Guest House Reservation Form 06sept2016

Statistical
Thermodynamics
of Alloys
Statistical
Therlllodynalllics
of Alloys
N. A. Gokcen
Albany Research Center
Bureau of Mines
U.S. Department of the Interior
Albany, Oregon
Plenum Press • New York and London

Library of Congress Cataloging in Publication Data
Gokcen, N. A.
Statistical thermodynamics of alloys.
Includes bibliographies and index.
I. Alloys. 2. Statistical thermodynamics. I. Title.
TN690.G673 1986 669'.94 86-6628
ISBN-13: 978-1-4684-5055-2 e-ISBN-13: 978-1-4684-5053-8
DOl: 10.1007/978-1-4684-5053-8
© 1986 Plenum Press, New York

Softcover reprint of the hardcover 1st edition 1986
A Division of Plenum Publishing Corporation
233 Spring Street, New York, N.Y. 10013
All rights reserved
No part of this book may be reproduced, stored in a retrieval system, or transmitted
in any form or by any means, electronic, mechanical, photocopying, microfilming,
recording, or otherwise, without written permission from the Publisher
TO
Altan Gokcen, Pianist

Selma Gokcen, Cellist
with devotion and admiration
Preface
This book is intended for scientists, researchers, and graduate students

interested in solutions in general, and solutions of metals in particular.
Readers are assumed to have a good background in thermodynamics,
presented in such books as those cited at the end of Chapter 1, "Thermo-
dynamic Background." The contents of the book are limited to the solutions
of metals + metals, and metals + metalloids, but the results are also appli-
cable to numerous other types of solutions encountered by metallurgists,
materials scientists, geologists, ceramists, and chemists. Attempts have been
made to cover each topic in depth with numerical examples whenever
necessary.
Chapter 2 presents phase equilibria and phase diagrams as related to
the thermodynamics of solutions. The emphasis is on the binary diagrams
since the ternary diagrams can be understood in terms of the binary diagrams
coupled with the phase rule, and the Gibbs energies of mixing. The cal-
culation of thermodynamic properties from the phase diagrams is not
emphasized because such a procedure generally yields mediocre results.
Nevertheless, the reader can readily obtain thermodynamic data from phase
diagrams by reversing the detailed process of calculation of phase diagrams
from thermodynamic data. Empirical rules on phase stability are given in
this chapter for a brief and clear understanding of the physical and atomistic
factors underlying the alloy phase formation.
Chapter 3 is a condensed and thorough summary of statistical ther-
modynamic methods necessary for pursuing the remaining chapters. Quan-
tum mechanical postulates evoked in deriving a number of useful equations
should not discourage the reader once he accepts the proposition that the
end justifies the means for such postulates. The methods of quantum
mechanics are therefore not always inductive and deductive, but often postu-
lative. Several excellent texts are cited for the reader interested in delving
deeper into statistical and quantum mechanics.
vii
viii Preface
Theories of solutions are presented in Chapter 4 in their general and

simpler forms that are believed to be free of incorrect earlier theories based
on unrealistic assumptions and methods. Additional theories, their
expansion, modification, and application, are given in the remaining chap-
ters. Abstract models such as cell theories have been left alone, as perhaps
they should be. Permutations of single bonds and half bonds are left to
their demise because they yield results inconsistent with the actual numbers
of configurations, which are enumerable for two-dimensional crystals. Addi-
tional results for the actual numbers of configurations in highly ordered
structures are necessary to obtain a firmer basis for the order-disorder
phenomena that are discussed in Chapter 5. However, the existing results
are already based on realistic numbers of configurations, as discussed in
detail in Chapter 5.
Interstitial solutions play important roles in theories and applications
of phase equilibria. Therefore, Chapter 6, the longest chapter in the book,
is devoted to this topic, and also contains computer calculations of the
Fe-C phase diagram. The temptation was great here to delve into hydrogen
storage in metals and alloys because in less than a quarter of a century a
considerable portion of our energy will be based on hydrogen generated
by solar energy. The Wagner theory on ternary interstitial solutions is
presented here rather than in Chapter 4. Unfortunately, it is applicable only
to interstitial or near-interstitial solutes dissolved in ideal binary metal
solutions. Progress in this field is necessary to account for large deviations
from symmetrical behavior in binary and dilute multi component solutions.
Chapter 7 deals with semiconductors within the area of expertise of
the author. A new interpretation of the thermodynamic behavior of electrons
and holes is given in this chapter. It was enticing to delve into p-n junctions
in general, and solar cells in particular. A long section on tandem solar
cells is presented to show that nearly perfect conversion of solar energy
into electrical energy is possible within the limitations imposed by the
second law of thermodynamics. The discussion of solar energy conversion
here and that of hydrogen storage in Chapter 6 complement each other in
the current atmosphere of intense research effort devoted to renewable
sources of energy.
Appendixes A and B contain recent and evolving theories of alloy
phase formation, and methods of estimation and correlation of thermody-
namic properties. Appendix C deals with the correlation of thermodynamic
properties in dilute solutions. The remaining appendixes contain selected
properties of the elements, selected phase diagrams and their thermody-
namic properties, a comprehensive list of symbols, and the periodic table.
The effort to write this book has been long and arduous. I am grateful
for the patience, tolerance, and endurance of my family during the long
Preface ix
hours that I spent reading numerous publications, writing, and correcting

the manuscript. I am indebted to several U.S. Bureau of Mines employees,
in particular Mrs. Sharon L. Brittain for her careful preparation of the
manuscript, Mr. George E. Daut for his meticulous proofreading, and Mr.
J. L. Wilderman, who kindly traced most of the figures as well as the
periodic chart at the end of the book.
Professor Marvin C. Y. Lee, formerly of the Beijing Institute of Aero-
nautics and Astronautics, People's Republic of China, and currently a
member of the research staff of the U.S. Bureau of Mines, kindly read the
entire manuscript and made valuable suggestions. He also read the proofs
and made the text as error-free as humanly possible. Very useful comments
and criticisms were offered on Chapter 6 by Professor Taiji Nishizawa of
Tohoku University, and on Chapter 7 by Professor Melvin Cutler of Oregon
State University.
Expert and invaluable help on my summary of the Engel-Brewer
Theories in Appendix A was very generously offered by Professor L. Brewer
of the University of California. Dr. A. R. Miedema of Philips Research
Laboratories in Holland kindly read and corrected Appendix B on the
estimation of the enthalpy of alloy formation, a technique that was
developed by his diligent group in Eindhoven. Finally, I am grateful to the
staff of Plenum Press for their outstanding effort in publishing this book ..
N. A. Gokcen
Albany, Oregon
Contents
CHAPTER 1 THERMODYNAMIC BACKGROUND

Introduction ........................................ 1
Consequences of Laws of Thermodynamics ............. 3
Helmholtz and Gibbs Energies ........................ 5
Partial Molar Gibbs Energy. . . . . . . . . . . . . . . . . . . . . . . . . . . 6
Activity... . .... ............. ........................ 7
Variation of Activity with Pressure and Temperature ..... 10
Depression and Elevation of Freezing Point. . . . . . . . . . . . . 11
Molar Properties. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Excess Molar Properties. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
Analytical Forms of Excess Partial Molar Properties ..... 15
Determination of G e and G~ of Ternary Systems. . . . . . . . . 23
Interaction Parameters. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
Other Definitions of Partial Molar Gibbs Energy. . . . . . . . . 27
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
CHAPTER 2 PHASE EQUILIBRIA AND PHASE DIAGRAMS

Single-Component Equilibria. . . . . . . . . . . . . . . . . . . . . . . . . . 29
Multicomponent Equilibria........................... 31
Phase Rule. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
Phase Diagrams ..................................... 36
Erroneous Diagrams ................................. 40
Lever Rule. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
Molar Gibbs Energy of Mixing-Phase Diagrams. . . . . . . . 43
tlG Diagrams for Other Phases. . . . . . . . . . . . . . . . . . . . . . . . 44
tlG Diagrams for Complex Systems. . . . . . . . . . . . . . . . . . . . 49
Calculation of Phase Diagrams from Thermodynamic Data 51
Empirical Rules on Phase Stability. . . . . . . . . . . . . . . . . . . . . 53
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
xi
xii Contents
CHAPTER 3 STATISTICAL THERMODYNAMICS

Distribution of Independent Particles. . . . . . . . . . . . . . . . . . . 61
Fermi-Dirac Statistics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
Bose-Einstein Statistics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
Boltzmann Statistics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
Distribution Laws. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
Entropy and Related Properties. . . . . . . . . . . . . . . . . . . . . . . . 69
Fermions, Bosons, and Boltzons .. . . . . . . . . . . . . . . . . . . . . . 71
Gibbsian Ensembles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
Vibrations of Atoms in Lattices. . . . . . . . . . . . . . . . . . . . . . . . 76
Potential Energy Functions. . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
CHAPTER 4 THEORIES OF SOLUTIONS

Regular Solutions-Zeroth Approximation. . . . . . . . . . . . . . 81
Preliminary Concepts. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
Distribution of Molecules. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
Equations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
First Approximation ................................. 101
Approximate Equations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
Dilute Binary Regular Solutions ....................... 106
Critique of Treatments Based on Permutation of Bonds. . . 107
Ternary Regular Solutions ............................ III
Regular Associated Solutions. . . . . . . . . . . . . . . . . . . . . . . . . . III
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
CHAPTER 5 LONG-RANGE ORDER

Ordering and Clustering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
Order-Disorder in Binary Alloys. . . . . . . . . . . . . . . . . . . . . . . 119
Long-Range Order Parameter ......................... 121
Gorsky, and Bragg and Williams (GBW) Approximation.. 122
Heat Capacity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
More General Cases ................................. 127
First Approximation ................................. 130
Enthalpy and Heat Capacity .......................... 139
Comments on Previous Approximations. . . . . . . . . . . . . . . . 141
Unequal Numbers of Atoms and Equal Numbers of Sites 143
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
CHAPTER 6 INTERSTITIAL SOLUTIONS

Introduction ........................................ 149
Pd-H System. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
Statistical Treatment, Zeroth Approximation ............ 154
Solid Interstitial Solutes .............................. 158
Application to Pd-H System. . . . . . . . . . . . . . . . . . . . . . . . . . 160
Contents xiii
Other Hydrogen-Metal Systems ....................... 163

Hydrogen Storage. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
Hydrogen Storage Metals and Alloys. . . . . . . . . . . . . . . . . . . 167
Ti, Zr, Hf and Their Alloys ........................... 168
Fe-C System. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
Wagner Model for Ternary Interstitial Alloys. . . . . . . . . . . . 181
Application of Equation (6.106) ....................... 187
Limitations of Wagner Mode!". . . . . . . . . . . . . . . . . . . . . . . . . 189
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
CHAPTER 7 SEMICONDUCTORS
Introduction ........................................ 195
Distribution of Electrons ............................. 199
Motion of Electrons and Holes. . . . . . . . . . . . . . . . . . . . . . . . 200
Dopants in Semiconductors. . . . . . . . . . . . . . . . . . . . . . . . . . . 202
Energy of Electrons and Holes ........................ 203
Concentration of Electrons. . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
Concentration of Holes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
Equilibrium Constant for Charge Carriers. . . . . . . . . . . . . . . 206
Correlation of Band-Gap Data. . . . . . . . . . . . . . . . . . . . . . . . 211
Derived Thermodynamic Properties. . . . . . . . . . . . . . . . . . . . 213
Ionization Equilibria. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
Fermi Energy in Doped Semiconductors. . . . . . . . . . . . . . . . 216
p-n Junctions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
Effects of Applied Current on p-n Junctions. . . . . . . . . . . . 218
Solar Cells. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
Simple Equivalent Circuit. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
Tandem Solar Cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
Calculation Procedure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
APPENDIX A ENGEL-BREWER THEORIES

Basic Principles .................................... 239
Nonintegral Electronic Configurations. . . . . . . . . . . . . . . . . 241
Transition Elements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242
Correlation of Bonding Energies. . . . . . . . . . . . . . . . . . . . . . 243
Electron Concentration Ranges in Alloys .............. 246
Effect of Pressure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
Multicomponent Diagrams. . . . . . . . . . . . . . . . . . . . . . . . . . . 251
Concluding Remarks. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
APPENDIX B ESTIMATION OF ENTHALPY OF ALLOY

FORMATION
Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
Empirical Coefficients .............................. 258
xiv Contents
Calculation of flH. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262

Alloys of Hydrogen ................................ 267
Ternary Hydrides .................................. 269
Ternary and Muiticomponent Alloys. . . . . . . . . . . . . . . . . . 272
Concluding Remarks ........ . . . . . . . . . . . . . . . . . . . . . . . 273
APPENDIX C CORRELATION OF THERMODYNAMIC

PROPERTIES IN DILUTE SOLUTIONS
Correlation of H~ and 57 in Binary Metal-Metal Systems 277
Estimation of flH· and flS· for Dilute Solutions of Oxygen
in Metals. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281
APPENDIX D SELECTED PROPERTIES OF THE ELEMENTS.... 287
APPENDIX E SELECTED BINARY PHASE DIAGRAMS AND

BINARY THERMODYNAMIC PROPERTIES........ 293
APPENDIX F LIST OF SyMBOLS............................... 311
APPENDIX G PERIODIC TABLE (Mendeleev Table of the Elements) 317
INDEX.......................................................... 321
Statistical
Thermodynamics
of Alloys
1
Thermodynamic Background
Introduction
The areas of thermodynamics with which this book is concerned are those
dealing with the interrelationships of energy and the changes in properties
of matter upon mixing various components to form various phases. A
component in this book is usually a pure element, but occasionally it may
be an intermetallic or a metal-metalloid compound of well-defined
stoichiometry. A single phase is a substance that has uniform physical and
compositional properties. At a phase boundary these properties change
abruptly.
The concepts of thermodynamics are based on empirical observations
of the macroscopic properties of matter and the results of these observations
are expressed in mathematical functions. Statistical thermodynamics of
solutions strives to obtain thermodynamic relationships based on the
molecular behavior of matter. The mathematical background needed for
our purposes will be developed as we proceed; a detailed background may
be found elsewhere.1.2 A basic knowledge of thermodynamics, as presented
in Refs. 1, 3, or 4, is essential in pursuing the subject matter in this book.
We shall define a term as representing a clear and concise phenomenon,
property, or concept, and designate it by a symbol. All the symbols in this
book are defined as they occur in the text, and are compiled and redefined
in Appendix F. A thermodynamic system, or briefly a system, is the substance
under investigation, separated from the surroundings by rigid or movable
boundaries that mayor may not exchange energy or matter or both with
the surroundings. The composition of a system is usually expressed in mole
fractions, Xi, but the molality, i.e., moles per kilogram solvent, is used for
solutions of electrolytes, and particles per cubic centimeter, for semiconduc-
tors. An extensive property {6 is a function of two of the three variables
2 Chapter 1
consisting of pressure P, volume V, and temperature T, as well as the number

of moles n i for an open system; thus,
{2J = {2J(P, T, n h n2 , ••• , n;) (1.1)
where there are i chemically non-compound-forming species. For brevity

we shall occasionally write n for all the components. The relationship
correlatng P, V, and T is called an equation of state. An intensive property
Table 1.1. Physical and Chemical Constants Q
Name Symbol Value and units
Fundamental constants:
Ice + water + vapor point
(triple point of water = 0.0100°C) 273.1600 K
Molar volume of perfect gas
we, 1 atm) 22,413.83 cm' mole-I,
Avogadro number 6.022045 x 1023 mole-I,
(molecules mole-I)
Gas constant R 8.31441 J mole-I K- ' ,
1.98719 cal mole-I K- ' ,
82.05684 cm 3 atm mole-I K -I,
Boltzmann constant k = R/ No 1.380662 X 10- 23 J K- '
Faraday constant F 96484.6 e g-equivalenC ' ,
23,060.4 cal
Velocity of light in vacuum , V-I g.equivalenC '

2.997925 x 108 m sec-I
Planck constant h 6.626189 x 10- 34 J sec
Proton charge (electrtonic charge) e+ 1.60219 x 10- 19 e
Permittivity of vacuum eo = 107/(47T,2) 8.85419 x 10- 12 e
m- I V-I
Defined constants, and conversion factors:

Thermochemical calorie cal 4.1840 J
Standard gravity gO 9.80665 m sec- 2
Atmosphere of pressure atm 101,325 oN" m- 2 (Pascals)
Torricelli of pressure Tor (or Torr) 1/760 atm
Newton oN" 105 dynes
Joule J 107 ergs
Milliliter ml 1.000028 cm 3
Electron volt eV 23,060.4 cal mole-I
= 96,484.6 J mole-I
Kelvins (formerly degrees Kelvin) K 273.1500 + °e
aFrom D. H. Whiffen, Manual of Symbols and Terminology for Physiochemical Quantities and Units, 1979
Ed., IUPAC, Pergamon Press (\979). See also Dimensions/ NBS, January 1974, a pamphlet by NBS.
Thermodynamic Background 3
G may be expressed by
(1.2)
where there are i - I composition variables because LXi = 1, and one of

the composition variables is not independent. We shall write L without its
limits for brevity when the subscript i, as in Xi, is used as the running index
for components. A single phase is defined to be homogeneous, in which P,
T, and Xi are uniform throughout. A process or a change in a thermodynamic
property is denoted by ~, which, for any property such as G,signifies that
MlJ = G(final) - G(initial) = G 2 - G\. The symbol ~ will always refer to
the final property minus the initial property. Equations (1.1) and (1.2) are
for any thermodynamic property in this section. The physical and chemical
constants useful in expressing various thermodynamic properties are listed
in Table 1.1.
Consequences of Laws of Thermodynamics
The first law of thermodynamics is a special consequence of the law

of conservation of energy in physics. We enunciate it by stating that there
is a single-valued function E, called the energy of a system, E = E (V, T, n),
so that when E changes from E\ = E( VI> TI> n) to E2 = E( V2, T2, n), then
~E = E2 - E\ is always the same irrespective of the process used in going
from (VI> TI> n) to (V2 , T 2 , n). For a closed homogeneous system in which
ni are not variables, we write
dE =. dq - P dV + dW + dW' ... (1.3)
where q is the thermal energy, -PdV is the work of compression or

expansion, and W, W', ... are other types of work. At constant volume, if
other types of work are absent for a closed system, dEv = dqv, and the heat
capacity at constant volume is (aEjaT)v = Cv, and therefore
dE = Cv dT - PdV; [Cv = (aEjaT)v] (1.4)
Thus, E is a function of T and V, hence E = E(T, V), because T and V

are present as dT and dV in equation (1.4). We add PV to both sides of
equation (1.3) without W, W', ... to write
d (E + PV) = dH = dq + V dp; (H = E + PV) (1.5)

4 Chapter I
where H == E + PV is the enthalpy by definition. At constant pressure,

dHp = dqp, and the heat capacity at constant pressure, CP' is defined by
Cp == (aH /aT)p = (aq/aT)p- Substitution of Cp in equation (1.5) gives
(1.6)
where H = H(T, P); therefore, H is a function of the more convenient

variables T and P than the variables T and V for E in equation (1.4).
The existence of entropy is an important consequence of the second law
of thermodynamics. The entropy S is given by
dS = dq(reversible) = dE + PdV (1.7)

T T
Consequently, the equations for dE and dH are
dE = TdS - PdV (1.8)
dH = d (E + PV) = T dS + V dP (1.9)
Comparison of equations (1.6) and (1.9) shows that for pure condensed
substances in their stable states at P = 1 atm = 101,325 Pa, the standard
entropy SO is calculated from
SO-SO(T=O)= i
o
T Co
-P-dT
T
(LlO)
where SoC T = 0) is taken to be zero for a pure stable single component in

a stable crystalline form on the basis of the third law convention. The third
law requires that So approach zero about as fast or faster than T approaches
zero. Some noncrystalline but annealable pure components also obey the
third law convention. If there are phase transitions at T\ and T2 with AH\ °
and AH2 °, equation (LlO) must be modified into
So = fT,
o
C °dT + (AH
-p-
T
-_\-
T\
0) + fT -p
T, T
__ 0) + fT -p-
C/o dT + (AH
2 CliO dT 2_
T2 T2 T
(Ll1)
where Cp 0, C~ 0, and C;o are the heat capacities in the temperature ranges
indicated by the limits of integration.
Helmholtz and Gibbs Energies
Equation (1.8) can be modified by subtracting d (TS) from dE; thus,
d(E - TS) = dA = -SdT- PdV; (A = E - TS) (1.12)
where A = E - TS is the Helmholtz energy, formerly called the work

function. Likewise, subtraction of d (TS) from dH in equation (1.9) gives
d(H - TS) = dG = VdP - SdT; (G= H - TS) (1.13)
where G = H - TS is the Gibbs energy. For an ideal gas at a fixed tem-

perature,
dG = V dP = RTd In P (1.14)
from which, by integration,
G = GO( P = 1) + RT In P (1.15)
where GO is the standard Gibbs energy at standard pressure so that GO is

a function of temperature only, i.e., GO = GO( T).
If a gas is not ideal, then a corrected pressure, called the fugacity, F,
is used to obtain the correct values of G:
dG = RTd InF; or G = GO + RT In F (1.16)
This is the definition of fugacity wherein GO refers to the unit fugacity, and
again GO is a function of T only. For example, if the equation of state is
PV = ZRT where Z is the compressibility factor, and if for moderate
pressures Z = 1 + aP, with a as a function of T, i.e., a = a(T), then the
substitution of V = ZRTI P = (1 + aP)RTI Pin dG = VdP yields dG =
RT(d In P + a dP); hence, with dG = RTd In F, we obtain din (FI P) =
a dP, and since for P ~ 0, FI P = 1, the integration of this equation from
P ~ 0 to higher values of P yields
In(FI P) = aP = Z-1 (1.17)
For example, at 300 K and toO atm, the correct value of F for hydrogen is
106.3 and equation (1.17), with Z = 1.061 at 100 atm, also yields 106.3. At
much greater pressures of hydrogen, e.g., hydrogen released from metals
and alloys into confined cracks or crevices, it is necessary to use Z as a
power series in P and carry out the foregoing procedure in detail to obtain
the values of F.
6 Chapter 1
Partial Molar Gib~s Energy
If a function f2f = f2f(P, T, nl> n2, n3) is homogeneous with respect to

its variables nt, n 2 , n 3 , then by the definition of homogeneous functions,
(1.18)
where a is the degree of homogeneity of f2f in n h n2, n3. For example,

f2f = ni + n~ is a second-degree homogeneous function, but f2f = nl + n~ is
not homogeneous. For extensive thermodynamic properties, a is one, but
f2f divided by nj or (nl + n2 + n3) is an intensive property so that G =
f2f/(n l + n2 + n3) is a molar property and a is zero for G; hence,
(1.19)
G is then a function of P, T, and two of the three independent composition

variables such as XI and X2. For example, f2f = AnI + Bn2, with A and B
as constants, is a homogeneous function of the first degree in n l and n2 but
G = f2f/(n l + n2) = A[nl/(nl + n2)] + B[n2/(nl + n2)] is a homogeneous
function of the zeroth degree. The coefficients of A and B are XI and X2,
respectively, and since Xl + X2 = 1, it is evident that G is a function of either
XI or X2, e.g., G = AXI + B(1 - XI). An interesting property of an extensive
thermodynamic function f2f is that
( 1.20)
where a = 1, since the degree of homogeneity of f2f is unity for an extensive

thermodynamic property. Equation (1.20) is the formal statement of the
Euler theorem on homogeneous functions as can be verified by using an
example such as f2f = Ani + Bn~ (for proof see Ref. 1). The total differential
of the extensive Gibbs energy f2f = f2f(P, T, nt, n2, n3) at constant P and Tis
(1.21)
If, purely for brevity, we write
( 1.22)
and call OJ the partial molar Gibbs energy of i, then equations (1.20) and
(1.21) respectively become
(1.23)
(1.24)
It is sufficient to write OJ as G j, but the use of a bar in OJ is an established

practice. Differentiation of equation (1.23) yields dG = I nj dOj + I Ojdn j,
and comparison with equation (1.24) yields the Gibbs-Duhem relation, i.e.,
(1.25)
Division of equations (1.23) and (1.25) by I nj yields. two important and

convenient equations:
(1.26)
( 1.27)
Equations (1.25) and (1.27), first derived by Gibbs (1875) and later indepen-
dently by Duhem (1886), are called the Gibbs-Duhem relations. It should
be noted that n j (i = 1,2, ... ) are the variables of state when a phase is open.
It is important to bear in mind that equations (1.18) to (1.27) are also
applicable to other thermodynamic properties such as E, H, S, and Cpo
Activity
For a pure liquid or solid (crystalline) phase, the Gibbs energy

[Gj(l, or c)] is equal to the Gibbs energy of its vapor Gj(g) at equilibrium;
hence, from equation (1.16),
G~(l, or c) = Gj(g) = G~(g) + RT In Ft (1.28)
where Ft refers to the fugacity over the pure condensed phase, and G7(g)
is a function of temperature only, i.e., G~(g) = G~( T) and G~(l, or c) =
G~(P, T). However, GW, or c) is so weakly dependent on P that for most
purposes this dependence for condensed phases will be generally ignored
at ordinary pressures. When the liquid or the solid is a solution containing
component i as one of the components, then
(1.29)
8 Chapter 1
where Fi without the superscript * is the fugacity over the solution. Solving
for G~(g) from equation (1.28), i.e., G~(g) = GW, or c) - RT In FT and
eliminating G~(g) in equation (1.29) leads to
Oi(l, or c) = G~(l;'or c) + RT In F;/ FT (1.30)
The ratio FJ FT, representing the fugacity of component i over a solution

divided by the fugacity over pure i, is called the activity of i in solution ai ;
therefore,
Oi(l, or c) = G~(l, or c) + RT In ai ; (1.31)
The term G~(l, or c) refers to the standard state, which is the pure stable
condensed phase at 1 atm. Therefore, G~ for brevity without the parenthetic
notations is a function of T only. For the standard state, the activity is
always unity.
The dimensionless quantity that relates the activity to the mole fraction
is denoted by Yi and called the activity coefficient; it is defined by
(1.32)
where the approximate equality is due to the close equality of F;/ FT and
PJ PT at low pressures.
A component in a solution obeys Raoult's law (1887) when its activity
and mole fraction are equal, i.e., a i = Xi, or Yi = 1. It has been observed
experimentally that Raoult's law is obeyed in the limiting case when Xi
approaches unity, i.e., ai ~ Xi or Yi ~ 1 when Xi ~ 1. The usefulness of this
limiting law is that Yi is unity in a small but finite range of composition
approaching Xi ~ 1 because otherwise Yi = 1 is the definition of the standard
state. The activity scale based on the pure component is called the Raoultian
activity.
The activity of component i becomes also proportional to its mole
fraction when Xi approaches zero, i.e., a i = Y~Xi for Xi ~ 0 with y~ becoming
a constant, hence the superscript (0). Combination of this equation with
the definition of activity yields
1>; = Ph~Xi = k'Xi (1.33)
where the last equality is the classical statement of Henry's law (1800), i.e.,
the pressure of i over a dilute solution of i is proportional to its mole
fraction if Fi = Pi and Ft = Pt, with k' as the proportionality constant.
Substitution of a i = Y~Xi in equation (1.31), with G~ == G~(l. or c), gives
Oi = G~ + RTln yO+ RTln Xi == G; + RTln Xi; (Xi ~ 0) (1.34)

where G; is given by
G; = G7 + R T In y7 (1.35)
As Xi increases, the deviation from the Henrian behavior of equation (1.34)

is accounted for by using the activity coefficient j; defined by
( 1.36)
a
where i is called the Henrian activity. Since Oi is independent of the choice
of standard states, Oi in equations (1.31) and (1.36) are equal at any
concentration; therefore,
• a· y.
G~ - G = -RTln~=
I I a; -RTln-'
/; ( 1.37)
The left-hand side is a constant at a given temperature; therefore, the ratio

of activities or activity coefficients of i based on these two different scales
is also a constant independent of composition. The left-hand side, from
equation (1.34), is -RT In y7; hence,
(1.38)
Therefore,
(1.39)
Thus, the actIvIty on one scale differs from that on another scale by a
constant factor at a given temperature when the concentrations are in mole
fractions. The activities and the activity coefficients based on Raoult's law
and Henry's law are summarized as follows:
Standard Reference
Basis Definition state state
Raoult's law p,/ Pt = ai = 'YiXi Pure i, i.e., Unnecessary

a, = I for Xi =I
Henry's law a, = p,/ ph7 = f,Xi ai = I for Xi = I 0 (for
Xi ....
hypothetical which f, = I)
(physically
nonexisting)
It should be noted that the standard state for both activity bases is the state
of unit activity. The state of unit activity for the Henrian scale is physically
10 Chapter 1
A
1.0,---r---,---,-----"
/ B
/
,---r------,;-----,---" 1.0
HENRY'S LAW/
"'v/
/
0.5 /
1.0
/ t
/ a, 05
/ 10'-
/ /
. /'\ 05
0,
/ /RAOULT'S
,b / / LAW
/
/ ____ __ __ ____
O~_~ __ ~ __ ~ __ ~_ 0 o~ ~ ~~ ~ ~o
o 05 1.0 o 05 10
Xz
Figure 1.1. Activity of component 2 in hypothetical binary solutions. (A) Deviation from
Raoult's law is positive; from Henry's law, negative. (B) Deviation from Raoult's law is
negative; from Henry's law, positive. Vertical right scale is for a 2 with standard state as pure
component 2, and vertical left scale for ti2 with reference state x 2 -> O. The activity curve for
component 1 is not shown, but it is similar to that for component 2 with a, = 1 at x 2 = 0, and
a, = 0 at x 2 = 1, and having a shape as dictated by the Gibbs-Duhem relation.
nonexisting, except in the unlikely case when the solution is ideal in the
Raoultian scale. Further, there is no need for the reference state for the
Raoultain scale, but for the Henrian scale, the reference state is the infinitely
dilute solution of i in a solvent. The activities based on these scales are
illustrated in Fig. 1. The deviation from Raoult's law is positive for com-
ponent i when 'Yi is greater than 1, and negative when 'Yi is smaller than 1.
Similarly, the deviation from Henry's law is positive when}; is greater than
1, and negative when}; is smaller than 1. Generally, when 'Yi> 1, then
}; < 1, and vice versa in the vicinity where the deviation from Henry's law
starts.
Variation of Activity with Pressure and Temperature
Equation (1.13) may be written in the same form for either Oi or G~;
hence, at constant T and composition,
a(Oj-On v _v~ = RT aaP

In aj= RT a In 'Yj
aP
(l.40)
aP I I
Both V; and V~ are small for condensed phases, and their difference is even
smaller; therefore, a j and 'Yj for real solutions vary insignificantly at pressure
changes of a few bars. For an ideal solution, 'Yj = 1; hence, \1= V~.
The activity a j varies with temperature according to
Ina·= OJ - O~ Fl; -
H~ Sj - S~
(1.41)
I RT RT R
where OJ = H - TSj and O~ = Hf - TSf. Differentiation of 0;/ T with respect

to temperature gives
( 1.42)
where the second and third terms after the first equal sign cancel out because
aH;/aT= Cpj, and as;/aT= Cp;/T. We write a similar equation for Of and
subtract it from equation (1.42) to obtain
aIn a aIn 'Yj

j Hj -H~ Rain a j
--=--=- or (1.43)
aT aT RT2 , a{l/ T)
For an ideal solution a j = X j and the left side of this equation is zero because
Xj is a variable independent of T in OJ = OJ(P, T, Xl> X2, .. .); hence, Hj =
Hf, and the substitution of this equality in equation (1.41) gives
OJ - Of = RT In Xj = -T(Sj -Sn (ideal) ( 1.44)
Depression and Elevation of Freezing Point
The freezing point of a dilute solution is defined as the beginning or

the onset of freezing. We designate a solvent as component 1, its freezing
point in pure state as Tfn a solute as component 2, and the onset of freezing
of solution as T, so that Tfr - T = .:l Tfr is the depression or the elevation
of the freezing point. We let X 2 be the mole fraction of component 2 in the
binary liquid phase and Y2, the mole fraction in the binary crystal phase.
For dilute solutions, the solvent obeys Raoult's law; hence,
12 Chapter 1
Rearrangement of G~(c) and G~(1) of this equation gives G~(c) - G~(l) ==

HHc) - H~(l) - T[S~(c) - S~(l)] == tlH rr - TtlS rr where tlH rr =
H~( c) - H~(l) is the enthalpy offreezing, and tlS rr = tlH rr / Tcr is the entropy
offreezing, and then the substitution for G~(c) - G~(l) from equation (1.45)
yields
In(1 - X2) - In(1 - Y2) =- rr [tl-Tcr]

tlH- - (1.46)
RTcr T
When X2 and Y2 are very small, the left side becomes Y2 - X2 and further,
T becomes very close to Tcr; therefore,
In 1 - X2 = Y2 _ X2 = tlH rr (tl ~cr) ( 1.47)

1- Y2 R Tcr
If the solubility Y2 is nearly zero, we obtain the familiar equation for the
depression of freezing point for many ambient solutions. For alloys,
however, when Y2 is greater than X2, the melting point is elevated, and when
Y2 is smaller than X 2, the melting point is depressed. Equation (1.45) can
be rewritten for liquid-vapor equilibria and a similar procedure can be
followed to obtain an equation identical in form with (1.47) for the elevation
of the boiling point.
Molar Properties
A solution of c components may be considered as formed according to
Xl(pure 1) +x2(pure 2) + ... ~ 1 mole of solution (1.48)*
For the solution, the molar Gibbs energy is G = L xlij, and for the pure
components, G = L XjG~ , so that the Gibbs energy of formation of solution
tlG for reaction (1.48) is
(OJ - G~ = RTln a j ) (1.49)
Likewise, for tlH, by using equation (1.43), we derive
tlH = ~t... x.(H -

I
-
I
0
H·)
I
aln~
= -RT2 ~t... x·--·
aT '
I
( ii. _ H~
I I
= -RT2 aIn
aT
')Ij)
(1.50)
*For alloys, 1 mole of solution is identical with 1 gram atom of solution.
The corresponding relationship for !lS is obtained by substituting equations

(1.49) and (1.50) into!lS = (!lH - !lG)/T:
!lS aIn 'Yi) .

+ Tx·--
x·(S· - S·) I.. x·ln
= -R '" a·I
- 0 (
= I..
'" I I I I aT '
I
- 0 aIn 'Yi)
S--S·=-R Ina·+T-- (
(1.51)
I I aT I
Generally, the accuracy of data on 'Yi for alloys does not justify a rigorous
partial derivative in equations (1.50) and (1.51); instead, a In 'YJaT::::
!lIn 'YJ!l is often used for solutions of metals to obtain !lH and !lS. A
much more accurate procedure is to obtain !lH by calorimetry at various
concentrations and then to calculate !lS from the data on a i by using
!lS = R L Xi In a i - (!lH/T). Figure 1.2 illustrates !lG, 0 1 - G~, O2 - G~,
!lH, and !lG for the binary system AI (component 1) and eu (component
2) obtained by plotting the data compiled by Hultgren et al. 5
0:
"-
<f>
<I
Lo
-4
o
~
Ci
U
.:£
-8
Figure 1.2. Selected thermodynamic

properties of liquid AI-Cu system at
1373 K. Subscript .l is for AI; 2, for·Cu.
Curves for t::.G, 0, - G'I, and O2 - G 2 -12 '--_..&._ _......_--I._...L........_ - - '
coincide where t::.G is minimum. Lower o 0.2 0.4 0.6 0.8 1.0
left vertical scale is for t::.H, t::.G, and
Oi-G~.
14 Chapter 1
The foregoing equations can be rewritten for an ideal solution by

observing that 'Yi = 1, independent of temperature and composition; there-
fore, for reaction (1.48),
dH(ideal) = 0; Ul;(ideai) - H~ = 0] ( 1.50a)
[Si(ideaI) - s~ = -R In x;] (1.51a)
Excess Molar Properties
The activity coefficient is a measure of deviation from ideality; there-

fore, certain thermodynamic properties related to the activity coefficients
are called excess thermodynamic properties as introduced by Scatchard
(1932). The excess Gibbs energy G e and the excess partial molar Gibbs
energy G~ are defined by
G e = G(reaI) - G(ideaI);
From equations (1.31) and (1.32) with 'Yi =;t. 1 for real solutions, and with
'Yi = 1 for ideal solutions, we derive
6~ = RT In a i - RT In Xi = RT In 'Yi (1.53 )
and with equation (1.26),
G e == He - TS e = RTI Xi In 'Yi (1.54)
We found earlier that Hi - H~ was zero for an ideal solution in conjunction

with equations (1.43) and (1.44), or Hi = H~; therefore, Hi(real) - Hi(ideal)
is the same as H;(real) - H~; and from equations (1.43) and (1.50),
He = fi _ HO = _ RT2 aIn 'Yi ( 1.55)

I I I aT
He = I - e =dH=-RT2 I xa·In- -'Yi

xH ~ (1.56)
I I aT I
ThermodYDamic BackgrouDd IS
where the molar enthalpy of formation tJ.H is the same as the excess enthalpy
of solution, He. Equations (1.53)-(1.56) and G~ = fi~ - TS~ lead to
S-ei = - R In 'Yi - a In 'Yi

R T InaT (1.57)
Se = - R ~
L..
(
x· In 'Y'
I I
aaT
In 'Yi)
+ Tx·--
I
(1.58)
Again a much more accurate procedure is to obtain He = tJ.H and fi~ by

calorimetry at various concentrations and then use the activity coefficient
in se = (He / T) - R L Xi In 'Yi, and S~ = (fie / T) - R In 'Yi to calculate Se
and S~.
Analytical Forms of Excess Partial Molar Properties
Binary Systems
The extensive excess Gibbs energy (lJe = (lJe(p, T, n h n2) for a binary
system is related to G~ by equation (1.22), i.e.,
(1.59)
where P, T, and n2 are explicitly shown as held constants during differenti-

ation. It is convenient to use the molar property Oe instead of the extensive
property (lJ by using (lJe = (nl + n2 )Oe in equation (1.59) to obtain
(1.60)
Differentiation of Xl = nl/(n l + n2) at constant n2 gives
Substitution of X2/ dXl for (n l + n2)/ dn l from this equation into equation
(1.60) gives
(1.61)
16 Chapter 1
Since G e is a function of one composition variable, e.g., Xl for a binary

solution, it is independent of n 2 ; therefore, the restriction required by the
constancy of n2 is unnecessary; hence,
(1.62)
where 1 - Xl has been substituted for X 2 to obtain symmetry in the subscripts.

The equation for G~ is obtained by replacing the subscripts 1 with 2 in
equation (1.62), i.e.,
-e
G2 = G e+ (1 - )(aGe)
X2 -- ( 1.62a)
aX2 P, T
Compiled values 5 of G e , G~, and G~ for AI(subscript l)-Cu(subscript 2)

at 1373 K are plotted in Fig. 1.3. All three curves intersect one another
-2
o
E
:; -4
u Figure 1.3. Excess molar Gibbs
-'" energy, G e , and excess partial molar
Gibbs energies, G~ and G~, for Al-Cu
system at 1373 K. Subscript 1 is for AI;
2, for Cu. The tangent line on the curve
-6 for G e at X 2 = 0.4 intersects the vertical
coordinate at X 2 = 0 to yield G~, and
at X 2 = 1 to yield G~, as shown by the
dashed lines with arrows.
Note that G~ is equal to G e plus
(1 - x 2 ) times the slope, aG e /ax 2 , as
required by equation (1.62a), and
0.2 0.4 0.6 0.8 1.0 similarly for G.. but the slope in this
case is positive because aGejax\ =
X2 -aG e /ax 2 •
where G e is a minimum as required by equations (1.62) and (1.62a). When

G e (or other excess molar properties) is plotted versus X 2, the method of
tangent-intercepts can be used to obtain G~ and G~ as indicated in Fig.
1.3, but this method is now obsolete because a polynomial data fitting on
a computer is superior (see Ref. 1). Equations (1.59) through (1.62) are
general in the sense that G, H, S, or other related thermodynamic properties
may be substituted for G e to obtain the corresponding equations for them.
The power series most often used for alloys are the Margules equation in
which e is the order of series; thus,
e
G e = xlxAal + a2x2 + a3x~ + ... + a3x;-I) = I AqX2(1 - xrl) (1.63)
q=2
where the relationships between ai and Aq can be obtained by eliminating
XI and setting the coefficients of x;
equal for equal values of i on the left
and on the right sides; however, the result is of no great importance for
our purposes, and we shall use only the equations involving Aq as follows:
G~ = I (q - 1)Aqx~ ( 1.64)
q=2
G~ = I Aq[l - qx~-I + (q - 1)xi] (1.65)

q=2
The boundary conditions for the Raoultian activity scale require that for
XI = 1, G~ = 0, and for X2 = 1, G~ = 0, and therefore G e is zero for XI = 1
or X2 = 1. Traditionally, G~ is often written in symmetry with G~ so that
e
G~ = I (t - 1)B,x; (1.66)
,=2
The coefficients Aq and B, must satisfy either the Gibbs-Duhem relation

XI dG~ + X 2 dG 2 =°[cf. equation (1.27)] or equation (1.63); therefore,
B - ( 1)' ~ q!Aq (1.67)

,- - ~ t I(
q-' . q
- t) I
.
where the values of t start from 2 and end with e. We stress here that when
any set of power series is selected for G~ and G~, they must satisfy the
Gibbs-Duhem relation, but when a power series satisfying the boundary
conditions is selected for G e , then G~, and G~ derived from such a G e
automatically satisfy the Gibbs-Duhem relation. For e = 6, the results for
B, in terms of Aq from equation (1.67) are
B2 = A2 + 3A3 + 6A4 + lOAs + 15A6

18 Chapter 1
4Bs = -4As - 24A6
(1.68)
It is evident that for e = 2, B2 = A2 and we have
( 1.69)
Another set of forms 6 •7 of Margules equations with e = 4 is
(1.70)
(1.71)
(1.72)
[In general, AI2 is not equal to A 21 ; and the term DJ2XIX2 may be written
as x l xz(D2l x l + DJ2X2 - E\2XIX2) if additional terms are required.] It is clear
that G~ can be obtained from G~ by interchanging the subscripts 1 and 2
and agreeing that DI2 = D21 since D21 does not appear in equation (1.70).
For DJ2 = 0, we have the cubic equations, and for DJ2 = 0 and AJ2 = A2h
we have the familiar quadratic equations of the type given by equation
(1.69) since A 21 x I + AJ2x2 = AJ2. Various additional equations are summar-
ized by Hala et al. 6 ; however, no specific equation can be claimed to have
much greater advantages over other equations in accurate representations
of data.
The effect of temperature on G e and G~ is adequately accounted for
by assuming that each coefficient Au, D ij, and so on is a linear function of
temperature, e.g., AI2 = aJ2 - f312 T. The constant term aJ2 times the compo-
sitional variables contribute to He and lif, and f32 times the same variables
contribute to se and S~. This is self-evident since a(G e /T)/a(1/T) = He,
and aGe/aT = _Se, and similarly for lif and S~.
EXAMPLE. Thermodynamic properties of the Mg-Cd system have been

determined by Castanet et al. 8 at various temperatures and concentrations,
by using a galvanic cell for measurements of G~(Mg). The results have been
fitted with the following equation:
G~(Mg) = (-25,413 + 43.949T)x~ + (9494 - 176.000T)x~

+ (-25,457 + 241.507 T)xi
+(22,421 - 106.594T)x~ (Jjmole)
where subscript 1 = Mg and 2 = Cd. Since aG~jaT = -S~, it is evident that

the coefficients of T in each term belong to S~, leaving the preceding
temperature-independent term for FI~. Therefore,
s~ = -43.949x~ + 176.000x~ - 241.507xi + 106.594x~

FI~ = -25,413x~ + 9494x~ - 25,457xi + 22,421x~
This equation corresponds to equation (1.64) with e = 5:
Hence, A2 = -25,413 + 43.549 T, A3 = 4747 - 88.000 T, and so on. It is there-

fore possible to express G 2 = B3X; + ... of equation (1.66) from equation
(1.68); the result is
G~(Cd) = (-6043.5 - 3.522T)xi + (-53,713.7 + 64.9513T)x;

+ (58,621.8 - 158:2205 T)xi + ( - 22,421 + 106.594 T)xi
The corresponding equations for FI~ and S~ can readily be obtained from
this equation. In addition, the equation for the molar Gibbs energy G e can
be obtained by using equation (1.63) from which He and Se can be obtained.
Therefore, the measurements of G~ = RT In 'Yl at various temperatures and
compositions yield G~, FI~, S~, G~, FI~, S~, G e, He. and se.
Ternary Systems
The equations selected for G e of a ternary system must again satisfy
the boundary conditions in that G e = 0 when Xi = 1 (i = 1,2,3). When this
condition is satisfied, the power series beginning with the second-order
terms automatically satisfy the Gibbs-Duhem relation. The derivation of
G~ from G e requires ee e
= (nl + n2 + n3) G and again writing
(1.73)
20 Chapter 1
where n \ signifies that all nj other than n l are held constant during differenti-
ation. The partial differential of XI is
The variable XI is an intensive property that may be written as

(nl/ n3)/[(n l / n3) + n2/ n 3) + 1] where the restriction of n\, or the constancy
of n2 and n3, is simply equivalent to the constancy of ratio r = n2/ n3 = X2/ X3
in terms of mole fractions. Substitution of (1 - XI) / dX I from equation (1.74)
into (1.73) for (nl + n2 + n3)/an l gives
(1.75)
If this process is repeated for a multi component system, the additional

restrictions in equation (1.74) are X3/ X4, X4/ XS, •..• In order to obtain the
analytical equation for G~ for a ternary system, a power series such as the
following equation is needed for Oe:
oe = x l xiA 21 x I + A I2 X2 - D 12 XI X2)
+ XI X3(A 3I XI + A13X3 - D 13 XI X3)

+ x2x 3(A 32 x 2 + A 23 x3 - D 23 X2X3)
+ XIX2X3(A21 + AI3 + A32 - CIX I - C2X2 - C 3X3) (1.76)
For X3 = 0, this equation becomes identical with equation (1.70); likewise,

for X2 = 0, it becomes Oe for the binary system 1-3, and for XI = 0, Oe for
the binary system 2-3. In addition, there are three ternary coefficients C h
C 2 , and C 3 • The order of equation (1.76) is four, i.e., e = 4. To derive the
equation for G~, it is necessary to substitute equation (1.76) into equation
(1.75) and carry out the differentiation with respect to XI' Care must be
exercised in this procedure because complications arise in obtaining the
correct derivatives unless the sum of the mole fractions is written as XI + X2 +
X3 = XI + xir + 1)/r = 1, and XI + (r + 1)X3 = 1, from X2/X3 = r; then from
dX I + [(r + 1)/ r] dX2 = 0, and from dXI + (r + 1) dX3 = 0, the derivatives are
correctly written as
( aX2) = __ X2 . (1.77)
aX3 r X2 + X3'
These equations may be rewritten in the following convenient forms:
(1.78)
The second of these equations can be obtained from the first by replacing
the subscript 2 with 3. The resulting equations for OJ (i = 1,2,3) by this
procedure are as follows:
O~ = x~[AI2+ 2xI(A21 - AI2 - D 12 ) + 3xiD 12 ]

+ xi[A\3 + 2x l (A 31 - A\3 - D\3) + 3xiD\3]
+ x 2x 3[A 21 + A\3 - A32 + 2xI (A31 - A l3 )
+ 2x3(A32 - A 23 ) + 3X2X3D23 - CIX I(2 - 3x l )
- C 2X2(1- 3x l ) - C3X3(1- 3xl )] (1.79)
O~ = x~[A23 + 2xiA32 - A 23 - D 23 ) + 3x~D23]

+ xi[A 21 + 2x2(A12 - A21 - D 12 ) + 3x~DI2]
+ x l x3[A 32 + A21 - A\3 + 2xiA12 - A 21 )
+ 2xI (AI3 - A 31 ) + 3XIX3D\3 - C 2x 2(2 - 3x2)
- C3X3(1-3x2) - C IXI(1- 3x2)] (1.80)
0; = xi[A 31 + 2x3(A\3 - A31 - D\3 + 3x~Dl3)]

+ x~[A32+2xa(A23 - A 32 - D23+3x~D23)]
+ x l x2[A\3 + A32 - A21 + 2x3(A 23 - A 32 )
+ 2xiA21 - A 12 ) + 3x IX2D I2 - C 3x 3(2 - 3x3)
- CIXI(1-3x3) - C 2xil-3x3)] (1.81)
An exchange of subscripts cannot yield O~ from OJ because a circular

rotation of subscripts is necessary. For example, to obtain O~ from O~ it
is necessary to use l(old) -+ 2(new), 2(0Id) -+ 3(new), 3(0Id) -+ l(new), and
to observe that D jj = D jj because there is only one D in equation (1.76).
The patterns of coefficients in equations (1.76) through (1.81) are now
established so that it is a simple matter to add the terms corresponding to
increasing values of E. If it is desirable to check the results, an additional
thermodynamic relationship can be used. For this purpose, it is necessary
22 Chapter 1
to differentiate oe=XlO~+X20~+X30; and use the Gibbs-Duhem rela-

tion Xl dO~ + X2 dO~ + X3 dO; = 0 to get dOe = O~ dX l + O~ dX2 + 0; dx 3.
At a fixed value of X3, dX3 = 0, then dX l = -dX2' and the equation of doe
becomes dOe = - O~ dX2 + O~ dX2; from this equation, it is evident that
( 1.82)
The exchange of subscripts 2 and 3 yields a similar equation for 0;. The
aditional terms in equation (1.76) contribute to 07 as required by equation
( 1.75).
The third-order Margules equation can be obtained from the foregoing
equations by setting Dij = 0 and C l = C 2 = C 3 = C. The results are given here
for convenience and as a check on equations (1.79)-(1.81):
oe = Xl X2(A 2I Xl + A12X2) + Xl X3(A 3I Xl + A 13 X3)

+ X2X3(A32X2 + A23X3) + XlX2X3(A2l + A13 + A32 - C) (1.83 )
O~ = x~[A\2 + 2x l (A 2l - A \2)] + X~[A13 + 2x l (A 3l - A 13 )]

+ X2x 3[A 2l + A13 - A32 + 2x l (A 3l - A 13 )
+ 2x3(A 32 - A 23 ) - C(l- 2x l )] ( 1.84)
O~ = X~[A23 + 2xiA32 - A 23 )] + Xi[A2l + 2x2(A 12 - A 21 )]
+ Xl x 3[A 32 + A2l - A13+ 2X2(A12 - A 21 )

+ 2x l (A 13 - A 31 ) - C(1- 2x 2)] (1.85)
0; = xi[A 3l + 2X3(A13 - A 31 )] + X~[A32 + 2x3(A 23 - A 32 )]

+ Xl x 2[A 13 + A32 - A2l + 2X3(A23 - A 32 )
+2X2(A2l - Ad - C(1- 2X3)] (1.86)
Equation (1.83) contains only one ternary term, C, which must be determined
from ternary data. It is possible to set C = 0 and use only the binary
coefficients to obtain 07 for the ternary alloys with success for liquid alloys,
but marginal results for solid alloys.
The second-order Margules equations can be obtained by setting Al2 =
A2i> A13 = A3i> A 23 = A 32 , and C = 0 in the preceding equations; the results
are
(1.87)
(1.88)
(1.89)
(1.90)
Equation (1.87) contains only the constants from the binary systems; there-
fore, it contains no terms with XIX2X3' The interchange of any set of two
subscripts does not change equation (1.87) since it assumed that Aij = Aji;
therefore, only in such sets of equations, G~ can be obtained from G~ by
interchanging the subscripts 1 and 2, and likewise, G;, by interchanging
the subscripts 1 and 3. Various other types of equations and computational
methods exist, as listed elsewhere in detail. l ,6,9,1O The foregoing equations
are also valid when other molar properties such as llH, IlS, and Se replace
Oe, after which the related partial molar properties can be derived by the
same procedure.
Determination of G e and G~ of Ternary Systems
Determination of the coefficients of the equation for G~ of one selected

component is sufficient for obtaining the equations for the remaining G~,
as well as for Oe as in binary systems. Consider the usual type of measure-
ments for a ternary system in which one of the components, labeled as
component 1, lends itself to convenient and accurate determinations of G~,
e.g., the vapor pressure of Plover the system is accurately measurable from
which G~ = RT In 1'1 = RT In[PI/(xIPt)] where Pt refers to the pure com-
ponent. If, for example, the equation necessary for adequately representing
G~ is that corresponding to e = 4, i.e., equation (1.79), then there are 12
coefficients which require a minimum of six well-spaced ternary values of
G~, the balance being three binary values for system 1-2, and three for
system 1-3 because component "1" can form only two binary systems over
which its partial pressure can be measured; if, however, at least two com-
ponents out of three are volatile so that their vapor pressures can be
measured accurately, a minimum of three data from the ternary system,
and the remaining nine, with three from each binary system, would be
sufficient for e = 4. All the foregoing equations may also be evaluated
entirely from well-spaced ternary data alone. Such ternary data are capable
of yielding the values of oe and G~ for the binary systems, but the data
on the binary systems alone cannot usually yield good values of O~ and
G~ for the ternary system although this approximation appears to yield
fairly good values for a few ternary systems in which the deviations from
the ideality are not large.
24 Chapter 1
The coefficients of equation (1.79) can be determined by computer if

the data are available for the activity coefficient of only one component.
Various computer programs are available lO •11 for this purpose, but the
essential point is simply solving a set of simultaneous equations in which
the unknowns are the coefficients in equation (1.79). If, however, all three
activity coefficients are known from experimental measurements, it is best
to compute G e at each set of compositions by G e = XI G~ + X2 G~ + X3 G~
and then solve for the unknown coefficients in equation (1.76). The foregoing
analytical method is far superior to the graphical integration of the Gibbs-
Duhem equation. 1,11
Interaction Parameters
Equations (1.88)-(1.90) for G~, G~, and G~ become considerably more

reliable as X2 and X3 approach zero, or XI approaches unity. For such dilute
solutions of components 2 and 3 in solvent 1, differentiation of these
equations yields
(1.91)
where the identity signs define eY) called the Wagner interaction parameters
or coefficients, 1,12 e~3) giving the effect of component 3 on G~, and e~2>, that
of component 2 on G~. This differentiation is easily carried out after
eliminating XI in equations (1.89) and (1.90), and then setting X2 and X3 to
zero. To generalize eY) for all values of i andj,including i = j, it is sufficient
to write eli) as (aGUax;); the results are
( 1.92)
(1.93)
For a binary solution of components 1 and 2, X3 = 0 in equation (1.89),

and for X2 ~ 0 or XI ~ 1,
(1.94 )
Equation (1.89), after substituting 1 - X2 - X3 for XI and neglecting the

second- and higher-order terms, gives
= RT In 'Y~ + £ e~i)x;
;=2
(1.95)
where e~n)xn is the nth term in an n-component system. Similarly,

_ n
0; = RTln 'Y3 = RTln 'Yi + L e~i)x; (1.96)
;=2
The Wagner interaction parameters and their generalized form have

been the subject of numerous papers. 13-16 The usefulness of these coefficients
lies in determining 'Y2 for multicomponent dilute solutions from 'Y2 for the
binary and ternary systems. Thus, e~2)! RT is the slope from a plot of In 'Y2
versus X2 for the binary system 1-2 and the linear portion of the plot near
X 2 ~ 0 gives aIn 'Y2! aX2; likewise, e~3)! RT is the slope of In 'Y2 versus X3 at
a selected fixed but low concentration of X2. For example, in a given solution
containing only X 2 = 0.02 as the solute, component 3 can be added in small
increments of 0.01 from 0.00 to 0.05 and then In 'Y2 is plotted versus X3. The
value of e~3) is then obtained from
RT (a In 'Y2) = RT In 'Yiat X3 = 0.05) -In 'Yiat X3 = 0.00) = e~3) (1.97)

aX3 X2 0.05 - 0.00
where it is assumed that the plot in this range is linear. Similar equations
for component 3 can be obtained by interchanging the subscripts 2 and 3.
A more interesting method, when experimentally possible, is to determine
G~ as a function of composition in sufficient ranges of concentration to
determine A;j in equations (1.88)-(1.90) and then use equations (1.91)-(1.93)
to determine eY).
EXAMPLE. The activities a; and activity coefficients and related ther-

modynamic properties for the binary systems of Hg(component 1), Sn(2),
and Zn(3) have been evaluated and compiled by Hultgren et a/. 5 The results
are extrapolated to 673 K by using the listed values of He and Se in
oe = He - TS e by assuming that He and Se are independent of temperature.
For example, the listed value for fig-Sn at 450 K are He = 211, and
se = -0.142 at XI = 0.5; therefore, oe = 211 + 673 x 0.142 =
307 cal! (g-atom of alloy) at 673 K, with a possible error of ± 100 cal! g-atom.
26 Chapter 1
The values of Oe (in calf g-atom) for the three binary alloys are fitted with
equation (1.70) as follows:
Oe(Hg-Sn) = xlx2(2180xl + 1080x2 -1530x l x2) in calfg-atom;

(A 21 = 2180; A12 = 1080; D12 = 1530) (1.98)
(A31 = 850; A13 = 1100; DI3 = 0)

(1.99)
Oe(Sn-Zn) = X2X3(1l60X2 + 2500X3 - 970X2X3);

(A32 = 1160; A 23 = 2500; D 23 = 970) (1.100)
The results for Hg-Zn by Kozin et al.17 are in agreement with equation
(1.99). The foregoing equations could have been represented as functions
of temperature but the ternary data to be used later here are available for
one temperature only,18 i.e., 673 K; therefore, equations (1.98)-(1.100), valid
for 673 K, are adequate for our purposes. The calculations show that the
values of 'Yi for each binary system vary by a factor of approximately 2
from Xi = 0 to Xi = 1; hence, the deviation from Raoult's law is not severe.
The values of a l for Hg in the ternary system Hg-Sn-Zn were deter-
mined by Nigmetova et al. 18 by measuring the vapor pressure of Hg over
various liquid alloys at 673 K. We wish to use equations (1.76) and (1.79)
with C i = 0, i.e., to obtain ternary equations by using the binary coefficients,
in order to test the resulting values with the experimental ternary data. 18
The results for oe and G~ are as follows:
(1.101)
G~ = x~(1080 - 860x l + 4590xi) + x;(1100 - 500xl)

+ x 2x 3(2120 - 500x I - 2680X3 + 2910x2x3) (1.102)
The result from this equation for G~ = RT In 'Yl (at Xl = 0.4, X 2 = 0.3,
X3 = 0.3) is 337.4 cal/g-atom, from which 'YI = 1.287. The experimental
result is 'Yl = 1.285 as read carefully from an appropriate figure by Nig-
metova et al. 18 The agreement is good because the deviation from ideality
is not severe.
Other Definitions of Partial Molar Gibbs Energy
The partial molar Gibbs energy OJ of component i is usually defined

by OJ = (aG/an j )P,T,n"n2,' but there are four additional definitions. The
energy of a closed system, p, underlined to show its correspondence with
G, is a function of entropy and volume since dp = T dS - P dY. For an
open system, the numbers of moles of components are also variables of
state; therefore,
(1.103)
The total differential of energy is then
dp = TdS - PdY + (a p ) , dn, + (a p ) . dn2 + . .. (1.104)

an, s,v,n' an2 s,v,n'
where n' means that the numbers of moles other than that inside the
parentheses are regarded as constants. The total differential of G =
p + pY- TS is
dG = dp + P dY + Y dP - T dS - SdT (1.105)
Substitution of the right side of equation (1.104) for dp in equation (1.105)

gives
dG = YdP - SdT + (a p ) , dn, + (a p ) , dn2 +... (1.106)

an, s,v,n' an2 s,v,n
From this equation, it is evident that an alternative definition of 0, is given

by
-
G-(a~
- ---
,- an, p,T,n' - an,
(aE) s,v,n'
(1.107)
Likewise, starting with ij = f(S, p, n" ... ), S = fW, Y. n" ... ), and .c\ =
f( y. T, n" ... ), and following a similar procedure, the following additional
definitions of 0, can be derived:
-,- (aH)
G - --=-
an, S,P,n' -
( S)
--T ---
an,
---
E,v,n' -
(aA)
an, V,T,n'
(1.108)
28 Chapter I
The definition given by OJ = (aflJ/anJp.T.no, is more convenient than the

remaining definitions because it is a property obtained under conveniently
attainable conditions of constant temperature and pressure. Therefore, the
additional definitions given in this section are seldom used in thermo-
dynamics.
References
1. N. A. Gokcen, Thermodynamics, Techscience, Hawthorne, California (1975).

2. H. Margenau and G. M. Murphy, Mathematics of Physics and Chemistry, Second Edition,
Krieger, Huntington, New York (1976).
3. D. R. Gaskell, Metallurgical Thermodynamics, Second Edition, McGraw-Hili, New York
(1981).
4. O. F. Devereux, Topics in Metallurgical Thermodynamics, Wiley-Interscience, New York
(1983); see also R. A. Swalin, Thermodynamics of Solids, Wiley·Interscience, New York
(1972).
5. R. Hultgren, P. D. Desai, D. T. Hawkins, M. Gleiser, and K. K. Kelley, Selected Values
of the Thermodynamic Properties of Binary Alloys, ASM, Metals Park, Ohio (1973).
6. E. Hala, J. Pick, V. Fried, and O. Vilim, Vapour-Liquid Equilibrium, Second Edition,
translated by G. Standart, Pergamon Press, Elmsford, New York (1967).
7. N. A. Gokcen, High Temp. Sci. 15,293 (1982).
8. R. Castanet, Z. Moser, and W. Gasior, Calphad 4(4), 231 (1980).
9. E. Hala, E. Wichterle, J. Polak, and T. Boublik, Vapor-Liquid Equilibrium Data at Normal
Pressures, Pergamon Press, Elmsford, New York (1968).
10. H. Renon and J. M. Prausnitz, 1. Am. Inst. Chem. Eng. 14, 135 (1968); see also J. M.
Prausnitz, Molecular Thermodynamics of Fluid Phase Equilibria, Prentice-Hall, Englewood
Cliffs, New Jersey (1969).
11. J. M. Prausnitz, C. A. Eckert, R. V. Orye, and J. P. O'Connell, Computer Calcuationsfor
Multicomponent Vapor-Liquid Equilibria, Prentice-Hall, Englewood Cliffs, New Jersey
(1967).
12. C. Wagner, Thermodynamics of Alloys, translated by S. Mellgren and J. H. Westbrook,
Addison-Wesley, Reading, Massachusetts (1952).
13. M. Ohtani and N. A. Gokcen, Trans. Metall. Soc. AIME 218, 533 (1960).
14. C. H. P. Lupis and J. F. Elliott, Acta Metall. 14,529 and 1019 (1966).
15. F. Neumann and H. Schenk, Arch. Eisenhuettenwes. 30, 477 (1959).
16. Z.-I. Morita and T. Tanaka, Iron Steellnst. 1. 23, 824 (1983).
17. L. F. Kozin, R. S. Nigmetova, and M. B. Dergache, lzd. Nauka, Alma-Ala (1977).
18. R. S. Nigmetova, N. A. Golovanova, and L. F. Kozin, Z. Fiz. Khim.53, 1450 (1979).
2
Phase Equilibria and Phase Diagrams
Single-Component Equilibria
The phase equilibria constitute a vast area of interest in metallurgy, geology,

chemistry, physics, and related fields. In a single-component system, such
equilibria occur between two phases, or three phases. The two-phase equili-
bria and the corresponding Gibbs energy relations for a component A are
A(Phase I) = A(Phase II); (2.1)
A rigorous derivation of equation (2.1) will be given later in this chapter.

Two phases may coexist over a wide range of temperature and the corre-
sponding range of pressure. Three phases may also coexist at a single
temperature and pressure for a pure component; therefore,
A(I) = A(II); A(II) = A(III); (2.2)
The equality of Gibbs energies in equation (2.1) requires that dGI = dGII,
and since dG = V dP - S dT, it is evident that
VI dP - Sl dT = VII dP - Sll dT (2.3)
Rearrangement of this equation with the substitution of Sll - Sl = aH / T

gives
dP aH
or - = - - (2.4)
dT TaV
The numerical values of aH and T are positive; therefore, if VII is larger

29
30 Chapter 2
than Vi, as in melting of most metals and vaporization of all elements and
compounds, then dT / dP is positive, i.e., an increase in P causes an increase
in T. On the other hand, when VII is less than Vi, as in melting of As, Sb,
Bi, and H 20, an increase in P causes a decrease in T.
A phase diagram for a single-component system is shown in Fig. 2.1.
Solid and vapor phases coexist along the curve AB; liquid and vapor phases,
along Be; and solid and liquid phases, along BO. The areas represent the
regions of stability for the single phases, and the lines, the coexisting two
phases. At B, all three phases are in equilibrium, and B is thus called a
triple point. The relative slopes of the lines at B are significant in thermody-
namics. The slope of BO is usually very steep because fl V is very small;
in Fig. 2.1 fl V, and hence the slope, is taken to be positive. At the triple
point B, the temperature is designated as TB and the slope of the sublimation
curve AB at T B is
(2.5)
where flHsub\ is the molar enthalpy of sublimation, and VII - Vi =

V(vapor) - V(solid) = V(vapor) = RT/ P, because the volume of vapor is
I
I
:~
~I
A '-I I
T-
Figure 2.1. Phase equilibria for one-component system.
Pbase Equilibria and Pbase Diagrams 31
usually about lOOO-fold larger than that of a condensed phase under equili-
brium conditions. Similarly, for the vaporization curve BC,
dP = (AHva p) P (2.6)
dT RT~ B
where AHvap is the enthalpy of vaporization. Equations (2.5) and (2.6) are
also valid at points other than B, i.e., at all possible values of P and T.
According to the first law of thermodynamics, the enthalpy of sublimation
at B is equal to the sum of the enthalpies of melting AHm and vaporization
AHvap, or AHsubl = AHm + AHvap; hence, AHsubl > AHvap. Therefore,
equations (2.5) and (2.6) show that the slope of AB is greater than the slope
of BC at their point of intersection. This requirement necessitates that the
phase boundaries intersect each other at angles less than 180° as shown in
Fig. 2.1. Consequently, the supercooled liquid and superheated solid must
have higher vapor pressures than the corresponding equilibrium phases.
For example, at TE , P E is greater than P F , and since Gg = GO,8 + RT In P,
then G8(over solid at E) > G8(over liquid at F) and for the coexisting
phases Gg(at E) = OS and G8(at F) = G 1; therefore, the preceding in-
equality is identical with G S > G 1 at TE • The solid at E must therefore
transform into liquid at F either directly or by vaporization and condensa-
tion, because the phase with the lower Gibbs energy is the liquid phase.
Thus, the extended portions of all the curves terminate in regions where
they represent nonequilibrium conditions. A pure substance may have
several additional triple points among its solid allotropes at various tem-
peratures and considerably higher pressures than the solid-liquid-vapor
triple point. It will be seen later that the intersecting curves at a triple point
in composition-temperature phase diagrams also obey the requirement that
they intersect one another at angles less than 180°.
Multicomponent Equilibria
Equilibria in multi component systems require the definition of degrees

of freedom and conditions of equilibrium. It was seen in Chapter 1 that
the Gibbs energy flf for an open multi component system consisting of
one phase is a function of pressure, temperature, and c numbers of moles,
i.e.,
(2.7)
32 Chapter 2
The partial molar Gibbs energy of component i was defined to be OJ =

aG/anj, and OJ is a function of P, T, and c - 1 composition variables which
may be taken as mole fractions, Xj; thus,
(2.8)
The number of independent intensive variables in this equation is called

the independent variables of states for a single phase, which consists of P, T,
and c - 1 composition variables, or simply c + 1 variables. Under specified
conditions, such as fixing P, T, and any number of composition variables,
and the coexistence of several phases, the number of independent variables
of state is reduced. The number of independent variables of state of any
system under equilibrium is called the degree of freedom, or variance.
If magnetic, electric, and gravitational fields were also variables of state
in equation (2.7), the number of independent variables would increase by
3 and the degrees of freedom would be 4 + c instead of 1 + c. Additional
variables of state do not present particular difficulties, and for simplicity it
is sufficient to consider the variables of state as given in equation (2.7). The
total differential of G for a phase is
aG aG - -
dG = - dP + - dT + G 1 dn l + G 2 dn2 + ...
aP aT
The first two terms represent the contributions to the Gibbs energy when
the number of moles of all components is constant, or when the system is
a closed system; hence, from dG = YdP - $ dT, aG / aP = $, it follows that
dG = Y dP - $ dT + 0 1 dn l + O2 dn2 + . . . (2.9)
where Y and $ are the volume and the entropy, respectively, both of which
are extensive properties.
The Gibbs energy G for a system consisting of 4> phases is the sum of
Gibbs energies of all phases, i.e.,
(2.10)
where the superscripts refer to the coexisting phases. The total differential
of G is then the sum of dG\ dG II , •.• ,each of which is written out according
to equation (2.9); therefore,
dG = (Y[ + yll + ... + y") dP - ($[ + $11 + ... + $<P) dT
+ 0: dn: + 0: 1 dn:[ + ... + ot dnt
+ O~ dn~ + O~[ dn~[ + ... + ot dnt
(2.11)
Phase Equilibria and Phase Diagrams 33
The system is in equilibrium under four specific conditions dictated by
(2.12)
where (1) the change in Gibbs energy is zero, (2) pressure and (3) tem-
perature are fixed and uniform from one phase to another, and (4) the
system must be closed, and therefore the amount of each component
n), n2, ... ,nc is fixed as indicated by the subscript n. The last condition
requires that
dnl = 0 = dnJ + dnJ' + ... + dnt
dn2 = 0 = dn; + dn;' + ... + dnt
dnc = 0 = dn~ + dn~[ + ... + dn~
Multiplication of the first equality in this set of equations by OJ, the second
by O~, ... , the last by O~, and subtraction of the results from equation
(2.11) after setting df(5, dP, and dT = 0, yields
(2.13)
For every possible value of each dn{, which is generally nonzero, equation
(2.13) is zero if, and only if, the coefficient of each dn{ is zero; therefore,
oJ = OJ' = ... = at; (¢ terms)
o~ = O~' = ... = at
[c( ¢ - 1) equations] (2.14)
G~ = O~' = ... = O~
34 Chapter 2
The method used to derive this equation is called Lagrange's method of

undetermined multipliers. l Equation (2.14) contains a very important ther-
modynamic statement first derived by 1. W. Gibbs (1875), that the chemical
potential of a component is the same in all phases under equilibrium
conditions. The term "chemical potential" is still used, but the appropriate
term is the "partial molar Gibbs energy". It is interesting to note that for
a single-component system, or for c = 1, 0:= G:, and 0:1 = G:\ and this
immediately gives a rigorous derivation of equation (2.1).
Phase Rule
The total number of intensive variables that can define each phase is
c - 1 composition variables, and pressure and temperature. For c/> coexisting
phases, therefore, the total number of intensive variables defining the system
is
Total number of variables = 2 + c/>(e - 1) (2.15)
the second term is the total number of composition variables, c - 1, for

each of the c/> phases. According to equations (2.14), the value of 0; for
any selected component i is the same in all the c/> phases, and each line in
equations (2.14) represents c/> - 1 independent equations, and further, e
lines represent c( c/> - 1) independent equations, i.e.,
Total number of independent equations = c( c/> - 1) (2.16)
The number of independent variables, called the degrees of freedom, and

denoted by Y, is equal to the total number of variables, minus the number
of independent equations, i.e., 2 + c/>( e - 1) - c( c/> - 1), or simply
Y=c-c/>+2 (2.17)
This equation is the formal statement of the well-known phase rule originally
derived by 1. W. Gibbs (1875). The degrees of freedom, or the variance, Y,
represents the number of unrestricted variables of state which may be a set
of variables out of P, T, and c - 1 compositions. It must be emphasized
that Y can be zero or a positive number, but never a negative number. It
should be remembered that the composition variables refer to each
individual phase and not to the bulk composition of a heterogeneous system
of two or more phases. The permissible number of composition restrictions
could be all in one phase or in several phases; for example, fixing the mole
fractions Xi and Xj of components i and j in one phase, or fixing Xi in one

phase and Xj in another phase, or fixing only Xi in two coexisting phases
decreases the degrees of freedom by two.
All the components were assumed to be soluble in all phases in the
derivation of the phase rule. However, in some cases the solubility of certain
components in some phases is very small or practically zero. If, for example,
the solubilities of kl components in Phase I, and k2 components in Phase
II are zero, the number of composition variables for Phase I decreases by
kt. and for Phase II, by k 2 • Therefore, when k i components are insoluble
in 4>j phases, the total number of variables in equation (2.14) decreases by
k i 4>j so that the right side becomes 2 + 4>( c - 1) - k i 4>j. The Gibbs energy
{25q is independent of n i in a phase q in which i is insoluble; hence, for
Oi = a{25q / ani = 0. Consequently, the number of independent equations in
equation (2.14) decreases by 4>j on each line of equation (2.14) and by k i 4>j
on all lines; the phase rule then becomes
(2.18)
Therefore, the absence of some components in some of the phases does not
alter the phase rule. It is often convenient to fix the pressure and decrease
the degrees of freedom by one in dealing with condensed phases such as
for substances with low vapor pressures. The phase rule then becomes
Y=c-4>+1 (constant pressure) (2.19)
A few examples illustrate some applications of the phase rule represen-

ted by equation (2.17). In a single-component system, for one phase,
Y = 1 - 1 + 2 = 2, and hence pressure and temperature may vary indepen-
dently; for two phases, Y = 1 - 2 + 2 = 1, either pressure or temperature
may vary independently, and if pressure or temperature is fixed Y = 0, or
the system is completely defined. If pressure and temperature are both fixed
in the last case, Y = -1; therefore, either the phase rule is violated, or an
unnecessarily large number of restrictions are imposed on the system. For
three phases, Y is zero and again the system is completely defined. If either
temperature or pressure is varied, one of the three phases must disappear
to permit this degree of freedom.
In a system consisting of three components, the maximum value of
Y is 4, since the minimum number of phases for any system is 1. When
4> = 3, then Y is 2 and Y may be chosen out of the following total
number of variables: 2(pressure and temperature) + 3(3 - l)(composition
variables) = 8.
36 Chapter 2
Phase Diagrams
A phase diagram represents a map of coexisting phase boundaries as

affected by the variables of state. The relationships between the phase
diagrams and the molar Gibbs energy diagrams are presented in detail in
the remaining sections of this chapter. We limit our discussion to condensed
phase diagrams in which pressure is held constant. The removal of the
restriction on pressure does not require a special treatment, but complicates
the phase diagrams. We shall consider the temperature-composition phase
diagrams, involving the solids and liquids, and limit our presentation largely
to binary phase diagrams. 2 - 5
A condensed binary phase diagram represented on the composition-
temperature coordinates consists of the single-phase regions separated by
the two-phase regions as shown in Figs. 2.2-2.4. The curves separating the
single phases from two phases represent the compositions of single phases.
When a homogeneous liquid phase freezes to form a homogeneous solid,
the phase diagram may be as in Fig. 2.2(A) with the phase boundary curves
spaced as indicated, or as in Fig. 2.2(B) with a common maximum point
for both boundary curves, or as in Fig. 2.2(C) with a common minimum
T A
B
T
t c
T Figure 2.2. Hypothetical binary phase
diagrams for liquid (L), solid (S), and
S + L phase fields. Upper curve in each
panel is "liquidus," and lower curve,
"solidus."
(1 ) x2 - (2 )
Figure 2.3. Hypothetical phase diagram for various transformations.
point. The curve showing the start of solidification, called the liquidus, is
tangent to the curve showing the completion of freezing, called the solidus,
at the same extremum point in Fig. 2.2(8) and (C). This requirement is
necessitated by the Gibbs-Konovalow theorem,t-6 which can be proved by
using the Gibbs energy G written as G = nlO I + n202, and functionally as
G = G(P, T, nJ, n2). The total differentials of these equations are
dG = n l dOl + n2 d0 2 + 0 1 dn l + O 2 dn2
dG = YdP - $dT+ 0 1 dn l + O 2 dn2
These equations yield Y dP - $ dT = n l dOl + n2 d02, and after division
with (nl + n2), and then imposing the constant-pressure restriction, we
obtain
(2.20)
38 Chapter 2
a + L
a + ABx ABx + E
CA) XB - CB)
Figure 2.4. Hypothetical phase diagram forcatatectic (metatectic) and eutectic transformations,
and for compound AB capable of dissolving its components to limited extents.
where S is the molar entropy. This equation, rewritten for two phases, yields
where OJ is the same for both phases for a given temperature. Rearrangement
of the preceding equations and division by dx~ yields
(2.21)
It is important to note that phases I and II may be liquid and solid,

respectively. This equation leads to the proof of the Gibbs-Konovalow
theorem in its concise and most useful form.
THEOREM. If a phase boundary curve for Phase I meets another phase

boundary curve for Phase II at a point (x~ = X~I, T2 ), both curves must become
horizontally tangent to each other.
PROOF. The compositions of both phases are equal, hence X!I - x! and
X11 - x~ in equation (2.21) are both zero; further, dOl! dx~ and d0 2! dx1
are both finite because 0 1 and O2 are continuous functions of T and X2,
and Sl - Sll is finite and nonzero; therefore, dT / dx~ must be zero. Equation
(2.21) can be divided by dX~1 instead of dx~, and the same argument can
be followed to show that dT / dX~1 is also zero. Therefore, both curves must
be horizontally tangent to each other and this completes the proof.
Eutectic-Type Reactions
The liquidus and solidus lines are often depressed as shown in the
upper left of Fig. 2.3 in such a way that at a particular composition and
temperature, the liquid and two solids, i.e., three phases, coexist. The
reaction upon cooling is liquid(LI) ~ solid(a) + solid(I3), which is called
the eutectic reaction. Above and below the eutectic temperature, one phase
must disappear according to the phase rule. There are other similar reactions
involving three phases with various possible combinations. Each reaction
is named according to the states of aggregation of phases and the number
of reactant phases. The suffix "tectic" is used for reactions involving one
Table 2.1. Three-Phase Equilibria in Binary Systems
Name Diagram Reaction upon cooling
Transitions involving liquid phases
Eutectic
" \~i ~ Liquid(L\) -+ solid(a) + solid(f3)
~
Monotectic Liquid(L3) -+ liquid(L2) + solid(e)
Peritectic Solid(f3) + liquid(L2) -+ solid(A)
Syntectic Ly /f3\ 42
Liquid(L\) + liquid(L2) -+ solid(f3)
Catatectic 7
(metatectic) ~ 'tI 4 SoIid(f3) -+ soIid( a) + liquid(L)
Transitions involving only solid phases
Eutectoid
1 'fl ~ Solid(f3) -+ soIid(a) + soIid(A)
Peritectoid
P M~ Solid( a) + solid(A) -+ solid( 1/)
40 Chapter 2
or two liquid phases, and the suffix "tectoid" for three solid phases. Various
types of three-phase reactions are presented in Figs. 2.3 and 2.4 and summar-
ized in Table 2.1 with the coexisting phases marked as in these figures.
These transformations consist of (1) eutectic-type, in which the reactant
is a single phase, and (2) peritectic-type, in which the reactants consist of
two phases. Thus, there are four eutectic-type and three peritectic-type
transitions in the preceding list.
A critical point exists when one phase dissociates into two phases at
a point where the phase boundary has a horizontal inflection point, i.e.,
when aT/ aX2 and a2T / ax~ are both zero as shown in Fig. 2.3 for L ~ LI + L2,
L ~ L2 + L3 , and 8 ~ 8 1 + 82. It should be noted that the maximum point
in the 17-region does not have an inflection point and the phase boundaries
for 17 and u are tangent to each other at their common maximum points.
The maximum point in the center of Fig. 2.4 is for a congruently melting
compound, AB, capable of dissolving its component elements to limited
extents as indicated by the phase region AB x.
The limits of solubility of one solid phase in another solid phase is
sometimes called the solvus curve, or briefly, the solvus. Thus, the E-phase
boundary below the eutectic in Fig. 2.4 is the solvus, which is the limits of
solubility of ABx in E. However, the E-phase boundary above the eutectic
is the solidus.
The preceding diagrams contain all the possible phase equilibrium
types encountered in the condensed phase binary systems. Various compila-
tions of phase diagrams exist,S-ll and recently evaluated diagrams are
published in the Bulletin of Alloy Phase Diagrams (a bimonthly journal
begun in 1980, published by ASM-NBS).
Erroneous Diagrams
A number of important aspects of the phase equilibria summarized in
Figs. 2.2-2.4 must be observed to avoid errors in drawing the phase
boundaries. All such errors violate (a) the phase rule, (b) the Gibbs-
Konovalow theorem, and (c) the requirement that the extended portions
of the phase boundaries terminate in the two-phase regions. Examples of
these violations are illustrated in Fig. 2.5 and discussed as follows.
a. Along Be, a, Lh L 2 , and f3 coexist and thus violate the phase rule
since the degree of freedom Y with four coexisting phases is -1 and Y
cannot assume a negative value. This error can be corrected by joining LI
and L2 at one point on the straight line Be. On D, a, 10, A, and f3 coexist
because at E there are more than two curves and one straight line intersecting
one another; the phase boundary curve ME must therefore not terminate
at E. Accordingly, there must be no more than one straight line and two
Figure 2.5. Errors in a hypothetical phase diagram violating phase rule or other thermodynamic
principles.
curves at a point of intersection of phase boundaries. On F, a eutectoid

and a peritectoid on the same horizontal line signifies that a, e, 8, and A
coexist and violate the phase rule. There are three phases on the straight
line J, and since J is not horizontal these phases coexist over a temperature
range and violate the phase rule since Y = -1. At K, there are two phases
A and L over a temperature range for pure component 2, and again Y = -1;
the phase boundaries must therefore intersect the vertical line for the pure
component at the same point.
b. At I, M, and N, the Gibbs-Konovalow theorem is violated, i.e.,
each pair of phase boundaries does not meet at an extremum point. At M,
the phase boundaries should not meet at all, and an entirely different
construction must be made to eliminate the error, particularly because of
the erroneous separation of {3 and A along a curve EM instead of a two-phase
field.
c. The extended portion of the phase boundary at B, shown by the
dashed extended curve, and the unstable liquid on the extended portion of
AC are in equilibrium. Since the liquid below C is unstable, the dashed
42 Chapter 2
curve below B must also be in an unstable region, not in a one-phase region

which is stable. Hence, the extended portion below B must terminate in a
two-phase region, or within the area BCDE. This is possible when the phase
boundary curves and the horizontal line BC intersect one another at an
angle smaller than 180°.
Lever Rule
Figure 2.6(a) shows a simple phase diagram in which solid phase A

dissociates at the critical point Q into two solids Al and '\2' The horizontal
line ACB ending on the phase boundaries at A and B is called a tie-line.
Point C represents the bulk composition Xc of a mixture of two phases. As
point C moves from A toward B, the compositions of Al and A2 remain
unchanged as XA and XB, respectively, but the mass of '\2 increases relative
to the mass of AI' All compositions Xi refer to the atomic fraction of the
a
o
.!: -50
o b
u
c5
<J -100
a 0.4 0.8 1.0

X2
Figure 2.6. (a) Hypothetical binary phase diagram for dissociation of solid A-phase into two
solid phases, AI + A2 ; critical point is Q; bulk composition is taken as C at temperature
corresponding to AB; (b) flO versus X 2 for temperature corresponding to ACB. Equation
(2.24) represents flO curve for 1 g-atom of alloy; A, C, and B represent corresponding
compositions in both panels.
second component. Let meA) and m(B) represent gram atoms of AI and
A2 at A and B, respectively, with the restriction that meA) + m(B) = 1.
Atomic balance requires that
xAm(A) + xBm(B) = xdm(A) + m(B)]
This equality can be used to derive
meA) _ XB - Xc CB
(2.22)
m(B) Xc - XA AC
which is known as the lever rule. Note that Xc representing the bulk
composition of two phases is not a composition variable for thermodynamic
properties of each phase such as dO and dGj • Equation (2.22) may be
transformed into a useful form by adding 1 to both sides of the first equality
and simplifying by using meA) + m(B) = 1 so that
(2.23)
where the second relationship is obtained from the first by using m(B) =
1 - meA).
Molar Gibbs Energy of Mixing-Phase Diagrams
The molar Gibbs energy of mixing refers to the process prescribed by

XI (pure component 1) + X 2 (pure component 2) ~ 1 gram atom of alloy.
A simple phase diagram and the molar Gibbs energy of mixing dO are
presented in Fig. 2.6. Variation of 6.0 with X2 refers to the temperature
represented by the tie-line ACB in Fig. 2.6(a). The curve in Fig. 2.6(b) is
drawn by assuming that dO is represented by
6.0 = 1000x2 + 700x~ - 2800x~ + l100x;

+ 600[(1 - x 2 ) [nO - x 2 ) + x21n x 2 ] (cal/mole) (2.24)
where the last term is the molar Gibbs energy of mixing of an ideal solution,
dO(ideaI) = RT(xl In XI + x21n X2) with RT taken to be 600 cal/mole
(301.93 K) as a convenient simple quantity. The remaining coefficients in
equation (2.24) were obtained by using one maximum point M and two
minimum points near A and near B. The tangent line ACB to the curve for
44 Chapter 2
flO intersects the vertical axis at X2 = 0 at the point corresponding to

flat = at - O~, and similarly, fl0 2 = O2 - O~, according to the tangent
intercept method. The tangent line ACB shows that flat = at -
O~ and
fl0 2 = O2 - O~ for At phase are the same as those for A2 phase.
The section of the straight line between A and B in Fig. 2.6(b) represents
flO for the mixture of two phases; thus, flO(at C) is
flO(at C) = XB - Xc flO(at A) + Xc - XA flO(at B) (2.25)

XB - XA XB - XA
which is in accord with the lever rule because flO of a mixture of two
phases is the sum of flO of its constituent phases. Equation (2.25) is linear
and it is represented by the straight line between A and B. Likewise, the
straight line DEF represents flO of two phases having the compositions
corresponding to D and F. The value of flO(at E) is larger than that of
flO(at C); therefore, any straight line joining two points on the curve AMB
has higher values of flO than those corresponding to the straight line ACB
at the same values of Xc; hence, a phase mixture along DEF is unstable
with respect to that along ACB for the same bulk composition.
The curve for flO in Fig. 2.6(b) has two inflection points, one at I
(X2 = 0.331) and the other at J (X2 = 0.791) as can be shown by substituting
these values of X2 in the second derivatives of equation (2.24) with respect
to X 2 • These points are called the spinodes, which are important in kinetics
of nucleation and growth of new phases from the supersaturated single
phases. The equations for the activities of components can be derived and
the results can be plotted versus X2. It can be shown that the maximum and
the minimum points in such activity versus composition diagrams for both
components coincide with the spinodes.
The shape of the curve in Fig. 2.6(b) changes with increasing tem-
perature as required by the phase diagram. Thus, A and B approach each
other and finally minimum, maximum, and the inflection points coincide
at the horizontal inflection point Q in Fig. 2.6(a) as required by the phase
diagram.
4.G Diagrams for Other Phases
The diagram shown in Fig. 2.6(a) is for a single solid phase decomposing
into two solid phases. For all other types of transformation, it is customary
to represent flO versus X2 for each phase at each selected temperature on
the same diagram. If the phases in equilibrium are solid and liquid, the
Phase Equilibria and Phase Diagrams 4S
convention for writing /lG for each phase through the entire range of
composition at a selected temperature T is as follows:
/lG(I) == G(I) - xl(I)GHstable phase at T)

- xz(I) G~(stable phase at T) (2.26)
/lG(s) == G(s) - xI(s)G~(stable phase at T)

- xz(s)G~(stable phase at T) (2.27)
where G(I) and G(s) are the molar Gibbs energies of the liquid and solid
solutions. Equation (2.26) at X2 = 1 becomes /lG(I) = G(I) - G 2(s) when
the stable pure phase for component 2 is solid at T, and since for X2 = 1,
G(I) is the same as G 2(1) for the pure liquid 2, then /lG°(I) is identical
with /lG2,m of melting 'for pure component 2. If, however, the stable phase
for component 2 is liquid, /lG(I) in equation (2.26) is then zero at X2(1) = 1.
Reconsider equations (2.26) and (2.27) at a temperature T greater than
the melting points of both components so that
/lG(I) = G(I) - xl(I)G~(I) - xz(I)G2(I) (2.28)
/lG(s) = G(s) - xl(s)Gi(l) - X2(S)G 2(I) (2.29)
Addition of xI(s)Gi(s) + xz(s)G 2(s) to the right side of equation (2.29) and
then subtraction of these terms from the same side, followed by a simple
rearrangement, yields
/lG(s) = G(s) - XI(S)G~(s) - X2(S)G 2(s)
(2.30)
where /l Gtm is the standard molar Gibbs energy of fusion of pure component
i. The first three terms after the equal sign can be transformed into
RTxl(s) In al(s) + RTx2(S) In a2(s) by using the definition of activity, i.e.,
Gj(s) = Gi(s) + RTln aj(s) and observing that G(s) = XI(S)G I + X2(S)G2.
Likewise, the three terms on the right side of equation (2.28) can be rewritten;
the results for equations (2.28) and (2.29) are therefore
/lG(I)= RTxI(l) In al(I) + RTxz(l) In az(1) (2.31)
/lG(s) = RTxI(s) In a1(s) + RTxz(s) In a2(s)
- x1(s)/lGi,m - X2(S)/lG 2,m (2.32)
A simple example for these equations is now presented, with the following
arbitrarily assigned values:
TI,m = 900K /lH':,m = 900R /lSi,m =R
46 Chapter 2
T2 ,m = 300K 6.H'2,m = 300R 6.S'2,m = R

where T;,m is the melting point of i. From the definition of 6.Go as 6.Go =
6.Ho - T6.So, with 6.C; = 0, it is evident that
6.G~,m = 900R - RT; 6.G'2,m = 300R - RT (2.33)
If, in addition, the solution is ideal so that the activities can be set equal
to the mole fractions, equations (2.31) and (2.32) become
6.G(l) = RTxt(l) In xt(l) + RTx2(l) In xiI) (2.34)
6.G(s) = RTxt(s) In xt(s) + RTxis) In xis)

- Xt(s)6.G~,m - X2(S)6.G'2,m (2.35)
Equations (2.34) and (2.35) are plotted in Fig. 2.7 for T = 1000 K. We note
that 6.G(s) for the solid in Fig. 2.7 is higher than 6.G(l) for the liquid;
-'G.~ 2.m
800
o
..,....... -.1G 1• m = 100 R
-800
-1600~---L----~--~----~--~
o 0.2 0.4 0.6 0.8 1.0
X2
Figure 2.7. Diagram for 40(1) and 40(s) at 1000 K from equations (2.34) and (2.35),
respectively. 40(S) > 40(1) throughout, and solid phase is thus unstable relative to liquid
phase. Phase diagram for this system is in lower portion of Fig. 2.8.
hence, the liquid phase is stable relative to the solid phase at 1000 K. The
difference between the vertical intercepts (at x\ = 1 and at X 2 = 1) for those
curves represent ~G~.frz = +100R and ~G~.frz = +700R where ~G~frz =
-~G~m with the subscripts frz and m referring to freezing and melting,
respectively. The positions of the curves below 300 K are reversed, i.e., the
curve for G(I) is above that for ~G(s) because the solid phase is stable
relative to the liquid phase below 300 K as can be shown by plotting a
different but similar figure.
Next to be considered are a set of curves at 550 K for which the liquid
and solid phases coexist at appropriate concentrations. These curves are
given by
~G(s) = G(s) - x\(s)G~(s) - X2(S)G~(s) - xis)~G2.m (2.37)
Substitution of aj = Xj and OJ = G~ + RT In Xj in these equations gives
~G(I) = RTx\(I) In x\(I) + RTxil) In xiI) + x\(I)~G~.m (2.38)
~G(s) = RTx\(s) In x\(s) + RTxis) In x2(s) - xis)~G~.m (2.39)
Substitution of equation (2.33) in these equations gives
~G(I) = 550Rx\(I) In x\(1) + 550Rxil) In xiI) + x\(I)350R (2.40)
~G(s) = 550Rx\(s) In x\(s) + 550Rxis) In X2(S) + xis)250R (2.41)
Equations (2.40) and (2.41) are represented in the upper portion of Fig.
2.8 by the curves marked LIQUID and SOLID respectively. The straight
line tangent to both curves gives the compositions of solid and liquid at
X2 = xis) = 0.450 and X2 = X2(I) = 0.709, respectively. Above the tangent,
a pair of phases have a higher value of ~G than the phases at the points
of tangency as discussed in conjunction with Fig. 2.6. The liquid phase is
stable from X2 = 0.709 to X2 = 1, and the solid phase is stable from X2 = 0.0
to X2 = 0.450, as required by the relatively lower values of ~G represented
by the lower sections of the curves. At the point of intersection of the curves,
~G(I) and ~G(s) are equal but 0\(1) and O\(s), as well as Oil) and Ois),
are not equal; therefore, there is no equilibrium at this point. The vertical
intercepts of the curves in Fig. 2.8 correspond to ~G~.m = 350R and
-~G2.m = 250R, as indicated in the figure. The solidus and liquidus points
given by the upper portion of Fig. 2.8 are shown in the lower portion. The
repetition of the foregoing procedure for OJ at various temperatures gener-
ates the entire phase diagram as shown in the lower portion of Fig. 2.8.
48 Chapter 2
:--- 0
600 +I1G, m = 350R
\ ' _I1G o2,m = 250R-

400
\
-J
0 200 \
....
~
\
\0
-J
< 0
U
C) ~
~ -200
~
-400
-600
'""'-
o
I
I
en l
I
on 01
'000 ~ '"':1
~ I 0 I
'" 1 "",1
XI xl L
750 ~ I I
I I
I~ I
I- I _~_
500
Figure 2.8. Upper diagram is for
s S + L "'"
~G(I) and ~G(s) at 550 K from
equations (2.36) and (2.37),
250 ~--~----~----~----~--~ respectively. Phase diagram in
o 0.2 0.4 0.6 0.8 1.0 lower portion is obtained from
X2 equations (2.43) and (2.44).
The phase boundaries in Fig. 2.8 have been drawn by equating OJ(l) =
O;(s) and then substituting various values of T to solve for Xj(s) and
Xj(l). In this region G~ is solid but G 2 is liquid, The liquid phase is
ideal, and for 0.(1) it is evident that 0.(1) = G~(1) + RT In x.(1), and like-
wise, O.(s) = G~(s) + RT In x.(s), and the equality of 0.(1) and O.(s)
yields
G.O( I) - G.O( s ) x.(1)

° = -RT In -(-)
= ~G.,m (2.42)
x. s
which is simply the equilibrium distribution ratio (or equilibrium constant)

for (component 1 in solid phase ~ component 1 in liquid phase). Substitu-
tion of AG~.m from equation (2.33) into equation (2.42) and simplification
gives
900 xl(l)
l--=ln-- (2.43)
T xl(s)
A similar equation is obtained for component 2 by an identical procedure;

the resulting final equation is
(2.44)
These equations represent the solidus and liquidus curves. For example, at
550 K, XI(l)/XI(s) = 0.5292, and xil)/x2 (s) = 1.5755, and then substitution
of 1 - xl(l) = x 2(l) and 1 - xl(s) = x 2(s) reduces the number of unknowns
in the first set of equations by two, leading to X2(S) = 0.450 and X2(l) = 0.709;
these points are indicated in Fig. 2.8.
Deviations from ideality modify the curves for AG represented in the
preceding figures; however, the principles involved in the representations
are basically the same. The curves similar to those in Figs. 2.6-2.8 for
nonideal solutions require substitution of al = I' IX1 and a 2 = I'2X2 in
equations (2.31) and (2.32), and expansion to generate a set of terms
expressed by G e = RT(xl In 1'1 + x2ln 1'2)' Appropriate analytical equations
for Ge(l) and Ge(s) must then be added to equations (2.34) and (2.35),
respectively, to represent them as functions of composition and temperature
for nonideal solutions. A detailed example of such a procedure will be
given in Chapter 6 in thermodynamic calculations of the iron-carbon phase
diagram.
aG Diagrams for Complex Systems
The diagrams of AG versus X 2 for complex phase diagrams may be

illustrated schematically by the eutectic system shown in Fig. 2.9. The AG
diagram for the liquid and a phases between the melting points of com-
ponents 1 and 2, e.g., at T(, is, in principle, the same as the diagram in Fig.
2.8, and therefore not presented. At Til, below the melting points of com-
ponents and above the eutectic temperature Te , the curves are shown in the
upper portion of Fig. 2.9. The straight lines tangent to each pair of curves
give the compositions of each pair of phases. Single phases are stable outside
50 Chapter 2
--_. --~ ··· --

: "":L~ L
a+p
Figure 2.9. Upper figure schematically

o 0.2 0.4 0.6 0.8 1 .0 shows t:J. G for liquid a and f3 phases
X2 at Til' Lower figure is phase diagram.
the two-phase regions where AG is lower for the stable phases than the
unstable phases. Figure 2.10 shows the diagrams at Te and TIll indicated
on the phase diagram in Fig. 2.9. At Tn three phases coexist as shown by
the single tangent line to the three curves. As the temperature decreases,
the curve for the liquid, AGO), moves up and the remaining curves move
down so that the a-phase at e is in equilibrium with the {3-phase at h for
TIll. Assume that the liquid and a phrases are supercooled as shown by
the extended dashed portions of the phase boundaries in Fig. 2.9. The liquid
at g, and the a-phase at f must then coexist as shown in Fig. 2.9 and in
Fig. 2.10, by the tangent line at f and g. The location of the curve for AG(l)
at TIll necessitates that the tangent at f and g be in the two-phase region,
and since the values of AG for the two phases at f and g are higher than
o 1.0
c d
Figure 2.10. Upper figure schemati-

cally shows t:.G for liquid a and f3
phases at T.. and lower figure, at Til!'
For TEO TUh and compositions, see o e 9 1.0
phase diagram of Fig. 2.9.
those at e and h, the phases at f and g are unstable. The extended portions
of the phase boundaries in Fig. 2.9, terminating in the two-phase regions,
are therefore properly constructed. This is possible when the angles between
the adjoining curves and the eutectic line at b and at d are less than 1800
as stressed earlier.
Calculation of Phase Diagrams from Thermodynamic Data
The criterion for equilibrium is expressed by the equality of 0 1 as well

as O2 in the coexisting phases. The required calculations for a binary system
52 Chapter 2
are to write two equations based on O\(a) - 0\({3) = 0 and 02(a)-

02({3) = 0 for two coexisting phases a and (3. The result for component 1
is
(2.45)
Substitution of 07 = RT In 'Yi and rearrangement of the result gives
IlG~(a ~ (3) - O~(a) + 0~({3) + RT In{[1 - xi(3)]/[1 - x2(a)]} = 0

(2.46)
IlG~(a ~ (3) - O~(a) + 0~({3) + RT In[x2({3)/xia)] = 0 (2.47)
These equations can be solved to obtain the values of x 2( a) and X2({3) at

various temperatures, representing the phase boundaries. There are various
types of computer programs presented and discussed for this purpose, \2-\7
depending on the complexity of functions chosen for 07.
Another method, called the minimization of Gibbs energy method, is
thermodynamically equivalent to the preceding method. Its basic principle
can be illustrated by using equations (2.24) and (2.25), and recalling that
dllG(at C) = 0; [IlG(at C) is a minimum] (2.48)
where the differential is for a closed system at constant pressure and

temperature. The total differential of a function u = U(XA' XB) is zero when
auf dX A and auf aXB are zero in du = (au/ aXA) dXA + (au/ aXB) dXB;
therefore,
_all_G---,-(a_t_C..:..) = o. (2.49)
aXA '
For this purpose we label X 2 in equation (2.24) once as X A = x 2 (A) and then
as XB = x2(B) to obtain the equations for IlG(at A) and IlG(at B), respec-
tively. These equations must then be substituted in equation (2.25) and then
in equation (2.49) to obtain two simultaneous equations with two unknown
mole fractions XA and XB' The solution of these equations yields XA = 0.170,
XB = 0.919 at RT = 600 cal/mole (301.93 K). The calculational detail\ shows
that Xc cancels out in setting equation (2.49) to zero.
Phase Equilibria and Phase Diagrams S3
The foregoing methods can be reversed to obtain thermodynamic

properties of mixtures, in particular G e , from the phase diagrams. The
computational procedure in this process is simpler because it does not
involve solving for the unknowns with the logarithmic terms. The accuracy
of the experimental methods for the determination of the phase boundaries,
particularly at high temperatures, is not sufficient for highly reliable values
of G e . The general principle based on experience in this respect can be
summarized as follows: Moderately accurate data for G e or G7 can generate
an excellent phase diagram, but a very highly accurate phase diagram is
necessary to obtain moderately reliable values of G e or en.
Empirical Rules on Phase Stability
Examination and generalization of binary phase diagrams have led to

interesting, and sometimes controversial empirical rules set forth by a
number of investigators. The most widely known such rules bear the names
of Hume-Rothery (1934) for substitutional solid solutions, and Hagg (1929)
for interstitial solid solutions. In substitutional solutions the atoms of metals
A and B occupy equivalent lattice positions, whereas in interstitial solutions,
the smaller atoms of B occupy the voids or the interstices in the lattice of
the larger atoms A, often by small expansion of the A-lattice. As more
binary phase diagrams became established in time, and more knowledge
of electronic properties developed, these rules were modified, not only by
the original proponents, but also by various investigators. The Hume-
Rothery rules l8 .19 have been stated and restated by himself and others, often
with differing interpretations. 20 The main reason for this is the existence of
various scales for electronegativities, atomic radii, and valencies. Therefore,
it is appropriate to begin the Hume-Rothery rules by eliminating one of
the original four rules, called the "relative valency effect," according to
which a metal of lower valency was regarded as more likely to dissolve a
metal of higher valency as a solute than vice versa. 18.20 This elimination,
recommended by Massalski/ o is also in accord with the more recent mono-
graph by Hume-Rothery et al. 18 Further, a statistical analysis by Gschneid-
ner21 shows that this rule is not useful, though it works out well for the
alloys of Cu, Au, and Ag with the B-subgroup elements. This leaves the
following Hume-Rothery rules accepted by recent critical reviewers.18.20.21
Rule 1, size-factor. Solid solubility of a metal A in B (or B in A) is
restricted to a few atomic percent (less than 5 at. % according to Gschneid-
ner22) if the difference between the atomic radii of A and B is more than
15%. The solubilities generally decrease further with increasing size differen-
ces in excess of 15%. For example, the atomic radius of Cu is 1.28 A, and
54 Chapter 2
that of Cd, 1.52 A, and Cd is therefore about 19% larger than Cu, and only
1.7 at. % Cd is soluble in Cu. In contrast, the atomic radius of Zn is 1.37 A,
which is 7% larger than that of Cu, and Cu-Zn form extensive solid
solutions. The radii recommended by Hume-Rothery are half of the closest
distances of approach of atoms in crystals of the pure elements, but in the
examples cited here, the conclusion is not significantly affected when other
atomic radii are used.
Within the favorable range of 15%, known as the favorable size-factor,
other properties of atoms play important roles; consequently, the favorable
size-factor is not sufficient for substantial terminal solubilities. For example
Mg (1.60 A) and Sb (1.61 A) have nearly identical atomic radii, but their
mutual solubilities are less than 0.04 at. %. This rule is therefore a negative
rule; i.e., if the size difference between two metals exceeds 15%, then the
mutual solubilities are limited, but if the size difference is less than 15%,
the mutual solubilities mayor may not be extensive. Theoretical justification
for this rule has been made by Friedel, Blandin, Eshelby, and others.2
Rule 2, electronegativity factor. Formation of stable intermetallic com-
pounds, more appropriately stable intermediate phases, will restrict terminal
solid solubilities. The possibility of formation of such compounds increases
with increasing differences in the electronegativities of component metals.
In general, when the difference in Pauling electronegativities exceeds
±0.4 volt, the solubilities are restricted even when the size-factor is favorable.
This rule is also known as the rule of electronegativity effect.
Rule 3, electron concentration factor. In many alloy systems an important
factor that determines the extent of solubilities in terminal and intermediate
phases is the electron concentration, which is usually expressed in terms of
electrons per atom of alloy, e/at. The compositional range of existence of
each phase having a particular crystal structure is frequently observed to
correspond to a narrow range of electron concentration.
Computation of the e/at. ratios is made by adding the electrons con-
tributed by each pure component element and then dividing the sum by
the number of atoms in the alloy. The number of electrons contributed to
e/at. by each element is not a universally accepted quantity. Usually, but
not always, the outermost sand p orbitals are considered to contribute to
the electron concentration. 24 Indeed, in the B-subgroup elements with paired
d-electrons, it is generally agreed that only the s- and p-orbitals contribute
electrons. (d-electrons are completely paired when there are ten such elec-
trons in the same orbital.) However, the convention is by no means universal
for the transition and noble elements in which the inner orbitals are assumed
to contribute to the electron concentration. Any method of selection is likely
to be controversial, but the main objective is to develop a consistent empirical
picture correlating the phase regions with the e / at. ratio. A set of assigned
values for selected elements, commonly used for computing e/at. ratios,
are listed in Table 2.2.
For a better fit of experimental results, numerous investigators have
presented arguments for assigning variable valencies to Cr, Mn, Fe, Co,
and Ni. When phase diagrams are plotted as temperature versus e/at.,
several interesting features and similarities among various systems are
exhibited. Unfortunately, there is no general agreement on the electron
concentration ranges for various phases encountered in various binary
diagrams. In addition, frequently a phase of known crystal structure in one
binary diagram does not appear in another binary diagram in the same
range of e/at. ratio. Evidently, this as well as other rules encounter greater
degrees of success for the elements in particular groups in the periodic
chart. Greater degrees of divergences and greater numbers of exceptions
occur when greater numbers of metals are considered. It is therefore advis-
able to apply these rules with appropriate modifications to limited selected
groups of elements for greater degrees of success.
The Cu-Zn diagram is shown in Fig. 2.11, in which Cu and Zn
contribute one and two electrons per atom, respectively. Two simple struc-
tures among six encountered in the Cu-Zn system are (1) the bcc ,a-phase
that exists at 1.36 to 1.55 in e/at. (this phase is sometimes written as CuZn
to show its roughly equiatomic composition), and (2) the cph e-phase that
occurs at 1.78 to 1.87 in e/at. (sometimes designated as CuZn3). Such phases
are frequently called electron phases, and they often show no definite
stoichiometry to be called electron compounds. A few examples of electron
phases are listed in Table 2.3 where (1) the limit of maximum solid solubility
of Cu corresponds to e/at. = 1.4, (2) bcc structure appears as Zn is added
in Cu when e/at. = 1.5, (3) y-phase boundary corresponds to e/at. = 1.6
to 1.65, and (4) cph boundary for the e-phase is variable but e/at. lies
about 1.8 for most binary alloys of the noble metals. For each group of
elements a similar scheme can be devised with different values of e/at. for
Table 2.2. Electron Contributions of

Selected Elements to Their Alloys
Group e/at.
IB: Cu, Ag, Au

IIA: Be, Mg, Ca, ... 2
lIB: Zn, Cd, Hg 2
IIIB: B, AI, Ga, ... 3
IVB: Si, Ge, Sn, Ph 4
VB: P, As, Sb, Bi 5
VIllA: Fe, Co, Ni, ... o
56 Chapter 2
WEIGHT PERCENT ZINC

10 20 30 40 50 60 70 80 90
Cu 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Zn

XZn
Figure 2.11. Cu-Zn phase diagram. (Adapted from Hultgren et a/. ll with permission.)
Table 2.3. Ranges of Electron/ Atom Ratios for Selected Electron Phases
fee
upper Minimum "y-phase a eph
Alloy boundary bee boundary boundary boundary (e)
Cu-Zn 1.38 b 1.48 1.58-1.66 1.78-1.87 c

Cu-AI 1.41 1.48 1.63-1.77
Cu-Si 1.42 1.49
Cu-Sn 1.27 1.49 1.60-1.63 1.63-1.75
Ag-Zn 1.38 1.58-1.63 1.67-1.90
Ag-Al 1.41 1.55-1.80
Au-Zn 1.31
"Complex cubic.
b Lower limit is 1.0.
'Terminal Zn-rich 1)-phase is also cph, and contains only 3 at.% maximum Cu, with e/at. = 1.97 minimum;
see Cu-Zn phase diagram (Fig. 2.\1).
various phases. The electron concentration factor is therefore not a definite

factor, but instead is a scheme that attempts to predict the existence of
various phases at certain ranges of values for electron concentration. Based
on the foregoing analysis, an acceptable form of electron concentration
factor may be stated as follows: the boundaries of each phase in the alloys
formed by each group of metals with the remaining metals occur at approxi-
mately the same ranges of electron concentration .
•c
1500
.. , ..
\
'
......
,.'
N.-Cr
1300
J •
1100 I
r\: . I
I I F. -v
, I Fc-Cr
I I ~-)(-)l-X C.-V
I I Q •• 0 Co-C.
NI.V
Fc-V~! :
• • • • • N:-Cr
-~I
Fc-Cr _ _ 1
900 ,
t
\ I
'. I
'-
',0 8,0 9'0 10·0

AVERAGE GROUP NUMBER
Figure 2.12. fcc solid solubility limits as solidus and solvus, plotted in terms of average group
number (AGN) values for Ni-Cr, Ni-V, Co-Cr, Co-V, Fe-Cr, Fe-V. In the systems Ni-Co,
Ni-Fe, and Co-Fe, continuous fcc solid solutions exist. (From Hume-Rothery et al. 18 with
permission.)
S8 Chapter 2
A different interpretation of the electron concentration effects is presen-

ted by Engel and Brewer as will be discussed in Appendix A. Other methods,
such as the average group number (AGN) versus temperature diagrams, IS
have been proposed to avoid the controversy regarding the values of e / at.
for the elements. For this purpose, AGN is taken to be the number of
electrons outside the inert gas shell of each metal; e.g., for an equiatomic
alloy of Ti-V, AGN is 4.5 since Ti contributes four electrons, and V, five
electrons per atom. The range of AGN is 1 to 10 for groups headed by K
to Ni. An interesting ploes of AG N versus temperature for the face-centered
cubic solid solubility limits of V and Cr in fcc ')1- Fe, {3-Co, and Ni is shown
in Fig. 2.12.
Interstitial phases and compounds: The interstitial alloy phases and
compounds are formed by small atoms of metalloids consisting of H, B, C,
N, and sometimes 0 and Si, and large atoms of transition elements that
provide the interstices, or voids, for the metalloids. The rule governing
interstitial solubilities is called the Hagg rule.
Rule 4, Hiigg rule. If the ratio of the atomic radius of a metalloid to
the atomic radius of a metal is smaller than 0.59, then the metal and metalloid
may form an interstitial solid solution in which the lattice of the metal is
usually stretched depending on the size difference and the concentration
of the metalloid. Frequently, intermediate phases and compounds may also
be formed. The stoichiometry of some of the intermediate phases roughly
corresponds to An B where A is the metal and n is often equal to 0.5, 1, 2,
or 4, but large deviations in stoichiometry may occur in all such phases.
The lattice strain about a metalloid atom is usually quite large; hence, the
solubilities of metalloids are generally small and usually decrease with
increasing sizes of the metalloid atoms. As the temperature increases, the
solvent metal tolerates greater degrees of lattice strains and the solubilities
generally, but not always, increase with increasing temperature. When the
radius ratio is more than 0.59, the solubilities of metalloids in metals become
extremely small, but intermetallic compounds may be formed, wherein the
crystal structures of solvent metals become highly distorted or modified.
References

2. A. Alper, editor, Phase Diagrams, three volumes, Academic Press, New York (1970); H.
C. Yeh in Volume I, p. 167; T. B. Massalski and H. Pops in Volume II, p. 221.
3. A. G. Guy and 1. 1. Hren, Elements of Physical Metallurgy, Third edition, Addison- Wesley,
Reading, Massachusetts (1974).
4. A. Reisman, Phase Equilibria, Academic Press, New York (1970).
5. Bulletin of Alloy Phase Diagrams, Vols. 1-5, ASM, Metals Park, Ohio (1978-1984).
6. 1. Prigogine and R. Defay, Chemical Thermodynamics, translated by D. H. Everett, Long-

mans, Green, New York (1954).
7. S. Wagner and D. A. Rigney, Metall. Trans. 5, 2115 (1974).
8. M. Hansen and K. Anderko, Constitution of Binary Alloys, McGraw-Hill, New York (1958).
9. R. P. Elliott, Constitution of Binary Alloys, First Supplement, McGraw-Hill, New York
(1965).
10. F. A. Schunk, Constitution of Binary Alloys, Second Supplement, McGraw-Hill, New York
(1969); J. F. Smith and Z. Moser, 1. Nuclear Mater., 59, 158 (1976).
12. M. Hillert, Physics 1038,31 (1981).
13. See, e.g., the articles by (a) M. Hillert, (b) P. J. Spencer, and (c) P. L. Lin, C. W. Bale,
and A. D. Pelton, in Calculation of Phase Diagrams and Thermochemistry of Alloy Phases,
edited by Y. A. Chang and J. F. Smith, Metal!. Soc. AIME (1979).
14. L. Kaufman and H. Bernstein, Computer Calculation of Phase Diagrams, Academic Press,
New York (1970).
15. H. Gaye and C. H. P. Lupis, Metall. Trans.6A, 1049 (1975).
16. C. W. Bale, A. D. Pelton, and W. T. Thompson. An on-line computer program (F*A*C*T)
is available for phase-diagram calculations through McGill University/Ecole Polytech-
nique de Montreal, Canada.
17. 1. Ansara, Int. Metals Rev. No.1, p. 20, Metals Society and ASM (1979).
18. W. Hume-Rothery, R. E. Smallman, and C. W. Haworth, The Structure of Metals and
Alloys, The Institute of Metals, London (1969).
19. W. Hume-Rothery, in Phase Stability in Metals and Alloys, edited by P. S. Rudman, J.
Stringer, and R. 1. Jaffee, McGraw-Hill, New York, p. 3 (1967).
20. T. B. Massalski, in Theory of Alloy Phase Formation, edited by L. H. Bennett, AIME,
Warrendale, Pennsylvania (1980).
21. K. A. Gschneidner, Ref. 20, p. 36.
22. K. A. Gschneidner, Ref. 20, p. I.
23. J. Friedel, Trans. Metall. Soc. AIME 230,632 (1964); see also pp. 8-9, and 38-39, in Ref.
20.
24. T. B. Massalski, in Physical Metallurgy, edited by R. W. Cahn, North-Holland, Amsterdam
(1965); see also C. S. Barrett and T. B. Massalski, Structure of Metals, McGraw-Hill, New
York (1980).
3
Statistical Thermodynamics
Statistical thermodynamics is the discipline that develops thermodynamic

relationships from the mechanics of submicroscopic particulates (for
brevity, particles) coupled with the laws of statistics. I ,2 The particles of
interest in this book are atoms and molecules, but we shall also have brief
remarks on photons, electrons, and nuclear components. Since classical
mechanics does not always lead to the most elegant results, we shall
use a limited number of rules and axioms from quantum mechanics.
The reader interested in basic quantum mechanics is referred to
standard texts 3 ,4; proficiency in this field is helpful but not essential for
our purposes. This chapter is intended to be a thorough background for
acquiring the important procedures of statistical mechanics as related to
thermodynamics.
Distribution of Independent Particles
The distribution of independent particles in various ways depends on

the following general axiom, based on actual enumeration of all the possible
arrangements for a given set of particles and conditions.
GENERAL AXIOM. If the first arrangement (e.g., selection, ordering) can

be made NI different ways, after which the second arrangement can be made
N2 different ways, independently of the first arrangement, then both arrange-
ments can be made in NI N2 different ways.
The result is based on actual construction of independent arrangements

with a few particles and then generalization to any number of particles. For
example, if we have two sets ofietters, (A, B, C) and (E, F), then the number
61
62 Chapter 3
of ways we can select one letter from the first set is 3, and from the second
set, 2, so that there are 3 x 2 ways of making both selections in the stated
sequence, i.e., AE, AF, BE, BF, CE, and CF. An obvious extension of this
axiom is for more than two sets when there are Nh N 2, N 3, ... independent
particles; the total number of arrangements is then N\ x N2 X N3 X ••••
The number of distinct ways of arranging N different objects all taken
together is N!, and this result is called the permutation of N different
objects. The deduction is easy for three letters A, B, C, because the first
letter can be selected three different ways, after which the second letter can
be selected two ways, leaving one.letter to be selected last so that we have
ABC, ACB, BAC, BCA, CAB, and CBA, or 3! = 6. Among the N objects,
if N\ were alike, then N\ ! of all the arrangements would be indistinguishable;
therefore, the total number of distinct arrangements (or distribution) is
(3.1)
We now consider a system containing N independent particles which

are identical. Let there be gh g2, ... ,gj boxes, each set of boxes correspond-
ing to each set of energies eh e2, ... , ej, respectively, called the energy
levels. The occupation number Nh N 2, ... , N of each box is the number
j
of identical particles contained in each box. Thus, we have
Boxes: gh g2,.··, gj
Energy levels: eh e2, ... , ej
Occupation numbers: Nh N 2, ... , N j
Each set of gj of the ith set of identical boxes differing from other boxes
represents the degeneracy of that level as given by quantum mechanics, i.e.,
the number of single-particle wavefunctions that yield the corresponding
single energy level. Note that gh g2, ... are the degeneracies, but eh e2, ...
are themselves not degenerate. The set Nh N 2, ... , N taken altogether is
j
called a distribution. A change in any or all of selected N j represents a new

distribution.
Fermi-Dirac Statistics
Let there be N\ identical particles in g\ identical boxes with the

restrictions that N\ ,,;;;; g\ and that each box has no more than one particle
Statistical Thermodynamics 63
in it. The number of boxes, each containing one particle, is Nt. and the
number of empty boxes is g\ - N\. The number of ways, Dt. for arranging
g\ boxes, of which N\ are alike and each contains one particle, and g\ - Nt.
which are empty and also alike, is given by equation (3.1), i.e.,
(3.2)
For a given set of occupation numbers, the total number of distributions is

the product of all D j ; hence,
D= n
j
j=\
D = n
j
j
,.
g.!
j=\ Nj!(gj - N;)!
(3.3)
We use the Stirling approximation, In g\! = g\ In g\ - g\ and similarly for

N\!, to write equation (3.3) as follows:
In D =.L j [
gj In (g.)
~ + N In (g. - N)]
-'--.-'
j (3.4)
,=\ g, N, N,
The particles obeying equation (3.3) or (3.4) are called fermions, typical
examples of which are electrons, neutrons, and protons which, according
to the Pauli exclusion principle, cannot occupy the same state with another
particle, or no two such particles can have the same quantum numbers.
Bose-Einstein Statistics
If we remove the restriction in the preceding section that each box may
contain no more than one particle, then the distribution becomes completely
different. Each box may now contain any number of particles; thus, for five
particles and four boxes, one of the possible arrangements is
A AAA A
Four boxes are generated by inserting the dashed, movable partitions, and
since there are three of these partitions, the number of movable partitions
is one fewer than the number of boxes so that we have g\ - 1 partitions.
The total number of objects to be arranged is 8, or g\ - 1 + N\. The number
64 Chapter 3
of ways of arranging g\ - 1 + N\ objects into g\ - 1 and N\ objects is also

given by equation (3.1), i.e.,
(3.5)
Again D is the product of all D; so that
D= n; (g;-I+N;)! (3.6)
;=\ N;!(g;-1)!
and when the Stirling approximation is used after neglecting 1 in (g; - 1),
we obtain
In D =L; {- g; In (g)
,=\
- .-'-. + N; In
g, + N,
(g + N)}
-'--.-'
N,
(3.7)
The particles (e.g., photons) that obey equation (3.6) are called bosons.
Boltzmann Statistics
Consider equation (3.2) for fermions when g; is very much greater than
N;, i.e., g; » N j ; hence, most of the boxes are not occupied and equation
(3.2) can be rewritten as
D\ = g$g\ - 1)(g\ - 2) ... (g\ - N\ + 1)[(g\ - N$!]

N\ !(g\ - N\)!
g\(g\ - 1) ... (g\ - N\ + 1)
(3.8)
N\!
There are N\ factors in the numerator, and since g\ - N\ + 1 and the

preceding factors are very close to g\ because g\ is very much greater than
Nt. equation (3.8) becomes
(3.9)
The total number of distributions, D, is therefore

. N
D=n~ (3.10)
;=\ N;!
Again for In D, we obtain
In D = L [N j In(gJ N j ) + NJ (3.11)
i=1
The particles obeying equation (3.10) are called the corrected boltzons, and
for the most part we shall be concerned with these particles. In the historical
development of equation (3.10) the right-hand side contained N! as a factor
immediately after the equal sign and such particles were simply called the
uncorrected boltzons. We shall have no specific use for uncorrected boltzons.
The preceding equations for D are functions of N j only since degeneracies
are generally not variables.
Distribution Laws
Consider a system consisting of N fixed number of particles with a

fixed total energy, E, and a fixed volume, V. Such a system was called a
microcanonical ensemble by 1. W. Gibbs. Statistical thermodynamic proper-
ties of this system are functions of N, E, and V. The requirements that
(first restrictive condition) (3.12)
and
(second restrictive condition) (3.13)
are called the restrictive conditions wherein E j is the energy of particle i.

We proceed to solve for the maximum value of In D within the restrictive
conditions specified by equations (3.12) and (3.13). We rewrite the differen-
tials of these equations as
(3.14)
where aEjaNj = Cj.

We next assume that the expression for D, within the constraints
imposed by the restrictive conditions and within a very close approximation,
is the same as the expression we shall obtain for the maximum term for D.
66 Chapter 3
This assumption states that only the maximum term of D in very close
proximity of E contributes significantly to D, or stated differently, if D
were plotted versus energy for a closed system, D would sharply peak out
at the prescribed value of E for the system. The justification for using the
maximum term for D is that the resulting statistical thermodynamic
equations accurately describe the observable thermodynamic properties.
For the maximum value of D, d In D is zero; hence,
i aln D
d In D = 0 = L - - dNi (3.15)
i~1 aNi
Multiplication of the first and second relationships in equation (3.14) with

the Lagrangian multipliers a and -{3, respectively, and combination of the
result with equation (3.15) yields
Li (alnD aN
--. + a - .- {3ei
)
dNi = 0 (3.16)
,=1 aN, aN,
Since each dNi is finite and nonzero, equation (3.16) is satisfied if, and
only if, the coefficient of each dNi is zero; hence,
aln D aN
- - + a - - {3e = 0 (3.17)
aNi aNi '
From equation (3.12), aN/aNi = 1, and
aln D
--+ a - {3e = O' (i=I,2, ... ) (3.18)
aNi "
Equations (3.12), (3.13), and (3.18) provide the simultaneous equations for
solving the unknowns N i , a, and (3.
For fermions, the first term on the left side of equation (3.18) is equal to
In(gi - N i ) -In N i ; because gi is independent of Nio therefore,
Ni 1
or - = ---:;:--- (fermions)
gi e -ex e{3e, + 1 '
(3.19)
Likewise,
(bosons) (3.20)
and
(boltzons) (3.21)
Equations (3.19)-(3.21) are called the distribution laws for systems consist-
ing of independent particles.
The values of a and f3 can now be obtained for boltzons. For this
purpose, we rewrite equation (3.12) by using equation (3.21) so that
(3.22)
For compact writing we use
(3.23)
and call q the molecular partition function. If the energy, Ej, is separable
to Ej and Ej' then q = qll; = (L gj e-/3£,) (L gj e-/3£j). For example, Ej and Ej
may refer to kinetic and potential energy, respectively. Equation (3.23) can
be substituted in equation (3.22) to write
eOq = N (3.24)
We solve for eO from this equation and substitute it into equation (3.21)
to obtain N j and then use equation (3.13) to obtain
(3.25)
where we assume that Ej is a function of volume and therefore q is a function

of f3 and V. Since E is known from measurements, gj and Ej are known
from molecular properties of a system such as an ideal gas, and f3 can then
be determined from equation (3.25). For this purp~se we substitute q and
the known value of N to obtain a from equation (3.24). The value of f3
can be obtained by using a simple ideal gas such as neon or argon. According
to the kinetic theory of gases, the molecules of an ideal gas are point particles
randomly moving in space without exerting attractive or repUlsive forces
on one another. Consider that there is 1 mole of an ideal monatomic gas
consisting of N molecules and occupying V cm 3 of volume. An isolated
cube of 1 cm contains N / V molecules of this gas. Let m be the mass of
each molecule and u its average velocity and assume that one-third of all
molecules move in each of the x, y, and z directions as shown in Fig. 3.1.
68 Chapter 3
~A~""'B--i-I- Y
e
x
1-1 cm-f
Figure 3.1. One-centimeter cube containing an ideal gas.
A molecule at "A" may move toward "B" to collide with the surface at
"B" and then rebound and travel to "e" and return to "A" again to repeat
its journey. In this process, each molecule travels a distance of 2 cm for
each collision on the selected surface dBe. When a molecule rebounds from
a surface, its average velocity changes from +u to -u and the change of
its momentum at the surface becomes mu - ( - mu) = 2mu. According to
Newton's laws of motion, the force exerted on a surface by a moving body
is equal to the rate of change of momentum and since this force acts on
dBe which has an area of 1 cm 2 , it also represents the pressure. The time,
t, necessary for each collision is t = (2 cm)/(u cm/sec) = 2/ u, and thus the
rate of change of momentum is 2mu/(2/u) = mu 2 which is the pressure
due to each molecule. For all the molecules moving in the direction of y,
i.e., N /3 V, the pressure is
P = (N /3 V)mu 2 (3.26)
MUltiplying both sides by V and observing that, from experiments, PV =

RT, it is readily seen that
PV = (2N /3)(mu 2 /2) = RT = NkT (3.27)
where k = R/ N is the gas constant per molecule, i.e., the Boltzmann

constant.
The kinetic theory states that the internal energy of a monatomic gas
consists entirely of the translations, or kinetic energy of its molecules;
therefore,
(3.28)
Combination with equation (3.27) gives
E = 1.5RT; E;(average) = E/ N = mu 2 /2 (3.29)
Consequently, E; in q = L g; e- f3E , is equal to mU7/2 where the velocity U;

is the quantized individual velocity component. The exponent of e must be
dimensionless; hence, f3 must be proportional to (mu 2 )-t, and since
(kT)-1 = (mu7!3)-1 from equation (3.27), then f3 is proportional to (kT)-t,
and in fact a more rigorous treatment beyond our scope also shows that
f3 = (kT)-I.
Entropy and Related Properties
The energies, E;, are functions of volume only. The reversible work
(dw) done on a system is given by dw = L N;( dEJ dV) dV = L N; dE;. From
E = N;E;, dE = L (E; dN; + N; dE;); however, from the first law of
thermodynamics, dE = dqth + dw where qth is the thermal energy exchange.
Consequently,
(3.30)
The second law of thermodynamics gives the existence of entropy S, by the

infinitesimal reversible thermal energy dqth divided by T, dS = dqth/ T;
therefore,
E·
dS = ~~dN (3.31)
'-T I
It was shown by equation (3.17) that
aln D
~ - - dN =
'- aN; I
~
'-
(f3E. dN - a dN)
I I I
(3.32)
The last term is zero because for a closed system a L dN; = a dN = o. The
left side of this equation is equal to L d In D where D is a function of N;
only. If the most probable value of D is the only significant contribution
to L d In D, then the summation can be replaced by the most probable
value of Dmp which we designated again as D so that
1 E·
d In D = f3 ~
'-
E dN
I I
= -k'-T
~ ~ dN
I
(3.33)
70 Chapter 3
Thus, from equations (3.31) and (3.33),
dS = kd In D (3.34)
The integratin of this equation yields the definition of the entropy, i.e.,
S = kIn D (3.35)
where it is assumed that for D = 1, S = O. For boltzons, we have
(3.36)
From equation (3.21), N j = gj e a e- f3 ", and from equation (3.24), e a = N / q,

so that gd N j = (q/ N) e f3 "; hence, with f3 = (kT)-t,
S=k[(ln~)LNj+L(f3Njej+NJJ =kNln !+~+kN

(3.37)
where the last equality was obtained by using equations (3.12) and (3.13).
The Helmholtz energy is given by
q
A = E - TS = -kNTln-- kNT (3.38)
N
and the Gibbs energy by G = A + PV = A + kNT; hence,
G = -kNTln!L (3.39)
N
where G (as well as A) at 0 K is taken to be zero for simplicity. It should

be noted from equations (3.24) and (3.39) that
a = G/(kNT) = G/kT (3.40)
where G = G / N is the Gibbs energy per particle. Similar equations can

be obtained for fermions, and bosons, starting with the appropriate D but
the resulting equations for G and a are identical with equations (3.39) and
(3.40).
Fermions, Bosons, and Boltzons
The electrons in metals and semiconductors obey the Fermi-Dirac

statistics, i.e.,
Nj e" e- f3e ,
(3.41)
gj e-" e f3e , +1 1 + e" e- f3e ,
where ex = GI kT and G here is often called the Fermi energy by physicists,

and denoted as Er by them. The equations corresponding to fermions are
g-e- e" e -f3 e,

E =I Njej =I -f3e (3.42)
1+ e e
I I "
i
(3.43)
A = E - (SI kf3) = NkTex - kT I In(1 + e" e- f3e ,) (3.44)
For fermions, ex = GI kT can be substituted into equation (3.41) to write
Nj
(3.45)
gj [e f3 (e,-G)] +1
As T approaches zero, or as f3 approaches infinity, we have three

possibilities, i.e., ej > G for which NJ gj ~ 0; ej = G, for which NJ gj ~ 0.5;
and for ej < G, NJ gj ~ 1. Therefore, all the levels for the first case are
empty, and those for the last case are full, but those for the second case
are half full. Equation (3.45) will be discussed in detail in conjunction with
semiconductors. Neglecting the small contribution from the second case,
the value of N is given by
£.=6
N =I Nj = 'I gj; (T~ 0) (3.46)
Ej=O
If for the fermions in the following equation
e f3 (G-e,)
1 + e f3 (G e,)
(3.47)
72 Chapter 3
ej is much larger than G, and f3 is finite and small, or the temperature is

sufficiently high, then the second term in the denominator is negligible and
we obtain equation (3.21) for boltzons, i.e.,
Nj f3(G-g) N -f3g
-=e '=-e' (3.48)
gj q
As a result of ej » G, we have N j « gj, which is the requirement for

applicability of the Boltzmann statistics.
The plus sign preceding the exponential term in the denominator of
equations (3.42) and (3.45), and after In in equations (3.43) and (3.44),
must be changed to minus to obtain the corresponding equations for bosons.
These results can be obtained by starting with equation (3.20) and following
an identical procedure; the result is
1
(bosons) (3.49)
Again for ej » G and high temperatures, this equation becomes the same
as equation (3.48) for boltzons.
Gibbsian Ensembles
Equations (3.12), (3.13), and (3.23) through (3.25) show that q refers
to a closed system described by constant E, N, and V. A microcanonical
ensemble is therefore a closed system defined by E, N, and V. (Canonical =
standard.)
Consider equation (3.38), which can be rewritten as
qN
A = -kTln- (3.50)
N!
From thermodynamics, A is a function of V and T and if we let the

Q = qN / N!, then we have
A = -kTln Q; (3.51)
Therefore, Q now refers to a system defined by N, V, and T and such a

system is called a canonical ensemble.
Subtraction of equation (3.51) from (3.39) yields
G - A = PV = kT In(NNQ/ qN) "" kTln B (3.52)

where E is called the grand partition function, considered to be a function

of V, T, and G. The system to which E refers is an open system, called the
grand canonical ensemble, although a more useful concept is the isothermal
isobaric ensemble 0 defined by
G = -kTln 0 = -kTln (~) N; (3.53)
where equation (3.39) is used on the right side, and 0, like G, is a function
of N, P, and T. We shall generally have very little use of E and 0 in the
topics considered in this book, and frequently use q and Q.
Quantum Mechanics of Free Particles

Ideal gases, electrons or other charged particles in rarified spaces, and
electrons in semiconductors can be treated as free particles. We consider
first an ideal gas, and then briefly the electron gas.
The partition function of an ideal gas is given by a simple form of
equation (3.23), i.e.,
(3.54)
where i is the index that counts the distinguishable and degenerate states
individually, one by one. According to quantum mechanics, the energy of
a free particle is given by
(3.55)
where h is Planck's constant, a, b, and c are the quantum numbers, and

A, B, and C are the dimensions of the container or box; hence, this system
is often called "particles in a box." Resemblance to the kinetic energy in
terms of the momentum, p = mu, i.e., e = p2j2m, is self-evident. The details
of intermediate steps leading to equation (3.55) involve postulates that are
accepted without question because the end results always justify the means
used in quantum mechanics. Therefore, we accept equation (3.55) without
presenting the details of postulates and procedures leading to this equation.
The values of the quantum numbers in equation (3.55) are so large and
close enough that they may be considered as continuous from zero to infinity.
The partition function given by equation (3.54) can therefore be written as
(3.56)
74 Chapter 3
where qa = L exp( -f3h 2a 2/8mA2), and so on. This summation can be

obtained as follows:
(3.57)
where the integral is listed in standard mathematical tables. The replacement

of the sum by the integral is permissible when the coefficient of a 2 in the
exponent is much smaller than unity, and this requirement is always met
for A = 1 cm, and practically all the experimentally attainable temperatures.
The result for q is
(ideal gas) (3.58)
where ABC = V is the volume of the box. The energy E is obtained from
E = -N (alnaf3 q) l.5N
(3.59)
v f3
The energy of an ideal gas is given by E = 1.5 NkT, hence, again f3 = 1/ kT.
The free electrons obey the same treatment except a spin degeneracy
factor of 2 (electron spin up and spin down) must appear in equation (3.58);
thus,
(electron gas) (3.60)
where m_ is the mass of a freely moving electron. This equation will be

used in conjunction with electrons and holes in semiconductors.
Rotation and Vibration of Diatomic Molecules

Two unsymmetrical masses mj and mj , separated by a fixed distance
of rij, have a moment of inertia of I = r~mjmj/(mi + mj). According to
quantum mechanics, the energy of two such masses rotating in three
dimensions is
J(J + 1)h 2
Ej = 8rr 2 I (J = 0, 1, ... ) (3.61)
where J is the rotational quantum number. The equation for qrot is similar
to equation (3.57) and can be integrated with J as the variable from zero
to infinity for sufficiently large values of T to obtain
(3.62)
For a symmetric molecule this equation must be divided by 2. The Gibbs

energy and the enthalpy per mole from equation (3.62) are
H rot = RT (3.63)
where Grot(at 0 K) and Hrot(at 0 K) are taken to be zero for simplicity. We

present these equations for completeness in the treatment of particles since
there will be only a brief reference to them later in this book.
Two particles, i and j, may vibrate with respect to each other by forming
a harmonic oscillator. The vibrational energy of such an oscillator with a
reduced mass of mr = mjmj/(mj + mj ) is also quantized. The energy is
given in quantums of energy L(hv), where L is the quantum number and
v is the frequency of vibration. The partition function is again
-Lhp/kT _ 1
L
_ 00
qvib - e - 1 -hp/kT (3.64)
L=O - e
The last equality is obtained easily because the summation represents the
sum of a geometrical series. For sufficiently high temperatures and small
frequencies, hv/ kT is small and qvib is very closely approximated by
qvib = kT/hv; (hv/ kT < 0.1) (3.65)
The Gibbs energy and enthalpy per mole from equation (3.65) are
H=RT (3.66)
The vibrational frequency, v, of a harmonic oscillator is proportional to the

inverse square root of the reduced mass, a/Jrn;, where a is a proportionality
constant; hence, for small values of m" e.g., for an electron and a nucleus,
mr is very small and v is large and qvib == 1, and the contribution to the
related thermodynamic properties is negligible.
76 Chapter 3
Vibrations of Atoms in Lattices
Consider a simple crystal consisting of N atoms of a single element.

Each lattice point is occupied by one atom vibrating in three possible
geometrical directions with a constant frequency of IJ. The restoring force
to the mean position of the atom is proportional to the displacement from
the mean position. Therefore, each atom is regarded as three independent
but localized harmonic oscillators having the same frequency in all three
directions. One might also picture the atoms attached by a lattice of short
rubber bands to their sites, and vibrating with increasing amplitude, with
increasing temperature. The energy of the crystal of N atoms is assumed
to be due entirely to vibration. At 0 K, the amplitude of vibration is so
small that the vibrational energy and the heat capacity are zero.
Each atom in a crystal acts as a three-dimensional oscillator, i.e., there
are 3N oscillators for a crystal consisting of N atoms. Therefore, equations
(3.66) need to be multiplied by 3; thus,
G = -3RT In qvib, H =3RT (3.67)
The heat capacity of a solid at high temperatures is 3R from dH I dT, in

accord with the law of Dulong and Petit. Application of these equations to
alloys requires a realistic representation of the frequency IJ as a function
of composition and temperature. Attempts to accomplish this have been
largely empirical.
Potential Energy Functions
The potential energy functions generally attempt to correlate the energy

as a function of the distance of approach of molecules from the gases to
the condensed states. Such functions are usually empirical or at best semi-
empirical. Two symmetric and neutral molecules, or atoms, attract each
other to decrease their energy per particle, E'I N, by -AI r6, where r is the
intermolecular distance. The force causing this attraction is called the
London dispersion force [F. London (1930)) arising from an attraction
between mutually induced dipoles in the interacting molecules. However,
when r becomes smaller, repUlsive forces come into play because of inter-
molecular repUlsion. An empirical approximation for the repulsive energy
is given by E"I No = BI r12; therefore, the total energy per particle is
E A B
-=--+- (3.68)
No 12 r6 r
This equation is known as the Lennard-lones potential (1924). The para-

meters A and B are related to ro, the minimum point in Fig. 3.2, by setting
dE / dr to zero; the result is A = 2B / r~. The minimum energy is designated
as -e = E(min)/ No, and substituted in equation (3.68) with A = 2B/r~, to
obtain
E(min) 2B B B
-e = = --+-=--
Nor~2 r~2 r~2
from which B = er~2, and then A = 2er~. Therefore, equation (3.68) becomes
(3.69)
This equation, known as the Lennard-lones 6-12 potential, has been used
extensively for gases and liquids (see Refs. 5 and 6). It represents an
intermolecular pair potential in gases, but the concept has been extended
to the condensed phases.
0.6.---",-.---.----.---.---.---.---.
0.4
0.2 r-
t O.S 1.0 1.2 1.4 1.6 I.S 2.0 2.2
i. 0
No
- 0.2
-0.4
-0.6
- 0.8
- 1.0
-1.2~--~--~--~--~--~--~--~--~
Figure 3.2. Lennard·Jones 6-12 potential for a pair of molecules. Vertical scale is based on
£ = J, and horizontal scale, TO = 1 in equation (3.69). Minimum point is at T = TO = 1 and
E(min)/No = -£ = -I.
78 Chapter 3
Machlin 7 has used a function similar to equation (3.68) for the structural
stability of alloy phases, lattice parameter defects, and other related
properties:
a {3
E=--+-
r4 r8
(3.70)
The parameters a and {3 are related to physically and mechanically measur-

able quantities. The pseudopotentials 8 will not be discussed here since brief
summaries are presented elsewhere. 8 They are not unique and not greatly
successful for the transition elements9 but often used by physicists.
Ionically bound crystals, such as NaCl, are also considered to have a
potential energy consisting of two terms. The attractive ionic forces con-
tribute E'I No = -Me 2 lr where E'I No is the energy per ion, M is the
Maedelung constant, and e is the electronic charge. The values of Mare
in the neighborhood of 1.7. The repulsive forces cause an increase in energy,
expressed by E"I No = B.I rm where m is an exponent that varies from
compound to compound, e.g., m = 6 for LiF and m = 8 for NaCI. The
parameter B. is related to M through the minimum point ro on the curve
for EI No versus r by B. = Me 2 r m - 1/m; therefore, the net potential energy
per particle is
-E=
No
(1
Me 2 - - +r---)
m I
r mrm
(3.71)
In general, all the potentials near the minimum energy point can be
closely approximated by an empirical quadratic function, i.e.,
(3.72)
where a is an empirical constant.

The overwhelming majority of potential energy functions are empirical
in nature. The validity of assumptions used in these potentials may often
be questionable but success has been obtained in a number of cases for
certain types of metals and their alloys.9
References
1. D. A. McQuarrie, Statistical Mechanics, Harper & Row, New York (1983).

2. F. C. Andrews, Equilibrium Statistical Mechanics, Wiley-Interscience, New York (1975).
3. A. S. Davydov, Quantum Mechanics, translated by D. Ter Haar, Pergamon Press, Elmsford,
New York (1976).
4. D. A. McQuarrie, Quantum Chemistry, University Science Books, Mill Valley, California

(1983).
5. J. O. Hirschfelder, editor, Intermolecular Forces, Interscience, New York (1967).
6. J. O. Hirschfelder, C. F. Curtiss, and R. B. Bird, Molecular Theory oj Gases and Liquids,
Wiley-Interscience, New York (1964).
7. E. S. Machlin, in Interatomic Potentials and Crystalline Dejects, edited by J. K. Lee, Metal!.
Soc. AIME, p. 33 (1981).
8. D. Stroud, in Theory oj Alloy Phase Formation, edited by L. H. Bennett, Metal!. Soc. AIME,
p. 84 (1980); see also W. A. Harrison, Pseudopotentials in the Theory oj Metals, Benjamin,
New York (1966).
9. R. Taylor, in Interatomic Potentials and Crystalline Dejects, edited by J. K. Lee, Metal!. Soc.
AIME, p. 71 (1981).
4
Theories of Solutions
The most general and simple theory intended to fit liquid and solid solutions
is the regular solution theory developed by vanLaar and Lorenz (1925),
Heitler (1926), and Hildebrand (1928). The term regular solution was
proposed by Hildebrand (1929) for solutions described by random
and yet nonideal behavior. This proposed behavior was properly called
the zeroth approximation to the regular solutions by Guggenheim,l as
distinct' from the first approximation, to be discussed later in this
chapter. Early developments in this field are presented in a number of
publications. 1- 3
Regular Solutions-Zeroth Approximation

The treatment involved in the zeroth approximation is based on the
assumptions that (1) the molecules* in solution are distributed randomly
despite nonzero enthalpy of mixing, (2) the component molecules are not
greatly different in size from one another so that the number of nearest
neighbors Z is the same for the pure components and for their binary
solutions, (3) the molecular interaction is limited to the nearest neighbors,
and hence the interaction beyond the nearest molecules is neglected, and
(4) the bond energy eij for a bond between unlike molecules i and j is
independent of composition and of temperature. The bond energy eij is
negative and refers to
Nj(gas) + ~(gas) = N(solution with ZN /2 bonds) (4.1)
where Nj and N j are the number of molecules of i and j, respectively, and

*Molecules and atoms are identical for solutions of metals on pp. 81-111.
81
82 Chapter 4
N is the sum of N j and ~,and for convenience, N is taken to be Avogadro's

number. There are Z/2 bonds per molecule and Z is the number of nearest
neighbors, which is also called the coordination number. Likewise, when
Nj(gas) and ~(gas) molecules condense into pure N j and ~ in the solid
or liquid state, they form ZNJ2 bonds of i-i type and Z~/2 bonds of j-j
type, respectively. We assume that we can assign Z values to the liquid in
the same way that we assign Z values to the crystals. The probability of
finding a molecule of i at a selected site is proportional to its mole fraction
Xj in solution, and that of finding another i next to the selected i is XjXj.
Likewise, the probability of finding a molecule of j and the probability of
finding a molecule i next to j is XjX j , and the sum of all probabilities for
all possible pairs is unity; thus,
(4.2)
For a multi component solution of band c components, equation (4.2) is

written as
b
I I XpXq = 1 (4.3)
p=1 q=1
The normalization of all probabilities to unity is therefore automatically

satisfied when the composition is expressed in mole fractions. The total
number of bonds, ZN /2, from equation (4.1) is distributed among the
probability terms in equation (4.2) and each term has its bond energy e jj ,
ejj, and ejj, so that the energy terms for the respective bonds are (ZN /2)x jx jejj ,
ZNxjxjejj, and (ZN /2)xjxAj, and the total energy of 1 mole of solution is
(4.4)
We assume that the energy E is nearly identical with the enthalpy H, because
the pressure-volume product PV in the definitional relationship H =
E + PV is negligibly small for condensed phases at ordinary pressures.
Likewise, the enthalpy for xjN molecules of pure condensed i is xjH~ =
(ZN /2)x j e jj where H~ refers to 1 mole of pure i. The corresponding enthalpy
for pure j is xjH'j = (ZN /2)xAj, and for the mixing process, i.e.,
Xj mole of pure i + Xj mole of pure j = 1 mole of i-j
the molar enthalpy of mixing (or solution) is
(4.5)
Theories of Solutions 83
where equation (4.4) is substituted for H to obtain the last equality, and
W;j, called the exchange energy, is defined by
(Wij = exchange energy) (4.6)
The enthalpy change, fl.H, is also called the molar enthalpy of formation
of solution. The arguments presented for the foregoing equations can be
readily extended to muIticomponent solutions to show that
where each term originates from each pairwise interaction. In equations

(4.5) and (4.7) fl.H is also the excess molar energy of solution, He. This
equation is identical in form with the second-order Margules equation
(1.87). Therefore, each term in equation (4.7) originates from each binary
system. The equation for fl.Hjo which is equal to H~, is identical to equation
(1.69), and for a binary solution,
(4.8)
The exchange energy Wij can be determined from a single experi-

mental value of fl.H or H~ at a convenient concentration Xj in equation
(4.5) or (4.8). The equations for a multicomponent solution can thus be
determined from the values of Wij' W;k,... for the constituent binary
systems.
The first assumption for the simple regular behavior, i.e., the random
distribution, signifies that the excess molar entropy Se be zero; therefore,
(4.10)
The equality of G~ and H~ signifies that, while H~ is not zero, S~ is zero,

and we observe immediately that S~ = 0 in G~ = H~ - TS~ is thermodynami-
cally suspect since H~ cannot be equal to G~ unless H~ is also zero, i.e.,
unless the solution is ideal. Further, a solution for which H~ is not zero
cannot have its molecules distributed randomly. The first approximation to
the regular solutions, to be discussed in the next section, confirms that
indeed S~ cannot be equal to zero unless H~ is also zero.
84 Chapter 4
Or---r---r---r---r-~
", .. 1..6
t -0.2
6G A
RT
-0.4
..........
1.0
t B
Gz 0.4
Figure 4.1. Regular solutions for various
values of a = NW / RT. Upper curves are for
0.2 e.G/ RT. Dashed horizontal line is tangent at
J and K to curve for e.G/ RT, with a = 2.6.
O~~~~~~--~~ L is critical point. Lower curves are for activity
o 0.2 0.4 0.6 0.8 1.0 G 2 of component 2; activity G, is symmetric
x
2
- to G 2 about vertical line through X2 = 0.5.
Equation (4.7) may also be treated according to equations (1.87)-(1.90)

for multicomponent solutions, after writing Oe = 6.H = He, to obtain
(4.11)
We return to equation (4.9) for a binary system and use 6.0 =

oe + O(ideal), with 6.0(ideal)* = RTxI In XI + RTx21n X2, to write
(a = NW/ RT; W=Wd (4.12)
where we use W instead of W 12 for simplicity in notation when we consider

only a binary system. It was assumed that W is independent of temperature;
consequently, a is inversely proportional to T, and the solution should tend
to ideality with increasing T. The dimensionless property 6.0/ RT is plotted
as a function of X2 at three different values of a in Fig. 4.l(A). The curve
* e.G(ideal) is sometimes unnecessarily called the configurational Gibbs energy and e.S(ideal),
configurational entropy.
for ll.G/ RT when a = 1 is concave up without maximum and inflection

points. The curve for a = 0 is very similar to that for a = 1, and therefore
is not shown. When a = 2, ll.G/ RT has an inflection point at L where the
first and the second derivatives of ll.G/ RT become zero at X2 = 0.5; thus,
(4.13)
(4.14)
are both zero for a = 2 at X 2 = 0.5; this is the critical point. At a > 2, the
curves for ll.G/ RT show two minima; e.g., for a = 2.6, the minimum points
are at J with X2 = 0.124 and at K with X2 = 0.876 and J and K are located
symmetrically with respect to each other. In the range of composition
between J and K, the solution separates into two phases, where the composi-
tion of each phase is constant but the relative proportions of these phases
vary from J to K. The tangent line to the curve at J and K is horizontal and
the intercepts with the vertical axes are given by ll.G 1/ RT = ll.G 2 / RT =
ll.G/ RT, with the values of ll.Gi calculated either at J or at K. The negative
values of a yield curves similar to that for a = 1 but with sharper dips.
Equation (4.14) for a> 2 is zero at the inflection points; e.g., for
a = 2.6, X2 = 0.252 and X2 = 0.748 are the inflection points, which are called
the spinodes.
The equation for ll.G2 can be derived from equation (4.12); the
result is
(4.15)
where a2 is the activity of component 2. The values of a2 from this equation

are plotted in Fig. 4.1(B) corresponding to the selected values of a. For the
positive values of a the deviation from Raoult's law is positive, and for
a = 0 the solution is ideal (Raoultian). The horizontal inflection point for
a = 2 is also indicated by L in Fig. 4.1(B). The curve for the activity of
component 1, al> is symmetric to that for a2 for each corresponding value
of a. The deviation from ideality is negative for the negative values of a
as expected, but not shown in Fig. 4.1(B). It can be shown that for a > 2,
the maximum and minimum values of a2 for X 2 versus a2 correspond to the
inflection points of the curve for X2 versus ll.G/ RT.
Preliminary Concepts
The first approximation to the regular solutions has been obtained in
its correct form by permutation of molecules. 4 The earlier approximation
86 Chapter 4
by Bethe 5 and by Guggenheim! were obtained by permutation of bonds,

leading to physically meaningless spliced molecules as shown in Fig. 4.2.
Thus, for example, for tetragonal crystals, such as certain silicon alloys for
which Z = 4, permutation of bonds of i-j type in Fig. 4.2(a) may formj-i
bonds in Fig. 4.2(b) and create physically impossible spliced molecules. A
critique of the method based on the permutation of bonds is given later in
this chapter after the presentation of the correct approximation based on
the permutation of molecules. A binary system i-j will be considered first
and then the results will be extended to multi component systems.
The total numbers of like and unlike bonds are useful in the treatment
of all types of solutions. In an ideal solution, the component molecules are
distributed randomly, and we recall in conjunction with equation (4.4) that
(ZN /2)x; = number of i-i bonds

(ZN/2)2x jxj = Z(NxjxJ = number of i-j bonds
(ZN/2)x; = numberofj-jbonds
For this distribution, we denote for brevity that
(4.16)
hence, the number of i-j bonds is equal to Zy* in an ideal solution.

Likewise, the total number of unlike bonds in a nonideal solution may be
denoted by
Total i-j bonds = ZY (4.17)
a b
Figure 4.2. Permutation of bonds in two·dimensional crystals. Bonds of i-j type in (a) form
spliced molecules of the types in (b) after permutation.
where Y is the parameter that gives the correct number of i-j bonds when
multiplied by Z. If we accept the concept of half-bonds for the moment,
then there are 2ZY half-bonds, half of which, i.e., ZY, have to be i half-bonds
emanating from the molecules of i, and the other half, ZY, have to be j
half-bonds emanating from the molecules of j. However, prior to mixing,
the pure i molecules have ZNj half-bonds of i type; therefore, after mixing,
the i half-bonds, not connected to the j half-bonds, are ZNj - ZY, and these
half-bonds must meet with each other to form i-i bonds. Likewise, the total
number of j half-bonds connected to the j half-bonds is Z~ - ZY. The
reader can verify these relationships by using unidimensional crystals, i.e.,
beads in necklaces as in Fig. 4.3. The total number of whole bonds of each
type in the mixture is therefore
i-j bonds = ZY
i-i bonds = (Z/2)(N j - Y) (4.18)
j-j bonds = (Z/2)(~ - Y)
It is very important to remember that the total number of unlike bonds is

equal to ZY. The total number of bonds before and after mixing is evidently
the same, i.e., NZ /2. Since the number of bonds per molecule is Z /2, then
the net number of molecules having unlike neighbors is 2 Y, the net numbers
/~"
Figure 4.3. Molecules in unidimensional pure crys·
tals i andj, and their solution i-j; Z = 2. i-j bonds =
Zy = 4 in the bottom configuration; hence, Y = 2,
N j - Y = 4, and ~ - Y = 2.
\.
--
i-j
()---()- --
.J
88 Chapter 4
of i and j molecules surrounded by like neighbors are (N; - Y) and

(Nj - Y), respectively. Figure 4.3 shows the various types of bonds for
unidimensional crystals for which Z = 2. To eliminate the error from a
limited number of molecules, the dangling terminal bonds in a linear crystal
are tied to form crystals similar to the beads in a necklace. For a large
number of molecules in a linear crystal, the end effect is negligible. The
upper drawings show pure i and pure j, and the lower drawing shows one
of the possible resulting alloys wherein the solid lines are the like bonds
and the dashed lines are the unlike bonds. We observe that ZY = 4 since
there are four i-j bonds, but Z = 2; hence, Y = 2. Therefore, the i-i bonds
are (6 - 2)Z/2 = 4, and the j-j bonds are (4 - 2)Z/2 = 2. In this case,
N; - Y = 4 and Nj - Y = 2 since Z = 2. It is also very easy to construct
such crystals for Z = 4 and obtain the same relationships as will be seen later.
The equilibrium among the net numbers of molecules having the like
and unlike bonds is
-i- + -j- = ···i··· + ...j ... (4.19)

I I
where one molecule of i with Z/2 of i-i bonds (or Z half-bonds) and one
molecule of j with Z/2 of j-j bonds (as shown by the solid lines for Z = 4
on the left side) react to form one molecule of i with Z/2 of i-j bonds and
one molecule of j with Z/2 bonds (as shown by the dotted lines on the
right side), so that altogether two molecules with opposite neighbors are
formed. Further justification for this reaction will be presented later. For a
solution formed by N; and Nj molecules of pure components i and j, the
foregoing reaction has the following numbers of species at equilibrium:
[(N; - Y)] + [(Nj - Y)] ~ 2Y (4.20)
where N; - Y and Nj - Yare the net numbers of reactant molecules in

reaction (4.19), and 2 Yare the net numbers of product molecules.
Distribution of Molecules
One-Dimensional Crystals
We shall first obtain the distribution of molecules for Z = 2 and then
extend it to Z = 4 to 6, where Z = 6 occurs in simple cubic and two-
dimensional hexagonal crystals. We take another set of arrangements shown
in Fig. 4.4 to derive the total number of arrangements for given values of
a b
Figure 4.4. Arrangements of molecules in unidimensional crystals (Z = 2). Y = 3 for (a),
Y = 2 for (b). Open circles are i molecules; black circles, j molecules.
Y, and Ni = ~ = 4. In Fig. 4.4(a), Y = 3 since there are 6 i-j bonds. There

are three locations I, II, and III, and one molecule of i is affixed on each
location to identify it. The number of locations is the same as Y. There is
one molecule of i next to I which can be placed next to II or III without
changing the value of Y and this is the number of distributable molecules
which is equal to Ni - Y. Likewise in Fig. 4.4(b) there are two locations I
and II; Y is therefore 2, and there are 2 molecules (Ni - Y = 2) which can
be placed next to II. The number of locations to which the molecules are
distributed is Y - 1 in both figures. We emphasize that the initial configur-
ation, or the configuration we have to start with is such that all the i beads
are placed together in I except the individual ones identifying II, III, ....
We assume at first that the locations I, II, ... are at fixed positions, but we
shall remove this restriction later. The number of ways of arranging Ni - Y
distributable molecules to the sites next to Y - 1 locations is
[( Ni - Y) + (Y - 1)]!/ (Ni - Y)! ( Y - 1)! (4.21)
The same formula is also applicable to the j molecules; hence, for both
permutations we have
(Ni -1)!(~ -1)!/(Ni - Y)!( Y -1)!(~ - Y)!( Y -1)! (4.22)
We remove the restriction on the system that I, II, and III have fixed
positions by rotating each configuration by one molecular position at one
time and repeat this procedure N times. When the rotations are constructed
on paper for all the configurations for a given value of Y, it is seen that
each configuration is unnecessarily repeated by a factor of Y. The number
of ways of arranging the molecules for a given value of Y is therefore N / Y
times the preceding expression, i.e.,
D = (Ni - 1)! (~ - 1)! N / Y! ( Y - 1)! (Ni - Y)! (~ - Y)!

= Ni!~!/(y!)2(Ni - Y)!(~ - Y)! (4.23)
90 Chapter 4
This equation is similar to but not identical with the Ising equation for
ferromagnetism. A stringent test for equation (4.23) is that the sum of D
for all the permissible values of Y is exactly equal to the sum of all the
possible configurations given by the random distribution, D rm :
yo
D rm = L D = N!/Nj!~! (4.24)
Y=m
where m is the lowest geometrically possible value of y, which is 1 for

linear crystals when the number of unlike bonds is 2, and the upper limit
is yo = ~ for N j > ~, i.e., all the j molecules have unlike neighbors. This
equation can readily be verified for the system N j = ~ = 4 of Fig. 4.4 as
follows: Y = 1, D = 8; Y = 2, D = 36; Y = 3, D = 24; Y = 4, D = 2; the
sum of all D's is 70, which is equal to N!/(Nj!~!) = 8!/(4!4!).
Two-Dimensional Crystals
A two-dimensional crystal lattice with a coordination number of 4 is
shown in Fig. 4.5. The molecules in the crystal occupy the points of
intersection of the lattice. It is also possible to construct a close-packed
hexagonal two-dimensional crystal with Z = 6 but we shall not be concerned
with it because of geometrical difficulties in enumerating various configur-
ations. The dangling bonds in a crystal having a limited number of molecules
cause a relatively large error in our enumerations but for a very large number
of molecules, again the error becomes quite negligible. We can eliminate
this error by tying the dangling bonds i, 2, 3, and 4 with the lattice points
1 2 3 4
T T T T
I I I I
I I I
a' I --- a
A 8
b' eI --- b
I
I
c' L
---I C
f 9
d' 0 C
-~ d
l' 2' 4'
Figure 4.5. Two-dimensional crystal lattice with Z = 4.
1',2', 3', and 4', respectively, away from the reader on the reverse side of
the page and in such a manner that we form three squares, e.g., 1,2, 1',2'
form a square. This process is best visualized if the lattice is stretched and
wrapped over a sphere so that 1,2, 1',2' become quadrangular in shape.
Similarly, a, b, c, and d are joined to the lattice points a', b', c', and d',
respectively, to form three more squares. The four corners A, B, C, and 0
form their square in this process so that there are altogether 16 squares for
16 lattice sites. (There are, likewise, N cubes for N lattice sites in a simple
cubic system for which Z = 6.) The corners A and C, as well as Band 0
are diagonally opposite to each other. This arrangement shows that any
configuration, such as the dashed L located on efg, returns to its original
position after four diagonal jumps along BO.
lt is convenient to formulate a procedure which can make the enumer-
ation of configurations clear and systematic. For this purpose, we take the
number of one of the component molecules either equal to, or larger than
the other without losing generality, e.g., N j ; " ~, and write a bond balance
for ~ molecules from equation (4.I6), i.e.,
j-j bonds + (0.5)(ZY) = (0.5)(Z~) (4.25)
This equation states that there are altogether Z~ half-bonds joined to the
j molecules, and two j-j of these are the j-j half-bonds and the remainder,
ZY, are the i-j half-bonds. Equation (4.25) is useful in computing Y from
the known numbers of j-j bonds. As an example, we consider N j = ~ = 8
in some detail. The number of configurations for all the permissible values
of Y has been counted and listed in Table 4.1. A stringent check on the
results is that the total number of configurations for all the possible values
of Y be equal to that given by equation (4.24).
When j-j is zero, Y is 8 and j molecules occupy diagonally opposite
positions, leaving the remaining positions for the i molecules. The sets of
positions for i and j molecules may be all interchanged once; therefore,
the total number of configurations is 2. The configurations for the values
of 1 and 2 for j-j are geometrically impossible as can be verified by
construction. The next configuration contains the cluster for which j-j = 3,
and there is only one such cluster as shown in Fig. 4.6{a). The four remaining
j molecules are surrounded entirely by the i molecules and they can be
arranged only one way. The entire configuration can be rotated four times
in 90° steps and placed in 16 different ways in the lattice, or stated differently,
Fig. 4.6{a) can be rotated in four ways and placed in 16 different positions.
The resulting number of configurations is 4 x 16 = 64. The configurations
containing the clusters of j-j = 4 are shown in Fig. 4.6{b)-{e). A selected
molecule in the cluster shown in Fig. 4.6{b) can be placed in 16 different
92 Chapter 4
I I
+ • • •
.L L'
• • • I I
a b c d e
........
L1
u
-1 -- ~ ]
0 ~
I
u
0
f 9 h I J
Figure 4.6. Arrangements of molecules for three, four, and five j-j bonds for N, = Nj = 8 and
Z=4.
lattice points and for each position the remaining three j molecules may
be arranged in four different positions and thus yield 4 x 16 = 64 configur-
ations. Figure 4.6{c) and (e) give 16 and 8 configurations, respectively,
because of symmetry, and Fig. 4.6{d) gives 32 configurations. The total
number of configurations for j-j = 4 is therefore 120. The configuration for
j-j = 5 is shown in the remaining parts of Fig. 4.6. Each cluster and its
mirror image in (f), (g), and (h) can be rotated four times and placed in
16 different sites to give 4 x 2 x 16 = 128 configurations for each figure and
384 configurations for three figures. The arrangement in (i) can be rotated
four times and thus yield {4 + 4 + 4)16 = 192 arrangements. The total for
j-j = 5 is then 576 configurations. The process of enumerating the configur-
ations for j-j = 6 is a very time-consuming task. There are three single
clusters of 6 j-j bonds involving squares with two attached bars that can
be rotated four ways and yield 3 x 4 x 16 = 192 configurations, and 15
one-piece clusters similar to (f) giving 1216 configurations. In addition,
there are nine two-piece clusters similar to (i) yielding 704 configurations,
all adding up to 2112 configurations. The results for j-j = 7 to 12 are listed
in Table 4.1 and it is readily seen that the total number of configurations
is equal to D rm • Similar enumerations for N j = 9 and ~ = 7 are also listed
in Table 4.1. The last column for the permutation of bonds, g, will be
discussed on page 107.
Table 4.1. Arrangements of Molecules for N j = N j = 8 and N j =9, ~=7 in

Two-Dimensional Lattice with Z = 4 a
Actual
arrange-
System j-j bonds Y y/y* ments D g
N j =8 12 2 0.500 8 392 257

Nj = 8 tt 2.5 0.625 0 l,ttO 1,482
y* =4 10 3 0.750 768 2,532 4,983
D rm = 12,870 9 3.5 0.875 1,600 3,837 10,169
8 4 1.000 4,356 4,900 12,870
7 4.5 1.125 3,264 4,934 10,169
6 5 1.250 2,112 3,920 4,983
5 5.5 1.375 576 2,442 1,482
4 6 1.500 120 1,176 2,574
3 6.5 1.625 64 427 24.4
2 7 1.750 0 tt2 1.12
7.5 1.875 0 20 0.20
0 8 2.000 2 2 7.8 x 10- 5
Total 12,870 25,624 48,995

N j =9 9 2.5 0.635 64 1,073 1,488
N j =7 8 3 0.762 624 2,240 4,838
Y* = 3.9375 7 3.5 0.889 1,920 3,588 9,478
D rm = tt,440 6 4 1.016 3,680 4,480 tt,403
5 4.5 1.143 3,136 4,386 8,446
4 5 1.270 1,392 3,360 3,801
3 5.5 1.397 512 1,993 1,005
2 6 1.524 96 896 147
1 6.5 1.651 0 293 10.41
0 7 1.778 16 64 0.27
Total tt,440 22,373 40,617
aD is the distribution of molecules; g, the distribution of bonds; D,m is given by equation (4.24).
The process of enumeration may be summarized as follows: (1) Con-

struct all the possible clusters for a given value of j-j, (2) eliminate duplica-
tions created by rotation, (3) place each cluster in the lattice and eliminate
those which do not fit or accommodate the other molecules, (4) eliminate
the duplications which occur when multiple piece clusters, such as those
in Fig. 4.6(c) and (e), are placed in various locations in the lattice, and (5)
check the results by comparing the sum of all configurations with Drm to
ensure that the duplications due to symmetry have been eliminated.
The results selected from enumerations for 20 additional systems 4 ,6 with
various values of N and ~/ N j are listed in Table 4.2 with the values of
94 Chapter 4
Table 4.2. Arrangements of Molecules for Various Values of N j and N j and for Z = 4
Actual Djactual
System j-j bonds Y YjY* arrangements arrangements
N j = 11 5 2.5 0.727 192 4.144

Nj = 5 4 3 0.873 688 2.093
Y" = 3.4375 3 3.5 1.018 1,664 1.161
D,m = 4,368 2 4 1.164 1,248 1.538
I 4.5 1.309 448 3.080
0 5 1.455 128 5.250
N, = IS IS 2.5 0.417 10 745.7
N j = 10 14 3 0.500 0
Y* = 6 13 3.5 0.583 2,300 34.74
D,m = 3,268,760
3 8.5 1.417 21,200 8.958
2 9 1.500 4,800 15.64
I 9.5 1.583 400 57.32
0 10 1.667 10 500.5
N j = 20 5 2.5 0.625 210 15.79
Nj = 5 4 3 0.750 1,650 5.182
y* = 4 3 3.5 0.875 8,850 1.884
D,m = 53,130 2 4 1.000 18,675 1.297
I 4.5 1.125 17,000 1.533
0 5 1.250 6,745 2.873
N, = 18 30 3 0.333 12 . 18,490
N j = 18 29 3.5 0.389 0
Y* = 9 28 4 0.444 3,456 1,204
D,m = 9.0751 x 10° 27 4.5 0.500 34,704 404
6 IS 1.667 9,780 113.5

5 15.5 1.722 1,152 206
4 16 1.778 504 82.6
3 16.5 1.833 144 40.3
2 17 1.889 0
17.5 1.944 0
0 18 2.000 2
N, =24 18 3 0.375 12 13,920
N j = 12 17 3.5 0.438 72 10,096
Y* = 8 16 4 0.500 8,820 298
D,m = 1.2517 X 109
N j = 84 24 4 0.298 100 1.047 x 107
N j = 16 23 4.5 0.335 218,000 3.556 x 105
Y* = 13.44 22 5 0.372 498,400 1.006 x 105
D,m = 1.3459 X 10 18
Table 4.2 (continued)
Actual D/actual
System j-j bonds Y y/y* arrangements arrangements
Ni =92 10 3 0.408 600 4,778

~=8 9 3.5 0.476 13,400 1,438
Y* = 7.36 8 4 0.544 284,450 373
D rm = 1.8608 X lOll 7 4.5 0.611 3,368,200 144
Ni =95 5 2.5 0.526 800 186
~=5 4 3 0.632 14,300 61.1
y* = 4.75 3 3.5 0.737 260,000 15.08
D rm = 7.5288 X 107 2 4 0.842 3.7726 x 106 3.553
1 4.5 0.947 2.2033 x 107 1.551
0 5 1.053 4.9208 x 107 1.240
Ni = 884 24 4 0.255 900 1.1707 x 1013
~ = 16 23 4.5 0.286 196,200 1.4660 x 109
Y* = 15.7156 22 5 0.318 8,805,600 7.0191 x 108
D rm = 7.7449 X 1033
Ni = 895 5 2.5 0.503 7,200 5,467
~=5 4 3 0.603 848,700 847
y* = 4.9722 3 3.5 0.704 2.3940 x 107 419
D rm = 4.8663 X 10 12 2 4 0.804 3.5389 x 106 30.18
1 4.5 0.905 2.0937 x lOll 4.068
0 5 1.006 4.6534 x 10 13 1.023
D from equation (4.23). A complete description of all enumerations is

presented elsewhere in detai1. 6 Accurate computations of D for the fractional
values of factorials require the use of numerical tables for the gamma
function which is defined by
I! = gamma function of (f + 1) (4.26)
where I is a fraction. For sufficiently large values off, e.g., I> 50, Stirling's
approximation may be used to obtain the numerical value of I!. The
maximum value of YI y* is given by (Nj + Nj)1 N j and decreases with
decreasing values of Njl N j for our convention of choosing N j ;;;. Nj. The
complexity of constructing and counting the configurations for high values
of both Nand Y I y* has limited our enumerations to relatively low values
of YI Y* for high values of N. The values of 10g(D/actual arrangements)
are plotted versus Y I y* as distinct points in Fig. 4.7. The results may be
96 Chapter 4
8
G N,' 84, NJ'16
0 N,' 90, Nj'IO I
I
'V N,' 92, Nj ' 8 I
0 N, '95, NJ' 5 / 6
ON,' 97, NJ' 3 I
~ N,' 884, NJ'16 I s
ON,' 890 NJ'IO I
9 N,' 892, NJ' 8 4~
SCALE ill u
en
D<N,' 895, Nj , 5
(} N,' 898, NJ' 2 /
I
/ 2
3
- =l. .I.oLF>J-=~: :" -
ZERO o
w
<i
u
2
en
Figure 4.7. Number of arrangements of molecules in two-dimensional crystals with Z = 4.

Vertical scales are log(D/actual arrangements) and horizontal scale is Y/ Y*.
summarized as follows: (1) All the points for a given value of N may be
fitted by a curve independent of the value of NJ ~. The dotted curves pass
through the actual points for N = 16 and N = 36 and the solid curves
ending in broken terminal portions on the left and on the right pass through
the points for N = 100 and 900. The dotted line for N = 25 is so close to
that for N = 36 that it has not been drawn except on the right side where
there is a small systematic difference between the two sets as shown by the
broken lines. The nature of the solid lines ending in broken lines for N = 16
and 36 will be discussed later, but in this paragraph the dotted curves for
N = 16 and 36 and the solid curves for N = 100 and 900 will be considered.
It is clear that D / actual arrangements is a function of Nand Y / Y* and
independent of NJ~. (2) The range of Y / Y* wherein the points follow
a concave-up curve becomes wider with increasing N on both sides of the
vertical line passing through Y/ y* = 1, or Y = y*. (3) The minimum point
at Y = Y* approaches the abscissa and becomes tangent to it with increasing
N. (4) The initial points on the left, marked as 1 close to the vertical axis
in Fig. 4.7, scatter more than the others because of the coarseness of the
clusters. A certain degree of scattering is also due to the fact that the values
of N j and ~ are rather small, and as they approach Avogadro's number
the scattering would undoubtedly smooth out.
We next consider the solid/broken lines for N = 16 and 36 in Fig. 4.7.
If we lower the dotted curves vertically so that they become tangent to the
abscissa at Y = Y*, we obtain very nearly the set of solid and broken
curves. We observe that as N increases, the distinction between the dotted
and solid curves disappears, and the dotted curve becomes tangent to the
abscissa as shown by the curves for N = 100 and 900. The solid portions
of these curves which extend from about 0.6 to 1.4 for Y / y* have actually
been obtained in the following manner. We correct equation (4.23) to obtain
Dc so that D/ Dc represents the solid portions of the curves, or Dc is the
actual number of arrangements for sufficiently large values of N. The
correction c, and the resulting corrected distribution function Dc are as
follows:
c = [(Z -l)(Z - 2)/Z]y b [(y/ Y*) -1] (4.27)
b = (Z - 1)/(2Z - I) (4.28)
D = ________~(_N~j_-_l~)!~(N~j-__l~)!_N__________
c (Y+c)!(Y+c-l)!(Nj - Y-c)!(~- Y-c)!
_ Nj!~!
(4.29)
- [( Y + c) !f( N j - Y - c)! (~ - Y - c)!
ijoth c and b were obtained by trial and error with the requirement that
c = 0 for Z = 2 and for Y / y* = 1. The empirical forms of c and b are not
unique because other similar formulas can be obtained for reasonably close
values of Dc. The correction given by equation (4.27) is negative for
Y/ Y* < 1, positive for Y/ Y* > 1, and zero for Y = y* because equation
(4.23) represents the correct number of configurations at Y/ y* = 1. This
98 Chapter 4
Table 4.3. Arrangements of Molecules for Simple Cubic Crystals; N = 512, Z = 6
Actual D/actual
Nj Nj D rm j-j y y/y* arrangements D arrangements
507 5 2.871 x 1011 3 4 0.8080 32,114,688 4.396 x 107 1.37

508 4 2.830 x 10° 3 3 0.7559 42,496 49,317 1.16
508 4 2.830 x 10° 2 10/3 0.8399 4,943,616 3,763,200 0.76
correction is imposed on the system purely by geometry when we proceed

from Z = 2 to 4. The functional form of c is appropriate in the range of
interest, Y / Y* from 0.6 to 1.4, and it is valid not only for Z = 2 and 4,
but also for Z = 6 as will be seen later. For Z = 2, c becomes zero and
equation (4.29) becomes identical with equation (4.23). The functional form
for c has been obtained emprically by fitting a large number of enumerations
listed in Tables 4.1 and 4.2. It is not important whether or not equation
(4.27) is rigorously correct and unique, but it is important for our purposes
+ 10
-10
-20
-30
-40
-50
_60~~ __ ~ ____ __-L____L -_ _- L_ _

~ ~
0.4 0.6 O.B 1.0 1.2 1.4 1.6

y/y*
Figure 4.8. Variation of numbers of arrangements of molecules with Y / y* in two-dimensional
crystals with Z = 4. Vertical scale is log[(D or g)/actual arrangements]; horizontal scale,
Y/ Y*. Upper curve is log(D/ Dc); circles are log (D/actual arrangements) from Fig. 4.7,
showing that Dc = actual arrangements.
that c is a very small and negligible fraction of each set of terms in each
pair of parentheses in equation (4.29) when N; and ~ become large enough
to be comparable to Avogadro's number. For example, at YI Y* = 1.4, c
is about 7.3% of Y + c for N; = ~ = 50, but for N; = ~ = 450, it is only
2.2 % of Y + c. Likewise, at Y1 y* = 1.2 the corresponding percentages are
4.1 and 1.2 for N = 100 and N = 900, respectively.
The configurations for the simple cubic system having Z = 6 are much
more difficult to enumerate. The error from the dangling bonds is eliminated
by a procedure similar to that for Z = 2 or 4, but for the limited number
of clusters we shall consider, it is important to observe that the number of
cubes is equal to the number of molecules and the total number of bonds
is ZN 12. A limited number of enumerations are listed in Table 4.3. The
results in the last column show that equations (4.27)-(4.29) are also satisfac-
tory for Z = 6. A greater number of enumerations with j-j = 2 to 5 and
~ = 4 to 6 is possible but quite time-consuming.
The variation of Dc with Y1 y* for N; = ~ = 450 and Z = 4, or an
equimolecular crystal of 30 x 30, is shown by the upper curve in Fig. 4.8.
The circles represent the actual points from Fig. 4.7 for D/actual arrange-
ments and show that DI Dc and D/actual arrangements are identical and
therefore Dc is equal to the actual number of arrangements. The lower curve
for log (g1actual arrangements) versus Y1 Y* will be discussed on page 107.
Equations
The energy of a system consisting of N; and ~ molecules is equal to

the sum of the bond energies for the sum of each type of bond. Therefore,
from equation (4.18) we obtain
(H = E) (4.30)
where e;;, ejj , and e;j are the energies for i-i, j-j, and i-j types of bonds,
respectively, and further, H = E. Substitution of equation (4.6) for W = W;j
in equation (4.30) gives
(4.31)
The configurational partition function corresponding to equation (3.51) on

page 72 is
(4.32)
100 Chapter 4
where Dc is number of possible configurations or arrangements, having the

same energy E as given by equation (4.31) for a given value of Y. For each
value of E, the corresponding value of Y is fixed, and the summation is
carried out over all values of Y corresponding to all the possible values of
E. According to Chapter 3, the sum in equation (4.32) may be replaced by
its maximum term, i.e.,
Q(max) = Dc e- E / kT = Dcexp[(-NiZeii/2 - ~Zejj/2 - YW)/kT]

(4.33)
The value of Y for which this equation is the maximum term in equation
(4.33) is obtained by differentiating In Q(max) with respect to Y and setting
the result to zero:
a In Q(max)/aY = (1 + e')[ -2In( Y + e) + In(Ni - Y - e)

+ In( ~ - Y - e)] - (W / kT) = 0 (4.34)
In this equation, e' is given by
e' = ae/aY = (1 + b)[(z -1)(z - 2)/z](y b / Y*) (4.35)
where the approximate equality holds for large values of Y; therefore, the
term b y b - I is not included in this equation. As N = Ni + ~ approaches
Avogadro's number, yb / Y approaches a value close to zero in the range
of 0.5 < Y / y* < 1.5. In addition, e in each of the logarithmic terms is
negligible in comparison with Y, Ni - Y, and ~ - Y; therefore, we could
have obtained an equation equivalent to (4.31) by using D of equation
(4.23) instead of Dc in equations (4.29) and (4.33). We rearrange equation
(4.34) and neglect e to derive
y2 = e- W / kT = e- NW/ RT (4.36)
(Ni - Y)(~ - Y)
where R = Nk is the gas constant. The left side of this equation is the law
of mass action or the pseudo-equilibrium constant for the following
reaction:
-i- + - j - = ···i··· + ...j ... (4.37)

I I : :
This reaction states that one molecule of i with Z half-bonds of i-i type
and one molecule of j with Z half-bonds of j-j type, both on the left and
shown with solid half-bonds, react to form one molecule of i with Z
half-bonds of i· .. j type and one molecule of j with Z half-bonds of j . .. i
type shown with dotted lines. The number of molecules of the reactants are
(Ni - Y) and (Nj - Y) and the products are Y and Y for the respective
species in reaction (4.37). The corresponding energy effect is clearly W =
(Z/2)(2eij - eii - ejj ). Reaction (4.37) is our short-range order-disorder
reaction.
First Approximation
The correct first approximation to the regular solutions may now be

obtained by starting with equation (4.36). We set, for brevity in notation,
n = e W / kT (4.38)
where n is always a positive quantity. We substitute n into equation (4.36)

to obtain
(1 - n) y2 - NY + NiNj = 0 (4.39)
There are two sets of values of Y satisfying this equation:
(+ sign unacceptable)
(4.40)
The set with the plus sign after N is not acceptable for two reasons: (1)
The maximum value of Y for Ni = Nj is Y = Ni for a perfectly ordered
crystal in which the number of j-j bonds is zero [(see equation (4.25)].
However, for n < 1, the plus sign would give Y > N i • For example, if
n = 0.64 and Ni = Nj = N /2, then Y = 2.5N (2) The value of Y cannot
be negative but for n > 1 it becomes negative for the plus sign in equation
(4.40) and positive for the minus sign. Therefore, we accept equation (4.40)
with the minus sign after N
The square root in equation (4.40) is inconvenient; hence, for simplicity
we introduce
(4.41)
102 Chapter 4
which transforms equation (4.40) into
(4.42)
where the last equality is obtained by substituting for (1 - n) from equation

(4.41). It is seen that for W = 0, f3 is unity and Y = NXiXj = y* as in a
random solution. We subtract Eii = N iZei;/2 and Ejj = ~Zej)2 from
equation (4.31) to obtain tlE "'" tlH in the same way that we obtained
equation (4.5); thus,
(4.43)
The equations for H~ and Hj can be obtained by substituting equation

(4.43) into (1.60); the results are
H- e. _ NWX· f3 + Xj - Xi. H- e. _ NWX· f3 + Xi - Xj

-
1 f3(f3 + 1) ,
- (4.44)
, J 'f3(f3+1)
It is evident that Hj can be obtained from H~ by interchanging the subscripts

i and j. For f3 = 1, these equations give the zeroth approximation form of
H~ = NWxJ, but f3 = 1 signifies that n = 1 or W = O. Therefore, it is not
possible to have a random solution when W is not zero. The zeroth
approximation is consequently inferior to the first approximation from a
theoretical point of view. The excess molar Gibbs energy of mixing, G e , is
obtained by integrating d( G e / T) = H e d(1/ T), i.e.,
(4.45)
The lower integration limit for the left side is zero because at very high
temperatures or for l/T = 0, Ge/T is zero. We wish to express d(T- 1 ) in
terms of f3 and for this purpose we differentiate equation (4.38) as follows:
dn = n(W/k)d(rl) (4.46)
We solve for n from equation (4.41) and then eliminate n from equation
(4.46) to obtain
(4.47)
where 1 - 4XiXj is equal to (Xi - XJ2 after we use (Xi + xj ? = 1 for a binary
system. The substitution of equation (4.47) into equation (4.45) gives
G=
_
T
e
Nk f
f3~1
f3
({3
4x·xf3
+ 1)({3 + Xi
IJ
df3
- Xj)({3 + Xj - Xi)
(4.48)
where T- 1 = 0 makes n = 1 in equation (4.38); hence, f3 = 1 from equation

(4.41). The quantity after the integral sign without df3 may be expanded into
J +
__ L + M = __-1 + XI·
+ X
J
f3+1 f3+~-~ f3-~+~ f3+1 f3+~-~ f3-~+~

(4.49)
where the numerators on the left side were determined by a well-known
algebraic method. Each term may now be integrated to obtain
Ge = NkTxi In [ f3 +
(I
X - X.]
/ + NkTxj In
[f3 +( JX - X]
)I (4.50)
Xi f3 + 1 Xj f3 + 1
This equation constitutes the first approximation to the regular solutions.

All the remaining equations in this section are derived from this equation.
The partial molar excess Gibbs energy of component i, 07, can be obtained
by substituting G e in equation (1.60) and carrying out the necessary
differentiation but this procedure is long and tedious. A simple and concise
procedure is to compare the terms of equation (4.50) with the right side of
G e = xi07 + xj07 and immediately obtain
- e
Gi = NkT In [f3 +( Xi - Xj]
); G-ej = NkT In [f3 +( Xj - Xi]
) (4.51)
Xi f3 + 1 Xj f3+1
Equation (4.50) is symmetrical in Xi and xj , i.e., the interchange of subscripts

leaves G e unchanged; hence, G e versus Xi is a symmetrical curve about the
line perpendicular to the Xi axis at Xi = Xj = 0.5.
The excess entropy Se is expressed by
j
S e = 2 NXiXj W - Nkx In [f3 + Xi - X] - N kx . In [f3 + Xj - Xi] (4.52)
T(f3 + 1) Xi(f3 + 1)
I J xj (f3 + 1)
For an ideal solution W = 0, f3 = n = 1, and equation (4.52) is zero as
expected, because se for a random solution is zero by definition. The
equations for 5~ and 5j are
- - [f3 + Xj -
s-e =NWxj Xi]
-Nkln [f3 + Xi - Xj] (4.53 )
I T f3 (f3 + 1) Xi (f3+ 1)
104 Chapter 4
(4.54)
Again, equation (4.54) can be obtained from (4.53) by interchanging i and j.

We proceed to test the success of the first approximation by setting
Xi = Xj = 0.5 and by using the available data on mixtures. The resulting
form of equation (4.50) is then
Ge = RT In [2/3/(1 + ~)] (4.55)
The corresponding form of equation (4.43) is
He = 0.5NW/(~ + 1) (4.56)
At Xi = Xj = 0.5, equations (4.38) and (4.41) give
2ln ~ = In n = W / kT = NW/ RT (4.57)
As a convenient test for equations (4.55) and (4.56), we take the experimental
value of G e and solve for ~ and then compute W from equation (4.57).
Next we substitute ~ and W into equation (4.56) to calculate He. The
reason for using this procedure is the simplicity of computation. The
calculated and the experimental values of He are compared in Table 4.4
for a number of randomly selected alloys.4,7 It is evident that the first
approximation is good when G e and He do not differ greatly, i.e., when
the entropy term in G e = He - TS e is small. There is a substantial degree
Table 4.4. Experimental and Calculated Values of He from First

Approximation to Binary Regular Solutions a
Alloy T(OK) Ge He(exptl) He(calc)
Ag-Au 800 -837 ± 150 -illi ± 50 -974

Bi+Pb 700 -300 ± 75 -265 ± 15 -326
Cd+Hg 600 -490±60 -645 ± 30 -559
Cd+Mg 923 -1l00±100 -1341 ± 100 -1306
Cd+Zn 800 478 ± 50 500 ± 20 378
Hg+ In 433 -495 ± 40 -540 ± 30 -580
In+ Pb 673 140 ± 50 230 ± 50 132
K+Na 298.15 160 ± 30 125 ± 50 131
Mg+Pb 833 -1630 ± 200 -2100 ± 200 -1980
Sb+Zn 823 -750± 100 -830± 200 -863
aOe and He are in calories for one gram atom (or mole) of solution at XI = x, = 0.5
(I cal = 4.184J).
Theories of Solutions lOS
of success in the results and in fact, for about half of the alloy systems for
which the data are reliable/ the calculated and the experimental values of
He agree fairly well.
Approximate Equations
Simple approximate equations can be derived in the narrow range of

-0.5 < W / kT < 0.5 by using {3 = (1 + 4axjxj )o.s = 1 + 2aX;Xj. Next, (3 is
substituted in 2/({3 + 1) = 1 - aX;Xj for use in equations (4.42) and (4.43);
then He is substituted in equation (4.45) to derive G e. The results corre-
sponding to equations (4.42), (4.43), (4.50), (4.51), and (4.52), in that
sequence, are as follows:
He = YW = NWxjxj [1 - XjXj ( W / kT)]
Ge = NWxjxJl - Xjxj (W/2kT)] (4.58)
(;~ = NWxJ[ 1 - XjXj (W / kT) + x~( W /2kT)]
The remaining partial molar properties can be derived from the molar
properties as shown in Chapter 1. It is important to remember that these
equations exclude the majority of alloys for which W/kT is larger than 0.5,
i.e., INWI > 0.5RT = 1000 cal/mole at 1000 K. However, it is likely that
for the alloys for which INW / RTI > 0.5, the criteria for regular behavior
cannot generally be met; hence, equations (4.58) may be useful for systems
in which the deviations from ideality are not great.
Short-Range Order
The relationship expressing Y* for the random distribution can be
obtained by setting the exchange energy W to zero in equation (4.38) to
obtain n = 1, leading to (3 = 1, from equation (4.41) and y* = Nx;xj from
equation (4.42). The ratio Y / y* is therefore
Y/ Y* = 2/({3 + 1)
106 Chapter 4
We define a short-range order parameter as by
y f3 - 1
a=l--=-- (4.59)
s y* f3+1
For a random solution f3 = 1 and as = 0, signifying that there is no order,

or the solution is random. Positive values of W yield n > 1 and also f3 > 1,
as > 0, indicating that the net number of atoms surrounded by the unlike
atoms is less than the random number, or there is a tendency for the like
atoms to cluster together. Negative values of W lead to the opposite
conclusion, i.e., when W < 0, then n < 1 and f3 < 1 and as < 0, or there is
a tendency for unlike atoms to surround each other. Very large negative
values of W lead to f3 approaching zero and as approaching -1. There are
other definitions of the short-range order which will not be discussed here.
Actually, Y 1 y* itself may be considered as an order parameter and reaction
(4.37) may be regarded as the order-disorder reaction. It will be seen later
that in ordered alloys, a different parameter, r, will be more convenient
than as.
Dilute Binary Regular Solutions
The equations for dilute binary regular solutions according to the first
approximation can be obtained by setting one of the mole fractions close
to zero. Since the equations derived in the previous sections are symmetric
with respect to composition, it is sufficient to consider the case for Xj -'» O.
The equations for He and H7 are zero as expected, and the equation for
H'} is
H'} = aHelaxj = NW; (4.60)
This result can be obtained by substituting Xj -'» 0 in equation (4.44) and

noting that f3 becomes 1 for Xj -'» O. Likewise, G e and 07 are zero, and O'}
can be obtained from equation (4.51); however, for Xj -'» 0 the resulting
equation contains In(OIO) and the indeterminate ratio can be obtained by
applying the L'Hopital theorem, i.e., by taking the derivative of the
numerator and then the denominator with respect to Xj and then setting
Xj -'» 0 in the resulting ratio. The derivative of the numerator, f3 + 2xj - 1,
is 2n, and that of the denominator, x(f3 + 1), is f3 + 1 = 2; therefore,
o'} = aGejaXj = NkTln n = NW; (4.61)

where n = e W/kT has been used to obtain the last equality. Since equations
(4.60) and (4.61) are equal, it is evident that
§e] = o., (4.62)
Consequently, in dilute solutions, the solute can have only one type of
configuration in which the j molecules are entirely surrounded by the solvent
molecules, and this arrangement corresponds to the random dispersion of
the j molecules in the solvent i molecules. The foregoing relationships can
also be derived by using equations (4.58).
Critique of Treatments Based on Permutation of Bonds
The first approximation that existed prior to 1970 was based on the
permutation of bonds as developed mainly by Bethe 5 and by Guggenheim.!
It was shown earlier in this chapter that such a permutation leads to
physically impossible spliced molecules. While the method has been the
subject of numerous papers, some based on abstruse mathematical argu-
ments, the most elegant and rigorous method is due to Guggenheim.!
The numbers of i-i, j-j, and i-j bonds used by Guggenheim are also
given by equation (4.18). The corresponding bonds for the random distribu-
tion involve y* of equation (4.16). The energy E is also given by equation
(4.30). However, Dc used in equations (4.29) and (4.32) are quite different,
since the bonds are used in permutation. Guggenheim assumed that an
observer is capable of distinguishing an i-j bond from a j-i bond and
therefore there are ZY/2 of i-j and ZY/2 of j-i bonds. There are altogether
ZN/2 bonds consisting of Z(N; - Y)/2, ZY/2, ZY/2, and Z(Nj - Y)/2
bonds of i-i, i-j, j-i, and j-j bonds, respectively. The permutation of these
bonds is given by
, (ZN/2)!
(4.63)
g = [Z(N; - Y)/2]!(ZYj2)!(ZY/2)![Z(Nj - Y)/2]!
This equation is inexact as admitted by Guggenheim because the sum of

g' for all the possible values of Y exceeds all the possible configurations
N!j(N;!Nj!) by a very large factor for Z greater than 2. To remove this
objection, Guggenheim introduces an arbitrary normalizing factor h =
h(N;, Nj) to obtain a corrected expression g = hg'. The assumption that h
is a function of N; and Nj and independent of Y is incorrect as will be
108 Chapter 4
seen later. The sum of g for all values of Y must be equal to all the possible
configurations, i.e.,
L hg' = N!/(Nj!~!) (4.64)

y
The sum in equation (4.64) is then replaced by its maximum term g(max) =
hg'(max). To accomplish this, hg' is differentiated with respect to Y and
the result is set to zero to solve for the corresponding value of Y. This
solution gives Y = Y* = Nxjxj, and when Y* is substituted in g(max) =
N!I (Nj! ~!) = hg', an equation is obtained for h. The substitution of h
into g = hg' gives
N! [Z(Nj - Y*)/2]!(ZY*/2)!(ZY*/2)![Z(~ - Y*)/2]!

g = Nj!~! X [Z(Nj - Y)/2]!(ZYI2)!(ZYI2)![Z(~ - Y)/2]!
(4.65)
It is very important to observe that equation (4.65) differs from equation

(4.63) by the factor h, which is assumed to be independent of Y. The
partition function Q is then expressed by
Q = L ge- E / kT (4.66)
y
where E is given by equation (4.30). We wish to replace g by g' to show

that the same final equations of Guggenheim can be obtained without h
and to prove that the use of the correction factor h is an exercise in futility.
Therefore, we write
Q' = L g'e- E / kT (4.67)

y
The sum in equation (4.67) can be replaced by its maximum term to obtain
the equation for Y. For this purpose, aIn Q'I aY is set to zero in the same
way that equation (4.34) was set to zero. The solution of the resulting
equation for a In Q'I aY = 0 yields
y2/(Nj - Y)(~ - Y) = exp(-2 WI ZkT)
= exp[ -(ejj + eji - ejj - ejJI kT] (4.68)

It is now evident that whether we had used g' or g = hg', we would still
obtain the same preceding equation because Guggenheim assumed that h
was independent of Y. The left side of equation (4.68) is the equilibrium
constant for the following reaction among the bonds:
i-i + j-j = i· .. j +j . .. i (4.69)
In contrast, the right side of equation (4.68) may be regarded as the

pseudo-equilibrium constant for reaction (4.69) where the exponential terms
give the energies corresponding to the same bond reaction. Therefore,
equations (4.68) and (4.69) reemphasize the independence of bonds as
permutable entities despite the attempt to normalize g' by using h. Permuta-
tion of bonds as independent entities leads to enormously large values of
g'; hence, g' cannot represent the order of magnitude of the correct distribu-
tion, Dc, for the molecules in solution. In addition, h is so small that g = hg'
quickly becomes smaller with increasing values of Y. Therefore, g cannot
represent Dc as shown in Table 4.1, and by the curve for g/actual arrange-
ments in Fig. 4.8. The values of g decrease sharply when Y/ y* becomes
either larger or smaller than unity as shown in Fig. 4.8. This is also evident
in the last column of Table 4.1 at high values of Y / Y* but the permissible
values of Y / Y* are not small enough for such a small number of molecules
to show a similar trend at low values of Y / y*. The curves for g / actual
arrangements at various values of N are not shown in Fig. 4.8, but they
shift lower relative to D / actual arrangements with increasing N. It is evident
that g does not represent the actual number of arrangements with a realistic
degree of approximation and therefore it cannot be used in deriving mean-
ingful statistical thermodynamic equations.
The ratio g/ Dc for Ni = ~ = L, Y* = NJ2, Y = NJ4, and for any
value of Z is given by
log(g/ Dc) = -0.057(Z - 2)L + 0.316Lb + 0.5 log L7T (4.70)
where Stirling's approximation for factorials has been used. The value of
D rm is expressed by
log D rm = 0.602L - 0.5 log L7T (4.71)
Therefore, g and Dc differ by a factor of e L or by about the same factor as

D rm • On the other hand, D/ Dc is given by
10g(D/ DJ = 210g[ Y/(L - Y)] = 0.316L b (4.72)

110 Chapter 4
where the expression in the center is for any value of Y obtained by using
In( 1 + a) = a and the last equality is for Y = L/4. It is seen that D and
Dc differ by a much smaller factor than g and Dc. For Z = 2, equation
(4.70) becomes comparable to equation (4.72) as log L is much smaller than
Lb. This is evidently due to the fact that L and the number of permutable
bonds become identical for Z = 2. However, the permutation of bonds
again creates spliced molecules.
The thermodynamic equations resulting from equation (4.68) require
that n of equation (4.38) must now be written as
n = e2W/ZkT (4.73)
g
When the procedure followed by equations (4.39)-(4.48) is repeated by

using the preceding n g , the result for G e is obtained, i.e.,
Ge(bond permutation) = - 2 -
ZNkT{ Xi In [b+X.-X]
(P ~ 1)J +
-Xi Xj
[b+X-X]}
In -xj (P J+ 1) I
(4.74)
where P, used only in this equation, is defined by p2 = 1 - 4xi xj(1 - ng ).

The success of equation (4.74) in correlating G e and He is considerably
less than equation (4.50) based on permutation of molecules. It should be
reemphasized that Guggenheim's method is the most rigorous and the
simplest treatment based on the permutations of the bonds. Therefore, all
the remaining treatments based on permutation of bonds add nothing
significantly new to the resulting thermodynamic equations; hence, they are
subject to the same criticism that the bonds are not permutable entities.
We next proceed to comment on a treatment called the "surrounded
atom" or the "central atom" model for binary alloys.8,9 The results can be
interpreted in terms of the pseudo-equilibrium constant as in equation (4.36)
and its reaction similar to (4.37). These relationships in terms of the notation
in this book are
y2 = e- W / ZkT .
(Ni - Y)( ~ - Y) ,
1 1 1 1
- (i-i) + - (j-j) = - (i . .. j) + - (j ... i) (4.76)
2 2 2 2
where the coefficient 1/2 in this reaction comes from the fact that W / Z
contains 1/2 as shown in equation (4.75). The difference between equation
(4.75) and Guggenheim's equation (4.68) is that the latter contains 2 W / Z =
Theories of Solutions III
(eij+ eji - eii - ejj ) in its exponent, or a quantity twice as much as in equation
(4.75). The mathematical procedure and some of the postulates in the
surrounded atom model are interesting; however, the resulting thermody-
namic relations, starting with equation (4.75), are subject to the same
criticism as equation (4.74) based on the permutation of bonds.
Ternary Regular Solutions

The zeroth approximation to multi component solutions was given by
equation (4.7) in which !1H = He = G e, and by equation (4.11). The first
approximation requires the additional pairwise terms. \0 For example, a
ternary system consisting of i-j-k components requires labeling {3 as {3ij'
{3ik, and {3jk, and then writing equation (4.50) for the pairwise terms as
follows:
(4.77)
where the last two terms are for the j-k system and identical in form with
the preceding terms. The equations for {3ij require writing Wij • .• for pairwise
interactions of net numbers of atoms. This relatively complicated equation
can be simplified by using equation (4.58), i.e.,
Liquid ternary alloys of some elements such as Mn, Fe, Co, and Ni are
expected to obey qualitatively this equation.
Regular Associated Solutions

The phase diagrams in which a compound semiconductor is a solid
phase, generally have a congruently melting solid, e.g., CuTe, SnTe, GaAs.
Such 1 : 1 or other intermetallic and metal-metalloid compounds in solid
state do not dissolve the component elements in significant concentrations.
112 Chapter 4
The liquidus for these binary systems at the melting point in the vicinity of
the compound has a temperature peak, the viscosity of liquid is a maximum,
and the conductivity is a minimum. Further, the enthalpy and entropy of
mixing show sharp minima in the vicinity of the compound. These properties
indicate the existence of a compound species such as AB in the liquid phase
of the binary system A-B; hence, the system may be considered as consisting
of three species, A, B, and AB in the liquid phase. Other compounds, ABno
may also exist, but the treatment presented here for the equiatomic com-
pound AB would require only minor modifications to extend it to other
intermetallic compounds, AB n • The existence of compound species along
with A and B forms the basis of associated solutions.
The solid phase, AB(s), the simplest compound to be considered here,
is in equilibrium with the liquid phase at the liquidus composition in which
the activities of AO) and BO) are denoted by a l and a2, respectively. The
accompanying equilibrium among A(I), B(I), and AB(s) is
AB(s) = A(I) + B(I);
where the activity of solid AB(s), denoted by a~, will be taken as unity,
because AB(s) is assumed to be a pure compound. Liquid ABO) is formed
by A and B in the liquid phase at the equiatomic composition, XI = X2 = 0.5,
according to
A(l) + B(l) = AB(I); ~GI = -RT In a3 (4.80)

a l (0.5) . a2(0.5)
We emphasize that aJ. a2, and a 3 without superscripts always refer to the
liquid phase. If the compound formation is also very strong in the liquid
phase so that the activity of AB(I) can be taken as unity only in equation
(4.80) at equiatomic composition, then the sum of the two preceding
reactions and their ~Go is
AB(s) = ABO); ~GO = RT I a l(0.5) . a2.(0.5) (4.81)

m n [al . a2][at liquidus]
The resulting ~G~ refers to the melting of 1 mole of AB(s) into ABO);
therefore,
0.5~G~ = ~H~ - T~S~ (4.82)

where tlH':n = Tm tlS':n refers to 1 g-atom of (A + B)or 0.5 mole of AB in

accord with the convention in thermochemistry of aHoy phases, and Tm is
the melting point. If it is assumed that the activity coefficients in the liquid
phase obey the regular zeroth behavior, i.e., In "1 = ax~ and In == axi, "2
then with a l = XI"I and a2 = X2"2, the logarithmic terms in equation (4.81)
become
0.5a + In 0.25 - ax~ - In XI - axi - In X2
= -In 4XIX2 - a(xi + x~ - 0.5); (Xi at liquidus) (4.83)
The elimination of XI by using 1 - X2 in the last term gives -2a(x2 - 0.5)2,

and substitution of this result and equation (4.82) into equation (4.81) yields
0.5RT In 4XIX2 - tlH':n + TtlS':n

a = (X2 - 0.5)
2 (4.84)
A single point on the liquidus yields one value of XIX2, from which a can
be calculated if tlH':n and tlS':n are known from measurements, e.g., from
calorimetry. After the evaluation of a, the activity coefficients can be
calculated from In "1 = aX2 and In "2
= axi. Calculations show that a varies
slowly with temperature along the liquidus of a number of binary systems.
This equation, first derived by Wagner, II was later used by Vieland,12
Thurmond,13 and other investigators. A modified treatment has also been
presented by Jordan.14
The regular associated solution model does not comply with the require-
ments for regularity because the compound AB is so much larger in size
than A and B, and the distinction between the A-B bond within the
compound AB and A-B bond in AB-A between the compound and A are
difficult to reconcile with the equality of bond energies for the same type
of bonds. The term regular in this case originates from Oe expressed as a
function of composition corresponding to the zeroth approximation to the
regular solutions, often modified by various investigators to enhance the
success of representation of experimental data.
A more accurate and elaborate treatment is presented by Hsieh et al. 15
correlating thermodynamic data l5 - 19 with the phase diagram. The results
for tlO and tlH of aHoy formation per gram atom of Sn-Te aHoy are shown
in Fig. 4.9, and the phase diagram, in Fig. 4.10.
The regular associated solution models can be simplified by dividing
the system into two pseudobinary systems because the solid phase is assumed
to be nearly stoichiometric with or without the assumption that the com-
pound AB exists in the liquid phase. If the solid phase is indeed
stoichiometric, then it can be shown that the discontinuity in the liquidus
114 Chapter 4
o :\. . . .(
~
Gibbs energy of mixing y
<:> 0 Nakamura et 01.

/,I~
~
t; Rakotamava et 01.
-5
\ 'il
Enthalpy of mixing
<:> Nakamura et 01.
'il Rakotamavo et 01.
/I ~
{iij
I
O\'i7v;z 'fJ.~
\:\ I
-10 0. ~
-15
c\t; ~ 'I
\\ f
-20 ~~ ,I,
\~~
'il~
-25
\y Figure 4.9. Gibbs energy and enthalpy
offormation ofSn-Te alloys at 1100 K.
- 27. 5 '--..L.-....I--.l---I.---JI...-.L.--"-.-L...--J..-.J (Courtesy of Y. A. Chang. 15 Data of
0.0 0.2 0.4 0.6 0.8 1.0 Nakamura et al. 16 and Rakotomavo et
Sn X Te - Te al. 17 )
1000- -
L
700 - 0 Biltz and

Mecklenburg
t; Rakatamava et 01. 674.0 0.85
D Le Bouteiller Figure 4.10.Sn-Te phase diagram
et 01.
according to Hsieh et al. 15 (Courtesy
600L-L-L-1~1L-~1~~1~~ of Y. A. Chang. 15 Data of Biltz and
00 0.2 0.4 0.6 0.8 1.0 Mecklenburg,18 Rakotomavo et al.,17
Sn X Te - Te and Le Bouteiller et al. 19 )
at AB is required both by the Gibbs-Konovalow theorem, and by Raoult's

law because the depression of freezing temperature has to be proportional
to the mole fraction of AB. Actually, even the solid compound AB deviates
slightly from stoichiometry and the liquidus shows smoother changes in the
slope with increasing deviations from the stoichiometry of AB(s).
Theories of interstitial solutions will be presented in Chapter 6 where
modified forms of the regular solution equations and the Wagner equations 20
will be discussed in detail.
References
I. E. A. Guggenheim, Mixtures, Oxford University Press, London (1952).

2. C. Wagner, Thermodynamics of Alloys, translated by S. Mellgren and J. H. Westbrook,
Addison-Wesley, Reading, Massachusetts (1952).
3. J. H. Hildebrand and R. L. Scott, The Solubility of Nonelectrolytes, Third Edition, Reinhold,
New York (1950).
4. N. A. Gokcen and E. T. Chang, 1. Chem. Phys. 55, 2279 (1971).
5. H. A. Bethe, Proc. R. Soc. London Ser. A ISO, 552 (1935).
6. N. A. Gokcen and E. T. Chang, A New Method for Enumerating Molecular Configurations
in Propellant Mixtures, Aerospace Report No. TR-OI72 (2210-\0)-1, The Aerospace Corp.,
EI Segundo, California (1971).
8. J.-c. Mathieu, F. Durand, and E. Bonnier, J. Chim. Phys.62, 1289, 1297 (1965); B. Brion,
J.-c. Mathieu, P. Hicter, and P. Desre, J. Chem. Phys.66, 1238, 1745 (1970).
9. C. H. P. Lupis and J. F. Elliott, Acta Metall. 15,265 (1967).
10. N. A. Gokcen, Scr. Metall. 16, 723 (1982).
II. C. Wagner, Acta Metall. 6, 309 (1958).
12. L. J. Vieland, Acta Metall. 11, 137 (1963).
13. C. D. Thurmond, J. Phys. Chem. Solids 26, 785 (1965).
14. A. S. Jordan, Metall. Trans. A/ME 1,239 (1970).
IS. K.-C. Hsieh, M. S. Wei, and Y. A. Chang, Z. Metallkd. 74, 330 (1983).
16. Y. Nakamura, S. Himuro, and M. Shimoji, Ber. Bunsenges. Phys. Chem. 84, 240 (1980).
17. J. Rakotomavo, M.-C. Baron, and C. Petot, Metall. Trans. A/ME 128,461 (1981).
18. W. Biltz and W. Mecklenburg, Z. Anorg. Chem. 64, 226 (1909).
19. M. Le Bouteiller, A. M. Martre, R. Farhi,and C. Petot, Metall. Trans. A/ME 88,339 (1977).
20. C. Wagner, Acta Metall. 21, 1297 (1973).
5
Long-Range Order
Chapter 2 dealt with the phase equilibria in alloys involving only the
first-order phase transitions. The Gibbs energy of a given multicomponent
system is a function of its variables of state P, T, n h n 2 , ••• , nc, i.e.,
121 = 6(P, T, n" n2, ... , nC>. This function is continuous for the first-order
phase transitions but its derivative, with respect to one of its variables,
becomes discontinuous upon a first-order transition. The variable of the
greatest importance is the temperature; therefore, we limit our discussion
to the derivatives of 121 with respect to T. A first-order transition is accom-
panied with a discontinuity in the first derivatives of 121; thus,
a(?i a(G/T)
-=-s·
aT '
or
a(1/ T)
= H (5.1)
would show a discontinuity in the entropy or enthalpy when these properties

are measured from a reference temperature such as T = 0, or often more
conveniently, T = 298.15 K. Condensation and freezing of pure components
provide some of the most elementary examples of first-order transitions. At
a second-order transition, the second derivatives of 121 exhibit a discon-
tinuity; i.e.,
or
a2 ( (?i/ T) = C (5.2)
a(1/ T)aT P
In summary, S, or H, is discontinuous for the first-order phase transitions

and Cp is discontinuous for the second-order phase transitions.
The order-disorder phenomena are second-order phase transitions in
which discontinuities are observed in Cp of metals and alloys. In pure
117
118 Chapter 5
elements, such as iron, order-disorder in the vicinity of 1043 K produces

magnetic order-disorder, but in alloys such as Cu-Zn at 742 K, the arrange-
ment in the crystal lattice causes order-disorder.
Theoretical treatments of order-disorder are basically applicable to
any type of second-order transition. However, we limit our presentation to
the order-disorder in binary alloys, with the tacit assumption that the
treatment of multi component systems do not present unusual difficulties.
Numerous statistical thermodynamic attempts have been made l - 1o to
formulate the behavior of long-range order in alloys after the significant
initial work of Gorsky,1 and Bragg and Williams 2 (GBW). The quasi-
chemical method, largely due to Bethe3 and to Guggenheim4 and based on
the permutation of bonds, is often erroneously claimed to be an improvement
over the GBW method. The permutation of bonds in the quasi-chemical
method is made to become the permutation of atoms for zero exchange
energy, or when the temperature is sufficiently high, through a dubious
normalization process. It has been shown conclusively by Gokcen and
Chang l l that (1) the normalization process of the quasi-chemical method
yields the same equations for the excess Gibbs energy of solution as without
normalization, and (2) the actual enumerations of configurations 11 prove
that the permutable entities are the atoms, not the bonds, as discussed in
the preceding chapter.
Ordering and Clustering
The number of AB bonds in a substitutional binary solution is more

than the random number (ZN /2)(2xAXB) when the exchange energy W =
(Z /2)(2eAB - eAA - eBB) is negative, and less when W is positive. Very large
negative values of W / kT may cause formation of one or more intermetallic
compounds in the same system. For positive values of W / kT, clustering,
or the association of like atoms, occurs, and in fact, when W / kT is
sufficiently large, the clustering causes separation of a phase into two
phases. On the other hand, if W / kT is negative to some optimal degree
that cannot always be quantified, A atoms may form nearly entirely A-B
bonds with virtually no A-A bonds if XA ..; X B• This is possible when certain
sites in the crystal lattice are occupied by A atoms and others by B atoms.
In such substitutional solid solutions, we shall see that the A atoms may
be regarded as having formed their own lattice, interpenetrating the lattice
formed by the B atoms. This phenomenon is known as the long-range order.
Since W itself may not often be sufficiently negative, the long-range order
is expected to occur at sufficiently low temperatures; however, the kinetic
Long-Range Order 119
energy of atoms may not often be sufficient to move the atoms into their
ordered state below ambient temperatures. The clustering may occur in
liquid and solid alloys but the long-range order may occur only in solid
solutions; therefore, this chapter is largely concerned with solid solutions.
Order-Disorder in Binary Alloys
We consider order-disorder phenomena in binary substitutional alloys

wherein the lattice structure of solid solution is retained upon transition
from the ordered state to the disordered state. It will be assumed that minor
changes in crystal structure upon transformation are not of primary import-
ance. This type of transition is therefore a second-order transition, which
may be called the Curie-type transition, or the coherent transition, but we
retain the term order-disorder transition as the more descriptive tenD for
our purposes. The treatment presented in this chapter can, however, be
readily extended to noncoherent transitions. For a body-centered cubic
(bcc) structure, shown in Fig. 5.1(a), the random distribution of A and B
atoms for an equimolar or equiatomic solution exhibits no particular prefer-
ence in the locations of A and B atoms. Each lattice site in such a solid
solution may therefore be occupied by either A or B whether the solution
is ideal or nonideal. However, if all the A atoms occupy only the corners
of the cube as shown in Fig. 5.1 (b), and all the B atoms, the body-centered
sites, then we have a perfectly ordered structure as in beta brass, Cu-Zn.
We designate the ordered sites by a for A atoms, and by b for B atoms
when the ordering is perfect, but when the ordering is partial, the sites still
retain their identities, even if they are occupied by a number of wrong
atoms. The identities of sites disappear when nearly half of the atoms of
each component occupy their right places, and the remaining half, the
wrong places for an equimolecular solution. We note that 1/8 of each corner
atom belongs to each cube, and that the nearest neighbors lie along the
diagonals of the cube by 0.866 of the cube edge. All the A atoms may be
regarded as forming their cubic lattice by interpenetrating a similar cubic
lattice formed by the B atoms, and for this reason an ordered structure is
said to form a superlattice. Tetrahedral crystals of the type exemplified by
ZnS, may be projected on two dimensions as in Fig. 5.1(c) to illustrate this
type of order in which each atom has four unlike nearest neighbors. The
picture is somewhat different for a close-packed two-dimensional ordered
structure for ABrtype phase as shown in Fig. 5.1(d). In this case, each A
atom is surrounded by six B atoms, but each B atom contributes one-third
to each hexagon with A in the center; therefore, the composition is AB 2 •
120 Chapter 5
a b
:0
I I
I
If)
~------ ',-- ~------
/ /
/ /
/ /
d
c
: A or B Oloms;. : A atoms; 0 -B OIOms
Figure 5.1. (a) Disordered structure in body-centered cubic crystals; each site is occupied
randomly by either type of atom. (b) Ordered AB type of alloy; corner atoms occupied by A
atoms form cubes interpenetrating cubes formed by B atoms occupying body-centered posi-
tions. (c) ZnS-type tetrahedral ordered structure projected on two-dimensional coordinates;
A = Zn, B = S. (d) Two-dimensional AB 2-type hexagonal close-packed structure. Each A atom
is entirely surrounded by B atoms, and every third atom in any direction is an A atom.
A face-centered cubic structure of AB3 type, such as AuCu3, has Au at the

cube corners and Cu at the face centers; corner atoms contribute one atom
and face-center atoms three atoms per cube.
The existence of order can be determined semiquantitatively, or in
favorable cases quantitatively, by X-ray diffraction. New diffraction lines
are observed when ordering initiates in a random alloy_ The intensities of
the new lines are than a measure of the degree of ordering. Electron and
neutron diffraction methods, and some of the physical properties also
determine the extent of order with various degrees of accuracy. Selected
reviews, discussions, and summaries on various aspects of ordering are to
be found in Refs. 5-10.
The atomic ratio of A to B may not always be a ratio of whole numbers,

and the numbers of a- and b-Iattice sites mayor may not be determined
by the crystal structure. As the composition of a perfectly ordered phase is
changed, the degree of order also changes because the excess number of
atoms of one of the components must be accommodated in the wrong sites.
Atomic vibrations increase with increasing temperature to such an extent
that, finally at a critical temperature, the degree of order disappears and
the solution becomes disordered in the sense that no measurable degree of
long-range order can be observed. We shall be concerned mainly with the
statistical thermodynamic treatment of long-range order, leaving physics
and physical metallurgy of this interesting type of phenomena to the
appropriate monographs and treatises.
Long-Range Order Parameter
Statistical thermodynamic treatment of ordering in alloys requires a

convenient definition of the long-range order parameter. There are several
definitions l - 6 but the most frequently used parameter, r, is defined by
A atoms on a-sites = N~ = 0.5 x all A atoms x (1 + r) (5.3)
where it is assumed that the number of A atoms is equal to or smaller than

the number of a-sites. For equal numbers of a- and b-sites, each designated
by L, and for equal numbers of A and B atoms, or XA = X B, this equation
is written as
N~ = x A (1 + r)L = 0.5(1 + r)L (5.4)
Thus, if all A atoms are distributed randomly, r is zero and the probability
of finding A atoms on a-sites is equal to its mole fraction. The parameter
r is unity for a perfectly ordered alloy and all the A atoms are on the
a-sites, i.e., N~ = N A = L. The recognition of two or more different types
of sites is essential in formulating the properties of ordered solutions. The
sets of a- and b-sites are sometimes called the a- and b-sublattice respectively.
Each site has Z neighbors, Z being the coordination number. The more
complicated cases of unequal numbers of A and B atoms and a- and b-sites
will be discussed later.
The parameter r for the preceding simple case with equation (5.4) and
XA = 0.5 leads to
A atoms on a-sites = N~ = 0.5(1 + r)L

122 Chapter 5
A atoms on b-sites = N~ = 0.5(1 - r)L

(5.5)
B atoms on b-sites = Nt = 0.5(1 + r)L
B atoms on a-sites = N~ = 0.5(1 - r)L
The second of these equations is obtained by subtracting the first equation

from the total number of A atoms, NA = L, because the remaining A atoms
must be accommodated on the b-sites. The third equation is the same as
the first, and the fourth equation is the same as the second as can be shown
by similar arguments. The ordering parameter r varies between zero and
unity, i.e.,
(5.6)
When r is zero, the solution is called "long-range disordered" or briefly

"disordered" but not necessarily random or ideal. The distinction between
"disordered" and "random" is therefore important in this chapter.
Gorsky, and Bragg and Williams (GBW) Approximation
The simplest treatment of order-disorder in alloys is the zeroth approxi-

mation due to GBW. 1- 5 This model is similar to the zeroth approximation
to the regular solutions. For simplicity, we first consider the body-centered
cubic (bee) lattice in our discussions with NA = NB = L as in the Cu-Zn
system. The nearest neighbors to an atom on the body-centered site are the
corner sites, and since Z = 8, we take into account the eight bonds emanating
from an a-site to the neighboring b-sites. The fraction of L sites occupied
by A atoms in N"A/L and this is also equal to the probability of finding an
A atom on the a-sites. The probability of finding an A atom on the b-sites
is N~/ L so that the probability of finding an AA pair from equation (5.5)
is
.
Probabtlty of AA = N"AN~ =
LL ( 1 - r 2)/ 4 (5.7)
Likewise,
Probability of BB = (1 - r2 )/4 (5.8)

The probability of AB pairs is the sum of two probabilities, i.e., the probabil-
ity of A on a-sites times Bon b-sites, i.e., 0.25(1 + r)2, plus the probability
of A on b-sites times B on a-sites, i.e., 0.25(1 - r)2; hence,
Probabilty of AB = (1 + r)2/4 + (1- r)2/4 = (1 + r2)/2 (5.9)
Multiplication of each probability term by the total number of bonds, ZL,

gives the total number of each type of bond, and multiplication of each set
of bonds by its bond energy eij gives the total energy of the alloy as in the
zeroth approximation to the regular solutions, i.e.,
E "'" H(r) "'" (ZL/4)(1 - r2)eAA + (ZL/4)(1 - r2)e BB + (ZL/2)(1 + r 2)eAB

(5.10)
For the disordered solution, r is zero and equation (5.10) becomes
H(r = 0) = (ZL/4)(e AA + eBB + 2eAB) (5.11)
For convenience, we rewrite the exchange energy of equation (5.10) as

follows:
(5.12)
Equations (5.10)-(5.12) give
H(r) - H(r = 0) = Lr2W/2 (5.13)
It is assumed that the atoms in each group as given in each one of

equations (5.5) can be rearranged randomly and this is the reason that the
model is called the zeroth approximation. The resulting distribution for all
the atoms on the a-sites is
L!
D=--------- (5.14)
a [(1 + r)L/2]![(1 - r)L/2]!
The distribution Db for the atoms on the b-sites is identical, i.e., Da = Db;
hence, the overall distribution Dr is
D =DD = [
L! ]2 (5.15)
r a b [(1 + r)L/2]![(1 - r)L/2]!
124 Chapter S
The entropy of solution is S = k In Dr; substitution of this entropy and H

of equation (5.10) in G = H - TS gives
G = H -Thin Dr (5.16)
The term In Dr can be obtained by using the Stirling approximation,
S(r) = 2k In Dr = Lk[21n 2 - (1 + r) In(1 + r) - (1- r) In(1- r)] (5.17)
where S( r) is the entropy of the solution. For r = 0, S( r = 0) = 2Lk In 2 as

expected since the distribution of atoms is random; therefore,
S(r) - S(r = 0) = -Lk[(I + r) + (I - r) In(I - r)] (5.18)
Equation (5.16) now becomes
ZL Lr 2 W
G(r) = 4 (eAA + eBB + 2eAB) + -2-
- LkT[21n 2 - (I + r) In(I + r) - (I - r) In(I - r)]
(5.19)
Likewise, from equations (5.13) and (5.18), or directly from equation (5.19),
we obtain
Lr 2 W
G(r) - G(r = 0) = -2- + LkT[(I + r) In(I + r) + (I - r) InO - r)]
(5.20)
The equilibrium state of the alloy corresponds to the value of r which makes
aG/ar zero:
aG
-ar = LrW + LkT[ln(I + r). - In(I - r)] =0 (5.21)
At the critical temperature, Te , the second derivative of G( r) with respect

to r is zero as r approaches zero, i.e., the order disappears; thus,
(iG(r)
ar2
.
--=LW+LkT - - + - - =0'
1+ r 1 - r '
(1 1) for r = 0 and T = Te (5.22)
Substitution of r ~ 0 yields
(W<O) (5.23)
The critical temperature must be positive; therefore, W must be negative,

i.e., the energy of 2AB bonds must be lower than the sum of the energies
of AA and BB bonds. The derivation of equation (5.23) from (5.22) is the
rigorous method of determination for Te , because, for a second-order
transition, the second derivative of 0 with respect to a variable of state
must be zero at the critical point.
Substitution of W = -2kTe, obtained from equation (5.23), in equation
(5.21) gives
1+
- r
-= exp (-rw)
- - = exp (2rTe)
-- (5.24)
I- r kT T
A relationship equivalent to equation (5.24) was first derived by Gorsky.i

This equation can be rearranged to obtain
-rw)
exp ( - - -1
r= kT == tanh(-rw) == tanh (rTe) (5.25)
-rw) 2kT T
exp (kT +1
where the identity sign defines tanh. The numerical computation of Tel T
from this equation is very simple. For this purpose, it is necessary to select
an arbitrary value of a = rTel T and read the value of tanh a from an
appropriate calculator or obtain it from the ratio immediately after the first
equal sign in equation (5.25). Then, tanh a is the value of r, and this value
substituted in a = rTel T yields the value of T from the experimentally
known value of the critical temperature. The reader can verify that for the
equiatomic Cu-Zn alloy having Te = 742 K, with a = 0.7, tanh 0.7 =
0.6044 = r, and then T = rTel a = 640.66 K.
Equation (5.25) has a solution that is r = 0 for any value of Tel T or
WI k. When T > Te , r = 0 is the only solution because above the critical
temperature, disorder prevails. At T < Te , there is another solution in the
range of 0 < r < 1, and this root corresponds to the minimum in O(r) as
can be shown by using equation (5.22) with (5.23) as follows:
a2 0(r)
- - = -2LkTe + LkT - - +--
ar2
(1
1+ r 1 - r
1) (5.26)
126 Chapter 5
1.0 ----
--Gokcen
'" "
----Zeroth '
.8
,,
approximation ",
.6
,
\
\
"- \
\
\
.4 \
\,
.2 Figure 5.2. Variation of long-range order param-
eter with temperature. Tc is 742 K. Curve for
quasi-chemical approximation is not shown
0 0.2 0.4 0.6 0.8 1.0 because it follows closely the zeroth approxima-
T/Tc tion curve. (From Gokcen. 12 )
The substitution of Tc = 742, T d 640.66, and r = 0.6044 yields 535Lk,

which is a positive quantity, and this value of r corresponds to a minimum
in G(r) - G(O). At 640.66 K, however, r = 0 corresponds to a maximum
in G(r) - G(O) because equation (5.26) becomes negative in value. The
equilibrium values of r and the corresponding temperature are plotted in
Fig. 5.2, as the reduced temperature T / Tc versus r.
Above the critical temperature, the alloy is disordered, and therefore
the equations derived for the zeroth approximation to the regular solution
become applicable if the GBW approximation is assumed to apply for the
ordered solution. The available data indicate that this approximation does
not predict the properties of ordered solutions satisfactorily,6.9 and for
ordered and disordered solutions of the same components, different values
of W have to be used to express G(r) - G(O) and the Gibbs energy of
formation of the disordered alloy, 11G. Nevertheless, because of its sim-
plicity, the GBW approximation has been used in a number of recent
publications.
Heat Capacity
The heat capacity C (= Cp ) can be obtained by differentiation of
equation (5.13) with respect to temperature:
aH(r) _ aH(r = 0) = C(r) _ C(r = 0) = LrW~ (5.27)

aT aT aT
where L W = -2LkTc = - RTc from equation (5.23). The expression for
ar / aT = dr I dT can be obtained by taking the logarithm of equation (5.24)
and then differentiating it; the result, substituted in equation (5.27), is
Rr2 T~( 1 - r2)
I1C = C(r) - C(r = 0) = 2 ( 2) (5.28)
T - TcT l-r
4 - ' - Experimental
---Gakcen
~
- - - - - Zeroth approximation
-.J
···········Quasi-chemical
~3 approximation
<l
2
\ '<:::::
04 1.1
T fTc
Figure 5.3. Variation of heat capacity with temperature for equiatomic Cu-Zn alloy. Tc is
742 K, and measured heat capacities of pure components were subtracted from that of alloy
to obtain experimental curve. t.C is the same as the excess heat capacity C e as shown on page
139. (From Gokcen. 12 )
where CCr = 0) is the heat capacity of disordered alloy. The heat capacity
C(r = 0) ofa random disordered solution is the same as the heat capacities
of its component elements. * The values of rand T are related by equation
(5.25) and shown in Fig. 5.2; therefore, for each value of r, T must be
computed and then the result must be substituted in equation (5.28) to
obtain de. The result for the previous example, for which T = 640.66 K,
Te=742K, or T/Te = 0.8634, and r=0.6044, is simply dC/R = 1.174.
Above the critical temperature, T> Te , r is zero and dC rapidly approaches
zero because of r2 in the numerator of equation (5.28). The values of dC
calculated from equation (5.28) are plotted in Fig. 5.3. The results will be
discussed in conjunction with those obtained by the refined methods to be
considered later in this chapter.
More General Cases
We have thus far considered the equal numbers of A and B atoms

distributed on equal numbers of a- and b-sites. The unequal numbers of A
and B atoms on equal numbers of a- and b-sites will be considered first
and then the unequal numbers of a- and b-sites with NA = a-sites = La, and
*t.H in equation (4.5) for a random disordered solution is independent of temperature; hence,
at.H/aT = C(r = 0) - C (component elements) = O.
128 Chapter 5
NB = b-sites = Lb will be outlined later. The most general treatment is not

given, but it can be derived by the same procedure.
Let L be the number of a- or b-sites for the case of equal a- and b-sites,
and 2L, the total number of sites, which is also the total number of atoms.
One of the atomic fractions has to be equal to or less than o.s to generalize
the range of composition for this case, and we take XA ~ 0.5 for this purpose.
The total numbers of A and B atoms are 2LxA and 2Lx B, respectively. In
a perfectly ordered alloy, all A atoms are on the a-sites, but all the B atoms
cannot be on the b-sites, because NB = 2LxB is greater than L; therefore,
the excess B atoms must occupy the a-sites. The order parameter r is defined
so that the A atoms on a-sites are given by XA(1 + r)L. Consequently, when
r is zero, on the average, one half of A atoms (or xAL) will be on the a-sites
and the other half on the b-sites, and when r is unity, all the A atoms are
on the a-sites. The A atoms on the b-sites are 2LxA - XA(1 + r)L =
XA(1 - r)L. The B atoms on the a-sites occupy the remaining a-sites, i.e.,
L - x A(1 + r)L = (XB - xAr)L. The remammg B atoms, 2LxB-
(XB - xAr)L = (XB + xAr)L, must then be on the b-sites. In summary,
A atoms on a-sites = N~ = x A (1 + r)L
A atoms on b-sites = N~ = XA(1 - r)L

(S.Sa)
B atoms on b-sites = N~ = (XB + xAr)L
B atoms on a-sites = N~ = (XB - xAr)L
Here the equation number (S.Sa) is used to indicate its correspondence with
equation (S.5).
The probability of A-A bonds is now N~N~/ L2 = xi(1 - r2), that of
B-B is (x~ - xir2), and that of A-B is x A(1 + r)(x B + xAr) plus XA(1- r) x
(XB - xAr) or a total of 2X AXB+ 2xir2. Multiplication of each of these
probabilities with the total number of all bonds, ZL, yields the number of
each type of bond, and multiplication with the corresponding bond energy
yields the total energy of each type of bond; the sum of all energies is then
E = H = LZ(xieAA + x~eBB + 2xAx BeAB) + 2Lxir2W

= H(r = 0) + 2LxAr 2W
The distribution function corresponding to equation (S.lS) is
(S.lSa)
The Gibbs energy difference G(r) - G(r = 0) corresponding to equation

(S.20) is
G(r) - G(r = 0) = 2Lx;"r 2 W + LkT[(xA + xAr) In(xA + xAr)

+ (XA - xAr) In(x A - xAr) + (xs + xAr) In(xs + xAr)
(S.20a)
Likewise, the relationships corresponding to equations (S.21) for any value

of r and the corresponding T, and (S.22) for r = 0 and T = T e , are as follows:
(S.21a)
for r = 0 (S.22a)
The last equation gives
(W< 0) (S.23a)
The next case to be considered is the unequal numbers of A and B

atoms, as well as a- and b-sites. However, we limit the treatment to N A =
a-sites = La, and Ns = b-sites = Lb, so that when the alloy is perfectly
ordered, all A atoms are on a-sites and all B on b-sites. The number of
a-sites is taken as La = xaN, and that of b-sites as Lb = xsN. Next, we take
N A < Ns or XA < Xs since one of the components has to be larger than the
other; this also signifies that La < Lb. The order parameter is defined so
that N~ = (XA + xsr)La, and when r = 1, or for perfect order, N~ = La, or
all A atoms are on the a-sites, and for r = 0, XALa atoms are on the a-sites.
The total number of A atoms, i.e., La, minus N~ gives A atoms on b-sites,
which are N~ = xs(1 - r)La, and for r = 1, there are no A atoms on b-sites.
The distributions of atoms on a- and b-sites are therefore
A atoms on a-sites = N~ = (XA + xsr)La

A atoms on b-sites = N~ = xs(1 - r)La
(S.Sb)
B atoms on b-sites = N~ = (xs + xAr)Lb
B atoms on a-sites = N~ = XA(1 - r)Lb

130 Chapter 5
The resulting equation for G(r) - G(r = 0) can be derived by using the
foregoing procedure; the result is
G(r) - G(r = 0) = NkTx'ir 2W + NkT[(x'i + xAxBr) In(x'i + xAxBr)
+ 2XAXB(1 - r) In(xAxB - xAxBr) + XB(XB + xAr)
(5.20b)
An ordered phase such as AuCu3 has XA = 0.25 and XB = 0.75 and the
equilibrium values of r are given by
aG = NkT{3In(1 + 3r)(3 + r) - 3ln [3(1 - r)2] + 2rW} = 0 (5.21b)

ar 16 kT
The critical temperature is related to W as required by
(PG = NkTc
ar2 8
(8 +~)
kTc
= O.
'
for r = 0 (5.22b)
from which Tc = - W j(8k). We give these relations for the sake of complete-
ness without making use of them.
First Approximation
The zeroth approximation to the long-range order is the GBW approxi-

mation which is analogous to the zeroth approximation to the regular
solutions. The first approximation to the long-range order attempts to correct
the random distribution assumption used in the GBW method, and to resolve
the discrepancy between the exchange en~rgy W for the ordered solutions
and that for the disordered solutions.
We consider a binary alloy of equiatomic composition for simplicity
in our formulation because the results can easily be extended to non-
equiatomic compositions. Again, N A = NB = L so that 2L is 1 g-atom of
an alloy, and in a perfectly ordered alloy, all NA are on the a-sublattice,
and all NB are on the b-sublattice. Let R be the actual number of A atoms
on a-sites and Q the remaining A atoms on b-sites, and let the number of
nearest neighbors to an atom be Z. The occupancy of B atoms is similar,
i.e., R is the number of B atoms on b-sites and Q is the remaining B atoms
on a-sites. The long-range order parameter r from equation (5.3) gives
a b _ L( )
NA=NB=R="21+r; a L1(
N bA =NB=Q="2 - )
r
(~~ R ~ L; 0 ~ Q ~~) (5.29)

If X is the net number of A atoms on a-sites whose neighbors are A atoms,

then R - X is the net number of A atoms surrounded by B atoms. In
summary, for all the atoms, we have
(i) A on a-sites with A neighbors =X

(ii) A on a-sites with B neighbors =R- X
(iii) A on b-sites with A neighbors =X
(iv) A on b-sites with B neighbors =Q- X
Total number of A atoms =R+Q=L
(5.30)*
(v) B on b-sites with B neighbors =X
(vi) B on b-sites with A neighbors =R- X
(vii) B on a-sites with B neighbors =X
(viii) B on a-sites with A neighbors =Q- X
Total number of B atoms =R +Q =L
The net numbers X, R - X, and Q - X represent the numbers of atoms

such that when each of these quantities is multiplied by Z/2, which is the
number of bonds per atom, we obtain the number of corresponding types
of bonds. This concept, based on previous publications,11 ensures that the
permutable entities are the net numbers of each type of atom, conceived
to have only one type of neighbor. The correction required to equate the
actual permutations to the permutations based on the net numbers of atoms
is small and will be discussed later. This concept, fully justified earlier, is
essential in pursuing the ensuing argument. The types of atoms shown in
equations (5.30) are illustrated in Fig. 5.4 for a unidimensioal crystal having
16 atoms and 16 bonds, with NA = NB = L = 8. The net number of atoms
in part (i) is illustrated in this figure. The A atoms on a-sites with A neighbors
generate the bonds numbered 1 through 4. The bonds belonging to all of
such atoms alone is half of four, because the other half belongs to the
neighboring atoms; therefore, X = 2. The remaining types of atoms in
equation (5.30) can be obtained by the same procedure.
The number of A atoms on a-sites is R and it consists of the sum of
(i) and (ii), the latter being X and R - X, respectively. Likewise, the number
of B atoms on the same a-sites is Q from (vii) and (viii), the latter being
*R is used here and in writing the equations for permutations (Da , Db, and D,); avoid
confusion with the gas constant; see also pages 143 and 145.
132 Chapter 5
0: A atoms l'= 0.25; R=5j Q=3

6: B atoms ( i) X=2
.: a -sites ( i i ) R-X=3
x: b-sites ( i i il X=2
NA=Ns=L=8 (iv) Q-X= I
Figure 5.4. Net numbers of different types of atoms on a- and b-sites for NA = NB = L = 8.
Four numbered bonds involve the A atoms on a-sites with A neighbors; half of these bonds
belong to A on a-sites; therefore. X = 2 atoms for part (i). Remaining parts in equation (5.30)
can be obtained similarly.
X and Q - X, respectively. The permutation of these atoms as in equations

(5.14) and (5.15) is
D = R! Q! (5.31)
a (R - X)!X! (Q - X)!X!
where Da is the distribution for the a-sites. The distribution, Db, for the
b-sites from [(iii), (iv)] and [(v), (vi)] is identical with Da; consequently,
the overall distribution Dr is
-DD - [
R!Q! ]2 (5.32)
Dr - a b- (R-X)!(X!)2(Q-X)!
The following Stirling approximations are useful to convert the second

order in equation (5.32) into the first order:
[(R - X)!]2 = (2R - 2X)!2 2X - 2R ; (X !)2 = (2X) !T2X

(5.33)
[(Q - X)!f = (2Q - 2X)!2 2X - 2Q
As a result, equation (5.32) becomes
D = _ _ _-'-(2_R....:.)-'!('-2~Q.;_)!_ __ (5.34)
r (2R - 2X)![(2X)!f(2Q - 2X)!
The elimination of Rand Q by using equation (5.29) in (5.34) gives
D = _ _---=[_L-'-(1_+---'r)-=..]-=..![L_(-'-:,-I_-_r--,-)=-]!_ __
(5.35)
r (L + Lr - 2X) ![(2X) !f(L - Lr - 2X)!
Comparison of Dr from this equation and Dr from equation (5.15) is now

appropriate. The use of relationships similar to those given by equation
(5.33), converts equation (5.15) into
D = _ _--.:.(2_L--'.)_!_ _ (5.15a)
r (L+Lr)!)(L-Lr)!
This equation states that R = O.5(L + Lr) and Q = O.5(L - Lr) can be
permutated for any value of R, Q, and r, irrespective of the value of X.
When r = I, then R = L; hence, both X and Q must be zero since there
are no A-A bonds, and Dr = I as expected from equation (5.35). Conversely,
if X = 0 or if there are no A-A bonds, R has to be equal to L by geometrical
requirements because when R is less than L, it can be easily shown by one-
and two-dimensional crystal constructions that X cannot remain zero if the
permutation given by equation (5.15a) were carried out for other values of
R than R = L. The GBW method assumes that Rand Q may have values
unrestricted by the values of X so that equation (5.15a) is valid for any
value of X, and this assumption is analogous to the random distribution
permitted in the zeroth approximation to the long-range disordered regular
solutions.
All A atoms have ZNA/2 bonds (or ZNA half-bonds) belonging to A
atoms. If for convenience we define Y such that ZY /2 are the A-B bonds
emanating from A atoms, then the remaining bonds are the A-A bonds;
hence, (A-A bonds) + ZY /2 = ZNA/2, but A-A bonds are equal to
(2X)Z/2 from (i) and (iii); consequently,
2X+ Y= NA = L (5.36)
For the disordered alloy, r is zero but the solution is not necessarily random,
and substitution of r = 0 in equation (5.35) gives
LlL!
D=------:: (5.37)
r [(L - Y)!Y!f
which is exactly equation (4.23) for equiatomic solutions for the first
approximation to the regular solutions.
134 Chapter 5
Equation (5.35) is not yet useful for our purposes when the solution
is highly ordered or Y is considerably high as was shown in equation (4.29).
The correction suggested there may be simplified as FL == [F(r)t in the
numerator of equation (5.35):
[L(t + r)]![L(l - r)]!F L
Dr = Dr(corrected) = (L + Lr - 2X)![(2X)!f(L - Lr - 2X)!
(5.38)
The requirement for F(r) is that F(r) = 1 for both r = 0 and r = 1, or
In F = 0 in both cases. The exponent Lover F = F(r) is for convenience
in deriving the succeeding equations from (5.38). The form of In F obeying
these requirements, and with one adjustable parameter,II.12 g, is
In F = g(l + r) In(l + r) + g(l - r) In(l - r) (5.39)
It will be seen later that without g, it is impossible to make the exchange
energy, W, the same in the ordered solution and the disordered solution of
the same alloy. Further, g is sufficiently small to make F a correction
equivalent to c in equation (4.29).
The energy E of the alloy is Ze hl /2 times each term in (i)-(viii) in
equation (5.30) with the subscripts hi representing AA and AB, or BB, and
ehh the corresponding bond energy; the result is
E = LZeAB - 2XW (5.40)

where W, the exchange energy, was defined by W =
(Z/2)(2eAB - eAA - eBB)' The partition function (P.F.) can be written by
using equations (5.38) and (5.40):
x
P.F. = L De- E / kT (5.41)
X~O
The maximum term in this equation is obtained by setting aIn(P.F.)/ aX to

zero; the result is
(L + Lr - 2X)(L - Lr - 2X) = 4X2 e- W / kT = 4X 2n; (n = e- W / kT )
(5.42)*
The value of X from this equation yields the maximum term in equation
(5.41), the term that replaces the summation in P.F. The solution for X
from this quadratic equation in X is
2X t ± J n + r2(l - n)
L 1- n
*Mathematical procedure requires that n here be 1/ n of Chapter 4; see page 101.

The minimum value of X is zero according to equations (5.30) when all

the A molecules are on the a-sites, i.e., r = 1; consequently, the positive
sign before the square root would yield an unacceptable nonzero value of
X. Therefore, we take the root with the minus sign preceding the square
root, and multiply the numerator and the denominator of the right side
with its complement, 1 + J n + r2(1 - n), and then simplify it to derive
2X
(5.43 )
L
where b is used for brevity to denote
(5.44)
We now proceed to derive the Gibbs energy G(r, T) at constant

composition and constant r by following the same procedure as that in
Chapter 4 (see also Refs. 4 and 11). We assume that the enthalpy H is nearly
identical with E of equation (5.40) for the alloys, and write
T_-' d[G(r,T)]=fT_-' Hd(r l ) (5.45)

f
T ,~o T T '~o
where the integration is carried out at constant composition and constant

r, and for simplicity T ~ 00 is written as T- 1 = O. The upper integration
limit for the left side is simply G(r, T)/ T. The lower integration limit for
the left side is
[ G(r, rl)]
T T-'~O
=-
T r'~o
(H)
( - I=0
-SrT
'
) (5.46)
Since H = E is finite according to equation (5.40), H / T is zero for T- 1 = 0,

and the entropy in this equation can be obtained from equation (5.38) by
writing S(r, T- 1 = 0) = k In Dr. The value of X for T- 1 = 0 can be obtained
from equation (5.42) by setting n = e- W/kT = 1 for T- 1 = O. The result is
(5.47)
Substitution of equation (5.47) in equation (5.38) and in S(r, T- 1 = 0) =

k In Dr. results in
S(r, r l = 0) = k In Dr
= Lk In F - Lk(1 + r) In(1 + r)
- Lk(1- r) In(1- r) + 2Lkln2 (5.48)
136 Chapter 5
For the right side of equation (5.45), the integration procedure is similar
to that used for the long-range disordered regular solutions with r = 0 as
in Chapter 4. The integration of the right side of equation (5.45) requires
expressing d(T- I ) by using b of equation (5.44) as follows:
d
W)
(- 2kT =
bdb
b 2 _ r2 (5.49)
Next, equations (5.40), (5.43), and (5.49) are substituted in equation (5.45)
to obtain
f T-'
T-'=O
H dT- I = f
T-'
T-'=O
[ZLe AB dr l +
(1
2Lk(1 - r2)bdb
+ b)(b + r)(b - r)
] (5.50)
The left side of equation (5.45) is equal to [G(r, T)/ T] + S(r, r l = 0) from
equation (5.46). The first term on the right side of equation (5.50) is ZLeAB/ T,
and the last term can be integrated by parts to derive
G(~ T) + S(r, rl = 0)
=
ZeAB b+r b- r +
Lk [ - - + (1 + r) In--+ (1- r) In-- - 2In--
1 bJ (5.51)
kT 1+ r 1- r 2
Substitution of equation (5.48) for the entropy on the left side and
rearrangement of the result gives
G(r, T) ZeAB
LkT = kT - In F + (1 + r) In( b + r)
+ (1 - r) In( b - r) - 2 In(1 + b) (5.52)
The equilibrium value of r is obtained by setting aG / ar = 0:
_l_aG=_alnF+ln(b+r) =0 (5.53)
LkT ar ar b- r
where all the remaining terms including those containing ab/ ar cancel out.
For r = 0, this equation requires that aF/ar be zero since F(r = 0) = 1.
Differentiation of equation (5.53) and thereafter substitution of r = 0 at the
critical temperature gives
For r = 0, equation (5.44) gives be = nO. s = e-W/2kTc, and the use of equation
(5.53) and the rearrangement of equation (5.54) yields
d 1n F
2
(W)
0.5 --;J;2 = exp 2kTe ; (5.55)
Equation (5.52) is also applicable to the regular disordered solutions

for which r = 0, because Dr of equation (5.38) becomes identical with
equation (4.29) for r = o. When r is set to zero, equation (5.52) assumes
the simple form given by
G(r=O, T) ZeAB W (1+

b)
---:'---~--"':'=-----2In (5.56)
LkT kT kT
For an ideal equiatomic solution the enthalpy H is equal to the enthalpy

of pure components, H = (eAA + eBB)ZL/2, and the configurational entropy
is the ideal entropy of the alloy so that S(ideal) = 2Lk In 2; hence, the ideal
Gibbs energy of solution is
G(ideal) Z
LkT = (eAA + eBB) 2kT -21n 2; (5.57)
Subtraction of this equation from equation (5.56) and substitution for W

from equation (5.12) yields the following excess Gibbs energy of solution,
G e for an equiatomic solution:
Ge ( 2e W/2kT )
-k-= 21n 2-2In(1 + b) =21n W/2kT (5.58)
L T I+e
This equation, where 2Lk is the gas constant, is identical 13 with equation
(4.55).
The correction term, In F, in equation (5.39) may now be substituted
in equation (5.55) to obtain the value of g for any system. We take, as an
example, the Cu-Zn system in the vicinity of equiatomic composition,
known as beta brass. The value of W / k is the only parameter permitted to
be determined from experimental data in any approximation to the regular
disordered solutions. The compilation of Hultgren et al. 14 extrapolated a
short distance to the equiatomic composition yields G e /2LkT = -1.202 at
773 K, which substituted in equation (5.58), gives W / k = -2678. The regu-
lar solution model assumes that W / k is independent of temperature. The
critical temperature Te is 742 K. The explicit functional form of F(r) is not
138 Chapter S
known and its determination would require extensive and laborious enumer-
ation of configurations for one- and two-dimensional crystals. I I Such a task
is expected to be very difficult and time-consuming. However, the function
F(r) suggested in equation (5.39) is a very small correction meeting the
conditions that it vanish for r = 1 and r = 0, and further, it is a much smaller
factor in Dr than any factorials that yield terms of the order of In M! =
M In M for any factorial term M!. We shall soon see that g is also a number
of considerably smaller than unity.
The successive derivatives of In Fare
aIn F 1+ r
--=gln--' (5.59)
ar 1 - r'
The last of these equations gives 2g at r = 0, and substitution in equation

(5.55) with WI k = -2678, and Tc = 742 K yields
g = exp (~)
2kTc
= exp(-1.8046) = 0.16454 (5.60)
The first derivative of In F, substituted in equation (5.53), yields the equili-

brium values of r at various temperatures:
l+r b+r
0.164541n-- = I n - - (5.61)
l-r b-r
A numerical example for this equation is useful. Let r = 0.80; the left side
of this equation is then 0.36153, so that
b + 0.8
1.4355 = - - (5.62)
b - 0.8
from which b = 4.4739, or b 2 = 20.0158 = 0.64 + 0.36n, with In n = 26781 T.

Equation (5.62) therefore yields n = 53.8217, and T = 671.91, or T I Tc =
0.9055. The points, such as this one, are plotted in Fig. 5.2. The values of
g play an important role in the result for WI k. For example, if we set
g = 0.1, instead of 0.1645, then equation (5.55) would yield Wlk = -3417
instead of -2678.
The curve for r versus T I Tc from the quasi-chemical treatment of
Guggenheim4 is close to that for the GBW method; therefore, it is not
shown in Fig. 5.2. The X-ray data of Chipman and Warren,15 not plotted
in Fig. 5.2, are fairly close to the curve obtained from equation (5.61).
However, a more stringent test for any theory is the success in representing
AC = C e versus T I Tc shown in Fig. 5.3 and discussed in the next section.
Enthalpy and Heat Capacity
The enthalpy of an ordered phase AB is given by equation (5.40), in

which equation (5.43) can be substituted to obtain
LW(1-r2)
E = H = LZeAB - ----'-----'- (5.63)
l+b
The total differential of H = H(r, T) is
dH = (aH) dT+ (aH) dr (5.64)

aT , ar T
which yields, after division by dT, the following useful equation:
dH = C = Ce = (aH) + (aH) dr (5.65)

dT aT , ar T dT
Since H(ideaI), as obtained by setting W = 0 in equation (5.63), is indepen-

dent of rand T, He = H - H(ideaI) can be substituted in equation (5.65)
to derive the same equation for C e, which is the excess heat capacity.
Substitution of equation (5.63) in (aH/aT), gives
( aH) = LW(1 - r2) (~) (5.66)

aT, (1 + b)2 aT,
Equation (5.44) can be rewritten as b 2 = r2 + (1 - r2)n, and at constant r,

with n = exp(-W/kT), its differential is 2bdb = (1- r 2)nWdT/kT2 ;
hence,
(5.67)
Substitution of this equation in equation (5.66) and division by 2Lk = R

gives
(5.68)
Similarly, the substitution of equation (5.63) in (aH/ar)T, and (abjar)y =

(r - rn)/ b, obtained by differentiation of b 2 , yields
~ (aH) = (W) [2rb + 2rb 2 + (r - r3 )(1 - n)] (5.69)

R ar T 2k b(1 + b)2
140 Chapter S
Equation (5.53), in which iJ In F / iJr is replaced by its equivalent in equation

(5.59), is
l+r)
-g In ( - - + In (b+r)
-- = 0 (5.70)
1-r b-r
The derivative dr / dT is obtained from this equation, after using b db =

r dr - rn dr - (1 - r2) n W dT / 2kT2 to eliminate db; the result is
(5.71)
Substitution of equations (5.68), (5.69), and (5.71) in (5.65) gives the final
equation for C e / R:
At 1000 K, r = 0, and C e / R = 0.295, which is the value for the regular

long-range disordered solution of Cu-Zn. For r = 0.404, b = 5.7406, n =
39.1873 from equation (5.61), and using - W/ k = 2678 in In n = 2678/ T,
we obtain T = 730.03 or T / Tc = 0.9839 with Tc = 742 K. Substitution of
these values with g = 0.16454 in equation (5.72) yields
(5.73)
This value, and others obtained similarly, are plotted in Fig. 5.3. It is evident
that the results are in fair agreement with the closely concordant and
independent experimental values of Moser,16 and Sykes and Wilkinson 17
plotted as C e / R = [C(alloy) - C(component elements)]/ R. The values
calculated from the quasi-chemical method are also plotted in Fig. 5.3,
indicating that these results are not significantly different from the zeroth
approximation of GBW.
We now justify the functional form of F(r) by the way of summary of
the foregoing procedure. Equation (5.39) for F = F(r) must satisfy (I) the
boundary conditions that In F be zero for r = 0 and r = 1, (II) equation
(5.53) for all the equilibrium values of r, and (III) equation (5.54) for r = 0
and T = Te. For example, In F = g(1- r) In(1 + r) would satisfy all the
requirements but not (II) for r = o. It may be stated that equation (5.39)
is not unique, but it is a useful and relatively small correction meeting the
foregoing requirements.
Further justification for equation (5.39) can be made by comparing Dr

calculated from equations (5.38) and (5.39) and from equation (4.29). The
latter can be written with its correction term c as follows:
D = _ _ _.....::['-L....:...(1_+----'r)-=.]-=.![_L-'-(1_-----=-=-r)-=-]!_ __ (5.74)
r (Y+Lr+c)![(L- Y-c)!f(Y-Lr+c)!
For r = 0, this equation reduces to equation (4.29) corresponding to

equiatomicsolutions. For example, when L = 500,000, r = 0.4, and b = 5.74,
then, from equation (5.43), 2X = 500,000 x 0.84/6.74 = 62,130, Y =
L - 2X = 437,690. In addition, Y* = 106 x 0.25 = 250,000, and Y/ Y* =
1.75 so that c = 1690. The substitution of these values in equation (5.74)
yields In Dr(5.74) = 357,200 whereas equations (5.38) and (5.39) yield
In Dr = 376,980, which is within 5% of the preceding value for In Dr(5.74).
These computations show that F in equation (5.38) and c in equation (5.74)
or (4.29) are of the correct order of magnitude, but the actual enumerations
are not yet available even for unidimensional crystals with various values
of the ordering parameter r. Therefore, equations (5.38) and (5.39) are to
be preferred for the supporting evidence that equations (5.53) and (5.54)
°
are satisfied, and that for r = and r = 1, equation (5.38) reduces to equation
(4.29) based on actual enumeration of configurations. It is also necessary
to emphasize that, in essence, F(r) is not only a correction factor for the
ordered solutions, but also for the existence of two sublattices for the com-
ponent atoms.
It is essential to reiterate here that unless equation (5.39) is used with
g adjusted to satisfy be at the critical point, as required by equation (5.55),
then it would be impossible to reconcile the value of the exchange energy
W from the regular disordered solutions, with W from equation (5.55) with
the correction required by equation (4.29) or (5.74) or any other similar
equation. Determination of c in equation (5.74) for values of Y/ Y* con-
siderably farther from unity, as in the preceding example, would require
enumeration of configurations for highly ordered systems. This is a very
difficult and time-consuming task, yet to be undertaken.
Comments on Previous Approximations

The quasi-chemical method is an early method that attempted to
improve on the GBW method. The main assumption of the quasi-chemical
method requires permutation of bonds that can be obtained from (i)-(viii)
of equation (5.30). This permutation requires that each factorial such as
X! be written as (ZX /2), a procedure that was criticized in Chapter 4. The
resulting Dr for the permutation of bonds is then normalized by an incorrect
142 Chapter 5
procedure as in the quasi-chemical method in long-range disordered regular

solutions in Chapter 4. In addition, it has been shown elsewhere conclus-
ively,llo12 and in Chapter 4, that the permutation of bonds is incorrect since
it creates impossible spliced molecules, and contradicts the numbers of
configurations obtained by actual enumerations. Further, it is a geometrical
fact that for Z = 2, or for unidimensional crystals, as beads on a necklace,
the long-range order exists; for example, when r = 1 and NA = N B, then
NA and NB occupy alternate succeeding sites as ABABA .... The quasi-
chemical method, however, predicts that there can be no order when Z = 2.
For these reasons, the quasi-chemical method, despite the claims by
numerous investigators, is not am improvement over the GBW method.
The value of WI k for an ordered solution is calculated from the critical
temperature Te in the GBW and quasi-chemical approximations. For the
same solution in the disordered state, WI k is determined from experimental
results, and we now proceed to show that each of these methods leads to
inconsistent results within itself. For Cu-Zn, the GBW method requires
2kTe = - W(GBW) from which W(GBW)I k = -1484, and the zeroth
approximation to the regular long-range disordered solutions, the analog
of the GBW approximation, yields W(zeroth)1 k = -3717 from G e /2LkT =
-1.202 with T = 773 and G e = 2LxAXB W = O.5L W The quasi-chemical
approximation for ordered Cu-Zn gives WI k = -1708 from WI k =
TeZ In[ (Z - 2)1 Z], and the same approximation for the disordered solution
gives Wlk = -3285 from G e /2LkT = -1.202 = (Z/2) In{2exp(WIZkT)1
[1 + exp( WI ZkT)]), with T = 773 and Z = 8. Similar calculations for other
binary systems show similar discrepancies. The results for r versus T I Te
and for C e /2Lk versus TITe, obtained by these methods, are presented in
Figs. 5.2 and 5.3 for comparison.
A significant improvement over the quasi-chemical method is the cluster
variation method developed and perfected by Kikuchi. 18 A simplified version
of the earlier development by Kikuchi has been presented by Guggenheim
and McGlashan. 19 The simplified version for the square as the cluster would
require writing quasi-chemical equilibria and their pseudoequilibrium con-
stants for pseudoreactions such as
A--A B--B A--B
1 1+ 1 1= 21 1
A--A B--B B--A
and accounting for all the possible lower hierarchy of configurations down
to single bonds. While with increasingly complex clusters the method would
minimize the error originating from the permutation of single bonds and
formation of unrealistic spliced atoms, clearly a great number of different

clusters must be used as shown in Ref. 12. Further, much larger and varied
clusters than those used by Kikuchi pose vastly increasing mathematical
complexities. 20
The mean field theory, as presented by Burley/i shows that the curve
for r versus T / Te is farther the left of the curves in Fig. 5.2, or somewhat
closer to the origins of coordinates. The calculations based on such curves
show decreasing degrees of success for representation of C e / R versus T / Te.
Unequal Numbers of Atoms and Equal Numbers of Sites
The first approximation can be extended to the case of unequal numbers

of A and B atoms with equal numbers of a- and b-sites. Again, we take one
of the numbers of atoms smaller than the other, i.e., NA < N B • The order
parameter r is defined as in equations (5.5a) and (5.29), and we repeat them
here for convenience:
(5.5c)
(5.5d)
Various types of atoms on a- and b-sites, corresponding to equations (5.30),

are
(i) A on a-sites with A neighbors = X

(ii) A on a-sites with B neighbors = R - X
(iii) A on b-sites with A neighbors = X
(iv) A on b-sites with B neighbors = Q - X
Total number of A atoms = R + Q = NA

(5.30a)
(v) B on b-sites with B neighbors = X'
(vi) Bon b-sites with A neighbors = R' - X'
(vii) B on a-sites with B neighbors = X'
(viii) B on a-sites with A neighbors = Q' - X'
Total number of B atoms = R' + Q' = NB

144 Chapter 5
These atoms are shown and summarized in Fig. S.S for NA = 8, NB = 12,
XI = 0.4, and L = 10, for sufficiently large numbers of atoms to avoid
objectionable fractional values for items other than r, X A , and XB in equations
(S.Sc), (S.Sd), and (S.30a). The determination of X is given in the figure
caption, and that of X' will now be presented as an additional example.
Atoms of B on b-sites have 8 bonds with B atoms, and the latter are on the
a-sites. Eight bonds are shared by B on b-sites, and B on a-sites; hence,
8/2 = 4 = X'. The remaining parts of equations (S.30a) are determined by
the same procedure, and the results are listed in Fig. S.S.
The a-sites contain R atoms of A, and Q' atoms of B, and the distribu-
tion of these atoms on the a-sites is similar to equation (S.31), i.e.,
R!Q'
D = ------"------ (S.31a)
a (R - X)!X!(Q' - X')!X'!
0: A atoms ( i) X=2
.6,.: 8 atoms ( ii) R-X=2
• a-sites (iii) X=2

x b -sites (iv) Q-X=2
NA = 8; Ne = 12 (v) X' = 4
L = 10·, r = 0.25 (vi) R' -X' = 3
R = 5; Q=3 (vii) X' = 4
R' = 7; Q' = 5 (viii) Q'-X = I
Figure 5.5. Net numbers of different types of atoms on a- and b-sites for NA = 8, NB = 12,
,L = 10, and r = 0.25. Four numbered bonds involve the A atoms on a-sites with A neighbors;
half of these bonds belong to A on a-sites; therefore, X = 2 atoms for part (i): the other four
half-bonds belong to A on b-sites with A as neighbors, and this constitutes part (iii). Remaining
parts in equation (5.30a) are summarized above and can be obtained similarly. All the numbers
are selected to avoid fractional atoms on a- and b-sites and fractional neighbors.
Likewise, Db is given by
R'!Q!
D - ------"------- (5.31b)
b - (R' - X')!X'!(Q - X)!X!
The overall distribution function Dr is the product of Da and Db, i.e.,

Dr = DaDb·
The energy of the system consists of Ze ij /2 times the number of bonds
of each type, summed up for all types of bonds. The procedure described
by equations (5.33)-(5.74) can then be pursued to obtain all the thermody-
namic properties of order-disorder in these types of alloys.
The preceding method can be generalized to any numbers of N A , N B ,
a-sites, and b-sites. For brevity, however, we limit our treatment to the case
described by equations (5.5b), i.e., NA < N B , La = N A, Lb = N b. Various
types of atoms are then summarized as follows:
N~ = Q = XB(1 - r)L a (5.5e)
(5.5f)
The types of atoms on a- and b-sites corresponding to equations (5.5e),

(5.5f), and (5.30a) are
(i) A on a-sites with A neighbors = X

(ii) A on a-sites with B neighbors = R - X
(iii) A on b-sites with A neighbors = X
(iv) A on b-sites with B neighbors = Q - X
Total number of A atoms = R + Q = La
(5.30b)
(v) B on b-sites with B neighbors = X'
(vi) B on b-sites with A neighbors = R' - X'
(vii) B on a-sites with B neighbors = X"
(viii) B on a-sites with A neighbors = Q' - X"
Total number of B atoms = R' + Q' = Lb
These atoms are shown and summarized in Fig. 5.6 for N A = 6 = La,
NB = 18 = L b , XA = 0.25, and r = 5/9. Again, these numbers are selected
to avoid fractional values for all the items in equations (5.5e), (5.5f), and
(5.30b). The enumeration R - X is illustrated in the caption for Fig. 5.6;
and this value also yields X because R is given by equation (5.5e).
146 Chapter 5
O:A atoms (i) X=I

~:B atoms ( ii) R-X=3
• : a-sites (iii) X= I
x : b-sites (iv) Q -X= I
NA = 6 = XAN =N/4 =La (v) X' =12
Na = 18 =XaN =3N/4 =Lb (vi) R'-X'=4

r =5/9 j R=4 (vii) X" = 2
Q =2; R'=16j Q'=2 (viii) Q'-X"=O
Figure 5.6. Net numbers of A and B atoms on a- and b-sites for NA = 6 = La' NB = 18 = L b ,
and r = 5/9. Six numbered bonds involve A atoms on a-sites with B neighbors; therefore,
R - X = 3 for part (ii). Remaining parts in equation (5.30b) are obtained similarly and
summarized above. All the numbers are selected to avoid fractional atoms on a- and b-sites
and fractional neighbors.
The distribution functions are similar to equations (5.31a) and (5.31b);

thus,
R!Q'
D=-----"'----- (5.31c)
a (R - X)!X!(Q' - X")!X"!
R'!Q!
D - ------"------ (5.31d)
b - (R' - X')!X'!(Q - X)!X!
Again, the overall distribution function, D" is given by Dr = DaDb'

The energy of the system can be obtained by multiplying Ze ij and the
net numbers of each type of atom in equations (5.30b) and adding up the
results; e.g., for the A atoms on a-sites with A neighbors, the energy term
is Ze AA X/2. Again, the procedure described by equations (5.33)-(5.74) can
be pursued to obtain all the thermodynamic properties of order-disorder
in these types of alloys. Some of the most interesting alloys of this category
are Ag2AI, AIFe 3, Au 3Cu, AUCU3, CoPt 3, CU3Pd, FePd 3, Fe 3Pt, FePt 3, Ni4Mo,
and Ni2Mo, to name a few. The preceding equations for NA < N B, La = Lb =
L, and for (NA = La) < (NB = L b ) have not yet been applied to such alloy
systems.
References
I. W. S. Gorsky, Z. Phys. 50, 64 (1928).

2. W. L Bragg and E. J. Williams, Proc. R. Soc. London Ser. A 145,699 (1934); 151,540 (1935).
3. H. A. Bethe, Proc. R. Soc. London Ser. A 150,552 (1935).
4. E. A. Guggenheim, Mixtures, Oxford University P.ress, London, Chapters IV and VII (1952).
5. R. Fowler and E. A. Guggenheim, Statistical Thermodynamics, Cambridge University Press,
London, Chapter XII (1956).
6. M. A. Krivoglaz and A. A. Smirnov, The Theory of Order-Disorder in Alloys, Macdonald,
London (1965); see also J. M. Cowley, Phys. Rev. 120, 1648 (1960).
7. R. M. White and T. H. Geballe, Long Range Order in Solids, Supplement 15, Solid State
Physics, Student Edition, Academic Press, New York (1983). .
8. L Prigogine, The Molecular Theory of Solutions, North-Holland, Amsterdam (1957).
9. D. de Fontaine, Solid State Phys. 34, 73 (1979); Acta Metall. 23, 553 (1975).
10. H. Sato, in Physical Chemistry: An Advanced Treatise, Volume X, edited by W. Jost,
Academic Press, New York, p. 579 (1970).
II. N. A. Gokcen and E. T. Chang, 1. Chem. Phys. 55, 2279 (1971); Scr. Metal/. 4, 941 (1970);
A New Method for Enumerating Molecular Configurations in Propellant Mixtures, Aero-
space Report No. TR-OI72 (2210-10)-1, The Aerospace Corp., El Segundo, California
(1971).
12. N. A. Gokcen, Scr. Metall. 17,53 (1983). (The treatment presented in this reference contains
minor initial statistical errors that have been corrected in this book. However, final equations
and conclusions are correct in this reference.)
13. N. A. Gokcen, Thermodynamics, Techscience, Hawthorne, California, Chapter XI (1975).
15. D. Chipman and B. E. Warren, 1. Appl. Phys. 21,696 (1950).
16. H. Moser, Phys. Z. 37, 737 (1936).
17. C. Sykes and H. Wilkinson, 1. Inst. Met. 61, 223 (1937).
18. R. Kikuchi, Phys. Rev. 81, 988 (1951); 1. Chem. Phys.60, 1071 (1974); and the intervening
papers.
19. E. A. Guggenheim and M. L McGlashan, Mol. Phys. 5, 433 (1962).
20. R. Kikuchi, 1. Chem. Phys. 60, 1071 (1974); R. Kikuchi and C. M. van Baal, Scr. Metall.
8, 425 (1974).
21. D. M. Burley, in Phase Transitions and Critical Phenomena, edited by C. Domb and M. S.
Green, Volume 2, Academic Press, New York, p. 329 (1972).
6
Interstitial Solutions
Introduction
It was stated in Chapter 2 that when the atomic diameter of a metalloid is

about 59% or less than that of a solvent metal, then the metalloid may form
an interstitial solid solution. Such metalloids are hydrogen, boron, carbon,
nitrogen, and oxygen, but silicon, phosphorus, and sulfur may also form
interstitial solid solutions in certain favorable cases. The interstitial solid
solutions of hydrogen and carbon in metals have received particular atten-
tion because they form some of the most interesting alloy systems. Metal-
hydrogen systems are very useful in hydrogen storage, hydrogen purification,
and isotope separation, and metal-carbon systems have unusual structural
and mechanical properties. We shall present and discuss the Pd-H and
Fe-C systems and then the Wagner theory on the solutions of interstitials
in binary metal solutions.
The interstitial lattice sites available for metalloids depend on the
crystal structure of the solvent metal. The number of such sites, e, is
considered to be 3 per metal atom for body-centered cubic structure
and 1 for face-centered cubic, close-packed structures, and liquid
metals. Let n be the number of interstitial atoms, and N, that of the
metal atoms; the total number of interstitial sites is eN, and the fraction
of sites occupied by the interstitial, y, and its atomic fraction x are
given by
n
y=-=_.
r
x=---=
n y
.
xle
y=-- (6.1 )
eN e' n+N y+l/e' I-x
where r = nl N is a convenient variable as will be seen later. The Henrian

149
150 Chapter 6
activity of the interstitial, d, may be written as
.. fy
a =yx = - - (6.2)
1-y
where y and f are the activity coefficients for their respective concentra-
tions, and the activity can be measured in terms of a gas phase potential
or pressure in equilibrium with the solution, such as H2(g) for dissolved
hydrogen [H], CH 4 /H 2, or CO/C0 2 gas mixtures for dissolved carbon [C]
as will be discussed later. We emphasize here that the symbols without
subscripts refer to the interstitial element unless they are qualified by immedi-
ately succeeding words in parentheses.
Diatomic gas molecules dissociate to dissolve as monatomic solutes in
metals. Small solubilities of a diatomic gas such as H 2(g) in metals of Group
VIA (e.g., Cr, Mo) and those to the right side in the periodic chart on
page 318 can be assumed to take place in two steps: (1) by dissociation of
H2(g) into monatomic gas, H(g) (or as superficially adsorbed H), requiring
a large positive value for tlH~ of dissociation, and (2) by dissolution of
H(g) in a metal, generally requiring a relatively small and usually negative
value for tlHll of solution. The value of tlH for the overall process is
tlH~ + tlHll , and this sum is a positive quantity because tlH~ is generally
much greater than ItlHlll; consequently, the solubility for a given pressure
increases with increasing temperature in accord with Le Chatelier's prin-
ciple. The elements in Groups II A-VA, as well as Pd, La, and Ta, dissolve
up to 105 times larger amounts of hydrogen than the remaining poor metal
solvents. For most of these good solvents for hydrogen, again H2(g) dissolves
as monatomic H, but tlH = tlH~ + tlHll is negative so that the solubilities
decrease with increasing temperature. [An italic H refers to enthalpy; a
roman H refers to dissolved hydrogen. Monatomic gaseous hydrogen is
always followed by (g), i.e., H(g).] Palladium dissolves large amounts of
hydrogen, and at certain pressures and temperatures up to 292°C, two phases
coexist with gaseous hydrogen. The phase low in hydrogen is the a-phase,
and that high in hydrogen, the ,8-phase. Above 292°C, known as the critical
temperature, Pd and H form a single solid solution. Hydrogen in palladium
is released without much difficulty even below ambient temperatures when
the pressure is decreased. For this and many other interesting reasons, and
for many technical applications, the solubilities of hydrogen and its isotopes
in palladium have been thoroughly investigated. 1- 5 It is therefore appropri-
ate to devote the next section to the thermodynamics of the Pd-H system
with more details than usual for other systems considered in this book.
Interstitial Solutions 151
_l
,""-1
I
100
I I ~ p
~;1.-
~.'-II
00
+' /
....
198·[
'.~OQ-~-I ~4---'~/ , I/ t 1
I;~~J"'~'~'-'-"'-"'~-i-J~q,'-i
P-'" •~ 188 .[
- _ __- / I
'-g
10 10
~
~ --..=t-.-.-r---.-.-.~~ -/
2oQO{Il-f--o-o-i--o-
143·[ O~o- - -....I
I / I
' t
C 0 100 ·C " / I
I{.-" l ii I /
,--
I 0--0
J
--_-0=0-0.-0
160·[
0--0-0-
¥ -..I
A-w- + -
I' / ! I
if \
110·[
0.1
i /
1---
I "";(.)(-::=\: )(->(-)(-
1 - - ' - )(.~)(!.:A. r-/+
~_IC"': 70·[ \ +
I \ !
, I
+-+ +-+-1
0.01 " ;{
1-;~ -+-1 10·[
1
o 01 02 03 O~ 05 06 07
Figure 6.1. Dissolved hydrogen as r = H/Pd atomic ratio versus gaseous hydrogen (H 2 )
pressure P at various fixed temperatures. For phase boundaries, see Table 6.1. (From Frieske
and Wicke4 with permission.)
Pd-H System
Hydrogen dissolves in palladium as shown in Fig. 6.1, where dissolved
hydrogen as r = HI Pd atomic ratio is plotted 5 versus gaseous diatomic
hydrogen pressure P in atmospheres. The a-phase on the left coexists with
the ,B-phase on the right at each temperature below 292°C and at each
pressure, which is known as the plateau pressure. Neutron diffraction and
other studies indicate that hydrogen occupies the octahedral sites in both
a- and ,B-phases. The number of octahedral sites is the same as the number
of Pd atoms; therefore, N is also the number of octahedral sites. Hydrogen
must also occupy other sites when nl N is greater than 1 at very high
pressures.
The a-phase has a face-centered cubic structure, slightly expanded by
dissolved hydrogen, and the ,B-phase has a highly distorted face-centered
cubic structure. The increase in volume of Pd upon transition from a to ,B
is about 1.6 cm 3I g-atom H. (The volume of pure Pd is close to 8.87 cm 3 / g-
atom.)
The most extensive recent study of the Pd-H system has been carried
out by Frieske 4 •5 * in the range of 20 to 300°C at hydrogen pressures from
0.01 to 140 atm. The solubilities and the phase boundaries were determined
*The results quoted in References 4 and 5 are from H. Frieske, Dissertation, University of
Munster (1972).
152 Chapter 6
by magnetic susceptibility and by weight change of Pd samples. The phase

boundaries at a given temperature are wider and the plateau pressures are
higher upon absorption, and lower upon desorption as shown in Fig. 6.2.
Despite numerous ingenious techniques and various precautions, this lack
of true equilibrium, referred to as the pressure hysteresis, has not been
eliminated. According to Flanagan et al.,6 the pressure hysteresis is due to
dislocation creation and plastic deformation during both hydrogen absorp-
tion and desorption. Neither absorption pressure nor desorption pressure
can therefore be the true equilibrium pressure. Further, the plateau pressures
for absorption and for desorption both tilt upward from left to right, as
shown in Fig. 6.2. However, many investigators have assumed that the
desorption pressure is closer to the equilibrium pressure, an assumption
that is considered to be controversial. Complete phase relationships necessi-
tate separate correlations for absorption and for desorption. We limit our
formulation to desorption, bearing in mind that a similar formulation can
also be obtained for absorption. The composition of a-phase in equilibrium
with {3-phase was obtained by extending the desorption plateau to intersect
the absorption curve as shown at the left in Fig. 6.2. The composition of
{3-phase in equilibrium with a-phase was obtained by extending the straight
portions of the magnetic susceptibility-composition curve and taking their
point of intersection as the equilibrium composition (not shown in Fig.
500 r--..,--,---Ir--..,-I-,---.------r---, 1000
~ D absorption
400 ~ • desorption X lloo
.~ 0 absorption ,:
~~ p ~
300 10
200
X· 1(f ~
cm3/mol ~ atm
100 1-;)oF----+---+-------1~-_+_ -+----+----1 0.1
'liD
of-----+----+------+----+----+-l'-~-=J~[T-flIIPl!!I 0.01
o 01 02 03 04 05 06 07
Figure 6.2. Magnetic susceptibility X and diatomic hydrogen pressure P, versus r = H/Pd
atomic ratio at 120°C. (From Frieske and Wicke' with permission.)
Interstitial Solutions IS3
6.2). No particular significance of interest will be attached in this chapter

to the intersection of the dashed lines on the right side in Fig. 6.2. The
compositions of coexisting phases determined by the foregoing procedure
are shown by the dashed concave-down curve in Fig. 6.1, and listed in
Table 6.1.
The plateau pressures, P, for 20 to 300°C in Fig. 6.1 are closely
represented by
4PdH 0 .5(,B-phase) -+ 4Pd(a-phase) + Hig) (6.3)
tlHO tlSo 4932

In P(atm) "'" - - -+- = --- + 11 .73'
RT R T ,
(tlHo = 9800 cal/mole H 2 ; tlSo = 23.3 cal/K-mole H 2 ) (6.4)
The activities of both Pd and PdHo.5 in reaction (6.3) decrease with increasing
temperature because both phases dissolve hydrogen; hence, P is nearly
equal to the equilibrium constant ofreaction (6.3); therefore, equation (6.4)
is simply an empirical equation without the activity corrections for PdHo.s
and Pd. Other equations for different ranges of temperature are available;
e.g., tlHo = 9325 cal/mole H 2, and tlSo = 21.8 cal/K-mole H2 have been
reported for -80 to +50°C by Wicke and Nernst. 7 (For lower temperatures,
see, e.g., Lynch and Flanagan. 8 ) The values of tlHO agree well with the
tlHo = 9440 cal/4 moles PdH o.5 based on the calorimetric data of Nace
and Aston. 9 Equilibria between a-phase and H 2 , or D2 have been investi-
gated in the range of approximately 0 to 100°C by Wicke and Nernse whose
results agree very well with those of Clew ley et al. lo :
Table 6.1. Pressure-Temperature-Composition in Two-Phase Field

in Pd-H System Q
°C T, a-phase T, ,B-phase P, atm H 2 (g) b
20 0.008 ± 0.002 0.607 ± 0.002 0.0061

70 0.017 ± 0.002 0.575 ± 0.002 0.071
120 0.030 ± 0.002 0.540 ± 0.003 0.44
160 0.046 ± 0.002 0.504 ± 0.003 1.41
200 0,075 ± 0.002 0.459 ± 0.004 3.7
243 0.117 ± 0.002 0.399 ± 0.005 8.8
288 0.21 ± 0.01 0.29 ± 0,0} 18.9
292 ± 2 c 0.250 ± 0.005 c 0.250 ± 0.005 c 19.7 ± 0.2c
"Composition is in r = H/Pd atomic ratio, and pressure of H2 (g), P, in atmospheres.

From Refs. 4 and 5.
bCalculated from equation (6.4) using the temperatures in the first column.
cCritical point; values based on experiment.
1S4 Chapter 6
Table 6.2. Values of AH(r) and AS for Equation (6.7)
Gas IlH(r), caljrnole H2 IlS, caljK-rnole H2
H2 23,970-21,500 r 25.6
D2 22,820-21,500 r 25.4
T2 22,050-21,500 r 24.6
0.5HAor D 2) ~ [H or D in a-phase]; Kp(H or D) = p:.5

1162 949
(H) =--642'
InKP T' , In Kp(D) = T - 6.40 (6.5)
Similar results for tritium in the range of 25 to 100°C, obtained by Schmidt, II

are represented by
832
In Kp(T) = T - 6.25 (6.6)
These equations refer to the a-phase within a limited range of temperature

and concentration.
The corresponding empirical equations based on the data of Sickingl2
and Wicke and Nernse for {3-phase desorption in a temperature range of
roughly -34 to +83°C are summarized by Wicke and Brodowsky5 by using
aH(r) and as in the following equation:
aH(r) r as
InP= - - - + 2 I n - - + - (6.7)
RT 1- r R
The results are listed in Table 6.2. for aH(r) and as.
The gas-phase equilibria, e.g., H2 + D2 = 2HD, can be combined with
these equations for use in isotope separation. The critical temperature for
the Pd-D system is 276°C (292°C for H 2), and the critical pressure is 35 atm
(19.7 atm for H 2).
The Pd-H and related equilibria and other properties have been dis-
cussed elsewhere in full detail. l-8
Statistical Treatment, Zeroth Approximation
The simplest statistical treatment of interstitial alloys, such as H in a

metal M, is based on the zeroth approximation to the regular solutions. We
start the alloy formation process with pure monatomic gaseous hydrogen
H(g) at 1 atm, in the same way as we start with pure metals at their standard
states in forming all metallic alloys. The basis of treatment is n atoms of
Hand N atoms of metals providing eN interstitial sites. Let ZII/2 and Z12
be the respective numbers of H-H and H-M bonds per atom of dissolved
H; ell and e12, the corresponding energies per bond; and eJ. the energy of
gaseous atomic hydrogen. The probability of finding a dissolved H atom
at a selected site is n/ eN and the probability of finding another H atom
next to the first atom is n 2/(eN)2. The total maximum number of bonds,
if all the interstices were occupied, is ZII cN /2, and multiplication with
n 2/ (eN)2 yields the actual number of existing H-H bonds, i.e., Zlln 2/(2eN).
The number of H-H bonds times ell, the H-H bond energy, yields
Zlle ll n 2/2cN, which is the energy due to the H-H bonds. The number of
H-M bonds per atom of H is Z12, and the total number of H-M bonds is
Z12n, which, multiplied by the bond energy e 12 , yields Z12e12n. The energy
of the metal is Z22Ne22/2 = Nh'2, and the enthalpy of the alloy is
where H(alloy) is the enthalpy of alloy, and h~ = e l + pv is the standard

enthalpy of one atom of hydrogen, H(g). The energy of metal M, Z22Ne22/2,
is written as its enthalpy Nh'2, which is assumed to be unaffected by the
dissolved interstitial. The entropy of hydrogen in solution is obtained in
three steps. Let s~ be the entropy per atom of gaseous hydrogen, H(g), at
1 atm; for the overall process of compressing and freezing monatomic
hydrogen to a hypothetical solid having the same volume as metal M, the
entropy decreases by s* to s~ - s*. The next process of random mixing by
the diffusion of solid hydrogen to the interstices increases the entropy of
solid hydrogen by distribution of n atoms of hydrogen in cN interstices.
This distribution, D, is given by
D = _(,-c_N--'..)_!_
(6.9)
(eN - n)n!
The resulting overall entropy of the alloy, S(alloy), is then
S(alloy) = n(s~ - s*) + Ns'2 + kin D ( 6.10)
where Ns'2 is the entropy of metal, which is assumed to be unaffected by

interstitial alloying. The Gibbs energy of alloy having n atoms of H, (cN - n)
156 Chapter 6
interstices, and N atoms of metal is O(alloy), which is obtained by sub-

stituting equations (6.8) and (6.10) in O(alloy) = H(alloy) - TS(alloy);
thus,
O(alloy) = Zlle ll n 2 /(2eN) + Zl2e12n + nh~ - nT[s~(g) - s*]
+ N(h~ - Ts~)
- kT[ eN In eN - (eN - n) In (eN - n) - n In n]

(6.11)
Here, h~ - Ts~ = O~I N is the standard Gibbs energy per atom of metal.
The last set of terms represent - kT In D, expanded by using the Stirling
approximation for factorials.
Equation (6.11) can be simplified by using the following notation:
0= NoO(alloy); (No = Avogadro's number)

2A = -eNOZlle ll ; O~ = No(h~ - Ts~); ( 6.12)
Oe = N O[ZI2e 12 + h~ - T(s~ - s*)]
Substitution of these equations in equation (6.11) results in
n2 e
0= -A-2-+ nO + NO~
eN
- RT[eN In eN - (eN - n) In(eN - n) - n In n] (6.13)
The partial molar Gibbs energy of dissolved hydrogen, G, is obtained

by partial differentiation of 0 with respect to n:
-
0= [a~
-
an P,T,N
=O
e
-2A
--+n R T l n -n- -
e eN eN-n
(6.14)
Substitution of r = nl N gives
- • 2A r
0=0 --2 r+ RTln-- (6.15)
e e-r
Dissolved hydrogen is in equilibrium with Hig) at its pressure P according

to
O.5H2(g) = [H in metal] ( 6.16)

Therefore, G is equal to 0.50(H 2) so that
( 6.17)
The use of this equation to eliminate G in equation (6.15), and a simple

rearrangement of the resulting relationship leads to
° 2A r
0.5RT In P = 0 - 0.50 0 (H2) - -2 r + RT I n - - ( 6.18)
c c-r
This is the equation frequently used by researchers in correlating P with r.

The unknown terms 0° - 0.50 0 (H2) and A must be determined by using
appropriate experimental data.
It is useful at this point to attach a thermodynamic significance to 0°
in equation (6.18). As y = n/cN approaches zero, the activity coefficient!
in equation (6.2) approaches unity, 2Ar/ c 2 in equation (6.18) becomes
negligible, and y/(l- y) = r/(c - r) approaches y = ric; therefore, re-
arrangement of equation (6.18) results in
(6.19)
where Kp = r/(cPOS) is the equilibrium constant for reaction (6.16) in the

dilute range where ! = 1. The values of Kp at various temperatures are
used in equation (6.19) to determine flHo and flSo in 0° - 0.50 0 (H2) =
flHo - TflSo, and substitute the result in equation (6.18) to obtain the
following alternative relationship:
° ° 2A r
O.5RTlnP = flH -TflS --2r+RTln-- (6.20)
c c-r
The remaining unknown parameter A can be determined from significantly

large values of r at the correspondingly large values of P of H2(g). A
relationship similar to equation (6.18) or (6.20) was first derived by Lacher. \3
Experimental results for M-H systems have been interpreted in terms
of three-dimensional oscillating H or H+ in metals, by using equation (6.18).
These interpretations are by no means universally accepted due to various
reasons, but chiefly because (1) the assumption that the zeroth approxima-
tion is valid, i.e., that hydrogen is dissolved randomly, cannot be fully
justified, particularly at high values of r, as discussed in Chapter 3, and (2)
the lack of true equilibrium, or the presence of the pressure-concentration
hysteresis as the two-phase region is approached, either from high- or low-r
ISS Chapter 6
region. The distribution of H atoms in the octahedral sites is indeed nonran-

dom as shown by extensive investigations. 14
Activity and Activity Coefficients

The value of the equilibrium constant for reaction (6.16) does not vary
with r. For sufficiently large values of r, we write Kp in terms of the activity
a and the activity coefficient f to obtain their dependence on concentration
and temperature:
K = _a_· = _---"f-'-y---;,-;: (6.21)

p po.s (1 _ y)po.s
Recall that the symbols without subscripts refer to the solute, and Prefers
to H2(g). Substitution of Kp and cy = r in equation (6.18) and elimination
of d - 0.5GO(H2) yields
• 2A y
6 - G = RTln a = - - y + RTln-- (6.22)
c l-y
Partial differentiation of equation (6.13) with respect to N and elimination

of r by using cy = r yields 6 2 for the solvent metal:
-af'f = G- 2 = + Ay + cRTln(l -
2
G~ y) (6.23)
aN
where 6 2- G~ = RT In a 2 , and the standard state for a 2 is the pure metal.
Solid Interstitial Solutes
We extend the derivation of the foregoing equations to solid interstitials.

Let a solid interstitial be C which may be the elements such as boron,
carbon, silicon, or phosphorus, for which s* is zero, and nh~ - nTs~ refers
to pure C. Consequently, equations (6.13)-(6.15) are valid and equation
(6.15) is the important relationship for 6 and for the activity ofC denoted
by u. We rewrite equation (6.15) by using y = ric and g =fY/(l- y) as
follows:
• • 2A y
6 + RTln a = G --y + RTln-- (6.24)
c l-y
On the Henrian scale for the activity, as y approaches zero, i.e., y ~ 0, then
f ~ 1, y/(1 - y) ~ y, and equation (6.24) becomes 6 = d + RT In y.
Comparison with the Henrian definition of activity 6 = GO + RT In y at
y ~ 0 and a ~ y shows that GO in equation (6.24) is the standard Gibbs
energy with the reference state asf ~ 1 at y ~ o. Equation (6.24) is therefore
identical with equation (6.22).
The activity on the Henrian scale in equation (6.24) is given by
6 - G" = RT In a for any concentration; however, the activity g on the
Raoultian scale, in which the pure solid is the standard state for the
interstitial, is obtained by subtracting GO from both sides of equation
(6.24):
- ° 2A
G - GO = RT In g(Raoult) = G - GO - - y + RTin -y- (6.25)
c 1-y
The equation for GO - GO, as well as for A, must then be obtained by fitting
the activity data with equation (6.25) at various temperatures and concentra-
tions. It is quite frequently possible to express GO - GO as a linear function
of temperature, i.e.,
(6.26)
where I:lHo and I:lSo are constant within experimental errors.

Substitution of a = fy / (1 - y) in equation (6.24) and rearrangement
of the result gives the equation for the activity coefficient:
- 2A
G e == RTlnf= - - y (6.27)
c
Equations (6.24) and (6.27) justify the definition of f by equation (6.2). It

can be shown that the corresponding equation for the solvent metal is
identical with equation (6.23):
(6.28)
As y decreases, Raoult's law is approached and Ay2 becomes negligible so

that this equation becomes 6 2 - G~ = RT In(1 - yr; hence, the concentra-
tion that must be used to correlate the activity a 2 and the activity coefficient
f2 is (1 - y r,
i.e., a2 = f2(1 - y) c. Substitution of a 2 in equation (6.28) as
6~ - G~ = RTlnf(1 - yr and the use of definitional equation 6~ =
RT Inf2 yields
(6.29)
160 Chapter 6
This equation can also be derived by using equation (6.27) in the Gibbs-
Duhem relation, dO~ = -[x/(l- x)] doe = 2Ay dy and integrating it from
y = 0 to y, observing that O~ is zero at y = 0 since a2 ~ 1 with y ~ o.
The molar excess Gibbs energy G e = xO e + (1 - x) O~ and the molar
Gibbs energy G = xO + (1 - x) O 2 can be easily obtained by substituting
the partial properties in these equations but we shall have no specific use
for such equations.
Alternative derivations of equations (6.24) and (6.27), e.g., that given
by Hillert and Staffansson/ 5 are also based on the zeroth approximation
to the regular interstitial solutions. In a number of applications, it is assumed
that A is a linear function of temperature, i.e., the bond energy ell is no
longer constant but varies linearly with temperature. This is a useful
empirical concept, as will be shown later in this chapter.
Application to Pd-H System
Palladium is face-centered cubic in crystal structure, which provides

one interstitial site for each Pd atom, i.e., c = 1. Equation (6.18) for the
Pd- H system is then
G- - 0.5GO(H 2 ) 2Ar I r
0.5 In P = -- + n-- (6.30)
RT RT 1- r
The miscibility gap between a and f3 phases disappears above the critical
temperature, Te , which is 565 ± 2 K in Fig. 6.1, where a horizontal inflection
appears in the P versus r curve. The first and second derivatives of equation
(6.30) with respect to r are zero at the critical point:
a In P 1 1 2A
0.5--=-+----=0; (6.31)
ar r 1 - r RT
a2 ln P 2r - 1
0.5 - - 2 - = ( = 0; (re = 0.5) (6.32)
ar 2)2
r - r
The first of these equations determines A in terms of Te and r = re, and the
second, the critical value of re , which is 0.5. Unfortunately, the experimental
value of re is 0.25, far below re = 0.5 from equation (6.32). An artifice
defended by a number of investigators is that there are two sets of sites
available for dissolved hydrogen: the first set is 0.6N of the total available
set of N, and the second set is O.4N. This artifice has been generally rejected l6
because all the N sites are equivalent. As evidence for the existence of two
such sites, a distinct change in the magnetic behavior of the Pd-H system
is often cited, i.e., Pd-H alloys change sharply from paramagnetic state to
diamagnetic state when r exceeds 0.6 or after 0.6N sites are occupied by
dissolved H. Substitution of 0.6N for N in the foregoing equations would
simply replace r by () = r/0.6. Equation (6.32) with () would then yield
(}c = 0.5 and rc = 0.3, which is not too far from the experimental value in
Table 6.1. The sites in excess ofO.6N, i.e., O.4N sites, are said to be available
for H at high pressures and low temperatures, and for this range it is
proposed that an empirical equation linear in r is obeyed:
In P = A + Br (6.33)
where A and B are independent of r, and linearly dependent on 1/ T, e.g.,

A = 1+ (m/ T). An alternative empirical relationship is equation (6.7). The
solubilities at high pressures for Pd-H and other systems are discussed in
much greater detail elsewhere/,17 and need not be considered here since
equations such as (6.33) are essentially empirical.
An alternative procedure to obtain an equation consistent with the
value of rc is to assume the existence of interactions beyond the closest
neighbors, leading to an additional term in r2/ RT in equation (6.18). This
term can be generated by writing the probability for a three-atom cluster
(n/ N)3, multiplied by its energy per cluster E' and the number of Pd atoms
N, i.e., E' n 3 / N 2 , and then substituting the result in equation (6.11). This
argument is not rigorous because it does not account for the corresponding
entropy effect. Subsequent differentiation as in equation (6.14) contributes
a term 3NoE'r2/ RT to equation (6.18); thus,
O· - 0.50 (H2) 2Ar

0 r 3NoE'r2
0.5 In P = --+In--+--=--- (6.34)
RT RT 1- r RT
The relationships corresponding to equations (6.31) and (6.32) at the critical

point are
2A 1 6NoE'r
--+--+ =0 ( 6.35)
RT r - r2 RT
(6.36)
The values of rc = 0.25 and Tc = 565 are now required for determination
of A and 6NoE'/ RT; substitution of these values in the preceding equations
yields
2A/ R = 5022, 3NoE'/ R = 4018 (6.37)

162 Chapter 6
The resulting relationship expressing In P as a function of r is
1140 r 5022 4018 2

0.5 In P == 6.35 ---+ I n - - - - - r +--r (6.38)
T l-r T T
where the first two terms on the right represent the solubility of H2 in the
a-phase for very small concentrations of r when In[r/O - r)) = In r, and
when the last two terms become negligible so that
d - 0.5GO(H2) r 1140
---R-T--'--''''- == In -po-.s == -T- - 6.35 (6.39)
Observe that the solubility decreases with increasing temperature according

to this equation for a fixed pressure, P.
Equation (6.39) represents the data for the a-phase in Fig. 6.1 well
within the errors indicated in Table 6.1 for desorption. It also represents
the data for the ,8-phase in a qualitative way but generally this phase requires
equations similar to equation (6.7) or (6.33) with empirical parameters as
mentioned earlier. Equations identical in form with those for a- and
,8-phases may also be derived for the absorption isotherms.
In the regions of sufficiently low values of r for the a-phase, and
sufficiently high values of r for the ,8-phase, the absorption and desorption
isotherms coincide without hysteresis, well within experimental errors;
consequently, equation (6.38) is obeyed by H in a-phase, and equation
(6.33) in ,8-phase for both absorption and desorption isotherms in the
single-phase regions.
There are other additional terms recommended to be included in
equation (6.30) on the basis of interpretation of experimental results. 16- 18
For example, the electrons from H atoms contribute to the 4d band of Pt,
and to the electron gas of Pd, and the H atoms in the octahedral sites are
conceived to be electronically screened protons. Therefore, when a third
component such as Cu is added in Pd, the solubility of hydrogen decreases
due to the contribution of one electron per Cu atom to the Pd-H solution,
in agreement with observations. 16 However, the number of electrons con-
tributed by the elements is not always predictable.
A pair of empirical equations similar to equation (6.30) have been
fitted by Evans and Everett I8 to their data on absorption and desorption
isotherms of the a-phase, and another equation, similar to equation (6.33),
has been fitted for the ,8-phase, and further, their treatment has been
extended to deuterium in palladium.
The first approximation to the regular solutions in its correct form has
not yet been applied to the interstitial solutions. If elI in equation (6.11) is
taken to be zero, then the random distribution given by equation (6.9) would
be valid, and the zeroth and first approximations would be identical.
Other Hydrogen-Metal Systems
The pressure-composition diagrams are more complex in systems hav-

ing more than one set of two phases. In such cases there are corresponding
numbers of plateaus, each plateau for each pair of coexisting phases.
Between two plateaus there is a single phase for which the pressure and
concentration both vary until the next plateau is reached. An interesting
example is the ErCorH system,t9 which contains two plateaus at 101°C,
with the first plateau terminating at about 1.7 atoms of H per mole of ErC0 3 ,
and the second, at about 4.5 atoms per mole of ErC0 3 •
The concentration-temperature diagrams are simple for some systems
such as the Pd-H system which contains a solubility gap with a critical
point, and can be constructed from the data in Table 6.1. However, the
majority of such diagrams are very complex as shown in Fig. 6.3 for the
Ta-H system. 20 •21
Hydrogen Storage
Hydrogen is a versatile fuel, readily generated by various methods and

easily converted to other forms of energy. Large solar farms and thermonu-
clear reactors of the future may generate enormous amounts of hydrogen
that can be used as a convenient and efficient source of energy. These
areas/ 2 •23 as well as the photovoltaic cells presented in Chapter 7, are active
fields of research on renewable sources of energy. It is therefore appropriate
to devote this section to hydrogen storage.
The storage of large amounts of hydrogen poses numerous problems.
Liquefaction wastes about 25% of the available energy of hydrogen, and
compression into cylinders requires large and heavy equipment that may
consume 15% of the energy of hydrogen. Certain alloys can store and
release hydrogen readily without consuming large amounts of energy; their
hydrogen capacities per unit volume may be in excess of that attained by
liquid hydrogen. A number of interesting alloys appear to be promising
and useful for practical applications; therefore, a brief presentation of this
topic, based largely on extensive investigations of Wiswall/ 4 Reilly/5 and
Hoffman et al.,26 is given in this section.
164 Chapter 6
100
350 + + + + + + a' + ·c
K
a 50
300 + 6 +
250 + + + + +
a. p -50
,200 + + + + I
t
T -100
150 + + + +
, .y
-150
100 + + + + Y
.
-200
....... 7,-
50 + + + + '
"
C+L 1L:
"
-250
""
"
01 02 03 04 05 0.6 07 08
r- H/Ta
Figure 6.3. Tantalum-hydrogen system; r = H/Ta atomic ratio. (From Kobler and Welter20
with permission.)
The plateau pressures in metal-hydrogen systems are generally linear

when In P is plotted versus 1/ T in K- 1 either for the absorption curve or
for the desorption curve as shown by equation (6.4). Figure 6.4 shows the
desorption pressures of a number of systems in which gaseous diatomic
hydrogen is in equilibrium with the indicated phases. 24 All such media
exhibit pressure-composition hysteresis as in the Pd-H system at tem-
peratures below the annealing range of the metal or alloy, and further, the
plateaus are seldom exactly horizontal, particularly for the alloys of two
metals. Two or more phases may appear at a given temperature in many
metal alloys and each pair of phases is at a different plateau pressure than
another pair of phases as shown in Fig. 6.5 for the FeTi-H system.
There are two types of reactions between a hydrogen-rich phase (or
hydride phase) AH, and a metal B, capable offorrning an alloy phase ABn:
(6.40)
(6.41)
t (ee)
250 150 60
F,Ti-F, Ti H
-
E
co
0..
0.1
0.01
0.001 ~~--~----~----~--~~--~~--~----~
1.8 2.2 2.6 3.0
IOOO/T
Figure 6.4. Dissociation pressures of metal-hydrogen systems. (From Wiswall 24 with per-
mission.)
In the first case, A and B form such a stable phase that hydrogen is displaced
as a gas, and in the second case, hydrogen and B share strong bonding with
A. In the latter case, hydrogen from ABnH, is released as follows:
(6.42)
166 Chapter 6
N
X
E
'0 10
W
a:
::>
(f)
(f)
w
a:
a..
z
0
I-
<t
U
0
(f)
(f)
0.1 '----'--:-'":::---'-"'---l---L.---l_-'------L----l._L-...L--L---l
0.2 0.4 0.6 0.8 1.0 1.2
ATOM RATIO H/(Fe + Til
Figure 6.S. Desorption pressure in FeTi-H system. (From Wiswa1l 24 with permission.)
where ABnHr-v may dissociate to release more hydrogen in the succeeding

stages. If B is a non-hydride-forming element, the hydrogen pressure over
ABnHr is often orders of magnitude larger than over AH r. Thus, according
to Libowitz et al./7 the hydrogen pressure over ZrNiH2 is about 10 10 times
greater than that over ZrH 2 at 250°C because Ni does not form a stable
hydride phase at moderate pressures. In some cases, other types of reactions
may occur as in the reaction involving Mg 2Cu:
( 6.43)
If, however, both A and B are stable hydride formers, more complex
reactions may often occur.
Hydrogen Storage Metals and Alloys
The alkali elements and their alloys form stable hydrides, unsuitable
for hydrogen storage at temperatures below 300°C. Alkaline earth elements
(Ca, Br, Sr) also form stable hydrides, but an alloy phase of calcium, CaNis,
is of possible interest because it stores hydrogen up to CaNisH6 and
dissociates at room temperature in the vicinity of 1 to 15 atm. Other calcium
alloys also have possible uses because they are similar to CaNis.
Hydrides of magnesium and its alloys are more promising media for
storage. Magnesium dihydride, MgH2' contains 7.65% by weight hydrogen
and dissociates at 287°C by adsorbing 17.8 kcalj mole H 2. The waste heat
from a combustion chamber or from a hot exhaust given off by a motor
can be used to supply the necessary heat for dissociation. The hydride is
re-formed at higher temperatures and pressures. The intermetallic phase,
26
24
I
• Pr"S5Y'"
[aIm/
22
20
18
16
12
t= 40'C
t = 20'C
2 3 4 s 6 7
Hydrogen Concentration
(at H/mol LaNiSI
Figure 6.6. Absorption-desorption isotherms for LaNis at various temperatures. (From
Kuijpers and van Mal 31 with permission.)
168 Chapter 6
Mg 2Cu, absorbs H2 according to the reverse of reaction (6.43). The equili-

brium pressure is about 1 atm at 239°C, lower than that for MgH 2. A similar
phase in the Mg-Ni system, Mg2Ni, reacts with H2 to form Mg 2NiH 4. Both
Mg 2Cu and Mg 2Ni are much more easily hydrided than Mg, but their
hydrogen contents are less than half as much as that for Mg. Some alloy
phases of Mg-AI might also be useful. Other alloys of magnesium appear
to have much fewer potential uses.
Scandium, yttrium, and the lanthanides form very stable dihydrides,
and less stable trihydrides. Dissociation pressures of dihydrides are very
low; hence, reversible storage is not possible. Some alloys of ABs type,
where A is Y, Th, or a lanthanide and B is usually Co or Ni, have interesting
properties. 28 -3o Results on pressure-temperature-hydrogen/metal atomic
ratio, P- T -r, are available for Y, Th, La, Ce, Pr, Nd, Sm, Gd as metals A
and Co and Ni as metals B. A number of ternary metal-alloys of ABs-x M
and AI_yCyB s types, where M is Pd, Ag, Cu, Fe, Cr, or Co and C is Er,
Y, Gd, Nd, Th, or Zr, have also been investigated. 24.2s
The results for lanthanum pentanickel by van Vucht, Kuijpers, and
co_workers 28 .31 are typical for the ABs-type phases as shown in Fig. 6.6,
where it is seen that LaNis may contain close to 7 atoms of easily recoverable
hydrogen at ambient temperatures where the hydrogen pressure, P, is about
2 atm. The concentration of H in the {3-phase is about 1.4% by weight
because the molecular weight of LaNis is rather high. The pressure hysteresis
is moderate and rehydriding is easily accomplished. When cobalt progress-
ively replaces nickel in LaNi s, the hydrogen pressure also decreases pro-
gressively. Further, when n in LaNi n increases from 4.9 to 5.5, the plateau
pressure increases by a factor of more than 3.3. While most hydride-forming
phases are sensitive to oxygen or moisture, LaNis is not. However, if carbon
monoxide is present, LaNis becomes "poisoned" or inhibited for hydrogen
absorption but a small amount of copper slightly reduces the poisoning
tendency.
Ti, Zr, Hf, and Their Alloys
Group IVA elements, Ti, Zr, and Hf, form binary hydrogen metal alloys
of considerable stability; therefore, they are not suitable for hydrogen
storage. Some metal binary alloys of the type TiFe have potential applica-
tions because they absorb H2 up to TiFeH2 and release it as shown in Fig.
6.5. The pressure hysteresis is high and the concentration of H is only 1.91 %
by weight. Powdered TiFe is not pyrophoric and not very expensive, and
for stationary storage it is possibly the best storage medium according to
Wiswal1. 24
Partial substitution of Fe by Ni or Co in TiFe results in considerable

changes in hydrogen capacity and pressure. For example, the hydrogen
desorption pressure for TiFeo.9Nio.! is about an order of magnitude lower
at 50°C than for TiFe. Both TiNi and TiCo form more stable hydride phases
than TiFe. Intermetallic compounds or phases such as TiFe2, TiC0 2, and
TiNi3 do not absorb significant amounts of hydrogen. The compound TiCr2
absorbs hydrogen at ambient temperatures at well above 100 atm.
Group VA elements, V, Nb, and Ta, form stable monohydrides of
narrow stoichiometry. Their dihydrides dissociate to monohydrides under
near-ambient conditions. Thus, for VH 2, dissociation into VH occurs at
l3°C and 1 atm, yielding 1.9% by weight hydrogen, without large pressure
hysteresis. The enthalpy of dissociation of VH 2 to VH is 9.6 kcaljmole H2
released. Addition of small amounts of Ti and Zr lowers the hydrogen
pressure; however, addition of Ta, Mo, Si, Fe, and possibly other elements
to the right side of V-A on the periodic chart, increases the hydrogen
pressure. Vanadium is expensive; therefore, it can be used in applications
requiring small quantities of hydrogen where the cost can be justified.
Other compounds such as TiV4 H s, Th 4 H!s, and VH3 are fairly stable
and expensive hydrides. The behavior ofThFes, ThCo s, and ThNis is similar
to the corresponding phases of lanthanum, but the dissociation pressures
of their hydrides are higher than those of LaBs phases where B is a
nonlanthanide transition metal. The properties of selected useful hydrides
are listed in Table 6.3. (See also Bambakidis. 32 )
Practical Applications
The processes requiring hydrogen storage and use are in their infancy.
Nevertheless, possible uses have already either been proposed or made in
Table 6.3. Thermodynamics and Hydrogen Contents of Selected Substances a
Reaction tlHfo, tlSfo, Density of %H g-atoms

with H2 kcal/mole H2 cal/K-mole H2 hydride, g/ cm' bywt H/cm 3
Mg .... MgH2 -17.8 -32.3 1.4 7.6 0.111

Ca .... CaH2 -41.7 -30.4 1.8 4.8 0.085
LaNis .... LaNisH6 -7.6 -26.0 8.3 1.4 0.113
Ti .... TiHl.97 -29.9 -30.0 3.8 4.0 0.147
Y .... YH 2 4.5 2.1 0.171
H 2(Iiquid)b 0.071 100 0.070
a Adapted from WiswalI. 24

b Oata for liquid H2 is given for reference, showing that all listed hydrides contain more hydrogen per unit
volume than liquid hydrogen.
170 Chapter 6
a few significant cases. Hydrogen concentration cells are the simplest power
generators that take advantage of a pressure difference between anode and
cathode. It is simply a concentration cell that generates power by consuming
hydrogen at a high pressure and releasing it at a lower pressure. Such cells
are already in use in space vehicles. The use of metal hydrides for providing
steady hydrogen pressures for the anode and cathode and thus storing or
generating considerable amounts of power in large- and small-size batteries
has been proposed.
Production and storage of hydrogen and its reuse in fuel cells may
provide a method for electric utility power leveling. A small experimental
model was developed by using TiFe for hydrogen storage, and an enlarged
research and development program ensued. The results appear to be
promising.
Experimental automotive propulsion has been achieved by using
TiFeH n • About 7.7 kg of useable H 2, stored in 1016 kg of TiFeH m has been
demonstrated to be capable of providing a range of 121 km, at a sustained
speed of about 80 km/hr for a bus weighing about 6.8 metric tons. The hot
exhaust gases were used to dissociate the hydride. The spent hydride was
recharged to 80% of its capacity in 15 min and full capacity in 1 hr. 24 This
performance is far superior to that obtained by acid battery automotive
propulsion.
Magnesium hydride, MgH2' has more than three times the available
hydrogen from TiFeHl.9 on the basis of hydrogen per unit weight of alloy.
It would permit longer range for automobiles with a lighter storage medium.
Its dissociation enthalpy and temperature are rather high, demanding special
engines with high exhaust temperatures for this purpose.
Ultrapurification of hydrogen, fractionation of H, 0, and T, can readily
be achieved by using metal diaphragms from hydride-forming metals. Metal
hydrides can be used for gas compression for power generation in turbines,
for heat storage, and even for air conditioning. Other uses are discussed by
Wiswall/ 4 Reilly/5 and Hoffman et al. 26
Fe-C System
The Fe-C system has been investigated by more than 200 researchers
because of the vast industrial importance of iron alloys. We cite a few
selected recent papers, summaries, and compilations from which the reader
may find all the remaining publications.
The phase diagram as computed by Ohtani et al. 33 from their thermody-
namic equations based on selected data for the activity of carbon and phase
boundary compositions is shown in Fig. 6.7. This diagram contains small
Mol Fraction
1800
L + Graphite
1700
1400
Fe 3 C
u
1300 1 1600
::.::
0
w-
1500
1200 w
0:: 0::
::> ::>
J-- Y 1400 ~
<{ 1100 0::
0::
W w
a... Y+Graphite or Fe 3C 1300 ~
~ 1000 w
w
J-- J--
1200
900 p 0.0223
p' 0.0206
S 0.764
s' 0.671 1100
800
740 0 C
700 Ifp~~s'=_~-=-=-=-=-=-=-=-~-=-=-=-=72::-7-=O::c-=-=-=-=--=-,=--:::I 1000
a + Graphite or Fe 3 C
I
I 900
600~--~--~~~~~~--~--~~
Fe 1.0 2.0 3.0 4.0 5.0 6.0 7.0
CARBON CONTENT, wt pct
Figure 6.7. Calculated phase diagram for Fe-C(gr) and Fe-Fe 3 C systems. (Adapted from
Ohtani el al. 33 with permission.) Broken lines are for Fe-Fe 3 C system.
improvements over the previous diagrams,35-41 but represents the best, and
in some respects the most recent results obtained by thermodynamic and
metallurgical investigations. The evaluation of the data and the resulting
summary in the form of thermodynamic relationships require a great deal
of experimental expertise and judgment. The more recent evaluations by
Nishizawa and co-workers/ 3.34 Agren,35 Harvig/ 6 and Chipman 37 are
required reading for the interpretation, evaluation, and selection of ther-
modynamic data and the construction of phase boundaries. We assume that
the necessary thermodynamic equations are available as given by Ohtani
et a/. 33 and proceed to compute the phase diagram from these equations.
We shall hrst discuss the Fe-C(gr) [iron-graphite] system in detail and then
outline the Fe-Fe3C [iron-cementite] system.
172 Chapter 6
a-Phase
The a-phase of pure iron is the body-centered cubic phase that exists
from ambient temperatures up to 1184 K. Its Curie temperature is 1043 K.
It dissolves very small amounts of carbon in equilibrium with graphite.
Therefore, with A == A", g(Raoult) = 1, and yl(l - y) """ y, equation (6.25)
can be rearranged to obtain grG O - "G" = RTlnSy" = -ll.H" + Tll.S", where
the last equality is from equation (6.26), and s on Sy" is for graphite
saturation. The recent results of Hasebe et a1. 34 , as reinterpreted here, are
• sxO'
grGo_"G """ RTlnSy" """ RTln-= -99,750+ 33.6T
3
1.52 X 10 10 4.80 x 1015)

+ ( 8170 - T2 + T4 ;
[Jig-atom of C (800 < T < 1200)] (6.44)
where ll.H" = 99,750, and ll.S" = 33.6, and all the energy units are in joules
in the remaining sections of this chapter. The last set of three terms in
equation (6.44) is a purely empirical correction representing the effect of
paramagnetism-ferromagnetism on the solubility of graphite in the range
of 800 to 1200 K. This correction is derived from H~(ferromag.)
H 2(paramag.) == ll.H 2(f.p.) determined by calorimetric measurements and
evaluated and summarized. 42 ,43 Nishizawa and co-workers 33 ,34 have pro-
posed that the contribution to equation (6.44) is ll.H 2(f.p.) x (-500/1043)
where -500 is the rate of change of Curie temperature in kelvins per atomic
fraction of carbon, and 1043 is the Curie temperature of pure iron. The
solubility of carbon in a-Fe is very small; therefore, -500 cannot be
determined accurately; hence, it must be regarded as a reasonable correla-
tion factor that relates ll.H 2(f.p.)/1043 to the Gibbs energy contribution in
equation (6.44). Other types of interpretation of magnetic effects on the
phase diagrams are discussed by Chuang et a1. 44 ,45 Equation (6.44) is based
on the experimental results from 823 K (0.0010 wt% C) to near the eutectoid
temperature of 1013 K (0.0206 wt% C) and the extended solubility calcula-
tions based on the activity of carbon in the 'Y-phase as will be seen later.
The a-Fe and graphite phase boundary is satisfactorily represented by
equation (6.44) from about 800 to 1013 K. At the eutectoid temperature of
1013 K, equation (6.44) yields Sy" = 0.000319 (0.0206 wt% C).
The a-phase (= ferrite) is in equilibrium with the 'Y-phase (= austenite)
from 1013 to 1184 K. The a I 'Y phase boundary in this temperature range
requires the activity of C(gr) in the 'Y-phase, to be discussed in the next
section. The equilibrium relations for dissolved carbon in these phases

require ad for carbon in the a-phase in terms of the Henrian standard
Gibbs energy aGo as follows:
ad = ad + RTlnya ( 6.45)
where ya is so small that the Henrian activity coefficient is unity, and further,
the linear term in ya in equation (6.25) is zero, i.e., A a = O. The superscript
a may be replaced with 8 to obtain the corresponding equation for the
8-phase which is also bcc, and stable from 1665 to 1809 K. The superscript
a in ad is not essential; it is used here for emphasis.
y-Phase
The 'Y-phase is face-centered cubic in structure, constituting the most
important area of the iron-carbon diagram. The activity of carbon, 9 'Y, has
been determined 46 - 49 by equilibration with gaseous mixtures of either carbon
monoxide plus carbon dioxide or hydrogen plus methane:
CO 2 + [C in Fe] = 2CO, KI=~ (6.46)

P a.
_ P2
P4
2H2 + [C in Fe]
II
= CH 4 , K =-- (6.47)
P g. p2
Here, Ph P2 , P4 , and P are the partial pressures of CO, CO 2, CH 4 , and H 2,

respectively, and 9 is the activity of carbon in the phase under investigation.
If the standard state is taken to be pure graphite, then K ~ and K ~I are
known from equilibria only with graphite, as listed in thermodynamic
compilations. 50 The results in the range of 1000 to 1800 K are closely
represented by
dG~ = -RT In K~ = 166,774 - 171.695T; (J/mole) (6.48)
dG~1 = -RT In K~ = -91,211 + 110.416T; (J/mole) (6.49)
The values of the activity are obtained from equation (6.46) or (6.47) by
using the experimental values of pU P2 or P4 / p2 and K~ or K~I calculated
from equation (6.48) or (6.49). The activity data are fitted with equation
(6.25) to obtain eGo - gr GO) and A 'Y, both of which are usually expressed
174 Chapter 6
as a linear function of temperature. The results obtained by evaluation of

the existing data 33 with equations (6.25) and (6.26) are represented by
yY
= 45,360 - 18.4T + (57,400 + l1.2T)yY + R T l n - -y ; (J/mole)
1-y
(6.50)
wherein
Yo- - groo = 45,360 - 18.4T (6.51)
AI' = -28,700 - 5.6T (6.52)
At the graphite saturation, yo

= groo, and the left side of equation (6.50)
is zero; then the solution of equation (6.50) for all the assigned temperatures
yields the austenite/ graphite (y-phase/ graphite) boundary. For example,
at 1013 K, the eutectoid temperature, equation (6.50) yields SyY = 0.0314
from which sx Y = SyY/(1 + SyY) = 0.03044 (0.671 wt% C), and at 1424 K,
the eutectic temperature, SyY = 0.0977 (2.06 wt% C).
ot I 'Y Boundary
The partial molar Gibbs energy "'0 of carbon in a-Fe from equation
(6.45) is equal to yo in y-Fe from equation (6.50) so that "'G- - groo +
RT In y'" is the left side of equation (6.50). The use of equation (6.44) to
eliminate gr 0° - '" 0- yields
RT In y'" = -54,390 + 15.2 T + (57,400 + 11.2 T)y l'

1.52 X 10 10 4.80 x 10 15 ) yY
+ ( 8170- 2 + 4 +RTln--y
T T 1- Y
(6.53)
where y'" is in equilibrium with y l' at each selected temperature. To check

the consistency of y'" and yY, we use yY = 0.0314 (0.671 wt% C) at 1013 K
to obtain y'" = 0.000319 (0.0206 wt% C) from equation (6.53).
Thermodynamic calculations of the a/ y phase boundary require the
equality of "'0 2 = 1'02 for iron. For this purpose, we use equation (6.28) to
write
"'0 2 = "'G~ + 3RT In(1 _ y"') (6.54)

( 6.55)
where A" is zero in accord with equation (6.45) and A Y is given by equation
(6.52). The equality of the left sides in these equations yields
YG~ - "G~ = 3RT In(1 - yet) + (28,700 + 5.6 T)(yY)2

- RT In(1 - yY) (6.56)
The left side of this equation is the standard Gibbs energy of a to 'Y
transformation for pure iron, which must include the magnetic effects. The
equations representing YG~ - "G~ are complex; therefore, we present the
values in Table 6.4 in close temperature intervals as calculated from the
lengthy equations given by Agren 35 (cf. also Orr and Chipman 43 ).
Equation (6.56) with the data in Table 6.4, and equation (6.53) can be
solved simultaneously for each temperature to obtain the equilibrium values
of yet and y Y and thus calculate the a / 'Y phase boundary. These equations
are transcendental and require a computer or a programmable calculator
for iterative computations. Equation (6.44) gives the ferrite/graphite satur-
ation, equation (6.50) with yo
= grG o gives the austenite/graphite satur-
ation, and equation (6.56), based on "0 2 = Y0 2 , gives the equilibrium
concentrations of iron in ferrite and austenite. Simultaneous solution of
these equations for T, yet, and yY yields the eutectoid temperature and the
compositions of the eutectoid phases. A simple iterative computation for
this purpose is to calculate yet and yY from equations (6.44) and (6.50),
respectively, for each temperature in 1 K intervals from 1010 to 1015 K; the
set of values for one of the temperatures satisfying equation (6.56) yields
the simultaneous solution of equations (6.44), (6.50), and (6.56). Thus, it
can be shown that yet = 0.000319 (0.0206 wt% C) and yY = 0.0314
(0.671 wt% C) at T = 1013 K also satisfy equation (6.56) whose left is
Table 6.4. Values of Standard Gibbs Energy of a ~ 'Y Phase Transition,

YG z - "G z = .1G z, in J/mole for Pure Iron at Various Temperatures, Calcu-
lated from Agren 35
T,K .:lGz,J/moie T,K .:lG z, J/moie T,K .:lGz,J/moie
800 1367 1000 338 1100 100

840 1116 1020 274 1120 71
880 884 1040 219 1140 45
920 675 1060 173 1160 23
960 491 1080 134 1180 4
980 411
176 Chapter 6
+296 J/mole as obtained by interpolation of the data in Table 6.4. Excessive

numbers of decimal places given here and later are for close checking of
the relevant equations; they are not intended to reflect the accuracy of
results because the experimental errors are about ±O.OOI wt% C for the low
ranges of carbon in iron, and up to about 0.01 wt% C for very high ranges
of carbon.
Liquid Phase
The ')'-liquid and 8-liquid phase boundaries require the activity of
carbon in the liquid phase. The liquid is considered to be close-packed in
structure for which c = 1. The equation for 10 recommended by Ohtani et
al.,33 based on selected sets of data on gas-phase and dissolved carbon
equilibria,51-53 in conformity with equation (6.25), is
I
10 = grG O + 24,000 - 16.4T + (52 + 50.6T)yl + RT In ~ (6.57)
l-y
where
AI = -26...,. 25.3T (6.58)
The equality of yo and 10 in equations (6.50) and (6.57) correlates yY and

yl and establishes the calculated austenite/liquid phase boundary:
21,360 - 2.0T + (57,400 + 11.2 T)yY + RT In ~

l-y
I
= (52 + 50.6 T)i + RT In - y- I (6.59)
l-y
The unknowns y Y and yl at each selected temperature require an additional

equation for their solution. This equation is obtained by setting Y0 2 of
equation (6.55) for ')'-Fe equal to 102 for the liquid phase. The latter is
(6.60)
wherein AI of equation (6.58) has been used. The result for the
austenite/liquid boundary is
IG~ - YG~ = (26 + 25.3 T)(yl)2 - RT In(1 - i)

- (28,700 + 5.6T)(yy)2 + RT In(1- yY) (6.61)
The left side of this equation is the standard Gibbs energy of melting for
pure y-Fe, which can be obtained from the compiled thermodynamic data42
by the following steps:
IC; = 46.02; IIC; = 41.84; 'YC; = 37.36 J/mole-K;

(at 1600 K and assumed constant within ±200 K) (6.62)
IH~ - IIH~ = 13,807 J/mole (at 1809 K)

IH~ - IIH~ = 13,807 + 4.18(T - 1809) = 6245 + 4.18T
IS~ - IIS~ = 13,80907 + f T eC; - 6C;)d In T = -23.720 + 4.181n T

18 1809
IG~ - IIG~ = 6245 + 27.900T - 4.18Tln T (6.63)
IIH~ - 'YH~ = 837 J/mole (at 1665 K); 8H~ - 'YH'2 = -6622 + 4.48T
6S~ - 'YS~ = -32.728 + 4.48 In T
6G~ - 'YG~ = -6622 + 37.208T - 4.48T In T (6.64)
Summation of equations (6.63) and (6.64) yields
IG'2 - 'YG'2 = -377 + 65.108T - 8.66T In T (6.65)
The hypothetical melting point of y-Fe is obtained by setting the left side
of this equation to zero and solving for T; the result is 1797 K (l524°C)
which will be used later.
Equation (6.65) provides the values for the left side of equation (6.61)
which can be solved simultaneously with equation (6.59) to obtain the
values of y'Y and / in the range of 1424 to 1767 K, and thus calculate the
y/( y + liquid) and (y + Iiquid)/Iiquid boundaries. For example, at 1500 K,
equation (6.65) yields 2286 J/mole, and it can be shown that y'Y = 0.0810
(1.71 wt% C) and yl = 0.1758 (3.64wt% C) satisfy both equations (6.59)
and (6.61).
A small and very likely error of ±50J/mole in IG~ - 'YG~ causes large
errors in the values of y'Y and yl calculated by using equation (6.61).
Therefore, a greater degree of accuracy is required in equation (6.65) for
pure iron to achieve a higher degree of accuracy in the results from equation
(6.61). A comparable error in (21,360 + 2.0T) of equation (6.59) causes
considerably lower errors in the calculated values of y'Y and yl.
178 Chapter 6
The eutectic temperature and compositions of coexisting phases can

now be calculated. There are two alloy phases in equilibrium with graphite
and three unknowns, i.e., T, y0r, and yl at the eutectic, requiring a simul-
taneous solution of equations (6.50) and (6.57) with "YO = groo = 10, and
equation (6.61), which was based on 102 = "Y02. A simple iterative method
for this purpose is to compute y"Y from equation (6.50) and / from equation
(6.57) at 1 K intervals from 1420 to 1430 and then substitute each set of
results in equation (6.61) to determine the temperature at which equation
(6.61) is satisfied. The result is T = 1424 K, y"Y = 0.0977 (2.06 wt% C), and
yl = 0.2093 (4.31 wt% C).
Peritectic Equilibrium
The liquid, austenite, and 8 phases are in equilibrium at the peritectic
temperature. The equalities of partial molar Gibbs energies 1l0( = "0) =
"YO = 10, and 1102 = "Y02 = 102 yield four equations to solve for T, yll, y"Y,
and /. The equation for 8-Fe is the same as that for a-Fe without the
magnetic correction terms; hence, 110 = "YO yields equation (6.53) without
the second, third, and fourth terms from the end, i.e.,
RT In yll = -54,390 + 15.2T + (57,400 + 11.2T)y"Y + RT In ~ (6.66)

l-y
This is the first equation. Further, "YO = 10 provides equation (6.59), which
is the second equation. For iron, 1102 = "Y02 yields the right side of equation
(6.56) with the superscript a replaced with 8, and the left side given by
equation (6.64), so that
6622 - 37.208T + 4.48Tln T = 3RTln(1- yll) + (28,700 + 5.6T)(y"Y)2

- RTln(1- y"Y) (6.67)
which is the third equation. The fourth equation is provided by equation

(6.61) whose left side is given by equation (6.65). It can be shown that these
equations are satisfied by T = 1767 K, yll = 0.00127 (0.082 wt% C), y"Y =
0.0079 (0.170 wt% C), and / = 0.0236 (0.505 wt% C). If the austenite +
liquid field boundaries were extended to zero carbon, they would coincide
with the hypothetical melting point of y- Fe, 1797 K, calculated earlier by
using equation (6.65).
The foregoing calculations are very sensitive to small errors in
thermodynamic properties of pure iron. Small differences between the
calculated phase diagrams of Ohtani et al. 33 and Agren 35 originate largely
in the selection of such data for pure iron (see also Schiirmann and
Schmid 54 ).
The 8/8 + austenite/ austenite boundaries were obtained by the same
procedure as that for the ferrite/ (ferrite + austenite) / austenite boundaries,
and the 8/ (8 + liquid)/Iiquid boundaries, as that for the austenite/
(austenite + liquid)/liquid boundaries by using the relevant preceding
equations.
Liquid/Graphite Boundary
The liquidus of graphite saturation, i.e., liquid/liquid + graphite phase
boundary in Fig. 6.7, is obtained by setting' G - gr 0° to zero in equation
(6.57) for each selected temperature. Thus, at 1650 K, y' = 0.2326
(4.76 wt% C), and at 1900 K, y = 0.2526 (5.15 wt% C). The compositions
of phases at the eutectic temperature were calculated earlier.
Cementite
The cementite phase, Fe3C, denoted as the 8-phase for brevity, is a
metastable phase, considered to have no deviation from stoichiometry as
the activities and temperature vary. Experimental evidence indicates that
deviations from stoichiometry exist, but all the phase diagram calculations
are based on perfect stoichiometry. The melting point of Fe3C is therefore
assumed to be congruent. The Curie point is 485 K and the structure is
orthorhombic. The solubility data for Fe3C in austenite and the CO-C0 2
equilibrium data 34.55 have been used to obtain the standard Gibbs energy
of formation of Fe 3C from pure '}'- Fe and graphite. Cementite, as a separate
phase, is in equilibrium with austenitic Fe and its carbon; therefore, from
the general relationship 0 = njGj + njGj,
(6.68)
where 0 is set equal to 80° because the 8-phase is assumed to be a pure

stoichiometric compound. Subtraction of 3 y O 2+ gr 0°from both sides yields
The left side of this equation is the standard Gibbs energy of formation of
Fe3C; the first and second sets of parentheses are given by equations (6.55)
and (6.50) respectively, wherein yY must refer to the Fe3C saturation of
austenite. The result derived by Ohtani et at. 33 is
80° - JY0 2 - groo = -8900 + 141.1 T - 18.99Tln T (6.70)

180 Chapter 6
A similar relation can be obtained for the solubility of cementite in ferrite

and the gas condensed-phase equilibria. Ancillary calorimetric data on Fe 3C
from near 0 K up to high temperatures and on Fe and C(gr) can also be
used 37 to check t::/So in !)/G = il 8 Ho - Til 8 so.
We assume that equation (6.70) has been evaluated by various methods
and proceed to outline the method of phase boundary calculation for the
Fe-Fe3C system. Substitution of equations (6.50), (6.55), and (6.70) in
equation (6.69) and rearrangement of the result yields
-54,260+ 159.5T-18.99Tln T = (57,400+ 11.2T)[y"Y -1.5(y"Y?]

+ RT In[y"Y(1 - y"Y)2] (6.71)
At the Fe-Fe3C eutectic temperature of 1417 K, this equation is satisfied

by y"Y = 0.0991 (2.09 wt% C) as shown in Fig. 6.7. The austenite-Fe3C
boundary can also be calculated from the foregoing equation.
Multiplication of equation (6.65) by 3, and subtraction from equation
(6.70) yields
(6.72)
Now, 8Go is formed by the liquid phase and the graphite so that equation
(6.68) must be rewritten as 8Go = 3'G2 + 'G. Subtraction of (3'G~ + graO)
from both sides of this last simple equation gives
(6.73)
We use the right side of equation (6.72) for the left side of equation (6.73),
and then substitute equations (6.60) and (6.57) for their respective terms
in parentheses on the right side of equation (6.73), and rearrange the result
to obtain
-31,769 - 37:824T+ 6.99Tln T = (52 + 50.6T)[y'-1.5(y')2]

+ RT In[y'O - y')2] (6.74)
This equation represents the liquidus for the cementite saturation of liquid
iron, as shown by the broken curve above 1417 K in Fig. 6.7. At 1475 K,
y' = 0.2516 (5.13 wt% C) is obtained from the preceding equation. At y' =
1/3 (x' = 0.25), which is the composition of Fe3C, this equation is satisfied
at 1524.1 K, and this is the melting point of Fe 3C. The eutectic temperature,
y"Y, and y' are calculated from equations (6.71), (6.74), and (6.61); the
results are T = 1417K,y"Y = 0.0991 (2.09 wt%), andy' = 0.2124 (4.37 wt%).
The cementite/ferrite equilibrium relationships can be obtained by a

similar procedure. For this purpose, it is necessary to add 3eG~ - "'G~) to
the left side of equation (6.70), and the numerical value of 3eG~ - "'G~)
from Table 6.4 to the right side of equation (6.70). The left side then becomes
eGo - 3"G~ - grG O which is the standard Gibbs energy of formation of
cementite from a-Fe and graphite. The O-phase is now formed by (Fe + C)
in the a-phase, Le., eGo = 3"'G2 + "'G; subtraction of 3"'G~ + grG o from
both sides of this last simple equation gives
(6.75)
The terms in the first set of parentheses are given by equation (6.54). The
remaining terms require writing equation (6.45) as "'G - grGo =
"'G- - grGo + RT In y"', wherein "'G- - grG O is given by equation (6.44). Then
the calculation of the ferrite/ cementite boundary follows the same procedure
as that for the austenite/cementite boundary. The ferrite/austenite boun-
daries remain unaltered since they do not involve either graphite or the
cementite above the eutectoid temperature. The calculation of the eutectoid
temperature and compositions with Fe3C is similar to that with graphite.
Despite enormous numbers of investigations on the Fe-C and Fe-Fe3C
systems, new and more precise thermodynamic and metallurgical data are
needed in various regions of these systems. Improved modern techniques
invite reinvestigation of discordant results recently discussed by various
investigators. 33 -38
Wagner Model for Ternary Interstitial Alloys
The Wagner model 56 deals with dilute solutions of an interstitially

dissolved component C in ideal binary solutions of A and B. We use C as
an interstitial, including carbon and Cig) as the diatomic gas. The interstitial
dissolution requires not only a favorable site-size but also the ability of the
solvent atoms to stretch out to accommodate the solute. The liquid metals
and alloys probably have a close-packed structure easily capable of stretch-
ing apart to dissolve the interstitial solutes. It is therefore not surprising
that the solubilities of interstitials are generally higher in the liquid phase
than in a solid phase of a metal or alloy. The interstitials in liquids are
assumed to occupy the octahedral site for which Z = 6. Additional qualita-
tive considerations are also possible. For example, according to Pauling,57
the radius of oxygen atom is 0.066 nm, whereas that of copper is 0.128 nm,
and Ag, 0.144 nm, so that, with a slight stretching of the lattice, oxygen
may occupy the interstitial sites. When the interstitials are dissolved, only
a partial transfer of electrons occurs because the metal or the alloy contains
182 Chapter 6
a fairly high density of conduction electrons. For a given oxygen pressure

below the formation of an oxide phase, more oxygen is dissolved in liquid
copper than in liquid silver. 58 - 61 Wagner 56 interprets that this phenomenon
as due to the greater transfer of electrons to oxygen in copper than in silver.
As copper is added in silver in increasing amounts, the extent of electron
transfer to oxygen also increases, thus increasing the solubility of oxygen.
The dissolution of a gaseous diatomic element, C 2 , such as H 2 , N 2 , or
O 2 may be written as the transfer of their monatomic species C(g), i.e.,
(6.76)
where C(g) is the monatomic gas, Vi is the vacant quasi-lattice site with;
atoms of Band Z - ; atoms of A, and C i is the interstitial atom occupying
this vacancy. The monatomic gaseous C is in equilibrium with its pre-
dominating diatomic species according to equation (6.16) for C 2 (g):
(6.77)
The change in energy for reaction (6.76) is simply E~ - E~s, where E~as is
taken as zero; hence, the energy with respect to gaseous atomic C is simply
E~. If C i at site; moves to site i + 1, with; + 1 atoms of B, Wagner writes
(6.78)
The energy change for this reaction is simply
(6.79)
Likewise, from; + 1 to ; + 2, we have /le(; + 1 ~ ; + 2). Wagner assumes

that
/lE(; + 1 ~ ; + 2) -/le(; ~ ; + 1) = h (6.80)
where h is a constant. Equation (6.80) signifies that as the interstitial atom

moves from one site to another site with one more B atom, it gains energy
as prescribed by h. Prior to this section in this book, it was assumed that
the energy per bond for each type of bond is equal, e.g., every ECA per bond
is equal to other ECA whether a C atom has one CA bond or as many CA
bonds as possible. Therefore, energy E~ of a C atom surrounded by i atoms
of Band Z - i atoms of A is
E~ = iEcB + (Z - i)ECA (6.81)

where all eCB are equal to each other, and all eCA are also equal to each
other. Recall that A and B form an ideal solution; hence, eAA, eAB, and eBB
are equal or they may be taken as zero without affecting our argument.
Substitution of equation (6.81) in equations (6.79) and (6.80) gives
[(i + 2)eCB + (Z - i - 2)eCA - (i + 1)ecB - (Z - i -1)eCA]
- [(i + l)ecB + (Z - i -l)ecA - ieCB - (Z - i)eCA] =h =0 (6.82)
Unfortunately, h = 0 does not realistically represent the activities of inter-

stitials in binary metal solvents. This means that the bond energy for each
type of bond has to vary according to the number of each type of bond
surrounding the interstitial solute, i.e., the equality of eCA bond energy in
one configuration to eCA in another configuration must be abandoned.
The Wagner equation for the energy of reaction !l.e = e~ for reaction
(6.76) must therefore be written in terms of h, after using equations (6.79)
and (6.80); thus,
(6.83)
For pure A and pure B as solvents, two boundary conditions exist for this
equation, Le., when i = 0 we have pure A and when i = Z we have pure
B, and therefore,
e~=o == ec; (C in pure A); (C in pure B) (6.84)
Chiang and Chang62 define a new variable Yi by
Yi = eci+l - i
ec (i = 0, 1, ... , Z - 1) (6.85)
where Yi is a very convenient parameter for the derivation of the Wagner

equation. Here i = Z would give e~+\ which is physically impossible;
hence, the upper limit of i is Z - 1 for i insofar as Yi is concerned. Equations
(6.83) and (6.85) yield
YHI - Yi = h (6.86)
Each successive value of this equation from i = 0 to i-I gives YI - Yo = h,

Y2 - YI = h, Y3 - Y2 = h, ... ; hence, the summation from i = 0 to i-I yields
i-I
Yi - Yo = L (YHI - yJ = ih (6.87)
i=O
184 Chapter 6
Substitution of equation (6.85) here and rearrangement of the result

yields
E~+\ - E~ = ih + Yo (6.88)
Resummation of both sides for i = 0 to i - I leads to
i-Ii~O i+l
(Ec - Ed ==
i
Bc -
i 0
Ec =
(i-\i~O ) +
•
lh
•
lYo (6.89)
The sum of ih is the sum of a simple arithmetic progression which is equal

to hi(i - 1)/2; therefore,
i i(i-1)h. 0
EC = 2 + 'Yo + Bc (6.90)
We note that for i = 0, this equation is obviously satisfied, and for i = Z,
z Z(Z -1)h
Ec = 2
+ Zyo+ EC (6.91)
Solving for Yo and substituting in equation (6.90), we obtain
(6.92)
[For the mathematically minded reader, it is to be noted that the result is

quadratic in i because the left-hand side of equation (6.83) is the differential
of two differences, and it is evident from calculus that a double differential
which is equal to a constant, leads to a quadratic function upon double
summation or double integration.] Equation (6.92) is derived by a different
procedure by Wagner, but the detailed derivation in this section is based
on the easier procedure by Chiang and Chang. 62
Equation (6.92) is equivalent to that postulated by Mathieu et ai.,63
that the bond energies vary by the numbers of different bonds surrounding
an atom. This is, at best, a postulate, but without such a postulate it is
difficult to account for the variation of activity coefficients in such alloys.
We follow a somewhat different procedure but use the essentials of
Wagner's arguments to obtain the equation for the activity coefficient of
the interstitial C as a function of the mole fraction X A of component A in
alloys of A-B. The partition function qc for C may be obtained by regarding
Interstitial Solutions 18S
the C atoms distributed in N interstitial sites as in boxes with various

energies [see equation (3.23)]. The probability of finding i atoms of A and
Z - i atoms of B at an interstitial site is D~x~xZ-\ and the total number
of such sites (or degeneracy) is N times this product where N = NA + N s .
The energy of each such site is e~, and the partition function is
qc = N
Z
L
i=O
D~X~-iX~· e-E~/kT; Z']
[D'z-- (Z-i)!i!
. (6.93)
The molar Gibbs energy of component C is given by equation (3.53) by

setting Nc equal to Avogadro's number, i.e.,
qc) Nc qc
G = -kTln ( - = -RTln-· (NckT = RT) (6.94)
c Nc Nc '
When e~ in equation (6.93) is zero, C atoms also distribute them-

selves randomly; the summation then becomes unity and then qc = N so
that
Nc
Gdrandom) = + RT In IV = RT In Xc; (e~ = 0) (6.95)
It is now necessary to express the mole fraction of C in pure A, x~ for the

general case when e~ is not zero. Equation (6.93) for pure A is obtained
by setting i = 0:
(6.96)
Equation (6.94) for the molar Gibbs energy of C in pure A, G~, is now
A)NC
G~=-kTln ( ~: =RTlnx~+Ncec (6.97)
This G~ is equal to Gdgas) of the monatomic gas in reaction (6.76) as

well as 0.5GC2 for the diatomic gas in reaction (6.77) at equilibrium. Since
186 Chapter 6
G C2 (gas) = G~2 + RT In PC2' from 0.5Gc /gas) = G~ we obtain
( 6.98)
If we take any arbitrary pressure of Cig) in equilibrium with x~ in pure

A as unity, i.e., PC2 = 1, which can be any pressure that introduces a very
limited number of C atoms in the alloy, and noting that 0.5G~2 is also a
constant for a given temperature, we obtain
[K' = exp(O.5G~/ RT)] (6.99)
where K' is a constant equal to exp(O.5G~/ RT). For C dissolved in pure

B at the same pressure of C 2 , an identical procedure gives
(6.100)
For C dissolved in the binary alloy AB, we equate G c in equation (6.94)

to 0.5GC2 = 0.5G~2 = RT In K', since PC2 = 1, and then write
z
Xc = K'qc/ N = K' I Dix~-iX~' e-E~/kT (6.101)
i=O
The reference state for a dilute solution of C in a binary alloy AB is usually

taken to be the dilute solution of C in one of the components; therefore,
we take dilute C in A as the reference state so that the activity of C, ci~, is
equal to x~, i.e.,f~ = 1 in a~ = I~x~. Combination of reaction (6.77) with
C(g) = [C in alloy] gives
0.5Cig) = [C in alloy]; (PC2 = 1) (6.102)
where Xc and Ie without superscripts refer to the alloy. This equation shows
that the activity of C over the entire range of binary composition must also
be the same for a fixed pressure of gaseous C 2 ; therefore, the activity
coefficient Ic in the alloy is defined by
or Ic =X~
-, (6.103)
Xc
Equations (6.99), (6.100), and (6.103) give
(6.104)
We substitute equation (6.92) in (6.101), and regroup the terms to obtain

Z
Xc = K' I Dixi-ix~[ei(e~-e~)/ZkT](e-e~/kT)[e(Z-i)ih/2kT] (6.105)
i~O
The term inside the first set of brackets is (f~)-i/Z from equation (6.104),
and that in the first set of parentheses is x~1 K' from equation (6.99).
Substitution of these terms in equation (6.105) with II fc = xci x~, yields
the following Wagner equation:
(6.106)
For Z = 4 this equation is
1
fc = x~ + 4X~XB(f~)-1/4. e3h/2kT + 6x~x~(f~)-1/2 . e 2h / kT
+ 4XAX~(f~)-3/4. e3h/2kT + x~(f~)-1 (6.107)
The parameter h is dependent on X A or X B in a complex way. It is empirically

determined by using a suitable experimental value of fc at such a composi-
tion that fc can be represented throughout the compositional range X A for
A-B.
If in equation (6.103) we write a~ = f~x~ for pure A, assuming that f~
f~ is not unity, then f~ would enter in equation (6.103) as f~ = f~x~/xc,
wheref~ is designated with a prime to distinguish it fromfc in the preceding
equations. The resulting equation for f~ is then
Z [ i XA
i~O D z (f~)I/Z
]Z-i[ (f~)I/Z ]i
XB
[e
(Z-i)ih/2kT
]
(6.108)
f~
This is the original Wagner equation for which we shall have no specific
application in this book, because it is much simpler and preferable to use
f~ = 1 as the reference state, taken to be an infinitely dilute solution of C
in pure A. An unusual and interesting derivation of equation (6.108) is also
given by Blander and Saboungi.64
Application of Equation (6.106)
The activity coefficient of an interstitial is conveniently determined by

using diatomic gases such as H 2 , N 2 , and investigating65 the equilibria in
reaction (6.102). For carbon, reactions (6.46) and (6.47) are investigated as
188 Chapter 6
00 0.2 OA 0.6 0.8 1.0

Ag Cu
Xc.
Figure 6.S. Variation of activity coefficient of sparingly dissolved oxygen,/o' in liquid Ag-Cu
alloys at 1373 and 1473 K. Reference state is oxygen in pure Ag for which I::
g = I. -A,
Tankins and Gokcen 60 ; - •• -, Block and Stiiwe58 ; solid curve is for hi k = 683.4 in equation
(6.106); all at 1473 K. -e, Tankins and Gokcen 60 ; 6, Fruehan and Richardson 59 : 0, Jacob
and Jeffes 6 '; solid curve is for hi k = 653.7; all at 1373 K. (Adapted from Chiang and Chang62
with permission.)
0
-2
\
\0
-
-4
~
0
c
-l
-6 ~
"'-9
-8
"<0"o~
...... ~Q)
0.0 1.0
Ag XPb
Pb
Figure 6.9. Variation of activity coefficients of sparingly dissolved oxygen,/o' in liquid Ag- Pb
alloys at 1273 K. Reference state is oxygen in pure Ag for'which I-;g = I. Data of Jacob and
Jeffes 6 '; solid curve is for hi k = 278.8 in equation (6.106). (Adapted from Chiang and Chang61
with permission.)
discussed earlier, and for oxygen and sulfur, the following reactions are
investigated:
(6.109)
(6.110)
The details of experimental techniques and an analysis of requirements for

attainment of equilibrium are presented elsewhere. 66 •67
Extensive applications of the Wagner equation to oxygen in all the
binary alloys for which data are reasonably good have been carried out
by Chiang and Chang. 62 Selected examples are shown in Figs. 6.8 and 6.9.
The results, obtained by linear regression analysis on a computer and
by using Z = 6, show that the Wagner equation is reasonably successful.
As an example, consider Fig. 6.8 for the Ag-Cu-O system studied by equi-
libration of reaction (6.109) at 1473 K. The value of fo can be obtained
by writing
(6.111)
The value of Kp for Cu is the same; therefore, for the same value of H 2/H 2 0,
f~u is given by
(1473 K)
Equation (6.106) can now be used with Z = 6 to obtainfo at XAg= Xcu = 0.5.
The result is fo(xcu = 0.5) = 0.00652, which is located on the upper curve'
in Fig. 6.8. Similar calculations for the Ag-Pb-O system are shown in
Fig. 6.9.
Limitations of Wagner Model
We present Wagner's discussions on the limitations of his equation

(6.108):
1. Equation (6.80) for the difference of differences in energy changes
is a heuristic approximation leading to e~ as a parabolic function of i as
in equation (6.92).
190 Chapter 6
2. The binary solvent alloy A-B is assumed to be random; therefore,

large deviations from ideality would make Dix~-iX~ in equation (6.93) a
poor approximation. However, other unpredictable factors may affect the
results favorably or unfavorably.
3. The effect of the second solvation shell surrounding the atoms
around the first solvation shell is ignored; this is also the case in nearly all
the statistical treatments.
4. Electron transfer from A-B in the interstitial site to the interstitial
atom C may change the interaction of A-B atoms surrounding C in the first
and second solvation shells (see Jacob and Alcock68 ).
5. If the molar volumes of components A and B are significantly
different, then the dependence of e~ on i, the number of B atoms around
C, becomes more complicated. This effect might be largely due to the change
in the distance between C and the first solvation atoms, and in addition, it
might affect the value of Z. For example, the molar volume ratio of Cu/Sn
is 7.1/16.25 for the solid elements at ambient temperatures, and the volume
effect is thus expected to be considerable in Cu-Sn alloys.
6. The effects of changes in the conduction electrons per gram atom
of an alloy, due to the changes in composition, have not been specifically
considered, but these may be at least partially attributed to h in equation
(6.92). Such effects are dominant in the solubility of hydrogen in Ag-Pd
and Cu-Ni alloys wherein the additional electrons contributed by Pd and
Ni increase the solubility of hydrogen. 69 - 71
7. The effects due to the changes in the vibrational changes of C-A,
C-B, and A-B bonds with the changes in the composition of A-B have not
been considered. We add, in summary, that the most important effect due
to the deviations from randomness is not considered. The Wagner model
is nevertheless an important step in statistical thermodynamics of interstitia Is
in solvent alloys deviating to small extents from ideality. In fact, deviations
from the ideality of the binary solvent are partly accounted for by selecting
the experimental value of h appropriate for the system.
The data for nitrogen in Co-Fe, Co-Ni, Fe-Ni, and a number of other
systems are in agreement with the Wagner interpretation. 62
Sulfur atoms are larger than carbon, nitrogen, and oxygen atoms and
slow diffusional data 72 for sulfur in solid Ag, Cu, Fe, and Ni indicate that
sulfur occupies the substitutional sites whereas rapid diffusion of H, C, N,
and 0 shows that they occupy the interstitial sites. Therefore, the value of
Z for sulfur in the liquid metals and their alloys should be 8 or 12 instead
of 6.
The foregoing treatment is not applicable to hydrogen according to
Wagner because the changes in the electronic constitution of the 'alloy
greatly affect the solubility of hydrogen 69 ,7o,73,74 in Ag-Pd, AI-Cu, Cu-Ni,
and Cu-Zn, and in Laves phases such as MgCU2, MgZn2' MgNh, and
MgZn2. This effect is due to the decrease in e~ with increasing screening
of protons (H+) by electrons in the alloys:
H(g) = H+ + e-(conduction electrons) (6.112)
Thus, when hydrogen enters in a metal having a greater density of electrons,

the proton H+ is more easily screened and a greater number of H+ are
accommodated. There are other factors involved in dissolution of hydrogen
that make the solubility in various metals and alloys difficult to explain by
simple attractive forces of the nearest neighbors.
References
1. G. Alefeld and J. Voelkl, editors, Hydrogen in Metals I, and II, Springer-Verlag, Berlin
(1978).
2. G. A. Lewis, The Palladium Hydrogen System, Academic Press, New York (1967).
3. W. M. Mueller, J. P. Blackledge, and G. G. Libowitz, Metal Hydrides, Academic Press,
New York (1968).
4. H. Frieske and E. Wicke, Ber. Bunsenges. Phys. Chern. 77, 50 (1973).
5. E. Wicke and H. Brodowsky, with H. Zuchner, in Hydrogen in Metals II, edited by G.
A1efeld and J. Voelkl, Springer-Verlag, Berlin (1978).
6. T. B. Flanagan, S. Kishimoto, and G. E. Biehl, in Chemical Metallurgy-A Tribute to Carl
Wagner, edited by N. A. Gokcen, Metall. Soc. AIME, p. 471 (1981).
7. E. Wicke and G. H. Nernst, Ber. Bunsenges. Phys. Chern. 68, 224 (1964).
8. J. F. Lynch and T. B. Flanagan, 1. Phys. Chern. 77, 2628 (1973).
9. D. M. Nace and J. G. Aston, 1. Am. Chern. Soc. 79, 3619,3623,3627 (1957); J. G. Aston,
Engelhard Ind. Tech. Bull. 7, 14 (1966).
10. J. D. Clewley, T. Curran, T. B. Flanagan, and W. A. Oates, 1. Chern. Soc. Faraday Trans.
1 69,449 (1973).
11. S. Schmidt, in Hydrogen in Metals II, edited by G. Alefeld and J. Volkl, Springer-Verlag,
Berlin (1978).
12. G. Sicking, Ber. Bunsenges. Phys. Chern. 76, 790 (1972).
13. J. R. Lacher, Proc. R. Soc. London Ser. A 161, 525 (1937).
14. M. Shamsuddin and O. J. KJeppa, 1. Chern. Phys. 71, 5154 (1979); W. A. Oates and R.
Ramanathan, in Proceedings, 2nd International Congress on Hydrogen in Metals, Paris,
1977, Paper 2All, Pergamon Press, Elmsford, New York (1978); G. Bourreau, O. J.
KJeppa, and K. C. Hong, 1. Chern. Phys. 67, 3437 (1977).
15. M. Hillert and L.-1. Staflansson, Acta Chern. Scand. 24, 3618 (1970); see also M. Hillert
and M. Jarl, Metall. Trans. AIME 6A, 553 (1975).
16. C. Wagner, Z. Phys. Chern. Abt. A 193, 386, 407 (1944).
17. See, e.g., B. Baranowski, Part II in Hydrogen in Metals, edited by G. A1efeld and J. VOlkl,
Springer-Verlag, Berlin (1978), p. 157.
18. M. J. B. Evans and D. H. Everett, 1. Less-Common Met. 49, 123 (1976).
19. T. Takeshita, W. E. Wallace, and R. S. Craig, Inorg. Chern. 13,2283 (1974).
20. U. Kobler and J. M. Welter, 1. Less-Common Met. 84, 225 (1984).
192 Chapter 6
21. U. Kobler and T. Schober, 1. Less-Common Met. 60,101 (1978); see also T. Schober and
H. Wenzl, in Hydrogen in Metals II, edited by G. Alefeld and J. Volkl, Springer-Verlag,
Berlin, p. 12 (1978).
22. T. N. Veziroglu and J. B. Taylor, editors, Hydrogen Energy Progress V: Proceedings of the
5th World Hydrogen Energy Conference, Toronto, Canada, 15-20 July 1984, Pergamon
Press, Elmsford, New York (1984).
23. J. O. Bockris, Energy: The Solar Hydrogen Alternative, Wiley, New York (1977).
24. R. Wiswall, in Hydrogen in Metals II, edited by G. Alefeld and J. Voelkl, Springer-Verlag,
Berlin, p. 201 (1978).
25. J. J. Reilly, Z. Phys. Chern. 117, 155 (1979).
26. K. C. Hoffman, J. J. Reilly, C. H. Waide, R. H. Wiswall, and W. E. Winsche, Int. J.
Hydrogen Energy 1, 133 (1976).
27. G. G. Libowitz, H. F. Hayes, and T. R. P. Gibb, J. Phys. Chern. 62,76 (1958).
28. J. H. N. van Vucht, F. A. Kuijpers, and H. C. Bruning, Philips Res. Rep. 25, 33 (1970);
H. H. van Mal, Philips Res. Rep. Suppl. 1 (1976).
29. H. H. van Mal, K. H. J. Buschow, and F. A. Kuijpers, J. Less-Common Met. 32,289 (1973).
30. J. L. Anderson, T. C. Wallace, A. L. Bowman, C. L. Radosevich, and M. L. Courtney,
Hydrogen Absorption by ABs Compounds, Los Alamos Sci. Lab., Rep. LA-5320-MS (1973).
31. F. A. Kuijpers and H. H. van Mal, J. Less-Common Met. 23, 395 (1971).
32. G. Bambakidis, editor, Metal Hydrides, Plenum Press, New York (1981).
33. H. Ohtani, M. Hasebe, and T. Nishizawa, Trans. Iron Steel Inst. Jpn. 24, 857 (1984).
34. M. Hasebe, H. Ohtani, and T. Nishizawa, Met. Trans. 16A, 913 (1985).
35. J. Agren, Metall. Trans. AIME lOA, 1847 (1979).
36. H. Harvig, Jernkontorets Ann. ISS, 157 (1971).
37. J. Chipman, Metall. Trans. AIME 3,55 (1972).
38. O. Kubaschewski, Iron Binary Phase Diagrams, Springer-Verlag, Berlin (1982).
40. M. Benz and J. F. Elliott, Trans. Metall. Soc. AIME 221,323 (1961).
41. M. Hansen and K. Anderko, Constitution of Binary Alloys, McGraw-Hili, New York
(1958); First Supplement by R. P. Elliott (1965); Second Supplement by F. A. Shunk (1969).
42. R. Hultgren, P. D. Desai, D. T. Hawkins, M. Gleiser, K. K. Kelley, and D. D. Wagman,
Selected Values of the Thermodynamic Properties of the Elements, ASM, Metals Park, Ohio
(1973).
43. R. L. Orr and J. Chipman, Trans. Metall. Soc. AIME 239,630 (1967).
44. Y.- Y. Chuang, Y. A. Chang, and R. Schmid, Acta Metall. in press.
45. Y.-Y. Chuang, R .. Schmid, and Y. A. Chang, Acta Metall. in press.
46. R. P. Smith, J. Am. Chem. Soc. 68, 1163 (1946).
47. H. Schenk, M. G. Frohberg, and E. Jaspert, Archiv. Eisenhiitenw. 36, 683 (1965).
48. K. Bungardt, H. Preisendanz, and G. Lehnert, Arch. Eisenhuettenwes. 35, 999 (1964).
49. S. Ban-ya, J. F. Elliott, and J. Chipman, Trans. Metall. Soc. AIME 245, 1199 (1969); Met.
Trans. lA, 1313 (1970).
50. L. B. Pankratz, J. M. Stuve, and N. A. Gokcen, Thermodynamic Datafor Mineral Technology,
Bureau of Mines Bulletin 677 (1984).
51. F. D. Richardson and W. E. Dennis, Trans. Faraday Soc. 49,171 (1953).
52. S. Ban-ya and Y. Matoba, Physical Chemistry of Process Metallurgy, Interscience, New
York, pp. 373-402 (1961).
53. T. Mori, K. Fujimura, H. Okajima, and A. Yamanchi, Tetsu To Hagane 54, 321 (1968).
54. E. Schiirmann and R. Schmid, Arch. Eisenhuettenwes. 50, 101 (1971).
55. E. Scheil, T. Schmidt, and J. Wiinning, Arch. Eisenhuettenwes. 32, 251 (1961).
56. C. Wagner, Acta Metall. 21, 1297 (1973).
IDterstitial SolutioDS 193
57. L. Pauling, The Nature of the Chemical Bond, Cornell University Press, Ithaca, New York
(1960).
58. U. Block and H. P. Stiiwe, Z. Metallled. 60, 709 (1969).
59. R. J. Fruehan and F. D. Richardson, Trans. Metall. Soc. A/ME 245, 1721 (1969).
60. E. S. Tankins and N. A. Gokcen, High Temp. Sci 4, 393 (1972); E. S. Tankins, Metall.
Trans. A/ME 1,2637 (1970).
61. K. T. Jacob and J. H. E. Je/les, Trans. /nst. Min. Metall. CSO, 32 (1971); see also 1. Chem.
Thermodyn. 3, 433 (1971),5,365 (1973).
62. T. Chiang and Y. A. Chang, Metall. Trans. A/ME 78,453 (1976).
63. J.-c. Mathieu, F. Durand, and E. Bonnier, 1. Chim. Phys.62, 1289, 1297 (1965); B. Brion,
J.-C. Mathieu, P. Hicter, and P. Desre, 1. Chim. Phys.66, 1238, 1745 (1970).
64. M. Blander and M.-L. Saboungi, in Chemical Metallurgy-A Tribute to Carl Wagner,
edited by N. A. Gakcen, Metall. Soc. A/ME, p. 223 (1981).
66. N. A. Gokcen, Trans. Metall. Soc. A/ME 206, 1558 (1956); 197, 191 (1953).
67. M. R. Baren and N. A. Gokcen, in Advances in Sulfide Smelting, V. 1 Basic Principles,
edited by Y. H. Sohn, D. B. George, and A. D. Zunkel, AIME, Warrendale, Pennsylvania
(1983) p. 41; see also G. Urbain, W. Burgmann, and M. G. Frohberg, C. R. Acad. Sci.
Ser. C 263(8), 595 (1966).
68. K. T. Jacob and C. B. Alcock, Acta Metall. 20, 221 (1972).
69. H. Brodowsky and E. Poeschel, Z Phys. Chem.44, 143 (1965).
70. H. Brodowsky and H. Husemann, Ber. Bunsenges. Phys. Chem. 70, 626 (1966).
71. F. G. Jones and R. D. Pehlke, Metall. Trans. A/ME 2, 2655 (1971).
72. S. J. Wang and H. J. Grabke, Z. Metalld. 61, 597 (1970).
73. W. Siegelin, K. H. Lieser, and H. Witte, Z Elektrochem. 61, 359 (1957).
74. H. Schnabl, Ber. Bunsenges. Phys. Chem. 68, 549 (1964).
7
Semiconductors
Introduction
The discovery of semiconductors and their practical applications have

played the greatest scientific and industrial role in this century. It is therefore
fitting to devote this chapter to the physical and thermodynamic properties
of semiconductors, and the simplest devices manufactured from them, with
a limited emphasis on solar cells.
The classical definition of semiconductors is that their resistivities are
about 10- 2 to 109 ohm-cm, whereas good metallic conductors have resis-
tivities on the order of 1O- 6 0hm-cm. Insulators have much greater resis-
tivities than semiconductors, i.e., roughly in excess of 10 14 ohm-cm. Resis-
tivities of semiconductors and insulators decrease, and those of metals
increase, with increasing temperature. These properties are much more
satisfactorily explained in terms of the band theory of solids.
The properties and usefulness of semiconductors were fully understood
when germanium and silicon were ultrapurified to obtain these elements
with impurity levels on the order of a few parts per billion by weight. The
background required for the physical properties of semiconductors is pres-
ented in standard texts l - 3 and monographs 4 - 9 ; it is therefore sufficient to
present a brief outline of their physical metallurgy, and then delve into
their statistical thermodynamics, and a few interesting uses.
Crystal Structure
The unit cell of a crystalline substance is the smallest volume of a solid

that contains all the crystallographic information of a macroscopic crystal
as shown in Fig. 7.1(a). The unit cell in this case is a cube, whose edge-length
is called the lattice constant. Consecutive repetition of unit cells of a solid
195
196 Chapter 7
in three dimensions generates the macroscopic crystal. The planes most

frequently encountered in semiconductor technology and indicated by Mil-
ler indices in parentheses are as follows: the (lOO)-planes are perpendicular
to the x axis as in Fig. 7.1 (b);. the (11 I)-planes intersect the x, y, and z axes
at equal distances from the origin as in Fig. 7.I(c); the (llO)-planes are
perpendicular to the x- y plane and intersect the x and y axes at equal
distances from the origin as in Fig. 7.1(d). The straight lines perpendicular
to these planes are called the crystal directions, and indicated by the same
numbers in brackets; e.g., the [lOO]-direction is perpendicular to (lOO)-plane
a b
c d
Figure 7.1. (a) Diamond lattice structure with outline of unit cell; (b) view of unit cell in [100]
direction; (c) view in [111] direction; (d) view in [110] direction. (From Green 7 with
permission.)
Semiconductors 197
shown in Fig. 7.1(b). Two other useful directions, [111] and [110], are
shown by (c) and (d), respectively. The crystal structure for Ge and Si is
called the diamond structure as indicated in Figs. 7.1(a) and 7.2(A). For
most of the useful Group III and Group V compounds, such as GaAs, the
crystal structure is quite similar, but designated as the zinc blende structure.
Figure 7.1 (b) indicates that these semiconductors are tetrahedral, i.e., each
atom has four closest neighbors and therefore the structure can be represen-
ted as shown in Fig. 7.2. Other interesting semiconductors such as CdS and
ZnS are in the wiirtzite structure, and PbS and PbTe are in the rock-salt
structure, but we shall not be concerned with the last two compounds in
this chapter since we wish to limit ourselves to the structures shown in
Figs. 7.1 and 7.2.
Band Structure
Silicon and germanium are the semiconductors to be discussed in

greater detail in this chapter. Each atom of these elements has four outer
electrons, two in the s-level and two in the p-Ievel in the gaseous state.
There can be two states (occupancy) in the s-level and six states in the
p-Ievel in each free atom. A possible band structure of the solid, formed
A B
• E leclron, eo • E leclron
eGe or Si eGo
GAs
Figure 7.2. (A) Structure of germanium and silicon; (B) structure of GaAs; CdS and ZnS are
identical with GaAs. Each atom shares four electrons with its neighbors. When an electron,
eo, jumps to the conduction band, it forms a free electron e- and leaves a hole e+ in the
valence band as shown in (A). Double lines from two neighboring atoms signify one bond of
previous chapters.
198 Chapter 7
o DISTANCE BETWEEN ATOMS

OJ------=.-=-==---=.::.:...:.....:....:..:.~:..::..
Conduction band
4 states/atom
empty at 0 K
p-Ievel
s -level
w
r~
(!)
a::
w
z Valence band
Figure 7.3. Schematic energy levels of Ge
w 4 states/atom
full at 0 K and Si as a function of distance between
atoms. Band-gap energy, E g , separates
Spacing in crystal valence and conduction bands.
by condensed atoms, is shown in Fig. 7.3. Half of the s- and p-Ievels merge
into a lower band filled with four electronic states, completely full at 0 K;
the resulting lower band is called the valence band. The remaining four
electronic states formed by the remaining four split levels merge to form
the conduction band. As the temperature increases beyond 0 K, or photons
of proper energy levels are injected into the semiconductor, electrons are
excited into the conduction band. The energy Eg required to excite an
electron into the conduction band is called the forbidden band-gap energy,
or briefly the band-gap energy, and often band-gap when a specific number
follows this term. The covalent bond energy of semiconductors such as Si
w
-----Ef------ BAND
Eg EDGES
o
1~(;;~!~f_'w*~r~m~!if~~jN1;~!~1~";
CRYSTAL DIMENSION
Figure 7.4. Simple band diagram for a semiconductor. Energy is measured from valence band
edge; hence, E g , E" and energy of e - are positive, and energy of e + is negative.
Semiconductors 199
and Ge is high; hence, their entropies of fusion are much larger than those
of metals since each atom in Fig. 7.2 contributes four electrons and shares
four electrons with four neighbors, and thus completes the octet. The
structure of most interesting semiconductors is tetrahedral, or the diamond
lattice type. Covalent compounds of Group II and Group VI (II-VI), e.g.,
CdS and ZnS, or III-V, such as GaAs and InSb, also have four electrons
per atom as in Si and Ge. For simplicity, the band structure is drawn in
block diagrams as shown in Fig. 7.4. The upper band is the conduction
band for occupancy by the electrons that carry electric current. The lower
band is for the valence electrons, and as will be seen later, the lower band
is also for the holes, which also carry electric current. Here the horizontal
axis is not important, but it is sometimes designated as the crystal dimension,
but the important point is the vertical axis which represents the energy
required to excite an electron to the conduction band.
Distribution of Electrons
The energy level of electrons at 0 K is less than the Fermi energy E r,

i.e., E_ < Ee (see Chapter 3). Therefore, all the electrons occupy the valence
band at 0 K as shown in Fig. 7.5. Some electrons gain energy in excess of
Er as the temperature increases and occupy the state corresponding to E_.
The probability 4>- = 4>_(E_) of occupation of an energy level E_ is given
by
N 1
g == 4>_(E_) = 1 + exp[(E _ _ E r)/ kT] "" exp[ -(K - Ee)/ kT] (7.1)
I
lLJ
~
z
Q
~ 0.5 . . . .
al
a::
I-
Figure 7.5. Distribution of fermions as CIl
a function of energy at 0 K and at is
T> 0 K. At E_ = Ee and T> 0,
<IdEe) = 0.5. At T = 0 K, <ldE_) = 1
for E_ < E e, and tP_(E_) = 0 for E_ > Ef
Ee· ENERGY, E
200 Chapter 7
where the equality holds for the general case, and the approximate equality
holds for the values of E_ - Er several times larger than kT so that "1" in
the denominator can be neglected. The electrons have a finite probability
of occupying the conduction band when their energy E is equal to or greater
than the band-gap energy Eg. The states for 0 < E_ < Eg are forbidden;
therefore, electrons with E_ < Eg have no allowed states and they must
return to the valence band. For pure and mildly alloyed (i.e., doped)
semiconductors, E_ - Er » kT is always satisfied, and the approximate
equality in equation (7.1) is valid; such semiconductors are called nondegen-
erate semiconductors. Conduction of electricity is due to the electrons in
the conduction band and the holes in the valence band. As the temperature
increases, the number of electrons and holes, hence the conductivity of
semiconductors increases, whereas the conductivity of metals decreases with
increasing temperature. When an electron leaves the valence band of a
semiconductor to occupy the conduction band, it leaves a positive charge
in the valence band that acts in every respect as a positively charged electron,
e +. Attempts to explain the properties of holes by any other scheme than
the positively charged particles have not been successful. The electrons and
holes are collectively called the carriers since both carry electricity. The
band-gap of a semiconductor such as Si is 1.12 eV, whereas the band-gap
of an insulator such as diamond is 7 eV. In a metal, such as Na or K, Er
is within one of the allowed bands for the valence electrons, and nearly all
the valence electrons move freely as an electron gas and conduct electricity.
Motion of Electrons and Holes
The treatment of the motion of electrons and holes is similar to that

of free particles in vacuum. The electrons in the conduction band and the
holes in the valence band move according to Newton's law; hence, the
kinetic energy E_ of an electron is a quadratic function of its momentum P_:
(7.2)
where E_ is the energy of an electron corresponding to p_, Ec is the energy

at the minimum point Pc in the conduction band, and E_(kin) is the net
kinetic energy. The minimum momentum in the conduction band is Pc, and
m_ is called the effective mass of electrons. Equation (7.2) and the corre-
sponding band-gap are illustrated in Fig. 7.6. The energy of holes is measured
Semiconductors 201
\ I
\ ,/
',.... ",,'"
>-
<.!)
lr
W
Z
W
Eg
Figure 7.6. Energy-momentum diagram
MOMENTUM
for indirect band-gap semiconductors.
For semiconductors such as silicon, a
direct band-gap may also exist as
between Py = 0 and D.
in the opposite direction with a relationship similar to equation (7.2):

2
Ey - E+ = P+ == E+(kin) (7.3)
2m+
where Ey is the energy of the valence band edge, E+ is the energy of the
hole, and m+, the effective mass of the hole. Here P is measured from the
top of the valence band which is taken to be zero. Note that the band-gap
energy Eg is given by
(7.4)
Solid curves in Fig. 7.6 are for an indirect band-gap semiconductor, e.g.,
silicon and germanium, for which Pc is not zero. In a direct band-gap
semiconductor such as GaAs, Pc = 0, i.e., the extrema of the two curves
occur at the same value of p, but again with Eg = Ec - Ey. The vertical
band-gap or direct band-gap energy for Si or Ge is about twice as high as
its indirect band-gap energy E g , but for direct-gap type semiconductors
such as GaAs, vertical band-gap is identical with the usual band-gap energy,
E g • We shall have no particular application for the vertical band-gap energy
for indirect gap semiconductors; hence, the band-gap will always refer to
the lowest value of Eg which may be direct (vertical) or indirect (nonvertical).
Actual momentum-energy diagrams for the possible values of p in different
crystal directions are complex and need not be discussed here for our
purposes. The electron and hole masses are the average masses that fit the
properties under consideration as will be seen later.
202 Chapter 7
Dopants in Semiconductors
A semiconductor in the purest possible state is called an intrinsic

semiconductor. Conduction electrons and holes in such a semiconductor
in a dark chamber are generated by thermal excitation, i.e., valence electrons
are excited into the conduction band, and holes, into the valence band by
thermal energy (with or without phonon energy as necessary). An alloying
element in very small quantities is called a dopant. Ionizing dopants increase
the conductivity drastically by yielding positive or negative carriers in
addition to the carriers originating from the excitation of valence electrons
from the semiconductor. A semiconductor containing an ionizing dopant
is called an extrinsic semiconductor.
A donor-type dopant such as phosphorus in the silicon lattice site
yields a positively charged stationary p+ ion and an electron e-, as shown
in Fig. 7.7. Phosphorus has five (s + p) electrons, and four of these are used
in bonding with the neighboring atoms, leaving one electron to move freely
in the conduction band and thus increasing the conductivity of silicon. For
a given temperature, and for light doping, the product n_n+, to be discussed
later in detail, is nearly the same as that for pure or intrinsic Si, but the
total number of carriers (n_ + n+) increases dramatically by doping with
P. For example, n_n+ "'" 1020 , or n_ = n+ "'" 1010 particles/ cm 3 for pure Si
at room temperature, but Si doped with 10 16 of P atoms/ cm3 has n_ "'" 10 16
and n+ "'" 104 , so that n_ + n+ is increased by a factor of about 500,000 upon
doping with P. The silicon doped with. Group V elements is called n-doped
or n-type silicon and P is called an n-dopant. In an n-doped semiconductor,
I n
Figure 7.7. (I) Phosphorus as ionized n·dopant; free electron e- given up by P is in the
conduction band. (II) Boron as ionized dopant; free hole e+ in the valence band is formed
upon acquisition of an electron from top of the valence band by B, which then forms B- ion.
Semiconductors 203
the electrons, e -, are the majority carriers, and the holes, e +, are the minority
carriers.
Addition of a dopant such as boron from Group III into a substitutional
lattice site has an opposite effect on the free electrons. Boron has only three
(s + p) electrons, and in bonding with the neighboring silicon atoms, a boron
atom attempts to borrow one electron from its neighbors and generate a
positive particle free to move in silicon. Thus, the borrowed electron on B
leaves a deficiency of one electron in the neighboring Si atoms and this
deficiency is the same as generation of a hole in the silicon valence band.
A p-dopant is also called an acceptor because it accepts electrons from the
semiconductor and thus creates holes in the valence band. Again, for a
given temperature and for light doping with B, the product n_n+ is nearly
the same as that for pure Si but the total number of carriers (n_ + n+)
increases dramatically by doping with B. The silicon doped with Group III
elements is called p-doped or p-type silicon and B is called a p-dopant.
Some elements may act as either p-, or n-dopant; e.g., Si as dopant in GaAs
may act amphoterically, i.e., if Si occupies the Ga lattice site in GaAs, it is
an n-dopant, whereas when it occupies the As lattice site, it is a p-dopant.
This can be paraphrased by Si ~ Si+ + e- when Si is an n-dopant in GaAs
and by Si ~ Si- + e+ when Si is a p-dopant. In both cases, the dopant
ionizes to yield an ion and a charge carrier. A Group IV element such as
carbon as a dopant may act as a neutral atom in Si. The ionization of a
dopant is a function of temperature, but beyond some range of temperature,
ionization is very nearly complete as will be discussed later. At 0 K, each
dopant captures its carrier and becomes neutral. For example, p+ captures
an electron and B- captures a hole to become neutral substitutional atoms.
The ionized dopant atoms above 0 K cannot move rapidly in a semI-
conductor to contribute significantly to electrical conductivity.
Energy of Electrons and Holes
The net kinetic energy of an electron was given by equation (7.2). The
electrons in the conduction band are free particles, and according to
quantum mechanics of free particles, they are subject to the Pauli exclusion
principle; this energy is given by
(7.5)
where h is Planck's constant and N_ is the density of quantum states for

electrons, or briefly, the density of states available for occupancy by electrons
204 Chapter 7
in numbers per cubic centimeter. This equation is obtained by replacing

the terms in the parentheses of equation (3.55) with (3 N /1T)2/3, which is
very close to N:13. A similar equation for the kinetic energy of a hole in
the valence band is
(7.6)
where N+ is the density of states for holes. The negative sign arises from
the fact that the energy of holes is measured downward from the top of the
valence band; hence, it is always negative.
Concentration of Electrons
The energy E_ of electrons in the conduction band has an additional
term which is the band-gap energy Eg; i.e., for the electrons in the conduction
band, Eg must be added to its net kinetic energy because the energy is
measured from the valence band, Ey; hence,
(7.7)
Differentiation of this equation after writing the left side as (E_ - Eg)I.S,
and rearrangement of the result yields
(7.8)
where p(E_) is the density of states per unit energy at E_ so that dN_ =
p(E_) dE_ is the number of states within a narrow energy dE_. The con-
centration of electrons n_ in the conduction band is then the integral of
cf>_dN_, Le.,
x (K - Eg)o.s exp[ -(K - E r)/ kT] dE (7.9)
Note that at energies E_ < E g , the electrons do not have enough energy to
hop in the conduction band. The integration is carried out by substituting
Semiconductors 20S
and observing that (E r - Eg) is constant for a given temperature, and further,
dE_ = d(E_ - Eg). Substitution of these relationships in equation (7.9)
yields
U fooE~Eexp (EgkT-
exp ( E - E )
.
4".
n_=J;3(2m_)\.5
r - E )
-
x (E_ - Eg)05d(E_ - Eg) (7.10)
The integrand is of the standard form [exp( -ox)]x°.5dx, whose integral in

mathematical tables is given as (11'10 )0.5/20; consequently, equation (7.9)
becomes
(7.11)
where the density of state N _ is the coefficient of the exponential term, i.e.,
(7.12)
and N_ is equal to ql V of equation (3.58).
Concentration of Holes
The Fermi distribution function for holes cfJ+ is symmetric to that for
electrons and symmetric with respect to Er in energy so that cfJ+ = cfJ+(E+)
is given by
(7.13)
Here E+ is measured from the top of the valence band; hence, the numerical
value of E+ is negative, and the approximate equality is valid when E r -
E+ » kT. The energy of a hole is the same as its kinetic energy because
holes need not overcome a barrier such as the band-gap energy to acquire
additional energy. The density of states p+ within a narrow energy difference
d(E+) is obtained by differentiating equation (7.6); the result is
(7.14)
206 Chapter 7
The density of holes, n+, is obtained by integration as in equation (7.9), i.e.,
n+ = fo
-00 p+cP+ dE+
47T
= f1 (2mS· 5
fO
-00 ( - E+)0.5 exp[(E+ - Er)/ kT] dE+
2
= h3(27Tm+kT)15exp(-ErlkT) (7.15)
where the integral is from -00 to 0 because the permissible energies of holes
are measured from the top of the valence band in the negative direction.
The product term before the exponential term is called the density of states
for holes, i.e.,
(7.16)
This expression is analogous to equation (7.12) for N_.

In an intrinsic semiconductor n+ is equal to n_; hence, the equality of
equations (7.11) and (7.15) leads to
At moderate temperatures, e.g., T = 500 K, with k = 8.617 X

10- 5 eV/particle-K, and m+/ m_ = 1.2 for GaAs so that the first term on the
left is about 0.012 eV, and since Eg for GaAs is about 1.35 eV, it is evident
that
(7.17)
For other semiconductors, the correction from the logarithmic term might
be an order of magnitude higher, but the preceding approximate equality
is still valid.
Equilibrium Constant for Charge Carriers
The charge carriers from an intrinsic semiconductor obey an equili-

brium relationship similar to the ionization of water. The reaction for the
formation of the charge carriers is
eo (in valence band) = e- + e+ (7.18)
where eo is the electron bound in the valence band that generates e- (i.e.,
electron) in the conduction band and e+ (i.e., hole) in the valence band,
Semicooductors 207
and eo acts in the same way as a 1-1 electrolyte in water. The electron eo
will always be referred to as the bound electron to avoid confusion with e -.
The band-gap of a semiconductor can be measured by the injection of
photons whose energy is gradually increased toward the band-gap of the
semiconductor. No change in the conductivity of semiconductors is detect-
able when the photon energy is less than the band-gap energy, but when
the photon energy is equal to or greater than the band-gap energy, the
conductivity of semiconductors sharply increases due to the generation of
charge carriers. The number of quanta of photons may be kept to a minimum
during the measurements so that the concentrations of e- and e+ at thermal
equilibrium are not significantly disturbed, and further, in principle, reaction
(7.18) can generate a photon or corresponding energy upon the reversal of
reaction (7.18) so that equilibrium can be assumed. The energy thus involved
is therefore diJ:ectly related to the formation of the reaction products from
one reactant eo; hence, aG = Eg. Another convincing argument that Eg
is a chemical potential or a Gibbs energy per particle is presented by
Thurmond,IO and paraphrased by the following equation:
Eg = (-
iJhVN~)
- = hv = aG- (7.19)*
iJn_ S,V;x
where N~ is the number of photon quanta and n_ is the number of electrons

generated at constant entropy (S), volume (V), and composition of e-,
denoted by x; ao is therefore a chemical potential per particle. 11 Since
each photon among N~ generates one electron, e -, it is evident that iJ N~I an_
is unity. Equation (7.19) is correct because in the limit, the photon energy
is absorbed adiabatically without changing the volume and the carrier
concentration, and hence Eg must represent a Gibbs energy per particle.
The equality of Eg to the standard Gibbs energy change, aGo, however,
can be proven rigorously by considering the equilibrium in reaction (7.18)
as will be seen later.
The usual derivation of the equilibrium constant is to equate the sum
of the molar Gibbs energies of one electron and one hole, G_ + G+, to the
molar Gibbs energy of one electron in the valence band, G( eo). We shall
return to this procedure later, but first show that the required relationship
can be obtained very simply by eliminating Er from equations (7.11) and
(7.15); the result is
n_ n+ Eg aGo
In-'-= --= --- (7.20)
N_ N+ kT kT
where aGo is the standard Gibbs energy for reaction (7.18).
*Precise measurement of band-gap energies may involve phonons, which will not be discussed
here.
208 Chapter 7
A rigorous treatment of the equilibrium for reaction leads to new

interesting ramifications. 12 This procedure requires rewriting the density of
states in terms of the partition functions as follows [see equation (3.60)]:
N_ = 2(21Tm_kTI h 2 )1.5 == q_1 V (7.21)

N+ = 2(21Tm+kTI h 2 )1.5 == q+1 V (7.22)
where q_ and q+ are the partition functions, and V is the molar volume of
the semiconductor in cubic centimeters. The concentrations of electrons
and holes may be expressed by n_1 N_ and n+1 N+ which are dimensionless
properties similar to mole fractions. 1O The particle Gibbs energies G_ and
G+ as functions of these concentrations are
n_
G- = GO- +kTln-
N_ (7.23)
(7.24)
where G~ and G~ are the standard Gibbs energies on the Henrian scale
for dilute solutions. 11 It should be noted that nl N for electrons and holes
are considerably smaller than unity for intrinsic and lightly doped semicon-
ductors so that Henry's law is obeyed, or the activity coefficients are unity.
This treatment is not concerned with heavily doped semiconductors for
which the activity coefficients differ from unity, and modified statistical
thermodynamic treatments become necessary. The standard state is the
hypothetically existing state for which nl N is unity and G = GO for _ and +
particles. It is important to stress here that in these hypothetical standard
states, both charge carriers are free particles, or boltzons, obeying classical
statistics because e- is above the Fermi level and e+ is below the valence
band edge.
It is necessary to assume here that the bound electrons participating
in reaction (7.18) are those right on the valence band edge, and their Gibbs
energy is Go = Gt + kT In Xo where Gt is a standard Gibbs energy and Xo
is a very small fraction of all valence electrons. If Xo is of the order of 10-4 ,
then a small fraction, such as 1O-5xo, ending up in the conduction band in
an intrinsic semiconductor cannot change the value of Xo significantly.
Therefore, the term kT In Xo can be added to Gt to write Go = Go for the
electrons participating in reaction (7.18). The equilibrium in reaction (7.18)
requires that
(7.25)
Semiconductors 209
Substitution of equations (7.23) and (7.24) in equation (7.25) yields
(7.26)
Equations (7.11) and (7.15) can be used to obtain the logarithmic term in
equation (7.26), eliminate E r, and thus show that
(7.27)
This equatipn confirms that AG of equation (7.19) is the same as AGo of

equation (7.26) for reaction (7.18). Consequently, equilibrium in reaction
(7.18) is necessary to prove equation (7.27).
The terms G~ and G~ in equations (7.23) and (7.24), or briefly G::',
refer to G± = G::' when n±/ N± = 1; hence, their partition functions for the
standard state are given by equations (7.21) and (7.22), so. that G~ =
- kT In( q±/ No) per particle according to equation (3.53), and
G~ - G~(at 0 K) = -kT In q± (7.28)

No
where ± is for "either + or -" but not both, No is Avogadro's number, and
G~(at 0 K) is zero as T approaches zero in - T In(q±/ No). Substitution of
appropriate values in q± yields
G~ + G~ - G~(O K) - G~(O K)
= -2kT(74.753 + 1.5In(m) + 1.5 In T + In V) (7.29)
where (m) = (m_m+)o.s is the geometric mean of the carrier masses, which
usually increases very slowly with temperature. The carrier masses used
here are called the density of state carrier masses. It will be assumed here
that the temperature dependence of (m) is negligible, in reasonable agree-
ment with the summary presented by Thurmond. lO
The electrons in the valence band move in a highly restricted volume
bound by the atomic core on one side and the electron cloud of the
neighboring atoms on the other side. The partition function of this restricted
cloud of particles is difficult to formulate but it is proposed here that
(7.30)
210 Chapter 7
where Vv is the restricted volume available for bound electrons very close
to the valence band edge, which is assumed to be a very small fraction A
of the semiconductor volume V. Further, Vv may contain other corrections
such as electron spin effects, and those due to the splitting of bands in the
semiconductor, and interactions of electrons within each band. It was shown
in Chapter 3 that the contributions of rotation of a bound electron about
an atomic core and its vibration with the core are quite negligible because
the moment of inertia of an electron and the distance between the electron
and the shielded proton are both very small. Thus, A V consists of very
thin peripheral ribbonlike volumes joined to each other in such a way that
the motion of bound electrons is very much limited and may involve
occasional exchange of positions with the neighboring peripheral volumes.
The corresponding Gibbs energy for a valence electron from equation
(7.30) is
G~ - G~(at 0 K) = -kT(74.753 + 1.5 In mo + 1.5 In T + In A V) (7.31)
Subtraction of this equation from equation (7.29) as required by the left

side of equation (7.26) and then substitution of the numerical value of mo
as the electron rest mass yields
=
(m)
kT ( 18.642 - 3 In - + In -A - 1.5 In T ) (7.32)
mo V
It is well to remember that !lGo refers to reaction (7.18) at constant pressure

and constant volume, which is generally the condition for the reactions that
take place in solids at ambient pressures.
The fraction of volume occupied by the electrons near the top of the
valence band is difficult to estimate. For silicon, it is tentatively estimated
here that A = 1.5 x 10-\ which is not unreasonable since eo is highly
constrained. Substitution of this value as well as (m)/ mo = 1.2 from the
summary of data for silicon by Thurmond lO and V = 12.05 cm 3 /mole
yields
Eg - EiO K) == !lEg = kT(6.801 - 1.5 In T)

= 5.860 x 1O- 4 T - 1.293 x 1O- 4 T In T (7.33)
where Eg is in electron volts and !leo = 1.5k-the coefficient of - Tin T -is

the change of heat capacities for reaction (7.18). The values of !lEg calculated
from this equation are within 6% of the values from a smooth curve
SemiconductoR 211
representing the experimental values ranging from 200 to 420 K for silicon.
Precise values are not yet available above 420 K.
Equation (7.33) is based on only one parameter, A, and it is quite likely
that the errors involved in (m), mo, and their temperature dependence as
well as V, Vv , and their temperature dependence, could introduce corrections
for the terms in equations (7.32) and (7.33). Therefore, it is assumed that
the coefficients in equation (7.33) are adjustable within reason if the
accuracy of band-gap measurements are indeed as accurate as correctly
stated by Thurmond 1o; consequently,
(7.34)
where A and B are adjustable parameters within the magnitudes prescribed

in equation (7.33). The success of this correlation with the best-fit curve for
the experimental data for each semiconductor will now be considered.
Correlation of Band-Gap Data

The data obtained by various investigators on the band-gap energies
of Si, Ge, GeAs, and GaP, and summarized by Thurmond, will be used to
illustrate the remarkable utility of equation (7.34). Two or more sets of
independent investigations for the same semiconductor agree so well that
a curve passing through or between the points on a plot of temperature
versus !lEg represents the data remarkably well for each semiconductor.
The set of data for GaP has slightly larger deviations above 800 K. Other
sets of data for diamond, SiC, InAs, InP, and so on are not sufficiently
accurate to test equation (7.34). The data for Si and GaP are plotted in Fig.
7.8.
The experimental data for silicon and germanium were obtained from
near 0 K to approximately 420 K and the analytical correlation presented
in this chapter is expected to be valid up to possibly 500 K. For Si, the
resulting analytical equation, based on equation (7.34), is
!lEg(Si) = Eg - 1.1700 eV = 4.546 x 1O- 4 T - 1.060 x 1O- 4 T In T;

(!lC O = 1.230k) (7.35)
This equation is represented in Fig. 7.8. Likewise for Ge, the corresponding
equation is
!lEg(Ge) = Eg - 0.7437 eV = 4.218 x 1O- 4 T - 1.205 x 1O- 4 T In T;

(!lC o = l.398k) (7.36)
212 Chapter 7
The coefficients of - TIn T as aco are presented in parentheses in equations

(7.35) and (7.36). Both equations have been obtained by using the smooth
curve based on experimental data and selecting the values of aEg at 200
and 400K.
The equations for aEg of GaAs and of GaP are as follows:
aEg(GaAs) = Eg - 1.519 eV = 3.086 x 10-4 T - 1.099 x 1O-4 T In T;

(a Co = 1.275k) (7.37)
aEg(GaP) = Eg - 2.338 eV = 6.318 x 1O-4 T - 1.537 x 10-4 T In T;

(aC O = 1.784k) (7.38)
400
0.0
0.0
-0.02
-0.1
-0.04
-0.2
-0.06
> >
~
0- -0.3 ~
w
w
<J -0.08 <J
-0.4
-0.10
-0.5
-0.12
-0.6
-0.14
0.0 1400
T/K
Figure 7.8. Variation of band·gap energy with temperature. Points are experimental data from
various sources, evaluated and presented by Thurmond 1o ; curves are from equation (7.35) for
Si, and equation (7.38) for GaP.
Semiconductors 213
Equation (7.38) is plotted in Fig. 7.8. Equations (7.37) and (7.38) have been
obtained by substituting the experimental best-fit data at 300 and 800 K
from a smooth curve. The experimental results for GaAs were obtained
from 0 to 1000 K, and for GaP, from 0 to 1273 K. Both sets of data are
perfectly represented by equations (7.37) and (7.38). Accurate representa-
tions of data are important in deriving the thermodynamic properties
obtained by differentiation of the equations for !lEg as functions of tem-
perature. However, the form of the selected equation is also very important. 12
Further justification for equation (7.34) will be dealt with in the remain-
ing sections. It is now significant to point out that below approximately
73 K, !lEg becomes zero for Si according to equation (7.35), in agreement
with experimental data, and similarly, !lEg is zero for Ge, GaAs, and GaP
below 33, 17, and 61 K, respectively. The data for Si particularly support
this conclusion as shown in Fig. 7.8. We call these points "characteristic
temperatures" for temperature independence of band-gap energies. This is
reminiscent of the disappearance of temperature dependence of a number
of other physical properties in solids. The lower temperature limit of physical
validity of equation (7.34) is therefore when the experimental values of !lEg
become independent of temperature. Nevertheless, equation (7.34) con-
tributes very little to !lEg as 0 K is approached because T appears as a
coefficient in both terms.
Derived Thermodynamic Properties
Differentiation of equation (7.34) with respect to temperature leads to
a!lGO
- -- = !lSo = B- A + B In T (7.39)
aT
where !lSO is the standard entropy change for reaction (7.18). The results
for the foregoing semiconductors are
!lSO(Si) = -3.486 x 1.060 x 10- 4 + 1.060 x 10-4 In T (7.40)
!lSO(Ge) = -3.013 x 10- 4 + 1.205 x 10- 4 In T (7.41)
!lSO( GaAs) = -1.987 X 10- 4 + 1.099 X 10- 4 In T (7.42)
(7.43)
214 Chapter 7
It is evident that as the temperature decreases below exp[(A - B)/ B], e.g.,
below 27 K for Si and below 12 K for Ge, the numerical value of t:.so
changes from positive to negative. Further, t:.so approaches -00 as T
approaches zero in accord with thermodynamic computation for free elec-
trons \3 and the fact that the standard states must be chosen as dictated by
the partition functions for boltzons in equations (7.21) and (7.22).
The values of t:.eo for the standard heat capacities are given in equations
(7.35)-(7.38), and they indicate that t:.eo is not greatly different from 1.5k
as required for the free particles or boltzons at constant volume.
The equation for the standard enthalpy change, t:.Ho, can be readily
obtained from the definitional equation t:.Go = t:.Ho - Tt:.So. The result
from equations (7.34) and (7.39) is
(7.44)
where t:.HO(O K) is the same as t:.GO(O K) or Eg(O K) since Tt:.SO is zero at

o K, and the identity sign == defines t:.t:.H.
Equation (7.34) represents the statistical thermodynamic interpretation
of the variation of Eg with temperature. This representation can be reconciled
with other interpretations of temperature dependence of E g , based on
various models and mechanistic interactions,IO,14,15 but not the derived
properties such as t:.t:.Ho, t:.So, and t:.eo. The main thesis of equation (7.34)
is its basis on sound statistical thermodynamic arguments, and its remarkable
success in representing the experimental data. Additional justifications for
equation (7.34) are given in Ref. 12.
It is feasible to estimate closely the unknown values of the band-gap
energy Eg of a semiconductor at various temperatures from a single value
of Eg at any temperature by using equation (7.33). Further, one more reliable
value of Eg at another temperature would permit the calculation of A in
equation (7.34) while retaining B = l.5k. The adjustment required for the
refinement of B necessitates at least one more well-spaced value of E g •
Ionization Equilibria
It was stated earlier that a dopant atom ionizes to yield a free or moving
carrier and a stationary charged ion. The energy levels of both the donors
and acceptors lie within the band-gap of semiconductors. The energy
required to obtain an electron is measured from the donor level to the
conduction level as a positive quantity, and this energy is a small fraction
of the band-gap, usually about 0.1 eV or less, as shown in Fig. 7.9. The
energy required to obtain a hole is measured from the valence band to the
Semiconductors 21S
Figure 7.9. Band-gap of a semiconductor

with donor or acceptor levels. Each level
may have closely spaced levels but for
simplicity they are generally shown as
single levels. Donors ionize to yield elec-
trons in the conduction band and accep-
tors take up an electron from the top of
the valence band to leave a hole in the
valence band. These levels are shallow,
i.e., much closer to band edges than
indicated here.
acceptor level as shown in Fig. 7.9 and it is usually about 0.1 eV or less.
Actually, each of these levels has closely spaced levels, but for simplicity
it is usually shown as a single level.
The ionization equilibria, as typified by the following reactions for
phosphorus and boron in silicon, are as follows:
(7.45)
(7.46)
where the neutral phosphorus atom po is partially ionized to yield p+ and

an electron in the conduction band whereas the neutral boron atom BO takes
on an electron from the valence band of silicon and thus leaves a hole (e +)
also in the valence band. The corresponding equilibrium constants are
related to the donor level Ed for P and for the acceptor level Ea for B, in
the same way as equation (7.26), i.e.,
(7.47)
(7.48)
where D;/ D~ is the ratio of the concentrations of ionized and nonionized

P, both in the same concentration units, and y_ is the product of the activity
coefficients of n_, D;, and D~. Equation (7.48) and its terms are similar
to equation (7.47). As the dopant concentration decreases, y_ or y+
approaches unity as expected. The dopant concentration is often known
216 Chapter 7
from chemical or other analyses, and the carrier concentration (e + or e-)

can be determined by electrical methods, e.g., conductivity measurements.
At sufficiently high temperatures, all dopants are ionized and at 0 K, they
are completely nonionized. The values of y± are usually less than one at
ambient temperatures for nondegenerate semiconductors, indicating that
attractive forces between the ions and the charge carriers cause negative
deviations in y±.16
Fermi Energy in Doped Semiconductors
It was shown by equation (7.17) that the Fermi level (Fermi energy)
in an intrinsic semiconductor lies in the close neighborhood of E g /2. In
n-doped semiconductors, the Fermi level varies with both doping level and
temperature according to equation (7.11) with (7.12) so that
(7.49)
Increasing the n-doping level, hence increasing n_, increases n_/ N_ toward
unity and Er increases toward E g • Further, as temperature decreases, the
right side of equation (7.49) decreases because n_/ N_ is less than unity for
reasonable degrees of doping and In( n_/ N _) varies slower than T; therefore,
decreasing temperature also increases Er toward E g •
An increase in the p-doping level has the reverse effect as shown by
using equation (7.15), i.e.,
(7.50)
As n+ increases, Er decreases, and as temperature decreases, Er tends to

zero, i.e., Er tends to coincide with the valence band.
p-n Junctions
A junction of p- and n-type semiconductors is called a p-n junction.

A silicon wafer doped with boron during the crystal growth is a p-type
semiconductor, and if a layer of phosphorus is diffused on one of its surfaces,
a homogeneous p-n junction is formed. Cadmium sulfide is an n-type
semiconductor, and if CU2S, a p-type semiconductor, is deposited on CdS,
it forms a heterogeneous junction.
Semiconductors 217
Two separate blocks of p- and n-type semiconductors are shown in

Fig. 7.1O(A). The difference between the Fermi levels of these semiconduc-
tors is called the contact potential or Helmholtz energy, which is usually 1
volt or less. The Fermi level of an n-doped semiconductor is higher than
that of a p-type semiconductor as shown by comparison of equations (7.49)
and (7.50), such as n-type Si and p-type Si; therefore, the electrons would
flow from the n-type to the p-type when these two blocks are joined together.
The electrons will continue to flow until the Fermi levels become the same
as shown in Fig. 7.10(B). As a result, the band edge of the n-side is depressed
with respect to the p-side as indicated in Fig. 7.10(B). A perfect junction
is made when the interface region is very thin, usually on the order of 1 jLm.
The migration of electrons from the n-side to the p-side leaves an
excess of positively charged ions on the n-side and creates an excess of
negatively charged ions on the p-side as shown in Fig. 7.10(B), and this
p-type -----Ef ----

A
+ f----
E n - type
+ c
DISTANCE FROM DISTANCE FROM
~J~U~N~C~T~IO~N~~O~__J~U~N~C~T~I~O~N___
Figure 7.10. (A) Fermi levels of n- and p-type semiconductors. (B) Fermi levels coincide after
junction formation; conduction and valence band are bent as shown. Near the junction,
negative ions e and positive ions EEl prevent excessive migration of carriers by establishing
the electrostatic potential shown in (C).
218 Chapter 7
condition creates a built-in electrostatic potential that prevents further

migration of electrons to the left side. The electric potential of a p- n junction
is shown in Fig. 7.10(C).
Effects of Applied Current on p-n Junctions
The net current flowing through an unbiased (bias = applied Voltage)

p-n junction is evidently zero because the Fermi energy is equal on both
sides of the junction. This can be analyzed as follows: The electron on the
p-side needs to jump from its Fermi level to the valence band by gaining
E = E; - E: as shown in Fig. 7.11, and roll to the right so that the electron
flow to the right, ](right), is
](right) = a exp[ -(E; - En/ kT]; (7.51)
where a is a constant dependent on such variables as diffusivity, and

diffusion length of electrons, but we shall not be concerned with these
details. The electron flow to the left requires energy acquisition by the
electrons on the other side, i.e., E' = E; - E"f + eVbi where Vb; is called
the built-in potential and e is the electronic charge taken to be a positive
quantity to avoid writing lei- The electron flow to the left is
](left) = a exp(-E'/kT) (7.52)
The energy E' = E is given by E' = E; - E: as indicated in Fig. 7.11;

hence, the electronic current to the left is identical with that to the right,
i.e., ](left) = ](right) and the net current is zero as expected.
The Fermi level of the n-side is depressed with respect to the p-side
when an external current is applied as shown in Fig. 7.12(A). In this case,
the p-n junction is reverse-biased and E"f is depressed with respect to E:
by e V and where e is taken to be positive and the numerical value of V is
negative.
Figure 7.11. Energy levels of electrons and built-in potential Vb; across an unbiased junction.
Semiconductors 219
-.+
/ ~e- I "'\
V<O
~r---------~~--------
z....t
__ 0
01
"--1 p - type j::1
UI
n-type
~
ZI
=>1
-:I .....
B
Figure 7.12. Energies in a reverse-biased p-n junction. Applied voltage, V, is considered to
be negative in this configuration; eV is the applied energy.
The electronic current flowing to the right, J(right), is again given by

equation (7.51):
J(right) = a exp[ -(E; - En/ kT] (7.53)
where the electrons gain enough energy over E:

to move to the right. The
electrons on the n-side need to gain (E; - E: - eV) above their Fermi
level Ef' to move to the left as indicated in Fig. 7.12(B). The electronic
current flowing to the left is therefore
J(Ieft) =a exp[ -(E; - E: - eV)/ kT] (7.54)
The net electronic current to the right side of the p-n junction is obtained
by subtracting equation (7.54) from equation (7.53); thus,
J(net, right) = -a exp[ -(E; - En/ kT][exp(eV/ kT) -1] (7.55)

220 Chapter 7
The conventional current, I, is in the opposite direction; hence, I =

-)(net, right). In general, E:
is considerably smaller than and further, E;,
it is possible to write
a exp
_(E; kT-E:) -_ exp(-EkT 10
g) =
- Irs (7.56)
where a exp(E: / kT) = 10 , and the right side of this equation is called the
reverse saturation current Irs; therefore,
(7.57)
This equation is for an ideal junction, i.e., ideal diode; however, two
adjustable parameters A and B are needed to make this equation more
general, i.e.,
(7.58)
where A and B are between 1 and 2, i.e., 1 :0;;; A :0;;; 2 and 1 :0;;; B :0;;; 2. This
equation is known as the Shockley diode equation. 17
It can be shown by using Fig. 7.13 that equation (7.55) is also valid
for a forward-biased p-n junction. In this case, the electrons on the right
+. -
/" I
e - , "'\
V>O
br---------~~--------~ 0
~
_I
P-type 1-1
ul
n - type .....-I
ZI
::::l...l
~I
A
f g::E;-Ef -eV
-t.---T
+ +
I :
Eg -Ef 1
+_____ ~ J-ev
Ef I _
Ef
B
Figure 7.13. Forward-biased p-n junction.
Semiconductors 221
side must overcome the energy from the Fermi level E r on the right to the
conduction band level of the left side, and this energy is equal to E; - E: -
eV because Er for the right side is now raised by ev'
The coefficient 10 in equation (7.58) is largely a property of electrons;
hence, it may be assigned l8 a value of 6.03 x 109 milliamp/cm 2, and for a
simple ideal diode, i.e., A = B = 1, equation (7.58) becomes
I (in mA/cm 2 ) = 6.03 x 109 e-E/kT(eeV/kT - 1);

(Irs = 6.03 x 109 e - E/ kT) (7.59)
(The assumption that A = B = 1 is used up to page 235.) For example, Irs

at 300 K for a semiconductor having E8 = 1.5 eV is
Irs = 6.03 X 109 exp( -1.5 X 1.60219 X 10- 19/1.38062 X 10- 23 X 300)
= 3.808 X 10- 16 mA/cm 2 (7.60)
where the conversion factor of 1 eV = 1.60219 x 10- 19 J, and k =

1.38062 X 10- 23 J/ K both have been used on a per particle basis. The term
DARK I-V-
v
o
ILLUMINATED I
J\ Voe
I-V"",/
Figure 7.14. Current-voltage charac-

teristics of a p- n junction. Solid
curve is for 1- Vofa dark (nonillumi-
-_/ " ./
nated) junction; dashed curve is for

1- V of an illuminated junction.
222 Chapter 7
exp( e V I kT) - 1 quickly approaches -1 for reverse bias of approximately

-0.2 to voltages often in the neighborhood of -50 volts. Thus, at V =
-0.2 volt,
exp(-0.2 x 1.60219 x 10- 19 /1.38062 x 10- 23 x 300) -1 = -0.9996
hence, the negative value of reverse saturation current of 3.808 x

10- 16 mAl cm 2 is quickly attained. For voltages in the range of -0.2 to
- 50 volts, the reverse current is the same. For a forward bias of + 1 volt,
the term exp(eVlkT) -1 = 6.315 x 10 16; and 1= 24.03 mA/cm 2 , which is
dramatically larger than I = -3.808 X 10- 16 mAl cm 2 for -1 volt of reverse
bias. As the value of Eg decreases, Irs increases and this has important
ramifications for solar cells to be discussed later. Thus, for Eg = 0.6 eV,
Irs = 0.502 mAl cm 2 which is considerably larger than that for Eg = 1.5 eV.
The I-V characteristic plot of equation (7.57) is shown by the solid curve
in Fig. 7.14. Since the reverse current is very small for Eg ~ 0.6, a p-n
junction is ideally suited for current rectification, switching, and other
remarkably useful purposes.
Solar Cells
The solar cells are p-n junction devices that convert solar energy
directly into electricity by the photovoltaic effect (PVE). This occurs when
the energy of photons is equal to or higher than the band-gap of the
semiconductor. Such photons are said to be ionizing radiation. Pure single-
crystal semiconductors are transparent to photons with energies smaller
than the band-gap energies. The ionizing radiation generates holes and
electrons in excess of those corresponding to the thermal equilibrium. The
excess electrons on the p-side in Fig. 7.10 are attracted by the positive ions
on the n-side of the junction and the excess holes generated on the n-side
are attracted by the negative ions on the p-side of the function and thus
the charges are separated at the p-n junction where a dipole layer (or
space-charge region) exists.
The charge carriers thus separated are the minority carriers, and their
separation increases the overall concentration of holes on the p-side and
electrons on the n-side. Consequently, the p-type semiconductor becomes
a source of positive electricity, and the n-type semiconductor, a source of
electrons. A shorted solar cell yields a current that is called the short-circuit
current, I se , but when a cell has no external load, it generates a voltage
called the open-circuit voltage Voe' The maximum power obtainable from
Semiconductors 223
a cell is the product of maximum power current Imp and maximum power
voltage Vrnp , which are situated at the knee of the I-V curve shown in Fig.
7.14. Note that this curve is in the fourth quadrant because the external
current flows in the same way as in the reverse-biased junction in Fig. 7.12
while the measured potential is opposite in direction. The potential
difference between the two ends of the space-charge region sets an upper
limit to the magnitude of voltage produced by photons. This difference is
determined by the difference in the Fermi levels of p- and n-semiconductors
and, hence, by the doping levels. The potential difference is high when the
doping levels are high and the junction profile is sharp in concentration
gradient. The lifetimes of the photogenerated carriers should be sufficiently
long so that they can be collected by the conductors attached to the p- and
n-sides.
A diagram of a typical silicon solar cell and a photograph of a commer-
cial cell are shown in Figs. 7.15 and 7.16. The top layer, exposed to the
solar radiation, has collector bars attached to the surface by various tech-
niques for collecting the electrons. The surface area covered by the bars is
usually of the order of 3 % of the top surface area. The bottom layer can
be made to have collector bars for tandem or stacked-up cells, but for single
cells, it is usually covered or soldered with a metallic layer. A space-qualified
silicon solar cell of 2 x 2 cm can generate about 55 mW at approximately
0.55 volt, and 100 rnA at air mass zero (AMO) radiation outside the atmos-
pheric region of the earth. Various aspects of solar cells, their manufacture
and uses are discussed well in a number of monographs,s-9 and in detail
in the IEEE Photovoltaic Specialists Conference Records. 19 The remaining
sections of this chapter are largely based on a paper by Gokcen and
Loferski. 18
p-type
Bock contoct/
Figure 7.15. Diagram of a solar cell.
224 Chapter 7
Figure 7.16. Silicon solar cell. (Courtesy of Solarex Corporation.)
Simple Equivalent Circuit
The I-V characteristics of a solar cell in the fourth quadrant are

generally represented in the first quadrant for the convenience of photovol-
taic specialists as shown in Fig. 7.17. The line amperage I, of a cell with a
resistive load R is given by
(A = elkT = 1,1604.5/ T) (7.61)
where A will be used hereafter for e/ kT for brevity. The derivation of this
equation is based on the fact that the I-V curve of a nonilluminated junction
in Fig. 7.14 is displaced downward by Ise, but otherwise the curve maintains
its shape. When I, = 0, V = Voe , and Ise is then correlated with Voe from
Semiconductors 225
Isc r------__
Imp
tI
Figure 7.17. Current-voltage characteristics of a

solar cell. Maximum power is Imp Vmp • which is
located at the knee of the curve. Imp Vmp/ Ise Voe 00
is called the fill factor. Ise here is - Ise of Fig. 7.14. v-
equation (7.61) by
Ise
e AVoe=-+1 (7.62)
Irs
The maximum power IImp is delivered to the extemalload when Rmp matches
what is known as the dynamic impedance Ro of the cell, given by a v/ aI, i.e.,
R __ aV _ 1
0- aI-IrsAeAv (7.63)
which is obtained from equation (7.61) by differentiation. The value of Rmp

is adjusted to be the same as Ro for which V in equation (7.63) becomes
the maximum power voltage Vmp , i.e.,
(7.64)
The equivalent circuit of such a cell is shown in Fig. 7.18. The maximum
SOLAR
-1
~R_D m_p~~mp
CELL
, -__ ___
Figure 7.1S. Simple equivalent circuit of a solar cell. Dynamic impedance of cell is RD. For
maximum power voltage Vmp • resistance at mp should match RD.
226 Chapter 7
power resistance obeys Ohm's law, i.e.,
Imp (mA/cm 2) = RVmp = Vmp I rs A e"'vmp (7.65)

mp
Substitution of this equation in equation (7.61) with I, = Imp yields
(7.66)
This is an important equation for our subsequent purposes. Solution of

e"'vmp from this equation and substitution in equation (7.65) yields
(7.67)
The maximum power TImp is Imp from this equation multiplied by Vmp in
volts; the result is
TI mp ( mW/ cm 2) = Imp Vmp = A V~p

(Ise + Irs) (7.68)
AVmp + 1
The ratio TImp/ (Ise VoC> is called the fill factor. It is now appropriate to
proceed to tandem or stacked solar cells.
Tandem Solar Cells
The efficiency of photovoltaic solar energy conversion can be increased

substantially by using a· tandem solar cell system l8 ,20-22 which is defined as
an arrangement of solar cells based on semiconductors having different
band-gaps with the largest band-gap cell exposed directly to the sun and
acting as a transparent window for the photons of smaller energy than its
band-gap; the next cell of lower band-gap in the sequence acts as a trans-
parent window for the succeeding cell of still lower band-gap, and so on.
The band-gaps, Eg (cell i), are therefore aranged as follows:
(7.69)
In order to simplify the notation in this chapter, the band-gap of the cell
being considered is denoted by E g , and that of the cell acting as a filter, by
E. It is evident that for a selected cell, Eg is fixed and E may assume various
values in excess of E g •
SemiCODductors 227
The reason for the efficiency increase provided by tandem solar cell
systems lies in the fact that they bring about a closer match between the
maximum energy photons hllmax absorbed in a given cell and the energy
gap Eg of the absorbing semiconductor. The excess energy hll - Eg of an
absorbed photon degenerates into heat. However, as the number of cells
increases, the photon flux available for absorption in anyone semiconductor
in the stack decreases which leads to a lowering of photovoltaic conversion
efficiency of each cell. The total photon flux, incident on the tandem cell
stack, can be increased by recourse to concentrating mirrors and lenses,
thus circumventing this last-named efficiency reduction. It becomes evident,
therefore, that the tandem cell concept is particularly compatible with solar
concentrator systems. This is potentially of great importance in view of
studies of large-scale solar energy conversion systems which show that the
combination of high levels of concentration with high-efficiency cells is one
possible way to achieve economically competitive photovoltaic solar cell
systems?O
The next section explores the theoretical limits on solar energy conver-
sion efficiencies at various temperatures, various concentration ratios, and
various numbers of cells in the tandem cell stack. Four temperatures (T =
200,300, 400, and 500 K), four concentration ratios (C = 1, 100, 500, and
1000 suns), and cell numbers up to 24 are considered. The energy gaps
of the photovoltaically active semiconductors which result in optimum
efficiencies are also presented.
Calculation Procedure
Solar Spectrum and Short-Circuit Current

The calculation for photovoltaic power generation is carried out for
the air mass zero (AMO) solar spectrum as described in Ref. 22, and
summarized here. The solar radiation is very close to that from the black
body radiation for 5750 K. Table 7.1 shows the integrated AMO photon flux
Nph(E) per cm 2 per second for energies between hll = 00 and hll = E,
obtained from
(7.70)
where nph(hll) is the photon flux (per cm 2 ) per hll. Table 7.1 also includes
the short-circuit current which would flow in a cell having an absorption
cutoff at hll = E if the collection efficiency were unity, i.e., if every photon
...
Table 7.1. Limiting Short-Circuit Current as a Function of Energy Gap Eg for Air Mass Zero (AMO) Solar Radiation a ~
Solar Solar
Isc(E) energy I,c(E) energy
E = hI! A (/Lm) 10- 14 Nph(E) (mA/cm 2) (mW/cm2) E = hI! A (/Lm) 10- 14 Nph(E) (mA/cm 2) (mW/cm2)
7.000 0.1772 0 0 0 2.000 0.6199 1223.8 19.606 51.53

5.000 0.2480 2.9 0.046 0.25 1.900 0.6526 1382.7 22.152 56.61
4.000 0.3100 31.8 0.510 2.24 1.800 0.6888 1568.6 25.134 61.87
3.900 0.3179 41.1 0.569 2.84 1.700 0.7293 1757.3 28.154 67.32
3.800 0.3263 53.4 0.856 3.59 1.600 0.7757 1966.9 31.512 72.94
3.700 0.3351 68.9 1.104 4.51 1.500 0.8273 2198.8 35.227 78.65
3.600 0.3444 85.9 1.377 5.51 1.400 0.8862 2445.0 39.171 84.43
3.500 0.3542 104.7 1.678 6.57 1.300 0.9539 2715.9 43.512 90.34
3.400 0.3647 125.2 2.006 7.71 1.200 1.0353 3020.5 48.391 96.35
3.300 0.3757 149.3 2.392 8.99 1.100 1.1289 3335.6 53.439 102.23
3.200 0.3875 174.8 2.800 10.30 1.000 1.2418 3674.9 58.875 107.89
3.100 0.3999 203.9 3.267 11.79 0.900 1.3789 4011.4 64.367 113.34
3.000 0.4133 249.2 3.993 14.00 0.800 1.5514 4413.5 70.709 118.60
2.900 0.4275 301.7 4.833 16.46 0.700 1.7724 4817.7 77.184 123.42
2.800 0.4428 358.5 5.743 19.09 0.600 2.0675 5175.4 82.915 127.14
2.700 0.4592 433.5 6.945 21.37 0.500 2.4803 5513.2 88.327 130.17
2.600 0.4769 518.6 8.309 25.99 0.400 3.0996 5854.2 93.790 132.64
2.500 0.4959 611.7 9.800 29.78 0.300 4.1333 6122.0 98.080 134.14
2.400 0.5166 711.8 11.404 33.71 0.200 6.2249 6325.6 101.343 134.95
2.300 0.5391 818.9 13.120 37.84 0.100 12.4177 6426.9 102.966 135.25
2.200 0.5636 935.5 14.987 42.08 0.000 0.0000 6476.6 103.762 135.30
2.100 0.5904 1068.3 17.116 46.66
aThe photon flux Nph(E), defined as the number of photons per em' per second having energy greater than E, is taken to be zero at E = 7 eY. Short·circuit current
l,c(E) = eNph(E). The column labeled "Solar energy" is the solar energy density delivered by photons having energies in excess of E = hll.
.1=-n
.
ooool
Semiconductors 229
having an energy hll > E contributed one pair of minority carriers to the
short-circuit current. Thus,
(7.71)
where e is the charge on the electron in coulombs. For example, Table 7.1
shows that for a cell in which the AMO sunlight is incident on a photovoltai-
cally active semiconductor with an energy gap Eg = 3.00 eV, the limiting
value of Ise is 3.993 mAl cm 2 •
If a cell based on a photovoItaicaIly active semiconductor having an
energy gap Eg has interposed between itself and the solar source a filter
which cuts off all photons with energy greater than E, where E > Eg, then
the limiting short-circuit current which this cell can produce is obtained
from the relation
(for cell with Eg) (7.72)
For example, again referring to Table 7.1, if E = 3.40 eV, and Eg = 3.00 eV,
then Ise(3.00 eV, 3.40 eV) = (3.993 - 2.006) mA/cm 2 = 1.987 mA/cm 2 • For
the required calculations, Ise(Eg, E) will be used in equations (7.66) and
(7.68) instead of I se , and Eg will be used in the relationship for Ise in
equation (7.59) to calculate Irs. The solar energy conversion efficiency TJ of
a single cell is
(7.73)
where 135.3 mWI cm 2 is the AMO solar spectrum intensity, which is the last
entry in Table 7.1, and C is the concentration ratio expressed by the number
of suns.
Calculation for an example with Eg = 1.5 at 300 K for a single cell and
C = 1 requires first Irs from equation (7.59), which was found earlier to be
Irs = 3.808 X 10- 16 mAl cm 2 • The value of Ise = 35.227 mAl cm 2 is listed in
Table 7.1 for E = Eg = 1.5 eV; therefore, Isel Irs = 35.227/3.808 X 10- 16 =
9.2508 X 10 16 ; also, A = el kT = 11,604.9/300 = 38.6829, and substitution
of these values in the logarithm of equation (7.66) yields
38.6829 Vmp + In(38.6829 Vmp + 1) = (38.0661 + 1) (7.74)
This equation can. be solved by successive approximation, but to make the

required steps shorter, we set Vmp = 1.00 for the second term on the left in
order that In(A V + 1) = 3.6809, and solve for Vmp of the first term to obtain
Vmp = 0.9147. This procedure can be repeated to obtain Vmp = 0.9170, and
one more repetition yields Vmp = 0.9169 which is close enough for computa-
tion with equation (7.68). The result from this equation yields n mp =
31.414mW/cm 2 , and then efficiency TJ
Table 7.2. Efficiency (%) of Tandem Solar Cells a ~
E.(eY)
E
(eY) 0.60 0.70 0.80 0.90 1.00 1.10 1.20 1.30 1.40 1.50 1.60 1.70 1.80 1.90 2.00 2.10 2.20 2.30 2.40 2.50 2.60 2.70 2.80 2.90 3.00 3.20 3.40 3.60 3.80 4.00 6.90
0.60 0.00
0.70 0.10 0.00
0.80 0.33 0.48 0.00
0.90 0.60 1.10 0.89 0.00
1.00 0.86 1.66 1.75 1.09 0.00
1.10 1.14 2.25 2.66 2.31 1.47 0.00
1.20 1.41 2.82 3.54 3.50 2.96 1.71 0.00
1.30 1.68 3.38 4.41 4.67 4.44 3.48 1.99 0.00
1.40 1.93 3.89 5.19 5.73 5.78 5.10 3.87 2.07 0.00
1.50 2.17 4.36 5.91 8.71 7.01 6.59 5.62 4.05 2.16 0.00
1.60 2.39 4.81 6.60 7.64 8.19 8.01 7.28 5.95 4.28 2.29 0.00
1.70 2.60 5.22 7.23 8.48 9.26 9.30 8.79 7.69 6.23 4.45 2.31 0.00
1.80 2.79 5.60 7.80 9.25 10.22 10.47 10.16 9.26 8.01 6.42 4.46 2.29 0.00
1.90 2.97 5.97 8.36 10.01 11.18 11.63 11.53 10.82 9.77 8.38 6.61 4.62 2.47 0.00
2.00 3.14 6.29 8.84 10.66 12.Dl 12.63 12.69 12.16 11.28 10.06 8.46 6.64 4.65 229 0.00
2.10 3.30 6.60 9.32 11.30 12.81 13.60 13.84 13.48 12.76 11.71 10.28 8.63 6.80 4.59 2.42 0.00
2.20 3.44 6.87 9.73 11.85 13.51 14.44 14.82 14.60 14.04 13.13 11.85 10.33 8.65 6.58 4.54 2.21 0.00
2.30 3.56 7.11 10.08 12.33 14.12 15.17 15.68 15.59 15.16 14.38 13.22 11.84 10.28 8.33 6.41 4.20 2.07 0.00
2.40 3.67 7.33 10.41 12.78 14.68 15.85 16.48 16.51 16.19 15.53 14.49 13.22 11.78 9.95 8.15 6.05 4.02 2.03 0.00
2.50 3.78 7.54 10.72 13.19 15.20 16.49 17.22 17.36 17.15 16.61 15.68 14.52 13.19 11.47 9.77 7.78 5.86 3.96 2.01 0.00
2.60 3.88 7.73 11.01 13.58 15.69 17.08 17.92 18.16 18.05 17.61 16.78 15.73 14.50 12.88 11.29 9.40 7.58 5.78 3.92 1.98 0.00
2.70 3.97 7.90 11.28 13.94 16.14 17.62 18.55 18.89 18.87 18.53 17.80 16.83 15.70 14.18 12.68 10.88 9.15 7.44 5.67 3.82 1.90 0.00
2.804.05 8.06 11.51 14.25 16.53 18.10 19.11 19.53 19.60 19.34 18.69 17.81 16.76 15.32 13.90 12.19 10.54 8.92 7.23 5.44 3.61 1.76 0.00
2.904.11 8.18 11.69 14.49 16.83 18.46 19.54 20.02 20.15 19.95 19.37 18.55 17.57 16.19 14.83 13.18 11.60 10.04 8.41 6.69 4.91 3.12 1.400.00
3.004.17 8.29 11.85 14.71 17.11 18.79 19.93 20.47 20.66 20.52 19.99 19.24 18.31 16.99 15.69 14.10 12.58 11.07 9.50 7.84 6.12 4.38 2.71 1.35 0.00
3.20 4.25 8.44 12.09 15.02 17.50 19.27 20.49 21.11 21.38 21.32 20.88 20.21 19.36 18.13 16.92 15.41 13.96 12.54 11.06 9.48 7.83 6.18 4.58 3.30 2.01 0.00
3.40 4.30 8.54 12.24 15.23 17.77 19.59 20.86 21.54 21.87 21.86 21.47 20.86 20.07 18.89 17.73 16.28 14.89 13.52 12.09 10.57 8.98 7.38 5.83 4.60 3.37 1.45 0.00
3.60 4.35 8.63 12.36 15.40 17.97 19.84 21.15 21.87 22.25 22.29 21.94 21.37 20.63 19.49 18.38 16.97 15.62 14.30 12.91 11.43 9.89 8.33 6.83 5.64 4.45 2.61 1.24 0.00
3.80 4.38 8.69 12.47 15.53 18.15 20.05 21.40 22.15 22.56 22.64 22.33 21.80 21.09 19.99 18.91 17.54 16.23 14.95 13.59 12.15 10.64 9.12 7.65 6.50 5.34 3.58 2.27 1.10 0.00
4.00 4.40 8.74 12.53 15.62 18.26 20.18 21.56 22.34 22.77 22.87 22.59 22.08 21.39 20.32 19.27 17.92 16.63 15.37 14.05 12.63 11.14 9.64 8.20 7.08 5.94 4.22 2.97 1.84 0.78 0.00 ('"l
6.90 4.44 8.81 12.63 15.76 18.43 20.39 21.80 22.61 23.08 23.22 22.97 22.50 21.85 20.81 19.79 18.48 17.23 16.01 14.71 13.33 11.88 10.42 9.01 7.92 6.82 5.17 3.99 2.93 1.94 1.23 0.00 DO
..,:r;-
aT = 300 K; A = 1. B = 1; solar concentration = C = 1. .
-.I
Semiconductors 231
T/ = (31.414/135.3)100% = 23.22%
This value is listed in Table 7.2 at the bottom of the column for Eg = 1.5.
The same value is also listed in Table 7.3 for a single cell. Table 7.2 has
been obtained by the foregoing procedure with Ise of equation (7.72)
calculated from Table 7.1. Each column after the first column gives the
efficiency of each cell for the impinging energy E on the first column. For
example, the efficiency of a cell with Eg = 1.5, having a filter of E ~ 2, is
10.06 as this filter would absorb radiation greater than 2.0 eV. A trial-and-
error method for two tandem cells on a computer would yield E = 2.0, and
Eg = 1.2 for the highest efficiency, E being the band-gap of the semiconduc-
tor which acts as a filter for the cell with Eg = 1.2, the latter receiving
radiation equal to or smaller than 2.0 eV. The efficiency for the filter cell is
19.79 from Table 7.2; moving horizontally to the left from the point Eg = 2.0
at the top of the column (efficiency 0.000), and reaching Eg = 1.2, we obtain
12.69; therefore, the total efficiency of two tandem cells is 19.79 + 12.69 =
32.48, as listed in Table 7.3. Any other combination of cells would yield a
lower efficiency.
Table 7.3 shows how sensitive the overall efficiency is to variations in
the selected band-gaps and the number of cells. The contribution to the
efficiency is dramatic up to six cells, after which it levels off as shown in
Table 7.3 and represented in Fig. 7.19.
Increasing the operating temperature of the cells decreases their efficien-
cies as shown in Fig. 7.19. The contribution to the efficiency from the smaller
Eg semiconductors decreases more rapidly with increasing temperature than
for larger Eg semiconductors, because the reverse saturation current, Irs.
Table 7.3. Variation of Optimum Efficiency with Numbers of Cells a
No. of cells 2 3 3 4 5
Eg(eV) 1.5 2.0,1.2 2.3, 1.6, 2.4,1.6, 2.5, 1.8, 2.6, 2.0, 1.6,
1.0 1.0 1.3,0.9 1.2,0.8
Efficiency 23.22 32.48 37.42 37.39 40.45 42.45
No. of cells 6 6 6 7
Eg(eV) 2.7,2.3,1.9, 2.8, 2.4, 2.0, 2.6, 2.2, 1.8, 2.9,2.5,2.1, 1.8,
1.5, 1.1, 0.8 1.6, 1.2,0.8 1.4, 1.0,0.8 1.5, 1.1,0.8
Efficiency 43.82 43.67 43.64 44.86
No. of cells 7 7 13 26
Eg(eV) 3.0,2.6,2.2, 2.8, 2.4, 2.1, 0.8 to 3.0 in 0.7 to 3.0 in
1.8, 1.4, 1.0, 1.8, 1.5, 1.2, intervals of intervals of 0.1,
0.8 0.8 0.2,3.4 3.2,3.4
Efficiency 44.70 44.67 48.13 50.09
"Band-gap range: 0.7 to 3.4; A = B = 1; 300 K; 1 sun.

232 Chapter 7
l
60
~
50
I 0-
0
/~ J0
- h~ ! ...-----0
40
I /,,/,-0
u o
J
0.
'itl30
J/I
-<>=
0.0-
I Sun
A 200 K
20
0/
¢
o
o
¢
300 K
400 K
500 K
100
4 8 12 16 20 24
No. OF CELLS
Figure 7.19. Variation of efficiency, T/ (%), with number of cells at various temperatures. Solar
intensity is one sun.
increases more rapidly for smaller values of E g • Therefore, the range of

band-gaps useful for operating at the higher temperatures is shifted to higher
Eg values. Thus, for 1 sun, the useful band-gap range is 0.7 to 3.4 eV at
300 K, 0.9 to 3.4 eV at 400 K, and 1.1 to 3.6 eV at 500 K. Figures 7.20 and
7.21 show the effects of 500-fold (C = 500 suns) and 1000-fold solar con-
centration (C = 1000), respectively, on the efficiency. Comparison with Fig.
7.19 is striking because not only the efficiency is higher, but also a corre-
spondingly smaller cell area is required for the same amount of power
generation, and further, the solar concentrators are very much cheaper than
solar cells.
Additional conclusions may be drawn from Fig. 7.22 where the effects
of temperature and numbers- of cells (as 6 and 24 cells) on the efficiency
are illustrated. For large concentrations (C ;;. 1000) and large numbers of
cells, e.g., 24 cells, the efficiency approaches 70% at 200 K, a temperature
Semiconductors 233
70
6/6
60
I ~O- 0
-u
50
!IO~O_
1:/°.___
0-
r:-
f/
40 0-
6
500 Suns
iff
200 K
30 0 300 K
~
o 400 K
0 500 K
20
0 4 8 12 16 20 24
No. OF CELLS
Figure 7.20. Variation of efficiency with number of cells at various temperatures. Solar intensity
is 500 suns.
achievable for solar cells on satellites, space laboratories, and space stations.
The efficiency of a Carnot engine operating between 5750 K and 200 K is
[(5750 - 200)/5750]100% = 96.5%, indicating that tandem solar cells are
not far from ideally achievable efficiencies. At solar concentrations greater
than 1000 suns and fewer than a dozen tandem cells, the minority carrier
concentration becomes comparable to the majority carrier concentration
and equation (7.58) and the derived relations are no longer valid.
The foregoing results are for AMO solar radiation. If the AMI or AM2
spectrum had been used (AMI being the radiation when the sun is vertically
above the horizontal cells at sea level), the results would have been similar
in the sense that high solar concentration and low cell temperature would
lead to efficiencies approaching 70%. However, the range of useful band-
gaps on the high band-gap side would have been cut off at smaller values
of Eg and the combination of Eg values required for optimized efficiency
would be different from that appropriate to the AMO spectrum, but the
234 Chapter 7
/-
7°r-"--II--TI--~I-r-I~I~~I==]I==I=~==~1
60 I-
I~o-
~ 50- !IO~___--~_
-
1000 suns
30~r
6. 200 K
-
o 300 K
o 400 K
<> 500 K
20 <> I 1 I 1 I L L J I 1 I
o 4 8 12 16 20 24
No. OF CELLS
Figure 7.21. Variation of efficiency with number of cells at various temperatures. Solar intensity
is 1000 suns.
Table 7.4. Variation of Efficiency with Number of Cells,

Temperature, and Solar Concentration, A = B = 2
No. of cells
2 3 6 12
1 sun
200K 25.3 35.4 40.6 47.6 51.4
300K 19.2 26.7 30.6 35.3 37.9
400K 14.4 19.7 22.7 25.7 27.3
100 suns
200K 30.1 42.0 48.3 56.2 60.0
300 K 25.2 35.1 40.1 46.7 49.9
400K 21.1 29.2 33.3 38.3 40.6
500 suns
200 K 32.1 42.5 51.1 59.5 63.2
300 K 27.7 38.5 44.0 51.0 54.2
400K 24.0 33.2 37.8 43.5 46.2
Semiconductors 235
400 500
~------~~------~~------~~-'70
60
,
",
"", , 50
- ,, -
,,
,
U U
Q. Q.
~50
~
" , 1000 suns 40 ~
" ~ f24 celis,
40
"
, ''(6
t ---t 100 suns
"
celis,
...!OO suns
30
"" " , """ ,
' ..... " ..... ,
30 20
200 300 400 500
TEMPERATURE. K
Figure 7.22. Variation of efficiency with temperature. For clarity, results for 500 suns are not
shown.
achievable efficiency becomes less sensitive to the Eg values as the numbers

of cells increase. This means that the variation of the solar spectrum which
occurs in the course of a day can be accommodated reasonably well by any
of a number of carefully selected energy gap combinations.
The efficiencies calculated here are limit efficiencies based on the
assumptions that there are no reflection losses, no cell internal series
resistance losses, no collection efficiency losses, and so on. Solar cells can
be designed to minimize such losses.
Fabrication of tandem cells of the type described here requires the
availability of semiconductors having continuously varying energy gaps as
those which can be achieved in alloys. Antypas and Moon23 have described
such systems composed of AIIIB v semiconductors while Loferski et al. 24
have shown how this principle can be extended to chalcopyrite semiconduc-
tors of the type AIBIIIC V1 • In both of these systems, energy gap, lattice
constant, and electron affinities can be varied independently, within limits.
236 Chapter 7
Consequently, it is possible in principle to produce solar cells having the

characteristics required by tandem cell systems from such alloys.
All the foregoing calculations are based on A = B = 1 in equation
(7.58). Table 7.4 shows calculations similar to the preceding results with
nonideal diodes having A = B = 2 in equation (7.58). Even in this case the
efficiency is 63.2% for C = 500 and T = 200 K. Solar cells could provide
as much as 30% of energy needs by the year 2020, with very little upkeep
and considerable efficiency by using a free and renewable source of energy.
Rapid progress is being made 19 to achieve this goal by various techniques.
References
1. C. A. Wert and R. M. Thomson, Physics of Solids, McGraw-Hili, New York (1970).

2. C. S. Barrett and T. B. Massalski, Structure of Metals, McGraw-Hili, New York (1980).
3. C. Kittel, Introduction to Solid State Physics, Wiley, New York (1976).
4. J. S. Blakemore, Semiconductor Statistics, Pergamon Press, New York (1962).
5. A. L. Fahrenbruch and R. H. Bube, Fundamentals of Solar Cells, Academic Press, New
York (1983); R. H. Bube, Electrons in Solids, Academic Press, New York (1981).
6. S. M. Sze, Physics of Semiconductor Devices, Wiley-Interscience, New York (1981).
7. M. A. Green, Solar Cells, Prentice-Hall, Englewood Cliffs, New Jersey (1982).
8. C. Hu and R. M. White, Solar Cells, McGraw-Hili, New York (1983).
9. S. J. Fonash, Solar Cell Device Physics, Academic Press, New York (1981).
10. C. D. Thurmond, J. Electrochem. Soc. 122, 1133 (1975); see also M. B. Panish and
H. C. Casey, Jr., J. Appl. Phys. 40, 163 (1969).
11. N. A. Gokcen, Thermodynamics, Techscience, Hawthorne, California, Chapter 6 (1975).
12. N. A. Gokcen, J. Chern. Phys.,83, 1240 (1985).
13. M. W. Chase, Jr., J. L. Curnutt, J. R. Downey, R. A. McDonald, A. N. Syverud, and E. A.
Valenzuela, J. Phys. Chern. Ref. Data 11,795 (1982).
14. M. L. Cohen and D. J. Chadi, in Handbook of Semiconductors, Volume 2, edited by
M. Balkanski, North-Holland, Amsterdam, p. 155 (1980).
15. P. B. Allen and M. Cardona, Phys. Rev. B 27,4760 (1983).
16. S. I. Chikichev, Russ. 1. Phys. Chern. (English translation of Z. Fiz. Khim.) 57,989 (1983).
17. A. K. Jonscher, Principles of Semiconductor Device Operation, Wiley, New York (1960).
18. N. A. Gokcen and J. J. Loferski, Sol. Energy Mater. 1,271 (1979).
19. IEEE Photovoltaic Specialists Conference Records (1964-1984),20 volumes.
20. J. J. Loferski, in 12th IEEE Photovoltaic Specialists Conference, 1976, p. 957 (1976).
21. N. S. Alvi, C. E. Backus, and G. W. Masden, in 12th IEEE Photovoltaic Specialists
Conference, 1976, p. 948 (1976).
22. Solar Array Design Handbook, JPL-SP43-38, Volume 1, NASA, Washington D.C. (1976).
23. G. Antypas and R. L. Moon, J. Electrochem. Soc. 120, 1579 (1973).
24. J. J. Loferski, J. Shewchun, B. Roessler, R. Beaulieu, J. Piekoszewski, M. Gorska, and
G. Chapman, in 13th IEEE Photovoltaic Specialists Conference, 1978, p. 190 (1978).
A
Engel-Brewer Theories
The electronic theories of metals explain numerous characteristics of metals

and alloys, and the existence of some of the phases at various compositions.
The Engel-Brewer (E-B) theories, or more appropriately the E-B rules,
attempt to correlate a very large number of phases existing in metals and
alloys with the electronic configurations, and to predict the existence of
phases for which experimental data are not yet available. Earlier work of
Engel l during 1939-1949 was subsequently clarified, amended, and greatly
expanded by Brewer. 2 - 11
The band theories of metals assume that the valence energy levels of
the free atoms in their ground states split into nearly continuous bands
upon condensation and crystallization. The electrons in a band of energy
levels are mobile and not localized on individual atoms. While the E-B
theory, based on the valence bond theory, does not directly contradict the
band theory, it assumes that the free atoms of an element may be first
excited to place their electrons to certain higher levels by absorption of
energy (promotion energy) and then the resulting atoms are condensed to
form a metal having an appropriate number of electrons in these excited
states and thus dictate the crystal structure. The electrons in their excited
and unpaired states are considered to playa major role in binding the atoms
in their condensed states.
The electronic structures of atomic gaseous elements in their ground
states are shown in Appendix G. (See also the first entry for each of the
selected elements in Table A.I.) The number of electrons exceeding one is
denoted by superscript numbers in this section (e.g., s, S2, p, p2, p3). The
s-orbitals are filled when there are two electrons whose spins are paired.
The p-orbitals are filled with six electrons, and the d-orbitals, with ten
electrons. The electrons in the p-orbitals begin to be paired when the number
of p-electrons exceeds three electrons, and those in the d-orbitals, when the
237
238 Appendix A
Table A.t. Electronic Configurations, Promotion Energies, and Crystal Structures
Promotion Promotion
Con- energy. Crystal Con- energy. Crystal
Element figuration keal/mole structure" Element figuration keal/mole strueture
Na s 0' bee Nb d"s 0 bee

Mg S2 0 dlsp 48
sp 63 eph Mo dSs 0 bee
AI S2p 0 Te d Ss 2 0
Sp2 83 fee d 6s 7
K s 0 bee dSsp 47 eph
Ca s2 0 Ru d 7S 0
sp 43 d 6sp 72 eph
as 51 bee Rh dBs 0
Se ds 2 0 d 6sp2 128 fee
dsp 45 eph Pd dlO 0
d 2s 33 bee d 7sp2 140 fee
Ti d 2s 2 0 Ag dlOs l 0
d 2sp 45 eph d Bsp2 133 fee
dls 19 bee Cd dlOS 2 0
V d l s2 0 dlOsp 87 eph
dlsp 47 Ba s' 0
d"s 6 bee ds 26 bee
Cr dSs 0 bee sp 35
d"sp 71 La ds 2 0
d 4sp2 165 dsp 38 eph
Mn d Ss 2 0 d 2s 8 bee
dSsp 53 Sp2 44 fee
d 6s 49 bee Hf d'S2 0
Fe d 6s 2 0 dls 40 bee
d 6sp 55 d 2sp 51 eph
d 7s 20 bee Ta d l s2 0
d SSp2 92 fee d 4s 28 bee
Co d 7S2 0 W d"S2 0
d 6 sP' 119 fee dSs 8 bee
d 7sp 67 eph Re d Ss 2 0
dBS 10 d 6s 34
Ni d Bs 2 0 dSsp 54 eph
dOs 1 Os d 6s 2 0
dBsp 74 d 7s 15
d 7sp2 120 fee d 6sp 67 eph
Cu dlOs 0 Ir d 7s 2 0
d Bsp2 120 fee d 7sp 75
Zn dlOS 2 0 d 6 sp2 162 fee
dlOsp 93 eph PI dOs 0
Sr s2 0 d 7sp2 156 fee
ds 52 bee Au dlOs 0
sp 41 dBsp' 159 fee
y ds 2 0 Hg dlos' 0
d 2s 31 bee dlOsp 106
dsp 43 eph
Zr d 2s 2 0
dls 14 bee
d 2sp 42 eph
aGround state has zero promotion energy; other values are the lowest energies for each configuration as given by Brewer.
·Stable erystal structure at ambient pressure; see page 318.
Engel-Brewer Theories 239
number of d-electrons exceeds five electrons. For example, 3d 6 structure for

iron signifies that 3 is the quantum shell and that there are six electrons in
the d-orbital, and further, one electron in excess of five is paired with one
of the five electrons, leaving four unpaired electrons for bonding upon
condensation. The paired electrons in pure elements are assumed to play
no role in bonding in solid state according to the valence bond theory. For
the first 20 elements in the periodic table (Appendix G), each orbital is
completed in sequence as an electron is added, but from elements 21 to 29
the s-orbital accepts electrons before the inner d-orbital is filled. Elements
21 through 28, for which the d-orbital accepts electrons in this manner, are
called the first series of the transition elements; they lie in the fourth series
in the periodic table and occupy Sc through Ni (IlIA-VIllA). The second
and third series of transition elements lie below the first series and their
s-orbitals also accept electrons before their d-orbitals are filled. In the
lanthanide and actinide series, the inner d- and I-orbitals are filled in the
same irregular way as the d-orbitals in the remaining transition elements.
The ground states of the monatomic gaseous elements considered in
this section are listed in Table A.l. The energy required to go from the
ground state to an excited state is called the promotion energy, and the
lowest of several values of the promotion energies corresponding to each
electronic structure are listed in Table A.I. For example, the ground state
ofTi is d 2 s 2 and the promotion energy to d 2 sp is 45 kcal/mole, and to d 3 s,
19 kcal/mole.
Basic Principles
The basic principle of the E-B theory lies in correlating the unpaired
s- and p-electrons with the crystal structures. An element crystallizes in
body-centered cubic (bcc) structure when it has one outer unpaired electron.
The alkali elements and Cr and Mo have one outermost s-electron each,
and they crystallize in bcc structure. In Cr and Mo, the ground state is dSs;
the next lowest level with an unpaired s-electron and one p-electron, d 4 sp,
requires 71 and 80 kcal/mole for promoting one d-electron to the p-orbital
for Cr and Mo, respectively. These are considered to be high promotion
energies, particularly because of no change in the total number of unpaired
d- plus s-electrons in d 5 s and in d 4 sp configurations; hence, Cr and Mo
cannot crystallize in any other structure than bcc, in agreement with observa-
tions.
For magnesium, the picture is different, because its ground state outside
the closed shell is 3s 2 , or briefly S2 (Table A.I); this state is not suitable
for bonding because S2 is a paired group of two electrons, unavailable for
240 Appendix A
bonding. The sp configuration obtained by promoting one s-electron to the

p-orbital with 63 kcal of promotion energy, Eprom, yields two electrons for
bonding. The excited atoms are then condensed to form close-packed
hexagonal (cph) crystal structure for Mg. The energy of sublimation of
crystalline Mg to monatomic gaseous atoms, AEatomizo is called the atomi-
zation energy by Brewer, despite the fact that it is the bond energy in
solid-state physics, and the sublimation from the solid state into the sp
gaseous atomic state, Esp (or E dsp when d-electrons are involved), is called
the bonding energy by Brewer/-4 not to be confused with the bond energy.
These processes are summarized as follows:
solid ~ gas (S2);

gas (S2) ~ gas (sp);
solid ~ gas (sp); Esp = AEatomiz + Eprom = bonding energy (A.1)

The solidification process in sequence is therefore S2 ~ sp ~ solid. This
process is favored by a -low value of Eprom and a high value of Esp with as
many available bonding electrons as possible. In fact, the promotion to a
reasonably high energy is justified if the promotion increases the number
of unpaired electrons available for metallic bonding in the solid; however,
the promotion of electrons from the completed noble gas shell is not
permitted because the required energy is prohibitively high.
The ground state for Al is S2p, and promotion of one of the two paired
s-electrons to the p-orbital yields three electrons for bonding with a promo-
tion energy of 83 kcal/mole for obtaining Sp2 configuration. The resulting
three electrons are then necessary for obtaining face-centered cubic (fcc) crystal
structure for aluminum.
The ground state for Si is (3s 2)(3p2), or briefly S2pZ. a state considered
to be unsuitable for bonding because S2 is a paired group of electrons. The
promotion of one s-electron to p yields Sp3 configuration with four unpaired
electrons available for binding, and this configuration gives the tetrahedral or
diamond crystal structure.
Engel postulated that all the unpaired d-electrons in transition elements
also take part in metallic bonding but they do not determine the crystal
structure; only the outer unpaired s- and p-electrons control the type of
crystal. This postulate has been criticized 12 - 14 on the basis that the d-
electrons in transition elements have strong bonding and directional proper-
ties; hence, they should also playa dominant role in determining the crystal
structure. However the d-electrons control the crystal structure indirectly
by contributing to the s- and p-shells according to the E-B postulates. 3 (See
also Altmann et al. 14a )
The foregoing considerations lead to the following E-B rules. (1) The
crystal structure depends on the total number of unpaired (s + p) electrons;
the crystal structures and the numbers of unpaired (s + p) electrons are as
follows:
• bcc: 1 electron
• cph: 2 electrons
• fcc: 3 electrons
• diamond structure: 4 electrons
We shall see later that these integral numbers of electrons per atom will be
modified to cover a range of fractional numbers for each crystal structure
encountered in some metals and in all alloys. For brevity, Brewer used I,
II, and III for bcc, cph, and fcc, respectively. (2) The bonding energy
depends on the number of unpaired do, So, and p-electrons available for
bonding. (3) The electronic structure in solid phases is close to that in the
gaseous atomic element promoted as necessary to yield the appropriate
numbers of (s + p) electrons. The electrons are promoted with as small
amounts of energy as possible for a given crystal structure with as many
bonding do, So, and p-electrons as feasible. The promotion is mainly justified
(i) on the basis of appropriate numbers of (s + p) electrons and (ii) on the
basis of maximum number of unpaired electrons obtained by reasonable
promotion energies.
The basic assumption that the excited atoms in gaseous state also exist
in condensed state cannot be proven by existing experimental techniques.
The cph crystal structure in Li and Na at low temperatures cannot be
explained by this theory because no reasonable promotion energy can
provide two (s + p) electrons for cph structure; to do so would require
promoting one electron from the closed noble gas shell to the p-orbital
beyond the outer s-orbital. Likewise, for Ca and Sr, fcc structure requiring
three (s + p) electrons cannot be reconciled with E-B rules. Despite these
and other shortcomings, the E-B concepts have achieved remarkable success
in correlating the behavior of metals and alloys, and in estimating the
compositional ranges of alloy phases as will be seen later.
Nonintegral Electronic Configurations
In its refined form, the E-B theory considers that the structures of all
the elements, and particularly the transition elements and their alloys, will
tolerate some deviation from the integral numbers of 1, 2, and 3, (s + p)
electrons for bcc, cph, and fcc structures, respectively. The postulated values
242 Appendix A
of (s + p) electrons per atom, based on observation of phase boundaries

in binary systems, are as follows:
• bcc(I): 1.0 to 1.5
• cph(II): 1.7 to 2.1
• fcc(III): 2.5 to 3.0
These ranges are in general agreement with the Hume-Rothery rules;
however, they are not rigid, and may be varied somewhat as justified by
the available data for binary and multi component systems as will be dis-
cussed later. Nonintegral ranges of numbers for electrons are an amendment
to the first E-B rule presented earlier, and essential for the alloy phases.
For the bcc structure of the elements, the fractional electronic structure is
dn-o.sspo.s for all the elements listed in Table Al with dn-1s structure,
except for Mo and W for which the usual structure is dSs, and for Nb, d 4 s
(here n is the same as the number of electrons outside the noble gas shell).
For the cph structure of the elements, only Ru is assigned d 6 .3 SpO.7 configur-
ation, and for the fcc structure, Mn is assigned d 4 . Ssp1.5, and Fe, d S.5 sp 1.5.
The bonding energies for the fractional configurations are computed from
Fig. Al as will be seen later.
Transition Elements
The unpaired d-electrons of the transition elements are generally less

bonding than the s- and p-electrons on a per electron basis. However, the
empty d-orbitals of one element may provide a sink for the excess electrons
from another alloying element and thus contribute strongly to bonding. The
d-electrons of the first transition series are less bonding than the remaining
two series. In fact, the contribution of d-electrons to bonding is so poor in
Fe, Co, and Ni that they are left largely undisturbed in the solid and thus
assumed to contribute to ferromagnetism. The promotion energies for the
first two elements of each series for s, and sp configurations are close enough
to justify qualitatively the existence of cph and bcc forms of Sc, Ti, Y, Zr,
and Hf. For V, Nb, and Ta, the promotion energies for the d 4 s configuration
are 6, 0, and 28 kcal/mole, respectively, and these values are low enough
to permit bcc structure (see Table AI). The bcc crystal structure with d 5 s
configuration for Cr and Mo does not require promotion energies, and W
requires only 8 kcal/mole. The promotion energy of 53 kcal/mole for dSsp
for Mn suggests that cph structure should exist, but it does not; however,
for d 6 s, the promotion energy is 49 kcal, and indeed bcc structure does exist
both at low and high temperatures. The observed existence of fcc Mn
requires d 4 sp2 configuration, which is difficult to reconcile l5 with the E-B
theory since the promotion energy is very high. For Tc and Re, the promotion
energies in Table A.l for d 5 sp are not too large, considering the fact that
there are seven bonding or unpaired electrons for d 5 sp instead of five for
d 6 s; hence, cph structure is favored, in agreement with observations. For
Fe, d 7 sand d 6 sp configurations have acceptable promotion energies but
cph structure corresponding to d 6 sp does not exist at ambient pressures.
For Ru, d 7 s configuration should yield bcc as in Fe, but this form of Ru
does not exist. However, the promotion energy for the d 6 sp configuration
for Ru is rather high, in fact higher than that for Fe, but the corresponding
cph structure is the observed form of Ru at all temperatures, whereas cph
structure is not observed for Fe at ordinary pressures. The bonding of extra
d- and p-electrons offset the high promotion energy of d 5 sp structure because
4d electrons have much greater bonding capability than 3d electrons as
shown in Fig. A.l.
The configuration for the cph form of Co is evidently d 7 sp since the
ground state d 7 S2 is not suitable for bonding. The promotion energy for
d 6sp2 for fcc is high and at high temperatures, solid Co is in fcc structure.
For Rh and Ir, the promotion energy for the d 6 sp2 configuration is high
but promotion to d 6 Sp2 is necessary for the fcc structure that is observed
at all temperatures. The configuration proposed by Brewer for Cu, Ag, and
Au is d 8sp2, i.e., three (s + p) electrons, for which the promotion energy is
fairly high. These metals exist only in fcc structure.
Correlation of Bonding Energies
The bonding energies of s- and p-electrons (in kilocalories per mole

of electrons) versus the elements, referring to the vertical and upper horizon-
tal axes, as devised by Brewer, are represented by the upper curves in Fig.
A.1. The ends of the upper curve in Fig. A.l(A) refer to Ca and Zn, which
have no bonding d-electrons; therefore, the bonding energy from equation
(A.l) divided by two (s + p) electrons yields these end points. Thus, for
Ca, Esp is the sum of Eprom = 43 and Eatomiz = 42, i.e., 85, and this value
divided by two (s + p) electrons is equal to 42.5 kcal/ mole electrons = Esp.
The corresponding value for Zn is 62 kcal. The initial sharp curvature on
the left can be explained on the basis that the bonding energies of s- and
p-electrons vary approximately with the inverse ratios of covalent atomic
radii in the periodic table. Thus, for Ca and Sc, the covalent radii are
1.74 and 1.44 A, respectively, and Esp = 42.5 (for Ca) x 1.74/1.44 =
51 kcal/mole electrons for Sc. The remaining portion of the curve up to Zn
is a smoothly joined and very nearly straight line ending at Zn. The influence
of the nuclear charge on (s + p) electrons is complicated by the screening
effect of the inner electrons; however, any error in the upper curve is
244 Appendix A
Co V Cr Mn Fe Co Ni Cu Zn
60
0
!l0
e
c
2 40
A ....
u
~
30
Q.
Ci 20
u
.a
10
0
0 234!143 o
No. of unpaired electrons per atom
Figure A.I. Valence· state bonding enthalpy (- energy) in kilocalories per mole of electrons.
Upper curves are for E,p, bonding enthalpies of s· and p·electrons, versus the elements on
upper horizontal axis. Lower curves are bonding enthalpies versus numbers of unpaired
d·electrons on lower horizontal axis. (From Brewer4 with permission.)
compensated by the energy assigned to the d-electrons by the lower curves;

therefore, the exact shape of the upper curve is not considered to be greatly
important for the succeeding computations. Also shown in Fig. A.l are the
bonding energies of the d-electrons, Ed, versus the number of unpaired
d-electrons as indicated by the lower curves, which refer to the lower
horizontal axis. Each point was obtained by calculating the value of total
bonding energy E dsp due to all unpaired do, SO, and p-electrons, then
subtracting Esp, obtained by multiplying the number of (s + p) electrons
with the value from the upper curve, and then dividing Ebonding - Esp by
the total number of unpaired d-electrons. The proximity of the curves for
bcc and cph indicates that an approximate curve that could have been
drawn slightly above the curve for cph might well have served for the fcc
structure. Recomputation of the bonding energies from these figures may
often show which phases have high enough sublimation energy to be the
most stable. For example, the ground state for Nb is d 4 s and Nb crystallizes
as bcc. From Fig. A.l(B), Esp = 52 x 1 (one s-electron) and (total Ed) =
36 x 4 (four d-electrons), so that the bonding energy Ebonding = 196, and
since the promotion energy is zero this is also equal to the sublimation or
atomization energy E subl ' The cph form of Nb has never been observed,
and for cph, the required configuration is d 3 sp, for which the promotion
energy is 48 kcaljmole. The values of the energies from Fig. A.l(B) are
Esp = 52 x 2 and (total Ed) = 35 x 3; hence, Ebonding = 209, and subtraction
Engel-Brewer Theories 24S
Sr Y Zr Nb Mo Tc Ru Rh Pd 49 Cd
60
I I
!5 s,~
"0
E
c
.....
::!
U 40
B ~
Q.
......
"0
30
20
o 2 3 4 4 3 2 o
No. of un~arred 4d electrons ~er atom
8e Le Ht Te W Re as Ir Pt Au Hg
70
"0
.,:LV-
5' 60
c: 6s,p
2
C .
~ 50
i 40
'0
u
"'" 30 oLe
o bee
No. of unpoired electrons per etom
Figure A.I (continued)
of 48, the promotion energy, yields Eatomiz = 161 kcaljmole, indicating that
the cph form is about 35 kcalless stable than the bcc form. Ifwe use d 3 .3 SpO.7
configuration, the lower limit of 1.7 (s + p) electrons for cph, the result
would be Ebonding = 52 x 1.7 + 35 x 3.3 = 204, and Eatomiz = 204 - 48 x
0.7 = 170, where 0.7 originates from the fact that 0.7 mole of d 3 sp and
0.3 mole of d 4 s make up d 3 .3 Sp O.7. Again, even with the use of the lower
limit for 1.7 (s + p) electrons, the result does not make the cph form as
246 Appendix A
stable as the bcc form. The foregoing figures indicate that as the number
of unpaired d-electrons decreases from five in each transition series, the
bonding energy per d-electron increases.
Electron Concentration Ranges in Alloys
Body-Centered Cubic Structure (bee, or I)
The electron concentration ranges in alloys can best be understood by

referring to the various binary phase diagrams in Appendix E whenever
possible. The binary diagrams hold the key to understanding the electronic
structures of alloy phases. The upper limit for the electron concentration
is assumed to be 1.5 (s + p) electrons per gram atom of alloy for the bcc
phases. For Cu-Zn alloys, the observed upper limit is 1.58. This is based
on the assumption that copper contributes one s-electron instead of three
(s + p) electrons as shown in Table A.l. The explanation of one s-electron
for Cu in this case by Engel and Brewer is that in bcc brass, which is
approximately equiatomic in composition, copper has so many zinc neigh-
bors that three unpaired (s + p) electrons cannot exist and dlOs structure
becomes energetically preferable. The observed upper limit for the bcc
structure in the Li-Mg system is 1.75 (s + p) electrons.
Group VA and VIA elements crystallize in bcc, and according to E-B
rules they should be miscible in their binary solutions, in accord with
observations. These elements may also dissolve elements other than those
in the preceding groups up to 1.5 (s + p) electron concentration. The
maximum solid solubilities of a number of elements in Cr, Mo, and W as
solvents for the bcc phases are given in Table A.2, showing good agreement
with the E-B rules. As an example, we compute the solubility of Rh in Mo,
in atom fraction x of Rh (d 6 sp2). The electronic contribution from Mo is
6 (d + s) electrons so that 6(1 - x) is the total contribution from Mo.
Rhodium has 9 (d + s + p) electrons per atom; hence, the contribution
from Rh is 9x, and the sum must correspond to dSspo.s, or 6.5 electrons
including 1.5 (s + p) electrons; thus, 6.5 = 6(1 - x) + 9x, and x = 0.167 or
16.7 at.% Rh in Mo. If the integral approximation of dSs were used for the
alloy, instead of dSspo.s, x would be zero. This example shows that although
d-electrons do not control the structure directly by remaining as d-electrons,
they do so indirectly by contributing electrons to p by unpairing the d-
electrons as from d 8 s to d 6 sp2 in Rh. For Ru(d 6 sp) as a solute, similar
calculations from 6.5 = 6( 1 - x) + 8x yield x = 0.25 as shown in Table A.2.
The values listed in Table A.2 refer to the temperatures at which the
solubility is a maximum or usually at the eutectic temperature. The disagree-
Table A.2. Maximum bcc Solid Solubility Limits of

Selected Elements in Cr, Mo, and W as Solvents"
Calculated
values in Experimental values
Cr, Mo,
Solute and W in Cr in Mo in W
Tc 50 50
Re 50 48.5 42 37
Ru 25 34 30 23
Os 25 31 19 18.5
Rh 16.7 -15 -20 -6
Ir 16.7 -12 16 -10
Pd 12.5 -5 6.5 -5
Pt 12.5 -10 15 4
"Values are concentrations in at. % of solute element in the first column.
ment between the calculated and experimental values becomes larger as the
difference between the group numbers of the solvents and solutes increases.
The transition elements with fewer than five d-electrons act as a sink
for the d-electrons of the elements having more than five d-electrons. In
addition, the (s + p) electron concentration may be lower than 1.5 so that
the structure for bcc alloys of Hf( d 3 s )-Ru( d 7 s) is d 4 .S s, for Hf as the solvent,
corresponding to 62.5 at.% Hf as calculated from 5.5 = 4x + 8(1 - x). For
Ru as the solvent, the structure is dS'ss, and 6.5 = 4x + 8(1 - x) yields
37.5 at. % Hf. Calculation of the appropriate numbers of (s + p) electrons
is discussed in detail in Ref. 4. If for a known binary system the electron
concentration and configuration are fitted, then for a nearby system the
same concentration and configuration can be used to estimate the unknown
solubility. In essence, the known maximum solubilities may dictate the
electronic structure, and the unknown maximum solubilities may be based
on the rules as adjusted according to the rules derived from similar binary
systems if they have been established by experiment.
The Laves phases, for which the ideal compositional range is AB2 to
AB s, generally having cubic, hexagonal, and complex structures, are formed
not only on the basis of electron to atom ratio, but also on the basis of
atomic radius ratio of A to B as discussed elsewhere in detail. 20,23,2S No
definite (s + p) electron concentration has been assigned to these phases
by Brewer. Some of these, such as LaNi s, are useful and interesting media
for hydrogen storage as was discussed in Chapter 6.
248 Appendix A
Close-Packed Hexagonal Structure (cpb, or II)

The electron concentration for cph structure is generally limited to the
range of 1.7 to 2.1 (s + p) electrons per gram atom of alloy (cf. Brewer3
and Hume-Rothery16). The lower limit is for the solubilities of solute
elements with one s-electron per atom in solvent elements with two or more
(s + p) electrons, whereas the higher limit is for those solutes having three
or more (s + p) electrons per atom in cph solvent elements. For example,
Mo and W have d 5 s structure, and as solutes, they should dissolve in solvent
metals Tc and Re, having d 5sp structure to the extent of d 5spO.7, or 30 at.%
as confirmed from 6.7 = 6x + 7(1 - x); the experimental values are
24 at. % Mo in Tc, and 20 at. % W in Re. To improve the results, Brewer
suggests d 5 . 1 SpO.7 corresponding to a maximum of 20 at.% Mo for the cph
structure. For Rh(d 6 sp2) as a solvent, Mo should dissolve until the solution
assumes d 6 sp1.1 corresponding to 0.30 mole fraction of Mo, but only
0.18 mole fraction of Mo is dissolved in Rh when cph structure appears.
For d 6 spO.7, 0.65 mole fraction Mo is dissolved in Rh; hence, Rh can no
longer be considered as the solvent. In this case, Mo is the solvent, but
since Mo is d 5 s, this configuration should be raised to d 5 spO.7, and assuming
that Rh( d 6 Sp2) contributes one d and three (s + p) electrons, then cph
phase should appear at x = 0.25 mole fraction of Rh, computed from
6.7 = 9x + 6(1 - x), but the experimental value is 0.43 Rh. Here, it may be
assumed that Rh cannot contribute three more electrons than Mo. On the
high electron side, and therefore Rh as solvent, Hume-Rothery's configu-
ration of d 6 .36 sp 1.1 corresponds to the experimental value of 18 at. % Mo.
Again, the degree to which the rules are adjusted depends on the systems,
and such rules may often appear to explain the solubilities and the electronic
structures rather than to predict the solubilities.
For the mutual solubilities of the elements capable of contributing
large numbers of electrons, clearly the values toward the upper limit of 2.1
must be used. Thus, for Ir(d 6 sp2) as a solute in Os(d 6 sp), with x as the
mole fraction of Ir, d 6 sp1.l structure requires 9x + 8(1 - x) = 2.1 + 6, which
yields x = 0.10, whereas the actual value of x is 0.38. Note that here 1.7
instead of 2.1 would give a negative value for x!
Face-Centered Cubic Structure (fcc, or III)

The range of electron concentration for the fcc structure was listed as
2.5 to 3.0 electrons per atom of alloy3,16. For example, the E-B calculation
requires that Os( d 6sp) as a solute can be dissolved in Ir( d 6Sp2) up to 50 at. %,
in close agreement with the measured value of 45 at. % Os. Other solutes
(in at. % ) in Ir as the solvent are as follows, with the calculated E-B values
in parentheses: Mo, 22 (17); W, 22 (17); Nb, 16 (12.5); Ta, 16 (12.5); Ti,

14 (10). The value for Mo(d 5 s) can be calculated by assuming that Ir
contributes nine electrons, and Mo, six electrons, and the alloy configuration
is d 6 spl.5 so that x = 0.17 for Mo in Ir as calculated from 6x+9(1-x) = 8.5;
hence, fcc structure exists from zero to 17 at. % Mo according to the E- B
postulates. The maximum measured concentration of Nb( d 4 s) in Ni( d 7 Sp2)
is about 14 at. % Nb which corresponds to d 6 .8 Sp1.5. Here d 7 sp1.5 configur-
ation for the alloy would yield a reasonably close value of 10 at. % Nb. The
contribution from the d-electrons is again based on how a series of alloys
for which an empirical set of electronic configurations can be assigned in
order to obtain the solubility close to the experimental results.
Other Phases
Several other phases exist in addition to I, II, and III, within the
electron concentration range up to 3.5. They follow, with increasing electron
concentration, the structures indicated by
These phases appear with increasing (d + s + p) electrons up to cph struc-

ture in numerous phase diagrams; they will be described and discussed
briefly in the following sections.
Cubic Cr3Si
The cubic Cr3Si structure (Strukturbericht A15)* is such that each Si
atom is surrounded by 12 Cr atoms without Si-Si bonds. In addition to Cr,
the elements of IVA to VIA may be substituted for Cr, and the elements
of VIlA, VIllA to VB may be substituted for Si. These phases are ordered
phases in which atomic radii should not differ by more than 15%. Excep-
tional A15 phases are Zn 4 Sn and MoTc. In Cr3Si-type phases, the (s + p)
electron concentration is 1.5 to 1.75 according to Brewer when the d
configuration is chosen properly. Thus, for Cr3Si, there are 7 (s + p) elec-
trons per 4 atoms or 1.75 (s + p) electrons per atom. For Cr3Pt, there are
6/4 = 1.5 electrons per atom, assuming that Pt does not contribute any d
atoms, but if Pt retains d 6 structure, or yields only one d-electron, then
1.75 (s + p) electron concentration still prevails.
*These structures are discussed in standard texts. l7 • IS See also Nevitt. 20 AI5 and 0' phases
compete with bcc and cph phases when atomic sizes are favorable.
250 Appendix A
a Phase
The a phase occurs at the atomic radius ratios ofO.93 to 1.15 according
to Nevitt. 20 It is formed by quasi-hexagonal layers of atoms, with a tetragonal
unit cell. The (s + p) concentration is generally 1.2 to 1.9, overlapping bcc
and cph structures. The range of (d + s + p) concentration is between 5.6
and 7.7 electrons per atom, with a very large number of phases occurring
in the vicinity of 6.6 electrons. The assignment of nonintegral configurations,
such as dS'sspo.s for Mn, d 6 . Sspo.s for Fe, retains the range of (s + p)
recommended by Brewer. For example, Cr-Mn, Mn-Fe, Mn-Re, Mn-Tc,
Mo-Tc, Re- W binary systems form a phases.
X or a-Mn Phase
The remaining phases, P, }.t, R, and X, are very closely related to the
a phase in their crystal structures. The P, }.t, and R phases are labeled as
the }.t phase by Brewer in his multicomponent diagrams. The electron
concentrations for these phases are determined from the phase diagrams
where they occur. The X phase for Mn is at 7 (d + s + p) electrons per
atom, and for the binary alloys, the range is 6.3 to 7.0 (d + s + p) electrons
per atom. Since all the transition elements forming X phases assume the d S
configuration, the (s + p) concentration range is from 1.3 to 2.0, overlapping
the a and cph phases.
Phase-Stabilizing Elements in Fe, Mn, Ti, and Zr

The transition elements to the left of iron (VIllA), except Mn, stabilize
a- and B-Fe (both bcc-Fe) with Fe as the solvent by enclosing the y-Fe
(fcc-Fe) inside a loop known as the y-Ioop. Nontransition elements to the
right of Fe forming substitutional solid solutions, e.g., AI, Si, also stabilize
bcc Fe. Aluminum stabilizes a-Fe by enhancing the pairing of d-electrons
in Fe by contributing the excess outer electrons in AI. This is also the case
for Si in Fe.
The following elements stabilize the y-Fe phase (fcc):
Mn (Fe) Co Ni Cu
Ru Rh Pd
Os Ir Pt Au
This is explained on the basis that the elements to the right, e.g., Ni, can
contribute electrons from the d-orbital to the s- and p-orbitals, and thus
stabilize the fcc (d X Sp2) structure.
Similar explanations are also applicable to bcc ~ fcc phase stabilization

in Mn-base alloysl5; i.e., the elements to the left of Mn in the periodic table
stabilize bcc-Mn, and those to the right, fcc-Mn.
The structure of Ti and Zr is cph at low temperatures and bcc at high
temperatures. The transition elements to the right of these elements VA to
VIllA, and Cu, Ag, and Au, all as solutes, lower the transition temperature
and favor the bcc forms of the Ti- and Zr-rich alloys because the empty
d -orbitals in Ti and Zr act as sinks for the (d + s + p) electrons. The presence
of vacant orbitals enhances unloading of d-electrons from such metals as
Ir and Pt and thus the formation of highly exothermic phases with Ti and
Zr.8 In the case of nontransition elements such as AI and Sn as solutes, the
excess available s- and p-electrons stabilize the cph structures of Ti and Zr.
Effect of Pressure
Brewer postulated that increased pressure will allow the nuclei to move
closer to permit the d-orbitals to overlap better and stabilize the crystal
structure corresponding to the largest number of bonding d-electrons. For
the metals having fewer than five d-electrons per atom, e.g., Ti in cph form
(d 2sp), increasing pressure would favor d 3s corresponding to the bcc form
of Ti, in agreement with experimental results. Similar results have been
verified for Zr and Hf. In general, for an element with fcc and d 3 Sp2 structure,
increasing pressure would favor d 3 sp2 ~ d 4 sp ~ d 5 s, or fcc ~ cph ~ bcc.
For d 4 sp2, increasing pressure would stabilize the structures corresponding
to d 5 sp which has the largest number of unpaired d-electrons. The elements
that have bcc structure at ordinary pressures, V, Nb, Ta, Cr, Mo, and W,
remain bcc up to the highest pressures investigated, because (1) for V, Nb,
and Ta, d 4 s structure cannot be changed to d 5 , and (2) for Cr, Mo, and
W, the d 5 s structure already has the optimum number of d-electrons. On
the other hand, for elements having more than five d-electrons per atom,
e.g., d 7s, increasing pressure would favor d 7s ~ d 6sp ~ d 5sp2 ~ , or bcc ~
cph ~ fcc. In summary, increasing pressure would tend to stabilize the
structures toward the maximum number of unpaired, or bonding d-elec-
trons, i.e., toward d m where m tends as close to five as possible. 3 ,15,21
Multicomponent Diagrams
Various phases and their maximum compositional limits in ternary and

multi component systems can be mapped by using the E-B postulates. The
method appears to have its best success for the second and third transition
252 Appendix A
Re (7J
Os (8)
W (6)
Ir (9)
Pt (10)
100 80 60 40 20 0
AT,% IN
Figure A,2. Optimum phase boundaries in alloys of W with Re, Os, Ir, and Pt, Numbers in
parentheses are (d + s) electrons. I = bee, II = cph, and III = fcc. (Courtesy of L. Brewer;
see also Reference 3.)
series of Group IVA, VA, and VIA elements alloyed with the remaining
groups of elements to the right,
One such diagram is shown in Fig. A.2 for the alloys of W, where the
horizontal axis represents at. % W. The vertical axis is the total outer electron
concentration, which increases from 7 for Re to 10 for PC A horizontal line
from Os to the left represents the binary W-Os system for which the
maximum compositional ranges of existence for various phases occur at
the intersections of this line with the boundaries for the single phases. The
areas marked I, II, and III are for bee, cph, and fcc, respectively; u and X
represent the phases designated by these letters.
Such diagrams are not isothermal since the maximum range of composi-
tion for one phase is generally not at the same temperature as that for
another phase. In the W-Os system, the range of composition for II is from
o to 53 at.% W, The upper horizontal line is for W-Re and the lower
horizontal line is for W-Pt. The range of composition for II in W-Re is 0
to 20 at. % W but this range refers to a different temperature than the
preceding range for II in W-Os. The horizontal line that can be drawn
starting halfway between Os and Ir refers to an electron concentration of
8.5 at 0 at. % Wand this line is for the ternary alloy of W-Os-Ir in which
the atomic ratio of Os to Ir is unity, This same line is also for the ternary
W-Re-Pt, and pentanary W-Re-Os-Ir-Pt system, for all of which the
electron concentration is 8.5 at 0 at. % W. The boundaries in this diagram
are largely obtained from the binary diagrams, and their validity for the
ternary and multicomponent systems has not yet been tested. Where experi-
mental data were absent, E- B rules were used to obtain the phase boundaries
in Fig. A.2. Since each point on a phase field refers to a different temperature,
it is not possible to draw the tie lines joining one point on one phase
boundary to another point on another phase boundary. It is also not possible
to designate the areas between the single-phase fields as the multiphase
fields for the ternary and multi component alloys.
Concluding Remarks
The basic E-B postulate that bcc, cph, and fcc correspond respectively
to one, two, and three (s + p) electrons is not applicable to cph forms of
Li and Na, and fcc forms of Ca and Sr. These metals have relatively
simple electronic structures; hence, the inability of E- B postulates to account
for these structures is fundamental. Nevertheless, the numbers of structures
and transition phenomena successfully explained by these theories are quite
impressive.
In addition to the electronic configuration, Brewer considered addi-
tional important factors, i.e., the size factor, and attractive and repulsive
force factors [see equation (B.19)]. All these factors, considered judiciously,
yield the appropriate solubility or stability limits for various structures and
their energies as discussed in Brewer's publications.
There are many aspects of the E-B approach that disagree with the
methods of Friedel,I3,14 Nevitt/o and Miedema et al. 22 It is nevertheless
generally felt that the E-B theories and postulates represent the most
comprehensive existing treatment of solubility ranges for various phases in
binary and multi component systems. The theories and schemes to be
developed in the future will likely take advantage of the E-B theories, along
with other theories, correlations, and postulates. The immediate problem
is to test the validity of the diagrams by obtaining experimental data on a
number of binary and multi component alloys of the transition elements for
which E-B theories claim the greatest success.
Additional recent publications by Brewer et al. 23 - 27 are recommended
to the reader interested in pursuing this topic further.
References
1. For a brief summary, see N. Engel, Powder Metall. Bull. 7, 8 (1954).

2. L. Brewer, in Electronic Structure and Alloy Chemistry of Transition Elements, edited by
P. A. Beck, Interscience, New York, p. 221 (1963).
3. L. Brewer, in High-Strength Materials, edited by V. F. Zackay, Wiley, New York, p. 12
(1965).
4. L. Brewer, in Phase Stability in Metals and Alloys, edited by P. Rudman, J. Stringer, and
R. Jaffee, McGraw-Hili, New York, pp. 39-61, 241-249, 344-346, 560-568 (1967).
254 Appendix A
5. L. Brewer, Acta Metall. IS, 553 (1967).

6. L. Brewer, Science 161, 115 (1968).
7. L. Brewer, in Plutonium and Other Actinides, edited by W. N. Miner, Metallurgical Society
of AIME, New York, p. 650 (1970).
8. L. Brewer and P. R. Wengert, Metall. Trans. A/ME 4, 83 (1973).
9. L. Brewer, in Transition Metal Alloys-A Chemist's View, Am. Inst. Phys. Conf. Proc.,
1972, edited by H. C. Wolfe, No. 10 (Part 1), p. 1 (1973).
10. L. Brewer, I. Nucl. Mater. 51, 2 (1974).
11. L. Brewer, Rev. Chim. Miner. 11,616 (1974).
12. J. Friedel, in Electronic Structure and Alloy Chemistry of Transition Elements, edited by
13. W. Hume-Rothery, Acta Metall. 13, 1039 (1965), 15,567 (1967).
14. J. Friedel, W. Hume-Rothery, R. Jaffee, L. Kaufman, W. M. Lomer, T. B. Massalski, and
W. B. Pearson, in Phase Stability in Metals and Alloys, edited by P. Rudman, J. Stringer,
and R. Jaffee, McGraw-Hili, New York (1967).
14a. S. L. Altmann, C. A. Coulson, and W. Hume-Rothery, Proc. R. Soc. London 240A, 145
(1957).
15. W. Hume-Rothery, Prog. Mater. Sci. 13,228 (1968).
16. W. Hume-Rothery, R. E. Smallman, and C. W. Haworth, The'Structure of Metals and
Alloys, Institute of Metals, London (1969).
17. C. S. Barrett and T. B. Massalski, Structure of Metals, McGraw-Hili, New York (1980).
18. C. Kittel, Introduction to Solid State Physics, Wiley, New York (1971).
19. L. Pauling, The Nature of the Chemical Bond, Cornell University Press, Ithaca, N.Y. (1960).
20. M. V. Nevitt, in Electronic Structure and Alloy Chemistry of Transition Elements, edited by
21. L. Brewer, in Structure and Bonding in Crystals, Volume I, edited by M. O'Keefe and A.
Navrotski, Academic Press, New York, p. 155 (1981).
22. A. R. Miedema, R. Boom, and F. R. deBoer, 1. Less-Common Met. 41, 283 (1975).
23. L. Brewer, Systematics of the Properties of the Lanthanides, edited by S. P. Sinha, D. Reidel,
Hingham, Massachusetts (1983).
24. L. Brewer, I. Chem. Ed. 61, 101 (1984).
25. L. Brewer, I. Materials Ed. 6 (5), 734 (1984).
26. J. K. Gibson, L. Brewer, and K. A. Gingerich, Met. Trans. A/ME ISA, 2075 (1984).
27. L. Brewer and D. G. Davis, Met. Trans. AIME ISA, 67 (1984).
B
Estimation of Enthalpy of Alloy Formation
Introduction
Various methods based on the properties of pure elements have been

developed for estimating the enthalpies of formation of a limited number
of alloys. We describe one of the more recent semiempirical methods
developed by Miedema et al.,1-5 applicable to a very large number of alloys.
This method is partly based on the Wigner-Seitz6 atomic cell theory for
pure metals, which will be described first since Miedema et al. assume that
this theory is also applicable to the alloys.
The Wigner-Seitz cells for pure metals are constructed by drawing
planes perpendicularly bisecting the lines joining the atoms in order to form
polyhedra. Each polyhedron for a pure metal contains one positively
charged atomic core, and the valence electrons of each atom; such cells are
shown schematically in Fig. B.I(a) for two-dimensional crystals of pure A
and pure B. At the cell boundary, the potential energy, and density of
electrons nws are at their minimum values. The binding energy Zeii/2 of
each atom forming Z bonds with its neighbors is equal to the energy of
sublimation of the atom, and this energy is assumed to be equal to the
energy of valence electrons outside the positive ion core. The total energy
of a crystal of N atoms calculated by Wigner and Seitz by introducing a
large number of corrections agrees well with the experimental data only
for the alkali metals and disagrees for the metals containing d-electrons.
The details of these calculations are not relevant for our purposes, but the
concept itself is useful as will be seen shortly.
The method of Miedema et al. I - 5 for the estimation of enthalpies of
formation, I1H, for alloys assumes that the Wigner-Seitz theory for pure
metals can be expanded to include the binary alloys. When two metals A
255
256 Appendix B
A A A A
A A A A
A A A A
+
A A A A
a
b c
Figure B.1. Wigner-Seitz cells for two-dimensional metals and alloys. (a) Cells for pure metals
A and B; they are mixed to form a mechanical mixture as shown in (b). The cell sizes (areas)
are identical in (a) and (b). (c) The mechanical mixture becomes an alloy phase and phase
boundaries move to smooth out the discontinuity of electron density in (b). In this figure
n~, < n~,; hence, the cell boundaries in (b) move to make the B cells larger as shown in (c).
Volume effects are related to the change transfer; hence, to the differences in the elec-
tronegativities.
and B form a hypothetical mechanical mixture of atoms, as shown in Fig.

B.I(b), the shapes of the cells change but the volume of each cell for each
type of atom remains unaltered. The cells are therefore drawn to leave no
void and no change in each atomic volume after mixing. The shapes of
cells in a mechanical mixture have no effect on the energies of atoms so
long as their cell volumes remain the same in pure and mixed states. In
this condition, f1H is zero since the alloy is a mechanical atomic mixture.
However, at the boundaries in Fig. B.I(b), a discontinuity in the Wigner-
Seitz electron density nws exists because the electron densities for A and B
are different, and we assume that n~s < n!s (superscripts identify the pure
metals). This discontinuity is smoothed out upon alloy formation when the
cells for B expand at the expense of those for A as shown in Fig. B.I(c).
The cell boundary is again where the density of electrons passes through
a minimum. Miedema et al. assume that the resulting change in the density
of electrons at the boundary of the Wigner-Seitz cell, f1nws> contributes a
Estimation of Enthalpy of Alloy Formation 257
positive term q(lln!!s3? to IlH where q is an empirical constant. This is the

well-known size effect in alloying, i.e., greater differences between the atomic
radii of A and B contribute greater positive values to 1lH.
The values of nws are in electrons per atomic unit cube with one atomic
unit cube equal to 0.529 A (0.1 nm = 1 A), i.e., nws = 1 corresponds to about
4 x 1022 electrons/ cm 3 ; they were found to be empirically related by n~s =
¢/V, where ¢ is the compressibility (or bulk modulus), ¢ = - V(ap/aV),
with P in units of 104 bars and V the molar volume in cubic centimeters
per gram atom. This relationship was originally obtained empirically by
plotting 0.5 In(¢/ V) versus In nws> which yielded a straight line for a number
of elements. 1 The results for nws> either calculated directly by the Wigner-
Seitz method or obtained by using ¢/ V, are listed in Table B.l, where the
values are empirically adjusted to obtain the best fit for the available data
on the enthalpy of formation per gram atom of alloy, IlH, for various alloys.
The adjusted values l - 5 differ slightly from the actual values of nws> but for
a limited number of elements they differ significantly; e.g., for Mn and U,
the assigned values of nws are higher by 44 and 36%, respectively. In most
cases, however, the actual and empiricallly adjusted values are considerably
Table B.t. Model Parameters for Transition and Noble Elements
n ws
l/3
<P V2 /3 n~/ <P V2 /3
Sc 1.27 3.25 6.1 Rh 1.76 5.40 4.1

Ti 1.47 3.65 4.8 Pd 1.67 5.45 4.3
V 1.64 4.25 4.1 Ag 1.39 4.45 4.7
Cr 1.73 4.65 3.7 La 1.09 3.05 8.0
Mn 1.61 4.45 3.8 Hf 1.43 3.55 5.6
Fe 1.77 4.93 3.7 Ta 1.63 4.05 4.9
Co 1.75 5.10 3.5 W 1.81 4.80 4.5
Ni 1.75 5.20 3.5 Re 1.86 5.40 4.3
Cu 1.47 4.55 3.7 as 1.85 5.40 4.2
Y 1.21 3.20 7.3 Ir 1.83 5.55 4.2
Zr 1.39 3.40 5.8 PI 1.78 5.65 4.4
Nb 1.62 4.00 4.9 Au 1.57 5.15 4.7
Mo 1.77 4.65 4.4 Th 1.28 3.30 7.3
Tc 1.81 5.30 4.2 U 1.56 4.05 5.6
Ru 1.83 5.40 4.1 Pu 1.44 3.80 5.2
aElectronegativity, cP, is in volts; electron density, nw>o is in electrons per (0.529 A)3; molar volume, V, is
in cm 3 at room temperature. For values related to compressibility, ¢, see, e.g., V. S. Fromenko and G. W.
Samsonov, Handbook a/Thermionic Properties, Plenum Press, New York (1966), and D. E. Eastman, Phys.
Rev. B 2, I (1970); for ¢, V, and nw>o see K. A. Gschneidner, Solid State Phys. 16,275 (1964), and V. L.
Moruzzi, J. F. Janak, and A. R. Williams, Calculated Electronic Properties 0/ Metals, Pergamon Press,
Elmsford, New York (1978).
258 Appendix B
closer to each other, but it must be emphasized that the adjustments in nws
are essential for the usefulness of the proposed semi empirical method. The
difference .:lnws is not known exactly, but it is assumed that .:In!!s3 for the
alloy is equal to [(n~s)1/3 - (n!s)1/3] for the pure components A and B.
The second effect is due to the electronegativity of elements, i.e., the
potential of electrons in metals, 4>. The more electronegative elements tend
to attract the electrons from the less electronegative elements upon alloying.
The electronegativity scales differ depending on the types of measurement,
such as the standard electrode potential method, ionization of monatomic
gases, and formation of halogen bonds. 7- 9 Since the electronegativity of
electrons in pure component A is different from that in pure component B,
the electrons tend to spend more time about the more electronegative atom
after alloying. Therefore, the potential of electrons is lowered upon alloying,
and consequently .:lH is lowered by -P(4)A - 4>0)2 = -p(.:l4>f where pis
an empirical constant. A term of this type has been used previously by
Pauling.7 Again, .:l4> for an alloy is not known, but it is assumed that
.:l4> = 4>A - 4>0 where 4>A and 4>0 refer to pure A and pure B, respectively.
The initial values of 4>, as used by Miedema et al., were the experimentally
measured values in a number of compilations cited in Table B.t. However,
these values have been adjusted slightly for empirical fitting of the experi-
mental values for .:lH. The adjustments are small in most cases,I-5 e.g., the
adjusted values for Ti and Zr are approximately 15% smaller than the
experimental values, but the remaining values for the transition elements
differ less than 15 % .
The contribution to .:lH resulting from the changes in electron con-
centration and electronegativity for certain classes of alloys is expressed by
.:lH = f(XA, V)g(XA' nws )[q(.:ln!!s3f - p(.:l4»2] (B.1)
where f(XA, V) is dependent on XA and volume V, and g(XA' nws) on XA
and nws as will be seen later, and p and q are empirical constants. The
next task is the determination of p and q.
Empirical Coefficients
The determination of empirical coefficients p and q requires experi-
mental data on .:lH, and in the absence of such data, binary phase diagrams
from which the algebraic sign of .:lH can be estimated. Initially required
information is whether .:lH is positive or negative. In the absence of
experimental data for .:lH, the sign of .:lH was obtained by the following
established empirical rules: (1) .:lH is negative for all the binary alloys in
which intermetallic compounds or ordered phases have been observed, and
(2) .:lH is positive for all the binary alloys when there are no intermetallic
Table B.2. Model Parameters for Nontransition and Nonnoble Metals·
n w,
'/3 4> V2 !.' n!:/ 4> V2 /3
HO) 1.5 5.20 1.42 B(3) 1.55 4.75 2.8

Li 0.98 2.85 5.5 AI 1.39 4.20 4.6
Na 0.82 2.70 8.3 Ga 1.31 4.10 5.2
K 0.65 2.25 12.8 In 1.17 3.90 6.3
Rb 0.60 2.10 14.6 n 1.12 3.90 6.6
Cs 0.55 1.95 16.8 C(4) 1.90 6.20 1.8
Ca 0.91 2.55 8.8 Si 1.50 4.70 4.2
Sr 0.84 2.40 10.2 Ge 1.37 4.55 4.6
Ba 0.81 2.32 11.3 Sn 1.24 4.15 6.4
Be(2) 1.60 4.20 2.9 Pb 1.15 4.10 6.9
Mg 1.17 3.45 5.8 N(5) 1.60 7.00 2.2
Zn 1.32 4.10 4.4 P 1.53 4.95 3.8
Cd 1.24 4.05 5.5 As 1.44 4.80 5.2
Hg 1.24 4.20 5.8 Sb 1.26 4.40 6.6
Bi 1.16 4.15 7.2
"For units, see footnote to Table B.1. Values of V 2 !> for H, Si, Ga, Ge, C, N, As, Sb, Bi are for hypothetical,
close-packed, metallic structures. The numbers in parentheses indicate the valence in the group headed
by that element.
compounds and the mutual solubilities are less than 10 at. %. For all the
binary alloys of the elements in Tables B.1 and B.2, except the alloys of
(Class a element + Class (3 element) in Table B.3, equation (B.t) is valid.
For I1H = 0, equation (B.t) requires that a plot of 11¢ versus I1n 1/3 should
leave positive I1H values, indicated by +, on one side of the straight line
passing through the origin, while negative values of I1H, indicated by
Table B.3. Values of Parameter rip·

Class a Class f3
Be B C N
lIA Transition metals, 1lIA- VIllA IB 0.4 1.9 2.1 2.3
Ca Sc Ti V Cr Mn Fe Co Ni Cu Mg AI Si
0.4 0.7 1.0 1.0 1.0 1.0 1.0 1.0 1.0 0.3 0.4 1.9 2.1 -
Sr Y Zr Nb Mo Tc Ru Rh Pd Ag Zn Ga Ge As
0.4 0.7 1.0 1.0 1.0 1.0 1.0 1.0 1.0 0.15 1.4 1.9 2.1 2.3
Ba La Hf Ta W Re Os Ir Pt Au Cd In Sn Sb
0.4 0.7 1.0 1.0 1.0 1.0 1.0 1.0 1.0 0.3 1.4 1.9 2.1 2.3
Th U Pu Hg n Pb Bi
0.7 1.0 1.0 1.4 1.9 2.1 2.3
"To obtain r/ p for a binary solid alloy formed by one Class a and one Class f3 element, multiply the
numbers under the elements. For liquid alloys rip is reduced by a faclor of 0.73. For all other alloys.
r/ p ; O. Class f3 elements. except those in the first column. contain p-type outer electrons.
260 Appendix B
should be on the opposite side. A mapping of this type for solid alloys is
shown in Fig. 8.2, wherein the slope of the straight line yields
tlrf> 12 (B.2)
1tln!fs3 = : = 9.4
For liquid alloys of the same types of components, a similar mapping exists,4
yielding again q/p = 9.4. Not all the alloys are on the correct side of the
straight line in Fig. B.2, and in a similar figure for liquid alloys, but there
are very few such alloys.
The corresponding analysis of the results on tlH for (Class a + Class
(3) alloys shows that an additional large term, r, is required in equation (B. 1):
The term r is justified heuristically as the hybridization or interaction energy

between the electrons of Class a and Class {3 elements, particularly when
a transition element with d-electrons interacts with the elements having
outer p-electrons on the right side of the periodic table. Introduction of r
makes the relationship between tlrf> and tln!fs3 hyperbolic for tlH = o.
Therefore, a plot of tlrf> versus tln!fs3 , as shown in Fig. B.3, should yield
(V!
-- -:~
Figure B.2. Mapping of t:..c/J in volts and t:..n!fsl

in (density units)·/l for positive, +, and negative,
-, values of t:..H according to equation (B.1).
Straight line is for t:..H = 0, with slope = (9.4)°·5.
Results are for binary solid alloys of (transition
or noble element) + (transition, noble, alkali, or
05 '0 alkaline earth). (From Miedema and de Chatel S
6n~"lldU ,'1)1--+ with permission.)
20
Zn,Cd,Hg AI,Ga,In, Tl
i 15
M
Ivl
10
05
°0 02
- 0 02 04 A 'I) 0.6
nws---+
5n,Pb AS,Sb,Bi
t
AQI
15
Ivl
10
0.5
-
02 04 06 o 02 04 06
An'3_ An "3
ws ws
Figure B.3. Mapping of f1cf> in volts and f1n!!; in (density units)!/3 for positive, +, and negative,
-, values of f1H according to equation (B.3), Solid curve separates + and - for binary solid
alloys of (Class a + Class (3), Class f3 elements are listed in each panel. (From Miedema and
de Chatel 5 with permission.)
262 Appendix B
the value of rip. Further analysis has shown that qlp is again 9.4; hence,
the straight lines with this slope that can be drawn through the origin are
asymptotes of the hyperbola at very large values of arl>. The results for rip
for solid alloys are obtained from the scheme in Table B.3, i.e., by mUltiplying
the number under the Class a element with the number under the Class 13
element; e.g., for Ag-Zn, rip = 0.15 x 1.4 = 0.21. The values of rip for
liquid alloys should be 0.73 x rip = r(l}1 p(l} where r(l} and p(l) refer to
the liquid alloys. Comparison of the four plots in Fig. B.3 shows that rip
increases with increasing number of outer p-electrons of Class 13 elements.
Calculation of ilH
The values of aH may now be calculated by using ql p = 9.4 for binary

alloys of any type, and rip computed from Table B.3 for (Class a + Class
f3) binary alloys. The recommended values 5 of p for liquid and solid alloys
are given in Table B.4. The values of p fall into three sets: those elements
classified as transition elements (trans.), noble elements Cu, Ag, and Au
(nob.), and the remaining elements (remn.).
Next, it is necessary to formulate f(x A, V} and g(XA' nws }. The energy
effects upon alloying are assumed to be generated at the contact surface
between dissimilar cells; therefore, it is further assumed that the surface
concentrations x~ and x~ are more relevant than the bulk concentrations
X A and X B , and x~ is defined by
(B.4)
where VA and VB are the molar volumes of the pure elements and x~ =
1 - x~. Since the cell volumes change upon alloying, VA for pure A, and
VB for pure B cannot be the same as VA(alloy} and VB(alloy} in the alloy,
Table B.4. Values of p for Solid and Liquid Binary AlloysQ
Alloy type p
(trans. or nob.) + (trans. or nob.) 14.1

(trans. or nob.) + (remn.) 12.3
(remn.) + (remn.) 10.6
"Transition elements (trans.) are shown in Table B.3; noble elements (nob.)
are Cu, Ag, and Au; others are designated as remaining elements (remn.).
Listed values, with q" n:!;, and y2/3 in Tables B.I and B.2, yield tlH in
kJ/g-atom (4.184 J = 1 cal).
but the use of VA and VB for the pure elements in equation (B.4) is assumed
to be valid for a large number of alloys. It is further assumed thatf(xA, V)
follows the zeroth approximation to the regular solutions with x~ substituted
for Xi:
f(XA, V) == f(x~, x~) = x~x~ (B.5)
The function g(XA, nws) =g is assumed to be
2XA V;:3 + 2XB V~/3

(B.6)
g = (n~s)-1/3 + (n!s)-I/3
where the denominator is called the screening length of electrons. An
approximate relationship exists between V; and n~s (i = A or B), and for
this reason g varies within a narrow range for all the alloys. However,
taking g as a constant leads to a less satisfactory correlation with the
experimental values of ilH. In the original treatment, l the product fg was
taken as equal to XAXB; therefore,the product of equations (B.5) and (B.6)
is a requirement4 largely based on empiricism, because fg provided a better
fit for ilH than XAXB. Substitution of equations (B.4)-(B.6) in equation
(B.3) yields
The partial molar enthalpy of solution of A, ilHA(XA ~ 0), when XA

approaches zero, can readily be derived from equation (B.7):
This equation justifies the use of V;:3 in equation (B.7), because ilHA(XA ~
0)/ ilHB(XB ~ 0) is roughly proportional to (VAi VB)2/3 in the majority of
cases considered by Miedema et al. although there are notable exceptions.
Comparison of equations (B.7) and (B.8) shows that
(B.9)
where XAX~ may have any value but ilH(XA ~ 0) refers to XA ~ O.

The surface concentration x~ should be expressed in terms of the
volumes in alloys V;(alloy) instead of V;(pure) listed in Tables B.I and B.2.
The crude empirical relation yielding V;.(3(alloy) is
(B.IO)
264 Appendix B
0.14 for monovalent A

0.10 for divalent A
UA = 0.07 for trivalent A
0.04 for higher valent and for
all transition and noble A
This equation is used only for large values of (CPA - CPB) and for alkaline
and alkaline earth metals and small-atom stongly electronegative elements
such as B(boron), C, H, and N. For example, VA(pure) = 5.5 and
VA (alloy) = 4.98 for A = Li, B = AI, and X B = 0.5. In contrast, VA (pure) =
4.8 and VA (alloy) = 4.68 for A = Ti, B = Fe, XB = 0.5, and UA = UB = 0.04;
the resulting difference in VA(pure) and VA (alloy) is considered to be
negligible for Fe-Ti. A few examples of calculated and experimental values
of aH are listed in Table B.5 for solid and liquid alloy phases of equiatomic
composition. For example, aH for liquid equiatomic Au-Zn alloy can be
Table B.S. Calculated and Experimental Values of tlH of Formation for 1 g-atom
of Equiatomic Solid and Liquid Alloy Phases. Results Are for Long-Range
Disordered Alloys, Except as Explained in Footnote a
kJ/g-atom kJ/g-atom
Alloy aH(ca\c) aH(exp) Alloy aH(calc) aH(exp)
Solid alloys Liquid alloys b

AI_Au" -37 -38.7" AI-Au -21 -33.8
AI_Nia -47 -58.8 a AI-Ni -22 -50.2
AI-Ti -44 -36.4 Au-Zn -16 -22.5
Co_Pta -11 -13.4 a Bi-Na -20 -30.1
Cu-zna -10.5 -12.1 "
Cu-Zn -5 -7.7 Cu-Fe +15 +8.9
Fe-Ti -19 -20.3 Fe-Si -18 -37.7
Fe-V -7 -8.4 Ga-Li -8 -23.0
Mn-Ni -8 -14.2 Hg-K -10 -21.3
Nb-Ni -31 -22.6 Hg-Pb +1 +0.2
Ni-Ti -38 -33.3 K-Pb -21 -19.8
Pd-Rh +2 +10 Ni-Ti -38 -38.5
"Ordered structures. Structure of AI-Au is uncertain, but assumed to be ordered in this calculation.
Magnitudes of !l.H cannot always predict the existence or the absence of long-range order.
bResults for liquid AI-Au, Au-Zn, Cu-Fe, Hg-Pb, and K-Pb and all those for solid phases were recomputed
for this book, and experimental data were obtained from R. Hultgren, P. D. Desai, D. T. Hawkins, M.
Gleiser, and K. K. Kelley, Selected Values of the Thermodynamic Properties of Binary Alloys, ASM, Metals
Park, Ohio (1973), and from O. Kubaschewski and C. B. Alcock, Metallurgical Thermochemistry, Fifth
Edition, Pergamon Press, Elmsford, New York (1979).
calculated by using p = 12.3, X A = XAu = 0.5, x~ = x~n = 0.4835, (n~J-l/3 +

(n!s)-1/3 = 1.395, and r / p = 0.3 x 1.4 x 0.73 = 0.3066; the result is
I1H = (2 x 12.3 x 0.5 x 4.7 x 0.4835/ 1.395)

x [9.4(1.57 -1.32)2 - (5.15 -4.l0? - 0.3066] = -16 kJ/ g-atom
The experimental value is -22.5 kJ / g-atom as listed in Table B.5. Large

differences between the calculated and experimental values are attributed
to unusual degrees of ordering in liquid alloys.
Ordered Phases
The value of x~ increases for ordered structures because a much larger
fraction of the surface due to B is in contact with A. A purely empirical
relation expressing this fraction, denoted by IB, is
IB = x~[1 + 8(x~x~?] (B.ll)
Note that for disordered solutions IB = x~, but equation (B.ll) must replace
x~ in equation (B.7). Substitution of equations (B.8) and (B.ll) in equation
(B.7) yields another form for I1H, i.e.,
(B.12)
where XA may assume any value for I1H; however, I1HA(xA ~ 0) after the
second equality refers to X A ~ O. For the ordered Cu-Zn alloy of equiatomic
composition at 700 K, I1H(exp) = -12.1 kJ/g-atom, whereas I1H(calc) =
-10.5 kJ/g-atom from equation (B.12), in contrast with I1H(exp) =
-7.7 kJ/g-atom and I1H(calc) = -5 kJ/g-atom for the disordered phase as
listed in Table B.5.
The average number o"solid ordered phases or compounds in a binary
system correlates roughly with the magnitude of I1H; thus, there is often
one compound when I1H is between -4 and -10; three compounds when
I1H is between -20 and -40; five compounds when I1H is more negative
than -75, all I1H in kJ/g-atom of A + B. Here the model proposed by
Miedema et al. cannot estimate I1H unless experimental results indicate
that the phases are either ordered or long-range disordered.
266 Appendix B
Alloys of C, Si, Ge, and N

The alloys of C, Si, Ge, and N require a special treatment,3,(O i,e" the
assumption that they exist in a hypothetical close-packed (cp) metallic
structural form, The volume of this cp form for each of these elements is
largely based on the volumes of closely neighboring cp elements, The
enthalpy of phase change, AHph , for the formation of each hypothetical
phase from the stable phase is as follows:
tJ.Hph , kJ/g-atom
C(graphite) --> C(cp metal) 100

0.5N 2 (gas) --> N(cp metal) 240
Si(diamond cubic) --> Si(cp metal) 33
Ge(diamond cubic) --> Ge(cp metal) 25
0.5H2(gas) --> H(cp metal) 100
Hydrogen is included here for comparison; it will be discussed in detail in

the next section, The derivation of these values is again semiempirical so
that the best-fit values can be obtained for AH of formation of alloys, The
values for N 2 (g) include the dissociation energy into atomic N(g), and then
condensation into a hypothetical cp metallic phase lO ; this is also the case II
for H 2 (g), As an example we compute AH for TiN (A = Ti, B = N), for
which n!!/ = 1.47(1.60), </> = 3,65(7,00), V2f3 = 4,8(2,2), where the values
outside the parentheses are for Ti, and those inside are for N(cp metal),
For Ti, V;'/3(alloy) = 4,478, and for N, Vif3(alloy) = 2.347 are obtained by
using UA = 0.04 in equation (B.I0), and from these values, x~ = 0.6561 and
x~ = 0.3439, because x~ (i = A or B) must be calculated by using Vi/3(alloy).
From equation (B.ll), IB = 0.484, and the substitution of the foregoing
values, as well as p = 12.3 and r/ p = 2.3, in equation (B.7) yields AH =
-273 kJ/g-atom for 0.5Ti(c) + O.5N(cp metal) = 0.5TiN. The value of
0.5AHph = 0.5 x 240 for 0.25N 2(g) ~ 0.5N(cp metal) added to -273 yields
AHo (from N 2) = -153 kJ/g-atom TiN for the enthalpy of formation from
gaseous N2 and crystalline Ti in their naturally existing states, i.e., their
standard states. The experimental value for AHO (from N 2) = AH; is
-169 kJ / g-atom TiN, in good agreement with the calculated value.
The value of AGO at the equilibrium pressure P(N 2 ) and a tem-
perature T( for N 2(g) + nA = An N2 can be calculated by writing AGo =
RT In P(N 2 ) = AH; - T( AS; from which ASg can be calculated. For the
simple case of P(N 2 ) = 1, it is evident that AS; = AH;/ T\. The resulting
value of AS;, and the experimental or calculated value of AH; can be used
as constants in AGo = AH; - T AS; to compute P(N 2 ) at other tem-
peratures within a possible range of 200 K above or below T(, assuming
Table B.6. Calculated and Experimental Values of AHo for Selected Ordered Binary
Phases of H, C, N, Si, and Ge a
Phase 4HO(calc) 4HO(exp) Phase 4HO(calc) 4HO(exp)
Tio.33 HO.67 -43 -42

VO.33 HO.67 -11 - 18
Ni o.67 HO.33 +4 -1
Zr0.33 HO.67 -63 - 55 CrN -40 -60
Nbo.33 HO.67 -28 -20 Fe2N -17 0
PdO.67 HO.33 +2 -7 HfN -169 -185
UIo.27 HO.73 - 51 -62 LaN -150 - 151
Lao.36 HO.64 - 55 -67 TaO.67 NO.33 - 88 -90
Hfo.37 HO.63 - 55 -42 TiN -153 -169
ThO.33 HO.67 -65 -49 VN - 86 -110
UH 3 -18 - 32 ZrN -184 -184
Fe2C -13 +8 FeSi -25 -39
HfC -79 -109 MnSi -42 - 30
M0 7C 3 -21 -13 ThSi -100 -63
MoC -38 -8 TiSi -79 -65
NbC -71 -69 ZrSi -100 -74
TaC -67 -72
Mg 2Ge -12 -38
TiC -75 -92
Ni 2Ge -20 -37
WC - 29 -19
UGe - 54 - 31
ZrC - 84 -96
• AHo is in kJ/g-atom of alloy formed from the elements in their standard states (Refs. 3, 10, II).
that the stoichiometry of An N2 remains unchanged and that the reactant

metal is pure A.
Selected values of ~Ho from the component elements in their standard
states for a number of carbides, nitrides, silicides, and germanides 3 ,JO are
given in Table B.6.
Alloys of Hydrogen
The values of ~H for metal hydrides can be calculated by following

the procedure given by Bouten and Miedema ll and by using the listed
values in Table R2 with rip = 3.9 for H, and ~Hph = lOOkJ/g-atom H.
The value of Vi{3 = 1.42 is based on the closest approach of approximately
2.3 A between two H-H atoms in binary compounds of hydrogen with
metals, wherein the atomic ratios of H to metal are high. However, Vi{3 is
again a best-fit semiempirical value for hydrogen, as well as the remaining
parameters. The value of rip for the alkaline-transition element alloys is
zero, but in contrast, despite the fact that H(cp metal) should also be an
268 Appendix B
alkaline metal, r / p is not zero for H but equal to 3.9 for the [H( cp metal) +
transition element] alloys. For example, r / p = 3.9 x 0.7 for the La-H alloys.
The calculations of V;';3(alloy) and fs for H(cp metal) are different
from those for other elements because these parameters are calculated by
a different procedure. The analytical form adopted for fs is identical with
equation (B.11) but equation (B.10) is rewritten as
(B.13)
The corresponding equation for Vi/\alloy) is obtained by interchanging

the subscripts A and B in this equation. The value of fs is not known since
V;';3(alloy) is not known for computation of x~, because x~ must be
computed by using V;';3(alloy) for the interstitial elements. Therefore, x~
is computed first by using Vi/3(pure) for pure A and pure B, in equation
(B.4) and then the result is substituted in equation (B.11) to compute the
first value offs. From this value offs, V;';3(alloy) is computed from equation
(B.13). This process is repeated once more to obtain closer values of fs and
Vi/3(alloy); however, in calculating the second value of Vi/3(alloy),
Vi/3(pure i) is always used on the right side of equation (B.13). For example,
for Tio.33Ho.67 (A = Ti, B = H), the parameters are: V;';3(pure) = 4.8;
V~/3(pure) = 1.42; (n;;s)~l/3 + (n!s)~l/3 = 1.347; ~n~/ = -0.03; cf>A - cf>s =
-1.55; UA = 0.04; and Us = 0.14. The calculated first and second values,
denoted as (I) and (II) after each symbol, are as follows:
x~(I) = 0.33 x 4.8/(0.33 x 4.8 + 0.67 x 1.42) = 0.6248;

x~(I) =1- x~(I) = 0.3752
fA(I) = 0.6248[1 + 8(0.6248 x 0.3752?] = 0.8995
fs(I) = 0.3752[1 + 8(0.3752 x 0.6248?] = 0.5402
V;';3(alloy, I) = 4.8(1 - 0.04 x 0.5402 x 1.55) = 4.639
Vi/3(alloy, I) = 1.42(1 + 0.14 x 0.8995 x 1.55) = 1.697
x~(II) = 0.33 x 4.639/(0.33 x 4.639 + 0.67 x 1.697) = 0.5738;

x~(II) =1- x~(II) = 0.4262
fA(II) = 0.5738[1 + 8(0.4262 x 0.5738)2] = 0.8483
fs(II) = 0.4262[1 + 8(0.5738 x 0.4262)2] = 0.6301
V~13(alloy, II) = 4.8(1 - 0.04 X 0.6301 X 1.55) = 4.612

Vi(3(alloy, II) = 1.42(1 + 0.14 X 0.8483 X 1.55) = 1.681
The third approximation is not necessary because the differences for
V;13(alloy, II) - V;13(alloy, I), (i = A or B), are already small and the
calculated third value for V;13(alloy, III) is very close to V;13(alloy, II).
The result for I1H is as follows:
I1H = 2 X 12.3 X 0.33 X 4.612 x 0.6301[9.4( -0.03? - (-1.55)2 - 3.9]/1.347

= -110 kJ/g-atom
I1HO[from H2(g)] = -110 + 0.67 x tOo = -43 kJI g-atom
The experimental value is -42 kJ/g-atom, in very close agreement with the
calculated value. The increase in the volume of Ti upon absorption of
hydrogen is calculated from
V(alloy) = 0.33 VTi + 0.67 VH = 4.73 (B.14)
where V(alloy) is the volume of the alloy, and the partial molar volume V;
is assumed to be the same as Vj(alloy, II). The volume of pure 0.33Ti is
3.47 cm 3; hence, 11 V = 4.73 - 3.47 = 1.26 which is the increase in the volume
of Ti after the dissolution of hydrogen, and the calculated value, 11 VI V =
1.26/3.47 = 0.36 or 36%, agrees well with the experimental value, 31 %.
Similar calculations for LaO.36HO.64 yield x~ = 0.705; x~ = 0.295;
V;.(3(alloy) = 7.727; Vi(3(alloy) = 1.826; IB = 0.3966; I1H = -119 kJ/g-
atom; and I1HO(from H 2) = -55 kJ/g-atom alloy, in fair agreement with
the experimental value of -67 kJ1g-atom. The calculated value of 11 VI V is
8%, in poor agreement with the experimental value l l of 19%. The results
for a number of binary hydrides selected from Bouten and Miedema l l are
listed in Table B.6.
Ternary Hydrides
Intermetallic compounds of a number of transition elements form stable

hydrides. The enthalpies of formation of such hydrides can be estimated
by a modification of the preceding method. 11
The enthalpy of formation of a ternary hydride ABn Hv+w from H2 and
ABn can be estimated from the standard enthalpies of formation of two
binary hydrides AHv and Bn H w , and the intermetallic phase or compound
AB n. Metal A and B are both transition elements, and A is assumed to form
a very stable solid hydride at a sufficiently low temperature, but B forms a
270 Appendix B
much less stable hydride than A. The generalized equation for estimation
of dH;(ABnHv+w) by reaction of ABn with H 2 (g), as indicated by the
subscript g on dH, is as follows ll :
+ w)H 2 + ABn = ABnHv+w
O.5(v
dH;(ABnHv+w) = dH;(AH v) + dH;(BnHw) (B.15)
- (1 - F) dH(AB n)
where v, w, and F depend on stoichiometric coefficient n and on the atomic
radius of A, i.e., whether A is one of the first or second group of metals as
shown in the first column of Table B.7. The empirical term F is a measure
of bonding between A and B atoms in the ternary hydride. Thus, when n
is greater than 5, and v is sufficiently high, H atoms completely surround
A and the bonding between A and B is greatly weakened so that F = 0; in
this case, equation (B.15) refers to the following reactions:
0.5(u + v)H 2 + ABn = ABnHu+v dHg(ABnH,,+w) = dH;(AH,,)

+ dH;(BnHw) - dH(AB n) (B.16)
This equality is possible only if dHH _H for the last reaction is zero.
Table B.7. Values of Parameters for Calculation of Enthalpies of Hydrides Formed

from Intermetallic Compoun.ds ADn
Metal AQ AB" AB"Hv+w v w F
Ti, Hf, Zr, V, Nb. Ta. Sc ABs ABsHs 2 3 0.1

AB3 AB3H4 2 2 0.2
AB2 AB 2H 3.S 2 1.5 0.4
AB ABH2 1.5 0.5 0.6
Lanthanides Y, Th. U. Pu ABs ABsH6 2.5 3.5 0.1

AB3 AB3HS 2.5 2.5 0.2
AB2 AB2H4 2.5 1.5 0.4
AB ABH 2.S 2 0.5 0.6
"The two groups of metals have considerably differing atomic sizes in aUoyed states.
The values of n, v, W, and F are listed in Table B.7. As an example,

we calculate .:lHg(AB5H6) = .:lHg(LaNi5H6) from the values listed in Tables
B.7 and B.8, wherein v = 2.5, W = 3.5, F = 0.1, .:lH;(LaH 2.S ) =
-185 kJjmole, and .:lH;(NisH3s) = +10 x 3.5 = 35 kJjmole. Calculations
by using equations (B.IO), (B.I1), and (B.12) yield .:lH(LaNi s) =
-70.5 kJjmole, and substitution of these three values of .:lH in equation
(B.14) with F = 0.1 gives .:lHg(LaNi5H6) = -87 kJjmole, ingood agreement
with the experimental value,13 .:lHg(LaNisH6) = -94 kJjmole. The com-
pound LaNis absorbs and desorbs hydrogen reversibly at ambient pressures
and temperatures and it contains twice as much hydrogen per cm 3 of alloy
upon saturation as pure cryogenic liquid hydrogen. 12- 14 This property makes
LaNi 5 an attractive medium for hydrogen storage.
Previous calculations 12 .13 were based on setting v = w, and using
equation (B.16) for a large number of hydrides, but the results from equation
(B.15) should be preferred.
The pressure of H2(g) over the binary and ternary alloys can be
calculated by using .:lso for the reactions in which 1 mole of H2(g) is the
Table B.S. Calculated Values of Standard Enthalpies of

Formation AH; Used for Calculation of AHg(ABn H v+ w)
AH; (kJ/mole) AH; (kJ/g-atom H)
A AH 2 .5 AH2 AH1.5 B BnHw(n>w)
Sc -192 -185 -158 V -37

Y -205 -188 -153 Nb -59
La -185 -173 -135 Ta -54
Ti -125 -136 -120 Cr -7
Zr -197 -188 -160 Mo -3
Hf -165 -161 -138 W +8
Th -218 -196 -158 Mn -25
U -95 -96 -80 Tc +I8
Pu -123 -123 -103 Re +26
V -12 -34 -42 Fe +7
Nb -74 -83 -78 Ru +21
Ta -65 -75 -72 Os +24
Co +9
Rh +12
Ir +21
Ni +10
Pd +5
Pt +13
272 Appendix B
reactant, e.g.,
Hz(g) + (l/3)LaNi s = (l/3)LaNisH6 (B.17)
For PH,=l, LlGo=RT In PH,=O, and (l/3)[LlHg (LaNi s H6)]=
LlH~(B.17)= T LlS~(B.17) for this reaction as indicated by equation (B.17).
Examination of the available data \Z for a number of hydrides indicates that
LlS~ = -125 ± 25 J/mole Hz-K for reaction (B.17) and other similar reac-
tions, each with a single metal at ambient temperatures, and this value is
very close to the negative value of the standard entropy, So, of H 2 (g),
indicating that the standard entropies of LaNis and LaNis H6 are not greatly
different from each other. This value of LlS~(B.17) corresponds to
T LlS;(B.16)=LlH;(B.16)=-37.5±7.5 kJ/mole Hz. Therefore, for the
binary and ternary hydrides PH,> 1 atm for LlH;(B.17»-37.5 kJ/mole
Hz, and PH ,<l atm for LlH;(B.17)<-37.5 kJ/mole Hz, i.e., the more
exothermic hydrides are more stable as expected. In general, the plateau
pressure or the equilibrium pressure PH" which is the pressure at which
two solid phases LaNi s and LaNisH6 coexist with Hz, does not exactly refer
to reaction (B.17), because the LaNi s phase is not pure, i.e., it dissolves a
small amount of hydrogen, and the number of H atoms in LaNisH6 decreases
with increasing temperature. Nevertheless, these effects for many hydrides
are not too great, and from a knowledge of LlH;(B.17) and LlS;(B.17)=
-125 J/mole Hz-K, it is possible to estimate the equilibrium pressure PH,
for various binary and ternary hydrides. A more exact calculation is possible
when PH, at one temperature, T\, and LlH; (B.17) for the reaction are known
from which LlS;(B.17) can be calculated from LlGo=RT In PH,=
LlH;(B.17)-T\ LlS;(B.17) and then the resulting values of LlH;(B.17) and
LlS; (B.17) can be used as constants to calculate PH, at higher and lower
temperatures in a possible range of roughly ±200 K.
Ternary and Multicomponent Alloys
The enthalpies of formation of ternary alloys other than those with H,

e, and N may be computed by using equation (B.7) wherein LlH may be
reidentified as LlHAB . The resulting equation for LlHABC for the ternary alloy
is given by
(B.18)
This is expected because LlHAB in equation (B.7) is obtained from pairwise

interaction of atoms through their electrons, similar to the pairwise interac-
tion of atoms in the zeroth approximation to regular solutions, except that
instead of the ordinary atom fraction, the surface atom fraction x~ is used
in equation (B.7), and x~ refers to the ternary composition [see equation
(B.7)]. It is evident that x~ must be replaced by fj if the ternary alloys are

ordered. Extension of equation (B.18) to multicomponent alloys is then
self-evident.
Concluding Remarks
The enthalpies of formation of the binary alloys, t:..H, from equation

(B.7) with r/ p = 0 are for the nonordered alloys, as well as for the liquid
alloys. The effect of crystal structures and temperature on t:..H cannot be
accounted for since this would unduly tax the method. The parameters used
in this method have been adjusted and readjusted in some cases so that the
results prior to Ref. 3 do not always agree with the more recent improved
values. It is possible to refine the method by readjusting these parameters,
particularly r / p, to obtain refined fits to specific groups of alloys, e.g., binary
alloys of Ti with the transition elements, or Ti with alkaline earth elements.
The overall success of this method is very good considering the fact
that it covers nearly all the possible alloys, although some important and
interesting alloys of 0, S, Se, and Te are not included; however, the
procedure can be extended to these elements when experimental results for
typical alloys of these elements become available. This method is 95%
successful in predicting the sign of t:..H according to Phillips. 15 A comparison
of the available values of t:..H for 40 binary systems formed by AI, Ti, Hf,
Cr, Fe, Co, Ni, Nb, Mo, and W with the calculated values by the Miedema
method shows that the differences generally do not exceed 8 kJ / g-atom, the
largest difference being 17 kJ / g-atom within a range of +47 to -62 kJ/ g-atom
for t:..H according to Kaufman. 16 The disagreement is large for silicides and
germanides, and for relatively small atoms alloyed with large atoms, leading
to large coordination numbers. I ?-19
The second term, p(t:..cP )2, in equation (B.t) is quite similar to the term
-96.487 b ('I' A - 'l'B? used by Pauling? with his electronegativities, 'l'i, and
the effective number of bonds, b, which is not well-defined for alloys. In
fact, there is an approximate linear relationship between the electronegativ-
ity scales cP and '1', and the latter is closely related to other electronegativity
scales as discussed in detail by Bennett and Watson. 8 On the average, cP is
larger by a factor of 2.5 than '1', as can be seen by comparing 'I' in the
periodic table of Appendix G with those for cP in Tables B.l and B.2.
A form of the regular solution equation proposed by Hildebrand and
Scott20 contributes a positive term to the enthalpy of alloy formation, i.e.,
t:..E A)0.5 (t:..E B )0.5]2

t:..H(kJ/g-atom of alloy) = (XA VA + X B VB) [( VA - VB X~X~
(B.19)
274 Appendix B
where x:';' = XA VA/(XA VA + XB VB) is the volume fraction in terms of the

atomic volumes of pure metals VA and VB and their atomic fractions XA
and XB. The term AEA/ VA is the energy of vaporization or sublimation of
the pure metal in kJ / g-atom divided by VA in cm 3, and the square root of
this ratio is denoted by 8A, which is called the solubility parameter. For
example, at 1000 K, AEAI = 305.2 kJ/ g-atom, VAl = 11.3 cm3, AEAI / VAl =
27.0 kJ/ cm 3 = 8t and the solubility parameter is 8A = 5.20. In the simple
case of VA = VB, equation (B.19) assumes the familiar form AH =
NWABXAX B given by equation (4.5). Boom et al. 4 have shown that there is
a linear relationship between log(AEJ V;) and log(n~s) (i = metals A or
B); therefore, the positive term in equation (8.1) is related to equation
(B.19). The mathematical handling of equation (8.19) is simplified if the
coefficient (XA VA + XBVB) is approximated by (VA + V B)/2. In addition, the
Pauling term may be added to the coefficient of x:';'x~ to write
AH(kJ/g-atom) = [ 2 VB (8 A - 8 B)2] x:';'x~

-96.487 b ('I' A - 'l'B) 2 + VA +
(8.20)
as proposed by Mott. 2l A two-term equation for AH has also been proposed

by Shimoji and Niwa. 22 Further, Predel and Sandig23 have shown that the
positive contribution to AH in liquid alloys is due to the differences in the
atomic volumes which is related to the second term in equation (B.1),
because Vi and n~s are also related. However, some metals with very small
size differences form alloys that have large positive enthalpies of formation.
Numerous other correlations have been attempted to account for attractive
(or negative), and repulsive (or positive) terms in equations containing two
terms for AH. Earlier methods have been presented in an excellent mono-
graph by Prigogine et al. 24
The compositional variables in equation (8.20) are the volume fractions
instead of atomic or mole fractions. Comparison of equations (B.20) and
(4.9)-(4.14) shows that
(B.21)
where the equal sign determines the critical point Tc and the inequality is
for the existence of immiscibility in the binary system (cf. also Fig. 4.1).
Motel has found that for an overwhelming majority of binary liquid alloys,
equation (8.21) is valid; therefore, it can be used to predict the presence
or absence of immiscibility in liquid alloys for which phase diagrams have
not yet been determined. The values of b for this purpose range from 1 to
6 as tabulated and discussed in detail by Mott.
A few more comments based on the more recent publications are

appropriate. Minor changes in <p and nws have been recommended by
Niessen et al. 25 for Ti, Zr, Nb, La, Hf, Re, U, Cu, and Ag. For example,
the largest change is 4% in <p for Ti, U, and Re, and 8% in n!!s3 for La,
and 4% for Ti. It is significant that the values of nws are very closely related
to the values of the interatomic electron densities from the self-consistent
band structure calculations. 26 (See Ref. 31 with comments.)
A more recent attempt has been made 27 to improve the success of the
Miedema method for the binary terminal solid solutions of transition metals
having 4d electrons. For this purpose, an elastic term, and a structural term
are added to AH of liquid alloy formation. Further, the Miedema method
has been extended to the phosphorus alloys by Niessen,28 and it is possible
to devise similar methods for the alloys of oxygen and sulfur.
Watson and Bennett29 have developed a method for estimation of AH
for the equiatomic binary alloys of transition elements. According to
Gachon, Charles, and Hertz/o this method, on the average, agrees well with
the Miedema method.
References
1. A. R. Miedema, 1. Less-Common Met. 32, 117 (1973).
2. A. R. Miedema, R. Boom, and F. R. deBoer, 1. Less-Common Met. 41, 283 (1975).
3. A. R. Miedema, 1. Less-Common Met. 46, 67 (1976).
4. R. Boom, F. R. deBoer, and A. R. Miedema, 1. Less-Common Met. 45, 237 (1976),46,271
(1976).
5. A. R. Miedema and P. F. de Chatel, in Theory of Alloy Phase Formation, edited by L. H.
Bennett, Metall. Soc. AIME, p. 344 (1980).
6. E. Wigner and F. Seitz, Phys. Rev. 43, 804 (1946).
7. L. Pauling, The Nature of the Chemical Bond, Second Edition, Cornell University Press,
Ithaca, N.Y. (1960).
8. J. A. Alonso and L. A. Girifalco, in Theory of Alloy Phase Formation, edited by L. H.
Bennett, Metall. Soc. AIME, p.484 (1980); see also L. H. Bennett and R. E. Watson, in
Theory of Alloy Phase Formation, p. 390.
9. C. H. Hodges, 1. Phys. F 7, L247 (1977); see also Ref. 5, p. 503.
10. P. C. P. Bouten and A. R. Miedema, 1. Less-Common Met. 65, 217 (1979).
11. P. C. P. Bouten and A. R. Miedema, 1. Less-Common Met. 71, 147 (1980).
12. H. H. van Mal, K. H. J. Buschow, and A. R. Miedema, 1. Less-Common Met. 35, 65 (1974).
13. A. R. Miedema, K. H. J. Buschow, and H. H. van Mal, 1. Less-Common Met. 49, 463, 473
(1976). See also K. H. J. Buschow, H. H. van Mal, and A. R. Miedema, 1. Less-Common
Met. 42, 163 (1975).
14. J. H. N. van Vucht, F. A. Kuijpers, and H. C. A. M. Bruning, Philips Res. Rep. 25, 133 (1~70).
15. J. C. Phillips, in Theory of Alloy Phase Formation, edited by L. H. Bennett, Metal!. Soc.
AIME, p.330 (1980).
16. L. Kaufman, Calphad 1,300 (1977).
17. K. C. Mills, Thermodynamic Datafor Inorganic Sulphides, Selenides, and Tellurides, Butter-
worths, London (1974).
276 Appendix 8
18. O. Kubaschewski, High Temp. High Pressures 4, 1 (1972).

19. L. B. Pankratz, J. M. Stuve, and N. A. Gokcen, Thermodynamic Datafor Mineral Technology,
Bureau of Mines Bull. 677 (1984).
20. J. H. Hildebrand and R. L. Scott, Regular Solutions, Prentice-Hall, Englewood Cliffs, New
Jersey (1962); see also The Solubility of Nonelectrolytes, Third Edition, Reinhold, New
York (1950).
21. B. W. Mott, 1. Mater. Sci. 3, 424 (1968).
22. M. Shimoji and K. Niwa, Acta Metall. S, 496 (1957).
23. B. Predel and H. Sandig, Z. Metallkd. 60, 208 (1969).
24. I. Prigogine, A. Bellemans, and V. Mathot, The Molecular Theory of Solutions, North-
Holland, Amsterdam (1957).
25. A. K. Niessen, F. R. de Boer, R. Boom, P. F. de Chatel, W. C. M. Mattens, and A. R.
Miedema, Calphad 7, 51 (1983).
26. F. R. de Boer, R. Boom, and A. R. Miedema, Physica 1018,294 (1980).
27. A. R. Miedema and A. K. Niessen, Calphad 7, 27 (1983).
28. A. K. Niessen, High Temp. High Pressures 14,649 (1982); see also 1. Less-Common Met.
82,75 (1981).
29. R. E. Watson and L. H. Bennett, Calphad S, 25 (1981).
30. J. C. Gachon, J. Charles, and J. Hertz, Calphad 9, 29 (1983).
31. F. v. d. Woude and A. R. Miedema, Solid State Commun. 39,1097 (1980); Physica 8100,
145 (1980); A. R. Miedema, private communication. According to these references, the
most recent view is that the volume effects are also related to the differences in the
electronegativities.
C
Correlation of Thermodynamic Properties

in Dilute Solutions
Several empirical attempts have been made to correlate the excess partial
molar enthalpies and entropies of solution for metals, and the standard
enthalpies and entropies of reactions of gases with their dilute solutions in
metals. We shall limit this appendix to the correlation of (1) the excess
partial molar enthalpy, aft = H~, with the excess partial molar entropy,
S~, in dilute solutions of solute metals i in solvent metals and (2) aH· with
as· in equilibria of gaseous oxygen with dissolved oxygen in metals as
discussed in two recent publications. I ,2 Thermodynamic behavior of a solute
metal A in a solvent metal B dictates the processes for effective removal of
A from B in order to purify B. Further, dilute solutions of hydrogen,
carbon, nitrogen, oxygen, and sulfur as solutes generally play important
technological roles in purification of metals. Therefore, in the absence
of data, it is useful to estimate thermodynamic properties of such dilute
solutions.
Correlation of jj~ and S~ in Binary Metal-Metal Systems
A reasonable correlation of H7 and S7 in dilute binary solutions of

metals i in solvent metals requires experimental data for sufficient numbers
of systems so that the resulting correlation can be used to estimate S~ from
the available data for either H~ or G~. A preliminary attempt of this type
has recently been made by Kubaschewski. 1 The data selected for this purpose
by Kubaschewski from various sources 3 - 12 are listed in Table C.l and plotted
in Fig. c.l. The scale in Fig. C.l(b) is expanded for clarity in plotting the
values of IH~I smaller than those in (a). The system deviating strongly from
277
278 Appendix C
Table c.l. Values of H7 and S~ for Dilute Binary Solutions of Solute i in Metals a
H~, S~, H7/ S~,

Solvent Solute State T,K kJ/g-atom J/g-atom-K in K x 10- 3 Ref.
Ag Au 800 -20.3 -5.7 3.57 7

Ag Cd 673 -26.8 -6.25 4.3 7
Ag Ge 1250 -12.5 -2.7 4.6 7
Ag Ni +76. +19.7 3.85 12
Ag Pb 1273 +10.5 +8.9 1.2 7
Al Fe 1873 -128.5 -38.7 3.3 7
AI Sn 973 +24.7 +9.4 2.63 7
AI Zn 653 (+15.7) (+6.0) 2.6 7
AI Zn 1000 +10.6 +4.2 2.53 7
Au Ag 800 -16.9 -5.8 2.93 7
Au Ag 1350 -20.6 -9.05 2.3 7
Au Cd 700 -65. -34. 1.9 7
Au Cu 800 -11.6 +2.55 7
Au Cu -15.6 +5.5 7
Au Fe 1123 +25.5 +21.8 1.17 7
Au Fe 1473 +65. +53.5 1.2 7
Au Ni 1150 +21.5 +5.1 4.2 7
Au Pd -46.7 -12.6 3.7 8
Au Pt s -2.65 -14.6 8
Bi Ag 1000 (+13.4) (+3.0) 4.47 7
Bi AI 1173 +19.9 +4.23 4.7 7
Bi Cd 773 +2.85 +3.65 0.8 7
Bi Cu 1200 +23.5 +9.8 2.4 7
Bi In 900 -5.76 -0.57 7
Bi Na 773 -32. +35.4 7
Bi Pb 700 -3.6 +1.34 7
Bi Sn 600 Nearly ideal 7
Bi TI 723 -22.2 -0.12 7
Bi Zn 873 +13.6 +6.9 7
Cd Bi 773 (+8.5) +9.95 7
Cd Mg 543 -13.2 +3.35 7
Cd Mg 930 -20.6 -4.2 4.9 7
Cd Na 673 -37. -32.5 1.15 7
Cd Pb 773 +16. +6.9 2.3 7
Cd Sb 773 +5.15 +21.6 7
Cd TI 750 +12.8 +5.8 2.2 7
Cd Zn 800 +8.8 +1.0 7
Co Cr 1473 +2.8 +1.75 1.6 8
Co Ni s Nearly ideal 8
Co Si 1873 -147. -25.5 5.75 6
Cr a-Fe s +21.5 +8.1 2.65 7
DermodYDamic Properties iD Dilute SolutioDS 279
Table Col (continued)
ii~, Sr. fir/ sr.

Solvent Solute State T,K kJ/g-atom J/g-atom-K in K x 10-3 Ref.
Cu Ag 1423 +16.3 +1.35 7

Cu AI -36.3 +24.5 7
Cu Bi 1478 +30.2 +12.1 2.5 3.5.10
Cu Ni 5 773 +6.2 -3.2 8
Cu Ph 1473 +36.2 +10.7 3.38 7
Cu Sn 1400 -33.5 +16.8 7
Cu 11 1573 +28.2 +1.3 7
Cu Zn 5 773 -23.0 +5.35 7
Fe AI 1873 -64. -10.6 7
1-Fe C(gr) 5 1426 +44.2 +17.7 2.5 7
Fe Cr a-Fe 1550 +25. +10.2 2.45 7
Fe Cr I 2000 +23.5 +10.2 2.3 7
Fe Cu I 1823 +47.6 (+7.5) 7
Fe Mn 1-Fe 1450 -14.1 -13.3 1.05 7
Fe Ni 1-Fe 1200 -15.1 -5.45 2.8 7
Fe Ni I 1873 -10.0 -2.15 4.65 8
Fe Si I 1873 -132. -17.6 7.5 6
Ga Al 1023 +2.2 +0.9 2.45 7
Ga Cd 700 +12.9 +1.41 7
Ga Zn 750 +5.45 +3.55 1.53 7
Ge Ag 1250 +16.4 +8.8 1.85 7
Hg Bi 594 +5.75 -2.47 7
Hg Cd 600 -7.87 -2.35 3.35 7
Hg Cs 298 (-133.) (-90.) (l.48) 7
Hg In 433 -9.02 -1.5 6.0 7
Hg K 600 -104. -40.6 2.56 7
Hg Na 648 -83.0 -29.3 2.83 7
Hg Pb 600 +5.63 -6.07 7
Hg Zn 608 +3.0 -1.1 7
In AI 1173 +21.0 +2.28 9.2 7
In Bi 900 (-6.65) (+1.62) 7
In Cd 800 +4.7 +1.18 3.98 7
In Cu 1073 (+4.25) (+1.57) (2.7) 7
In Na 713 ( -17.6) ( -23.5) 7
In Sb 900 -12.3 +4.5 7
In 11 723 +2.9 -1.6 7
In Zn 700 +11.0 +4.3 2.56 7
Mg AI ( -9.9) ( -3.87) (2.56) 7
Mg Cd 5 543 -16.9 -3.2 5.28 7
Mg Cd 923 -23.4 -8.06 2.9 7
Mg In -34.5 -9.6 3.6 7
(continued)
280 Appendix C
Table C.I (continued)
fi~, Sf, fif/ Sf.

Solvent Solute State T,K kJ/g-atom l/g-atom-K in K x 10-3 Ref.
Mo C(gr) 2000 +143. +52.5 2.73 II

Mo Cr s 1471 +20.4 +3.1 6.57 7
Na K 384 +3.65 +0.9 4.05 7
Ni Cr -17.6 -2.83 6.2 8
Ni Co Nearly ideal 8
Ni Cu s +16.5 +4.7 3.5 8
Ni Fe -y-Fe 1200 -24.3 -5.0 4.85 8
Ni Fe I -32.3 -9.25 3.5 8
Ni Mn s 1050 (-60.) (-7.2) 7
Ni Si I 1873 -188. -23. 8. 6
Pb Ag 1273 +11.7 +3.27 3.58 7
Pb AI +26.4 -3.8 7
Pb Bi 700 -3.52 +1.1 7
Pb Cd 773 +9.35 +1.95 4.8 7
Pb Cu 1473 +27.5 +5.6 4.9 7
Pb Fe +103. +22. 4.7 12
Pb In 676 +4.0 +3.44 1.16 7
Pb K 848 -51.6 -18.8 2.75 7
Pb Mg 973 <-21. -3.9 (5.) 7
Pb Na 700 -38.5 -7.2 5.35 7
Pb TI s 523 -4.9 -6.95 0.7 7
Pb TI -3.45 -2.55 1.35 7
Pb Zn 923 +20.0 +4.54 4.4 7
Pd Cu s (-42.) ( -27.3) 7
Pd -y-Fe 1273 ( -75.) (+6.85) 7
Pd Ni 1273 (-11.4) (+4.8) 7
Pt Au (+24.8) (-3.8) 9
Pt Cu 1350 -28.8 +3.9 7
Pt Pd s -12.7 -7.1 9
Sb Ag 1250 ( +8.4) ( +11.) 7
Sb Cd 773 -3.0 +5.65 7
Sb Sn 905 -4.2 +2.75 7
Sn Ag 900 +4.4 +2.35 1.87 7
Sn AI 1000 +14.6 +6.65 2.2 7
Sn Cd 773 +6.6 +4.5 1.47 7
Sn Cr (+68.5) (+37.) (1.85) 12
Sn Hg 423 +1.85 -3.25 7
Sn In -0.85 -3.0 7
Sn Sb 905 -4.5 +2.43 7
Sn TI 723 +4.02 -1.38 7
Thermodynamic Properties in Dilute Solutions 281
Table C.I (continued)
H~, S~, ii~/ S~,

Solvent Solute State T,K kJ/g-atom l/g-atom-K in K x 10-3 Ref.
Sn Zn 750 +8.7 +6.0 1.45 7

11 Ag 975 +15.0 +4.0 3.75 7
11 K 798 -42. -11.5 3.65 7
11 Mg 923 (-11.7) (+10.) 7
11 Na 673 -33.5 -2.5 7
Zn AI 653 +25.9 +12.3 2.1 7
Zn AI 1000 +11.1 +2.95 3.77 7
·Solute and solvent in pure states have the same state of aggregation. Greatly uncertain values are in
parentheses. States of aggregation are denoted by I for liquid and s for solid.
the proposed correlation have been labeled as Au-Fe, Bi-Na, Cd-Au, and
so on. The values of ii~ and S~ calculated from Margules-type equations
for concentrated solutions and extrapolated to infinite dilution have been
disregarded. All the data are for the systems in which both the solute and
the solvent have the same state of aggregation. The arithmetic average of
ii~1 S~ = K, listed in the seventh column of Table C.l, was found to be
K = 3400 K and the line showing the empirical relation
(Jig-atom) (C.l)
is plotted in Fig. C.I. Substitution of this equation in G~ yields
- = -( 1--;;T) = -
G~ H~ S~(K - T); (Jig-atom) (C.2)
Therefore, if G~ is known, then both ii~ and S~ can be estimated, or if ii~

is known, then both S~ and G~ can be estimated from equation (C.2).
Equation (C.l) is approximate, and attempts to refine it by using other
methods in Appendixes A and B have not yet been made, although it might
well be an interesting subject to be pursued in the future.
Estimation of 4H· and 48· for Dilute Solutions of Oxygen in Metals
The equilibria between diatomic gaseous oxygen and dissolved oxygen

have been investigated in sufficient degrees of accuracy for a number of
282 Appendix C
Solution Melting point of solvent metal

<1450 K > 1450 K
+60
..
solid a (,,)
liquid CD 8
Note: symbols in brackets above denote +50 Au-Fe
the more uncertain value•.
+40
CD
Bi-Na +30
•
ID
+20 Au-Fe
CD
Cu-Sn
-120 -SO +SO + 120

ll.ii,/kcal (g alom)- ,
~
,..
e !:.
Fe-Si -20
e • :>:,
Co-Si
Au-Cd '"
CD
-30 .0
a;
ID Cd-Au
CD -40 ~
a
(J) ID
+20 Cd-Sb Au-Fe
eD
Cu-Sn
(J) +10 •
8
Pd-Fe
-SO -60 -40 eD
,
• (J)
+40 +60
• '"
Pd-Cu CD
Pt -Au 8<ll Pt-AI ll.ii;lkcal (g alom)-'
e -10
Fe-AI
8
I.
'"
_~I
Au-Pd _/1)
e.
:>:
-20
• a;
Cu-Pd
..
~"I
0
~
Cd-Au
(I)
b
Figure c.t. Correlation of jj~ and S~ for infinitely dilute binary solutions of i in solvent
metals. (~jji = jj~.) (From Kubaschewski' with permission.)
metals 2 ,13,14 such as Ag, Co, Cu, Fe, Ni, Pb, and Sn yielding I1H- and I1S-
of adequate accuracy to devise a method of estimation.
The usual method of equilibration is either with (H 2 + H 2 0) or (CO +
CO 2 ) mixtures (see Chapter 6) to obtain
• Xo··
O.50ig) = [0]; I1G o = -RT In rn- = I1Ho - TI1S o (C.3)
VP02
where the concentration of oxygen is expressed in atomic fraction Xo

(numerous publications use weight percent or atomic percent). The values
of I1H~ and I1S~ are assumed to be constant within the temperature ranges
of investigation. The results for the aforementioned metals are listed in
Table C.2 as linear functions of temperature. In general, the disagreement
among various sets of data for I1H~ and I1S~ is larger than that for I1G~
because I1G~ is the directly determined property.
Attempts have been made 16- 18 to correlate I1H- with I1Hrz98 of the
metal oxide expressed as the standard enthalpy of formation per gram atom
of oxygen. If a selected metal forms several oxides, then the oxide yielding
the largest value of I1Hrz98 is used for this purpose. The empirical correla-
tions proposed by Fitzner2 for I1H~ and I1S o for oxygen dissolved in metals
with full d-orbitals are
Table C.2. Standard Gibbs Energy Change for O.50 2 (g) = [0] in Selected Metals a
-aH~98' cal/g-atom [0]

aG;(exp), Range cal/g-
System cal/g-atom [0] for T atom [O]b <lG~(estm) <lG~(exp)' <lG~(estm)'
Ag-O -3,375 + 10.405T 1246-1433 7,420 -693 + 9.153 T +1l,192 +12,121

Co-O -14,730 + 4.241 Td 1823-1973 56,870 -19,929 + 4.622 T -6,672 -1l,147
Cu-O -20,710 + 10.871 T 1373-1773 40,800 -18,185 + 10.388 T -5,491 -3,642
Fe-O - 28,890 + 6.177 Td 1823-1973 65,000 -26,584 + 5.664 T -17,154 -15,822
Ni-O -1l,920 + 4.289 Td 1773-1973 57,300 -20,254 + 4.667 T -3,771 -1l,387
Pb-O -28,283 + 12.053 T 773-1473 52,340 -28,345 + 11.977T -1l,409 -1l,577
Sn-O -43,610 + 15.281 T 1173-1473 69,410 -45,737 + 15.709T -22,217 -23,744
aConcentration of oxygen is in atom fraction in pure metal.

b From Pankratz et al. ls
cFor Ag-D, Co-O, Ph-D, and Sn-D, numerical values are for 1400 K; remaining values are for 1900 K.
dExperimental results are from Tankins et al. ls for elements having unfilled d-orbitals; others, for elements with filled
d-orbitals, are from a summary by Chiang and Chang l4 where extensive references to the original investigations are given.
284 Appendix C
where -9.151 is for the change in the standard state from atomic percent
to atomic fraction, tlH~98 and tlH~ are in cal/ g-atom oxygen, and tlS~ is
in cal/g-atom-K. The equations for the transition elements in which the
d-orbitals are partially filled are
tlH~ = -5.252 x 1O-6(tlHf~98? + 1.6 x 1O-11(tlH~98)3; (unfilled d) (C.6)
tlS~ = -3.151 - 4.141 x 1O-9(tlH~)2 - 2.2 X 1O-14(tlH~)3; (unfilled d) (C.7)
The difference between equations (C.5) and (C.7) is 6.0 cal/ g-atom-K, which
must be considered as an empirical term because theoretical justifications
for this value are tenuous. The numerical values obtained from equations
(C.4)-(C.7) are listed in Table C.2, indicating that the experimental and
estimated values of tlG~(exp) and tlG~(estm) are in good agreement, except
for the Ni-O system. The estimated equations for 18 transition metals are
listed by Fitzner/ whose entropy term must be changed as follows to bring
his results in line with the notation used in this appendix: tlSE(Fitzner)
-9.151 = tlS~(this appendix). The estimated results for tlG~ of oxygen in
most of the transition metals are probably within 15 kcal/ g-atom[ 0] of the
experimental value for tlG~. However, the available data for some of these
metals are scanty and greatly scattering. Therefore, new experimental data
are needed for oxygen in the transition metals to test the foregoing empirical
equations and, if necessary, to modify them to attain greater degrees of
success for prediction of thermodynamic properties of dissolved oxygen.
The method may also be extended to other interstitial solute elements such
as H, C, N, and S.
References
1. O. Kubaschewski, High Temp. High Pressures 13,435 (1981).

2. K. Fitzner, Thermochim. Acta 52, 103 (1982).
3. T. Azakami and A. Yazawa, Can. Metall. Q. 15(2), 111 (1976).
4. C. Bernard and I. Ansara, Report LTPCM-1974-TM02, Ecole Normale Superieure d'Elec-
trochimie et d'Electrometallurgie, St.-Martin d'Heres, France (1974).
5. J. Bode, J. Gerlach, and F. Pawlek, Erzmetall 24, 480 (1971).
6. T. G. Chart, High Temp. High Pressures 5, 241 (1973).
8. O. Kubaschewski and C. B. Alcock, Metallurgical Thermochemistry, Fifth Edition,
Pergamon Press, Elmsford, New York (1979).
9. O. Kubaschewski and J. F. Counsell, Monatsh. Chem. 102, 1724 (1971).
10. B. Predel and A. Emam, Z. Metallkd. 64, 496 (1973)
11. L. L. Seigle, C. L. Chang, and T. P. Sharma, Metall. Trans. AIME lOA, 1223 (1979).
12. D. A. Stevenson and J. Wulff, Trans. Metall. Soc. AIME 221,271 (1961).
13. See, e.g., E. Fromm and E. Gebhard, Gase und Kohlenstoff in Metallen, Springer-Verlag,
Berlin (1976).
14. T. Chiang and Y. A. Chang, Metall. Trans. AIME 7B, 453 (1976).
15. E. S. Tankins, N. A. Gokcen, and G. R. Belton, Trans. Metall. Soc. AIME 230, 820 (1964).
16. F. D. Richardson, 1. Iron Steel Inst. London 166, 144 (1950).
17. K. Fitzner, K. T. Jacob, and C. B. Alcock, Metall. Trans. A/ME 88, 669 (1977).
18. L. B. Pankratz, J. M. Stuve, and N. A. Gokcen, Thermodynamic Data/or Mineral Technology,
Bureau of Mines Bull. 677 (1984).
D
Selected Properties of the Elements

N
Condensed phase, gg
298.15 K Phase transitions Ideal gas, 298.15 K
Atomic
Element weight Cop So Tm IlH;" Tv IlH~ Cop So IlHfo IlGfo
Ac 227.0278 6.5? 1324 3480? 4.975

Ag 107.8682 6.070 10.17 1235.08 2.70 2440 59.9 4.968 41.321 68.090 58.802
AI 26.98154 5.82 16.776 933.602 2.580 2798 70.1 5.112 39.302 78.8 69.102
Am 243? 6.178 13.023 1449 3.440 4.968 46.473 67.90 57.927
Ar 39.948 83.798 2.81 87.29 1.558 4.968 36.983 0 0
As 74.9216 5.892 8.534 876 6.621 876 4.968 41.611 72.12 62.258
At 21O? 575?
Au 196.9665 6.075 11.330 1337.58 2.957 3130 80.0 4.968 43.116 87.5 78.023
B 10.81 2.65 1.410 2365? 12.0 4275 117.0 4.971 36.649 132.80 122.293
Ba 137.33 6.713 14.918 1002 1.852 2171 33.8 4.968 40.663 43.0 35.324
Be 9.01218 3.93 2.27 1562 2.919 2745 69.9 4.968 32.545 77.40 68.374
Bi 208.9804 6.10 13.56 544.592 2.700 1837 36.200 4.968 44.669 49.5 40.225
Bk 247? 1256
Br 79.904 18.09 36.379 265.90 2.527 332.6 7.065 8.616 58.64 7.388 0.751
C 12.011 2.036 1.372 4130 4.981 37.76 171.29 160.441
Ca 40.08 6.05 9.94 1113 2.04 1757 36.626 4.968 36.992 42.5 34.434
Cd 112.41 6.20 12.38 594.258 1.48 1040 23.809 4.968 40.066 26.73 18.475
Ce 140.12 6.44 17.2 1071 1.305 3700 99.0 5.515 45.807 101.0 92.471
Cf 251? 1213
CI 35.453 172.18 1.531 239.10 4.878 8.111 53.29 0 0
Cm 247? 6.617 1613 3.500 6.717 47.158 92.60 83.668
Co 58.9332 5.93 7.18 1768 3.7 3200 90.0 5.503 42.879 101.5 90.856
Cr 51.996 5.58 5.65 2133 4.047 2945 82.3 4.968 41.635 95.0 84.271
Atomic weights are from Pure Appl. Chern. 55, 1101 (1983). Melting points are from Bull. Alloy Phase Diagrams 2(1),146 (1981) where references to original sources
...,
...,>
are given. Remaining data are from Pankratz et 01., Thermodynamic Data for Mineral Technology, Bureau of Mines Bull. 677 (1984), and from R. Hultgren, P. D.
Q,
..=
Desai, D. T. Hawkins, M. Gleiser, K. K. Kelley, and D. D. Wagman, Selected Values of the Thermodynamic Properties of the Elements, ASM, Metals Park, Ohio
(1973), corrected for 1968 temperature scale. Melting points of C and Fe are author's values. C; and So are in cal/mole·K; !lH"", is in kcal/mole. !lHfo and !lGfo ><'
refer to formation of ideal stable gas from stable cOHdensed phase. For H, N, 0, and halogens, the gas phase is diatomic; all others are monatomic. t::'
Cs 132.9054 7.695 20.37 31.54 0.50 952 16.198 4.968 41.942 18.32 11.888 rn
!!.
Cu 63.546 5.841 7.924 1358.62 3.12 2839 71.9 4.968 39.743 80.60 71.113 ~
Oy 162.50 1685 ;-
co.
Er 167.26 1802 "CI
Es 2521 1093 a
'1:1
Eu 151.96 1095
..
F 18.9984 53.48 0.122 85.02 1.562 7.481 48.443 0 0 i·
<II
Q
Fe 55.847 5.970 6.52 1809 3.30 3135 83.6 6.136 43.112 99.3 88.390 ...
Fm 2571 18001 1;
Fr 2231 3001
..
l""l
;'
Ga 69.72 6.250 9.758 302.9241 1.336 2478 61.8 6.058 40.375 66.2 57.072
Gd 157.25 8.860 16.24 1586 2.403 3540 85.9 6.584 46.416 95.0 86.003 II
.a
Ge 72.59 5.58 7.43 1211.5 8.83 3107 79.1 7.345 40.104 89.5 79.758 f:
H 1.00794 13.81 0.028 2039 0.2158 6.892 31.207 0 0
He 4.00260 3.5 0.005 4.22 0.020 4.968 30.125 0 0
Hf 178.49 6.150 10.41 2504 5.68 55001 1581 4.972 44.643 148.0 137.793
Hg 200.59 6.687 18.14 234.314 0.549 629.81 14.151 4.968 41.792 14.67 7.618
Ho 164.9304 1747
I 126.9045 13.011 27.758 386.7 3.709 458.4 10.021 8.814 62.277 14.919 4.627
In 114.82 6.389 13.82 429.784 7.80 2346 55.3 4.979 41.508 58.15 49.895
Ir 192.22 5.996 8.481 2720 6.247 44001 134.71 4.968 46.241 159.0 147.742
K 39.0983 7.050 15.457 336.34 0.558 1043.7 19.038 4.968 38.297 21.31 14.500
Kr 83.80 115.765 0.391 119.75 2.158 4.968 39.191 0 0
La 138.9055 6.48 13.6 1191 1.481 3730 98.9 5.438 43.564 103.0 94.066
Li 6.941 5.887 6.954 453.7 0.717 1638 35.160 4.968 33.143 38.410 30.602
Lr 2601 19001
Lu 174.967 6.40 12.18 1936 4.457 4.986 44.142 102.2 92.671
Md 2581 11001
Mg 24.305 5.950 7.810 922 2.139 1363 30.250 4.968 35.501 35.00 26.744
Mn 54.9380 6.280 7.650 1519 2.882 2335 54.0 4.968 41.493 67.10 57.010
Mo 95.94 5.750 6.850 2896 7.777 49191 140.81 4.968 43.461 157.50 146.584
- - - - - - - - ----- - --- -
(continued) :
N
Condensed phase, ~
~
298.15 K Phase transitions Ideal gas, 298.15 K
Atomic
Element weight C·p S· Tm t.H:;' Tv t.H~ Cop S· t.Hf o t.Gf·
N 14.0067 63.1458 0.172 77.36 1.335 6.961 45.770 0 0

Na 22.98977 6.730 12.298 371.0 0.622 1177 23.285 4.968 36.714 25.755 18.475
Nb 92.9064 5.880 8.70 2742 6.302 5200? 163.0 7.208 44.490 175.2 164.529
Nd 144.24 1294
Ne 20.179 24.563 0.08 27.07 0.422 4.968 34.947 0 0
Ni 58.69 6.210 7.140 1728 4.100 31.87 88.5 5.583 43.519 102.80 91.954
No 259? 1100?
Np 237.0482 910
0 15.9994 54.361 0.106 90.19 1630 7.021 49.005 0 0
Os 190.2 5.905 7.8 3306 7.6 5285 178.32 4.968 46.00 189.0 177.611
P(wh) 30.97376 5.698 9.820 317.29 0.157 550 2.908 4.968 38.980 75.620 66.926
Pa 231.0359 6.601 12.400 1848 2.950 5.476 47.308 145.0 134.592
Pb 207.2 6.370 15.490 600.652 1.147 2023 42.5 4.968 41.890 46.750 38.879
Pd 106.42 6.188 9.013 1828 3.970 3237 85.4 4.968 39.902 90.4 81.190
Pm 145? 1315
Po 209? 527
Pr 140.9077 6.560 17.670 1204 1.646 3785 70.9 5.105 45.339 85.0 76.750
Pt 195.08 6.180 9.950 2042.1 4.696 4100 121.85 6.102 45.960 135.1 124.364
Pu 244? 7.850 13.420 913 0.675 3503 82.14 4.984 42.317 82.500 73.884
Ra 226.0254 973
Rb 85.4678 7.424 18.35 312.63 0.524 974.5 17.23 4.968 40.626 19.33 12.688
Re 186.207 6.035 8.730 3459 7.9 5869 170.85 4.968 45.131 184.00 173.147
Rh 102.9055 5.958 7.542 2236 6.356 3970 117.89 5.022 44.387 133.10 122.115
Rn 222? 202
Ru 101.07 5.730 6.839 2607 9.200 4423 142.34 5.144 44.550 153.6 142.356 '"C:I
-=.>
:0
S 32.06 5.425 7.661 388.37 0.413 5.658 40.086 66.20 56.533
~.
Sb 121.75 6.030 10.880 903.905 4.750 4.968 43.059 62.70 53.106 '"
C
Sc 44.9559 1814
[FJ
Se 78.96 5.987 10.144 494.3 1.472 4.990 42.210 56.25 46.689
Si 28.0855 4.780 4.50 1687 12.082 5.318 40.121 107.70 97.080 ....
Sm 150.36 1347 ~
co.
Sn(wh) 118.69 6.450 12.236 505.118 1.680 2876 70.7 5.081 40.244 71.99 63.639 "C
Sr 87.62 6.393 12.500 1042 1.960 1654.1 32.730 4.968 39.323 39.20 31.203 'r::I
Ta 180.9479 6.060 9.920 3293 7.560 5731 177.61 4.985 44.242 186.90 176.667
..a
::l
~.
1b 158.9254 6.91 17.52 1629 2.580 3496 79.1 5.895 48.552 92.90 83.648
0
Tc 98? 8.0? 2477 4.970 43.248 164.5? 153.38? ...
Te 127.60 6.140 11.880 722.72 4.180 4.970 43.640 50.60 41.131 ........
Th 232.Q381 6.532 12.760 2031 3.30 5061 122.96 4.969 45.425 142.73 132.991 ~
Ti 47.88 5.980 7.320 1943 3.30 3562 100.6 5.839 43.066 113.00 102.342
...a
Tl 204.383 6.290 15.340 577 0.990 1746 39.215 4.968 43.226 43.25 34.936
1:
"
Tm 168.9342 6.460 17.690 1818 4.025 2220 45.6 4.968 45.412 55.50 47.235
U 238.0289 6.612 12.000 1407 2.185 4407 110.92 5.663 47.724 127.00 116.349
V 50.9415 5.950 6.915 2202 5.461 3694 106.8 6.217 43.544 123.20 112.279
W 183.85 5.800 7.800 3695 8.5 5828 196.92 5.092 41.549 203.40 193.338
Xe 131.29 161.3918 0.55 165.Q3 3.02 4.968 40.530 0 0
y 88.9059 6.34 10.62 1795 2.724 6.181 42.869 100.7 91.085
Yb 173.04 6.39 14.30 1092 1.830 1467 30.800 4.968 41.352 36.35 28.284
Zn 65.38 6.070 9.950 692.73 1.750 1180 27.565 4.968 38.451 31.17 22.672
Zr 91.22 6.186 9.320 2128 5.000 4682 139.11 6.367 43.315 148.30 138.164
...~
E
Selected Binary Phase Diagrams and Binary

Thermodynamic Properties
The binary phase diagrams selected 1•2 in this appendix are for convenient
reference in pursuing the topics covered in this book. The objective is to
present a representative number of various types of diagrams with various
types of phase equilibria.
Selected thermodynamic data from Hultgren et ae are for the excess
gram atomic (or molar) enthalpy and Gibbs energy of formation from which
the corresponding partial molar properties and activities can be calculated.
The procedure yields values as accurate as those listed by Hultgren et al.
at the composition intervals of 0.1 if a power series of the type gIVen by
equation (1.63) with five terms is used for He and for se, and the results
are then combined in G e = He - TSe to obtain five terms dictated by the
data for G e • We illustrate the procedure by using only two parameters:
(E.!)
Substitution of two selected values of He at the corresponding compositions

yields the values of A2 and A 3 • The equation for Se is similar, i.e.,
(E.2)
Substitution of equations (E.!) and (E.2) in G e = He - TSe gives
Next, two values of G e at the corresponding compositions yield the

coefficients in equation (E.3) and since T is known, then C 2 and C3 can
293
294 Appendix E
be calculated. The use of equations (1.64), (1.66), and (1.68) yields
(E.4)
(E.5)
The same procedure also yields the equations for G~ and G~, as well as S~
and S~.
WEI GHT PERCENT COPPER

5 10 15 20 30 40 50 60 70 80 90
I I I I I I I I I I
1135~
1300
1200 ~
1234- L
./
V I
"\~ I"'----.. ./
~
V
(Cu)
"
1100
(Ag) ;0.141
~ ...-'" 1052· ",
0.399 0.951\
1000
/ \
900
800 I
700
I
Ag 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Cu
XCu
XCu: 0.1 0.3 0.5 0.7 0.9, at 1423 K

HO: 455 921 1014 829 352 }
GO: 309 calf g-atom
305 701 841 714
Figure E.1. Ag-Cu system.

Selected Binary Phase Diagrams 295
WEIGHT PERCENT COPPER

20 40 60 70 80 90 95
1 I I I I I
~;:;t.
130n 13'--
L
It{)
r-X
1200
110 0
1; ~ p
\.-
1000 / Y2 (Cu)
~33.25°
/
V
897-
Erj
900 / 864·
80 0
\8~
0.025 .0.17.3.
I--- - e.~~T ~
8 ~~ O.BOE
I'\(AI) I~I
~8 TJi
70 0 0.671
.60.. 636-
~2 0.504
60 0
~
500
AI 0.1 0.2 0.3 OA 0.5 0.6 0.7 0.8 0.9 Cu

XCu
XCu: 0.1 0.3 0.5 0.7 0.9, at 1373 K
He: -459 -1445 -2163 -1979 -800}
G8 : -836 -2289 -3278 -3240 -1474 cal/g-atom
Figure E.2. AI-eu system.

296 Appendix E
WEIGHT PERCENT ZINC

10 30 50 60 70 80 90 95
'" -------
I I I I I
9~3~~O I
II
900 ~
----
I--- L
---
800 ~
r-- I---
r--
692.6550
700
..........
655 0 ~~
(AI), 624.6 0 {All z /0.665 0.887"'" p.976\
600 ,../
0.395 -.....
(Zn)~
.....- ~/ 548"
,/ 0.594 0.986
500
400 I
(
I
I
300
200
Al 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Zn

XZn
xZn: 0.1 0.3 0.5 0.662* 0.978*, at 653 K

He: 326 715 825 839 136}
94 cal/g-atom
Ge : 251 549 620 578
XZn: 0.1 0.3 0.5 O} 0.9, at 1000 K

He: 225 516 614 526 234}
Ge : 144 344 388 313 154 cal/g-atom
Figure E.3. AI-Zn system. Asterisks indicate phase boundary.

Wf. "10 Go
10 ~O 30 40 SO 60 70 8f' 9.0
00
GoA$
1~38·
UqJd
I~ 00
~
/V ~
Vr
\
\
10 00
817 GOA
~
8 00
• 810·
6 00
I
A"" GoA.r Liq. + Ga.At
\
4 00
00
GaA$~Go ~9'5.
_1- I
o 10 ~O 30 40 SO '070 80 90/00
Ar.". Ga
Figure E.4. As-Ga system.
298 Appendix E
WEIGHT PERCENT PLATINUM

10 20 30 40 50 60 70 80 90
II II II II (I
L 20~
2000
,,'
- - l--- I---
.... 7
~ .... ,,
1800
V"'"
/"
/::---
/'
1600
~
1400
V - 1~2S· -
0.60 r---......
r1336.15. (Au, Pt)
1200 V '\
Au 0.1
/
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
\
Pt
XPt
xPt: 0.11* 0.20 0.30 0.42* 0.79* 0.80 0.90, at 1423 K
G8 : 572 863 1121 1368 1123 1084 657, calfg-atom
Figure E.5. Au-Pt system. Asterisks indicate phase boundaries; data are for single solid phase
region with solid Au and Pt as standard states.
WEIGHT PERCENT TELLURI UM
OK ,
10 20
I
30 40 50 60 70 80 90
",,1364-
1300
L ,/
/
1200
V
......V ~
/ ~
'" \
1100
1000
I {l--
\
900 i\
800
722" \
700 1'0.99-
597· (Te)-
600
594.18-
f-(Cd)
Cd 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Te
X Te
Figure E.6. Cd-Te system.
300 Appendix E
WEIGHT PERCENT ZINC

(0 20 30 40 50 60 70 80 90
I I I I
500 I
450H/'---+--+--t-----i:--_+--+--t---+--+_---1
400~-4-_+-~-~-+--+--~_;-_+-~
Cd 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0:9 Zn

XZn
XZn: 0.1 0.3 0.5 0.7 0.9, at 800 K

He: 185 421 500 430 192} cal/g-atom
192
G e: 171 399 478 414
Figure E.7. Cd-Zn system.

10 30 60 70 80 90
"'- --
~
~S·
ISOO
-
t;:,.
1400
b/' // I
/
~LiqUid-
I
I
1\
,,,
\
IJOO
« + Liquid
\
1200 1-1'
I
I
\sl
1100 I
I 083 0
1000
900
1 + e
\
800 a
~
--- --- - - - 1----
~agl;;(l ~/;;;,,~~ - - --- --- ---
700
600
aL
SOO
0 10 20 30 40 SO 60 70 80 90 100
Wt.% Cv
X Fe : 0.1 0.3 0.5 0.7 0.9, at 1823 K

He: 804 1789 2132 1917 950} cal/g-atom
Ge : 712
735 1589 1837 1581
Figure E.S. eu-Fe system.

302 Appendix E
WEIGHT PERCENT NICKEL
--
10 20 30 40 50 60 70 80 90
-- -
II I II I I II II 1726"
--
L ~ ~-
1600
L-- ~
-- -- - -~
1400
~56.55°
1200
{Cu, Nil
1000
....... (Cul
800
6310
600
./
-. /
400 -
CURIE TEMP.
/
./
200
/'/
/'
Cu 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Ni

XNi
X Ni : 0.1 0.3 0.5 0.7 0.9

He: 74 265 425 449 232}
295 call g-atom at 973 K
Ge : 213 527 673 622
Ge : 229 596 725 609 261, at 1823 K
Figure E.9. Cu-Ni system.

At. % Sn
·c /0 cO ,JO 40 SO 60 70 80 90 100
1100
~
\\ \
1000
900
800
1\ 798.1\
700
\ "'~ P ~
7SS'
~
Liquid
"I / ~ 640~
600 <X
6' \) -I~ f4 ~
~
~\;f~
'r-SJ £ I.iquid ~
""-
SeO ~
SOO
L 41S·
"'" r\
\
400
/ lSO'
/
! \
,JOO
1J ee7'
cOO 189' 186°-
:
ISO 1'-17' 1
o 10 cO ,JO 40 SO 60 70 80 90 100
Wt % Sn
XSn: 0.1 0.3 0.5 0.7 0.9, at 1400 K

He: -666 -934 -475 -97 +52} calfg-atom
GO: -1018 -1495 -1238 -784 -291
Figure E.IO. Cu-Sn system.

304 Appendix E
Wt%Si
·c S 10 IS ..0 "S JO 40 SO 60 10 80 90
1600
L/fIUld FJ..Si FO~rj Fflj' I. FoS;"IU FOS,.. IJ}
ISOO "- I I 1 I
~ ~ ~
~~ + ~
1410°
\'\ /
1400
V
\
"'" \'' ii'O./ '.
I\. /'
1275°
IJOO
/ ~ IS-,L
..... I 02 d 1212° 08'
"'. t
, .. 00
r-£'9f
1/00
Y
1/ '\ I-Fo.. S,
1'090·
,-FoS, f-- Fo 2IH}
1000
II J '07rl ~FoSSiJ
r2!L
I 99~
I at 1/; 0&,
960"
900
1/ I 82S. FoS,.. IL}

800
700
600
I--
,,
S40· \
,
SOO
\
\
400 \
o 10 ..0 JO 40 SO 60 70 80 90 100
At%Si
0.1 0.3 0.5 0.7 0.9, at 1873 K

He: -3079 -7987 -9049 -6662 -2441 }
Ge : -2319 -5511 -5745 -4005 -1474 caljg-atom
Figure E.Il. Fe-Si system.

WEIGHT PERCENT ANTIMONY

10 2) 30 40 50 60 70 80 90
I I I I I I I I
,.....904
L
800· /
~
800 773
".--
700 V 0.683
600
'/
500 I
t,..42S.76°
'r-I'
428-
400
:.... (In) (Sb)'"
In 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Sb

XSb
XSb: 0.1 0.3 0.5 0.7 0.9. at 900 K

He: -283 -718 -769 -569 -211}
-257 cal/g-atom
GO: -364 -848 -928 -694
Figure E.12. In-Sb system.

306 Appendix E
WEIGHT PERCENT TIN

10 30 50 60 70 80 85 90 95 98
1043.7·
/ ' ~r--..
1000
/
""
922 0 L
/
'"
900 t...
i\ ""V 834.4°
800 \11 o.
K~·0345
(Mg) .~
-
700
{J
I""
"
600
500 476.7- 5OS.06

CSn)-
Mg 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Sn
XSn
0.1 0.3 0.5 0.7 0.9, at 1073 K

-1607 -3291 -3385 -2410 -852. cal/g·atom
Figure E.n. Mg-Sn system.

WEIGHT PERCENT ANTIMONY

10 20 30405060708090
900 ."...,.
L
~
~ !
800
./
/ I
700 I
V (Sbl-f
SOO.s- ./
or
600
~V 524.4- 1
500
~o.o58
(P1bl
\
\
400
Pb 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Sb
XSb
0.1 0.3 0.5 0.7 0.9, at 905 K

0 -8 -16 -12 -2} cal/g-atom
-41 -95 -112 -94 -40
Figure E.14. Pb-Sb system.

308 Appendix E
WEIGHT PERCENT TIN

10 20 30 40 50 60 70 80 9:)
I I I I I I I
600.6"
r-- r--.
600
"'"
'"
L
.............
~
i'-..... ~ 5)5.06
500
\ 456 0 ~ ~ .- ~ u(.~
(Pb)
V 029
0.74 0.985
If
400
Pb 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 09 Sn

XSn
XSn: 0.1 0.3 0.5 0.7 0.9. at 1050 K

He: 130 285 327 274 120 }
156 calfg-atom
GS : 323 604 583 416
FigureE.15. Pb-Sn system.

WEIGHT PERCENT ZINC

5 10 15 20 30 40 506070 90
OK 0.72 10153 0
-.....
,
~
~ ~
iOOO
\
V
900
L
V
6"9. J/,
800
700 II 690.81·
692.6S~~
III~ISI 591.J.
~9977
I
600 I
0.022.
(Z7)-
500 f--CPbl
Pb 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Zn
XZn
0.1 0.22* 0.97. * at 923 K

455 918 168}
359 716 199 cal/g-atom
Figure E.16. Pb-Zn system. Asterisks indicate phase boundary.
References
1. C. 1. Smithells and E. A. Brandes, Metals Reference Book, Sixth Edition, Butterworths,

London (1983). The diagrams for As-Ga, Cu-Fe, Cu-Sn, and Fe-Si are reproduced from
this source with permission.
2. R. Hultgren, P. D. Desai, D. T. Hawkins, M. Gleiser, and K. K. Kelley, Selected Values of
the Thermodynamic Properties of Binary Alloys, ASM, Metals Park, Ohio (1973). The
remaining 12 diagrams in this appendix are reproduced from this source with permission.
F
List of Symbols
1. Latin Symbols
(Symbols not defined here are empirical constants usually used on only one
or two pages.)
A Helmholtz energy (A = E - TS)

A amperes; used only after a number or a prefix
A 12 , A l3 , •.• empirical coefficient, equations (1.70)-( 1.72)
A,B, ... components A, B, ...
Aq empirical coefficient, equations (1.64), (1.65), (1.67), (1.68)
AGN average group number in alloys. For Ti- V equiatomic alloy,
AGN = 4.5, since Ti = 4 and V = 5
a, activity of component "i" based on pure "i" as the standard,
equation (1.31); fJ for interstitials in Chapter 6
activity of "i" on Henrian scale, equation (1.36)
metal B
empirical coefficient in equation (1.66), cf. equations (1.67)
and (1.68)
b empirical correction term, equation (4.28)
bcc body-centered cubic structure
C solar intensity in number of suns, equation (7.71); used in
this context only in Chapter 7
heat capacity; C v and Cp , heat capacities at constant volume
and pressure, respectively; equations (1.4) and (1.6)
coulomb; used only after a number or a prefix
temperature in Celsius. (Although use of the degree sign is
not currently recommended, it is used in this book to avoid
confusion with other C.)
311
312 Appendix F
c empirical correction term, equation (4.27)

c number of interstitial sites per solvent metal atom in Chapter
6; c = 1 for bcc, C = 3 for close-packed crystals
c subscript for number of components in Chapter 2
(C) crystalline state or phase
¢ bulk modulus on page 257
cph close-packed hexagonal structure
D actual number of configurations
De corrected distribution of particles, equations (4.27) and (4.29)
D rm random distribution of particles; D a, Db, and Dr are defined
on page 132
d total differential
d, S,p atomic orbitals (shells) in Appendix A
E molar energy of a system
Ee energy level of conduction band in semiconductors, Chapter 7
Er Fermi energy, Chapter 7
Eg band-gap energy, Eg = Ee - E y , Chapter 7
Ey energy level of valence band, Ey = 0, Chapter 7
eij bond energy for each bond between i and j atoms
eo electron on valence band edge
e - ,e + electron in conduction band and hole in valence band, respec-
tively
eV electron volts
F fugacity defined by equation (1.16)
Ft fugacity over pure condensed" i"
I activity coefficient defined by equation (6.2)
j; activity coefficient of "i" on Henrian scale, equation (1.36)
I! gamma function, i.e., factorials offractional numbers, equation
(4.26)
I(x A , V) function defined by equation (B.1)
fcc face-centered cubic structure
fr, frz subscript for freezing
o any intensive property; molar Gibbs energy (0 = H - TS)
G any extensive thermodynamic property; extensive Gibbs energy
OJ partial molar Gibbs energy of component "i," equation (1.22)
o· standard molar Gibbs energy on Henrian scale, equations (1.34)
and (1.35)
standard molar Gibbs energy, based on pure component "i,"
i.e., Raoultian scale
o~ standard Gibbs energy of valence band electron
g distribution of bonds, equation (4.65); g' is given by equation
(4.63)
List of Symbols 313
g defined by equation (5.39)

gi degeneracy of ith level
g(XA, nws) function defined in equation (B.1)
H molar enthalpy (H = E + PV)
He excess molar enthalpy, identical with tlH of formation of 1
gram atom of alloy from its pure components, e.g.,
xA(pure) + (l - x) B(pure) = A - B (phase with x atom frac-
tion of A)
partial molar enthalpy of Hi", equal to Fr
Planck's constant
moment of inertia, equation (3.61)
current density, Chapter 7
line current in a solar cell circuit, Chapter 7
reverse saturation current for a diode, Chapter 7
short-circuit current in solar cells, Chapter 7
rotational quantum number in Chapter 3 only
electron flux density, Chapter 7
temperature in kelvins, follows numbers and T
equilibrium constant
Boltzmann's constant, i.e., gas constant per particle, R/ No
liquid in phase diagrams
vibrational quantum number only in Chapter 3
liquid after atomic and compound symbols, and thermo-
dynamic properties
In natural logarithm, base e
log common logarithm, base 10
M molecular weight or mass
M Maedelung constant
m mass of a particle
m meter, used only after a number or a prefix
m subscript for melting, i.e., fusion
(m) geometrical mean mass of carriers equal to (m_ m+ )0.5
effective masses of electrons and holes, respectively, Chapter 7
atomic masses of i and j
reduced mass, mr = mimj / (mi + mj )
number of particles in a system; Avogadro's number in
Chapter 4
densities of quantum states for electrons and holes, respectively,
Chapter 7
occupation number of particles for state i in Chapter 3; number
of atoms of component i
Avogadro's number; sometimes used as N
314 Appendix F
n nano; used only as a prefix after a number

n quantity defined by equation (4.38); n > 0; ng is defined by
equation (4.73)
n number of interstitial atoms in Chapter 6
nj number of interstitial atoms in Chapter 7
nws Wigner-Seitz electron density in electrons per cm 3 , Appendix B
n_,n+ numbers of electrons and holes per cm 3 of a semiconductor
P partial pressure of H2(g) in Chapter 6 only
P,Pj pressure in atm (1 atm = 101, 325 Pa); Pj, partial pressure of i
Pt vapor pressure of pure" i"
p empirical parameter in equation (B.1) in Chapter 7 only
p-,p+ momentum of electrons, holes
Q configurational partition function in Chapter 4, equation (4.32)
Q partition function fOl\ canonical ensemble, Chapter 3
q. dissipative energy, i.e., heat, equation (1.3)
q partition function for microcanonical ensemble, equation (3.23)
q empirical parameter in equation (B.1)
q empirical coefficient in Appendix B
q-,q+ partition functions in Chapter 7
q,h thermal energy exchange, equation (3.30)
R gas constant
RD dynamic impedance, equation (7.61)
maximum power resistance for a solar cell circuit, Chapter 7
,
Rmp
empirical parameter in equation (B.3)
" '0
rot
distance between atoms, equation (3.69) and Fig. 3.2
subscript for rotation, Chapter 3
S solid in phase diagrams
S molar entropy defined by equation (1.7)
Sj partial molar enthalpy
(s) solid state
subl subscript for sublimation
T temperature in kelvins
time in seconds in Chapter 3; temperature in Celcius elsewhere
U speed (velocity)
UA empirical parameter in equation (B.10)
V molar volume
V volume of semiconductor in cm 3 , Chapter 7
V molar volume in cm 3 in Appendix B
V mp maximum voltage in a solar cell, Chapter 7
Voc open-circuit voltage in a solar cell, Chapter 7
V subscript for constant volume, e.g., C~
v,vap subscript for vaporization
List of Symbols 31S
(v) vapor state

vib subscript for vibration in Chapter 3
W, W', Wit various types of work in Chapter 1
~j exchange energy, equation (4.6)
Xi, X A , Xs mole fraction or atomic fraction of i, A, B
X~ surface atomic fraction of A atoms in Appendix B
Y, Y* net numbers of molecules with unlike neighbors, equations
(4.16) and (4.17)
yo highest value of Y
Y concentration of interstitial atoms as fraction of interstitial sites,
i.e., Y = nl eN, equation (6.1); used without subscript
Yi mole fraction of "i" in the second phase when Xi is used for
the first phase; also, YA and Ys for A and B
Yi function defined by equation (6.85)
Z compressibility factor, PVI RT; used only in Chapter 1
Z coordination number, i.e., number of nearest neighbors of an
atom in solid and liquid states
2. Greek Symbols
a phase designation
a defined by equations (3.24) and (3.40), i.e., a = 01 kT
a NWI RT in equation (4.12)
as short-range order parameter defined by equation (4.59)
f3 phase designation
f3 II kT in Chapter 3 only
f3 defined by equation (4.41) and used only in Chapters 4-6
'Yi activity coefficient, equation (1.32); 1'7 when Xi ~ 0
.y activity coefficient in equation (6.2)
.:l difference; final state minus initial state
a partial differential
8A solubility parameter 8~ = .:lEAl VA where .:lE is enthalpy of
vaporization or sublimation
ei energy level, only in Chapter 3; energy of an interstitial atom
in Chapter 6
e\j) Wagner interaction parameter, equation (1.91)
71 efficiency, percent, equation (7.71)
K empirical parameter equal to 3400 K, equation (C.1)
A defined as el kT in Chapter 7
A phase designation
316 Appendix F
/I frequency of vibration, per second

partition function for grand canonical ensemble
n product of terms
TImp maximum power in Chapter 7
peE) density of states per unit energy at E, equation (7.8)
L summation
sigma phase in Appendix A only
degrees of freedom, equation (2.17)
electronegativity in volts; Appendix A only
Fermi distribution function, equations (7.1) and (7.2)
Pauling's electronegativity
G
Periodic Table
(Mendeleev Table of the Elements)
IA
I H
0.115
0.46 CPH
6Hv, (I)",
2.20 FCC
TIA
26 Fe
1. ' Alomic 99.3 BCC }cryotal
radius (2)- 1.28 FCC
3 Li 4
38.41
Be , ~.83 BCC
struclure, (5)
Bce 11.4 BCC 3d64S~

1.57 1.13
0.98 CPH 1.57 CPH Elect" .. ~- Electron
1,'2,' 1,'2,2 olivily, (3) configuration, (4)
II No 12 Mg
25.75 35.0
1.92 BCC 1.60 CPH
0.93 CPH 1.31
3,' 3,' IDA riA Y.A lQA ilIA / Y.III A",
19 K 20 Co 21 Sc 22 Ti 23 V 24 Cr 25 Mn 26 Fe 27 Co
21.31 42.5 BCC 113.0 123.2
90.3 BCC BCC 95.0 67.1 BCC 99.3 BCC 101.5 FCC
2.3B BCC 1.97 FCC 1.64 1.47 1.36 BCC 1.28 BCC 1.12 FCC 1.28 FCC 1.25 CPH
0.82 1.00 1.36 CPH 1.54 CPH 1.63 1.66 1.55 CUBIC 1.83 BCC 1.88
4.' 40' 3d ' 4. 2 3d 24. 2 3d l 4.' 3d 54.' 3d54.' BCC 3d'4,2 3d!4"
37 Rb 38 Sr 39 Y 40 Zr 41 Nb 42 Mo 43 Tc 44 Ru 45 Rh
19.33 39.2 BCC 100.7 14B.3 175.2 157.5 158 153.6 133.1
2.51 BCC 2.15 1.81 BCC 1.60 BCC 1.47 BCC 1.40 BCC - CPH 1.34 CPH 1.34 FCC
FCC CPH CPH
0.82 0.95 1.22 1.33 1.60 2.16 1.90 2.20 2.28
5.' 5,2 4d'5.' 4d 25.' 4d45. ' 4d 55. ' 4d 55s' 4d!5s' 4d'5,'
55 Cs 56 Bo [\ 72 Hf 73 To 74 W 75 Re 76 Os 77 Ir
18.32 43.0 148.0 BCC 186.9 203.4 184.0 111.61 159.0
2.70 BCC 2.24 BCC \ 1.59 CPH 1.47 BCC 1.41 BCC 1.38 CPH 1.35 CPH 1.35 FCC
0.79 0.89 1.30 1.50 2.36 1.90 2.20 2.20
("9..z. \
t \\
6.' 6,2 4t'45d 26s' 4f '4 5d l Ss' 4f'45d 4SS2 4f '4 5d 5S.' 4f '45d 66,2 4f'45d!S,'
87 Fr 88 Rol\
-
-
-
- \%
0.70
78'
0.90
78'
"90 \ ~ 57 Lo 58 Ce 59 Pr 60 Nd 61 Pm 62 Sm
- 8CC
..A \ 103.0 BCC 101.0 BCC 85.0 78.3 49.4
FCC 1.82 FCC 1.83 BCC 1.82 BCC - CPH - BCC
~ 1.87 CPH CPH
'" 1.10 CPH 1.12 CPH 1.13 1.14 1.13 1.17 RHOM
\~
\01 89 Ac 90 Th 91 Po 92 U 93 NP 94 Pu
FCC 142.73 8ce 145.0 127.0 BCC 111.1 BCC? 82.5 BCCa,_
\ 1.10
1.80
1.30
FCC BCC
1.50 BC-TElR
1.38 TETR
1.38 ORTR
- TETR
1.36 ORTR
- FCC y ,.
1.28 NON.,11
4
5f 6d'7. 2 5f 8 782
NOTES: (1) ilHv (in kcal/mole) is for vaporization or sublimation to

monatomic gas at 298.15 K, except for H 2, N 2, O 2, F 2, C1 2, Br2, and 12 to
diatomic gases at their boiling points; from Refs. 6.42 and C.18. For diatomic
gases at low temperatures, see Rossini et aI., Selected Values of Chemical
Thermodynamic Properties, Circular 500, NBS (1952). (2) Goldschmidt
atomic radius (in A) at lowest temperature; from Ref. E.l. (3) Pauling's
electro negativity from Ref. 8.7, and Periodic Table of the Elements by
Sargent- Welch Scientific Co. (1979) as sactioned by NSRDS. Most values
are reliable to only one decimal place. See also McGraw-Hill Encyclopedia
of Science and Technology (1977) where (4) electron configuration is also
given. (5) Stable crystal structure in nearest geometrical shape at ambient
conditions; structure at highest temperature on top and that at lowest
temperature on bottom. From: Hultgren et al., Selected Values of the
2 He
0.02 BCC
- FCC
- CPH
ms liS is irS illS I,'
5 B 6 C 7 N 8 0 9 F 10 Ne
132.8 RHOM 111.29 1.33 1.63 CUBIC 0.41
0.91 TETR 0.11 HEX 0.11 HEX 0.60 HEX ~ MOl2 1.60 fCC
2.04 RHOM 2.55 3.04 CUBIC 3.44 MON 3.98 NON I -
1"2,'pl 1,'2s'p' 1"2.l p3 Is'2s'p4 1,"2sl p' 1,'2i,'
13 AI 14 5i 15 P 16 5 17 CI 18 Ar
18.8 101.1 15.62 ORTR 6S.2 4088 1.54 CPH
1.43 FCC 1.11 DIAN 1.04 NON 1.01 ORTR 1.92 FCC
1.09 CUBIC
1.61 1.90 2.19 2.58 ORTR 3.IS -
ilIIA IS TIS 3.'pl 3,'p' 3s'p' 3,"p4 3,',' 3I'p'
28 Ni 29 Cu 30 Zn 31 Ga 32 Ge 33 As 34 5e 35
1.01
Br 36 Kr
102.8 80.5 31.11 6S.2 89.5 12.12 56.25 NOI 2.IS fCC
1.25 FCC 1.28 FCC 1.31 CPH 1.35 ORTR 1.39 DIAM 1.25 RHOM I.IS MOl 1.19 ORTR 1.91
1.91 1.90 1.65 1.81 2.01 2.18 2.55 HEX 2.96 -
3d 4s'
l 3d 4.1
1D ID
3d 4.' 3d ID4,'pl 3d l'4,'p' 3dl~,'p' 3d1D4,"p4 3d104,'pl 3d I04';,'
46 Pd 47 Ag 48 Cd 49 In 50 5n 51 5b 52 Te 53 I 54 Xe
90.4 68.09
58.15 11.99 62.1 50.S
26.13 10.02 3.02 fC- BC- FCC
1.31 fCC 1.44 fCC 1.52 CPH 1.51 TETR 1.58 TETR 1.61 RHOM 1.43 TRI US ORTR 2.18
2.20 1.93 1.69 1.18 1.96 DIAN 2.05 2.10 2.6S -
4d lO 4d10s.1 4dI05.' 4d 105,'pl 4d 105,'p' 4d 10s,'p' 4~10s,'p' 4d I05,'pl 4d I05,'p'
78 Pt 79 Au 80 Hg 81 TI 82 Pb 83 Bi 84 Po 85 At 86 Rn
135.1 81.5 14.S1 BC-TETR 43.25 BCC 4S.15 49.5 - - 3.92
1.38 fCC 1.44 FCC 1.55 RHOM 1.11 1.15 fCC 1.82 RHOI 1.40 CUBIC -
CPH
- fCC
2.28 2.54 2.00 2.04 2.33 2.02 2.00 2.20 -
4f 14 5d,&,1 4f 145dIO S,1 4f I4 5dI06,' 4f145d"61'pl 4f l'5dloS,'p' 4f145dl0&s'P' 4t1 45dIOS,",' 4fl'5dI06s',s 4f 145d I06,'p'
63 Eu 64 Gd 65 Tb 66 Dy 67 Ho 68 Er 69 T~170 Yb 71 Lu
55.5 BCC, 36.35 BCC 102.2 BCC
44.9 95.0 92.9 69.4 11.9 BCC 15.8
BCC BCC BCC
2.04 BCC 1.80 1.11 1.11 1.16 CPH 1.15 CPH 1.14 CPH 1.93 fCC 1.13 CPH
1.20 1.20 CPH 1.20 CPH 1.22 CPH 1.23 1.24 1.25 1.10 CPH 1.21
4f'6.' 4f 15d l6s' 4f'6,' 4f 106,' 4f"6s' 4f"6,' 4f ll6s' 4f 6,'
14 4f"sd IS,'
95 Am 96Cm 97 Bk 98 Cf 99 Es 100Fm 101 Md 102 No 103 Lr
61.9 92.6 - -
fCCr --
- - - -
- fCC, - fCCI - fCC' - - - - -
1.30 CPH 1.30 CPH 1.30 CPH! 1.30 CPH? 1.30 1.30 1.30 1.30 -
5fll" 5f 76d l 1s' 5f l 1l' 5f l0 1l' 5f"18' 5f l'1I' 5f l'1I' 5f141s' 5f146b,'
Thermodynamic Properties of the Elements, ASM, Metals Park, Ohio (1973);

H. W. King, Bull. Alloy Phase Diagrams 3(2), 276 (1982); F. L. Oetting,
M. H. Rand, and R. J. Ackermann, The Chemical Thermodynamics ofActinide
Elements and Compounds, Part 1, The Actinide Elements, IAEC, Vienna
(1976); K. A. Gschneidner and L. R. Eyring, editors, Handbook on the
Physics and Chemistry of Rare Earths, Volume 1, Metals, North-Holland,
Amsterdam (1978).
BCC, body-centered cubic: CPH, close-packed hexagonal; DIAM,
diamond; FCC, face-centered cubic; HEX, hexagonal; MON, mono-
clinic; ORTR, orthorhombic; RHOM, rhombic; TETR, tetragonal; TRI,
trigonal.
A recent controversial convention recommends designating the groups
consecutively from 1 to 18.
Index
Acceptor level, 215 Band-gap

Acceptors, 203 data
Activity, 7 GaAs, GaP, 212
in Ag-Cu-O, 188, 189 Ge, Si, 211
in Ag-Pb-O, 188-189 energy, 198, 209
in AI-Cu, 13, 16 Band structu re, 197
coefficient, 8, 9 diagram, 198,215,217-220
definition of, 7-9 Bias (applied voltage), 218
and excess properties, 15 Boltzmann
Henrian, 8, 9 constant, 2
Raoultian, 8 statistics, 64
reference state, 9 Boltzons, 65-71
standard state, 8, 9 Bond energy, 81
variation with pressure, 10, II Bonding energy, 240
variation with temperature. 10, II Bose-Einstein statistics, 63
See also alloys of Hg Bosons, 64, 71
Ag-Cu-O, 188 Built-in potential, 218
Ag-Pb-O, 188 Bulk modulus, 257
AI-Cu, 13, 16, 295
AI-Zn, 296
Alloy formation, definition, 12 C-Fe, 170-179
Alloys, see specific systems in alphabetical Canonical ensembles, 72
sequence, e.g., AI-Cu Carriers, see Semiconductors
Arrangement of molecules (atoms), see Catatectic, 39
Distribution Cd-Mg,18
As-Ga, 297 Cd-Te, 299
Atomic masses (weights), 288-291 Cd-Zn,300
Atomization energy, 240 Cementite-Fe, 171, 179-181
Au-Cu, 294 Chemical constants. 2
Au-Pt,298 Chemical potential, 34
Average group number, 57, 58 Closed systems, 3, 27
Avogadro's number, 2 Clustering, 106, 118
321
322 Index
Components, definition, 1 Electron

Compressibility, 257 -to-atom ratio, 54
factor, 5. compounds, 55
Concentration concentration factor, 54
of electrons, 204 contribution, 54, 55
of holes, 205 gas, 74
Conduction band, 198 mass, effective, 200
Congruent melting, 40 geometric mean, 209
Constants, table of, 2 in semiconductors, 197-220
Conversion units, 2 Electron concentration factor, 54
Coordination number, 81, 82 Electronegativity, 257
Critical point, see Temperature Electronegativity factor, 54
Crystals Elements
one-dimensional, 88 electronic structure (configuration), 237, 318
structures, see Appendixes A and G phase transitions, 288-291
three-dimensional, 89, 90 selected properties, 288-291; see also
two-dimensional, 90 Appendixes A and B
Cu-Fe, 301 Elevation of freezing point, 'II
Cu-Ni,302 Empirical rules (on phases), 53, 54
order-disorder in, 127, 137-140 electron concentration factor, 54
Cu-Sn, 303 electronegativity factor, 54
Cu-Zn, 56 Hiigg rule, 58
Curie point, 172 relative valency effect, 53
Current-voltage characteristics, 221 size factor, 53
dark, 221 Energy, 3
illuminated, 221 atomization, 240
band-gap, 201
bond, 81, 82, 240
D-Pd, 153 bonding, 240
Defined constants, 2 definition of, 3
Degeneracy, 62 of electrons, 203
Degrees of freedom (variance), 32, 34 exchange, 83
Densities of states, 203, 206 of holes, 203
Dilute solutions, 24, 106 of ideal gases, 68
correlation of thermodynamic properties, kinetic, 68, 200
277-285 of electrons, 200, 203
regular, 106 of holes, 201, 203
Distribution levels, 62
of fermions, 199 potential, 76
of independent particles, 61 promotion, 238, 239
laws, 65 rotational, 74
of molecules, 88 translational, see kinetic
in one-dimensional crystals, 88 vibrational, 75
in three-dimensional crystals, 98, 99 Engel-Brewer theories, 237-254
in two-dimensional crystals, 89, 90 correlation of bonding energies, 243
ratio, 49 effect of pressure, 251
Donor level, 215 electron concentration, 246
Donors, 202 electronic configurations, 241
Dopants, 202 multicomponent diagrams, 251
Dulong and Petit, law of, 76 Ensembles, 65-72
Index 323
Enthalpy, 3, 4 Free particles, 73

definition, 3 Freezing point
of estimation of alloy fonnation, 255-276 depression of, 11
estimation of Il.Ho, 281 elevation of, 11
excess molar, 14 Fugacity, 5, 11
excess partial molar, 14 of H2(g), 5
correlation with S7, 277 Fundamental constants, 2
of fonnation, 12
of mixing, 82 Gamma function, 95
partial molar, H" 12-14 GaAs, 212
of solution, 82 GaP, 212
of vaporization, 288-291 Gas constant, 2
Entropy, 4 Ge,211
configurational, 84 Gibbs-Duhem relation, 7
definition, 4 Gibbs energy, 5
estimation for Il.S·, 281 and band-gap, 209
excess molar, 14 chemical potential, 34
excess partial molar, 14 definition, 5
correlation with "7, 277 detennination of, 23
of fonnation, Il.S, 13 diagrams, 42, 44
of melting (fusion), 288-291 excess molar, 14, 15
of phase change, 288-291 excess partial molar, 14. 15
and related properties, 69 of mixing, 14, 43
of vaporization, 2118-291 of fonnation of solutioll, 12
Equilibrium minimization of, 52
among phases, 29-57 of mixing and phase diagrams, 43
constants for charge carriers, 206 molar, 12
diagrams, see alphabetical list of phase partial molar, definition, 6
diagrams, e.g., AI-Cu alternative definitions, 27
multicomponent,31 Gibbs-Konovalow theorem, 37, 38
Euler's theorem, 6 Gibbs phase rule, 34
Gibbsian ensembles, 65-72
Eutectic and eutectoid points, 39
reactions, 39 Gorsky-Bragg and Williams approximation
Excess molar properties, 14 (GBW) , 118
analytical fonns of, 15 Grand canonical ensembles, 73
Exchange energy, 83
H-FeTi,166
Extensive properties, definition, 1
H-LaNi s, 167
Extrinsic semiconductors, 202
H-Pd, 150-162
Faraday constant, 2 H-Ta, 163, 164
Fe, 169-178 Hligg rule, 58
Fe-Fe3C, 171, 179-181 Hannonic oscillator, 75
Fe-Si,304 Heat, see Enthalpy
Fenni-Dirac statistics, 62 Heat capacity
Fenni energy (level), 71, 199, 216 at constant pressure, 4
Fennion, 63, 71 at constant volume, 3
First law of thennodynamics, 3 Dulong and Petit, law of, 76
First-order phase transition, 117 excess (in order-disorder), 127, 139
Free energy or free enthalpy (obsolete tenns order-disorder, 126, 140
for Gibbs energy), see Gibbs energy Helmholtz energy, 5
324 Index
Henrian activity, 8, 9 Machlin potential, 78

Henry's law, 8, 9 Maedelung constant, 78
Hg-Sn,25 Majority carriers, 203
Hg-Sn-Zn, 26 Margules equations, 17-23
Hg-Zn, 25 binary systems, 17-19
Holes, 197 ternary systems, 19-23
Homogeneous function, 6 Metatectic (catatectic), 39
Hume-Rothery rules, 53 Method of intercepts, 17, 44
Hydrides, 269; see also various hydrogen- Mg-Cd, 18
metal systems Mg-Sn, 306
Hydrogen, 5 Microcanonical ensembles, 65
in metals, see various hydrogen-metal Miller indices, 196
systems Minority carriers, 203
Hydrogen-metal systems (other than those Molar volume, ideal gas, 2
cited above), 163-170,267-272 Molar properties, 13
Hydrogen storage, 163, 167 excess, 14
Momentum, 68, 200, 201
Ideal diode Uunction), 220 Monotectic, 39
Ideal gas, 5, 67, 68 Multicomponent (n-component) alloys
Ideal solutions, 14, 84 (systems), 25, 251, 272
In-Sb, 305 diagrams, 251
Inflection points, 42, 84 equilibria, 31
Intensive properties, definition, 2
Interaction parameters, 24 N (nitrogen) in metals, 266
Interstitial alloys (solutions), 149-193 n-type semiconductor, 202
statistical treatment, 154 Nearest neighbors (Z), 81
Wagner model, 181 Number of nearest neighbors (Z), 81
zeroth approximation, 154
Intrinsic semiconductors, 202 o (oxygen) in metals and alloys, 281-284
Ionic crystals, 78 Occupation numbers, 62
Ionization equilibria, 214 Ohm's law, 226
Isotope separation, 154 Open systems, 2, 27
Order, 118
-disorder, see Long-range order
Joule, 2
short-range, 105
Junctions, see Semiconductors
of transition, 117
Kelvin scale of temperature, 2 P (phosphorus) in metals, 275

Kinetic energy, 68, 200, 203, 204 phosphides, 212
p-n junctions, 216
Laws of thermodynamics, 3 Partial molar properties, 6
Leonard-Jones potential, 77 definition, 7
Lever rule, 42 enthalpy, 12
Liquidus line, 37 entropy, 13
Long-range order, 117-147 Gibbs energy, 6, 27
cluster variation, 142 Partition functions, 67, 72, 73
first approximation, 130 Pb-Sb, 307
GBWapproximation, 122 Pb-Sn, 308
heat capacity, 126, 140 Pb-Zn,309
parameter, 106, 121 Pd-H, 150-162
Index 325
Pd-T, 154 Resistance, 195, 225, 226

Periodic table, 318 Rotation of molecules, 74
Peritectic, 39
Peritectoid, 39 Second law of thermodynamics, 4
Phase Second-order phase transition, 117
definition, I, 3 Semicond~ctors, 195-236
diagrams (various types), 30, 36-38,42, acceptor level, 215
48, 50, 294-309 (specific phase acceptors, 203
diagrams are alphabetically listed, e,g" band diagrams, 198, 215-221
Ag-Cu) band-gap data (GaAs, GaP, Ge, Si), 211,
calculation from thermodynamic data, 51 212
critical points, 40, 42 band-gap energy, 198
erroneous, 40 and Gibbs energy, 209
hypothetical, 30, 36-38, 42, 48, 50 band structure, 197
liquidus, 37 bias (applied voltage), 218
solidus, 37 built-in potential, 218
solvus,40 carrier concentration, 204-205
tectic reactions, 39 carrier density, 202
equilibria, 29 conduction band, 198
for multicomponents, 31 crystal structures, 195, 197
for single components, 29 current-voltage characteristics, 221, 225
rule, 34 densities of states, 203, 206
stability rules (empirical), 53-55 direct band-gap, 202
transformations, 117 donors, 202
triple point, 30 levels, 215
Photons, 64, 207 dopants, 202, 203
Photovoltaic effect, 222 amphoteric, 203
Physical and chemical constants, 2 electrons, 197-220
Planck's constant, 2 effective mass, 200
Potential geometric mean mass, 209
energy functions, 76 energy-momentum, 201
Leonard-Jones, 77 equilibrium constant for carriers, 206
Machlin,78 extrinsic, 202
Probability, 82 Fermi energy (level), 71,199,216
Promotion energy, 238, 239 holes, 197
ideal diode (dark and illuminated), 220, 221
Quantum mechanics, 61, 73 indirect bandgap, 201
of free particles, 73 intrinsic, 202
ionization equilibria, 214
Raoult's law, 8, 9 kinetic energy of carriers, 200-203
Raoultian activity, 8 n-type,202
Reference state, 9 p-n junction, 216
Regular solutions, 81 p-type,203
associated, III reverse saturation current, 220
dilute, 106 valence band, 198; see also Solar cells
first approximation, 101 Short-range order, 105
ternary, III disorder reaction, 101
zeroth approximation, 81; see also Long- parameter, 106
range order Si (silicon), see Semiconductors; Solar cells
Relative valency effect, 53 Single-component phase equilibria, 29
326 Index
Size factor, 53 Statistics

Sn-Te, 114 Boltzmann, 64
Solar cells, 222-236 Bose-Einstein, 63
current-voltage characteristics, 225 Fermi-Dirac, 62
dynamic impedance, 225 Symbols, list of, 311-316
fill factor, 225, 226 Syntectic, 39
maximum power, 225 System, I
current, 225 closed, 3, 27
resistance, 225, 226 open, 2, 27
voltage, 225
open circuit voltage, 221, 224, 225 Temperature, Celsius and Kelvin, 2
photovoltaic effect, 222 critical, 37,40,42, 124-127, 136-138,
power generation, 227 151, 153, 154
short circuit current, 221, 224, 225 Ternary systems (alloys, solutions), 111,269
silicon, 224 Theories of solutions, 81
simple electric circuit for, 224 Thermodynamic background, I
tandem, 226 Thermodynamic properties, 2-28
Solar energy conversion efficiency, 229-236 of clements, 288-291
Solar radiation, 228 of selected binary alloys, 293-309
photon flux, 227, 228 Thermodynamics, laws of, 3
spectrum, 228 Third law of thermodynamics, 4
wavelength, 228 Transitions (first- and second-order), 117
Solidus line, 37 Triple point, 30
Solutions
dilute, 24 Unit cell, 195
regular, 106 Units and conversion factors, 2
enthalpy of, see Enthalpy
entropy of, see Entropy Valence band, 198
ideal, 14, 84 Variables of states, 32
interstitial, 149 Variance (degrees of freedom), 32, 34
regular, 81-115 Velocity
approximate equations, 105 of light, 2
associated, III of molecules, 68
based on bonds, 107 Vibration
dilute, 106 of atoms in lattices, 76
first approximation, 101 of diatomic molecules, 74
ternary, III Vibrational energy, 75
zeroth approximation, 81; see also Long-
range order Wagner interaction parameters, 24
substitutional, 53, 118, 119 Wavelength, solar spectrum, 228
theories of, 81 Wigner-Seitz cells, 255
Solvus line, 40 Work function (obsolete term for Helmholtz
Specific heat, see Heat capacity energy), 5
Spinodes, 44, 85
Standard state, 8, 9 Zeroth approximation, see Regular solutions
Statistical thermodynamics, 61 Zinc blende, 197

Guest House Reservation Form 06sept2016

Uploaded by

Copyright:

Available Formats

Guest House Reservation Form 06sept2016

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Guest House Reservation Form 06sept2016

Uploaded by

Copyright:

Available Formats

Statistical

Plenum Press • New York and London

© 1986 Plenum Press, New York

Altan Gokcen, Pianist

This book is intended for scientists, researchers, and graduate students

Theories of solutions are presented in Chapter 4 in their general and

hours that I spent reading numerous publications, writing, and correcting

CHAPTER 1 THERMODYNAMIC BACKGROUND

CHAPTER 2 PHASE EQUILIBRIA AND PHASE DIAGRAMS

CHAPTER 3 STATISTICAL THERMODYNAMICS

CHAPTER 4 THEORIES OF SOLUTIONS

CHAPTER 5 LONG-RANGE ORDER

CHAPTER 6 INTERSTITIAL SOLUTIONS

Other Hydrogen-Metal Systems ....................... 163

APPENDIX A ENGEL-BREWER THEORIES

APPENDIX B ESTIMATION OF ENTHALPY OF ALLOY

Calculation of flH. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262

APPENDIX C CORRELATION OF THERMODYNAMIC

APPENDIX D SELECTED PROPERTIES OF THE ELEMENTS.... 287

APPENDIX E SELECTED BINARY PHASE DIAGRAMS AND

APPENDIX F LIST OF SyMBOLS............................... 311

APPENDIX G PERIODIC TABLE (Mendeleev Table of the Elements) 317

consisting of pressure P, volume V, and temperature T, as well as the number

{2J = {2J(P, T, n h n2 , ••• , n;) (1.1)

where there are i chemically non-compound-forming species. For brevity

Table 1.1. Physical and Chemical Constants Q

Name Symbol Value and units

Velocity of light in vacuum , V-I g.equivalenC '

Defined constants, and conversion factors:

where there are i - I composition variables because LXi = 1, and one of

Consequences of Laws of Thermodynamics

The first law of thermodynamics is a special consequence of the law

dE =. dq - P dV + dW + dW' ... (1.3)

where q is the thermal energy, -PdV is the work of compression or

dE = Cv dT - PdV; [Cv = (aEjaT)v] (1.4)

Thus, E is a function of T and V, hence E = E(T, V), because T and V

d (E + PV) = dH = dq + V dp; (H = E + PV) (1.5)

where H == E + PV is the enthalpy by definition. At constant pressure,

where H = H(T, P); therefore, H is a function of the more convenient

dS = dq(reversible) = dE + PdV (1.7)

Consequently, the equations for dE and dH are

dE = TdS - PdV (1.8)

where SoC T = 0) is taken to be zero for a pure stable single component in

Helmholtz and Gibbs Energies

Equation (1.8) can be modified by subtracting d (TS) from dE; thus,

d(E - TS) = dA = -SdT- PdV; (A = E - TS) (1.12)

where A = E - TS is the Helmholtz energy, formerly called the work

d(H - TS) = dG = VdP - SdT; (G= H - TS) (1.13)

where G = H - TS is the Gibbs energy. For an ideal gas at a fixed tem-

from which, by integration,

where GO is the standard Gibbs energy at standard pressure so that GO is

dG = RTd InF; or G = GO + RT In F (1.16)

In(FI P) = aP = Z-1 (1.17)

Partial Molar Gib~s Energy

If a function f2f = f2f(P, T, nl> n2, n3) is homogeneous with respect to

where a is the degree of homogeneity of f2f in n h n2, n3. For example,

G is then a function of P, T, and two of the three independent composition

where a = 1, since the degree of homogeneity of f2f is unity for an extensive

If, purely for brevity, we write