm ajel x
1
hNel i
e
x > 0
2
e0
These functions are only applied to the potential region near the PZC. One new
parameter namely, ajel , which is the characteristic decay length for the jellium
plasma, is introduced in these functions. Now, the surface energy Us (equation
(10.5.14)) is minimized by introducing the functions (10.5.15) into the appropriate
terms and varying ajel . It turns out that ajel depends only weakly on hNel i and has
a value close to 23 nm1 for the jellium model without pseudopotentials. Now, the
various surface properties may be calculated, including the surface energy Us , the
surface potential wm , and the electronic work function Wm
el . Attention is focused
on wm in the following discussion.
According to the jellium model as described above, the surface potential at the
PZC is given by
wm
0
e0 hNel i
e0 a2jel
10:5:16
542
Thus, the surface potential is positive, and since ajel is approximately constant, it
is approximately proportional to the average electron density. In the case of
mercury for which hNel i is 81.6 electrons nm3 , wm
0 is equal to 2.79 V. It is positive
because it is measured from vacuum into the interior of the metal, which is
positive with respect to the electron cloud at the metal surface. When the metal
is charged, the electron overspill changes in response to the sign and magnitude of
the charge density on the metal. If the metal is positively charged, the extent of
electron overspills is reduced and wm is smaller in magnitude. On the other hand,
if the metal is negatively charged, the electron overspill extends further from the
metal and wm is larger.
In order to extend the above treatment to the metal j solution interface, one
must consider the eect of the solvent molecules adsorbed on the metal on the
electronic overspill. Because the solvent molecules are polarizable, an induced
dipole moment is established in the solvent monolayer, which acts to reduce the
extent of overspill. As a result, the dipolar potential due to the metal is reduced by
a factor corresponding to the optical permittivity of the monolayer, eop . Recalling
that this dipole potential is designated as gm , one has at the PZC
gm
0
e0 hNel i
eop e0 a2jel
10:5:17
In the case of water, eop is 1.776, so that gm

0 is equal to 1.57 V. In assessing this
result one should remember that the total surface potential has two opposing
contributions, namely, gm and gs : gs is also large and contains the contribution
from orientation of solvent dipoles in the monolayer at the interface.
By measuring the change in gm with charge density on the metal, the metals
contribution to the inner layer capacity can be estimated. Thus,
1
@gm e0 hNel i @ajel
Cm @sm
eop e0 a3jel @sm
10:5:18
Cm is negative, reecting the fact that gm decreases in magnitude as sm increases

from negative to positive values. Leiva and Schmickler have estimated Cm to be
47:6 mF cm2 at the PZC for the Hg j aqueous solution interface in the absence
of ionic adsorption [31].
The parameters discussed here have been estimated for a number of metals at
which double layer studies have been carried out [26, 30]. However, their signicance to the experimental data cannot be properly assessed without considering
also the contribution of the solvent monolayer to interfacial properties. This
subject is considered later in section 10.7 in the discussion of the solvents role
in the double layer.
10.6 The GouyChapman Model of the Diffuse

Double Layer
The model of the diuse layer which is still used today was developed by Gouy
[21] and Chapman [22] in the early part of the twentieth century. It is based on a
543
one-dimensional solution of the PoissonBoltzmann equation to yield the electrostatic and thermodynamic properties of the diuse layer. It preceded the much
better-known solution of the same equation by Debye and Huckel who used it to
develop the rst theory of electrolyte solutions thirteen years later (section 3.8).
The fundamental assumptions of the two theories are the same: the ions are
assumed to be point charges with no volume, and the ionic distribution is determined by the laws of electrostatics with consideration of the randomizing eects
of thermal motion. In addition, the solvent is considered to be a dielectric continuum with a uniform relative permittivity equal to that of the pure solvent, es.
The polarizable electrode is assumed to be a homogeneous conductor in which the
charge distribution is unaected by the presence of discrete ions in the adjacent
solution.
As was seen earlier, the Poisson equation, which is a combined statement of
Coulombs and Gauss laws, may be written as
rz
10:6:1
r2
es e0
where is the local potential at a given point in the medium, rz, the charge
density due to ions at that point, and e0, the permittivity of free space. In the
GouyChapman (GC) solution of this equation, the variation in inner potential is
only considered in a direction perpendicular to the interface. Variation in in the
(x; y)-planes which are parallel to the interface is ignored. This is equivalent to
considering the point charges in any given (x; y)-plane as being smeared out. It
also means that the local potential calculated at any distance z from the interface is the average potential fz estimated without consideration of the discrete
charge located on the corresponding (x; y)-plane. It follows that the Laplacian in
can be replaced by the derivative d2 fz =dz2 . Thus, equation (10.6.1) becomes
d2 fz
rz
2
es e0
dz
10:6:2
The Boltzmann part of the model involves development of an expression for

the charge density rz. For any ion i, its electrochemical potential in the diuse
layer must be equal to that in the bulk of the solution, that is,
~si
~zi m
m
10:6:3
~si ,
~zi is the electrochemical potential at position z in the diuse layer, and m
where m
that in the bulk of the solution. Since variation in potential is only considered in
~zi is
the z-direction, the expression for m
~zi mi RT ln czi zi Ffz
m
10:6:4
where mi is the standard chemical potential, czi , the concentration of ion i at
position z, fz the average potential at position z, and zi, the charge number of
ion i. The corresponding equation in the bulk of the solution is
~si mi RT ln csi zi Ffs
m
10:6:5
where csi is the bulk concentration of ion i and fs , the average potential in the
bulk. In the following discussion, fs is set equal to zero so that all potentials in the
544
double layer are referred to that in the bulk. It is emphasized that the assumptions
made in writing equations (10.6.4) and (10.6.5) are equivalent to neglecting the
eects of non-ideality due to the ionic atmosphere which is present on a given
plane in the diuse layer. In addition, by setting fs equal to zero, the eect of nonideality in the bulk of the electrolyte solution is neglected.
Combining equations (10.6.3)(10.6.5), one obtains the result
czi csi expzi ffz
10:6:6
where f F=RT. This is the Boltzmann equation for ion i in the diuse layer.
The charge density due to this ion at position z is
rzi z zi Fczi
10:6:7
The total charge density is obtained by adding up the contributions from each
type of ion:
X
X
X
rzi z
zi Fczi
zi Fcsi expzi ffz
10:6:8
rz
i
Finally, the PoissonBoltzmann equation for the diuse layer is

X zi Fcsi
d2 fz

exp zi ffz
e
e
dz2
s
0
i
10:6:9
In order to solve this dierential equation both sides are multiplied by 2 dfz/dz
to obtain

z
X zi Fcsi
dfz d2 fz
d df 2
z df
10:6:10
2
2
exp

z
ff
i
dz dz
dz dz2
es e0
dz
i
The equation to be solved is thus
z 2
X 2zi Fcsi
df
d

exp zi ffz dfz
dz
e
e
s
0
i
10:6:11
This equation can now be integrated from position z in the diuse layer to the
bulk of the solution. The quantity dfz =dz is the electrical eld in the diuse layer
perpendicular to the interface. This eld goes to zero in the bulk of the solution.
After integration, one obtains
z 2 X
df
2RTcsi
exp zi ffz 1
10:6:12
dz
e
e
s
0
i
This result can be used to estimate the eld at any position in the diuse layer. It is
the fundamental equation of the GC model and is used to derive all the equations
which follow.
In the case of a simple 11 electrolyte, equation (10.6.12) has a much simpler
form which is often seen in presentations of this model. If the bulk concentration
of the electrolyte is denoted as ce, one may write
z 2
df
2RTce
4RTce
exp ffz exp ffz 2

cosh ffz 1 10:6:13
dz
e0 es
e0 es
Using the identity

cosh ffz 1 2 sinh2
z
ff
2
the expression for the eld becomes

z
z
dfz
8RTce 1=2
ff
2k
ff

sinh
sinh
dz
e 0 es
2
2
f
545
10:6:14
10:6:15
where k is the DebyeHuckel reciprocal distance introduced in section 3.8. It

should be noted that equation (10.6.13) is a quadratic equation and therefore
has two roots. Only the negative root is physically admissible. Thus, equation
(10.6.15) shows that when the potential fz is positive, the eld dfz/dz is negative.
Equations (10.6.12) and (10.6.15) are now used to derive the other important
results from the GC model.
A. The Potential Drop Across the Diffuse Layer
The GC theory is used most often to estimate the potential drop across the diuse
layer. This quantity is important in colloid phenomena and electrode kinetics, for
example. Noting that the inner boundary of the diuse layer is located at distance
zd from the interface, the square of the electrical eld at this location is given by
z 2 X
i
df
2RTcsi h
exp zi ffd 1
10:6:16
dz d
es e0
i
where fd is the potential drop across the diuse layer. This is a non-linear equation in fd which can be solved by iterative methods, provided that the value of the
eld at the inner boundary, (dfz/dz)d, is known. The eld is not a quantity which
is obtained experimentally so that fd cannot be calculated without further
analysis.
In the absence of specic ion adsorption the potential drop across the diuse
layer may be related to the charge density on the electrode sm by means of Gauss
law. For the present system the relationship is
z
df
sm e0 es
10:6:17
dz d
It follows that the electrode charge density is related to the potential drop by the
equation
i
X h
s2m
2GC
ci expzi ffd 1
10:6:18
i
where
GC 2RTe0 es 1=2
10:6:19
GC is called the GouyChapman constant.

In the case of a simple 11 electrolyte at a concentration of ce M, equation
(10.6.18) simplies to
546
s2m 2
2GC ce cosh ffd 1
or using equation (10.6.14)
ffd
sm 2AGC sinh
2
10:6:20
!
10:6:21
where AGC
GC ce1/ 2. If the concentration ce is expressed in M,
GC is equal to
5.8687 mC cm2 in water at 25 C. This is the well-known result relating the electrode charge density to the potential drop across the diuse layer for a 11
electrolyte in the absence of specic ionic adsorption. Since fd has the same
sign as sm, only the positive root of equation (10.6.20) is physically admissible.
An analytical equation may also be written for fd as a function of sm:
2
!1=2 3
2RT 4 sm
s2m
d
5
ln
f
10:6:22
1
2AGC
F
4A2GC
EXAMPLE
Estimate the potential drop across the diuse layer for an electrode charge
density of 10 mC cm2 and a 11 electrolyte concentration of 0.1 M at 25 C.
The GC constant AGC is given by
AGC 5:8687 0:11=2 1:856 mC cm2
10:6:23
The ratio sm / (2AGC) is

sm
10
2:694
2AGC 2 1:856
10:6:24
From equation (10.6.22) one obtains

fd 2 0:0257 ln2:694 2:6942 11=2 0:0883 V
10:6:25
Thus the potential drop across the diuse layer is 88.3 mV.
Plots of the potential drop across the diuse layer are shown as a function of
electrode charge density in g. 10.18 for a 11 electrolyte with concentrations in
the range 0.011 M. As electrolyte concentration increases, the absolute value of
fd decreases for constant charge density sm. This demonstrates that the screening
ability of the electrolyte increases with the increasing concentration.
The GC results are compared in g. 10.18 with Monte Carlo calculations of
Boda et al. [32]. These were carried out assuming that the electrolyte ions are hard
spheres with a diameter of 300 pm in a dielectric continuum. The estimates of fd
using the Monte Carlo technique fall below the GC estimates. They demonstrate
the importance of including nite ion size in a model of the diuse layer.
B. The Differential Capacity of the Diffuse Layer
The diuse layer capacity is an important quantity often used in the analysis of
double layer capacity data. In the absence of specic adsorption, the diuse layer
capacity is dened as
547
Fig. 10.18 Plots of the potential drop across the diuse layer fd against the electrode
charge density for a 11 electrolyte at the concentrations indicated. The lled points
show the results of Monte Carlo calculations assuming an ionic diameter of 300 pm [32].
Cd
dsm
dfd
10:6:26
On the basis of equation (10.6.18), the diuse layer capacity is equal to

Cd
2GC X
z fc expzi ffd
2sm i i i
10:6:27
For a 11 electrolyte a much simpler expression may be derived on the basis of

equation (10.6.21),
!
ffd
Cd fAGC cosh
10:6:28
2
Using the identity
cosh
!
!
d
ffd
2 ff
sinh
1
2
2
10:6:29
and equation (10.6.21), the diuse layer capacity may also be written as
!1=2
s2m
10:6:30
Cd fAGC 1 2
4AGC
This result demonstrates that Cd increases with electrode charge density but that it
is approximately independent of electrolyte concentration when sm is large.
EXAMPLE
Estimate the diuse layer capacity for an electrode charge density of

10 mC cm2 and a 11 electrolyte concentration of 0.1 M at 25 C. Repeat the
calculation at the PZC.
548
From the earlier example, under these conditions AGC is equal to

1.856 mC cm2 and sm / (2AGC), to 2.694.
It follows from equation (10.6.30) that
Cd 38:92 1:8561 7:2581=2 207:6 mF cm2
10:6:31
When sm is equal to zero, the result is

Cd 38:92 1:856 72:24 mF cm2
10:6:32
These results demonstrate that Cd is usually much greater than the experimentally observed capacity. Only when the concentration is very low so that the
GC constant AGC is also low, does Cd reach values which are less than the
experimentally observed capacity.
Plots of the diuse layer capacity Cd against electrode charge density are shown
in g. 10.19. At low concentrations, Cd rises sharply from its minimum value at
the PZC and reaches values which are more than 10 times the experimentally
observed capacity. Cd is low and close to the experimentally observed capacity
only in the region of the PZC. Thus, in very dilute solutions the position of the
minimum in the experimental capacity can be used to locate the PZC if ionic
adsorption is absent.
Using estimates of Cd based on equation (10.6.27) or (10.6.28), the experimental capacity C may be corrected for the diuse layer contribution to obtain the
capacity of the inner layer Ci. Since Cd is usually much greater than C, errors in
the GC estimate of Cd are not important except in the vicinity of the PZC, where
Cd falls to its lowest values. When the experimental capacity is obtained at constant electrode charge density and as a function of electrolyte concentration, the
GC estimates of Cd are often used to determine whether ionic specic adsorption
Fig. 10.19 Plots of the diuse layer capacity Cd according to the GouyChapman model
gainst the electrode charge density sm for a 11 electrolyte at the concentrations indicated.
549
is present using equation (10.4.2). In this test, the experimental values of 1/C are
plotted against the GC estimates of 1/Cd. If a plot with unit slope is obtained, it is
concluded that the inner layer capacity Ci is independent of electrolyte concentration, and that ionic adsorption is absent. The value of Ci may be found from the
intercept. These plots are known as ParsonsZobel plots. Another application of
the same plot is to determine surface roughness of a solid metal electrode. With
these systems the apparent surface area is used to estimate the specic dierential
capacity C. This quantity may dier from the true surface area because of surface
roughness. In this case, the slope of the ParsonsZobel plot is not unity, but equal
to a value which reects the ratio of the true surface area to its apparent value.
Examination of experimental data for liquid metals such as mercury reveals
that errors in the GC estimates of Cd are not important near the PZC. However,
this simple model certainly overestimates Cd further from the PZC. According to
Monte Carlo calculations [32], Cd reaches a maximum and then becomes negative
when sm is suciently large. Studies of the diuse layer capacity using more
detailed models are the subject of current research.
C. The Potential Prole in the Diffuse Layer
The potential in the diuse layer is determined as a function of distance from the
oHp by integrating the eld as given by equation (10.6.12) or (10.6.15). An
analytical expression is only obtained for the case of symmetrical electrolytes.
The derivation presented here is for 11 electrolytes for which the result obtained
below is most often applied.
On the basis of equation (10.6.15) the dierential equation to be solved to
obtain the potential prole in the diuse layer for a 11 electrolyte is

z
ff
2k
z
csch
dz
10:6:33
df
f
2
The solution to this equation is
z
f
z
z
ff
2k
z
csch
df
dz
f
2
fd
10:6:34
zd
where the lower limit for integration is the oHp, which is located at a distance zd
from the geometrical metal j solution interface. Integration gives
!
z
ff
ffd
ln tanh
10:6:35
ln tanh
kz zd
4
4
This result may also be expressed as
!
z
ff
ffd
tanh
tanh
expkz zd
4
4
10:6:36
550
When the argument of the tanh function is small, this function may be replaced by
the argument. Under these circumstances the potential prole is given by

4RT
fd
fz fd expkz zd
10:6:37
F
It follows that fd decays exponentially with distance from the oHp under these
conditions.
The potential prole is the least reliable feature of the GC model. Certainly,
Monte Carlo calculations in which the ions are represented as charged hard
spheres in a dielectric continuum show that the GC potential prole is seriously
in error at high electrolyte concentrations. However, it is sometimes used at very
low concentrations to obtain an approximate idea of potential variation in the
diuse layer.
D. The Ionic Surface Excesses in the Diffuse Layer

In order to estimate the surface excess of any ion in the diuse layer its concentration prole must be integrated in the diuse layer. Thus, a general denition of
the surface excess is
1
i
czi
csi dz
zd
expzi ffz 1dz
csi
10:6:38
zd
In order to integrate this equation, one must make use of the general relationship
for the eld in the diuse layer; that is, equation (10.6.12). In the following
derivations only 11 electrolytes are considered, so that the expression used for
the eld is equation (10.6.15).
The surface excess of the monovalent cation in a 11 electrolyte is given by
fce

2k
0
fd
expffz 1 z
df
sinh ffz =2
10:6:39
Factoring out exp(f fz =2) in the numerator, one obtains

fc
e
k
0
f
"
!
#

ffz
2ce
ffd
z
exp
exp
df
1
2
k
2
10:6:40
This is usually expressed in terms of sd , the excess charge in the diuse layer due
to cations. Thus,
sd
"
!
#
"
!
#
2Fce
ffd
ffd
exp
F
1 AGC exp
1
k
2
2
10:6:41
551
It is easily shown in a similar way for the monovalent anion that

sd
"
!
#
"
!
#
2Fce
ffd
ffd
exp
F
1 AGC exp
1
k
2
2
10:6:42
Plots of sd and sd against electrode charge density are shown in g. 10.20.
The surface excess of the ion whose charge is opposite to that on the electrode is
large and somewhat smaller in magnitude than that of the electrode charge. As
electrolyte concentration increases, the magnitude of this surface excess decreases.
On the other hand, the surface excess of the ion with the same sign as that of the
charge on the electrode is small in magnitude and approximately constant for low
electrolyte concentration and high electrode charge densities. It has the same sign
as the excess for the predominant ion, indicating that it is actually in decit in the
diuse layer. Thus, a negatively charged electrode causes an excess of cations and
a decit of anions in the diuse layer. As electrolyte concentration increases, the
magnitude of this decit increases.
In summary, since the surface excesses are obtained by integrating over the
whole diuse layer, the GC estimates are assumed to give reasonable values for
these quantities. The estimates are often used in the analysis of double layer data
involving ionic adsorption in the inner layer.
Considerable eort has been made in recent years to improve the GC model.
Early work [33] was carried out at the primitive level with the solvent represented
as a dielectric continuum and the ions as hard spheres. The integral equation
approach was one method applied to this problem. This work was followed by
Monte Carlo studies [32]. The general result of these studies is that the GC model
overestimates the magnitude of the diuse layer potential drop (see g. 10.18).
Fig. 10.20 Plots of the ionic surfaces excesses sd (*) and sd (^) in the diuse layer
against electrode charge density sm for a negatively charged electode. Data are shown for a
11 electrolyte at 0.1 and 1 M.
552
10.7 The Structure of the Inner Layer in the Absence

of Adsorption
When adsorption is absent, the inner layer is populated by solvent molecules only.
The change in number and orientation of these molecules with electrode potential
or electrode charge density is considered to be the main reason for the experimental
variation in inner layer capacity, Ci. The rst system to be examined extensively in
the laboratory was the Hg / aqueous solution interface with NaF as electrolyte [34].
At lower temperatures, the inner layer capacity curve possesses a maximum at a
small charge density just positive of the PZC (g. 10.21). Minima are observed at
sm 10 mF cm2 and at 11 mF cm2. At the extremes of polarization, the inner
layer capacity rises. As the temperature increases, the central maximum disappears, and at suciently high temperature the maximum and minimum at positive
potentials are no longer apparent. This behavior has been interpreted as an indication of solvent reorientation, the maximum on the capacity curve corresponding to
the charge density at which net polarization in the solvent layer is zero. As the
electrode is made more negative or positive, the solvent dipoles tend to orient with
their positive or negative ends, respectively, toward the metal. This leads to a drop
in inner layer capacity. The increase in capacity at the extremes of polarization can
be attributed to ionic specic adsorption.
Signicant changes in inner layer behavior are observed when water is replaced
by a non-aqueous solvent in which electrolytes are soluble. In the case of the Hg |
methanol interface, a deep minimum is observed on the inner layer capacity curve
at negative charge densities [35]. In the case of the amide solvents [35], a variety of
behaviors is observed. The protic solvents, formamide, and N-methyl formamide,
possess capacity maxima at negative charge densities (g. 10.22). The other fea-
Fig. 10.21 Inner layer capacity against electrode charge density for the Hg | aqueous
solution interface in the presence of NaF for temperatures in the range 085 C: (*) 0 C;
(~) 25 C; (!) 45 C; (^) 65 C; (g) 85 C [34].
553
Fig. 10.22 Dierential capacity of the inner layer against electrode charge density for the
Hg | formamide (KPF6 as electrolyte), Hg | N-methylformamide (KClO4 as electrolyte) and
Hg | dimethylformamide interfaces (KClO4) at 25 C [35].
tures of these curves are similar to those found for water; minima are found on
either side of the maximum and the capacity increases at the extremes of polarization. On the other hand, the capacity curve for the aprotic solvent, dimethylformamide is relatively featureless. At positive charge densities, a maximum is
observed; then, the capacity falls to very low values at negative charge, where
there is a shallow minimum. This type of capacity curve is found for a great
variety of aprotic solvents, including propylene and ethylene carbonate, acetonitrile, acetone, dimethylsulfoxide, and sulfolane [35].
Relatively simple models in which the solvent molecules are represented as
hard spheres with embedded dipoles were developed after capacity data for the
metal | solution interface became available in a variety of solvents. More recently
the MSA has also been applied at the non-primitive level to explain inner layer
properties, especially near the PZC. More details about these models are presented below.
A. Solvent Monolayer Models for the Inner Layer
The simplest model of solvent structure in the inner layer is that proposed by
Watts-Tobin [36]. In this treatment, the inner layer is assumed to consist of a
monolayer of solvent dipoles represented as hard spheres; the dipoles may assume
one of two orientations, that is, with the electrode eld or against it (g. 10.23). By
estimating the relative concentrations of the two orientations, the potential drop
across the monolayer, and its dierential capacity may be found as a function of
electrode charge density. In order to calculate the electrochemical potential of a
dipolar molecule, one must determine the local eld at the position of the molecule. In general, the local eld Ee is given by the sum of the eld due to the charge
554
Fig. 10.23 Models of the solvent

monolayer at the electrode | solution
interface according to (a) the two-state
model with spherical solvent dipoles in
either the up (") or down
orientations (#); (b) the three-state model
in which a third state with solvent dipoles
parallel (!) to the interface has been
added; (c) the cluster model with clusters
in the up (%&%) and down (&%&)
orientations, and single molecules up
(") and down (#).
on the electrode and the reaction eld due to surrounding dipoles; in order to
calculate the later contribution, dipoledipole interactions must be estimated in
detail. A simpler procedure is to relate the local eld to the total potential drop
across the monolayer. Accordingly, Ee is equal to
Ee d
m f
zd
10:7:1
where dmf is the potential drop across the inner layer and zd, its thickness which
corresponds to the diameter of the spherical solvent molecules. The electrochemical potential of a molecule in the up orientation with its positive end toward
the electrode may be written
m~ u m0 Ur u kB T ln Nu pEe
10:7:2
where m0 is the standard chemical potential, Ur u, the energy of non-electrostatic

interaction between the up dipole and the metal, Nu, the number of up
dipoles per unit area, and p, the magnitude of the dipole. Equation (10.7.2) is
written in molecular rather than molar units. The corresponding equation for the
down dipoles is
m~ d m0 Ur d kB T ln Nd pEe
10:7:3
~d and the relative amounts of the two species is obtained

~u m
At equilibrium, m
from the relation

kB T ln
Nu
Ur d Ur u 2pEe
Nd
2p f
Ur d Ur u d m
zd
The partition function for the monolayer is dened as

X
ai expbpi Ee au expbpEe ad expbpEe
qm
555
10:7:4
10:7:5
where ai expbUr i . If pi is the component of the dipole moment in a direction

perpendicular to the interface from the solution toward the metal, then the concentration of either species is given by
Ni NT ai expbpi Ee =qm
10:7:6
NT is the total number of solvent dipoles per unit area in the monolayer. In the
case of hexagonally close-packed spheres, it is given by the relationship
2
NT p 2 Nu Nd
3z d
10:7:7
The potential drop across the monolayer is composed of a contribution due to

the charge on the electrode and one due to the net polarization of the dipoles. In
the Watts-Tobin model, this is written as
d m f
sm zd pNu Nd
ei e0
ei e0
10:7:8
where ei is the relative permittivity of the inner layer and e0, the permittivity of
free space. The parameter ei is introduced in order to account for distortional
polarization due to molecular polarizability. Rearranging equation (10.7.8), the
charge density may be expressed as a function of dmf:
sm
ei e0
d ln qm
d m f kB TNT
zd
dd m f
10:7:9
where
d ln qm
p
a expbpEe ad expbpEe
dd m f qm zd kB T u
10:7:10
Dierentiating equation (10.7.9) with respect to dmf, one obtains an equation

for the solvent contribution to the inner layer capacity :
Csolv
ei e0
d2 ln qm
kB TNT
zd
dd m f2
10:7:11
where

d2 ln qm
p2
d ln qm 2

dd m f
dd m f2 z2d k2B T2
10:7:12
Csolv is a maximum when d ln qm/d (dm f) 0, that is when Nu Nd. On the

basis of equation (10.7.4), this condition is met when dmf zd (Ur d Ur u)/2p. A
556
plot of Csolv against sm for typical values of the parameters dened in this model
is shown in g. 10.24. Since only one extremum, the maximum, is predicted, the
model is not capable of accounting for all the features on the experimental curves
(gs. 10.21 and 10.22). However, the simple two-state model provides a reasonable explanation for the maximum found experimentally for most systems provided the other extrema are attributed to physical phenomena other than solvent
reorientation.
A more reasonable description of a solvent monolayer at the interface should
include solvent dipoles in all possible orientations. However, if one is only interested in the potential drop due to the monolayer in a direction perpendicular to
the interface, and chemical interactions are ignored, it would be sucient to
consider three states, namely, those which result from vector resolution of all
possible orientations. These orientations are the two considered above, up
and down, and a third orientation with the dipole vector parallel to the interface. The three-state model (g. 10.23) has been applied to describe the structure
of aprotic solvents at charged interfaces [37]. If one extends the two-state model of
Watts-Tobin, the electrochemical potential of the third state is given by
m~ p m0 Urp kB T ln Np
10:7:13
where Urp is the chemical (non-electrostatic) interaction between the parallel

dipole and the metal, and Np , the number of parallel dipoles per unit area. By
combining this equation with either equation (10.7.2) or (10.7.3), the relation
between the concentration of parallel dipoles, and up or down dipoles,
respectively, may be obtained. The condition for total coverage of the surface
becomes
Fig. 10.24 A plot of inner layer capacity due to the solvent, Csolv against electrode charge
density, sm according to the two-state model (a) and three-state model (b). The values of
the parameters for the two-state model are zd 0.48 nm, p 3.0 debye (1.0 1029 C m), ei
9.0, T 298 K, Ur1 5kT, Ur2 0; the parameters are the same for the three-state
model except that zd 0.30 nm and Ur3 7kT.
NT Nu Nd Np
It follows that the partition function of the monolayer is given by
X
qm
ai expb pi Ee au expb pEe ad expb pEe ap
557
10:7:14
10:7:15
The equations for the potential drop dmf and inner layer capacity Ci are the
same as those for the two-state model (equations (10.7.8) and (10.7.11)). If the
fraction of the monolayer composed of dipoles in the parallel orientation is
relatively large, the shape of the inner layer capacity curve changes. This follows
from the fact that the numerical values of the partition function qm and the
derivative d2 ln qm/[d(dm f)]2 are quite dierent when the third component is
included in the model. The latter quantity is now given by the equation

d2 ln qm
2 p2
d ln qm 2
a
exppE
a
exppE

10:7:16
u
e
d
e
dd m f
z2d
dd m f2
where
au expbpEe ad expbpEe qm
10:7:17
It follows that d2ln qm/[d(dm f)]2 is a minimum when dln qm/d(dm f) 0, that
is, when Nu Nd. The inner layer capacity curve calculated with the parameters
chosen previously for the two-state system and with a low value of Urp is also
shown in g. 10.24. As predicted, a minimum occurs at the position of the maximum on the curve for the two-state system. At charge densities suciently far
from the minimum, maxima are observed. The three-state model is able to
account for inner layer capacity curves in a variety of solvents such as methanol,
ethylene carbonate, and dimethylformamide [35].
In the case of strongly structured solvents such as water, formamide, and Nmethyl formamide, an improved description of inner layer solvent structure is
obtained when it is assumed the clusters of solvent molecules are present at the
interface. A cluster is assumed to contain two or more molecules held together by
hydrogen bonding such that the net dipole moment of the unit adsorbed at the
interface is signicantly less than that of a monomer in the direction perpendicular
to the interface. The clusters are assumed to have two orientations, one with the
net dipole vector pointing toward the metal (up) and one with the net dipole
vector in the opposite direction (down). Furthermore, they predominate in the
inner layer at low electrical elds. The central maximum on the inner layer capacity curve occurs approximately at the charge density where the populations of
clusters in the up and down orientations are equal. As the charge density is
made more positive or negative, the concentration of down or up clusters,
respectively increase, and the inner layer capacity decreases. Considering the fact
that the electrochemical potential of a monomer changes much more with electrode charge density than that of a cluster, the concentration of monomers in a
given orientation increases, with respect to that of the clusters as the charge
increases. This leads eventually to an increase in capacity at the polarization
extrema. Thus, capacity curves for strongly structured solvents are characterized
by a maximum and two minima (see gs. 10.21 and 10.22). It has been shown
558
[3840] that cluster models provide reasonable descriptions of the inner layer
capacity curves for water, formamide, and N-methylformamide. The development
of the theory with inclusion of dipoledipole interactions is rather complex, since
species of dierent size populate the monolayer. The interested reader is referred
to the original literature [3840] for further details of these models.
It should be noted that the method of dealing with dipoledipole interactions in
the monolayer presented here is that originally due to Watts-Tobin [36]. The chief
criticism that one can make of the Watts-Tobin model is that it introduces an
eective relative permittivity for the monolayer, ei, to account for the eects of
distortional polarization. In a completely molecular description, these eects can
be dealt with by considering the polarizability of the solvent molecules [37, 41].
More detailed models for the role of hydrogen bonding at the interface have also
been described [42].
10.8 The Specic Adsorption of Ions

An ion is considered to be specically adsorbed at a polarizable electrode when its
concentration in the double layer is greater than one would expect on the basis of
electrostatic considerations only. Pragmatically speaking, this denition of specic adsorption implies that the surface excess of the given ion is greater than that
predicted by diuse layer theory for a given electrode charge density and bulk
concentration. Up to the present, all experimental data have been analyzed
assuming the GouyChapman model for the diuse layer. On the basis of data
in the literature, most anions are specically adsorbed at Hg and other polarizable
metals from aqueous solutions. On the other hand, most inorganic cations are not
adsorbed, or else, they are weakly adsorbed at large negative charge densities.
In the following discussion some important features of the experimental
results for anion adsorption are presented. Emphasis is placed on data obtained
at mercury electrodes in aqueous solutions. This simply reects the fact that a
large fraction of the existing data were obtained for these systems. This is
followed by a presentation of the theory applied to ion adsorption at polarizable
interfaces.
A. Experimental Data for Anion Adsorption
The presence of ionic specic adsorption may be conrmed by examining the
dependence of the PZC on electrolyte concentration. From the data presented
in g. 10.6, it is seen that the PZC measured with respect to a constant reference
electrode is approximately independent of electrolyte concentration when ionic
adsorption is absent. The classic example of such a system is NaF in water at an
Hg electrode. However, when anion adsorption occurs the PZC shifts in the
negative direction with increase in electrolyte concentration. This is most pronounced for the I anion in the case of the halides. When cations are adsorbed,
the PZC shifts in the positive direction. The interpretation of the shift of the PZC
with electrolyte concentration can be put on a sound thermodynamic basis when
the electrode potential is measured with a reference electrode reversible to one of
559
the ions in solution (E or E potential scale). Details of the exact thermodynamic

analysis have been given by Delahay [G2].
In order to determine the extent of ionic specic adsorption, several assumptions must be made in analyzing thermodynamic data for the polarizable interface. If anions are adsorbed, it is customary to assume that cations are not
adsorbed in the same potential range. Furthermore, it is assumed that the surface
excess of cations is equal to that estimated by GouyChapman theory. In order to
illustrate the method of data analysis, the cell
Hg j x KI in H2 O j AgI j Ag
10:8:1
is considered. Since the reference electrode is reversible to the anion of the electrolyte, the Gibbs adsorption isotherm is given by
dg sm dE dme
10:8:2
It is often more convenient to analyze the data at constant electrode charge

density sm rather than constant potential E . Introducing the Parsons function
x g smE, one may write
dx dg sm dE E dsm E dsm dme
10:8:3
Noting that dme RT dln ae , the charge density due to cations is given by

@x
10:8:4
s F f
@ ln ae sm
Experimentally, one measures the interfacial tension, g and surface charge density, sm as functions of electrode potential E and electrolyte activity ae. This is
most commonly done by measuring the dierential capacity C as a function of E
and ae . The values of the potential of zero charge and interfacial tension at this
potential are also required, so that one may integrate C, once to obtain sm , and a
second time to obtain g. When interfacial tension data are not available, the
second integration constant may be estimated in a potential region where ionic
adsorption is absent using the GouyChapman theory. Parsons function x is
then calculated for constant sm as a function of salt activity. Numerical dierentiation of these data according to equation (10.8.4) allows one to estimate s.
On the basis of the electroneutrality condition, one may also calculate s
(s sm s .
As noted above, further analysis of the data requires the assumption that the
K cations are not specically adsorbed. Then, the surface excess of cations is
located in the diuse layer and equal to that predicted by GouyChapman theory,
sd . Combining equations (10.6.41) and (10.6.42), it is easily shown that the
surface excess of anions in the diuse layer is given by
sd
AGC sd
sd AGC
10:8:5
The charge due to specically adsorbed anions is equal to

sad s sd sm sm sd
AGC sd
sd AGC
10:8:6
560
In the presence of specic adsorption, the potential drop across the diuse layer is
given by
2
!1=2 3
2
2RT
s
s
s
ad
m
ad
5
ln4 m
10:8:7
fd
1
F
2AGC
4A2GC
The extent of ionic adsorption may be classied as strong, moderate, or weak.
As the electrode potential is made more positive, the charge on the electrode
increases in the positive direction. At the same time, the surface excess of
adsorbed anions increases. If the sum sm sad is always negative for electrolyte
concentrations greater than 0.01 M, so that the potential drop across the diuse
layer is also negative, then the ionic adsorption is strong. Examples of anions
which are strongly adsorbed on Hg are I , CN , and CNS . When the sum sm
sad changes sign from positive to negative with increase in electrolyte concentration above 0.01 M for electrode charge densities positive of the PZC, then anion
adsorption is moderate. Good examples of moderately adsorbed anions are NO
3
and Cl . In the case of weak adsorption, the potential drop across the diuse
layer remains positive for positive electrode charge densities. This is the most
dicult situation to analyze experimentally because the estimate of sad depends
strongly on the GC model used to estimate the ionic surface excesses in the diuse
layer.
An important aspect of analyzing the double layer data in the presence of
specic adsorption is the determination of the dielectric properties of the inner
layer. In the Grahame model for ionic adsorption [G1], the adsorbed ions are
assumed to have their charge centers located on the inner Helmholtz plane (iHp).
Furthermore, the iHp is closer to the electrode surface than the oHp. This is due
to the fact that the adsorbed ions replace solvent molecules on the electrode
surface, whereas the counter ions on the oHp do not. Another feature of the
following treatment is that the charge on the adsorbed ions is assumed to be
located on the iHp. Accordingly, the potential drop across the inner layer is
given by
d m f
sm sm sad
Kma
Kad
10:8:8
where Kma is the integral capacity of the region of the inner layer between the
metal and the iHp, and Kad that between the iHp and oHp (see g. 10.14). Noting
that the integral capacity of the inner layer is that resulting from two capacitors
Kma and Kad in series, one may write
1
1
1
Ki
Kma
Kad
It follows that equation (10.8.8) may be rewritten as
m sad
d m f
Ki Kad
10:8:9
10:8:10
On the basis of this relationship, the potential drop across the inner layer is zero at
the PZC in the absence of ionic adsorption. Under these circumstances fd is also
561
zero. These conditions provide the denition of the rational potential scale used
by Grahame [G1]. In order to estimate the inner potential of the electrode fm, one
must measure the PZC in a system in which specic adsorption is absent, for
instance, the aqueous NaF system against a given reference electrode, Er0 and
subtract that potential dierence from the potential of the polarizable electrode
measured against the same reference electrode, Er ; thus,
fm Er Er0
10:8:11
Plots of dm f against sad for constant sm using data obtained for the KI system
[43] are shown in g. 10.25. It is apparent that the plots are approximately linear.
The slope of the plot gives the reciprocal of the integral capacity of the outer
region of the inner layer (equation (10.9.10)); according to the above results, Kad
varies from approximately 80 mF cm2 at high positive electrode charge densities
to 70 mF cm2 at the more negative values. From the values of dm f in the limit
that sad 0, one may calculate the integral capacity of the inner layer, Ki. This
quantity varies from 34 mF cm2 at high positive charge densities to 18 mF cm2 at
the most negative values. These results are very similar to those obtained in the
absence of specic adsorption in the aqueous KF system [G1]. However, the
analysis is clearly approximate, since both Ki and Kad vary with sm, and to a
lesser extent with sad. In addition, the contribution of the dipole potential and its
possible change with adsorbed ionic charge density are ignored in the above
analysis.
An alternative method of studying ionic specic adsorption involves the use of
electrolyte solutions of constant ionic strength. An electrolyte with no adsorbing
ions such as KF is added to one in which one of the ions adsorbs, for instance, KI
Fig. 10.25 Potential drop across the inner layer, m d , against the charge due to
specically adsorbed iodide ions for a mercury electrode in KI solution at 25 C. The
plots are made at constant electrode charge density, the charge in mF cm2 being shown
adjacent to each curve [43].
562
to keep the total ionic concentration constant, and the relative amounts of the two
electrolytes are varied. The corresponding electrochemical cell is
Hg j xKI; I xKF in H2 O j K specific electrode
10:8:12
where I is the ionic strength. The Gibbs adsorption isotherm for this system is
dg sm dE I dmKI F dmKF
10:8:13
Introducing Parsons function, x g smE , one obtains

dx E dsm I dmKI F dmKF
10:8:14
Since the ionic strength is held constant, one may assume to a rst approximation
that the activity coecients of the individual ions do not depend on solution
composition. Since the concentration of K ions is constant, it follows that
dmKI RT d ln aKI RT d ln cI
10:8:15
dmKF RT d ln aKF RT d ln cF
10:8:16
and
where cI and cF are the concentrations of iodide and uoride ions, respectively.
These concentrations are related to the ionic strength in such a way that one may
write
cI
d ln cI
10:8:17
d ln cF d lnI cI
I cI
Thus, the Gibbs adsorption isotherm becomes

cI
RT d ln cI
dx E dsm I
I cI F
10:8:18
At a given electrode charge density, the surface excess of an anion in the diuse
layer is proportional to its bulk concentration. Since uoride ion is not adsorbed,
the total surface excess F may be equated to that in the diuse layer, dF . It
follows that the surface excess of iodide in the diuse layer is given by
cI
dI

10:8:19
I cI F
Thus, the Gibbs adsorption isotherm may be written
dx E dsm ad
I ln cI
sad
d ln cI
E dsm
f
10:8:20
where ad
I is the surface excess of specically adsorbed iodide ions. Having determined values of x at constant sm for varying bulk iodide concentration, the
adsorbed charge density sad is obtained by dierentiation as follows:

@x
10:8:21
sad f
@ ln cI sm
The eects of ionic strength on anion adsorption have been studied for a
number of systems involving both anions and cations. When adsorption is strong,
563
the magnitude of sad increases with increase in ionic strength. This is due to a
decrease in the magnitude of fd (see equation (10.8.7)). In the case of strong
anionic adsorption, fd is negative, so that the double layer becomes less repulsive
for adsorbing anions as the ionic strength increases. This has been demonstrated
for the adsorption of I , which has been studied at several ionic strengths [44, 45].
In the case of anions which are moderately adsorbed, a change in ionic strength
can act in both directions. For example, in the case of NO
3 adsorption, when the
d
concentration
is
low,
f
is
positive.
Thus,
an
increase
in ionic strength
bulk NO
3
results in a decrease in the magnitude of fd so that the double layer becomes less
attractive for anions. As a result, the magnitude of sad decreases with increase in
ionic strength. On the other hand, fd is negative for higher NO
3 concentrations
and the eect of ionic strength is the same as that observed for the I system.
The above method of determining ionic specic adsorption from solutions of
constant ionic strength has been applied extensively for both cations and anions.
One reason for its popularity is that the correction for the ionic excess in the
diuse layer is made without explicitly introducing the GC model. However, the
assumptions made in deriving equation (10.8.20) can lead to serious errors in the
estimation of sad, especially when adsorption is moderate or weak. The assumption subject to most criticism is that the individual ionic activity coecients do not
vary with solution composition [45, 46]. In a system such as KI KF, it is clear
that the average cationic atmosphere around an anion does not change greatly
with solution composition; however, the anionic atmosphere around the K
cation does change as the relative amounts of KI and KF are changed. A correction for this eect can be made on the basis of models for ionic activity coecients
in electrolyte mixtures [45, 46].
Another problem arises in the case of weak or moderate adsorption [46]. When
the sign of the potential drop across the diuse layer changes with solution
composition or electrode charge density, there is a change in the nature of the
predominant ion at the oHp. Since dierent ions have dierent sizes, the position
of the oHp also changes. Ionic size eects are not considered in the GC model of
the diuse layer. Thus, use of the model based on equation (10.8.10) must consider the possibility that Kma and Kad vary with adsorbed charge density sad for
constant charge density sm on the electrode [46].
The discussion in this section is based to a large extent on the theory of ionic
adsorption developed by Grahame [G1]. In the next section, other factors
involved in ionic adsorption and the underpotential deposition of metals are
considered.
B. Partial Charge Transfer and the Dipole Moment of
the Adsorbate
In Grahames model of specic adsorption [G1] the charge on the adsorbed ion is
located at a discrete site on the ion, normally considered to be the center of the ion
when it is monoatomic. However, specic adsorption is assumed to involve the
formation of a covalent bond between the ion and the metal atoms of electrode.
This has led to the suggestion that charge is transferred between the adsorbed ion
and the electrode. The process of ionic adsorption is then described as
564

z
+Az
Azb rWad lz e (
ad rWb
10:8:22
Az is an ion with charge number z replacing r water molecules at the interface, W,

a water molecule, and lz , the fractional charge involved in the adsorption process.
The subscripts b and ad refer to the bulk of the solution and adsorption at
the interface, respectively. The fraction lz can be both positive and negative. It
cannot be measured by thermodynamic methods.
An eective way of assessing the interaction of an adsorbed ion with its environment is to compare its dipole moment in the adsorbed state with the corresponding value for the same ion adsorbed on a metal from the gas phase. When
the ion is adsorbed on the conducting metal, it forms an image charge which is
located in the metal at a distance ri from the interface, where ri is the radius of the
ion (see g 10.14). Thus, in the gas phase, a dipole is established with a magnitude
pad e0 ri
10:8:23
In the case of the iodide ion with a radius of 206 pm, the value of pad for gas phase
adsorption is 3.310-29 C m, or 10.0 debyes. For adsorption from solution, the
dipole moment can be estimated when the charge density due to adsorbed ions sad
is equal and opposite to the charge density on the metal sm. On the basis of
equation (10.8.8), the potential drop across the inner layer under these circumstances is
sad
10:8:24
d m f
Kma
Noting that the diuse layer potential drop fd is zero when sad is equal to sm,
this is also the potential due to the dipoles established by the adsorbed ions and
their images in the metal. The corresponding dipole moment of the adsorbed ions
is
ze e
pad i 0 0
10:8:25
Kma
From the work of Grahame [43], the value of Kma obtained from a study of iodide
ion adsorption at Hg from solution of KI in water at the pzc is 47.6 mF cm2. The
resulting estimate of pad is 0.9 D for the iodide ion. This is much smaller than the
value obtained for adsorption from the gas phase. Values of pad for other ions
obtained in studies of ionic adsorption at polarizable metals are summarized in
table 10.3. They are all much smaller than estimates of pad for the corresponding
process from the gas phase.
The low values of pad recorded in table 10.3 are generally attributed to strong
screening of the ionic charge by the solvent and the electrons in the metal [47, 48].
First, the solvent dipoles surrounding the adsorbed ion are polarized so that they
align with their dipoles opposite to the dipole created in the adsorption process
(see g. 10.14). Second, the electronic charge in the metal which spills out into the
solution to an extent which depends on the metals properties partially surrounds
the adsorbed species and further screens the adsorbed charge. Partial charge
transfer would further decrease the value of pad. However, on the basis of quantum-mechanical considerations it is not expected to be large for the halide and
alkali metal ions considered in table 10.3. This conclusion is supported by the
565
Table 10.3 Dipole Moments of Ions

Adsorbed on Polarizable Electrodes from
Aqueous Solution at 25 C [47, 48]
Ion
Metal
Dipole Moment
pad /debyes
Hg
Hg
Hg
Hg
Hg
Ag(110)
Au(311)
1.18
0.97
0.81
1.01
1.01
0.35
0.49
Anions
Cl
Br
I
N3
SCN
Br
Br
Cations
K
Rb
Cs
Rb
Cs
Hg
Hg
Hg
Ga
Ga
1.21
1.22
1.18
0.27
0.27
observation that the magnitude of pad increases slightly for the halide ions as the
extent of adsorption increases at more positive electrode charges [48]. If partial
charge transfer were important, the value of pad should decrease as the metal
becomes more electrophilic.
Quite dierent adsorption characteristics are formed in the case of the underpotential deposition of metal ions [49]. As described in section 10.2, underpotential deposition (UPD) is the process by which a metal ion adsorbs on a
dierent metal substrate at a potential more positive than that at which it is
electrodeposited on itself. UPD occurs when there is a signicant dierence
between the work functions of the depositing metal phase M and the substrate
metal S. Consider as an example the electrodeposition of Pb2 on Au. The work
function of polycrystalline Au is approximately 300 mV greater than that of
polycrystalline Pb (see table 8.2). This also means that the PZC of the Au electrode is positive of that for Pb. As a result, Pb2 adsorbs on Au more readily than
it does on Pb. The adsorption process is accompanied by signicant charge transfer. In fact, an estimate of pad for Pb2 on Au is close to zero, indicating that the
cation is essentially discharged.
Very interesting studies have been made of the pattern for UPD on single
crystal metal substrates [49]. A commensurate pattern is usually observed for
the formation of a partial monolayer. The theory of UPD and formation of the
initial monolayer is an area of active research. Of course, the phenomenon of
UPD is restricted to formation of a monolayer. Once this has formed, the deposition process reverts to one of metal ion Mn on metal M. By comparing the UPD
process with anion-specic adsorption, the role of partial charge transfer in these
processes is claried.
566
C. The Adsorption Isotherm

The adsorption isotherm is derived here for process (10.8.22) for the case that
partial charge transfer is negligible (z 0). The condition for equilibrium is
~A r~
~ad
mad
m
A rmW m
W
10:8:26
~ad
m
A
~A , the value
is the electrochemical potential of the ion in the adsorption site, m
~ad
in the bulk, m
W , the electrochemical potential of water in the adsorption site, and
mW, its chemical potential in the bulk. The electrochemical potential of the ion in
the solution is given by equation (8.6.9). The electrostatic contribution is
expressed through the activity coecient of the ion which is needed to relate its
concentration to its activity. The chemical potential of the water molecules in the
bulk is given by equation (8.6.10). Thus, the electrostatic contribution to this term
is considered to be negligible.
The electrochemical potential of the adsorbed ion is given by
ad;
ad
~ad
m
A mA RT ln A zA F
10:8:27
is the standard chemical potential of the adsorbed ion, A is its surface

excess, and ad , the micropotential experienced by the ions at the adsorption
site. The micropotential ad diers signicantly from the average potential fad
in the plane on which the charge center of the ions is located (see g. 10.14). This
is mainly because the adsorption site is so close to the conducting electrode. As a
result, the eects of images formed in the electrode and of the shielding due to the
electron overspill must be taken into account. Finally, on the basis of the discussion regarding interfacial water in section 8.6, the electrochemical potential of
water molecules at the adsorption site is
ad;
mA
ad;
~ad
m
W mA RT ln W hpW iEe
10:8:28
ad;
is the standard chemical potential of an adsorbed water molecule, W ,
where mA
the surface excess of water molecules, hpW i, the average dipole moment perpendicular to the interface, and Ee, the local eld experienced by these molecules.
When the expressions for the individual electrochemical potentials are introduced into equation (10.8.26), the following result is obtained:
RT ln A r RT ln W Gad RT ln zA Rt ln aW zA Fad rhpw iEe

10:8:29
where
ad;
ad;
Gad mA
r mW mA r mW
10:8:30
The two terms on the left-hand side of equation (10.8.29) describe the entropic
eects resulting from the replacing of adsorbed ions by water molecules at the
interface. On the basis of the analysis given in section 8.6, this may be re-expressed
in terms of the maximum surface excess of adsorbed ions Am to obtain
ln A r lnAm A r ln r ln A r ln W
10:8:31
where
r Am r A W
10:8:32
567
In the case that one water molecule is replaced by one ion in the adsorption
process, the result simplies to
10:8:33
ln A lnAm A ln A ln W
In order to understand the dependence of ion adsorption on the electrical
properties of the interface, the micropotential ad is expressed in terms of fad ,
the average potential on the adsorption plane or inner Helmholtz plane, plus a
term accounting for the local departure of the potential from its average value:
ad fad zad
10:8:34
ad
z is called the discreteness-of-charge potential; it depends on the distance of the

adsorption plane from the metal phase, and the number and location of ionic
images in this phase. On the basis of Gauss law the average potential drop
between the adsorption plane and the oHp is
s sad
fad fd m
10:8:35
Kad
Estimation of the discreteness-of-charge potential zad requires a detailed model of
the interfacial region which gives its dielectric properties and the location of
images in the conducting metal phase. According to the model developed by
Levine [50], the value of zad is
C Kad Ki sad
zad d
10:8:36
Cd Ki Kma Kad
Since the diuse layer capacity Cd is usually much greater than the capacities
observed in the inner layer, an approximate expression for zad is
Ki
s
10:8:37
zad
Kma Kad ad
Under most circumstances, the discreteness-of-charge potential has a sign opposite to that of the average potential fad. This means that the electrostatic work
done to adsorb an ion on the iHp is considerably less than the value estimated on
the basis of the average potential on this plane. This result reects the stabilizing
inuence of the images formed in the conducting metal which reduce the electrostatic repulsion between adsorbed ions with the same charge.
After substituting in the expression derived for ad the adsorption isotherm
may be written as
ln A r lnAm A ln Bad aA zA ffd Aad sad
10:8:38
G
z f
r ln r r ln aW rhpw iEe A sm
RT
Kad
10:8:39
where
ln Bad
and
Aad

zA F Kma Ki
Kad
Kma
10:8:40
RT ln Bad is the standard Gibbs energy of adsorption for a given charge density
on the electrode. The last term in sm gives the most important contribution to the
dependence of Bad on the electrical characteristics of the interface. Bad also
568
depends on the electrostatic contribution of the electrochemical potential of water

(rhpW iEe ) but this contribution changes much less with sm than the last term.
Finally, Aad is the interaction coecient which describes the repulsive interactions
between adsorbed ions on the adsorption plane. It depends on the dielectric
properties of the inner layer, the discreteness-of-charge component being an
important factor in determining its magnitude.
When experimental data are only available for low coverages (A Am , the
entropic term on the left-hand side of equation (10.8.38) may be simplied as
follows:
r lnAm A r ln Am
r A
Am
Then the adsorption isotherm becomes

s
ln ad ln B0ad aA zA ffd A0ad sad
zA F
10:8:41
10:8:42
where
ln B0ad aA ln Bad r ln Am
10:8:43
and
A0ad Aad
r
zA FAm
10:8:44
In equation (10.8.42) which is known as the virial isotherm, the surface excess of
adsorbed ions A is expressed in terms of their charge density sad using the charge
on one mole of ions zA F.
The virial isotherm describes the adsorption of monoatomic ions such as the
halides on mercury and other metals. In order to test its applicability to experimental data it is plotted in the form

sad
10:8:45
ln
zA ffd ln B0ad A0ad sad
z A aA
Accordingly, the quantity on the left-hand side of equation (10.8.45) should be a
linear function of sad, the charge density due to adsorbed ions, for constant
charge density on the electrode. Typical results obtained for the adsorption of
iodide ion at Hg [43] are shown in g. 10.22. Linear plots are obtained for a wide
variation in the bulk iodide ion concentration, in this case from 0.025 M to 1 M.
The isotherm parameters are found from the slope and intercepts of these plots.
For example, at sm 0 mC cm2, the slope of the isotherm plot in g. 10.26 is 0.28
cm2 mC1. For monoatomic ions like iodide, the value of Am is expected to be
large, a reasonable estimate of zAFAm being 100 mC cm2. As a result, the
contribution of the term r/(zAFAm) in equation (10.8.44) is negligible. From
the plots of dm f against sad at sm 0 mC cm2 in g. 10.25, the estimates
of Ki and Kad are 28.9 and 73.5 mF cm2, respectively. On the basis of equation
(10.8.40) the estimate of the interaction parameter Aad is 0.21 cm2 mC1. This
compares favorably with the experimental result. In assessing this isotherm and
its application to experimental data it should be remembered that the model used
to derive the isotherm is not based on a molecular description of the solvent at the
569
Fig. 10.26 Plots of experimental data for the adsorption of iodide ion on Hg according to
the virial isotherm using data obtained from solutions of varying ionic strength [46]. The
iodide ion activity was set equal to the mean ionic activity. The electrode charge density in
mC cm2 is indicated adjacent to each plot.
interface. Further development of this topic should be based on a completely

microscopic description of interfacial properties.
Other adsorption isotherms have been used in the literature to analyze experimental data. However, it can usually be shown that they are limiting forms of the
general isotherm derived here (equation (10.8.38)). For example, in early work the
importance of the term in the diuse layer potential (zAf fd) was not recognized.
By using equation (10.8.7) for this contribution, the exact dependence of this term
on sm and sad is obtained. Equation (10.8.7) can be simplied in limiting cases
and the form of the isotherm without an explicit dependence on fd obtained.
10.9 The Adsorption of Molecules at Electrodes

Water-soluble organic molecules are usually preferentially adsorbed at the metal |
solution interface. Thus, small molecules such as methanol, acetone, and acetonitrile have a polar group which results in their being water soluble, and non-polar
groups which prefer to be at the interface out of the aqueous environment. The
accumulation of the organic solute at the interface results in a lowering of the
interfacial tension to an extent which depends on the surface excess of the adsorbing compound. Typical results obtained for the adsorption of 2-butanol from
aqueous solutions containing 0.1 M Na2SO4 are shown in g. 10.27. It is apparent
from these results that the extent of adsorption depends on both the electrode
potential and the concentration of 2-butanol. In this system, the PZC which occurs
at the maximum value of the interfacial tension shifts in the positive direction with
adsorption of the alcohol molecule. The exact features of the electrocapillary curve
depend on the dipole moment of the adsorbing molecule and its orientation at the
570
Fig. 10.27 Interfacial tension of the Hg | aqueous solution interface against electrode
potential E for 0.1 M Na2SO4 with varying amounts of 2-butanol at 25 C: (*) 0 M,
(^) 0.105 M, and (~) 1.00 M [51].
interface. As the electrode potential is made more positive or negative, and consequently, the electrical eld at the interface increased, the organic molecules at the
interface are replaced by solvent molecules. As a result the interfacial tension is
independent of solution composition at the extremes of polarization.
The other common way of studying molecular adsorption is to measure the
dierential capacity of the electrode | solution interface. Typical capacity data
obtained for a similar system, namely, t-butanol in water are shown in g. 10.28.
In the potential region when the organic molecule is adsorbed, the capacity is
signicantly reduced from the value observed in the absence of adsorbate. In
addition, the extent of capacity lowering increases with increase in adsorbate
concentration. Maxima are observed on the capacity curves in the potential
regions where desorption of the organic molecules occurs. They result from the
rapid change in surface coverage with potential in these regions. At high bulk
concentrations the maxima often involve very high capacities which are frequency
dependent. In this case the adsorption process is not at equilibrium, so that the
measured capacity is not thermodynamically relevant.
Experimental data relevant to molecular adsorption have been obtained for a
wide variety of organic compounds, including alcohols, aldehydes, ketones,
amines, and amides [G5]. Included are studies of liquids which are polar solvents
themselves, for example, acetonitrile, acetone, ethylene carbonate, formamide,
and methanol. Studies have also included organic acids which are weakly dissociated in water.
A. Analysis of the Thermodynamic Data
An ideal thermodynamic cell which could be used to study the adsorption of
adsorbate A at mercury is the following:
571
Fig. 10.28 Dierential capacity of the Hg | aqueous solution interface against electrode
potential ESCE for 0.1 M Na2SO4 with varying amounts of t-butanol at 25 C: (*) 0 M, (^)
0.108 M, and (~) 0.540 M [52].
Hg j xA; yNaF; H2 O j Hg2 F2 j Hg
10:9:1
x is the concentration of the adsorbate A which is varied in order to study the

eect of bulk concentration on the interfacial surface excess; y is the concentration
of the electrolyte whose activity is kept constant. The corresponding electrolyte
concentration is kept reasonably high to provide electrical conductivity to the
solution. For low values of x, the electrolyte concentration is constant.
However, at higher concentrations of organic solute, the activity coecient of
the electrolyte varies with organic solute concentration. Thus, in general, the
concentration of the electrolyte must also be varied in order to keep its activity
constant. It is also important that the ions of the electrolyte not adsorb on the
electrode to a signicant extent.
The condition for equilibrium at the polarizable interface is
dg sm dE dme A dmA w dmw
10:9:2
where E is the electrode potential with respect to the uoride reference electrode,
, the surface excess of the Na cation, A, that of the organic adsorbate, w,
that of water molecules, me, the chemical potential of the electrolyte, mA, that of
the adsorbate, and mw, that of water. The experiments are carried out for constant
me and the relative activities of the adsorbate and water changed so that
xA dmA xw dmw 0
10:9:3
where xA and xw are the mole fractions of adsorbate and water. It follows that the
relative surface excess of adsorbate w
A can be found by measuring the change in
interfacial tension with chemical potential of the adsorbate holding the electrode
potential E and the electrolyte activity ae constant. Thus,
572
w
A

xA
@g
A w
@mA E ;me
xw
10:9:4
In the following discussion the relative nature of the surface excess is neglected
and the role of water in the thermodynamic analysis ignored.
In dilute solutions, the eects of the organic solute on the electrolyte activity
can be ignored so that experiments can be carried at constant electrolyte concentration. The chemical potential of the electrolyte is given by
me me RT ln ye ce
10:9:5
where ye is the mean activity coecient on the concentration scale. Thus, the
change in electrolyte activity with organic solute concentration is zero for constant electrolyte concentration, provided the activity coecient ye is also constant.
However, for the concentration range of interest, signicant changes in ye are
expected (see g. 10.29). The increase in electrolyte activity for constant electrolyte concentration is known as the salting out eect. It is due to the fact that the
number of water molecules available to solvate the ions of the electrolyte
decreases as the concentration of the organic solute increases. This eect was
usually ignored in early studies of organic adsorption, so that the results obtained
with constant electrolyte concentration are signicantly in error. On the other
hand, the task of determining the value of the electrolyte concentration needed
to maintain constant electrolyte activity is formidable, so that another procedure
for analyzing the thermodynamic data would help facilitate these studies.
Consider now experiments carried out with a constant electrolyte concentration ce in the cell
Hg j xA; ce NaF; H2 O j Hg2 F2 j Hg
10:9:6
Fig. 10.29 Plots of the change in ionic activity coecients determined by specic ion
electrodes against acetamide concentration cAM for 0.25 M NaF: Na ion activity coecient (!); F ion activity coecient (~); mean activity coecient for NaF(*) [53].
573
The thermodynamic data are used to calculate the function # [53], which is
dened as
# g sm E me
10:9:7
If the original experimental data involve capacity measurements as a function of
electrode potential E and organic solute activity aA , the data must be integrated
in the usual way to obtain electrode charge density sm and interfacial tension g. In
addition, the electrolyte activity must be determined as a function of aA ; this can be
accomplished easily using specic ion electrodes for most electrolytes commonly
used in these studies. Taking the total derivative of # , the result is
d# dg sm dE E dsm me d dme
10:9:8
Substituting this into the GAI, one obtains

d# E dsm me d A dmA w dmw
10:9:9
It follows that the relative surface excess of the organic adsorbate is given by

@#
w
10:9:10
A
@mA sm ;
If the electrolyte is not specically adsorbed then the surface excess of the cation is
constant for constant electrode charge density. It follows that introduction of the
function # provides an easy route for analyzing experimental data obtained at
constant electrode charge density. It also means that experimental data are not
easily analyzed at constant electrode potential. Further details about the thermodynamics of these systems are given elsewhere [53, 54].
B. The Adsorption Isotherm
The process of molecular adsorption at the polarizable interface involves replacement of water molecules in the inner layer which are solvating the electrode. Thus,
(Aad rWb
Ab rWad +
10:9:11
where the subscript ad designates molecules at the interface and the subscript
b molecules in the bulk. The same process can be used to describe the adsorption of molecules at the aqueous solution | air interface as described in section 8.6.
The adsorption isotherm derived earlier is also applicable here, namely,
ln A r lnAm A ln Bad aA hpA i rhpw i
Ee
RT
10:9:12
The experimentally controlled variables which are explicit in equation (10.9.12)

are the activity of the adsorbate aA and the temperature T. The experimentally
observed quantity is the surface excess A. The dependence of the adsorption on
the electrical state of the interface is expressed through the local eective eld Ee
perpendicular to the interface and the average dipole moments of the adsorbate
hpAi and water molecules hpWi in the same direction. The contribution of the last
term is much larger under most circumstance for adsorption at the electrode |
solution interface than at the solution | air interface. As a result, further treatment
of the two problems is quite dierent.
574
The local eective eld Ee cannot be determined experimentally, and must be

estimated in terms of other electrical properties of the interface. This has led to
considerable discussion in the literature [54] concerning the appropriate experimental electrical variable. Two choices directly available from experiment are the
electrical potential E and the surface charge density sm. The choice of sm is
clearly more appropriate because most experimental are measured at constant
electrolyte concentration and therefore must be analyzed at constant sm as
described above. Thus, the adsorption isotherm is written as
ln A r lnAm A ln Bad aA
Aad A
Am
10:9:13
where
ln Bad
Gad
r ln r r ln aW
RT
10:9:14
RT ln Bad is the eective standard Gibbs adsorption energy, which depends on

electrode charge density. This dependence can often be described by the equation
RT ln Bad RT ln Bad;max bad sm sm;max 2
10:9:15
where Bad;max and sm;max are the values of Bad and sm , respectively, at which Bad
is a maximum. Aad is the interaction coecient which also can depend on the
electrical state of the interface. Positive values of Aad signify repulsive interactions
between the adsorbing molecules, whereas negative values indicate that the interactions are attractive. Aad is the interaction coecient which also can depend on
the electrical state of the interface. The eects of these interactions on adsorption
for typical conditions at constant electrical state of the interface are shown in g.
10.30. When interactions are absent the value of the fractional coverage y rises
sharply with bulk concentration and then levels o as coverage increases. The
presence of repulsive interactions results in a signicantly slower rate of increase
in y over the concentration range in which a comparison is made. On the other
hand, when these interactions are attractive, y increases more rapidly. However,
the presence of these interactions is not readily apparent from visual examination
of y against aA data for a single system. This feature of the adsorption data can
only be assessed when the value of the maximum surface excess Am is known.
Some methods for determining this quantity were considered in section 8.6.
Typical experimental results obtained for the adsorption of acetamide on mercury at negative charge densities [55] are shown in g. 10.31. As the electrode
charge density is made more negative, the increase in surface coverage with bulk
acetamide concentration is retarded. This demonstrates that the value of the
interaction parameter Aad is increasing, that is, becoming more repulsive. At
more positive charge densities (not shown) the plots of Am against cAm do not
dier signicantly from the plot for sm 5 mC cm2.
Further analysis of the experimental data requires determination of the maximum surface coverage Am. Various methods are available to estimate Am. In the
present case, the best value in the region of maximum adsorption is 0.32 nmol cm2
and some variation with electrode charge density is observed. The latter observation is attributed to reorientation of the polar adsorbate in the electrodes eld.
575
Fig. 10.30 Plots of surface coverage y A/ Am against bulk activity of the adsorbate aA
with a moderately adsorbed species (RT ln Bad 0) for the cases of no interaction (Aad
0), attractive interaction (Aad 1), and repulsive interaction (Aad 1). The geometrical
factor r was set equal to unity.
The experimental data for acetamide are plotted according the Frumkin isotherm (equation (10.10.13) with r 1) in g. 10.32. This plot was constructed
using the acetamide concentration rather than its activity. This is equivalent to
assuming that the activity coecient of the adsorbate is constant in the bulk of the
solution over the concentration range considered. The slopes of the isotherm plots
are negative at the most positive electrode charge densities (3 and 6 mC cm2 )
indicating that the acetamide molecules interact attractively on the adsorption
plane at the interface. However, the interaction coecient decreases in magnitude
Fig. 10.31 Plots of the surface excess of acetamide Am against bulk acetamide concentration cAm for three electrode charge densities: (*) 5; (!) 10; and (^) 15 mC cm2 [55].
576
Fig. 10.32 Plots of experimental data for the adsorption of acetamide at mercury according to the Frumkin isotherm (equation (10.10.13)) with r 1 and y A/Am). The activity
of acetamide aA was set equal to its concentration cA. Data are shown for sm 3 (*); 6
(&); 9 (~) ; 12 (!) and 15 mC cm2 (^).
as the electrode charge density is made more negative and eventually goes to zero.
The changes in the nature of the interaction with electrical state of the interface
observed for acetamide are typical for polar molecules at the mercury | solution
interface [55].
In early work the question of isotherm congruency was considered in the
development of the theory of interfacial adsorption [54, 55]. A congruent isotherm
is one for which the surface excess is specied once the values of the adsorbate
activity aA and standard Gibbs energy of adsorption RT ln Bad have been specied. This theory was developed for the cases that either the electrode potential E
or the electrode charge density sm is the appropriate electrical variable. More
extensive examination of the experimental data has led to the conclusion that the
concept of congruency is only a rst approximation. In fact, when a completely
molecular approach ia applied to the problem of organic adsorption, the adsorption isotherm is clearly not congruent [56].
The most recent experimental work has involved studies of organic adsorption
at the single crystal faces of polarizable solid metal electrodes [57]. These experiments provide details of the role of the metal in organic adsorption. By examining
these data within the context of the new molecular descriptions of interfacial
adsorption the theory of this important process will be greatly advanced.

The discussion in this chapter has focused on the metal | solution interface. The
interfacial property considered in most detail was the dierential capacity. It was
577
shown that the contributions of both the metal and the solvent play an important
role in determining the change in capacity with electrode potential. The adsorption of ions and solute molecules was also discussed in detail. These interfaces
have been studied experimentally in great detail, both at dierent metals, and for
a wide variety of solvents. In addition, considerable eort has been directed to
developing theoretical models for the electrical double layer. Other systems which
have been studied include the semiconductor | solution interface and the metal |
molten salt interface.
Double layers are also important in colloid chemistry. When a colloid particle
is composed of an ionic crystal, it often preferentially adsorbs one of its component ions, thereby acquiring a charge. As a result the colloid particle is surrounded by a double layer. The interfacial properties are very important in
determining a variety of colloidal properties, including electrophoresis and electroosmosis. It also plays a role in colloid stability and coagulation phenomena.
The eects of the electrical properties of the interface are well known in colloid
chemistry. The description of colloid phenomena is a well-developed area of
physical chemistry which is often important in industrial processes.
Finally, double layers are important in a variety of biological phenomena,
especially those occurring at cell walls. Although these systems are more complex
in their description, the fundamental concepts applied here are also applicable in
the biological systems. For example, ion transport phenomena in membranes
depend on the electrical state of the membrane interfaces. Thus many concepts
from the physical chemistry of polarizable interfaces and colloids are also applicable to charged interfaces in biological systems.
General References
G1. Grahame, D. C. Chem. Rev. 1947, 41, 441.
G2. Delahay, P. Double Layer and Electrode Kinetics; Wiley-Interscience: New York,
1965.
G3. Damaskin, B. B.; Petrii, O. A. An Introduction to Electrochemical Kinetics (in
Russian), 2nd ed.; Vysshaya Shkola: Moscow, 1983.
G4. Bockris, J. OM.; Conway, B. E.; Yeager, E., eds. Comprehensive Treatise of
Electrochemistry; Plenum Press: New York, 1980; Vol. 1.
G5. Lyklema, J.; Parsons, R. Electrical Properties of Interfaces Compilation of Data
on the Electrical Double Layer, National Bureau of Standards Document NBSIR
83-2714, US Department of Commerce, 1983.
References
1. Mohilner, D. M. In Electroanalytical Chemistry; Bard, A. J., ed.; Marcel Dekker: New
York, 1966; Vol. 1, Chapter 4.
2. Lipkowski, J.; Schmickler, W.; Kolb, D. M.; Parsons, R. J. Electroanal. Chem. 1998,
452, 193.
3. Heyrovsky, J. Chem. List 1922, 16, 256.
4. Lippmann, G. Compt. Rend. 1873, 76, 1407; Pogg. Ann. 1873; 149, 561; J. Phys. 1874,
43.
5. Payne, R. In Techniques of Electrochemistry; Yeager, E., Salkind, A. J., eds.; Wiley
Interscience: New York, 1973; Vol. 1, Chapter 2.
6. Barradas, R. G.; Kimmerle, F. M. Can. J. Chem. 1967, 45, 109.
578
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
42.
43.
44.
45.
46.
47.
48.
49.
Gouy, G. Ann. Chim. Phys. 1903, 7, 145.

Grahame, D. C. J. Electrochem. Soc. 1951, 98, 343.
Pajkossy, T. J. Electroanal. Chem. 1991, 300, 1.
Fawcett, W. R.; Kovacova, Z.; Motheo, A. J.; Foss, C. A. J. Electroanal. Chem. 1992,
326, 91.
Lipkowski, J.; Stolberg, L. In Adsorption of Molecules at Electrodes; Lipkowski, J.,
Ross, P. N., eds.; VCH: New York, 1992; Chapter 4.
Gale, R. J. Spectroelectrochemistry, Theory and Practice; Plenum Press: New York,
1988.
Abruna, H. D. Electrochemical Interfaces: Modern Techniques for In-Situ Interface
Characterization; VCH: New York, 1991.
Wieckowski, A. Interfacial Electrochemistry, Theory, Experiment and Applications;
M. Dekker: New York, 1999.
Foley, J. K.; Korzeniewski, C.; Daschbach, J. L.; Pons, S. Electroanal. Chem. 1986, 14,
309.
Faguy, P. W.; Marinkovic, N. S. Anal. Chem. 1995, 67, 2791.
Marinkovic, N. S.; Hecht, M.; Andreu, R.; Fawcett, W. R. J. Phys. Chem. 1977, 99,
6760.
Marinkovic, N. S.; Calvente, J. J.; Kovacova; Fawcett, W. R. J. Electroanal. Chem.
1997, 421, 49.
Magnussen, O. M.; Ocko, B. M.; Wang, J. X.; Adzic, R. R. J. Phys. Chem. 1996, 100,
5500.
von Helmholtz, H. F. L. Ann. Physik 1853, 89, 211; 1879, 7, 337.
Gouy, G. Compt. Rend. 1910, 149, 654.
Chapman, D. L. Phil. Mag. 1913, 25, 475.
Stern, O. Z. Elektrochem. 1924, 30, 508.
Trasatti, S. In Trends in Interfacial Electrochemistry; Silva, A. F., ed.; Dordrecht:
Boston, 1984; p. 25.
Badiali, J. P.; Rosinberg, M. L.; Goodisman, J. J. Electroanal. Chem. 1983, 143, 73;
1983, 150, 25.
Schmickler, W. Chem. Rev. 1996, 96, 3177.
Trasatti, S.; Lust, E. Mod. Aspects Electrochem. 1999, 33, 1.
Frumkin, A. N. Potentials of Zero Charge; Nauka: Moscow, 1979.
Smith, J. R. Phys. Rev. 1969, 181, 522.
Schmickler, W.; Henderson, D. J. J. Chem. Phys. 1984, 80, 3381.
Leiva, E.; Schmickler, W. J. Electroanal. Chem. 1987, 228, 39.
Boda, D.; Fawcett, W. R.; Henderson, D. J.; Sokolowski, S. J. Chem. Phys. 2002, 116,
1770.
Schmikler, W.; Henderson, D. J. Prog. Surf. Sci. 1986, 22, 323.
Grahame, D. C. J. Am. Chem. Soc. 1957, 79, 2093.
Fawcett, W. R. Israel J. Chem. 1979, 18, 3.
Watts-Tobin, R. J. Phil. Mag. 1961, 6, 133.
Fawcett, W. R. J. Phys. Chem. 1978, 82, 1385.
Parsons, R. F. Electroanal. Chem. 1975, 59, 229.
Damaskin, B. B. J. Electroanal. Chem. 1977, 75, 359.
Fawcett, W. R.; Levine, S.; de Nobriga, R. M.; McDonald, A. C. J. Electroanal. Chem.
1980, 111, 163.
Levine, S.; Bell, G. M.; Smith, A. L. J. Phys. Chem. 1969, 73, 3534.
Guidelli, R.; Schmickler, W. Electrochim. Acta 2000, 45, 2317.
Grahame, D. C. J. Am. Chem. Soc. 1958, 80, 4201.
Dutkiewicz, E; Parsons, R. J. Electroanal. Chem. 1966, 11, 100.
Fawcett, W. R.; McCarrick, T. A. J. Electrochem. Soc. 1976, 123, 1325.
Fawcett, W. R.; Sellan, J. B. Can. J. Chem. 1977, 55, 3871.
Schmickler, W; Guidelli, R. J. Electroanal. Chem. 1987, 235, 387.
Schmickler, W. J. Electroanal. Chem. 1988, 249, 25.
Leiva, E. Electrochim. Acta. 1996, 41, 2185.
579
50. Levine, S. J. Colloid Interface Sci. 1971, 37, 619.

51. Nakadomari, H.; Mohilner, D. M.; Mohilner, P. R. J. Phys. Chem. 1976, 80, 1761.
52. de Battisti, A.; Abd El Nabey, H. B.; Trasatti, S. J. Chem. Soc., Faraday Trans. 1.
1976, 72, 2076.
53. Fawcett, W. R.; Champagne, G. Y.; Komo, S. and Motheo, A. J. J. Phys. Chem. 1988,
92, 6368.
54. Damaskin, B. B., and Kazarinov, V. E. In Comprehensive Treatise of ElectroChemistry, Bockris, J. OM., Conway, B. E., Yeager, E., eds.; Plenum: New York,
1980; Vol. 1, Chapter 8.
55. Fawcett, W. R.; Champagne, G. Y.; Motheo, A. J. J. Chem. Soc., Faraday Trans.
1990, 86, 4037.
56. Parsons, R. Chem. Rev. 1990, 90, 813.
57. Trasatti, S., Doubova, L. M. J. Chem. Soc., Faraday Trans. 1995, 91, 3311.
Problems
1. Given the following cell:
Hg j K2 SO4 in H2 O j Hg2 SO4 j Hg
derive an expression from the Gibbs absorption isotherm for the change in
interfacial tension at the Hg | solution interface with potential drop across the
cell, and chemical potential of K2 SO4 .
2. The following surface tension data were obtained for the Hg | 1.0 M CsCl
interface:
E/V
g/mJ cm2
E=V
g=mJ cm2
0
0:1
0:2
0:3
0:4
0:5
0:6
34.50
37.64
39.71
41.05
41.87
42.26
42.29
0:7
0:8
0:9
1:0
1:1
1:2
1:3
41.99
41.40
40.56
39.51
38.21
36.92
35.36
Determine the PZC and values of sm for 200 mV increments in E in the

range 0:2 to 1:2 V.
3. The following surface tension data were obtained in LiCl solutions of varying
concentration at a constant potential on the E scale (1:0 V). Determine the
surface excess of the cation for all concentrations except the rst and last. y
is the mean molar activity coecient.
Conc/M
g=mJ cm2
sm =mC cm2
0.03
0.10
0.30
1.00
3.00
0.84
0.78
0.72
0.72
0.96
40.69
40.50
40.06
39.69
39.43
8:4
8:9
10:1
11:0
11:6
580
Use the GC model to estimate s and s assuming no ionic specic adsorption, and compare these estimates with those obtained from the thermodynamic analysis.
4. The following data were obtained at the PZC for the Hg | aqueous NaF
interface for concentrated solutions:
Conc/M
Density/g cm3
E/V
g=mJ cm2
0.1
0.25
0.5
1.0
4.0
6.25
8.0
10.0
1.001
1.008
1.020
1.045
1.189
1.285
1.36
1.440
0.773
0.710
0.670
0.645
0.792
1.098
1.48
2.132
0:355
0:333
0:318
0:300
0:261
0:238
0:228
0:207
42.65
42.69
42.73
42.77
43.03
43.28
43.67
44.30
Estimate the relative cationic surface excess at the PZC and plot it as a
function of the mole fraction ratio xs =xw . Is there any evidence for ionic
specic adsorption? If not, can an estimate of w be made from these data?
Comment on the results of your analysis.
5. In his experimental study of the adsorption of iodide ion at mercury,
Grahame obtained the following results at an electrode charge density of
10 mC cm2 :
Conc. KI/M
Electrode Potential
E/V
Cationic Surface Excess

s =mC cm2
0.025
0.1
1.0
0:5412
0:5896
0:6714
8.87
12.43
18.86
Calculate the adsorbed charge density due to I ion. Construct a plot of

E fd against sad and determine the capacity of the outer region of the
inner layer, Kad . Given that the PZC on the above potential scale in the
absence of adsorption is 0:472 V, determine the integral capacity of the
inner layer, and thus, the capacity of the inner region of the inner layer, Kma .
6. Using the above data construct an adsorption isotherm plot on the basis of
equation (10.8.45) and determine the isotherm parameters.
7. The following surface pressure data were obtained for the cell
Hg j xM DMA; 0:15 M Na2 SO4 ; H2 O j Hg2 SO4 j Hg
at a constant potential E of 0:7 V at 258C, where DMA is dimethylacetamide:
Conc DMA/M
p g g0
mJ cm2
0.04
0.063
0.1
0.16
0.25
0.4
0.63
1.0
1.60
0:575
0:645
0:848
1:135
1:135
1:753
2:101
2:506
2:947
581
Determine the relative surface excess of DMA for concentrations in the range
0.10.63 M, assuming the activity coecient of Na2 SO4 is independent of
DMA concentration.
8. Measurements of the activity coecient of Na2 SO4 in the above solutions for
a constant concentration of 0.15 M yielded the following results:
Conc. DMA/M
y (0.15 M Na2 SO4 )
0.01
0.10
0.16
0.25
0.4
0.63
1.0
1.60
0.396
0.428
0.434
0.444
0.462
0.506
0.582
0.726
Correct the above results for the variation in y to obtain a better estimate of
the relative surface excess of DMA. Assume that the charge on the electrode,
sm , is constant and equal to 6 mC cm2 .
9. Derive the form of the Frumkin isotherm
ln lnm
A
ln bc
m
which is appropriate when m . The following data were obtained for the
adsorption of thiourea from ethylene carbonate solutions:
Conc. Thioureau/M
Surface Excess
=molec cm2
0.0095
0.058
0.26
1.06
3 1013
6 1013
9 1013
12 1013
Determine the isotherm parameters ln b and A=m .
Appendix A
Mathematical Background
In this appendix some important mathematical methods are briey outlined.

These include Laplace and Fourier transformations which are often used in the
solution of ordinary and partial dierential equations. Some basic operations
with complex numbers and functions are also outlined. Power series, which are
useful in making approximations, are summarized. Vector calculus, a subject
which is important in electricity and magnetism, is dealt with in appendix B.
The material given here is intended to provide only a brief introduction. The
interested reader is referred to the monograph by Kreyszig [1] for further
details. Extensive tables relevant to these topics are available in the handbook
by Abramowitz and Stegun [2].
A.1 Laplace Transforms

The method of Laplace transforms is applied to the solution of ordinary and
partial dierential equations. It falls into the general category of operational
calculus, and greatly simplies the problem to be solved. When a dierential
equation is solved a family of functions satisfying the equation is found. Using
the boundary conditions, a specic function is selected from the family. Methods
of solving dierential equations is a whole subject in itself which is not discussed
here. However, when the Laplace transform is applied to the dierential equation,
it is transformed into an algebraic equation. The algebraic equation is usually
much easier to solve.
582
MATHEMATICAL BACKGROUND
583
The Laplace transform of a function ft is dened as

fs
est ftdt
A:1:1
Accordingly, a new function fs is dened which depends on s as independent

variable. This transformation may also be written as
fs Lfftg
A:1:2
where the operator L designates Laplace treansformation. Inverse Laplace transformation leads back to ft:
ft L1 ffsg
A:1:3
A simple example of Laplace transformation is given now for the function eat :
1
Lfe g
at
at st
e e
0
1

1
ea st
dt
as
0
A:1:4
Substituting in the limits one obtains

Lfeat g
1
sa
A:1:5
Other examples are given in section 6.5 and in table A.1.

Table A.1 Laplace Transforms of Some Common
Functions [1]
ft
a
t
tn
t
eat
fs
a
s
1
s2
n!
sn1
a 1
sa1
1
sa
cos ot
s
s2 o2
sin ot
o
s2 o2
cosh at
s
s2 a2
sinh at
a
s2 a2
584
A.2 Fourier Transforms

Fourier series are powerful tools used to describe periodic functions. Suppose that
fL x is a periodic function that is dened in the interval L 0 L. A simple
example is a square wave which is centered at x 0, and repeated continuously
for values of x greater than L and less than L. Such a function may be represented as an innite sum of cosines and sines, so that
1 h
np
npi
X
fL x a0
An cos
bn sin
A:2:1
L
L
n1
The technique used to generate Fourier series is easily extended to develop
expressions for non-periodic functions. The Fourier series then becomes a
Fourier integral. The Fourier integral expression for a function fx is
1
Aw cos wx Bw sin wxdw
fx
A:2:2
The coecients Aw and Bw are given by

1
Aw

1
Bw

fv cos wv dv
A:2:3
fv sin wv dv
A:2:4
1
1
1
The Fourier integral expression is simpler in the case of odd or even functions.
For example, if the function is odd,
fx fx
A:2:5
fx
Bw sin wx dw
A:2:6
On the other hand, if fx is an even function,

fx fx
A:2:7
fx
Aw cos wx dw
A:2:8
The integrals which arise can often be evaluated making use of the Laplace transforms given in table A.1.
Fourier transforms are important integral transforms that can be used in the
solution of dierential equations. Fourier transformation is also used in modern
spectroscopy to obtain spectral data of improved precision. If fx is an even
function its Fourier transform is dened as
585
1=2 1
2
F c w
fx cos wx dx

A:2:9
Inverse Fourier transformation involves estimation of the following integral:

1=2 1
2
fx
F c w cos wx dw

A:2:10
In the case of odd functions a Fourier sine transform is used. Thus, if fx is an

odd function, the Fourier transform is
1=2 1
2
F c w
fx sin wx dx

A:2:11
and the inverse Fourier transform is

1=2 1
2
fx
F cw sin wt dw

A:2:12
Fourier transforms of some simple functions are given by Kreyszig [1]. However,
usually the Fourier transform is more commonly expressed in complex form
without specifying the nature of the function to be transformed. This type of
Fourier transform is considered in the following section after a discussion of
complex numbers and functions.
A.3 Complex Numbers and Functions

Complex numbers are often used to deal with phenomena involving periodic
functions. A complex number z is usually written as
z x jy
A:3:1
where x ispthe
real part of z and y is the imaginary part. The imaginary unit j is
equal to 1 . The usual mathematical operations may be performed with complex numbers. Thus, addition gives
z1 z2 x1 j y1 x2 j y2 x1 x2 jy1 y2
A:3:2
Multiplication of two complex numbers gives

z1 z2 x1 jy1 x2 jy2 x1 x2 jx1 y2 x2 y1 j2 y1 y2
x1 x2 y1 y2 jx1 y2 x2 y1
A:33
The square of j is 1, so that the product y1 y2 becomes part of the real term
in z1 z2 . Division of one complex number by another illustrates the use of the
complex conjugate of the denominator. Thus,
586
z1 x1 j y1 x1 j y1 x2 j y2 x1 x2 y1 y2
x y x1 y 2
j 2 12
z2 x2 j y2 x2 j y2 x2 j y2
x22 y22
x2 y22
A:3:4
Important complex functions are based on the Euler formula, that is,
e jy cos y j sin y
A:3:5
Thus, the exponential of a complex number z is given by

ez ex cos y j sin y
A:3:6
e jz cos z j sin z
A:3:7
This can also be written as
It follows that
cos z
ejz ejz
2
A:3:8
sin z
e jz ejz
2j
A:3:9
The complex hyperbolic functions are

e jz ejz
cos z
2
A:3:10
e jz ejz
j sin z
2
A:3:11
cosh j z
sinh jz
Now, the Fourier transform can be written in complex notation. The Fourier
transform of any function fx is written as
1=2 1
2
F w
fxejwx dx

A:3:12
1
The exponential in jwx can be expanded into a term in cos wx and one in j sin
wx. Thus, the above formula is general and can be applied to both odd and even
functions. Inverse Fourier transformation gives
1=2 1
2
fx
F we jwx dw

A:3:13
1
A.4 Power Series

It is often helpful to express a function in terms of a power series. A very simple
power series is
1
X
an xn a0 a1 x a2 x2

587
A:4:1
n0
The coecients, ai are often related to one another in a simple way which is
determined by the nature of the function. An important method of expressing
functions in a power series is the Taylor and Maclaurin expansions. In a Taylor
expansion the function fx is expanded about a given point x0 and the coecients
are related to the values of the derivatives of the function at x x0 . Thus, the
Taylor expansion of fx is
!

df
x x0 2 d2 f
fx fx0 x x0
dx xx0
2!
dx2
xx0
A:4:2

x x0 n dn f
dxn xx0
n!
The Maclaurin expansion is the special case in which the function is expanded
about x 0. Thus, the Maclaurin expansion of fx is
!

2
df
x2 d2 f
fx f0 x
dx x0 2! dx

x0

xn dn f
n! dxn x0
A:4:3
A list of important functions expressed as power series follows:

eax 1 ax
a2 x2
an xn

2!
n!
1
1 ax a2 x2 a3 x3

a ax
sin ax ax
cos ax 1
a3 x3 a5 x5 a7 x7
3!
5!
7!
a 2 x 2 a 4 x 5 a 6 x6
2!
4!
6!
A:4:4
A:4:5
A:4:6
A:4:7
ln1 ax ax
a2 x2 a3 x3 a4 x4
2
3
4
A:4:8
sinh ax ax
a3 x3 a5 x5 a7 x7
3!
5!
7!
A:4:9
cosh ax 1
a 2 x 2 a 4 x 4 a 6 x6
2!
4!
6!
A:4:10
These expansions are especially useful when ax 1. Under these circumstances

the function can be approximated by the rst few terms in its expansion.
588
References
1. Kreyszig, E. Advanced Engineering Mathematics, 8th ed.; John Wiley: New York,
1999.
2. Abramowitz, M.; Stegun, I. A. Handbook of Mathematical Functions; Applied
Mathematics Series 55; National Bureau of Standards, Washington, DC, 1964.
Appendix B
The Laws of Electricity

and Magnetism
In this appendix, the basic laws of electricity and magnetism are reviewed. These
laws are conveniently expressed in terms of Maxwells equations. They lead to the
fundamental equations describing the propagation of electromagnetic radiation.
In addition, they involve vectors and vector calculus. Thus, it is also convenient to
review the basic operations of vector calculus in the following summary. In this
textbook, vectors are denoted in bold type. Two examples of vectors arising in
electricity and magnetism are the electrical eld E and the magnetic eld B. The
magnetic eld B is usually called the magnetic induction.
Electrostatics gives a description of the forces between charges distributed in a
known manner in space. This description is greatly facilitated by introducing the
electrical eld E which is the force per unit charge q. Both the force F and the eld
E are vectors whereas the charge q is a scalar quantity. The relationship between
F and E is
F qE
The eld due to a point charge in a vacuum is
q
E
4pe0 r2
B:1
B:2
where r is the distance from the charge to the point where the eld is measured,
and e0, the permittivity of free space. In the SI system of units, e0 is equal to
8.8541012 F m1. The force between two charges q1 and q2 separated by a
distance r12 is
589
590
q1 q2
4e0 r212
B:3
This is Coulombs law.

Gauss law relates the eld due to a collection of charges to the magnitude of
charges. Thus,
X qi
E n da
B:4
e0
i
S
S is a surface which surrounds the charges, for example, a sphere. n is a vector

perpendicular to the surface. The integration is carried out over the surface area a.
Both Coulombs law and Gauss law have been written here for charges in a
vacuum. This law may be used to estimate the eld due to a uniform distribution
of charges. A well-known example is the eld due to charge on the surfaces of a
parallel-plate capacitor. For this system, the eld is
s
E
B:5
e0
where s is the charge density on the plates measured in C m2. It is important to
remember that the eld points in the direction that a positive test charge would
move. The vector E points from the positively charged plate to the negative one.
The electrical potential is dened as the work done to move a test charge from
one point in a eld to another. Suppose the potential dierence is estimated
between a point located a distance r1 from a point charge q to a point located
a distance r2. The dierence in potential is

q
q
1
1
f f
dr

4pe0 r2 r1
4pe0 r2
r2
B:6
r1
The minus sign expresses the fact that work is done and the energy of the system
increases when the positive test charge moves against the eld. The general expression for the potential due to a point charge q is
q
B:7
f
4pe0 r
The electrical potential is a scalar quantity.
The general result relating the electrical potential to the electrical eld is
E grad f rf
B:8
where grad is a vector operator dened as follows:

gradf
@f
@f
@f
i
j
k
@x
@y
@z
B:9
i, j, and k are unit vectors pointing in the three directions dening the Cartesian
coordinate system. On the basis of equation (B.8), the eld vector points in the
direction that the potential decreases, all directions in space being considered.
The vector operator grad appears often in the description of systems which are
not at equilibrium. A well-known example in chemistry is a concentration gradi-
THE LAWS OF ELECTRICITY AND MAGNETISM
591
ent. In this case the gradient of the concentration, grad c, denes the direction in
which a concentration change occurs, and can be used to estimate the diusion of
the species whose concentration is changing. Another example is a temperature
gradient in a system in which thermal energy is being redistributed.
At this point it is convenient to introduce two other vector operations, namely,
the scalar product and the vector product. The scalar product of two vectors f and
g is denoted as f g and is given by
f g fg cos y
B:10
where f and g are the magnitudes of f and g, and y is the angle between f and g.
The scalar product may also be written in terms of the components of f and g in
the Cartesian system. Thus,
f g fx gx fy gy fz gz
B:11
where fx , fy , fz are the components of f in the Cartesian directions, and gx , gy , and

gz the components of g.
The vector product between f and g is denoted as f g and results in a vector h
pointing in a direction perpendicular to the plane containing f and g. The value of
h in terms of the Cartesian components of f and g is
h fy gz fz gy i fz gx fx gz j fx gy fy gx k
B:12
If the vector f points in the x-direction (fy fz 0) and the vector g, in the
y-direction gx gz 0), then the vector h points in the z-direction.
Now the vector operators dening the derivatives of a vector are introduced.
The divergence operator div simply gives the derivative of the vector in terms of
its Cartesian components. Thus,
div f r f
@fx @fy @fz
@x @y @z
B:13
According to this equation the divergence of a vector is given by the scalar

product of the vector operator r and the vector. The other operator dening a
vector derivative is curl. It is dened as follows:

@fy @fx
@fx @fy
@fx @fz

curl f r f
i
j
k
B:14
@y @z
@z @x
@x @y
Returning to electrostatics, a general result following from equation (B.4) is
div E
rz
e0
B:15
where rz is the charge density in space. Combining this with equation (B.7), one
obtains
div gradf r rf r2 f
rz
e0
B:16
This is Poissons equation in vacuum. In a medium with relative permittivity es ,

Poissons equation is
592
r2 f
rz
e0 es
B:17
The operator r2 is called the Laplacian. In Cartesian coordinates the Laplacian of

f is given by
r2 f
@2 f @2 f @2 f
@x2 @y2 @z2
B:18
When the charge density in the medium is zero, Poissons equation reduces to
Laplaces equation, that is,
r2 f 0
B:19
Much has been written about the solution of Laplaces equation in both Cartesian
and spherical coordinates.
Several important relationships are relevant to the properties of a dielectric.
The vector P dened for a dielectric is the net dipole moment per unit volume.
When it is combined with the electrical eld, one obtains the denition of the
electric displacement D. Thus,
D e0 E P
B:20
D e0 es E
B:21
Alternatively, one may write

where es is the relative permittivity of the dielectric medium; it is a dimensionless
quantity which is also called the dielectric constant. Combining equations (B.20)
and (B.21), it follows that
P e0 Ees 1
B:22
div D rz
B:23
Finally, it can be shown that

This is one of the fundamental Maxwell equations. The other Maxwell equation
for stationary charges is
curl E 0
B:24
When charges move, one must consider the associated magnetic eects. In the
case that a point charge q is moving, the force on it is
F qE v B
B:25
where v is the velocity of the charge and B, the magnetic induction. B is the
fundamental magnetic quantity just as the electrical eld E is the fundamental
electrical quantity. The SI unit for B is the tesla, or weber m2.
The magnetic induction at a distance r from a charge q moving with velocity v
is given by the BiotSavart law:
m vr
B 0q 3
B:26
4p jr j
m0 is the permeability of vacuum and is equal to 4p 107 N C2 s2, or 4 107
Tesla m A1. This can be used to derive an expression for the magnetic induction
THE LAWS OF ELECTRICITY AND MAGNETISM
593
due to a current I in an innitely long straight wire in vacuum at a perpendicular

distance r:
B
m0 I
2pr
B:27
Another one of Maxwells equations can be derived from the BiotSavart law,
namely,
div B 0
B:28
The magnetic dipole moment m can be dened from a constant current I moving in a circular loop. The magnitude of m is given by
m Ia
B:29
where a is the area of the loop. The magnetic induction due to the current
measured at the center of the loop in a direction perpendicular to its area is
B
m0 I
2r
B:30
where r is the radius of the circular loop. The energy of interaction between an
external eld B and the magnetic dipole moment is
u m B
B:31
Magnetic media are classied as diamagnetic, paramagnetic, and ferromagnetic. A diamagnetic medium contains no permanent magnetic dipoles but magnetic dipoles can be induced in an external magnetic eld. A paramagnetic
medium contains permanent magnetic dipoles which are randomly oriented in
the absence of an external magnetic eld. A paramagnetic medium contains permanent magnetic dipoles which are randomly oriented in the absence of an external magnetic eld. In the presence of an external magnetic eld, the dipoles in a
paramagnetic material align with the external eld. A ferromagnetic medium
represents an extreme case of paramagnetism in which the magnetic dipoles are
permanently aligned with respect to one another in domains of nite dimensions.
This property of magnetic materials can only be treated by quantum mechanics.
The ferromagnetic eect decreases with temperature and reverts to paramagnetism at temperatures above the Curie temperature.
In a magnetic medium, the vector M represents the sum of the magnetic dipole
moments in unit volume. In a diamagnetic or paramagnetic medium, M is proportional to the external magnetic induction B. The proportionality constant is
negative in the case of diamagnetism and positive for paramagnetism. The relationship between B and M in a magnetic material in which a current density Jel is
owing is

B
curl
M Jel
B:32
m0
The quantity (B=m0 M is dened to be the magnetic eld H, so that
B m0 H M
B:33
594
Thus equation (B.32) can also be written as

curl H Jel
B:34
A more general form of equation (B.34) derived by Maxwell is

curl H Jel
@D
@t
B:35
Finally, one can dene the relative permeability of a magnetic medium km by the
relationship
B km m0 H
B:36
Maxwells equations are equations (B.23), (B.24), (B.28), and (B.35). Equation
(B.21) gives the relationship between E and D, and equation (B.36), that between
B and H. With these results all of the fundamental relationships in electricity and
magnetism can be derived.
Appendix C
Numerical Methods of
Data Analysis
The proper analysis of experimental data requires careful consideration of the

numerical techniques used. Real data are subject to experimental error which
can have an eect on results derived from the analysis. Often, this analysis
involves tting a curve to experimental data over the whole range or over part
of the range in which experimental observations have been made. When
thermodynamic data are involved, the relationship between the independent
and dependent variable is usually not known. Then, arbitrary functions such as
polynomials in the independent variable are often used in the data analysis.
This type of data analysis requires consideration of the level of error in both
variables, and of the eects of the error on derived results.
In this appendix, methods of numerical analysis used for interpolation,
integration, and dierentiation of numerical data are discussed. Since these
procedures involve curve tting, regression analysis is also considered.
Emphasis is placed on methods in which a minimum of assumptions are made
regarding the functional relationship between the independent variable x and
the dependent variable y.
C.1 The Principle of Least Squares

Two types of error are generally recognized in discussions of experimental data.
One is determinate or systematic error, that is, error which arises from the design
of the experiment. The second is indeterminate or random error which arises
because the experimental conditions cannot be perfectly controlled. It is assumed
595
596
that the latter type of error can be dealt with using statistics. It is the main focus of
the discussion in this section.
Determinate error usually results from experimental equipment which is faulty.
Students usually rst meet this concept in the analytical chemistry laboratory in
determinations of weight and volume. The quality of the equipment used is
reected in the accuracy of the results obtained. Accuracy is a measure of how
close the experimental result is to the truth. For example, if one wishes to make up
a solution of accurately known concentration in a volumetric ask of 100 mL,
both the ask and the balance used must be carefully calibrated. The ask is
calibrated by weighing it empty and then lled with distilled water at a known
temperature. On the basis of the weight of water, and the known density of the
water, one may calculate an accurate volume for the ask when it is properly lled
to the mark. Calibration of the balance is based on the use of standard weights
which do not corrode and which cover the range in mass used in the experiment.
The accuracy of the standard weights and the quality of the volumetric ask
determine the accuracy of the concentration of the solution which is made.
The operators inability to control perfectly all the variables in an experiment
leads to indeterminate error. For example, errors in the concentration of the
solution made up in a volumetric ask can result from uctuations in the room
temperature. Changes in humidity in the room can also aect the weights involved
in the experiment, for example, the weight of a solute which can adsorb water
vapor. Indeterminate or random error aects the precision of the experiment. This
type of error is treated by the laws of probability. If one determines the concentration of a solution each day on ve successive days, the results are the same only
within the precision of the experiment. On the basis of experience, the operator
averages the results, and then estimates the standard deviation in order to have an
indication of the precision of the experiment.
When a given quantity x is observed an innite number of times, the observed
values follow a Gaussian or normal error distribution. Thus, the probability of
observing a value x for the quantity being measured is given by
"

#
1
1 x x 2
exp
Px
2
s
2p1=2 s
C:1:1
where x is the average value of x, and s the standard deviation. The probability
distribution has the well-known shape shown in g. C.1 with a maximum at x
equal to x . The probability function is normalized so that the total probability
found by integrating the area under the curve is unity. The height and width of the
distribution curve depend on the standard deviation, becoming lower and
broader, respectively, as s increases. When the observation deviates from the
mean, or most probable value, by the standard deviation s the probability of
its being observed drops to e-1/2 Pmax, that is, 0.607Pmax. When x x is 2s, this
fraction is 0.135, and when x x is 3s, 0.011. Since the probability of observations with x greater than 3s is less than 1%, such observations are often rejected
from the data pool on the basis that the observed error is not random in nature.
When the same quantity is observed N times, the probability of the resulting
NUMERICAL METHODS OF DATA ANALYSIS
597
Fig. C.1 Plot of the probability of observing x, Px, against the value of x with respect to
its average value x.
distribution of values of xi is given by the product of the probabilities of each

observation
"
#

N
Y
1
1 X xi x N 2
PN
Pxi
exp
C:1:2
2 i
sx
2p1=2 sx
where x N is the estimate of the most probable value of x based on N observations
and sx , the corresponding estimate of the standard deviation. These quantities
dier from the true values x and s, which are only found if x is observed an
innite number of times. In order to determine x N , one must nd the maximum in
the function PN . This occurs when there is a minimum in the negative argument of
the exponential function. Therefore, the condition for a maximum in PN is
Xxi x N 2

minimum
C:1:3
sx
i
This is the principle of least squares applied to the determination of a single
quantity x. The sum of squares is minimized, allowing x N to have the optimum
value. Thus, one has the condition
@
2 X
2
x x N 0
C:1:4
@x N sN i i
Since the summation is carried out over N values of xi , the result is
x N
X xi
i
C:1:5
It follows that the best estimate of x is simply its average value, a result that is
considered intuitively obvious.
598
Two other quantities must be dened in order to complete the description of

the statistical parameters for a single observed quantity. The rst is the variance
vx , which is given by
X
vx
xi x N 2
C:1:6
i
This simply is the sum of the squares of the deviation of the individual observations xi from the average value for the population x N . Expanding the expression
in the summation, the variance may also be written as
!2
X 2
X
xi
xi =N
C:1:7
vx
i
The estimate of the standard deviation is obtained from the root mean square of
the variance so that
v 1=2
x
sx
C:1:8
N1
The factor N 1 rather than N is required in the denominator because one degree
of freedom in the data set has been lost due to the estimation of the average, x N .
The laws for combination of errors may be used to obtain an estimate of the
error in the derived average x N . If y is a function of x, then the variance of y is
given by
2
dy
vy vx
C:1:9
dx
If y is a function of several variables, x1 ; x2 ; c3 ; . . . ; then this relationship is
generalized to give
X @y 2
vy
v xi
C:1:10
@xi
i
Now, applying this to the equation for x N (equation (C.1.5)), one obtains
X vx
v
vx
C:1:11
x
2
N
N
i
The standard deviation of the average is
sx
sx
N1=2
C:1:12
Thus, the standard deviation of the average decreases with the square root of the
number of observations. The above treatment is easily extended to data sets in
which the individual observations have dierent levels of error. In this case the
standard deviation is designated si because it varies with index i. Under these
circumstances a weighted average must be calculated using the relationship
X
X
x N
wi xi =
wi
C:1:13
i
where
wi
1
1
;
vi
s2i
599
C:1:14
vi being the variance of observation i. More details about weighted least-squares

procedures can be found in the monograph by Bevington and Robinson [1].
C.2 Linear Regression

Curve tting is applied when a dependent variable y is measured experimentally as
a function of an independent variable x. The simplest relationship between these
variables is a straight line. Under some circumstances, x and y do not come
directly from experiment but are calculated from the experimental data on the
basis of a theory which predicts that y should be linear in x. In the present
context, the independent variable x is always that with a negligible level of
error, whereas any signicant experimental error is only associated with the
dependent variable y. Under these circumstances the principle of least squares
is applied to estimating the best straight line through the points with respect to the
random error in y.
The equation of the straight line is written as
y a bx
C:2:1
where a is the intercept on the y-axis, and b, the slope. The best values of a and b
are determined in the regression analysis with the requirement that the sum of the
squares of the deviations between the estimated value of y, yci , and the experimental value yi be a minimum. This sum is dened by the equation
X

yi a bxi 2 minimum
C:2:2
i
where the estimate of y is

yci a bxi
C:2:3
The minimization of is performed by optimizing the parameters a and b.

Mathematically, this leads to two equations called the normal equations. In the
present case, they are
X
@
2
yi a bxi 0
@a
i
C:2:4
X
@
2
yi a bxi xi 0
@b
i
C:2:5
and
After simplication by placing the summation operator with each term, one
obtains
600
Na
X
bxi
axi
yi
C:2:6
xi yi
C:2:7
bx2i
where N is the number of values of xi and y. In order to simplify the symbols, this
is written as
Na Sx b Sy
C:2:8
and
Sx a S2x b Sxy
C:2:9
P
P
where Sx equals i xi , Sy equals i yi , and so on.
The equations (C.2.8) and (C.2.9) constitute two equations in the unknowns a
and b. The solution of these equations is
a
Sy S2x Sxy Sx
Qx
C:2:10
and
b
Qxy
Qx
C:2:11
where
Qx NS2x S2x
C:2:12
Qxy NSxy Sx Sy
C:2:13
and
The dierence Qx is the variance of x multiplied by the number of points N; Qxy is

the covariance of x and y also multiplied by N. Quantities like these turn up
frequently in the solution of the normal equations associated with regression
analysis.
Another important quantity associated with the t of the best straight line to
the data is the estimate of the standard deviation in y. From equations (C.2.2),
(C.2.6), and (C.2.7) one obtains
X

yi a bxi 2 S2y aSy bSxy
C:2:14
i
The estimate of the standard deviation in y is then

1=2

sy
N2
C:2:15
where the factor N 2 in the denominator indicates that two degrees of freedom
were lost in the estimation of the intercept a and slope b. One may now also
estimate the errors in these quantities. For the intercept a,
X @a 2
2
s2y
C:2:16
sa
@yi
i
601
On the basis of equation (C.2.10),

@a S2x xi Sx
@yi
Qx
C:2:17
so that
s2a
s2y X 2
Sx xi Sx 2
Q2x i
sy2 X 2
2
Sx2 2xi Sx S2x x2i S2x
Qx i
C:2:18
Simplifying, the result is

s2a
S2x s2y
Qx
C:2:19
In the case of the slope b, from equation (C.2.11),

@b Nxi Sx
@yi
Qx
C:2:20
so that
s2b
X @b 2
i
@yi
s2y
X Nxi Sx 2
i
Q2x
s2y
C:2:21
After simplication, the result is

s2b
Ns2y
Qx
C:2:22
The nal quantity required in order to assess the quality of the t is the
correlation coecient r. It arises from a comparison of the results of a t of y
as a function of x with those from a t of x as a function of y. The denition of the
correlation coecient is
r
Qxy
Qx Qy 1=2
C:2:23
where
Qy NS2y S2y
C:2:24
The value of r is zero when there is no correlation and reaches a value of 1 when
there is perfect correlation. The sign of r is the same as that of the slope b and has
no signicance in assessing the value of the t. Further interpretation of the
correlation coecient is not straightforward without more detailed knowledge
of the statistical properties of the random error in the data sample considered.
In an approximate way, the value of r2 gives the percent of the observed variation
in y with x that has been explained by the correlation obtained.
602
EXAMPLE
Using the following set of data obtain the equation of the best straight line
relating the value of y to the independent variable x. Estimate the standard
deviation of y, and of the slope and intercept of the resulting linear equation.
Finally, estimate the correlation coecient r.
x
1
2
3
4
5
6
7
8
9
10
19.5
9.2
4.0
2.5
0.0
14.2
25.0
26.0
28.0
40.4
First, it is necessary to calculate the sums dened in the development of the

normal equations and the expression for . These are
Sx 55; S2x 385; Sy 98:4; S2y 4405:94; Sxy 1064:3
C:2:25
Now, the values of Qi are calculated:

Qx 10 385 552 825
C:2:26
Qy 10 4405:94 98:4 34376:84
C:2:27
Qxy 10 1064:3 55 98:4 5231
C:2:28
It is now possible to estimate the best values of the intercept and slope:
a
98:4 385 1064:3 55

25:0
825
C:2:29
5231
6:3
825
C:2:30
and
b
The next step in the analysis is to calculate the error estimates and the
correlation coecient r. On the basis of equation (C.2.14), the value of is
4405:94 25:0 98:4 6:34 1064:3 120:91
and
sy

120:91 1=2
3:9
8
C:2:31
C:2:32
The estimate of the error in a is

sa 3:9

385 1=2
2:7
825
C:2:33
603
and in b,
10
sb 3:9
825
1=2
0:4
C:2:34
Finally, the value of r is

r
5231
0:982
825 34;3771=2
C:2:35
Thus, the tted line explains approximately 96% of the observed variation in y
with x.
The data are plotted in g. C.2, which also shows the best straight line from
the least-squares regression. It should be noted that the calculation must be
carried out using all of the signicant gures. This follows from the fact that
the intermediate steps in the calculation often involve subtraction of two very
large numbers to give a small result (see equations (C.2.26) (C.2.29)). Thus,
regression analysis is often carried out in computer calculations using double
precision arithmetic. This is especially important in multiple linear regression
which is discussed in the following section.
Under some circumstances, the linear relationship between x and y involves
simple proportionality so that the value of y is zero when x is zero. The equation
of the straight line is then
y bx
C:2:36
so that only one parameter, namely, the slope b is determined in a regression

analysis. The least-squares analysis is then considerably simpler. The condition
for a minimum in the sum of squares of the deviations between the observed and
calculated values of y is now
Fig. C.2 Plot of the data used in the example showing the best straight line t by least
squares.
604
X
y1 bxi 2 minimum
C:2:37
This minimum is obtained by optimizing the value of b. It follows that

X
@
2
yi bxi xi 0
C:2:38
@b
i
or
xi yi b
x2i
C:2:39
Using the notation introduced earlier, the solution giving the value of b is
b
Sxy
S2x
C:2:40
Since there is only one parameter to be determined in the regression analysis, there
is only one normal equation (C.2.39).
The error in y may now be estimated using equation (C.2.37). The result is
X

yi bxi 2 S2y bSxy
C:2:41
i
with equation (C.2.39) being used to simplify the expression after expanding the
terms in the summation. The estimate of the standard deviation in y is now

1=2

sy
C:2:42
N1
where N is the number of observations. Since only one parameter is estimated in
this analysis, the factor in the denominator of equation (C.2.42) is N 1. The
estimate of the error in the slope b is found using the equation
X @b 2
s2b
s2y
C:2:43
@y
i
i
On the basis of equation (C.2.40),
@b
x
i
@yi S2x
C:2:44
It follows that
s2b
s2y X
S2x 2
x2i
s2y
S2x
C:2:45
Finally, the correlation coecient r is estimated as before, using equation

(C.2.23).
There are two points that must be remembered in applying the above analysis
to experimental data. The rst is that the analysis assumes that the random error
is associated only with the dependent variable y. Thus, it is important that the
variables be chosen with this fact in mind. In practice, this means that the independent variable x is that quantity which has a lower level of error than that
605
estimated for the dependent variable y. The second point is that the analysis
assumes that the level of error in the dependent variable y is independent of its
magnitude. This is not always the case, either because the level of experimental
error changes over the range that the variable y is observed, or because a quantity
derived from the variable y is used in the regression analysis. Simple examples
arise in the analysis of kinetic data for chemical reactions. In these experiments,
one normally observes the concentration of a reactant or a product as a function
of time. If the reaction is rst order, analysis of the data to obtain the rate
constant requires that the logarithm of the concentration be plotted as a function
of time. If the level of error in the concentration is independent of time, then the
level of error in ln c changes with time. Under these circumstances, weighted least
squares must be used to analyze the data properly. Details describing how to use
weights in regression analysis are available elsewhere [1].
C.3 Multiple Linear Regression

The analysis presented for tting a straight line to experimental data is easily
extended to a curve or to a system with more than one independent variable on
the basis of the principle of least squares.
Consider rst a system where the observed quantity y depends on two independent variables, x1 and x2 . The resulting relationship describes the equation of
a plane in three dimensions and may be written as
y a bx1 cx2
C:3:1
Three parameters, namely, a, b, and c are required to specify the relationship. The
equation describing the principle of least squares is
X
yi a bx1i cx2i 2 minimum
C:3:2

i
The sum is now minimized with respect to each of the adjustable parameters.
This leads to the following three equations:
X
@
2
yi a bx1i cx2i 0
C:3:3
@a
i
X
@
2
yi a bx1i cx2i x1i 0
C:3:4
@b
i
X
@
2
yi a bx1i cx2i x2i 0
C:3:5
@c
i
After simplication, and using the notation introduced earlier, the resulting
normal equations are
Na bSx1 cSx2 Sy
C:3:6
aSx1 bSx2 cSx1 x2 Sx1 y
C:3:7
aSx2 bSx1 x2 cSx2 Sx2 y
C:3:8
606
This system of linear equations is usually solved using determinants, and matrix
diagonalization techniques. Computer programs to do this are easily available.
The important point to remember is that the calculation should be performed in
double precision arithmetic because it often involves subtraction of large numbers
which are close to one another. In determinant notation, the solution to these
equations is
0
1
Sy
S x1
Sx2
@ Sx1 y Sx2 Sx1 x2 A
1
Sx2 y Sx1 x2 Sx2
2
C:3:9
a
D
0
1
N
Sy
Sx2
@ Sx1 Sx1 y Sx1 x2 A
Sx2 Sx2 y Sx2
2
b
C:3:10
D
0
1
N
Sx1
Sy
@ Sx Sx2
S x1 y A
1
1
Sx2 Sx1 x2 Sx2 y
C:3:11
c
D
where
N
D @ Sx1
Sx2
S x1
S x2
1
S x1 x2
1
Sx2
Sx1 x2 A
Sx2
C:3:12
Once the coecients a, b, and c have been found, it is a simple matter to estimate
using equation (C.3.2), and hence the value of the standard deviation y :

1=2

sy
C:3:13
N3
where N is the number of observations. Estimates of the errors in the coecients
sa ; sb , and sc can then be obtained using the same procedure outlined for linear
least squares. The results are
!1=2
Sx2 Sx2 S2x1 x2
1
2
ss sy
C:3:14
D
!1=2
NSx2 S2x2
2
C:3:15
sb sy
D
!1=2
NSx2 S2x1
1
sc sy
C:3:16
D
The same type of analysis applies when data are tted to a quadratic equation
by least squares. The relationship between the dependent variable y and the
independent variable x can be written as
y a bx cx2
and the least-squares condition is
X
yi a bxi cx2i 2 minimum
607
C:3:17
C:3:18
The normal equations are now

Na bSx cS2x Sy
C:3:19
aSx bS2x cS3x Sxy
C:3:20
aS2x bS3x cS4x S2x y
C:3:21
The solution of these equations expressed

0
Sy
Sx
@ Sxy Sx2
Sx2 y Sx3
a
D
0
N
Sy
@ Sx Sxy
Sx2 Sx2 y
b
D
0
N
Sx
@ Sx1 Sx2
Sx2 Sx3
c
D
where
N
D @ Sx
Sx2
Sx
S x2
S x3
in terms of determinants is
1
Sx2
Sx3 A
Sx4
1
Sx2
Sx3 A
Sx4
1
Sy
Sxy A
Sx2 y
1
Sy2
Sx3 A
Sx4
C:3:22
C:3:23
C:3:24
C:3:25
Since the values of xi are raised to the fourth power in this calculation, a sum such
as S4x can be very much smaller or very much larger than the sum Sx . As the
operations of determinant diagonalization demonstrate, this means that double
precision arithmetic must be used in solving the normal equations.
Having obtained values of the coecients a, b, and c estimates of the errors
may be made. The value of sy is again given by equation (C.3.13). The expressions for the errors in the coecients are
!1=2
S2x Sx4 S2x3
C:3:26
sa sy
D
sb sy
NSx4 S2x2
D
!1=2
C:3:27
608
sc sy
NSx2 S2x
D
!1=2
C:3:28
It is clear that the above methods can be easily extended to equations involving
more than two independent variables or to polynomials involving terms of order
higher than x2. Each time a term is added to the right-hand side of the equation
relating the independent and dependent variables, the order of the determinants
dened from the normal equations increases. Although the calculations rapidly
become extremely tedious if carried out by hand, they are easily carried out using
available computer routines.
Unless there is a fundamental theory which gives a functional relationship
between x and y, tting a polynomial over the whole range of data available
should be avoided. This is especially true with thermodynamic data for which
an analytical function relating x and y is usually not available. Numerical analysis
of these data is best carried out using methods in which a segment of data containing a few points are t to a curve. These techniques are described in the remaining
sections of this appendix.
C.4 Numerical Methods

Numerical methods include those based on nite dierence calculus. They are
ideally suited for tabulated experimental data such as one nds in thermodynamic
tables. They also include methods of solving simultaneous linear equations, curve
tting, numerical solution of ordinary and partial dierential equations and
matrix operations. In this appendix, numerical interpolation, integration, and
dierentiation are considered. Information about the other topics is available in
monographs by Hornbeck [2] and Lanczos [3].
The numerical procedures used to deal with experimental data are illustrated
with results from an experiment in chemical kinetics recorded in table C.1. The
reaction is rst order but that will be ignored in the analysis of the data. Note that
the data points have been obtained for equal increments in the independent variable time. It turns out that numerical techniques are especially easy to apply when
this is the case. The second feature of this data set is that the precision of the time
data is much higher than that of the concentration data.
The same data are plotted in g. C.3. It is apparent that there is some scatter of
the results about a smooth curve tted through the experimental points. The
smooth curve was obtained by tting a third-degree polynomial to the data by
least squares. However, since the reaction is rst order, this is not a correct
description of the relationship between c and t. The third-degree polynomial
represents only an approximate description of these data in the range that they
are available and illustrates that curve tting over the whole data range is often
not a good idea. Serious errors in interpreting the results would be expected if this
curve were extrapolated to values of time greater than 130 s.
In order to avoid making assumptions about the functional relationship
between the two variables involved in the experiment, it is recommended that
609
Table C.1 Values of the Fraction of

Reactant Remaining as a Function of
Time in a Solution Kinetics Experiment
Time
t=s
Fraction
Unreacted
c=c0
0
10
20
30
40
50
60
70
80
90
100
110
120
130
1.000
0.923
0.810
0.734
0.670
0.617
0.539
0.500
0.460
0.397
0.368
0.339
0.298
0.279
numerical procedures involve only a small segment of the data. This is illustrated
by applying these techniques to a segment involving ve points. A low-order
polynomial, specically, a quadratic equation, can be t to the three central points
explicitly, or to all ve points by least squares. Calculations such as interpolation
and dierentiation are limited to the region around the central point. One moves
the segment of data considered through the array adding one point on one side
and dropping one point from the other. For example, suppose a segment of three
points is chosen. The rst calculation involves data obtained at 0, 10, and 20 s
Fig. C.3 Plot of the fraction unreacted against time using the data in table C.1.
610
with emphasis on results near or at 10 s. The calculation can then be carried out
for the interval from 10 to 30 s with emphasis on results near or at 20 s. One
moves through the data set segmentally adding one point at a higher value of time
and dropping one at a lower value. If smoothing is required, the segment should
contain ve points with a quadratic t to these data by least squares. If results are
limited to the region of the central point, then the rst calculation with smoothing
is only available at or near 20 s in the present example. This segmental technique is
now illustrated for several numerical procedures.
C.5 Numerical Interpolation

The problem of numerical interpolation is considered rst. Suppose that an estimate of the fraction unreacted is required at 52 s. Since data are available at 50 s
and 60 s the simplest way to estimate c=c0 is to draw a straight line between the
two adjacent points and thereby estimate c=c0 . However, one sees clearly from the
graph that the data do not fall on a straight line, so an improved estimate is
obtained by using the three closest data points and tting them to a quadratic
equation. In order to keep the following discussion general, time is referred to as x
and c=c0 as y. Thus the equation to be t to the three points is
C:5:1
y a bx cx2
Estimation of a, b, and c requires that three equations based on the data at 40, 50,
and 60 s be solved. The problem can be made simpler by redening the coordinate
system for the interval in question so that the central point corresponds to (0, 0) in
a new Cartesian system. Thus, new variables x and Z are dened with
x x x0
C:5:2
Z y y0
C:5:3
and
where x0 and y0 are the values of x and y in the center of the interval. This
transform is illustrated for the ve points centered at 50 s in table C.2.
Table C.2 Transformation of
Data Centered at 50 s to a New
Coordinate System
Position
2
1
0
1
2
30
40
50
60
70
0.734
0.670
0.617
0.539
0.500
20
10
0
10
20
0.117
0.053
0
0.078
0.117
611
Since the curve goes through (0, 0) in the new coordinate system, the quadratic
equation becomes
Z bx gx2
C:5:4
Now, only two parameters need to be obtained, namely, b and g. On the basis of
the data at position 1 (40 s), one has
Z1 bx1 gx21
C:5:5
Z1 bx1 gx21
C:5:6
and at position 1 (60 s),
These equations are now solved to obtain the values of b and g. The results are
b
x21 Z1 x21 Z1

x1 x21 x1 x21
C:5:7
x1 Z1 x1 Z1
x1 x21 x1 x21
C:5:8
and
For the present example, the estimates are

100 0:053 100 0:078
6:55 103
b
10 100 10 100
C:5:9
and
g
10 0:078 10 0:053
1:25 104
10 100 10 100
C:5:10
It is now a simple matter to estimate the value of y at 52 s. In the new

coordinate system this corresponds to x equal to 2. Thus, from equation (C.5.4)
Z 6:55 103 2 1:25 104 4 0:0136
C:5:11
Transforming back to the original coordinate system,

y y0 Z 0:617 0:0136 0:603
C:5:12
Thus, the estimate of y at 52 s is 0.603.

When the values of x and y are reported for equal increments in x, the
equations dening b and g are simpler. In this case,
x1 x1 h
where h is the value of the increment. Then the value of b is given by
Z Z1
b 1
2h
C:5:13
C:5:14
and that of g by
g
Z1 Z1
2h2
C:5:15
When the level of error in the dependent variable is high, the interpolation
procedure should be carried out with smoothing. This could involve tting the ve
612
points centered near the point of interest to a quadratic equation by least squares.
Using the change of variables discussed above, this t may be realized by tting
equation (C.5.4) to the data in table C.2 by least squares. The normal equations
giving the values of b and g by this technique are
SxZ bSx2 gSx3
C:5:16
Sx2Z bSx3 gSx4
C:5:17
and
where the sums involve the values of x and Z recorded in table C.2. These equations are easily solved to obtain b and g.
When the values of x are equally spaced, the least-squares analysis based on the
data with transposed coordinates gives particularly simple results. First of all, the
sums Sx and Sx3 are zero. Thus, the normal equations immediately give the result
that
b
SxZ
Sx2
C:5:18
and
g
Sx2 Z
Sx4
C:5:19
Expanding the sums and simplifying, the estimates of b and g are

b
2Z2 Z1 Z1 2Z2

10h
C:5:20
4Z2 Z1 Z1 4Z2

34h2
C:5:21
and
For the example being considered, the estimates are

2 0:117 0:078 0:053 2 0:117
5:99 103
10 10
C:5:22
4 0:117 0:0078 0:053 4 0:117

7:353 106
34 100
C:5:23
b
and
g
On the basis of equation (C.5.4), the estimate of Z at x 2 is 0.0120 so that the

estimated value of y is 0.605. This does not dier greatly from the previous
estimate. The dierence reects the level of error in the original data.
C.6 Numerical Integration

Now, the procedure for numerical integration is examined. Since random error
tends to cancel out in the additive procedure associated with numerical integration, smoothing routines are normally not used. Suppose that integration is
613
carried out in a segmental fashion involving three points per segment. On the
basis of the analysis presented above, the integral involving data points i 1,
i, and i 1 can be written as
i1
zi1 zi1
a bx cx2 dx
C:6:1
i1
where zi is the value of the integral at point i. This integral can also be written in
terms of the transformed equation (C.5.4). It then becomes
i1
yi bx gx2 dx
zi1 zi1
C:6:2
i1
After integration one obtains

b
g
zi1 zi1 yi xi1 xi1 x2 i1 x2 i1 x3 i1 x3 i1
2
3
C:6:3
Given the values of yi , b, and g for the interval being considered one can easily
calculate the increment in the integral, zi1 zi1 .
When the data points are equally spaced with respect to the independent variable, equation (C.6.3) can be considerably simplied. Recalling that xi1 is equal
to h and xi1 to h, this equation becomes
2h3
g
3
C:6:4
1
yi1 2yi yi1
2h2
C:6:5
zi1 zi1 2hyi

On the basis of equation (C.5.15), g is given by
g
Substituting this expression into equation (C.6.4) and simplifying, one obtains
h
zi1 zi1 yi1 4yi yi1
3
C:6:6
This result is known as Simpsons rule, which provides a very simple and convenient way to perform numerical integration.
The above result may be generalized to a wider interval in the data set provided
the number of data points is odd. The procedure involves adding up the contributions to the integral from segments of the data involving three points. After
calculating the area associated with the rst three points, points 1 and 2 are
dropped and points 4 and 5 added. One then adds the contribution under the
new segment to that from the previous one. In this way, the total integral is
evaluated. Consider an interval involving 2n 1 equally spaced points starting
with the nth point, and increasing through 0 to the nth point. The value of the
integral is then
614
h
zn zn yn 4yn1 2yn2 4yn3

3
2yn2 4yn1 yn
C:6:7
When the integration involves an equal number of equally spaced data points,
a special procedure must be devised to estimate the area associated with the last
segment in the data set. Suppose the last three points are designated n 2, n 1,
and n, where n is an even number. Simpsons rule may be applied in the normal
way up to the (n 1)th point. In order to estimate the area associated with the last
segment located between the (n 1)th and nth point, a quadratic equation is tted
to the last three points. Using the above denitions of b and g, this yields the
following values:
b
Zn2 Zn
2
C:6:8
Zn Zn2
2
C:6:9
The value of the integral at the nth point is then

n
zn zn1
yn1 bx gx2 dx
C:6:10
n1
Since the value of xn1 is 0 and xn is h, the result is

zn zn1 hyn1
b h2 g h3
2
3
C:6:11
Numerical integration techniques are easily applied in the analysis of experimental data which are acquired digitally. Thus, values of the electrical current
which are obtained as a function of time for a xed interval between observations
are easily converted to electrical charge by numerical integration.
The technique for numerical integration applied here is one of the simplest and
best known. However, there are a variety of other techniques which have been
used and which may be more suitable in certain applications [24].
C.7 Numerical Differentiation

Numerical dierentiation of real experimental data is a dicult procedure
because of the experimental error which is typically present. This means that
data smoothing is usually required before the rst derivative is calculated. On
the basis of the segmental technique described here, the rst derivative of y with
respect to x is given by
dy
b 2cx
dx
C:7:1
where the quadratic equation t to the data in the given segment has been used. In
terms of the transformed data, this result may be written
dy dZ
b 2gx
dx dx
615
C:7:2
Calculation of the derivative should be limited to values of x close to or at the

central point in the interval.
Suppose one wishes to estimate dy=dx for the data in table C.1 at 50 s. In the
rst calculation performed above, a quadratic equation was t to three points
with the central point at 50 s and values of b6:55 103 and g1:25 104
estimated without smoothing. Since x 0 at t 50 s, the estimate of dy=dx on the
basis of equation (C.7.2) is 6:55 103 s1 . This corresponds to the negative of
the rate reaction in terms of the fraction unreacted.
This procedure was carried out in a sequential fashion moving through the
data set, making the rst calculation at 10 s, and the last at 90 s. The results are
plotted in g. C.4. The level of error in the rst derivative is obviously quite high,
approximately 10 times that in the observed value of y. This suggests that estimation of dy=dx should be carried out with some smoothing.
When data smoothing is introduced on the basis of a quadratic t to ve
adjacent points, the values of b and g are estimated on the basis of equation
(C.5.16) and (C.5.17). These results are also presented in g. C.4. The scatter in
the estimates of the derivative is signicantly reduced when local smoothing of the
data is used.
When the data points are equally spaced as is the case in the present example,
the rst derivative may be estimated in a much simpler way. When smoothing is
not used and the derivative is estimated at or close to the central point of these
values of x, then the relationships derived earlier apply. The values of b and g are
estimated using equations (C.5.14) and (C.5.15) and used in equation (C.7.2) to
obtain the value of dy=dx. When smoothing is applied, and a quadratic equation
is tted to ve data points, the derivative at or near the central point can be
Fig. C.4 Plot of the rst derivative of the data in table C.1 estimated using a segmental
quadratic t without data smoothing (*) and with smoothing (*).
616
estimated using values of b and g obtained by applying equations (C.5.20) and

(C.5.21).
In summary, the techniques discussed here are designed to perform simple
numerical procedures with experimental data while making the minimum of
assumptions regarding the relationship between x and y. They are especially
important in the analysis of thermodynamic data where procedures such as
numerical dierentiation are often called for. They are very easy to set up on a
computer using programming languages such as BASIC and FORTRAN. Since
high-order polynomials are not involved in the analysis, no special precautions are
needed regarding the precision of the data at intermediate steps in the data analysis. Further information may be found in the following references.
References
1. Bevington, P. R.; Robinson, D. K. Data Reduction and Error Analysis for the Physical
Sciences; McGraw-Hill: New York, 1992.
2. Hornbeck, R. W. Numerical Methods; Prentice-Hall: New York, 1975.
3. Lanczos, C. Applied Analysis; Pitman: London, 1957.
4. Press, W. H.; Flannery, B. P.; Teukolsky, S. A.; Vetterling, W. T. Numerical Recipes,
The Art of Scientic Computing; Cambridge University Press: Cambridge, 1989.
Index
acceptor number, 191

activated complex, 324
activation energy, 323
for electron transfer, 348
activity coecients, 38, 116
electrolyte, 112, 120, 464
Henrys law , 35
ion pair, 141
Raoults law, 21, 26, 31
regular solution, 26
Wilson model, 31
adiabaticity of electron transfer, 347
adsorption at interfaces
of ions, 558
of molecules, 569
adsorption isotherm
liquid | gas interface, 401
metal | solution interface, 566, 573
unimolecular lms, 436
atom transfer reaction, 321
Born model, 102
Bronsted equation, 375
capacitance bridge, 519
chemical exchange reaction, 358
chemical potential, 14, 78, 273
chemical shift, 219, 221, 223, 224

compensation potential dierence, 408, 417,
461
computer simulations, 73
molecular dynamics, 74
Monte Carlo, 74
concentration, 4
conductivity of ions, 274
molar, 277
in non-aqueous media, 294
proton, 298
in water, 283
continuity equation, 257
coordination number, 85, 210, 244
correlation functions, 61
direct, 70
overall, 69
pair, 62
partial pair, 68
total, 64
correlation parameter, 161
Debye, P. J. W., 148
Debye equation, 157
DebyeHuckel model, 121
constants, 126
extended law, 126
617
618
INDEX
DebyeHuckel model (continued)

limiting law, 127
reciprocal length, 123
DebyeOnsager model, 288
constants, 292
dielectric relaxation, 169, 180
dierential capacity, 519
diraction experiments, 65, 206
neutron, 66, 206, 210
surface X-ray, 528
X-ray, 65, 206
diuse double layer, 531, 542
dierential capacity of, 546
ionic surface excesses of, 550
potential drop across, 546
potential prole of, 549
diusion, 258, 264
from a wall, 266
diusion coecient, 258, 264, 283
diusion controlled reactions, 329
DimrothReichardt parameter, 191
dipole moment, 53
of adsorbate, 563, 565
dipole potential, 533, 536
distribution functions, 62
Donnan potential dierence, 484
donor number, 193, 225
elastic deformation of crystals, 514
electrical displacement, 156, 170
electrical polarization, 155
distortional, 169
orientational, 169
electrical potential, 590
discreteness-of-charge potential, 273, 567
inner potential, 272, 375
macropotential, 121, 273
micropotential, 121, 273
of point charge, 53, 104
of point dipole, 53
outer potential, 395
surface potential, 395
electroanalysis, 474
electrocapillary curve, 513
electrocapillary maximum, 518
electrochemical cell, 448
concentration cell, 468
electrochemical potential, 271
elecrode potential, 451
absolute, 460
standard, 454
electrode reaction, 449

type I, 454
types II and III, 456
electrolytes, 112
strong, 112
weak, 115
electron density in metals, 539
electron transfer reactions, 314
heteronuclear, 314
homonuclear, 314, 316
inner sphere mechanism, 315
outer sphere mechanism, 315
theory of, 346
electrophoretic eect, 289
electrostatic equilibrium, 508
EMF, 449
ensemble, 48
canonical, 48
grand canonical, 50
microcanonical, 50
enthalpy, 8
activation, 325
excess mixing, 22
mixing, 18, 25
solvation, 100
Born model, 105
MSA model, 106
transfer, 188
entropy, 8, 49
activation, 325
excess mixing, 22
hard spheres, 83
mixing, 17
solvation, 100, 103
Born model, 104
MSA model, 108
enzyme catalysis, 311
equilibrium constant
determination of, 466
ion association, 136
ion partition, 428
weak electrolyte dissociation, 115
femtochemistry, 338
ash photolysis, 338
FloryHuggins isotherm, 406
uid
hard sphere , 79
LennardJones, 59
uorescence decay time, 345
ux of mass or energy, 256, 270
INDEX
Fourier transform, 584

frequency dispersion, 521
Friedman equation, 178
Frumkin, A. N., 383
Frumkin isotherm, 406
Galvani potential dierence, 396, 428, 472
Gauss theorm, 256
Gibbs adsorption isotherm, 510
GibbsDuhem relationship, 10, 39, 117
Gibbs energy, 9, 50
activation, 325, 348
excess mixing, 22, 30
hard spheres, 83
mixing, 17, 26
solvation, 100, 103
Born model, 104
MSA model, 107
transfer, 185
Gibbs model for interfaces, 391
GouyChapman model, 531
Guggenheim model for interfaces, 392
Hamiltonian, 50
Hammett equation, 376
heat capacity, 49
Helmholtz energy, 9, 49, 77
hard spheres, 82
Helmholtz layer, 530
Henderson, D. J., 45
Henderson equation, 480
Henrys law, 35
Hildebrand, J. H., 3
hydrolysis reaction, 317
indicator electrode, 474
infrared spectra of solvents, 232
acetone, 237
acetonitrile, 236
dimethylformamide, 238
formamide, 235
methanol, 234
water, 233
infrared spectroscopy, 227
attenuated total reection, 228
of electrolyte solutions , 242
of non-electrolyte solutions, 239
inner Helmholtz plane, 532
inner layer, 552
integral equations, 70
hypernetted chain, 71
619
mean spherical, 72
OrnsteinZernike, 71
PercusYevick, 71
interface, 383
liquid | liquid, 426
metal | solution, 422
non-polarizable, 425, 431
polarizable, 425, 431, 510
solution | air, 401, 427
interfacial spectroscopy, 437, 524
interfacial tension, 385, 389, 517
intermolecular forces, 52
intermolecular potential, 63, 71
internal energy, 7, 48, 76
hard spheres, 82
ionion interactions, 111
DebyeHuckel model, 121
MSA model, 130
ion selective electrodes, 494
composite, 500
crystalline, 499
glass, 497
liquid membrane, 499
ion solvation, 100
in non-aqueous solvents, 184, 223
in water, 103, 219
ionsolvent interactions, 100
Born model, 202
MSA model, 106
ionic association, 135
Bjerrum model, 138
equilibrium constant, 136
Fuoss model, 140
MSA model, 140
ionic conductivity, 286, 295
ionic radii, 97
Pauling, 97
ShannonPrewitt, 97, 99
ionic specic adsorption, 531, 558
ionic strength, 123
ionophores, 491
irreversible processes, 255
isomerization reactions, 343, 365
jellium model, 533, 539
Kalman, E., 204
Kenrick, F. B., 410
Kirkwood correlation parameter, 161, 166
Kohlrausch equation, 283
KoppelPalm equation, 198
KrygowskiFawcett equation, 196
620
INDEX
Lamor precessional frequency, 215

Langmuir trough, 435
Laplace transform, 267, 582
laser spectroscopy, 338
laws of electricity and magnetism, 589
LennardJones interaction energy, 59
Lewis acids and bases, 191
LewisSargent equation, 481
ligand exchange reaction, 319
NMR spectroscopy and, 363
limiting molar conductance, 284
linear Gibbs energy relationships, 375
linear regression, 599
liquid junction, 449, 471
liquid junction potential, 472, 477, 483
LorentzLorenz equation, 157
magnetic eld, 215, 359, 592
magnetogyric ratio, 214
Marcus, R. A., 304
Marcus inverted region, 349
maximum bubble pressure experiment, 388
Maxwells equations, 589
medium eects in solution kinetics, 366
dynamic eects, 369
static eects, 367
membrane potentials, 484
membranes, 484
glassy and crystalline, 485
liquid, 491, 499
mobility, 277
molar polarization, 157
molar refraction, 157
multiple linear regression, 605
NavierStokes equation, 261
Nernst, W. H., 447
Nernst equation, 451, 456
neutralization reaction, 317
Nikolsky equation, 489
nuclear magnetic resonance (NMR)
spectroscopy, 213
and chemical exchange reactions, 358
numerical methods, 595, 608
numerical dierentiation, 614
numerical integration, 612
numerical interpolation, 610
Onsager, L., 254
Onsager reciprocal relations, 257
osmotic coecient, 117
Ostwald, F. W., 95
Ostwald dilution law, 287
outer Helmholtz plane, 531
packing fraction, 73, 80, 132
Palinkas, G., 204
Parsons, R., 508
Parsons function, 559
partial charge transfer, 563
partition equilibria, 427
partition function, 48, 75
congurational, 52, 74
internal, 51
translational, 51
Pekar factor, 353, 373
p-functions, 502
pH, 502
phenomenological coecient, 257, 280
plastic deformation of crystals, 514
Poggendorf compensation potentiometer,
450
PoissonBoltzmann equation, 122
potential energy, 52, 76
potential of mean force, 63, 71
potential of zero charge, 423, 521, 535, 537
power series, 586
precursor complex, 313
pre-exponential factor in rate constant 323,
355
pressure, 8, 49, 77, 80
internal, 18
vapor, 15
principle of least squares, 595
protolysis reaction, 317,
proton transfer reactions, 317
NMR spectroscopy and, 359
Raoults law, 15
Raman spectroscopy, 229
random error, 596
rate of entropy production, 289
reaction coordinate, 324
reaction order, 306
reaction rate constant, 306
real potential, 396
reference electrode, 474
relaxation techniques, 332
eld jump, 338
pressure jump, 337
temperature jump, 336
relaxation time, 343
INDEX
621
Debye, 152, 171, 180, 262

longitudinal, 173, 181
reorganization energy, 351
inner sphere, 351
outer sphere, 353
surface excess, 391, 404

surface potential, 395
solution | gas interface, 412, 414, 416
surface pressure, 403
surface selection rule, 525
saturated calomel electrode, 476

second harmonic generation, 438
selectivity equilibrium, 488
silver | silver chloride electrode, 452, 458,
476
solvatochromic basicity, 193
solvatochromism, 245
solutions,
density, 5, 119
distribution functions, 88
ideal, 15
ideally dilute, 33
non-ideal, 18
regular, 24
solvent monolayer models, 553
cluster, 557
three state, 556
two state, 554
solvent permittivity, frequency dependence,
171
solvent polarity, 198
solvent properties, 149
dielectric, 152
molecular, 153, 198
thermodynamic, 150
spinlattice relaxation, 217, 359
spin quantum number, 214
spinspin relaxation, 217, 359
standard deviation, 598
standard electrode potential, 454
standard hydrogen electrode, 452
static solvent permittivity, 153
Debye model, 154
electrolyte solutions, 176
Kirkwood model, 161
MSA model, 162
Onsager model, 160
stickiness parameter, 164, 166
Stokes and anti-Stokes lines, 231
StokesEinstein equation, 280
Stokes law, 260
Stokes radius, 280, 294
Stokes shift, 347
structure factor, 64, 67, 206
structure function, 68, 206
TATB assumption, 185, 427

time-of-relaxation eect, 289
time resolved solvation studies, 342
transfer of ions, 184
enthalpy, 188
Gibbs energy, 185
transition state theory, 323
transport number, 276, 479
ultrasound experiments, 338
ultraviolet-visible spectroscopy, 245
underpotential deposition, 530
unimolecular lms, 433
variance, 598
vector calculus, 589
vibrational frequency shift, 239
ion induced, 243
solvent induced, 239
vibrational spectroscopy, 226
infrared, 227
interfacial, 524
Raman, 229
vibrational sum frequency spectroscopy,
440
virial adsorption isotherm, 568
viscosity, 259
Volta potential dierence, 397, 461
experimental determination of, 408
metal | solution interface, 424
volume, 6
excess mixing, 22
excess molar, 13
mixing, 18
molar, 7
partial molar, 10
water structure, 84
Wertheim parameter, 107, 163, 168
work function, 398
electrons in metals, 398, 400, 535
ions in solution, 417, 420, 421
YoungLaplace equation, 386

Liquids, Solutions, and Interfaces: From Classical Macroscopic Descriptions To Modern Microscopic Details (Topics in Analytical Chemistry) (Fawcett 2004)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Liquids, Solutions, and Interfaces: From Classical Macroscopic Descriptions To Modern Microscopic Details (Topics in Analytical Chemistry) (Fawcett 2004)

Uploaded by

Copyright:

Available Formats

Liquids, Solutions, and Interfaces

This page intentionally left blank