Statistical Thermodynamics and Microscale Thermophysics

Many of the exciting new developments in microscale engineering are based on

the application of traditional principles of statistical thermodynamics. This book
offers a modern view of thermodynamics, interweaving classical and statistical
thermodynamic principles and applying them to current engineering systems. It
begins with coverage of microscale energy storage mechanisms from a quantum
mechanics perspective and then develops the fundamental elements of classical
and statistical thermodynamics. Next, applications of equilibrium statistical ther-
modynamics to solid, liquid, and gas phase systems are discussed. The remainder
of the book is devoted to nonequilibrium thermodynamics of transport phenomena
and an introduction to nonequilibrium effects and noncontinuum behavior at the
microscale.
Although the text emphasizes mathematical development, it includes many
examples and exercises that illustrate how the theoretical concepts are applied to
systems of scientific and engineering interest. It offers a fresh view of statistical
thermodynamics for advanced undergraduate and graduate students, as well as
practitioners, in mechanical, chemical, and materials engineering.

Van P. Carey is a Professor in the Mechanical Engineering Department at the

University of California, Berkeley. The main focus of his research is development
of advanced computational models of microscale thermophysics and transport in
multiphase systems.
Statistical Thermodynamics
and Microscale Thermophysics

VAN P. CAREY
University of California, Berkeley

CAMBRIDGE
UNIVERSITY PRESS
PUBLISHED BY THE PRESS SYNDICATE OF THE UNIVERSITY OF CAMBRIDGE
The Pitt Building, Trumpington Street, Cambridge, United Kingdom

CAMBRIDGE UNIVERSITY PRESS

The Edinburgh Building, Cambridge CB2 2RU, UK http://www.cup.cam.ac.uk
40 West 20th Street, New York, NY 10011-4211, USA http://www.cup.org
10 Stamford Road, Oakleigh, Melbourne 3166, Australia

© Cambridge University Press 1999

This book is in copyright. Subject to statutory exception

and to the provisions of relevant collective licensing agreements,
no reproduction of any part may take place without
the written permission of Cambridge University Press.

First published 1999

Typset in Times Roman 10/12 pt. in L9TEX2 £ [TB]

A catalog record for this book is available from

the British Library.

Library of Congress Cataloging-in-Publication Data

Carey, V. P. (Van P.)

Statistical thermodynamics and microscale thermophysics / Van P.
Carey.
p. cm.
ISBN 0-521-65277-4 (hb). - ISBN 0-521-65420-3 (pb)
1. Statistical thermodynamics. 2. Thermodynamics. I. Title.
QC311.5.C36 1999
621.402'l-dc21 98-45449
CIP

ISBN 0 521 65277 4 hardback

ISBN 0 521 65420 3 paperback

Transferred to digital printing 2004

To Lee F. Carey, Bart Conta, and Dennis G. Shepherd,
three of the best applied thermodynamicists I ever met
 Contents

Nomenclature x
Preface xv

1 Quantum Mechanics and Energy Storage in Particles 1

1.1 Microscale Energy Storage 1
1.2 A Review of Classical Mechanics 5
1.3 Quantum Analysis Using the Schrodinger Equation 8
1.4 Model Solutions of the Time-Independent Schrodinger Equation 11
1.5 A Quantum Mechanics Model of the Hydrogen Atom 17
1.6 The Uncertainty Principle 20
1.7 Quantum Energy Levels and Degeneracy 23
1.8 Other Important Results of Quantum Theory 25

2 Statistical Treatment of Multiparticle Systems 29

2.1 Microstates and Macrostates 29
2.2 The Microcanonical Ensemble and Boltzmann Statistics 30
2.3 Entropy and Temperature 37
2.4 The Role of Distinguishability 43
2.5 More on Entropy and Equilibrium 50
2.6 Maxwell Statistics and Thermodynamic Properties
for a Monatomic Gas 61

3 A Macroscopic Framework 71
3.1 Necessary Conditions for Thermodynamic Equilibrium 71
3.2 The Fundamental Equation and Equations of State 75
3.3 The Euler Equation and the Gibbs-Duhem Equation 80
3.4 Legendre Transforms and Thermodynamic Functions 83
3.5 Quasistatic and Reversible Processes 90
3.6 Alternate Forms of the Extremum Principle 93
3.7 Maxwell Relations 95
3.8 Other Properties 99

4 Other Ensemble Formulations 107

4.1 Microstates and Energy Levels 107
4.2 The Canonical Ensemble 108
4.3 The Grand Canonical Ensemble 117
4.4 Fluctuations 127
4.5 Distinguishability and Evaluation of the Partition Function 137

vn
viii Contents

5 Ideal Gases 145

5.1 Energy Storage and the Molecular Partition Function 145
5.2 Ideal Monatomic Gases 145
5.3 Ideal Diatomic Gases 149
5.4 Polyatomic Gases 160
5.5 Equipartition of Energy 164
5.6 Ideal Gas Mixtures 167
5.7 Chemical Equilibrium in Gas Mixtures 170

6 Dense Gases, Liquids, and Quantum Fluids 179

6.1 Behavior of Gases in the Classical Limit 179
6.2 Van der Waals Models of Dense Gases and Liquids 182
6.3 Other Models of Dense Gases and Liquids 196
6.4 Analysis of Fluids with Significant Quantum Effects 205
6.5 Fermion and Boson Gases 212

7 Solid Crystals 225

7.1 Monatomic Crystals 225
7.2 Einstein's Model 228
7.3 Lattice Vibrations in Crystalline Solids 229
7.4 The Debye Model 233
7.5 Electron Gas Theory for Metals 238
7.6 Entropy and the Third Law 241

8 Phase Transitions and Phase Equilibrium 245

8.1 Fluctuations and Phase Stability 245
8.2 Phase Transitions and Saturation Conditions 264
8.3 Phase Equilibria in Binary Mixtures 271
8.4 Thermodynamic Similitude and the Principle of Corresponding States 284

9 Nonequilibrium Thermodynamics 297

9.1 Properties in Nonequilibrium Systems 297
9.2 Entropy Production, Affinities, and Fluxes 298
9.3 Analysis of Linear Systems 301
9.4 Fluctuations and Correlation Moments 304
9.5 Onsager Reciprocity of Kinetic Coefficients 309
9.6 Thermoelectric Effects 311

10 Nonequilibrium and Noncontinuum Elements of

Microscale Systems 325
10.1 Basic Kinetic Theory 325
10.2 The Boltzmann Transport Equation 337
10.3 Thermodynamics of Interfaces 346
10.4 Molecular Transport at Interfaces 351
10.5 Phase Equilibria in Microscale Multiphase Systems 357
Contents ix

10.6 Microscale Aspects of Electron Transport in Conducting Solids 375

10.7 The Breakdown of Classical and Continuum Theories
at Small Length and Time Scales 384

Appendix I Some Mathematical Fundamentals 391

Appendix II Physical Constants and Prefix Designations 397
Appendix III Thermodynamics Properties of Selected Materials 399
Appendix IV Typical Force Constants for the Lennard-Jones 6-12 Potential 409

Index 411
Nomenclature

a normalized bulk velocity, = wo/{2NpJi^/M)1^2

aR Redlich-Kwong constant
av van der Waals constant
AE* interface area
^R Redlich-Kwong constant
by van der Waals constant
B second virial coefficient
c particle speed
cp molar specific heat at constant pressure
cv molar specific heat at constant volume
C third virial coefficient
eQ electron charge
E energy
Ex electric field in the x direction
f fugacity
fie) average occupancy of a microstate with energy e
/(W, 25, t) fractional particle velocity distribution
/eP distribution function for an ensemble of particles
/es ensemble number density
/pv particle velocity distribution
f>c fraction of molecules with speeds greater than c
F Helmholtz free energy
8 degeneracy
g(r) radial distribution function
g(v) frequency distribution function
g*(e) distribution function for electron quantum states
8e gravitational acceleration
8 molar Gibbs function
gi degeneracy of energy level /
G Gibbs function
h Planck's constant
h molar specific enthalpy
h molar specific enthalpy of saturated liquid
K molar specific enthalpy of saturated vapor
n = h/2ix
H enthalpy
H Hamiltonian
I moment of inertia
j number flux of molecules
jm mass flux of molecules
Nomenclature xi

Jax flux of species a in the x direction

Js,x flux of entropy in the x direction
Ju,x flux of energy in the x direction
Ja flux of species a
Js flux of entropy
Ju flux of energy
&B Boltzmann constant
kt thermal conductivity
kx force constant for linear restoring force
K wavenumber, = 2n/X
Kc equilibrium constant
Kp equilibrium constant
L characteristic system dimension
Ltj kinetic coefficients
L ms mean spacing between particles
Lvr temperature gradient length scale, = T/WT
Lz Lorentz number, = kt/a&T
m mass per particle
M mass per molecule for polyatomic molecules
M molecular mass
rij number of ensemble members in energy level £j
(ni) mean number of particles in microstate I'
N number of particles
Na number of species a particles
NaJ number of species a particles in energy level /
A^A Avogadro's number
Na number of species a particles per unit volume
N'a instantaneous number of species a particles
p momentum
Pi generalized m o m e n t u m
px m o m e n t u m in the x direction
P =Pr~l
p fluctuation probability density function
pj generalized momentum
P pressure
Pa partial pressure for species a
Pc critical pressure
PY reduced pressure, = P/Pc
Psat equilibrium saturation pressure
P probability
q molecular partition function
qa molecular partition function for particle species a
qe partition function for electronic energy storage
qi generalized coordinate
qmi partition function for internal energy storage
#nuci partition function for nuclear energy storage
qTOt partition function for rotational energy storage
xii Nomenclature

*7rot, nucl partition function for rotational and nuclear energy storage
<7tr partition function for translational energy storage
<7vib partition function for vibrational energy storage
generalized coordinate
Q canonical partition function
re equilibrium bubble or droplet radius
R universal gas constant, = JVA&B
Ru Rydberg constant
s molar entropy
molar internal energy for saturated liquid
h molar entropy for saturated vapor
s system entropy
s entropy per unit volume
S' instantaneous entropy
surface excess entropy
t = Tv-l
T temperature
Tc critical temperature
Tm equilibrium melting temperature
Tm temperature in phase m
Tr reduced temperature = T/ Tc
7-sat equilibrium saturation temperature
u velocity in the x direction
u molar specific internal energy
Uf molar specific internal energy for saturated liquid
ug molar specific internal energy for saturated vapor
U system internal energy
U internal energy per unit volume
U' instantaneous internal energy
u potential function in quantum system
surface excess internal energy
V velocity in the _y direction
V mass specific volume
group velocity
V molar specific volume
Vc critical molar specific volume
Vr reduced specific volume, = v/vc
V = 0r-l
V system volume
w velocity in the z direction
w0 bulk velocity normal to interface
W number of microstates corresponding to a particular macrostate
X Cartesian spatial coordinate
xt mole fraction of species /
y Cartesian spatial coordinate
7 Cartesian spatial coordinate
z compressibility factor, = Pv/RT
Nomenclature xiii

Z partition function for the microcanonical ensemble

Za partition function for the microcanonical ensemble for species a
Zyv configuration integral for single-species system
ZNatNb configuration integral for a binary mixture

Greek
P =l/kBT
/3j coefficient of thermal expansion
y analysis parameter
y ratio of specific heats, = cp/cv
ya activity coefficient for species a
Fa bulk motion correction factor for a > 0
r_ a bulk motion correction factor for a < 0
Vf* surface excess mass
8 liquid film thickness
8Q differential heat interaction (positive into system)
8W differential work interaction (positive out of system)
A^iv molar latent heat of vaporization, = hy — h\
e energy
Si energy level i
Sf energy microstate /'
£F Fermi energy
So characteristic energy in the Lennard-Jones 6-12 pair potential
sA absolute Seebeck coefficient of material A
SAB thermoelectric power for a thermocouple of materials A and B
0 angular coordinate in spherical polar coordinates
#D Debye temperature, = hv^/k^
#E Einstein temperature, =
#F Fermi temperature, =
0YOt rotational temperature, = /z 2/(87T 2/& B)
0rotm = 9TOt for linear molecules,
= (#rot,A#rot,##rot,c)1/3 for nonlinear polyatomic molecules
# vib vibrational temperature, = hv/kB
KT isothermal compressibility
X wavelength
A,c mean free path between collisions
A thermal de Broglie wavelength
A lattice wave amplitude
jia chemical potential (per molecule) for species a
/xo Fermi energy
pia molar chemical potential for species a, = /jLaNA
/xm absolute viscosity
v frequency
V>D Debye cutoff frequency
VE Einstein characteristic frequency
yc mean collision frequency
xiv Nomenclature

total number of translational and rotational energy storage modes

grand canonical partition function
Peltier coefficient for a junction of material A and material B
pN number density of particles
ac collision cross section
ae electrical conductivity
<7i interfacial tension
OLJ characteristic distance in the Lennard-Jones 6-12 pair potential
oig interfacial tension at a liquid-gas interface
as symmetry number
ax standard deviation in variable x
ol variance of variable x
a accommodation coefficient
r time interval
rr relaxation time
fA Thomson coefficient for material A
(rc) mean time between collisions
0 intermolecular pair potential
4> electric potential
0 angular coorinate in spherical polar coordinates
4>a fugacity coefficient for species a
<&N potential energy in system containing JV particles
x/r time-independent wave function
^ wave function
co angular frequency, = 2TTV
co acentric factor
Q total number of system microstates
 Preface

The structure of this book is designed to facilitate coherent development of classi-

cal and statistical thermodynamic principles. The book begins with coverage of microscale
energy storage mechanisms from a modern quantum mechanics perspective. This infor-
mation is then incorporated into a statistical thermodynamics analysis of many-particle
systems with fixed internal energy, volume, and number of particles. From this analysis
emerges the definitions of entropy and temperature, the extremum principle form of the
second law, and the fundamental relation for the system properties. The third chapter takes
the concepts derived from the statistical treatment and uses mathematical techniques to
expand the macroscopic thermodynamics framework. By the end of the third chapter, the
full framework of classical thermodynamics is established, including definitions of all com-
monly used thermodynamic properties, relations among properties, different forms of the
second law, and the Maxwell relations.
In the fourth chapter, statistical ensemble theory is covered, building on the initial statisti-
cal treatment in Chapter 2 and the expanded macroscopic framework developed in Chapter 3.
The canonical ensemble and grand canonical ensemble formalisms are developed, and the
relations developed from these formalisms are used to explore the significance of fluctu-
ations in thermodynamics systems. By the end of the fourth chapter all the fundamental
elements of classical and statistical thermodynamics have been established. Chapters 5-7
deal with applications of equilibrium statistical thermodynamics to solid, liquid, and gas
phase systems.
The final three chapters of the text cover thermal phenomena that involve nonequilib-
rium and/or noncontinuum effects. Throughout these chapters, implications of the statistical
thermodynamics theory based on microscale characteristics of the system are considered in
tandem with predictions of classical thermodynamics. Chapter 8 presents an integrated treat-
ment of multiphase systems, which includes classical aspects of phase transitions and phase
equilibria as well as nonequilibrium features associated with metastable states and intrinsic
stability limits. It draws upon the van der Waals model to illustrate how the fundamental
equation together with necessary conditions for equilibrium define the saturation condi-
tions and the limits of intrinsic stability for a pure fluid or a binary mixture. In this chapter,
thermodynamic similitude and near critical behavior are also examined from statistical and
classical thermodynamic viewpoints.
Chapter 9 covers nonequilibrium thermodynamics of transport phenomena. Here again,
classical as well as microscale statistical perspectives are considered. The relations among
fluxes and potentials are obtained from classical theory, while the key Onsager reciprocity
relations are obtained by considering the statistics of microscale fluctuations. Applications
involving thermoelectric effects are discussed in detail. Chapter 10 uses the tools of statis-
tical and classical thermodynamics developed earlier in the text to analyze nonequilibrium
effects and noncontinuum behavior at the microscale. There are many important examples
in which these effects are important. The intent in Chapter 10 is to introduce these effects
by exploring a limited number of important examples. Kinetic theory and the Boltzmann

xv
xvi Preface

transport equation are developed because they are commonly used as the basis for analysis
of nonequilibrium or noncontinuum systems. Nonequilibrium and noncontinuum aspects
of multiphase systems and electron transport in solids are examined in detail. The final
section of Chapter 10 examines length scales and time scales at which classical and contin-
uum theories become suspect. This defines the range of conditions for which classical and
continuum theories are expected to be accurate. Although limited in its coverage, Chap-
ter 10 is intended to provide an introduction to microscale aspects of nonequilibrium and
noncontinuum phenomena and to illustrate how they relate to the theoretical framework
developed in the preceding chapters.
Many aspects of microscale thermophysics can be fully understood only after a basic
understanding of statistical thermodynamics, nonequilibrium thermodynamics, and kinetic
theory is attained. Chapter 10 is therefore afittingcapstone to this text. It brings the presenta-
tion full circle in the sense that the treatment begins with microscale energy storage, develops
equilibrium classical and statistical thermodynamics theory for macroscopic systems, and
then returns to microscale systems to explore the limitations of macroscopic theory and to
examine how tools from statistical thermodynamics and nonequilibrium thermodynamics
can be used to analyze microscale thermophysics and transport phenomena.
Virtually all other statistical thermodynamics texts develop statistical thermodynamics for
a system containing one species of particle. This makes it difficult to expand the analytical
framework to handle systems that contain a mixture of different particle or molecular
species. In this text, a binary mixture of two species of particles or molecules is considered
throughout the analytical development. By so doing, the pure system behavior is captured
by setting the number of particles of one species to zero, and the interactions between
different particles in a binary mixture are indicated directly. Furthermore, the symmetry
of the terms in the resulting relations for the two species generally makes it clear how
the equations should be extended to handle a mixture of three or more molecular species.
This presentation should better prepare engineers for the increasing number of important
applications involving multicomponent mixtures.
Although the text emphasizes mathematical development, it includes examples in each
chapter and exercises at the end of each chapter that illustrate how the theoretical concepts
are applied to systems of scientific and engineering interest. An effort has been made to
include examples and problems that span a wide variety of modern applications. Some of
the exercises ask students to develop and run computer programs to tackle more challenging
problems. Inclusion of these problems and examples is motivated by the author's belief that
a full understanding of these concepts is best developed by working with them directly. Ex-
tensive use of the van der Waals model is made in the examples throughout the text. This is
done for two reasons. First, the van der Waals model can be related back to molecular char-
acteristics through the statistical thermodynamics framework. This connection illustrates
how relations among properties are connected to properties of the molecules themselves.
Second, although very idealized, the van der Waals model exhibits most of the features of
real fluids, including the existence of a critical point and first-order phase transitions. The
examples demonstrate how a model of this type can be used with the results of statistical
and classical thermodynamics theory for quantitative engineering analysis.
The topic sequence here is somewhat unconventional for a statistical thermodynamics
text. I have therefore included a brief introductory remarks section at the beginning of each
chapter of the text to provide the reader with an overview of the contents of the chapter and
a sense of how that chapter fits into the flow of ideas in the text.
Preface xvii

The structure of the book makes it suitable as a primary text for a graduate-level course in
advanced engineering thermodynamics or as a source of foundation material in an advanced
specialty course on microscale transport phenomena. For a graduate course on advanced
engineering thermodynamics, the first seven chapters would be the primary source material
for the course and some special topics from the remainder of the text could be included. For
an advanced graduate course on microscale transport phenomena, the material in Chapters
1-7 could be reviewed briefly and the material in Chapters 8-10 could provide source
material for an introduction to the thermodynamics of microscale systems and microscale
transport phenomena. The foundation thus provided could then be followed by an in-depth
exploration of microscale transport phenomena in different system types.
Thanks are due to reviewers of the early manuscript versions of this text: Richard Buckius
of the University of Illinois, Edward McAssey, Jr. of Villanova University, and Donald
Fenton of Kansas State University. Their many valuable criticisms and suggestions are
greatly appreciated.
 CHAPTER 1

Quantum Mechanics and Energy

Storage in Particles

This text differs from most statistical thermodynamics textbooks in that it does
not deal exclusively with statistical aspects of thermodynamics. Instead, it attempts to
weave together statistical and classical elements to develop the full theoretical framework
of thermodynamics. Chapter 1 begins this development by establishing the basic features
of energy storage at the atomic and molecular levels. It contains a very short introduction
to basic aspects of quantum mechanics. The quantum models discussed in this chapter
are models of energy storage modes found in common molecules. Conclusions regarding
energy levels and their degeneracy for these modes of energy storage are cornerstones of
the statistical thermodynamic theory developed in later chapters.

1.1 Microscale Energy Storage

Since this text is designed for graduate-level engineering instruction, it is likely that
the reader has already encountered some elements of thermodynamics in previous courses
and very likely that he or she has some idea of the usefulness of thermodynamic analysis
for systems of scientific and technological interest. Mechanical or chemical engineers who
are thoroughly versed in classical equilibrium thermodynamics may wonder what a statis-
tical development of thermodynamics has to offer beyond the tools provided by classical
thermodynamics.
In response to such an inquiry, we can identify two main benefits of developing a statis-
tical thermodynamic theory. First, by design, statistical thermodynamics theory provides a
link between macroscopic "classical" thermodynamic analysis of system behavior and the
microscopic characteristics of the atoms, molecules, or subatomic particles that make up the
system. To better understand the nature of establishing such a link, consider the length-scale
and time-scale ranges shown in Figure 1.1. The coordinate axes in this diagram span just
about the entire range of length and times scales associated with systems and processes
in the known universe. The shortest length scales are associated with the interior structure
of atoms and the dimensions of subatomic particles, while the largest length scales are
associated with intergalactic distances.
In a similar manner, the shortest time scales are associated with microscale interactions
of atoms and/or fundamental particles, and the longest is the age of the universe. Relativity
theory requires that matter and energy cannot move faster than the speed of light. Since
natural processes generally are the result of the motion of matter or photons, processes
having very large length scales and very small times scales cannot be attained. Hence the
upper left region of the diagram is in some sense a forbidden zone. Statistical thermodynam-
ics is conceptually important because it provides the theoretical link between microscale
interactions and systems undergoing processes at much larger length and time scales.
The second benefit of developing statistical thermodynamic theory is that it provides ana-
lytical tools for engineering and scientific analysis. With increasing frequency, engineers
and scientists are seeking to work with systems whose behavior lies outside the bounds of

1
 1 I Quantum Mechanics and Energy Storage in Particles

10"" i

18
io -
evolution ol
the universe

1012" j

geologic

i
io6- processes
on earth Y

common processes
AL
(m) - ^> detectable by
human senses
>
V
s relativistic limit
10-6-
1
*i 1
molecular,
atomic and
subatomic
-12 interactions
10
-15
10 18
to" 1 5 io- 1 2 10" 1 10 10 10
At (s)

Figure 1.1 Length and time scales of known processes.

classical thermodynamic analysis. The areas of microelectro-mechanical systems (MEMS),

fusion reactor development, and ultrafast laser processing of materials are just three exam-
ples of engineering applications in which important phenomena are beyond the bounds of
classical thermodynamics theory. Engineers specializing in the thermosciences will need
tools from statistical thermodynamics and kinetic theory to deal with these and other new
applications areas.
In developing the theoretical foundation of thermodynamics, we first seek to define how
energy is stored in systems composed of large numbers of particles. In this text, the term
particles is meant to include subatomic particles, atoms, or molecules. As a first step in
dealing with such systems, we must first establish how individual particles themselves store
energy. Some insight into this issue can be obtained by considering the simple models
of atomic and molecular structures shown in Figure 1.2. In general, such particles may
store energy as translational energy or by modifying their internal state. Internal energy
storage is directly linked to the internal structure of the particle. In a single atom, internal
energy storage or release is accompanied by changes in nuclear or electronic energy levels.
Molecules composed of two or more atoms may also store energy internally by altering the
nuclear state or electron energy levels of one or more of the atoms. In addition, molecules
may store energy in kinetic energy of rotation about the molecule's center of mass and in
1.11 Micro scale Energy Storage

O O

N9 H2O CO9

Figure 1.2 Models of typical molecular structures.

,\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\
OOOO
8 §o

O§
8 „
AWWWWW A\\\\\\\\\
Figure 1.3 A typical system of molecules.

vibration of the molecule. The latter mode of energy storage results when atoms oscillate
about their equilibrium positions in the molecular structure.
Our ultimate goal is to develop analytical tools that can be used to analyze a system of
particles, like that shown schematically in Figure 1.3. The particles shown in the figure are
molecules. The system contains a mixture of diatomic molecules and molecules containing
three atoms. Note that translation of the molecules leads to collisions that cause energy to
be exchanged among the different modes of energy storage. At the microscopic level, the
distribution of energy within each molecule and within the system is constantly changing
with time.
In analyzing systems like that shown in Figure 1.3, we will assume that the mass of
each species of particle and energy are conserved at both the microscopic and macroscopic
level. Conservation of energy is a key foundation element in both classical and statistical
thermodynamics. For that reason, conservation of energy is generally referred to as the first
law of thermodynamics. The volume V and the number of each species of particle in the
system Na, Nt,, . . . are taken to be intrinsic properties of any system. We will assume that
they can be measured if necessary and, if desired, we can specify their values.
The internal energy of the system, U, is defined as the total energy of the particles in
the system, including the translational and internal energy, and any energy stored in force
interactions between the particles. If we can isolate the system so no energy crosses its
boundary, the first law requires that the internal energy is constant, independent of time,
even though particle collisions may alter the way in which the energy is stored within the
 11 Quantum Mechanics and Energy Storage in Particles

system. For an isolated system at fixed U, V, Na, Nt>, . . . , a system will macroscopically
attain an equilibrium state that is unchanging with time. At the microscopic level, the system
is constantly changing its configuration. The system thus is in a state of dynamic equilib-
rium. We will see in later sections of this text that the nature of this dynamic equilibrium
strongly influences how the system reacts to changes in the constraints on the system. This
issue is central to thermodynamic analysis of systems, since as engineers and scientists,
we often want to predict how a system will change if a change occurs in one or more of
the macroscopic constraints on the system (such as the specified values of U, V, or particle
numbers).

Example 1.1 The mean speed of a helium atom in helium gas at 27°C is about 1,020 m/s.
Use this information to estimate the internal energy of the gas per kmol.

Solution The mass per helium atom m is known to be 6.65 x 10~~27 kg. Assuming
the internal energy is due only to kinetic energy of the molecules, we can estimate the mean
kinetic energy per atom as

ek = \mc2 = (±) 6.65 x 10- 27 (l,020) 2 = 3.46 x 10~21 J.

Multiplying by Avogadro's number, we obtain the internal energy per kmol:

u = (6.02 x 1026)3.46 x 10~21 = 2.082 x 106 J/kmol = 2,082 kJ/kmol

Pioneering efforts to construct statistical theories of the behavior of large collections of

particles were made by Maxwell, Boltzmann, and Gibbs in the last half of the nineteenth
century. Because these efforts preceded the development of quantum theory, simple classical
mechanics treatments of molecule energy storage mechanisms were incorporated into these
initial statistical mechanics models. In fact, the statistical mechanics models of Maxwell and
Boltzmann were highly controversial when they were first proposed because the existence
of atoms and molecules was not widely accepted by scientists at that time. The work
of Planck, Einstein, and others in the early part of the twentieth century established the
existence of atoms, molecules, and eventually the subatomic structure of matter. Quantum
theory provided the means for predicting the manner in which energy is stored in atoms,
molecules, and subatomic particles. This made it possible to extend the statistical mechanics
concepts developed by Boltzmann and Gibbs to a wide variety of system types.
In this text, the current knowledge of the atomic structure of matter is taken as the starting
point for development of macroscopic thermodynamics. Macroscopic equilibrium thermo-
dynamics is a consequence of the nature of microscale energy storage and the statistical
behavior of very large numbers of particles. We will therefore first establish the basic ele-
ments of quantum theory necessary to determine the rules of energy storage and energy
exchange in atoms, molecules, and subatomic particles. We will then construct a statistical
analysis of a system composed of a large number of particles that obey such rules. Once
these two components are blended together to form a statistical thermodynamic framework,
we will explore the application of the framework to a variety of systems.
1.2 IA Review of Classical Mechanics

1.2 A Review of Classical Mechanics

Classical mechanics analysis of the motion and mechanical energy storage of a
particle can be cast in at least three ways. The Newtonian formulation of classical mechanics
can be stated as

where p is the particle momentum and F is the force exerted on the particle.
The Lagrangian formulation of classical mechanics is stated in terms of the Lagrangian
L defined as

L(x, y, z, u, v, w) = K(u, v, w) - U(x, y, z), (1.2)

where U(x, y, z) is the potential energy of the particle and K is the kinetic energy of the
particle given by

K(u, v, w) = ™[u 2 + v2 + w2]. (1.3)

In the above relation, m is the mass of the particle. Noting, for the x direction, that

dL dK dL dU
— = — =mu = p x, — = -— = F x, (1.4)
au au ox ox
Newton's equation can be written as

dt\duj dx
If we generalize the notation as
x=q\, y = q2, z = q$, u = q\, v = q 2, w = q?>, (1.6)

the equations of motion can then be written as

= ^ , 7 = 1,2,3. (1.7)
dt\dqj

An extraordinary and very useful property of Lagrange's equations of motion is that they
have the same form in any coordinate system (useful since it is sometimes easier mathe-
matically to define U in a particular coordinate system). A set of equations like (1.7) can
be written for each particle in a multiparticle system.
A third formulation of classical mechanics is the Hamiltonian formulation. To construct
the Hamiltonian formulation of classical mechanics, we begin by defining generalized
momenta as
dL
Pj^—r-, 7 = 1,2, ...,37V (for a system of TV particles). (1.8)
dqj

Each generalized momentum pj is said to be conjugate to coordinate qj. Since each momen-
tum is linearly proportional to the velocity qj, it is a simple task to replace any velocity terms
in the formulation with the corresponding generalized momentum pj. We do so throughout
 11 Quantum Mechanics and Energy Storage in Particles

the definition of the Lagrangian and define the Hamiltonian H for a system of N particles
to be
37V
i L( p2, •.., P3N,q\,42, ..

(1.9)
In general, for systems of particles we expect that the total kinetic energy of the particles is
given by
3/V

^ = E ^(<7i,<72,<73,-..,43/v)<7 7 2 > (1.10)

y=i
and by definition
dL dK
Pj = TT- = — = 2ajqj. (1.11)
dqj dqj
The coefficient a} varies depending on whether the generalized velocity is linear or angular.
Substituting this result and the definition of L into the relation defining H yields
H = K + U. (1.12)
Note that the Hamiltonian equals the total system energy. If the Lagrangian is not an explicit
function of time, we can differentiate the relation defining H to obtain
3/V 3N Q r 3/V ~T
d
M Y ^ W E j
d
E d
(L13>
Using the following relations obtained above:
dL dL
^=PJ>
d dqj^ = P j
qj

the relation for dH becomes

37V 3N

j j

Since H = H{p\, P2, ••.,P3N, q\»<?2, •••, #3/v) it also can be mathematically stated that

By equating coefficients of the dpj and dqj terms in Eqs. (1.15) and (1.16), we obtain
Hamilton's equations of motion

These 6TV first-order equations completely describe the dynamics of a system of TV

particles. Given the initial conditions, we could try to solve these equations to predict the
1.2 IA Review of Classical Mechanics

Figure 1.4

macroscopic behavior of a system of molecules. However, to do so is undesirable for two

reasons. First, solving so many equations is very difficult, and second, at the microscopic
level quantum effects are not included. Hamilton's equations have been presented here
because they demonstrate that the energy of the system is a function of appropriately chosen
coordinates and conjugate momenta. This concept will be central to our development of a
statistical view of the thermodynamics of the system in subsequent chapters.

Example 1.2 The mass in the mass-spring system in Figure 1.4 oscillates about the
position x — 0. The spring exerts a restoring force F that is linearly dependent on x, obeying
the relation
F = -kx,
where k is a constant that characterizes the spring. Find the Lagrangian and Hamiltonian
for the system and show that Eq. (1.5) reduces to Newton's equation of motion.

Solution At any instant, the kinetic energy K is (l/2)mu2, where u =dx/dt.

The potential energy is
px px Y
U = - F dx = - -kxdx = -kx.
Jo Jo 2
The Lagrangian and Hamiltonian are therefore given by

L = \mu2 --\kx2
and
H = \mu2 + \kx2.
Substituting the relation for L into Eq. (1.5), we obtain
d fdL\ dL
dt \du J = 37'
d
— (mu) --
= -kx = F.
dt
Thus, Eq. (1.5) reduces to the time derivative of momentum being equal to the applied force,
which is essentially Newton's law of motion.
 11 Quantum Mechanics and Energy Storage in Particles

1.3 Quantum Analysis Using the Schrodinger Equation

As aresult of the work of Planck and others, by 1920, the dual nature (wave/particle)
of radiant energy was widely accepted. Radiant energy is carried in packets called photons,
each with energy /zv, which have particle and wavelike qualities. Here v is frequency
and h = 6.63 x 10~34 Js is Planck's constant. In 1924, de Broglie argued that if radiant
energy could have both wave and particle properties, then perhaps matter particles also
have wavelike qualities. By analogy, he postulated that if a particle has total energy s and
momentum /?, then it has a wavelength A and frequency v associated with it such that

s = hv (1.18a)

and
p = h/k. (1.18b)
For a particle traveling with momentum mv this implies that
X = h/mv. (1.19)

Example 1.3 In 1927 Davisson and Germer demonstrated the wavelike scattering of
an electron beam reflected from a nickel surface. The wavelength of the electron beam
determined from the interference pattern and the known spacing of the nickel atoms agreed
well with that predicted by de Broglie's formula. The mass of an electron is 9.109 x 10~31 kg.
If they travel at 0.05 times the speed of light, it follows that

663
* 1 0 " 3 4 = 4 . 8 5 x 10-»m =
9.109 x 10-31(-05 x 3 x 108)

For visible light X = 4,000-8,000 A. Thus an electron beam can provide higher image
resolution since its wavelength is much shorter than visible light. This is the motivation
behind the development of the electron microscope.

The coordinates q\, q2, • • and

• momenta p\, P2, • • that• quantify the dynamics of an N-
particle system define a multidimensional space referred to as phase space. The wavelike
nature of particles implies that they are really smeared in phase space with no definite
position and momentum. It can be argued, based on the de Broglie relations, that for
any measurement, the uncertainty in a particle's position, Ax, and in its corresponding
momentum, Apx, are constrained by the relation

\AxApx\ > — = -, (1.20a)

where

h = —. (1.20b)
Lit
Equation (1.20a) is a quantitative statement of the Heisenberg uncertainty principle. We
will consider this issue further in Section 1.6.
131 Quantum Analysis Using the Schrodinger Equation

Because matter has some wave characteristics, it is expected that some wave equation
governing its behavior should exist. If we consider the following mathematical representa-
tion of a one-dimensional (1-D) wave:
V = C exp{/[2;o:A) - 2jrvt]}, (1.21)
and differentiate, it is easy to show that
avj/ h
-m— = -* = Pxv (i.22)
OX A
and
avi/
ih— =hvW = eV. (1.23)
ot
This suggests the identification of —ih(d/dx) with p x and ih(d/3t) with e. In a nonrela-
tivistic, classical (3-D) system, the total Hamiltonian (total energy) for a particle is given
by
D2
H = K + U = — + U(r), (1.24)
2m
where
P2 = p2x + P2y + P2z, (1.25a)
r = xi + j j + zk. (1.25b)
The above arguments imply that

p1 = - i » A f-!»!-) = -n2^-r (1.26c)

3z \ 3z/ dzz
Substituting Eqs. (1.26) and (1.25a) into (1.24), the right-hand side becomes the Hamiltonian
operator
2
Id2 32 32 1
\ * l * \ (1.27)

The Hamiltonian H is interpreted as equaling the total energy s, which, in turn, is identified
with iTi(d/dt). Replacing H with ih(d/dt), and applying the operators on both sides to the
wave function ^ , yields
gvji — h2
iTi = V 2 ^ + U(r)V. (1.28)
dt 2m
With a little rearranging this can be written as

This is the Schrodinger equation for the system wave function *I>.
10 II Quantum Mechanics and Energy Storage in Particles

The wave version of quantum mechanics theory is built upon the following postulate:
• The stateof a quantum-mechanical system is completely specified by a function
^ ( r , t) that depends on the coordinates of the particles and on time. This function,
called the wave function or state function, has the property that *I>*(r, t) *I>(r, t)
dx dy dz is the probability that the particle lies in the volume element dx dy dz
located at location r at time t. (Note that *I>* is the complex conjugate ofW.)
The wave function ^ is determined by solving the Schrodinger equation.
For *I> to have the property indicated in the postulate, ^ must satisfy

/ W*dV = 1. (1.30)
Jv
If solution of the Schrodinger equation is postulated to be of the form
vi/(r, t) = f(t)f(r) (1.31)
then substituting into Eq. (1.29) and rearranging leads to

Note that the left side of the above equation is a function only of time whereas the right side
is a function only of position. Hence, each side must be a constant, which we will designate
as C o . Setting the left side equal to Co yields
/ = /o exp{-iC 0 f/ft}, (1.33)
where /o is the initial value of / at t = 0. Substituting this result into Eq. (1.31) yields
vl/(r, t) = foi/r(r) exp{-iCot/n}. (1-34)
Differentiating gives

ih— = C 0 ^ = eV, (1.35)

ot
from which we conclude that the separation constant Co is the energy,
C0 = s. (1.36)
The relation for / is therefore
(1.37)
Thus, the time variation is oscillatory with fixed amplitude. Boundary conditions and po-
tential energy effects are manifested in the spatial variation. Setting the right side of the
separated Schrodinger equation equal to e yields
h2 ,
U(r)i/ VV = sty. (1.38)
2m
The above equation is the time-independent Schrodinger equation. The time-independent
Schrodinger equation with appropriate boundary conditions generally forms a Sturm-
Liouville system for which s is an eigenvalue. For such a system it is known that solutions
exist only for discrete values of s. Solution of the time-independent Schrodinger equation
will specify the allowable quantum energy levels for the system.
1.4 I Model Solutions of the Time-Independent Schrodinger Equation 11

1.4 Model Solutions of the Time-Independent Schrodinger Equation

Several simple model solutions of the time-independent Schrodinger equation can
be used as models of energy storage in atoms and molecules. For this reason we will examine
these simple model solutions in this section.

A Free Particle in a Box -1-D Case

For the one-dimensional particle-in-a-box system, the potential function that con-
fines the particle to a finite x domain is defined as
U = 0 forO < x < L, (1.39a)
U -> oo at x = 0 and x = L, (1.39b)
(see Figure 1.5).
Note that U becomes infinite at the walls due to the instantaneous reversal of momentum
at each wall. Since a particle with finite energy cannot exist where U —• oo, the probability
of finding the particle must approach zero at x — 0 and x = L. Assuming \jr is continuous,
we adopt the boundary conditions

=0 atjc = = L- (1.40)
and take U = 0 over the entire interval. For the 1-D case, the time-independent Schrodinger
equation becomes an ordinary differential equation:

h2 \ d2^ _
(1.41)
'2m)
By direct substitution, it is easily shown that the solution must have the form

/ [2m~^ \
\[r = A sin W —^-x . (1.42)

To satisfy the boundary conditions it is necessary that L ^2msx/h2 be integer multiples

of 7i \

n x = l , 2 , 3, ...,

!
I
x= L x=0

Figure 1.5
12 II Quantum Mechanics and Energy Storage in Particles

or

(L43)

where nx is the quantum number. Substituting and requiring that

V = 1

yields A = J2/L. The full wave function solutions are therefore

x[r = J=- sin ( ^ ^ V nx = 1 , 2 , 3 , . . . . (1.44)

A Free Particle in a 3-D Box

For the 3-D case, the time-independent equation becomes

Extending the reasoning used for the 1-D case, we have taken (7 = 0 inside the box and ^
is taken to be zero at all walls:
^ = 0 a t x =0 , x =L , y=0 , y =L , z=0 , z =L . (1.46)
Using separation of variables, it can be easily shown that solutions satisfying the boundary
conditions and the requirement that

[ W * d V = f f f W*dxdydz =l (1.47)
Jv Jo Jo Jo
are of the form
/2\ ' fnnxx\ {7znyy\ (7tnzz\
= sin s[n sm (L48)
+ {i) {-r) {-r) {-ir)>
where nx, ny, and nz must be chosen from the set {1, 2, 3,...}. It follows directly from this
solution form that

n +
) ' "l> + n%- (L49)

Note that a given e = £/ may correspond to a number of different combinations of nx, ny,
and nz. The wave function (eigenfunction) for each combination will be different but the
total energy of the system will be the same. Hence the system may occupy a given energy
level Si in a number of different ways. The number of modes of occupancy for the ith
energy level is called the degeneracy gj of that level. If g, = 1, the particle is said to be
nondegenerate.

Example 1.4 Translational energy storage levels for molecules in an ideal gas are usually
assumed to correspond to those predicted by the quantum solution for particles in a 3-D
14 I Model Solutions of the Time-Independent Schrodinger Equation 13

box. Use the particle-in-a-box solution to determine the separation between the first two
translational energy levels for neon atoms in a system with a volume of 1,000 cm 3.

Solution The energy difference between the lowest two nonzero energy levels
specified by Eq. (1.49) is given by

Substituting h = 6.63 x 10~34 Js, V = 0.001 m3, and m = 3.35 x 10~26 kg for neon yields

This illustrates that the typical separation between translational quantum states is so small
for macroscopic systems as to be unmeasureable. We will show later that the mean trans-
lational energy of an atom in an ideal gas is {3/2)k^T, where k^ = 1.38 x 10~23 J/K
and T is the absolute temperature. (We will define absolute temperature more precisely
later. For now we will accept it as a suitably defined temperature scale in the units kelvin,
abbrevated as K.) At 290 K (near room temperature), the mean translational energy of a
molecule is 6.0 x 10~21 J, which is much larger than the separation between levels computed
above.

The Harmonic Oscillator

Consider a particle moving about an equilibrium position at x = 0 with a restoring
force linearly dependent on x:
F = -kx. (1.50)
The simplest example of such a system is the mass-spring system shown schematically in
Figure 1.4 in which the mass can move in one dimension. The potential energy stored in
the spring is given by
x
fx kx2
Fdx= / kxdx = . (1.51)
/
Jo 2
The standing-wave Schrodinger equation then becomes

The probability of finding the particle at x —>•±oo must be zero since U —> oo at those
locations. This implies the boundary conditions
i/r(+oo) = ^ ( - o o ) = 0. (1.53)
Defining
P = Ime/h2 (1.54a)
and
a2 = km/h2 (1.54b)
14 II Quantum Mechanics and Energy Storage in Particles

we can write Eq. (1.52) as

d
^ f = 0. (1.55)

Executing the transform

%=al/2x, (1.56a)
2
x/r = 2(§)exp{-§ /2} (1.56b)
Eq. (1.55) becomes

Note that the exp{—^ 2/2} factor automatically forces the solution to satisfy the boundary
conditions at x -» ±oo. Equation (1.57) is the Hermite equation, which has solutions only
when
n "I

- - l \ =2n, n = 0, 1, 2, 3, . . . . (1.58)
J
The resulting solutions are Hermite polynomials, defined as

//„(§) = ( - l ) V 2 — - ( ^ ) . (1.59)

Substituting into the transform relation for \// and invoking the normalization condition
(1.30) yields wave solutions of the form
-.1/2
Hn(al/2x)exp{-ax2/2}. (1.60)

Expressing Eq. (1.58) in terms of the original variables, we obtain the following relations
for the quantum energy levels:

1
"* ' « = 1,2,3,.... (1.61)

From a classical analysis of a harmonic oscillator of this type, {\/2n)^/k/m can be

identified as the frequency of the oscillator, which we denote as v:
e=(n + \)hv, n = 1,2,3,..., (1.62)
v = (l/27t)y/k/m. (1.63)
For the harmonic oscillator, the levels of a quantum-mechanical harmonic oscillator are
often represented graphically in a diagram like that shown in Figure 1.6.
Note that the ground-state energy (for n = 0) is not zero. The energy of the oscillator
can be written
£ H O = (pl/2m) + (kx2/2). (1.64)
A zero value of Euo would require both px and x to be exactly zero, violating the uncertainty
principle.
1.4 I Model Solutions of the Time-Independent Schrodinger Equation 15

\ "=5 /
\ n=4 /
=3 /
2
\_n = /

V
Figure 1.6

m m z
moment of inertia / = l 2 Q

Figure 1.7

For the rigid rotor shown in Figure 1.7, the potential energy U is 0 for the free
rotation in two angular directions. Expressing the Laplacian in spherical coordinates, the
standing-wave Schrodinger equation then becomes

1 9 1 92^ 2/
2
(1.65)
sin 0 dO V S m " dO sin
Postulating a separation of variables solution,
(1.66)
we substitute, taking m2 to be the separation constant, to obtain
d 2®
= —m (1.67)

1 21 e m2
sin P=0. (1.68)
sin OdO V dO J ' \ h2 sin 2 0
The 27r periodicity of the solutions requires that

= <D(2TT),

(1.69)

/ U \ / Z7T

Solutions to Eq. (1.67) are of the form

O = A sin m<p + 5 cos me/), (1.70)
16 II Quantum Mechanics and Energy Storage in Particles

where m = ± 1 , ±2, ±3, . . . to satisfy the periodic boundary conditions. Transforming the
independent variable in the second Eq. (1.68) to § = cos 0 yields
'lie m2 \

For integer values of m, this is Legendre's equation. For a given m, solutions can be obtained
for an infinite number of e values specified as
+2
s = —1(1 + 1), / = integer > \m\. (1.72)

Solutions are the associated Legendre polynomials

h2\m/2 jl-\-m

(L73)
I'll
If these results are back-substituted into the separation of variables solutions, and the
normalization condition ( / / / yj/^yjr dV = 1) is imposed, the wave functions take the form

- • > '' l; 'iirr (nn,, »)«••*

»)«•
Note that each / corresponds to one s (energy level) and / is therefore the principal (rotational)
quantum number. For each / there can be any m in the range specified by / > \m\. There
are 2/ + 1 values of m permissible for each / and hence for each e. Thus, the degeneracy gi
of energy level E\ is
g,=2l + \. (1.75)
Increased degeneracy corresponds to increased degrees of freedom for energy storage.

Example 1.5 The rigid-rotor solution is commonly used to model rotational energy storage
in molecules in an ideal gas. Use the rigid-rotor solution to determine the separation between
the first two energy levels for nitrogen molecules in an ideal gas.
Solution The energy difference between the lowest two nonzero energy levels
specified by Eq. (1.72) is given by

Ae = (—] [2(2 - h i ) - 1(1 + 1)] = 4 (—) .

The nuclei of the atoms of nitrogen with mass 2.33 x 10~26 kg are separated by a mean
distance of about 1.6 x 10~10 m. The moment of intertia of the nitrogen molecule is given
by
/ = x-mzl = 1(2.33 x 10"26)(1.6 x lO" 10 ) 2 = 3.0 x 10~46 kg m2.
Substituting the result for / and Ti = 6.63 x 10~34/27r Js in the relation for Ae yields
A /(6.63X10-34/2TT)2\ n A n IO_23T
Ae = 4 46 = 7.42 x 10 J.
V 2(3.0 x 10- ) J
We will show later that the mean rotational energy of a nitrogen molecule in an ideal gas
is kBT, where T is the absolute temperature and kB = 1.38 x 10~23 J/K. By comparison, at
1.5 IA Quantum Mechanics Model of the Hydrogen Atom 17

290 K (near room temperature) the mean rotational energy of a nitrogen molecule is 4.0 x
10~21 J, which is about 50 times larger than the separation between levels computed above.

The simple model systems for which we have developed quantum mechanical solutions
to the Schrodinger equation will be used in later sections to model specific energy storage
mechanisms within atoms and molecules. Before proceeding further it is illuminating to
examine how an entire atom can be treated quantum mechanically. The simplest treatment
of this type is the model of the hydrogen atom described in the next section.

1.5 A Quantum Mechanics Model of the Hydrogen Atom

The Schrodinger equation described in previous sections can be used to describe
the hydrogen atom as a whole. For the hydrogen atom the nucleus consists of one proton
and one electron is in orbit around the nucleus (see Figure 1.8).
The attractive force between electron and nucleus is given by Coulomb's law:
F = Cce2Jr2,
where ee = 1.602 x 10~19 coulombs is the charge of one electron. We therefore define the
potential energy for the system as

U = - f Fdr = - —, d = Cce2e. (1.76)

r
Joo
The negative sign is applied because stored energy increases as r —> oo. Note that the mass
of an electron = 9.1 x 10~31 kg and the mass of a proton = 1.67 x 10~27 kg. Since the
proton mass mp is much larger than the electron mass me, we idealize the nucleus as being
stationary, with the electron orbiting around it. The Schrodinger equation then takes the
form

+ (e + —
r
W = 0. (1.77)
2me V ,
Postulating a solution of the form xfr — Y(0, 0)/?(r), substituting, and executing the usual
separation of variable technique results in the following equations for Y and R:

1 W J n / l
' " " ^ = -(^W (1.78)
sin 0 J dO V dO J \sin2 0

• ^ + fc)U^-^l|« = 0, (1.79)
dr I \Ti2 J I r 2/

electron

Figure 1.8
18 II Quantum Mechanics and Energy Storage in Particles

where
I=mer2. (1.80)
The equation for Y is the rigid-rotor equation (with the same periodic boundary conditions)
solved above. It follows directly that solutions for Y are

)e-t (L81)

To solve the radial Eq. (1.79), we define

rj = -, L(rj)=at]-lR(ari)ell/2. (1.82)
a
For this system, the energy zero point is taken to correspond to infinite separation of the
electron and nucleus. The definition of a above includes the minus sign inside the square
root because the energy levels for finite separation are negative. Invoking these definitions,
the equation for R becomes
d2L dh
r]-—rl + (k + 1 - ri)— + n LL = 0, (1.83)
dr\ dr\
where

* = 2/ + l, nL = ^ - - l - l . (1.84)
h
This equation, the associated Laguerre equation, has solutions known as the associated
Laguerre polynomials'.
nL-k . 1

Rearranging the relation for nL and substituting for a yields

s= -m*C* . (1.86)
2Ti (nL+l + l)2
2

Setting n = nL + I + 1 yields

The negative sign on the right side of the relation for e is consistent with the requirement
that increasing n correspond to increasing energy. Solving for R(r) yields

a88)
")-
Note that a is a function of n and
1.5 IA Quantum Mechanics Model of the Hydrogen Atom 19

For a given n, I can vary from zero to n — 1 and give a valid radial solution. As shown
for the rigid-rotor solution, for each /, m can have any of 2/ + 1 values. Hence, for this
solution:
n = principal quantum number (specifying energy level),
/ = angular quantum number,
m = magnetic quantum number,
and for a given n, the limits on / and m are:

0 < / < n - 1, -I <m <l.

It follows directly that the total degeneracy for a given n value is given by
n-\
gn = ^ T ( 2 / + 1) = n2. (1.90)

Radiation emission or absorption results in a change of energy level for the hydrogen atom.
For emission, the photon energy hv is given by
_ 2ix2meC\ I" 1 1
2 2
h [n n\
This relation can be written in terms of the wavenumber k\2 as

where R\\ is the Rydberg constant given by

2n2meC}

Based on experimental determinations of the (Baimer) series of emission spectral lines

for hydrogen, Johannes Rydberg, in a study done in the 1890s, concluded that Rn =
109,720 cm" 1 . The quantum-theory relation indicated above predicts Ru = 109,680 cm" 1 .
The agreement of the theoretical prediction with the measurements is indeed impressive.

Example 1.6 The Balmer series of emission spectral lines for hydrogen corresponds to
electrons falling from levels above n = 2 down to the n = 2 level. Show that the Balmer
series corresponds to wavelengths between 3,640 and 6,563 A.

Solution Since the wavenumber k\2 = 1 Ai2> Eq. (1.92a) can be written

For the Balmer series, «2 = 2, and so

l
1 [1 1
20 II Quantum Mechanics and Energy Storage in Particles

The largest wavelength in the series corresponds to n\ =3. Substituting and using Ru =
109,720 cm" 1 , we find

Xl2 = 6.562 x 10~5 cm = 6,562 A.

The smallest wavelength in the series corresponds to n\ —> oo. In this case, the relation for
X12 yields

A.12 = 3.646 x 10"5 cm = 3,646 A.

Thus, the Baimer series spans the wavelength range from 3,646 A to 6,562 A.

1.6 The Uncertainty Principle

To further explore the nature of the uncertainty principle, we now wish to reconsider
the one-dimensional particle in a box discussed earlier. For that system, we found that

( ^ ) », = 1,2,3 (1.93)

Because V*^ dx is interpreted as a probability, it can be used to calculate averages and

variances. For the particle in a 1-D box, xjr has no imaginary part and is therefore its own
complex conjugate:

'nf—— J. (1.95)
It follows that

= ( - 1 sinz ( ^ ^ )dx (1.96)

is the probability that a particle is found between x and x + dx. The average value of x, the
mean position of the particle, is obtained by integrating x, weighted by its probability, over
the box dimensions:

MT)-
Evaluation of the integral yields

(x) = - for all/i*. (1.98)

We can also evaluate the spread of the variation of x by computing the variance of x (from
the mean) defined as

aj= / (x-{x))2\J/*i/dx. (1.99)

1.6 I The Uncertainty Principle 21

Expanding, we obtain

fL fL fL
2 2 2
<y = x \ff*\J/dx — 2{x) / x\l/*\jf dx + {x) / \\r*\\rdx
Jo Jo Jo
2 2
= (x ) - 2(xf + (x)
= (x2) - {x}2. (1.100)
{x2} is computed as

Substituting (1.98) and (1.101) into (1.100), we get

9 o i ( L \ f 7t2n2 \
a2 = {x2) - (x)2 = I j f ^ - 2 1 . (1.102)

The standard deviation in x is therefore given by

2
(1.103)

Similarly computing the average energy or momentum is a bit trickier because these quan-
tities are represented by operators:
h2 d2
H = - — -~-z (energy), (1.104)
2m dx
d
px = —ih— (momentum). (1.105)
dx
We now consider the question: In integrating to get the average, does the operator work
on \lf*(x)\j/(x) or on \// or ^* alone? To address this question we consider the following
equation, which results from the definition of the Hamiltonian,

sf. (1.106)
Multiplying both sides by x/r* and integrating, we obtain

/ \l/*Hirdx = £ / }/ dx = e. (1.107)
Jo Jo
This indicates that to get e for (s) we must sandwich the operator between \j/* and \jr\

(e) = / f*Hfdx. (1.108)

Jo
Based on this same reasoning, we define

(Px) = J f*pxfdx = J r(-'fi-^)fdx. (1.109)

22 II Quantum Mechanics and Energy Storage in Particles

Substituting for the wave function yields

/
7tnxx\ ( _ d \ 12 . / ixnxx
S
L VL L
(1.110)

which leads to
(Px)=0 (for all nx). (1.111)
2
Similarly, for {p x)\
rL
( d \2
(p*)= 1 rp2x \jr* i—ih— I i/sdx (1.112)
JO
L
o V dxj
2
I" fl fnnxx\\ ? d f p2 fnnxx\]
\ - sin —— ( - ^ 7 - 2 V T sin - 7 — dx,
JO LV L V ^ 7J ^2LV L V ^ yJ
which reduces to
2 2*,2
ff
(^) = ^ . (1.113)

Combining these results, we obtain

°v = (P!) - (p*)2 = ^-jf-

or equivalently

oP = ^ . (U14)
If we substitute this result into the relation for ox to eliminate L, we obtain
I 1/2

which implies that

/
fi\ [n2nl
2 , • „ -• - d-116)
For any nontrivial wave function solution nx > 1, which implies that the term in square
brackets is greater than one. It follows directly that

(1.117)

which is the mathematical statement of the Heisenberg uncertainty principle. We can in-
terpret this result as follows. Since ap = Ttnxn/L, for a free particle (L -> oo) op is zero.
A free particle thus has a definite momentum. However, if we localize the position of the
particle within a dimension L, it no longer has a definite momentum, and the uncertainty
in its momentum is given by ap = itnxn/L. In the limit of exactly specifying the particle's
location, the uncertainty in momentum is infinite. The uncertainty principle states that the
product of the two uncertainties is of the order of Planck's constant.
Although we have derived the uncertainty principle for the specific case of a particle in
a 1-D box, it can be shown to be generally valid. This result implies that a particle does
not have an exact point representation in phase space. For the 1-D particle in a box, the
1.71 Quantum Energy Levels and Degeneracy 23

• :APx,min

•HI2

Figure 1.9

most we can say is that it is somewhere inside a region having an area of order Ti/2 (see
Figure 1.9). For large systems (large L) with large momentum, phase space is virtually a
continuum since n is a very small number.

1.7 Quantum Energy Levels and Degeneracy

In the previous sections of this chapter, we have used basic quantum mechanics
theory to explore the energy storage characteristics of several model systems. In later chap-
ters the particle-in-a-box solution will provide the basis for evaluating translational energy
storage in atoms, molecules, and other particles. Similarly, the rigid-rotor and harmonic-
oscillator solutions will be used as models for rotational energy storage and vibrational
energy storage in molecules. The hydrogen atom model considered in Section 1.5 is an ex-
ample of yet another energy storage mechanism in atoms and molecules: electronic energy
storage.
It is important to recognize that in all the models considered in this chapter, the energy
levels accessible to the system are quantized - they can take on only specific values corre-
sponding to integer values of appropriate quantum numbers. We have also found that the
harmonic-oscillator energy levels can be occupied in only one way, whereas the energy
levels of the rigid rotor and the hydrogen atom exhibit varying amounts of degeneracy.
We now will examine the issue of degeneracy for the particle-in-a-box model. For the 3-D
particle in a box, the quantum energy levels are given by

8 = (1.118)
8mV2/3
where V = L3 and nx, ny, nz are chosen from {1, 2, 3, }. In this case, the degeneracy
g(e) is dictated by the number of ways that R2 = 8mV2/3e/h2 can be written as the sum
of three squared positive integers. Thus, g is an erratic function of R2{s) for small R values
but becomes smooth at large R. Consider a three-dimensional space spanned by orthogonal
coordinates nx,ny,nz. There is a one-to-one correspondence between energy states for the
particle and points located at integer values of nx, ny, nz (see Figure 1.10). The equation
8mV2/3s
R2 = = (n2x +n2y+ n2z) (1.119)
h2
represents the surface of a sphere in this space. For large R, the states are so close together
that the total number of states with s < R2h2/8mV2/3 is essentially equal to the volume of
24 11 Quantum Mechanics and Energy Storage in Particles

surface for specified

R2 =n} 2
?

nodes correspond to
quantum states

Figure 1.10

one octant of a sphere of radius R. If the number of such states is denoted as n e, we have

"« = 8 ( ~ 3 ~ ) = 6 ( h2 ) • (1.120)

The number of states in the interval s to s + 8 s is interpreted as the degeneracy of energy

level e:
dne n
•x'28s. (1.121)
ds 4 \ h1 )
This relation allows us to evaluate the degeneracy for a particle in a box. For common
circumstances, this relation predicts an extremely large degeneracy (see example below).

Example 1.7 The mass of electrons is 9.109 x 10 31 kg. Suppose an electron has an
energy equivalent to a velocity of 0.05 times the speed of light and is confined in a volume
of one cubic millimeter. The energy is
e = mv2/2 = 9.1 x 10"31(0.05 x 3 x 10 8 ) 2 /2 = 1.0 x 10~16 J.
If we take 8s = 0.0l£, the degeneracy is

= 5.3 x 1026.
Clearly, the degeneracy of this energy state is very large. For a system having a large number
of such particles, the system degeneracy would be enormous!
1.81 Other Important Results of Quantum Theory 25

Table 1.1 Summary of Relations for Energy Quantum States, Degeneracy, and
Selection Rules

Energy Quantum States Degeneracy and Selection Rule

Particle in a 3-D box s =

<+<+<
nx,ny, nz contained Selection rule: none
in {1,2, 3,....}
Harmonic oscillator s = [n + | ) hv g(e«) = 1
n = 0, 1,2, 3 , . . . Selection rule: Dipole
moment must vary with
oscillation and An = ±1
Rigid rotor gist) = 2/ + 1
/ = 1,2,3,... Selection rule: Rotor must
have a dipole moment
and A/ = ±1
-meC\
Hydrogen atom _ 2Ti2n2 g(sn) = n2
n = \,2,3,... Selection rule: none

As we construct a theoretical framework for thermodynamics in later chapters we will

have to account both for the quantized energy levels of particles and the degeneracy of the
levels. The quantum energy levels and degeneracy for the systems considered in this chapter
are summarized in Table 1.1.

1.8 Other Important Results of Quantum Theory

In addition to the quantum theory specification of energy levels and degeneracies,
there are other predictions of quantum theory that are relevant to developments in later
chapters. First, quantum theory not only predicts the discrete energy level accessible to
the system but also predicts the permissible transitions between energy levels. The allowed
transitions from one energy state to another are dictated by selections rules, which generally
vary from one system to another.
For the particle in a 3-D box, there are no restrictions on transitions from one state to
another. The particle in any state, with any combination of n x,ny, and nz, can absorb or
emit a photon and undergo a transition to any other state accessible to the particle. Thus,
there is no selection rule for this system.
In contrast, the rigid-rotor and harmonic-oscillator systems do obey selection rules.
For the rigid rotor, the selection rule requires that the rotor (which is usually a model for
a molecule) must have a permanent dipole moment and only transitions associated with
emission or absorption of a photon corresponding to A/ = ±1 are permitted.
For the harmonic oscillator, the selection rule requires that emission or absorption of a
photon can occur only if the dipole moment of the oscillator varies during vibration and only
for An = ± 1 . As we will discuss further in later chapters, molecules may simultaneously
be rotating and vibrating. As a first approximation, the energy storage modes are usually
26 II Quantum Mechanics and Energy Storage in Particles

modeled as separate independent energy storage modes. The rotational and vibrational
energy of a diatomic molecule may then be given by

£vib,rot = U + - ) hv0 + —/(/ + 1). (1.122)

The energy levels associated with vibration are generally more widely spaced than those
for rotation. In this case the respective selection rules for each mechanism described above
still apply (An = ± 1 , A/ = ±1). We will discuss the effect of these selection rules on
emission spectra in a later chapter.
For the simple hydrogen atom system discussed in Section 1.5, there are no selection
rule requirements on transitions between electronic energy states. This is not generally
true for other atoms and molecules, however. When additional electrons are present and
when both vibrational and electronic transitions may occur, selection rules specific to each
circumstance must be obeyed.
Derivation of selection rules for different system types requires an analysis of photon
interaction with the molecule using the full time-dependent Schrodinger equation. Although
analysis of this sort is beyond the scope of this introductory presentation, knowledge of
the selection rules is important to the understanding of absorption and emission spectra
in gases. For our purposes, we will incorporate the selection rules into our analysis as
accepted results of quantum theory. Selection rules for the simple systems considered in
this chapter are summarized in Table 1.1. The reader interested in more detailed treatment
of this issue may consult the references by Liboff [1] and McQuarrie [2] cited at the end of
this chapter.
Another noteworthy result of quantum theory is that all known particles generally fall
into one of two categories: fermions and bosons. This categorization is associated with
the quantum property of spin. Fundamental particles that have half-odd-integral spin are
fermions. Particles that have integral spin are bosons. Electrons and protons are fermions. In
addition, a compound particle containing an odd number of fermions will itself be a fermion
if the binding energy of the compound particle is high compared with other transitions. An
example of such a particle is helium 3 (He3). Photons and deuterons have integral spin and
are therefore bosons. The full significance of the spin property can only be appreciated
by examining the quantum mechanics theory in more detail. We will not pursue such an
examination here. It is sufficient for us to note that quantum theory requires fermions to
obey the Pauli exclusion principle, which states that in a system of TV fermion particles,
no two fermions may occupy the same quantum state. In a system of N boson particles,
there are no restrictions on how many particles may occupy a given microstate. Atoms and
molecules for which there are no restrictions on microstate occupancy are also generally
categorized as bosons.
In developing a statistical treatment of energy storage in a system of N particles, we
therefore must allow for two particle types: fermions and bosons. In the next chapter we
will begin to explore the impact of particle type on macroscopic thermodynamic behavior
of multiparticle systems.

Exercises
1.1 Using laboratory instrumentation, if we can determine the position of a helium atom to within
1 nm (10~9 m), what is the quantum limitation on the uncertainty of any measurements of
its velocity?
Exercises 27

1.2 We will show later that the mean speed of a molecule or atom in an ideal gas is given by
(UBT/7tm)l/2, where T is the absolute temperature and kB = 1.38 x 10"23 J/K. Using this
relation, determine and plot the variation of the mean quantum wavelength for a helium
atom (m — 6.64 x 1CT27 kg) in an ideal gas between absolute temperatures of 5 K and 300 K.
What does this suggest about the effect of temperature on quantum effects in an ideal gas?
1.3 As mentioned in Example 1.4, the mean translational energy for an atom in an ideal gas is
(3/2)£B T, where T is the absolute temperature and k# = 1.38 x 10~23 J/K. Assuming the 3-D
particle-in-a-box solution applies to the translational quantum levels for neon molecules in
an ideal gas, determine the order of magnitude of the largest quantum number among nx,ny,
and nz for atoms in a system at T = 290 K. Is this value large or small compared to one?
1.4 CO2 is a linear molecule with a carbon atom (m = 1.99 x 10~26 kg) at the center and an
oxygen atom (m =2.66 x 10~26 kg) at each end. The separation between the carbon and
oxygen atoms is about 2.77 x 10~10 m. Use the rigid-rotor quantum solution to determine
the separation between the first two energy levels for CO2 molecules in an ideal gas.
1.5 Use the rigid-rotor solution to determine the first three rotational energy levels of N2O.
This molecule is linear and unsymmetric (N—N—O) with a moment of inertia of 6.69 x
10- 46 kgm 2 .
1.6 We will show in a later chapter that the vibrational energy storage in a nitrogen molecule
can be modeled using the harmonic-oscillator quantum solution if the mass of the oscilla-
tor is taken to be the reduced mass mr = m/2 (m being the mass of a nitrogen atom). A
suitable value of the spring constant k must also be specified. A value of 2,250 N/m has
been suggested for the effective spring constant. Use this value and the harmonic-oscillator
solution to predict the separation of the first two vibrational energy levels for the nitrogen
atom. We will also show later that in an ideal gas at absolute temperature T, the mean trans-
lational energy for a nitrogen atom is (3/2)kBT where kB = 1.38 x 10~23 J/K. Compare
the separation between the two lowest vibrational levels to the mean translational energy
of a nitrogen molecule in an ideal gas at an absolute temperature of 290 K. (Note that for
nitrogen, m = 5.65 x 10~26 kg.)
1.7 For the harmonic oscillator discussed in the text, the wave function for the lowest (ground
state) energy level is given by Eq. (1.60) with n — 0. (Note that Ho(a l/2x) is identically
one for all arguments.) An oscillator in this state has an energy s = Ti^/k/m/2.
(a) For a classical harmonic oscillator with specified k, m and total energy e, derive a
relation for the maximum amplitude of the oscillation jcmax,ci-
(b) For the quantum-mechanical oscillator, determine the probability that the amplitude
of the oscillation will exceed xmaxc\. This behavior is characteristic of quantum-
mechanical systems and is responsible for the phenomenon called the tunnel effect in
which the system is said to have tunneled into a classically forbidden zone.
1.8 The three-dimensional harmonic oscillator shown in Figure 1.11 is a possible model for
an atom in a solid crystal lattice. The restoring force in each of the directions x, y, z is
linear and represented by a "spring" constant for each direction. You may assume that the
oscillations are small enough that the energy stored in any one spring is independent of that
stored in any other.
(a) Write down the appropriate form of the Schrodinger equation for this problem and list
the boundary conditions.
(b) Develop a solution for the wave function and determine the quantum energy levels
for the 3-D case. (Hint: Where possible, make use of the results obtained for the 1-D
harmonic oscillator.)
(c) Do the energy levels of this system have a degeneracy greater than 1? Briefly explain
your answer.
28 1 I Quantum Mechanics and Energy Storage in Particles

Figure 1.11

1.9 Using the results of the model analysis in the text, calculate the ionization energy of the
hydrogen atom. Note that the ionization energy is the energy required to take the electron
from the ground state to a totally unbound condition (far from the nucleus). Express your
answer in joules.
1.10 We will later show that in an ideal gas at absolute temperature T, the mean translational en-
ergy for a helium atom is (3/2)kBT. Use the results of the particle-in-a-box quantum model
to predict the degeneracy at s values within 1% of (3/2)kBT for a system with a volume of
0.1 m3 and an absolute temperature at 280 K. (Note that for helium, m = 6.64 x 10~27 kg.)
1.11 As mentioned in Example 1.5, the mean rotational energy for a nitrogen molecule in an ideal
gas is kBT, where T is the absolute temperature and kB = 1.38 x 10~23 J/K. Assuming the
rigid-rotor solution applies to the rotational quantum levels for nitrogen molecules in an ideal
gas, determine the quantum number / and the degeneracy g at the mean energy kB T for nitro-
gen molecules in a system at T = 290 K. Are these values large or small compared to one?
1.12 Carbon dioxide is a linear molecule with a carbon atom (m = 1.99 x 10~26 kg) at the center
and an oxygen atom (m = 2.66 x 10~26 kg) at each end. The separation between the carbon
and oxygen atoms is about 2.77 x 10~10 m. We will show in a later chapter that the mean
rotational energy for a CO2 molecule in an ideal gas is kBT, where T is the absolute temper-
ature and kB = 1.38 x 10~23 J/K. Assuming the rigid-rotor solution applies to the rotational
quantum levels for CO2 molecules in an ideal gas, determine the quantum number / and the
degeneracy g at the mean energy kBT for CO2 molecules in a system at T = 290 K. Are
these values large or small compared to one?

References
[1] Liboff, R. L., Introductory Quantum Mechanics, Addison-Wesley Publishing Company,
Reading, MA, 1980.
[2] McQuarrie, D. A., Quantum Chemistry, University Science Books, Mill Valley, CA, 1983.
 CHAPTER 2

Statistical Treatment ofMultiparticle Systems

Using the basic features of microscale energy storage discussed in Chapter 1,

in Chapter 2 we develop the foundations of statistical thermodynamics. In doing so, we
introduce the concepts of microstates and macrostates and properly account for the fact that
particles in fluid systems are generally indistinguishable. The development of the theoretical
framework in this and subsequent chapters considers a binary mixture of two particle types.
The statistical machinery is applied first to a microcanonical ensemble of systems, each
having a specified volume, number of particles, and total internal energy. Definitions of
entropy and temperature emerge from this development. Application of the results to a
monatomic gas is discussed.

2.1 Microstates and Macrostates

In this chapter we will construct a general statistical mechanics foundation on
which we will develop a full equilibrium thermodynamic theory for systems composed of
a large number of particles. In doing so we will make use of the information about energy
storage derived from quantum theory in the previous chapter.
In analyzing systems of particles, we can deal with the state of a system at two levels:
the microstate of the system and the macrostate of the system. The system microstate is
the detailed configuration of the system at a microscopic level. To specify the microstate
we would have to specify the quantum state (including the position) of each particle in the
system. If we observe a system at a macroscopic level, we can, at best, distinguish some of
the gross characteristics of the system. Here, the term macroscopic refers to observations
that are averaged in some sense over time and length scales that are much larger than
those characterizing molecular motion. A nontechnical analogy of this circumstance is an
impressionist painting viewed from across the room. At that distance, the eye averages
together the microscale brush strokes. In a macroscopic sense, the image is well defined,
even though we cannot discern every individual brush stroke.
An additional point can be extracted from the analogy with impressionist paintings.
Because the eye averages the contributions of individual brush strokes, different microscopic
arrangements of brush strokes may average to produce essentially the same macroscopic
image when seen from across the room. Thus, one macrostate can correspond to many
microstates. The same is true of thermodynamic systems composed of many particles. The
macrostate is a macroscopically distinguishable system characteristic that corresponds to a
collection of system microstates that macroscopically "look" the same.
Equilibrium in classical thermodynamics has an intrinsically macroscopic viewpoint. It
implies that the macrostate of the system is unchanging with time. However, we expect that
the microstate will continuously change with time owing to the motion of the particles and
collisions among particles. If the macrostate changes with time, the associated phenomena
lie in the domain of nonequilibrium thermodynamics - a topic we will explore in later
sections of this text. From a microscopic perspective, macroscopic equilibrium results when

29
30 21 Statistical Treatment of Multiparticle Systems

a system is constrained in such a way that it spends the overwhelming majority of the time
in microstates that correspond to one macrostate.
The statistical treatment of systems of many particles is based on the following postulate:
• All microstates consistentwith the constraints on the system are equally probable.
Since, as discussed above, each macrostate corresponds to many microstates, it follows
directly that:
• The macrostate having the largest number of (equally probable) microstates is
most probable (i.e., most likely to be observed).
These concepts are central to the development of a statistical treatment of multiparticle
systems. We will expand them in more quantitative ways later in this chapter. We now
must bring in appropriate computational machinery to make the treatment functional. This
is done in the next section. Some of the computational machinery is based on results of
combinatorial analysis and basic probability theory, which are reviewed in Appendix I.
Readers unfamiliar with these areas are urged to read Appendix I before proceeding to the
next section.

2.2 The Microcanonical Ensemble and Boltzmann Statistics

The use of ensemble concepts was developed to a refined methodology by J. Willard
Gibbs in the early 1900s. We could pursue a statistical analysis that focuses on the energy
states of individual particles within a system, as Boltzmann did in his pioneering work on
statistical mechanics. One drawback of such an approach is that it becomes more compli-
cated to use when the particles in the system are interdependent in such a way that the
total system energy cannot be cleanly divided among the particles. The ensemble formal-
ism has the advantage that it can be extended in a straightforward way to treat systems of
interdependent particles.
For our purposes, an ensemble is defined as a collection of a very large number of
identical systems that are all subject to the same macroscopic constraints. The analysis we
will develop is referred to as ensemble theory. Each system in the ensemble is identical
in its imposed macroscopic constraints, but each system in the ensemble is in a different
microstate. In a single real system with the imposed macroscopic constraints, the microstate
will vary with time owing to random fluctuations at the microscopic level resulting from
particle motion and particle collisions. The ensemble of systems can be viewed as a list of the
microstates from which the real system is randomly selecting its instantaneous microstate.
This interpretation is the basis of the ergodic hypothesis, which may be stated as follows:
• The ensemble exhibits the same average properties in space (over the ensemble)
as a single system exhibits in time.
It follows from this hypothesis that the ensemble average of a property is the same as the
time average of that property for the single system. Each system in the ensemble must
be in one of the microstates consistent with the imposed constraints. As a result, different
constraints define different types of ensembles.
In this chapter we will consider a specific type of ensemble in which all systems consist of
Na particles of species a and Nb particles of species b confined within a specified volume V.
Each system in the ensemble has the same total internal energy U. We denote this energy
2.2 I The Microcanonical Ensemble and Boltzmann Statistics 31

1 2 3

\\ \ \
\ \x\x\
\ Na \ N \ Na \
\
\\ b \\
a
\
b \ \\\
\ N
\ b
N N
\
\ V \ V \\\ V \
u \
\ \
\
\ \\ u \ \\ U \
\

Figure 2.1 The microcanonical ensemble.

content as internal energy to differentiate it from kinetic energy associated with bulk motion
of the entire system or potential energy associated with the bulk system. Specifically, the
internal energy is the sum total of the energy associated with translation of the particles
and the energy stored within the particles. The motivation for picking a system containing
a binary mixture of two components is that it will allow us to examine how the statistical
framework is affected by the presence of two species. Note that by setting Na or Nb to zero
we should be able to obtain results that are applicable to a single-component system.
Each ensemble member system is isolated from all other systems so that Na, Nh,V, and
U values are fixed. As indicated schematically in Figure 2.1, there are n systems in the
ensemble, where n is very large. In particular, n must be large enough so that the ensemble
includes every microstate accessible to the system within the imposed constraints.
For this type of ensemble, there are no work interactions, no heat transfer, and no mass
exchange among the systems. In later chapters we will consider other types of ensembles
where one or more of these interactions among systems will be possible. For now, however,
we will specifically focus on the microcanonical ensemble.
Our objective in this section is to use the microcanonical ensemble as a framework to
explore Boltzmann statistics. In the systems in the ensemble, particles of each species all
have access to the same set of quantum energy states for that particle type under the specified
system constraints. Each system in the ensemble is in one of a large number of microstates
that are possible for the system with the specified Na, Nt,,V, and U values. We define
Q = the total number of system microstates consistent with the values of Na,
Nb,V, and U specified for the ensemble member systems.
If we had the ability to examine one of the systems at a microscopic level, we could, in
principle, determine the energy level in which each particle resides and we could count the
number of particles that reside in each energy level for that specific system in the ensemble.
At the end of this process, we would have two sets of occupation numbers {Naj} and {Nbj},
where each Naj is the number of particles of type a in energy level / having energy £;, and
each Nbj is the number of particles of type b in energy level j with energy Sj. Note that
in general, the energy levels £/ and £j will have degeneracies gi and gj, respectively. Each
energy level may therefore correspond to many microstates.
The sets of occupation numbers {Naj} and {Nbj} must satisfy the following constraints
on the system:
oo

EN
i=0
aJ = Na, (2.1)

(2.2)
7=0
32 21 Statistical Treatment of Multiparticle Systems

and
OO 00

" " :Nbi=U, (2.3)

where the summation limit oo denotes a summation over all possible energy levels. If we
could similarly determine the occupation numbers for all the systems in the ensemble, we
would find that, in many instances, systems will have the same pair of occupation number
sets. This is a consequence of the fact that a given set of occupation numbers can correspond
to many different microstates.
The sets of {Naj} and {Nbj} values dictate how energy is distributed in the system, since
Na,i /Na is the fraction of type a particles with energy £/ and Nbj/Nb is the fraction of the b
particles with energy Sj. Because the total system energy is limited, there is a finite limit to
the energy level attainable by a single particle and therefore the occupation number values
must approach zero as / —>• oo or j: ^ oo. A specific pair of {Naj} and {Nbj} distributions
thus indicates, in a gross or macroscopic way, the state of the system. We therefore designate
each pair of occupation number sets {Naj} and {Nhj-} satisfying the above global constraints
(2.1)-(2.3) as being a macrostate of the system.
We assume that we can distinguish individual particles of both species from one another
(i.e., they are distinguishable). In a later section we will reconsider the validity of this
assumption, but for now we will accept it as correct for the system we are considering.
Based on the arguments described above, a macrostate is presumed to correspond to a pair
of occupation number sets {Naj} and {Nbj} satisfying the global constraints (2.1)—(2.3).
For species a, the number of ways of dividing the Na distinguishable particles among the
energy levels in conformance with the occupation number set {N a,i} is equal to the number
of ways of dividing Na distinguishable objects into groups of Na^, Na,\,..., Naj,
From basic combinatorial analysis (see Appendix I) this is given by

(number of ways of dividing Na particles

les ) TT 1
{among energy levels as dictated by {Naj} J f_-Jj
f_Jj Naj!

However, for species a, each energy level £/ has degeneracy gi, which means that the N aj
particles in that energy level may be in any of gi microstates. As discussed in Section 1.8,
we also know that particles may be bosons or fermions. Here, we will assume that our
particles are bosons, and hence they are free to occupy any microstate without restriction.
(We will return to consider fermions in a later section.)
To get a relation for the number of microstates corresponding to an occupancy number set,
we must multiply the contribution of each energy level in the above equation by the number
of ways that the Naj particles can be distributed among the gt microstates for that energy
level. As described in Appendix I, the number of ways of distributing Naj distinguishable
particles in gi microstates is equal to gi aJ. Multiplying this by the corresponding term in
Eq. (2.4) for each energy level, we obtain the following relation for the number of ways of
arranging the species a particles among the microstates for a specified occupancy number
set:

{
number of ways of arranging
species a particles among > =Na!TJ——. (2.5)
l=0

microstates dictated by gi and {Naj}

221 The Microcanonical Ensemble and Boltzmann Statistics 33

Based on identical arguments, we can write a similar relation for species b:

{
number of ways of arranging \ ^ uhj

species b particles among > =Nh! TT — J-—. (2.6)

I ./V A i !
y=0
microstates dictated by gj and {Nbj} J
For each arrangement of the a particles, the number of possible arrangements of b particles
is given by Eq. (2.6). The total number of combined arrangements of the Na and Nf, particles
is therefore equal to the product of the expressions on the right-hand sides of Eqs. (2.5) and
(2.6). This total number of arrangements is the total number of microstates for the specified
pair of {Naj} and {Nbj} distributions. Designating this number of microstates as W, we
can therefore write

(
oo Nai \ / oo Nbj \

Note that for real systems W « O(NA\)« <9(1026!). It follows that the total number of
possible microstates consistent with the system constraints is equal to the sum of the W
values for all possible sets of occupation numbers:

W({Na ./J. {Nbj

n
OO Nai

= EE Nad
1=0
Si '
Najl
(2.8)

where the notation

EE
indicates a sum over all possible sets of occupation numbers consistent with constraints

We will denote the probability of an observed state as P (observed state). From the above
observations it follows that

P(macrostate{NaJ}, {NbJ}) =

P (any given microstate) = —.

^L
From these results we can also interpret W as
P (the specified macrostates)
W = —= .
P (any given microstate)
W is sometimes interpreted as a disorder number or a randomness number. In general, W
is very large because P (any given microstate) is small.
Since all microstates are equally probable, the macrostate with the largest number of
corresponding microstates is most probable and hence is most likely to be observed in a real
system constrained to specified Na, Nb,V, and U values. Consider as a common example'
34 2 / Statistical Treatment of Multiparticle Systems

\wwx\w\wwwww\w\\
nitrogen and oxygen gas
fixed NNl,NOl,V,U

wwwwwwwwwwwvvv
Figure 2.2 An example system.

of such a thermodynamic system a mixture of oxygen and nitrogen (similar to air) in a

closed, rigid, perfectly insulated container (Figure 2.2).
Macroscopic measurements indicate that at equilibrium, the macrostate observed for the
system does not change detectably over long periods of time. It is clear that during such long
time intervals, the system microstate changes many times due to motion of the molecules
and collisions. If all available microstates are equally probable, a single macrostate is likely
to be observed for long time intervals only if the number of microstates for that macrostate
is much larger than the number for any other macrostate.
In terms of our ensemble, this implies that the overwhelming majority of the member
systems will be in microstates that correspond to one macrostate (or equivalently, one pair
of occupation number sets). For the system considered here, we therefore seek to find the
pair of occupation number sets {Naj}, {Nb,j} that maximizes W subject to the constraints
(2.1)—(2.3). We expect that this set of occupation numbers are those that characterize the real
system at equilibrium. Once determined, this distribution will provide the means to relate mi-
croscopic characteristics of the particles to macroscopic properties of the system at equilib-
rium. These results will form the foundation of our development of thermodynamics theory.
The paired sets of occupation numbers that maximize \nW also maximize W. Since
attempting to maximize In W is mathematically simpler than attempting to maximize W,
we will adopt the objective of finding the sets of occupation numbers {Naj}, {Nhj} that
maximize In W subject to the constraints (2.1)—(2.3).
Taking the natural log of both sides of Eq. (2.7) yields

\nW= lnNa\
/=0 /=o 7=0
(2.9)

Since Na and Nb are typically very large, we make use of Stirling's approximation (see
Appendix I),

\nNl=N\nN -N (for large N). (2.10)

Using (2.10) to evaluate the In Na and In Nb terms, the relation (2.9) for In W becomes

In W = Na InN a - Na + ^ NaJ I n g t - 2_^(NaJ InN aJ - Naj) + Nb InN b

;=o /=o

- Nh b,j In gj (2.11)
y=0
2.2 I The Microcanonical Ensemble and Boltzmann Statistics 35

Using constraints (2.1) and (2.2), this reduces to

oo oo
N ln
InW = Na lnNa + J2 aJ S/ ~J2
;=o /=o

(2.12)
7=0 7=0

Differentiating both sides of the above equation yields

oo oo

<i(ln W) = dNa \nNa + dNa + V^ dNaj Ing/ — 2_.(dNaj ln Naj + dNa,i)

/=o /=o
CO OO

+ dNb\nNb + dNb + ^d/V^jlngy — Y/(dNbj\nNbj + dNbj),

7=0 7=0

(2.13)

and since constraints (2.1) and (2.2) imply that

oo oo

/ dNa ( = dNa = 0 , y dNb .- ^ dNb = 0,

/=0 7=0

Eq. (2.13) reduces to

oo oo oo oo
m
d(ln W) = 2_2 dNaj ln gj — ^ J dNaj ln Na,i + ^ J ^ ^ , 7 In gy — ^ J ^ ^ , 7 ^,7 •
/=0 /=0 7=0 7=0
(2.14)

At the maximum that we seek, d(\n W) = 0. Imposing this requirement on Eq. (2.14) gives
00 00 00 00

/=o /=o 7=0 7=0

(2.15)

Differentiating the constraint relations yields

00

YdN<*,i =dNa = 0 , (2.16a)

/=o
00

YdNbj = dNb = 0, (2.16b)

and
00 00

Y SidNa,i + ] T ejdNbj = dU = 0. (2.16c)

/=o 7=0
36 21 Statistical Treatment of Multiparticle Systems

We now make use of the method of Lagrange multipliers. (See Appendix I for a description
of this technique.) We multiply the constraint equations by Lagrange multipliers Ai, A2, and
A3 respectively:
00

Xx^dNaJ=0, (2.17a)
/=o
00

X2^dNbJ=0, (2.17b)
y=o

SjdNbj = 0. (2.17c)
/=0 7=0

We then add these equations and subtract Eq. (2.15) to get

00 00 00
£ dN
dNaJ +X3Y, ' aJ ~ Y,dNaJ Ingi + ^ dNaJ In NaJ
i=0 i=0 i=0 i=0
0 0 0 0 0 0 00

8 dN dNb ln
+ X2 Y^ dNbJ + A3 J^ i *J ~ Yl 'J 8j + J2 dNb'Jln N
*>J = °'
7=0 7=0 7=0 7=0

which reduces to
OO OO

^T(A! + A3£/ - ln^- + \nNaj) dNaJ + ^ ( A 2 + A3£y- - \ngj + l n ^ - ) dNbJ = 0.

/=o 7=0
(2.18)
At the maximum, this relation must hold for arbitrary choices of dNaj and dNbj, which
implies that the coefficients of the differential terms must all be zero:
A! + X38i - ln gi + \nNaj = 0, (2.19a)
A2 + k3sj - In gj + \nNhJ = 0. (2.19b)
Solving these relations for Naj and Nbj, we obtain
NaJ=gie-^e-^£', (2.20a)
x
NhJ=gje-^e- ^. (2.20b)
Using the constraint relations (2.1) and (2.2) together with Eq. (2.20) yields
O O O O OO

/=0 i=0 i=0

OO OO OO
Nb
Yl >->
7=0
i =:
/
7=0
Sj e 2{
\2 y~^
7=0
(2
"21b>

Combining Eqs . (2.20) and (2.21), we find that

Naj gv gie~k*
(2.22a)
2.3 I Entropy and Temperature 37

' '
The distributions above are generally written in the forms

8
^ = A^ (2.23a)
and

Nb Zb
where
oo

Za = ^2gie-k3ei, (2.24a)
;=o
oo

Zb = J28Je~k3eJ' ( 2 - 24b >

Equations (2.23a) and (2.23b) above are the distributions that maximize In W subject to the
imposed constraints. They are generalized Boltzmann distributions in terms of the as yet
undefined constant A3. The parameters Za and Zb are termed partition functions. In a single-
component system only one partition function would be defined, usually just designated as
Z. We will see that a partition function is itself a thermodynamic property because it is a
unique function of the macrostate of the system.

2.3 Entropy and Temperature

In the analysis in the previous section, note that once sets of N aj and Nbj values
are specified, W can be calculated using Eq. (2.7). Thus, W is a macroscopic property of this
system, which indicates the macrostate of this system. For the same reasons, In W may be
considered to be a macroscopic property of the system. Either W or In W could be used as
an indicator of the macrostate for a system with specified Na,Nb,U, and V. Although they
both contain the same information, using In W as a property is advantageous. To understand
why, we first adopt the following definition for a property we call the entropy of the system:

S = kB In W. (2.25)
In the above definition, kB is a numerical constant that we set to define the units for this
property. In accordance with convention, &B is referred to as the Boltzmann constant in
recognition of Boltzmann's pioneering work in statistical mechanics on which the derivation
of Eq. (2.25) is based.
Now we will explore how this property behaves when we consider an isolated system
composed of two subsystems (see Figure 2.3).
When the entire composite system is at equilibrium, the overall W and S values for each
subsystem are uniquely determined:

Si=kB\nWu (2.26)
S2 = kB\nW2. (2.27)
38 2 / Statistical Treatment of Multiparticle Systems

impermeable wall

subsystem 1 subsystem 2

XXXXXXXXXXXXXXX

Figure 2.3

By definition, W\ is equal to the number of ways we can divide Ni distinguishable particles

among the energy microstates of system 1, and for each way of dividing up the N\ particles in
subsystem 1, there are W2 ways of dividing up the N2 particles of subsystem 2 among its
energy microstates. It follows that the total number of ways in which the two subsystems
can divide N\ + N2 particles among their respective energy microstates is W\ W2\
(2.28)
Hence, for the overall system
S1&2 = kB \n{Wx W2} = In Wx In W2 = S2. (2.29)
Clearly the entropy of the composite system is equal to the sum of the entropies of the
two subsystems. Since this analysis can easily be extended to any number of subsystems by
considering them two at a time, we conclude that the entropy of a composite system with an
arbitrary number of subsystems is additive over the constituent subsystems. This additivity
property is intuitively attractive and practically useful. When conducting a thermodynamic
analysis of a system composed of several subsystems, the definition of this entropy property
ensures that the entropy of the overall system is just the sum of the entropies of the compo-
nent systems. We therefore adopt the definition (2.25) for a new macroscopic property of
systems called entropy and include it in our development of the theoretical framework of
thermodynamic s.
It is important to note that the definition of entropy given in Eq. (2.25) is applicable
to any macrostate (with any sets of occupation numbers). If we designate the occupation
number sets that maximize W for the binary system considered in the previous sections as
{Naj}* and {Nbj}*, then the value of entropy for the system at equilibrium is given by
S = kB In WmSiX (at equilibrium), (2.30a)
where
(2.30b)
We now wish to return to the issue of evaluating the constant A3 in Eqs. (2.23a,b). We
begin by substituting Eq. (2.12) into Eq. (2.25) to obtain

— =Na\nNa
kB
/=o ;=o
oo oo

^ Nbj In gj — ^^ Nbj In Nbj • (2.31)

2.3 I Entropy and Temperature 39

In (2.31) we next evaluate Na,t and Nhj terms in the summation using Eqs. (2.23a,b),

and reorganize the summations to obtain

—
k = yVfl InNa + -Z2 . V fte-^< (In ft + X3e,- - InNa - In f t + In Za)
U
-^y] gje-^' (In gj + \}£j -lnNh- In gj

a +
"a 1=0 a
1=0

oo , ,• oo

Zb
1=0 j=o

b b
^ ^ - ^ - ^ ^ _|_ -u—v X ^ .^-^3£; (2 33)

Using Eqs. (2.24a,b) to evalate the last two summations in Eq. (2.33) yields

7=0
(2.34)

Equations (2.23a,b) indicate that the expressions in brackets in the / and j summations
above are just Naj and Nbj, respectively. We can therefore write Eq. (2.34) as

S ( \
r = A J 2 > "'i + J2 J 'J + N"ln Z" + Nh ln Zb-
X e N s Nh (2 35)
-
KB
\i=o y-=o /
Equation (2.3) implies that the term in brackets is equal to U. Substituting U for this term
in (2.35), we obtain

^ (2.36)
40 2 / Statistical Treatment of Multiparticle Systems

We next differentiate (2.36) with respect to A3 to get

Z dk Z
3 / Na,Nb,V " \ 3 / Na,Nh,V b \ 0*3 / Af^jVj.V

(2.37)

The last two partial derivatives on the right side can be evaluated using the definitions of
Za and Zb (Eqs. (2.24a,b)):

(2-38)

239
Na,Nh,V \ 7-0 / Na,Nh,V

In general, the energy levels and the degeneracies for the particles in the system are expected
to be functions of the volume of the system and the number of particles of each species.
The particle-in-a-3-D-box quantum solution, which is a model of the translational energy
storage in the particles, indicates that the quantum energy levels associated with translation
depend directly on the system volume (see Table 1.1). Dependence of the energy levels
on system volume and number of particles may also occur if the particles are electrically
charged or if longer-range molecular attractive forces are important. In general, it is therefore
reasonable to expect that the energy levels and degeneracies are functions only of Na, Nh,
andV:

Si =£i(Na9Nb,V), (2.40a)
8i=gi(Na,Nb,V), (2.40b)
ej=ej(Na,Nb,V)9 (2.41a)
gj=gj(Na,Nb9V). (2.41b)

Furthermore, since Na, Nb, and V are fixed for the system under consideration, the
Si, gi, £j, and gj arefixedand can be considered as constants in taking the partial derivative
on the right side of (2.35). Evaluating of the derivatives in Eqs. (2.38) and (2.39) and adding
the equations together then yields

Z
Na,Nb,V b V 9 ^ 3 / Na,Nb,V Z
J

(2.42)

Using Eqs. (2.3) and (2.23a,b) it can be easily shown that the right side of the above equation
is just —U:

N.(3Z.\ +*>(*?>) =_(/. (2 . 43)

Za \ 9 ^ 3 / Na,Nh,V Z
b \ <^3 / Na,Nh,V
23 I Entropy and Temperature 41

Substituting Eq. (2.43) into Eq. (2.37) and solving for A3 yields
A
3= , /qrnn ! N"M'V - (2.44)

Since from basic multivariable calculus

,iV,,,V 1
(2.45)

Eq. (2.44) can be converted to

kB(dU/dS)NatNb,v
Equation (2.46) is the sought-after relation for A3. Obviously the derivative
(dU/dS)Na,Nb,v plays an important role in the statistical thermodynamics framework we
are developing. Because of its importance, we define a special symbol T to denote this
derivative:
T = (dU/dS)NatNb,v. (2.47)
Since T is usually the symbol that designates temperature in thermodynamics texts, the
reader may wonder whether this is, in fact, the temperature. Although we have not demons-
trated it here, this is an appropriate definition of temperature. In a later section, we will
show that this definition is consistent with the usual intuitive understanding of temperature.
With this definition, the relation for A3 becomes

^3 = T V (2.48)
k BT
The equilibrium distributions of the particles among the available energy states is then
given by

(2.49a)
Na Za
NbJ -ej/kBT
(2.49b)
Nh zb
where

Za =
oo
,-.„,. (2.50a)
i=0
oo
zb = E^ (2.50b)

Replacing X3 with \/kBT in Eqs. (2.36) and (2.43) then gives the following relations
among thermodynamic properties of the system:

^ ^ (2.51)

• (2-52)
Na,Nb,V
42 2 / Statistical Treatment of Multiparticle Systems

Now that we have developed a definition of temperature, a word about units is in order.
Four temperature scales are commonly used. The Kelvin scale uses units defined so that
100 units on the Kelvin scale (K) correspond to the difference in temperature between
the boiling point and freezing point of water at one atmosphere pressure. On this absolute
scale, the temperature at the triple point of water is 273.16 K. The other common scale
is the Rankine scale (°R), which uses units defined such that 180 units on the Rankine
scale correspond to the difference in temperature between the boiling and freezing points
of water at atmospheric pressure. The triple point of water corresponds to 491.688°R on
this scale. Temperatures in the Celsius (°C) and Fahrenheit (°F) scales are related to Kelvin
and Rankine scale values as:

°C = K-273.15,
°F = ° R - 459.67.

To this point we have not discussed the magnitude or units for the Boltzmann con-
stant. Equations (2.49a,b) imply that the units should be energy divided by temperature.
For consistency with the SI system, the appropriate value for the Boltzmann constant
is

kB = 1.3805 x 10~23 J/K.

Example 2.1 A system contains distinguishable species a and species b boson parti-
cles at temperature T = 300 K. The species a particles can occupy only the energy levels
k^0a, 2k^0a, or 3k#0a whereas species b particles can occupy only the levels k^Ob, 2&B#/>,
and 3kB6b- 0a = 60 K, Ob — 30 K and each energy level has a degeneracy of one. The system
contains NA = 6.02 x 1026 particles of each species. Determine the distribution of energy
levels occupied at equilibrium and the mean energy of a particle in the system.

Solution The partition function for the species a particles is given by Eq. (2.50a):
oo
Za — V ^ gie~ Sl/kBT = e~kB6a/kBT + e-uBea/kBT + e-3kBea/kBr

i=0

—0.2 i —0.4 i —0.6 o A^5 o

= e + e +e = 2.038.
A similar calculation using Eq. (2.50b) yields Zb = 2.464. The distribution of particles
among available energy states for species a is given by Eq. (2.49a). Substituting for the first
energy level, we obtain

N a l esi/kB: e-kBeai lkBT s>-0-2

Na Za Za 2.038

Similar substitutions for the second and third energy levels yield

Na 2 e~0A Na3 e~06

- 0 .329, — KJ.ZOy.
Na 2.038 Na 2.038
2.4 I The Role of Distinguishability 43

The values of Naj /Na thus determined indicate the fraction of particles that, on the average,
occupy the /th energy level for species a. For species b, using Eq. (2.49b) in a similar manner
yields
NhX e~0A Nb2 e~02 Nb3 e~03
—— = = 0.367, —— = = 0.332, —— = = 0.301.
Nb 2.464 Nb 2.464 Nb 2.464
For species a, the mean particle energy (sa) is obtained by taking a weighted average over
all accessible levels:

(ea) = V -^-si = 0A02kB0a + 0329(2)kB0a + 0.260(3)^^

jri Na
= \MlkB0a = 1.867(1.3805 x 10~23)60 = 1.546 x 10~21 J.
The same method is used for species b:
3

(sb) = ^2 ~^l£j = 0361 kB0b -h 0.332(2)*B% + 030l(3)kB0b

= l.934kB0b = 1.934(1.3805 x 10"23)30 = 8.010 x 10~22 J.

Thus for this system the mean energy values for the two species differ by about a factor of
two.

2.4 The Role of Distinguishability

The analysis in the previous section is based on the premise that the particles are
distinguishable. This is appropriate for solid crystals where each atom is localized near a
specific point in the crystal lattice. Unfortunately, it is not appropriate for fluids in which
the particles (atoms, molecules, or fundamental particles) are free to move about within the
system volume. In systems with freely moving particles, the particles of a given species
are indistinguishable. Our task in this section is to examine how our analysis from the
previous section must be changed to account for the indistiguishability of the particles in
fluid systems.
In considering modifications to the analysis to account for indistinguishability, we will
also account for restrictions on microstate occupancy. As discussed in Section 1.8, virtually
all particles fall into two categories: bosons and fermions. There are no restrictions on the
number of bosons that may occupy any given microstate. Most atoms and molecules behave
as bosons. One exception is 3He, which is a fermion. For a system of fermions, no more than
one fermion may occupy a given microstate. Electrons and protons are perhaps the most
important examples of particles that behave as fermions. The restriction on the occupancy
of microstates for electrons is a consequence of the Pauli exclusion principle.
We first consider systems of bosons. As in the last section we are considering member
systems in a microcanonical ensemble having specified N a, Nb,V, and U. If we consider
one energy level occupied by Naj particles of species a, the number of ways of selecting the
44 2 / Statistical Treatment of Multiparticle Systems

Naj particles from the total of Na particles in the system is just 1, because the particles are
indistinguishable. No matter which Naj we choose, the particles occupying the energy level
would "look" the same. However, we can permute the particles among the gi microstates for
the energy level in a number of ways. Because the particles are bosons, there are no restric-
tions on the way that the particles may occupy the microstates. From basic combinatorial
analysis (see Appendix I) we can determine the number of these permutations as

(number of ways of arranging Naj 1 (gt + Naj — 1)!

{indistinguishableparticles among gi microstates J (gi — l)\Naj\
(2.53)

Because the arrangements within energy levels are independent of one another, the total
number of ways of arranging the a species particles among the microstates is the product
of such terms for all energy levels:

(number of ways of arranging "j ^

< species a particles among > = |T — — . (2.54)
~_n (gi ~~ 1)-Na,i-
l
{microstates dictated by gi and {Naj} J ~°

A similar relation must apply for species b:

( number of ways of arranging ")

J • u ,• / I rr (gj+NbJ-1)1
< species b particles among >= I I — . (2.55)
{microstates dictated by gj and {Nbj} J J=o
For each arrangement of the a particles, the number of possible arrangements of b particles
is given by Eq. (2.55). The total number of ways of combining the arrangements of the two
species is therefore equal to the product of the right sides of Eqs. (2.54) and (2.55). We
designate this total number of microstates for the specified pair of distributions {N aj} and
{Nbj}, as

wbos =ft<*' +_^; -l)-ft<*J + ">J- !) ! ,

i=0 l
' a,i- j j ,j •
(2.56)
= Q

Alternatively, the particles may be indistinguishable fermions. Again, because the par-
ticles are indistinguishable, selection of any Naj particles of species a for energy level i
amounts to the same occupancy. The fermions may be permuted among the gi microstates,
but no more than one may occupy a given microstate. Thus a microstate is either occupied
by one particle or unoccupied. Combinatorial analysis again provides the means to calculate
the number of such arrangements (see Appendix I):

{number of ways of selecting Naj microstates { Ol.

\Z.J I)
{to be occupied from gi total microstates J Najl(gi — Naj)\
Because the arrangements within the energy levels are independent, the total number of
system microstates is the product of terms like the right side of Eq. (2.57) for all energy
2.4 I The Role of Distinguishability 45

levels:
(number of ways of arranging "| ^
I species a particles among ( =Y[ ^1 AJ ' (2 58)
'
u
{microstates dictated by gi and {Naj\ J '~
The corresponding equation for species b is
( number of ways of arranging "j ^
< species b particles among ( =J1 Tr— *'__ M • (2'59)
J
\microstates dictated by gj and {Ntj} )
As in the case for bosons, the total number of microstates for the two-species system is the
product of the terms on the right sides of Eqs. (2.58) and (2.59). To indicate that this result
is for fermions, we denote the total number of microstates as
oo . oo .

W(er = TT it FT it . (2.60)
Nl
l (SNVf}N\(gN)\
The total number of microstates for either bosons or fermions can be represented by the
relation

tf,/}' \Nb,j}) = I I 7 ,,T ,.AT—: II

(2.61)
where W for indistinguishable bosons corresponds to
rj = 1, £= 0 (indistinguishable bosons), (2.62)
and for indistinguishable fermions we set
K) = 0, §= 1 (indistinguishable fermions). (2.63)
For systems containing indistinguishable bosons or indistinguishable fermions, we again
seek to find the occupation number sets {Naj}, {Nt,j} that maximize \nW subject to the
constraints (2.1)—(2.3). As before, we interpret the pair of occupation number sets that
maximizes In W as being the equilibrium distribution for the system. Taking the natural log
of both sides of Eq. (2.61) and rearranging yields
oo oo

lnV^b/f = y j l n { [ g ; + rj(Naj — 1)]!} — V^ln{[g z — r\ — ^Naj]l]

i=o ;=o
oo oo

— \^\n{Naj\} + V^ln{[^ 7 + rj(Nbj — 1)]!}

00 OO

j=0 j=0
46 2 / Statistical Treatment of Multip article Systems

Since all the factorials on the right side of Eq. (2.64) are expected to be large, we apply
Stirling's approximation to each:

$} aii - 1)} - [gi + r](Naj - 1)]]

/=o
oo

i=0
oo

;=o j=o

x ln{gj + n(NbJ - 1)} - [gj + r](NbJ - 1)]] -

j=0

Nh,j \nNhJ - NbJ].

(2.65)

Using the constraint relations (2.1) and (2.2), the above relation simplifies to

lnWb/{ = Y)lgi + Tl(Na<i - l ) ] l n { # + ii{Naj - 1)} - \g, - j, - $NaJ]

i=0

x ]n{gi - r \ - $Na,i) - NaJ \nNaJ + (1 - rj -

bJ - 1)] ]n{gj + y?(^,y - 1)} - [gj -V- $NbJ]

7=0

x ln{ gi - n - $NbJ] - NbJ \nNbJ + (1 - r\ - £)#*,;]]. (2.66)

Differentiating, we obtain

d(ln Wb/f) = Y^[vdNa,i \n{g, + r,(NaJ - 1)}

i=0

- n - $NaJ] + $dNaJ - dNaJ In NaJ - dNaJ

bj ln{gj + n(NbJ - 1)}

bj \n{gj - n - $NbJ] + $dNbJ

- dNbJ In NhJ - dNbJ + ( ! - » ; - § ) dNhJ]. (2.67)

2.4 I The Role of Distinguishability 47

Setting J(ln Wb/f) = 0 and simplifying yields

oo

^ | > 7 l n { g / + rj(Naj — 1)} -f § ln{g/ — rj — %NaJ} — lnNaj

i=0

gj + ri(Nbj - 1)} + H Mgj ~ n - $NbJ} - InN bj ] dNbJ = 0.

(2.68)

Using the method of Lagrange multipliers, we multiply the differential constraint relations
(2.16a), (2.16b), and (2.16c) by Xi, A2, and A3, respectively, add the resulting equations,
and subtract Eq. (2.68). The final result is
oo

^[kidNa,i + k3eidNati - [ri\n{gi + fi(NaJ - 1)}

/=o
00

+ | Into - r, - SNaJ) - \nNaJ] dNaJ] + J } ^ ^ , ; + tejdNbj

- [r, ln[gj + r,(NhJ - 1)} + $ Info - r, - $NbJ) - lnNbJ] dNhJ] = 0,

(2.69)

which simplifies to
oo

Y^lki + k3et - r) \n\gi + r](Naj - 1)} - § ln{# - r\ - t;NaJ} + lniVflf/] dWfl,/

/=o

oo

7=0

=0. (2.70)
At the maximum, the coefficients of the dNaj and dNbj terms must all be zero, which
implies that

A.! + X3£i - n ln{gi + n(Na,i - 1)} - H ln{# - »? - t ^ , , } + In Wo,,- = 0,

(2.71a)
A-2 + ^£j ~ n ln{g7- + v(Nbj ~ 1)} - H Info - »; - ?Wft,;} + lnNbJ = 0.
(2.71b)

For bosons, we set r\ = 1 and $ = 0 in Eqs. (2.71a,b) and rearrange to obtain

) (2.72a)

Nb : = gje-k2e-liS> (1 + ^ - — ) (for bosons). (2.72b)

V gj 8jJ
48 21 Statistical Treatment of Multiparticle Systems

For fermions, we set r\ = 0 and § = 1 in Eqs. (2.71a,b) and obtain

(2.73a)

NbJ = gje-kle-k'£J 1 ^ (for fermions). (2.73b)

V 8j J
Equations (2.72a,b) and (2.73a,b) can be solved explicitly for Na,i and A^ j , but the resulting
relations are complicated and it is difficult to carry the analysis further. Instead, we note
that for systems of particles with moderate to high energies, the degeneracy of the most
populated energy levels is enormous (see Example 1.4). For such circumstances, the number
of microstates for each energy level is much larger than the number of particles in the system
occupying that level and very few of the available microstates are occupied. Systems in
which this is true are said to exhibit dilute occupancy. For conditions that result in dilute
occupancy, Naj/gi, Nbj/gj, 1/g;, and \/gj are negligible compared to one. If we neglect
these ratios compared to one in Eqs. (2.72a,b) and (2.73a,b), we find that both sets of
equations reduce to

Na,i=gie-Me-X3e', (2.74a)
Xl XlS
NbJ=gje- e- '. (2.74b)

Using the constraints (2.1) and (2.2) on Naj and Nbj, we can eliminate the multipliers X\
and A.2 and obtain

(2 75a)
'
,jor indistingUishable bosons or fermions\
\in the limit of dilute occupancy J
(Z75b)

which is identical to the Boltzmann statistics result obtained for distinguishable particles
with no restrictions on microstate occupancy.
Although the form of the distributions (2.75a,b) is the same as Boltzmann statistics for
distinguishable particles, the relations for W for indistinguishable bosons and fermions are
different from the relation for W that applied to distinguishable bosons. We must therefore
reexamine how the change in the relation for W affects the thermodynamic properties for
the system. Since by definition S = k^ In W, we set S/k^ equal to the right side of Eq. (2.66),
and use the distributions (2.75a,b) to evaluate Naj and Nbj. In addition, because we have
dilute occupancy, we neglect Naj/gi, N^j/gj, 1/g/, and l/gj compared to one, where
appropriate. For both the boson and fermion particle types, the resulting equation for S/ k^
reduces to

-^ = X3U + Na \nZa - Na lnNa + Nb \nZb - Nb \nNb + Na + Nb. (2.76)

To evaluate X3 we use the same approach as was used in Section 2.3 for distinguishable
particles. We differentiate with respect to A.3, evaluate the derivatives and solve for A3. The
2 A I The Role of Distinguishability 49

result of this manipulation is

1 1
(2.77)

It follows directly that for a system that exhibits dilute occupancy, the distributions and the
partition function definitions for indistinguishable bosons or fermions are identical to the
Boltzmann statistics case:
Na,i g -ei/kBTT

Na Za '
(2.49a)
,-Sj/kBT
Nbj 8
Nb zb (2.49b)
where
oo
(2.50a)
Z a- gi
/=0
oo
(2.50b)

It follows from these relations and the energy constraint (2.3) that the relation for U is
identical to that for Boltzmann statistics of distinguishable particles:
_ NakBT2 fdZa\ NbkBT2 (dZh\
U — —~ I ~7^r I H ~ I ~r^r I • (l.jl)
Z
Za \O1 ) Na,Nb,V b V "^ / Na,Nh,V
Substituting A3 = \/kBT into Eq. (2.76) yields the following relation for the entropy:

a
kB k BT \Na) \Nb)
Note that this relation differs from the entropy relation obtained for distinguishable bosons.
The net effect of particle indistinguishability is to reduce the entropy of the system. Equations
(2.49), (2.50), (2.52), and (2.78) thus provide the linkage between microscale energy storage
and the macroscopic thermodynamic properties for a system of indistinguishable bosons or
fermions in the moderate to high energy limit where dilute occupancy occurs.
What about low-energy systems? It turns out that in very cold systems we cannot invoke
the dilute occupancy approximation and we must allow for low degeneracy and quantum
effects in the analysis of the statistical behavior of such systems. We will return to examine
such systems in more detail in Section 6.4. Fortunately, in engineering applications, the
overwhelming majority of system types and particle energy levels encountered do lie in
ranges where dilute occupancy occurs. The results summarized above therefore provide a
useful foundation for thermodynamic analysis of this broad range of system types.

Example 2.2 Initially a system contains one kmol (A^A = 6.02 x 1026 molecules) of water
molecules in a liquid phase at atmospheric pressure and 373.2 K. Energy is added to the
system at constant pressure until it contains only a vapor phase at 373.2 K. Evalute Zg/Zf,
50 21 Statistical Treatment of Multiparticle Systems

where Zf is the partition function for the initial liquid state and Z g is the partition function
for the final gaseous state.

Solution For a single-component system, Eq. (2.78) reduces to

-kB = —
k BT
+Na\
Since Na — NA, the resulting S and U values are those for one kmole of water, s and w,
respectively:

kB
Evaluating this relation at the initial liquid state and the final gaseous state and subtracting
the two equations yields
^g ~ 5f u g-uf
*B k BT
Rearran^*ing, we get
Sg-Sf ug - M f

which can be solved for Z g /Zf to give

Zf [ NAkBT J
where Sfg = sg — Sf and wfg = wg — Mf. Values of 5fg and Mfg are tabulated in standard steam
tables:
at T = 373.2 K: sfg = 108.9 kJ/kmol K, ufg = 3.758 x 104 kJ/kmol.
Substituting these values into the above relation for Zg/Zf, we obtain
373.2(108.9)-3.758 x l O 4 \ _ ^ , vin,58
6XP
Zf ~ \ (6.02 x 1026)1.3805 x 10-^(373.2)
Conversion from a liquid to a vapor phase has resulted in an enormous increase in the
partition function. From its definition, Eq. (2.50a), it is clear that an increase in Z results
from an increase in degeneracy or an increase in the number of accessible energy levels that
are comparable to kBT. Conversion of liquid to vapor increases both the degeneracy and
the number of accessible energy levels by a large amount.

2.5 More on Entropy and Equilibrium

Entropy
Entropy is such an important property in the thermodynamics theory we are devel-
oping that it is useful to clearly document the characteristics of entropy that have emerged
from our analysis so far. There are four important characteristics of the entropy property that
2.5 I More on Entropy and Equilibrium 51

we have identified in our development. First, we have defined entropy so that the entropy
of a composite system is additive over the constituent subsystems.
Two additional important properties of the entropy relate to its functional dependence on
other properties for a system at equilibrium. We have shown that if we specify the energy
U, volume V, and the number of each species of particles in a system, at equilibrium, the
value of the system entropy is uniquely defined as the maximum of S for all the macrostates
consistent with the imposed constraints. Thus equilibrium can be viewed as a condition
dictated by an extremum principle whereby the system seeks a state of maximum entropy.
This is sometimes referred to as the entropy maximum principle. In addition, since the
resulting maximized S value depends on the specified U,V,Na, and A^, for equilibrium
states, S must be a function of U, V, Nai and A^:
S = S(U,V,Na,Nb). (2.79)
A relation among U,V, Na, Nb, and S is termed a fundamental relation for a thermodynamic
system. We will show in the next chapter that all conceivable thermodynamic information
about a system can be obtained from the fundamental thermodynamic relation.
The fourth characteristic of entropy relates to its dependence on internal energy U.
As the energy of a system increases, it will increasingly occupy higher quantum energy
levels. For the 3-D particle-in-a-box quantum solution considered in the first chapter, it was
shown that the degeneracy of the energy levels increases monotonically as the energy level
increased. Because the particle-in-a-box solution is an appropriate model of translation
energy storage in particles, we expect that in any system with translating particles, the
system will increasingly occupy energy levels with higher degeneracies as the system energy
increases. Furthermore, Eq. (2.31) clearly indicates that entropy will increase monotonically
with increasing energy level degeneracy. Taken together, these observations imply that
entropy will be a monotonically increasing function of internal energy.
In reaching this conclusion, we have considered only translational energy storage. In
systems of molecules that have rotational and vibrational energy storage in addition to
translation, these additional mechanisms can be modeled with the rigid-rotor and harmonic-
oscillator solutions discussed in Chapter 1. We will see in Chapter 5 that the vibrational
energy storage in polyatomic molecules can be modeled as a combination of independent
normal modes of vibration, each of which can be considered to be a simple harmonic oscil-
lator. Depending on the structure of the molecule, the vibrational energy storage levels for
the molecule may be nondegenerate or its vibrational energy levels may have a degeneracy
greater than one. The rigid-rotor model solution implies that the degeneracy of rotational
energy levels becomes increasingly large as the energy level increases. The overall effect
of combining rotational energy storage and/or vibrational energy storage with translation is
essentially the same as translation alone: System entropy will increase monotonically with
system energy.
What if there is no translation, as is the case in a crystalline solid? In that case, the atoms
are bound to surrounding molecules at a specific mean location. We will examine such a
system in detail in a later chapter. Although the simple one-dimensional harmonic oscillator
considered in Chapter 1 exhibits energy levels that are nondegenerate, in a three-dimensional
lattice of atoms, the number of normal oscillatory modes accessible to the system generally
increases as the energy of the system increases. This trend produces a rapid increase in
number of possible system microstates as the system energy rises. Equation (2.25) then
implies that entropy will steadily increase with system internal energy.
52 21 Statistical Treatment of Multiparticle Systems

The overall conclusion is that, for solid or fluid systems, the entropy of the system will
increase monotonically with increasing system internal energy. This characteristic is a direct
result of the fact that, at higher energies, the system generally can span a wider range of
energy levels and the degeneracy of the energy levels increases with increasing energy. The
important characteristics of the entropy function are summarized below:
(i) Equilibrium corresponds to a maximum entropy for a system at specified U,V,Na,
and Nb. This is the entropy maximum principle.
(ii) The entropy of a composite system with an arbitrary number of subsystems is
additive over the constituent subsystems. This is the additivity property.
(iii) At equilibrium, the value ofentropy is a function oft/, V, Na, and Nb. This function
is referred to as the fundamental relation for the system,
(iv) Entropy is a monotonically increasing function of energy.
One final aspect of the definition of entropy is worth noting. In the previous section it
was shown that for a system at equilibrium the entropy is given by
S = &B In WmSiX (at equilibrium), (2.30a)
where WmSiX is given by Eq. (2.30b). It was also argued that for the system to achieve a
macroscopically stable equilibrium, Wmax must be very much larger than the W value for
any other macrostate. As indicated by Eq. (2.8), the total number of accessible microstates
Q must be the sum of the W values for all macrostates. If we designate Wt as the W value
for macrostate /, then Q, is computed as
M
y
Q = }T/Wi = Wi+W2 + W3 + '- + WM, (2.80)
i=\

where M is the total number of macrostates consistent with the system constraints. In the
above expression for £2 there are M positive terms, the largest of which is WmSLX. The sum
must be at least as large as WmSiX and is certainly less than M times WmSiX:
Wmax <Q< MWmax.
Taking the natural log of all terms, we obtain
In WmSiX < In £2 < InM + In Wmax.
With real systems in which the number of particles is on the order of Avogadro's number,
the number of microstates W is far greater than the number of macrostates M. In the above
relation, In M is therefore negligible compared to In W. It follows that to a high degree of
accuracy
(2.81)
and the entropy of the system at equilibrium is given by
S = kBlnQ. (2.82)
The significance of this result is that it links the entropy of the system at equilibrium
to the total number of microstates accessible to the system. It follows from the fact that at
equilibrium Q is virtually equal to WmSiX because WmSiX will be so much larger than any other
contribution to Q. Qualitatively this result implies that, to a high degree of accuracy, we can
2.5 I More on Entropy and Equilibrium 53

say that if the number of accessible microstates is doubled, the entropy will increase by a
factor of In 2. Conversely, the magnitude of the entropy is a direct indicator of the number
of microstates accessible to the system.

Example 2.3 As in Example 2.2, we here will consider a system that contains one kmol
of water molecules (NA = 6.02 x 1026 molecules) initially in a liquid phase at atmospheric
pressure and 373.2 K. Energy is added to the system at constant pressure until it contains
only a vapor phase at 373.2 K. Evalute Qg/ £2f, where Qf is the total number of accessible
microstates for the initial liquid state and £2g is the total number of accessible microstates
for the final gaseous state.

Solution For any system at equilibrium, Eq. (2.82) requires that

S = kB\nQ.
Since we are considering a system containing one mole the entropy values in the inital
liquid and final vapor states are the molar values, Sf and sg, repsectively. It follows from the
above equation that
Sf = /:Bln£2f and sg = kBlnQg,
where Qf is the number of microstates accessible to the system in the initial liquid state and
Qg is the number of microstates accessible to the system in the final vapor state. Combining
these equations yields

sg - Sf = kB In I — I ,

which can be solved for £2g/ £2f to give

^ = eXP { it
where Sfg = sg — Sf. Values of Sfg are tabulated in standard steam tables:
at T = 373.2 K: ^fg = 108.9 kJ/kmol K.
Substituting these values into the above relation for Qg/ Qf, we obtain
^g [108.9x10001 34xl027
O^ = eXP
\ 1.38 x 10-23 ) = 1
° •
Conversion from a liquid to a vapor phase has clearly resulted in an enormous increase
in the number of microstates accessible to the system.

Conditions at Equilibrium
With the knowledge of entropy obtained thus far, we now will more fully explore the
macroscopic requirements for equilibrium. Consider the isolated system shown in Figure 2.4
consisting of two subsystems, I and II. Each contains particles of species a and species b,
54 2 / Statistical Treatment of Multiparticle Systems

f porous plate (free to move)

\\\\\\\\\\\\\N

subsystem / subsystem //

Figure 2.4 A composite system.

and the wall separating the two subsystems is a thin porous metal plate that allows free
exchange of heat and species exchange between the subsystems. The plate is free to move
so that its x position is unrestricted. The outer boundary of the overall system is rigid and
insulated so that no energy or mass exchange with the outside can occur. The values of
U9V,Na, and Nb for the total system are therefore fixed.
For each subsystem, S at equilibrium is a function of U, V, Na, and Nb in the subsystem.
Each subsystem could equilibrate individually at various U,V,Na, and Nb values that would
not violate the specifications of U,V, Na, and Nb for the overall system. The extremum
principle requires that, at equilibrium, the subsystems must reach equilibrium conditions
that maximize S for the overall system subject to the overall system constraints on U, V, Na,
and Nb- Additivity of S requires that

S = S 1 + S U. (2.83)

Differentiating this relation yields

(2.84)

The interpretation of the above equation is as follows. Once the overall system has reached
equilibrium, molecular-level fluctuations may briefly produce differential exchanges of
energy or mass or volume between the subsystems, which will produce changes in the
entropy of the subsystems. However, since S for the overall system is at a maximum, the
resulting differential change in overall S is zero. Equation (2.84) therefore implies that

(2.85)

At equilibrium for each subsystem

Sl = S1(U\v\Nl,N1b)9 (2.86a)
ll ll l l l
S = S (U \V \Nl\Nl ). (2.86b)

For subsystem I, we can therefore write the following first-order expansion:

+ dN
)) dm (2.87)
bJ Na,V,V
2.5 I More on Entropy and Equilibrium 55

We can similarly write for subsystem II

dN*. (2.88)
BNb Na,U,V

We further note that since

and U,V, Na, and Nb are fixed constants, differentiating these relations yields
dUl = -dUl\ (2.89)
l l
dV = -dV \ (2.90)
dN\ = -dNl\ (2.91)
J ^ = -dNf. (2.92)
Substituting Eqs. (2.87) and (2.88) into (2.85) and using Eqs. (2.89)-(2.92), we obtain

ds dUl
dS V dvj
dVl
Na,Nb,V U,Na,Nb U,Na,Nb

dS
dNla
a/ Nb,U,V SNa Nh,U,V

dNlh = 0. (2.93)
dNb Na,U,V

If dS = 0 for all possible exchanges, the perturbation coefficients in square brackets must
each be equal to zero. It follows directly that at equilibrium we must have

(2.94)
Na,Nb,V

(2.95)

/as V dS
\dNa Nb,U,V Uiy
a/ Nb,U,V
I
ds dS \ n
(2.97)
dN
Na,U,V b) Na,U,v'
Using Eq. (2.47),

(2.47)
V^ / Na,Nb,V
56 21 Statistical Treatment of Multiparticle Systems

Eq. (2.94) can be written

1 _ 1

or equivalently
Tl = T11. (2.98)
Thus to achieve equilibrium with respect to internal energy exchange with volume held
constant, we require that the T property defined by Eq. (2.47) must be the same in both
subsystems. This is the characteristic that we intuitively associate with the concept of
temperature:
• When two closed systems of fixed volumeare brought into contact, no net exchange
of internal energy will occur if the two systems are at the same temperature.
We therefore conclude that Eq. (2.47) does in fact provide an appropriate definition of
temperature T. The temperature so defined is termed the absolute temperature.
To analyze the meaning of the other relations (2.95)-(2.97), we first note the following
basic fact from multivariable calculus. If y is a function of x and z, then

dz)y \dy)\dz)x
For constant Na and Nb, S is a function of U and V and Eq. (2.99) implies that
dS
I i w , „ iw , • ( 2 - 10 °)
/ Na,b

We define a property P as

P = -[ — ) • (2.101)

Using (2.100) and this new property definition, Eq. (2.95) can be written as
pi pii

— = —. (2.102)

Since we have already shown that the temperatures must be equal at equilibrium, this
requirement reduces to
Pl = Pu. (2.103)

Note that this requirement is associated with exchange of energy between the subsystems
due to motion of the separating boundary, which causes one system to contract while the
other expands. This requirement implies that if two systems are separated by a boundary
surface that offers no resistance to applied forces, the two systems are in mechanical equi-
librium if this P property is the same in both systems. This notion is consistent with our
intuitive understanding of pressure or normal stress exerted by a fluid or solid on a system
boundary. We therefore adopt Eq. (2.101) as a definition of pressure.
2.5 I More on Entropy and Equilibrium 57

For constant V and A^, S is a function of U and Na and Eq. (2.99) implies

= . (2.104)

For constant V and Na, S is a function of U and Nt> and we similarly find that

= ( ) . (2.105)
V,Na,Nh \oiy/b/ V,Na,S T \dNb ) v,Na,S
We define properties \xa and fih as

Using Eqs. (2.104) and (2.105) and these new property definitions, Eqs. (2.96) and (2.97)
can be written as
u1 u11
ff = ^ (2-108)

Since the temperatures must be equal at equilibrium, these requirements reduce to

Hla = M?> (2.H0)

Mb = M"- (2.111)
These requirements are associated with species exchange between the subsystems. Equa-
tions (2.110) and (2.111) imply that if two systems are separated by a permeable boundary
surface, the two systems are in mass transfer equilibrium for each species if this \x property
is the same in both systems. The properties defined by Eqs. (2.106) and (2.107) are termed
the chemical potential for species a and the chemical potential for species b, respectively.
In a mixture there is a corresponding chemical potential for each species present.
In summary, in terms of these newly defined properties, for the system in Figure 2.4 we
can state the necessary conditions for equilibrium as being:

It should be noted that these are necessary conditions for equilibrium in the specific system
shown in Figure 2.4. If the characteristics of the boundary between the subsystems were
changed, the necessary conditions for equilibrium would generally be different. We will
explore this issue in more detail in Chapter 3. The analysis in this section serves to illustrate
that the definitions of temperature, pressure, and chemical potential make it possible to
concisely state the necessary conditions for equilibrium in the system shown in Figure 2.4.
We will see in Chapter 3 that for a wide variety of systems, it is often most convenient to
specify the necessary conditions for equilibrium in terms of these properties.
58 2 / Statistical Treatment of Multiparticle Systems

We have already discussed units for temperature. Pressure must have units of energy per
unit volume, or equivalently, force per unit area. The latter is consistent with the common
interpretation of pressure as the force per unit of surface area on the walls or boundary
surfaces that enclose a system. In the SI system, the units would be N/m 2. As defined here,
chemical potential has the units of energy per molecule. In the SI system, the units would
be J. The chemical potential per kmol jla and the chemical potential per unit mass fia are
also sometimes used in thermodynamic analysis. These are related to the chemical potential
per molecule as

jXa = fiaNA,

where NA is Avogadro's number and Ma is the molecular mass of species a. In the SI

system the units of jla and fLa are J/kmol and J/kg, respectively.

Negative Absolute Temperatures

In Section 2.5, we concluded that for common thermodynamic systems contain-
ing large numbers of atoms, molecules, or other particles, the entropy is a monotonically
increasing function of internal energy. Most common thermodynamic systems containing
atoms and molecules in a fluid or solid state have an infinite number of energy levels. As
the temperature of the system is increased, more particles are raised to higher energy levels.
This is accompanied by increasing disorder as the particles are distributed over more and
more states. As the energy increases, entropy increases and since, by definition, temperature
is given by

T = (dU/dS)Na,Nb,v (2.47)
it follows that the (absolute) temperature is positive (T > 0) for such systems.
From the definition (2.47) it is clear that the absolute temperature could only be negative
if the entropy of the system decreased as its internal energy increased. The discussion above
implies that this cannot occur in a system that has an infinite number of energy levels. In
such a system, an increase of temperature produces increasing occupancy of higher energy
levels. However, Eq. (2.49a) indicates that the ratio of the mean number of particles (of type
a) in two adjacent energy levels / and / + 1 is given by

c ^ (2.112)
N
a,i gi
In a system with infinitely many energy levels, if the system energy isfinite,the higher energy
levels of the system must be less populated than lower energy levels. Since gi+\/gi is slightly
larger than one and £/+i > £/, Naj+\/Naj can be less than one only if the temperature T is
positive. As T approaches infinity, Naj+\/Naj approaches one. This ratio could equal one
only if the system had infinite energy. Equation (2.112) implies that negative T would require
that Naj+\/Naj be greater than one, which would require more energy than in the limit
of infinite temperature. We therefore conclude that negative temperatures are impossible in
a system that has an infinite number of energy levels. Since most common systems are of
this type, negative temperatures are regarded as an impossibility for virtually all systems
encountered in nature and in technological applications.
2.5 I More on Entropy and Equilibrium 59

However, suppose we found a system that had only a finite number of energy levels. It
may then be possible to cause a population inversion in a system withfiniteenergy, resulting
in Naj+\/Naj > 1, which would imply that the system has a negative temperature. A system
having only two possible energy states is analyzed in Example 2.4. This type of system is
shown to have positive temperatures at low energy levels and negative temperatures at high
system energy levels in which the majority of the particles are in the higher energy state.
Thus, in a system with only two energy states, the population inversion that exists at higher
energies results in negative absolute temperatures.

Example 2.4 Particles in a system are distinguishable and can exist in one of only two
possible energy levels, each having a degeneracy of one. The energy of the lower ground
state is taken to be zero. The upper state has energy s\. The system contains N a particles.
Determine how the entropy and temperature vary with energy for this system.

Solution The internal energy of the system is zero if all the particles are in the
ground state. If Naf\ is the mean number of particles in state 1, then Na — Na,\ must be the
mean number in the ground state. It follows that the internal energy of the system must be
given by
U = NaAsl.
With NGii particles in state 1, the number of microsates W is the number of ways of putting
Na distinguishable particles into two bins such that there are A^,i in one and N a — NOi\ in
the other. It follows from Eq. (2.4) that

Taking the log of both sides and using the fact that S = k^ In W yields
S/kB = \n[NJ] - ln[(Na - NaA)\] - ln[iVa>1!].
Using the Stirling approximation, this relation becomes
S/kB = Na \nNa -Na- (Na - NaA) ln(Na - NaA) - (Na - NaA) - NaA In A ^ - NaA.
Since U = Na,\S\, we replace Na,\ with U/e\. Doing so and rearranging somewhat, we
obtain

NakB V NasiJ V NaeiJ \NasJ \NaeiJ

The variation of entropy with internal energy for this system predicted by the above relation
is shown in Figure 2.5.
It can be seen that at low energies, the entropy increases with increasing energy. At higher
energies, the trend reverses, however. At zero energy, the system is perfectly ordered with all
particles in the ground state. At the highest possible energy for the system, U = N ae\, the
system is also perfectly ordered with all particles in the higher energy level. The entropy is
zero for both of these perfectly ordered states. The entropy is maximum when the particles
are evenly divided between the two accessible energy levels. The temperature is the inverse
of the slope of the curve in Figure 2.5. At low energies the slope is positive and therefore
the temperature is positive. T increases as U increases at low energies, approaching +oo
60 2 / Statistical Treatment of Multiparticle Systems

{du\

T<0

Figure 2.5

as U -> (\/2)Nae\. At high energies, the slope and hence the temperature are negative. T
increases as U increases. As U decreases toward {\/2)NaS\, T approaches —oo. Thus, the
system has positive temperatures at low energies and negative temperatures at high energies,
and U = (l/2)Nas\ is a singular point that corresponds to T = ±oo.

As noted above, most common systems contain particles with an unlimited number
of increasing energy levels, and absolute temperatures for such systems are positive. If a
system with unbounded energy levels at a positive temperature is brought into contact with a
system at a negative temperature, the equilibrium temperature reached when the systems are
brought into contact must be positive because the composite system will have unbounded
energy levels.
For the system considered in Example 2.4, negative temperatures correspond to higher
system energies than positive temperatures. When a system of the type considered in the
example at a positive temperature is brought into contact with a similar system at a negative
temperature, energy will transfer from the system having a negative temperature to the sys-
tem having a positive temperature. This transfer increases the entropy of the composite
system, which must be maximized at equilibrium. Thus, negative absolute temperatures are
hotter than positive absolute temperatures. The scale of absolute temperature therefore runs
from +0 K to +oo K / - 0 0 K to - 0 K.
The two-energy-state system considered in Example 2.4 can be used as a model of
the nuclear spin energy storage in a crystal exposed to a magnetic field. In the presence
of a magnetic field, the lowest nuclear state may be split into a finite number of nuclear
magnetic states. For this model to apply, the energy storage in the nuclear spin states must be
decoupled from energy storage elsewhere in the crystal lattice. The nuclear spin storage must
reach equilibrium quickly and transfer of energy between nuclear spin storage and other
storage modes must be weak. When these conditions are met, the nuclear spin subsystem
behaves like as a finite-state system like that considered in Example 2.4 and may exhibit
negative temperatures. Abragram and Proctor [1] conducted experiments with LiF crystals
that meet these conditions. Their investigation indicated that the lithium and fluorine nuclei
2.61 Maxwell Statistics and Thermodynamic Properties for a Monatomic Gas 61

may act as two separate subsystems that exhibit different negative temperatures. Further
information on the thermodynamics of systems that exhibit negative absolute temperatures
can be found in the references by Ramsey [2], Klein [3], Abragam and Proctor [1], Proctor
[4],andMuschik[5].

2.6 Maxwell Statistics and Thermodynamic Properties for a

Monatomic Gas
Up to this point in this chapter we have talked in general terms about the energy
storage in a system of particles. In this section we want to consider a specific system type. We
will specifically consider a system containing two species of boson-type particles that store
energy only by translational kinetic energy. We will limit the model to conditions where
the system exhibits dilute occupancy. This is a good model of a monatomic ideal gas for
moderate temperatures where electronic and nuclear energy storage effects are negligible.
We begin with the Boltzmann distribution relations generated in Section 2.3:

^ = ^-= , (2.49a)
Na Za
%C B
' *1 , (2.49b)
Nh Zb
where

Za = YJ^~e'IH\ (2.50a)
/=0
oo
Zb = J2gje-£J/kBT. (2.50b)
y=o
For translational storage only, the degeneracy is given by Eq. (1.121) from the particle-in-
a-box quantum solution in Section 1.7:

The energy levels for translational storage become more closely spaced as the energy
increases. For translation of atom-sized particles at room temperature or above, the spacing
of the quantum energy levels is extremely close. For practical purposes, we can consider
energy to be a continuous variable. In this continuum limit, the summations in the relation
for the partition functions become integrals:
°° ^ r°° ir /8m V 2 / 3 \ 3 / 2
J2 * «
1=0
=E
alls
'Jo 4
(2.113a)

Zb= f^W £ ^ ri(*^\3/2 W

j=0 alle
(2.113b)
62 21 Statistical Treatment of Multiparticle Systems

Evaluating the integrals on the right side of the above relations yields

h2

(2.114b)

In the continuum limit, Naj, the number of a particles in energy level /, is replaced with
dNa, the differential number of particles in the interval from s to s + ds, and 8s in the
degeneracy relations becomes a differential ds. A similar transformation applies to the
distribution for b particles. The discrete distributions (2.49a,b) are thereby transformed into
the following continuous distributions:

(2.115a)
Na \4ZaJ V h2
3/2
2 (2.115b)
b \4ZbJ \h J
Substituting the partition function relations (2.114a,b) into (2.115a,b), we obtain

(2.116a)

<£ _ (4=) JLr^-u, an*)

Thus both species obey the same energy distribution at equilibrium. Note that the right side
of the above equations are the probabilities that a particle has energy between s and s + ds.
We can determine the mean energy of a particle as

elw
)= / e(— ) vle- de.

Evaluating the integral yields

(e) = lkBT. (2.117)

Thus the mean energy per particle is the same for species a and species b, regardless of
differences in particle mass.
If we denote the speed of the type a and type b particles as c, the energy of the species
a particles is related to their speed as

e = \mac2, (2.118a)

while for species b, the energy is given by

e = \mbc2. (2.118b)

The energy distributions can be converted to speed distributions by substituting the ap-
propriate energy relation (2.118a) or (2.118b) into Eqs. (2.116a,b). The resulting speed
2.61 Maxwell Statistics and Thermodynamic Properties for a Monatomic Gas 63

distributions for the two species are

(2.119a)

(2.119b)
Nh 27TkBT

Note that if the masses of the two particle types are different, they will have distinctly
different speed distributions, as dictated by the relations above. The above equations are
two embodiments of the Maxwell speed distribution, which was first derived as part of
the kinetic theory of gases developed by James Clerk Maxwell in the 1860s. This relation
emerges as a special case of the Boltzmann distribution.

Example 2.5 Derive a relation for the most probable speed of an atom in a monatomic
gas. Use the results to determine the most probable speed for helium atoms at 290 K.

Solution Equation (2.115a) can be interpreted as the probability that a randomly

selected atom in a monatomic gas has a speed between c and c + dc:
3/2
c2e~"

The right side of the above relation exhibits a maximum at a particular c value. This
most probable value cmv,a is that which maximizes dNa/dc. Solving the above relation for
dNa/dc, differentiating, and setting the result equal to zero yields

ma
This relation indicates that the location of the peak in the distribution curve shifts to higher
speeds as the temperature increases. This shift is indicated graphically in Figure 2.6.
For helium (m = 6.64 x 10~27 kg) at 290 K, we therefore have

2(1.38 x 1Q- 23 )(29O)\ 1/2

= 1095 m/s.
m 6.67 x 10~27 )

dNa
dc

increasing T

Figure 2.6
64 2 / Statistical Treatment of Multiparticle Systems

The same line of analysis applied to species b yields the relation

2kBT\l/2

mb J
It is clear from the relations for cmPiO and cmptb that the most probable speed will differ if the
mass of the two particle types is different. Thus, in a mixture of neon (m = 3.35 x 10~26 kg)
and helium (m = 6.64 x 10~27 kg), the most probable speeds for the two types of atoms
will differ by more than a factor of 2.

Having evaluated the partition function, we are now in a position to evaluate other
thermodynamic properties for the system we are considering. In Section 2.3 we obtained
the following relation for the system internal energy:

(2.52)
ai
Na,Nh,V ^b \ / Na,Nh,V

Substituting the partition function relations (2.109a,b) into the above equation, and rear-
ranging somewhat, the relation for internal energy becomes

U = \{Na+Nb)kBT. (2.120)

We also derived the following relation for the entropy of the system in Section 2.4:

S U
(2.78)
kB kBT
Substituting Eq. (2.120) and the partition function relations (2.114a,b) into the above equa-
tion yields, after a little rearranging,
S 5
SAT , XM , 3M i \2nmaV^kBT\ 3 1 / 2nmbV2/3kBT \
a b) T aLLL 2 b L L L 2
kB~2^ "" 2 \ h2N aP j ^ 2 \ h N2b/3 V
(2.121)

We now solve Eq. (2.120) for T and substitute into Eq. (2.121) to obtain

S _
— = -(N +Nb) + -A^aln<^ 2 ^ > +-^ln<^ ^ >.
^ 22 22 {3h2NZ
\3h Na'
/3 (Na + Nh)j
(NN)f -2 \u«-«
3h2N2bt/3(Na + Nb) J
(2.122)

Note that this equation is of the form S = S(U, V, Na, Nb), and it is therefore a fundamental
equation for the system. For a system with fixed Na and Nb, the form can be made a bit
cleaner if we define a reference state as corresponding to specific volume and temperature
values VQ and To. It follows that the internal energy and entropy at these reference conditions
are given by

Uo = l(Na+Nb)kBTo, (2.123)
2.61 Maxwell Statistics and Thermodynamic Properties for a Monatomic Gas 65

Nb) + -Na In <^ TTr- > + -Nb In

2 \3h2N^\Na+Nb)) 2 ;}•
(2.124)
The fundamental equation can then be cast in terms of the reference state values as

^.|W.+Wh((^.)W(£)}. (2.25,
kB 2 {\VoJ \UoJ)
This relation is plotted in Figure 2.7.
In engineering analysis, generally only the change of system entropy is of interest. For
such cases, the reference values for S, V, and U cancel from the calculations when using the
above equation. Hence the reference state is arbitrary and can be chosen for convenience.
Regardless of the choice, the change in entropy between two states 1 and 2 is given by
2/3
u2
,Vi
2/3
7V (2.126)
vj
One additional interesting result can be obtained from this analysis. Solving the funda-
mental Eq. (2.125) for U yields the relation
Zli
V 2/3
r, rr ,(V\-
Y f 2{t
2(S-S 0 ) (2.127)
U = Uo[ — ) exp
Also, by definition, we know that

(2.101)

Evaluating the derivative above using Eq. (2.127) yields

_ (dU\ _2 V~5/3 f( 2(5-
O (C C \ ^
P = L\b — oo)
< > ,
[ 3(Na + Nb)fCB J

2
v/vn

Figure 2.7
66 2 / Statistical Treatment of Multiparticle Systems

initial state final state

\ subsystem 1 subsystem 2
$ V\.TitNa S Vi+V2,Na,Nh ^

Figure 2.8

which, using Eq. (2.127), can be rewritten as

? = -(- (2.128)

3\V

Finally, combining Eqs. (2.120) and (2.128) to eliminate U, we obtain

?L9 (2.129)

which the reader may recognize as the ideal gas equation of state. Thus, our model predicts
that this system of boson particles that store energy only by translational motion will obey
the ideal gas equation of state.

Example 2.6 Initially a composite system consists of two subsystems as shown in

Figure 2.8. One has volume V\ and contains Na particles of species a and the other has
volume V2 and has Nb particle of species b. Initially the two subsystems are in equilibrium
at the same temperature Tt. The partition separating the subsystems is removed, allowing
the particles to mix. Find the change in temperature and entropy for the composite system
if these particles behave as bosons that store energy by translation alone. You may assume
that dilute occupancy applies.

Solution Using Eq. (2.120), we have initially

Ux = lNakBTi, U2 = l
and therefore

In the final mixed state

Since the composite system is isolated, there can be no change in the total internal energy,
implying that Uf = U[. Equating the above relations for U\ and Uf and rearranging, we
obtain

Tf-T[ = 0.
Exercises 67

Thus, there is no change in temperature for the system. Using Eq. (2.121), we can obtain
the following relations for the initial entropy of subsystems 1 and 2:

Si 5,, , 3 A 7 ,_ (2nmbV22/'

Additivity requires that

5i = Si + S2
and for the final state Eq. (2.121) applies:
Sf 5 _ 3 f
N
2H

The change in entropy of the composite systems is therefore given by

Sf — Si = Sf — S\ — 52.

Substituting the relations for Si, 52, and Sf into the above equation and simplifying yields
the following relation for the change in entropy:

Note that this change of entropy due to the mixing of two pure gases is always positive
since V\ + V2 is always larger than either V\ or V2. This increase in entropy due to mixing
alone is sometimes referred to as the entropy of mixing.

The statistical analysis of systems in a microcanonical ensemble presented in this chapter

has led us to several important conclusions regarding macroscopic properties and equilib-
rium conditions in such systems. The interested reader can find further discussion of these
topics in the references [6]-[9] listed at the end of this chapter. In the next chapter we will
more fully develop a framework for macroscopic thermodynamic analysis of systems.

Exercises
2.1 Determine the possible distributions among energy states of three simple harmonic os-
cillators such that their total energy is 13hv/2. All three harmonic oscillators have the
same characteristics frequency v. Is the set of occupancy numbers {1, 0, 1, 1, 0, 0,...} an
acceptable distribution?
2.2 Determine the number of ways of distributing six distinguishable atoms into four energy
levels such that one atom is in the first level, one is in the second, three atoms are in the
68 21 Statistical Treatment of Multiparticle Systems

third, and one is in the fourth level. (Note that the order within an energy level does not
matter.) Repeat the calculation for indistinguishable atoms.
2.3 A system contains five indistinguishable bosons of a single species that can occupy one of
five quantum energy levels that all have the same degeneracy. Determine and plot the total
number of microstates for this type of system for energy level degeneracy values of 1, 10,
and 100 and plot the number of microstates as a function of the degeneracy. (Hint: You may
want to work in terms of the log of the number of microstates.)
2.4 A system contains five indistinguishable fermions of a single species that can occupy one
of four quantum energy levels that all have the same degeneracy. Determine and plot the
total number of microstates for this type of system for energy level degeneracy values of 2,
5,10, and 50 and plot the number of microstates as a function of the degeneracy.
2.5 A system contains two indistinguishable bosons of species a and two indistinguishable
bosons of species b. For both species, each particle can occupy one of three quantum
energy levels. All the energy levels for both species have the same degeneracy. Determine
and plot the total number of microstates for this type of system for energy level degeneracy
values of 1, 10, and 50 and plot the number of microstates as a function of the degeneracy.
(Hint: You may want to work in terms of the log of the number of microstates.)
2.6 A system of two energy levels with energies £o and S\ is populated by N particles at
temperature T. The degeneracy of both levels is one. The particles populate the microstates
according to Boltzmann statistics.
(a) Derive an expression for the average energy per particle.
(b) Determine the limiting behavior of the average energy per particle as T —> 0 and

2.7 A system contains boson particles with two internal energy levels: a ground state of degen-
eracy g\ and a low-level excited state of degeneracy gi and energy As above the ground
state. Dilute occupancy is a good approximation for this system. Derive a relation for the
average energy per particle in the system in terms of g\, g2, As, and temperature T.
2.8 Gibbs theorem states that the entropy for a binary mixture ideal gas is the same as the sum
of the entropies each gas species would have if it occupied the total system volume V alone
at temperature T. Use the results of Section 2.6 to show that this is true for a mixture of
two ideal monatomic gases.
2.9 The standard deviation of an energy distribution
dNa
= fa(s)ds
Na
is given by
ae = [(s2) - < £ ) 2 ] 1/2 ,
where the angle brackets denote the weighted average of the enclosed quantity,

(Y(s))= /
-L
Jo
y(s)fa(s)ds.

Derive a relation for oe for the translational energy of an atom in a monatomic gas at
temperature T. What happens to oe at low temperatures (71 -> 0)?
2.10 For a molecular speed distribution
References 69

initial state final state

\ subsystem 1 subsystem 2 subsystem 3 "

; v 1 + v 2 +v 3 $
^ VltTi,Na
5 Na,Nb,Nc ^

Figure 2.9

xxxxxxxxxx

Nxwxwxwwww
Figure 2.10

the root mean square (rms) speed CnnS>fl is given by

1/2

Derive a relation for cms,fl for a monatomic gas at temperature T. Use the result to evaluate
Crms,a for helium atoms at 290 K and compare cm^a to the most probable speed.
2.11 Initially a composite system consists of the three subsystems with different volumes as
shown in Figure 2.9. Initially all three subsystems are in equilibrium with the same temper-
ature T[. Each subsystems contains a different monatomic gas species a, b, or c. The two
partitions separating the systems are removed, allowing the different species to mix. Find
the change in temperature and entropy of the composite system between the initial config-
uration and the final state. (Hint: Extend relations obtained in the two-subsystem problem
discussed in Example 2.6 to a composite system having three subsystems.)
2.12 The cylinder shown in Figure 2.10 initially contains neon gas at 200 kPa and 300 K. The gas
expands from its initial volume of 0.1 m3 in a constant-entropy process until its volume is
0.2 m3. Using results from Section 2.6, find the change in system internal energy for the
process and the final system temperature.
2.13 For a monatomic gas that undergoes a constant entropy process from state 1 to state 2, use
results of Section 2.6 to show that
p T/V3 n T/5/3

2.14 For a monatomic gas that undergoes a constant-entropy process from state 1 to state 2, use
results of Section 2.6 to show that
5/3 r 5/3

Px

References
[1] Abragram, A. and Proctor, W. G., "Experiments on Spin Temperature," Physical Review, 106:
160, 1957.
[2] Ramsey, N. R, "Thermodynamics and Statistical Mechanics at Negative Absolute
Temperature," Physical Review, 103: 20-8, 1956.
70 21 Statistical Treatment of Multiparticle Systems

[3] Klein, M. J., "Negative Absolute Temperatures," Physical Review, 104: 589, 1956.
[4] Proctor, W. G., "Negative Absolute Temperatures," Scientific American, 239(2): 78-85, 1978.
[5] Muschik, W., "Thermodynamical Algebra, Second Law and Clausius' Inequality at Negative
Absolute Temperatures," Journal of Nonequilibrium Thermodynamics, 14, 173-198, 1989.
[6] Callen, H. B., Thermodynamics and an Introduction to Thermostatistics, 2nd ed., John Wiley &
Sons, New York, 1985.
[7] Kittel, C. and Kroemer, H., Thermal Physics, 2nd ed., W. H. Freeman and Company, New York,
1980.
[8] Robertson, H. S., Statistical Thermophysics, Prentice-Hall, Englewood Cliffs, NJ, 1993.
[9] Tien, C. L. and Lienhard, J. H., Statistical Thermodynamics, Hemisphere Publishing
Corporation, New York, 1979.
 CHAPTER 3

A Macroscopic Framework

The basic elements of statistical thermodynamics were developed in Chapter 2.

In this chapter, we digress briefly from development of the statistical theory to expand the
theoretical framework using mathematical tools and macroscopic analysis. By doing so we
more strongly link the statistical theory to classical thermodynamics and set the stage for
alternative statistical viewpoints considered in Chapter 4.

3.1 Necessary Conditions for Thermodynamic Equilibrium

In the previous chapter, we have derived several important pieces of information
about thermodynamic systems. The goal of this chapter is to expand the framework of
macroscopic thermodynamic theory so that it can be applied effectively to a variety of system
types. We will begin by summarizing the important ideas developed in the last chapter.
So far, we have taken the volume V, internal energy U, and particle numbers N a and
Nt, to be intrinsic properties for any system we may consider. We subsequently defined the
properties entropy S, temperature T, pressure P, and chemical potentials \x a and /x^. Our
analysis of the statistical characteristics of thermodynamic systems has led to the conclusion
that for a system with fixed U, V, Na, and Nb, equilibrium corresponds to a maximum value
of the system entropy. This is referred to as the entropy maximum principle. The entropy of
a composite system with an arbitrary number of subsystems is additive over the constituent
subsystems. This is the additivity property of entropy. The value of entropy at equilibrium
is a function of U, V, Na, and Nb. This function is the fundamental relation for the system.
We further concluded that entropy is a monotonically increasing function of energy. Finally,
we considered two subsystems I and II separated by a boundary that is free to move and
permits exchange of heat, species a, and species b. For such a system, we showed that the
necessary conditions for equilibrium are:

Observation of the physical environment around us reveals that matter exists in different
forms that we call phases. A fluid phase is characterized by the free motion of molecules,
atoms, or other particles within the system boundaries. In general, fluids are said to be gases
or vapors when their density is low whereas at high density they are categorized as liquids.
When the particles become bound together in such a way that their motion is limited to a
very small region of space, the matter in the system is acknowledged to be in a solid phase.
Because liquids and solids are characterized by high density, they are sometimes jointly
referred to as condensed phases. The surface that defines the boundary between regions
containing different phases is usually termed an interface.
We note that the boundary between the subsystems in Figure 2.4 has characteristics iden-
tical to those at the boundary between two different phases in a multiphase system. Specif-
ically, this boundary is permeable to all species in the system, it permits heat interactions

71
72 3 IA Macroscopic Framework

r fixed porous plate that is

adiabatic and permiable to
species a but not to species b

Figure 3.1

between the phases, and the boundary can move in response to pressure changes on either
side. The conditions specified in Eq. (3.1) are therefore the necessary conditions for equi-
librium in a system containing two species and two phases with a flat interface between
the phases. (The effects of interface curvature will be considered in a later chapter.) These
equilibrium conditions are sometimes referred to as the conditions for phase equilibrium.
The necessary equilibrium conditions specified in Eq. (3.1) are specific to a system in
which the boundary between the subsystems has the following characteristics:

(1) It is permeable to both species a and species b particles.

(2) It permits heat exchange between the subsystems through the boundary.
(3) The boundary can move in response to pressure changes within the subsystems.

It is important to note that the necessary equilibrium conditions will change if any of these
characteristics are altered. As an example, we will consider the system shown in Figure 3.1
in which the boundary between the subsystems has the following characteristics:
(1) It is permeable to species a but not to species b particles.
(2) It permits no heat exchange between the subsystems through the boundary (the
solid material in the porous wall is adiabatic).
(3) The boundary cannot move in response to pressure changes within the subsystems.
The first step of thermodynamic analysis for a system like that in Figure 3.1 must be to
determine the necessary conditions for equilibrium. The analysis required to determine
the necessary conditions for equilibrium begins with the fundamental relation for each
subsystem:

Sl = Sl(U\ V\ Nla, Nlh), (3.2a)

S11 = Su (Ul\ Vl\ W°, Nll). (3.2b)

Expanding each of these relations to first order in each variable and using the definitions of
7, P, \ia, and fib, we obtain

± (3.3a)

1 Pn ;/ n
11 n
S - —HTf -L —rlV - i-2- 11 -jkdN". (3.3b)
> - dU + dV Tll
These relations dictate the small change in subsystem entropy that results from small changes
3.11 Necessary Conditions for Thermodynamic Equilibrium 73

in U, V, Na, and A^. Since at equilibrium

dS = dSl + dSu = 0 (3.4)
it follows that, at equilibrium conditions, small changes in U, V, Na, and Nb in the subsys-
tems that result from microscopic fluctuations must satisfy

Ijf/ 1 + £dV* - &dNl - £dNl + ±dU* + ^dV* - £dN? - &dN? = 0.

(3.5)
The analysis to this point is essentially identical to that in Section 2.5 for the specific systems
shown in Figure 2.4. In fact, the above relation applies to any composite system with two
subsystems and two species. The specific constraints associated with the boundary wall
between the subsystems are invoked in the next step of the analysis. The nature of these
constraints will differ for different boundary conditions.
For the system considered here, the location of the permeable wall separating the sub-
systems is fixed. It follows that the volumes of the subsystems are fixed and hence
dVl = dVu = 0. (3.6)
Also, since the porous boundary between the subsystems is impermeable to species b, N^
is fixed and Nl = 0. It follows that
dN\ = dNf = 0. (3.7)
Because the total number of species a particles in the system is fixed, if afluctuationresults
in a transfer of such particles between the subsystems, it must be true that
dN] = -dN?. (3.8)
Thefinalconstraint on the system is associated with energy transfer between the subsystems.
The solid structure of the wall between the subsystems is specified as being adiabatic. It
might appear at first glance that this implies that fluctuations of internal energy in the
subsystems are not possible. This conclusion is not correct, however. Because species are
exchanged between the subsystems, the internal energy of the subsystems can fluctuate.
Nevertheless, conservation of energy does require that
dUl = -dUu. (3.9)
Combining Eqs. (3.5)-(3.9), we obtain

Since this must be satisfied for arbitrary dUl and dN\ at equilibrium, it follows that the
necessary conditions for equilibrium are

Tl = Tl\ Ui\=Hll (3.11)

Note that, for this system, there are no requirements involving the pressure or the chemical
potential for species b. These properties may take on whatever values are necessary to
satisfy the specified values of Nb and V in the subsystems.
74 3 IA Macroscopic Framework

subsystem porous piston of mass mp

/ and cross sectional area Ap
/ that is permiable to species a
and species b
X3D
UJ_O_OLJ\
S

Figure 3.2

For an arbitrary system, determination of the necessary conditions for equilibrium re-
quires simultaneous consideration of the dS = 0 relation and the constraints on the system.
Equations (2.93) and (3.5) are examples of appropriate forms of the dS = 0 relation for
this type of analysis. It should be noted, however, that sometimes a more complicated form
is necessary to fully account for all the physics of the system. This circumstance is illus-
trated in Example 3.1. In the next section we examine equations of state that can be used
together with the necessary conditions for equilibrium and other constraints to determine
the equilibrium state of a composite system like that in Figure 3.1.

Example 3.1 Two parts of a completely isolated system are separated by a porous piston
of mass mv as in Figure 3.2. (Completely isolated implies no heat or work interactions with
the surroundings are permitted.) The space on either side of the piston contains a binary
gas mixture of species a and species b. The piston, which is free to move in the vertical
direction, has a cross section Ap and is permeable to both species a and species b. Determine
the necessary conditions for equilibrium.

Solution For a system with two subsystems and two species, the following dS = 0
relation must apply at equilibrium to satisfy the entropy maximum principle:

Na,Nb,V u,Na,Nh

^r) dNl + dUn + ^ dVn

dNb Na,U,V a,Nb,V

dNl1 = 0.
SNb

Because both the total volume and the total quantity of each species are fixed, any random
perturbation must satisfy

dVl = -dVl\ dNla = -dN1^ dN\ = -dNf.

Using these relations and the definitions of P, \ia, and /x^, we can manipulate the above
3.2 I The Fundamental Equation and Equations of State 75

dS = 0 relation into the form

-V

Energy in this system can be stored as internal energy in the gas in each subsystem and
as potential energy associated with the vertical position of the piston. For any energy
perturbation of the subsystems, conservation of energy requires

where ge is the gravitational acceleration at the earth's surface. Combining this relation and
the definition of T with the relation above yields

Since this must hold for arbitrary perturbations, each of the prefactors of the differentials
must be zero, which implies the following necessary conditions for equilibrium:

3.2 The Fundamental Equation and Equations of State

The results obtained in the previous chapter implicitly embody the fact that macro-
scopically distinguishable states exist that are characterized completely by the internal en-
ergy U, volume V, and the number of particles Na and Nt,. In this chapter we will consider
the more general case of a system containing r species of particles. The extension of results
from the two-species case considered in the previous chapter to the case of an arbitrary
number of components is generally quite transparent. In the general case, the fundamental
equation has the generic form
S = S(£/, V , W i , N 2 , . . . , N r ) . (3.12)
Note that conservation of mass, energy, and physical space (for nonrelativistic systems)
dictates that additivity applies to the particle numbers, internal energy, and volume, as
well as entropy. In examining thermodynamic properties of a system at equilibrium, those
whose values are proportional to the number of particles in the system are termed extensive
properties, while those whose values are independent of the mass in the system are said
to be intensive properties. For additivity to apply to entropy, S must be a homogeneous
first-order function of the extensive variables U, V, N\, N2,..., Nr. This implies that for
an arbitrary constant f
, V,NUN2,..., Nr). (3.13)
The fact that S is a monotonically increasing function of U ensures that the absolute
76 3 IA Macroscopic Framework

temperature T is positive. If we postulate that S is continuous and differentiable in addition

to being a monotonic function of U, we can, in principle, invert the fundamental relation to
get U as a function of S, V, N\, N2, . . . , Nr •Thus
U= U(S,V,NuN2,...,Nr),
S= S(U,V,NuN2,...,Nr)
are equivalent forms of the fundamental relation. Because the fundamental relation is ho-
mogeneous and of first-order, it follows that for a system containing a single species of
particle
S(U, V, N) = (N/NA)S(U/(N/NA), V/(N/NA), NA),
where NA is Avogadro's number. For such a system, if we define molar specific properties
as
u = U/(N/NA), (3.14)
v = V/(N/NA), (3.15)
s = S/(N/NA), (3.16)
then the fundamental relation can be written in the form
s = s(u, v, NA) = s(u, v). (3.17)
By convention, the basic principles of thermodynamics are organized in terms of three
laws of thermodynamics. The first law of thermodynamics embodies basic conservation
of energy. If we limit our considerations to simple systems that are macroscopically ho-
mogeneous, isotropic, and not subject to the effects of electric charge, chemical reactions,
electromagnetic fields, or surface effects, then conservation of energy requires that changes
in internal energy of a system must equal the net energy exchange across the boundary
of the system. Generally we categorize system energy exchanges as either being work or
heat. In differential form, the first law of thermodynamics for a simple closed system can
therefore be stated as
dU =8Q-8W, (3.18)
where 8Q and 8W represent differential quantities of energy crossing the system boundary
as heat and work, respectively. Here, heat (8Q) is taken to be positive when it transfers
energy into the system and work (8W) is taken to be positive when it transfers energy out
of the system. Work and heat are not properties and hence are not exact differentials. From
experience, we know that it is possible to make the system walls from materials that poorly
transmit heat. In the limit of zero heat exchange with the system, the system walls are said to
be adiabatic. If the walls of the system are virtually adiabatic, changes in U between states
can be experimentally determined by measuring the work required to achieve the change
of state.
The second law of thermodynamics embodies the extremum principle. There are several
ways of stating the second law, but for our purposes we can state it as follows:
• For a closed isolated system, the values of theextensive properties established at
equilibrium are those that correspond to the macrostate having the maximum S of
all macrostates accessible to the system.
3.2 I The Fundamental Equation and Equations of State 11

We will develop other equivalent ways of stating the second law in later sections of this
text.
The third law of thermodynamics states that the entropy of a pure crystalline solid system
vanishes as the absolute temperature approaches zero:

) . (3.19)
T->0 / pure solid crystal

The full justification for the third law is best understood from the perspective of quantum
statistical mechanics. We will return to discuss this point when we consider the thermo-
dynamics of crystalline solids in Chapter 7. For the present we will simply adopt this as a
plausible means of setting the zero point of entropy for pure crystalline solids.
We now will further explore the relationships among thermodynamic properties that
evolve from the fundamental equation. Expanding the fundamental relation

U = U(S, V, Nu N2,..., Nr) (3.20)

in a Taylor series and retaining only first-order terms, we obtain

fdU\

Using generalized versions of the definitions of T, P, and the chemical potential in Eq. (2.47),
(2.97), and (2.102) in the Chapter 2,

) , (3.22)
'V,NuN2,...,Nr

y) , (3.23)
/S,NuN2,...,Nr

(3.24)

Eq. (3.21) can be written

r
dU = TdS- PdV + 5Z Mi dNt. (3.25)

The above equation for dU applies to changes between two equilibrium states and hence
we can integrate it only along a path composed of a sequence of equilibrium states. Such a
path is termed a quasiequilibrium path. A process that follows a quasiequilibrium path is
referred to as a quasistaticprocess. The properties T,P, and M; are all intensive properties.
This is demonstrated as follows. Consider a system consisting of f identical subsystems.
By definition, the temperature of the composite system is given by

" " ~,SV,SNu...,SNr). (3.26)

In the above relation S, V, N\,..., Nr are the properties of the identical subsystems and,
by additivity, £S equals the entropy for the composite system. Since U is also additive over
78 3 IA Macroscopic Framework

the subsystems,
£ / ( ? S, ? V , ? # , , . . . , S N r ) = CU(S, V,Nu.-.,Nr) (3.27)
and therefore

Tcomp = J l ^ $ U ( S , V,NU..., Nr). (3.28)

The factors of f cancel to yield

^ , N l , . . . , N r ) . (3.29)

The right side is just the definition of the temperature for a subsystem. Thus the temperature
in each subsystem is equal to that for the system as a whole. Temperature is therefore
independent of the quantity of mass in the system, and it is therefore an intensive property.
It is left as an exercise to show that the same line of analysis demonstrates that P and /x,
are also intensive. For a system containing a single species of particle, the definitions of the
intensive properties P and T can be cast in the form

T = (— . (3.31)

For a closed system of fixed composition, no changes in particle numbers can occur, and
for any changes in state dNi = 0. Equation (3.25) then reduces to
dU = T dS-PdV. (3.32)
Since by definition P, T, and /x/ are partial derivatives of a function of S,V, N\,..., N r,
they themselves are functions of these variables:
T = T(S,V,Nu...,Nr), (3.33)
P = P(S,V,Nu...,Nr), (3.34)
M =in(S, V,Nu...,Nr). (3.35)
Relations of this type, expressing intensive properties as functions of independent extensive
properties, are called equations of state. Knowledge of an equation of state does not imply
knowledge of all properties of a system. Since a fundamental equation is homogeneous
and of first-order, it follows that equations of state are homogeneous and of zeroth order.
Consequently, multiplication of each independent extensive property by a scalar constant
f leaves the function unchanged. Thus, for example

, f V, f N i , . . . , SNr) = T(S, V,NU..., Nr). (3.36)

The equation of state for a pure ideal gas,
PV
T = —, (3.37)
7V£
is homogeneous and of zeroth order since multiplication of both V and iV by the same
3.2 I The Fundamental Equation and Equations of State 79

constant leaves the resulting value of T unchanged. It can also be written as

Pv
T = T, (3.38)

where R — ./VA^B is the universal gas constant.

As an equation of state for real gases, the ideal gas equation of state is inadequate in two
regards. First, the equation of state implies that v = RT/P. But if P -> oo or T —> 0,
0 cannot go to zero as the ideal gas relation predicts because the molecules themselves
occupy some space. We can patch up the formula by introducing a constant b v representing
this residual volume per molecule. It follows that

v = RT/P +NAby, (3.39)

which can be rearranged to give

P = ~, rr-r-- (3-40)
v - NA by
The second inadequacy is that attractive forces in a real gas reduce the pressure exerted
by gas on walls of the containment below that which would exist in the absence of such
forces. Attractive forces are often proportional to r~ 6 , where r is the separation distance
between molecules. In a gas the mean separation distance is proportional to v~ l/3 and we
therefore expect that the reduction in pressure due to such forces will be proportional to
v~2. This suggests that we account for this effect with a relation of the form

v — NAbv vl

where av is a constant. Equation (3.41) is van der Waal's equation, a prototype equation of
state for liquids as well as dense gases. Different constants must be used for different gases
as the size and attractive forces for each molecular species will be different. Although it
is not quantitatively exact over wide ranges for some gases, van der Waal's qualitatively
represents the behavior of the liquid and gas phases of many pure substances.

Example 3.2 For the monatomic gas model (indistinguishable bosons with dilute occu-
pancy) discussed in Section 2.6, derive the equations of state for the chemical potential.

Solution The entropy of the gas is given by Eq. (2.122) as

S 5 3 f 47tmaV2/3U } 3 f 47tmhV2/3U
— = -(N a + Nb) + -Na In { ^ } + -Nb In < ^ f

Since by definition fia = (dU/dNa)v,s,Nh> w e could invert this relation to get U as a func-
tion of the other variables and differentiate to determine \x a. However, a somewhat easier
alternative is to note that for a function y of x and z we have

dy
dzjy \dy
80 3 IA Macroscopic Framework

and therefore we can write that

ds \ /ds\ f du
a JV,UM
But by definition thefirstpartial derivative on the right side above is just \/T and the second
is \xa, and so

Differentiating the entropy relation above with respect to Na and setting the resulting
equation to —fi a/T and rearranging, we obtain

^ T ) l n ( — 4 7 t ™2 a V/3 U
—1.
2 ) [3h N5 (Na+Nb))
Since the internal energy of the gas is given by

we can solve this relation for T and substitute into the relation for \x a. This yields the
following equation of state:
U
.3h2Na' (Na+Nb).
The identical reasoning can be applied to species b to obtain
U \ f 4jtmhV2/3U 1
y

3.3 The Euler Equation and the Gibbs-Duhem Equation

The fundamental relation in terms of U is homogenous and of first order, implying
that
, V,NU...9 Nr). (3.42)
Differentiating both sides of this relation with respect to £ and applying the chain rule, we
obtain

, J L W + V —f^U[/ (343)
which reduces to
du du ^A at/
S+ V+ V Nt = U. (3.44)
Since Eq. (3.44) must be valid for any choice of £, we take f = 1:
3.3 I The Euler Equation and the Gibbs-Duhem Equation 81

Using relations (3.22)-(3.24), the partial derivatives can be replaced with the corresponding
intensive properties, converting the Eq. (3.45) to
r

U = TS-PV + ]T/x/ N/ . (3.46)

Equation (3.46) is referred to as the Euler equation. Differentiating both sides of Eq. (3.46)
yields
r r

dU = TdS + SdT - P dV -VdP + ]T ^i dNt + ^ N( dm. (3.47)

Replacing dll in the above equation with the right side of Eq. (3.25) and eliminating terms
that cancel, we obtain the Gibbs-Duhem equation
r

SdT -V dP + Y^Ni d
^i = °- ( 3 - 48 )

For a single-component system this reduces to

SdT -VdP + Nd^ = 0. (3.49)
Equation (3.49) can also be written in terms of a molar chemical potential /x = NA/JL:

djl = -sdT + vdP, (3.50)

which is the commonly cited form of the single-component Gibbs-Duhem equation. For
a single-component system u = u(s, v) and it follows that T = T(s, 0), P = P(s, v),
[I = pi(s, v). We may use these three relations to eliminate s and 0, yielding one relation
among three variables T,P, and ft. For a binary system U/N2 = U(S/N2, V/N2, N\/N2),
from which we can derive four equations of state:
T = T(S/N2, V/N2, Ni/N2),
p= P(S/N2,V/N2,N{/N2),
fix =in(S/N2,V/N2,Ni/N2),
h,V/N2,Ni/N2).
These four relations can be used to eliminate three variables S/N 2, V/N2, and N\/N2,
thereby obtaining a relation among T, P, \i\, and \i2. If we add one more species, we get
one more variable to eliminate and one more equation of state to use to eliminate it. Hence,
in principle, it is always possible to get a functional relation among the intensive properties
T, P, and chemical potentials /x/. The Gibbs-Duhem equation is the differential form of
the relation among these properties.

Example 3.3 For a binary mixture of monatomic gases, derive the relation among T, P,
and the chemical potentials fia and / ^ .

Solution The internal energy of the gas is given by Eq. (2.120) as

U = UNa+Nb)kBT.
82 3 IA Macroscopic Framework

We can solve this relation for T and substitute into the relation for iia obtained in Exam-
ple 3.2. This yields the following relation:

The identical reasoning can be applied to species b to obtain

3 (f|

This system also obeys the ideal gas law

PV = (Na+Nb)kBT.
Rearranging, we can write this relation in the form

V / * B T \ (Na + Nb)/NA
Na V P ) Na/NA PXa'
where Xa is the mole fraction of species a, denned as
Na/NA
" (Na+Nb)/NA
Substituting the right side of the above equation for V/N a into the relation for fia, and
rearranging, we obtain

Applying the same analysis to species b yields

\+kBT\nXb.

Since Xb = 1 — Xa, the above relation for /u,/, can be written

This relation for /i^ can be combined with the above relation for [ia to eliminate Xa. Doing
so yields the following relation among T, P, fia, and \xb:

l - c x p | — - ( -
3.4 I Legendre Transforms and Thermodynamic Functions 83

It is also noteworthy that the relation

= B
~V2j "I Wp~W >+k*T lnXa

can be rearranged to the form

'5
a,ret

where Pa is the partial pressure of species a, defined as

p V P
r
a — ^a r i

Pa, ref is an arbitrary reference value for Pa, and

/ .3/2 f_2 \3/5

Corresponding relations (with the as changed to bs) apply to species b. This form of the
relation for fia illustrates that, for a monatomic ideal gas at constant temperature, there
is a direct functional relation between the partial pressure of the species and its chemical
potential.

3.4 Legendre Transforms and Thermodynamic Functions

Legendre Transforms
The original extremum principle that has emerged from our statistical analysis
indicates that the equilibrium state is that which maximizes S for a specified system total
internal energy. We now want to develop other formulations of the extremum principle. To
emphasize the mathematics, we generalize the fundamental relation
S= S(U,V,Nu...,Nr)
to the form

Y = y(xo,x1,x2,...,x,) (3-51)
and we define

We want to express the information contained in the fundamental relation in terms of

derivatives §*, which in our thermodynamic analysis are intensive properties. We shall first
consider the simplest case where Y is a function of one variable X:

The derivative § indicates the slope of the tangent lines to the curve Y = Y(X) as shown
in Figure 3.3.
84 3 IA Macroscopic Framework

Figure 3.3

The tangent line at each point is uniquely defined by its slope, £, and its intercept of
the Y axis, x//. The family of tangent lines defined by \j/(%) defines an envelope to the
original curve Y(X) and thus contains the same information. Both can be considered to be
representations of the fundamental relation.
How do we determine xj/{%) when given Y (X)? This type of manipulation is known as
a Legendre transformation. For the simple one-variable case, since
Y -xjr
(3.53)
X-0
it follows that
(3.54)

To execute the transform, we therefore proceed as follows:

(1) Differentiate Y(X)to get § (X).
(2) UseV = y ( X ) - £ ( X ) X t o g e t ^ ( X ) .
(3) Eliminate X to get f (§).
The function \j/(%) is termed the Legendre transform of F(X). The inverse process is also
of interest. Since \\r = Y — £X, differentiating yields
dis =dY -t;dX (3.55)

Using the fact that § = dY/dX it follows that d^r = - X di=9 or equivalently

(3-56)

So if we start with ^ (§), we can invert the transformation in the following steps:
(1) Differentiate ^r(?) to get X(£).
(2) Usey=?X(§)-^)togety(§).
(3) Eliminate § to get 7 (X).
The symmetry of the transform and its inverse can be seen in the comparison below:
Legendre Transform ofY(X)
Y = Y(X)
§ = dY/dX Eliminate X and Y to get
3.4 I Legendre Transforms and Thermodynamic Functions 85

slope Pj

Figure 3.4

Inverse Transform

* = W)
= -df/d% ) Eliminate § and \j/ to get Y(X)

Generalization of the transform to three variables Y = F(Xo, X\) requires that we

consider a surface as being defined by an envelope of tangent planes. A plane is defined by
the intercept of the plane on the Y axis, denoted as ifr, together with §o and £i, which are the
slopes of traces of the plane on the Xo-Y and X\-Y planes (see Figure 3.4). The function
^ = VK^o, £i) thus defines a family of planes that form an envelope about the surface.
We extrapolate the transform process to an arbitrary number of variables as follows. The
relation
Y = (3.57)
represents a hypersurface in a (t + 2)-dimensional space and

(3.58)

represents the partial slope of an associated hypersurface. The family of tangent hyperplanes
that defines the hypersurface is mathematically specified as

(3.59)
k=0
Taking the differential yields

(3.60)

Using the definition (3.58) of §* and a Taylor expansion of Y,

(3.61)
*=o
86 3 IA Macroscopic Framework

we can replace dY in Eq. (3.60) to obtain

t

d\jr = -^2,Xk d%k. (3.62)

k=0
A n d s i n c e a T a y l o r e x p a n s i o n o f ^ ( ? o > §i > • • • > & ) y i e l d s
(3 63)
(lrW -
k=o V 0 ^
it must be true that
-Xk = ^-. (3.64)

To execute a Legendre transformation, we use the following t + 3 equations:

y = y(xo,Xi,x2,...,x,), (3-57>
^ (3.58)

(3.59)
£=0

to eliminate the variables Y and Xo, Xi, X 2, . . . , Xt (t + 2 unknowns) and obtain ^ =

o> £i ,•••>&)• To execute the inverse transformation, we use the t + 3 equations
), (3.65)

,2,...,r), (3.64)

Y
= f+Y.^Xk (3 66)
-
to eliminate the variables i/r and §o9§i? §2, • ••,%t (t + 2 unknowns) and obtain 7 =
y(Xo,Xi,X2,...,X,).
We need not transform a function Y = F(Xo, Xi, X 2 , . . . , Xt) with respect to all vari-
ables Xo, Xi, X 2, . . . , Xf. We can treat some as fixed constants. This is termed a partial
transform. We denote the partial transform with respect to XQ, XI, X 2 , . . . , Xn on func-
tion Y = 7(X 0 , Xi, X2, . . . , Xt) as Y = Y[%0, §i, £2, • ••, §/i, Xw+ i, . . . , X f ]. To execute a
partial transform we use the n + 3 relations
y = y(xo,x1,x2,...,xr), (3.67)

-U- (* = 0 , l , 2 , . . . , / i ) , (3.68)

o, §i, & , . . . , §n, X n + i , . . . , X,] = y ( X 0 , . . . , X,) - ] £ k x * (3.69)

to eliminate the variables Y and Xo, Xi, X 2, . . . , Xn {n + 2 unknowns) and obtain Y =

y[§o> §i, §2. • • >• ?«. X n + i, . . . , X,]. To execute the inverse transformation, we use the n-\- 3
3.4 I Legendre Transforms and Thermodynamic Functions 87

equations

Y = Y[$0, £ i , h , •••,£„, Xn+U ..., X , ] , (3.70)

Xk = -3Y[So,...,Sn,Xn+l,...,Xt]/d$k (Jt = 0, 1,2 «), (3.71)
t

Y(X0,..., X,) = r [ § 0 , . ••, Xt] + y52&Xk (3.72)

£=0

to e l i m i n a t e t h e v a r i a b l e s y [§o, . . . , X , ] a n d §o, §1, §2, • ••>£«(w + 2 u n k n o w n s ) a n d o b t a i n

The Legendre transformation is also useful in formulating the mechanics of multibody

systems. As described in Chapter 1, the Lagrangian

L = L(vuv2, ...,vr,quq2, •. •, r)q

characterizes the dynamic state of a system in terms of generalized velocities i>/ and gener-
alized coordinates qi. Another formulation of dynamics characterizes the state of the system
in terms of the Hamiltonian

H = H ( p i , p 2 , ..., p r , q u q 2 , . . . , # r ) ,

which is a function of generalized momenta pi and coordinates qi. In Chapter 1, it was also
noted that
dL

It follows directly that the Lagrangian for the system can be converted to the Hamiltonian
by executing a partial Legendre transformation with respect to the generalized velocities.

Thermodynamic Functions
We will now explore the application of Legendre transforms to thermodynamic
relations. For the fundamental relation U = U(S,V, N\,..., Nr) the derivatives of U
with respect to the independent variables correspond to the intensive parameters T, —P,
and \i\, . . . , /x,-. Functions obtained by using a Legendre transform to replace one or more
extensive properties with an intensive parameter are called thermodynamic potentials.
The Helmholtz free energy denoted as F (or A) is obtained by the partial Legendre
transform replacing S by T:

Y = U = U(S,V,Nu...,Nr),

k=0
 3 IA Macroscopic Framework

As the only & is T and the only Xk is S, the above relation becomes
F = U -TS. (3.73)
Differentiating gives
dF =dU -TdS -S dT. (3.74)
Using the following relation obtained previously,
r
dU = TdS - PdV + ] T At, dNt, (3.25)
/=i

and substituting to eliminate dU, yields the following relation for dF:
r
dF = -SdT - PdV + ^2iiidNi. (3.75)

The enthalpy, denoted as //, is obtained by the transform replacing the volume with the
pressure:
Y = U= U(S,V,Nu...,Nr),

H = U -(-I
which simplifies to
H = U + PV. (3.76)
Differentiation yields
dH =dU + PdV + V dP. (3.77)
Combining Eq. (3.77) with Eq. (3.25), we obtain
r
dH = TdS + VdP + ]T/x/ dN(. (3.78)

The Gibbs function, denoted as G, is obtained by the transform replacing S by T and V

by P:
U = U(S,V,Nu...,Nr),
dU_ dU
L i J i

G = U -TS-(-l
The relation for G simplifies to
G = U -TS + PV. (3.79)
3.4 I Legendre Transforms and Thermodynamic Functions

Differentiating and using Eq. (3.25) to replace dU yields

r
dG = -SdT + VdP + ^ /x/ dNt. (3.80)
i=\

For a system containing a single particle type, we can obtain a function referred to as the
grand canonical potential, denoted as U[T, V, /x], by executing the transform that replaces
S by T and N by /x:
U = U(S,V,N),
dU dU
T (181)
= Jf' »=M>
U[T,V,ii] = U -TS-iiN.
If we replace S, V, and all the 7V/S with the corresponding intensive properties, we obtain
the complete Legendre transform of U, given by

U[T, P, /xi, . . . , /xr] = U - TS + PV - > /x/ AT/.

If we use the Euler equation (3.46) to evaluate U, the above equation reduces to
)
,Mi,...,Mr] = 0. (3.82)
Thus, the complete Legendre transform of U is identically zero.
Although we shall not do so here, it is possible to perform analogous transforms to the
entropy form of the fundamental relation S = S(U, V, N\,..., Nr). The resulting properties
are referred to as Massieu functions. Massieu functions provide an alternate, and equivalent,
means of formulating equilibrium thermodynamic theory.

Example 3.4 Prove that for a system containing a single species of particle, the molar
specific Gibbs function is equal to the chemical potential per mole of particles.

Solution For a system with only one species of particle, the Euler equation (3.46)
requires that

Solving this relation for /xfl, dividing by Na, and multiplying by Avogadro's number, we
obtain
U + PV-TS A ^ ^
^aNA = ^T7^ =u + Pv-Ts.
Na/NA
Since by definition G = U + PV — T S, the right side of the above equation is just the molar
specific Gibbs function g and the left side of the above equation is the chemical potential
per mole of particles, jla. Thus
90 3 IA Macroscopic Framework

Figure 3.5 S (U, V) for a single component system of fixed N.

3.5 Quasistatic and Reversible Processes

The fundamental equation of a simple system S = S(U, V, N\,..., Nr) or U =
U(S,V,N\,...,Nr) defines a surface in a thermodynamic configuration space. For the
simple case of a system containing N particles of a single species, N is fixed and S =
S(U, V) and the fundamental relation is a surface in three space similar to that shown in
Figure 3.5.
For a composite system consisting of two or more subsystems, a similar interpretation
applies, except that S is a function of more variables. Each point on the hypersurface repre-
sents an equilibrium state. An arbitrary curve drawn from state A to state B on the surface
represents a sequence of equilibrium states. Such a process could occur only if, at any
instant as it traversed from A to B, the system deviated from an equilibrium state by only
an infinitesimal amount. Processes that conform to this idealization are termed quasistatic
processes or quasiequilibrium processes. In most systems it is possible to identify a char-
acteristic relaxation time required to reestablish equilibrium when the system is perturbed
away from equilibrium by a small amount. In a quasistatic process, the relaxation time of a
system is short compared to the time over which the process occurs. The differentials dS,
dU, etc., considered previously represent changes along a quasistatic path.
Consider now a real system in equilibrium state A. Very often the release of an internal
constraint allows a system to access a new state B of higher entropy than the state it was
in at the instant the constraint was released. Since our extremum principle mandates that
the system entropy must be the highest possible at equilibrium, we expect that release of
the constraint will initiate a spontaneous process terminating in state B with higher entropy
than state A. Perhaps the simplest example of a situation of this type is the free expansion of
a gas into an evacuated subspace of a system when removal of a barrier allows the particles
to enter the previously empty space (see Figure 3.6).
For a system containing N particles of a pure monatomic gas, it is a simple matter to
show that the change in entropy from state A to state B is given by

SB - SA = NkB In (3.83)
VA
From the above relation, it is clear that if VB > VA then SB > SA- Obviously it is also possible
to similarly release an internal constraint on energy or number of particles.
3.5 I Quasistatic and Reversible Processes 91

initial state A final state B

Figure 3.6 Free expansion of a gas.

To reverse the process resulting from the removal of the constraint, the system would
have to spontaneously reduce its entropy, achieving an S value less than the original value
and hence less than its maximum. This would violate the extremum principle. Hence, the
transition from A to B resulting from the release of an internal constraint is an irreversible
process. The idealized limiting case in which the increase in S is vanishingly small can
be reversed without violating the extremum principle. It is therefore termed a reversible
process.
It is possible to reverse state from B to A if the system is coupled appropriately to
another system. However, to invert such a process and to decrease the entropy of a system,
a corresponding or larger increase in S must occur in some coupled system. Why is this so?
This must be true because if we consider the original system and the coupled system together
to be one system, the interaction that reduces S in the original system is the spontaneous
result of removing an impediment to such interactions. We have already argued that such
spontaneous processes occur because the system is undergoing a transition to a state of higher
entropy. Thus the entropy of the combined system must increase, or in the limit, remain
the same. This directly implies that if the entropy of the original system decreases by an
amount SB — SA, the entropy of the coupled system must increase by that amount or more.
An increase in S is an indication that the spontaneous process is irreversible, and an
irreversible process always results in an increase in S. Hence, for any real process in a
closed system
dS>0. (3.84)
For a closed simple compressible system with afixednumber of particles of each species,
Eq. (3.32) indicates that

dU = T dS-PdV. (3.32)

The first law requires that for differential changes of state

dU = 8Q -8W. (3.18)
Note that Eq. (3.32) is valid for changes along a quasiequilibrium path. It is therefore
restricted to quasistatic processes. If we combine Eqs. (3.32) and (3.18) to eliminate dU,
the following relation is obtained:

= 8Q-8W + PdV. (3.85)

For systems of fixed volume and no work interaction, it follows that

8Q = TdS. (3.86)
92 3 IA Macroscopic Framework

This relation is valid for quasisteady heat exchange at constant temperature for a system of
fixed volume.
By considering energy exchanges with a simple compressible system, we can deduce
several important consequences of the extremum principle. We first define a thermal reser-
voir as being a system enclosed by a rigid impermeable wall for which all heat transfer
interactions are quasistatic with
dU = 8Q = TresdSres (3.87)
and transfer of heat to or from the reservoir has a negligible effect on its temperature. We
now consider a quasistatic process in which heat is transferred from a thermal reservoir
to a system held at constant volume so that there is no work interaction. For the system,
Eq. (3.86) requires that
dSsys = 8Q/Tsys. (3.88)
By additivity, the change of entropy of the composite system composed of the original
system and the reservoir is given by

dScomp Sys = dSsys + dSYQS = — —. (3.89)

-* sys I res

Note that here we have adopted the convention that heat exchange is positive when it flows
into the system. For this spontaneous process within the composite system, Eq. (3.84)
implies that dScomp sys > 0. Combining this requirement with Eq. (3.89) it is clear that for
such a process to occur, we must have
Tres > Tsys. (3.90)
Thus heat exchange spontaneously occurs between two systems only if their temperatures
differ, and heat spontaneously flows from a high temperature system to a lower temperature
one. It also follows that reversible transfer of heat can only be achieved if the temperatures
of the two systems are the same.
Reversible heat transfer therefore requires that the transfer occur quasistatically and at
constant temperature. These are exactly the requirements for the validity of Eq. (3.86). We
can therefore state that

dS = — (3.91)

for a reversible heat addition process when 8W — P dV. Note that this requires perfect
conversion of the energy associated with the P dV effect to mechanical work, or vice versa,
depending on the direction of the volume change. Because of friction and other effects, such
conversion is never perfect. For work done on the system, 8W and P dV are both negative
and P dV is greater (less negative) than 8W owing to imperfect conversion. For work done
by the system, 8W and P dV are both positive and 8W is less than P dV owing to imperfect
conversion. In either case, mathematically we can state that for any work interaction
PdV-8W>0. (3.92)
In the above relation the equal sign applies to the limiting case of a quasistatic work
interaction with perfect conversion. Such a process would be reversible. The equal sign in
Eq. (3.92) therefore applies to a reversible work process.
3.61 Alternate Forms of the Extremum Principle 93

Solving Eq. (3.85) for P dV — 8W and substituting into the above inequality (3.92), we
obtain
TdS -8Q>0. (3.93)
Solving this relation for TdS yields the following relation, which applies to any real process:
8Q
dS > — (for any real process). (3.94)

In Eq. (3.94), the equal sign applies in the limit of a reversible process. Note that if a process
is reversible and adiabatic, Eq. (3.94) implies that dS = 0. Thus, a reversible and adiabatic
process for a simple compressible system offixedcomposition is a constant-entropy process.
Such processes are also said to be isentropic.
This line of analysis naturally leads to a discussion of heat engines, heat pumps, and
the Carnot cycle. The interested reader may wish to consult the references by Howell
and Buckius [1] and Modell and Reid [2] for further discussion of these topics. At this
point, however, we will explore other aspects of the theoretical framework of macroscopic
thermodynamic s.

3.6 Alternate Forms of the Extremum Principle

The extremum principle on which our development of macroscopic theory has
been based has thus far been stated in terms of entropy. This statement of the princi-
ple arose from our consideration of a system held at fixed volume and energy. We can,
however, obtain other, completely equivalent statements of the extremum principle by con-
sidering other system constraints. All these statements are equivalent forms of the second
law of thermodynamics. The entropy maximum form of the second law can be succinctly
stated as
• For a system held at constantV, U, and N(, at equilibrium, unconstrained param-
eters take on values that maximize the entropy of the system.
A more mathematical statement is
dS = 0, d2S < 0 (at equilibrium for a system at fixed V, U, and Nt). (3.95)
We now consider the more general case of an isolated system offixedvolume containing
a pure substance. For any spontaneous process, the first and second laws in the entropy
maximum formulation require that
dU = 8Q-8W, (3.96)
80
dS>^. (3.97)

Combining these relations yields

TdS >dU + 8W. (3.98)
If the only possible work mode is pressure-volume work, then 8W = P dV. Since V is
fixed, 8W = 0 and upon rearranging (3.98) becomes
dU<TdS. (3.99)
94 3 IA Macroscopic Framework

Now rather than imposing a condition of fixed energy, we allow heat exchange so as to keep
S fixed. It follows that dS = 0 and Eq. (3.99) reduces to
dU < 0. (3.100)
Thus, for a system with fixed V and 5, spontaneous processes always reduce the system
energy. When the system reaches a condition of minimum U, spontaneous processes will
produce zero change in U. The equilibrium condition thus corresponds to minimum U for
a system at constant V and S:
dU = 0, d2U > 0 (at equilibrium for a system at fixed V, 5, and Nj).
(3.101)
Instead of the entropy maximum principle, we have an energy minimum principle:
• For a system held at constantV, S, and N(, at equilibrium, unconstrained param-
eters take on values that minimize the internal energy of the system.
We now consider an isolated system held at constant V, T, and Nt. For any spontaneous
internal process the first and second laws take the form of Eqs. (3.96) and (3.97). Combining
these relations to eliminate 8 Q gives
TdS >dU+SW.
Because V is fixed, 8W = PdV = 0 , which reduces the above equation to
TdS>dU. (3.102)
Differentiating F = U — TS and rearranging, we obtain
dU =dF + TdS + S dT. (3.103)
Substituting this result into the T dS inequality (3.102) yields, after a little rearranging,
dF<-SdT. (3.104)
For fixed 7\ dT = 0, whereupon Eq. (3.104) reduces to
dF < 0. (3.105)
Hence any spontaneous process must result in a decrease in F. It follows that at equilibrium,
F must be a minimum. Mathematically we state this condition as
dF=0, d2F>0. (3.106)
We generalize this as the Helmholtz potential minimum principle:
• For a system held at constant V, T, and N(, at equilibrium, unconstrained parameters
take on values that minimize the Helmholtz free energy of the system.
We can similarly develop an extremum principle for a system held at constant P and S.
With these constraints, the first and second laws require that for a spontaneous process
dH < 0. (3.107)
It follows that any spontaneous process reduces H. At equilibrium, H must therefore be a
3.71 Maxwell Relations 95

minimum. We state this condition mathematically as

dH = 0, d2H>0. (3.108)
This leads us to the Enthalpy minimum principle:
• For a system held at constantP, S, and N(, at equilibrium, unconstrained parameters
take on values that minimize the enthalpy of the system.
A final circumstance of interest is a system held at constant T and constant P. For these
constraints, the first and second laws require that for any spontaneous process
dG < 0. (3.109)
Any spontaneous process thus reduces G, implying that G must be a minimum at equilibrium
with
dG = 0, d2G>0. (3.110)
This can be stated as the Gibbs function minimum principle:
• Fora system held at constant P, T, and N(, at equilibrium, unconstrained parameters
take on values that minimize the Gibbs function of the system.
It is important to note that all forms of the extremum principle are equivalent. Generally it is
most advantageous to select the form that corresponds to the imposed boundary conditions
on the system of interest.

3.7 Maxwell Relations

Additional relations among thermodynamics properties can be obtained purely
from mathematical requirements. We consider here the fundamental equation in the energy
form for a system containing a binary mixture:
U = U(S,V,Na,Nh). (3.111)
Expanding U in a Taylor series and truncating after the first-order terms, we obtain

dU=["-- dS+["--) dV+[^-\ dNa+[-^- dNh.

(3.112)
We take the derivative in the first term of Eq. (3.112) and differentiate with respect to
V and take the derivative in the second term and differentiate with respect to S, which
yields

(^(™) ) ^ (3.113)

() ) . (3.114)
ds\dvJ / dSdv
But since the order of differentiation does not matter, the cross derivatives in Eqs. (3.113)
96 3 IA Macroscopic Framework

and (3.114) are equal. Consequently,

= (±(dJL\ I . (3.115)

However, by definition

— =T, (3.116a)
$$ /Na,Nb,V

= -/>, (3.116b)
J b

) =fia, (3.116c)
aJNb,S,V

) =Vb, (3.116d)
Na,s,v

and Eq. (3.112) can be written as

dU = T dS-PdV +iia dNa + fib dNb. (3.117)

Using Eqs. (3.116a) and (3.116b), we can write Eq. (3.115) as

) =() . (3.118)
dV d S
Js,Na,Nh \ /Na,Nb,V
Note that the above equation is a relation among thermodynamic properties known as a
Maxwell relation. This line of analysis can be applied to any two of the partial derivatives
that appear in Eq. (3.112). Doing so, the following additional relations among properties
can be obtained:

Na,V,Nb

(3.120)
\dNbJs,V,Na \ dS
/Nb,V,Nh
dP

, (3.122)

() () . (3.123)

Other Maxwell relations can be derived by considering the Legendre transformed fun-
damental relations. In Section 3.3 we showed that the Helmholtz free energy F could be
related to T, V, Na, and Nb by executing a partial Legendre transform that replaces S by T
in the fundamental relation U = U(S, V, Na, Nb). The resulting relation is of the form
F = F(T,V,Na,Nb) (3.124)
3.71 Maxwell Relations 97

and the Legendre transform yielded the relation

dF = -SdT - PdV +iiadNa+ nbdNb. (3.125)
Expanding the relation (3.124) for the Helmholtz function in a Taylor series and retaining
only first-order terms, we obtain

f u v
V,Na,Nb \ /T,Na,Nb \uiya/T,V,Nb \uiyb/T,V,Na

(3.126)
A comparison of Eqs. (3.125) and (3.126) clearly shows that

= -S, (3.127)
V,Na,Nb

— = -P, (3.128)

= Ma, (3.129)

= iib. (3.130)

We now can proceed in the manner described above to obtain an additional set of Maxwell
relations. For each relation, we select a pair of partial derivatives in Eq. (3.126), differentiate
each with respect to the differentiation variable of the other, equate the cross derivatives,
and substitute appropriate definitions from among Eqs. (3.127)-(3.130). This process yields
the following six additional Maxwell relations:

r) =( — ) , (3.131)
/T,Na,Nb \3T JV,Na,Nb

(3.132)
V,Na,Nb

ds
(3.133)

(3.134)
V,Na,Nh

(3.135)

| . (3.136)

The partial Legendre transform that replaces volume with pressure provides a relation
for enthalpy,
H = H(U,P,Na,Nb), (3.137)
 3 IA Macroscopic Framework

and subsequent differentiation yielded the relation

dH = TdS + VdP + fia dNa + ^b dNh. (3.138)
Using the same procedure as in the case of the Helmholtz function gives the following
Maxwell relations:

^ ) =(*-?-) , (3-139)
3? )S,Na,Nb V $$ / P,Na,Nh

(3.140)
P,Na,Nb

(3 141)
\ x* I ' -

dN UI
a/S,P,Nb \ /S,Na,Nh

(3.143)

(3.144)

The partial Legendre transform that replaces entropy by temperature and volume with
pressure yields a relation for the Gibb's function
G = G(T,P,Na,Nb), (3.145)
for which
dG = -SdT + VdP + na dNa + fib dNh. (3.146)
Using the same procedure as in the case of the Helmholtz function yields the following
Maxwell relations:

= (^f) , (3-147)
T,Na,Ni, y"1 /P,Na,Nb

MaJT,P,N> \dT )PM

(3.149)

(3.150)

(2U51)
™ bJT<P<Na \dPJTM
(3.152)
dN
bJT,P,Na \dNa/T,P, Nb
3.81 Other Properties 99

Using the definition of the grand canonical potential

U[T, V, na9 fib] = U -TS- iiaNa - tibNb, (3.153)

for which it can be shown that

dU[T, V, tia, fib] = -SdT -PdV -Na d[ia - Nb dfib, (3.154)

we can obtain the following Maxwell relations:

(w) -(£) •
< 3 1 5 6 )
• -

dS\ _(dNh\

(3.159)

It should be clear at this point that this methodology can be applied to any of the Leg-
endre transformed forms of the fundamental equation. Other Maxwell relations can be
derived in this manner by considering the partial transform that replaces only Na by \xa
(U[S, V, /j.a, Nb]), the partial transform that replaces Na by /j,a and Nh by \ib (U[S, V,
l^a, Hb\), a n d the partial transform that replaces V by P, Na by iAa, and Nb by fit,

3.8 Other Properties

Several other properties are of interest in thermodynamic analysis of specific two-
component systems. The isothermal compressibility KJ of a two-component system is de-
fined as

P JTNaNh
Another property of interest is the coefficient of thermal expansion fij defined as

Two additional properties that are frequently of use in thermodynamic analysis are the molar
specific heat at constant volume cy and the molar specific heat at constant pressure dp. These
100 3 IA Macroscopic Framework

are defined as
\ /3£A ^ (3 i 6 3 )

V\ ' (3-164)
where NA in the above relations is Avogadro's number. These defintions are equiva-
lent to

The equivalence of (3.164) and (3.166) is demonstrated in Example 3.5. Note that setting
either Na or Nb to zero one recovers the definitions of the above parameters for a pure
substance.

Example 3.5 Show that the definitions (3.164) and (3.166) for cP are equivalent.

Solution For a differential change in properties, Eq. (3.32) indicates that, for a
system with fixed Na and Nb,
dU = T dS-PdV.
By defintion, H = U + PV and it follows that dH = dU + PdV + VdP. Substituting
the above relation for dU and simplifying yields
dH = T dS + V dP.
Dividing both sides by dT, this relation becomes
dH
dT \dT) \dT
At constant P,Na, and Nb, the last term on the right vanishes and we can interpret dH /dT
and dS/dT as partial derivatives. The above relation then becomes

dT
Jp,N ,Nb b
p,Naa,N \dTd T //P,Na,Nh
The above result indicates that the enthalpy derivative in Eq. (3.164) can be replaced by
the right side of the above relation. Doing so yields Eq. (3.166). This implies that the
defintions (3.164) and (3.166) for dp are equivalent. A similar line of reasoning can be used
to demonstrate that the defintions (3.163) and (3.165) for dy are equivalent.

Example 3.6 Derive a general relation for the difference cP — cv.

3.8 I Other Properties 101

Solution Using a partial Legengdre transform that replaces U with T, a funda-

mental equation in the form S = S(U, V, Na, Nb) can be converted to S = S(T, V, Na, Nb).
For a system with fixed Na and Nb, it follows from basic calculus that

Using the Maxwell relation (3.131), this equation can be written in the form

I
)
/V,Na,Nh
(£) \
- 1
d I
/V,Na,Nh
dV.

We can similarly consider a partial Legendre transform that replaces U with T and V
with P, thus converting the fundamental equation from the form S = S(U, V, Na, Nb) to
S = S(T, P, Na, Nb). For fixed Na and Nb, basic calculus dictates that the corresponding
relation for dS is

T,N..Nt

Using the Maxwell relation (3.147), this equation becomes

Y
)
/P,Na,Nb
(£) \ d I
/P,Na,Nh
Equating the right side of the above equation to the right side of the equation for dS obtained
above, we obtain

U 1 U 1 U 1
V,Na,Nh \ /V,Na,Nh \ / P,Na,Nh \ / P,Na,Nb

Using the defintions (3.165) for cv and (3.166) for cP, this equation can be written in the
form
fdP\ \Na + Nb] fdV
cv dT + I — j dV = \cPdT - ( —

Solving for dT, we find

L(Na + Nb)(cP - cv) J (Na+ + NN)(

[l(N b)(cP - cv)) \
For a system with fixed Afa and Nb, there exists an equation of state of the form P —
P(T, V), or equivalently T = T(P, V). Hence, it follows from basic calculus that

d
dT=( ^-\ dV+(d^i dP.

For this equation and the one above to both be valid for arbitrary differential changes, the
respective coefficients of dV and dP must be equal. Equating the coefficients of either dV
102 3 IA Macroscopic Framework

or dP and solving for cP — cv yields

_ ( NAT \ fdP\ fdV

Cp Cv =
~ \ ) ) "
As discussed in Appendix I, for a function z = z(x, y)

Therefore

d T
\ Jp,Na,Nb\dV /T,Na,Nb
m
Substituing the right side of the above equation for (dP/dT)v,Na,Nb the relation for
dp — dy, we obtain

NA T

Using the definitions (3.161) and (3.162), this relation can be written more compactly as

Cp Cv =
~ (Na+Nh)KT'
The above general equation relates cP — cv to the isothermal compressibility and the co-
efficient of thermal expansion.

Exercises
3.1 V = 1.67 v = 1.80 V — 1.92

s = 7.35 u = 2506 s = 7.48 u = 2544 s = 7.61 u = 2582

s = 8.37 u = 3048 s = 8.52 u — 3131 s = 8.70 u = 3241
s = 9.06 u = 3665 s = 9.22 U — 3807 s = 9.45 u = 4005
3
(Units: v ~" m /kg, s ~-kJ/(kgK), M '- kJ/kg.)
Consider the tabulated properties above for steam. Note that u, v, and s are defined on a
per unit mass basis and that u = u/M, v = v/M, and s = s/M, where M is the molecular
mass.
(a) For each of the three values of specific volume, curve fit the relation u(s) with a
function of the form u = u^s1 + U\S + UQ, where UQ, U\, and u^ are constants. (Note
that the constants will differ for each v value.)
(b) For the variations of UQ,U\, and u2 with v, develop curve fits of the form u2 =
a2v2 + a\V + «3, u\ = b2v2 + b\V + b^, UQ = c2v2 + c\v + C3.
(c) Assemble the results of parts (a) and (b) together to obtain a single function u(s, v).
Is this function a proper fundamental equation? Explain your answer.
(d) Use the relation obtained in part (c) to predict the temperature and pressure for s = 8.52
J/kg K and v = 1.80 m3/kg.
Exercises 103

3.2 A system is divided into three subsystems by two rigid walls that are permeable to both
species of molecules in the system. The walls are fixed so that they cannot move in response
to pressure changes. The system is isolated so that no work or heat interactions with the
surrounding are possible. The three subsystems have different volumes and initially contain
different amounts of species a and species b molecules. After a long time, the overall system
reaches equilibrium. Determine the necessary conditions for equilibrium in this composite
system.
3.3 The system to be considered in this problem is identical to that in Figure 3.1 except that
the porous plate separating the two subsystems is free to move horizontally in response to
pressure changes in the subsystems. Determine the necessary conditions for equilibrium in
this system.
3.4 The system to be considered in this problem is identical to that described in Exercise 3.1
except that the piston is permeable to species a and species b molecules. Initially there are
no species b molecules in subsystem I and Nbti molecules of species b in subsystem II.
Determine the necessary conditions for equilibrium in this system.
3.5 The system to be considered in this problem is identical to that in Figure 3.1 except that the
plate separating the subsystems is adiabatic, is impermeable to both species, and is free to
move horizontally in response to pressure changes in the subsystems. Find the one neces-
sary condition for equilibrium in this system and explain why there is only one necessary
condition.
3.6 Three tanks containing helium gas each have their own shut-off valve. Initially the valves
are closed. Tank I has a volume of 0.010 m3 and an initial pressure of 200 kPa. Tank II
has a volume of 0.005 m3 and an initial pressure of 500 kPa, and tank III has a volume of
0.015 m3 and an initial pressure of 50 kPa. The tanks are connected to a common pipe and
the valves are opened, allowing the pressure to equalize. Find the final pressure in the tanks.
You may neglect the volume of the common pipe relative to the tank volumes.
3.7 For a single-component system, liquid and vapor may coexist under specific pressure and
temperature conditions. Such circumstances are said to be saturation conditions. As the
pressure is increased, the difference in density and other properties for the liquid and vapor
phase decreases until a pressure is reached where the difference between the two phases
vanishes. This is known as the critical point for the substance. The van der Waals equation
exhibits this type of behavior and is therefore an appropriate prototype equation of state for
a system that undergoes liquid-vapor phase change.
(a) If PC,TC, and 0c are the pressure, temperature, and volume per kmol at the critical
point, show that the van der Waals equation can be written in the form
p_ ^ 3
r
3 0 , - 1 Cr2'
where Pr = P/ Pc, Tr — T/Tc, and 0r = v/vc.
(b) Determine the values of vr that correspond to PY — 0.03 and T r = 0.85. What is the
physical interpretation of each value?
3.8 In Example 3.2 it was shown that the relation for chemical potential in a binary mixture of
monatomic gases can be written

where Pa is the partial pressure of species a defined as

Pa = XaP,
104 3 IA Macroscopic Framework

Pa,rcf is a n arbitrary reference value for Pa, and

p3/2, 2 \3/5

4 /3 /
~5j3 I •

(a) For Pajef = 101 kPa, determine the value of 0^a for helium, neon, and argon.
(b) For Pa = 101 kPa, determine and plot the variation of [ia with T for helium, neon, and
argon. What happens to \ia as T approaches zero? (Note that we will show in Chap-
ter 4 that the dilute occupancy assumption from which these results were derived is
inaccurate at very low T, and so this behavior is not expected to be physically realistic.)
3.9 Show that the equations of state for a binary mixture of monatomic ideal gases satisfy the
Euler equation (3.46).
3.10 We will demonstrate in Chapter 6 that the fundamental equation for a van der Waals fluid
can be written in the form
s = R In {(£> - NAby)(u + N2Aay/vfv} + s 0,
where i-v and so are fixed constants for a given fluid.
(a) Show that the van der Waals equation of state (3.41) can be derived from this relation.
(b) Show that for a van der Waals fluid the enthalpy per kmol is given by
v \ 2N2Aaw

3.11 A piston and cylinder device contains an ideal monatomic gas. The cross-sectional area of
the piston is 3.14 cm2. Initially the distance between the piston and the end of the cylinder is
2 cm, the temperature of the gas is 300 K, and the pressure is 100 kPa. The gas is compressed
reversibly and adiabatically until the distance between the piston and the cylinder is 1 mm.
Find the pressure and temperature of the gas at the end of the process and determine the
work done on the gas.
3.12 The piston and cylinder system in Figure 3.7 contains a binary mixture of two monatomic
ideal gases with Na = 0.2 kmol of species a and Nt, = 0.5 kmol of species b. Initially the
system has volume V\ = 0.001 m3 and pressure Pi = 101 kPa. Heat interaction between
the gas and the piston transfers energy into the gas and the piston slowly moves, increasing
the volume to V2 = 0.003 m3. The resulting constant-pressure (isobaric) expansion process
in the gas is reversible. Determine the work done by the gas and the energy transferred as
heat during the process.
3.13 An insulated rigid cylinder with a total volume of 0.015 m3 is divided into two chambers
(side 1 and side 2), each containing a binary mixture of monatomic gases. The volume of
side 1 is twice that of side 2. The chambers are separated by a rigid membrane that is perme-
able to component a but impermeable to component b. Heat transfer may also occur across
the membrane. Initially, on side 1: Na = 0.6 moles, Nh = 0.8 moles, and T = 400 K,

Figure 3.7
References 105

and on side 2: Na = 1.2 mole, Nt, = 0.6 moles, and T = 300 K. After equilibrium is
established, what are the values ofT,Na, and P on sides 1 and 2?
3.14 A simple compressible system undergoes a reversible constant-pressure (isobaric) process
from state 1 to state 2. In general the process may involve simultaneous work and heat
interactions with the system. Show that the heat exchanged during the process is equal to
the change in enthalpy H2 — H\.
3.15 Determine KJ(P) and fir(T) for a mixture of three monatomic ideal gases.
3.16 Determine the value of cp/cy for a binary mixture of two monatomic ideal gases.
3.17 Show that the definition (3.165) for cv is equivalent to Eq. (3.163).

References
[1] Howell, J. R. and Buckius, R. O., Fundamentals of Engineering Thermodynamics, 2nd ed.,
Chapter 9, McGraw-Hill, Inc., New York, 1985.
[2] Modell, M. and Reid, R. C , Thermodynamics and Its Applications, 2nd ed., Chapters 5-7,
Prentice-Hall, Inc., Englewood Cliffs, NJ, 1980.
 CHAPTER 4

Other Ensemble Formulations

In this chapter, we explore alternate statistical viewpoints of equilibrium systems.

We do so by defining alternative system ensembles with different combinations of system
boundary constraints. We also examine information on system property fluctuations that can
be extracted from the statistical analysis of the ensembles. This information on fluctuations
provides important insight into phase transitions considered later in Chapter 8.

4.1 Microstates and Energy Levels

The purpose of this first section is to clarify the relationship between microstates
and energy levels and to emphasize how they are represented in the statistical thermodynam-
ics theory developed in this text. In the Boltzmann statistics analysis presented in Chapter 2,
we considered the set of occupation numbers that indicated how the particles in the system
were distributed over the available energy levels. This led to the following relations for the
partition function:

(2.50a)
i=0
oo

7=0
The analysis can also be formulated in terms of occupation numbers that indicate how the
particles are distributed over the accessible microstates. The summations in the partition
functions would then be over all microstates rather than over the energy levels, and the
summation would be written as

i'=0
oo

where each /' or j ' designates a different microstate. If the degeneracy of energy level / is gi,
then there will be gi identical terms ofe~£i^kBT in the summation in Eq. (4.1a). In the energy
level formulation, these terms are represented by a single term gie~e^kBT. The microstate
summation indicated in Eqs. (4. la,b) is therefore identical to the corresponding energy level
summation in Eqs. (2.50a,b). The only difference is the method of bookkeeping.
For the most part, in the statistical models in this text we will consider distributions over
energy levels. The reason for this choice is that distributions over energy levels can often be
macroscopically detected, whereas it is impossible to detect how a system is distributed over
its many microstates. In particular, spectroscopic measurements can reveal how molecules

107
108 4 I Other Ensemble Formulations

in a system are distributed over the accessible energy levels. Distributions over energy levels
determined from statistical thermodynamics are valuable in interpreting the results of such
measurements. Despite this preference for the energy level formulation, it will sometimes
be convenient to consider the statistics of the system in terms of microstates. When we elect
to do so, primed indices such as /' or j ' will be used to emphasize that we are doing the book-
keeping in terms of microstates. Unprimed indices will indicate energy levels. The reader
should keep in mind that the physics is the same regardless of the way we do the bookkeeping.

4.2 The Canonical Ensemble

In his pioneering work on statistical mechanics, J. Willard Gibbs [1] explored
the statistical analysis of many-particle systems in terms of different ensemble types. One
of these was the microcanonical ensemble, which we have examined in Chapter 2. Two
other ensemble types that are particularly useful are the canonical ensemble and the grand
canonical ensemble. The former we will examine in detail in this section and the latter we
will consider in the next section.
As indicated schematically in Figure 4.1, a canonical ensemble is a collection of systems
that all have the same volume, the same number of particles of species a, and the same
number of particles of species b. (We are again considering systems that contain a binary
mixture of particles to explicitly include mixture effects.) The systems are all assumed to
be in perfect thermal contact with one another so that at all times they are at the same
temperature. The ensemble is perfectly insulated from the rest of the universe so that the
total internal energy of the ensemble is fixed at UE- Each system may be in any system
microstate consistent with the specified V,Na,Nt>, and T for each system in the ensemble.
The total number of systems in the ensemble, nE, is presumed to be so large that essentially
every possible microstate is represented in the ensemble. The energy levels accessible to
the systems in the ensemble, Ej, and the degeneracies, g 7 , are assumed to be a function of
Na,Nb,mdV:

gj=gj(Na,Nb,V).

Before proceeding to the canonical ensemble analysis, some comments regarding the
advantages of the canonical ensemble are in order. First, note that for the canonical ensemble
wefixthe temperature of the systems whereas for the microcanonical ensemble we specified

\\\x\\\\\\ \\w\\\\\w\\\\x\\\\\\\\\xxxx\x\\\xx\\
\\ Na
1 2
Na 3 Na 4 Na 5 Na 6 Na \\
\\ N
b N
b N
b N
b N
b N
b s\
\ V V V V V V \\
\ T T T T T T ss
\\ 7 \
\\ Na 8 Na 9 Na "£ Na \
\ N
b N
b N
b »b \
s
s V V V V \\
\\ T T T T \\
\\
A\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\W
\\

Figure 4.1 The canonical ensemble.

4.2 I The Canonical Ensemble 109

the energy of the systems. At a macroscopic level we can closely approximate isolating a
system so that its energy is fixed. However, even the best insulation can allow microscale
energy exchanges with the system. In that sense, we can never exactly construct a completely
isolated system with energy fixed at an exact value. On the other hand, it is often easy
to envelop a system in a larger body at a fixed temperature. One example would be to
immerse the system in a large well-stirred pool of water. The system and the pool will reach
an equilibrium at a temperature essentially equal to that of the pool. Microscale energy
fluctuations may momentarily transfer energy to or from the pool and lower or raise the
energy of the system, but over time, microscale fluctuations will keep the temperature of
the system fixed to a high level of precision. In addition, it is often easier to quantify the
temperature and volume for a system of interest than its energy and volume. As a result it is
sometimes more useful to have a statistical thermodynamics model for a system held at fixed
volume and temperature than one held atfixedvolume and energy. These arguments provide
some hints that the canonical ensemble formulation of statistical thermodynamics theory
may be particularly useful in some applications. We will see later that this is indeed the case.
We begin by considering the set of occupation numbers {nj} that indicates how the
member systems in the ensemble are distributed over the system energy levels. Each nj
in the set indicates the number of systems in energy level Sj. It follows directly that the
distribution must satisfy
oo

J2nj=nE, (4.2)
7=0
OO

Y^njEj = UE, (4.3)

7=0

where n^ is the total number of systems in the ensemble and UE is the total energy in the
ensemble. Based on the ergodic hypothesis (see Section 2.2), we assume that the average
properties of the ensemble are equivalent to the time average properties of the system being
considered. We therefore seek to determine the set of occupation numbers for the ensemble,
which we can then use to evaluate the statistical properties of the ensemble. For fixed Na,Nb,
T, and V, the energy of a real system at equilibrium is observed to be virtually constant with
time. Since particle collisions are continuously changing the microstate of the system, the
virtually constant system energy implies that, in the equilibrium distribution, the majority
of the ensemble members are in microstates with energies very close to a single value. In
terms of our ensemble, this implies that equilibrium corresponds to an occupation number
set that produces the greatest number of ensemble microstates. We are therefore seeking
the occupation number distribution that maximizes the number of ensemble microstates W
subject to the ensemble constraints indicated in Eqs. (4.2) and (4.3).
The systems in the ensemble are distinguishable. The number of ways of dividing the n^
distinguishable systems in the ensemble among the system energy levels as dictated by the
occupation number set {nj} is equal to the number of way of dividing n^ distinguishable ob-
jects into groups of n0, n\,... ,n}•, From basic combinatorial analysis (see Appendix I),
this is given by

(number of ways of dividing systems 1 -^ 1

< > = «E! I I —• (4.4)
{among energy levels as dictated by {nj}) -M: n-3!
110 41 Other Ensemble Formulations

However, each system energy level Ej has degeneracy gj, which means that the rij systems
in that energy level may be in any of gj microstates. To get a relation for the number of
microstates corresponding to an occupancy number set, we must multiply the contribution
of each energy level in the above equation by the number of ways that the nj systems can be
distributed among the gj microstates for that energy level. As described in Appendix I, the
number of ways of distributing nj distinguishable systems over gj microstates is equal to
gjJ. Multiplying this term for each energy level by the corresponding factor on the right side
of Eq. (4.4), we obtain the following relation for the total number of ensemble microstates
for a specified occupancy number set:

(4.5)
j=onj'

We will denote the probability of a system being in the jth energy level as Pj. From the
above observations it follows that
p nj J2lnj]nJW({nj})

For the ensemble considered here, we therefore seek tofindthe occupation number set that
maximizes W subject to the constraints (4.2) and (4.3). We expect that this set of occupation
numbers are those that characterize the statistics of the real system at equilibrium. Again,
for convenience, we elect to maximize In W subject to the constraints. Taking the natural
log of both sides of Eq. (4.5) yields
oo oo

inW =]nnEl + ^2nj]ngj -^lnw,-!. (4.7)

7=0 j=0

Since nE and rij are typically very large, we make use of Stirling's approximation (see
Appendix I) to evaluate the In nE\ and In nj! terms. The relation (4.7) then becomes
oo oo

\nW = nElnnE - nE + ^ n y l n g y - ^(w y -ln« y - - rij). (4.8)

7=0 7=0

Using constraint (4.2), this reduces to

00 OO

In W = nE In nE + V^ rij In gj — V^ nj In rij. (4.9)

7=0 7=0

Differentiating both sides of the above equation gives

oo oo

d(\nW) = dnElnnE + dnE + ^2dnj In gj -^(dnjlnnj +dnj). (4.10)

7=0 7=0

Since constraint (4.2) implies that

oo

2^ dnj = dnE = 0,
7=0
4.2 I The Canonical Ensemble 111

Eq. (4.10) reduces to

oo oo

d(lnW) = 2_^dnj \ngj — \Jdfty Innj. (4.11)

7=0 7=0

The maximum in In W implies that d(\n W) = 0. Imposing this requirement on Eq. (4.11)
yields
oo oo
dn lnn
Y dnj In gj - ^ J J =°- ( 4 - 12 )
7=0 7=0

Differentiating the constraint relations then gives

oo

\2dnj=dnE = 0 (4.13a)
7=0

and
oo

] T Ejdrij = dUE = 0. (4.13b)

7=0

We now make use of the method of Lagrange multipliers. (See Appendix I for a descrip-
tion of this technique.) We multiply the differentiated constraint equations by Lagrange
multipliers a and /3 respectively:
oo
dn =
a ^2 j °' (4.14a)
7=0
oo
E dn =0
$Y2 J J - (4.14b)
7=0

Adding these equations, subtracting Eq. (4.12), and collecting terms within a single sum-
mation yields
oo

^2 (a + PEJ ~ m#7 + lnnj)dnj = 0. (4.15)

7=0

At the maximum, this relation must hold for arbitrary choices of dnj, which implies that
the coefficients of the differential terms must all be zero:
a + pEj - Ingj + ]nrij = 0. (4.16)
We can now solve this relation for rtj. In doing so, we will designate the resulting
rij distribution as ft* to serve as a reminder that this distribution applies to a system at
equilibrium. The resulting relation is

Using the constraint relation (4.2) together with Eq. (4.17) yields
oo oo oo
n e ae Ej
nE = ^2 ) = Y^ 8j ~ ~^ = ~ Yl 8je~^Ej-
e a
( 4 - 18 )
7=0 7=0 y=0
112 4 I Other Ensemble Formulations

Combining Eqs. (4.17) and (4.18) leads to

(4.19)

The distribution above is generally written in the form

(4.20)
Q '
where

(4.21)
j=o

In the above relations, Q is termed the canonical partition function. Note that the summation
in Eq. (4.21) is over all system energy levels.
Two tasks remain to complete the canonical ensemble formulation: We must evaluate the
multiplier /3 and we must relate the partition function to other thermodynamic properties. To
accomplish these tasks, we will take advantage of information already obtained for systems
in a microcanonical ensemble. We first note that for the canonical ensemble as a whole, the
total energy and the total number of systems are fixed.
Furthermore, since the volume of each system is fixed and the number of systems is fixed,
the volume of the ensemble is fixed. These fixed conditions are the same as those imposed
on systems in a microcanonical ensemble. Thus, if we consider the systems in the canonical
ensembles to be just a special kind of "particle," with energy levels and degeneracies equal
to those for the canonical ensemble system, the canonical ensemble as a whole can be
treated as if it is one system in a microcanonical ensemble (see Figure 4.2). Note that the
systems in the canonical ensemble are free to exchange energy through random fluctuations
just as the particles could in the microcanonical ensemble. The results of microcanonical

Microcanonical Ensemble
2 3 n

\W\\\\\\\\\\\\\\\\\\\\V\X\X
\
\ 1 Na 4 5 6 AT \
2
Na 3 Na Na
Na \
s, N N Nb Nb Nb N
\
s
b b b \
s V V V V V V ^
T
777

T T T T T ^
ZZA

1 8 AT 9
Na Na Na HE
Na $
\ N
b N
b Nb ^ s
< V V V V \
T T T T \

Figure 4.2 A canonical ensemble as one system in a microcanonical ensemble.

42 I The Canonical Ensemble 113

analysis in Sections 2.2 and 2.3 apply to the canonical ensemble as a whole if we assign
variables as follows:
Na is set to zero since we consider only one type of system.
Nb is set equal to nE.
Particle energy level Sj with degeneracy gj corresponds to system energy level Ej
with degeneracy gj.
The particle occupation numbers Nbj correspond to the system occupation num-
bers nj.
Invoking these correspondence arguments, we can modify Eq. (2.31) accordingly to
obtain the following relation for SE, the total entropy of the canonical ensemble:
~ OO 00

-^
kB= nE\nnE + V* ttj In gj - Y^ nj \nrij. (4.22)
U U
Consistent with the usual assumptions for the microcanonical ensemble, this relation should
apply to an ensemble of distinguishable systems that have reached equilibrium by achieving
a system energy distribution that maximizes the number of ensemble microstates. Into
Eq. (4.22) we substitute the equilibrium distribution for the ensemble given by Eq. (4.20)
to obtain

(4.23)
Rearranging, this becomes

I = nElnnE - ( | ) ^ ^ j (|)

(|)
j=o \ ^ / j=0

°°
j2sje-'EnnQ. (4.24)
j=o

Canceling terms and using Eq. (4.21) to evaluate some of the summation terms, we can
simplify this relation to

i = ~P ( f ) E EJSJ<-'
JSJ Ei
+ »E In Q- (4.25)

We define the ensemble average of any property as being the summation over all energy
levels of the value of that property, for the system energy level, weighted by the probability
that the system exists in that energy level. Designating an arbitrary property as 7, the
definition of its ensemble average of (Y) is
oo oo —fiE
(Y) = Y, Y(Ej> V> ^a, Nb)P(Ej) = Y, Y(Ej, V, Na, Nh)^——. (4.26)
7=0 7=0 ^

The canonical ensemble analysis presumes that the average property value computed
using definition (4.26) is the value of that property that would be observed macroscopically
114 41 Other Ensemble Formulations

for a system at equilibrium at the Na, Nb,V, and T values specified for the ensemble. The
internal energy and entropy values observed at equilibrium would therefore be given by

EJ^--—, (4.27)

oo —0E-
^-—. (4.28)

Substituting Eq. (4.27) into Eq. (4.25) we can write

p (4.29)
The total entropy for the canonical ensemble, by additivity, is just the sum of the entropies
of the member systems. We write this sum in the form

SE = J2 S(EJ> v> Na, Nh) n). (4.30)

In the above relation, the entropy for each level is multiplied by the number of systems in
each level, n*9 and the terms for all levels are added, thus accounting for the contribution
of all systems in the ensemble. Using the occupation number distribution (4.20) to evaluate
«*, Eq. (4.30) becomes

Combining Eqs. (4.28) and (4.31) we obtain

SE = nE(S). (4.32)
Substituting the right side of Eq. (4.31) for SE converts Eq. (4.28) into

^ = -j8(£/)+lnG. (4.33)

Since the ensemble average values of U and S are taken to be equal to the equilibrium
properties for the system considered, Eq. (4.33) can be written

A = -pu+\nQ. (4.34)
£B
We next differentiate (4.34) with respect to p at constant Na, Nh, and V:

The last partial derivative on the right side can be evaluated using the definition of Q,

mm-±
Eq. (4.21):
4.2 I The Canonical Ensemble 115

As discussed in Section 2.3, the energy levels and the degeneracies for the particles in the
systems in the ensemble are expected to be functions of the volume of the system and
the number of particles of each species. It follows directly that the energy levels and res-
pective degeneracies for the systems are also functions of those variables. Since Na, Nh,
and V are fixed for these systems, the E-}and gj are fixed and can be considered as constants
in taking the partial derivative on the right side of (4.36). Evaluation of the derivative in
Eq. (4.36) then yields

hCD =±£-Bj8je->E'=-U. (4.37)

Q\W )Na,Nb,V Q~^
Substituting Eq. (4.37) into Eq. (4.35) and solving for f$ gives
(dS/dP)NatNb,v = 1
k(dU/dP)M,v kB(dU/dS)Na,Nb,v'
Since, by definition, T — (dU/dS) Na,NbjV, the relation for ft becomes

P = v^f- (4.39)
JcBT
Having evaluated ft, we can substitute this result into Eqs. (4.21) and (4.34) to obtain
the following relations for the partition function and entropy:

A = JL+ing. (4.41)
£B kBT
Upon examining Eq. (4.27), it can easily be verified that

V,Na,Nb

Upon using Eq. (4.39) and the chain rule to convert the derivative, we obtain the following
relation for U:

v,Na,Nb

Relations for other properties are derived as follows. Solving Eq. (4.41), for In Q, we obtain
lnQ fk ^ fk T . (4.44)
B B

Differentiating this relation yields

(4.45)
k B T2
But in Chapter 3 we showed that for any system containing a binary mixture of two species
of particles
dU = TdS - PdV + iiadNa + fibdNb. (4.46)
116 41 Other Ensemble Formulations

Substituting the right side of Eq. (4.46) for dU in Eq. (4.45), we obtain

+ ~S- ( 4 - 47 >
It is clear, however, from Eq. (4.40) that since gj and Ej are functions of V, NA, and
Nb, In Q is a function of T, V, Na, and Nb. Mathematically expanding In Q to first order in
these variables yields

V,Na,Nb V 9V /T,Na,Nb

«.. (4.48,
V 9^a /T,V,Nh V
For both (4.47) and (4.48) to be valid for all differential changes, the corresponding coeffi-
cients of the differential terms must be equal. Equating the coefficients of the dT terms leads
to Eq. (4.43). Equating the coefficients for the other differential terms gives the following
relations for the pressure and the chemical potentials:

, (4.49,
T,Na,Nb

a = -kBT[ , (4.50)

The relations (4.40), (4.41), (4.43), and (4.49)-(4.51) provide the link between microscale
energy storage in the system and macroscopic thermodynamic properties. In addition, by
substituting Eq. (4.39) into Eq. (4.20), we also obtain the following relation for the proba-
bility of finding a system in system energy level j :

(4.52)
Q

Example 4.1 Evaluate the canonical partition function Q for a monatomic ideal gas and
show that Eq. (4.49) is consistent with the ideal gas equation of state.

Solution Equation (4.44) links the partition function to entropy, temperature,

and internal energy:
_ S U
kB kBT'
For a monatomic gas, the internal energy and entropy are given by Eqs. (2.119) and (2.120).
Substituting these relations for S and U into the above relation yields

ln<2 = ^Na\n{ ,:,, 9 / , \ + -Nb\n{ /_,„» } + Na + Nb.

4.3 I The Grand Canonical Ensemble 117

Differentiating with respect to V, we obtain

'BQnQ)\ = ^ < ^
dV
JT,Na,Nb
Finally, we substitute the right side of the above relation into Eq. (4.49) to get
'3(lng)
P =kBT -kBT.
dv T,Na,Nh v

The final result above demonstrates that Eq. (4.49) is consistent with the ideal gas equation
of state.

4.3 The Grand Canonical Ensemble

We will now consider the statistical behavior of systems in a grand canonical
ensemble.
As indicated schematically in Figure 4.3, a grand canonical ensemble is a collection of
systems that all have the same volume. We are again considering systems that contain a
binary mixture of species a and species b particles to explicitly include mixture effects.
The systems are all assumed to be in perfect thermal contact with one another so that they
are all at the same temperature T. The ensemble is perfectly insulated from the rest of the
universe so that the total internal energy of the ensemble is fixed at £/E. The walls of each
system are permeable so that both species of particles may freely move among the systems.
The systems will thus reach mass transfer equilibrium, with the result that the chemical
potential for species a and species b will be the same in all systems. The boundary of the
ensemble is impermeable to both species and the total number of species a and species b
particles in the ensemble are fixed at Na$, and Nb,E> respectively.
Each system may be in any system microstate consistent with the specified V and the
values of Na and Nb for that specific system. The total number of systems in the ensemble,
«E> is presumed to be so large that microstates spanning all possible values of N a and Nb are
represented in the ensemble. The energy levels accessible to the systems in the ensemble
and the associated degeneracies are assumed to be a function of N a, Nt>, and V. For systems
with a specific pair of Na and Nb values, Eua,Nb,j and gNa,Nhj are a function of V:

8Na,Nb,j = gNa,Nb,j(V).

i2 * Ha
Hb
Ha
Hb
Ha
Hb
Ha
Hb
I Hb V V V V
T i| v
T | r T T T

A^a I A^a I A*a Ha

Hb
V \ V \ V V
T \ T \ T T

Figure 4.3 The grand canonical ensemble.

118 41 Other Ensemble Formulations

We begin by considering the set of occupation numbers {n^a^bj} that indicates how
the member systems in the ensemble are distributed over the system energy levels and the
ranges of particle numbers. Each n^a^b,j in the set indicates the number of systems in
energy level ENarNhj with particle number values of Na and A^. It follows directly that the
distribution must satisfy
oo oo oo

Z Z Y,nN«N>»j= n*> (4 53)

-
Na=0 Nb=0 7=0

ZZ Z Z^-'^^.^ =u*> (4 54)

oo oo oo

-
Na=0 Nh=0 y=
oo oo
oo oo
oo

Z Y,nNa,Nb,jNa = Na,E, (4.55)

Na=0 Nh=0 7=0
oo oo oo

Z Z J2nNaMJNb = NbtE, (4.56)

Na=0 Nh=0 7=0

where n^ is the total number of systems in the ensemble and UE is the total energy in the
ensemble.
Based on the ergodic hypothesis (see Section 2.2), we assume that the average properties
of the ensemble are equivalent to the time average properties of the system being considered.
We therefore seek to determine the set of occupation numbers for the ensemble so that we
evaluate the statistical properties of the ensemble. For fixed \± a, \ib, T, and V, the Na, Nb,
and U values for a real system at equilibrium are observed to be virtually constant with time.
Since the microstate of the system is continuously changing, the virtually constant values of
Na, Nb, and U imply that, in the equilibrium distribution, the vast majority of the ensemble
members are in microstates corresponding to one value of energy and one combination of
Na and Nb values. In terms of our ensemble, this implies that equilibrium corresponds to an
occupation number distribution that produces the greatest number of ensemble microstates.
We are therefore seeking the occupation number distribution that maximizes the number of
ensemble microstates W subject to the ensemble constraints indicated in Eqs. (4.53)-(4.56).
The systems in the ensemble are distinguishable. As discussed in Appendix I, from basic
combinatorial analysis, we know that

{
number of ways of putting n^0 distinguishable objects] ^

in boxes with no objects in box 0, n\ objects in box 1, > =ftdo-TT —•

7=0
n 2 objects in box 2, . . . , rtf objects in box i, . . . J
Here, each combination of Na, Nb, and j is a bin in which one or more systems may reside.
Hence, there must be a factorial term like those in the above product expression for each
combination of Na, Nb, and j . For a given set of occupation numbers {riNa,Nhj} we can
therefore write
number of ways of dividing n% systems
among groups corresponding to specific
combinations ofNa, Nb, and j as dictated .
(4.57)
4.3 I The Grand Canonical Ensemble 119

The triple product above indicates that we multiply the factorial terms for all combinations
of Na, Nt,, and j . Each system energy level E^a^bj has degeneracy gNa,Nbj, which means
that the nNa,Nbj systems in that energy level may be in any of gNa,Nb,j microstates. Note
that, in general, the degeneracy may depend on Na, Nh, and j . For simplicity in notation,
we will denote the degeneracy as gNa,Nbj, keeping in mind that it implicitly depends on
Na, Nb, as well as j . To get a relation for the number of microstates corresponding to an
occupancy number set, we must multiply the contribution of each energy level in the above
equation by the number of ways that the nNa,Nhj systems can be distributed among the
gNa,Nhj microstates for that energy level. As described in Appendix I, the number of ways
of distributing n^a^hj distinguishable systems over gNa,Nbj microstates is equal to g^^l') •
Multiplying this term for each energy level by the corresponding factor on the right side of
Eq. (4.57), we obtain the following relation for the total number of ensemble microstates
for a specified occupancy number set {n^a^bj}:
n
OO OO OO Na.NbJ

Na=0 Nb=0 j=0

We will denote the probability of a system being in the 7 th energy level with particle
numbers Na and Nb as PNa,Nbj- The above observations imply that

~ nNaMJ T,{nNa,NbJ}nNa,NbjW({nNaiNbj})
F N ,N J
a b = = ^ 7 7 7 T ( f x (459))
(

For the ensemble considered here, we therefore seek to find the occupation number
set that maximizes W subject to the constraints (4.53)-(4.56). We expect that this set of
occupation numbers are those that characterize the real system at equilibrium. Again, for
convenience, we elect to maximize In W subject to the constraints. Taking the natural log
of both sides of Eq. (4.58) yields
0 0 0 0 0 0

\nW = 2_, /-^ / , (nNa,Nb,j:^ngNa,Nb,j:~ ^nnNa,Nh,j]-) + ln^E-- (4.60)

Na=0 Nb=0 j=0

Since nE and nNa,Nbj are typically very large, we make use of Stirling's approximation (see
Appendix I) to evaluate the \nnEl and lnnNayNbj\ terms. The relation (4.60) for \nW then
becomes
OO OO OO

In W = y^ y^ y^ in™ M, ,- In gM M, j — n^ M. ,- In ftjv Nh / + nN Nh /) + nE In nE — n E.
Na=0Nb=0 j=0

(4.61)
Using constraint (4.53), this reduces to
OO OO OO

inW = } y 2_, \nNa,Nbj^-ngNa,Nbj ~~ nNa,Nbj^nnNa,Nb,j) + nElnnE.

S.-OWJ-O • ' • • ' { 4 6 2 )

Differentiating Eq. (4.53) yields

OO OO OO

^2 $Z 5Z dnNa,Nbj = dnE = 0. (4.63)

Na=0 Nb=0 j=0
120 41 Other Ensemble Formulations

Differentiating both sides of Eq. (4.62) and using Eq. (4.63), we obtain
oo oo oo

d(\nW)= Y Y Y (dnNaMJlnSNa,NbJ - dnNaiNbj]nnNaiNbJ). (4.64)

Na=0 Nb=0 7=0
Since d(\n W) = 0 at the maximum we seek, Eq. (4.64) further reduces to
oo oo oo

y y y [dn^a^hj\ng^a^hj — dn^a^bj\nn^a^hjj =0. (4.65)

Na=0 Nb=0 7=0
We once again employ the method of Lagrange multipliers. To do so, we first differentiate
the constraint relations and multiply both sides of each by a Lagrange multiplier to obtain
oo oo oo
adn
Y Y Y ^MJ = 0, (4.66)
Na=0 Nb=0 7=0
oo oo oo

E V
/_-/
V ^ PEN
/—/ r
Nh
"a,"b,J
jdnN Nh
"a,"b,J
j — 0, (4.67)
Na=0 Nb=0 7=0
oo oo oo
Y Y Y YaNadnNaMJ = 0, (4.68)
Na=0 Nb=0 7=0
oo oo oo
Y Y Y YbNbdnNatNbj = 0. (4.69)
Na=0 Nb=0 7=0

We then add Eqs. (4.66)-(4.69) and subtract Eq. (4.65). After a little rearranging, the
resulting relation takes the form
oo oo oo

Y Y Y (a + PENa,Nb,j + YaNa + YbNb ~ In gNa,NbJ + ^nnNa,Nbj)dnNa,NbJ = °-

Na=0 Nb=0 7=0
(4.70)

At the maximum, this relation must hold for arbitrary dn^a^h,j, which can only be true if
the coefficients of the differential terms are all zero:
a + PENaiNbJ + YaNa + YbNb - \ngNaMJ + \nnNaMJ = 0. (4.71)
The distribution that we obtain by solving Eq. (4.71) will be designated as n*N^Nhj to denote
that it is expected to apply to a system at equilibrium. The resulting relation for n*N^Nbj is

" L A W = 8Na,Nbj exp(-of - YaNa - YbNb - PENaMJ). (4.72)

Applying the constraint relation (4.53) we find
oo oo oo

Na=0 Nb=0 7=0

oo oo oo

= e~" Z ) E E SNa,NbJ exp(-yflM, - YbNb - pEN.,NtJ). (4.73)

Na=0 Nb=0 7=0

Combining Eqs. (4.72) and (4.73) yields the following relation for the distribution:
n
k,NbJ = 8NatNbJ &w(-YaNa - YbNb - PENatNbJ)
nE S
43 I The Grand Canonical Ensemble 121

where S is the grand canonical partition function defined as

00 00 OO

S = J2 Y Y SNa,Nh,J^V{~yaNa ~ YbNb - PENaMJ). (4.75)

Na=0 Nh=0 7=0

Our next task is to determine the unknown multipliers P,ya, and yb. We begin by noting
that each subset of systems within the grand canonical ensemble that have the same Na
and Nb values are, taken together, a canonical ensemble. In addition to having the same
Na and Nb values, they all also have the same volume and are in thermal equilibrium with
each other at the same temperature. Within this subset, all energy microstates should be
represented, and thus this subset fulfills all the requirements to be considered as a canonical
ensemble. The probability that a system within this subset is in system energy level j is
given by

PNa,NhJ = 5 5 *
,7=0 nNa,Nh,j
Using Eq. (4.74) to evaluate n*Na,NbJ- in the above relation, we obtain
5 _ {h)"EgNa,NbJ exp(-YaNa - ybNb - PENaiNbJ)
* Na,NhJ — v-^OO / 1\ / A T \T nn \ ' ^ '
^ {i)"8bJ exp(-y a iV f l - ybNb - PENa%Nb%j)
which reduces to

But we have shown in the previous section that for a canonical ensemble with specified Na
and Nb, the probability that a system is in energy level j is given by

je
Pj = ^ - ^ , (4.79)
where

7=0
Clearly, Eqs. (4.78) and (4.79) are equivalent only if ft = l/k^T. Since this must be true
for all the canonical ensemble subsets within the grand canonical ensemble, we therefore
conclude that for the grand canonical ensemble as a whole

To evaluate ya and yb, we return to the definition of the grand canonical partition function,
Eq. (4.75). For convenience, we define
Ya = YakBT, (4.81)
Yb = YbkBT. (4.82)
Substituting (4.80)-(4.82) into Eq. (4.75), the definition of the grand canonical ensemble
becomes
OO OO OO

s
= Y Y Y 8Na,NbJexp[-(yaNa + ybNb + ENaMj-)/kBT]. (4.83)
Na=0 Nb=0 7=0
122 41 Other Ensemble Formulations

We note that for specified Na, Nb, and j , the energy levels and degeneracies are expected
to be at most functions of the volume V. The other parameters that appear in the definition
of S are T,ya9 and yb. If we consider the natural log of S, it too would be a function of
V, T, ya, and %. Expanding In 3 to first order in these variables, we obtain

V,ya,yh V 3V /T,ya,yh

± d b (4 84)
+ "1T^ *?* + \ ^ Y' '
d
V Ya JT,V,yh V $Yb JT,V,ya
We must bring to this analysis some additional information on properties. The ensemble
average of any property is computed by multiplying the value of the property at each com-
bination of Na, Nb, and energy level j by the probability that a system has that combination
and summing such terms for all possible combinations of Na, Nb, and j . Designating an
arbitrary property as 7, this definition of its ensemble average (Y) is written as
oo oo oo

(Y)= J2 £ ^2Y(EN.,N>,j,V,Na,Nb)P(ENa,NbJ,Na,Nb)
Na=0 Nh=0 7=0
oo oo oo

Na=0 Nh=0 7=0

8Na,NbJ exp[-(j>flyva + %Nb + ENaMj-)/kBT]

x — . (4.85)

The grand canonical ensemble analysis is based on the premise that the average property
value computed using definition (4.85) is the value of that property that would be observed
macroscopically for a system at equilibrium at the /xa, \ib, V, and T values specified for
the ensemble. The ensemble average values of Na, Nb, internal energy U, and pressure P
are therefore given by
oo oo oo
N Na,NhJexiP[~(Ya^ h + YbNb + ENC,M,j)/kBT]
y y
Na=0 Nb=0
(4.86)
ex

Na=0 Nb=0
EN- b
Na,NbJ p [ ~ " ( / a ^ h + YbNb + ENct,Nhj)/kBT]

(4.87)
oo oo oo k
8Na,NbJQXP[~{ YaNa + ybNb +ENa,Nb,j)/ BT]
V V'
2^ 2^. <Na,Nb
3
Nn=0Nh=0 (4.88)
oo oo oo
P(ENaMJ,V,Na,Nb)
Na=0 Nb=0 7=0
RN Nh i ^X P I"— (9aNa + YbNb + EN Nh j) /kBT]
X *Na'Nb>] n Va a-rn b^ Na,Nb,j J / B J^ ^ ^

We will now use Eq. (4.83) together with Eqs. (4.86)-(4.89) to evaluate the partial deriva-
tives in Eq. (4.84). The derivatives with respect to T, ya, and yb are fairly straightforward
4.3 I The Grand Canonical Ensemble 123

to evaluate.
r
V ^-^ /
oo OO 00
1

exp[-(yaNa + yh Nh + ENaMJ)/kBT]
X

as
oo oo oo

g^,wfc,y exp[-(yaAfa + j>fcAffc + ENatNbJ)/kBT]

- ^ . " (4.91)

3(lnB)

(X) OO OO

Na=0 Nh=0 7=0

KknT
\ B

gNa,Nbj exy[-(yaNa + j/^^ + ENaMJ)/kBT]

( W
(4.92,

The derivative with respect to V is a bit trickier to evaluate:

/3(lnS)
V 9V
t M oo oo oo
1 d \-^ v—

say ^ ^«
/Ya=0 Nfc=0 7=0
x exp[— (yaNa + %Ni, + ENa,Nhj)/kBT~\
-. OO OO r, OO

^ /V a =0 ATfe=O 7=0
00 OO

f; f; --«>.w.^»/h.r3g(^.^.v.r)j
1

(493)
124 41 Other Ensemble Formulations

At this point we note that Eq. (4.49) from our analysis of the canonical ensemble implies
that the V derivative of Q in the above equation is just Q/ kBT times the pressure for a sys-
tem at specified Na, Nb,V, and T. We can therefore write the relation for the V derivative
of In S as
/3(lnB)\ _ I y - y - P(Na, Nb,V,T)Q(Na, Nb,V,T) -{%Na+9bNb)/k^
k T
V 9V JT,ya,9b~Z t o k o »
(4.94)
The factor P(Na, Nb, V, T) in the above relation is equal to the ensemble average pressure
for a canonical ensemble at the specified Na, Nb,V, and T. It follows from the definition
of the canonical ensemble average properties (4.26) that

ge
P(Na, Nb9 V,T) = J2 p{ENa,NhJ, V, Na, Nb)^ . (4.95)
Substituting this result into Eq. (4.94) yields
OO OO OO

Na=0 Nb=0 j=0

8Na,Nb,j exp[-(j>aA^a + %Nb
x ^

(4.96)
Comparing the right side of the above equation with Eq. (4.89), we see that the triple sum-
mation in Eq. (4.96) is just the ensemble average pressure (P) for the grand canonical
ensemble. Replacing the triple summation with {P) simplifies Eq. (4.96) to

(4.97)

Substituting Eqs. (4.90)-(4.92) and (4.97) for the partial derivatives in Eq. (4.84), we
obtain the following relation:
?b(Nb) , (U) \ ^ , (P) ^ (Na) (Nb)
+ +
+ 2 )dT + j-^dV
j^dV - -r^dy
r^dyaa - -—-dy b.
2 2 2 kT kT kT
k BT k BT k BT ) k BT k BT kBT
(4.98)
We add d({U)/kBT), d(ya{Na)/kBT), and d(yb(Nb)/kBT) to both sides and rearrange to
get
r, , (U) , 9a(Na) ?b(Nb)\ d(U)
kkT
BT kT
kBT kT
k BT ) kkT
BT k BT k BT kBT
(4.99)
As a final step, we drop the ensemble average brackets, acknowledging that the ensemble
average properties are expected to equal the real physical properties for the system of interest:

Tf + TV + W) " FT + T^dV + TT"1' + TT

k BT k BT k BT ) kBT kBT k BT k BT
(4.100)
4.3 I The Grand Canonical Ensemble 125

We now compare Eq. (4.100) with the following equation, developed in Chapter 3, which
applies to a binary mixture of two species:
TdS = dU + PdV - /xadNa - iibdNb. (4.101)
Equation (4.101) can be written in the form

(4.102)
k BT k BT
Equations (4.100) and (4.102) both must be valid for any choices of the differential terms.
If we set dNa and dNb both to zero, the right sides of both equations are equal and it follows
directly that

k BT )
Since the differentials on the left side of Eqs. (4.100) and (4.102) must be equal, the right
sides must also be equal for any choices of the differentials. For this to be true, the coefficients
of corresponding differentials on the right sides must be equal. Equality of the coefficients
for the dNa and dNt, terms requires that
Ya = -»a, (4.104)
Yb = -fib- (4.105)
Equations (4.104) and (4.105) can then be used with Eqs. (4.81) and (4.82) to evaluate the
Lagrange multipliers ya and yb:
Ya = ~^a/kBT, (4.106)
(4.107)
Substituting Eqs. (4.80), (4.106), and (4.107) into Eqs. (4.74) and (4.75), we obtain the
following relations for the occupancy distribution and the grand canonical partition function:
n
h,Nb,j = 8Na,NbJ exp[(^flNfl + iibNb - ENatNbJ)/kBT]
HE

where 3 is the grand canonical partition function defined as

oo oo oo
s
= ^2 5Z 5Z 8Na,Nbj exp[(tiaNa + iibNb - ENaMJ)/kBT]. (4.109)
Na=0 Nh=0 7=0

The distribution (4.108) can also be interpreted as the probability PNa,Nbj that a system at
equilibrium with specified T, V, \ia, and \ib will, at an arbitrarily chosen time, contain Na
species a particles, Nb species b particles and be in energy level j :
pN N . = SNa,NhJ exp[(/xaJVa + jibNb - ENaMJ)/kBT] ^ ^^

Having evaluated all of the Lagrange multipliers, we now will turn to the problem of
obtaining relations for thermodynamic properties. Using Eqs. (4.104) and (4.105) to evaluate
Ya and yb, the differential relation (4.103) can be integrated to obtain
U ^aNa
S = kB In S + — —
126 41 Other Ensemble Formulations

In the above relation So is a constant of integration. We note, however, that in the limit of
Na and Nb both going to zero, the system has only one accessible microstate: that with zero
internal energy and zero particles. In that limit, the partition function defined by Eq. (4.109)
goes to one and In S is zero. With the exception of the So term, the other terms on the right
side of the above equation also vanish as Na and Nb approach zero, implying that So is the
limiting value of entropy as the total number of particles approaches zero. In general our
definition of entropy has been based on the notion that it is an indicator of the number of
accessible microstates for the system. This notion suggests that a perfectly ordered system
with only one microstate should have zero entropy. We therefore set the integration constant
So equal to zero. Our relation for entropy then becomes

* k in-H-+ U ^aNa flbNb

(dun
S = kB\na + — ——. (4.111)
Replacing the ensemble average properties with the actual physical properties and using
Eqs. (4.104) and (4.105) to evaluate % and yb, Eq. (4.90) becomes

dT
\ ) T T T
The right side of Eq. (4.112) is identical to the last three terms on the right of Eq. (4.111).
We replace those three terms by the left side of Eq. (4.112) to obtain

(4.113)
V,/ia,fih

Equation (4.113) explicitly relates the system entropy to the grand canonical partition
function.
In Eqs. (4.91), (4.92), and (4.97), we similarly replace ya with —/x a, replace % with
—[ibt a n d replace each ensemble average property with the corresponding system physical
property. Rearranging the resulting equations a bit yields the following additional property
relations:

J , (4.114)

9(lnS)
^ (4.115)

. (4.116)

Solving for U, Eq. (4.111) becomes

U = TS - kBT In 3 + fiaNa + fibNb. (4.117)
In Chapter 3, we derived the Euler equation, which takes the following form for a system
containing a binary mixture of components a and b:
U = TS- PV + iiaNa+fihNh. (4.118)
Comparing Eqs. (4.117) and (4.118), we see that both relations can be valid only if
PV =kBT \nE. (4.119)
44 I Fluctuations 127

Solving Eq. (4.112) for U and using (4.114) and (4.115) to evaluate Na and Nh, we also
obtain
[ fd(\nE)\ fd(\nE)
U=kBT\/xa +/
(4.120)
Equations (4.113)—(4.116), (4.119), and (4.120) provide the link between the grand canon-
ical ensemble statistics and macroscopic thermodynamic properties.

Example 4.2 Determine the value of the grand canonical partition function 3 for a binary
mixture of monatomic ideal gases containing one kmol of each species.

Solution Since the mixture must obey Eq. (4.119),

PV =kBT In 3 ,
and the ideal gas equation of state,
PV = (Na + Nb)kBT,
we can combine these relations to obtain
In 3 = Na + Nb.
This relation can be solved for S, yielding

Substituting for Na and Nh, we obtain

r^ _ ^(l + l)6.02xl026 _ JQ5.2X!0 2 6

Clearly the numerical value of 3 is extraordinarily large, reflecting the fact that the number
of accessible system energy microstates with energies comparable to k#T is enormous.

4.4 Fluctuations
In this section we are going to examine microscale fluctuations in a system at
macroscopic equilibrium. We will consider fluctuations in a system that contains a binary
mixture of species a and species b particles. The degree to which a system fluctuates about
the mean value of a thermodynamic property Y can be quantified in terms of its variance
<jy (where ay is the standard deviation):
aF2 = (Y - (7)>2. (4.121)
Here we will specifically consider fluctuations of Na in a grand canonical ensemble (for
which V, r , /xa, and fit, are held fixed). Applying Eq. (4.121) to fluctuations in Na yields
< = (Na ~ (Na)f = (N* - 2Na(Na) + (Naf) = (N%) - 2(Na)(Na) + (Na)2
= (N%) - (Na)2. (4.122)
128 41 Other Ensemble Formulations

The ensemble average of a property Y for the grand canonical ensemble for a two-species
system is

oo oo oo

Na=Q Nb=0 7=0

oo oo oo

a=O Nb=0 j=0

~ ENa,NhJ)/kBT]
X

where the partition function for the two-species system is

oo oo oo

Y Y 8Na,Nh,jQXv[(VaNa + fibNb - ENaMJ)/kBT]. (4.123)

Na=0 Nb=0 y=0

Treating N% as the property and applying this relation, we have

^aNa + [ihNh - ENatNbJ)/kBT]

Na=0 Nb=0 j=0

OO OO OO , rr,

k BT
Y Y A^
Na=0 Nb=0 j=0

We simplify this relation by manipulating the right side as follows:

oo oo oo

Na=0 Nb=0 j=0

9 /NqgNqMJ exp[(/xQNa + iibNb - ENaMJ)/kBT]

OO OO OO

a =0 Nb=0 7=0
&cp[(fiaNa + iibNb-ENaMj)/kBT] | ^ B ^ 95

^ ^ NagNaMJ exp^aNq + fjLbNh - ENa,NhJ)/kBT]

x
2^0 ^1= 0 2
a 7=0
^Br as
+
= kBT— +kBT— (Na). (4.124)
9/x d[i
4 A I Fluctuations 129

Using Eq. (4.114) from the previous section to simplify the second term on the right of
Eq. (4.124), we obtain

Substituting the right side of the above equation for (N%) in Eq. (4.122) yields

a2 kBT(^) +(Na)2-(Na)2,
V °^a Jv,T,nb
which reduces to
2 , 1,fHNa)\
cr^ =kBT[ — . (4.125)

It should be obvious that a similar relation applies for the variance for species b, which can
be obtained by switching the roles of a and b in Eq. (4.125). Equation (4.125) links the
mean magnitude of the fluctuations in Na to the derivative (dNa/d/jLa)vj,fib. This provides
information about the fluctuations in the number of particles for one species in the system.
We may also be interested in fluctuations in the total number of particles in the system
N = Na + N(j. With some additional effort, the line of analysis above can be extended to
show that the variance in the total number of particles in the binary system is given by

| (d(Nh)

(4.126)

With a similar (but longer) line of analysis, it can be shown that the variance in the system
internal energy for the binary system is

du\ 2 f du \ 2 2 / du \2

(4.121 a)

where
(d(Na)\ (d(Nh)\
/ J /X
<D = -, r ^ , ^ \ :' ' « , r . (4.127b)
(d(Na)\ fd(Nb)\ _ fd(M\ fd(Nb)\

We will now explore the interpretation of Eq. (4.125) in terms of other thermodynamic
properties. From basic calculus, for a function z = z(x, y) we know that (see Appendix I)
dz\ fdz\ fdx

Holding T and \xt,fixed,we can therefore write

^_(dJ^l) (W_) . (4.128)

130 41 Other Ensemble Formulations

At constant (Na) and T, we expand (dV/d/jia)^a)j^h using the chain rule

dv
(4 129)
ZP I \z I • '
O r
(Na),T,iih \ / {Na),T,tih Ko^a J (Na),T,fih
Substituting the right side of Eq. (4.129) for (dV/dixa){Na)J^h in Eq. (4.128) gives

¥-) • (4.130)
VJ,nh
The first and third partial derivatives on the right side of the above relation can be evaluated
using the following two Maxwell relations:

fd(Na)\

Replacing the first and third derivatives in Eq. (4.130) with the derivatives indicated in
Eqs. (4.131) and (4.132), we obtain

) (™) ( B™) . (4.133)

Solving Eq. (4.119) for In S and substituting the resulting expression for In S in Eq. (4.114)
yields

Jv,T,fib

(4.134)

It is clear from this result that the first partial derivative in Eq. (4.133) is just N a/V.
Using this result to replace this derivative in Eq. (4.133) leads to

() () . (4.135)

To evaluate the rightmost derivative in the above equation, we proceed as follows. First, we
rewrite Eq. (4.134) in the form

• (4.130

Differentiating both sides of Eq. (4.136) with respect to V, we obtain

4.4 I Fluctuations 131

Switching the order of differentiation yields

,d^a
K3^)
L\ v* / P.T.UH-iy V,T,M,

(4.138)

Since Eq. (4.134) implies that (dP/9/xfl)v,r,/z& — Na/V, the above equation can be written

(4.139)

Substituting the right side of Eq. (4.139) for the rightmost derivative on the right side of
(4.135), we obtain

Replacing the derivative in Eq. (4.125) with the right side of Eq. (4.140) and replacing the
ensemble average properties with system physical properties, the relation for the variance
of Na becomes

Dividing both side by A^ and taking the square root, we obtain the following relation for
the fractional standard deviation of Na from its mean value:

The relations obtained for a binary system apply to single-species system if we set Nt>
to zero and drop fib as a parameter. For a single-species system, here we will drop the a
subscript and designate the number of particles and chemical potential simply as N and /x,
respectively. Conversion of relations derived earlier thus requires that we set Nb to zero and
replace Na and \ia with Af and /x respectively. It follows that for a single-species system,
the relation for the fractional standard deviation in the number of particles Af is given by

. (4.143)
-(dP/dV)NJv2
Using a similar strategy to simplify Eq. (4.127), it follows that for a single-species system
the variance of the internal energy about its mean value is given by

L
where cr^ is the variance of TV, which can be computed using Eq. (4.143). With some minor
manipulation of Eq. (4.144), the following relation for the fractional standard deviation of
132 4 I Other Ensemble Formulations

U can be obtained:
1/2
k B T2
(4.145)
u
At this point, a comment on the significance of these results is warranted. Note that
the analysis of fluctuations presented in this section was based on results for the grand
canonical ensemble formulation. The grand canonical ensemble formulation applies to an
open system in thermal and mass exchange communication with a host of other systems
subject to the same constraints. Within in a large body of fluid, such as the air in a typical
room, if we define a subsystem with a specified volume that is small compared to the overall
size of the body of fluid, that subsystem is subject to the constraints imposed on systems
in a grand canonical ensemble. For that reason, the results of the statistical analysis of the
grand canonical ensemble provide useful tools for thermodynamic analysis of fluidelements
within a larger body of fluid.
Our previous analysis of the grand canonical ensemble led to relations for the number of
ensemble members for each particle-number-energy level combination. The distribution
we found for a single-component system has a sharp maximum at a particular combination
of N and U, which we interpret as being the equilibrium values for a system at the volume,
temperature, and chemical potential values for the ensemble. The shape of the distribution is
shown schematically in Figure 4.4. The relations for a^ /N and cru/U derived in this section
directly indicate the sharpness of the peak in the distribution. If the standard deviation from
the mean property value is very small, the peak is very sharp, and the probability that the
system will be observed with a value of U or N other than the mean value is very remote.

Figure 4.4
4.4 I Fluctuations 133

However, if the fractional standard deviation is significantly greater than zero, there is a
nonnegligible chance that the system may be observed to wander away from the mean U
or TV value.
The sharpness of the distribution for TV can be easily assessed for an ideal gas. Since the
equation of state for an ideal gas is PV = NkBT, it follows that

-(8P/dV)Nj = - y r - (4.146)

Substituting therightside of Eq. (4.146) for -{dP/dV)NJ in Eq. (4.142), we obtain

°i = ^ (4.147)
Thus for an ideal gas, as the number of particles increases, the fractional level of fluctuations
decreases. For a system containing Avogadro's number of particles, the fractional deviation
from the mean number of particles is of the order of 10"13 . It is therefore not surprising that
any attempts to measure the number of particles in a system of macroscopic size will always
give a result in close agreement with the mean value of N, and the measured value for an
equilibrium system will not measurably change with time. Furthermore, to detect deviations
from the mean in macroscopic systems, we must have instrumentation that can resolve such
measurements to better than one part in 10 13. This level of precision is generally far beyond
common instrumentation. These results imply, however, that if the system is small so that
N is not large, the fluctuations may be a significant fraction of the mean value.

Example 4.3 Estimate the level of density fluctuations in a cubic centimeter of air at
normal atmospheric pressure and room temperature.

Solution For normal atmospheric pressure and room temperature we take P =

101 kPa and T = 293 K. Using the ideal gas equation of state we find
PV (101,000)(10-2)3
N 19
= W = 1.38 x l Q - * (293) = 2 - 5 Q

1
^ = -i= = , = 2.00 x 10- 10 .
19
N V^V V2.50 x 10
This implies that the standard deviation in the density in the 1 cm3 volume is only two parts
in 1010, an amount so small it is virtually impossible to measure.

Example 4.4 Estimate the level of density fluctuations in a spherical control volume with
a diameter of 200 A at conditions typical of the earth's atmosphere at 3,000 meters altitude.

Solution For the specified altitude, typical temperature and pressure conditions
are P =10 kPa and T = 268 K. The volume of the control volume is given by
134 41 Other Ensemble Formulations

Using the ideal gas equation of state we find

PV (70,000)(4.2 x 10-24)
1.38x10-23(268)
= 4= = 4== o . i i .
N VN V79
Thus, density shifts of 11% over regions of this size are common in the atmosphere at
this altitude. These fluctuations are significant because the variation in density results in
a variation in the dielectric constant, and the variation of the dielectric constant causes
the region to scatter light passing through the atmosphere. Visible light corresponds to
values of wavelength A between 4,000 and 7,000 A. For scattering objects that are small
compared to the wavelength of the radiation, the fraction of the incident radiation that is
scattered is strongly wavelength dependent, being proportional to X~4. This regime is known
as Rayleigh scattering. Rayleigh scattering does occur from regions of anomalous density
in the atmosphere that result from density fluctuations. Because of its strong wavelength
dependence, shorter wavelength blue light is scattered to a greater degree than the other
colors in the visible range, making the sky appear blue. The blue sky is a result of the
combined effects of density fluctuations and Rayleigh scattering.

Equation (4.142) also implies that fluctuations may be significant if — (dP/dV)Nj is

small. For pure fluids, — {dP/dV)^j becomes small in the vicinity of the critical point
and for metastable superheated liquid or supersaturated vapor. We will discuss metastable
fluids in more detail in a Chapter 8. However, it is worth noting here that the large density
fluctuations that occur when — {dP/dV)^j becomes small may raise the density of a vapor
nearly to that of a liquid phase or may lower the density of a liquid near that for a vapor
phase. Large density fluctuations that may initiate a change of phase are sometimes referred
to as heterophase fluctuations.

Example 4.5 A system contains saturated argon vapor at atmospheric pressure and 87.3 K.
Treating the vapor as an ideal gas, estimate the probability that a heterophase fluctuation
may occur within a spherical region of the system having a diameter of 5 nanometers.

Solution We consider subsystems of the overall system that permit mass and
energy to cross their boundaries with each subsystem held at the temperature and chemical
potential of the overall system. The subsystems are modeled as members of a grand canonical
ensemble. The probability that a subsystem has Na argon atoms is obtained by setting Nb = 0
in Eq. (4.110) and summing over all energy levels j . Doing so yields

7=0

where 3, the grand canonical partition function defined by Eq. (4.109) with Nh set to zero, is
oo oo
3
= E E gN.je^N--E>"-»
Na=0 j=0
4.4 I Fluctuations 135

Since the summation in the relation for P^a is over j , it can be rewritten as

7=0
The summation in the above relation is equal to the canonical partition function for an
ensemble of systems having volume V, Na particles, and temperature T. The above relation
is therefore equivalent to

E(T,V,/jLa)
In this equation, the dependencies of the partition functions have been indicated explicitly.
Taking the natural log of both sides converts the relation to

In this model analysis, we will treat the vapor as if it is an ideal gas. We will therefore
use relations derived previously for a monatomic ideal gas to evaluate /ji a,\nQ, and In S
in the above relations. Equation (4.119) is used to evaluate In S:
In 3 = PV/kBT.
For In Q, we use the relation derived in Example 4.1 with Nt> set to zero,

ln
G = (I
and the relation obtained in Example 3.3 for a monatomic gas is used to evaluate the chemical
potential,

In this analysis, the subsystem is a control volume with volume V. The values of \i a and T
for the subsystem are taken to be the same as those in the overall bulk system. In calculating
\ia we therefore use T and V/Na values associated with the bulk system. The overall system
is assumed to obey the ideal gas equation of state PV = NakBT, and to get the \ia value
for the bulk, we can replace V/Na in the above relation with kBT/P since T and P are
specified for the bulk system:
2/
J2nma '\

Substituting this relation for the chemical potential and the equations above for In S and
In Q into the equation for PNa,we get, after some simplification,

\nPNa =Naln[ — — +tffl- — .

\NakBT J \kBT J
In the subsystem, the chemical potential and the temperature are taken to be fixedat the bulk
system values, but we expect that the number of atoms in the volume may fluctuate with
136 41 Other Ensemble Formulations

time about a mean, which is the equilibrium value of Na. Since the gas in the subsystem
volume obeys the ideal gas law, and the mean pressure in the volume is expected to equal
the bulk system pressure P, we interpret P V/k^T as being equal to the mean number of
atoms in the subsystem (Na). We can therefore rewrite the above relation as

In P Na = Na In I - ^ i I + Na - <Na).

Taking the exponential function (inverse natural log) of each side of the above equation,
we get the following relation for the probability that the subsystem volume will contain N a
atoms:

p
Na = e

In a spherical volume with a diameter of 5 nm, the mean number of argon atoms in
saturated vapor at atmospheric pressure and 87.3 K is computed as

(Na) = PV/kBT = (101,000)(TT/6)(5 X 10~9)3/(1.38 X 10"23 x 87.3)

= 5.50 molecules.

The density of saturated argon liquid at these conditions is 1,394 kg/m 3 (see Appendix III).
Dividing this mass density by the molecular mass of 39.9 kg/kmol for argon and multiplying
by Avogadro's number yields a density of 2.10 x 1028 molecules/m3. To fill the subsystem
with molecules at the liquid density would require a fluctuation such that

Na = (2.10 x 10 2 8 )(TT/6)(5 X 10~9)3 = 1,380 molecules.

Putting the calculated values of (Na) and Na into the relation for PNa, we find that the
probability that such a fluctuation would occur is
1,380

Thus, the probability that random fluctuations in the vapor will produce near liquid
density in a region nanometers in dimension is virtually zero. The accuracy of this prediction
is limited because we treat the vapor as an ideal gas, and its behavior will deviate from that
for an ideal gas as it approaches saturation. The result nevertheless suggests that formation
of a liquid phase in a tiny subregion of the system by random fluctuations in the vapor is
very unlikely, even at saturation conditions. This observation is qualitatively correct for most
fluids. We will return to consider this issue in more detail in Chapter 8. In that chapter, we
will show that if the system is pushed into a supersaturated state, heterophase fluctuations
may have high enough probability to be likely initiators of a phase transition.

The ensemble formulations discussed so far in this chapter have considered system energy
levels. To apply the results of these formulations to real systems of atoms, molecules, and
other particles, we need to establish the link between molecular or particle energy levels
and system energy levels. We will examine this issue in detail in the next section.
4.5 I Distinguishability and Evaluation of the Partition Function 137

4.5 Distinguishability and Evaluation of the Partition Function

We now wish to turn our attention to the problem of evaluating the canonical and
grand canonical partition function for systems of atoms, molecules, or other particles. As
discussed in Chapter 2, particles commonly encountered in real systems fall into one of two
categories: fermions and bosons. For bosons, there are no restrictions on the manner in which
the system of particles distributes itself over available energy microstates. The behavior of
fermions is limited in that no two identical fermions can occupy the same particle energy
microstate. Analysis of fermions gives rise to Fermi-Dirac statistics, whereas consideration
of bosons leads to Bose-Einstein statistics.
To apply the ensemble statistical framework to a system of N particles, we must first
determine the complete set of energy microstates accessible to the system. If the system
is governed by quantum theory, this would generally require the solution of the Af-body
Schrodinger equation, which is a virtually impossible task. Fortunately, there are many
systems for which the Af-body Hamiltonian operator can be written as a sum of the individual
Hamiltonians. The total energy can then be written as a sum of individual energies for all
N particles:

(4.148)

The most common example of a system that conforms to this behavior is an ideal gas.
Another example is the decomposition of the Hamiltonian of a polyatomic molecule into
the contributions for its various degrees of freedom:
=
*~* ^translation ~T ** rotation T" ^vibration i ^electronic- (4.14;/)

For many systems amenable to analysis by classical mechanics and quantum mechanics
treatments, H, by a proper selection of variables, can be written as a sum of individual
terms. Pseudoparticles are sometimes associated with terms in this sum even if there are
no real particles. These are sometimes also called quasipartides. These include photons,
phonons, plasmons, magnons, and rotons. For a systems of many distinguishable particles
that behave this way, the total system energy can be computed as

If the system is held at fixed V and T, the canonical partition function is given by

Yl (4.151)
J'=° JLJ'b>Jc>-

where a,b,c, ... designate the different molecules and eya indicates energy microstate
j'a for particle a, eyb indicates energy microstate jrb for particle b, etc. Note that we have
elected, at least initially, to write the relation for the partition function as a sum over all
microstates rather than over energy levels. The notation j' a , j' b , j'c, ... in the summation
denotes a summation over all possible combinations of these indices. Equation (4.151) can
138 41 Other Ensemble Formulations

be rewritten as
oo oo oo

Note that we have cast this relation as a product of summations over all the energy microstates
for each particle in the system. If the degeneracy of each energy level for particle a,b, c,...
is gja, gjb, gjc, . . . respectively, we can write Eq. (4.151) as the product of summations over
all energy levels for the particles:

p~£h • Ik T \ ~ ^
Q = c
Ja=O Jb=O Jc=O

Each summation in the above relation is, in fact, a partition function for a specific particle.
Since we will most often apply this concept to molecules, we will define a molecular
partition function q as
oo

q = Y1 8je~£j/k*T- (4.154)

The above definition is written as a sum over the energy levels of the particle or molecule.
When it is useful, we can also write it in terms of a summation over energy microstates:

(4.155)
j'=o
With this definition we can write the relation (4.153) for Q as
Q = qaqbqc ---qN (for N particles), (4.156)
where the subscripts denote the particles for which the molecular partition function is
defined. Note that this implicitly assumes that the molecules are distinguishable. If the Af
molecules or other particles are all of same type, this relation reduces to
Q(N, V, T) = [q(V, T)f. (4.157)
If the particles are of two types, a and b, and the system contains Na particles of type a and
Nb of type b, the relation for Q becomes
Q(Na, Nb, V, T) = [qa(V, T)]N<[qb(y, T)f". (4.158)
If we now specifically consider molecules for which the molecular Hamiltonian can
be approximated by a sum of Hamiltonians for the various modes of energy storage in
the molecule, by the same line of reasoning outlined above, we can break the molecular
partition function into factors associated with each mode of energy storage,
<7molecular = <?tr<7rot<7vib<7e, (4.159)

where qiY, qTOt, qv{\>, and qQ are the partition functions for translational, rotational, vibrational,
and electronic energy storage, respectively. These are defined as
oo
<7tr = £ > r , ^ - £ » ^ r , (4.160)
4.5 I Distinguishability and Evaluation of the Partition Function 139

qmt = ^gtOtje-er«.j/*B'9 (4.161)

j=o
oo

q Y i b = J2 gvib,je~£"lb<j/kBT, (4.162)

oo

j=o

Note that the above definitions are written as summations over energy levels. Equivalent
definitions in terms of a summation over energy microstates also apply.
In this manner, not only can we reduce an TV -particle problem to a one-particle problem,
but we can reduce it further to the individual energy storage modes of a single particle.
While this is an attractive result, there is one major problem with it: Atoms and molecules,
in general, are not distinguishable. One exception is a system of atoms in a crystal lattice.
Because the mean position of each atom is fixed in space, each particle is distinguishable.
The results obtained above for distinguishable particles are therefore directly applicable to
solid crystals, and we will explore how we can apply them to solid crystals in a later section.
In a liquid or gas, the particles are free to move about within the system and any of
the particles may occupy a given location. For a given microstate of a system containing
a fluid, the particles have specific positions, velocities, and internal energy microstates. If
we exchanged two particles of the same species, the microstates would be different if the
particles were distinguishable, but we would not be able to tell them apart if the particles
are indistinguishable. If we treated the particles as distinguishable when in fact they are
not, we would overcount the number of microstates. Thus, the relation between Q and the
molecular partition function obtained for distinguishable particles is not correct for systems
containing a gas or liquid.
Without a means to link the canonical partition function to energy storage in indistin-
guishable particles or molecules, the canonical ensemble formulation of statistical ther-
modynamics would be of little value as an analysis tool. Fortunately, there is a fairly
straightforward way to establish such a link. In our analysis of the canonical ensemble, we
derived the following relation, which links the partition function Q to the internal energy,
entropy, and temperature.

In Chapter 2 we also derived the following equation for a system in a microcanonical

ensemble that contains a binary mixture of distinguishable particles:

f ^+NalnZa+Nb\nZh. (2.51)
£B k BT
Comparing Eqs. (4.41) and (2.51), it is clear that both can be valid for a system containing
a binary mixture of distinguishable particles only if

In Q = Na In Za + Nb In Zb. (4.164)
140 41 Other Ensemble Formulations

The definitions of the partition functions

oo

Z« = 5> e ~ e ' / ** 7 '' (2.50a)

/=0

are identical to that for the molecular partition functions defined by Eq. (4.154). Note that
both are summations over all accessible particle or molecular energy levels. It follows
directly that
Za=qa, (4-165)
Zb = qb. (4.166)
Substituting Eqs. (4.165) and (4.166) into Eq. (4.164) and solving for Q yields
Q = q^aq^h (for distinguishable particles). (4.167)
This result is identical to Eq. (4.158) obtained earlier in this section by writing the total
system energy in the canonical partition function as the sum of individual energies for all
particles in the system.
In Chapter 2 we also derived the following relation, which is valid for indistinguishable
bosons or fermions in the limit of dilute occupancy:

+Na+Nh. (2.78)
*B kBT " {Naj • - {N
Comparing Eqs. (4.41) and (2.78), it is clear that both can be valid for a system containing
a binary mixture of indistinguishable particles only if

inQ = Naln!^\ + Nbln(^-\ + Na + Nb. (4.168)

Rearranging the right side, Eq. (4.168) can be written in the form
In Q = Na \nZa + Nb \nZb - (Na lnNa - Na) - (Nb \nNb - Nb). (4.169)
Using Stirling's approximation in reverse, the terms in parentheses on the right side of Eq.
(4.169) are given by
Na\nNa-Na =\nNa\, (4.170)
Nb\nNb-Nb = lnNhl. (4.171)
Substituting the right sides of Eqs. (4.170) and (4.171) into (4.169) yields
In Q = Na \nZa + Nb lnZb - \nNa\ - lnNb\. (4.172)
Since Za = qa and Zb = qb, we can substitute and solve Eq. (4.172) to obtain the following
relation for Q:
for indistinguishable
6 particles \
F
. (4.173)
in the limit of dilute occupancy/
4.5 I Distinguishability and Evaluation of the Partition Function 141

Thus, by requiring consistency between the results for the microcanonical ensemble and
the canonical ensemble we have derived a relation between the molecular partition function
for distinguishable bosons and for indistinguishable bosons or fermions in the limit of
dilute occupancy. Furthermore, if we compare Eq. (4.167) for distinguishable particles with
Eq. (4.173), we note that Q for the binary mixture of indistinguishable particles in the high
temperature limit is lower by a factor of l/(Na\Nb\). This reflects the reduced number of
distinguishable system microstates when the particles are indistinguishable.
It should be transparent from the form of Eq. (4.173) that if the system contains r species,
the relation for Q becomes

n i=\
qNi
—
Nf •!
/for indistinguishable particles \
\ in
i n the
. . .
thp» limit
l i m i t of
r»f dilute
Hi1nt«=» occupancy/
,
nrpnnQTirv /'
(4.174)

where qt and Ni are the molecular partition function and number of particles for the ith
species.
The importance of Eq. (4.173) cannot be overstated. It provides the link between molec-
ular energy storage characteristics and the system partition function. We now have a clear
sequence of steps to determine system thermodynamic properties from particle energy
storage characteristics in the high temperature limit:
(1) Determine the partition functions for the individual modes of energy storage for
each species of molecule or other particle.
(2) Construct the molecular partition function for each species as <7moiecuiar =

(3) Construct the system canonical partition function using Eq. (4.173) (or (4.174)).
(4) Use the relations derived in Section 4.2 to determine thermodynamic properties
from the canonical partition function Q.
It should be noted that Eq. (4.173) and the methodology defined in the steps above are
applicable if the system meets the following criteria:
(1) The particles are indistinguishable.
(2) The total system energy can be written as the sum of the energies associated with
each particle.
(3) The system exhibits dilute occupancy.
Chapter 5 of this text will be devoted to applying the methodology defined in the steps
above to systems containing different ideal gases.
In closing this section, we take on the task of defining more precisely the high temperature
limit in which dilute occupancy is attained. Based on the quantum analysis of a single particle
in a box, in Chapter 1 we found that at moderate to high particle energies, the number of
translational states with energy less than a specified value s was given by

6 \ h2 )
If the mean translational energy of a particle in a system is (s), then n^e) is a good estimate
of the number of microstates accessible to the particle. Furthermore, in Section 2.6 we
showed that in a system of particles that store energy only by translation, the mean energy
142 4 I Other Ensemble Formulations

Table 4.1 Assessment of the Dilute Occupancy Criterion

for Several System Types

6N ( h2 V / 2
Fluid T(K) 7tV \\2mkBT )
Liquid helium 4 1.6
Gaseous helium 4 0.10
Gaseous helium 20 0.0020
Gaseous helium 100 4.0 x 10-5
Liquid neon 27 0.011
Gaseous neon 100 3.0 x 10~6
Liquid argon 86 5.1 x 10- 4
Gaseous argon 86 2.0 x 10- 6
Electrons in sodium 300 1,500

per particle is related to the temperature as

{£) = | * B r .
Dilute occupancy implies that
(4.176)
Taking e = (3/2)kBT in Eq. (4.175) and combining Eqs. (4.175) and (4.176) yields the
following criterion for dilute occupancy:
3/2
6N /
(4.177)
\l2mkBT
This condition is favored by high molecular mass, high temperature, and low particle density
(N/V).It turns out that this condition is satisfied for all but the lightest molecules at low
temperatures (see the typical values in Table 4.1).
Strictly speaking, the criterion described above was derived for systems containing par-
ticles that store energy by translation only. However, the results are also reasonably accurate
for most polyatomic molecules, since translational states usually account for much of their
energy. In any case, additional internal storage modes only increase the number of acces-
sible microstates and so if a system satisfies the inequality condition (4.177), additional
internal modes will only further dilute the occupancy. Thus (4.177) is a sufficient condition
for dilute occupancy as long as the particles are free to translate within the system.
At high system temperatures, the mean energy per particle will be high and the separation
of the quantum energy states will be small compared to the energy of the particle. As a result,
the energy can be considered to be a continuous variable, and the behavior of the system
is accurately predicted by classical mechanics models. At high temperatures, the statistical
behavior of such systems can also be treated by the statistical mechanics model developed
by Boltzmann, which is based on a classical mechanics view of energy storage in particles.
For that reason, in the limit of dilute occupancy, the particles are said to obey Boltzmann
statistics (for indistinguishable particles). Because their original formulation was based on
a classical mechanics view of energy storage, the Boltzmann statistics behavior at high
Exercises 143

temperature is also sometimes called the classical limit. Further discussion of ensemble
theory and the dilute occupancy limit can be found in References [2]-[5].

Exercises
4.1 The following relation defines an approximate partition function for a pure dense gas:

t V,T) = — (2nm^2TY {V _
N\ \ h J
In this relation, a and b are different constants for each molecular species. From this partition
function, derive an equation of state of the form P = P(V, N,T).
4.2 For a single-species system we can take Nb = 0, drop Ub as a parameter, and drop the a
subscritpts on Na and fia. Equation (4.125) then becomes

V d» Jv,T
Use this result to derive Eq. (4.145) for a single-species system held at constant V, /x, and T.
4.3 Show that Eq. (4.143) can be written in the form
r / T I1/2
GN kBT
N [-(dP/dv)T(N/NA)v2_
where v is the molar specific volume in m3/kmol.
4.4 Show that for a single-species system held at constant V, N, and T, the variance of the
energy is given by

•*-•*'(£)„
4.5 As noted in Chapter 3, for a single-component system, the van der Waals equation of state
can be written in the form

30r - 1 v 2'
where Pr = P/Pc, Tr = T/Tc, and vT = v/vc, with Pc, Tc, and vc being the pressure,
temperature, and volume per kmol at the critical point. For 0r = 1, plot the variation of
(aN/N)(PcvcN/NAkBTc)l/2 with Tr predicted by the van der Waals equation of state for
1 < Tr < 2. What is the physical interpretation for the behavior of the fluid as Tr —> 1?
4.6
(a) At constant T and V, the chemical potentials for a binary mixture are functions of the
number of particles of each species:
[ia = fia(Na, Nb), iih = iu,b(Na, Nh) (at constant T and V).
Show that this implies that

^a / V,T,nb V dNa ) V,T,Nb V dNb J V,T,Na ^ \dNbJ V,T,Na V dNa ) V,T,Nb

144 41 Other Ensemble Formulations

(b) A binary mixture of two monatomic ideal gases a and b has temperature of 300 K, a
pressure of 50 kPa, and a concentration of species a of one part per trillion (i.e., a mole
fraction of 10~12). You take a small sample of the binary mixture at one instant of time
that has a volume of one cubic millimeter. Assuming that the fractional uncertainty
in the measured concentration from this sample is equal to crNa/Na, use the results of
part (a) to assess the accuracy of bulk concentration measurements from a sample this
small.
4.7 Derive Eq. (4.126) for the variance of N = Na + Nb in a binary mixture of two particle
species a and b.
4.8 For 0.018 kg of steam at 100°C, use the van der Waals equation to estimate how far the
pressure must rise above the equilibrium saturation pressure before aN/N is of order one.
4.9 For 0.001 kmol of liquid nitrogen at 120 K, use the van der Waals equation to estimate how
far the pressure must fall below the equilibrium saturation pressure before crN/N is of order
one.
4.10 For saturated neon vapor at 27.1 K and atmospheric pressure, estimate the probability that a
heterophase fluctuation will raise the density to that of saturated liquid in a spherical region
with a diameter of 5 nm. Note that neon is a monatomic substance with a molecular mass
of 20.18 kg/kmol and a saturated liquid density of 1,205 kg/m3 at 27.1 K.
4.11 Evaluate the parameter

6N / h2
TTV \ 12mkBT
for steam at 100°C and assess the validity of the dilute occupancy assumption for this
system.
4.12 In a gas of protons the mean kinetic energy of each proton is 1.04 x 10~19 J. The mass of a
proton is 1.66 x 10~27 kg. Evaluate the dilute occupancy assumption for this system.
4.13 Evaluate the parameter
3/2
6N ( h2
TtV \\2mkBT
for electrons in a copper wire at room temperature and assess the validity of the dilute
occupancy assumption for this system.

References
[1] Gibbs, J. W., Elementary Principles in Statistical Mechanics, reprinted by Ox Bow Press,
Woodbridge, CT, 1981.
[2] Hill, T. L., An Introduction to Statistical Thermodynamics, Dover Publications, New York, 1986.
[3] Kittel, C. and Kroemer, H., Thermal Physics, 2nd ed., W. H. Freeman and Company, New York,
1980.
[4] McQuarrie, D. A., Statistical Mechanics, Harper and Row, New York, 1976.
[5] Robertson, H. S., Statistical Thermophysics, Prentice-Hall, Englewood Cliffs, NJ, 1993.
 CHAPTER 5

Ideal Gases

In Chapter 5 we apply the statistical theory developed in earlier chapters to ideal

gases. We consider a binary mixture of two gas species and note that pure component
relations can be recovered by setting the number of particles of one species to zero. The
necessary conditions for chemical equilibrium in reacting gas mixtures are also examined.

5.1 Energy Storage and the Molecular Partition Function

We now wish to turn our attention to the problem of evaluating the partition function
for a system containing a pure ideal gas. In doing so we will make use of the methodology de-
veloped in Section 4.5 that relates the canonical partition function to the molecular partition
function for indistinguishable particles under conditions of dilute occupancy. The major task
in applying this analysis to systems of ideal gas molecules is to derive relations for the molec-
ular partition function from available information about energy storage in the molecules.
In general, energy can be stored in a molecule as translational, rotational, or vibrational
motion, and as a result of electronic or nuclear transitions. For the gases considered in this
chapter, these energy storage mechanisms will be assumed to be independent. We therefore
expect that we can decompose the Hamiltonian for a polyatomic molecule into its various
degrees of freedom

" — ^translational ~r ^rotational ~r ^ vibrational T~ ^electronic ~r tl nuclear w«I)

and, as a result, the molecular partition function can be written as the product of factors
associated with each of the energy storage modes,

q(V,T)= 4tr<7rot<7vib<7e<7nucl. (5.2)

To evaluate each of the factors on the right side of Eq. (5.2), we will analyze each of the
energy storage modes that are active for a given type of molecule. We will begin in the next
section by considering monatomic gases. We then consider more complicated diatomic and
polyatomic molecules in subsequent sections.

5.2 Ideal Monatomic Gases

We now will apply the analysis tools developed in the previous chapter to an
ideal monatomic gas. "Ideal" implies that the gas is dilute enough that intermolecular
interactions can be neglected except for their effect during very brief collision processes
between molecules. This is generally true for monatomic gases at pressures below a few
atmospheres and temperatures greater than 20°C.
The number of independent energy storage modes is sometimes referred to as the number
of degrees of freedom for the molecule. A monatomic gas atom has translational, electronic,
and nuclear energy storage modes. To a very good approximation, the translational, nuclear,

145
146 51 Ideal Gases

and electronic Hamiltonians are separable. Based on the results obtained in Section 4.5, this
implies that
q(V,T) = qtiqeqnuc\> (5.3)

The Translational Partition Function

In Section 2.6, we determined the partition function for a particle that stores energy
by translation alone. The definition of the partition function Z a in that section is identical
to the definition of qtY:

^ (5.4)
1=0
It follows directly from the analysis presented in Section 2.6 that the translational partition
function for a monatomic gas is given by

The above relation can be written in the form

<7tr="^, (5.6)
where

A =
27tmkBT
The factor A that occurs in the translational partition function has units of length. The usual
interpretation of A is based on the following. The mean translational energy of a particle is
computed as

8trj€
<etr) = V sttJ . (5.8)

With a little rearranging, it is easy to show that

(5.9)

Substituting the above relation (5.5) for qtY yields

(Str) = (j) kBT. (5.10)

Since £tr = p2 /2m, where p is the particle momentum, the average momentum is propor-
tional to (mkftT)1!2. Thus A is essentially h/p, which is equal to the de Broglie wavelength
associated with the thermal motion of the particle. As a result, A is termed the thermal
de Broglie wavelength. Note also that the condition for the validity of dilute occupancy,
Eq. (4.177), can be stated in terms of A as

(5 U)
V/N • -
5.2 I Ideal Monatomic Gases 147

Since the fraction on the right side is close to one (0.72), the criterion can be more simply
stated as
A
« 1. (5.12)
V/N
This implies that the thermal de Broglie wavelength must be small compared to the mean
volume per molecule in the system containment. Note that this is similar to the condition
that quantum effects decrease as the de Broglie wavelength becomes small compared to the
physical system dimensions.

Example 5.1 Calculate the thermal de Broglie wavelength and assess the assumption of
dilute occupancy for nitrogen gas at 300 K and 101 kPa.

Solution The mass of a nitrogen molecule (N2) is 4.65 x 10~26 kg. Substituting
into Eq. (5.7), we obtain
_ / h2 \1/2 _ / (6.63 x 10~34)2 1/2

\2nmkBT) ~ V2^(4-65 x 10-26)(1.38 x 10~23)300,

= 1.91 x 1 0 " n m .
Treating the gas as ideal, we can evaluate V/N using the ideal gas law
V _ kBT
N ~ P
Substituting the system temperature and pressure yields
V (1.38 x 10-23)300
= 4.10 x 10~26 m3 (per molecule).
N 101,000
It follows from these results that
A3 (1.91 x 10" 11 ) 3
= 1.70 x 10" 7 .
V/N 4.10 x 10~26
This ratio being much less than one implies that the dilute occupancy idealization is appro-
priate for this system.

The Electronic Partition Function

For our analysis here, we write the electronic partition function as a sum over
energy levels:
oo
£ l/k T
qe==J2gzje- - », (5.13)
/=o
where gej is the degeneracy and £e,/ is the energy of the /th electronic level. We fix the
arbitrary zero of the energy such that £e,o = 0. In doing so we effectively measure all
electronic energy levels relative to the ground state:
•••, (5.14)
148 5 I Ideal Gases

where A£e,/o is the energy of electronic level / relative to the lowest energy level (ground
state). These A£e,/o terms are typically of the order of electron volts. In general, As e jo/kBT
is quite large at ordinary temperatures and only thefirstterm in the summation is significantly
different from zero. Consequently, we need retain only the first term in most cases. In a few
cases, such as for halogen gases, the second and third levels are close to the ground state.
For such cases, the second- and third-level terms should also be retained.
For all monatomic gases, the electronic energy levels and degeneracies are known from
measurements and/or quantum theory models. Hence, we write the electronic partition in
the form

<7e = ge,0 ~\~ g e , \ e C

'° B • (5.15)
Possible inclusion of additional higher energy level terms needs to be considered only at
very small changes A£e,/o or at extremely high temperatures.

The Nuclear Partition Function

Because nuclear energy levels are separated by millions of electron volts, at room
temperature only the first term in a summation for qnuc\ need be considered:
<7nucl = gnucl,0, (5.16)
where gnuc\,o is the degeneracy of the lowest energy level.

Properties
Assembling the parts, we have the following relation for the molecular partition
function for a monatomic ideal gas:

q = <7tr<7e4nucl

y fge Q _|_ g e ie~ASe-l0^kBT)gnuc\ 0. (5-17)

Applying Eq. (4.174) we obtain the following relation for the canonical partition function:

^— V(g e0 + gele~Aeelo/ *B r)gnucio • (5.18)

2
V h J ' '
We will usually find that it is appropriate to take gnuci,o = 1- We will therefore do so in
evaluating the properties of the gas. It follows directly from our analysis of the canonical
ensemble that
_2/3(lnG)\ 3 T^NgeAAeeAOe
U = k T
* XT =-NkBT + Ae ,k T , (5.19)
V dT JNV 2 ge + -As^k*T

N,V

= NkB\n{[ ,_," ) T U, (5.20)

5.3 I Ideal Diatomic Gases 149

where

S, _

Equation (5.20) is referred to as the Sackur-Tetrode equation. We also can obtain a relation
for the pressure

which, not surprisingly, is the ideal gas equation of state. Note that inclusion of the electronic
storage terms does not affect this equation of state.

5.3 Ideal Diatomic Gases

As a next step, we will consider an ideal gas composed of diatomic molecules.
In addition to translational and electronic degrees of freedom, diatomic molecules possess
vibrational and rotational degrees of freedom as well. Diatomic molecules represent the
next step upward in complexity above a monatomic gas. Many technologically and envi-
ronmentally important gases are diatomic. Examples include H2, N2, O2, CO, NO, CI2, and
HC1.
A general procedure to determine the energy levels of a diatomic molecule would be to
set up the Schrodinger equation for the two nuclei and n electrons and solve it for the set of
eigenvalues of the diatomic molecule. Exact application of this method has been done for H2,
but it becomes increasingly difficult as the number of electrons in the molecule increases.
Approximate methodologies have been developed to circumvent difficulties associated with
full solution of the Schrodinger equation. One way of determining electron orbitals in the
molecule is the so-called LCAO-MO method. The acronym "LCAO-MO" stands for linear
combination of atomic orbitals-molecular orbital. This method is based on the notion that
orbitals for the molecule can be predicted as linear combinations of orbitals in the individual
atoms. Using this method (or variations of it), total electron density maps can be generated
for virtually all diatomic molecules.
The simpler model that we will adopt here makes use of the Born-Oppenheimer approx-
imation. This approximation is based on the fact that the nuclei are much more massive
than the electrons. The electrons are therefore considered to move in a field produced
by the nuclei fixed at some internuclear separation. The motion of the nuclei in the elec-
tronic potential field established by the surrounding electron cloud varies with the elec-
tronic state of the molecule. Calculation of the nuclei interaction force potential </>j(r)
for even the ground state is a difficult n -electron calculation. As a result, semiempiri-
cal approximate relations for 0/(r) are often used (where r is the internuclear separation
distance).
It is a simple matter to show that the motion of two masses in a spherically symmetric
potential can be rigorously separated into two separate problems by the introduction of
center-of-mass and relative coordinates. For example, for the analogous 2-D case of two
masses m\ and mi, moving in the x-y plane, the total energy is given by

E (*?+tf) + &+%+*<* * * y£ <523>

150 5 I Ideal Gases

Introducing the new variables

m\X\+m2x2 ynxyxJrm2y2
(5.24)
m\ + m2 m\ + m2

x\2 = x\ — x2, y\2 = y\ — y2, (5.25)

we can reduce this two-body problem to two one-body problems. Substituting to execute
the above transformation of variables, we can show that the Hamiltonian for this system is
given by
m m 2
H =E = ^ \ x + Y2) + y (x22 + y22) + 0(* 12 , v12), (5.26)

where mr, the reduced mass of the two particles, is

mxmi
mY= . (5.27)
mi + ^2
Using
dH _ . 3# _

equations of motion for this system can be derived that can be divided into two groups:
(1) those that govern the translation of the center of mass having effective mass m\ + m 2
and (2) equations that govern kinetic and potential energy interactions resulting from the
relative motion of the two particles. Note that this transformation of the problem can also
easily be extended to the 3-D case. If we lump the relative motion modes of storage together
as internal storage, the Hamiltonian can be written as
H = HtY + H-mt, (5.29)
where HtT is the contribution associated with translation of the center of mass and Hmt
is the contribution associated with relative motion of the atomic nuclei in the molecule.
The latter contribution includes vibrational and rotational energy storage. From quantum
analysis, this decomposition of the Hamiltonian will result in energy eigenvalues that also
separate into two independent contributions,
£ = £tY + e^ (5.30)
The partition function of a diatomic molecule is therefore expected to be given by
q=qtrqint. (5.31)
Since the center-of-mass motion is analytically the same as that of a monatomic particle,
we also expect that

(5.32)

The density of the translational states alone is so high that we expect dilute occupancy. It
follows that
aNaN
,T) = ^ ^ . (5.33)
AM
5.3 I Ideal Diatomic Gases 151

force - -d<p/dr

d<pldr>0 => attractive

d({>/dr<0^> repulsive

Figure 5.1 Potential function for model system.

When decoupled from translation, as described above, the relative motion of the two
bodies can be interpreted as one body of reduced mass mr = m \m2/{m \ + m^) moving about
the origin. The potential function representing the force acting on the particle of reduced
mass 0(r) is of the form shown in Figure 5.1. The relative motion of the nuclei in the
molecule consists of rotary motion about the center of mass and relative vibrational motion
of the two nuclei. This is mathematically represented by the rotary motion and vibrational
motion of the body of reduced mass in the potential 0(r). Because the amplitude of the
vibrational motion is very small it is a good approximation to consider it as representing
the angular motion of the effective mass at a fixed internuclear distance r e . The equilibrium
internuclear distance is presumed to correspond to the zero force (d</>/dr = 0) condition at
the bottom of the potential well. Expanding 0(r) about r e we get

d<f>
dr

The linear term in the 0(r) expansion vanishes because d(p/dr =0 at the bottom of the
well. The expansion can then be written as
= 0(re) + \k(r - r e ) 2 + • •
•, (5.35)
with k being the effective force constant defined as

(5.36)

The approximation embodied in this treatment of the relative motion storage modes is
called the rigid-rotor-harmonic-oscillator approximation. It follows directly from these
arguments that we can write
(5.37)
and
(5.38)
<7int — (5.39)
152 5 I Ideal Gases

If we extend this treatment to include the electronic and nuclear degrees of freedom, for
the complete molecule we have

H = HtT + HTOt + //vib + HQ + ^nuch (5.40)

e — c _|_ c _J_ c ., _l_ o _1_ c , f S ZL1 ^

q = q q q -hq q \ (5 42)

We have already evaluated g tr. The electronic partition function will essentially be the
same as that for a monatomic gas (although the energy levels will be those for the specific
molecule):
qQ = ge o -f- gQ^e ' . {JAJ)

As for the monatomic gas, we will adopt the approximation that qnuc\ = 1 (i.e., the molecule
remains in the nuclear ground state, which presumably has a degeneracy of one). Note that
the implication in the above relations that these modes are independent is an idealization
that is not exactly correct. It does, however, serve as a useful approximate model. Within
this approximation, the partition function is given by

Q{N,V,T)= m

To fully evaluate Q, we therefore must develop a means of evaluating qTOt and gvib.

The Vibrational Partition Function

Modeling the vibrational mode as a harmonic oscillator, our previous quantum
analysis indicated that the energy levels and degeneracies are given by

f f o r « = 0 , 1,2,..., (5.45)
gvib,n = I

where the characteristic frequency v is related to the reduced mass and force constant as
1 / h \ l/^

v= — (— ) (5.46)
In \mrj
Since the above relation predicts that the lowest (ground) state has finite energy, assigning
zero energy to that state would be inconsistent. Instead, we therefore assign zero energy
to the bottom of the potential well for the lowest-energy (ground) electronic state. The
oscillator will then not actually be able to reach the bottom of the well but will have a
lowest energy of (l/2)hv as a ground state (see Figure 5.2).
As discussed in Chapter 1, transitions from one vibrational level to another can be induced
by electromagnetic radiation. For a molecule to change its vibrational state by absorbing
radiation it must (1) change its dipole moment when vibrating and (2) obey the selection
rule: An = ±1. Note that the frequency of the absorbed or emitted radiation for a transition
between levels is !
. — c 1 / h \ /2
(5.47)
5.3 I Ideal Diatomic Gases 153

0(r) for the lowest electronic state

lowest energy level = (l/2)/iv

zero energy •

Figure 5.2 Lowest energy electronic ground state.

Consequently, the vibration spectrum for a diatomic molecule will consist of one line.
Vibration lines typically occur in the infrared at wave numbers (l/k = v e/c\) near
1,000 cm" 1 . Based on the above arguments, the vibrational partition function becomes

5.48)
n=0 n=0
In this case, the summation converges to a result that can be expressed in closed form:
^2, 1
forz-^1, (5.49)
1I -—z 7- 1 x
«=0
and hence
e-hv/2kBT

(5 50)
-
The constraint on convergence of the sum is satisfied for any value of absolute temperature.
The quantity hv/ k^T is larger than 1 at low to moderate temperatures. However, if the
temperature is high enough, hv/k^T may be much less than 1, and because the energy
levels are close together, the summation for gVib can be replaced with an integral:

qwib = • (5.51)
7o hv
This result is identical to the limit of the closed-form relation for qY[\y for hv/k^T —> 0.
Using ^vib(^) we can evaluate the vibrational contribution to the gas internal energy,

v
V1U •* ' "".D-"" rv rri a
\ r\ Q IT 1
(5.52)
oT \ 2 ^vib/J _ i
where the vibrational temperature 0v^ is defined as
_ hv
Vlb (5.53)
" £B '
The vibrational contribution to the molar heat capacity is

NA\ fdUvih\ ... /6U\2 <*"•

(5.54)
154 5 / Ideal Gases

Table 5.1 Rotation, Vibration, and Electronic Constants

for Diatomic Molecules
n
frot #vib
(K) (K) (J/molecule) £e,0
19
Cl2 2 0.351 808 4.029 x 10- 1
H2 2 85.3 6215 7.607 x 10~19 1
h 2 0.054 308 2.497 x 10-19 1
N2 2 2.88 3374 1.589 x 10"18 1
O2 2 2.07 2256 8.363 x 10- 19 3
CO 1 2.77 3103 1.800 x 10~18 1
HC1 1 15.2 4227 7.400 x 10- 19 1
HI 1 9.06 3266 5.128 x 10"19 1

v,vib

N k
AB

room temperature ~ 0(0.1 # v ^

Figure 5.3 Vibrational component of the specific heat for a diatomic gas.

Note that as T -» oo, L^b -> NICBT and cy,vib ~^ ^ A ^ B - The variation of cV,vib with tem-
perature has the form shown in Figure 5.3. Vibration-related constants for some diatomic
molecules are listed in Table 5.1.

The Rotational Partition Function for a Heteronuclear

Diatomic Molecule
Modeling the rotational mode as a rigid rotor, our previous quantum analysis
indicated that the energy levels and degeneracy are

for 7 = 0 , 1 , 2 , . . . . (5.55)
gj = 2 . /
Consistent with the above result, the zero of the rotational energy is taken to be the 7 = 0
state. As discussed in Chapter 1, transitions from one rotational level to another can be
induced by electromagnetic radiation if (1) the molecule has a permanent dipole moment
and (2) A7 = ± 1. The frequency of radiation emitted in the process of changing from level
j + 1 to j is given by

Ve = £z±Lli^ = _A_( 7 - + i) for 7 = 0 , 1 , 2 , . . . . (5.56)

5.3 I Ideal Diatomic Gases 155

This implies that radiation is absorbed at frequencies given by multiples ofh/4n2l, resulting
in a set of equally spaced spectral lines, often found in the microwave region.
Based on the above results, the rotational partition function is given by

7=0 «=0

where
h
(5.58)

is the rotational constant for the molecule. We further define

h2
(5.59)

as the characteristic temperature for rotation. Unlike the vibrational case, the summation for
qYOi cannot be written in closed form. We can replace the summation with an integral if the
energy levels are close together ((£/+i — £j)/kBT <^l). This condition can be equivalently
stated as
2(7 + \)0rot/kBT < 1. (5.60)
This condition is satisfied at moderate temperatures except for large j values. However, the
terms with large j are so small compared to those for low j values that slight inaccuracy
in evaluating them has a negligible effect on the overall result. For high temperatures, we
therefore write
O
(2j + De-./0+iKWr dj= (5.61)
/
Thus for <9rot «C T
Sn2IkBT
(5.62)

The accuracy of this result increases as the temperature increases. Noting that the summation
for qrot can be written as

_ (5.63)
n=0
it is clear that if T is small compared to 0YOt the first few terms should suffice to evaluate qYOt:

qYOt = 1 + 3e~26mt/T + 5e~66mt/T + le~lWrot/T + • •

•. (5.64)
The number of terms shown above is sufficient to determine the sum to within 0.1% for
#rot >0.1T. For 0YOt<0.1T, but not small enough for the integral to give a good approxima-
tion, an intermediate approximation is needed. Using the integral result as the lowest-order
approximation, higher-order correction terms can be generated using a series expansion
technique. The following expansion can be obtained with this method:

1
<7rot = (5.65)
156 5 / Ideal Gases

For most substances, however, 0TOt is very low and the high-temperature approximation will
be accurate. For such gases the rotational contribution to the internal energy is

= NkBT (5.66)
dT
It follows that the contribution to the specific heat is

(5.67)

Another feature of interest is the fraction of molecules in the y'th rotational energy level

N qTOt
This variation exhibits a maximum that can be determined by differentiating with respect
to j and setting the result equal to zero. This yields
1/2
7max — (5.69)

The variation of Nj/N with j for CO at 300 K is shown in Figure 5.4. The plot clearly
indicates that most molecules are in excited rotational levels at ordinary temperatures. Note
that all of the above analysis of the rotational storage mode presumes that we can treat the
atomic nuclei as distinguishable particles. In fact, this is not appropriate for homonuclear
molecules.

The Rotational Partition Function for a Homonuclear

Diatomic Molecule
The wave function of a homonuclear diatomic molecule must have a certain sym-
metry with respect to the interchange of the two identical nuclei in the molecule. The nature
of the required symmetry is dictated by the spins of the nuclei. If the two nuclei have inte-
gral spins, the wave function must be symmetric under interchange of the nuclei (they are
bosons). If the two nuclei have half-integral spins, the wave function must be antisymmet-
ric (they are fermions). Quantum analysis of the effect of these restrictions is somewhat
complicated. However, in the high temperature limit, the net effects are easily quantified.
Of particular note is that the restrictions on the overall wave function link the nuclear and

0.1-

0.08-

0.06-

0.04-

0.02-

0-
10 15 20

Figure 5.4 Variation of Nj /N with j for CO at 300 K.

5.3 I Ideal Diatomic Gases 157

rotational effects in such a way that the combined nuclear and rotational partition function
does not factor into qnuc\qvot. (This is true, in general, for either the boson or fermion case.)
For either the half-integral or integral spin case, it can be shown that for the high temperature
limit the rotational-nuclear partition function reduces to

<7rot,nucl = (2l + I) 2 ( T— J , (5.70)

where / is the nuclear spin quantum number. In the high temperature limit, we can therefore
consider
<7rot,nucl — #nucl#rot> (5.71)
where
tfnucl = (2/ + lf, (5.72)
T
<7rot = (5.73)
26n
For these relations to be valid, T must be greater than about 5# rot. The high temperature
result for the heteronuclear and homonuclear cases are often represented together as

?«* = - ^ - , (5.74)

where as = 1 for the heteronuclear case and a s = 2 for the homonuclear case. Because its
deviation from 1 is linked to symmetry considerations, a s is termed the symmetry number. It
is frequently assumed that the same symmetry number correction applies to the expansion
relation that can be developed for small 0YOt/T:

1 1
1+ 15 315
(5.75)

Note from Table 5.1 that 0YOt for virtually all molecules except H 2 is so low that almost all
circumstances of practical interest correspond to the high temperature limit. For H 2 , two
nuclear spin configurations are possible: (1) parallel nuclear spins, which is termed ortho-
hydrogen, and (2) opposed nuclear spins, which is para-hydrogen. The strong nuclear spin
effect for hydrogen at low temperatures causes the thermodynamic properties of ortho-
hydrogen to be distinctly different from those of para-hydrogen.
Using the general relation above for qTOt, the rotational contributions to U, dy, and S are
determined to be

Urot = NkBT (5.76)

1+ (5.77)
45 (T
i
(5.78)
— In
Before putting together the pieces of the overall partition function, it is useful to reexamine
the electronic contribution. We argued previously that the electronic partition function for
158 5 / Ideal Gases

a diatomic molecule should essentially be of the same form as a monatomic one,

(5.79)

In treating diatomic molecules it is common, however, to take the zero of the electronic
energy to be the separated, electronically unexcited atoms at rest. When the atoms are
brought together to the bonded configuration, the electronic energy decreases (following
the same potential energy curve that governs the vibrational mode). If the depth of the
electronic ground state potential well is £>e then the energy of the electronic ground state
for the molecule is — DQ.
Note that DQ is the energy required to break the molecule into two atoms (i.e., the energy
of dissociation). We also define the quantity Do as the energy difference between the lowest
vibrational state and the dissociated molecule:

-hv. (5.80)

As indicated in Figure 5.5, if the electronic state is altered to the second level, the potential
well in which the nuclei vibrate may, in general, also change. Based on the above-defined
zero point for the electronic state, the electronic partition function is given by

He — <5e,0^ i <5e,l^ i *' ' • ^J.olJ

Having considered each contribution to the total partition function q, we can now assemble
the parts to obtain the harmonic-oscillator-rigid-rotor approximation for q:

= 2
hvi2kKT (5.82)
sh

Figure 5.5 Definition of the electronic ground state.

531 Ideal Diatomic Gases 159

This relation applies provided that #rot <^ T, only the ground electronic state is important,
and the energy zero points are defined as described above. Using this relation for q, we can
obtain thermodynamic properties for the ideal diatomic gas:

(5.83)

phv/lcBT
hv
c v = NA k B (5.84)

27T(mi + m2)kBT\3/2Ve5/2
S =

hv/kBT _e-hv/kBT}
(ehv/kBT _
(5.85)

PV =NkBT. (5.86)

Although beyond the scope of the present discussions, this treatment can be extended to
account for mode coupling, centrifugal distortion, and anharmonic effects. In spite of its
simplicity, the relations obtained from the harmonic-oscillator-rigid-rotor model gener-
ally agree well with measured properties for most diatomic gases. The additional effects
mentioned above are generally important only at very high temperatures.

Example 5.2 A piston and cylinder device contains one kmol of N2 gas initially at T\ =
300 K with initial volume V\ = 0.002 m3. The gas is compressed reversibly and adiabatically
to a volume of 1.0 x 10~5 m3. Determine the final temperature T2 of the gas.

Solution The reversible and adiabatic compression is a constant entropy process.

Setting 5*2 = S\ and using Eq. (5.85) to evaluate the entropy values, we obtain

»'* MM! A2 ) N

hv/kBT{ _e-hv/kBTn

= NkB In ln
h2 N

?e0

Substituting for T\, V\, and V2 and simplifying, we can reduce this relation to

} = 20.66.
160 5 I Ideal Gases

where #vib is the vibrational temperature. #vib = hv/k& = 3,392 K for nitrogen. Iteratively
solving this equation yields

T2 = 2750 K.

5.4 Polyatomic Gases

Treatment of polyatomic molecules makes use of the same analytical tools used to
analyze diatomic molecules. Where advantageous and appropriate, we may use the classical
limit results to evaluate contributions to the partition function. As in the case of diatomic
molecules, the Schrodinger equation that describes the motion of the nuclei in the potential
field of the molecule can be transformed into center-of-mass and relative coordinates. This
allows us to break down the Hamiltonian into a component associated with translation of
the center of mass and a component associated with other internal modes,

H = Htr + // int , (5.87)

from which it follows that

s = etT + fiint, (5.88)

q = qtrqinu (5.89)

where

and M is the total mass of the molecule. We again expect that the density of translational
energy states alone is sufficiently large that the dilute occupancy approximation is valid,
and hence
aNaN
Q= ^ ^ . (5.91)

We further expect that

<7int = <7rot<7vib<7e<7nucl- (5.92)

Each of the n atoms in the molecule requires three coordinates to locate. The total required
for the molecule is therefore 3n total coordinates, which, in the modified treatment, fall into
three groups:

Number of Coordinates To Specify

3 Center-of-mass location
2 or 3 (2-linear, 3-nonlinear) Molecule orientation
3n — 5 or 3n — 6 Relative locations of n nuclei

We will again make use of the rigid-rotor-harmonic-oscillator approximation to allow us to

separately evaluate qrot and q^^. We choose, as the zero of energy, all n atoms completely
5.4 I Polyatomic Gases 161

separated in their ground electronic states. It follows that the energy of the ground electronic
state for the molecule is —D e and the electronic partition function is
qQ = geOeD*/k»T + ••
•. (5.93)
The two or three coordinates specifying orientation of the molecular represent rotational
degrees of freedom. The 3n — 5 or 3n — 6 internal coordinates represent vibrational degrees
of freedom. By selectively defining the internal coordinates, the oscillatory behavior can be
represented as a combination of independent harmonic oscillators. Internal coordinates so
defined are termed normal coordinates. The Hamiltonian can then be written as the sum of
contributions for each oscillator. The total energy is then given by

(5.94)

where
1/2
(5.95)
mT
and a = 3n — 5 for a linear molecules and a = 3n — 6 for a nonlinear molecule. The a
fundamental frequencies Vj can be obtained from a normal coordinates analysis of the
molecule. In practice they are often determined spectroscopically. The oscillatory modes
for a linear triatomic molecule such as CO2 are shown in Figure 5.6.
For a nonlinear triatomic molecule such as water, there are three modes. The modes
for the water molecule are shown in Figure 5.7. Each normal mode of vibration makes an
independent contribution to gVit>- Treating each as a harmonic oscillator yields

(5.96)

(1) symmetric stretch

/^V^/^\ Co\-+- W asymmetric stretch

Q ) (3&4) bending mode

(in two orthogonal planes)

Figure 5.6 Vibrational modes for CO2.

H) (H) (H) (H) (H

(1) (2) (3)

Figure 5.7 Vibrational modes for water molecules.

162 5 I Ideal Gases

where
OyibJ=hVj/kB. (5.97)

The rotational properties of a polyatomic molecule depend on the general shape of the
molecule. If the molecule is linear (e.g., CO2, C2H2) the problem is the same as for a
diatomic molecule. The energy levels are given by

J 21
\ fovj = 0 , 1 , 2 , . . . , (5.55)
8j = 2y + 1 J
where for a polyatomic molecule
n

I = ^2mjsj, (5.98)

with
Sj = the distance of the 7th nucleus from the center of mass of the molecule,
nij•= the massof the 7th atom in the molecule.
In the high temperature limit, the rotational partition function for a linear polyatomic
molecule becomes
lk
BT T
<7rot = 73 = T-, (5.99)
ash2 as6TOt
where

(5.100)

In Eq. (5.99), as is a symmetry number, which is equal to the number of ways that the
molecule can be rotated into a configuration indistinguishable from the original. Classically,
a s is a factor included to avoid overcounting indistinguishable configurations in phase space;
crs = 1 for asymmetric molecules such as N2O and COS, whereas a s = 2 for symmetric
molecules such as CO2 and C2H2.
For a nonlinear molecule, the result depends on the principal moments of inertia IA , IB ,
and Ic. With respect to these, we can define

(5.101)

The simplest case is the spherical top, corresponding to IA = IB = IQ. For this case, the
energy levels and degeneracy are

ej = ^ ± i > , (5.102)

gj = (2j +1)2, y=0,l,2,.... (5.103)

In the high temperature limit, we compute qI0l as

qrot = 1 / (2j + ife-JU+m^iksT djt (5-104)

^s Jo
5.4 I Polyatomic Gases 163

where as is the symmetry number for a polyatomic molecule (the number of ways the
molecule can be rotated "into itself). crs — 2 for H2O, as = 3 for NH3, etc. In the high
temperature limit, the integrand in Eq. (5.104) is small except at j values that are large
compared to 1. The relation for qrot then becomes

qrot = - f" 4fe-J^/2ikBT dj = ^ _ ft* IkBT\ ( 5 1Q5)

With a bit more analytical effort, it can be shown that in the high temperature limit for the
general case of IA ^ IB ^z Ic

_ n^_ ( r3 1/2

#rot — I Q p a
^s \ "rot, A "rot, B "rot, C

Assuming that the nuclear storage mode is always in the nondegenerate ground state,
gnucl = 1 and complete relations for q can be assembled as follows:
for a linear polyatomic molecule:

T
7-1

for a nonlinear polyatomic molecule:

1/2 / 3«-6
%/kBT
#rot,A#rot,fi#rot,C ,
\ j —i /

(5.108)

where the vibrational and rotational temperatures are given by Eqs. (5.97) and (5.100) or
(5.101).
Using the property relations for the canonical ensemble, the following relations can be
derived from the above relations for q:
for linear polyatomic molecules:
3n-5
U 5 e^j/T 1 De
(5.109)
NkBT 2 _ IT 9 u •IT 1 1 T '

3n-5

+E A ] )
5 2 e<U,j/T
Cy b,A
\(9m r
(5.110)
NAkB 2 (p@v\b,j/T 1^2
7=1

\ 3/2
S {(2nMkBT Ve^X f Te
i *•
In
NkB ) N f 1 asemt
3n-5 r Q
)
IT
£e,0, (5.1H)
*-l
PV=NkBT; (5.112)
164 5 I Ideal Gases

Table 5.2 Constants for Polyatomic Molecules

Molecule as (K) (K) (J/molecule) ge>0

18
CC14 12 0.0823 (3) 310 (2), 450 (3), 660, 1120(3) 2.189 xlO- 1
CH4 12 7.54(3) 1870 (3), 2180 (2), 4170, 4320 (3) 2.914 xlO- 1 8 1
CO2 2 0.561 954(2),1890,3360 2.702 xlO- 1 8 1
H2O 2 13.4,20.9,40.1 2290,5160,5360 1.613 xlO- 1 8 1
NH3 3 8.92, 13.6, 13.6 1360,4800,4880 (2), 2330 (2) 2.067 xlO- 1 8 1
NO2 2 0.603 850(2), 1840,3200 1.586 xlO- 1 8 2
SO2 2 0.422, 0.495, 2.92 750, 1660, 1960 1.797 xlO- 1 8 1

for nonlinear polyatomic molecules:

3«-6
u = 3+ 7^ +
De
(5.113)
NkBT 2T

3n-6
Cy
(5.114)

1/2 ^
= ln y }+ln( —
^A
2
* j
(5.115)
/ ^

PV = NkBT. (5.116)
Values of rotational and vibrational temperatures for various molecules are listed in
Table 5.2. Note that when two or more modes have the same 6mi or #vib value, the multiplicity
is indicated in parentheses.

5.5 Equipartition of Energy

We now will examine how energy is distributed among modes of molecular energy
storage for a particular common system type. To do so, we will begin by considering each
molecule as a system in a canonical ensemble. This is consistent with the fact that in a real
system each molecule is confined to a specified volume and collisions with other molecules
essentially keep it in equilibrium with a system offixedtemperature. Suppose the molecular
energy microstate is specified by a relation of the form

(5.117)
/=0

The first term on the right side of Eq. (5.117) is the contribution associated with energy
storage that depends quadratically on a spatial coordinate or generalized velocity £. The
second term on the right side of Eq. (5.117) is the sum total effect of all other modes of
5.5 I Equipartition of Energy 165

energy storage in the molecule. The quadratic £ storage mode is assumed to be independent
of all other modes, from which it follows that
q=q$qa, (5.118)
where
oo

ft = V e~y^f/kBT, (5.119)

*P [-^Si'j/kBT . (5.120)
»'=0 V /=0 /
From the definition of the partition function, it follows that the probability of a particular
microstate /', j ' is given by
5 _ N,,r _ exp(y^,/kBT ELo Sr,,/kBT)
( }
'•'•''
~ "AT ~ q •
To get the probability of a specific %y value, irrespective of /', we sum Pv^i over all /':

»-g/w-£iL,w t .r) (5l22)

i'=0 /'=o ^

Using Eqs. (5.118) and (5.120) and interpreting Py as Ny/N, we can rewrite Eq. (5.122) as

^r = • (5.123)
N ft
Passing to the continuum limit for energy levels close together, t-y passes to § and Ny/N
becomes (l/N)dN/di-. In this limit, the equivalent form of Eq. (5.123) becomes

where the relation for ft can be written in integral form,

poo
ft = / r y ^ r ( / f . (5.125)
Jo
In terms of the continuum formulation, the mean energy of a molecule associated with the
generalized coordinate or velocity £ is given by

= ^f. (5.126)
Jo
Substituting Eq. (5.124) yields

which, upon evaluating the integrals, becomes

<£?) = * ^ . (5.128)
166 5 I Ideal Gases

We have thus demonstrated that each energy mode of a particle that is an independent
quadratic function of one of the coordinates or velocity components of the particle con-
tributes an average energy of k^T/2 per particle to the system. This result is referred to
as the theorem of equipartition of energy because it implies that system energy distributes
itself equally, on the average, over quadratic energy storage modes.
Note that validity of the equipartition theorem requires that (1) the Boltzmann distribution
is valid (the system is in equilibrium), (2) energy storage for each particle is independent
of that for other particles, (3) the spacing between energy levels must be close enough that
the continuum model is a reasonable representation, (4) the total energy of each particle
can be split into energies for storage modes that are independent, and (5) the energy storage
modes considered must be quadratic functions of position or velocity. If some particle
energy storage modes are independent and quadratic while others are not, the equipartition
theorem applies to the quadratic modes, if the other requirements listed above are met.
It follows from the above discussion that if particles in a system have B independent
quadratic energy storage modes, the molar internal energy u is given by

u = NABkBT/2 (5.129)

and the molar specific heat cy is

du

where R = A^A^B is the universal gas constant.

Example 5.3 Assuming that the proper requirements are met, use the equipartition the-
orem to determine cy for a monatomic ideal gas and a diatomic ideal gas at moderate
temperatures.

Solution At moderate temperatures, molecules of a monatomic ideal gas store

energy by translation only. The energy is therefore given by

s = -mu2 H—mv 2 -\—mw 2.

The energy storage thus consists of three independent modes, each of which is a quadratic
function of the velocity component u, v, or w. Thus B = 3 in Eq. (5.130) and cy = 3R/2.
Since dp =dy + R for an ideal gas, it follows that dp = 5R/2 and dp/dy = 5/3. These
results agree well with measured values for monatomic gases such as argon.
Diatomic molecules have the "dumbbell" configuration shown schematically in
Figure 5.8. Each molecule has three independent quadratic translational modes, just as

z i axis normal
to x-y plane

Figure 5.8
5.61 Ideal Gas Mixtures 167

single atom species do. In addition, at moderate temperature, each molecule can store sig-
nificant energy in rotation about the y and z axes. None is stored in rotation about the x axis
because the atomic masses are located right on the axis. The rotational energies associated
with rotation about the two orthogonal axes are quadratic functions of angular velocity
and are independent. Hence there are a total of five independent quadratic energy storage
modes, and the specific heat at moderate temperatures is c v = 5R/2. At high temperatures,
diatomic molecules also store energy in vibrational motion of the bond between the atoms.
As a first approximation, if the bond is considered elastic, the structure stores energy like a
one-dimensional harmonic oscillator. There are actually two independent quadratic storage
modes associated with this motion. One is the potential energy stored by compressing or
stretching the bond. This mode is quadratic in the deviation of the interatomic spacing from
its equilibrium value. The other mode is the kinetic energy associated with the relative mo-
tion of the atoms. This mode is quadratic in the relative velocity of the atoms with respect
to each other. When energy storage in vibration is fully active, the molecule has seven
independent quadratic energy storage modes: three in translation, two in rotation, and two
in vibration. Equipartition then dictates that cy = 1R/2.

Example 5.4 Determine the final temperature of the gas for the compression process
described in Example 5.2 assuming that the modes of energy storage obey the equipartition
theorem.

Solution Classical theory for an ideal gas with constant specific heat predicts
that for a constant entropy process

T2 = Tl( — ) , y = — =
V2) Cy Cy

If we assume that three translational modes, two rotational modes, and two vibrational
modes of energy storage are all fully quadratic and conform to the requirements of equipar-
tition, cv = 1R/2, y = 1.40, and the above relation predicts T2 = 2,498 K for the values of
T\,V\, and V2 specified in Example 5.2. This is about 250 K lower than the prediction of
fully statistical theory. Alternatively we could assume that the vibrational modes do not par-
ticipate to a significant degree with the result that there are only five quadratic modes: three
translational modes and two rotational modes. Then, cv = 5R/2, y = 1.29, and the above
relation predicts T2 = 1,395. This prediction is obviously much lower than the prediction
of the full statistical theory. This is clearly a case where the full statistical theory provides
a better prediction than that provided by a constant specific heat model.

5.6 Ideal Gas Mixtures

To this point in this chapter we have focused on the properties of a system con-
taining a single molecular species. However, many systems of scientific and engineering
interest contain mixtures of ideal gases. Fortunately it is a relatively simple matter to extend
the analysis framework developed thus far to systems containing mixtures. In Chapter 4 we
showed that for a binary mixture the internal energy and entropy are related to the canonical
168 5 I Ideal Gases

partition function as
/am
(4.43)
Q ,a,b

f = JL+lnQ. (4.41)
kB k BT
These relations can be rearranged to obtain the forms

( 5
\ XT I ' '131)
d l
\ / V,Na,Nb

+ kBlnQ. (5.132)
V,Na,Nb

We also showed in Chapter 4 that for a binary mixture

V ;
\NalJ\Nb\J
Substituting the right side of Eq. (4.173) for Q in Eqs. (5.131) and (5.132) and simplifying
by using Stirling's approximation (In AM = N \nN + N) yields the following relations for
U and S:
/ Anns! \ \ /^nn/7,\\
(5.133)
dI
V,Na,Nh \ /V,Na,Nb

S = kBTNa (3(l^a)) + kBNa \nqa - kBNa \nNa - kBNa

d T
\ /V,Na,Nb

+kBNh\nqb-kBNh\nNh-kBNh. (5.134)
V,Na,Nh

We define ila and ub as the internal energy per mole of species a and b, respectively, if they
occupied the full system volume V at temperature T:

ua=kBTzNA[ — , ub=kBTlNA[ — .
dI dI
\ Jv \ Jv
(5.135)
With these definitions, the relation for the mixture internal energy becomes
U = (Na/NA)ua + (Nb/NA)ub. (5.136)
Note that this implies that the total internal energy is equal to the sum of the molar internal
energies for each species, with each being weighted by the number of moles of that species
in the mixture. Using the definition developed in Chapter 3, we can also derive a relation
for the specific heat of the mixture:

Cy =
NA fdU\
(Na+Nb)\dTjVNaM
N. fdHA Nh (Mh\
Na (Na+Nb)\dT JVM
5.61 Ideal Gas Mixtures 169

Denoting the pure component molar specific heats as

V,Na
(w)
\ dT
/V,Nh
(5.138)

we can write the relation for the mixture specific heat in the form
N Nh
(Na + Nb) (Na + Nb)
Thus, the molar specific heat of the mixture is sum of the molar specific heats for the
constituents, with each being weighted by the mole fraction of that species in the mixture.
We also define an entropy per mole for each of the two species at the system volume and
temperature:
Sa = JcBTNA (dllnq"d^'T)]} + kBNA \nqa(V, T) - kBNA InNA - kBNA,
(5.140)

sh = kBTNA[ '"I' + kBNAlnqb(V, T) - kBNAlnNA - kBNA.

(5.141)
In terms of these molar entropies, the entropy of the binary mixture can be written as
S = (Na/NA)[sa - R \n(Na/NA)] + (Nb/NA)[sb - R \n(Nb/NA)]. (5.142)
The equivalence of Eqs. (5.134) and (5.142) can be verified by substitution of Eqs. (5.140)
and (5.141) into Eq. (5.142). The usefulness of Eqs. (5.136), (5.139), and (5.142) is that
they make it possible to use pure gas relations to evaluate the properties of mixtures. Their
forms also clearly indicate how the relations must be extended for mixtures of more than
two ideal gases.

Example 5.5 In a prototype combustion engine, at the end of the combustion process,
the working fluid is a mixture of H 2O, CO2, and N2 gas at a pressure of 600 kPa and a
temperature of 1,800 K. The mole fractions of H2O, CO2, and N2 in the mixture are 0.190,
0.095, and 0.715, respectively. In the engine, this mixture expands until its pressure is
101 kPa. Modeling the expansion as reversible and adiabatic, determine the temperature of
the gas at the end of the expansion process.
Solution The entropy of the mixture can be represented as
S = SH2o(NH2o, T, V) + S C o 2 (^co 2 , T, V) + SN2(NNl, T, V),
where Sn2o is evaluated using Eq. (5.115), SQO2 is evaluated using (5.111), and S^2 is
evaluated using Eq. (5.85). The reversible and adiabatic expansion is a constant entropy
process. Hence, for the mixture S2 — S\ = 0, which can be expressed as

where
), T2, V2) -
r 2 , v 2) -
A S N 2 = SN 2 (NN 2 > T2, V2) — SN 2 (NN 2 > T\,
170 5 I Ideal Gases

Using Eqs. (5.85), (5.111), and (5.115) and the ideal gas law for the mixture (PV = NkBT)
together with appropriate values of M, 6rot, #vib, and <rs for the three gases, we obtain the
following relations:

2290/Tl
\n{T2Px/TxP2) + \ \a(T2/T{) +
2290/7 1

2290/ Tx 5160/T 2 5160/Ti 5360/ T2

5360/7!
£5360/7i _ 1 — InI V 1 - ^ - 2 2 9 ° / r i A 1 - 6>- 5160 / r i
_ £-5360/7 2

_ £-5360/7i

^In(r 2 /7i) + \K(J2PXITXP2)

954/7 2 954/Ti \ 1890/T2 1890/^

+ +
' £954/72 _ ! ~ £954/7! _ - '

3360/T2 3360/Ti \(\ - e'^'12 y(\ - ,

m
~^ £3360/72 _ 1 ^*n/T 7 — '

l_£-3360/72O
1 _£-3360/7i J(

ASN2=NN2kB \-\n(T2/Tx)-

3?>14/T2 3314/Ti fl-.

£3374/72 _ ! £3374/7j _ J \ 1 - £-3374/7,

Per mole of the gas mixture, the number of molecules of each species can be calculated from
the specified mole fractions as NH2o = 0.190NA, NCo2 =0.095N A , and AfN2 =0.715A^A.
Using these relations and the specified Pi, T\, and P2 values, we sum the right-hand sides
of the three species AS relations and set the result equal to zero. The resulting equation can
be iteratively solved for T2. Doing so yields

T2 = 1230 K.

5.7 Chemical Equilibrium in Gas Mixtures

We showed in Chapter 4 that for a binary mixture

^ ) ( — V (4-173)
Na\ \Nb\
5.71 Chemical Equilibrium in Gas Mixtures 111

For a mixture of four gas species a,b,c,d, extension of the analysis in Chapter 4 yields

(5.143)
Na\J\Nb\J \NC\J \Nd\
We will consider the following general gas phase reaction among the four species in the
mixture:
vaA + vbB ^ vcC + vdD. (5.144)
For a mixture in a vessel of fixed volume iield at a specified temperature T, this reaction
governs conversion of species until an equilibrium concentration of reactants and products
is established. Suppose we somehow force the reaction in the forward direction by a small
amount resulting in a change dNa in species a. The reaction (5.144) requires that
Vb Vc VA
7 A T (y I A T TAT U | X T 7 A T k* 7 A T / C *t A ^ \

dNb = —dN a, dNc = dNa, dNd = dNa. (5.145)

va va va
We designate the ratio of the change in the amount of species a to its stoichiometric
coefficient as
dy = — - . (5.146)

Equations (5.145) and (5.146) together require that

dy = dN1 = dN, = _dN, = _dNl
va vh vc vd
Thus the parameter y (=Na/va) can be interpreted as a reaction coordinate, changes in
which indicate the extent to which reactants A and B have been converted to products C
andD.
We showed in Chapter 3 that the Helmholtz free energy change for a multicomponent
system can be written as

dF = -SdT -PdV + ^ UndNq, (5.148)

where the summation is taken over all species rj in the system. In the four-species system
of interest here, rj can be any member of the set {a, b, c, d). Because the system is held at
fixed V and T, Eq. (5.148) simplifies to

dF = fiadNa + nbdNb + VcdNc + fiddNd. (5.149)

Using Eq. (5.145), we can reorganize the right side of Eq. (5.149) to obtain
dF = (vafia + vbnb - VCJXC - vd\xd) dy. (5.150)
For a system held at constant T and V, equilibrium corresponds to a minimum in the
Helmholtz free energy with respect to all possible changes in the reaction coordinate dy.
At the minimum in F, dF = 0. This is true for arbitrary dy if and only if the collection of
terms multiplying dy in Eq. (5.150) is identically zero:
= 0. (5.151)
172 5 I Ideal Gases

Equation (5.151) is a necessary condition for chemical equilibrium in a system held at

constant T and V. The chemical potential for each species in the gas is given by

i ^^(lnG)\ , . 1 W
fir, = -kBT , (5.152)
diy/
V n / NY^,V,T
where r] can be any member of the set {a, b, c, a1}. Treating the system as a mixture of ideal
gases, it follows from the extension of our binary mixture results that
°a \((lb \ ( a
c \ ( a
d \

Na\)\Nb\)\Nc\)\Nd\)'
Taking the natural log of both sides of this relation yields
In Q = Na \nqa -\nNa\+Nb \nqb - ]nNb\ + Nc lnqc - \nNc\ + Nd \nqd - lnNd\.
(5.153)

Using Stirling's approximation: inNl = N\nN + N, this relation can be reorganized to

In Q = Na ln(qa/Na) + Nb \n(qb/Nb) + Nc ln(qc/Nc)

+ Nd ln(qd/Nd) + Na + Nb + Nc + Nd. (5.154)
Noting that qa =qa(V, 7), qb=qb(V, T), qc=qc(V, T), and qd=qd(V, T), we can dif-
ferentiate this relation for In Q to obtain

Na). (5.155)

It similarly follows that

Vb = ~kBT ln(qb/Nb), [ic = -kBT \n(qc/Nc), fid = -kBT \n(qd/Nd).

(5.156)
After substituting and some rearranging, we can write the requirement for chemical
equilibrium (5.151) as

Noting that the number concentration of species rj is p^^ = Nn/ V (in number of species
t] molecules per m 3 ), we can write the above relation in the form

The left side of Eq. (5.158) can be taken as the definition of the equilibrium constant KC(T)
of the reaction

For ideal gas mixtures, it is sometimes convenient to define an equilibrium constant in

terms of the partial pressures of the reactants and products at equilibrium. For the reaction
5.71 Chemical Equilibrium in Gas Mixtures 173

considered here, this equilibrium constant is defined as

Kp(T) =
(P./Po)HP>/Po)*'
where Pv is the equilibrium partial pressure of species r\ and PQ is the standard reference
pressure, which is usually taken to be 1 atmosphere or 1 bar. For each ideal gas species, the
partial pressure is related to the number density of that species as
P, = PN,,kBT. (5.161)
Using Eqs. (5.161) and (5.159), it is easily shown that Eq. (5.160) implies that
,Vc + Vd-Va
v (T\lkBl \
PN,aPN,b
and hence KP(T) can be evaluated as
= (Pc/PoY'(Pd/POrd = (qc/Vyc(qd/VYd
(Pa/Po)v"(Pb/Poy>> (qa/V)v°{qb/Vy> \
(5.163)
For a reaction involving an arbitrary number of reactant (nr) and product (np) species,
v i A i + v 2 A 2 + •••v +
n r A n i •&v[Bi + v'2B2 + •••v'np+Anp, (5.164)
the above analysis can be generalized to obtain

jrm n"L,PJ n"u(qj

K{T) = = (5 165)
-
_ n"L.{PJ

Equation (5.165) or (5.166) provides a theoretical link between the reaction equilibrium
constant and the molecular partition functions for the species in the ideal gas mixture.

Example 5.6 Consider the reaction below involving diatomic gases in the temperature
range 550 K to 1,000 K:
H2 + l2 O 2HI.
Derive a relation for the equilibrium constant for this reaction.
Solution The definition of the equilibrium constant (5.165) dictates that for this
reaction

Kc(T) = =
(qn2/V)(qh/V) q^qi2'
Substituting the molecular partition function relation for diatomic gases (5.82) with appro-
priate constants for the three gases involved, we obtain the following relation:
m
KC(T)=> ™ V^^WWiAd-.
mH2mh J V C H I / 1 -
,m-Do,H2-Do,i2)/kBT)
174 5 I Ideal Gases

Note that this relation specifies the temperature dependence of the equilibrium constant
for this reaction. We have implicitly assumed that translational, rotational, vibrational, and
electronic storage modes are all important.

Additional insight into the application of statistical thermodynamics theory to ideal gases
is provided by the treatment in References [1] and [2].

Exercises
5.1 The O—H bond length in water is 0.96 A and the H—O—H bond angle is 104 degrees.
Lay out a drawing of the water molecule graphically and determine the location of its center
of mass and its three moments of inertia. From these results, verify the values of 0TOt given
for water in Table 5.2.
5.2 What is the most probable value of the rotational quantum number for N2 gas molecules at
300 K? What is the most probable vibration quantum number for this same situation?
5.3 Determine the value of cy for NH3 at 300 K. You may neglect electronic and nuclear storage
effects. Does the value agree with that predicted by the equipartition theorem?
5.4 Use the equipartition theorem to determine the value of cv for a mixture containing half
water and half ammonia vapor by mole at 300 K. You may neglect electronic and nuclear
storage effects.
5.5 Determine the value of dy for a mixture of water vapor and nitrogen gas at 2,200 K. In the
mixture the mole fraction of water vapor is 0.3. Electronic and nuclear storage effects may
be neglected.
5.6 A somewhat more accurate expression for the vibrational energy of a diatomic molecule is

sn= f n + - j hv - xA n + - j hv,
where xe is called the anharmonicity constant. The additional term here represents the first
correction for deviations from strictly harmonic behavior. Treating xe as a small param-
eter, derive relations for P, U, and S that include the anharmonic effect at least to first
order in xe. Specifically explore the nature of such relations in the high temperature limit
(T/0vih = 0(1)).
5.7 A cylinder in a compressor contains one kmol of O2 gas initially at T\ = 300 K with initial
volume V\ = 0.002 m3. The gas is compressed reversibly and adiabatically to a volume of
1.0 x 10~5 m3. Determine the final temperature T2 of the gas assuming that the modes of
energy storage obey the equipartition theorem.
5.8 In a prototype combustion engine, at the end of the combustion process, the working fluid
is a mixture of H2O, CO2, and N2 gas at a pressure of 800 kPa and a temperature of 2,000 K.
The mole fractions of H 2O, CO2, and N2 in the mixture are 0.190, 0.095, and 0.715, re-
spectively. In the engine, this mixture expands until its pressure is 120 kPa. Modeling the
expansion as reversible and adiabatic, determine the temperature of the gas at the end of
the expansion process.
5.9 Using the results of Example 5.6, plot the variation of the equilibrium constant with
temperature for the reaction H2 + I2 4> 2HI over the temperature range from 500 to
1,000 K. Show that the variation is nearly a straight line when plotted as Kc versus l/T.
Exercises 175

(Inr

air to/from shelters

Figure 5.9

5.10 For a manned expedition to Mars, the Brayton cycle system shown schematically in
Figure 5.9 is to be used to provide electrical power and heat. The atmosphere of Mars
is essentially pure CO2 gas at about 1 kPa. In its simplest embodiment, the system would
use CO2 gas from the atmosphere as a working fluid for the system. The gas would be
compressed to fill the system. In its design operating mode, carbon dioxide gas at state 1
would be compressed to high pressure (state 2) and the gas would be heated using a small
nuclear reactor to temperature r 3 . The high temperature, high pressure gas is then expanded
through a gas turbine to pressure P4. The high temperature, low pressure gas is then sent
through a heat exchanger to recover energy to heat structures in which the crew will live. The
working fluid is then recycled to the inlet of the compressor. The system design operating
conditions are summarized below:
Tx = 303 K, Px = 80 kPa, T3 = 1,500 K, P3 = 400 kPa, P4 = 80 kPa,
compressor efficiency = Qi2S — h\)/(Jt2 — h\) = 0.87,
turbine efficiency = {h4 — h3)/(h4s — h3) = 0.91.
For this system design, analyze the following performance issues:
(a) One concern about this system is that by heating the CO2 up to 1,500 K, the gas may
partially decompose into CO and O2 according to the equation

Consider one cubic meter of CO2 gas at equilibrium and assess whether significant
decomposition is likely to occur at 1,500 K. Refer to the De values in Tables 5.1 and
5.2 in formulating your analysis. How hot would the CO2 have to get before 1% of it
has decomposed?
(b) Derive relations for the enthalpy and entropy of carbon dioxide that are valid over the
range of temperatures encountered in this system.
(c) Write one or more computer programs to accomplish the following:

(i) Use the property relations to determine the mass flow rate of CO2 required to
generate 100 kW of electrical power, assuming that the electrical generators are
85% efficient.
(ii) For the resulting flow rate determined in (i) determine the rate of heat delivered
to waste heat recovery unit qh.
(iii) Determine how the efficiency of the system (wt — wc)/qm varies with T3 for T3
values between 1,300 and 1,700 K. Plot the resulting variation and assess the
effect of reducing or increasing T3 on the system performance.
176 5 I Ideal Gases

(d) Compare your results obtained in task (c.i) with those obtained with an analysis for
a constant specific heat and constant ratio of specific heats. Use cP = 37.0 kJ/kmol K
and y = cp/cv — 1.289 (the values for 300 K) in these calculations.
(e) Adding He to the CO2 working fluid may improve heat exchanger performance by
increasing the thermal conductivity of the working fluid. A proposal has been made
to modify the design to a system that runs under the same conditions using a working
fluid that is 50% He and 50% CO2 by mole. Repeat parts (b) and (c) for the system
running on this binary mixture working fluid. Using your results, assess the effect on
performance of changing to a binary He-CO 2 mixture for the working fluid.
5.11 Appendix III contains a copy of the standard psychrometric chart used in analysis of air-
conditioning systems. This chart is constructed for mixtures of dry air (22% oxygen plus 78%
nitrogen by mole) and water vapor. The chart provides values of specific volume (f), specific
enthalpy (h), and relative humidity (0) as functions of humidity ratio (W) and temperature
(T). Units to be used in this problem for these properties are specified in the table below:

Property Units
v m3/(kg of dry air)
h kJ/(kg of dry air)
W (kg of water vapor)/(kg of dry air)

(a) Using results from Chapter 5, show that Dalton's law applies to mixtures of O 2, N2,
and H2O if the mixture behaves as an ideal gas. Dalton's law states that the mole
fraction of any species is equal to the ratio of its partial pressure to the total pressure.
The partial pressure is the pressure the species would have if it occupied the system
volume alone at the same temperature.
(b) Using results from Chapters 3-5, derive relations for specific enthalpy, relative humid-
ity, and specific volume as functions of humidity ratio and temperature. Correct the
enthalpy equation with appropriate constants so that in the limit of zero water vapor
content, the enthalpy goes to zero at a temperature of 0°C and in the limit of no dry
air content, the ratio h/W goes to 2,501 kJ/kg.
(c) Write a computer program to predict h, v, and 0 for specified values of system pres-
sure, temperature, and humidity ratio.
(d) Use the program developed in part (6) to construct a psychrometric chart for 1 atmo-
sphere pressure. Your chart should show lines corresponding to
h = 0, 20, 40, 60, and 80 kJ/kg dry air,
v = 0.8, 0.85, 0.9, and 0.95 m3/kg dry air,
(j) = 20, 40, 60, 80, and 100%.
The chart should span temperatures from 0°C to 50°C and W values from 0 to 0.03.
(e) Repeat (d) at a pressure of 0.687 atmospheres (the pressure at 10,000 feet). Give an as-
sessment of the differences between the charts. Are they large enough to be important?
5.12 A composite system consists of two chambers connected by a short pipe with an insulated
valve in it. Each chamber has a volume of 0.5 m3 and the composite system is insulated so
it is thermally isolated from the surroundings. Initially the valve is closed and one cham-
ber contains 0.5 moles of N2 gas at 400 K while the other contains 0.5 moles of Br2 gas
at 800 K. The valve is opened and the gases are allowed to mix within the composite
system. Determine the final temperature of the gas in the composite system after it reaches
equilibrium. In these circumstances, you may neglect electronic and nuclear storage.
References 111

5.13 Four kmol of hydrogen (H2) gas and 0.1 kmol of nitrogen (N2) gas are fed into an initially
evacuated chamber having a volume of 0.9 m3. The gas in the chamber is held at 400°C by
heating or cooling the chamber walls as required. The gases react according to
N2 + 3H2<S>2NH3.

(a) Assuming the system reaches thermodynamic equilibrium, estimate the concentration
of ammonia (NH3) in the mixture.
(b) If we want to increase the concentration of ammonia in the mixture, should we increase
or decrease the temperature? Justify your answer.
In analyzing this system, take the degeneracy of the electronic ground state to be one.

References
[1] McQuarrie, D. A., Statistical Mechanics, Chapters 5, 6, and 8, Harper and Row, New York,
1976.
[2] Tien, C. L. and Lienhard, J. H., Statistical Thermodynamics, Chapter 7, Hemisphere Publishing
Corporation, New York, 1979.
 CHAPTER 6

Dense Gases, Liquids, and Quantum Fluids

The ideal gases considered in Chapter 5 are arguably the simplest fluids encoun-
tered in real systems. The behavior of dense gases, liquids, and quantum fluids deviates
strongly from that of an ideal gas. In this chapter, we examine how the thermodynamic
framework developed in earlier chapters can be applied to such fluids. The van der Waals
model for dense gases and liquids is explored in detail for pure and binary mixture sys-
tems. In doing so, we demonstrate that the statistical thermodynamic framework provides a
link between microscopic characteristics of the molecules or particles and the macroscopic
behavior of these fluids.

6.1 Behavior of Gases in the Classical Limit

In Chapter 5, we observed that if the temperature is high enough, we can replace
the summation in the definition of the partition function with an integral to obtain the
limiting form the partition function at high temperature. This reflects one of the fundamental
characteristics of quantum theory, which is that classical behavior is attained in the limit
of large quantum number. At high temperature, the average energy per molecule increases
and this implies that the average quantum number for each energy storage mode is higher.
Thus at higher temperatures, the mean behavior of the system is classical in nature.
In the previous sections of this text we have attacked the problem of determining the
partition function by considering the problem from a quantum perspective. We now will
explore how we can determine the partition function in the high temperature limit by
considering the energy storage from a classical viewpoint.
The basic definition of the molecular partition function that we developed earlier can be
written in terms of a summation over all quantum states / :

where /3 = 1/k#T. In the classical limit we expect that the summation can be replaced by
an integral. Here we first consider independent particles for which the energy microstates
are not affected by the presence of other particles in the system. For a single molecule
the energy in a particular microstate is equal to the Hamiltonian evaluated at particular
values of generalized momenta py and position coordinates qy that specify the microstate.
Thus

2,...,pH,q\,q2,...,q!:) ^ {6.2)

where £ is the number of momenta and coordinates necessary to completely specify the
motion and position of the molecule in a particular microstate. Summing over all possible

179
180 61 Dense Gases, Liquids, and Quantum Fluids

microstates, when the microstates are very close together, must, in some sense, be equivalent
to integrating over all possible momenta and coordinate values:

fe dp2 ... dpK dqx dq2...

However, the integration above over the coordinates and momenta results in a dimensional
quantity and q is, by definition, dimensionless. We expect, therefore, that a proportionality
constant is required that renders the expression for q dimensionless. This suggests the
following definition for the molecular partition function in the classical limit:

q ==C
C qqjjj7.
j 7 . •.fe-^(M2,.^M^2,.,^dpi dp2 %%.dp% dh dfa .. . ^ . (6.3)
Note that the integrals are taken over all accessible values of the parameters. The momenta
range from — oo to oo and the spatial coordinates span the volume of space accessible to
the particle.
To determine the proportionality constant Cq, we consider the specific case of the classical
limit for a monatomic ideal gas. In the absence of electronic and nuclear effects, molecules
of a monatomic ideal gas store energy by translation only. The Hamiltonian is therefore
given by

Substituting this relation into Eq. (6.3) yields

qtt = Cqjfj Jjje-rt+K+to^dp; dpy dPz dx dy dz.

In the above equation the spatial integrals are taken over all accessible values of x, y, and
z. The integration over these variables yields the volume of the container,

qtt = CqVjJJ e-^+K+PVW dpx dpy dp7.

The momentum integrals are evaluated from — oo to oo, which results in the following
relation:

qtT = CqV(2nmkBT)3/2. (6.4)

In Chapters 2 and 5, we derived the following relation for the translational component of
the molecular partition function:
Vilnmk^Tf
q«= ^ • (5-5)

Equivalence of Eq. (5.5) with the relation (6.4) obtained from the integral definition of q
requires that
6.1 I Behavior of Gases in the Classical Limit 181

It can be easily shown that if the particle is restricted to translational energy storage in only
two dimensions, consistency between the integral and summation definitions for q requires
that Cq = l/h2. This implies that, in general,

Thus, for the summation and integral definitions to be consistent,

q = jtjjj'' -Je-^^^-^^-^dp, dp2... dp^ dqx dq2

(6.5)

Note that since Planck's constant h has units of momentum (kg m/s) multiplied by length
(m), setting Cq = 1/ h5 makes the integral expression for q dimensionless.
We can extend the results above to a system of N indistinguishable particles using the
Eq. (4.173) obtained in Chapter 4 for the partition function for the canonical ensemble. For
a single-component system this relation reduces to

Q= (6 6)
Tn: -
Substituting the right side of Eq. (6.5) for the molecular partition function yields

For independent particles, the coordinates and momenta in the integrals for each molecule
are independent of one another. The N-fold product of 2£ -dimensional integrals is therefore
equivalent to a single integral over all 2t-N momenta and coordinates:

..J fe
N

n dA/,i
Since the momenta and coordinates are independent, we replace the N sets of § momenta
variables with a single set of N% momenta: px x to px ^ becomes px to p^, p2X to p2^
becomes p^+x to /32£> etc. We similarly replace the N sets of § coordinates with a single
set of coordinates ranging from 1 to N§ . The above relation for Q then becomes

Q = —-— / / / • • •/V^1^^"-1^1'"-'^^^ (6.7)

^ j J J j = l

In the exponential term in the above equation, the sum of the H j for all Af particles is
equal to the total system energy H(px,..., pN^, qx,..., q^^) for the system microstate
182 61 Dense Gases, Liquids, and Quantum Fluids

corresponding to the specified set of pj and qj values:

Q
= N^IJf ••Je-^"-M'•-«»•>
JldPjdtj. (6.8)
J— 1

This relation links the Hamiltonian H of the N molecule system to the canonical partition
function for independent particles in the classical limit. As suggested by the above relation
the system Hamiltonian will, in general, be a function of the position and momenta of all
the particles in the system.
Ideal gases are an important category of substances that can be accurately modeled by
assuming that the particles are independent. Dense gases and liquids do not conform to
this model. The molecules in such systems interact with each other through long-range
attractive forces and short-range repulsive forces. As a result, the Hamiltonian for the
system is a much more complicated function of the positions and momenta of the particles.
Equation (6.8) relates the total system Hamiltonian to the system partition function. Thus
although this equation was derived for independent particles, it can also be applied to a
system of interacting particles if the system Hamiltonian can be computed as a function of
the coordinates and momenta of the particles.
Although we have not proven that Eq. (6.8) is a correct definition for the partition function
in the classical limit for a system of interacting particles, this equation is, in fact, correct for
a system of this type. Although we will not do so here, it is possible to start with the quantum
sum definition for Q and derive Eq. (6.8) as the classical limiting behavior for h —> 0 (see
the discussion in the reference by McQuarrie [1] cited at the end of this chapter).

6.2 Van der Waals Models of Dense Gases and Liquids

Single-Species Systems
In dense gases and liquids, the molecules are close enough that they are constantly
interacting with neighboring molecules. This is fundamentally different from ideal gases in
which molecules travel through space without interacting with other molecules until a brief
(and infrequent) collision occurs. In this section we will first consider a system containing
one species of interacting molecules. To evaluate the partition function for such a system
using Eq. (6.8),

Q=
Nh«llI' •Je-P^-M^JldPjdQj, (6.8)
we must develop a relation for the system Hamiltonian, H. For a system containing in-
teracting particles, we write the system Hamiltonian as the sum of three components, as
indicated below:

> <7AU> • • •. fat). (6.9)

62 I Van der Waals Models of Dense Gases and Liquids 183

The first term on the right side of Eq. (6.9) is the sum of the translational energies of the
molecules, which depends only on the momenta of their centers of mass. The second is
the sum of the rotational energies of the molecules, which depends on rotational momenta
of the molecules. The third term, O^, is the potential energy due to intermolecular force
interactions, which is a function only of the position coordinates of the molecules. Here
we specifically will model the molecules as diatomic, linear polyatomic, or nonlinear poly-
atomic with three different moments of inertia. The translational energy is a function of three
Cartesian momenta and the rotational energy depends on two or three momenta. Hence,
the total number of momenta in the Hamiltonian per molecule, §, is either five or six. The
integral for Q in Eq. (6.8) is over all molecular momenta and coordinates, which are taken to
be independent variables. We can therefore group the integrations into three factors, the first
being integrals over the translational momenta, the second being integrals over the rotational
coordinates and momenta, and the third being integrals over the translational coordinates:

N £-3

n J JJ /=i /=i

e
III'"I ~"* z_
(6.10)

Since the translational Hamiltonian for each molecule is the same and has the usual
quadratic form, the factor inside the first set of square brackets in Eq. (6.10) is easily
evaluated. Doing so, one finds that it equals the translational component of the molecu-
lar partition function divided by V to the Nth power, (qtTms/ V)N = (2TTMkBT/h2)3N/2.
The rotational Hamiltonian is also assumed to be the same for each molecule. It follows
that the factor in the second set of square brackets in Eq. (6.10) can be decomposed into
the product of N sets of integrals, each of which is equal to the rotational component
of the molecular partition function qmt. The factor inside these square brackets therefore
equals qrNoV
The factor inside the third set of square brackets in Eq. (6.10) is the classical configuration
integral Z^:

-JJH' (6.11)

With this definition, the above relation for Q can be written as

Q = (6.12)
N[h3N

As discussed in Chapter 5, grot varies with the molecular structure. For a linear diatomic
or polyatomic molecule
T
(6.13)
184 61 Dense Gases, Liquids, and Quantum Fluids

whereas for a nonlinear polyatomic molecule with three different moments of inertia

nX
X,2 / j X3/2

<7rot = (6.14)
7
rot,m

where

) (6.15)

The relation (6.12) for Q for either case can be written as

Q=
N\hiN 7
rot,m
zN, (6.16)

where § is the total number of translational and rotational energy storage modes and it is
understood that 0TOUm = 0rot for linear molecules.
To complete our model analysis we must evaluate ZN. To do so using Eq. (6.11) we
must develop a relation for the potential energy associated with particle force interactions,
<!>#. In dense fluids like liquids, one approach to dealing with the system potential energy
associated with force interactions among neighboring molecules is to make use of a radially
symmetric distribution function g(r) defined such that

the number density of molecules with centers between \

4nr2(N/V)g(r)dr=
r and r + dr measured relative to a specific molecule /

Qualitatively, the distribution function looks like the curve shown in Figure 6.1. The distri-
bution function can be determined using Monte Carlo simulation methods or experimentally
using x-ray scattering. The product g{r)dr can be thought of as a correction factor that when
multiplied by the system mean density gives the local (time-averaged) density.
We choose any molecule in the bulk fluid as the central molecule shown in Figure 6.2. In
the type of model considered here, the intermolecular potential function 0(r) is idealized
as being spherically symmetric. A spherically symmetric potential is likely to be a good

3-
g(r)

2 -

/ \ — -

i i i

D ID 3D 4D
r

Figure 6.1 Figure 6.2

6.2 I Van der Waals Models of Dense Gases and Liquids 185

Hard Sphere Potential

oo forr<D
0 forr>£>

Figure 6.3

approximation for nonpolar uncharged molecules with a reasonably spherically symmetric

structure. A more general formulation is required for molecules having a permanent dipole
moment or grossly unsymmetric structure or electron charge. For a spherically symmetric
potential function it follows that
/ the intermolecular potential energy between theN
2
Anr (j){r){N/V)g{r)dr = central molecule and other molecules with
y centers at distances between r and r + dr
The intermolecular potential can be determined from quantum mechanical analysis or mod-
eled empirically. Two common models, the hard sphere and Lennard-Jones 6-12 models,
are shown in Figure 6.3.
The total potential energy of the fluid is obtained by integrating over all values of r,
multiplying by N, since any of N molecules may be central, and multiplying by 1/2 so that
each interacting pair is counted only once. The total potential energy is therefore given by

N f
- 4nr2(l>(r)(N/V)g(r)dr. (6.17)
^ JO
Considering Figure 6.1, we can make the following crude, but reasonable approximation
forg(r):
0 forr < D,
g(r) = (6.18)
1 forr>£>,
where D is the molecular diameter. With this idealization, the relation (6.17) for <$>N can
be rearranged to obtain
2nN2
r2(f)(r)dr. (6.19)

Note that in adopting Eq. (6.17) we are implicitly assuming that the intermolecular potential
is radially symmetric. It follows that if we define

av = -In \ r2c/)(r)dr, (6.20)

JD
186 61 Dense Gases, Liquids, and Quantum Fluids

we obtain the following simple relation for <&N:

<$>N = - ^ L . (6.21)

Substituting Eq. (6.21) into Eq. (6.11) simplifies the integration because O^ is independent
of any of the coordinates. The exponential term can therefore be moved outside the integrals:

.J JJ
The multiple integral is equal to the integral of unity over all accessible values of x, y 9 and
z raised to the power N. At first glance it might appear that this multiple integral should
equal VN. However, in dense gases and liquids, the molecules may occupy a significant
fraction of the physical system space. This occupied space is inaccessible to any given
molecule and should not be included in the multiple integration. Consequently the quantity
(V — Nbv)N is a better approximation for the multiple integral, where bv is a constant rep-
resenting the mean volume occupied by a molecule in the system. The resulting relation for
ZN is

ZN = ea»N2/vk»T(V - Nbv)N. (6.23)

Substituting the above relation for ZN into Eq. (6.16), we obtain the following relation
for the partition function:

2
eavN /VkBT

(6.24)

Since most thermodynamic properties are related to the natural log of Q (see Chapter 4) it is
useful to take the log of both sides of Eq. (6.24) and simplify using Stirling's approximation
to obtain
\27tMkBT(V -Nbv)2

Substituting this result into Eq. (4.49), it is easily shown that for the model fluid in our
analysis

P = -»**£--**. (6.26)
V-Nbv V2 K }

The above relation is the van der Waals equation of state discussed previously. Thus, our
model analysis has led us to the partition function for a van der Waals fluid. Now that we
have the partition function, we can generate relations for other properties as well. From
6.2 I Van der Waals Models of Dense Gases and Liquids 187

Eqs. (4.50) and (4.43) and the definition of enthalpy //, one can also obtain the following
relations for our model fluid:

Nbv p \ \2nMkBT(V -Nbv)

In
kBT V-Nbv \2

ln7t+lna] lnf (6 27)

ln7r+lnffs ln (6 2?)
2 J 2 Uot,J VkBT' -
avN2
^ (6.28)

Using Eq. (4.41) we can also obtain the following relation for the entropy:

(630,

Using Eq. (6.28) to eliminate T from the above equation, we obtain the fundamental equation
for a van der Waals fluid:

a Nl
= #*B { - + 1 ) + —^~ ^ { ZTTTJT^ IU+ ~y-

(6.31)

Although it incorporates many idealizations, this model analysis provides a workable frame-
work for predicting properties of dense gases and liquids that relates macroscopic thermo-
dynamic properties to molecular characteristics. Note in particular that it relates the van
der Waals constants av and bv to the physical size of the molecules and the intermolecu-
lar potential function, which characterizes force interactions among molecules. Values of
these constants for selected molecules are listed in Table 6.1. A plot of the fundamental
equation for nitrogen obtained using Eq. (6.31) with constants from Table 6.1 is shown in
Figure 6.4.
The model analysis for a fluid of interacting molecules presented in this section obviously
incorporates several idealizations. It is based upon the central idealization in van der Waals'
doctoral thesis in 1873 [2] that a vapor and a liquid are really just extremes of a continuous
fluid phase. The van der Waals model described here can be improved by increasing the
accuracy of the idealizations in the model. Specifically, more accurate approximations for
the intermolecular potential function and the distribution function g(r) could be incorpo-
rated into the model. For some molecules, an accurate equation of state requires that the
188 61 Dense Gases, Liquids, and Quantum Fluids

Table 6.1 Van der Waals Constants Determined from Critical Data for
selected Fluids^

Molecule 1049 x av (Pa m6/molecule2) 1029 x bv (m3/molecule)

C2H2 (acetylene) 12.2 8.47

NH3 11.7 6.20
CO2 10.1 7.09
CO 4.04 6.54
CF4H2(R134a) 27.7 15.9
H2 0.682 4.40
CH4 6.31 7.09
N2 3.76 6.40
o2 3.78 5.23
H2O 15.2 5.05

Adapted from Wark [5].

fundamental equation for nitrogen (van der Waals model)

Figure 6.4

model include a nonsymmetric intermolecular potential function. Discussions of alternate

intermolecular potential function models can be found in Hirschfelder, Curtis, and Bird [3],
and Tien and Lienhard [4],
The property variations predicted by the van der Waals model exhibit all the features of
those for real liquids and gases. As a predictive tool it is fairly good for low and moderate
gas densities, but it generally becomes increasingly inaccurate as the density approaches
that of the liquid phase. At liquid densities, it is not uncommon for the property predictions
to be off by as much as 100%. The accuracy of the van der Waals property relations can
be improved over a limited range of conditions if the constants are chosen to provide an
optimal fit to data in that range.
6.2 I Van der Waals Models of Dense Gases and Liquids 189

The van der Waals model clearly has limited usefulness as a predictive tool. Yet it
remains valuable because it illuminates the linkage between the molecular characteristics
of the substance and its macroscopic thermodynamic properties, and it does so at high
densities characteristic of liquids and low densities characteristic of vapors. Because it is
applicable to both liquid and vapor phases, in Chapter 8 we will use the van der Waals
model to explore thermodynamic features of phase transitions from liquid to vapor for a
pure substance. We will see that the transformation from a low density to a high density
fluid (or vice versa) is often not a simple continuous transition for real fluids and that the
van der Waals model predicts such behavior.

Example6.1 For nitrogen (N2) 0 ^ = 3.75 x 10~10 m and So = 1.31 x 10~21 J are recom-
mended values for the constants in the Lennard-Jones 6-12 intermolecular potential. Use
these values to predict the van der Waals constants av and bv for nitrogen.

Solution Substituting the Lennard-Jones 6-12 potential into Eq. (6.20) we obtain
the following relation for av:
/>oo / \ 12 / \ 6
av = -In I r24s0 (— ) - (— ) dr.
JD [\ r ) \r )
Execution of the integral yields

Here we take D = a y where the potential becomes strongly repulsive. Substituting a y =

3.75 x 10~10 m and £o = 1.31 x 10~21 J for nitrogen, we obtain

av = 3.86 x 10" 49 Pam6/molecule2.

The parameter bv is interpreted as the volume of space made inaccessible by the presence
of each of the other molecules. We therefore compute this as

bv = 2TTD3/3.

Again we take D = a Lj and substitute to obtain

bv = 11.Ox 10~29 nvVmolecule.

The values of av and bv obtained in this way do not exactly match those in Table 6.1 for
nitrogen. Although the value for av is close, the bv value in the table is only about 60% of
that computed here.

Example 6.2 Using the relations for the van der Waals model, determine the value of the
specific heat at constant pressure for saturated liquid nitrogen and saturated nitrogen vapor
at 100 K and 778 kPa. Compare the results to the recommended values of 64.8 kJ/(kmol K)
for liquid and 41.3 kJ/(kmol K) for vapor.
190 61 Dense Gases, Liquids, and Quantum Fluids

Solution By definition cP = [(dH/dT)P N]N=NA.FOYH = H(V, N, T)mdP =

P (V, N, T), it can be shown by considering changes in H along a line of constant P that

Differentiating Eqs. (6.26) and (6.29) as required and substituting yields

,rIAJ x [ 2av bvkBT 1 / N Ak B \
A
' \ M 1 L(V/NA)2 (V/NA-bv)2\ \V/NA-bJ

To compute the value of dp, the av, bv, and § values for nitrogen from Table 6.1 are used.
We first compute V/N& for the liquid and vapor phases at the specified temperature and
pressure conditions using Eq. (6.26). (These are the lowest and highest of the three solution
values for V/NA.) Doing so, we find that V/NA = 9.99 x 10~29 m3/molecule for the liquid
phase and V/NA = 1.536 x 10~27 m3/molecule for the vapor phase. Substituting into the
above equation then yields

cP = 48.9 kJ/(kmol K) for the liquid

and

cP = 29.7 kJ/(kmol K) for the vapor.

These values are within 30% of the corresponding recommended values cited above.

An Extended van der Waals Model for Binary Mixtures

We will now explore how the type of model discussed above for pure fluids can be
extended to a binary mixture. For a binary mixture containing Na molecules of species a
and Nh molecules of species b, we showed in Chapter 4 that the canonical partition function
is given by
N
/ nN a \ / n b\

If we substitute the right side of Eq. (6.3) for the molecular partition function for each
species, the above relation becomes

" 1=1

1 ^ 1 r r r r
Tj—T 1 / / / / —RfJ /v /v /v
p j
x I I —— / / / • • /• e dp: i .. .dp,; * dq< \ .. .dq\ t (6.33)
Nhl h^b J J I I
In the above relation %a and £/> are the number of momenta and coordinates required to specify
the Hamiltonian of species a and species b molecules, respectively. Since the momenta and
coordinates for each molecule (regardless of species) are independent and we are integrating
6.2 I Van der Waals Models of Dense Gases and Liquids 191

over all possible values, we can combine the two lists into one and reorder them to run from
1 to Nai=a + NbZb- Doing so yields

(6.34)

But when combined, the two summations in the exponential term are just the total Hamil-
tonian for the system

1=1 j=\

Substituting yields

As was the case in our initial treatment of a single-species system above, this relation has
been derived using relations that were derived for particles for which energy storage is
independent of the presence of the other particles. We have arrived at a result that depends
only on the overall system energy storage. To explore the behavior of this mixture, we
will assume here that this relation for Q also is valid for a binary system of interacting
particles.
Continuing the analysis, we write the system Hamiltonian as the sum offivecomponents.
The first two are the sums of the translational energies of the molecules of each species,
which depend only on the momenta of their centers of mass. The second two are the sums
of the rotational energies of the two species, which depend on rotational momenta and coor-
dinates of the molecules. The fifth component is the potential energy due to intermolecular
force interactions, which is a function only of the position coordinates of the molecules.
Here we specifically will model the molecules as diatomic, linear polyatomic, or nonlinear
polyatomic with three different moments of inertia. The translational energy is a function
of three Cartesian momenta and the rotational energy depends on two or three momenta.
Hence, the total number of momenta in the Hamiltonian per molecule, %a or §/,, is either
five or six. The resulting relation for the Hamiltonian is
Na Nb

#sys = 2-/#trans,/(At,/> Py,i> Pz,i) + /2 ^ trans >) (Px,j> Py,j> Pzj)

1=1 ;=1

Na Nb

otj(Pe,n • •
•) + J^ H^Pe^ ''0
7=1
i , Z i , . . . , XNa+Nb, yNa+Nb, ZNa+Nb)- (6.37)

The integral for Q in Eq. (6.36) is over all molecular momenta and coordinates, which
are taken to be independent variables. We can therefore group the integrations into five
factors, the first two being integrals over the translational momenta, the third and fourth
192 6 I Dense Gases, Liquids, and Quantum Fluids

being integrals over the rotational coordinates and momenta, and the fifth being integrals
over the translational coordinates. Thus we have

N h\
7=1

Nb #6-3

w»-
• /
» XNa+Nh, yNa+Nh, ?Na+N

Na+Nh
x II (ixy-, ayj dzj (6.38)

Since the translational Hamiltonian for the molecules of each species is the same and has
the usual quadratic form, the factor inside the first square brackets in Eq. (6.38) is easily
evaluated. Doing so one finds that it equals the translational component of the molecular
partition function divided by V to the Na power: (qtrms/ V)Na = (27tMakBT/ h2)3Na/2. Sim-
ilar reasoning applied to the second integral indicates that it equals {27tM hk^T/h2)3Nh/2.
Since the rotational Hamiltonian for the molecules of each type is also assumed to be the
same, the factor in the third set of square brackets can be decomposed into the product of
Na sets of integrals for each molecule, each of which is equal to the rotational component
of the molecular partition function for species a. The factor inside these square brackets
therefore equals q^a. The same reasoning implies that the fourth group of integrals is q^ ib.
The factor inside the fifth set of square brackets is the configuration integral for the binary
mixture:

h, yNa+Nh, zNa+Nh)/kBT]

dxj,dyjdz (6.39)

With this definition, the above relation for the partition function can be written as

_ (6.40)
6.2 I Van der Waals Models of Dense Gases and Liquids 193

To complete our determination of the partition function, we need to develop a relation for
the configuration integral. To do so, we first define the distribution functions gaa(r), gab(r),
gba(r), and ghb(r) such that

(
the number density of type a molecules with
centers between r and r + dr measured relative
to a specific a molecule
/ the number density of type a molecules with
2
^7zr {Na/V)gah{r) dr = centers between r and r + dr measured relative
y to a specific b molecule

(
the number density of type b molecules with \
centers between r and r -\- dr measured relative
to a specific b molecule )

(
the number density of type b molecules with \
centers between r and r + dr measured relative
to a specific a molecule )

We also define 0 a a (r), 0^(r), and 0a/>(r) as the potential functions representing the force
interaction between two molecules of type a, two molecules of type b, and one type a and
one type b molecule, respectively. As in the pure fluid system considered above, we will
idealize these potential functions as being radially symmetric. It follows that

r
v
/the intermolecular potential energy ^
between a central type a molecule
and other molecules with centers at
\distances between r and r -\- dr )
N
4nr2\(j)hh(r)gbb(r)( - ^ j + (j)ab(r)gab(r) ( y-a ) \dr

/the intermolecular potential energy \

between a central type b molecule
and other molecules with centers at
\distances between r and r + dr )

To get the total intermolecular potential energy for the system we integrate the above
relations from r = 0 to infinity and multiply each integral by the number of corresponding
molecules divided by 2 to properly account for the total effect of the molecules of that type.
194 61 Dense Gases, Liquids, and Quantum Fluids

Adding the results for the two species types together yields the relation
/•OO

o L ~ Vv J \v J\
+
T" / ^^ ^bb^gbb^ ( y" ) + ^b(r)gab(r) I -y j Ur. (6.41)
Defining
/•OO

flWiflfl = - 2 ; r / r2cj)aa{r)gaa(r)dr, (6.42a)

Jo
/o
poo
po
av,ab = -7T / r2cj)ab(r)[gba(r) + g t e (r)]dr, (6.42b)
Jo
r
r2<t>bb(r)gbb(r)dr, (6.42c)
av,bb = -27T /
Jo
we can rearrange this relation to obtain
_ ay,aaN2 _ 2av,abNaNb _ av,bbNl
}
V V V '

Because O is independent of the integration variables, the relation (6.39) for the configura-
tion integral can be written

ZNa,Nb = exp [{a^aaNJ + 2av,abNaNb + aVtbbN%)/VkBT]

X
J J J"'J I I dxj^dyjdzj. (6.44)

The multiple integral is equal to the integral of unity over all accessible values of x, y, and
z raised to the power Na + Nb . In dense gases and liquids, the molecules may occupy
a significant fraction of the physical system space. If we ignored this effect, the multi-
ple integral would equal VNa+Nh . Since this occupied space is inaccessible to any given
molecule and should not be included in the multiple integration, a more accurate expression
for this multiple integral would be (V — Nabv,a — Nbbv,b)Na+N\ where bv^a and bv^ are
constants representing the mean volume occupied by a species a and a species b molecule,
respectively. The resulting relation for Z^a,Nh is

ZNaM = exp [{av,aaN2a + 2av,abNaNh + av,bhN2h)/VkBT] (V - Nabv,a - Nbbv,b)N°+N».

(6.45)

Consistent with our treatment of linear or nonlinear molecules described earlier in this
section, we take
- j
, (6.46)
t,m)a/

Substituting the above relations for ql0^a, <2W,£> and Z^a^b into Eq. (6.40), we obtain
6.2 I Van der Waals Models of Dense Gases and Liquids 195

the following relation for the partition function:

\2nMakBT(V -Nabv,a-
=
h2
27tMbkBT(V -Nabv,a-
X
Nb\J[ h2
(|«-3)/Va/2 itb-5)Nbl2 / T \ V;b-3)Nb/2
N N
a \ (a \ I b \ (a \
&s,a \ V"rot,m)a / <7S b \V (7 rot,m/fe
x exp[(av,aaN* + 2av,ahNaNb + av,bbN2h)/VkBT]. (6.48)
Taking the natural log of both sides and using Stirling's approximation yields
\27tMakBT(y-Nabv,a-.

2nMbkBT(V - Nabv,a -

- lnas,fl + Nb\
2
T
2 \(0rot,m)aJ 2
, av,aaNl + 2aVtabNaNb + aVtbbNl
VkBT
Using Eqs. (4.41), (4.43), and (4.49) with the above relation for In Q, we obtain the following
property relations:

Hf-MH- - Nabv,a -

- Nabv,a -

+ Na | ]nn - \ncrStO \ + Nb\

-^Nu / T \
(6.50)

+ + a bbNl
U = ^NakBT + | ^ f a r - ( " - ^ ^ ' ^ - ), (6.51)
_ (Na+Nb)kjj,T _ aViagNl + 2av,abNaNb + a^^iV^

2 2 V - Nabv,a - NbbVtb
2
av,aaN a+2av,abNaNb
196 61 Dense Gases, Liquids, and Quantum Fluids

Relations for the chemical potential for each species can also be obtained using Eqs. (4.50)
and (4.51).
The van der Waals type model described here is obviously highly idealized. It grossly
simplifies the interaction between the two species of molecules in the binary mixture. For
some systems, this model is too idealized to be of much use as a general predictive tool. It
is presented here primarily because it illustrates how a microscale model of the molecular
interactions can be used to generate property relations for a binary mixture in either a
liquid or vapor phase. Its main value is that it illuminates how the interdependencies among
thermodynamic properties that occur in real binary fluid mixtures are related to microscopic
features of the system. In Chapter 8 we will use this model to explore the thermophysics
of liquid-vapor phase equilibrium conditions in a system containing a binary mixture. This
type of binary mixture model may also be useful as a starting point for development of
empirical tools for predicting thermodynamic properties.
Although it embodies idealized treatments of excluded volume effects and particle force
interactions, the van der Waals model remains attractive because of its simplicity. Over the
past thirty years, there have been numerous efforts to develop augmented or generalized
van der Waals equations of state. These efforts have typically sought to develop improved
treatments of excluded volume effects and molecular force interactions within the van der
Waals framework. This has led to a series of generalized van der Waals models that are
simple and accurate enough to be useful for engineering analysis. Further information on
efforts to develop generalized van der Waals models can be found in references by Kac et al.
[6], Alder and Hecht [7], Rigby [8], Hoover et al. [9], Lebowitz and Waisman [10], Song
and Mason [11], Nordholm et al. [12-14], Hooper and Nordholm [15-17] and Penfold et al.
[18]. The earliest of these efforts focused on equations of state for pure systems, whereas
more recent efforts have included development of a van der Waals model for vapor-liquid
equilibrium in simple binary mixtures. Other approaches to modeling dense gases and
liquids are discussed in the next section.

6.3 Other Models of Dense Gases and Liquids

The Redlich-Kwong Equation of State

Since the late 1940s numerous efforts have been made to develop an equation of
state for pressure having better accuracy than the van der Waals model. In 1949 Redlich and
Kwong [19] proposed an improved equation of state that is substantially more accurate than
the van der Waals equation. This was originally introduced as an empirical refinement of
the van der Waals equation of state relating P,V9 N, and T. To gain a broader perspective
on this model we will reconsider the model analysis presented in the previous section.
The analytical development of the van der Waals model for a pure fluid in the previous
section lead us to Eq. (6.17), which can be written as
OO "I

r2(P(r)g(r)dr\. (6.54)

We subsequently introduced idealizations regarding 0 (r) and g(r) that led to the conclusion
that the factor in the square brackets was a constant for a given fluid. To improve the model,
we now relax this idealization by admitting the possibility that the term in square brackets
6.3 I Other Models of Dense Gases and Liquids 197

is a function of T and density. To do so, we postulate that

/•OO

-lit \ r2(j)(r)g{r)dr = aRy(T)r](V/N), (6.55)

Jo
where aR is a constant for a given fluid and y(T) and rj(V/N) are unknown functions of
temperature and inverse density, respectively. Note that for y = rj = 1 this treatment reverts
to the van der Waals model. Replacing the term in square brackets, we can write the relation
(6.54) for <&N in the form
aRyr]N2
<$>N = . (6.56)

Following the same line of analysis as for the van der Waals model in the previous section,
the relation for the configuration integral becomes

ZN = ea^N2/vk^T (V - NbR)N, (6.57)

where now the mean effective volume occupied by a molecule is denoted as bR. Substituting
into Eq. (6.16) and taking the log of both sides yields

Since P =k^T (d In Q/dV)r,N> differentiating the above relation yields the following equa-
tion of state:

V-Nb* V2 [V N\dVjy l
'
The following relation was proposed by Redlich and Kwong [19] as an improved equation
of state:
p = NkBT *RJV
X2
V -NbR T ' V(V +NbK)'
Comparing Eqs. (6.59) and (6.60), it is clear that the fluid modeled in our analysis will obey
the Redlich-Kwong equation of state if we require that
Y = T~l/2, (6.61)

^ d ^ !
(6.62)
V N\dVJ V+NbR
The differential Eq. (6.62) is easily solved in closed form. In doing so, we set the integration
constant to zero to ensure that as the mean separation distance between molecules becomes
large (V/(NbR) —> oo), the effect of <& on the partition function vanishes so that the
resulting asymptotic behavior of Q is consistent with ideal gas behavior. The resulting
relation for r\ is

UJJln H (6 63)
-
198 61 Dense Gases, Liquids, and Quantum Fluids

Substituting (6.61) and (6.63) into Eq. (6.58) yields a relation for the log of the partition
function for a fluid that obeys the Redlich-Kwong equation of state:

3N\ \27tMkBT(V-NbR)2^

(£3W / T \ a RN fV+NbR\
+ z In I + 3 2 In . (6.64)
2 \0rot,mJ bR k B T / \ V J

Using Eqs. (4.50) and (4.43) and the definition of enthalpy, we obtain the following addi-
tional property relations for a Redlich-Kwong fluid:

fi NbR /3\ \2nMkBT{V -

-NbR \2J I N2/3h2
T
In 7r — In a s | z— In (
\Orot,m )

(6.65)

(6.66)

)NkT In +
2 V-NbR,
(6.67)

Using Eq. (4.41) we can also obtain the following relation for the entropy:

S = Nk

The relations obtained for U and S reflect the fact that this is an empirical model that
has a limited range of applicability. Specifically, the last terms on the right side of these
relations become infinitely large as T —> 0, which is physically impossible. It is expected,
therefore, that the accuracy of this model becomes increasingly poor as the temperature is
reduced at low temperature. Nevertheless, at moderate temperatures, this model is generally
substantially better than the van der Waals model. Recommended constants for the Redlich-
Kwong model for some fluids are summarized in Table 6.2.
Because Eq. (6.66) cannot be solved explicitly for T, it is not possible to combine
Eqs. (6.66) and (6.68) to obtain a closed-form fundamental relation for a Redlich-Kwong
fluid. However, these two relations provide a parametric representation of the fundamental
relation for this fluid. The resulting fundamental equation surface for nitrogen is shown in
Figure 6.5.
6.3 I Other Models of Dense Gases and Liquids 199

Table 6.2 Redlich-Kwong Constants Determined from Critical Data for

Selected Fluids^

Molecule 1048 x aR (Pa m6K1/2/molecule2) 1029 x bR (m3/molecule)

co 18.0 4.93 5
CO2 4.81 4.56 5
CF4H2(R134a) 54.9 11.0 6
CH4 8.97 4.93 6
N2 4.34 4.45 5
o2 4.84 3.65 5
H,0 39.7 3.51 6

Adapted from Wark [5].

fundamental equation for nitrogen (Redlich-Kwong model)

0^0 4?
molar

Figure 6.5

Since the Redlich-Kwong equation was proposed, a number of other P-V-T equations
of state have been developed in an effort to improve the accuracy of property prediction for
dense gases and liquids. Generally these are empirically based and contain more constants
that must be specified to optimize the fit to measured property values. Descriptions of these
other equations of state can be found in the references by Wark [5] and Reid, Prausnitz, and
Poling [20].

Example 6.3 To make popcorn kernels pop, they must be heated to approximately 204°C
(400°F). Because the kernel is hard, there is little, if any, change in volume during heating,
and the rise in temperature of liquid water inside the kernel is accompanied by an increase
in pressure. (11-14% water content is considered optimal for popcorn kernels.) When the
pressure becomes large enough, the hard shell of the kernel cracks, dropping the pressure
200 6 I Dense Gases, Liquids, and Quantum Fluids

suddenly. The superheated liquid inside the kernel thenflashesto vapor explosively, causing
the soft interior of the kernel to swell, producing a "popped" kernel. Estimate the pressure
inside the kernel when it reaches 204° C.

Solution Initially the kernel is presumed to be at room temperature, 20°C, and

atmospheric pressure, 101 kPa. To determine the corresponding molar specific volume, we
use the Redlich-Kwong equation of state (6.60). If we set N = NA, then V is the molar
specific volume, which we designate as v. The equation of state then becomes.

p = NA k B T
v-NAbR
For water, aR = 3.97 x 10~47 Pa m6K1/2/molecule2 and bR = 3.51 x 10~29 m3/molecule.
Substituting these values of the constants and setting P = 101 kPa and T = 293 K, we can
determine the value of specific volume of the liquid in the kernel by iteratively solving the
above equation for 0. (The lowest of the three roots corresponds to the liquid.) For liquid
at these conditions, this process predicts that
v = 0.0243 m3/kmol.
Since the amount of liquid in the kernel and the volume of its shell do not change significantly
during the heating process, we idealize the process as occurring at constant specific volume.
The pressure at 204°C is found by substituting the initial specific volume determined above
into the Redlich-Kwong equation of state:
_ (6.02 x 1026)(1.38 x 10-23)(204 +
0.0243 - (6.02 x 1026)(3.51 x 10~29)
(3.97 x 10-48)(6.02 x 1026)2
~ (204 + 273)1/2(0.0243)[0.0243 + (6.02 x 1026)(3.51 x 10~29)]
= 6.36 x 108 Pa = 636,000 kPa.
Thus, the pressure inside the kernel is estimated to reach more than 6,000 atmospheres
if the kernel is heated to 204°C. The generation of such high pressures is a consequence
of the fact that (dP/dT)^ is large for water at these conditions. By differentiating the
equation of state and evaluating the derivative at 20°C and v = 0.0243 m3/kmol, it is found
that (dP/dT)^ = 3.2 x 106 Pa/K for water. The high pressure that results from heating the
kernel cracks the shell of the kernel, which is the first step in the cascade of events described
above, which ultimately results in a "popped" kernel.

Example 6.4 Using the relations for the Redlich-Kwong model, determine the value
of dp for saturated liquid nitrogen and saturated nitrogen vapor at 100 K and 778 kPa.
Compare the results to the recommended values indicated in Example 6.2.

Solution As noted in Example 6.2, it can be shown by considering changes in

H along a line of constant P that

—\ - f—\
dTjVtN \dVjTtN(dP/dV)T,N\N=NA
63 I Other Models of Dense Gases and Liquids 201

Differentiating Eqs. (6.67) and (6.60) as required and substituting Avogadro's number for
N yields

,
NAbR\'
dH\ ( 1 V
T,N \V-NAbR
_ aRNA T3 / 1 \ _3/
[2 \V + yVA^R; _ 2 V V /_ (^ +
2 VV
+

NA k B T aRRNJ
J aR
aN\
R\
(V ~ NAbR)2 TV22VV
TV 22
(V+N AbR)
(V+Nb) TV2V(V+NAbR)2'
For nitrogen, aR = 4.34 x 10"48 Pam6K1/2/molecule 2, bR = 4.45 x 10"29 m3/molecule, and
^ = 5. We first compute V/NA for the liquid and vapor phases at specified temperature and
pressure conditions using Eq. (6.26). (These are the lowest and highest of the three solution
values for V/NA.) Doing so, we find that V/NA = 9.99 x 10~29 m3 for the liquid phase
and V/NA = 1.536 x 10~27 m3 for the vapor phase. Substituting into the above equation
then yields

cP = 86.0 kJ/(kmol K) for the liquid,

cp = 38.5 kJ/(kmol K) for the vapor.

The liquid value is within 35% of the recommended value for liquid cited in Example 6.2.
The vapor value is within 10% of the recommended value for vapor.

Example 6.5 One kmol of methane gas initially occupies a volume of 0.32 m3 and has
a temperature of 290 K. The gas expands reversibly and adiabatically until its volume is
0.64 m3. Find the final temperature and pressure of the methane after the expansion process.

Solution Evaluating the right side of Eq. (6.68) at the initial conditions, T\ and
V\, and at the final conditions, T2 and V2, subtracting the results for the two cases, and
rearranging yields

^-^^in^K^Blni^^}
•m
, fV2 + NbR\ , aRN , fVi+NbR\
+
2bRTl>2'" \ VJ r {
V v2 J ' 2 fc R r, 3 / 2 - V Vi J )
202 6 I Dense Gases, Liquids, and Quantum Fluids

Because the process is reversible and adiabatic, the entropy of the system is unchanged and
the right side of the above equation must equal zero. Since the system contains one kmol,
we set N to NA. Substituting NA for N and the initial values for T and V, we obtain
0.64-
290J "An»M\032-NAbRj
/0.64 + NAbR\ aRNA
2 + m
m
\ 064
0.64 ) 2^(290)3/2
2^R(290)3/2 V 0.32

For methane, aR = 8.97 x 10~48 Pa m6K ^/molecule 2 , 6R = 4.93 x 10~29 m3/molecule,

and § = 6. Substituting these values and iteratively solving for T2 yields
T2 = 224 K.

The Virial Expansion

In the remainder of this section we will discuss several other approaches to model-
ing dense gases and liquids. An alternative to the van der Waals model for dense gases is the
virial expansion. This approach is based on the observation that the ideal gas equation of
state is expected to be valid in the limit of low density N/V -> 0. As the density becomes
larger, it is expected that the ideal gas relation will require progressively larger correction
for the increasing effects of molecular interaction. The virial expansion is therefore postu-
lated as an expansion in powers of density with the ideal gas relation as the lowest order
approximation:

In the above relation, the B, C, . . . prefactors are referred to as the virial coefficients.
Implementation of this approach requires that we develop a means of determining the virial
coefficients. We can derive a relation for the first virial coefficient as follows. From the
results of Chapter 4 we know that

, • (4-49)
kBl \ dV J T N
From Eq. (6.16) we can see that the V dependence of In Q resides only in the configuration
integral Z^. It follows that

=( (6.70)
k BT V 9V
where

• ^p[

(6.11)
6.3 I Other Models of Dense Gases and Liquids 203

To first order, the potential energy is postulated to be equal to the sum of the potential
energy contributions for each pair of molecules 0/7 in the system,
. N N

Substituting Eq. (6.71) in (6.11) we get

For small deviations from ideal gas behavior, the potential energy interaction between
molecules will be small compared to kBT. Thus, each exponential factor in the integrand
in Eq. (6.72) is very close to one. We therefore define
f(j = e-^/k*T - 1 (6.73)
and rewrite the integrand in Eq. (6.72) as

Because the fij terms are small compared to one we expand the product and neglect quadratic
terms to obtain

=
exp I - - 2^ 2^ I FT ) \l +
? 1^ Zjtftij* • (6J5)

Inserting this result in the relation for ZN yields

z III lix + x " "

III "J\ 2 •
Integration of the first term in the integrand gives simply V N. Substituting the definition of
fij we then obtain

~ z/. (6.77)

Each term in the integrand summation is assumed to depend only on the separation distance
between the /th and y'th molecules. Integration of that term over all the N — 2 other sets of
dxu dyi, and dz\ results in a prefactor of VN~2:

(6.78)
We replace the set of xt, yi, z/ coordinates by polar coordinates in which r is the separation
distance between molecules / and j and integrate immediately over the angles since 0/7
204 61 Dense Gases, Liquids, and Quantum Fluids

depends only on r. Integration over the Xj, yj, zy coordinates, on which 0/y does not depend,
yields another factor of V. The net result is

ZN = VN + J2J2\ [ (e-^(r)/kBT - 1) Anr2 dr) , (6.79)

2 m
/=1 j=l \Jo )
where we have set the upper integration limit to infinity since the integrand vanishes over a
distance that is small compared to the extent of the system. Assuming the potential function
is the same for all molecular pairs, we drop the subscript andfindthat the double summation
is (N2 — N) times the resulting integral:

ZN = VN + VN~l ({N~N)) [ (e-^r)/k»T - l)47tr2 dr (6.80)

Since N is much larger than one, we can further simplify this to

ZN = VN \l + 2n(N2/V) [ (e-W*7 - l)r2dA. (6.81)

Substituting the right side of the above relation into Eq. (6.70), differentiating, and using
the fact that the second term in the square brackets is small compared to one, it follows that
to first order in this small term, the relation for P/k^T is

Comparing this result with the virial expansion (6.69), it is clear that the second virial
coefficient is given by
0
poo
= -2n (e-*^k»T-l)r2dr. (6.83)
Jo
Equation (6.83) provides a link between thefirstmacroscopic correction for nonideal gas
effects and the intermolecular potential function. To develop accurate relations for the co-
efficients for higher order terms in the virial expansion, simultaneous potential interactions
among more than two molecules must be considered and higher-order N/V terms must
be retained throughout the analysis. Development of higher-order corrections associated
with the interaction of clusters of more than two molecules is beyond the scope of this text.
The interested reader can find a further discussion of this type of analysis in the texts by
Hirschfelder, Curtis, and Bird [3], and McQuarrie [1] cited at the end of this chapter.

Other Models of Liquids

A variety of approaches have been proposed to develop microscopic models for
liquids. As in the case of the van der Waals model described above, the usual goals of such
modeling efforts are (1) to combine the model with statistical thermodynamic results to
obtain macroscopic properties and (2) to facilitate an understanding of how microscopic
features of the liquid affect macroscopic properties. An extensive discussion of microscale
models of liquids is beyond the coverage of this text, but a sampling of other types of models
will be described here.
6.4 I Analysis of Fluids with Significant Quantum Effects 205

The virial coefficient approach attempts to model a dense fluid as a very imperfect
gas in which multiple particle collisions are frequent. In the analysis described above we
considered only the first correction in the virial expansion, which depends only on pair
interactions. By extending the analysis to consider higher-order terms and multiple particle
interactions, it is possible to derive relations that are reasonable models of fluid behavior at
higher density. Models of this type have been used in the study of condensation theory, but
efforts to apply them at densities characteristic of liquids have yielded only limited useful
information.
An alternate approach is to view the liquid as an imperfect crystal. Two versions of such
models have been developed. Models of the first type are termed cell theories. Models of
this type treat the liquid as a distorted crystal with a molecule located at or near each lattice
point. The liquid nature of the system is reflected in a loss of long-range order in the system.
Cell theories typically relate liquid properties to the lattice energy and the free volume in
the lattice. The second category of models are hole theories. These attempt to account for
the fact that liquids differ from crystals in that motion of the molecules would leave some
of the lattice sites vacant. A discussion of cell and hole theories and equations of state that
can be derived from them can be found in Hirschfelder, Curtis, and Bird [3], and Hansen
and McDonald [21].
With the rapid increase in computing power over the past two decades, efforts to model
liquids have increasingly focused on molecular simulation tools that computationally ac-
count for energy interactions among all the molecules in a model system. There are now
very broad and active ongoing efforts to develop Monte-Carlo type stochastic simulations
and deterministic simulation methods that can be used to explore the relationship between
molecular characteristics and macroscopic liquid properties. The interested reader may
find useful summaries of the foundations of such techniques in the texts by Hansen and
McDonald [21], Allen and Tildesley [22], and Haile [23].

6.4 Analysis of Fluids with Significant Quantum Effects

In Chapter 4 we simplified the formulation of statistical thermodynamics by con-
sidering conditions in which dilute occupancy applied. We now will reconsider the more
general case of a system in which quantum effects may play an important role. To explore
the thermodynamics of fluids in which quantum effects are important, we will develop our
analysis in terms of the grand canonical ensemble. We consider a binary mixture of particles
of type a and b and define

Ej>(Na, Nb, V) = the energy microstates available to a system with Na particles

of type a, Nt, particles of type b, and volume V,
£// = energy of quantum state V for particles of type a,
ni'(Ej') = the number of type a particles in particle microstate V when
the system is in system microstate Ey,
ek, = energy of quantum state kr for particles of type b,
nic'(Ej') = the number of type b particles in particle microstate k' when
the system is in system microstate Ey.
206 61 Dense Gases, Liquids, and Quantum Fluids

Together, the sets {ni>} and {n^} specify the system quantum state. It follows that
OO

Ey =E /'=0 k'=0
(6.84)

oo
Na
= v^
/'=0
nv, (6.85a)

oo
Nh =y nk>, (6.85b)
kf=0

and the canonical partition function for the system is given by

(6.86)

which, using Eq. (6.1), we can write as

(6.87)
{«/'}*{«*'}* /'=0 k'=0

The {nrY and {riv}* notation in (6.87) implies that we sum over all sets of n\> and n^ that
satisfy Eqs. (6.85a) and (6.85b). By definition, the grand canonical partition function for a
binary system is given by

S(V, T, Ha, fib) = Q(Na,Nb,V,T). (6.88)

Na=0Nh=0

We define

(6.89)
(6.90)

and substitute Eq. (6.87) for Q to get

oo oo

s(V,r, A4a . At6 )=EE

l'=0

{«,'}*

(6.91)
k'=0
6.4 I Analysis of Fluids with Significant Quantum Effects 207

Using Eqs. (6.85a) and (6.85b), we replace the exponents Na and Nb by the summations
over all nr a n d ^

a(V, T, ua, ah) — /^ hal'~° ' y^ exp

Na=0 l'=0
°° oo r OO -i
x ex £ >n k T
E^ ^ E P - E k k'l * -
L J
(6-92)
Nb=0 {nk,}* *'=0

We then reorganize Eq. (6.92) and write the resulting relation in the form
S(V, T, fia, iih) = %a%b, (6.93)
where
oo oo
(6-94)

E I I (^^^ / M B r )^ • (6-95)
Nb=0 {«,/}* *'=0

In the relation for §a we are summing over all values of Na from zero to infinity and,
consequently, each nr ranges over all possible values. As a result, the double summation in
Eq. (6.94) can be reorganized to obtain
«l,max «2,max «3,max OO

s. = EEE-IK^""T- (6-96)
m=o «2=o n3=o /'=o
This result can be written in the form
s r B1 S2/ Br 2
w
l,max ^2,max ^3,max

& = E (x ae - '^ ) E (^«- * )" E (^

/ti=0 n2=:0 «3=0
or in the even more compact form

^=n E (^
OO «/',max

r=o
Note that we have allowed for the fact that each nr may have a maximum value that it may
attain, nrfTaax. We can, however, let w/',max be infinity, when appropriate. Some understanding
of why this transformation of the relation for %a is valid can be obtained by considering
Example 6.6.

Example 6.6 Show that the transformation from Eq. (6.94) to (6.98) is valid for V = 1, 2
with possible n\ and ni values of 0, 1 and 2.

Solution If we let
208 61 Dense Gases, Liquids, and Quantum Fluids

Eq. (6.79) above can be written as

OO OO

Na=0{ni,}*l'=0
For these conditions, §fl is determined as follows:

^ = (£ cw £? E
V Na=0 Na=2

{«/'}* {«/'}*

Note that for each term in the summation, the sets of n\ and n^ must be chosen so n\ +
= Na. Evaluating the sums yields

Factoring the polynomial, we obtain

2 2
$a=(i+?,+s?) ( i + f t + ? D = n E «'')"'' •
/'=l W/ ,=0
£l /kBT
which for fZ/ = Xae- ' has the general form of Eq. (6.98).

The rearrangement to convert Eq. (6.94) to Eq. (6.98) can also be done to the
relation (6.95) for £&. Doing so produces the relation

=n
n
OO k',m

(6.99)
jfc'=0«,/=0

Substituting Eqs. (6.98) and (6.99) into Eq. (6.93), we obtain the following relation for the
partition function:

n
n
V.max k',max

/'=0
£ (6.100)
k'=0
Equation (6.100) is a general result that we can apply to fermions or bosons. Because
these results can be applied to systems whose behavior is dictated by quantum effects, they
are generally termed quantum statistics. Fermi-Dirac statistics are obtained by applying
these results to fermions. For fermions ny and n^ can only be either 0 or 1, since no two
fermion particles may be in the same quantum state. Equation (6.100) for S then reduces
to

E(V, T, fia, fxh) = TT (1 + kae-e'l/kBT) TT (1 + A p ' / % r

/'=0
(6.101)
6.4 I Analysis of Fluids with Significant Quantum Effects 209

Bose-Einstein statistics are based on an absence of restrictions on quantum state occupancy.

For bosons, the nv and n^ can be 0, 1, 2 , . . . and «/',max and nk>ymax are infinite. The relation
(6.100) for S therefore becomes

/'=0 L«/'=o
n
(6.102)
e k T E k T
We will assume that each of the Xae i'l * and A./,6 k'/ ^ terms in the summations is less
than one. Using the fact that for y < 1

(6.103)
m=0

we can then write Eq. (6.102) as

E(V, T, na, nh) = f[ (1 - Xae-e<'/k°T) (for bosons).

/'=0 k'=0
(6.104)
These two fundamental distributions of the statistical mechanics of a binary mixture of
independent particles can be written in the following compact form:

Sf/b = I ] (1 ± Ke ± (6.105)
l'=0 k'=0

where, as the notation suggests, the upper sign applies for fermions and the lower sign
applies for bosons.
In our analysis of the grand canonical ensemble of systems containing a binary mixture,
we showed that

Na = (6.106)
v,T

Nb = k B T (6.107)
v,T
and since

(6.108)
, /'=o /'=0

Nh = (Nb) = I (6.109)
f
k =0

it follows that
oo

y^lnr) =kBT (6.110)

/'=0 V,T

oo

t) = k BT (6.111)
210 61 Dense Gases, Liquids, and Quantum Fluids

The definitions of Xa and Xt, dictate that

'ain3\ /ainS
(6.112)
°^a J v,T a J V,T
91nS // 3
3 In
In 33 \ (6.113)
V,T V d^b ) yj
Using Eqs. (6.112) and (6.113) to replace the derivatives with respect to chemical potential
with derivatives with respect to ka and A,/,, Eqs. (6.100) and (6.101) become

dlnE
(6.114)
a J y,T
/'=0
OO

(6.115)

Substituting the relation (6.105) into Eq. (6.114), we obtain

In
/'=o _l'=0
OO

_/'=0

V,T

o-si'/kBT
(6.116)

Similar substitution of Eq. (6.105) into Eq. (6.115) and evaluation of the derivative yields

™f Xhe ""** \ {6ni)

/ -J \ 1 -4- \ p—e kiIk^T I
V 7
*'=0 k'=0 ^

The right and left sides of Eqs. (6.116) and (6.117) are equal only if the respective terms in
the sums are equal. This indicates that the average number of type a particles in quantum
state / ; and the average number of type b particles in quantum state k' are given by

(6.118)
p-eki/kBT

(6.119)

As discussed in Chapter 4, in the limit of high temperature or low density, the number
of available particle quantum states is much greater than the number of particles. For such
6.4 I Analysis of Fluids with Significant Quantum Effects 211

conditions, the average number of particles in any state is very small (i.e., we have dilute oc-
cupancy). These arguments imply that N / V —> 0 for fixed T or T —• oo forfixedN/V cor-
responds to (ni>) —> 0 and (w^> —> 0. From Eqs. (6.118) and (6.119) it is clear that («//) -» 0
must correspond to Xae~£'f/kBT -» 0 and (n^> ->•0 must correspond to Xbe~£k'/kBT -> 0.
Hence, in the low density/high temperature dilute occupancy limit
(nr)=Xae-£>'/k»T, (6.120)
£k /kBT
(nk,) =Xbe- ' . (6.121)
Summing both sides of Eq. (6.120) over V and both sides of (6.121) over k' yields

Kqa, (6.122)
/'=0 l'=0
oo oo
)=Nb = (6.123)
k'=0 k'=0

Combining Eq. (6.120) with (6.122) and Eq. (6.121) with (6.123) yields
(nr) e-
(6.124)
qa
(6.125)
Nb qb
Taking the natural log of both sides of Eq. (6.105), we obtain
oo oo

In 3 = ± ^ l n {1 ± Xae-£'f/kBT} ± ^ l n {l ± khe-*«/k*T}. (6.126)

/'=0 k'=0

Expanding the summation terms in aTaylor series for small Xae~£''^kBT and small Xbe~£k'^kBT
and retaining only the first term in each, the above equation transforms to

r oo oo
Xb^2e~£k'/kBT
In S == ± \±Ky^e-er/kBT ± (6.127)
[ /'=0 k>=0

which reduces to
lnS == Kqa + hqb (6.128)
or equivalently
(6.129)
Replacing the exponentials with the series representation
oo
X
- (6.130)

Eq. (6.129) can be written in the form

\Xbqb) \ _ v ^ y ^ lka<la \ l kb a
b
a b
Na=0 " ' ) \Nb=0 " ' ) Na=0Nb=0
(6.131)
212 61 Dense Gases, Liquids, and Quantum Fluids

We showed previously that for a grand canonical ensemble of systems containing a binary
mixture of particles

Na=0Nb=0 Na=0Nh=0

Comparing this result with that immediately above, it is clear that in the dilute occupancy
limit

<bKh>
/ nNa \ / Nb \

Equation (6.133) is identical to the relation (4.173) for a system exhibiting dilute occupancy.
Thus, in the dilute occupancy limit, the Fermi-Dirac and Bose-Einstein statistics results
obtained in this section are consistent with the results obtained in Chapter 4 for systems
exhibiting dilute occupancy. Equations (6.105), (6.118), and (6.119) provide a starting point
for the thermodynamic analysis of fluids in which quantum effects are important.

6.5 Fermion and Boson Gases

Having developed a statistical framework for fluids that exhibit quantum effects
in the previous section, we now will explore how quantum effects influence the behavior
of different gases. We will consider two categories of quantum gases. The first is a gas
comprised of indistinguishable fermions. (Recall that fermions have half-integral spin and
no two particles of this type may occupy the same quantum state.) Particles of this type
include electrons and protons.
The second type of gas to be considered here is one containing indistinguishable bosons.
(Recall that bosons have integral spin and have no restrictions on state occupancy.) Particles
of this type that exhibit quantum effects include photons and deuterons (nuclei of the
deuterium atom consisting of one neutron and one proton).
Results obtained in the previous section indicate that for either of these gas types con-
taining one species of particle
00

\±Xe-e*'/kBT)±l, (6.134)
k'=0

where
k = en/k*Tm (6.136)

Where a ± appears in Eqs. (6.134) and (6.135) the upper sign applies for fermion gases and
the lower for boson gases. It follows immediately from the above results and Eq. (4.119)
that for either a fermion or boson gas

(/»V)f/b = ±kBT J2ln 0 ± *-e-St'/kBT). (6.137)

k'=0
6.5 I Fermion and Boson Gases 213

Also, since the sum over all n'k must equal N,

(6.138)

It was shown in the previous section that in the classical (dilute occupancy) limit,
the relations for Fermi-Dirac and Bose-Einstien statistics reduce to those obtained us-
ing classical Boltzmann statistics. Systems that exhibit deviations from classical behavior
are said to be degenerate. Strongly degenerate thus implies large quantum effects, whereas
weakly degenerate implies weak but significant quantum effects.

A Weakly Degenerate Fermion Gas

We wish to consider the behavior of the general relations

N =V — , (6.139)

PV = kBT ^In (1 + Xe-£k'/kBT) (6.140)

k'=0
in the limit of small but significant X for a fermion gas of particles having translational
energy storage only. From our analysis in Chapter 1, we know that for translation of a
particle in a 3-D cubical box, the quantum energy levels are given by
h2
£ = 8 m y 2 / 3 {n\ + n) + n]), nx, ny, nz = 1,2,3, .... (6.141)

This relation indicates that the translational energy levels are close together for large V.
We therefore invoke a high volume limit to allows us to treat s as a continuous variable. As
shown in Chapter 1, the degeneracy of the energy levels given by the above relation is

(6.142)

For the large V limit, we therefore replace the summation over all quantum states with
an integral over all energy values, with a multiplying factor included to account for the
degeneracy of the energy distribution:

/•oo

PV = kBT / g(s) In (1 + Xe-£/k*T) de. (6.144)

Jo
Substituting for g(s) yields

PV = ^M. (^f)V
2 ^f)V2V r
rsxl2\n{\+ke-elk»T)de. (6.146)
4 \ hh/ / Jo
2
J v
214 61 Dense Gases, Liquids, and Quantum Fluids

The integrals in the above expressions cannot be evaluated in closed form. However, we note
from the results in the previous section that the high temperature classical limit corresponds
to \e~Elk*T —>• 0. Since high temperatures correspond to e~£^kBT near one at the lowest
energy levels, this implies that the classical limit corresponds to X -> 0. We therefore expand
the integrands in the above relations as power series in X and integrate term by term to obtain
the following expansions for N/V and P:

A w , 3/2 , (6.147)
A 3 4-f / 3/2

(6.148)

where
/ h2 \ 1/2

(6.149)

An expansion for P/kBT as a power series in Af/V is postulated as having the form
P N

The unknown coefficients are determined by substituting the above expansions for
and N/V and equating coefficients for each power of X. This process yields the following
series:

+ + (6 150)
) -
The above expression is in the form of a virial expansion for pressure. Using a similar
approach, the following expansion can be obtained for the internal energy:

Note that although there are no intermolecular forces, the particles experience an effective
interaction through the quantum symmetry requirement of the N-body wave function. For
fermions, the interaction is essentially repulsive since the first correction tends to increase
the pressure. These results clearly indicate the increase in quantum effects with increasing
thermal de Broglie wavelength A.

A Strongly Degenerate Fermion Gas

We will now consider an ideal gas of indistinguishable fermions at low temperature
and/or high density. Analysis of this type of gas is useful as a model of the valence electrons
in a metallic crystal. The electron gas model is a reasonable approximation because although
the electrons do interact with nuclei in the metal, the interaction is due to Coulombic forces
that are so long range that any individual electron experiences a nearly uniform electronic
potential within the bulk metal. The success of the model as an analytical tool stems from
the fact that at least some of the macroscopically observable properties of the metallic
6.5 I Fermion and Boson Gases 215

crystal are due more to quantum statistical effects than to the details of electron-nucleus
and electron-electron interactions.
For the fermion case, Eq. (6.120) dictates that the average number of particles in the ^th
quantum state is given by

Using the definition (6.136) of X, we can rewrite this relation as

(nk<) = (6.153)
\
The energy quantum states are presumed to be dictated by the particle in a 3-D box solution.
Assuming V is large, we again treat the energy s as a continuous variable and convert the
above relation to an energy distribution

f{e) = / ,., _, (6.154)

J

where f(s) is the average occupancy of a microstate with energy e. (Note that for fermions,
the occupancy must be either 0 or 1.) As illustrated in Figure 6.6, in the limit of T —> 0,
/x —>• /xo and the distribution equals one for s < //o, and it equals zero for s > /xo. This implies
that at T = 0 , all the states with s < /xo are occupied and those with £ > /xo are unoccupied.
The 3-D particle-in-a-box solution dictates that the degeneracy for the particle transla-
tional energy levels is given by Eq. (6.142). For electrons, the degeneracy g*(e) is actually
a factor of two larger than that for a simple particle because the electron can be in one of
two spin states (±1/2) for each translational state. Thus

g*(e) de = An (^] Vsl/2 ds. (6.155)

At T = 0 , we can integrate the distribution over all energies to get N, the total number of
particles in the system:

N - f
(6.156)
Rearranging yields the following relation for

(6 157)
-

n0
Figure 6.6
216 6 I Dense Gases, Liquids, and Quantum Fluids

H0/kBT=100
H0/kBT=l0

Figure 6.7

For an electron gas in a metal, /xo is termed the Fermi energy. The value of this energy
is typically 1 to 5 eV or 0.2 to 0.8 fJ. It can be seen in Figure 6.7 that even for room
temperature levels where JJLQ/ k^T is on the order of 100, the distribution / (s) is still virtually
a step function. Thus, in cases like this, where the Fermi temperature 6? = /xo/ kB is large
compared to the system temperature T, f(e) is well approximated by the step function
1 for s <
(6.158)
0 for s >
Since the Fermi temperature for most metals is on the order of 20,000 K, the step function
is a good approximation for most temperatures encountered in engineering applications.
For low to moderate temperatures (T/0p -> 0), the occupancy distribution can therefore
be used to determine the internal energy and pressure of the gas as follows:
poo /2m\3/2 f^°
Uo= f(e)g\e)ede = 47t[--)
2 V s3/2ds, (6.159)
Jo \h ) Jo
poo
P0V=kBT / f(e)g*(e)]n(l+ke-e/k»T)de
Jo 3 2
/
£/B
-}' \n[\-\-Xe )ds. (6.160)
= 4»f=£j V
Evaluating the integrals and using the definition of the Fermi energy yields
Uo = iN/xo, (6.161)

(6.162)
5\ V
Uo and Po are the zero point energy and pressure, respectively, for a fermion gas. Equa-
tion (6.161) implies that at low temperature, the electron "gas" in the metal contributes
nothing to the heat capacity since (dU/dT)y -> 0 there. Equation (6.162) indicates that
quantum effects cause the pressure to approach a finite limit as T —>•0. With a bit more
mathematical effort, the following expansion for U at low temperatures can be obtained:
Sir2 (kBT
= U0 (6.163)
12 V Mo
6.5 I Fermion and Boson Gases 217

which implies that at low temperatures the molar specific heat for the electron gas varies
linearly with temperature:

^ ) (6.164)
Mo

A Weakly Degenerate Boson Gas

For a weakly degenerate ideal boson gas, the following previously obtained forms
of Eqs. (6.122) and (6.123) apply:
OO

PV = - £ B r ^ l n ( l - keek'/k*T), (6.165)

V (6.166)

Note that these relations are valid only if the summands are finite continuous functions for
all values of energy s. To avoid singularities, we must require that

0 < X < e£o/kBT,

where £Q is the ground state energy. This assures that Xe~£k'^kBT is less than one for all states
except the ground state where it may be close to one. We again invoke the idealization
that for large V the energy is virtually a continuous variable, replace the summation over
all quantum states with an integral over all energy values, and use the degeneracy for the
3-D particle-in-a-box solution. In doing so, we separate the ground state term from the
summation so that we can avoid large values of the integrand when Xe~£o/kBT is close to
one. We also redefine the zero energy as corresponding to the ground state so that £0 = 0.
The resulting relations for N/V and PV are

N X 7 r / 8 m \ 3 / 2 f00 Xsl/2e-£/kBT ds
V V(l-X)
3/2
/

For the weakly degenerate case, X is small compared to one and for large V the first term on
the right side of the above expressions can be neglected. Following the line of analysis as
for the weakly degenerate fermion gas to evaluate the integrals, we can obtain the relations
fort/ and P:
N A 3 / A 2
^ (6.169)

(6-170)

Thus the first correction for quantum effects for a weakly degenerate boson gas is the same
as that for a weakly degenerate fermion gas but with opposite sign.
218 61 Dense Gases, Liquids, and Quantum Fluids

A Strongly Degenerate Boson Gas

Equations (6.167) and (6.168) also apply to a strongly degenerate boson gas in the
limit of large V. For convenience, we rewrite those relations in the form
N IN(X)
V A3
+ - (6.171)
( 1 - •A.)
> IP(k) In •X)
v . (6-172)

where

(6.173)

is. (6.174)

Equation (6.135) indicates that for a boson the average number of particles in state k! is

{nk>) = ——Tjry. (6.175)

This implies that the average number of particles in the ground state (e 0 = 0) is

(n0) = -^—. (6.176)

1 — A
For the number of particles in the ground state to be finite, it is necessary that 0 < k < 1.
Note that no similar restriction on A applies to a fermion gas. We can combine Eqs. (6.171)
and (6.173) to generate an equation of state if we can solve Eq. (6.171) for X in terms of the
other variables. Over the range 0 < X < 1, IN(X) is a bounded monotonically increasing
function of X. The limit of /^(A) at X = 1 is also well defined, with INW being equal to
2.612. Equation (6.171) can be rewritten as

i_. (6,77)

Analysis of this equation reveals that the right side grows monotonically with X for
0 < X < 1, and hence A3N/V increases monotonically with X. For A3N/V < 2.612,
X is small and the second term in Eq. (6.177) can be neglected for large V. X is then given
to good accuracy by the root of

IN(X) - A3— = 0 for A 3 A7V < 2.612. (6.178)

For A3N/V > 2.612, X is close to one and we can use the value of 1^ at one to approximate
the solution for X as

U 1
for ^ / y > 2.612. (6.179)

The condition A3N/V = /^v(l) = 2.612 is a special condition for a gas of this type in
6.5 I Fermion and Boson Gases 219

the sense that it marks a transition between high-A and low-A. behavior. By definition
h2 y/2N
A"- =
V l27TmkBT\ V
and A3N/V is therefore a function of temperature for fixed density. It follows that for
low temperatures, A3N/V is large and Eq. (6.179) provides an approximate relation for X.
Substituting this relation into Eq. (6.176), we can obtain a relation for the mean occupancy
of the ground state. Doing so and neglecting terms of order V~l compared to one for large
V, we obtain

(6.180)

Defining a temperature To as the temperature at which A3N/V = 1^(1) it follows that

2 V2
Nf h \ _I
N
V \27tmkBTo) ~
This relation can be used to eliminate IN (1) in Eq. (6.180). After rearranging, the resulting
form of this relation becomes
3/2
(6.181)
To)
Note that the definition of To implies that the transition condition A3N/V = 2.612 corres-
ponds to T = TQ. Temperatures above To correspond to A3N/V < 2.612 and the value
of X determined from Eq. (6.178) will be less than one and will decrease rapidly toward
zero as T increases. This observation and Eq. (6.176) implies that {no) —> 0 as T increases
above TQ. Thus when T ^> 7b, the number of molecules in the ground state is essentially
zero, with bosons distributed smoothly over the many quantum states available to each.
Equation (6.181) implies that, as the temperature is lowered past 7Q, the occupancy of the
ground state abruptly increases to a significant level, with the occupancy increasing until, at
T = 0, all the bosons are in the ground state. This trend is illustrated in Figure 6.8. The pref-
erence of the bosons for one state out of the many available to them that is exhibited at T = To
is analogous to the phase transition from a gas to a condensed phase. Because of this analogy,
this abrupt enhanced occupancy of the ground state is termed Bose-Einstein condensation.

1.5-,

0 0.5 1 1.5

Figure 6.8
220 61 Dense Gases, Liquids, and Quantum Fluids

With a bit more effort, equations for P/kBT and the internal energy can be obtained
for temperature ranges above and below To. Doing so yields the following relations for the
internal energy:

f/ = f(§)*BrG&)/,w forr>r 0)
N
l(i)*Br(sk)/,(D forr<r0,
where Ip{\) = 1.342. Differentiating these relations and taking TV = NA yields the fol-
lowing expressions for the molar specific heat:

cv ((^)(^W(!)^ 0

To determine cy for a given temperature in the low temperature range, Eq. (6.179) must
first be used to determine X for the specified temperature. The integrals I N (X) and IP (X)
can then be evaluated and substituted into Eq. (6.183) to determine cv.
The above relations indicate a variation of cy with T that exhibits a peak with a discon-
tinuous slope at T = 7b- The variation of cy for helium 4 at its lambda transition point is
similar to the trend indicated by Eq. (6.183). Calculation of T$ for helium yields 3.14 K,
which is comparable to the lambda transition temperature of 2.18 K. The lambda transi-
tion is not a purely Bose-Einstein condensation phenomenon, but quantum statistical effects
appear to be an important contributing factor in this behavior of helium at low temperatures.
Although the occurrence of Bose-Einstein condensation at low temperatures was pre-
dicted in the 1920s, experimental observation of this phenomenon in a low temperature gas
has only recently been achieved. Creation of Bose-Einstein condensate by cooling rubidium
gas was achieved by scientists at the Joint Institute for Laboratory Astrophysics in Colorado
in 1995. These investigators used a combination of laser cooling and evaporative cooling in
a magnetic trap to cool rubidium atoms in the gas to less than 100 nK. This extraordinarily
low temperature must be achieved to create a collection of atoms in the ground state with
the lowest allowable energy for the atoms in the magnetic trap. Atoms in the ground state
are confined by the magnetic field to a region near the center of the trap. Since the energy,
and hence the momentum, are fixed for atoms that drop down to the lowest level, the region
they occupy must be of finite size to satisfy the uncertainty principle.
The behavior of atoms in the Bose-Einstein condensate observed in these experiments
can be thought of as being analogous to that of photons in a laser beam. In a laser beam
the wave behavior of all the photons is the same. In the Bose-Einstein condensate, the
quantum mechanical wave functions that describe physical characteristics of the atom, such
as position and velocity, are all the same. The coalescence of the particle wave function into
a single wave is the essential feature of Bose-Einstein condensation. Since the experiments
that created Bose-Einstein condensate in rubidium gas, other investigators have succeeded
in creating Bose-Einstein condensate in lithium and sodium.
The macroscopic properties of Bose-Einstein condensates are largely unknown at this
point, but it is anticipated that they may exhibit behavior that is much different from classical
fluids. It may be possible that matter waves of a Bose-Einstein condensate can be reflected,
focused, diffracted, and modulated in frequency and amplitude. A Bose-Einstein condensate
is expected to have compressibility like a fluid, but the quantum coherence may give the
fluid other unusual characteristics. There has been some speculation that the viscosity of
Exercises 221

the condensate will be extraordinarily low, making it a sort of superfluid. Exploration of the
properties of the condensate is expected to be one of the main objectives of future research
in this area. Further information on recent experimental investigations of Bose-Einstein
condensation in low temperature gases may be found in the references by Anderson et al.
[24], Andrews et al. [25], and Cornell and Wieman [26].

Exercises
6.1 For methane (CH4) au = 3.82 x 10"10 m and s0 = 2.04 x 1(T21 J are recommended values
for the constants in the Lennard-Jones 6-12 intermolecular potential. Use these values to
predict the van der Waals constants av and bv for methane. How do the resulting values
compare to those in Table 6.1?
6.2 For oxygen (O2), use the tabulated values of av and bv in Table 6.1 to estimate the depth of
the potential well in the interaction potential £0 and the approximate effective diameter of
the molecules D = ou •
6.3 The chemical potential of a pure van der Waals fluid as a function of temperature, volume,
and number of molecules is given by Eq. (6.27). Use this relation with Eq. (6.26) to plot
the variation of fi/k^T as a function of pressure for one kmol of water at 100°C. In your
plot, show the variation for pressures between 10 kPa and 500 kPa.
6.4 Demonstrate directly using results of Section 6.2 that G = JJLN for a pure van der Waals
fluid.
6.5 A system containing one kmol of dense fluid a executes a change of state from V\, T\ to
V2, T2 resulting in a change in internal energy (U2 — Ui)a. Another system containing one
kmol of dense fluid b also executes a change of state from V\, T\ to V2, T2 resulting in a
change in internal energy (U2 — U\)b. The sizes of the species a and b molecules are the
same but for the molecular interaction potentials, the depth of the potential well SQ> is larger
for species a than it is for b. Which do you expect to be larger, (U2 — U\)a or (U2 — U\)bl
Justify your answer.

6.6 Derive relations for the chemical potentials \xa and /x& as functions ofT,V,Na, and Nt, for
the van der Waals binary mixture considered in Section 6.2.
6.7 Derive the equation for cP = [{dH/dT)PyN]N=NA cited at the beginning of Example 6.3.
(Hint: Consider changes of H along a line of constant P and assume H = H (V, N, T) and
P = P(V, N, T) as in Eqs. (6.26) and (6.29).)
6.8 Use the van der Waals model to predict the molar specific heat cP for saturated liquid oxygen
and saturated oxygen vapor at 104 K and 352 kPa. Compare the results to the recommended
values of 54.4 kJ/(kmol K) for liquid and 33.6 kJ/(kmol K) for vapor.
6.9
(a) Beginning with the differential relation

use the defintion of the specific heat at constant pressure and the Maxwell relation

dP
J T,N \" A / P,N
222 61 Dense Gases, Liquids, and Quantum Fluids

to derive the following general relation for the differential change in enthalpy for a
system of fixed mass:

= cPdT+ \v-T — dP.

(b) Use the result of part (a) to derive the following relation for the Joule-Thomson
coefficient:

(c) For many throttling flow processes, enthalpy is constant while the pressure drops. The
Joule-Thomson coefficent indicates whether the corresponding temperature change
is positive, negative, or zero. Consider a throttling process involving nitrogen at an
initial temperature of 140 K. Use the van der Waals model results for nitrogen to
determine whether the throttling process will increase or decrease the temperature for
the following two initial pressures: 200 kPa and 4,000 kPa.
6.10 Oxygen gas is compressed reversibly and adiabatically from an initial condition of 20° C
and 101 kPa to a final pressure of 5,000 kPa. Determine the final temperature and the work
required (enthalpy change) per kmol of gas using (a) the van der Waals model and (b)
relations for an ideal gas with constant specific heats. For part (b) use the equipartition pre-
diction (assuming rotation and translation storage only) for oxygen: dp = 29.1 kJ/(kmol K).
Compare the results for the two models and comment on the reasons for the differences.
6.11 Nitrogen gas expands reversibly and adiabatically from an initial condition of 20° C and
8,000 kPa to 4,000 kPa. Determine the final temperature and the enthalpy change per kmol
of gas using (a) the Redlich-Kwong model and (b) relations for an ideal gas with constant
specific heats. For part (b) use the equipartition prediction (assuming rotation and translation
storage only) for nitrogen: cP = 29 A kJ/(kmol K). Compare the results for the two models
and comment on the reasons for the differences.
6.12 Use the equation for dp cited at the beginning of Example 6.3 and the relations for a van
der Waals fluid to calculate and plot the variation of cP with temperature for nitrogen gas at
10 MPa between temperatures of 150 and 300 K. Compare your results to the predictions
of equipartition theory.
6.13 Use the equation for cP cited at the beginning of Example 6.3 and the relations for a van
der Waals fluid to calculate and plot the variation of cP with temperature for methane gas
at 9 MPa between temperatures of 200 and 400 K. Compare your results to the predictions
of equipartition theory.
6.14 In Chapter 3, we obtained a relation for dp — cv that can be written in the form

CP ~Cy =

where aT is the isothermal compressibility and fiT is the volume coefficient of thermal
expansion defined as

For oxygen at a temperature of 90 K, use the van der Waals relations obtained in Section 6.2
to compute and plot the variations of otj and (dp — cv)/N^k^ with v between v = 0.010
References 223

m3/kmol and v = 10.0 m3/kmol. Assess and comment on the physical significance of the
trends in the plots.
6.15 The so-called square-well interaction potential is defined as

oo for r < a,
(p{r) — ^ — £Q for a < r < Xa,
0 for r > Xa.

(a) Determine a relation for the second virial coefficient as a function of temperature for
molecules that obey the square-well potential defined above.
(b) For nitrogen, recommended values of the constants in the square-well potential are

o = 3.28 x 10- 10 m, X = 1.58, e0 = 1.31 x 10~21 J

Using these values with the relation determined in part (a), plot the variation of B with
temperature between 100 and 400 K.
6.16 The hard-sphere interaction potential is defined as

r oo for r < a ,
10 for r > cr.

(a) Determine a relation for the second virial coefficient for molecules that obey the
square-well potential defined above. How does B vary with temperature for this case?
(b) For methane a value of 3.36 x 10~10 m is recommended for a. Use this value and the
result of part (a) to determine the percent correction to PV/NIc^T provided by the
second virial coeffcient for methane at T = 200 K and P = 101 kPa.
6.17 Determine the next term in each of the series (6.150) and (6.169) for P / &B T. Based on your
results, assess the symmetry of quantum effects for fermions and bosons to order (N/V)3
in these expansions.

6.18
(a) An application requires use of gaseous nitrogen (N2) at 77 K and 76 kPa. Determine
whether quantum effects will introduce a significant departure from ideal gas behavior
for this system.
(b) Hydrogen is a possible alternative to the use of nitrogen in this system. Assess the
importance of quantum effects for gaseous hydrogen (H2) under the same conditions.

References
[1] McQuarrie, D. A., Statistical Mechanics, Chapter 12, Harper and Row, New York, 1976.
[2] Van der Waals, J. D., On the Continuity of the Gaseous and Liquid States, edited by J.S.
Rowlinson, Studies in Statistical Mechanics, Vol. XIV, Elsevier Publishing, New York, 1988.
[3] Hirschfelder, J. O., C. F., Curtis, and R. B., Bird, Molecular Theory of Gases and Liquids, J.
Wiley & Sons, New York, 1954.
[4] Tien, C. L. and J. H., Lienhard, Statistical Thermodynamics, Chapter 9, Hemisphere
Publishing Corporation, New York, 1979.
[5] Wark, K., Advanced Thermodyanmics for Engineers, Chapter 5, McGraw-Hill, New York,
1995.
224 61 Dense Gases, Liquids, and Quantum Fluids

[6] Kac, M., G. E., Uhlenbeck, and P. C , Hemmer, "On the van der Waals Theory of the
Vapour-Liquid Equilibrium I. Discussion of a One-Dimensional Model," /. Math. Phys.,
4: 216-228, 1963.
[7] Alder, B. J. and C. E., Hecht, "Studies in Molecular Dyanmics VII. Hard Sphere Distribution
Functions and an Augmented van der Waals Theory," /. Chem. Phys., 50: 2032-7, 1969.
[8] Rigby, M., "The van der Waals Fluid: a Renaissance," Q. Rev. Chem. Soc, 24: 416-32, 1970.
[9] Hoover, W G., G., Stell, E., Goldmark, and G. D., Degani, "Generalised van der Waals
Equation of State, /. Chem. Phys., 63: 5434-8, 1975.
[10] Lebowitz, J. L. and E. M., Waisman, "Statistical Mechanics of Simple Fluids: Beyond van der
Waals," Physics Today, March, 1980, pp. 24-30.
[11] Song, Y. and E. A., Mason, "Statistical Mechanical Theory of a New Analytical Equation of
State,"/. Chem. Phys., 91: 7840-53, 1989.
[12] Nordholm, S. and A. D. J., Haymet, "Generalized van der Waals Theory I. Basic Formulation
and Application to Uniform Fields," Aust. J. Chem., 33: 2013-27, 1980.
[13] Nordholm, S., M., Johnson, and B. C, Freasier, "Generalized van der Waals Theory III. The
Prediction of Hard Sphere Structure," Aust. J. Chem., 33: 2139-50, 1980.
[14] Nordholm, S., J., Gibson, and M. A., Hooper, "Generalized van der Waals Theory VIII. An
Improved Analysis of the Liquid/Gas Interface,"/. Stat. Phys., 28: 391-406, 1982.
[15] Hooper, M. A. and S., Nordholm, "Generalized van der Waals Theory II. Quantum Effects on
the Equation of State," Aust. J. Chem., 33: 2029-35, 1980.
[16] Hooper, M. A. and S., Nordholm, "Generalized van der Waals Theory IV. Variational
Determination of the Hard Sphere Diameter," Aust. J. Chem., 34: 1809-18, 1981.
[17] Hooper, M. A. and S., Nordholm, "Generalized van der Waals Theory XII. Curved Interfaces
in Simple Fluids,"/. Chem. Phys., 81: 2432-8, 1984.
[18] Penfold, R., J., Satherley, and S., Nordholm, "Generalized van der Waals Theory of Fluids.
Vapour-Liquid Equilibria in Simple Binary Mixtures," Fluid Phase Equilibria, 109: 183-204,
1995.
[19] Redlich, O. and J., Kwong, "On the Thermodynamics of Solutions. V. An Equation of State.
Fugacities of Gaseous Solutions," Chem. Rev., 44: 233, 1949.
[20] Reid, R. C , J. M., Prausnitz, and B. E., Poling, The Properties of Gases and Liquids, 4th ed.
McGraw-Hill, New York, 1986.
[21] Hansen, J. P. and I. R., McDonald, Theory of Simple Liquids, 2nd ed. Academic Press,
Orlando, FL, 1986.
[22] Allen, M. P. and D. J., Tildesley, Computer Simulation of Liquids, Oxford University Press,
Oxford, UK, 1989.
[23] Haile, J. M., Molecular Dynamics Simulation: Elementary Methods, J. Wiley & Sons, New
York, 1992.
[24] Anderson, M. H., J. R., Ensher, M. R., Matthews, C. E., Wieman, and E. A., Cornell,
"Observation of Bose-Einstein Condensation in a Dilute Atomic Vapor," Science, 269:
198-201, 1995.
[25] Andrews, M. R., C. G., Townsend, H. -J., Miesner, D. S., Durfee, D.M., Kurn, and W, Ketterle,
"Observation of Interference between Two Bose Condensates," Science, 275: 637-41, 1997.
[26] Cornell, E. A. and C. E., Wieman, "The Bose-Einstein Condensate," Scientific American, 278:
40-5, 1998.
 CHAPTER 7

Solid Crystals

Chapter 7 demonstrates the application of statistical thermodynamics theory to

crystalline solids. Because of its relevance to electron transport in metallic crystalline solids,
the electron gas theory for metals is also described in this chapter. This chapter provides
only an introduction to the microscale thermophysics of solids. Readers interested in more
comprehensive treatments of solid state thermophysics should consult the references cited
at the end of this chapter.

7.1 Monatomic Crystals

Our objective here is to use statistical thermodynamics tools to evaluate thermo-
dynamic properties of solid crystals. Our first goal is to derive a relation for the partition
function Q. In doing so, we will specifically consider the structure of a monatomic crystal.
One approach is to model the crystal as a system of regularly spaced masses and springs as
indicated schematically in Figure 7.1. The springs represent the interatomic forces that each
atom experiences. The mean locations of the masses are at regularly spaced lattice points.
Actually, each atom sits in a potential well whose minimum is at a lattice point. The
potential well for each atom is usually very steep. Each atom vibrates about its equilibrium
position with a small amplitude, which suggests that we can work with a Taylor series
representation of the potential valid near the equilibrium point.
Rather than working with the potential for a single atom, we will consider the potential for
the crystal as a whole, which we will designate as <£>. The total potential energy is a function of
the displacement of all N atoms in the crystal from their equilibrium position at the minimum
of the potential well. The displacement of each atom has three Cartesian coordinates and
the total potential is therefore a function of all 3N such coordinates. Designating each of
the coordinates as £•,it follows that
(7-1)
Expanding O in a Taylor series for small displacements yields
3N 3N

(7.2)
Since O is a minimum when the §,- = 0, the first derivatives are all zero, and we can write
l 3N 3N
= O(0, 0,..., 0) + - J2 J2kiJ^j +••
•, (7.3)

where

225
226 7 / Solid Crystals

Figure 7.1

are force constants. The important results here are that:

(1) O is a quadratic function of the displacements.
(2) The minimum O is a function of the lattice spacing, which is a function of the
density N/V. To emphasize this, we write this term as <£(0, . . . , 0, N/V).
(3) This expansion represents a system of coupled harmonic oscillators. The coupling
is in the cross terms §/§y where / ^ j .
Analysis of a large number of coupled harmonic oscillators is a difficult problem. How-
ever, for a polyatomic molecule, we previously argued that the complex vibrational motions
of the molecule could be decomposed into a set of independent harmonic oscillators by in-
troducing normal coordinates. Noting that a crystal containing N atoms can be considered
to be just a large polyatomic molecule, this implies that we can also decompose the vibra-
tions of the crystal into a set of independent harmonic oscillators by applying such a normal
coordinate analysis.
Considering the 37V degrees of freedom for the crystal, three are associated with trans-
lation of the whole crystal and three more are concerned with rotation of the whole crystal.
The remaining 3N — 6 degrees of freedom are attributable to vibration. Thus, if we applied
a normal coordinate analysis, we would obtain 3N — 6 vibrational frequencies vy, where

1 / k- 1/2
= —1^^) , 7 = 1,2 3N-6. (7.5)
27T\mTjJ
Note that &y is an effective force constant and mYj is an effective reduced mass for the 7 th
vibrational mode. The mYj and kj are expected to depend on the atomic mass, the bond
force constants, and the equilibrium lattice spacing (~(V/N) 1^3). For a given crystal type,
it follows that the VjS are dependent only on density (N/V).
Since the crystal does not translate or rotate, the complete partition function is given by
3N-6
Mb,/. (7.6)
7.11 Monatomic Crystals 227

where gVib,y is the vibrational partition function associated with the 7 th vibrational fre-
quency. Note that there is no factor of N! in the denominator of this relation. Since each
atom stays nominally at a fixed lattice point, we could, in principle, label each point and
thus distinguish individual atoms. It is therefore not necessary to divide by AM to avoid
including contributions of identical configurations more than once.
For a harmonic oscillator with a specified characteristic frequency v, we have already
shown in our treatment of diatomic gases that
e-hv/2kBT

Using this result for each vibrational mode, we have

3N6 -hv,/2kBT
£
7= 1
Since there are 3N vibrational characteristic frequencies, it is plausible to assume that they
are essentially continuously distributed. We therefore introduce a distribution function g(v)
such that g{v)dv equals the number of normal frequencies in the interval from v to v + dv.
Since N is of the order of 1026, the number of vibrational degrees of freedom is essentially
3N. Taking the log of both sides of Eq. (7.8) for Q yields

For the many closely spaced frequencies Vj, we replace the summation by an integral by
treating v as a continuous variable and multiplying by g(v) dv:
<D(0, . . . , 0 , N/V) po
-ln£ = Z/C
#B^ Jo
JO L B^ J

(7.10)
where /•OO

/ g(v)dv = 3N (7.11)
Jo
Jo
since there are a total of3N normal frequencies.
If we can determine g(v), we can develop a relation for Q and subsequently calculate
the thermodynamic properties of the crystal. In fact, using the relation for U derived for the
canonical ensemble and the above relation for In Q, it can easily be shown that
e~hv/k*T
j
poo
2 2 hv/k T
cv = Nk
NAkB // (hv/k(h/kT)
BT) e- » (l - e-hv/k*Ty2g(v)dv. (7.13)
Jo
The main issue, then, is how to evaluate g(v). If each of the N atoms of a crystalline solid
behave as harmonic oscillators with three degrees of freedom (each having two modes of
energy storage), classical (equipartition) theory predicts that each atom contributes 3 x 2 x
(1/2)&B = 3&B to the specific heat:
cv = 3NAkB = 24.9 kJ/(kmol K) = 6 cal/(mole K). (7.14)
228 7 / Solid Crystals

This result is known as the law ofDulong and Petit. It is in good agreement with the mea-
surements for many crystalline solids at high temperatures, but it fails at low temperatures.
As T —> 0, dy —• 0 proportional to T 3 , for common metallic solids such as aluminum, gold,
and silver. Einstein's model described in the next section was the first to predict a deviation
from the equipartition prediction due to quantum effects as T -> 0.

Example 7.1 Recommended values of the specific heat for copper and aluminum at 25°C
are 0.396 kJ/kg K and 0.900 kJ/kg K, respectively. Compare these values to the predictions
of the law of Dulong and Petit.

Solution The law of Dulong and Petit requires that for copper
cv = cv/M = 3NAkB/M = 3(6.02 x 1026)(1.38 x 10~23)/63.5
= 0.392 kJ/kg K,
whereas for aluminum
cv = 3NAkB/M = 3(6.02 x 1026)(1.38 x 10~23)/27.0
= 0.923 kJ/kg K
The values predicted by the law of Dulong and Petit are within 3% of the recommended
values.

7.2 Einstein's Model

Einstein was the first to propose a theoretical explanation for the low temperature
variation of cy for crystalline solids. His explanation was based on the quantum theoreti-
cal arguments used earlier by Planck in his treatment of electromagnetic radiation. As we
have done above, Einstein argued that the crystal could be treated as a set of 3N indepen-
dent harmonic oscillators. He further postulated that all the harmonic oscillators had the
same characteristic frequency vE. In our formulation this is equivalent to assuming that the
distribution function g(v) is a delta function with one frequency
g(v) = 3N8(v - VE). (7.15)
Substituting this relation into the above equation for cy, after a little manipulation, we can
write the result in the form

where
0E = hvE/kB. (7.17)
Note that vE and hence 0E (the Einstein temperature) are presumed to be characteristic
values for the particular solid. The variation of the specific heat with temperature predicted
by the Einstein model is plotted in Figure 7.2.
Agreement of Einstein's model with data is qualitatively good. However, there is a
problem with its prediction of the specific heat variation at low temperatures. As T -> 0,
7.3 I Lattice Vibrations in Crystalline Solids 229

1.0

0.8

0.6

0.4

0.2

0
0 0.5 1.5

Figure 7.2

the above relation for cy varies as

(7.18)

This is not consistent with data for monatomic crystals, which imply that dy -» 0 propor-
tional to T3 as T —> 0. To avoid this deficiency, a better treatment of energy storage in the
crystal at low temperatures must be included in the model. We will examine the features of
energy storage in a crystal in detail in the next section.

7.3 Lattice Vibrations in Crystalline Solids

The Einstein model in the previous section grossly idealizes the nature of vibra-
tional motions in the solid in that it assumes that all oscillations have the same frequency.
In real crystals, the frequencies of lattice vibrations can span a wide range. To place the
energy storage of the crystal in proper perspective, in this section we will examine some
of the basic features of lattice vibrations in a crystal. This will provide a foundation for
development of an improved relation for the specific heat in the following section.
The structure of the crystal is specified in terms of its lattice and its basis. The basis is
one or more atoms with specific spacing. The lattice is a periodic array of points in space.
The structure of the crystal is generated by repeatedly hanging the basis at each lattice point.
The smallest basis that will generate the crystal structure is termed the primitive basis. A
cell defined by the position vectors of atoms in the primitive basis is a primitive cell for
the crystal. Possible types of primitive cells include simple cubic, body-centered cubic,
face-centered cubic, and hexagonal. These are shown schematically in Figure 7.3. The
body-centered cubic structure is characteristic of chromium, molybdenum, and tungsten.
Crystalline forms of carbon (diamond) silicon, tin, gold, silver, copper, and sodium chloride
have a face-centered cubic structure. In a compound such as sodium chloride, the basis
contains one atom of each species. For pure materials such as diamond and silicon, the
basis for the face-centered cubic structure also contains two atoms. Crystalline forms of
zinc, magnesium, and titanium exhibit a hexagonal close packing structure.
There are many interesting and important features of the crystal structures described
above. However, the main point of interest to us here is that the structure is spatially
periodic. We therefore seek to determine how such a periodic structure stores energy as
230 7 / Solid Crystals

simple body-centered face-centered

cubic cubic cubic hexagonal close packing

Figure 7.3

7+1

mean position
of atom

Figure 7.4

oscillatory motion of the atoms about their mean lattice locations. To explore the basic
features of such lattice vibrations we will specifically consider a crystal with a monatomic
basis and focus on vibrational behavior in one dimension. The situation of interest is shown
schematically in Figure 7.4. Here we assume that the forces exerted on each atom are
linear in the displacement of the atom relative to its neighbors. Thus, for the 1-D motion
represented in Figure 7.4, the force on atoms in plane j is given by
fj = - Hj) + j-x - §,-), (7.19)
where C/ is the force constant between adjacent atoms. It follows that the equation of
motion for atoms in plane j is

ni- (7.20)
dt2
where m is the mass of the atom. We expect solutions of this equation to have sinusoidal
spatial and time dependence, implying that the solution is of the form
$j=$oe-iiiKa+a*). (7.21)
In the above relation, to and K are the angular frequency and wavenumber of the wave,
respectively. These are related to the frequency v and wavelength A, of the wave as
to = 2TTV, (7.22)
K = 2TT/\. (7.23)

Substituting the displacement relation (7.21) for the terms in Eq. (7.20), we obtain, after
a little rearranging, the dispersion relation relating the frequency and wavenumber for the
7.3 I Lattice Vibrations in Crystalline Solids 231

(ACflm)

3i I a

Figure 7.5

waves:
CO = (2Cf/m)(\ - cos Ka). (7.24)

A plot of co verses K for — n/a < K < n/a is shown in Figure 7.5. For K values beyond the
range shown in this figure, the variation of co with K repeats with period 2n/a. Although
the wave solution of the displacement amplitude mathematically varies in a continuous
way with z, only its magnitude at the planes corresponding to integer multiples of a is of
physical relevance. Using the solution (7.21), it is easily shown that the ratio of displacement
amplitudes at adjacent planes is

-iKa (7.25)
=e

Thus, the displacement ratio for adjacent atoms is periodic in K with period In/a, and all
physically distinct variations of the displacement differences are represented by wavenum-
bers in the range — n/a < K < n/a. This range of K, which represents the entire range
of possible wave motions for the atoms in the crystal, is referred to as the first Brillouin
zone. Note that if K is outside the first Brillouin zone, it is always possible to convert it
to a physically equivalent value K inside the zone by subtracting an appropriate integer
multiple of 2n /a .A value of K outside the first Brillouin zone reproduces the lattice wave
motion of the corresponding K within the zone. This periodic dependence of frequency
on wavenumber for the lattice wave motion is distinctly different from wave motion in an
elastic continuum.
Another noteworthy feature of the lattice wave motion represented by Eq. (7.21) is the
group velocity obtained by differentiating the dispersion relation
dco
Va = (7.26)
~dK'

Using (7.24) and (7.26), we obtain

]/2
Vg=(^-) cos(Ka/2). (7.27)
V m)
232 7 / Solid Crystals

The group velocity is obviously periodic in K. In the limit of the wavelength being long
compared to a, Ka is small and cos(Ka) is well approximated by 1 — (Ka)2/2. Substituting
this approximation for cos(Ka), we find that as Ka —> 0 Eq. (7.24) becomes

co2 = (Cf/m)K2a2. (7.28)

Taking the cosine factor in Eq. (7.27) to be one for small Ka, the group velocity is given by
1/2
(7.29)

Combining Eqs. (7.28) and (7.29) to eliminate a yields

vg = — (7.30)

for the small Ka limit, which is consistent with the theory of elastic waves in a continuum.
It should be noted that in a three-dimensional lattice, waves of three different polarizations
are possible: oscillation in one of two orthogonal directions normal to the direction of
propagation (transverse waves) and oscillation in the direction of propagation (longitudinal
waves). In general there will be a separate dispersion relation for transverse and longitudinal
wave modes in the Brillouin zone.
If the basis for the crystal lattice contains two or more atoms, multiple dispersion relations
linking frequency and wavenumber in the Brillouin zone can exist for each polarization.
When there is ionic bonding of different atoms in the lattice, as in a NaCl crystal, the atoms
carry opposite charges and vibrations of the lattice may be induced by the electric field of a
light wave. A branch of the dispersion relation that represents lattice vibrations that respond
to this type of excitation is termed an optical branch. A branch that corresponds to waves
generated by purely mechanical interaction is termed an acoustic branch.
The discussion above provides only a brief introduction to the physics of lattice vibra-
tions. Several important features are evident, however. First, as postulated in the Einstein
model in the previous section, the lattice waves are quantized. The energy of a given mode
of vibration with angular frequency co is taken to be that for a simple harmonic oscillator

IV / l\hco
,7.31,

When a mode is occupied to quantum number n, the mode is interpreted as being occupied
by n pseudoparticles of vibrational energy that are termed phonons.
Given the dispersion relation behavior described above, it is also clear that lattice vi-
brations will occur over a range of frequencies dictated by the dispersion relations within
the first Brillouin zone. They will not have a single frequency, as postulated in the Einstein
model. In general, prediction of the distribution of frequencies g(y) discussed in the previ-
ous section is not a simple matter. To do so we must consider the features of oscillations in
the lattice in detail. However, we have also demonstrated that in the limit of the wavelength
being long compared to the interatomic spacing a, the wave behavior for the acoustic branch
approaches that for waves in a continuum. In the next section we will use this information
to construct the Debye model for energy storage in a crystal, which improves upon the
Einstein model described in the previous section.
7.4 I The Debye Model 233

7.4 The Debye Model

In quantum theory, the energy level of a harmonic oscillator is directly proportional
to its frequency. Because the lower energies are most populated at very low temperatures,
the inadequacy of Einstein's model at low temperatures is likely due to inaccurate treatment
of low frequency (or long-wavelength) modes. The theory of the heat capacity of crystals,
later developed by Debye, treats the long-wavelength frequencies of the crystal in an exact
manner and consequently predicts the correct low temperature variation of c v.
Debye reasoned that modes having wavelengths that are long compared to the atomic
spacing do not depend strongly on the detailed atomic-level structure of the solid. He further
argued that as a consequence, long-wavelength modes can be calculated by assuming that
the crystal is a continuous elastic body. This is a viable approach because, as we noted in the
previous section, the lattice vibrations approach the behavior of waves in a continuum in the
long-wavelength limit. The Debye theory treatment of the distribution of mode frequencies
includes calculation of the modes as standing waves within the extent of the crystal. We
therefore let the imaginary part of

A = Ae/(k"r-ft>0 (7.32)

represent a wave of amplitude A traveling through the crystal in vector direction k with
frequency co = 2nv. Here k is called the wave vector, and its magnitude is 2TV/X. The
velocity of this wave is given by v = co/\k\ = vX. A standing wave can be obtained by
superimposing two waves traveling in opposite directions. Doing so yields

A = 2Aeikr cos cot. (7.33)

To make this a confined standing wave, we require that the imaginary part of the above
relation vanish at the edges of the crystal. If we take the crystal to be a cube of length L,
this requirement becomes
kxL = nxn, (7.34)
kyL = ny7t, (7.35)
kzL = nzn, (7.36)
where
nx,ny,nz = 1,2,3,.... (7.37)
The magnitude of the wave vector for any combination of ns is therefore given by

(7.38)

Since we expect the ns to typically be large, the number of standing waves with a wave
vector of magnitude less than |k| is essentially equal to one eighth of the volume of the
sphere defined by the above equation (see our earlier quantum analysis of energy level
degeneracy for a particle in a 3-D box in Chapter 1):
W V|k|3 ^ ^
7T
234 71 Solid Crystals

The number between |k| and |k| -h d\k\ is obtained by differentiating this relation:

(7.40)

We can convert this result to a frequency distribution g(v) dv by using the relation
v = vw/X = i;w|k|/27T (7.41)
and the equivalent differential form

d\k\ = — dv, (7.42)

where vw is the wave velocity. Substituting (7.41) and (7.42) into (7.40) yields
4nVv2
g(v) dv = — dv. (7.43)

This relation accounts for one type of wave. As we noted in the previous section, two types
of waves are possible: (1) transverse waves in which the medium vibrates perpendicular to
the direction of propagation (i.e., the direction of |k|) and (2) longitudinal waves in which
the medium vibrates in the same direction that the wave propagates.
Since transverse waves can vibrate in two normal directions, there are three wave po-
larizations possible for the medium: two transverse waves and one longitudinal wave. All
three of these waves make a contribution equal to g(y) to the overall distribution func-
tion g(v). Combining the three contributions yields the complete expression for the Debye
approximation to g(v) dv:

( 2 1\
_ + ~)47tVv2dv, (7.44)
W V\ )
where vt and i>i are the transverse and longitudinal wave velocities, respectively. Defining
a mean wave velocity as
1 / 3
2 1\
± + L) (7.45)

we can write the relation for g{v) dv in the form

g(v) dv = — \ ~ v 2 dv. (7.46)

This relation is exact in the limit of low frequencies where we expect that crystal behaves
like a continuous elastic body. Although this relation is not expected to be accurate at high
frequencies, the Debye theory uses this distribution for all the normal frequencies. However,
the total number of frequencies must equal 3N. Debye therefore defined a cutoff frequency
vD SO that the integral of g(v) dv from 0 to vD equals 3N:

/ g(v)dv= — — v 2dv = 3N. (7.47)

Jo Jo K
7.4 I The Debye Model 235

Executing the integral for the above g(v) relation yields

yD = (^-] vm. (7.48)

\AV J
The Debye theory frequency distribution can then be stated as

{ 0
^-v2dv forO < v < vD,
for v > VD-
(7.49)

This distribution can be inserted into the previously derived relations for thermodynamic
properties of the crystal. Doing so yields the following relation for the specific heat:

/6>D\~3 f°D/T ri*e"

cv=9NAkB(^-) / —J——dn, (7.50)
V T ) Jo (<?v - 1)
where 0D is the Debye temperature defined as

0D = Y±- (7.51)

The integral cannot be evaluated in terms of simple functions and must be evaluated nu-
merically. It is traditional to define the Debye function £>(£) as

so that the heat capacity is given by

cv=3NAkBD(T/6D). (7.53)
Tabulations of the Debye function are available and it can be evaluated using any number
of numerical methods. Values of #D that give the best overallfitto experimental cv data are
listed in Table 7.1 for a variety of monatomic solids.
The limiting behavior of the Debye dy relation can be evaluated as follows. At high
temperatures, the upper limit of the integral in the cy relation becomes very small. Since
we then need to evaluate the integrand for small rj, we can use an expansion of it:

(7.54)

Substituting into the cy relation yields

cv = 3NAJcB (for T -> oo), (7.55)

which is in agreement with the law of Dulong and Petit. The low temperature limit can be
obtained by letting the upper limit of the integral go to infinity:

[°»/T rfe« A f°° r]4e« A 4n4

/ drj = / ^dn = . (7.56)
2 2 ;
Jo (e«-D ' Jo ( ^ - D 15
236 7 / Solid Crystals

Table 7.1 Representative Values of Debye and

Fermi Temperatures for Various Monatomic
Crystalline Solids

Number of Valence
Solid #D(K) 0F(K) Electrons
Na 158 37,500 1
K 91 24,600 1
Cu 343 81,200 1
Ag 225 63,600 1
Au 165 63,900 1
Mg 400 82,700 2
Zn 327 109,000 2
Cd 209 86,600 2
Al 428 134,900 3
Pb 105 108,700 4

0.1

0.01

0.001

0.0001
0.01 0.1 10

Figure 7.6

Substituting, we find that

12TT4
Cy = — (asT^O), (7.57)

which agrees with the T3 variation observed experimentally. The variation of the specific
heat with temperature predicted by the Debye model is plotted in Figure 7.6. The propor-
tionality of dy to T3 at low temperature can be seen in this plot.
The Debye model is based on a very idealized treatment of the distribution of lattice
vibration frequencies. It is a successful model because the resulting distribution is accurate
at very low temperatures and because the specific heat and other properties depend on
integrals of the distribution function, which weakens their sensitivity to inaccuracies in the
distribution at higher frequencies. In general, the Debye temperature varies with temperature
7.4 I The Debye Model 237

0.5
prediction of
lattice vibration
model of Fine [1]
for tungsten

0.2-

0.1-

frequency v in terahertz (THz)

Figure 7.7 Frequency distribution of lattice vibrations predicted by the lattice vibration
analysis of Fine [1] for tungsten. Also shown is the frequency distribution predicted by the
Debye model for 6fo = 367 K, which is typical of tungsten.

for a given material. The variation is usually weak, however, usually no more than about
10% between 0 and 100 K, and assuming a constant value of the Debye temperature is often
a reasonable approximation for engineering applications.
The discussion in the previous section indicates that the frequency distribution for real
crystals may be much more complex. For real crystals, a more accurate representation of
the frequency distribution can be obtained either from detailed analysis of the vibrational
characteristics of the lattice or from experiments. Figure 7.7 shows a comparison of the
Debye distribution and the variation predicted by a lattice vibration analysis for tungsten.
In principle, a more realistic distribution could be incorporated into the above analysis to
get an improved prediction of the specific heat variation with temperature. The interested
reader can find more information on frequency distributions for real crystal lattices in
references devoted to modern treatments of solid state physics (see, for example, the text by
Kittel [2]).

Example 7.2 Solid pure silver has a density of 10,500 kg/m3. Reported propagation veloc-
ities for longitudinal and transverse waves in silver are 3,650 and 1,610 m/s, respectively. Use
these values to estimate the Debye cutoff frequency and the Debye temperature for silver.

Solution The Debye mean wave velocity is computed using Eq. (7.45),
-1/3 / o i \ -1/3
= 3 i/3(_L
3
^_i_) =1,817 m/s.
V(l,610) (3,650)V
The cutoff frequency is obtained by using Eq. (7.48). Using the fact that the mass density
p is (N/V)(M/NA), we fmd
1/3
f 3N \ 1 / 3 /M\l/3
fpNAA/M\
fpN
J \ 4JT/3
2 1/3
(10,500)6.02 x 10 7107.9
1,817 = 4.37 x l 0 l z Hz.
238 71 Solid Crystals

From Eq. (7.51) we then obtain

o hvD 6.63 x 10- 34 (4.37 x 1012) 210 K

= -
kB 1.38 x 10- 23
This value of the Debye temperature is close to the value #D = 225 for silver in Table 7.1.

7.5 Electron Gas Theory for Metals

Many pure metals, such as iron, copper, and gold, have a crystalline structure
and the energy storage associated with lattice vibrations can be appropriately treated with
the model described in the previous sections. In a metal, the valence electrons in the outer
shells are only weakly bound to the nucleus of the atom. These become conduction electrons,
which move about freely in the volume of the metal. Because the valence electrons have
escaped, the atoms at the lattice points in the metal effectively have a net positive charge.
Conduction electrons do interact with these ions as they move about in the metal. To analyze
the contribution of these valence electrons to the thermodynamics properties of the metal,
we model them as a fermion gas occupying the volume of the crystal. To use the results
of the fermion gas analysis presented in the last chapter we must determine the quantum
energy states and degeneracy for the electrons.
For a cubic crystal with characteristics dimension V1/3, energy levels of electrons are
modeled as a particle in a cubic 3-D box with side Vl^. In Chapter 1, we found that for
this model the energy levels are given by

where nx,ny, nz are contained in {1, 2, 3, } and the number density of the energy states
between s and s + de (degeneracy) is given by

Here we will designate the number density of electron quantum states as g*(s)de. For
fermions, no two particles can be in the same quantum state. Two electrons can occupy
each energy state in the particle-in-a-box solution, one with spin up and the other with spin
down. Consequently, the density of states for the electrons is twice that for the particle-in-
a-box solution:

If electrons were classical particles we would attribute to them a quadratic component of

kinetic energy for translation in three orthogonal directions, and the equipartition theorem
would dictate that the mean kinetic energy of each would be (3/2)kBT. It follows that for
a kmol (containing NA atoms), with each atom having one electron in the valence shell,
the contribution to the molar heat capacity would be (3/2)NAkB. This is not consistent
with experimental measurements, which indicate that the electronic contribution to the heat
capacity varies linearly with temperature. Equipartition predictions are not valid because
the electrons in our model gas are not classical particles.
7.5 I Electron Gas Theory for Metals 239

The internal energy in one kmol of the electron gas is given by

Mel = / eg*(e)f(e)de, (7.61)

JO

where f(e) is the equilibrium occupancy of energies between s and £ + de for a fermion
gas. In the previous chapter we showed that this distribution is
1
(7.62)

At low to moderate temperatures, the chemical potential is virtually the same as that attained
in the limit of T -> 0. Designating this value as the Fermi energy e F, the distribution f(s)
can be written
1
x _|_ e(s-sF)/kBT ' (7.63)

The product g*(s)f(s) represents the density of accessible states. Its variation with s is
shown in Figure 7.8.
By definition, the electron gas molar specific heat is given by

- f «•<»# ))< • •
#

We further note that the definitions of g* and / require that the integral of their product
must equal the total number of electrons in the system. Thus, for a system containing one
kmol of electrons
poo
NA= / g\e)f{e)de. (7.65)
Jo
Differentiating both sides of the above relation with respect to temperature yields

For convenience, we multiply both sides of Eq. (7.66) by £F and combine the result with

l-
/u-—Tie ,,-0.01
a8
"l
8*f

K
0.4

if \ r/9
f V
0.2 Tie _ - 1
P -0.1

0 1

Figure 7.8
240 7 / Solid Crystals

Eq. (7.64) to obtain

r * (df\ J ,n£n
cv = / (£ — £F)^ (s) I — ds.
el (7.67)
Jo \dT)
If kBT < £ F (T < # F = £F/&B), which is often the case, df/dT is nonzero only when
s = £p- Hence, the relation (7.67) is well approximated as
r
(£-£F)l-L\ds. (7.68)
l
Differentiating Eq. (7.63), we obtain
£ £ k T
'df\ fs-sF\ e( -^»
k 2
(7.69)
dT ) V BT ) [1 +
Substituting the right side of Eq. (7.69) for df/dT and changing the integration variable to
x] — (s — £p)I'k^T transforms Eq. (7.68) into
2 r)

For ^ B ^ <^C £F the lower limit of the integral may be taken to be — oo. The resulting definite
integral equals 7T2/3, and the relation for the specific heat reduces to
cV.ei = (7i2/3)klTg*(eF). (7.71)
To evaluate g* (£p) we use Eq. (7.60) together with the relation (6.157) for the Fermi energy
derived in Chapter 6. Substituting the result into Eq. (7.71) and rearranging, the relation for
the specific heat becomes

Q^ (7.72)
where 0? = £F/^B is the Fermi temperature.
For metallic crystals, contributions of both lattice vibrations (phonons) and electrons
must be included. For temperatures below the Fermi temperature, we combine Eqs. (7.50)
and (7.72) to obtain

^ y ( f ) I -Ji-^dr,, (7.73)
where rve is the number of valence electrons per atom in the crystal. At temperatures below
both the Fermi and Debye temperatures, this relation reduces to

(7.74)

Experimental data for pure metals at low temperature are consistent with the variation
indicated by this equation in that a bestfitis provided by the combination of a linear and
a cubic temperature term. In some cases the prefactors predicted by the above model are
a goodfitto the data. In others, the model prefactors do notfitthe data well. The main
cause of the disagreement in many cases is that the model assumes that the electrons do
not interact with the lattice. In fact they may do so, and phonon-electron interactions may
7.6 I Entropy and the Third Law 241

have a significant effect on energy storage in some metals. In real crystals, the potential
field imposed by the lattice is spatially periodic. It can be shown that this spatial periodicity
changes the accessible energy levels for the electrons, resulting in bands of energy that they
can occupy and bands that are inaccessible. A full discussion of these issues is beyond the
scope of this text. Further discussion of electron interactions with the lattice and electronic
energy bands in solids can be found in the references by Kittel [2] and Hook and Hall [3].

Example 7.3 Assess the importance of the electronic contribution to the specific heat of
aluminum at 70 K.

Solution The specified temperature is well below the Debye and Fermi temper-
atures for aluminum. We therefore use Eq. (7.74), which can be written in the form

NAkB 2 \OpJ 5
Substituting r ve, #p, and #D values from Table 7.1 yields

cv _ 7T 2 / 70 \ 12TT4/70\3
+
NA k B " T V134, 900 J ~T~ V428)
= 0.0077 + 1.023 = 1.030.
In this case the electronic contribution is slightly less than 1% of the specific heat value.
More than 99% is due to phonon energy storage.

7.6 Entropy and the Third Law

Now that we have examined thermodynamic characteristics of gas, liquid, and solid
phases at high and low temperatures, we are in a position to draw some general conclusions
about how the entropy of substances varies at low temperatures. First, for a Debye crystal,
we substitute

N,V
from Chapter 4 and the relations
[ ^3- v2 dv for 0 < v < vn,
g(v)dv={»° - - (1.49)
0 for v > vn.

nv/liBl
- In C = — + // I ln{l - e } + —— | g(v) dv
J0 L Z/Cnl
(7.10)
for a Debye crystal into the relation for entropy obtained in Chapter 4,

7^-= 7 ^ 7 + I n C- (4-41)
kB kT
242 7 / Solid Crystals

100-

10,

1
NAkR
0.1

0.011 z
0.001

10-4;

10

0.01 0.1 10

Figure 7.9 Entropy versus T/6& for a solid Debye crystal.

Doing so yields the following relation for the entropy of a monatomic crystal:
-OD/T

(- (4.75)

This relation is plotted in Figure 7.9. As T -> 0, the integrals in the above relation approach
constants, with the net result that S ->•0 as T -> 0 proportional to T 3 . This is fully con-
sistent with the following statement of the third law of thermodynamics presented without
justification in Chapter 3:
• The entropy ofa pure crystalline solid vanishes as the absolute temperature ap-
proaches zero: (lim r _+ 0 S = 0)pure solid crystal-
Note that this suggests that there is an absolute scale of entropy for a pure crystalline solid
that is referenced to absolute zero temperature. The work of Walther Nernst in the early part
of the twentieth century implied this limiting behavior. Since that time, the work of other
investigators indicates that only pure crystalline solids have zero entropy at absolute zero
temperature. If we substitute the single-species form of relation (6.127) for the partition
function for fermion or boson gases into the entropy relation obtained in Chapter 4,
a In 3
= kkBB \nE+ (4.113)
V
we obtain the following relation for the entropy of a fermion or boson gas:

I _\_ Jfi-£if)/kBT\ , v-
(4.76)
/'=0
Analysis of the limiting behavior of the right of this relation would allow us to determine
the limiting value of entropy as T —> 0 for either a fermion or boson gas. At first glance
this appears to be a significant result. However, virtually all substances undergo transition
to a liquid and/or a solid phase as the temperature approaches zero. The fermion/boson
gas model thus becomes inapplicable to virtually all substances at temperatures near zero.
Determination of the limiting entropy value for an arbitrary substance at zero temperature
Exercises 243

requires very detailed analysis of the substance's behavior near zero. Analyses of this type
indicate that the entropy approaches a finite value for substances other than a pure crystalline
solid. Further discussion of the third law can be found in the references by Zemansky and
Dittman [4], Lay [5], and Goldstein and Goldstein [6].

Exercises
7.1 For a monatomic crystal of the type considered in Section 7.1, show that the crystal obeys
the law of Dulong and Petit in the high temperature limit for any g(v) that has the following
characteristics:
(1) g(v) = 0 for v greater than some cutoff frequency vmax.
(2) g(v) is finite for 0 < v < vmax.
7.2 Show that in the limit T -> 0 that the entropy of a Debye crystal is given by S = (4/5)7i4NkB

7.3 Derive a relation for the entropy of an Einstein crystal and compare its predictions to that
of the relation for the Debye crystal as T - • 0.
7.4 We argued in Section 7.4 that in the limit T —• 0 the integral in Eq. (7.50) approached
4JT2/15 and consequently that the specific heat approached the cubic variation specified by
Eq. (7.57) as T ->•0 . Rewrite the integral in two parts:

Evaluate the second integral approximately for large r\ and derive a next correction term for
the specific heat in the limit of small but finite temperature T.
7.5 Solid pure zinc has a density of 7,100 kg/m3. In most metals, the propagation velocity for
longitudinal waves is about twice that for transverse waves. Use this fact to estimate the
velocity of longitudinal lattice waves in zinc.
7.6 For diamond, the Debye temperature is 2,230 K. Will the specific heat agree with the law
of Dulong and Petit at room temperature? Quantitatively justify your answer.
7.7 Reported propagation velocities for longitudinal and transverse waves in titanium are 3,320
and 1,670 m/s, respectively. Solid titanium has a density of 7,300 kg/m 3. Use these data to
estimate the Debye cutoff frequency and the Debye temperature for titanium.
7.8 You are designing a mechanical component and you want the material for the component
to have the lowest possible average acoustic velocity. If you only had density and specific
heat data available, would you select a metal for the component with high or low values of
density and specific heat? Briefly justify your answer.
7.9 A block of pure copper has a mass of 30 kg. Write a computer program to numerically
determine the energy required to raise the block's temperature from 20 to 400 K.
7.10 A tank with aluminum walls contains 6.0 kg of nitrogen gas at 800 kPa and 290 K. The walls
of the tank have a total mass of 20 kg. The tank is dipped into and out of liquid nitrogen,
which suddenly cools the tank's walls to 77 K, while the gas inside is still at 290 K.
Heat exchange then occurs between the tank walls and the interior gas. Assuming that no
exchange occurs between the walls and the air outside the tank, determine the equilibrium
temperature in the gas and in the tank wall. For the composite system, consisting of the
nitrogen gas and the tank walls, write a computer program to determine the net change in
entropy between the conditions after chilling the walls and the final equilibrium.
244 7 / Solid Crystals

7.11 Estimate the specific heat of lead at 40 K. Assess the importance of the electronic contri-
bution to the specific heat at this temperature and compare the overall value at 40 K to the
value predicted by the law of Dulong and Petit.
7.12 Derive a relation for the fraction of the specific heat that is due to phonon energy storage in
a pure metallic crystal. Plot the variation of this fraction with temperature between 0 and
300 K for aluminum. Based on your plot, how low must the temperature be before electron
contributions are significant?
7.13 Based on the theory presented in this chapter, experimental measurements of the heat
capacity of metals at very low temperature are often fit with a curve of the form
cv = ycT+pcT3.
Measured data indicate that yc — 2.08 J/(kmol K2) for potassium and yc = 0.69 J/(kmol K2)
for copper. Compare each of these values to the corresponding value predicted by simple
electron gas theory. What does the comparison suggest about the accuracy of the electron
gas model?
7.14 Write a computer program to compute the specific heat of gold as a function of temperature
from 0 to 200 K and plot the results. At what temperature are electronic and phonon energy
storage about equally important?
7.15 Initially a 0.6 kg block of pure solid gold is at 30 K and a block of pure aluminum with a mass
of 1.0 kg is at 90 K. The two masses are brought into contact with one another. Assuming
the blocks do not thermally interact with the surroundings, determine the temperature in
each block after equilibrium is reached.
7.16 Initially a 2.5 kg block of pure solid gold is at 20 K, another 1.0 kg block of gold is at
40 K, and a block of pure aluminum with a mass of 1.0 kg is at 90 K. The three masses are
brought into contact with one another. Assuming the blocks do not thermally interact with
the surroundings, determine the temperature in each block after equilibrium is reached.

References
[1] Fine, P. C , "The Normal Modes of Vibration of a Body-Centered Cubic Lattice," Physical
Review, 56:355-9, 1939.
[2] Kittel, C , Introduction to Solid State Physics, 6th ed., John Wiley & Sons, New York, 1986.
[3] Hook, J. R. and H. E. Hall, Solid State Physics, 2nd ed., John Wiley & Sons, Chichester,
England, 1991.
[4] Zemanksy, M. W and R. Dittman, Heat and Thermodynamics, Chapter 19, McGraw-Hill, New
York, 1981.
[5] Lay, J. E., Statistical Mechanics and Thermodynamics of Matter, Chapter 5, Harper and Row,
New York, 1990.
[6] Goldstein, M. and I. F. Goldstein, The Refrigerator and the Universe, Understanding the Laws
of Energy, Chapter 14, Harvard University Press, Cambridge, MA, 1993.
 CHAPTER 8

Phase Transitions and Phase Equilibrium

Chapter 8 approaches the topics of phase equilibrium and phase transitions from
a microscale perspective. Specifically, the roles of fluctuations and system stability in the
onset of phase transitions are examined in detail. Using aspects of statistical thermodynamics
theory developed in earlier chapters, the van der Waals model is used to demonstrate how
fluctuations and system instability give rise to phase transitions in fluid systems. Binary
fluid systems are considered, with pure fluid results being recovered when the mass of
one species is set to zero. It is shown that critical exponents and the law of corresponding
states can be deduced from the van der Waals model for pure fluids. Microscale aspects of
solid-liquid transitions are also considered.

8.1 Fluctuations and Phase Stability

In the development of thermodynamics presented in the preceding chapters, we
have identified different categories of substances according to the density of the substance
and the nature of molecular interactions in the substance. Gases have low density and the
molecules spend most of the time traveling through space with momentum and energy being
exchanged between molecules only through brief collisions. In liquids, the molecules are
free to roam about within the system but the density is much higher than in gases, with
the mean distance between adjacent molecules being only one to two molecular diameters.
Because the molecules are close to their neighbors, they continuously are subject to force
interactions with nearby molecules, resulting in continuous exchange of momentum and
energy.
In solids the atoms have fixed average locations in space and they interact with adjacent
atoms by stretching or compressing the bonds with the adjacent atoms. Lattice vibrations
facilitate storage and transport of energy within a crystalline solid. The solid, liquid, and
gas categories are termed phases of a material. A phase is defined as a quantity of matter
that is macroscopically homogeneous throughout. Thus far, we have said nothing about
how transitions from one phase to another can occur, or how the microscale characteristics
of the system affect such transitions. In this chapter we will explore these issues in detail.
In particular, we want to explore the roles of microscale fluctuations and system stability in
the onset of phase transitions. In doing so we will use the van der Waals model to examine
features of phase stability and phase transitions. Although this is a very simplistic model,
it exhibits the main features of phase transitions observed in most real systems. We will
also examine ways that we can generalize the thermodynamic framework to treat phase
transitions in all real fluids.

Fluctuations
In Section 4.4 we examined the predictions of statistical thermodynamics theory
regarding microscale fluctuations in a system at macroscale equilibrium. We specifically

245
246 8 I Phase Transitions and Phase Equilibrium

• composite system
boundaryrigidand
subsystem / impermeable to mass and
heat exchange
f \
\ subsystem // »
\ /

subsystem rigid boundary defines

fixed volume but is permeable to mass
and energy

Figure 8.1

considered the variance of properties, defined for an arbitrary property Y as

a2 = (Y - (Y))2. (8.1)

The square root of the variance of cry is the standard deviation of Y from its mean value.
Here we will consider fluctuations in a permeable control volume within a larger system
at equilibrium. The region within the control volume we will designate as subsystem II,
whereas the region outside the control volume but within the larger system is designated as
system I. The walls of the composite system are rigid and permit no heat or mass exchange,
effectively isolating it from the rest of the universe. The system contains two species a and
b. This situation is shown schematically in Figure 8.1. The subsystem II is in equilibrium
with the surrounding system I, and hence it is at the same pressure, temperature, and
chemical potential values as system I. For a system held at fixed P, T, /xa, and fit>, the
grand canonical ensemble provides the model for statistical analysis. For a system of this
type we demonstrated in Section 4.4 that the variance in the number of type a particles N a
in the system is given by

(4.125)

We similarly noted that the variance of the total number of particles N = Na+ Nb and the
variance of the internal energy within the system are given by
(d(Nb)\
CTN = kBT
V,T,^b V d^b ) V,T,tia V V,T,fia

(4.126)

*i'W\ d U
)
dT
/V,Na,

(4.127a)
where
(d{Na)

(4.127b)
fd(Na)\ fd(Nb _ (d{Na)\ (d{Nb)\
\\ d/jL
^ a /J T/
V T,1 ,jib I \dfjib
^
8.1 I Fluctuations and Phase Stability 247

Given these relations, we can draw some conclusions regarding possible levels of fluctua-
tions in the system within the control volume in Figure 8.1. First, we note that in Eq. (4.127a)
we can equivalently write the relation for (dU/dT)v,Na,Nb a s
dU\ fd(Nil/NA)\ N fdu\ Ncv
T N {Z)
,d J v,Na,Nb \ dT /v,Na,Nb A\dTjVNaNh NA
and hence the relation for a?r can be written

The first term on the right side of Eq. (8.3) is well behaved as long as the molar specific heat
is positive and finite. The derivatives (dU/dNa)yj^b and (dU/dNb)v,T,na represent the
amount by which the system internal energy increases when one particle of type a orb are
added to the system. For a system with a large number of particles, this is expected to be a
finite number comparable to the mean energy per molecule in the system. The effect of the
second and third terms therefore depends primarily on o^a and a^h, respectively. Equation
(4.125) indicates that cr^ is finite and positive as long as {dixa/dNa)yjillh is positive.
If (dijLa/dNa)v,T,flh approaches zero, cr^ and <J^ will become very large, implying that
there will be very large fluctuations in the molecular density and internal energy within the
control volume. The derivatives that appear in O and cr^ also clearly may affect the level
of fluctuations in the system. Note that one of the Maxwell relations is

and so there are actually three derivatives that dictate the magnitude of O and a^:
(d/JLa/dNa)v,T,tibi (d/Jib/dNb)v,T,ijLa, an
d (dlJ<a/dNb)v,T,iib- If these are all finite and the
molar specific heat is finite, the magnitude of fluctuations in N, N a, and U are finite and
well defined. However, as any of these approaches zero, the system mass and energy fluc-
tuations become very large. When the fluctuations become very large, the system no longer
has a well-defined equilibrium state in the usual thermodynamic sense. One of two things
may happen in such circumstances. Either the system remains in a choatic state with high
fluctuation levels or one of the large fluctuations may cause the system to change to a new
state in which fluctuation levels are low enough that the state is stable and well defined in
the usual thermodynamic sense.
The above relations apply to a binary mixture. For a system containing a pure substance,
these relations reduce to simpler forms. To apply them to a single-species system we set Nt,
to zero, and drop /x^ as a parameter. For a single-species system, here we will also drop the
a subscript and designate the number of particles and chemical potential simply as Af and
//, respectively. It follows that for a single-species system, the relation for the variance in
the number of particles Af is given by
k T (84)
V/ '
In Section 4.4 we showed that this relation can be converted to the form
2 N2kBT
248 8 I Phase Transitions and Phase Equilibrium

For a single-species system <£> reduces to one and the relation for the variance in the internal
energy becomes

These relations indicate that for a pure system the subsystems within the control volume
in Figure 8.1 will have well-defined mean properties with a low level of fluctuations about
the mean values only if cy and — (dP /dV)^j are positive and finite. For any physically
realistic system, N and U are always positive and they have positive mean values within the
control volume. For such systems, the definition of the variance makes it mathematically
impossible for a^ or a^ to have a negative value. This implies that — (dP/d V)Nj cannot
be negative for physically realistic systems. Although Eq. (8.6) indicates that cy could be
negative and result in a positive cr^, this also seems unrealistic since it would imply that
adding internal energy to a system would decrease its temperature.
For virtually all real systems, dy is finite and greater than zero and its value does not
contribute to high or unrealistic fluctuations in the system. However, as we shall see, in
some real systems — (dP/dV)uj can become small enough to produce high fluctuation
levels in subregions of a system like that in Figure 8.1.
The above results indicate that for a pure or binary system, it is possible to encounter
circumstances in which the necessary conditions for equilibrium are met but fluctuations
in energy or particle number are large. Although we have assumed that the system is in
equilibrium, we have not explored the stability of the equilibrium within the system. The
question of stability is especially relevant to systems that exhibit high levels of density
or energy fluctuations. These observations suggest that if the system is not stable, such
fluctuations could cause the system to undergo a transition to some new configuration. To
examine this issue we will now consider the stability of a thermodynamic system in more
detail.

Intrinsic Stability
We now will examine the stability of the overall system shown in Figure 8.1. Be-
cause the system is isolated and its wall are rigid, its mass, volume, and internal energy are
fixed. In Chapter 3 we showed that the entropy maximization principle leads to a specifica-
tion of the necessary conditions for equilibrium between two subsystems like subsystems I
and II in Figure 8.1. This result was obtained by requiring dS = 0 for the composite system
at equilibrium. The stability of the system is actually dictated by additional requirements
that result from a more complete specification of the entropy maximum condition.
We know from our analysis of thermodynamic systems in Chapter 4 that there will be
fluctuations in the system in Figure 8.1. If the system is perfectly isolated, its total mass,
volume, and energy are fixed. However, internal fluctuations could cause changes in the
number of particles or energy in subsystems I and II, provided that there are no net changes
in these parameters for the system as a whole. In our analysis here, the fundamental equation
is assumed to apply to each of the subsystems. The control volume defining subsystem II
is assumed to be fixed, and since the volume of the total system is fixed, it follows that the
volume of subsystem I is also fixed.
8.1 I Fluctuations and Phase Stability 249

For perturbations 8Ul, 8Nl, and 8Nl in system I, the resulting change in entropy for
system I is obtained by expanding the fundamental equation in a Taylor series

AS1 = 8Sl + (^] 82Sl + f^j 83Sl + • •

•, (8.7)

where the 8Sl, 8nSl terms are a shorthand way of representing the terms of successive order:

£
W= £ E
^- E r]=Ul,NlNlh $

Identical reasoning results in corresponding relations for subsystem II:

AS 1 1 = SSa + (±) S2SU +(±) 83SU + ••-, (8.11)

?II
= E £
(8-13)

E E E
The additivity property of entropy dictates that the change in the entropy of the total
system equals the sum of the changes in the subsystems. We can therefore write the total
change in entropy for the composite system A S as
AS = 8S + 52S + <53S + • •
•, (8.15)
where
8S = 8Sl + 8Sn, (8.16)
n l 11
<TS = 8 S + (TS , w = 2, 3, 4 , . . . . (8.17)
For the subsystems, since the mass and energy of overall system isfixed,fluctuationsin
these quantities for the subsystems must satisfy
8Ul = -8Ul\ (8.18)
SNl = -8N*, (8.19)
8Nl = -8N%. (8.20)
250 8 I Phase Transitions and Phase Equilibrium

Substituting Eqs. (8.8) and (8.12) into (8.16) and using Eqs. (8.18), (8.19), and (8.20) yields

8S = 8Ul

SNln
U,Nh,V

ds y / ds y1 (8.21)
dN
bJu,Na,V \dNbJu,Na,V
Since by definition
(dS/dU)Na,Nb,v = T,
(dS/3Na)UtNbjV = -f
(dS/dNb)UiNa,v =
it follows from the equilibrium requirements that the chemical potential values and temper-
ature be equal in systems I and II that the first variation of S is zero,
8S = 0 (at equilibrium). (8.22)

The condition 8S = 0 reflects a zero slope condition at equilibrium. Basic calculus dic-
tates that a necessary condition for a maximum in S is that (8nS)m[n < 0, where (8nS)m[n
is the lowest-order nonvanishing variation of S for n = 2, 3, 4, . . . . The maximum in S as-
sures stability because for equilibrium at the maximum, fluctuations away from equilibrium
result in a decrease of the system entropy. Spontaneous subsequentfluctuationsthat increase
the entropy back toward the maximum are then most likely to occur, tending to restore
the original equilibrium. To determine the necessary conditions for a stable equilibrium,
we therefore seek to determine the conditions that result in a maximum value of entropy at
the equilibrium point. Thus, for the composite system we are considering here, the criteria
for equilibrium and stability may be stated as
8S = 0 criterion of equilibrium, (8.23)
8nS < 0 for the smallest
•criterion of stability. (8.24)
« = 2 , 3 , 4 . . . at which 8nS ^ 0

The above criteria indicate that we must first consider 82S to determine the necessary
conditions for the system to be stable. Summing the rights sides of Eqs. (8.9) and (8.13),
we obtain

+2
2
+2 / d s y SUlSNl + 5^S\n d2S
8.11 Fluctuations and Phase Stability 251

+ \dNadNb
(8 25)
-
Because the terms in the above equation are part of a Taylor series, the partial derivatives
are evaluated at the preperturbation state, which is the same in both subsystems. For a given
state, however, the derivatives of extensive properties scale with the size of the system
Nl = Nl+ Nl because at equilibrium S, U, Na, and Nh do.
As an example, consider
Nl i
( YY
\dUdNa) Nl xNl
Thus, for the two subsystems of different size at the same equilibrium state, it follows that
N
Y/r ^ ) (8(8 26)
6)
or equivalently
)I )
d2S _ Nl ( d2S
~ Wl \dUdN~a
where A^n = TV" + NlJ. Similarly, it can be shown that

N" (
'
d2s Nl d2s
(\dUdN )h
Y' (
\dUdN )b
Y ' ( }

(830)
dNadNb dNadNb
We next modify Eq. (8.25) by using Eqs. (8.18)-(8.20) and (8.27)-(8.30) to eliminate
terms with II superscripts. Equation (8.25) then becomes

l
(8.31)

Since A^1 and A^n are both positive definite numbers, the sign of 82S is dictated by the term
in square brackets in Eq. (8.31). Inside the square brackets all the terms are evaluated for
subsystem I. We will therefore drop the I superscript with the understanding that these terms
252 81 Phase Transitions and Phase Equilibrium

are evaluated in subsystem I. We can say that the equilibrium will be stable if 82S* is less
than zero, where 525* is defined as

+2
(^rr^r)8U8Na+2 (Jr&-)8U8Nh
\dUdNJ \dUdNJ
+ 2 f--ll—]sN aSNb]. (8.32)
\3NadNbJ J
Rearranging the terms on the right side of the relation for <52S*, we obtain

025/31/2)
(d2s/dUdNa)(d2s/dUdNh)
dNadNb (d2S/dU2)
Q2S (d2s/dUdNa)2
dwj (d2s/du2)

where

d2s (8.34)
ac/ 2 '
(d2S/dUdNa)2
(8.35)

r d2s (d2s/dUdNa)(d2s/dUdNh) 12
2 2 2
dS (d S/dUdNh) [dNadNb (d2s/du2) \
S3 2 2
dNl (d s/du ) d2S (d2S/dUdNa)2
(d2S/dU2)

In Eq. (8.33) the squared terms in curly brackets are positive definite. The conditions under
which 82S* is negative are dictated by the prefactors of these terms. To determine the
conditions for which <52S* is negative, we must therefore evaluate the partial derivatives in
the prefactors of the squared terms. Using results from Chapters 2 and 4, we evaluate the
derivatives as follows. Since by definition

( ^ . (8.37)

Using Eq. (8.2) to evaluate the derivative in (8.37) we obtain

?^ (8.38)
Using the relations (3S/dNa)Nb,u,v = —/^a/T, (dS/dNb)Na,u,v = —^b/T, it can similarly
8.1 I Fluctuations and Phase Stability 253

be shown that

(dNa\ fdNb\ _ fdNa_\

V d^ ) V,Tifib V ^ / V,7>fl V d^ ) V,T,fia

The stability criterion is satisfied if §i, ^2. and §3 are all negative. Examination of Eqs.
(8.38)-(8.40) indicates that the parameters that dictate stability for a binary system are
cv, (d/JLa/dNa)v,T,flb, (d/JLb/dNb)v,T,(jLa, a n d (d/jba/dNb)v,T,fih' Since the Maxwell relations
require that (d/jLa/dNb)v,T,(jLb — (9/WdA^)v,7\/za> w e need not consider both cross deriva-
tives. Reexamination of the relations derived above for cr^, aft, and o^ reveals that the
parameters that dictate stability are exactly the same as those that dictate the magnitude of
fluctuations in the system. It is readily seen that — &B/§I is the first term in Eq. (8.3) for a^
and that

fe = .*(^>) = —, (8.41)
T \SNaJVtTttlb aNa
fc = — ^ - . (8.42)
°Nh
This is a particularly noteworthy result because it provides insight as to the physical
causes of system instability. For example, our analysis above indicates that stability is
assured if £1, §2, and §3 are negative. Negative £3 requires that (d(jLh/dNh)Vj^a be positive.
If the value of this derivative were decreased until it passed through zero and became
negative, the stability of the system would be in doubt. Our analysis offluctuationsin the
system indicates that the fluctuations in Nb and U would be infinite if the value of this
derivative reached zero. A system with infinite fluctuations is certainly not stable in the
usual thermodynamic sense. This strongly implies a link between the stability of the binary
system and the magnitude of the fluctuations in the system.
Since the choice of a and b species designations is arbitrary, the requirement that §3 be
less than zero is equivalent to the requirement that (3iia/dNa)vj,iib and (d/jLb/dNb)v,T,fMa
be greater than zero. Using the Maxwell relation for the cross derivative and doing a bit of
rearranging, we can state the requirement that £2 be less than zero as

> Q

1 _

Our analysis implies that for a stable system, (diLa/dNa)v,T,iib anc* (d/^b/dNh)Vj^a are
positive. For a stable system, since the above inequality is also satisfied, the second term
in the denominator is less than one. If conditions change so that (dfjia/dNa)v,T,iib or
(d/jib/dNb)v,T,iia decreases toward zero, this second term can only become smaller, and the
inequality (8.43) will be violated only when (dfia/dNa)Vj^h becomes negative. Stability
is therefore assured if (d/jia/dNa)v,T,fih and (dfjLb/dNb)v,T,na are both positive, and the
additional requirement (8.43) is redundant. For the binary system considered here, we can
254 8 I Phase Transitions and Phase Equilibrium

therefore state that stability is assured if the following conditions are met:
cv > 0, (8.44)

> 0. (8.45)

These are both necessary and sufficient conditions for stability of a binary mixture system.
If these conditions are met, the system is said to be intrinsically stable.
The analysis developed in this section for binary mixtures can be extended to systems
containing more than two species. In principle, such an extension is straightforward, but the
mathematics becomes increasingly complicated as the number of components increases.
Further information on analysis of thermodynamic stability for systems containing more
than two species can be found in the text by Modell and Reid [1].
The analysis described above must apply to a single-species system in the limit of
Nt, -» 0. As we noted in connection with our analysis of fluctuations, to apply the binary
mixture results to a single-species system we set Nt, to zero and drop /x^ as a parameter.
We also drop the a subscript on // and N. Equation (8.44) still applies to a single-species
system, but the second requirement becomes

>0 <8 46)

- -
In Chapter 4 we showed that for a binary mixture

(4 140)
(iir) —£{JF) NT • -
which implies that for a single-species system

) • (8-47)

Since V and N are positive definite, the condition (8.46) is equivalent to (dP/dV)N,T < 0.
We can therefore state the necessary and sufficient conditions for intrinsic stability of a pure
system as
cv > 0, (8.48)

dV)NT
Because it relates to thermal energy storage, the first condition is sometimes referred to as
the criterion of thermal stability. Similarly, because the second condition is associated with
the mechanical compressibility of the system, it is sometimes referred to as the criterion
of mechanical stability.

Implications for Dense Gases and Liquids

As discussed in Chapter 6, the van der Waals equation provides a prototype model
of the behavior of pure dense gases and liquids. At low temperatures, the van der Waals
8.1 I Fluctuations and Phase Stability 255

model predicts isotherms that look like those in Figure 8.2. At low values of the molar
volume v = VNA/N , points along the isotherms correspond to a liquid phase, whereas at
high 0, the fluid is in a gaseous phase.
To explore the behavior of fluids at low specific volume, we consider the fluid inside
the piston and cylinder apparatus shown in Figure 8.3. This apparatus initially contains
vapor phase fluid with high specific volume at temperature T\. The state point for the
system is on isotherm T\ at the far right in Figure 8.2. We move the piston to slowly
decrease the volume of the system exchanging heat with an isothermal reservoir so as to
keep the system temperature constant at T\. As this process occurs, the state point follows
the isotherm to the left. At first this results in an increase in pressure. If the system follows
the isotherm, the van der Waals model predicts that the pressure eventually would pass
through a maximum and then a minimum before rising steadily at low v. These extrema
in the pressure occur only for temperatures below a certain value referred to as the critical
temperature, Tc. We note that at the maximum and minimum (dP/dv)T = 0, which implies
that (dP/d V)N,T = 0 and that a^ -» oo at these points. This implies that the system would
exhibit very large fluctuations near these conditions. Furthermore, in the region between the
pressure minimum and maximum, (dP/dv)T > 0, which implies that (dP/dV)Nj > 0.

Vapor

Liquid

= V /(N /NA)

Figure 8.2

reservoir at T = 7 1

Spiston
s
\
ss pure fluid
s
s 1888881
388888
\ VWWWWV

Figure 8.3
256 81 Phase Transitions and Phase Equilibrium

It follows from our analysis above that the requirements for intrinsic stability are not satisfied
in this region.
The above observations lead to the conclusion that stable thermodynamic states can-
not exist in the range of conditions between the pressure minimum and maximum on the
isotherm. As v is decreased isothermally if the system passes through a sequence of equilib-
rium states, the observed behavior of a real system does, in fact, deviate from the sequence
of states predicted by the van der Waals isotherm. At some point before reaching the max-
imum pressure on the isotherm, the pressure levels off and stays constant as 0 drops. The
system condition bifurcates with a portion remaining at the vapor state at which the pres-
sure first leveled off and the remaining portion exhibiting a higher density liquid state. The
pressure at which this bifurcation occurs is the saturation pressure Psat , which varies with
temperature for a pure fluid: P^t = Aat(T)-
At the saturation pressure the gas and liquid phase can coexist, with the former referred to
as saturated vapor and latter as saturated liquid. For the vapor and liquid phase subsystems
to be in equilibrium within the overall system, we know that the temperature, pressure, and
chemical potential must be the same in both phases. In Chapter 6 we derived Eqs. (6.26) and
(6.27) for the pressure and chemical potential of a pure van der Waals fluid. These relations
can be written in terms of the molar specific volume as

v — N A bv
\i NAbv
k BT v-NAby \2
- 5) , , 1 (g - 3), f T \ 2avNA
-—lnjr-ln<r s —In j
2 J 2 0mt,m
Clearly for a given fluid, the chemical potential is a function of T and 0. For a given
temperature T below the critical temperature, the saturation conditions are dictated by the
following five relations:

k BT
-5)
[ 2 ^ \Vrot,m /

Mv NAl 3\ l \27tMkBT(vv -
k BT 2) I N2A/3h>
ln,-lnJ-^lnm-^, (8.53)

fly = jL6l, (8.54)

Oi - NAbw vf
N M
P= - ^1, (8.56)
8.1 I Fluctuations and Phase Stability 257

where for v and \x the subscript 1 denotes a saturated liquid property and the subscript
v denotes a saturated vapor property. These equations embody the requirements that the
vapor and liquid satisfy the van der Waals equation of state and chemical potential relations
and that the pressure and chemical potential are the same in both phases at equilibrium.
There are five equations for the five unknowns /JLV, fii, 0v, OI, and P. Although the equations
are nonlinear, they can be solved to determine the five unknown saturation properties for
a specified value of temperature. The saturation conditions dictated by these relations are
indicated as the saturated vapor and liquid curves in Figure 8.2 and Figure 8.4.
Figure 8.5 shows the variation of the saturation pressure as a function of temperature
predicted by solving Eqs. (8.52)-(8.56) using the van der Waals constants for nitrogen from
Table 6.1. Also shown is the variation indicated by standard thermodynamic tables based

f Isotherm for T < Tc

P i * Critical Point
• Saturation Curve
Metastable
Supersaturated Vapor

- Metastable Superheated Liquid

Figure 8.4

3500

50 100 150
T (K) r

Figure 8.5
258 81 Phase Transitions and Phase Equilibrium

on experimental measurements. Qualitatively, the model prediction is consistent with the

variation in the measured data, but the numerical agreement becomes increasingly worse
as the temperature decreases. This is not surprising since the constants in Table 6.1 were
obtained by matching nitrogen properties at the critial point.
The curves in Figure 8.4 denoting the limit of intrinsic stability for the vapor and liquid
are termed the vapor spinodal limit and the liquid spinodal limit, respectively. It has been
widely demonstrated that liquid can be superheated and vapor can be supercooled (super-
saturated) beyond their equilibrium saturation conditions without a phase transformation
occurring, even though the system equilibrium state would lie within the saturation "dome"
in Figure 8.4 at equilibrium. Vapor that is supercooled below its equilibrium saturation tem-
perature and liquid that is superheated above its equilibrium saturation temperature exist in
a nonequilibrium condition referred to as a metastable state. Possible metastable states lie
between the spinodal and saturation curve, as indicated in Figure 8.4.
The portions of the van der Waals isotherms in the metastable regions of the P—v diagram
represent metastable equilibrium states that may undergo a transition to a stable thermody-
namic equilibrium state with two phases present if the system experiences a fluctuation of
sufficient magnitude. Experiments indicate that metastable states of this type do occur in
real systems and that the behavior indicated by the van der Waals model for such systems
is at least qualitatively correct. As the state point of a system penetrates further into the
metastable region, the van der Waals model predicts that —{dP/dv)T becomes progres-
sively smaller, approaching zero at the spinodal. We know that the level of fluctuations
increases as — {dP/dv)r decreases, making it increasingly likely that the system will jump
to a stable equilibrium state. The large density fluctuations that occur when —{dP/dV)^j
becomes small may raise the density of a vapor nearly to that of a liquid phase or may lower
the density of a liquid near that for a vapor phase. Large densityfluctuationsthat may initiate
a change of phase are sometimes referred to as heterophase fluctuations. Thus, jumping to
a new equilibrium may initiate the formation of a new phase. Spontaneous formation of a
new phase within a preexisting phase is referred to as homogeneous nucleation. Because
fluctuations become enormous and the intrinsic stability condition is not satisfied at the
spinodal curve, the spinodal condition is interpreted as a limiting condition beyond which
a phase transition must occur.
The relationship of the chemical potential variation to the stable and metastable regimes
can be seen in Figure 8.6. In this figure the variation of n/k^T as a function of pressure is
plotted for three different isotherms. This was done parametrically for nitrogen by computing
/JL/ &B T and pressure over a range of molar density (N / VNA) using the following rearranged
versions of Eqs. (6.26) and (6.27):

(8-57)
- (N/VNA)NAby
(N/VNA)NAbv /3\, \2nMkBT[l-(N/VNA)NAbv
k BT 1 - (N/VNA)NAh

•-5)
i n , - i n a] - ^ In
k BT
(8.58)
8.11 Fluctuations and Phase Stability 259

T (K)

-15
20

Figure 8.6 Chemical potential variation for nitrogen (van der Waals model).

For a specified temperature, the computed value of ix/k^T = fl/RT at each value of
(N/VNA) was plotted as a function of the corresponding pressure. The resulting curves
shown in Figure 8.6 thus indicate the variation oijl/RT with P along an isotherm predicted
by the van der Waals model for nitrogen. In Figure 8.6, the letters on the 100 K isotherm indi-
cate states that are the same as the corresponding locations in Figure 8.4. Segment AB is sub-
cooled liquid, BC is superheated liquid, DE is supersaturated vapor, and EF is superheated
vapor. The segment CD is inaccessible because of stability considerations. As shown in
Figure 8.6, as the temperature approaches the critical point (126 K), the cusps in the chemical
potential curve disappear, leaving a smooth variation of /x with P at temperatures above Tc.
Another useful perspective can be obtained by considering the variation of n/ k^T
with molar density (N/VN&) and temperature as predicted by Eq. (8.58). This variation
for nitrogen is shown in Figure 8.7. We showed above that when (dji/dN)yj —> 0 the
level of fluctuations in the system becomes enormous and intrinsic stability is no longer
assured. In Figure 8.7, this derivative will approach zero at locations where a line of constant
temperataure on the fi/kBT surface has zero slope. It can be seen in this figure that at
high temperature (>140 K) the slope of the surface along a line of constant temperature
is steep at high densities, becoming less steep at low densities, but never going to zero.
However, at lower temperatures (<60 K) there clearly exist two locations where the slope
of the surface along an isotherm goes to zero. These correspond to the two points where
the isotherm intersects the liquid and vapor spinodals. The wavy nature of this surface at
low temperatures and densities thus indicates a region where the substance lacks intrinsic
stability. This region is inaccessible to the system and attempting to traverse it generally
will result in a phase transition.
260 81 Phase Transitions and Phase Equilibrium

van der Waals model for nitrogen

(N/V)/NA

Figure 8.7

fundamental equation for nitrogen (van der Waals model)

UmolK
u = UNA I N
(MJ/kmol)
VNAI N
(m3/kmol)

Figure 8.8

Useful insight can also be obtained by examining the fundamental relation. In our analysis
above, we considered a closed system with afixednumber of molecules. For such a system, at
equilibrium, the entropy must be at a maximum, which mathematically implies that 8 2 S < 0.
This mathematical constraint on the second variance of the entropy is equivalent to the
geometric requirement that the fundamental relation have negative curvature at equilibrium.
Figure 8.8 shows the fundamental equation surface for nitrogen obtained using the van der
Waals model in Chapter 6. Note that the surface in this plot represents the relation among s,
w, and v obtained when the fundamental Eq. (6.31) for a van der Waals fluid is rearranged to
a form containing only molar specific properties. Curvature trends in the specific properties
are equivalent to those for the corresponding extensive properties S,U, and V for a system
with fixed N.
8.11 Fluctuations and Phase Stability 261

In Figure 8.8, stable portions of the fundamental surface with negative curvature are
concave downward. It can be seen that in Figure 8.8, the surface is concave downward
everywhere, except for a small region at low v and u. This corresponds to the range of
conditions where we have shown that intrinsic stability is not assured. This, of course, cor-
responds exactly to the region between the spinodal curves in Figure 8.4. Stable equilibrium
cannot exist in regions of upward curvature on the fundamental surface because fluctuations
become very large as they are approached, and at the location where the curvature changes
from positive to negative, intrinsic stability is no longer assured. The phase transition that
occurs when the system state attempts to penetrate this region can be viewed as the sys-
tem bifurcating, with a portion jumping across this region of positive curvature to a stable
equilibrium on the other side with negative curvature.
It should be noted that although we have frequently used the van der Waals model in
examples, the conclusions reached in this section are applicable to any real fluid. This is
based on the expectation that relations among the properties like those derived for the van
der Waals fluids exist for all real fluids, although we may not be able to easily write them
down in closed form.

Example 8.1 Use the results for the van der Waals model for a pure substance to predict
the temperature at the limit of intrinsic stability for liquid nitrogen at atmospheric pressure.
Also determine the level of density fluctuations in one cubic micron of the liquid at 1 K
below the temperature at the limit of intrinsic stability.

Solution The necessary condition for intrinsic stability of a pure substance,

Eq. (8.49),
dP\
— <0
dv)NT
can be expressed in terms of the molar specific volume as

The limit of intrinsic stability corresponds to (3P/dv)T — 0. Differentiating Eq. (8.50), we

find that the specific volume at the stability limit must satisfy
dP\ _ N Ak BT 2avNj
dv)T~ (v-NAby)2 v3
Combining this equation with the van der Waals equation of state to eliminate 7\ we
obtain quation with tl

1(0 - 2NAby)

For the specified value of pressure, this cubic equation has three solutions for v. Sub-
stituting ay = 3.76 x 10"49 Pa m6/molecule2 and by = 6.40 x 10~29 m3/niolecule for
nitrogen from Table 6.1, we iteratively compute solutions of v = 0.07941, 1.120, and
— 1.20 m3/kmol. The negative value is not physically possible and is ignored. The larger
positive value is the spinodal limit for supersaturated vapor. The value of interest here for
262 81 Phase Transitions and Phase Equilibrium

superheated liquid is the lower positive value v = 0.07941 m3/kmol. Substituting this value
into the above equation for the spinodal condition and solving for T yields a value of 107 K
for the temperature at the limit of intrinsic stability. Note that this is well above the saturation
temperature of 77 K for nitrogen at atmospheric pressure.
To evaluate the level of density fluctuations at 106 K, we first use the van der Waals
equation of state (8.50) to iteratively determine the specific volume for liquid at 101 kPa
and 106 K. There are three solutions to the resulting cubic equation. The lowest, v = 0.07065
m3/kmol, corresponds to superheated metastable liquid. Noting that (dP/dV)Nj = (NA/N)
(dP/dv)r we can rewrite Eq. (8.5) as
1/2

N l-(dP/dv)Tv2N_
In this equation, N is the number of molecules in one cubic micron, determined as

v(io- 6 ) 3 _ io- 18
N = "" / = ~,~ —=26 = 8.52 x 10 9 molecules.
v/NA ~ 0.07065/6.02 x IO
Using the equation noted above for (dP/dv)T,
dP\ N Ak BT 2ayNJ
dv)T~ (v-NAby)2^ v3 '
we substitute to obtain
6.02 x 1026(1.38 x 10-36)106
dv JT [0.07065 - (6.02 x 1026)6.40 x 10~29]2
2(3.76 x 10~49)(6.02 x IO26)2 7
+
(O07065P = -8-°° X 10 P

Substituting the above results and the value of Af into the relation for a^/N, we find that

8308(106) , 1/2 _ . . ^ _ 5
7 2 9 = 1.6 x 10"
8.00 x 10 (0.07065) 8.52 x IO
Thus the fractional standard deviation in the number of molecules in a cubic micron is very
small, even within 1 K of the intrinsic limit of stability. The system must get very close to
this limit before the level of fluctuations becomes large.

Solid-Liquid Phase Transitions

The discussion above focused on fluctuations within the system and their impact
on phase stability and phase transitions. The nature of such fluctuationscan be analyzed in
a straightforward manner for gases and liquids. Fluctuations are clearly different for solids,
however, and to understand how microscale features affect solid-liquid phase transitions, we
must consider the system from a different perspective. In this section we will specifically
consider the possibility of a transition from a liquid to a crystalline solid structure, or
vice versa. The nature of processes that take place when a solid melts continues to be the
8.11 Fluctuations and Phase Stability 263

subject of ongoing research. A simplistic but useful perspective on such processes can be
developed by considering the basic characteristics of energy storage in solid crystal. We
argued in Chapter 7 that energy is stored by lattice vibrations in a crystal, which are basically
concerted vibrations of atoms about their lattice locations in the crystal. Considering these
to be classical harmonic oscillators, we would expect that such vibrations could be sustained
in a stable manner as long as their amplitude does not become too large. For a cubic lattice
structure, if v is the molar specific volume, the mean spacing 8\ between lattice sites occupied
by atoms must be
&i = (v/NA)l/3. (8.59)
The energy in each classical oscillator with a maximum displacement of zmax is
We expect that the crystal lattice will begin to shake itself apart when zmax becomes close
to 8\. If we interpret this as the onset of melting, we expect that the melting point will
correspond to conditions where the mean energy of the most energetic oscillations is close to
(1 /2)ks(v/NA)2^ •Since the mean energy per oscillator scales with (1 /2)&B T, it follows that
the melting point temperature Tm should approximately be given by a relation of the form.
(l/2)* B r m = C0(l/2)A:s(i)/WA)2/3, (8.60)
where Co is an unknown constant. Our analysis of a harmonic oscillator in Chapter 1 also
indicated that the effective spring constant is related to the mass per atom m a and the
frequency of the oscillator as
ks = ma(27tv)2. (8.61)
Picking the Debye cutoff frequency VD as the characteristic frequency of the most energetic
oscillations, the above relation becomes
ks=ma(27TvD)2. (8.62)
Substituting the right side of (8.62) for ks in (8.60) and using the definition of the Debye
temperature 0D = hvD/kB, we can rewrite Eq. (8.60) as
O/JVA)2/3
• (5.03)

For a species with atomic weight M, the mass per atom is M/NA. We can therefore replace
raa with M/NA to obtain
2 2/3
M / W D \ / v \
= L0——— TT • (8.64)
Nk \ h J \NAJ
By absorbing the numerical factors into the unknown constant, which we now designate as
C m , we can write the above relation for the meting point temperature as

T —C M fkBoDy / v Y
<8 65)
wi (nr) U ) • -
This heuristic model provides a prediction of how the melting temperature of a crystalline
solid varies with molecular weight, Debye temperature, and molar specific volume of the
crystal. In Figure 8.9, the melting temperature for several metallic crystals is plotted as a
function of the term in square brackets in Eq. (8.65). Most of the data do lie close to a straight
264 8 I Phase Transitions and Phase Equilibrium

10,000

5,000 10,000 100,000

2/3
M

Figure 8.9

line, which corresponds approximately to Cm = 0.06. Some metals, such as zinc and tin,
deviate significantly from the line. Part of the reason for the scatter is that the cited Debye
temperature is usually a bestfitto energy storage in the crystal at low temperatures, and it may
not be representative of the crystal's behavior at higher temperatures near the melting point.
Another interpretation of the above analysis can be obtained by rearranging Eq. (8.65)
to the form

(8.66)
M UB^D/
Since this relation represents the conditions for incipient melting, it implies that the onset
of melting will occur when the dimensionless group on the left side of Eq. (8.66) reaches
the threshold value of C m . This, in a sense, is a stability limit for the crystal, beyond which
it will vibrate so severely that its crystalline structure will begin to break down. The plot
of data for various metals in Figure 8.9 indicates that breakdown begins to occur when the
value of this dimensionless group approaches 0.06. The scatter in the data also indicates
that the exact threshold value may vary significantly from one material to another.

8.2 Phase Transitions and Saturation Conditions

First- and Second-Order Phase Transitions

The discussion in the previous section indicates that for a pure substance, a phase
transition from liquid to vapor or vapor to liquid will occur at constant pressure and temper-
ature but will result in a change in the molar specific volume. This is also generally true for
8.2 I Phase Transitions and Saturation Conditions 265

solid-liquid phase transitions in a pure substance. For a van der Waals fluid in particular,
Eq. (6.30) can be rearranged to obtain the following relation for the molar entropy of the fluid:

r^q (867)
This relation implies that, for fixed 7\ a change in molar specific volume will produce a
change in entropy for a van der Waals fluid. In general, a change of phase for a pure sub-
stance is accompanied by a change in specific volume and specific entropy. In Chapter 3 we
demonstrated that for a pure substance the molar Gibbs free energy g is equal to the molar
chemical potential ft = NA/JL. The Gibbs-Duhem equation for a pure substance, Eq. (3.50),
can therefore be written in the form
dg = vdP -sdT. (8.68)
Because P and T are constant during a solid-liquid or liquid-vapor phase change of a pure
substance, g is unchanged during the process. The differential relation (8.68) implies that
g can be considered a function of P and T. It follows mathematically that

dT. (8.69)

Since both Eq. (8.68) and Eq. (8.69) are valid, we find that

= -s. (8.70)
p
The conclusion that v and s change discontinuously during the phase change implies that
these first-order derivatives of the Gibbs function change discontinuously. A phase change
in which the first-order derivatives of the Gibbs function change discontinuously is known
as a phase change of first order. A phase change in which second-order derivatives of the
Gibbs function change discontinuously is categorized as & phase change of second order.
If we consider a system containing N& molecules or atoms, the definition of enthalpy
H = U + PV implies the following relation among molar specific properties:

h = u + Pv. (8.71)
Differentiating this relation we obtain
dh = du-h P dv + vdP.

Equation (3.32), dU — TdS — PdV, when applied to a system containing N& molecules,
can be written as
du = Tds - Pdv. (8.72)
In Eqs. (8.71) and (8.72) u and h are the molar specific internal energy and enthalpy,
respectively. Combining Eqs. (8.71) and (8.72) to eliminate du yields
dh = Tds + i)dP. (8.73)
266 81 Phase Transitions and Phase Equilibrium

Dividing both sides of this relation by dT and considering a process at constant pressure,
in which the ordinary differentials become partial differentials, we obtain

-
With this result we can write the definition (3.164) for the specific heat at constant pressure
for a pure system as

)
Because s changes by a finite amount with no change in temperature, (ds/dT)P is infinite
and it follows that the molar specific heat cP is infinite when both phases are present in the
system. Note that cP is finite for a system containing only one of the saturated phases.

The Clapeyron Equation

We now will consider what happens when a system containing a pure substance
undergoes a change from an initial saturated condition to another saturated condition for
which the pressure and temperature differ by dP and dT, respectively. Because the initial
and final states are both saturation states, the molar Gibbs function for the two phases must
be the same before and after the change. The changes in g for phase I and for phase II during
the state change must therefore be the same:
dgl = dgu. (8.76)
Using the Gibbs-Duhem equation to replace the differential Gibbs function changes, we
obtain
vl dP - sl dT = v11 dP - s11 dT. (8.77)
Noting that dT and dP correspond to changes along the saturation curve, we rearrange this
relation to the form

Equation (8.78) is the general form of the Clapeyron equation, which relates changes
in volume and entropy to the slope of the saturation P-T curve for a pure substance. We
showed above that at constant pressure dh = Tds, which implies that for the phase change
at constant T and P, sn - sl = (h11 - hl)/T. The Clapeyron Eq. (8.78) can therefore be
written in the equivalent form

<879)

Using the shorthand notation

Ah = hu - h\ Av = v11 - 01 (8.80)
we can write the Clapeyron equation in the form
fdP\ Ah
- (8.81)
\dTjsat TAv'
8.2 I Phase Transitions and Saturation Conditions 267

This form of the Clapeyron equation is particularly useful because it relates the slope of
the saturation curve to measurable quantities. The volume change can be measured directly
and the enthalpy change associated with the phase change can be measured using standard
calorimetry techniques.
For the case of a liquid-vapor phase change away from the critical point, we can invoke
two idealizations: (1) at saturation, the liquid volume is negligible compared to the vapor
volume and (2) the vapor behaves as an ideal gas. This allows us to replace Av in the
Clapeyron equation with NAkBT/P\

(^1 =^ - (8.82)

This relation is known as the Clausius-Clapeyron equation. Over a limited temperature

range, the enthalpy of vaporization may be approximated as being constant. The Clausius-
Clapeyron equation can then be integrated to obtain

Ah
lnP sat = - _ _ + Co, (8.83)
NA k B T
where Co is a constant that must be determined from measured data or a model of the fluid.
This relation suggests that a plot of In Psat versus \/T would result in a straight line over
a limited range of temperature. In fact, a plot of data for many substances produces very
nearly a straight line from the lowest temperature where a liquid and vapor coexist (the triple
point) all the way to the critical temperature. This makes this relation particularly useful
in fitting vapor pressure data for real substances. The reasons for the better-than-expected
accuracy of Eq. (8.83) will be examined further in Section 8.4.

Example 8.2 If Eq. (8.82) is integrated assuming Ah is a linear function of temperature,

a relation of the form

ln/>sat = A - - +

is obtained for the vapor pressure curve. This form has been suggested as an improved rela-
tion for fitting vapor pressure data. Show that the van der Waals model for a pure substance
discussed in Section 8.1 predicts a relation of this form if the following idealizations are
invoked: (1) the liquid specific volume is constant and (2) the vapor behaves as an ideal
gas.

Solution The saturation conditions are dictated by Eqs. (8.52)-(8.56) with the
additional idealizations indicated above. Since the vapor is to be treated as an ideal gas, the
ay and bv terms in Eqs. (8.53) and (8.56) are set to zero, which reduces them to the forms

= ln\2itMk Tv^)
- U) h H '
/3\ B [(§-5)

(? - 3) .

P = NAkBTvw.
268 81 Phase Transitions and Phase Equilibrium

Setting the right side of the above equation for jiy/k^T equal to the right side of Eq. (8.52),
substituting NAk^T/P for 0v, simplifying, and rearranging, we obtain
/ Pi - NAby \ _ NAbv 2ayNA
n
\NAkBT/PJ ~ v{-NAbY
This relation can be further rearranged to the form

ln P
= \ ^FT~ ~ l n ( ^ ~ ^A^V) + MNAkB) + In T.

If the molar specific volume of the liquid is taken to be constant, this relation clearly has
the form ln Psat = A - (B/T) + C ln T.

The Gibbs Phase Rule

We now will consider the conditions at equilibrium in a general system that contains
m phases and r species. In doing so we consider each phase that is present to be a subsystem.
It follows from results obtained in Chapter 3 that, at equilibrium, the temperature, pressure,
and chemical potential of each species must be the same in each phase. Denoting phases
with roman numerals and species with number 1 to r, we can state these requirements
as
ri = r n = . . . = r m , (8.84a)
/>! = />„ = . . . = />m, (8.84b)

(8.85)

At equilibrium, the fundamental equation must apply to the system as a whole and any
subsystems we define. As discussed in Chapter 3, we can represent the fundamental equation
in several different forms. Beginning with the energy form U = U(S, V, N\,..., Nr), we
can execute the partial Legendre transform that replaces 5 with T and V with P to obtain
the Gibbs function representation G = G(T, P, N\,..., Nr), where by definition
G = U -TS + PV. (8.86)
Since U,S, and V are additive over subsystems and T and P must be same in all subsystems,
the above relation implies that G is additive over the subsystems.
We define the mole fraction of species / as
Xt = Ni/N, (8.87)
where Af is the total number of particles of all species,
r

N = J2Ni' (8 88)
-
Equation (8.87) implies that in the Gibbs function representation of the fundamental
8.2 I Phase Transitions and Saturation Conditions 269

equation, we can replace Nt with X(N:

G = G(T, P, XiN, X2N,..., XrN). (8.89)
But we know that G is additive over any subsystems. If we consider four subsystems with
N particles each, the value of G for the total system of 4N particles must be four times that
of each subsystem having N particles. This line of reasoning can be applied to any number
of subsystems, which implies that G must be linearly proportional to the total number of
particles in the system. It follows directly that
G(7\ P, XiN, X2N,..., XrN) = NG(T, P, XUX2,..., Xr), (8.90)
and hence the Gibbs function for any system can be determined as
G = NG(T, P, XUX2,..., Xr). (8.91)
This result implies that the ratio G/N is a function of T, P, and the mole fraction of each
species in the system,
G/N = f(T, P, Xu X2,..., Xr). (8.92)
Differentiating G/N with respect to the number of species / particles and using the chain
rule from basic calculus yields

= ( ) ( ) ( ) ( ) . (8.93)

Since N = N\ + N2 H \-Nr, dN/dNj = 1. Using this result and the fact that by definition
Hi = (dG/dNj)T,p,Nj#, we can rearrange the above relation to the form

Since for any species j , Xj = Nj /N and Af is the sum of all the species numbers, each mole
fraction Xj is a function of all the species numbers. Accounting for the Nt dependence of
all the mole fractions, the derivative in Eq. (8.94) can be rewritten using the chain rule as

f ^ \ 3Xj J T , P . N ^ ^

From the definitions of X, and N, it can be easily shown that

= , (8-96)
N
P.NV
where 5/y is the Kronecker delta. Substituting this result into Eq. (8.95) we obtain

(8 97)
-

This rather lengthy manipulation has led us to an important conclusion. Since G/N is
a function of T, P, and the mole fractions, the right side of Eq. (8.97) is a function only
of these parameters. We have therefore demonstrated that the chemical potential for each
270 81 Phase Transitions and Phase Equilibrium

species can, in principle, be written as a function of T, P, and the mole fraction of each
species in the system:
to=ln(T,P,Xl,X2,...,Xr). (8.98)
Note that this is a relation among intensive parameters only. A relation of this type exists for
each species / in each phase in the system. Thus, in a system with m phases and r species,
there are m x r relations of this type that must be satisfied at equilibrium. The equality
of chemical potential in each phase for each species embodied in Eq. (8.85) results in an
additional (m — 1) x r equations that must be satisfied. For each phase it must also be true
that the sum of the mole fractions must equal one, which results in m additional equations.
The total number of equations to be satisfied is therefore
total number of equations = m x r + (m — 1) x r -\- m = 2mr — r + m.
The intensive properties involved in these equations are P, T,r chemical potentials for each
of m phases, and r mole fractions in each of m phases. The total number of variables is
given by

total number of variables =2+mxr+mxr= 2mr + 2.

The number of independently variable intensive properties in the overall system is the
difference between the total number of variables and the number of equations they must
satisfy. Designating the number of independently variable intensive properties as r\, it follows
from our above analysis that

rj = 2mr +2- [2mr - r + w],

which simplifies to
rj = r - m + 2. (8.99)
Equation (8.99) is known as the Gibbs phase rule. It permits us to determine the minimum
number of independent thermodynamic properties from the set {T, P, X\, X2, . ••, Xr} that
must be specified to specify the equilibrium state of a system containing r species and m
phases. For a pure single-phase system (r = 1, m = 1) it indicates that two independent
intensive properties must be specified to establish the state of the system. For a pure ideal
gas, specifying pressure and temperature is sufficient to determine the state of the system,
as we already know from the ideal gas equation of state. The Gibbs phase rule further
stipulates that, for a binary ideal gas mixture, it is necessary to specify three independent
properties from the set {7\ P, X\, X2}.
For a single-species system with two phases present, only one intensive property can be
independently varied. The presence of two phases is a saturation condition. As we discussed
in the previous section, the requirements for equilibrium lead to a relation between temper-
ature and pressure for a van der Waals fluid at saturation. The same is true for any real pure
substance, with the result that the pressure and temperature cannot be chosen independently.
The relation between pressure and temperature for saturated mixtures of liquid and vapor is
referred to as the vapor-pressure curve for the substance because it indicates the equilibrium
pressure of vapor above a liquid surface at a specified temperature. It is also possible for
three phases of a pure substance to coexist in a system at equilibrium. The Gibbs phase rule
then indicates that the number of independently variable intensive properties from the set
8.3 I Phase Equilibria in Binary Mixtures 111

100,000
critical point •
10,000

1,000 liquid
P
r
sat 100 solid
(kPa)
10 -g

0.11

0.01
100 200

Figure 8.10

{T, P} is zero. This implies that there is a unique combination of the intensive properties
from this set that correspond to the equilibrium state with three coexisting phases. Such
a state is termed a triple point for the substance. Triple points for substances are often
specified as reference points for properties or instrument calibration because they are fixed
states that can be uniquely reproduced. Figure 8.10 shows the primary saturation curves
for water (it does not show possible solid-solid phase transitions) and indicates that single
combination of pressure and temperature corresponding to the triple point.
It is noteworthy that the solid-liquid coexistence curve for water in Figure 8.10 has a
negative slope. For most common solids, the slope of this curve is positive. This behavior
is associated with the fact that, unlike most substances, water expands when it freezes, with
the result that solid ice at saturation has a higher specific volume than saturated water at
the same pressure. The Clapeyron Eq. (8.79) indicates that (dP/dT)SSit must be negative if
solidification reduces the enthalpy and increases the specific volume of the substance. For
most substances, freezing decreases both the enthalpy and specific volume, resulting in a
positive slope to the solid-liquid saturation curve. For water the negative slope implies that
if the pressure on ice is increased at constant temperature, the ice will melt. This behavior
facilitates ice skating by causing a liquid film to form under the skate where the pressure
is elevated by the weight of the skater acting on the skate blade. This liquid film provides
lubrication, allowing the skater to glide smoothly over the ice surface. Refreezing of the
liquid film after the pressure is relieved is sometimes referred to as revelation.

8.3 Phase Equilibria in Binary Mixtures

Systems containing a liquid and vapor phase and more than one species are com-
monly encountered in engineering practice. The simplest system of this type is one con-
taining a binary mixture with coexisting liquid and vapor phases. In this section we will
explore the features of such a system in some detail.
The Gibbs phase rule indicates that for a binary mixture (of species a and b) with two
phases present, specification of two properties from the set {T, P, Xa, Xb} will establish the
equilibrium state for the system. We can examine how specification of two properties dictates
272 81 Phase Transitions and Phase Equilibrium

the state by considering the van der Waals model for a binary mixture discussed in Chapter 6.
Using the relations fia/kBT = — (3 In Q/dNa)v,T,Nh and/jLb/kBT = — (^^Q/dNb)VjjNa,
we can differentiate Eq. (6.49) and rearrange the result to obtain the following relations for
the chemical potentials:

k RT v - Xa
'3\ \2nMakBT(v-XaNAbv,a-

($„ - 5),In n — .In <J 1\ (| a -3) .In ( T

Sa

(8.100)

ICRT v — XaN\byta •

\27tMbkBT(v-XaNAbVya-.

. 1 1 fe3)i
In7r - In aStb I 1 In
2 >'"J 2
2av,bbNAXb + 2av,abNAXb
(8.101)

where 0 is the molar specific volume and Xa and Xb are the mole fractions defined as
v = VNA/(Na+Nb), (8.102)
Xa=Na/(Na + Nb), (8.103)
Xb = Nb/(Na+Nb). (8.104)
If we specify the temperature and pressure in the system, as required by the Gibbs
phase rule, we can determine other properties for the equilibrium state by enforcing the
necessary conditions for equilibrium. We therefore take the temperature, pressure, and
chemical potential for each species to be the same in both phases. In addition, in each phase
the sum of the mole fractions must equal one, which allows us to replace Xb with 1 — Xa for
this mixture. We obtain two equations by evaluating the chemical potential for each species
in each phase and setting them equal, and we obtain two more by requiring that the pressure
equation of state be satisfied in both phases. After a bit of simplifying, these relations can
be written in the forms
NA k B T
P =

(8.105)
8.3 I Phase Equilibria in Binary Mixtures 273

p = NA k B T

(8.106)

ex 1 A. v & ^^h 1 A v h) v ~^^& v A v ^7 ^^/? v A v /?

v yk BT v\kBT
_ • (Ol Xg%\NAbyia XbtlNAbytb)Xgty 1
_(OV — Xa,yNb X A ^ ^ ) ^ J '

Vy —

ta - XbiiNAbytb)Xbty 1

With the additional two constraints

XflfV + Xft>v = 1, (8.109)
Xflti + Xft>1 = 1 (8.110)
we have a complete set of equations that can be used to determine the equilibrium conditions.
If T and P are specified, Eqs. (8.105)—(8.110) can, in principle, be solved for the six
unknowns v\, 0v, Xat\, Xb,\, Xfl)V, and Xb,v. Because the equations are nonlinear, this is not
a simple task. However, if we idealize the vapor phase as having ideal gas behavior, we can
simplify the solution process by neglecting the van der Waals terms for the vapor phase
in the equations for the first iteration. A solution can then be generated with the following
algorithm:
(1) For the specified P and T, determine 0v using the ideal gas law: vv = NAkBT/P.
(2) Guess Xath
(3) Compute Xbti using Eq. (8.110).
(4) Solve Eq. (8.106) iteratively to determine £>i.
(5) Solve Eqs. (8.107) and (8.108) simultaneously for X^ v and XbtV.
(6) If the values of Xa?v and XbiW do not satisfy Eq. (8.109), the guessed XOi\ value
is corrected and steps (3)-(5) are repeated. When Eq. (8.109) is satisfied 0v is
recomputed by iteratively solving Eq. (8.105) using the most recent values of XGi\9
Xb,u Xa,v, and XbjV and steps (2)-(6) are repeated. When successive values of
the mole fractions and specific volumes do not change significantly, the algorithm
terminates.
One additional aspect of this model must be considered. The parameters byja, bVjb, av,aa,
and ayfbb are taken to be the pure component van der Waals constants for species a and b.
274 8 I Phase Transitions and Phase Equilibrium

oxygen and nitrogen mixtures at 101 kPa

(extended van der Waals model)
95
|
• bubble point
90 # dew point

85 m

80

75
0.2 0.4 0.6 0.8
oxygen mole fraction

Figure 8.11

The constant aYtab, however, represents interactions of a type molecules with b type
molecules. Accurate specification of this parameter requires knowledge of the potential
force interaction between a type and b type molecules (see Chapter 6).
To demonstrate the use of the analysis described above, we will specifically consider a
binary mixture of oxygen and nitrogen at low temperatures. Figure 8.11 shows the phase
equilibrium diagram for such a mixture at atmospheric pressure predicted by the extended
van der Waals model and computational algorithm outlined above. Species a and b were
chosen to be nitrogen and oxygen, respectively, and the pure component van der Waals
constants were taken to be
aWiaa = 3.76 x 10"49 Pa m6/molecule2, b^a = 4.90 x 10"29 m3/molecule,
(8.111)
49 6 2 29 3
aWibb = 3.78 x 10" Pam /molecule , b^h = 3.98 x 10~ m /molecule.
(8.112)
The aWfaa and aWibb values are the same as those in Table 6.1. The bYfCl and bWib values were
altered from those listed in Table 6.1. The values used here were chosen so that the pure fluid
equilibrium saturation temperature at 101 kPa matched established values of 77 K and 90 K
for nitrogen and oxygen, respectively. To evaluate the interspecies interaction constant, the
following plausible, if ad hoc, relation was used here:

flt
M = '"|fl>'B. (8.113)

Although this extended van der Waals model is crude in many respects, the binary phase
diagram in Figure 8.11 is remarkably close to that indicated by measured data for nitrogen-
oxygen mixtures.
The upper curve in Figure 8.11 is the dew point curve, where condensation first oc-
curs when a vapor mixture of fixed concentration is slowly reduced in temperature. The
lower curve is termed the bubble point curve because it represents conditions where vapor
8.3 I Phase Equilibria in Binary Mixtures 275

first forms when a liquid mixture of fixed concentration is slowly heated. Combinations
of T and oxygen mole fraction below this line correspond to a liquid phase only. Com-
binations above the dew point line correspond to vapor phase only. At points in the lens-
shaped region between the bubble point and dew point lines, a liquid and vapor phase
coexist. In such a two-phase system, the liquid concentration is the value on the bubble
point line at the specified temperature, and the vapor concentration is the value on the
dew point line at the specified temperature. Thus, for this mixture, the concentrations in
the vapor and liquid phases are different at equilibrium, except at the pure species end-
points.

Example 8.3 Use the extended van der Waals model discussed above to estimate the
limits of intrinsic stability for a liquid binary mixture of oxygen and nitrogen at atmospheric
pressure.

Solution As discussed in Section 8.1, the necessary and sufficient conditions

for intrinsic stability for a binary mixture are that (dixa/dNa)v,T,^h > 0 with either species
taken as species a. It can be shown from basic mathematical considerations that

SNjVTN\dNa)VTNJ \dNhJVTNa'
Also, one of the Maxwell relations requires that

The necessary and sufficient conditions for intrinsic stability can therefore be stated as the
requirement that for either fluid taken as a or b

v,T,Nh

For the oxygen and nitrogen mixture considered here, we designate oxygen terms with an
o subscript and nitrogen terms with an n subscript. The necessary and sufficient conditions
for stability can then be stated as

v,T,Nn

dN
n V,T,N0

The three partial derivatives in the above relations are evaluated as follows. First, the right
side of Eqs. (8.103) and (8.104) are substituted into Eqs. (8.100) and (8.101) to replace the
mole fractions with Na and A^. The resulting equations for [ia and \i\, are then differentiated
with respect to Na and Nb. Arbitrarily taking a to be oxygen and b to be nitrogen, we obtain
276 8 I Phase Transitions and Phase Equilibrium

the following relations for the derivatives in the stability criteria relations:

kBT
- XoNAby,o - XnNAbVfn

(Av,0) _
2 2
( - XXNb
(v o A v ,o
N b - XXNb)
nNAbw,n) Xo vNAkBT I'

k BT
N0 + Nn [v-X0

+ (v - XoNAbVtO - XnNAbVtn)2 vNAkBT J'

V,T,N0 No + Nn[v- XoNAby,o - XnNAbv,n

2
( J V A / V , ) L
(0 - XoNAb^o - XnNAby,n)2 Xn vNAkBT \'

To determine the stability limits, the following procedure is used. The pressure is fixedat
one atmosphere, and the mole fraction of oxygen is specified. The mole fraction of nitrogen
is calculate simply as Xn — 1 — Xo. Beginning at a very low value of specific volume, the
above relations for the derivatives are used to compute the left side of the stability condition
inequalities for progressively larger values of specific volume v. For each 0 value, the
extended van der Waals equation of state is used to compute the temperature,

p = NAkBT a,,ooN2AX2o + 2ay,onNJXoXn + av,nnN2AX2n

v - X0NAby,0 - XnNAbYjn v2

Beginning well below the bubble point for the mixture, as the specific volume increases,
T increases and the computed values for the left sides of the stability criterion inequalities
decrease, eventually reaching zero. The point where one of these expressions equals zero
is interpreted as the limit of intrinsic stability. In these example calculations, the same
van der Waals constants were used as in the calculations represented in Figure 8.11. The
resulting limits of stability are plotted for several mixture concentrations in Figure 8.12.
Note that these results imply that the mixtures may be superheated substantially above
the equilibrium bubble point without a loss of intrinsic stability. The fluctuation relations
(4.125) and (4.126) imply, however, that the level of concentration and density fluctuation
in the liquid will become very large as the stability limit is approached.

In some binary systems, the interactions among the two molecular species are such that
the concentrations in the vapor and liquid phases are the same for a mixture at a specific
temperature and pressure. The phase diagram then looks like the plot in Figure 8.13. A
two-phase binary mixture in which the concentrations in the two phases are the same is
termed an azeotrope or an azeotropic mixture. In general, the conditions corresponding to
8.3 I Phase Equilibria in Binary Mixtures 277

bubble point
limit of intrinsic
stability

0 0.2 0.4 0.6 0.8 1

oxygen mole fraction

pressure = P

vapor

liquid

bubble point line azeotropic point

1.0

pure less-volatile pure more-volatile

component b component a

Figure 8.13

the two-phase region in the binary phase diagram vary with pressure, as illustrated in Figure
8.14 for nitrogen-oxygen mixtures.
Figure 8.14 was generated using the extended van der Waals model described above
to predict the phase equilibria at three pressure levels for nitrogen-oxygen mixtures. The
shaded lens-shaped regions are the cross section of a three-dimensional region in which a
liquid and vapor phases coexist. Combinations of T, P, and system oxygen mole fraction
above this region in Figure 8.14 correspond to a single-phase system containing only a vapor
phase. For combinations below the two-phase region, the system contains only a liquid
phase. Although the van der Waals model is very idealized, the variation in the binary phase
diagram indicated by Figure 8.14 is close to that observed for real nitrogen-oxygen mixtures.
278 8 I Phase Transitions and Phase Equilibrium

nitrogen-oxygen mixtures
(extended van der Waals model)

A
110- Ay A
7(K)

100-

//; r /
V /
90-

300

80-

70-
() 0.5
oxygen mole fraction

Figure 8.14 Binary phase diagrams for three pressures as predicted by the extended van der
Waals model for nitrogen—oxygen mixtures.

As discussed in Chapter 6, the model analysis used in this section to predict vapor-
liquid equilibria in binary mixtures represents a simplistic extension of the classical van der
Waals theory. Use of extended generalized van der Waals models to predict vapor-liquid
equilibria in simple binary mixtures has been explored by Penfold et al. [2]. For simple
fluids their results indicate that, as predictive tools, such models may be competitive with
fully empirical equations of state.

Example 8.4 In a cryogenic system, an uninsulated portion of a pipe carrying liquid

nitrogen at atmospheric pressure is in contact with dry air at atmospheric pressure and
20° C. The liquid nitrogen inside is at 77 K and because of heat exchange with the sur-
roundings, the outside wall of the pipe is at the slightly higher temperature of 85 K. A
layer of liquid condensate forms on the outside of the bared pipe. If local thermodynamic
equilibrium exists at the surface of the liquid layer, estimate the oxygen concentration in
the liquid.

Solution In this analysis we idealize the surrounding air as a binary mixture of

oxygen and nitrogen with an oxygen mole fraction of 0.21. At atmospheric pressure, the
phase equilibrium diagram in Figure 8.11 indicates that for local thermodynamic equilibrium
to exist at the liquid-gas interface at 85 K, the local oxygen concentration in the vapor X 02,v
8.3 I Phase Equilibria in Binary Mixtures 279

(on the dew point line) and the liquid Xo2\ (on the bubble point line) must be

This implies that the concentrations in the vapor and liquid phases tend toward these values if
the pipe, and the fluid adjacent to it, are held at 85 K. Thus, the liquid condensate on the pipe
is expected to have an oxygen concentration of approximately 0.78. If the liquid drips off of
the pipe and lands in a warmer location, it will evaporate. Because of the liquid's higher O2
concentration, this may result in a local oxygen concentration that is much higher than the
normal atmospheric concentration of 0.21 at the location where the liquid evaporates. Since
the flammability of many substances increases as O2 concentration increases, condensation
on the liquid nitrogen lines can be a fire hazard if the liquid condensate drips off of the lines
into locations where flammable materials exist.

Empirical Models
An alternative approach to constructing a model of binary phase equilibria is to
make use of empirical models. If the vapor is idealized as being a mixture of independent
ideal gases and the mixture therefore obeys Dalton's law, the partial pressure of each
component Pa is equal to the total pressure multiplied by the vapor mole fraction:
Pa = PXa,v. (8.114)
Two idealized models that relate the vapor partial pressure of a component to its con-
centration in the liquid phase are Henry's law and Raoulfs law. Henry's law states that the
partial pressure Pa is proportional to the mole fraction of component a in the liquid,
Pa=CHXa,l. (8.115)
In the above relation, CH designates the Henry's law proportionality constant. Henry's law
is usually a good approximation for components having low liquid concentrations Xa,\.
Raoult's law states that the partial pressure for component a is given by
Pa = Psx,aXa,l, (8.116)
where Psat,a is the saturation pressure for pure component a at the specified system temper-
ature r , and Xai\ is the liquid mole fraction of species a. Raoult's law is generally a good
approximation for values of XQt\ near one.
If Raoult's law is assumed to apply to a binary system, it follows that
Pa = Psat,aXa,\, Pb = Psat,bXb,\> (8.117)
Since Dalton's law requires that the sum of the partial pressures must equal the system
pressure P,
P = Psat,aXa,\ + Psat,bXb,l- (8.118)
Using the fact that Xa,\ + Xb,\ = 1, the above relation can be rearranged to obtain

X
a,\=~ — 7^7- (8.119)
° C n " ( O
280 8 I Phase Transitions and Phase Equilibrium

In Eq. (8.119) the functional dependence of the pure component vapor pressures on temper-
ature has been explicitly indicated. Thus, if the binary mixture conforms to Raoult's law, the
bubble point curve Xa\(T) on the phase diagram for a specified pressure can be predicted
from the pure component vapor pressure curves using Eq. (8.119). If, in addition, Dalton's
law applies to the vapor mixture at equilibrium, then the partial pressure indicated by
Dalton's law and that indicated by Raoult's law must be equal for each species. It follows that
p Y PY (8.120)
Rearranging the above relation yields

av — ^a lv-* /• ^O.IZIJ

Thus, having determined X a j ( r ) as described above, Eq. (8.121) can be used to determine
the dew point curve Xa^{T) for a specified pressure, completing the phase diagram. The
phase diagram shown in Figure 8.15 was constructed by applying this idealized method
to a mixture of refrigerants R-ll and R-113 at atmospheric pressure. The predictions of
this model are actually fairly close to the actual observed mixture behavior for these fluids
at atmospheric pressure. However, some mixtures deviate strongly from the predictions of
Raoult's law and Dalton's law, and this type of model will not be accurate for such mixtures.

Phase Equilibria in Binary Mixtures - General Considerations

Systems containing two phases and two components arise commonly in appli-
cations and many of these do not conform to the extended van der Waals model or the
ideal mixture model described above. To deal with such systems, we must consider them
from a more basic perspective. The temperature, pressure, and chemical potential for each
component must be the same in both phases. The main difficulty in applying these require-
ments is that we need to relate the chemical potential constraints to other, more measurable

Rl 1/R113 Binary Mixture at P = 101 kPa

(ideal mixture model)
325

320 vapor

315 dew point line

310

305
bubble point line
300
liquid
295
0.0 0.2 0.4 0.6 0.8 1.0

mole fraction of R-11

Figure 8.15
8.3 I Phase Equilibria in Binary Mixtures 281

properties. The extended van der Waals model discussed above accomplishes this, but some
mixtures are not accurately treated by this model. For pure substances, the Gibbs-Duhem
equation can be used to relate molar chemical potential to changes in temperature and
pressure:
djl = -sdT + vdP. (8.122)
Integrating this relation between a reference state (designated with a 0 subscript) and an
arbitrary state, we obtain

(i(T,P) = (i0- [ sdT+ [ vdP. (8.123)

Use of this relation requires that we have a means of evaluating the chemical potential at the
reference temperature and pressure. There is no fundamental thermodynamic requirement
that allows us to specify the absolute value of the chemical potential at some particular
point. As we have shown, statistical thermodynamics allows us to predict an absolute value
of chemical potential if we have an appropriate model of the fluid, such as the van der Waals
model discussed above. However, this is only approximate for many substances. Because
we only desire the ability to compute changes in chemical potential, we elect to evaluate the
reference condition in an arbitrary, but consistent manner for the substances we consider.
If we consider an ideal gas, we can use the ideal gas equation of state to evaluate 0 in
Eq. (8.123) and integrate along a line of constant temperature to obtain

A(7\ P) = Ao + RT ln(P/Po), (8.124)

where R = NAkB is the universal gas constant. This result implies that, for an ideal gas, the
change in chemical potential along a line of constant pressure is simply related to the log
of the pressure ratio. To generalize this relation so it applies to substances other than ideal
gases, G. N. Lewis defined the fugacity f such that for any substance

£(7\ P) = Ao + RT l n ( / / / 0 ) . (8.125)
In this defining relationship either Ao °f /o m a v be chosen arbitrarily, but once one is
chosen, the other must be fixed to satisfy Eq. (8.125) at the reference state. In the limit
of low pressure, if the real substance approaches ideal gas behavior, we expect that / / / 0
approaches P/PQ. The ratio / / / 0 is termed the activity and is often denoted by a separate
symbol a. This term arises from the fact that a = f/fo indicates the difference between the
chemical potential in an arbitrary state and that in the standard state, when both states are
at the same temperature. The associated difference in the chemical potential is presumed to
be an indicator of how "active" the substance will be with respect to mass transfer.
It follows from the definition of fugacity that, at fixed T,

d(l(T, P) = RTd(\n / ) . (8.126)

Substituting into Eq. (8.122), we obtain (for T fixed)
d()nf) = vdP/RT. (8.127)
For fluids in general, we define the compressibility factor Z as
Z = Pv/RT. (8.128)
282 8 I Phase Transitions and Phase Equilibrium

The compressibility factor quantifies the deviation of the fluid from ideal gas behavior. If Z
is close to 1, the behavior is essentially ideal. In general Z differs significantly from one for
dense gases and liquids. The value of Z can be predicted using the virial expansion or van
der Waals models discussed in Chapter 6. Using the definition of Z, we can rewrite (8.127)
as

^ . (8.128)

Subtracting J(ln P) from both sides of Eq. (8.128), we obtain

d[ln(f/P)} = (Z - l)y-. (8.129)

We integrate both sides from zero pressure, where / = P, to an arbitrary state along a line
of constant temperature:

ln(//P) =\ f (Z - 1 ) ^ 1 . (8.130)
L JO * J r=constant
This relation can be solved explicitly for the fugacity to give

U
p
dP}

(Z-l)—\ . (8.131)
* ) Inconstant
Thus, for a pure substance, if we know Z(P, T) we can generate / ( P , T) by executing the
integration indicated in Eq. (8.131) with T held constant.
For each component in a two-phase two-component system, a different relation between
chemical potential and fugacity exists for each of the two phases:
Aa = Aa.O + RT l n (/i//i.o). (8-132)
A? = A?,o + *rin(/ a n // a n o ). (8.133)
Identical relations apply for species b. Equating the right sides of Eqs. (8.132) and (8.133)
yields
A^.o + RT In (fl/flo) = A"o = RT ln (/iV/"o) • (8-134)
Since Eq. (8.125) applies to any two states for a specific substance at the same temperature,
it can be used to relate reference states for the two phases if they are at the same temperature
but have different pressures and concentrations:

Ai,o = A°o + RT In (/j, o //°o) • (8-135)

Substituting the right side of this relation for p}a 0 in Eq. (8.134) and simplifying leads to

fl = fa- (8-136)
Using identical arguments for species b converts the equality of chemical potential to the
condition
fl = fll- (8-137)
Thus if we define the chemical potential reference state in all phases at the same temperature,
the equality of chemical potential in all phases can be replaced by the equality of fugacity
8.3 I Phase Equilibria in Binary Mixtures 283

for each component at equilibrium. In some applications, statement of the equilibrium

conditions in terms of fugacities is advantageous because of their close link to pressure.
Regardless of whether equilibrium is specified in terms of chemical potential or fugacity,
to generate a binary mixture phase diagram like Figure 8.11 we must develop a relation for
fugacity or chemical potential as a function of temperature, pressure, and concentration.
We have shown above that, for a pure substance, / can be determined from the dependence
of the compressibility factor on T and P. Unfortunately, extending this approach to predict
the fugacity of a component in a mixture is generally a very difficult task. For gas mixtures,
it is common to define the fugacity coefficient for component a as

4>a = T ^ , (8-138)

where Xa^ is the mole fraction of species a in the vapor phase. For liquid mixtures, it is
conventional to define the activity coefficient for component a as

Ya = r r ^ , (8.139)

where Xa\ is the liquid mole fraction of species a and fa$ is the fugacity of component a
at some convenient reference condition. With these definitions, the equality of fugacity for
species a in the liquid and vapor phases can be written as

(t>aXa,yP = YaXa,lfa,0. (8-140)

With this formulation, we can establish relations among Xa>v> Xfl>i, / \ and T at equilibrium
if we can develop relations for (pa and ya as functions of these variables:
0a=0fl(7\/>,Xfl,v,XM), (8.141)
Ya = Ya(T9P,Xa,hXb,i). (8.142)
If the vapor is a mixture of ideal gases, the definition of fa dictates that (j)a = 1. An ideal
liquid mixture is taken to correspond to ya = 1 with fa$ equal to the saturation pressure of
pure liquid species a at the given system temperature. If the gas and liquid phases are both
ideal mixtures, Eq. (8.140) reduces to
Xa,vP = XaAP^a. (8.143)
Since the left side of (8.143) is just the partial pressure of vapor component a, this relation is
equivalent to Raoult's law. Note that identical reasoning will generate relations equivalent
to Eqs. (8.138)—(8.143) for species b. The value of this formulation is that it provides a
framework to account for the deviation of real mixtures from the ideal behavior embodied
in Dalton's law and Raoult's law. The deviation of (/>a, (/>b,Ya, a n d Yb fr°m one indicate the
degree to which these mixtures deviate from the ideal behavior.
It is possible to obtain phase equilibrium data experimentally and use it directly to
establish a phase equilibrium diagram or to develop empirical relations for the fugacity and
activity coefficients as functions of temperature, pressure, and concentrations. Analytical
determination of relations for </>a, <j>b, ya, and Yb f° r nonideal mixtures requires that we
incorporate methods to accurately model the force interactions between molecules in the
vapor and liquid phases. Often this is a challenging task because the nonideal behavior is
a consequence of complex force interactions among the molecules resulting from factors
284 8 I Phase Transitions and Phase Equilibrium

such as the nonspherically symmetric nature of the molecules and the polar nature of the
molecules. The effects of these molecular characteristics on fugacity and activity coefficients
are beyond the scope of this text. The interested reader may find a detailed discussion of
these issues in the text by Prausnitz et al. [3].

8.4 Thermodynamic Similitude and the Principle of Corresponding States

Models such as the van der Waals model and Redlich-Kwong model discussed
in Chapter 6 can be used to predict fluid properties in either the vapor or liquid phase.
An interesting feature of these models is that they suggest that values of thermodynamic
properties of different pure materials can be treated in a unified way when the property is
normalized with its value at the critical point. Beginning with the general form of the van
der Waals equation of state,

P = J^L- - ^#, (8.144)

v — N\bv v1
we can evaluate the constants ay and by by requiring that the slope of the van der Waals
isotherm and its curvature be zero at the critical point. The latter requirement establishes
an inflection point there. Invoking these requirements results in the following relations for
the van der Waals constants:

(O45)
3^ = 'IF'
(8.146)

We define reduced properties, denoted with an r subscript, by normalizing each with its
value at the critical point:
Tr = T/Tc, PY = P/Pc, vr = v/vc. (8.147)
Replacing the physical properties in Eq. (8.144) with the product of the reduced property
and its value at the critical point, we can write the van der Waals equation of state in terms
of reduced properties as

Pv=^rJL o- (8.148)
3vr - 1 v]
In addition, it is also easily shown that the compressibility factor Z = Pv/RT can be
expressed in terms of reduced properties as

Z = ZC^, (8.149)

where for a van der Waals fluid, the critical compressibility factor Zc = Pcvc/RTC is
0.375. Since in principle, the above two relations (8.148) and (8.149) can be combined
to eliminate vr as a variable, they imply that the compressibility factor can be consid-
ered to be only a function of reduced pressure and temperature Z = Z(Pr, TY). For a van
der Waals fluid this functional dependence can be determined by specifying P x and Tr,
solving Eq. (8.148) iteratively for physically possible vx values, and then computing Z
8.4 I Thermodynamic Similitude and the Principle of Corresponding States 285

0 1 2 3 4 5 6 7

1.0
—•———««
— -| Tr =2.0
Tr = 1-5 - ^-—•
0.8
s T i q

i \ 1

\ 1^V
0.6 \ \

i \ -\
'• « N T ' = 1 . 1 . •y
0.4

0.2
= 1.0
curvefits to data r
i

0 1 2 3 4 5 6 7

(b)
Figure 8.16

from (8.149) with Z c = 0.375. The variation of Z = Z(P r , TT) computed in this manner
for a van der Waals fluid is shown in Figure 8.16a. The compressibility data for some
real fluids can, in fact, be collapsed onto a single curve in this way. The prediction for a
van der Waals fluid is qualitatively the same as for many real fluids, but the quantitative
accuracy varies considerably. Figure 8.16b shows the variation of Z = Z(F r , TT), which
closely fits data for water, CO2, nitrogen, and a variety of hydrocarbons presented by
Su [4].

Example 8.5 Use the compressibility charts to determine the density of R134a at 120°C
and 4.0 Mpa.
286 81 Phase Transitions and Phase Equilibrium

Solution From the Table Ills in Appendix III, the properties of R134a at the
critical point are
Tc = 101.0°C, Pc = 4.056 kPa, vc = 0.00195 m3/kg.
Since the molecular mass is 90 for R134a, it follows that
0c = Mvc = (90)(0.00195) = 0.1755m3/kmol.
The compressibility factor at the critical point is therefore determined as
6
Z = HsUl = 4-056 x l O
O ^^ . 1755) =
c
RTC (8308)(101+273)
For the conditions of interest, the reduced temperature and pressure are
PY = p/pc = 4.0/4.056 = 0.986,
TY = T/Tc = (120 + 273)/(101 + 273) = 1.050.
Using Figure 8.16b, for the above values of PY and TT, we find that
Z = 0.64.
Solving Eq. (8.149) for 0r, and substituting, yields
= ZT\_ = 0.64(1.05) = ^
Vr
Zc P r 0.229(101+273)
The molar specific volume and density are then computed as
0 = vrvc = 2.97(0.1755) = 0.521 m3/kmol,
p = \/v = 1/0.521 = 1.92kmol/m3.
The molar density indicated above is within 14% of the value obtained by reading the mass
density from the R134a P-h diagram in Appendix III and converting it to molar density by
dividing by the molecular mass.

For a van der Waals fluid, the difference between jl/T and its value at the critical point
jlc/Tc can also be expressed as a function only of reduced temperature and volume:

In7 r - In
vr- 1 2 2 \ 2 J 4 \vYTY
(8.150)
where

Ac 2nMkBTc(vc-NAby)2/3
In
vc-NAby
8.4 I Thermodynamic Similitude and the Principle of Corresponding States 287

This implies that when we invoke the necessary conditions for equilibrium in a two-phase
system, the requirement of equality of chemical potential can be expressed in terms of
reduced properties. For a van der Waals fluid, the deviation of the molar specific entropy
from its value at the critical point can also be expressed as a function of reduced properties:
S
~ "- = | lnrr + In t3^-^] , (8.152)
NAkB 2
where

NA k B \2
-5
In it — In <7S (8.153)
7

For a system containing the liquid and vapor phase of a pure van der Waals fluid
Eqs. (8.148) and (8.150) can be used to state the necessary conditions for equilibrium
in terms of reduced properties. It follows from these equations and the definitions of the
reduced properties that at equilibrium the reduced pressure and temperature must be the
same in both phases. Using Eq. (8.150) to evaluate the chemical potential in each phase and
setting the resulting expressions for the liquid and vapor chemical potentials equal to each
other and using the fact that the reduced temperature is the same in both phases yields

— ln(30ri — 1) — — — = — - — ln(30 rv — 1) — — — ,
*^^r,l A ^ ^r, 1 -* r, sat ^^r,v *

(8.154)
where 0r>i and vr^v are the reduced molar specific volumes in the liquid and vapor phases,
respectively. The van der Waals equation of state must also be satisfied in both phases:

A,sat = *Tx^\ ~ ^ - , (8.155)

3ur,i - 1 v;Y
87r sat
/>rsat = ' - 4". (8.156)
3ur,v - 1 i^ v
For specified Tr sat, the above three nonlinear equations can be solved for Fr,sat, 0 r j, and
0 r v . The variations of PY^at with rr>sat and 0rj and 0 r v with T rsat obtained by iteratively
solving these equations are shown in Figures 8.17 and 8.18. In Figure 8.17, F r s a t is plotted
as a function of \/Tr on a semilog plot to demonstrate that the van der Waals model predicts
a nearly linear variation of ln(F rsat ) with 1 / Tr, which was suggested by our consideration of
the Clapeyron equation in Section 8.2. In Figure 8.18, the vapor values of reduced specific
volume correspond to vr > 1 and liquid values correspond to 0r < 1.
It is also noteworthy that the relation for the enthalpy of a van der Waals fluid derived
in Chapter 6 can be manipulated to obtain the following relation for the change in molar
specific enthalpy for conversion of saturated liquid to saturated vapor:
AJ, = r r 3 ^ _ 3 ^ 1 »r. _ n (8,57)

The pairs of liquid and vapor reduced specific volumes on the right side of the above
relation are functions of only TY. This implies that Ah\y/NAkBTc is a function only of
288 8 I Phase Transitions and Phase Equilibrium

A
0.8

A
0.6
O van der Waals fluid
• nitrogen
0.4 A water
• methane
T argon
0.2 .—1 • I 1 , •—, • • i . ,

0.2 10 100 200

Figure 8.17

1
7 1
O van der Waals
fluid
P + nitrogen
r
r,sat
A water
• methane

0.1- \ T argon
i
\
A
9
\

A \
0.01-

0.004- A
1 1.5 2 2.5 3
\ITr

Figure 8.18

reduced temperature for a van der Waals fluid. For the values of Tr indicated in Figure
8.18, corresponding pairs of 0r values computed using the van der Waals model can be used
to determine A/iiv/'N\k^Tc. The resulting variation of £Ji\w/NpJz^Tc with TY is shown in
Figure 8.18.
In addition to the computed results for the van der Waals model, points representing
established properties for several real pure fluids are plotted in reduced parameter form in
Figures 8.17-8.19. It can be seen that, although the qualitative behavior of the van der Waals
model is consistent with that of the real fluids, the predictions of the van der Waals model
differ substantially from the data for realfluids,particularly at low temperatures. The trends
in the data for real fluids in these figures is linked to the nature of the molecules. For argon,
8.4 I Thermo dynamic Similitude and the Principle of Corresponding States 289

10
O van der Waals
fluid
• nitrogen
• water
• methane
•
6--
T argon
•

v
1 ii
0.2 0.4 0.6 0.8

Figure 8.19

nitrogen, and methane, the molecules are nearly spherically symmetric geometrically and in
terms of their potential interactions with neighboring molecules. The points for these fluids
lie essentially on the same curve in these figures. The points for water differ significantly
from those for argon, nitrogen, and methane. Water molecules have a nonspherical struc-
ture and their polar nature results in nonspherically symmetric potential force interactions
with neighboring molecules. The trends in these data suggest that the corresponding states
framework implied by the van der Waals model works well when the molecules are close
to spherically symmetric, but deviations from such symmetry cause the thermodynamic
similarity to break down.
In Chapter 6, we derived the van der Waals equation of state from an interaction potential
model that contained two parameters: the depth of the potential well and a distance parameter
a L j. It is perhaps not surprising then that we can characterize the behavior of a van der
Waals fluid in terms of two macroscopic parameters P c and Tc (or equivalently aw and
bY). The behavior of some real fluids is well characterized with two parameters. Generally,
this is true for molecules that nominally exhibit spherical symmetry in their shape and in
their force interactions with other molecules. For many substances, two parameters is not
enough to provide useful correlation accuracy. Molecules that show the strongest deviation
from the two-parameter corresponding states behavior are most commonly those having
geometry and interaction potentials that are extremely nonspherical and those with strong
dipole moments. Intermediate molecules that are slightly nonspherical and/or have weak to
moderate dipole moments behave in a manner similar to that predicted by the two-parameter
corresponding states model, but they exhibit significant deviations from its predictions for
some states.
An obvious means to patch up the corresponding states model so as to account for
nonsymmetric and polar behavior is to include an additional characterizing factor. One way
of doing so is to include the critical compressibility Z c = Pcvc/RTC as a third parameter.
Values of Z c vary among different fluids. For a van der Waalsfluid,we noted that Z c = 0.375.
290 8 I Phase Transitions and Phase Equilibrium

The values of Z c computed from measured critical properties for argon, nitrogen, and water
are 0.291, 0.290, and 0.236, respectively. Separate Z = Z(PY, TT) charts established for
different Z c values do provide improved accuracy in the predicted fluid properties.
An alternate means of adding a third parameter to the corresponding states representation
is to include the acentric factor co defined by Pitzer et al. [5] as

co = -1.0 - log10(Pr,sat)rr=0.7. (8.158)

The rationale for this parameter is as follows. For molecules that exhibit a high degree of
spherical symmetry (such as Ar, N 2, and CH4), the variation of Fr?sat with TT shown in Figure
8.17 indicates that Pr>sat equals 0.1 at TT = 0.7(1 / TY = 1.43). Thus, for symmetric molecules
the base ten log of P r s a t is —1. The acentric factor, co, which is the difference between —1
and Iog10(/>r?sat)7;=o.7? is therefore an indicator of the deviation of a given molecular species
from behavior exhibited by spherically symmetric molecules. In this sense it is a logical
choice as an additional parameter to quantify nonspherical and/or polar characteristics of
molecules. Values of co for a wide variety of fluids are tabulated in the text on gas and liquid
properties by Reid, Prausnitz, and Poling [6]. As suggested by the water points in Figure
8.17, water has a positive value of co of about 0.34. For many common fluids, recommended
values of co are positive and less than 0.5. A few, such as neon, have recommended values
that are negative. Pitzer et al. [5] proposed a relation for the compressibility factor having
the form

Z = Z0(Tr, Px) + coZx{Tx, PT), (8.159)

where the ZQ function applies to molecules exhibiting spherical symmetry and coZ\ is a
correction for nonsymmetric behavior. Pitzer et al. [5] tabulated values of Z o and Z\ as
functions of Tr and Pr. Updated and extended versions of these tables are provided by Reid
et al. [6]. Further discussion of extensions of the corresponding states framework may also
be found in the text by Reid et al. [6].

Near Critical State Behavior

In our examination of phase equilibria for a pure system we found that the molar
specific volumes of a coexisting liquid and vapor phase change with saturation temperature.
As the temperature is increased in a pure two-phase system of fixed volume, the specific
volume of the vapor decreases and the specific volume of the liquid increases. As the
temperature approaches the critical temperature, the molar specific volume of both phases
approaches the specific volume at the critical point. The critical state is of special importance
to the thermodynamic characteristics of the substance for two reasons. First, it defines the
ranges of temperature and pressure where a liquid-vapor phase transition is possible. In
addition, the critical state is of special interest because the thermodynamic characteristics
of a pure fluid system become quite extraordinary near the critical point. If a system molar
specific volume is equal to vc and the temperature is higher than the critical temperature,
the state of the substance is said to be supercritical. If the temperature of the system is
gradually lowered toward the critical temperature, properties behave in an anomalous way
as the critical point is approached. The critical isotherm has zero slope and zero curvature
(an inflection point) at the critical point. Equations (8.5) and (8.6) imply that density and
energy fluctuations become infinite as (dP/dv) T -> 0. At the critical point the system
8.4 I Thermodynamic Similitude and the Principle of Corresponding States 291

thus does not have a well-defined equilibrium state in the usual thermodynamic sense. The
resulting density fluctuations cause light passing through the fluid to be scattered so that the
fluid intermittently looks opaque in localized regions within the system. This phenomena
is referred to as critical opalescence.
In addition, because (dP /dv)r approaches zero as the critical point is approached, the
isothermal compressibility KT = - (\/V)(dV/dP)T,N = - (l/v)(dv/3P)T approaches
infinity there. This implies that the small hydrostatic pressure variation in the system will
produce a very large density gradient from the top to the bottom of the system. Any systems
on the earth that are close to the critical state strongly depart from our usual definition of an
equilibrium state, and our usual definitions of thermodynamic properties break down in this
range of conditions. In addition to the theoretical complexities associated with this behavior,
the high levels of property fluctuations and the gravity-induced variation of density makes
experimental determination of physical properties very difficult near the critical point.
A widely used means of quantifying property behavior near the critical point is to examine
how properties deviate from their value at the critical point in the near-critical region. The
usual approach is to postulate a power law variation and quantify the behavior in terms of
the exponent of the power law dependence. For a fluid system of the type considered here,
it is common to define pressure, temperature, and volume parameters as

(8.160)
t = TY-l = (T- Tc)/Tc, (8.161)
v = vY-l = (v - % ) / % . (8.162)
These parameters are obviously defined so that their deviation from zero indicates the
fractional deviation of the associated physical property from the critical value. When t is
positive (T >TC), only a single value of v is possible, whereas when iis negative (T < Tc), v
can take on two values: v\ for the liquid phase and vv for the vapor phase. To characterize
the behavior near the critical point, we are interested in the power law variations:

p ~ v 8, (8.163)

(8.164)

(8.165)

(8.166)

The exponents in the power law variations above are termed critical exponents for the
system. They can be determined explicitly for a van der Waals fluid. Using Eqs. (8.160)-
(8.162) to replace PT, TY, and vY in the van der Waals equation of state (8.148), we obtain
AM _l_ t\ 1
1. (8.167)

Expanding the right side to third order in the small variables t and v we find

p = --v3 + f(4 - 6v + 9v2) + • •

•. (8.168)
292 81 Phase Transitions and Phase Equilibrium

This implies that 8 = 3 and the critical isotherm (t = 0) varies as p = -(3/2)u 3 near the
critical point. Differentiating this relation, we can also show that

KT = \PJ-\ (8.169)
6
which implies that y = y' = 1.
As discussed above, the values of molar specific volume for saturated liquid and vapor
are dictated by the requirements that temperature, pressure, and chemical potential must
be the same in both phases. In the near-critical region we invoke the requirement of equal
pressure and temperature by evaluating the right side of Eq. (8.168) for the saturated liquid
at t and v\ and setting it equal to the right side of Eq. (8.168) evaluated for saturated vapor at
t and vw. Doing so and retaining terms to third order yields, after a little rearranging we get

(v3y - tf) + 4t(vy - d) - 6? (v2 - v2) = 0. (8.170)

Dividing through by (uv — v\), we can reduce this to
(v2 + vivy + v?) + 4t-6t(vy + CO = 0. (8.171)
To invoke the requirement of equal chemical potential and temperature, we use Eqs. (8.160)-
(8.162) to replace the PY, TY, and 0r terms in Eq. (8.154) and expand the resulting terms for
small t, v\ and vw. Rearranging the resulting equation, we obtain
(v2 + v\vy + vx2) + At - At2 - 4t(vv + DO = 0. (8.172)
In principle, Eqs. (8.171) and (8.172) solved simultaneously dictate v\ and vY for specified t
near the critical point. These relations can be rearranged to the following equivalent forms:

3(vy+ Vl) 2 + (W - vx)2 -f 1 6 f - 2At(vv Hh ^0 = 0, (8.173)

2
vx)2- r-(t)v - vx)2 •+•1 6 f - 16f - 16f(i5v Hh Vx) = 0. (8.174)
When two phases are present, 0v is larger than vc and Oi is smaller than 0c. It follows directly
that vy must be positive and v\ must be negative. As a result vy + v\ is much smaller than
t)v — v\. We therefore neglect the 3(i)v + vx)2 term in the above equations as well as the t
term in (8.174) (compared to the t term) and combine Eqs. (8.173) and (8.174) to eliminate
t>v + v\. After simplification, the resulting relation becomes
Dv - U! = 4 ( - f ) 1 / 2 , (8.175)
which implies that the critical exponent f$ is 1/2. Using similar reasoning together with the
other property relations for a van der Waals fluid it can be shown that the critical constants
a and a' are both zero for a van der Waals fluid.
For real purefluids,the critical exponents generally span a range of values. The exponents
a and ot' generally lie between —0.2 and 0.3. For real fluids, ft is typically between 0.3 and
0.5 and 8 is typically between 4 and 5. For the isothermal compressibility of purefluidsthe
critical exponents generally lie in the ranges 1.2 < y < 1.4 and 1.0 < y' < 1.3. Figure
8.20 shows the computed variation of t)v — v\ with t indicated by the iteratively computed
saturation values of molar specific volume for a van der Waals fluid plotted in the figure.
Also shown are the established data for nitrogen and the near-critical variation for a van der
Waals fluid indicated by Eq. (8.175). The near-critical relation agrees well with computed
Exercises 293

1 oo-
~r
nitrogen data
van der Waals fluid

approximate
asymptote
for nitrogen

Figure 8.20

results for the van der Waals fluid even at large deviation from the critical temperature. The
data for nitrogen exhibit a slope that implies a /3 value of 0.45.
In addition to the pure fluid system examined here, critical phenomena arise in a number
of other physical systems, including multicomponent systems and systems in which electro-
magnetic effects are important. The limited discussion here only considers a small portion
of the larger statistical physics framework that has been developed to analyze characteristics
of near-critical systems. The interested reader can find further information on this topic in
the texts by Callen [7], Stanley [8], and Yeomans [9].

Exercises
8.1 Derive Eqs. (8.39) and (8.40) from Eqs. (8.35) and (8.36).
8.2 Use the van der Waals model to determine the variation of the chemical potential with
pressure for methane at T = 160K. Use a computer program to determine /x/kBT as a
function of v and simultaneously compute P for each value of v along the isotherm. Plot
your results and interpret the nature of each segment of the isotherm. Also compare the
equilibrium saturation pressure indicated by your model results to the established value of
1,592 kPa at this temperature.
8.3 Use the van der Waals model to determine the variation of the chemical potential with
pressure for argon at T = 130 K. Use a computer program to determine JJL/ &B T as a function
of 0 and simultaneously compute P for each value of v along the isotherm. Plot your results
and interpret the nature of each segment of the isotherm. Also compare the equilibrium
saturation pressure indicated by your model results to the established value of 2,027 kPa at
this temperature.
8.4 If liquified natural gas (methane) leaks from containers on a ship it may spill onto the surface
of water near the ship, which is at a temperature of 15°C. The liquid methane at atmospheric
pressure that contacts the water may be heated close to 15°C. Liquid methane is known
to vaporize explosively when heated close to the intrinsic limit of stablity. Use the results
for the van der Waals model for a pure substance to predict the temperature at the limit of
intrinsic stability for liquid methane at atmospheric pressure. Repeat the calculation using
the Redlich-Kwong model. Use model constants from the tables in Chapter 6 for these
294 8 I Phase Transitions and Phase Equilibrium

calculations. What are your conclusions regarding the possibility of explosive vaporization
of methane when it leaks from container ships?
8.5 Use the results for the van der Waals model for a pure substance to predict the temperature
at the limit of intrinsic stability for water at atmospheric pressure. Repeat the calculation
using the Redlich-Kwong model. Use model constants from the tables in Chapter 6 for
these calculations.
8.6 Using the Redlich-Kwong model, estimate the minimum temperature at which water vapor
is intrinsically stable at atmospheric pressure.
8.7 A vessel containing saturated liquid oxygen at 140 K is slowly depressurized in such a
way that the temperature of the contents remains constant. The liquid oxygen is expected
to vaporize violently when the liquid reaches the intrinsinc limit of stability. Estimate the
pressure at which the liquid will vaporize violently.

8.8 A vessel containing saturated nitrogen vapor at 80 K is slowly compressed in such a way
that the temperature of the contents remains constant. The nitrogen vapor is expected to
suddenly initiate formation of liquid droplets when the vapor reaches the intrinsic limit of
stability. Estimate the pressure at which droplet formation is expected to be observed.

8.9 The equation of state (6.60) and chemical potential relation (6.65) for the Redlich-Kwong
model discussed in Chapter 6 can be written in terms of molar specific volume v and molar
chemical potential (/x) as follows:

= NAkBT __ aRNj
0-NA6R Tl/H(v + NAbR)'
- 5)

(g-3) / J_\ _ aKNl (v + NAbK

n
2 \ e ) (Nb)NkTyt "

NAbR)

These equations can be used to predict the saturation conditions for a fluid if the constants
for the Redlich-Kwong model are known. At equilibrium, the pressure, temperature, and
chemical potential must have equal values in the liquid and vapor phase. The saturation
values of P,v\, and 0v for a specified T can be calculated as follows:
1. A value of pressure is guessed to initiate the scheme.
2. For the value of pressure and temperature Eq. (i) is iteratively solved for £>i and 0v.
3. The values of Oi and T are inserted into Eq. (ii) to compute /xb and the values of 0v and
T are inserted into Eq. (ii) to compute /xv.
4. If /2V equals /xl9 the guessed value of P is correct, as are the values of v\ and vv computed
in step 1. If /xv does not equal jlx, a correction to P is generated using a Newton-Raphson
scheme with fix — /2V as the error function. The calculation scheme then returns to step 2.
Write a computer program to use the algorithm above to determine Tsat(P) for R 134a at
atmospheric pressure andfiveother points between the critical pressure and one atmosphere.
Use values of aR and bR for R134a from Table 6.2 with f = 6 and a s = 1. The value of 9roUm
for R134a is unknown, but a value of 5 K can be used here without loss of accuracy. Run
the program and tabulate and plot your results. Compare the predictions of this model with
saturation data for R134a tabulated in Appendix III.
References 295

8.10 Air in the atmosphere at 101 kPa and 20° C has a water vapor mole fraction of 0.02. Consistent
with Dalton's law, the water vapor behaves as if it alone occupied the volume accessible
to the gas at its partial pressure and the mixture temperature. This implies that when the
system is in equilibrium, condensation of water vapor may occur when the partial pressure
of the vapor equals the saturation pressure of water at the system temperature. Similarly,
for a metastable mixture, the limit of intrinsic stability is expected to correspond to the
partial pressure of water vapor being equal to the pressure at the intrinsic stability limit for
pure water vapor at the system temperature. If the mixture is cooled at constant (overall)
pressure, at what temperature will the intrinsic limit of stability be reached? (Hint: Use the
van der Waals model to predict the intrinsic limit.)
8.11 Derive Eq. (8.100) from Eq. (6.49) from Chapter 6.
8.12 Use the extended van der Waals model discussed above to estimate the limit of intrinsic
stability for a binary mixture of oxygen and nitrogen vapor at atmospheric pressure with a
nitrogen mole fraction of 0.3.
8.13 Use the extended van der Waals model to estimate the temperature limit of intrinsic stability
for pure oxygen vapor at atmospheric pressure. To reduce the possibility of condensation
when the gas is cooled at constant pressure, it is suggested that nitrogen vapor be added to
make a mixture with a nitrogen mole fraction of 0.1. Evaluate whether this will have the
desired effect.
8.14 Derive Eq. (8.150) from Eq. (8.51).
8.15 Derive Eq. (8.154) from Eq. (8.150).
8.16 Derive Eq. (8.157) from Eq. (6.29).
8.17 (a) Derive the reduced equation of state Pr — Pr(Tr, 0r) for the Redlich-Kwong model,
(b) For the Redlich-Kwong model, derive the relation that embodies equivalence of chemical
potential in coexisting liquid and vapor phases at equilibrium in terms of reduced properties.
(This should look similar to Eq. (8.154) derived for the van der Waals model.)
8.18 Derive Eq. (8.168) from Eq. (8.167) for small t and v.
8.19 Derive Eq. (8.169) from Eq. (8.168) for small t and v.
8.20 Plot the saturation data for methane from Appendix III near the critical point and estimate
the value of 0 in Eq. (8.165).

References
[1] Modell, M. and R. C , Reid. Thermodynamics audits Applications, 2nd ed., Prentice-Hall,
Englewood Cliffs, NJ, 1983.
[2] Penfold, R., J. Satherley, and S. Nordholm. "Generalized van der Waals Theory of Fluids.
Vapour-Liquid Equilibria in Simple Binary Mixtures," Fluid Phase Equilibria, 109: 183-204,
1995.
[3] Prausnitz, J. M., R. N. Lichtenthaler, and E. Gomes de Azevedo. Molecular Thermodynamics of
Fluid-Phase Equilibria, Prentice-Hall, Englewood Cliffs, NJ, 1986.
[4] Su, G-J. "Molecular Law of Corresponding Sates for Real Gases," Industrial and Engineering
Chemistry, 38:803, 1946.
[5] Pitzer, K. S., D. Z. Lippmann, R. F. Curl, C. M. Huggins, and D. E. Petersen. "The Volumetric
and Thermodynamic Properties of Fluids. II. Compressibility Factor, Vapor Pressure and
Enthalpy of Vaporization,"/. Am. Chem. Soc, 77:3433, 1955.
296 8 I Phase Transitions and Phase Equilibrium

[6] Reid, R. C , J. M. Prausnitz, and B. E. Poling. The Properties of Gases and Liquids, 4th ed.
McGraw-Hill, New York, 1987.
[7] Callen, H. B., Thermodynamics and an Introduction to Thermostatistics, 2nd ed. John Wiley &
Sons, New York, 1985.
[8] Stanley, H. E., Introduction to Phase Transitions and Critical Phenomena, Oxford University
Press, New York, 1971.
[9] Yeomans, J. M., Statistical Mechanics of Phase Transitions, Oxford University Press, New
York, 1992.
 CHAPTER 9

Nonequilibrium Thermodynamics

The statistical and classical thermodynamics framework developed in Chapters

1-8 of this text is based on analysis of systems at equilibrium. In Chapter 9 we explore the
extension of this framework to systems that are not in equilibrium. This chapter focuses
on systems that exhibit steady spatial variations of properties. Systems of this type are
modeled as having local thermodynamic equilibrium and obeying a linear relation between
fluxes and affinities. Analysis of microscale features of such linear systems is shown to
link correlation moments and kinetic coefficients. The Onsager reciprocity relations are
subsequently derived. Thermoelectric effects are examined as an example application of
the nonequilibrium linear theory developed in this chapter.

9.1 Properties in Nonequilibrium Systems

The thermodynamic theoretical framework developed in previous chapters of this
text is limited to analysis of equilibrium states. Often, however, it is the process that takes
the system from one state to another that is of primary interest. Overall changes accom-
plished during the process can be determined by analyzing the initial and final states using
equilibrium thermodynamics. If the process is very slow, it may be well approximated by a
sequence of equilibrium states, and a quasistatic model may adequately predict the outcome
of the process.
In many real processes, the departure from equilibrium is so severe that the quasistatic
model is too inaccurate to be useful. The objective of this chapter is to develop thermo-
dynamic tools that can be applied to irreversible processes in nonequilibrium systems. In
doing so, we also want to examine how such processes relate to the microscale features of
nonequilibrium systems.
In the equilibrium theory developed thus far, the system considered for analysis is pos-
tulated to be macroscopically (spatially) homogeneous. Real systems often are not in equi-
librium and are not spatially homogeneous. An example of a system in which a steady
nonequilibrium condition exists is shown in Figure 9.1. In this system, the opposite walls
of a container of gas are held at different temperatures. The separation distance between
the walls is 0.1 m and the pressure in the gas is one atmosphere.
If a small temperature sensor is used to measure the temperature at several locations
between the opposing walls, the readings would indicate a linear variation of temperature
with distance between the walls having different temperatures. Measurements of this type
are meaningful if the sensor is small enough that molecules that interact with it have energy
distributions close to the Boltzmann distributions characterizing an equilibrium system at
the temperature indicated by the sensor. Three conditions must be satisfied for this to be
true:

(1) The region containing molecules interacting with the sensor must be small enough
that variation of the mean properties of the molecules in the region is small.

297
298 9 / Nonequilibrium Thermodynamics

(2) The region interacting with the sensor must contain enough molecules so that a
large number of molecules interact with the sensor within a time r m that is short
compared to macroscopic time scales associated with the process. This ensures
that the sensor reading accurately reflects the mean properties of the molecules
in the region and that a well-defined property exists within the time available to
measure it.
(3) The region interacting with the sensor must be large compared with the mean free
path of the molecules so that the sensor is responding to collisions with molecules
that underwent their last collision within the region of interest.
In the next chapter, we will further explore departures from thermodynamic equilibrium.
For now it is enough to acknowledge that if all these conditions are met, measurement of
local properties will be meaningful within subregions of the size specified.
These observations suggest that in general we may expect that local thermodynamic
properties can be defined for regions satisfying the three conditions above. Continuum
models of systems in which conditions are not spatially uniform are generally based on
the assumption that differential regions within the system satisfy the above three condi-
tions. Along with this assumption of local thermodynamic equilibrium, it is also generally
assumed that the functional relations among thermodynamic properties that hold for equi-
librium systems also hold among local thermodynamic properties. This allows us to use
all the analytical machinery developed in previous sections to relate local properties in our
analytical treatment of nonequilibrium systems. While these idealizations are not fully valid
in all cases, they provide a useful analytical framework for many real systems.

9.2 Entropy Production, Affinities, and Fluxes

We begin our development of nonequilibrium thermodynamics by considering a
differential volume of a continuous system that is not in equilibrium. The system contains
a binary mixture of two particle types a and b. We adopt the idealization of local thermody-
namic equilibrium discussed in the previous section. It follows directly that the fundamental
relation relates local properties in the differential volume:
= S(U,V,Na,Nb). (9.1)
9.2 / Entropy Production, Affinities, and Fluxes 299

For convenience, we elect to work in terms of properties per unit volume. Since the
fundamental equation is first order, we can divide both sides by the volume of the local
differential system to obtain a relation of the form
S = S(U,Na,Nb), (9.2)
where the S, U, Na, and Nb are the entropy, internal energy, and number of molecules per
unit volume. Expanding Eq. (9.2), retaining first-order terms, and using the definitions of
temperature and chemical potential yields
~ dt) [ia ~ \ih ~
dS= — - ydN a - ydNb. (9.3)
The significance of Eq. (9.3) is that it implies that transport of energy or mass into the volume
changes U, Na> a n d Nb a n d therefore changes the entropy of the system. Thus, the rate of
change of entropy in the system can be linked to the rates of change of energy and mass in
the system. We therefore treat entropy as a transportable quantity and define J 5 , Ju, Ja, and
Jb as vector fluxes of entropy, energy, species a, and species b in the continuous system,
respectively. Note that these are vectors which have units of entropy, energy or number
of molecules per unit area per unit time. For the differential volume with total differential
surface area A, it follows directly from these definitions that

d§= [ Js-dA, (9.4)

JA

dU = 3v dA, (9.5)
IA
dN*a=
a = f 3aa- dA,
I 3 (9.6)
IA
JA

dNb= [ 3b-dA. (9.7)

JA
In the above relation dA is a vector with magnitude equal to the differential area element in
the integration and direction normal to the surface element. The integral is carried out by
summing the integral contributions of the six faces of the volume element in Figure 9.2.
Substituting Eqs. (9.4)-(9.7) into Eq. (9.3) yields, after combining integral terms,
l a
T T i ^
JS — — JU -r — .

This relation is valid for an arbitrary differential volume only if the integrand is zero. It
follows that

3s = \3u ~ y J . - yib. (9.9)

Equation (9.9), which relates the entropy flux at an arbitrary location tofluxesof energy and
mass, provides a starting point for analysis of transport in systems that are out of equilibrium.
Treating entropy as a transported quantity, the total rate of accumulation (substantive
derivative) of entropy within the differential volume (fixed in the laboratory reference frame)
is given by
300 9 / Nonequilibrium Thermodynamics

system boundary

Figure 9.2

Conservation of energy and of each species similarly require that

(9.11)

%-%+*•>-<>• (9.12)
DNb 8Nb
+ V •Jfe = 0. (9.13)
Dt dt
Taking the divergence of Eq. (9.9), substituting it into (9.10), and using Eqs. (9.11)—(9.13)
yields
DS
(9.14)

Expanding the vector notation in Cartesian coordinates, we can write Eq. (9.14) as
DS

[la
dy\T

- — I — )Jb,x - — I
dx\T J dy\
In Eq. (9.15), each term on the right side is the product of a gradient term and a corres-
ponding flux term for one of the Cartesian coordinate directions. The prefactor for each
flux term is referred to as an affinity. Each affinity can be viewed as a driving force for
the corresponding flux. The reasons for this follow from equilibrium thermodynamics. If
adjacent regions of the system are at equal temperature, there is no temperature gradient
and the affinity terms for temperature are zero. Since the temperatures in adjacent regions
9.3 /Analysis of Linear Systems 301

are the same, they are thermally in equilibrium and no net flux of internal energy will occur.
Hence, the flux of internal energy is directly associated with nonzero values of the affinity
factors. Similar reasoning applies to the terms involving chemical potential and mass flux.
The total rate of entropy generation is thus seen to be the sum of terms, each of which is
the product of an affinity and a flux. This can be written more generically for our binary
mixture system as

i=\
where the J[ are three energy fluxes and six mass fluxes associated with the orthogonal
Cartesian coordinate and at is the corresponding affinity for flux //.
Evaluation of the rate of entropy generation within the differential volume obviously
requires that we determine the fluxes of energy and each species at the point of interest. The
relation for the flux depends on the nature of the molecules and their state in the system.
From a macroscopic viewpoint, each flux may, in general, be expected to be a function of
the system state variables P, 7\ \ia, and /z& and the affinities at. In fact, in most real systems
the fluxes are, to a good level of accuracy, functions of the affinities alone:

Ji = Ji(a1,a2,...,a9). (9.17)
Since each affinity is viewed as a driving force for the corresponding flux, we expect that
each flux vanishes if the affinities go to zero. One approach to generating a functional form
for the fluxes would be to develop a Taylor series expansion for each flux 7/:
9 1 9 9

Ji = J2 iJ J + 2 5Z 5Z LUkajak + ' •' ,

L a
(9-18)
j=\ j=\ *=1

where

L,j = 1^-, (9.19a)

daj
d2J-
(9.19b)
daj
Fortunately, in many systems of practical and scientific interest, the system behavior is
well approximated by a linear relation between fluxes and affinities. Linear systems are
considered in more detail in the next section.

9.3 Analysis of Linear Systems

Treating the binary mixture system considered in the previous section as a linear
system, we retain only the linear terms in Eq. (9.18):

Although the above linear relation may be applicable to a variety of different system types,
here we will specifically consider a system in which energy and each species are transported
only in the x direction. Replacing each affinity term aj with the equivalent gradient term
302 9 /Nonequilibrium Thermodynamics

implied by the equivalence of Eqs. (9.15) and (9.16), and taking Jt as —J a,x and — Jb,x for
species a and b, respectively, we can reduce the linear flux relations to

In Eq. (9.21), JUx is the flux of total internal energy in the material. However, we are often
more interested in heat flow. By analogy with Eq. (3.86) for quasisteady heat exchange in
a system of fixed volume,

8Q = TdS, (3.86)

we define a heat flux JQ ?JC by the relation

JQ,X = TJS,X. (9.24)

The x-direction component of the vector Eq. (9.9) indicates that

J
Js,x = j Ju,x ~YJ^~Y ^' < 9 ' 25 )
Combining Eqs. (9.24) and (9.25) yields

JQ,X = Ju,x ~ fiaJa,x - ^bJb,x- (9.26)

Substituting the right sides of Eqs. (9.21), (9.22), and (9.23) forJUx, Ja,x, and Jb,x in
Eq. (9.26), the resulting relation and Eqs. (9.22) and (9.23) can be written as

(9.27)

(9.28)

(9.29)
ox \i / \i y ox \i y ox

where

LQQ = L{ju + l^a(LaU 4" L\Ja)

Lab) + /x^Lflfl + filLbt, (9.30)

LaQ = Lau + \laLaa + l^bLab, (9.32)

^G* = Lub + /XfrL^ + /x«L^, (9.33)
LbQ = Lbu + VbLbb + MflL^. (9.34)
9.3 /Analysis of Linear Systems 303

Note that the affinities for the fluxes / e > x , — Ja>x, and —J b,x are now different than for
the original internal energy and mass fluxes. In this formulation of the analysis they are

Although linear, the above flux relations (9.27)-(9.29) are awkwardly cast in terms of
chemical potential gradients. Generally, however, it is often possible to express the chemical
potential as a function of other properties, making it possible to recast the relation in a more
useful form. For example, if we consider a binary mixture of van der Waal fluids, the
relations for the chemical can be written in the forms

Ma (Pa+Pb)bva / 3 \ \2nMakBT(l - pabv,a-Pbbvb)2/3

= I - In
kB T 1 — pabVja — Pbb Vjb \2 J _ |
(£a — 3)
In n — In a s a I In

(9.35)

{pa + Pb)K,b f^\) \2jiMbkBT(\ - pabVia - t

— Pabv,a — Pbby,b \ 2 ,

In 7t — In aStb In
2 J 2 \\PTou'njb/
4-
—, (936)
k BT
where pa and ph are molecular number densities of each species defined as

pa = Na/ V, pb = Nbf V. (9.37)

These relations indicate that jxa and jib are functions of pa, pb, and T. We can therefore
use the chain rule to reorganize the flux relations to the forms

Qx
' ~ L T2 T 3T T 8T \ dx dpa
[ T dp T 3pa\ dx
\LQa dfia LQb d^l dpb
[ TT ddpb T dpb J dx

] dT
ox
\Laa djia Lab
\_ T apa T opa

I *^aa & fta '-'ab

T dpb T dph\ dx

_ _ \_l±o_ Lbb_dji^ L^a/Xfli ar^ rz^a/x^ L ^ a ^ i ap«

M
~ [ r2 T dT T dT J dx [ T dpa T dpa\ dx

T dpb \ dx
304 9 /Nonequilibrium Thermodynamics

The form of these relations is consistent with those associated with Fourier heat con-
duction, which links heat flux to temperature gradient, and Fick's law, which links mass
diffusion flux to concentration or density gradients. They indicate, however, that contri-
butions to the heat flux may result from a concentration gradient as well as a temperature
gradient. Transport of thermal energy associated with mass concentration gradients is some-
times referred to as the diffusion thermo effect or the Dufour effect. The above relations also
indicate that contributions to the mass flux can result from a temperature gradient as well as
a concentration gradient. Transport of mass induced by a temperature gradient is sometimes
termed the thermal diffusion effect or the Soret effect.
For linear systems, the parameters L/7 in Eq. (9.20) play a central role in the linkage
between property gradients and fluxes of mass and energy in a system. These parameters
are called kinetic coefficients. In the next two sections we will explore the nature of these
coefficients in more detail.

9.4 Fluctuations and Correlation Moments

An important characteristic of the kinetic coefficients can be understood by con-
sidering the behavior of fluctuations in the system. Specifically, we are interested in the
microscale fluctuations in a system containing two species a and b that is in contact with
a reservoir that holds the temperature and chemical potential of each species in the system
constant. These intensive properties are fixed for the system, while the extensive variables
U, Na, and Nb may fluctuate because of exchanges of mass and energy with the reservoir
in contact with the system. In our analysis, the volume of the system is assumed to be
fixed (although the analysis can be extended to also consider changes in volume). The
system considered here is equivalent to a member system in a grand canonical ensemble
(see Chapter 4). More importantly, this system is equivalent to a control volume within a
much larger system. Thefluctuationswithin this type of system are therefore equivalent to
the local fluctuations in a control volume in a large system at equilibrium. For an arbitrary
thermodynamic property 7, in our analysis here we will denote the instantaneous values as
Y1. The mean value of the property, which we take to be equivalent to the thermodynamic
property at equilibrium, we will designate simply as Y.
For the type of system of interest here, at equilibrium, the fundamental equation relates
the entropy to the volume and the mean values of £/, Na, and Nb'.
S = S(V,U,Na,Na). (9.41)
When U, Na, and Nb fluctuate, their instantaneous values will generally differ from their
respective mean values. The value of entropy computed for the system with the fundamental
equation using instantaneous values of U, Na, and Nb we will interpret as the instantaneous
entropy of the system S':
S' = S(V,U',N'a,Nfb) (9.42)
By definition (see Chapter 2), the entropy is related to the number of microstates W for the
system as
S = kB In W. (9.43)
This relation holds at equilibrium, and we can also use it to determine the number of
9.4 /Fluctuations and Correlation Moments 305

microstates W that correspond to the instantaneous entropy for a particular set of fluctuating
U\N'a, and Rvalues:
S' = kB]nW. (9.44)
Combining Eqs. (9.43) and (9.44), we can obtain the following relation:
W' = WeAS/kB, (9.45)
where W is the number of microstates for the equilibrium condition at the mean values of
U, Na, and Nb and
AS = Sf - S(V, U, Na, Nb). (9.46)
Since we are interested in systems that behave classically, the properties U', N'a, and Nb
are taken to be continuous and we use Eq. (9.45) to construct a probability density function
p defined such that p dU'dN'adN'b is the probability of a fluctuation with combinations
of these variables that lie within intervals U' to U' + dUf, N'a to N'a+dN'a, and Nb to
Nb + dNb. This probability must be proportional to the number of microstates W and must
be properly normalized. We therefore define the probability density function as

(9.47)
JoT/o™ W'dU'dN'adN'b f/oT
Since W is a constant for the equilibrium condition about which the fluctuations occur, this
can be simplified to
Bt ( 9 - 4 8 )

where p0 is given by

Note that if p is integrated over all possible U', N'a, and N'b, the result is unity, as is necessary
for a normalized distribution function.
We now will use Eq. (9.48) to explore correlations among fluctuations. For small fluc-
tuations in the extensive properties U, Na, and Nb about the equilibrium (mean) values,
the resulting fluctuating entropy can be represented as a Taylor series expansion about the
equilibrium state values. For simplicity in notation, we set Y\ = U, Y^ = Na, and Y^ = Nt>.
The Taylor series expansion can then be written as

S' = S(V, YU Y2, Y3) + £ ( ! £

where the zero subscript implies that the derivatives are evaluated where Y[ — F; = 0 for
all /. The fluctuating entropy must be a maximum at equilibrium to ensure that fluctuations
away from equilibrium tend to be restored by spontaneous processes that tend to increase
entropy. This is a necessary condition for the equilibrium to be stable. Because the fluctuating
306 9 / Nonequilibrium Thermodynamics

entropy exhibits a maximum at equilibrium, it follows directly that the first derivatives in
the expansion above must be zero. The expansion therefore simplifies to

(9.51)
Since the second derivatives in the double summation above are evaluated at the equilibrium
condition and are nonzero, we take them to be equal to the corresponding second derivative of
S with respect to equilibrium values of extensive properties. The above relation then becomes

(9.52)
For convenience, we next introduce the following variables:
AS = S' - S(V, U, Na, Nb), (9.53)
£i = Y{-Yi. (9.54)
With some simple rearrangement, Eq. (9.52) can be expressed in terms of these variables as

••£/+ •••. (9.55)

If we now truncate the above series and insert the right side into Eq. (9.48) for AS, we obtain
the following relation for the probability density function for fluctuations with extensive
property deviations Y[ — Y\ = e\, Y'2 — Y2 = s2, and F3' — F3 = £3:

(956)
Y^y^^r
where

==
P o poo TOO Tod A C/K j ~j ~j • (V.J /)
Loo /-oo /-oo e S/ B dslds2de3
Note that by changing the variables of integration in Eq. (9.57), the limits of integration are
changed to span all possible fluctuations. Equation (9.56) is sometimes called the Einstein
fluctuation formula. This relation reflects the fact that, to first order, AS is a quadratic func-
tion of the deviation of extensive properties from their equilibrium values. Furthermore,
because the fluctuating entropy is a maximum for zero deviation from the equilibrium state,
the second derivatives in the double sum must be negative. It follows that AS is zero at £i =
s2 = £3 = 0 and becomes increasingly negative as £1, e2, and £3 deviate from zero. As £1, £2,
and £3 approach ±00, AS -> —00. Einstein's fluctuation formula indicates that the proba-
bility of fluctuations approaches zero as the absolute values of £1, e2, and £3 become large.
For each of the extensive variables associated with the equilibrium state there is a driving
potential associated with changes in that variable: l/T for U, fia/T, for Na and ^b/T for
Nb. For each external property Yt the corresponding potential is (dS/dYi)Yj^. We now will
9.4 / Fluctuations and Correlation Moments 307

extend this definition to the fluctuating properties we have defined here as follows: We
define a driving potential 0- for fluctuating external property Y{ as

y,^. (9.58)

Since (dS'/dY'^y, is a function of the external properties F/, we construct a Taylor series
expansion for 0- about the equilibrium state:

(9 59)
) F
>-^)+-' -
where the zero subscript again indicates that the derivatives are evaluated at the equilibrium
state. Defining a[ as the deviation of 0- from its value at the equilibrium state,

it follows that

where the zero subscript denotes the value of the derivative in the limit Y[ — Yt -+ 0. Since
(dS'/dYl)0 = 0, this reduces to

Also, since S is a constant for the equilibrium state about which the fluctuations occur,
Eq. (9.53) implies that

'?) i9^) (9-63)

,3i7; v a*/;'

and since Eq. (9.53) indicates that £r = Y( — Yt, the above relation can be written

Combining Eqs. (9.62) and (9.64) we find that

Note that the definition of a •is such that

if Yt = U then a- = — , (9.66)

if 7- = Afa then a- = ( — 1 -, (9.67)

and

- ^ . (9.68)
308 9 /Nonequilibrium Thermodynamics

Before proceeding further a comment regarding fluctuation of intensive and extensive

properties is in order. In our development of statistical thermodynamics we considered the
grand canonical ensemble in which each member system is in contact with reservoirs (or
other member systems) that exchange mass and energy in such a way that the temperature
and chemical potential for each species are the same for all systems. In this ensemble, the
extensive variables U, Na, and Nb fluctuate (vary among the ensemble members) about the
mean value, but within this statistical model the conjugate driving potentials 1 / T, \ia IT, and
IJLb/T cannot vary. Real systems are not subject to the restrictions of this statistical model,
however. For example, sensors that measure temperatures operate by reaching thermal
equilibrium with the surrounding portion of a system. If the energy of the system fluctuates,
the sensor output will usually exhibit a corresponding fluctuation in temperature.
We therefore interpret fluctuations of the conjugate driving potentials 1/T, /j,a/T, and
fib/ T in the following way: In the control volume offixedvolume considered here, fluctu-
ations inU, Na, and Nb with interacting reservoirs cause the state to fluctuate and the state
variation for the system can be parameterized either in terms offluctuationsin the external
properties U, Na, and Nb or in terms of the changes in the driving potentials that result from
fluctuations in U, Na, and Nb.
The correlation moment of an extensive variable fluctuation £; and a driving potential
fluctuation a'- can be computed as
pOQ pOO pOO

= / / sia/jpd£ide2d£3, (9.69)
J/—oo
—oo J—oo
J—oo JJ—oo
—oo

where p is the probability density function defined by Eq. (9.48). Substituting for otj and p
using Eqs. (9.65) and (9.48) gives
•»OO pOO pOO / OA C \
<£,-«;.) = / / / e, (°-p) /V AS/ * B deide2de3. (9.70)
J—oo J—oo J—oo \ O£j J

With some straightforward manipulation, this relation can be rearranged to the form
poo /»oo poo cj
(sia'j)=kBp0 / / Si — (e AS/k'i)dslds2ds3. (9.71)
J—oo J -oo J—oo v£j

We now consider two possibilities: / = j and / ^ j . First, for / = j we can reorganize

Eq. (9.71) in the form
poo poo
po poo
(eia'j)=kBp0 /// //
—oo J—oo
J—oo J—
poo poo poo
-ksPo // // eAS/kBdSide2de3. (9.72)
J -oo J —oo J -oo

From Eq. (9.57) it follows that the second term in Eq. (9.72) equals - £ B . In the first term,
/ must equal 1, 2, or 3, and the integration with respect to et can be done first to obtain

f
J—
—oo J —oo
(9.73)

As discussed above, AS approaches negative infinity as £; approaches ±oo, and therefore

the term in square brackets is zero at both integration limits. The first term in Eq. (9.73) is
9.5 / Onsager Reciprocity of Kinetic Coefficients 309

therefore zero and for i = j , the correlation moment reduces to the simple result
(eta'j) = -kB. (9.74)
For i ^= j , the integration on the right side of Eq. (9.71) can be done first with respect to Sj
to obtain

-oo J-oo
Here again, AS approaches negative infinity as £j approaches dboo, and therefore the term
in square brackets is zero at both integration limits. Thus, for / ^ j , (stoi'j) = 0. We can
summarize the results for both cases as

These results imply that fluctuations in extensive variables are correlated with fluctuations
in the corresponding driving potential, but not with fluctuations in other driving potentials.
As will be shown in the next section, this impacts the interrelationship among kinetic
coefficients.

9.5 Onsager Reciprocity of Kinetic Coefficients

In classical thermodynamics, the extensive properties for a system in contact with a
reservoir for that property are taken to be fixed constants. As noted above, this is not strictly
true. If energy or mass is permitted to flow freely between a system and a reservoir, at a
microscopic level the exchange will produce fluctuations in the property. In large systems,
these fluctuations are small and rapid and the value of the property stays close to the average
for the system.
On rare occasions a large fluctuation may occur, which, for example, might alter the
energy in the system by a nonnegligible amount. If the system were decoupled from the
reservoir just after such a fluctuation, the energy and temperature of the system would be at
least locally different. If the system were not decoupled, the fluctuation would decay by the
spontaneous flow of energy between the reservoir and the system. Onsager [1] realized that
this spontaneous flow is in some sense driven by the temperature difference generated by the
original fluctuation. He connected macroscopic processes to nonequilibrium thermodynam-
ics by assuming that the decay of a spontaneous fluctuation is identical to the macroscopic
flow of energy or mass between a reservoir and a system that are not in equilibrium.
The usefulness of Onsager's assumption can be seen by considering the following analy-
sis. We consider again the constant-volume control volume system discussed in the previous
section. The system is in contact with a reservoir with which it can exchange energy and
particles of species a and b. For convenience, we will continue to use the notation defined
in the previous section. In the previous section we discussed correlation moments. For two
extensive properties 7/ and Yj, the correlation moment is defined as
poo poo po
SiSj) = / / iSjpd8id82ds3, (9.77)
—oo JJ —oo
J —oo —ooJ J—oo
—

where p is the probability density function defined by Eq. (9.48). An extension of this
concept is the delayed correlation moment (£;£ 7 (r)), which is the average product of the
310 9 / Nonequilibrium Thermodynamics

deviations e£- and 6j, with Sj being observed a time r after £t is observed. In the absence of
magnetic field effects, the physical laws governing the motion of the particles are assumed
to be symmetric with respect to the reversal of time. This implies that in the absence of a
magnetic field, the delayed correlation moment must be unchanged by the replacement of
r by —r:
(SiSjir)) = (SiSji-T)). (9.78)
Because only the relative time interval between the samples is important, the right side is
equivalent to (£;(T)£ 7 ). It follows that

j j (9.79)
Subtracting (£;£/) from both sides of Eq. (9.79) and dividing by r, we obtain

(9.80)

This relation must hold for all r, and we therefore expect it to be valid in the limit r -» 0.
Taking this limit, we can write Eq. (9.80) in terms of time derivatives as

Consistent with the discussion above, we assume that the decay of a fluctuation S( is governed
by the same linear laws as macroscopic processes. We can then relate £t and Sj to fluctuations
of thermodynamic potentials a[ as
3

ki = YMi<*'i> ( 9 - 82 )

(9.83)
1=1

Using Eqs. (9.82) and (9.83) to replace £; and Ej in (9.81) and interchanging the average
and summation operations yields

i=\

Equation (9.76) indicates that the bracketed terms in the above equation are nonzero only if
the indices are the same (/ = / or / = j). Using Eq. (9.76) to evaluate the terms in Eq. (9.84)
leads to

which simplifies to
Lij = Lji. (9.85)

The above relationship among the kinetic coefficients is known as Onsager reciprocity.
In our analysis of this relationship we have considered a system with no electromagnetic
effects. If a system with such effects in considered, it can be shown that a broader statement
9.6/ Thermoelectric Effects 311

of this relation is that the value of kinetic coefficient L/y measured in external magnetic
field Bext is equal to the value of Lji measured in external magnetic field — Bext*
Li7(Bext) = L y/ (-B e xt). (9.86)
(See, for example, Callen [2].)
Clearly, if there is no magnetic field, the statements of Onsager reciprocity (9.85) and
(9.86) are equivalent.
These results imply that for the kinetic coefficients in Eqs. (9.21)-(9.23), Ly a = Lau,
Ljjb = Lt>u, and La^ = Lba. Inspection of Eqs. (9.30)-(9.39) reveals that the validity of these
reciprocity relations for the mass and internal energy fluxes assures that reciprocity holds for
the kinetic coefficients associated with the heat and mass fluxes in relations (9.27)-(9.29):
Lqa = LaQ, LQI> = LbQ, and Lab = Lba- The analytical framework for nonequilibrium
thermodynamics developed in this chapter is particularly useful for analyzing thermoelectric
effects. We consider such effects in the next section.

9.6 Thermoelectric Effects

In this section we will consider a conductor in which one-dimensional flow of
heat and electric current occur. The electric current is assumed to be a result of electron
motion in the conductor. We therefore apply Eqs. (9.27) and (9.28), taking species a to be
the electrons and neglecting the species b terms:

(9 87)
) '

where
LQQ = LJJU + /xa(LUa + LaU) + fi2aLaa, (9.89)
LQa = LUa + llaLaa, (9.90)
LaQ = LaU + ixaLaa. (9.91)
There are positive ions in the conductor, but they are assumed to be immobile and therefore
do not contribute to the energy or current flow.

Electrical and Thermal Conductivities

In considering flow of electrons we break the overall chemical potential for the
electrons into two parts: a chemical portion /xch and an electrical portion /xe:
V>a = Mch 4- Me- (9.92)

In the absence of a magnetic field, if the charge on an electron is e e, then

Me = *e0e, (9.93)
where 0 e is the electrostatic potential. The chemical portion of the chemical potential is a
function of the electron concentration and the temperature. The electrochemical potential
312 9 / Nonequilibrium Thermodynamics

per unit charge is /xfl/ee- It follows that the gradient of the electrochemical potential per
unit charge is the sum of two components

V(/i«/e e ) = (l/e e )V/ie + d / ^ V / i c h . (9-94)

The first term on the right side of Eq. (9.94) is the gradient of the electrostatic potential,
which is the electric field. The second term is a driving force arising from a concentration
gradient.
The electrical conductivity aQ is defined as the ratio of the electric current density —e eJa^x
to the potential gradient V(/z fl/ee) in a isothermal system:

d
Substituting the right side of (9.88) for —J a,x f° r zero temperature gradient converts the
above relation to

a e = ^ . (9.96)

The thermal conductivity is similarly defined as the heatfluxper unit temperature gradient
for zero electric current:

^f- for Ja,x=0. (9.97)

oxx
Setting the right side of Eq. (9.88) to zero we obtain

T) dx L dx \TJ
Combining this equation with Eq. (9.87) to eliminate (l/T)(d/jLa/dx) and substituting the
expression for —J a,x into Eq. (9.97) yields
L
kt = QQLaaLQ"LaQ ( 9 9 9 )

LaaT

The Seebeck Effect

The circuit composed of two different conducting materials shown schematically
in Figure 9.3 is known as a thermocouple. Thermocouples constructed with two different
metals are commonly used for measuring temperature. The junctions where the two dis-
similar metals are connected are at different temperatures T\ and T2, and the impedance
of the voltmeter is high enough that negligible current flows in the circuit. The terminals
of the voltmeter are presumed to be at the same temperature Tm. The production of an
electromotive force in a thermocouple under conditions of zero current flow is termed the
Seebeck effect. Here we designate the two materials in the circuit as A and B.
Assuming that the chemical potential and temperature are only a function of distance
along the conductors, the zero current requirement (9.97) can be rearranged to the form
d
J± = h}L*L. (9.100)
dx TLaa dx
9.6 / Thermoelectric Effects 313

We can multiply through by dx and integrate this relation for each of the conductor segments
in Figure 9.3 to obtain

(9.101)

a -a - [Ml (9.102)

L
_ f aQ dj
=
(9.103)
a l am
~ <JTm TL%a

Combining these relation to eliminate /xa>i and /ia>2» we obtain

rT2 / LA LB \
H /la m = (9.104)
' ~ JTl \J^a~~Tl4a)dT'

Since the temperature is the same at both terminals of the voltmeter, the voltage difference
measured by the meter is A0 m = (/xfl>m+ - /x a m _)/^ e . Thus the voltage difference at the
meter is given by

^aQ
dT. (9.105)
e,LAa

This relation relates the voltage measured by the meter to the temperatures at each of the
thermocouple junctions. If one junction is held at a reference condition, the second junction
temperature can be inferred from the measured voltage. A common way of establishing
a reference condition for one junction is to immerse it in an ice-water bath, which has a
saturation temperature of 0°C at atmospheric pressure.
It is conventional to define the thermoelectric power of the thermocouple sAB as the
change in voltage A0 m per unit change of temperature difference T2 — Ti. The sign of sAB
is chosen so that it is positive if the voltage change tends to drive current from material A
314 9 / Nonequilibrium Thermodynamics

to material B at the hot junction. It follows that

dA(pm L Q L Q /nm^
SAB = —Tzr- = T jA - T jB • (9.106)
dT2 TeeL*a TetLBaa
The thermoelectric power &AB is also sometimes referred to as the Seebeck coefficient.
Because SAB is equal to the difference of two terms, each of which is associated with a
specific material, we also define the absolute thermoelectric power or absolute Seebeck
coefficient of a substance A as

§ (9 107)
* =^7X- '

The thermoelectric power for the thermocouple is then given by

SB-£A- (9.108)
Note that if the variation of the thermoelectric absolute powers with temperature is known
and T\ is fixed, the variation of thermocouple voltage with temperature can be computed as

(sB -eA)dT. (9.109)

This is the usual way that the voltage output is determined for thermocouples having different
combinations of materials.
The original transport Eqs. (9.87) and (9.88) contained three kinetic coefficients LQQ,
Laa, and L Qa = La Q. If we know these three parameters we can calculate the electrical con-
ductivity, the thermal conductivity, and the thermoelectric absolute power using Eqs. (9.96),
(9.99), and (9.107). Moreover, the reverse is also true. We can use these equations to cal-
culate the three kinetic coefficients if we determine a e , ku and £A by measurement.

Example 9.1 You are told that for copper at 300 K, the absolute thermoelectric power
is 1.83 /xV/K, the thermal conductivity is 400 W/mK, and the electrical conductivity is
5.88 x 107 ^2~1m~1. Use these data to determine the kinetic coefficients LQQ, Laa, and
LaQ for copper at this temperature.

Solution Solving Eq. (9.86) for Laa yields

el

Substituting ee = —1.602 x 10~19 coulombs for the charge on one electron and the given
values of <re and temperature, we obtain
(-1.602 x 10- 19 ) 2 _ IQ47 K K
5.88 x 107(300) mVAs2 mJs
We can also rearrange Eq. (9.107) to solve for LaQ.

Lao = -eATeQLaa.
9.6/ Thermoelectric Effects 315

Substituting, we find that

LaQ = -1.83 x 10~ 6 (300)(-1.602 x 10"19)6.87 x 1047 = 6.05 x 1025 — .

ms
Finally, we invert Eq. (9.99) and substitute the previously determined values of L aa and
LaQ to determine LQQ:

!k = 400(300,= . 3.60 x 10'

6.87 x m

Example 9.2 In real thermocouple temperature measurement systems, it is common to

connect the leads of the potentiometer to the thermocouple using two copper lead wires as
shown schematically in Figure 9.4. Show that this provides an accurate measurement of the
voltage difference </>e,5 — 0e,2 if T\ = T6 and T5 = T2.

Solution The chemical potential difference between locations 1 and 6 can be

written
Ma,6 ~ Ma,l = ( / V 6 - Mfl,5 (/V2 ~ Ma,0-
Multiplying both sides of Eq. (9.100) by dx, using (9.107) to replace the kinetic coefficients,
and integrating from 1 to 2 yields

cu dT.

Executing a similar integration from 5 to 6, we obtain

a,6 eesCudT.
J 7s

Substituting the right side of the relations for jxa2 — H>a,i and fia^ — \x a^ into the above
relation for /x a?6 — Ma,i we can write
rT2
l
a,6 ~ Ma,l =
316 9 / Nonequilibrium Thermodynamics

We can reorganize the integrals in this equation into the form

rT6 rT
=
/V6 - /Vi (/V5 - /V2) - /I eec8sCudT + / ee8CudT.
J T\ J T2

Finally, since in each conductor the chemical component of the chemical potential is uni-
form, the change in chemical potential within the conductor is equal to the change in elec-
trostatic potential <?e0e, where 0 e is the voltage. We can therefore write the above relation as
fTe ~ fTs
ee e ,6 eQ Cfl _ et e ,5 et e ,2 J^ etsCu J^ ee£Cu
Canceling eQ throughout, we can reduce this to
pTt pT$
0e,6 - 0e,l = (0e,5 ~ 0e,2) ~ / ^Cu dT + / £CudT.
JTX JT2
Clearly, if T\ = T6 and T5 = T2, the contribution of the integrals vanishes and the measured
voltage difference at the potentiometer is identical to that between points 2 and 5. If these
equalities are not true, the measured voltage will be in error by an amount dictated by the
integral terms.

Example 9.3 In the arrangement of thermocouples shown schematically in Figure 9.5,

junctions 2, 4, and 6 are held at a reference temperature Tref and junctions 3, 5, and 7
simultaneously sense the same temperature Tp. This circuit configuration is known as a
thermopile. Show that for this arrangement, the voltage difference 0e,8 — 0e,2 is three times
that for a single thermocouple of the same type under the same conditions.

Solution Multiplying both sides of Eq. (9.100) by dx, using (9.107) to replace
the kinetic coefficients, integrating across each segment of the circuit from 2 to 9, and
combining the resulting equations yields
fT-i pTi, pTs pT(y
Ma,8 ~~ Ma,2 — ~~ / ^e^A dT — I £ e£fl dT — / e QsA dT — j e t £B dT
JT2 JT3 JT4 JT5
pTi pT%
- / eesA dT - / eeeB dT.

tSSsSs^ssx:

Figure 9.5
9.6 / Thermoelectric Effects 317

Using the facts that T2 = T4 = T6 = 7ref and T3 = T5 = T7 = Tp, we can write the above
relation as
fTP fTref rTp rTKf
M«,8 - /V2 = - / eesA dT - eJB dT - eJA dT - ,JB dT
JTref JTp JTKf JTp

e,sAdT - I eeeBdT,

which reduces to
fTp
ee(eA - sB) dT.
JTr*

Since in each conductor the chemical component of the chemical potential is uniform, the
change in chemical potential within the conductor is equal to the change in electrostatic
potential ee</>e, where 0 e is the voltage. Substituting ee</>e for [ia in the above equation and
canceling et throughout, we obtain

fTp (£A ~SB)dT.

0e,8 - 0e,2 = 3 /
The integral in the above equation is equal to the voltage difference across a single ther-
mocouple of the same type having one junction at Tv and the other at r ref . Thus the voltage
difference across the entire thermopile is three times that amount.

The Peltier Effect

Figure 9.6 shows the junction between two different materials that conduct electric
current and heat. The absorption or evolution of heat when electric current flows steadily
across an isothermal junction of two materials is referred to as the Peltier effect.
As current flows across the junction, conservation of electrons requires that the flux of
electrons must be the same in both materials. From Eq. (9.26) we know that for either
material (neglecting b terms since we are only considering electron motion)

JlJ,x = (9.110)
Since at the junction /xa is the same in both materials and Ja,x is continuous across the
junction, jJLaJa,x is the same in both materials at the interface. It follows, then, from

A n .
I u I
Figure 9.6
318 9 / Nonequilibrium Thermodynamics

Eq. (9.110) that

or equivalently

Thus, the energy flux may change by an amount equal to the change in heat flux across
the junction. For an isothermal junction, setting the temperature gradient terms to zero and
combining Eqs. (9.87) and (9.88) for material A yields
L
A Qa A

aa
Using Eq. (9.107), we can rewrite (9.113) as
j£x = TeJAJ*x. (9.114)
Identical reasoning for material B yields
(9.115)
Subtracting the right and left sides of Eqs. (9.113) and (9.115) and using the fact that
J
a,x = Ja,x = Ja,x, we obtain

JQ,X - JL = TGB - eA)eeJa,x- (9.116)

Thus the difference in the heat fluxes, which equals the heat absorbed or rejected per unit
junction area, is related to the electric current flux and to the difference in the values of
the absolute thermoelectric power for the materials. In connection with this process, the
Peltier coefficient ft AB is defined as the heat that must be supplied to the junction per unit
of electric current passing from material A to material B:
jB _ jA
Qx Qx
rtAB = ' ' = T(sB - sA). (9.117)
^e^a,x
Thus the Peltier coefficient can be determined for a junction between two materials if
the junction temperature and the absolute thermoelectric power for each material at that
temperature are known.

Example 9.4 A junction between a copper and a nickel wire is held at a temperature of
300 K. The diameter of the wire is one millimeter. If a current of 20 A flows from the copper
to the nickel, determine the rate at which the Peltier heat must be added or removed to hold
the junction temperature constant.

Solution Taking material A to be the copper and B to be the nickel, we find the
total Peltier heat for the junction QP by multiplying both sides of Eq. (9.116) by the wire
cross-sectional area Ac:
9.6 / Thermoelectric Effects 319

At 300 K, s for copper is 1.83 /xV/K and e for nickel is -19.5 /xV/K. Noting that etJa,xAc
is the total current flow, which equals 20 A here, we substitute to compute Q P:
Qp = 300(-19.5 x 10~5 - 1.83 x 10"6)20 = -0.128 W.
The negative value of Qv indicates that the heat must be removed from the junction to
maintain a constant temperature.

The Thomson Effect

An additional thermoelectric effect, known as the Thomson effect, arises when
electric current flows through a temperature gradient. To explore this effect, we consider the
situation shown in Figure 9.7 in which heat flows through a material from a hot reservoir to a
cold reservoir. In the absence of electrical current flow, a temperature gradient is established
in the material in accordance with the temperature dependence of the kinetic coefficients for
the material. At each point along the conducting material, we now bring a thermal reservoir
into contact with the material having a temperature equal to the local temperature in the
material. Since each reservoir is in thermal equilibrium with the material at each location,
no heat is exchanged with the reservoirs.
Having brought these reservoirs into contact with the material, we now impose an elec-
trical potential difference between points 1 and 2 to drive electrical current through the
conducting material. At any location along the conducting material, the energy flow will
differ from that established initially by a temperature gradient alone. The first law requires
that any change in the total energy flow in the conductor must be accompanied by an energy
exchange with the reservoir in contact with the material at that point. The heat exchange
per unit volume of material q"' with the reservoir must equal the divergence of the energy
flux V •Jv. Using Eq. (9.103) for this one-dimensional case, we obtain

" ~ . (9.118)
dx dxy ^ ' ™ "'" dx dx ax
''
Combining (9.80) and (9.81) yields the following relation for JQy.
_ (LqQL a a — L Q a L a Q \ d f \ (9.119)
J Q,x —

1
;2

Vreservoirs at local !

temperature T

Figure 9.7
320 9 /Nonequilibrium Thermodynamics

We then substitute the right side of Eq. (9.119) for JQx and use (9.88) to eliminate d/j,a/dx
in Eq. (9.118) to obtain

/// & \(LQQLaa — LQaLaQ\ d / 1 \ L Q a 1

q =7- -j 7- - - Ja,x
dx L\ Laa ) dx \T) Laa J

^-Kx - ^y^^T (T)} Jax' (9 120)

-
Using Eqs. (9.96), (9.97), and (9.107) we next replace the kinetic coefficients with the
conductivities and the absolute thermoelectric power:

dx [ e \
(9.121)

The temperature variation, which is now held fixed by the reservoirs, was determined for
the case of no heat exchange and no electric current. Setting q'" and Jax to zero in the above
relation reduces it to the following relation which dictates the temperature variation in the
material:

0
-£["*£(?)]•
The term on the right side of the above relation appears on the right side of Eq. (9.121). In
accordance with Eq. (9.122), we set this term to zero in Eq. (9.121) to obtain, after a bit of
rearranging,

ds 1
q'" = TeeJ0,x— - -{ej^f. (9.123)
OX 1

Since the absolute thermoelectric power is a function only of local temperature, we can
write this result in the form

q'" = etJa,xT (^-) ^ - -(etJa,x)\ (9.124)

\dTJ dx ae
The second term on the right side of Eq. (9.124) is the Joule heat that is produced by
the flow of electric current through the material. The negative sign on this term indicates
that heat must be removed volumetrically from the material to maintain a constant local
temperature. Note that this well-known resistance heating of the material occurs even when
no temperature gradient exists in the material.
The first term on the right side of Eq. (9.124) is known as the Thomson heat. The Thomson
heat is the amount of heat exchanged (volumetrically) with the local reservoir to maintain
a steady temperature when current eQJa,x traverses temperature gradient dT/dx in the
material. From classical thermodynamic reasoning, Sir William Thomson, who later became
Lord Kelvin, predicted this result in 1855. Thomson's deduction was experimentally verified
for a number of materials long before the microscale aspects of thermoelectric effects were
understood. It is conventional to define the Thomson coefficient f as the Thomson heat per
9.6 / Thermoelectric Effects 321

unit current and per unit temperature gradient:

_ e*Jg,xT(•£)%; _ fds_
T
~ e j (*L\ ~ L I dT

The Thomson coefficient is therefore directly related to the temperature derivative of the
absolute thermoelectric power. By convention it is defined so that, if it is positive, the Thom-
son heat must flow into the material to keep the temperature constant. If we differentiate
both sides of Eq. (9.117) with respect to temperature, we obtain
dsA\
dT ° " " \dT dT ) (9.126)
Using the definition of the Thomson coefficient (9.125) we can write Eq. (9.126) as

^ § B - e A + iB-tA. (9.127)
di
Note that Eq. (9.127) relates coefficients associated with the Peltier, Seebeck, and Thomson
effects. In acknowledgment of Thomson's pioneering work on thermoelectric phenomena,
Eqs. (9.127) and (9.117) are known as Kelvin's first relation and Kelvin's second relation,
respectively.
Both the Seebeck coefficient SAB and the Peltier coefficient it AB are defined for a pair
of materials connected at a junction, but the Thomson coefficient and the absolute ther-
moelectric power are properties of a single material. For a superconducting material, it is
impossible to sustain an electrostatic potential within the material because infinite current
flow would result. Hence the absolute thermoelectric power must be zero since a tempera-
ture gradient cannot induce a finite electrostatic potential. If we pick a superconductor as
material A, then §AB = eB—s A= eB and the absolute thermoelectric power of material B
can be determined from measurements of the rate of change of thermocouple voltage with
temperature (at 2 in Figure 9.4) using Eq. (9.106):

eB = ^ . (9.128)
dT
Experiments have, in fact, been done in which a junction between lead and a superconductor
was formed and the voltage versus temperature difference was determined for temperatures
up to 18 K. It can be argued based on the third law of thermodynamics that the Seebeck
coefficient must be zero when the junction temperature reaches absolute zero, and low
temperature measurements are consistent with this line of reasoning.
We can also integrate Eq. (9.125) for an arbitrary material to compute the absolute
thermoelectric power as

e= -dT + sx. (9.129)

The Thomson coefficient has been determined experimentally for lead from 20 K to room
temperature. The constant £\ at T\ can be determined from the low temperature s data men-
tioned above and the r data can be used to numerically evaluate the integral in Eq. (9.129)
to determine values of the absolute thermoelectric power for lead from 20 K to room tem-
perature. With this procedure it is possible to establish values of the absolute thermoelectric
power for lead over a broad range of temperatures. The absolute thermoelectric power of
322 9 /Nonequilibrium Thermodynamics

any other conductor can then be determined simply by pairing it with lead in a thermocou-
ple and using Eq. (9.106) to determine e — £iead from the slope of the measured variation
of thermocouple voltage with temperature. This approach makes it possible to determine
absolute thermoelectric power of any conductor of interest. Once determined, values of
the absolute thermoelectric power can be used to analyze the thermoelectric phenomena
described in this section.
In closing this section, it should be noted that we have considered only the simplest
of the interactions among the generalized fluxes that may arise in physical systems. Other
thermoelectric and thermomagnetic effects can be defined and analyzed using methods that
are basically the same as those developed in this section. The interested reader may find
further discussion of thermoelectric effects in the texts by Zemanksy and Dittman [3] and
Callen [2], the paper by Callen [4], and the handbook edited by Rowe [5].

Exercises
9.1 Derive Eqs. (9.27)-(9.29) from Eqs. (9.21)-(9.23) and (9.26).

9.2 Using the fact that /xa and /x^ are functions of pa, pb, and 7\ derive Eqs. (9.38)-(9.40) from
Eqs. (9.27)-(9.29).

9.3 You are told that for nickel at 300 K, the absolute thermoelectric power is -19.5 /zV/K, the
thermal conductivity is 91 W/mK, and the electrical conductivity is 1.43 x 107 ^~ 1 m~ 1 .
Use these data to determine the kinetic coefficients LQQ, Laa, and LaQ for nickel at this
temperature.

9.4 For chromium at 300 K, the absolute thermoelectric power, the thermal conductivity, and the
electrical conductivity are estimated to be 21.8 /zV/K, 94 W/mK, and 7.80 x 106 £2~lm~l
respectively. Use these data to determine the kinetic coefficients LQQ, Laa, and LaQ for
chromium at this temperature.

9.5 A copper-nickel thermocouple is to operate between 273 and 310 K. The variations of the
absolute thermoelectric power for these materials over this range are given approximately
by

sm = - 19.5 + 0.055(300 - T), eCu = 1-83 - 0.0032(300 - T),

where T is in K and s^i and SQU are in /xV/K. If the reference junction is held at 0°C, compute
and plot the variation of the voltage measured at the potentiometer with temperature T for
temperatures between 290 and 310 K.

9.6 A thermocouple is to be used to measure temperatures between 273 and 320 K. For nickel,
copper, and chromium, the variations of absolute thermoelectric power for these materials
over this temperature range are given approximately by

eNi - -19.5 + 0.055(300-7),

sCu = 1.83-0.0032(300-7),
sCr = 2 1 . 8 - 0 . 0 8 4 ( 3 0 0 - 7 ) ,

where T is in K and s^i and §cu are in \i V/K. To get the highest sensitivity for the resolution of
the potentiometer (most voltage difference per degree kelvin), which two of these materials
should be used for the thermocouple?
Exercises 323

9.7 Copper and chromium are to be used in a thermopile like that shown in Example 9.3 with
the reference junctions held at 0°C. For the purposes of this analysis, assume that over the
temperature range of interest
sCu = 1-83 - 0.0032(300 - T), sCr = 21.8 - 0.084(300 - T),
where T is in K and sCu and sCr are in /xV/K. Determine the variation of output voltage
difference with temperature for the thermopile for temperatures between 280 and 320 K.
Plot your results.
9.8 Two wires, one of nickel and the other of chromium, meet at a junction immersed in an
ice-water mixture, essentially holding the junction at 0°C. Between 200 and 300 K the
variations of absolute thermoelectric power for these elements are given approximately by

sm = -19.5 + 0.055(300 - T), sCr = 21.8 - 0.084(300 - T),

where T is in K and £NI and £& are in /xV/K. If a current of 50 A flows from the nickel
to the chromium, determine the rate at which the Peltier heat is transferred to or from the
ice-water mixture.
9.9 The junction between a chromium wire and an aluminum wire is held at 0°C by immersing
it in an ice-water mixture at atmospheric pressure. Between 270 and 300 K, the variations
of the absolute thermoelectric power for these materials are given approximately by

Scr = 21.8 - 0.084(300 - T), sAl = -1.66 - 0.0027(300 - T),

where T is in K and eCr and sA\ are in /xV/K. If electric current flows through the junction,
how large must it be and which way must it flow for the Peltier heat delivered to the ice
water to equal 0.1 W?
9.10 Electric current is to be driven through a junction formed by two of the following materials:
copper, aluminum, chromium, nickel, or iron. The junction is held at 0°C by immersing it in
an ice-water mixture. Between 270 and 300 K, the variations of the absolute thermoelectric
power for these materials are given approximately by

ecu = 1.83 - 0.0032(300 - T), eM = -1.66 - 0.0027(300 - T),

£Cr = 21.8 - 0.084(300 - T), eNl = -19.5 + 0.055(300 - T),
£Fe = 1 5 . 0 - 0 . 0 1 7 ( 3 0 0 - 7 ) .
where T is in K and the absolute thermoelectric power values are in /xV/K. Which combi-
nation of materials will maximize the Peltier coefficient for the junction?
9.11 The table below lists data indicating the variation of the absolute thermoelectric power with
temperature for nickel and chromium. Fit each set of data with a second-order polynomial in
temperature. Use the resulting relations to determine and plot the variation of the Thomson
coefficient for each material between 200 and 400 K.

T (K) £Ni (/xV/K) sCr (/xV/K)

100 -8.5 5.0

300 -19.5 21.8
500 -25.8 16.6

9.12 The table below lists data indicating the variation of the absolute thermoelectric power with
temperature for aluminum and iron. Fit each set of data with a second-order polynomial in
324 9 /Nonequilibrium Thermodynamics

temperature. Use the resulting relations to determine and plot the variation of the Thomson
coefficient for each material between 200 and 400 K.

T (K) sM (/xV/K) £Fe (MV/K)

100 -2.2 11.6

300 -1.66 15.0
500 -1.96 3.0

References
[1] Onsager, L., "Reciprocal Relations in Irreversible Processes," Phys. Rev., 37: 407, 1931.
[2] Callen, H. B., Thermodynamics and an Introduction to Thermostatistics, 2nd ed., Chapter 14,
John Wiley & Sons, New York, 1985.
[3] Zemanksy, M. W. and Dittman, R., Heat and Thermodynamics, McGraw-Hill, New York, 1981.
[4] Callen, H. B., "Application of Onsager's Reciprocal Relations to Thermoelectric, Thermomag-
netic and Galvanomagnetic Effects," Phys. Rev., 73: 1349, 1948.
[5] Rowe, D. M., ed., CRC Handbook ofThermoelectrics, CRC Press, Inc., Boca Raton, FL, 1995.
 CHAPTER 10

Nonequilibrium andNoncontinuum Elements

ofMicroscale Systems

There are many important systems that exhibit nonequilibrium or noncontinuum

behavior. Thisfinalchapter examines some important examples of such systems. In doing so,
we have two objectives. Thefirstis to understand how, and under what conditions, the system
behavior may deviate from the idealizations embodied in equilibrium theory or continuum
theory. The second is to demonstrate theories and methods that are commonly used to
model nonequilibrium and noncontinuum systems. Because they are commonly used to
analyze such systems, kinetic theory and the Boltzmann transport equation are introduced.
Nonequilibrium and noncontinuum phenomena associated with multiphase systems and
electron transport in solids are examined in detail. The final section of Chapter 10 uses
results from previous chapters to examine length scales and time scales at which classical
and continuum theories become suspect. Doing so defines the range of conditions for which
we expect classical and continuum theories to be accurate models of real physical systems.
Although limited in its coverage, this chapter provides an introduction to microscale aspects
of nonequilibrium and noncontinuum phenomena and serves to illustrate how they relate
to the theoretical framework developed in the preceding chapters.

10.1 Basic Kinetic Theory

With increasing frequency engineers are dealing with microscale systems in which
the applicability of classical macroscopic equilibrium thermodynamics becomes question-
able. Generally, the applicability of classical equilibrium theory breaks down because the
system is far from equilibrium and/or the system behavior deviates from a continuum model.
The departure from equilibrium may be due to spatial nonuniformity of properties in the
system or it may be because the system is in a metastable state or because the system has
not had sufficient time to relax to equilibrium. A departure from continuum behavior may
be due to dimensions in the system being comparable to the mean spatial separation of the
molecules, giving the system an intrinsic granularity, or it may be due to the presence of
phases separated by an interface.
As we noted in the previous chapter, the assumption of local thermodynamic equilibrium
is widely used to extend classical equilibrium thermodynamic analysis to nonequilibrium
systems. In some cases, however, even this approach does not provide an accurate model
of the thermophysics of the system. In this final chapter of this text we will explore ex-
amples of microscale systems in which nonequilibrium and/or noncontinuum effects cause
important deviations from classical thermodynamic predictions. In doing so our goals will
be (1) to define the limits of applicability of classical theory and (2) to develop additional
analysis tools from statistical thermodynamics and kinetic theory that can be applied to
such systems. As a first step in developing additional analysis tools for nonequilibrium
systems, in the first section of this chapter we will establish the basic framework of kinetic
theory.

325
326 10 I Nonequilibrium and Noncontinuum Elements of Microscale Systems

The Maxwell Distribution

We now consider a fluid composed of molecules that are free to move about in
space. The total energy of a microstate occupied by a molecule is given by
£ = £trx + £117 + £trz + Mother- (10.1)

The first three terms on the right side of Eq. (10.1) are the kinetic energies associated with
translation in the x, y, and z directions. The fourth term represents the energy associated
with all other energy storage modes. It follows directly that the molecular partition function
is given by
Q = <7to<7tr^trz<7other- (10.2)

The translational contributions are given by relations of the type

n'=0
The summation is over the quantum energy states as dictated by the 1-D particle-in-a-box
quantum solution,

< ) 2
^ ' "x-1'2'3'---- (10-4)
Substituting yields
oo

qta = '52e-hW*mLlk*T. (10.5)

n
'x=°

Considering the translational storage in the x direction, independent of other storage

modes, the probability that a molecule is in a translational microstate between nfx and
n'x + A < is
Yn*+A< 2 2 2
e-h (n') /8mL kBT

V-oo

Assuming we are at high enough temperature that the n'x values are large, the spacing of
the quantum levels is relatively close and we can replace the summations with integrals:
K + A < . e-h2(n')2/8mL2kBT ^ /
p , - Jn'=n'x
* ~" f°° e-h2in'x)2l%mL2kBT dn, '

Letting Anx = dn' in the limit and evaluating the integral in the denominator yields

P n, =
27tmLzkBT
Solving (10.4) for n'x and differentiating gives

dn'x = \l-1^re-^dsux. (10.9)

10.1 I Basic Kinetic Theory 327

Substituting (10.4) and (10.9) into Eq. (10.8), we obtain a relation for the probability that
the x direction translational energy is between stYX and etYX + dstTX:

P tar* tostTX + dstTX) = J-*—;S^ /2e-£trx/kBT dsixx. (10.10)

y TCkBl
Since siYX = mu2/2, the probability that stTX lies between strx and siYX + d£t™ equals the
probability that \u\ is between \u\ and |u| + d\u\. It follows that dsiYX = d(mu2)/2 and

|«| to \u\+d\u\) =J -—--e- m^2l2k*T(m/2)d(\u\2). (10.11)

y 7tm\u\2kBT
By symmetry, the probability that u is between u and u-\-du must be one half the probability
that \u\ is between |w| and \u\ + rf|w|. Thus

1 / 2 mu2/2k T
-J —— e- * (rn/2)d(u2). (10.12)
2 y itrnwkT
Interpreting P (u to w + Jw) as being equivalent to the fraction of molecules with u velocities
between u and u + du, dNu/N, we can write Eq. (10.12) in the form

( 1 C U 3 )

An identical analysis for the v and iu velocity components yields

**T d w

Interpreting the relations (10.13)—(10.15) as the probabilities that a molecule has a velocity
within a differential amount of u, v, or w, it follows that the probability that a molecule
simultaneously is within differential amounts of specific u, v, and w values is,
dNuvw fdNu\fdNv\fdN.
N \ N J\ N J\ N
3/2

27tkBT
Equations (10.13)—(10.16) are various forms of the Maxwell velocity distribution. We
are also interested in the number density of molecules with speed c in the interval c to c + dc
regardless of direction. We therefore convert the above relation to spherical coordinates by
using
u = csin#cos0, (10.17)
v = csin(9sin0, (10.18)
w = ccos6>, (10.19)
c = [u2 + v2 + w2]l/2. (10.20)
328 10 I Nonequilibrium and Noneontinuum Elements of Microscale Systems

Figure 10.1 Relation between polar and Cartesian coordinates.

dc

Figure 10.2 A differential change in c as a spherical shell in Cartesian velocity space.

and the corresponding differential forms of these relations. Executing the change of variables
yields
3/2
dNce<t> _ ( m \ 2 mc2/2k T
c e- » si (10.21)
N " \27TkBTj
In this relation the subscript on dN has been changed to c0</> to reflect the fact that this
is now the number of molecules with speeds between c and c + dc and directional angles
between 0 and 0 + dO and 0 and 0 + d(p (see Figures 10.1 and 10.2).
The number of molecules with speeds in the range c to c + dc irrespective of direction
can be found by integrating the above equation over all values of 0 and </>:

^/T£ dNc0(p.

Substituting the number distribution relation (10.21), we obtain

(10.22)

dNr =4TTN (10.23)

27zkBT
This is the Maxwell distribution of molecular speeds obtained for monatomic gases in
Chapter 2. This distribution is plotted in Figure 10.3. Associated with the Maxwell dis-
tribution we can define the most probable speed, c mp, the root mean square (rms) speed,
(c 2 ) 1/2 , and the average speed, (c). The most probable speed cmp is the speed that maxi-
mizes dNc/dc. Solving the Maxwell distribution relation for this quantity, differentiating,
10.11 Basic Kinetic Theory 329

dN/dc
1

0.8 >*
-
// \
0.4 i \\
1
J
0.2 •••

0-
0 0.5 1 1.5 2.5
V2
(m/2/cBT) c

Figure 10.3 The Maxwell speed distribution.

and setting the result equal to zero yields

(.0.24,

The mean speed (c) and the rms speed (c2)1^2 are computed by weighting the appropriate
parameter with the fraction of the molecules in each speed interval and integrating over all
possible speeds:

[°° c (dNc\ d c fSkBT\l/2

= [ 1JT~ ) = [ (10.25)

(10.26)

Note that these three representative molecular velocities are all proportional to {k^T/m) l/1
with the proportionality constant ranging from 2 1 / 2 to 3 1 / 2 .
The fraction of molecules with speeds greater than a given value c is determined by
integrating dNc/N from c to infinity. Designating the fraction as f>c, we have

Upon evaluating the integral, the resulting relation can be written as

2c
f>c = exp{-(c/c mp ) } - erf{c/cmp} (10.28)

where cmp = (2kBT/m)l/2.

Example 10.1 Determine the fraction of air molecules that would have velocities greater
than the escape velocity at 300 K on the earth and the moon. What do the results imply
about the tendency for these bodies to hold a gaseous atmosphere near their surface?
330 101 Nonequilibrium and Noneontinuum Elements of Microscale Systems

Solution Assuming that the mean molecular mass of the gas molecules is the
same as that for air we have:
29kg/kmol 26
m = 6.02 x 1026 molecules/kmol = 4.8 x 10~ kg/moloecule.
For the most probable speed
cmp = (2kBT/m)l/2 = (2 x 1.38 x 10~23 x 300/4.8 x 10~ 26 ) 1/2 = 415 m/s.
For the earth, the escape velocity is 1.1 x 104 m/s. To determine the fraction of the molecules
that exceed this value we first determine c/cmp = 26.5; we then substitute this ratio into
Eq. (10.28) to determine f>c. In doing so, we noted that for large values of its argu-
ment 1— erf x approaches zero more rapidly than exp(—x 2). The relation for f>c then
becomes
2c
f>c =
7TC,
mp

which is an extraordinarily small number (~10 30°). On the moon, the escape velocity is
only 2,000 m/s. Thus c/cmp = 4.82 and it follows that

^ {_(4<82)2} =

This fraction is significant given that the number of molecules might be ~10 40 . Thus, on
the moon there would be a steady loss of molecules escaping gravity into space as a result
of their thermal motion. Eventually the entire atmosphere would escape into space.

We now turn our attention to the following question: What is the flux of molecules
through (or striking) a flat surface in a gas? Considering the region within the box shown
in Figure 10.4, we want to know the number of molecules in the box that strike the shaded
surface E* per unit area, per unit time. Regardless of i; and w, a molecule with a given u
must be within uAt at the beginning of the time interval At to pass through the surface. We
know from previous results that the fraction of molecules with u velocities in the interval u
to u + du is

(10.29)

Figure 10.4
10.11 Basic Kinetic Theory 331

We also know that of the molecules in the box with velocity w, only those within a fraction
of the total box volume equal touAt/Lx will pass through the surface in time At. We define
the differential flux dju as the number of molecules with velocity u passing through surface
£ * per unit time, per unit area. It follows from the above discussion that dju must be given by

(10.30)

where V =LxLyLz. Integrating to get contributions to the total flux from all possible u
velocity values, we obtain the following relation for the total flux j of molecules across
surface J^*:

(10.31)

which reduces to

(10.32)

Thus j is proportional to the density of molecules (N/V) and the mean speed (c).
We consider now the escape of molecules of a gas through an opening in a wall of a
container, as indicated in Figure 10.5. If the size of the opening is large compared with the
mean spacing between molecules, a bulk flow of the gas results that can be analyzed with
continuum fluid dynamics theory. If the region outside the container is evacuated, choked
sonic flow occurs through the opening. The mass flux, from compressible flow theory, is
given by

_ frnN\ /ykBT (10.33)

Jm,large — I
T/ I\ >
\ V J V m
where y = cP/cv is the ratio of specific heats. If, however, the opening size is smaller
than the mean spacing of the molecules, the mass flux is j multiplied by the mass per
molecule:

_1fmN\ IUBT (10.34)

Jm,small — "7 I T7 I\ •
4 \ V J V 7TW7

Figure 10.5
332 10 I Nonequilibrium and Noneontinuum Elements of Microscale Systems

By taking the ratio of these two relations, we find

Jm, large
(10.35)
Jm, small

Thus, the mass flux for the small hole is about a factor of three lower than the macroscopic
compressible flow prediction.

Collision Processes
In this section we will develop a simplistic model of molecular collisions processes.
In doing so we will consider a system containing a single kind of molecule. The average
number density of molecules (N)/V will be designated as p^. In this simple model, all
molecules move with the same speed c m, although the direction of the velocity vector is
uniformly distributed in all directions. We begin by considering one specific target molecule
in the system and define g as the mean relative speed of other molecules in the system
with respect to this one molecule. In our simple model, the molecules are presumed to
be spherically symmetric with diameter D. Molecules collide if their motion brings their
centers to within a distance less than their diameter D (see Figure 10.6). This is essentially a
hard sphere model. The collision cross section ac presented by the target molecule is given
by
ac = 7tD2. (10.36)
Molecules that will collide with the target molecule in time dt have their centers within a
circular cylinder of length g dt and volume acg dt. The number of molecules in this volume
is pN<ycg dt. It follows directly that the mean frequency of collision with the target molecule
yc is given by
yc = pNcrcg. (10.37)
To determine the mean relative speed, we consider the velocity vectors of two arbitrarily
chosen molecules 1 and 2, as indicated in Figure 10.7. From the geometry, the magnitude
of the relative velocity vector |g 12 | is given by

|gi2l = [|v 2 | 2 sin 2 0 - |v 2 |cos^) 2 ] 1 / 2 . (10.38)

Equation (10.38) can be rearranged to obtain

(10.39)

Jcos0
Figure 10.6 Figure 10.7
10.11 Basic Kinetic Theory 333

Taking g as being the root mean square value of \gn\, we have

8 = (Igi2|2)1/2 = (Iv2|2 + IvH2 - 2|v,||v 2 |cos0> 1 / 2 . (10.40)
In our model the magnitude of the velocity vector for any molecule is cm. Hence | v 11 = cm
and |V2| = cm and the relation for £ reduces to

8 = <lgi2|2>1/2 = (2c2m-2c2mcose)l/2. (10.41)

If we take the average over all values of 0 from 0 to 2n, the cosine term in Eq. (10.41)
averages to zero, and the relation for g becomes
8 = V2c m . (10.42)
Substituting this result into Eq. (10.37) for the collision frequency yields
yc = V2pNaccm. (10.43)
The mean time between collisions (rc) is just the reciprocal of the collision frequency:

<rc) = 1 = — . (10.44)
v ^2
The mean free path A.c is defined as the mean distance traveled by a given molecule between
collisions. The mean free path is taken to be equal to the product of the mean time between
collisions (rc) and the mean speed for a given particle (c):
Xc = (rc)(c). (10.45)
Taking (c)=cm and substituting the relation (10.44) for (r c ), we obtain the following
relation for the mean free path:

^ ( 1 0 . 4 6 )

If ac = nD2, the relation for the mean free path becomes

Ac = — -. (10.47)
V2npND2
Using the ideal gas equation of state to evaluate the density, the relation for Ac from our
simplistic analysis can be written

<1048)

Example 10.2 For nitrogen gas at atmospheric pressure and 300 K, estimate the mean free
path and the mean time between collisions if the effective diameter of the molecules is 4 A.

Solution The mean free path is computed using Eq. (10.48). Substituting the
specified values we get
k BT 1.38 x l O " 2 3 (300) o
A.c = -j= — = — — = 5.76 x 10 8 m.
V2nD2P 10 2
V2TT(4.0 X lO" ) (101,000)
334 10 I Nonequilibrium and Noncontinuum Elements of Microscale Systems

The mean speed is given by Eq. (10.25):

'8kBT\l/2 ((8) 1.38 x 10~ 2 3 (300)\ 1 / 2

im ) \ TT(28/6.02 X 10~26) /
The number density is computed from the ideal gas law as
P 101,000 22
PN = =
k^T 1.38 x 10~ 23 (300) = 2M X 10

Finally, we use these results to compute the mean time between collisions using Eq. (10.44):

<*> = •
= 1.21 x 10~7s.
V2(2.44 x 1022);r(4.0 X 10- 10 ) 2 476

Transport Properties
We now wish to explore the relationship between the mean free path Xc and
transport in gases. To explore momentum transport, we consider the classic problem of
flow between two parallel plates shown in Figure 10.8. For the circumstances shown in
Figure 10.8, (u) varies from 0 to U between the plates. From our knowledge of macroscopic
(laminar) fluid mechanics we expect (u) to vary linearly between the plates. It follows that
the bulk momentum varies as

(Px) = — . (10.49)
a
Although equally many molecules from above and below cross a plane at a given height y,
the ones from above carry more momentum down than upward-moving ones carry upward.
As a result, there exists a net downward flux of momentum.
Let j equal the flux of a transported quantity (here it is momentum). The number of
molecules crossing a horizontal plane per unit area, per unit time, with velocity c = (u, v, w)
is equal to the number of molecules in a parallelepiped with a base of unit area and a height
equal to | v |. The volume of the parallelepiped is | v | and the number of molecules in it with
the specified velocity c is \V\PN f(u, v, w) du dv dw.
Note that if v is positive, particles come from below, and if it is negative they come from
above. Because these are equally probable, the net flux is zero. A molecule crossing the

y = a y////////,
i W/////////AV////// moving upper plate
/ - —^u
gas JJy_
y («> a

1 fc A. f) V
y/////////////, '////Ay/, tixed lower plate

Figure 10.8
10.11 Basic Kinetic Theory 335

horizontal plane has, on the average, traveled a distance Ac since its last collision. It follows
that yf, the average y location of its last collision, is
y = y-vXc/\ci (10.50)
where
|c| = [ w 2 + V + u; 2 ] 1/2 . (10.51)
We assume that during the last collision, the molecule came into local equilibrium with the
environment and therefore carries the momentum associated with height y'. Expanding the
x momentum in a Taylor series, we obtain

{Px)y=y, = {Px)y=y -
The net flow of momentum (px) is given by integrating the product of (px)y=y> and the
number of molecules crossing the surface from height yr over the entire range of velocities:
'00 /»OO
/•OO /»OO

/ / (px)y = yvpNf(u,v,w)dudvdw. (10.53)

—oo J-oo J—oo

If we substitute the Taylor series expansion (10.52), the relation for j(y) becomes
•OO /•OO /»OO
d{px) vXc~
' (Px)y=y> ~
—oo J —oo J —oo

(10.54)
Using the fact that {px)y = yf = mUy/a and substituting Eq. (10.51) for |c|, we obtain
/.oo roo /.oo VmUyl
= / / I I VPN f(u, v, w)dudv dw.
J—oo J—oo J—c
L aJ
, fd(px)\
PNK f°° f°° f°° \ v2f(u,v,w)
- \n^) L L L V*+*+wi*
dudvdw.
(10.55)
For these conditions

exp / m[(u - (u))2 + v2 + u>2]\ (ia56)

{ w 1'
If (u) is small compared to the molecular velocities (since molecular speeds are compa-
rable to the speed of sound this implies low Mach number), we can simply neglect it in the
above expression for / . Evaluation of the triple integrals using this reduced relation for /
is then straightforward. The first integral in Eq. (10.55) is zero and evaluation of the second
yields
1 l
"" (10.57)

where (c) is the mean speed defined by Eq. (10.25). Noting that (px) = m{u) this can be
written as
L
{^\^f (10.58)
336 10 I Nonequilibrium and Noncontinuum Elements of Microscale Systems

or more generally as
1
JOO = -( I)PNK(C)V(U). (10.59)

We note the similarity between the above relation for j and Newton's law of viscous flow,

^ (io.6O)
dy
which implies that for the gas

Mm = (l)pNmXc{c). (10.61)

Using previously derived relations for the mean free path and the mean speed and using the
ideal gas relation to evaluate the density, we can convert this relation to

It can be seen that /xm is independent of density and pressure.

Note that in the above analysis we did not need to specify the form of (p x) to derive the
above relation for j(y). When (px) is interpreted as momentum, j is the transport (flux) of
momentum. The same analysis can be applied to heat transfer if (p x) is replaced with (e),
the average energy per molecule. Setting (px) = (e) yields

= c^
K }
dy dy dT dy NNA
where cv is the molar heat capacity. Substituting in the above analysis yields

where j is now a heat flux vector. Comparing this with Fourier's law,

j G = -*tV7\ (10.65)

we conclude that

= /1 \ f^M^L) = ^L. (10 .66)

V3yV ^ J N
The above result implies that iimcy /'mN\kt = 1. A more refined analysis indicates that this
ratio is constant but equals a value somewhat different from 1.
Although we will not pursue it here, this analysis can also be extended to mass dif-
fusion in multicomponent gas mixtures (to predict the mass diffusion coefficient). More
rigorous kinetic theory models can be constructed that treat the collision process in a more
sophisticated manner. The interested reader can find more information on such treatments
in References [l]-[3] listed at the end of this chapter.
10.2 I The Boltzmann Transport Equation 337

10.2 The Boltzmann Transport Equation

System Behavior in Phase Space

In the previous section we have interpreted the distribution function as unchanging
in time. This is consistent with our development of the statistical thermodynamics frame-
work, which implicitly assumes that the system is in an equilibrium state characterized by
an equilibrium distribution that is unchanging in time. In general, however, we expect that
if the system is initially in a nonequilibrium state, the distribution will be different from
the relation that applies at equilibrium. As the particles in the system move and collide, the
distribution will change with time as the system configuration evolves, ultimately achieving
the equilibrium distribution as time goes to infinity.
To begin our consideration of these issues we consider a classical system of N molecules.
Each molecule has s degrees of freedom so that the number of coordinates needed to
specify positions of all N molecules is l = Ns. Note that the coordinates may include
angular coordinates and relative positions as well as translation coordinates. The / spatial
coordinates qi and / corresponding momenta pj completely specify the classical mechanical
state of the system. The initial qt and pj together with the equations of motion completely
determine the future (and past) course of the classical system.
We now include in the analysis a conceptual Euclidean hyperspace of 2/ dimensions,
with a coordinate axis for each of the 2/ coordinates and momenta. Gibbs termed this
conceptual space a phase space for the system. The state of the classical TV-body system
at any time t is completely specified by the location of one point in phase space, referred
to as a phase point. The evolution of the system state with time is completely described by
the motion or trajectory of the phase point through phase space. The trajectory of the point
is dictated by the equations of motion of the N bodies, as prescribed by Newton's laws
and the force interactions among the molecules. These, in principle, can be integrated if the
initial location of the point is specified. In practice, integration of such a large system of
equations is not feasible.
We now consider a microcanonical ensemble of systems in phase space. There are
A systems in the ensemble, each having the same N, V, and U. The classical state of
each system in the ensemble is represented by a point in phase space. As each sys-
tem evolves in time, each point will trace out its independent trajectory. It is impor-
tant to note that the trajectories are independent because the systems in the ensemble
are themselves isolated and independent. The postulate of equal a priori probabilities
(see Chapter 2) requires that there is a representative phase point in phase space for
each and every set of coordinates, consistent with the fixed macroscopic constraints (N,
V, U). Associated with this ensemble we define a number density / e s (p, q, t) such
that

(the number of systems in the ensemble that \

/ es (p, q, 0^pdq =
have phase points in dp dq about p, q at time tj
V (10.67)

In the above definition p and q are vectors px, p2,..., P\ and q\,qi, . . . , § / , respec-
tively, dp = dpx dp2 . . . dph and dq = dq\ dq2 . . . dq\. From this definition, it follows
338 10 I Nonequilibrium and Noncontinuum Elements of Micro scale Systems

directly that

allp,q
The ensemble average of any function ^ (p, q) of the momenta and coordinates of the
system is defined by

(t) = j Iff -'[ ^(p, q)/es(p, q, 0 dp dq. (10.69)

allp.q

Gibbs postulated that this ensemble average is, in fact, the corresponding thermodynamic
function.
Since the equations of motion determine the trajectory of each point, they must also
determine the density of points at any time, if the dependence of / e s on p and q is known
at some initial time t0. Thus, for a classical system, the time dependence of / e s is dictated
by the laws of classical mechanics.
We consider a small volume of phase space 8px8p2 . . . 8pl8q\8q2 ...8qi about the point
px, p2, - - •,Pi, qi, qi, • ••,qi- The number of system state points inside this volume of phase
space at any instant is
&n = fes{Pi,P2> •-.,P i , q u q i , . . . , q i , t ) 8 p l 8 p 2 . . .SpfiqiSfc, ...8qh
(10.70)
Note that 8n will change with time as points enter and leave through faces of the differential
volume element. Consider the two parallel surfaces of the element that are perpendicular
to the q\ axis located at q\ and q\ + dq\. The number of phase points entering through the
first face (at q\) per unit time is given by

8n^ = feS8pl8p2...8pl8qi8q2...8qi-±-. (10.71)

bqi
Since the volume element is of differential dimensions, the rate at which points pass out of
the element through the other face can be expressed as
/es + ~^roqi 8 p x 8 p 2 ... 8 p l 8 q l 8 q 2 ... 8qt \ q x + —<$<?i —-.
dq\ J \ dq\ J dqx
(10.72)
where / e s and the derivatives are evaluated at q\. Subtracting 8h^+d^ from 8h^ yields the
net flow of phase points in the q\ direction into the volume element, which reduces to

q2...8qi. (10.73)

In a similar manner, the following relation can be derived for the net flow into the volume
in the px direction:

8nPl -8nPl+dPl =-[ — P i + / e s — )8p l8p2.. .8pl8ql8q2 .. .8qh

(10.74)
10.2 I The Boltzmann Transport Equation 339

Similar relations can be derived for the net flow in all the other coordinate directions. The
total net flow into the volume through all faces of the element is then found by summing
contributions from all directions in phase space.
d
^ 3/es • h 3/es • pj\
« +/ + P +f JSp8p . . . 8pl8ql8q2 . . . 8qh
(10.75)
This total net inflow must equal the change of 8n with time, which implies that

(10.76)
In Chapter 1 we noted that, for a system having many coordinates and momenta, the
Hamiltonian H satisfies
dH dH
1 <10J7)
*-->'• wr' -
Differentiating the first relation with respect to pj and the second with respect to qj and
equating the expressions for the cross derivatives of H yields
dqj dpi
-LL = ^. (10.78)
dqj dpj
This result implies that the factor multiplying / e s in Eq. (10.76) is zero. Using this fact and
using Eq. (10.70) to eliminate 8n transforms Eq. (10.76) to

Equation (10.79) is the Liouville equation, which is an important cornerstone of classical

statistical mechanics.
Note that we can apply this analysis for an ensemble of systems to a distribution for a
system of noninteracting particles in the following manner. We simply consider each par-
ticle in our system of interest to be a "system" in our ensemble. This is possible because
a particle or molecule has a defined set of coordinates and momenta that describe its state
just as the systems in the ensemble did. Each molecule in the system then can be viewed
as moving through phase space as its microstate changes with time in the system due to
its motion through physical space. Note that because we assumed that the systems in the
ensemble did not interact, the Liouville equation can be applied only to particles that do not
interact.
Designating the distribution function as / e p for an ensemble of particles, we replace / e s
in Eq. (10.79) with / e p and rearrange slightly to obtain

Since / e p = Uv(q\, qi, • •

•,<?/, plt p2, •••, Pi, t) Eq. (10.80) is equivalent to the total
340 10 I Nonequilibrium and Noncontinuum Elements of Microscale Systems

differential of / e p with respect to t being equal to zero:

d
-dT=0-
This is essentially a conservation equation for "convective" transport of particle state points
through phase space. This result leads to several important conclusions. First, this equation
implies that the density of particle phase points in the neighborhood of any selected moving
phase point is constant as that phase point moves along its trajectory. Gibbs referred to this
as the principle of conservation of density in phase. Also, if (po, qo, to) a r e the coordinates
of a particle's phase point at time to, Liouville's equation implies that for all subsequent t

/es(p, q, 0 = /es(Po, qo, to) (10.81)

for that phase point as it moves along its trajectory through phase space. All subsequent
points traversed by a particle in phase space are linked to a unique original point at time Jo.
Since all the trajectories are dictated by the same equations of motion, two phase points
could occupy the same location in phase space only if they started at the same point, which
is not possible since no two particles can occupy the same physical location. It follows that,
for noninteracting particles, trajectories of phase points can never cross, and the particle
state points must move through phase space along nonintersecting trajectories.

The Boltzmann Transport Equation

By applying Eq. (10.80) to an ensemble of particles, the phase space corresponds to
the combination of momenta and coordinates accessible to the particle. Since each momenta
is always linearly related to corresponding velocity, we can define generalized coordinates
Zj and generalized velocities Wj such that

qj=2j, (10.82)
pj =rhjWj, (10.83)
where the rhj are generalized inertia factors. Using these relations to replace p j with Wj
and q} with zy- and interpreting the distribution function in terms of velocities as

(the number of particles in the ensemble that

fp,{vf,z,t)dvdz=[ ,....„.:_.. , • .) dO.84)
\have phase points in dwdz about w, z at time t
we can transform Eq. (10.80) to

—— + y] Zj —-j— + y^ Wj -—j— = o. (io.85)

j=\ J j=\ J

In Eq. (10.84), w and z are vectors w\, W2, . . . , wi and z\, Z2, . . . , z/, respectively, dw =
dw\dw2 .. .dwi, and dz = dz\dz.2 . . . dz\. Since along particle flowlines in phase space
Zj = xbj, we can further modify Eq. (10.85) to
10.2 I The Boltzmann Transport Equation 341

Thus, the Liouville equation indicates that if we follow the particles in a volume element
along a flow line in phase space without collisions, the distribution is conserved. It follows
that
/ pv (z + dz, w + dw, t + dt) = / pv (z, w, 0 . (10.87)

If collisions do take place, the distribution / p v will change over a differential time interval
dt by an amount (3/pv/90coii dt, and therefore
/pv(z + dz, w + dw, t+dt) — / pv (z, w, t) = (3/pv/aOcoii dt, (10.88)
which rearranges to
/pv(z + dz, W + dw, t+dt)- / pv (z, W, 0 _ ( 9/pv ^ Q

a* /coll
The left side of Eq. (10.89) is dfpv/dt, which from basic calculus is given by

f 4 t
Replacing the left side of Eq. (10.89) with the right side of Eq. (10.90) we obtain

() . (10.91)
ft jt dwj V dt Ao.i
Equation (10.91) is the Boltzmann transport equation. This equation plays a central role
in the development of kinetic theory for nonequilibrium systems. Note that if we define a
fractional distribution of the particles in the system as
/ the fraction of system particles in the ensemble that
/(w, z, t)dwdz = I
\have points in dw dz about w, z at time t
(10.92)
then
/(w,z,0 = ^ l ^ , (10.93)

and by substitution into Eq. (10.91), it is clear that the fractional distribution also satisfies
the Boltzmann transport equation,

Thus, we are free to work with either a number density distribution for particles or a
fractional distribution in our analysis of the time evolution of the system.
For many systems, the collision term in the Boltzmann transport equation (d//dOcoii
may be treated with the introduction of a relation time r r defined such that

(1
3t • ' °- 9 5 )
342 10 I Nonequilibrium and Noncontinuum Elements of Microscale Systems

where / 0 is the equilibrium distribution for the system. In general rr may be a function of
velocity (or energy) and position in the system. Combining Eqs. (10.94) and (10.95) we
obtain the Boltzmann transport equation with the relaxation time approximation:

h> u);
J
h > Wj
J
= . (10.96)
dt ^ dzj J
^ dwj J
rr
7=1 7=1

As an example, we consider a spatially uniform nonequilibrium distribution of veloci-

ties that is established in a system by external mechanisms that are suddenly removed. The
system is expected to relax to its equilibrium distribution in the absence of external mech-
anisms and forces. If no forces act on the particles, the acceleration terms in Eq. (10.96)
are zero. The spatial gradient terms are also zero since the distribution is spatially uniform.
Using the relaxation time approximation, the Boltzmann transport equation reduces to
d
-l = -l^A. (10.97)
dt rr
If the initial distribution is known to be fu the above equation can be solved by elementary
methods to give

f = fo + (fi-fo)e-'/T'. (10.98)

Example 10.3 A system contains atoms with mass m in a gaseous state. The state of the
system may be assumed to be spatially uniform and effects of external force fields may be
neglected. Initially the fractional distribution function / is given by

{ ^3

0
for — vc < u < vc, —v c < v < vc, —v c < w <

for vc < |w|, vc < \v\, vc < \w\.

v c,

Use the Boltzmann transport equation with the relaxation time approximation to explore
the time evolution of the distribution function.

Solution When the system reaches equilibrium, it will establish the Maxwell
distribution
m
\ -m(u2+v2+w2)/2kBT
ZTTK

Multiplying f\ and / 0 by (l/2)mc2, integrating over all u, v, and w values, and equating
the results, wefindthat conservation of energy requires that the final temperature is related
to vc as follows:

vc =
V m) '
If we substitute into Eq. (10.98), the variation of / with time can be expressed as
10.2 I The Boltzmann Transport Equation 343

0.3 v = w=0 •v\ t > T = 0, 0.5, 1, 2 10

0.25 / .

0.2 / , -
0.15
I
V
/
1
0.1

0.05

0 , , (MOpM •*••»—i—i—
-3 -2
ulvc

Figure 10.9

where f[ is the step function defined above. Note that / is a function of time and the
three velocities that define phase space. To gain some insight into the temporal variation,
the variation of / with u along the v = 0, w = 0 plane is plotted for different times in
Figure 10.9. The plot indicates that the time required for the gas to establish its equilibrium
distribution is about 5 to 10 times r r. If rr is on the order of the mean time between
collisions, this implies that it takes an average of about 5 to 10 collisions per molecule to
reach equilibrium.

Moments of the Boltzmann Equation

The link between the distribution of velocities and macroscopic transport phe-
nomena is established by considering moments of the Boltzmann transport equation. Here
we will limit our attention to a system with no external forces fields. We further limit our
analysis to a system of particles having transportable quantities that are at most a function of
particle translational velocity components. There are then only three translational velocity
components to be considered. In such a system, the acceleration terms in the Boltzmann
transport equation are zero, reducing it to the form

(10.99)
coll

Let ir(ibj) be any function of translational velocity components but independent of location
and time. We multiply both sides of Eq. (10.99) by if(vbj) and integrate over all possible
vbj values to obtain

r r r / £i -F \

z
= / / /VKwy)(—-)
j J J J \ d r
/ coll
y=i
t
(10.100)
344 10 I Nonequilibrium and Noncontinuum Elements of Microscale Systems

When the transportable quantity of interest is a power law function of velocity, as is the case
for momentum or kinetic energy, conversion of Eqs. (10.99) to (10.100) amounts to taking
a moment of the Boltzmann transport equation. We reorganize Eq. (10.100) by reversing
the order of differentiation and integration to obtain

d f f f . . . v^ 3 r r r
ft J J J ~1 3Zj J J J
ootooo J~l —ootooo

= (1 f f fxlffdw dw d (10.101)
Wl W2
\dt J J J ^
—ootooo
coll

Since / is the fraction of particles in a differential volume of phase space around (w\, W2, w^),
it can also be interpreted as the probability offindinga particle in the system with this com-
bination of velocity components. It follows directly that the mean value of x// is given by

(x/f) = i/fdwidw2dw3 (10.102)

—ootooo

and the mean value of the Wjxjf product is

(xbjif) = ffwjdwidw2dw3. (10.103)

—ootooo

Using Eqs. (10.102) and (10.103) to replace thefirsttwo integrals in Eq. (10.101), we obtain

(10.104)
o t dz u l
j = l j \ /coll

The relation (10.104) applies to any \j/ that is a function of particle velocities. For
concreteness, we will analyze this equation further for the specific case where \\r is the
kinetic energy of a particle:
= \mc2, (10.105)
where
c2 = w\ + w\ + w\. (10.106)
Substituting in Eq. (10.104) we get

2 2
Ii i ^ ? f ^ ) (10.107)
2 dt 2 r^dzj \dt y coll

For elastic collisions, the total energy of the particles in the system does not change with
time, and therefore the mean energy per particle in the system does not change with time.
It follows that the right side of Eq. (10.107) is zero. It can therefore be written

^-(c2) + y2^r(wjC2)=0. (10.108)

10.2 I The Boltzmann Transport Equation 345

The mean flux of kinetic energy in the Zj direction is given by

UQJ) = PNWj(l/2)m(wl + wl + wl)fdwidw2dw3, (10.109)

—ootooo

where the number density of particles in the system p^ = N / V is taken to be constant. The
right side of the above equation is therefore {\/2)mp^ times (vbjC2) and hence

. (10.110)
Combining Eq. (10.110) with (10.108), we find

£TUQJ) = 0. (10.111)
zj

For a system of particles that stores energy by translation only, we demonstrated in Chapter
2 and in the previous section that

2W 00.12,
m 2M M
If the system obeys the ideal gas law, cy = dp — R = dp — PM/mp^T and

M mpm

Substituting for (c2) in (10.111) and simplifying yields

J
7= 1

If the pressure in the system is constant, this further reduces to

3
mppjCp dT yr-^ 9 ,1A1in
= \~ / —Z-(jQi)=Q> (10.H5)
z
M dt -^ oZj
Equation (10.115) is one form of the well-known energy transport equation for a system
of particles in the absence of bulk motion. If, as argued in the previous section, the energy
flux is related to the temperature gradient as

we obtain an energy transport equation that is equivalent to Fourier's Law, discussed in

Chapter 9 from the point of view of macroscopic nonequilibrium thermodynamics,

o. (10.117)
M dt

Note that the kinetic theory analysis from which the transport equation was derived
embodies that assumption that the distribution / is the Boltzmann distribution derived for
an equilibrium system, Thus, although properties may vary spatially throughout the system,
346 10 I Nonequilibrium and Noneontinuum Elements of Microscale Systems

this analysis is based on the assumption that, locally, / is the Boltzmann distribution for the
local temperature and density of the system. This is essentially equivalent to the macroscopic
assumption of local thermodynamic equilibrium used widely in macroscopic analysis of
nonequilibrium systems. If locally / deviates substantially from the Boltzmann distribution
for the local conditions, the validity of the analysis is questionable.
With some additional effort, the energy transport equation with bulk motion effects can
be derived in this manner, as well as relations for transport of momentum and transport
of species in multicomponent systems. Each of these involve different moments of the
Boltzmann transport equation. The interested reader can find the details of such derivations
in References [ 1 ]-[3]. These derivations provide the conceptual link between the microscale
analysis tools of kinetic theory and macroscopic transport theory.

10.3 Thermodynamics of Interfaces

Macroscopic thermodynamic and fluid mechanics treatments of the boundary

between two phases often assume a sharp discontinuity in density and/or composition
across the boundary. This sharp boundary is termed an interface between the phases. In
considering the details of phase equilibria and physical processes at the interface, it is
important to recognize that changes in properties, which can sometimes be quite large,
actually occur over a very thin region centered about the nominal position of the inter-
face. Figure 10.10 schematically shows the density distribution near a liquid-vapor in-
terface. In the liquid, the density is lower in the interface region than in the bulk liquid
phase. As described in Chapter 6, the molecules attract each other at moderate separa-
tion distances, and energy must be supplied to move them apart. Hence, the mean energy
per molecule associated with this type of interaction is greater in the interfacial region
than in the bulk liquid. The system thus has an additional free energy per unit area of

bulk liquid bulk vapor

distance normal to interface

Figure 10.10
10.3 I Thermodynamics of Interfaces 347

bulk fluid I bulk fluid II

interfacial region

U i
1 ' 1
1 / l
1 1
1 i 1
1
1
1
y
II
V
1 ' 1
1 [ \
1 1

1 , 1
1
\ - ^ 1
1 1

Figure 10.11

interface owing to the slightly larger mean spacing of the molecules in the interfacial
region.
To explore these issues in more detail, we specifically consider the interfacial region
between two pure fluids that is presumed to be bounded by two surfaces T>i and En that
are parallel to the nominal interface surface E* but are located in the corresponding bulk
phases. As shown in Figure 10.11, the mean specific volume and internal energy actually
vary continuously across the interfacial region. Because the continuous transition in mean
property values between the bulk phases occurs over a region of finite thickness, the actual
mass and internal energy of the region may differ from values that would be calculated
by assuming that the bulk phase extended all the way to E*. This difference is called the
surface excess of the property. The surface excess mass Tf* is defined
dy Jn
if = (10.118)
-y\

and the surface excess internal energy (/e is given by

yu fj
u dy
(10.119)
-yi

Note that both of these properties are defined per unit area of surface E*. Surface excess
values of other thermodynamic properties can be similarly defined. Note that these properties
have been give a superscript E* because they depend on the location of surface E*. This
concept facilitates a link between the idealized concept of a two-dimensional interface and
the actual three-dimensional character of the interfacial region.
To examine the relationships among properties in the interfacial region and properties
in the surrounding bulk phases at equilibrium, we consider the system in Figure 10.12. The
348 10 I Nonequilibrium and Noncontinuum Elements of Microscale Systems

overall system isolated

Figure 10.12

entire system, consisting of the interfacial region between Sj and En and bulk fluids I and
II in Figure 10.12, is taken to be completely isolated from the surroundings. If the shape of
the interfacial region (and hence its volume) is assumed to be fixed, its internal energy may
be taken to be a function only of the number of molecules and its entropy, U = U (N, S). It
follows from results obtained in Chapter 2 that
dU = iidN + T dS, (10.120)
where

S
//V,shape \ d y V / 1 S,shape

We consider possible perturbations that leave bulk fluid / unchanged while permitting
exchange of energy and mass between the interfacial region and bulk fluid / / . Because the
overall system is isolated, we must have
09 (10.122)
0, (10.123)
fjikdNiY + TirdSir + fJindNu + TudSu = 0. (10.124)
Combining Eqs. (10.122)-(10.124) yields
(Mir - Mil) dNiT + (r ir - r n ) dSiY = 0. (10.125)
Because Eq. (10.125) must be satisfied for all possible perturbations of Afir and Sir, necessary
conditions for equilibrium are /xir = /xn and Tir = 7 n . An identical set of arguments can be
presented for perturbation interactions between the interfacial region and bulk phase I. The
net result is that, at equilibrium, the temperature and chemical potential must be the same
in all three regions:
Mir = Mi = Mil, 7]r = Ti = r n . (10.126)
Equation (10.120) applies to the interface region if the interface shape is fixed, regard-
less of the variation of thermodynamic properties within the region at equilibrium. Taking
Eq. (10.120) as representing the actual property variations in the interfacial region and
dropping the subscripts on \x and T since they are the same in all regions, we have
dUir = LidNir + T dSir. (10.127)
103 I Thermodynamics of Interfaces 349

An identical relation would apply if the properties equaled those of the respective bulk
phases all the way to the interface surface X*. Designating properties in that fictitious case
with a prime we have
dU;T = fi dN[x + T dS[T. (10.128)
Subtracting Eq. (10.127) from Eq. (10.128) we obtain
d(u!Y - Uir) = nd(N[x - Mr) + Td(S[Y - Sir). (10.129)
But by definition, the terms in parentheses in (10.129) are excess properties,
Uf = U!r - Uk, N?'=N!r-Nir, Sf'=S'ir-Sk, (10.130)
and so Eq. (10.129) can be written
dUf* = [i dNf + T dS?* (fixed interface shape). (10.131)
If we now relax the fixed-shape restriction and allow deformation of the interface surface
X*, in general both its area and its curvature can change. If the principle radii of curvature
for the surface are much greater than the interfacial region thickness, the effect of changing
curvature on thermodynamic properties is expected to be very small. Since this is usually the
case, we will neglect curvature effects on properties here. However, we do want to account
for the effect of changes in interfacial area resulting from deformation of the interface. We
therefore assume that in addition to being a function of N^* and Sf*, U^P* is also a function
of the area of surface X*: f/eE* = Uf* (Nf*, Sf\ AE*). Thus, mathematically, it follows
that

(10.132)
and since Eq. (10.132) must reduce to (10.131) if dA^* is zero, we conclude that

We similarly define the interfacial tension ax as

. cam,
which allows us to write Eq. (10.132) as
dUf* = ixdNf* +T dSf +o{dAi:\ (10.135)
For the interfacial region, the Helmholtz free energy F is given by
F = U-TS. (10.136)
Subtracting from Eq. (10.136) the corresponding equation that would apply if the two parts
of the interfacial region were occupied by bulk fluids I and II yields an analogous relation
for the surface excess free energy:
F?* =UF -TSF. (10.137)
350 10 I Nonequilibrium and Noncontinuum Elements of Microscale Systems

Differentiating Eq. (10.137) and substituting the right side of Eq. (10.135) for dUf* yields
dF?* = -Sf* dT +fidN?* + Oi dA*\ (10.138)
However, if we consider FeE* = Fes* (T, N?\ As*) from a purely mathematical point of
view, we can write

(10.139)
Comparison of Eqs. (10.138) and (10.139) clearly indicates that the interfacial tension is
related to the surface excess free energy as

Thus ox is equal to the change in surface excess free energy produced by a unit increase in
interface area A x*.
An additional useful relation can be obtained by considering a shift in the nominal
interface surface E* uniformly toward region II by some amount dy. The position of E* is
somewhat arbitrary because of the continuous property variations in the interface region.
But, if our analysis is to yield meaningful results, the values of properties in the three regions
in Figure 10.12 should be independent of where E* is located. In allowing E* to shift by
dy, we therefore expect that the overall free energy for the system as a whole is unchanged
because no change occurs in the physical state. From Eq. (3.75) obtained in Chapter 3, we
know that for the pure bulk fluids

dF = -SdT + iidN - PdV. (10.141)

Using Eq. (10.141) for the bulk fluids and Eq. (10.138) for the interfacial region, we can
write the total change in F for the three regions in Figure 10.12 as
dF = -SfdT + ii dNf* + cri dA^
-SidT + /x dNi - Pi dVi - Sn dT + /x dNu - P\\ dVu = 0. (10.142)
Because the total number of molecules and total volume do not change and there is no
change of temperature associated with the position shift,
dN = dNi + dNu + dN^ = 0, (10.143)
dV = dVi + dVn = 0, (10.144)
dT = 0. (10.145)
Combining Eqs. (10.142)-(10.145), we can reduce the resulting relation to

(10.146)
dV\
Here we characterize the surface E* by two radii of curvature measured in two perpen-
dicular planes containing the local normal to E*, as indicated in Figure 10.13. The changes
10.4 I Molecular Transport at Interfaces 351

fluid II

Figure 10.13

in AE* and V\ for a shift in X* of dy are given by

JA E * = § 2 ^ i + £ i d § 2 , (10.147)
JVj = fa2dy. (10.148)
Using simple geometric relations between sides of similar triangles, it can be argued that

d\~x = - dy, d& = — dy. (10.149)

r\ r2
Combining Eqs. (10.147)-(10.149) yields

^ =i + i. U0..50,
dVi rx r2
which can be combined with Eq. (10.146) to obtain

\ (10.151)

Equation (10.151) is usually called the Young-Laplace equation. It relates the pressure
difference across the interface to the interfacial tension and the geometry of the interface
at equilibrium. We have shown that the interfacial tension is related to the surface excess
free energy as indicated by Eq. (10.140). This links the interfacial tension to the molecular
properties that dictate the variation of the molecular spacing (density) and other properties
across the interfacial region. The analysis in this section has considered two fluid phases
of a pure substance. This type of analysis can also be extended to interfaces in multicom-
ponent systems with slight modifications. Further discussion of interfacial phenomena in
multicomponent systems can be found in References [4] and [5].

10.4 Molecular Transport at Interfaces

The mass and energy transfer during vaporization and condensation processes are
important in a variety of technological applications. To fully understand the characteristics
of such processes, we must consider the liquid-vapor interface at the molecular level. To do
so, we will use some of the results from kinetic theory developed in Section 10.1 to interpret
the motion of vapor molecules near a liquid-vapor interface. We consider a plane surface
in the vapor phase but immediately adjacent to the interface as shown in Figure 10.14.
352 10 I Nonequilibrium and Noncontinuum Elements of Microscale Systems

I—^y vapor at Tv, Pv

z • •
Jl Jv
surface 2* just
inside vapor —7 k I

Figure 10.14

Even if no net vaporization or condensation occurs at the interface, a dynamic equilibrium

is established in which the molecules from the vapor phase that enter the interfacial region
and become part of the liquid phase are balanced, on the average, by an equal number of
molecules that escape the liquid into the vapor region. When (net) condensation occurs, the
flux of vapor molecules joining the liquid must exceed the flux of liquid molecules escaping
into the vapor phase. When (net) vaporization occurs, the opposite must be true.
At least initially, we admit the possibility that the respective values of pressure and
temperature in the two bulk phases shown in Figure 10.14 may be different. For the system
shown in this figure, the net number flux of molecules across surface £* immediately
adjacent to the interface must be equal to the difference between the number fluxes j \ and
7v passing through surface XI * in opposite directions just inside the vapor region:

;'net = 7 v - 7 l . (10.152)

To facilitate the evaluation of jv and 71 in Eq. (10.152), we will assume that here the flux of
molecules j \ is characterized by T\ and P\ whereas jw is characterized by TN and Pv. With
this assumption, it would appear that jw and 71 can be obtained by using Eq. (10.32), which,
when combined with the ideal gas relation, can be written in the form

where M is the molecular weight of the substance. However, use of Eq. (10.153) to evaluate
jv and j \ is inadequate in two respects. First, Eq. (10.153) predicts the flux of molecules
in a stationary gas, whereas in the system in Figure 10.14, the vapor must have a net bulk
velocity w = w$ in the z direction as a result of the phase change at the interface. The
bulk velocity is toward the interface (u>0 > 0) for condensation and away from the interface
(tuo < 0) for vaporization.
By extending the analysis presented in Section 10.1, it can be shown from kinetic theory
(see Schrage [6]) that when the gas moves normal to a planar surface with a speed VOQ, the
flux of molecules through the plane in the direction of bulk motion is

/ M \ *^ P
yW = r a - (10.154)
\27tNAkBT) m
and the flux of molecules in the direction opposite to that of the bulk motion is
U- = r-a ( J* ) -, (10.155)
10.4 I Molecular Transport at Interfaces 353

Figure 10.15

where
w0 (10.156)
a = (2N k T/M)l/2'
A B
The factors Ta and F_ a , which correct for the effects of bulk gas motion, are given by the
following relations:
Ta = exp(a2) + anxl\\ + erf(a)] (for a > 0), (10.157)
2 1/2
F_fl = exp(a ) - an [1 - erf (a)] (for a < 0). (10.158)
The variations of Ta and F_fl with a are shown in Figure 10.15.
Because the vapor in Figure 10.14 is expected to be in bulk motion relative to the surface
Ef, it is more appropriate to compute jY using Eq. (10.154) or (10.155). As suggested by
Schrage [6], we assume that the surface Ef is an infinitesimally small distance from the
interface so that there is no bulk motion effect on molecules emerging from the liquid and
passing through the surface E*. The net flux of molecules from the liquid j \ is therefore
calculated using Eq. (10.153).
The second inadequacy in using Eq. (10.153) alone to compute fluxes at the interface is
that doing so implicitly assumes that all molecules crossing the surface E* in the negative
z direction are actually from the liquid phase. Actually only a fraction a e is actually due
to vaporization from the liquid. The remaining fraction 1 — ae is due to "reflection" of
vapor molecules that strike the interface but do not condense. Note that in this context
"vaporization" is used to describe molecules escaping the liquid phase and "condensation"
is used to describe molecules from the vapor phase being absorbed into the liquid phase. The
fraction of molecules crossing the surface E* in the positive z direction that condense and
are not reflected is designated as ac. If no phase change occurs at the interface, equilibrium
requires that
= Cfc = O (at equilibrium).
Usually <7e and &c are assumed to be equal even for the dynamic case when (net) phase change
occurs at the interface, although the validity of this assumption is suspect. As a result, a
is variously referred to as a vaporization coefficient, evaporation coefficient, condensation
354 10 I Nonequilibrium and Noncontinuum Elements of Microscale Systems

coefficient, or accommodation coefficient. The term accommodation coefficient has been

perhaps most widely used for & in recent years and will be adopted here.
Allowing for the bulk motion of the vapor and the role of the accommodation coefficient
as described above, the portion of the incident molecular flux that actually enters the liquid
phase upon striking the liquid, jYC, is given by

{ & jw+
J +
f
for condensation,

crjw- for evaporation,

. (10.159)

and the portion of the molecular flux crossing surface S* that is due to molecules that
actually emerged from the liquid phase, j \ e , is given by
j]s = &j. (10.160)
The net molecular flux to or from the interface as a result of the phase change j \ is just equal
to the difference between jvc and j\e:

ji = M-h- (10.161)
Substituting Eqs. (10.159) and (10.160) into (10.161), evaluating j w + or jw_ at PY and Tv
and j at P\ and 7], and rearranging yields

a ( M \l/2\ TPv
m\2nNAkBJ [T1/2 T' J (10-162)

Macroscopically, the heat flux to the interface JQ\ must equal the net mole flux jxm/M
multiplied by the latent heat of vaporization A/*iv. Using Eq. (10.162) to evaluate j[, we can
write the resulting relation for the heat flux as

For constant ambient pressure, surrounding vapor must move to replace condensing vapor
or it must move away to make room for vapor generated by evaporation. This implies that
the bulk velocity of the vapor is related to the net molecular flux at the interface as

wo = —- = — — , (10.164)

which implies that

2
yy" -
Since JQ^ — j\mAh\v/ M, it also follows that

mAhlvPy \2NAkBTy
Because F is a function of a, Eq. (10.162) is actually an implicit relation for j u and
Eq. (10.163) is actually an implicit relation for heat flux. If the liquid and vapor phases are
in equilibrium (and the interface is planar), the pressure in each phase is expected to equal
10.4 I Molecular Transport at Interfaces 355

the saturation pressure at the temperature in that phase. If we assume that this is also true
when net condensation or vaporization occurs, Eqs. (10.162) and (10.163) can be written

With this assumption we have linked the molecular flux and the heat flux at the interface to
the temperatures in the bulk phases Tv and T\.
For vaporization and condensation processes at high temperatures, a is often small.
For example, condensing or evaporating saturated pure water at 100°C at a heat flux of
100 kW/m2 corresponds to an a value of 1.3 x 10~4. In contrast, the values of a for va-
porization or condensation processes at cryogenic temperatures (below 100 K) may be
much higher. In the limit of small a, Ta and T_a in Eqs. (10.157) and (10.158) are well
approximated by
r = l+tf7r 1 / 2 . (10.169)
Substituting this relation into Eqs. (10.167) and (10.168) and using Eqs. (10.165) and
(10.166), with a little manipulation, we obtain the following relations for the molecular flux
and heat flux:
*•*•*• i i •* N a L V - * - v / •» Na.LV-*-i / i /I A infW

(10.171)

An alternate form of the heat flux relation can be obtained if the relations
APsat,vl = PsatWv) ~ P^TO, (10.172)
ATvl = TY-T{ (10.173)
are substituted into Eq. (10.171) to eliminate P\ and T\\
Id 1 Ah\y
(2-<r)J (2TT MNAkB)l/2 [ TYl/2
(10.174)
We next expand the second term in the square brackets in terms of ATV\/Ty using the
assumption that APV\/PY <^ 1 and ATV\/Ty <^ 1 so that terms of order (APy\/Py)2 may
be neglected. This leads to the relation
2a

Using the Clapeyron equation (8.81) to evaluate AF satv i/Ar v i as

;at TyAOiv'
356 10 I Nonequilibrium and Noncontinuum Elements of Microscale Systems

we can further simplify Eq. (10.175) to

2a 1 (Ahlv)2 I" _ Aat(r v )Ai) lv

J
(2-<r)J (2jtMNAkBTyy/2TvAvlw [[ 2Ahlw
(10.176)

Equation (10.176) is an explicit relation between the heatfluxassociated with a condensation

or vaporization process and the temperature difference across the interface that is required
to drive the process. Note that this result applies for small a.
The above analysis of interfacial transport assumes that the fluxes of condensing and
vaporizing molecules can be derived from kinetic theory for each flux separately and the
results superimposed to obtain the net heat flux. The analysis does not consider nonequi-
librium interactions between the molecules leaving the interface and those approaching the
interface, which may have different mean energy levels because of the difference between
the vapor and liquid temperatures. Inclusion of these effects makes analysis of the system
considerably more difficult (see Wilhelm [7]).
Despite its approximate nature, the results of the above analysis provide a useful frame-
work for assessing the effects of interfacial transport on vaporization and condensation
processes. It should be noted that the equation relating JQ-U and A7vl derived above applies
equally well to vaporization and condensation with the convention that JQ\ is positive for
condensation and negative for vaporization. In engineering analysis of vaporization and
condensation processes, it is usually assumed that the vapor and liquid temperatures are
both equal to the saturation temperature at the local pressure. The above results indicate
that there is some deviation from this assumption of local thermodynamic equilibrium.
Fortunately for most cases of practical interest, this deviation is very small.
Inspection of the relations derived above clearly indicates that the relation between
heat flux and driving temperature difference depends directly on the value of the accom-
modation coefficient a. Quoted values of a in the literature vary widely. The tabulation
assembled by Paul [8] lists values ranging from 0.02 to 0.04 for some liquids to values
very near 1.0 for others. Mills [9] has suggested that molecular accommodation should be
imperfect only when the interface is impure. However, recent molecular dynamic simu-
lation studies by Yasuoka et al. [10] imply that the accommodation coefficient for a pure
water interface should be near 0.4. Paul [8] noted that in almost every case where a was
found to be appreciably less than one, the average condition of the vapor molecules was
different from that for the liquid phase because of association, disassociation, or poly-
merization. This suggests that if a fluid is virtually pure and if association, disassociation,
and polymerization do not occur to any significant degree, the accommodation coeffi-
cient should be close to unity. Because extreme purity is unlikely in most engineering
systems, a value of a less than one is expected in common applications. Values reported
in the literature should therefore be regarded as typical values of o rather than constants
applicable to all systems since system contamination may alter values for a particular
system.
Equation (10.176) indicates that a finite temperature difference is required to drive heat
flow across an interface in the presence of vaporization or condensation. This is com-
monly interpreted as an additional resistance to heat flow. In many cases, this resistance
is low and may justifiably be neglected in engineering analysis of such systems. However,
Eq. (10.176) indicates that this resistance may be large at high heat flux levels and/or at
10.5 I Phase Equilibria in Microscale Multiphase Systems 357

low accommodation coefficient values. The importance of this resistance can be estimated
from Eq. (10.176) with estimates of the system accommodation coefficient.

Example 10.4 In the latter stages of a vaporization process in an evaporator, a thin liquid
film of water flows along the walls of the evaporator tube with mostly vapor flowing in the
central core of the tube. The pressure in the tube at the location of interest is 100 kPa. A
heat flux of 200 kW/m2 is applied to outer surface of the tube and is conducted through
the tube wall and the liquid film to the interface. For vaporization at this heat flux level,
estimate the temperature difference across the interface if the accommodation coefficient is
taken to be 0.5.

Solution For water at 100 kPa, Ah\y = 40.6 MJ/kmol, Av\Y = 30.1 m3/kmol, and
we take F sat (r v ) = 100 kPa and Tv = r sat (100 kPa) = 373 K. Solving Eq. (10.176) for A7 vl
and substituting for the properties, we obtain

= T \2-&] 2{2itMNAkBTy)1'2'.
J c>
2<T J (AAiv)[2AAiv - /> sat (r v )At) lv ]
"2-0.5"
= 200,000
2(0.5)
1026)(1.38 X 10~23)373)1/2 373(30.1)
2(2TT(18)(6.02 X
X
(40.6 x 10 )[2(40.6 x 106) - (100,000)30.1]
6

= 0.040 K.

10.5 Phase Equilibria in Microscale Multiphase Systems

Small Bubbles in Pure Liquid

In Chapter 6, we discussed the fact that liquid may be superheated beyond its equi-
librium saturation temperature at a specified pressure. We further noted that with increasing
superheat, system density and energy fluctuations increase in magnitude. In superheated
liquid, density fluctuations in the liquid may result in localized regions where the molecular
density has been lowered to nearly that for saturated vapor. In real systems, extreme fluctu-
ations of this type may result in the formation of an embryo vapor bubble within the bulk
fluid. This process is known as homogeneous nucleation. Experimentally, homogeneous
nucleation is observed to occur in superheated liquid, forming a very small bubble in the
liquid. This type of nucleation process is depicted in Figures 10.16a & b. In the first portion
of this section we will explore the necessary conditions for the bubble to be in equilibrium
with the surrounding liquid and we will assess the stability of such an equilibrium.
To examine the stable equilibrium conditions for a spherical bubble, we consider the
system shown in Figure 10.16b. Note that the liquid pressure is held constant in this system.
At equilibrium, the temperature in the vapor and the liquid must be the same, and the
chemical potential in the two phases must be equal,
Al = Ave. (10.177)
358 10 I Nonequilibrium and None ontinuum Elements of Mieroseale Systems

initial state after bubble formation

Because of the curvature of the interface, however, the pressures in the vapor and in the
liquid are not equal. For a spherical bubble or radius r e, the Young-Laplace Eq. (10.151)
requires that
(10.178)

Integrating the Gibbs-Duhem equation for a pure substance,

djl = -sdT + vdP, (3.50)
at constant temperature from P = Psat(T\) to an arbitrary pressure P, we obtain

\i - Msat = / vdP. (10.179)

We next use the ideal gas law (0 = RTV/P) with Tv = T\ to evaluate the integral on the right
side of Eq. (10.179) for the vapor phase. This yields the following relation for the chemical
potential of the vapor at its equilibrium pressure Pve:
Pve \
Ave = (10.180)

For the liquid phase, we use Eq. (10.179) again, but this time, because the liquid is
virtually incompressible, 0 is taken to be constant and equal to the value for saturated liquid
at 7i, 0 = Oi. Evaluation of the integral in Eq. (10.179) for P = P\ then yields
Al = Asat,l + V\[P\ ~ (10.181)
Note that for the "normal" flat interface saturation condition, equilibrium requires that
Asat,v = Asat,i- Using this fact while combining Eqs. (10.179), (10.180), and (10.181) we
obtain, after a little rearranging, the following relation for the vapor pressure inside the
bubble at equilibrium:

/"ve = (10.182)

For superheated liquid, P\ must be less than PSdx(T\), as is evident for the isotherm shown
in Figure 8.4. Consequently, the exponential term in Eq. (10.182) is less than one and P ye
10.5 I Phase Equilibria in Micro scale Multiphase Systems 359

is less than Psat(7i). Combining Eqs. (10.178) and (10.182) to eliminate P ve , we obtain the
following relation for r e :

2a;
(10.183)
)
/>sat(7])exp{i> 1[/ i-JPsa - P{

Thus only bubbles having a radius equal to that given by the above relation will be in
equilibrium with the surrounding superheated liquid at T\ and P\.
Alternatively, if Eqs. (10.178) and (10.182) are combined to eliminate P\, the relation
for Pve becomes

(10.184)

In most cases, PYQ — Psat(T\) is small compared to 2cf[/re. Hence, Pve is well approximated
for most systems by

(10.185)

Because the exponential term in the above equation is less than or equal to one, Eq. (10.185)
implies, as did Eq. (10.182), that PyQ < Ps&t(T\). The situation regarding the state points of
vapor and liquid can be seen more clearly by considering Figure 10.17. If the liquid state
point is on the superheated liquid curve at point 1, the vapor state point corresponding to an
equal value of chemical potential /xve = jl\ must lie on the superheated vapor curve at point
2. Hence, for a bubble of finite radius, equilibrium can be achieved only if the liquid is super-
heated and the vapor is superheated relative to the normal saturation state for a flat interface.
Note also that the steep slope of the superheated vapor line results in values of PWQ that are
much closer to Psat(7i) than to P\. Because F ve — P\ = 2<7i/re, this provides justification
for the assumption that Psat(T\) — Pve <£ 2o\/rQ, which was used to obtain Eq. (10.185).

Isotherm for T=T,

AB - subcooled liquid
BC - superheated liquid
EF - supercooled vapor
FG - superheated vapor

Figure 10.17
360 10 I Nonequilibrium and Noncontinuum Elements of Microscale Systems

Example 10.5 Determine the size of a steam bubble in equilibrium with superheated
liquid water at atmospheric pressure and 200°C. Also determine the vapor pressure inside
the bubble.

Solution For water at 200°C, Fsat(7i) = 1,554 kPa, vx = 0.02083 m3/kmol, and
ax = 0.0304 N/m. The size of the bubble at equilibrium is given by Eq. (10.183),
2cr{
r - Pi
2(0.0304)
l,554,000exp{.02083[101,000 - l,554,000]/8308(200 + 273)} - 101,000
= 4.22 x 10~8 m.

The pressure in the vapor is given by Eq. (10.182),

RT\ J
.02083[101,000 - 1,554,000] 1 _
8308(200 + 273) J ~

The pressure in this case is only slightly different from the normal saturation pressure at
200°C.

We now wish to examine the stability of a bubble that satisfies the equilibrium require-
ment discussed above. We begin by considering the Helmholtz free energy of the system
after a bubble of radius r has formed. By definition, F = U — TS. Since U and S are additive
over subsystems in a composite system, and T is taken to be the same in all subsystems, it
follows that F is additive over the subsystems. The Helmholtz free energy for the system
as a whole is equal to the sum of contributions associated with the liquid, the vapor, and the
interface:

F = Fi + F v + Fi, (10.186)

where

fl = #i(fii - 7i$i), (10.187)

F v = iVv(Kv - TiSy), (10.188)
2
Ft =47tr a{. (10.189)

In the above relations, N\ and Nv are the number of moles of liquid and vapor in the system,
respectively. Combining Eqs. (10.186)—(10.189), we obtain

F = N\(iii - T\si) + NY(uy - Tisy) + Anr2^. (10.190)

10.5 I Phase Equilibria in Microscale Multiphase Systems 361

By definition, the Gibbs free energy is given by

G = U -TS + PV. (10.191)
Since F = U — TS,we can write
G = F + PV. (10.192)
Since the volume of the system V must equal the sum of the vapor and liquid volumes,
V = Nivi + Nyvv,
and we can combine Eqs. (10.190) and (10.191) to obtain
G = Ni(u\ - Ti§i) + NY(uv - T\sy) + Anr2a{ + P\(N\V\ + Nwvy). (10.193)
By definition, the specific Gibbs free energies for the liquid and vapor are given by
gl = ui-Tlsi + Pi0u (10.194)
gy = uv - Tifv + Pvvv. (10.195)
Reorganizing Eq. (10.193) and using Eqs. (10.194) and (10.195) we can write the relation
for the system Gibbs free energy in the form
G = Nvgv + N{gx + Anr2a{ - Nyvv(Py - PY). (10.196)
The Gibbs free energy of the system before the bubble forms is given by
G0 = (Ny + Nl)gl. (10.197)
The change in Gibbs free energy associated with the bubble formation is therefore
AG = G - G o . (10.198)
Combining Eqs. (10.196)—(10.198), we can write the relation for AG in the form
AG = Nv(gy - gx) + Anr2a{ - Nvvy(Pw - Pi). (10.199)
For a pure fluid, we demonstrated in Chapter 2 that jX = g. Thus, for a bubble at equilibrium
with the surrounding liquid, /2V = /2 b which, for a pure system, implies that gY = gx.
In addition, at equilibrium, the Young-Laplace equation for the spherical bubble requires
that

The product Nvvy is the total volume of the bubble, which for a spherical bubble is related
to the radius as
Nwvy = (4/3)jrr 3 . (10.200)
It follows that at equilibrium, Eq. (10.199) reduces to
AG e = (4/3);rre2ai, (10.201)
where the radius of the bubble in equilibrium with the surrounding liquid is denoted as r e .
Substituting the right side of Eq. (10.200) for Nwvy allows us to write Eq. (10.199) as
AG = Ny(gy - gx) + 47tr2a{ - (4/3)7rr3(PV - Px). (10.202)
362 10 I Nonequilibrium and Noncontinuum Elements of Microscale Systems

Because T\ and P\ are fixed, gx is a function of T\ and P\, and | v is a function of T\ and
Py, Eq. (10.202) implies that AG is a function of Pw, flv, and r. Assuming that the vapor
obeys the ideal gas law (i)v = RT\/PV) and the bubble is in mechanical equilibrium and
satisfies the Young-Laplace equation (Pv — P\ = 2<7i/r) regardless of size, the dependency
of AG on Py and vv can be removed, with the result that AG is only a function of r. With
these idealizations, we can explore the stability of the vapor bubble with a radius near the
equilibrium radius r e by expanding AG in a Taylor series about r = r&:
f dAG\ 1 f d2AG\ 9
AG = AG e + ( ) (r-rc) + -l — j (r - re)2 + • •
•.
(10.203)
Using the fact that the Gibbs-Duhem equation implies that

= Vy (10.204)

together with the ideal gas law and the Young-Laplace equation, it can be shown directly
by differentiating Eq. (10.202) that

=0, (10.205)

Using these results to evaluate the derivatives in the expansion (10.203) for AG together
with Eq. (10.201), we obtain

A
-naxrl - ^ - 12 + - I (r - r e ) 2 + • •
•. (10.207)

Using Eq. (10.201) we can rearrange this equation to the form

x 2

+ ••
•. (10.208)
AG e (4/3)7^ L 1 + 2aire/PlJ \re
Note that the term in square brackets in the expansion is approximately equal to 2 for most
systems because usually 2axrtIP\ <$C 1.
The expansion (10.207) for AG indicates that AG has a local maximum at r = r e , as
indicated in Figure 10.18. It can also be argued that in the limit r -> oo, all the metastable
superheated liquid would undergo transition to superheated vapor. Note in Figure 10.17 that
a transition from superheated liquid to superheated vapor results in a decrease in /2, which
implies a decrease in g and in the system Gibbs free energy. Thus AG must be negative at
large r. Because we also know that AG must approach zero as r -> 0, the overall variation of
AG with r must look like the solid-line representation of Eq. (10.208) plus the broken-line
extensions shown in Figure 10.18.
The variation of AG with r shown in Figure 10.18 has several noteworthy features. We
noted in Chapter 3 that for a system held at constant temperature and pressure, the Gibbs
free energy must be at a minimum for a stable equilibrium. It follows that AG must be
at a minimum for stable equilibrium in the system considered here, and it is clear from
10.5 I Phase Equilibria in Microscale Multiphase Systems 363

Figure 10.18

Figure 10.18 that a bubble of radius r =re is in an unstable equilibrium. We know from
our analysis in Section 10.4 that at the interface there is a continual exchange of molecules
between the vapor and liquid phases. If a bubble is in equilibrium with r = r e, the loss of one
molecule from the bubble decreases its radius into the range 0 < r < r e where dAG/dr
is positive. Further reduction of r is expected in this range because spontaneous processes
that decrease the Gibbs free energy are most probable. The end result is that once entering
this range of r, the bubble is likely to collapse completely. If a bubble of radius r = re gains
one molecule and enters the range rQ < r, decreases as r increases and the bubble is most
likely to spontaneously grow.
The above observations provide insight into the nature of homogeneous nucleation of
vapor in a superheated liquid. If density fluctuations in a metastable superheated liquid
produce an embryo bubble of radius r <re, the bubble is likely to collapse. Conversely,
if the embryo bubble has a radius greater than r e, it is expected to spontaneously grow,
resulting in homogeneous nucleation of the vapor phase in the system.
In the above analysis, the assumption of mechanical equilibrium embodied in the use of
the Young-Laplace equation is reasonable if the bubble radius readjusts rapidly to mechan-
ical force variations. However, if the mechanical forces change too rapidly for the liquid to
readjust, then mechanical equilibrium may not be maintained. For such conditions, a better
assumption would be that the values of chemical potential in the two phases are equal. Use
of this idealization, along with the others noted above, results in an expansion for AG that
is identical to Eq. (10.207) except that the term in square brackets is simply equal to 3.
Zeldovich [11] and Kagen [12] have shown that this modified form of the expansion for
AG is more appropriate for cavitation processes in which a superheated liquid condition
is created by a sudden (isothermal) decrease in pressure. Regardless of whether chemical
or mechanical equilibrium is assumed, the stability characteristics of the vapor bubble are
qualitatively the same and quantitatively only slightly different.
The thermodynamic analysis presented in this section forms the foundation for analysis
of the kinetics of bubble embryo formation in superheated liquids. The interested reader
can find more information on analysis of the kinetic limit of superheat in the references by
Carey [13] and Debenedetti [14].

Small Droplets in Pure Vapor

The vapor phase of a pure substance may be brought to a supersaturated state
by either cooling it by transferring heat through the walls of a containing structure or by
364 10 I Nonequilibrium and Noneontinuum Elements of Microscale Systems

system held at
constant Tv and Pv

Figure 10.19

rapidly dropping the pressure and temperature of the gas in an adiabatic process. For either
type of process, once a saturated state is reached, further reduction in the system pressure
and temperature or further removal of heat may initiate condensation of some of the vapor
to liquid if nuclei of the liquid phase are present in the system. Often there are adsorbed
molecules in a near-liquid state on the containment walls or on dust particles suspended in
the vapor that serve as nuclei. However, in the absence of these types of nuclei, the processes
mentioned above may reduce the pressure or cool the vapor to a state point in the metastable
vapor region of Figure 8.4.
Heterophase fluctuations of density in a metastable supersaturated vapor may result
in the formation of an embryo droplet that may initiate the condensation process. Such
a homogeneous nucleation process will, as in the case of vaporization, depend on the
thermodynamic stability of droplet embryos in the vapor. To explore these matters further,
we will specifically consider the system shown in Figure 10.19 containing a liquid droplet of
radius r in equilibrium with a surrounding vapor held atfixedtemperature Tv and pressure F v .
At equilibrium, the chemical potential in the vapor and the droplet must be equal:

/xv = (10.209)
The temperature in both phases must be the same and the pressures in the vapor and liquid
must be related through the Young-Laplace equation,

(10.210)

Using the constant-temperature integrated form of the Gibbs-Duhem equation derived

above,

vdP, (10.179)
JPsA
we evaluate the integral on the right side using the ideal gas law (v = RT V/P) for the vapor
to obtain

Ave = (10.211)
I ^satUl) J
For the liquid phase inside the droplet, the chemical potential can again be evaluated
using Eq. (10.179). The liquid is taken to be incompressible, with v equal to the value for
10.5 I Phase Equilibria in Micro scale Multiphase Systems 365

saturated liquid at r v , which we designate as v\. With this assumption, evaluation of the
integral in Eq. (10.179) for P = F le yields

Ale = Asat,l + ~ fis (10.212)

Equating the values of /2ve and /2le given by Eqs. (10.211) and (10.212) to satisfy Eq. (10.209),
and using the fact that /xsat>v = Asat,i> w e obtain the following relation:

Mfie ~ (10.213)
RTy

Comparison of the above relation with Eq. (10.182) reveals that the relation (10.213) for the
equilibrium vapor pressure surrounding a small liquid droplet is similar in form to that for
the pressure inside a small vapor bubble in equilibrium with a surrounding liquid pool. As
seen in Figure 10.20, if the vapor state point is on the metastable supercooled vapor curve
at point 1, the liquid state corresponding to equal /2 must lie on the subcooled liquid line at
point 2. For a liquid droplet with finite radius, equilibrium can therefore be achieved only if
the liquid is subcooled and the vapor is supersaturated relative to the normal saturation state
for a flat interface. Equation (10.213) also indicates that if P v >JPsat(77v), then P\e must also
be greater than PsaX(Tw), which is consistent with the state points indicated in Figure 10.20.
If we substitute Eq. (10.210) to eliminate Fie, Eq. (10.213) becomes

= J P sa t(7;)exp< (10.214)
RTW
In most instances, the steep slope of the supercooled vapor line in Figure 10.20 requires that
the value of Py be much closer to PSax(Tv) than to P\Q. When this is the case, PY — Psat(Tv) is
small compared to 2ai/r e = P\e — Pw, and the relation (10.214) for PY is well approximated
by
2fliQj \
(10.215)
~r,RTy J '

Isotherm for T=TV

E
A
/
A
Vie 2
B,F

AB - subcooled liquid
BC - superheated liquid
EF - supercooled vapor
FG - superheated vapor

G,

• - — P le -1
D
P

Figure 10.20
366 10 I Nonequilibrium and Noncontinuum Elements of Microscale Systems

Equations (10.214) and (10.215) can also be inverted to solve for the equilibrium droplet
radius for a given set of vapor pressure and temperature conditions:

r (ia216)
/vx) in{/y/> sat (r v )} - pv + p s a t(r)'
(r v )'
for P
^ l v - ^atOV) « 2ai/r e . (10.217)
RTv\n{Pv/Psai(Tw)}
The stability of an embryo droplet that forms in a system initially containing supercooled
vapor can be analyzed in a manner almost identical to that used above to analyze the stability
of a bubble in superheated liquid. Such an analysis again considers the change in the Gibbs
free energy AG associated with the formation of a droplet in a system held at constant
temperature Tw and constant pressure Pw, as indicated in Figure 10.19. Proceeding in the
manner described earlier in this section for a bubble in superheated liquid, it can be shown
that the change in the Gibbs free energy of the system for formation of a droplet with just
the right radius r e to be in equilibrium with the surrounding vapor is

AGe = \noxrl. (10.218)

The above result is in fact, identical to the corresponding result obtained in the analysis
of the bubble in superheated liquid. Following the same line of analysis used above in
analysis of the bubble, it can be shown that for the droplet formation process depicted in
Figure 10.19, the leading terms in a Taylor series expansion for AG about r = r e are

AG = \no{rl - AitG{{r - r e ) 2 + • •
•. (10.219)

These results imply that the variation of AG with r is qualitatively similar to that shown
in Figure 10.18. The expansion indicates that AG has a local maximum at r — r e . In the
limit r —>> oo, all the metastable vapor would undergo a transition to subcooled liquid. Note
in Figure 10.20 that a transition from supercooled vapor to subcooled liquid results in a
decrease in /z, which implies a decrease in g and in the system Gibbs free energy. Thus,
AG must be negative at large r. Because AG must approach zero as r —>• 0, the overall
variation of AG with r must again look like the curve indicated in Figure 10.18.
The interpretation of the variation of AG with r for the embryo droplet is also similar
to that for the embryo bubble considered above. At constant temperature and pressure,
the Gibbs free energy must be at a minimum for a stable equilibrium. It follows from the
expansion (10.219) that a droplet of radius r = r e is in an unstable equilibrium. If a droplet
is in equilibrium with r = r e, the loss of one molecule from the bubble decreases its radius
into the range 0 < r < r e where dAG/dr is positive, and further reduction of r is favored
because of the associated decrease in the Gibbs free energy. Once entering this range of r,
the droplet is likely to evaporate completely.
If a droplet of radius r = r e gains one molecule and enters the range r e < r, AG decreases
as r increases, which favors further growth of the droplet. Hence, if density fluctuations in
a metastable vapor result in formation of an embryo droplet of radius r <r e, the droplet is
likely to lose molecules and disappear. If its radius is greater than r e, it will grow sponta-
neously, initiating the condensation process in the vapor via homogeneous nucleation.
The results of this type of thermodynamic analysis form the foundation for modeling
of the kinetics of droplet embryo formation in supersaturated vapor. The interested reader
70.5 / Phase Equilibria in Microscale Multiphase Systems 367

pressure control system

/ solid I
d

Figure 10.21

can find more information on analysis of the kinetic limit of supersaturation for vapors in
the reference by Carey [13].

Ultrathin Liquid Films

We now will turn our attention to very thin liquid films on solid surfaces. If a solid
surface is initially in contact with a gas or vapor phase and a small quantity of liquid is
brought into contact with the surface, the liquid may bead up into a droplet if it poorly
wets the surface, or it may spread over the surface to form a thin liquid film. The latter
behavior generally results when there is a strong attractive force interaction between the
molecules in the liquid and the molecules of the solid near the interface. To analyze more
fully the thin film that occurs for such circumstances, we consider the system shown in Fig-
ure 10.21 in which liquid is in contact with and fully wets the horizontal upward-facing solid
surface.
A housing with a trapped volume of gas inside is brought into close proximity to the
solid surface so that a thin film of liquid of thickness 8 exists between the vapor-liquid
and liquid-solid interfaces. The liquid does not wet the inside of the housing and a tube
connects the vapor space inside the housing to a system that holds the pressure constant.
If the pressure in the gas inside the housing is Pg and the liquid-gas interface is flat, then
at equilibrium, Pg must balance the liquid pressure across the interface in the liquid. If 8 is
large, the pressure across the interface in the liquid film P { will just equal the local ambient
pressure in the liquid P\ given by
gQzhM
— r ra (10.220)

However, if the housing is brought very close to the solid surface, the pressure inside the
housing must not only balance the local liquid pressure P\, it must also counteract the
attractive forces between the liquid molecules and the solid surface, which otherwise would
draw liquid into the film to establish a thicker film of liquid on the surface. When the film
is very thin, these attractive forces act to pull liquid into the layer as if the pressure in the
layer were reduced below the ambient pressure P\ by an amount Pd, which is known as the
368 10 I Nonequilibrium and Noneontinuum Elements of Microscale Systems

500-

400-
: )
v
(kPa)
300- \
\
\
200-

100-

0-
^ _ "—- ~ I I I

0.5 1 10 100 1000 5000

8 (nm)

Figure 10.22 Variation of disjoining pressure with film thickness for carbon tetrachloride on
glass at 77°C (from Ref. [15]).

disjoining pressure. By convention, if the affinity of the liquid for the solid draws liquid
into the film, Pd is taken to be negative.
For the circumstances shown in Figure 10.21, the local pressure P\ and the disjoining
pressure effect (due to solid-liquid attraction) act in tandem to thicken the film.To maintain
a thin liquid film, the gas pressure force in the housing must balance both effects, which
implies that at equilibrium
Pg = pY + \pd\ = P{ - pd9 (10.221)
where P\ is given by Eq. (10.220). Note that in Eq. (10.221) because Pd is negative for a liquid
that is attracted to the solid surface, — Pd is positive for this system. Because the attractive
forces between liquid molecules and those of the solid are expected to be stronger for liquid
molecules that are closer to the interface, the disjoining pressure difference — Pd required
to counteract them is expected to increase as 8 gets smaller. This is reflected in the variation
of disjoining pressure with 8 indicated in Figure 10.22 for carbon tetrachloride on glass.
If we consider the process of slowly bringing the housing progressively closer to the
solid wall while adjusting Pg to maintain equilibrium, the pressure required to thin the layer
becomes continually larger until finally the last monolayer of liquid molecules is removed
from the surface. The molecules of gas are then in contact with those of the solid to within
an interfacial separation <50, which is less than the thickness of the liquid monolayer just
removed. The work done per unit area of the solid surface, Wu is given by

= /° -Pd(S)d8. (10.222)
•J do

Note that in removing the liquidfilmon a unit area of solid surface we have created a unit area
of solid-gas interface and destroyed a liquid-solid interface and a liquid-gas interface. In
Section 10.3, we identified the excess free energy associated with a gas-liquid interface as

(10.141)
dT
We can extend this concept to define the surface excess free energy associated with a unit
area of the solid-liquid and solid-gas interfaces:

(10.223)
10.5 I Phase Equilibria in Microscale Multiphase Systems 369

(10.224)
T,Nes

It follows that if our process of removing the liquid film from the surface is reversible,
the work done must equal the free energy gain associated with creation of the solid-vapor
interface minus the free energy losses associated with destruction of the liquid-vapor and
liquid-solid interfaces. We can state this mathematically as

= a s g - Ogi - as\. (10.225)

This provides a theoretical link between interfacial tensions and the work required to re-
move a liquid film from a surface, which we also have related to the concept of disjoining
pressure. The right side of Eq. (10.225) is also referred to as the spreading coefficient for
the liquid since it equals the net free energy gain associated with the spreading of liquid
over a solid surface. Consistent with the second law, such spreading will be spontaneous
(in a system at constant T,N, and V) if the net free energy change is positive. This type of
spontaneous spreading is exhibited by some hydrocarbon liquids on metal surfaces.
For a thin liquid film on a solid surface, like that shown in Figure 10.23, the affinity of
the liquid for the solid may affect the equilibrium vapor pressure near the interface. More
information about disjoining pressure effects in such circumstances can be obtained by
examining the features of a liquid film on a solid surface at a molecular level. We consider
first the interaction between a single fluid molecule and the molecules of the solid. This
circumstance is shown schematically in Figure 10.24.

vapor
gas-liquid interface

liquid film

Figure 10.23

Figure 10.24
370 10 I Nonequilibrium and Noneontinuum Elements of Microscale Systems

We assume here that the separation distance between the fluid molecule and the solid
molecules is sufficiently large that only the long-range attractive portion of their pair po-
tential is important. We therefore take the form of the potential to be

0(r) = - — . (10.226)
rn

Note that if we take this to be equivalent to the outer portion of the Lennard-Jones 6-12
potential,

0(r) = (10.227)
r J Vr / J
then
Cj, = n = 6. (10.228)
We further assume that the force interactions between the single fluid molecule and
the molecules in the solid are additive. It follows that the potential energy of the system
associated with attractive forces is the sum of the energy increase for each solid molecule
as it is brought from infinitely far away (z -» oo) into its position in the solid such that the
fluid molecule is a distance z = D from the surface. Consistent with our sign convention
(work is positive if done by the system) this must equal the negative of the work done by
the system to bring the solid molecules from infinitely far away into place in the solid wall.
The work done by the system can be computed by summing the contributions for each
molecule in the solid. The number of molecules in a differential ring element of the solid
in Figure 10.25 is Inp^^x dx dz. Each of these is at a distance r = \lx2 + z2 from the
fluid molecule. Weighting the corresponding interaction potential (10.226) by the number
of solid molecules in the differential ring volume and integrating, we obtain the following
relation for the work done by the system to bring the solid molecule from infinitely far
away from the surface to its place in the solid with its surface a distance D from the fluid
molecule:

rZ=OO ,*=OO x

NS
dxdz. (10.229)
' ^ Jz=D Jx=0 (X2+Z2)n/2

solid

fluid
molecules

Figure 10.25
10.5 I Phase Equilibria in Microscale Multiphase Systems 371

The integral over x can be evaluated from standard integral tables. Doing so and evaluating
the results at the indicated limits yields

WD = -2npN,sCt [ \7 ^ r r l dz- ( i a 2 3 °)
2
JZ=D LO2-2)z*~ J
Executing the integration over z, gives

02-2)02-3)
Note that for n > 3, WD is negative. The work done by the system is therefore negative,
which implies that work must be done on the system by some outside agent to move the
solid molecules away from the fluid molecule, or conversely, to move the fluid molecule
away from the surface.
We now consider a thin slab of fluid molecules a distance delta from the solid surface
and having a differential thickness d8, as indicated in Figure 10.25.
Assuming that we can obtain the net effect by summing the contributions of the fluid
molecules in the layer, the work that must be done on the system per unit area to move the
thin layer of liquid molecules from distance 8 from the surface to infinity is

^f-"N,ld8. (10.232)
02-2)02-3)
If the solid was initially in contact with a semi-infinite region of liquid, the work required
to remove liquid to a thickness 80 is obtained by integrating the above relation:

= pN,xd8. (10.233)
O?-2)O2-3)
J8o
Note, however, that we previously obtained Eq. (10.222) for Wf from our macroscopic
analysis,
poo
Wf = -Pd(8)d8. (10.222)

Clearly, for both of the above equations for Wf to be valid, the disjoining pressure must be
given by

(n - 2 ) (/i - 3 )
It is conventional to define the Hamaker constant AH as
AH = 7r2p/v,sp^vjC0. (10.235)
With this definition, the disjoining pressure relation can be written

it{n — 2)02 — 3)
If the long-range attractive force potential is equivalent to the long-range portion of the
Lennard-Jones 12-6 potential, then n = 6 and the above relation reduces to

^8-\ (10.237)
6
372 10 I Nonequilibrium and Noncontinuum Elements of Microscale Systems

Equations (10.235) and (10.237) link disjoining pressure effects to the intermolecular long-
range attractive forces between molecules in the liquid near the solid surface and molecules
within the solid.

Example 10.6 For carbon tetrachloride (CCI4), the constants £0 = 4.51 x 10 21 J and
orLJ = 5.9 x 10~10 m are suggested for use in the Lennard-Jones 6-12 pair potential. Use
these data to estimate the dependence of disjoining pressure on liquid film thickness for
carbon tetrachloride on glass (SiO2).

Solution As an idealized model, we will assume here that the long-range force
interaction between a CCI4 molecule and a molecule in the solid is the same as that between
two CCI4 molecules. It follows that the long-range portion of the pair potential for CCI4-
solid molecule interactions,

0(r) = - ^ , (10.226)

must be equivalent to the outer portion of the Lennard-Jones 6-12 potential (10.227) for
CCI4 pairs interactions. This requires that n = 6 and from Eq. (10.228)
C 0 = 4eO0Lj = 4(4.51 x 10~21)(5.9 x 10~10)6 = 7.61 x 10~76 Jm6.
The density of the fluid, CC14, and the solid, SiO 2, are 1,590 kg/m3 and 2,210 kg/m3,
respectively. These are converted to number densities by dividing by the corresponding
molecular mass and multiplying by Avogadro's number. This gives

pN ! = ®-NA = (6.02 x 1026) = 6.22 x 1027 molecules/m3,

M\ 154.0
pN s = ^NA = ^ ^ ( 6 . 0 2 x 1026) = 2.21 x 1028 molecules/m3.
Ms 60.1
For n = 3, Eq. (10.236) becomes
-Pd = A8B,

where A = AH/67T and B = —3. Using Eq. (10.235), we then compute A as

A=AH = * pNlf)NsC^ = :n {622 x 1027)(2.21 x 1028)7.61 x 10~76

O7T O 6

= 5.48 x 10" 20 Pam~ 3 .

This idealized analysis indicates that A = 5.48 x 10~20 Pa m~3 and B = —3.0 in the disjoin-
ing pressure relation —Pd = A8B. This agrees well with data for thick films (8 > 10~8 m)
examined by Potash and Wayner [15]. However, for film thicknesses less than 10~8 m, these
investigators indicated that A = 1.782 Pa m~ 06 and B = -0.6 provides a better fit to data.
This implies that the Lennard-Jones-type pair potential considered in the simple analysis
in this section is somewhat inadequate for very thin films.

The equilibrium conditions for a very thin film in contact with a solid surface and its
vapor can be examined by considering the system shown in Figure 10.23. At a specified
10.5 I Phase Equilibria in Microscale Multiphase Systems 373

vapor ambient pressure Pwe, the system is at equilibrium at temperature TQ. Two necessary
conditions for equilibrium are that the temperature and chemical potential in the liquid and
vapor phases must be equal. As shown earlier in this section, integration of the Gibbs-
Duhem equation along a line of constant temperature from saturation conditions to an
arbitrary point on the isotherm in the vapor and liquid regions yields the following two
relations:

Ave = Asat,v + ^ e In j p ™ J , (10.238)

Al = AsatJ + V\[P\ ~ ftatOTe)]. (10.239)

Note that we have taken the isotherm temperature to be TQ. Setting /2ve = (Lx and solving
for P ve , we obtain
r f),rp, — p AT w ^
(10.240)

The analysis above links the disjoining pressure to the film thickness 8 by a relation of the
form
Pd = -A8~B. (10.241)
(See also the discussion of disjoining pressure in References [15] and [16].) It follows from
the above relation that for the system in Figure 10.23
/>, = p v e - A8~B. (10.242)
Substituting this expression into Eq. (10.240) yields
ve sat
p — p (T \ P Y n J ^ e' I (AC\lA/\\
I yg — * sat\-'6/ ^ r i i " ^lu.Zfi'Jj

For a specified film thickness, this relation can be used to determine the equilibrium
vapor pressure for a given temperature. Alternatively, for a specified system pressure and
film thickness, the equilibrium temperature can be determined. By rearranging this relation,
we can also obtain an expression for the equilibrium film thickness for a specified set of
vapor pressure and temperature conditions:

(10.244)

At a given vapor pressure, this equation indicates that if the film is thin enough, it can
be superheated (Te > r sat (P v )) and be in equilibrium with the vapor. This means that the
liquid is in a superheated metastable state and the vapor is also superheated, as indicated in
Figure 10.17. It also implies that heating the solid surface, and hence the liquid film, to a
temperature slightly above its normal saturation temperature will not necessarily evaporate
the liquid film completely. Instead, an equilibrium thin film may remain on the surface with
a thickness that satisfies Eq. (10.244).
Equation (10.244) can be interpreted as indicating that a thin film can exist on a solid
surface even when its vapor pressure in the surrounding gas is below the normal saturation
pressure for the system temperature. This is consistent with the observed adsorption of thin
374 10 I Nonequilibrium and Noneontinuum Elements of Microscale Systems

liquid films onto surfaces such as metal, even when the vapor pressure of the adsorbed
species is below the normal saturation value at the system temperature.
The liquid film thickness generally must be exceedingly small for the equilibrium satu-
ration conditions to differ significantly from normal flat-interface saturation conditions. For
carbon tetrachloride on glass, Potash and Wayner [15] recommend the values A = 1.782
Pa mB and B — 0.6. Using these values with thermodynamic data for CCI4, Eq. (10.231)
was used to predict the fractional change in the equilibrium saturation pressure with film
thickness at a system temperature of 77°C. The resulting variation is shown in Figure 10.26.
For a liquid film thickness of 1 mm (1,000 fim) the fractional change in the vapor pressure
is less than 10~4, or less than 0.01%. A 1% change in the vapor pressure would require a
film thickness of about 0.1 /xm. For films thicker than 1 /xm, the deviation from normal
equilibrium conditions is insignificant.
For highly wetting liquids (those that tend to spontaneously spread over a surface),
the above considerations also affect the thermophysics at the location where a pool of
liquid comes into contact with the wall of a container. Such circumstances are shown in
Figure 10.27. A highly wetting fluid establishes an extended meniscus at the wall, which

tetrachloride at 77 °C
10-5
lO"3 1(T2 lO' 1 1 10 10 2 10 3
8 (Jim)

Figure 10.26

wall at Tw > Tsat

Figure 10.27
10.61 Microscale Aspects of Electron Transport in Conducting Solids 375

includes an intrinsic meniscus region due to interfacial tension effects as well as a thin
film that results from the liquid's tendency to spontaneously spread over the surface. If
the wall temperature is above the normal saturation temperature, heat may be conducted
across the liquid film in this region, resulting in evaporation at the interface. However,
because the interface is curved in the intrinsic meniscus region, the wall superheat will have
to exceed that required for equilibrium with the interface curvature present. In addition,
because of the disjoining pressure effects described above, beyond the intrinsic meniscus a
thin film can exist on the wall at equilibrium, even when the wall is superheated above the
normal saturation temperature for the local vapor pressure. This thin film can exist without
evaporating if its thickness is equal to that specified by Eq. (10.244).

Closing Remarks
For all of the microscale multiphase systems considered in this section, the ther-
mophysics involves dynamic equilibrium with molecular exchange at the interface, and one
of the phases exists in a metastable state. Our statistical thermodynamics results in earlier
chapters indicates that the fluctuations associated with the metastable state will be low for
weak departures from normal saturation conditions. Note that this permits coexistence of
a metastable phase with another stable phase in these microscale multiphase systems. Our
analysis also indicates, however, that there are limits as to how far the metastable phase
can depart from normal saturation conditions. If we consider a bubble in equilibrium with
superheated liquid at a fixed temperature, for example, Figure 10.17 indicates that if the ra-
dius of the bubble becomes progressively smaller, eventually the equilibrium liquid pressure
becomes so low that it goes beyond the limit of intrinsic stability for the liquid. When this
limit is exceeded, the system cannot achieve an equilibrium condition. Similar limitations
apply to the cases of a droplet in equilibrium with supersaturated vapor and equilibrium for
a thin liquid film on a solid surface.

10.6 Microscale Aspects of Electron Transport in Conducting Solids

We now will examine the relationship between microscopic thermophysics of
electron behavior in solids and the macroscopic transport of heat and electric current. To
do so, it is necessary to determine the heat flux and current using a model that accounts for
the velocity and energy distributions of the electrons in the solid. Such an analysis is based
on the Boltzmann transport equation developed in Section 10.2.
The system of interest is shown schematically in Figure 10.28. We consider first a metal
in which the electron transport can be modeled as transport in an electron gas. In Section
10.2 we demonstrated that the Boltzmann transport equation applies to the distribution / pv
defined so that fvvdu dz is the number of particles in the ensemble (system) that have points
at a specific location of phase space (which defines a microstate). Here, we consider each
combination of u, v, and w to be a microstate. For an electron gas, we noted in Chapter 6
that, at equilibrium, the mean number of electrons in any microstate with energy s is given
by the Fermi distribution
(10245)
A -
For a specified set of w, v, and w values, the mean occupancy of the velocity microstate is
taken to equal the value predicted by the Fermi distribution at s = (l/2)me(u2 + v2 + w1).
376 101 Nonequilibrium and Noneontinuum Elements of Microscale Systems

material with electrical conductivity oe

and thermal conductivity kt
/

Z TX>T2

electric field Ex={<t>2-^)lL

temperature gradient dT/dx=(T2- T{) IL

Figure 10.28

To properly convert between the velocity component microstates and the energy microstates
on which the Fermi distribution is based, we must properly normalize the distribution
function. We therefore take the equilibrium distribution for the electrons to be

fnO = ' +e(£SF)/kBT ' (10.246)

If we integrate over all possible values of the velocity components, we account for the
contributions of all microstates. If we consider a unit volume of the conducting solid,
integrating fn0 over all w, v, and w must equal the number of electrons in the unit volume,
which is the number density of electrons pN:
/•OO /»OO pOQ

PN = / / / fnodudvdw. (10.247)
J — oo J—oo J—oo

To convert the triple velocity integral to an integral over energy, wefirstconvert the Cartesian
velocities to spherical coordinate representation using

u = csinO cos0, (10.248a)

v = c sin 0 sin 0, (10.248b)
w = ccosO, (10.248c)

where the angles are defined in Figure 10.1. Equation (10.247) then becomes
pin pn
roo pin
pN = / / / fn0c2 sin OcdO d(j)dc. (10.249)
Jo Jo Jo
Aside from the sine term, the integrand is independent of 0 and 0. We therefore execute the
integrations over these variables to obtain
poo
pN = 4TT / fn0 c2dc. (10.250)
Jo
We next use the fact that

c = (10.251)
10.6 I Micro scale Aspects of Electron Transport in Conducting Solids 377

to convert the variable of integration from c to s. Doing so yields

Since the Fermi distribution is virtually a step function except at very high temperatures,
we evaluate the integral as

AfS ' de = —5-75-Af£ F . (10.253)

me Jo 5me
Solving the above equation for Af, we obtain

e
A f= ™ 2, (10.254)

from which it follows that the equilibrium distribution is

(10.255)

When the system departs from equilibrium, we expect that the distribution obeys the
Boltzmann transport equation:

h > w; h > Wi = — . (10.256)

J J
d< fy Mj U ^j \ 9'/con
In our model analysis here, we will make the assumption that the distribution function
with temperature gradients and electric fields present is not very much different from
the equilibrium distribution, and the collision term can be represented by a relaxation
time rr:

where / n o is the equilibrium distribution for the system. Combining Eqs. (10.256) and
(10.257), the Boltzmann transport equation becomes.

dO.258)

When only an electric field in the x direction of magnitude E x is present, the Boltzmann
equation reduces to

eEA %U _ ^ _ Z A o > (1Q 2 5 9 )

dx \ me ) du rr
where we have replaced the acceleration in the x direction u with the electric field force
—e eEx divided by the mass of the charge carrier me.
378 10 I Nonequilibrium and Noncontinuum Elements of Mieroseale Systems

Based on our assumption that the distribution function is close to the equilibrium distri-
bution for the system fn0, we let /„ = nf0 on the left side of Eq. (10.259) and solve for /„
to obtain

^Y^-r^. (10.260)
du dx
Differentiating Eq. (10.255) to obtain dfn0/dT and dfn0/de yields
d_U _ ( 3m*'2PN \ d r ^
BT ~ )

Combining the above two equations to eliminate the exponential terms, we obtain

fn0 nno*i\
(ia263)
df
Using the chain rule we can write

(10 264)
T^^T- -
dx oT dx
Substituting the right side of Eq. (10.264) for dfn0/dT in Eq. (10.263) yields
. (10.265)
dx
We can similarly use the chain rule to obtain

^ = ^ ^ , (10.266)
du ds du
and since
s = (l/2)me(u2 + v2 + w2) (10.267)
it follows that
ds
— = m eu (10.268)
du
and Eq. (10.266) can be written as

—— =meu——. (10.269)
du ds
Using Eqs. (10.265) and (10.267), we can reorganize Eq. (10.260) to the form
feeExrT\ dfn0 d \s-s¥]dfnOdT nn07m
fn = //i0+ ^w- TrukBT— - r - = - -7— - j - . (10.270)
\ me J ds dT I kBT J 9£ Jx
Executing the differentiation on the right side then gives

(?)§]
10.61 Micro scale Aspects of Electron Transport in Conducting Solids 379

For the system considered here, the electron flux in the x direction is given by
pOQ rOO rOO

JeiX = - / / eQufndudvdw. (10.272)

J—oo J —oo J — oo

Substituting the right side of Eq. (10.271) for /„ in Eq. (10.272) yields
/•OO pOO />OO /»OO pOO /»OO

Je,x = - // / eeufnodudvdw - / / / ^e^2

J—oo J —oo
—oo J—oo J —oo J—oo J—oo

df
n0 x (e\dT d feF\dT]
x — x+ - -—+ T—[ — ) — \dudvdw. (10.273)
de \T J dx dT\T ) dx \

Because /wo is an even function of w, f, and w, and w is odd, the first triple integral in
the above relation is zero, whereupon the relation reduces to

,oo
,oo ,oo ?8fn0\ ( s\dT d(
±(sAdT_
e,x = - / / eQrYu —— \e eEx + - ) — + T—-( dudv dw.
J-ooJ-ooJ-oo 9^ L \T J dx dT\ J dx
dT\T
(10.274)

We first convert the Cartesian velocities to spherical coordinate representation using Eqs.
(10.248a)-(10.248c). Equation (10.274) then becomes

,oo r2n p VnO \ r t ( S \ dT _ d fs¥\dT] 4

Je,x = - / / e e r r -— \eeEx+ - - j - + T— — H ~ k 4
Jo 7o Jo 9£ L \T J dx dT\Tjdx\
x sin3 6> cos2 (pdOdcpdc. (10.275)

Executing the integrations over 0 and 0, we obtain

3 70 9s L \TJ dx dT\Tjdx\
(10.276)

Using the relation (10.251), we next convert the variable of integration from c to e. Doing
so yields

J dx dT \T
T J dx
d JJ
(10.277)

This relation can be rearranged to the form

JJ = '11111^1*17 4 . T — I — \'H-\ I T ~Jn() P 3 / 2

e,x /2 e r
Afmesl L dT \T J dx \ Jo de
xx—s 5/2de. (10.278)
s
380 10 I Nonequilibrium and None ontinuum Elements of Mieroseale Systems

The heat flux in the x direction is similarly given by

poo /»oo poo
J Q X = I j j usfndudvdw. (10.279)
J — oo J —oo J — oo

Converting the integrals to polar coordinate form, integrating over the angles, and converting
the integral over c to an integral over energy gives
d
PN [£AdT] f°° d
fn0 5/2,
JQ,X = — — — / r—— s"'1 ds
dT \T J dx J Jo r ds
(10.280)
Afmeel/2 \Tj dx Jo
For convenience, we define the following integrals:
1 f°° dfno
h/2 = --T / Tr—— (10.281)

1 Z*00 9/^0 5/2 ,

(10.282)
A fJ 0 ds
1 p 3/nO 7/2^ (10.283)
AfJ0 ds
With these definitions, the electron flux and heat flux relations can be written
a i s-p\dl \ I \ \ al
etEx + T (10.284)
df\j)dI.
dT
372
(10.285)
TJdx\ \T J dx
The electrical conductivity for an isothermal conductor, defined as
tfe = (Je,x/EX)T (10.286)
is determined by manipulating Eq. (10.284) to give

PNelh/2 (10.287)
Ge =
,3/2 *
mP
Note that in the above equations, the electron number density p^ is computed as the
atom number density in the metal multiplied by the number of valence electrons per atom.
If we set Jex to zero and combine Eqs. (10.284) and (10.285), we obtain the following
relation for the zero-current heat flux:
PN
JQ,X = — 3/2™
(10.288)
mesF T
h/2 J dx
The definition of the zero-current thermal conductivity is

(10.289)
dT/dx'
10.61 Micro scale Aspects of Electron Transport in Conducting Solids 381

Using Eq. (10.288), the thermal conductivity is determined to be

(10.290)
mee5v'zT \ h/i
Computational evaluation of the fluxes and conductivities requires that we evaluate
the integrals defined in Eqs. (10.281)-(10.283). Since fn0 is known, the main difficulty is
accounting for the relaxation time r r. In general, rr is proportional to the time between colli-
sions and is therefore a function of the energy of the particle. All the integrals are of the form

(10.291)
AfJ0 ds
We first expand the relation (10.262) in a Taylor series about ep- Retaining terms to second
order, we find

- — + •••, (10.292)

where
$ = (£-eF)/kBT. (10.293)
Since to second order
e~xl = 1 — JC2 H , (10.294)
we choose to modify Eq. (10.292) to the form

^ = - ( 3 2 * S t T)e~*2/4> (10 295)

-
where we expect this approximation to be accurate to second order in £. Also, by definition
sn = (sF + kBT^)n = ej(l + kBT^/s¥)n. (10.296)
We expand the second factor on the right of the above relation for small k BT^/sF to obtain

(10.297)
2

To account for the variation of r r with £, we also expand rr in a Taylor series about sF:

(e ep)2+
- • • • • (i °- 298)
Assuming a power law variation of r r with s,
rY = Ar£-y, (10.299)
we can evaluate the derivatives in the above equation and the expansion can be manipulated
to the form

Tr = Tr,F (10 .300)

382 10 I Nonequilibrium and Noneontinuum Elements of Microscale Systems

Inserting the right sides of Eqs. (10.295), (10.297), and (10.300) into the appropriate loca-
tions in (10.291), multiplying the series, and keeping terms to second order, we obtain

1 + (n - y) + yn +

(10.301)
SF

The integral can be evaluated term by term using standard integral tables, which yields the
following relation:

(10.302)
We can use this relation to evaluate the integrals in the flux and conductivity relations.
Another interesting result is obtained by considering the ratio known as the Lorenz
number Lz:

Lz = (10.303)
creT'
Substituting the relations (10.287) and (10.290) for the electrical and thermal conductivities
yields

Lz=\ •'/2'3? + li/2)% (10-304)

Equation (10.302) can be used to evaluate I>$/2,15/2, and Iq/2 in Eq. (10.296). Doing so
indicates that to second order in k#T/e$, the factor in parentheses in Eq. (10.304) is 4,
regardless of the value of y. This implies that

Lz = 4- ,2 '
(10.305)

A more rigorous treatment indicates that the factor in parentheses in Eq. (10.304) is about 3.3.
The value of Lz{e\jkl^) for most metals is, in fact, close to 3.3, as indicated in Figure 10.29.

w
n
- K Fe pt
: Na • Ag Pb
r?h > n ?| D
Au
: Al OCu
Ni

50 100 150 200 250

molecular mass

Figure 10.29
10.6 I Micro scale Aspects of Electron Transport in Conducting Solids 383

Although approximate, our analysis here indicates a value that is within about 25% of that
for most metals. This relation between the thermal and electrical conductivity for metals is
known as the Weidermann-Franz law.
This line of analysis can be applied to semiconductors, but the analysis becomes much
more complicated if the accessible electron energies have a band structure. If the energy
distribution is well approximated by a continuous relation, the analysis is somewhat easier.
The Fermi distribution used above applies to metals where the number density of free elec-
trons in the gas is high. In semiconductors, the number density of electrons becomes low
enough that the allowable energy microstates are not densely populated. In these circum-
stances the system exhibits dilute occupancy and the distribution deviates from that for a
fully degenerate fermion gas, approaching the classical form of a Maxwellian distribution
with a mean energy approximately equal to the Fermi energy. As a first approximation,
semiconductors have been modeled using a Boltzmann distribution centered on the Fermi
energy. Such a distribution would have the form

fno = Cf — e~^lk*T, (10.306)

\27imkT )
where Cf is an appropriate normalization constant. Note that this is consistent with our
observation in Chapter 6 that fermion gas behavior approaches that of a classical gas in the
limit of dilute occupancy. Constructing a model of this type with a power-law dependence
of rc on s yields results similar to those above for metals, except that Lz(el/k\) depends
on the power-law exponent y.
The analysis of electron transport presented in this section can also be extended to derive
information about thermoelectric effects. For example, setting J€jX — 0 in Eq. (10.284), we
obtain the relation for the electric field induced by the presence of a temperature gradient:

dx V h/i )
In Chapter 9, we defined the absolute thermoelectric power as the change in electric potential
per unit change in temperature along a conductor. Thus, for material A,
A0 (d(f)/dx)Ax
8A — 1 (I

AT (dT/dx)Ax (dT/dx)
and since basic electrostatics dictates that Ex = d(p/dx, we can combine Eqs. (10.307) and
(10.308) to obtain

Using Eq. (10.302) to evaluate and 1^/2 and 15/2 leads to

(10.310)

Note that since k^T/s^ is the Fermi temperature #F> this can be written in the form
384 10 I Nonequilibrium and Noncontinuum Elements of Microscale Systems

The ratio kB/ee is equal to 86.1 /xV/K. Because the Fermi temperature for most metals is
typically greater than 10,000 K (see Table 7.1), the absolute thermoelectric power for most
metals is very low, being on the order of 1-10 /xV/K. With some additional effort, relations
for other thermoelectric properties can also be obtained.

Example 10.7 Use the results of the analysis in this section to estimate the absolute ther-
moelectric power for aluminum at 300 K. Assume that the relaxation time r r is independent
of energy.

Solution For copper, the Fermi temperature is approximately 134,900 K (see

Table 7.1). Since the relaxation time is assumed to be constant, we use Eq. (10.311) with
y=0:

Substituting the parameter values, we obtain

The value predicted by this relation is only slightly different from the value of about
— 1.7 /x V/K indicated by measured data.

This line of analysis using the Boltzmann transport equation provides a linkage between
the microphysics of the electron behavior in the conductor and the macroscopic properties
of the material. It also provides a kinetic theory foundation for the linear transport models
discussed in Chapter 9. It should be noted that in this model we have neglected interactions
between electrons and the ions in the metal lattice as well as the scattering of electrons off
of contaminants and imperfections in the lattice. Because energy storage in the lattice is
associated with phonons (vibrations of the lattice; see Chapter 7), the interaction of electrons
with lattice waves is modeled as electron-phonon collisions. Scattering of electrons due
to imperfections in the lattice and collisions with phonons inhibit electron transport and
neglecting these effects generally overestimates the electrical conductivity. Phonons as well
as electrons transport heat in the metal. In metals, the electron concentration is large and
the electron contribution dominates the thermal conductivity of the material. Although the
kinetic theory model developed here is highly idealized, its predictions are comparable to
the behavior of real metals, and the model provides useful insight into how free electron
behavior affects the thermoelectric properties of the metal. The interested reader can find
further information about the microphysics of transport phenomena in conducting materials
in References [17] and [18] at the end of this chapter.

10.7 The Breakdown of Classical and Continuum Theories at Small Length

and Time Scales
The purpose of this final section is to summarize conclusions regarding the break-
down of classical and continuum theories in systems characterized by small length or times
scales. A primary objective throughout this text has been to construct theoretical machinery
10.71 The Breakdown of Classical and Continuum Theories 385

that can be applied to macroscopic system processes from information about the microstruc-
ture of matter. In doing so we have taken advantage of two features. The first is that quantum
effects vanish at high temperatures and/or large system sizes. The second is that the statistics
of a system of particles becomes increasingly deterministic as the number of particles in
the system becomes large.

Length Scales
We have identified two length scales that play a central role in assessing the
appropriateness of classical, continuum models of thermal phenomena. The first is the
thermal de Broglie wavelength A defined as

A= (10312)
y^rT-
As discussed in Chapter 4, if the characteristic dimension of the system, L, is comparable
to or smaller than A, quantum effects are important and must be included in any model
analysis. Thus
L =A (10.313)
defines one lower bound on the system size for applicability of classical thermodynamics
theory. Our analysis in Chapter 4 indicated that L < A is likely only for very low temper-
atures and/or systems containing particles of very low mass.
In gaseous systems, the mean free path of the particles between collisions Ac is another
limiting length scale. If the length scale of the system is comparable to or smaller than
Ac, the particles interact with a boundary of the system and then, on the average, travel a
distance more that L before interacting with another particle in the system. At mean free
path values comparable to L, the system cannot be accurately modeled as a continuum.
Depending on the type of boundary conditions on the system, the energy distribution within
the system may depart substantially from the Boltzmann distribution and consequently the
behavior of the system may depart from predictions of equilibrium theory.
In condensed matter systems such as solid crystals or liquids, the particles continuously
interact and so the concept of mean free path for the atoms or molecules is inapplicable.
However, it reappears when the electron gas model is applied or energy storage or trans-
port in the lattice is modeled in terms of phonons. When the system characteristic length
becomes comparable to the mean free path of the electron or phonon, local thermodynamic
equilibrium may not be achieved and the behavior of the system will deviate from the mean
behavior for a continuum.
Another intrinsic length scale for a condensed matter systems is the mean spacing be-
tween atoms or molecules L ms . This can be computed approximately from the molar specific
volume as
Lms = (v\/NA)l/3. (10.314)
When the dimensions of the system approach L ms , the number of molecules becomes too
small to effectively average in statistical theory. Also, as noted in our consideration of fluctu-
ations in systems of molecules, in general, the relative magnitude of fluctuations in density
and energy generally increases as the number of molecules becomes small. The system
therefore will not have well-defined average properties in the usual thermodynamic sense.
386 10 I Nonequilibrium and Noncontinuum Elements of Microscale Systems

Interfacial regions at a liquid-gas interface and thin adsorbed liquid films are examples
of cases where the system dimension is beginning to approach the mean separation dis-
tance between molecules. We have extended our thermodynamic analysis to such systems
by introducing a macroscopic treatment of the special microscale features of the system.
Clearly, as this length scale is approached, classical continuum theory becomes suspect and
its accuracy must be reassessed.
In two-phase gas-liquid systems we have seen that if the dimensions of the dispersed
phase become small, the conditions for equilibrium become much different from those
in which length scales of both phases are large. The departures from normal (flat inter-
face) equilibrium conditions become large as the limit of intrinsic stability is approached.
In superheated liquids, this limit corresponds approximately to about 0.9 times the critical
temperature. Applying Eq. (10.183) at this temperature, we conclude that the departure from
normal equilibrium conditions becomes substantial when the size of a bubble approaches
r n , given by

r)} P{ '
Similar considerations apply to the limiting dimensions of liquid droplets in supersaturated
vapor and thin liquid films in superheated vapor.
In addition to the physical length scale for the system, the length scale over which
temperature changes significantly is also important in assessing the viability of classical
continuum theory. This length scale L^T is defined as

LAT = — . (10.316)

If boundary conditions are established in a system that result in L A j being comparable

to the mean free path of the particles, the continuum model will not accurately represent
the system behavior. From the definition, this clearly is likely to occur only if the system
temperature is low and/or the temperature gradient is very large. When such conditions
exist, use of classical continuum models must be reconsidered.

Time Scales
In gaseous systems, collisions redistribute energy in the system, which facilitates
its evolution toward equilibrium. Equilibrium statistical thermodynamics provides useful
predictions of physical properties in gaseous systems if the time scale over which the system
is permitted to reach equilibrium is much longer than the mean time between collisions (r c ).
Fortunately, in most systems (r c) is very small and this requirement is satisfied. However,
in some systems this is not satisfied and the system may exist in a nonequilibrium state for
long periods on time, making it impossible to analyze the system behavior with classical
thermodynamics tools.
Within gaseous systems, the number of collisions per molecule required to reach an
equilibrium distribution may be different for different modes of energy storage. In some
systems, for example, the translational energy may achieve a Boltzmann distribution in
about 5 collisions per molecule, while it may take an average of 20 collisions for rotational
energy and an average of 50 collisions per molecule for the vibrational energy to achieve
equilibrium distributions. This difference occurs because rotational and vibrational energy
10.71 The Breakdown of Classical and Continuum Theories 387

are not exchanged in every collision. In high speed gas flow across a shock wave, the
residence time of the gas in the system may be small compared to the time required for the
gas to reach full equilibrium, resulting in a gas flow downstream of the shock that is far
from equilibrium. If the rotation and/or vibration modes are very slow to equilibrate, they
are sometimes said to be "frozen" in the distributions that existed before the gas passed
through the shock.
In solids, the time constants of interest are the mean time between collisions for phonons
and electrons. In an analogous manner to gases, these time constants dictate the rate at which
energy is transported and distributed in the material. In a solid crystal, the characteristic
time associated with lattice vibrations is on the order of the mean of the inverse of the
frequency of lattice vibrations:

(10.317)

At very low temperatures the lower energy, lower frequency modes contain most of the
energy and the time between phonon collisions is longer than at higher temperatures where
higher frequencies contain more energy resulting in a higher mean time between phonon
collisions. In metals where a substantial portion of energy transport is by free electrons, the
mean time between electron collisions with electrons and phonons is a limiting time scale
for equilibration of the system and transport of heat.
Very short time scale pulsed laser heating of solids is now possible in which energy is
delivered to a localized region of the solid over a time interval that is comparable to the mean
time between collisions of phonons and electrons. Processes of this type cannot be modeled
using classical tools based on local thermodynamic equilibrium because the system cannot
react fast enough to reach local equilibrium. These time scales thus define limits to the
applicability of classical continuum tools that are commonly used in larger-scale slower
processes.

Other Considerations
Another feature of systems that can cause them to deviate from conventional
classical thermodynamic behavior is high levels of property fluctuations. We have shown
that pure liquids and gases may exist in metastable states in which —(dP/dv)r may be
small, resulting in very large density and energy fluctuations. When this is so, the properties
vary significantly with time. This is quite unlike systems in which —(dP/dv)r is large
and the system properties are virtually invariant with time. The classical definition of an
equilibrium state does not exist for systems that exhibit severe property fluctuations. As we
have seen in previous chapters, this type of nonclassical behavior is exhibited near the limits
of intrinsic stability and near the critical point. Small —{dP/dv)j thus represents another
limit of the range of applicability of classical thermodynamic theory. We have also shown in
Section 10.4 that at high vaporization or condensation rates, the conditions at liquid-vapor
interfaces may deviate significantly from those for local thermodynamic equilibrium.

Other Analysis Tools

The above discussion describes several instances in which classical thermodynam-
ics and continuum transport theory based on local thermodynamic equilibrium no longer
388 101 Nonequilibrium and Noncontinuum Elements ofMicroscale Systems

provide accurate models of system behavior. The obvious question at this point is: What
can be done to model systems that exhibit nonclassical or noncontinuum behavior? There
are at least three avenues to pursue. The first is to adapt the statistical thermodynamics
framework described here to handle nonequilibrium systems. Nonequilibrium statistical
mechanics continues to be an area of ongoing research. Much progress has been made,
and work in this area continues to develop the analytical machinery that can be applied to
nonequilibrium systems.
The second avenue of analysis for nonequilibrium systems involves solving the Boltzmann
transport equation. In this chapter we have developed only an approximate analysis for sys-
tems that deviate only slightly from equilibrium. Solving the Boltzmann transport equation
for a system that departs strongly from equilibrium is a challenging problem, as was rec-
ognized by Boltzmann himself. Although Boltzmann derived the equation in 1872, fruitful
attempts to solve the equation did not come until much later. In 1905 Lorenz applied
Boltzmann's transport theory to the electron gas model of electrical conduction. In 1917
David Enskog and Sydney Chapmann independently developed the mathematical theory
of nonequilibrium gases based on Boltzmann's transport theory. Harold Grad solved the
Boltzmann equation by expanding in a series of orthogonal polynomials in 1949. Since the
1950s, computer-based methods to solve the Boltzmann transport equation have also been
developed and have been used successfully to determine transport in nonequilibrium sys-
tems. Computer-based methods for solving the Boltzmann transport equation show promise
as a means of analyzing nonequilibrium systems that are far from equilibrium.
The third alternative is to use computer molecular dynamic simulations to predict sys-
tem thermophysical properties and/or transport in nonequilibrium systems. This category
of techniques includes both deterministic molecular dynamic simulations and stochastic
Monte Carlo simulation schemes. Deterministic schemes have the advantage that they ac-
count for complex molecular interactions in the system in a relatively straightforward way.
Stochastic methods use probability distributions derived from kinetic theory to represent the
interactions among molecules. The use of such distributions tends to average the character-
istics of such interactions but offer improved computational efficiency. With the projected
continued increase in accessible computing power, molecular simulation schemes are likely
to be used with increasing frequency for analyzing systems that depart strongly from equi-
librium conditions. More information on such methods can be obtained from References
[19] and [20] at the end of this chapter.

Exercises
10.1 Estimate the mean free path and mean time between collisions for nitrogen gas at atmo-
spheric pressure and 80 K. You may assume that nitrogen molecules have an effective
diameter of 3.7 A.
10.2 The thermal conductivity of nitrogen gas at atmospheric pressure and 300 K is reported to
be 0.0259 W/mK. Use the results of the simple kinetic theory developed in Section 10.1
together with the known behavior of ideal gases to estimate the mean effective diameter of
a nitrogen molecule.
10.3 If the gas in Example 10.3 is argon at atmospheric pressure and 300 K, estimate the time
required for the gas to reach equilibrium if the relaxation time is taken to be that required
for each atom to undergo three collisions.
10.4 Derive Eq. (10.150) from Eq. (10.32).
Exercises 389

10.5 Film condensation of pure oxygen vapor on a cooled, flat surface occurs at atmospheric pres-
sure. The heat flux removed at the solid wall is a constant and uniform value of 100 kW/m 2.
Assuming an accommodation coefficient of 1.0, estimate the temperature difference across
the interface during this process.
10.6 A thin film of liquid nitrogen flows along the walls of the evaporator tube with mostly
vapor flowing in the central core of the tube. The pressure in the tube at the location of
interest is 208 kPa. A heat flux of 50 kW/m2 is applied to the outer surface of the tube and is
conducted through the tube wall and the liquid film to the interface. For vaporization at this
heatfluxlevel, estimate the temperature difference across the interface if the accommodation
coefficient is taken to be 0.9.
10.7 Determine the equilibrium vapor pressure for nitrogen vapor in equilibrium with droplets
having a diameter of 0.5 /xm at 80 K.
10.8 From Eq. (10.202), the ideal gas law, and the Young-Laplace equation, derive Eqs. (10.205)
and (10.206).
10.9 In superheated water at atmospheric pressure and 460 K, spontaneous fluctuations create a
bubble with a radius of 4.0 x 10~8 m. What do you expect to happen to this bubble? Briefly
explain your answer. Properties of water at 460 K are: v\ = 0.0205 m3/kmol, Psat = 1,172
kPa, and 07 = 0.0407 N/m.
10.10 Estimate the number of molecules needed to make up a bubble in equilibrium with super-
heated liquid nitrogen at atmospheric pressure and 100 K.
10.11 A water droplet with a radius of 2.0 nm (nanometers) spontaneously forms in supersaturated
steam at atmospheric pressure and 80° C. What do you expect to happen to this bubble?
Briefly explain your answer. Properties of water at 80°C are: v\ = 0.0185 m3/kmol, Fsat =
47.4 kPa, and at = 0.0629 N/m.
10.12 In the last stages of the vaporization process inside the tubes of an evaporator, the two-phase
flow consists of 5 jim diameter droplets of R-134a liquid entrained in the flowing vapor.
If the vapor were in equilibrium with the droplets at a vapor pressure of 315 kPa, by how
much would the equilibrium temperature differ from the standard (flat interface) saturation
temperature?
10.13 Droplets of liquid water in atmospheric fog have a mean diameter of 10 fim. If the air
temperature is 14°C, estimate the partial pressure of water vapor in the air. (Hint: Treat the
air and water vapor as a mixture of ideal gases.)
10.14 Derive the equation for JQtX at zero current (10.288) from Eqs. (10.284) and (10.285).
10.15 Derive Eq. (10.302) for /„ from Eq. (10.291) using Eqs. (10.295), (10.297), and (10.300).
10.16 Use Eqs. (10.304) and (10.302) to show that Eq. (10.305) is valid to second order in kB
T/s¥.
10.17 Use the results of the analysis in Section 10.6 to estimate the absolute thermoelectric power
for gold at 300 K. Assume that the relaxation time r r is independent of energy.
10.18 Equation (10.311) predicts that the absolute thermoelectric power varies linearly with tem-
perature. Assuming that the relaxation time rr is independent of energy, compute and plot
the variation for copper between 250 and 350 K and compare the results to the following
curvefit to experimentally determined values:
sCu = 1.83 - 0.0032(300 - T), (10.318)
where T is in K and the absolute thermoelectric power is in /xV/K.
390 10 I Nonequilibrium and Noncontinuum Elements of Microscale Systems

10.19 Use the results of the analysis in Section 10.6 to estimate the absolute thermoelectric power
for aluminum at 300 K. Determine the value of y that matches the model prediction to the
reported value of —1.66 /xV/K.
10.20 For liquid methane at atmospheric pressure, estimate the size of a bubble in equilibrium
with surrounding superheated liquid at which classical theory is expected to fail because
large density and energy fluctuations occur in the liquid.

References
[1] Chapman, S. and T. G. Cowling, The Mathematical Theory of Non-Uniform Gases, 3rd ed.,
Cambridge University Press, Cambridge, UK, 1970.
[2] Gombosi, T. I., Gaskinetic Theory, Cambridge University Press, Cambridge, UK, 1994.
[3] Hirschfelder, J. O., C. F. Curtiss, and R. B. Bird, Molecular Theory of Gases and Liquids,
J. Wiley & Sons, New York, 1954.
[4] Adamson, A. W., Physical Chemistry of Surfaces, 4th ed., J. Wiley & Sons, New York, 1982.
[5] Miller, C. A. and P. Neogi, Interfacial Phenomena, Marcel Dekker, New York, 1985.
[6] Schrage, R. W, A Theoretical Study of Interphase Mass Transfer, Columbia University Press,
New York, 1953.
[7] Wilhelm, D. J., "Condensation of Metal Vapors and the Kinetic Theory of Condensation,"
Argonne National Laboratory Report ANL-6948, 1964.
[8] Paul, B., "Compilation of Evaporation Coefficients," ARS Journal, 32:1321, 1962.
[9] Mills, A. F., "The Condensation of Steam at Low Pressures," Technical Report NSF GP-2520,
series no. 6, issue no. 39, Space Sciences Laboratory, University of California at Berkeley,
1965.
[10] Yasuoka, K., M. Matsumoto, and Y. Kataoka, "Molecular Simulation of Evaporation and
Condensation I. Self Condensation and Molecular Exchange," Proc. ASMEIJSME Thermal
Enginerring Joint Conf, 2:459, 1995.
[11] Zeldovich, Y. B., "On the Theory of New Phase Formation: Cavitation," Acta Physiochem.
URSS, 18:1, 1943.
[12] Kagen, Y, "The Kinetics of Boiling of a Pure Liquid," Russian J. Phys. Chem., 34:42, 1960.
[13] Carey, V. P., Liquid-Vapor Phase-Change Phenomena, Taylor and Francis, New York, 1992.
[14] Debenedetti, P. G., Metastable Liquids, Concepts and Principles, Princeton University Press,
Princeton, NJ, 1996.
[15] Potash, M., Jr. and P. C. Jr. Wayner, "Evaporation from a Two-Dimensional Extended
Meniscus," Int. Journal of Heat and Mass Transfer, 15:1851, 1972.
[16] Deryagin, B. V. and A. M. Zorin, "Optical Study of the Adsorption and Surface Condensation
of Vapors in The Vacinity of Saturation on a Smooth Surface," Proc. 2nd Int. Congress on
Surface Activity (London), 2:145, 1957.
[17] Kittel, C , Introduction to Solid State Physics, 6th ed., John Wiley & Sons, New York, 1986.
[18] Ziman, J. M., Principles of the Theory of Solids, 2nd ed., Cambridge University Press, 1972.
[19] Haile, J. M., Molecular Dynamics Simulation, Elementary Methods, John Wiley & Sons, New
York, 1992.
[20] Frenkel, D. and B. Smit, Understanding Molecular Simulation, From Algorithms to
Applications, Academic Press, San Diego, CA, 1996.
 APPENDIX I

Some Mathematical Fundamentals

I.I Probability Distributions

Let u be a variable that can assume M discrete values u\, U2,..., uM with corr-
esponding probabilities P(u\), P(u2), •. •,P(UM)- The variable u is said to be a discrete
variable and P(w) is said to be a discrete distribution. The mean value of u is computed
as

(1-1)

Since the probability distribution must be normalized,

the relation for the mean value of u reduces to

M

The mean of any function f(u) is given by

M

(f(u)) = ^2f(uj)P(uj). (1.4)

7=1

If the random variable u is continuous rather than discrete, we define P{u)du as the prob-
ability that the random variable u lies between u and u + du. If the probability density
function P(M) is normalized so that

f P(u)du = \

the mean value of u is given by

(1.5)

uP(u)du, (1.6)

where the integration limits in Eqs. (1.5) and (1.6) imply integration over the entire accessible
range of u values. In a similar fashion, the mean value of a function of u is computed
as

</(")>= I f(u)P(u)du, (1.7)

•7Wmin

where again integration is executed over the entire range of u.

391
392 I / Some Mathematical Fundamentals

Gaussian distributions with um - 1.0

P(u)

I
1 ^^ c7 = 0.1

\
= 0.5

< :
-0.5 0.5 1.5 2.5

Figure I.I

An important and frequently encountered continuous probability distribution is the

Gaussian distribution defined as

P(u) = (1.8)

where um is the mean value of u computed using Eq. (1.6). The quantity a 2 , termed the vari-
ance, dictates the width of the Gaussian distribution. As indicated in Figure I.I, the smaller
the value of a, the narrower the peak of the distribution. In the limit a -> 0, the Gaussian
distribution becomes a delta function.
Further discussion of probability distributions and their use can be found in References
[1] and [2].

1.2 Some Results of Combinatorial Analysis

The development of statistical thermodynamics presented in this text makes heavy
use of results from a branch of mathematics known as combinatorial analysis. Derivation
of these results is beyond the scope of this text. However, in this section some useful results
are summarized and examples are given to illustrate their use.
For TV distinguishable objects
(the number of ways of arranging}
= N\ = N(N - l)(N - 2 ) . . . (2)(1) (1.9)
\N distinguishable objects J
and
' the number of ways of putting N distinguishable
objects into r distinguishable boxes so that there N\
are N\ in box 1, N2 in box 2, ..., Nr in box r, and (Nll)(N2\) •••
(Nr\)
order in any box does not matter
(1.10)
1.2 / Some Results of Combinatorial Analysis 393

AB AC AD AE BC 1

CDE BDE BCE BCD ADE -2

— 10 ways
BD BE CD CE DE 1

ACE ACD | ABE ABD ABC -2

Figure 1.2

To illustrate use of Eq. (1.10), suppose we have objects A, B, C, D, and E and we are to
put them in two boxes (1 and 2) so that there are two in box one and three in box two. Thus
N = 5,Ni=2, and N2 = 3 and
5.4-3.2.1
{the number of ways) = = 10.

The ten different ways are illustrated in Figure 1.2.

Additional useful results for distinguishable objects are
(the number of ways of selecting N distinguishable G!
(1.11)
{objects from a set of G distinguishable objects (N\)(G-N)l
and
(the number of ways of N distinguishable objects^
= GN. (1.12)
[can be put in G distinguishable boxes
To illustrate use of Eq. (I.I 1), suppose we have objects A, B, C, Z), and E and we are to
select three objects. Thus G = 5, TV = 3, and
5.4.3.2.1
{the number of ways} = = 10.

The ten different combinations are

ABC, ABD, ABE, ACD, ACE, ADE, BCD, BCE, BDE, CDE.
To illustrate use of Eq. (1.12), we consider objects A, B, and C, which we are to put in box
number 1 and box number 2. Thus G = 2, N = 3, and
{the number of ways] = (2)3 = 8.
The eight different ways are illustrated in Figure 1.3.
A useful result for indistinguishable objects is
(the number of ways ofN indistinguishable objects (N + G- 1)!
(1.13)
\can be put in G distinguishable boxes N\(G- 1)!
To illustrate use of this relation we consider four indistinguishable objects A, A, A, A,
which we are to put in two boxes. It follows that N = 4, G = 2, and

{the number of ways) = = 5.

(4-3 .2-1X1)
394 I / Some Mathematical Fundamentals

ABC AB AC

ABC C B
— 8 ways
BC A | C B

A BC AB AC

Figure 1.3

AAAA AAA A AA
— 5 ways
| AAAA A AAA AA

Figure 1.4

The five different ways of putting the four objects into the two boxes are illustrated in
Figure 1.4.
More detailed discussion of the basic elements and application of combinatorial analysis
can be found in References [3] and [4].

1.3 The Method of Lagrange Multipliers

We wish to address the problem of maximizing a function of several variables
f(x\, JC2,..., xr) when other constraints, expressible as g\ (x\, jt2,..., xr) = 0, g2{x\, *2>
. . . , xr) = 0, etc., are imposed. If it were not for the constraints, the maximum of f(x\, x2,
..., xr) is dictated by the requirement that for arbitrary changes in the independent variables

(1.14)
J=\
In the above relation, the zero subscript indicates that the equation equals zero only when
all partial derivatives are evaluated at the maximum (or minimum) of / .
If we have an additional constraint g\(x\, X2, ..., xr) = 0, the constraint relation can be
differentiated to obtain

SXJ = 0. (1.15)

We next multiply both sides by a constant X\ and subtract the resulting relation from
Eq. (1.14) to obtain

(1.16)

Since this must hold at any combination of 8XJ values, it follows that
d
f\ , fdSi\ = 0 > l<j<r. (1.17)
K
j / o
)
x
jJo
1.4 / Stirling's Approximation 395

We thus have a set of r equations that, in principle, can be solved for r values of the Xj at
the extremum: Xj = Xjj0. The values of x7>0 thus obtained will depend on Xi. In practice,
the multiplier value is usually dictated by physical requirements of some sort.
Lagrange's method is easily extended to multiple constraints. Expressing the constraints
as gt (JCI , x 2 , . . . , xr) = 0, we differentiate to yield for each

-— &*i = 0 - (1.18)
7= 1

Multiplying each relation of this type by Xt and subtracting it from the 8f relation yields

f-^ ^ ^=o. (1.19)

It follows that the x7?0 values at the extremum are determined by simultaneous solution of
the equations

(1.20)

As in the single-constraint case, the multipliers A.,- will generally be dictated by physical
considerations.

1.4 Stirling's Approximation

Because AM = - 1)(JV - 2) • • (2)(1),
• lnA^! is given by
N

(1.21)
m=\

In the plot in Figure 1.5, the summation in Eq. (1.21) is equal to the sum of the areas in the
rectangles. The function lnx forms an envelope to the rectangles such that the area under
the curve approximates the area in the rectangles. This approximation gets more accurate
as TV and x increase. This implies that
N
lnA^! = = / Inxdx = [xlnx — x]^. (1.22)
m=\

3.5-n

lire 3-j
2.5-=
2\
1.5-=
1^
0.5-i

0 5 10 15 20 25 30
x

Figure 1.5
396 / / Some Mathematical Fundamentals

Neglecting 1 compared to TV for large N, we find Stirling's approximation for large N:

\nNl = N\nN -N. (1.23)

1.5 Relations among Partial Derivatives

In our thermodynamics analysis, we are frequently interested in functional relations
among three variables such as x = x (y, z) •For differential changes in the variable, it follows
from basic calculus that

If the relation among the three variables can be inverted to the form y = y(x,z),it follows
from the same reasoning that

Substituting the right side of (1.25) into (1.24) for dy and rearranging yields

Equation (1.26) must hold for arbitrary values of two of the three variables x, y, and z. If we
choose x and z as the independent variables, Eq. (1.26) must hold for arbitrary dx and dz.
This is true only if the terms in square brackets in Eq. (1.26) are identically zero. Setting the
expressions in square brackets equal to zero, we obtain the following two relations among
partial derivatives that must be satisfied:

*).©.-•

References
[1] Kreysig, E., Advanced Engineering Mathematics, 5th ed., John Wiley & Sons, New York, 1983.
[2] J. L. Devore, Probability and Statistics for Engineering and the Sciences, 3rd ed., Duxbury
Press, Belmont, CA, 1991.
[3] Riordan, J., An Introduction to Combinatorial Analysis, Princeton University Press, Princeton,
NJ, 1980.
[4] Slomson, A. B., An Introduction to Combinatorics, Chapman and Hall, London, 1991.
APPENDIX II

Physical Constants and Prefix Designations

Table II. 1 Physical Constants'1

Avogadro's number, NA = 6.02252 x 1026 kmor 1

Boltzmann's constant, kB = 1.38054 x 10~23 J/K
Electron rest mass, me = 9.1091 x 10~31 kg
Electron charge, ee = 1.6021 x 10"19 coulomb
Earth normal gravitational acceleration, gc = 9.80665 m/s2
Planck's constant, h = 6.6256 x 10"34 J s
Stephan-Boltzmann constant, <JSB = 5.6697 x 10"8 W/m2-K4
Universal gas constant, R(= NA/kB) = 8.3143 kJ/kmol-K
Atomic mass unit, uam = 1.66043 x 10~27 kg
Speed of light in vacuum, q = 2.997925 x 108 m/s
a
Source: "New Values for the Physical Constants," Phys.
Today, 17:48-9, Feb. 1964.

Table II.2 Prefix Designations

Fraction Prefix Abbreviation

1
IO- deci d
io- 2 centi c
io- 3 milli m
io- 6 micro ti
IO- 9 nano n
io- 1 2 pico P
io- 1 5 femto f
io- 1 8 atto a

Multiple Prefix Abbreviation

10 deka da
IO2 hecto h
103 kilo k
IO6 mega M
IO9 giga G
IO12 tera T
IO15 peta P
IO18 exa E

397
 APPENDIX III

Thermodynamic Properties
of Selected Materials

Figures and tables in this appendix are reprinted by permisson of the American
Society of Heating, Refrigerating and Air-Conditioning Engineers, Atlanta, Georgia, from
the 1997 ASHRAE Handbook - Fundamentals (SI edition).

Table III. 1 Argon - Properties of Saturated Liquid and Vapor

Enthalpy Entropy
Density Volume
Temp. Press. kg/m3 m3/kg kJ/kg kJ/(kg K)
K MPa Liquid Vapor Liquid Vapor Iliquid Vapor

**83.80 0.068950 0.24746 1415.0 -121.82 42.417 1.3250 3.2848

85 0.078975 0.21846 1407.8 -120.46 42.892 L.3410 3.2628
90 0.13362 0.13476 1377.0 -114.74 44.735 1.4060 3.1779
95 0.21321 0.08758 1345.2 -108.98 46.317 L.4676 3.1023
100 0.32401 0.05938 1312.0 -103.15 47.594 1.5266 3.0340
105 0.47259 0.04167 1277.3 -97.202 48.520 L5835 2.9713
110 0.66575 0.03006 1240.8 -91.099 49.943 L.6388 2.9128
115 0.91049 0.02217 1202.0 -84.795 49.100 1.6931 2.8574
120 1.2139 0.01663 1160.4 -78.234 48.610 1.7467 2.8038
125 1.5835 0.01263 1115.4 -71.340 47.464 1.8004 2.7508
130 2.0269 0.009661 1065.6 -63.998 45.495 L8548 2.6970
135 2.5527 0.007393 1009.2 -56.016 42.431 L.9112 2.6404
140 3.1704 0.005600 941.94 -47.017 37.749 1.9720 2.5775
145 3.8921 0.004103 853.13 -36.042 30.180 :>.O433 2.5000
* 150.73 4.865 0.00187 535. 1.87 1.87 :121 2.27

** Triple point.
* Critical point.

399
400 III / Thermo dynamic Properties of Selected Materials

<
I
h-
Z
LU

•S "
ii

goc

li
11

I! -

E1
Ill / Thermodynamic Properties of Selected Materials 401

aanssaud
402 III / Thermodynamic Properties of Selected Materials

Table III.2 Methane - Properties of Saturated Liquid and Saturated Vapor

Enthalpy Entropy
Density Volume
Temp. Press. kg/m3 m3/kg kJ/kg kJ/(kg K)
K MPa Liquid Vapor Liquid Vapor Liquid Vapor

**90.68 0.011719 3.9781 451.23 -357.68 185.75 4.2894 10.2823

100 0.034495 1.4769 438.89 -326.63 203.44 4.6147 9.9154
110 0.088389 0.6234 424.89 -292.28 221.11 4.9408 9.6080
115 0.12357 0.42971 417.58 -274.79 229.29 5.0954 9.4787
120 0.19189 0.30565 410.05 -247.07 236.97 5.2450 9.3620
125 0.26933 0.22330 402.27 -239.12 244.06 5.3900 9.2555
130 0.36800 0.16688 394.23 -220.89 250.51 5.5311 9.1572
140 0.64196 0.098424 377.15 -183.40 261.15 5.8036 8.9789
150 1.0405 0.061244 358.26 -144.02 268.02 6.0677 8.8146
160 1.5918 0.039387 336.61 -101.79 269.82 6.3299 8.6525
170 2.3266 0.025614 310.47 -55.066 264.21 6.5992 8.4773
180 3.2820 0.016266 276.00 -0.242 245.79 6.8937 8.2605
* 190.555 4.5950 0.006166 262.2 132.3 132.3 7.572 7.572

** Triple point.
* Critical point.

Table III. 3 Nitrogen - Properties of Saturated Liquid and Vapor

Enthalpy Entropy
Density Volume
Temp. Press. kg/m3 m3/kg kJ/kg kJ/(kg K)
K MPa Liquid Vapor Liquid Vapor Liquid Vapor

**63.15 0.012530 1.4817 867.78 -150.45 64.739 2.4271 5.8381

65 0.017418 1.0942 860.78 -146.79 66.498 2.4841 5.7688
70 0.038584 0.52685 840.77 -136.67 71.058 2.6338 5.6042
75 0.076116 0.28217 819.22 -126.39 75.275 2.7750 5.4664
77.35 0.101325 0.21680 808.61 -121.53 77.113 2.8384 5.4090
80 0.13699 0.16409 796.24 -116.02 79.065 2.9078 5.3486
85 0.22903 0.10174 771.87 -105.56 82.334 3.0333 5.2455
90 0.36071 0.06631 745.99 -94.914 84.982 3.1530 5.1531
95 0.54090 0.04491 718.38 -83.991 86.890 3.2684 5.0680
100 0.77886 0.03132 688.65 -72.666 87.901 3.3811 4.8973
105 1.0842 0.02228 656.20 -60.785 87.791 3.4926 4.9078
110 1.4671 0.01602 620.04 -48.119 86.203 3.6048 4.8258
115 1.9390 0.01150 578.14 -34.247 82.471 3.7211 4.7358
120 2.5133 0.008035 525.12 -18.105 74.996 3.8495 4.6251
125 3.2099 0.004863 431.03 6.015 55.882 4.0342 4.4331
* 126.20 3.400 0.003184 314.0 30.70 30.70 4.2270 4.2270

** Triple point.
* Critical point.
Ill / Thermodynamic Properties of Selected Materials 403

i
Q.

LU

5
18
o<
fi
|o

|J

oi O

ii 3
404 III / Thermodynamic Properties of Selected Materials

o X

LU

ooc

It
li
cc ^

11 }
Ill / Thermodynamic Properties of Selected Materials 405

D)

a.
<

LU

fi
ifgi
2§

S|
406 III /Thermo dynamic Properties of Selected Materials

Table III.4 Oxygen -Properties of Saturated Liquid and Vapor

Density Volume Enthalpy Entropy

Temp. Press. kg/m3 m 3 /kg kJ/kg kJ/(kg K )
K MPa Liquid Vapor Liquid Vapor Liquid Vapor
**54.36 0.000146 96.500 1306.8 -193.50 49.174 2.0931 6.5571
60 0.000726 21.460 1281.5 -183.92 54.266 2.2608 6.2303
70 0.006262 2.8910 1236.6 -167.35 63.127 2.5161 5.8083
80 0.030124 0.68011 1190.2 -150.62 71.541 2.7393 5.5161
90 0.099348 0.22749 1141.9 -133.68 79.170 2.9381 5.3031
90.188 0.101325 0.22341 1141.0 -133.36 79.304 2.9417 5.2996
95 0.16308 0.14418 1116.7 -125.10 82.572 3.0302 5.2163
100 0.25400 0.09568 1090.7 -116.43 85.629 3.1183 5.1389
105 0.37853 0.06593 1063.7 -107.62 88.285 3.3032 5.0689
110 0.54341 0.04685 1035.4 -98.634 90.479 3.2853 5.0045
115 0.75561 0.03412 1005.5 -89.417 92.141 3.3654 4.9441
120 1.0223 0.02535 973.78 -79.909 93.184 3.4440 4.8864
130 1.7490 0.01457 902.39 -59.663 92.932 3.5998 4.7736
140 2.7876 0.008541 813.20 -36.684 88.067 3.7610 4.6521
150 4.2188 0.004660 675.87 -6.672 72.654 3.9543 4.4832
*154.581 5.0429 0.002293 436.1 33.09 33.09 4.205 4.205

** Triple point.
* Critical point.

Table III.5 Refrigerant 134a (1,1,1,2-tetrafluoroethane) - Properties of Saturated Liquid

and Saturated Vapor

Enthalpy Entropy
Density Volume
Temp. Press. kg/m 3 m 3 /kg kJ/kg kJ/(kg K )
K MPa Liquid Vapor Liquid Vapor Liquid Vapor
**-103.30 0.00039 1591.2 35.263 71.89 335.07 0.4143 1.9638
-80.00 0.00369 1526.2 4.2504 99.65 349.03 0.5674 1.8585
-60.00 0.01594 1471.0 1.0770 123.96 361.51 0.6871 1.8016
-40.00 0.05122 1414.8 0.36095 148.57 374.16 0.7973 1.7649
-30.00 0.08436 1385.9 0.22596 161.10 380.45 0.8498 1.7519
-26.07 0.10132 1374.3 0.19016 166.07 382.90 0.8701 1.7476
-20.00 0.13268 1356.2 0.14744 173.82 386.66 0.9009 1.7417
-10.00 0.20052 1325.6 0.09963 186.78 392.75 0.9509 1.7337
0.00 0.29269 1293.7 0.06935 200.00 398.68 1.0000 1.7274
10.00 0.41449 1260.2 0.04948 213.53 404.40 1.0483 1.7224
20.00 0.57159 1224.9 0.03603 227.40 409.84 1.0960 1.7183
30.00 0.77008 1187.2 0.02667 241.65 414.94 1.1432 1.7149
40.00 1.0165 1146.5 0.01999 256.34 419.58 1.1903 1.7115
50.00 1.3177 1102.0 0.01511 271.59 423.63 1.2373 1.7078
60.00 1.6815 1052.4 0.01146 287.49 426.86 1.2847 1.7031
70.00 2.1165 995.6 0.00867 304.29 428.89 1.3332 1.6963
80.00 2.6331 927.4 0.00646 322.41 429.02 1.3837 1.6855
90.00 3.2445 836.9 0.00461 343.01 425.48 1.4392 1.6663
100.00 3.9721 646.7 0.00265 374.02 407.08 1.5207 1.6093
*101.03 4.0560 513.3 0.00195 389.79 389.79 1.5593 1.5593
Ill / Thermodynamic Properties of Selected Materials 407

Table III.5 Refrigerant 134a Properties (cont'd)

Spec. H e a t ' CP Vanor Sound Speed Viscosity Thermal Cond. Surface

Temp. Press. kJ/(kg K) c
x
m/s /xPas mW/(m K) Tension
°C MPa Liquid Vapor Vapor Liquid Vapor Liquid Vapor Liquid Vapor mN/m
**-103.30 0.00039 L.147 0.585 1L.163 1135. 127. 2186.6 6.63 28.15
-80.00 0.00369 L.211 0.637 L.151 999. 134. 1109.9 7.57 — — 24.11
-60.00 0.01594 L.220 0.685 ]1.146 904. 139. 715.4 8.38 121.1 — 20.81
-40.00 0.05122 L.243 0.740 1.148 812. 144. 502.2 9.20 111.9 8.19 17.66
-30.00 0.08436 L.260 0.771 1.152 765. 145. 430.4 9.62 107.3 9.16 16.13
-26.07 0.10132 1.268 0.784 L.154 747. 146. 406.4 9.79 105.4 9.52 15.54
-20.00 0.13268 1.282 0.805 1.157 719. 146. 373.1 10.05 102.6 10.07 14.63
-10.00 0.20052 L.306 0.842 L.166 672. 147. 326.3 10.49 98.0 10.93 13.16
0.00 0.29269 L.335 0.883 1.178 626. 147. 287.4 10.94 93.4 11.79 11.71
10.00 0.41449 1.367 0.930 1.193 579. 146. 254.3 11.42 88.8 12.66 10.30
20.00 0.57159 L.404 0.982 1.215 532. 145. 225.8 11.92 84.2 13.57 8.92
30.00 0.77008 1.447 1.044 L.244 484. 143. 200.7 12.48 79.6 14.56 7.5
40.00 1.0165 L.500 1.120 L.285 436. 140. 178.2 13.10 75.0 15.64 6.27
50.00 1.3177 L.569 1.218 L.345 387. 137 157.7 13.83 70.4 16.84 5.01
60.00 1.6815 L.663 1.354 L.438 338. 132. 138.6 14.71 64.8 18.19 3.81
70.00 2.1165 1.806 1.567 1.597 287. 126. 120.3 15.85 61.2 19.72 2.67
80.00 2.6331 :Z.069 1.967 1.917 235. 118. 102.1 17.46 56.6 21.46 1.63
90.00 3.2445 :2.766 3.064 :>.832 178. 108. 82.6 20.15 — — 0.72
100.00 3.9721 - — 105. 94. 53.0 28.86 — — 0.03
•101.03 4.0560 X) OO (X) 0 0 — — 00 oo 0

** Triple point.
* Critical point.
408 III / Thermo dynamic Properties of Selected Materials

I
APPENDIX IV

Typical Force Constants for the

Lennard-Jones 6-12 Potential

Constants listed are inferred from second virial coefficients.

Atom or Molecule £o/kB (K)

Ar 120 3.41
Ne 35.6 2.75
N2 95 3.70
o2 118 3.58
CO 100 3.76
NO 131 3.17
CO2 198 4.33
CH4 149 3.78
N2O 189 4.59
CF4 153 4.70
CH 2 =CH 2 199 4.52
C2H6 243 3.95
C3H8 242 5.64
rc-C7H16 282 8.88

^Source: Hirschfelder, J. O., Curtiss, C. E, and Bird, R. B., Molecular Theory of Gases and Liquids,
John Wiley, New York, 1954.

409
 Index

accommodation coefficient, 353-354 Debye cutoff frequency, 234

acentric factor, 290 Debye model, 233-238
activity, 281 Debye temperature, 235-236
activity coefficient, 283 degeneracy, 12, 23-25
additivity property, 38, 52 degrees of freedom, 145
affinities and fluxes, 299-304 dense gases, 182
associated Legendre polynomials, 16 dew point curve, 274
azeotropic mixture, 276-277 diffusion thermo effect, 304
dilute occupancy, 48
Baimer series, 19 disjoining pressure, 367-375
Boltzmann constant, 37 dissociation energy, 158
Boltzmann distributions, 37 distiguishability, 32, 43-49, 137, 140-143
Boltzmann statistics, 30, 142, 209 distribution function, 184
Boltzmann transport equation, 337-346 Dufour effect, 304
Born-Oppenheimer approximation, 149
Bose-Einstein condensation, 219-221 Einstein fluctuation formula, 306
Bose-Einstein statistics, 137 Einstein's model, 228-229
boson gases, 212, 218-221 Einstein temperature, 228
bosons, 26 electron gas model, 214-217, 238-241
bubble point curve, 274 electron transport, 311-312, 375-384
electronic partition function, 147
canonical ensemble, 108-117 energy levels, 107
canonical partition function, 112, 184, 186 energy minimum principle, 94
cavitation, 363 energy transport equation, 345
cell theories for liquids, 205 ensemble theory, 30, 108-127
chemical equilibrium in ideal gases, 170-174 enthalpy, 88
chemical potential, 57 enthalpy minimum principle, 95
Clapeyron equation, 266-267 entropy
classical limit, 141-142, 179 and the third law, 241
classical mechanics, 5 definition, 37, 50
Clausius-Clapeyron equation, 267 maximum principle, 51-52, 93
coefficient of thermal expansion, 99 of mixing, 66-67
collision cross section, 332 production, 298
collision processes, 332-334 equations of state, 78-80
combinatorial analysis, 32, 392-394 equilibrium, 53, 71-75, 298
compressibility factor, 281, 284-286 equilibrium constant, 172-174
configuration integral, 183 equipartition of energy, 164-167
correlation moments, 304-309 ergodic hypothesis, 30, 109, 118
critical compressibility, 284, 289-290 Euler equation, 80-81
critical exponents, 291-293 excess properties, 347
critical opalescence, 291 extensive properties, 75
critical temperature, 255 extremum principle, 51
crystal lattice
frequency distribution, 227-228, 235, 237 Fermi-Dirac statistics, 137, 208
primitive basis, 229 Fermi energy, 216, 239
primitive cell, 229-230 fermion gases, 212-217
structure, 229 Fermi temperature, 216, 236, 239-240

411
412 Index

fermions, 26 Joule-Thomson coefficient, 221-222

Fick's law, 304
first Brillouin zone, 231 kinetic coefficients, 301-304, 309-311
first law of thermodynamics, 3, 76 kinetic theory, 325-346
fluctuations, 127, 245-248, 304-311
Fourier heat conduction, 304, 345 Lagrange multipliers, 394-395
fundamental equation Lagrangian, 5
definition, 51 Laguerre polynomials, 18
monatomic ideal gas, 64 lattice vibrations in crystalline solids, 229-234
van der Waals fluid, 187 acoustic branch, 232
fugacity, 281-284 group velocity, 231-232
fugacity coefficient, 283 lattice waves, 230-232
optical branch, 232
gas thermal conductivity, 336 phonons, 232
gas viscosity, 336 Legendre transforms, 83-87
Gibbs-Duhem equation, 80-81 length scales, 1,384-388
Gibbs function, 88 linear systems, 301-304
Gibbs function minimum principle, 95 Liouville equation, 339
Gibbs phase rule, 268-271 liquid spinodal limit, 257-258
grand canonical ensemble, 108, 117-127 local thermodynamic equilibrium, 298
grand canonical partition function, 121 Lorenz number, 382

macrocanonical ensemble, 30
Hamaker constant, 371
macrostate, 29
Hamiltonian, 6
macrostate probability, 33
harmonic oscillator, 13
Massieu functions, 89
Helmholtz potential minimum principle, 94
Maxwell distribution, 325-330
Henry's law, 279
Maxwell relations, 95-99
Hermite polynomials, 14
Maxwell statistics, 61
heteronuclear diatomic molecules, 154
mean free path, 333-334
heterophase fluctuations, 134, 258
mean time between collisions, 333-334
homogeneous nucleation, 258, 357
metastable states, 258
homonuclear diatomic molecules, 156
method of Lagrange multipliers, 394-395
hydrogen atom, 17 microstate probability, 33
microstates, 29, 107
ideal gases molecular energy storage, 1
chemical equilibrium, 170-174 molecular partition function, 138
diatomic, 149-160 molecular simulation tools for liquids, 205
mixtures, 167-174 moments of the Boltzmann transport equation,
monatomic, 145-149 343-346
polyatomic, 160-164 monatomic crystals
indistinguishable particles, 43 general, 225-228
intensive properties, 75 law of Dulong and Petit, 228, 235
interfaces partition function, 227
accommodation coefficient, 353-354 monatomic gas, 61-67
excess properties, 347 mean energy, 62
interfacial tension, 349 speed distribution, 63
molecular transport, 351-357
thermodynamics, 346-351 nonequilibrium systems, 297
Young-Laplace equation, 351 nuclear partition function, 148
intermolecular potential, 184-185
internal energy, 3, 31 Onsager reciprocity, 309-311
irreversible processes, 91
isentropic processes, 93 particle in a box solutions, 11-13
isothermal compressibility, 99 partition functions, 37, 112, 121, 184, 186
Index 413

partial derivative relations, 396 values, 154, 164

partial Legendre transforms, 86 Rydberg constant, 19
partial pressure, 83
phase equilibrium Sackur-Tetrode equation, 149
azeotropic mixture, 276-277 second law of thermodynamics, 76, 93-95
binary mixtures, 271-284 selection rule, 25-26
bubble point curve, 274-277 Schrodinger equation, 8
dew point curve, 274-277 Soret effect, 304
empirical models, 279 specific heat at constant pressure, 99-100
general, 71-72, 280-284 specific heat at constant volume, 99-100
microbubbles in liquid, 357-363 spreading coefficient, 369
microdroplets in vapor, 363-367 Stirling's approximation, 34, 395
ultrathin liquid films, 367-375 supercritical states, 290
van der Waals model for, 256-261, 272-274 symmetry number, 157, 162
phase space, 8, 23, 337
phase stability, 245-254 temperature
criteria for, 254 Celsius scale, 42
phase transitions definitional, 55-56
first and second order, 265 Fahrenheit scale, 42
liquid-vapor, 254-262 Kelvin scale, 42
solid-liquid, 262-264, 271 negative absolute values, 58
phonons, 137,232 Rankine scale, 42
photons, 137 thermal diffusion effect, 304
physical constants, 397 thermal reservoir, 92
potential force constants, 409 thermodynamic potentials, 87
prefix designations, 397 thermodynamic properties
pressure, 56 Debye crystal, 235
probability distributions, 391-392 general, 115-117, 125-127, 140-141
property tables and charts, 399-408 ideal gas, 148-149, 159, 163-164
psychrometric chart, 176, 408 ideal gas mixtures, 168-169
Redlich-Kwong fluids, 198
quantum fluids, 205 van der Waals fluids, 186-187, 195
quantum mechanics, 8-25 thermodynamic similitude, 284-290
quantum statistics, 208 thermoelectric effects
quasiequilibrium path, 77 electrical conductivity, 311-312, 380
quasiequilibrium process, 77 Joule heat, 320
quasistatic processes, 90 kinetic coefficients, 311
microscale aspects, 375-384
Raoult's law, 279 Peltier coefficient, 318
Redlich-Kwong fluid model, 196-202 Peltier effect, 317-319
reduced mass, 150 Seebeck coefficient, 314
reduced properties, 284 Seebeck effect, 312-317
regelation, 271 thermal conductivity, 312, 380-381
relaxation time, 90, 341-343 Thomson coefficient, 320
reversible heat addition, 92 Thomson effect, 319-322
reversible processes, 90-91 Thomson heat, 320
rigid rotor, 15 Wiedermann-Franz law, 382-383
rigid-rotor-harmonic-oscillator approximation, third law of thermodynamics, 77, 241-243
151 time-independent Schrodinger equation, 10
rotational constant, 155 time scales, 1,384-388
rotational partition, function translational partition function, 146
diatomic molecules, 154-157 transport properties, 334
polyatomic molecules, 162 tunnel effect, 27
rotational temperature
definition, 155, 162 uncertainty principle, 8, 20
414 Index

van der Waals constants, 185-186, 188 vibrational temperature

van der Waals equation of state, 79 definition, 153, 161-162
van der Waals model values, 154, 164
binary mixtures, 190-196 viral coefficients, 202, 204
pure fluids, 182-190 viral expansion, 202-204
vapor spinodal limit, 257-258, 275-276
vibrational partition function Wiedermann-Franz law, 382-383
diatomic molecules, 152
polyatomic molecules, 161 Young-Laplace equation, 351