Computerized Thermodynamics

for Materials Scientists and Engineers

by
Mats Hillert and Malin Selleby
Computerized Thermodynamics for
Materials Scientists and Engineers

Mats Hillert and Malin Selleby

Stockholm, 2018
 Mats
c Hillert and Malin Selleby
Dept Materials Science and Engineering
KTH Royal Institute of Technology

Typeset in LATEX

ii
Introductory remarks

Basic thermodynamics is simple.

Applications may become very complicated.
Especially if one wants to express relations analytically or evaluate results by analytical meth-
ods.
Numerical methods could be even more laborious.
Previously, there was no other alternative.
Today, applications can be greatly simplified by using numerical methods through computer
software.
The computer software obeys the rules of basic thermodynamics.
Properties of phases (substances or solutions) are stored as parameters in well defined models.
The models are analytical expressions and are based on physical models, which may be more or
less correct.
A thermodynamic data bank contains a set of databases and is equipped with modules for com-
puting equilibria and diagrams.

iii
iv
Introduction

Classical thermodynamics deals with states of equilibrium but is also used to shed light on
spontaneous changes between two states. One can compute the state of equilibrium or the
driving force for a change by searching for the extremum of a characteristic state function. It is
most common to use a function called Gibbs energy, which must be minimized under constant P ,
T and content of matter in the system under consideration. From this criterion one has derived
a large number of equations under various conditions of equilibrium. In applications it is often
necessary to solve a set of such equations simultaneously. Modern computers offer a different
possibility. Software has been developed for the automatic minimization of Gibbs energy. It is
no longer necessary to define or find the proper set of equilibrium equations. It is sufficient to
define the conditions for the equilibrium state to be computed. Advanced software is not limited
to minimizing the Gibbs energy but makes it possible to use other conditions than constant P ,
T and content of matter.

In this new situation it seems that students of thermodynamics should not be forced to handle
the wealth of formulae common in classical textbooks of thermodynamics. The main emphasis
should be on the modelling of the Gibbs energy of various kinds of phases and on different ways
of defining the conditions for various kinds of equilibria. This should lead to a new strategy of
teaching thermodynamics and it requires close contact with computers. Exercises on the use of
equilibrium equations in various situations should be replaced by computer-operated exercises
mainly concerned with an analysis of what factors might control the equilibrium conditions
in a given situation. The student should thus be required to acquire a better fundamental
understanding of thermodynamics and less skill in manipulating formulae.

The present compendium represents a first attempt to present thermodynamics in this new
way. Hopefully, the text is self-consistent and logical but much shorter than in textbooks with
derivations of many equations and relations between various quantities. However, in order
to acquire any working ability in thermodynamics it is necessary for the student to practice
repeatedly on a computer as s/he works her/his way through the text.

v
List of important terms

Associate Real or hypothetical group of atoms, to be treated as

a species
Component Atom or group of atoms used for representing the
composition of the system
Concentration Amount per volume
Constituent A species in a certain phase or sublattice
Constitution Distribution of constituents on the sublattices
Content Amount in system or phase
Dormant Sleeping, a phase excluded from taking part in the
equilibrium
E
Excess Gibbs energy G, Gibbs energy in excess of a model regarded as ideal
Isobaric Under constant P
Isobarothermal Under constant P and T
Isoplethal Under constant amounts of or ratio between some
components
Isothermal Under constant T
Molar content Amount (in mole) per mole of phase or system, i.e. a ratio.
Molar Gibbs Energy Gm , Gibbs energy per mole of atoms or formula units
Mole fraction, xj Fraction of a component j in a system. Same as
“molar content”
Reciprocal system A special kind of four-component system
Reference state Unique state of an element at some selected P and T ,
used as reference for Gibbs energy
Site fraction, yj Fraction of constituent j in a sublattice
Species Atom or group of atoms of some stability, e.g. molecule
Standard state State of a pure element at the current P and T
Stoichiometric Composition/Phase expressible by formula with small
integer coefficients
Suspended Phase/Species/Component/Constituent that is excluded
from the system
System The portion of the universe for which equilibrium is
considered

vi
Contents

1 Basic Thermodynamics 1

1.1 First law of thermodynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.2 Second law of thermodynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.3 The combined law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.4 Driving force and dissipation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.5 Variable Composition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

1.6 Expressing chemical potentials... . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

1.7 Gibbs–Duhem relation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

1.8 Phases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

1.9 Gibbs phase rule . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

1.10 Introduction of new components . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

1.11 Gibbs energy of formation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

1.12 Properties of the Gibbs energy function . . . . . . . . . . . . . . . . . . . . . . . 17

1.13 Adiabatic changes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

1.14 State of reference and standard states . . . . . . . . . . . . . . . . . . . . . . . . 19

1.15 Duhem’s theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

1.16 Characteristic state function and Gibbs energy model . . . . . . . . . . . . . . . 20

vii
CONTENTS

2 Solution Models 23

2.1 Constitution and constituents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

2.2 Ideal solutions and related non-ideal solutions . . . . . . . . . . . . . . . . . . . . 24

2.3 Chemical activity and activity coefficient . . . . . . . . . . . . . . . . . . . . . . . 26

2.4 Excess Gibbs energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

2.5 Dilute solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

2.6 Phases with sublattices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

2.7 Interstitial solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

2.8 The ideal gas model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

2.9 Vapour pressure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

2.10 Thermal vacancies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

2.11 Solutions with associates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

viii
 CONTENTS

3 Phase and property diagrams 45

3.1 Introduction to phase diagrams . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

3.2 Unary systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

3.3 Binary system with almost stoichiometric phases . . . . . . . . . . . . . . . . . . 48

3.4 Binary system with solution phases . . . . . . . . . . . . . . . . . . . . . . . . . . 51

3.5 Molar phase diagrams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

ix
CONTENTS

4 Properties of important substances 55

4.1 Elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

4.2 Oxides . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

x
List of Figures

1.1 For an internal variable, the equilibrium is found by minimizing G if P and T are
constant. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.2 The chemical potentials of a binary alloy of composition xA , xB are found from
the intercepts made on the axes by the tangent. . . . . . . . . . . . . . . . . . . . 6

1.3 The equilibrium between binary phases is found from the common tangent. Its
intercepts on the axes give the chemical potentials. . . . . . . . . . . . . . . . . . 9

1.4 The chemical potentials are not defined by a phase of constant composition, a
so-called stoichiometric phase. It is here approximated by a very sharp parabola. 9

1.5 The Ag-Cu phase diagram at 1 bar. . . . . . . . . . . . . . . . . . . . . . . . . . 13

1.6 Construction to find the driving force to form one mole of a binary phase from
the pure components. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

1.7 (a) Construction to find the initial driving force D for precipitating a binary φ
phase from a solution phase of an initial composition xα◦B . (b) Diagram showing
α/φ
the final composition, xB , of the parent phase at the α + φ equilibrium. Dint is
the integrated driving force for the complete precipitation process. . . . . . . . . 16

2.1 Molar Gibbs energy diagram for an ideal binary α phase, illustrating the line
of reference and the construction to obtain the chemical potentials for a given
composition. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

2.2 Molar Gibbs energy diagram for a ternary phase, illustrating the plane of reference
and the construction to obtain the chemical potentials for a given composition. . 25

2.3 a) The binary regular solution model for a series of values of the interaction energy.
b) The activity of A in the regular solutions illustrated in a). c) The activity of
B in the regular solutions illustrated in a). . . . . . . . . . . . . . . . . . . . . . . 29

2.4 Gibbs energy diagram illustrating the construction to find a miscibility gap. The
Gibbs energy curve is taken from Fig. 2.3 . . . . . . . . . . . . . . . . . . . . . . 29

2.5 Comparison between the ideal and dilute solution models for a regular solution
with L = −2RT . The dilute solution model is based on the tangent at infinite
dilution, i.e. Henry’s law. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

2.6 Surface of reference for a reciprocal phase in a quaternary system. . . . . . . . . 33

xi
LIST OF FIGURES

2.7 Composition tetrahedron for a quaternary system with the composition square
for a reciprocal solution phase, (A, B)1 (C, D)1 . . . . . . . . . . . . . . . . . . . . 34

2.8 Difference between the SER reference and the reference for oxygen based on pure
O2 gas at the current temperature. The difference depends strongly on the tem-
perature. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

2.9 The equilibrium partial pressures of various sulphur species over liquid sulphur.
S8 is the most abundant species. Monatomic S is very rare and falls below the
diagram. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

2.10 Excess Gibbs energy accounted for with the associate solution model using a series
of values of the standard Gibbs energy of formation of AB associates. . . . . . . 43

3.1 (a): Potential phase diagram of Fe. (b) and (c): Property diagram of Gibbs energy
of Fe as function of temperature at the pressure P1 marked in the potential diagram. 46

3.2 Property diagram from Fig. 3.1(c), now rotated to demonstrate that the phase
with the lowest Gibbs energy is the stable phase. . . . . . . . . . . . . . . . . . . 47

3.3 Property diagrams of enthalpy and entropy of Fe, illustrating discontinuous changes
at the temperature of transformation. . . . . . . . . . . . . . . . . . . . . . . . . 47

3.4 (a). Potential diagram of Fe with the high pressure phase  included. (b) and (c):
Property diagram of Fe at the pressure P1 , marked in the potential diagram. . . 48

3.5 Potential phase diagram of the binary W-C system at a pressure of 1 bar . . . . 49

3.6 Potential phase diagram of the binary Cr-C system at 1 bar, showing the phase
fields of various carbides. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

3.7 Left: Potential phase diagram of Fe-C illustrating the phase fields of two solution
phases at 1 bar. Right: Property diagram at the temperature marked in the phase
diagram. It illustrates that Gibbs energy is here discontinuous at the transfor-
mation because the phases have different compositions. Note the difference from
Fig. 3.1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

3.8 Left: Phase diagram of the Fe-C system with one molar axis, illustrating that
the two-phase field is no longer a line but has an extension in the direction of
the molar axis. It is filled with tie-lines connecting the two phases that are in
equilibrium with each other. Right: Property diagram of the Gibbs energy as
function of composition at the temperature marked in the phase diagram. The
compositions of the phases in equilibrium are indicated with black dots. Cf. Fig.
1.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

3.9 The fractions of α and γ at the temperature marked in the phase diagram of Fig.
3.8 as function of composition, illustrating the linear variation along a tie-line. . 52

3.10 Same diagram as the right part of Fig. 3.8 but with the common tangent included
and showing the whole composition axis. The intersections of the common tangent
give the chemical potentials. Cf. Fig. 1.3. . . . . . . . . . . . . . . . . . . . . . 53

3.11 (a) Potential phase diagram of the ternary Fe-Cr-C system at 1 bar and 1500 K.
(b) Molar phase diagram of the same system. . . . . . . . . . . . . . . . . . . . . 54

xii
 LIST OF FIGURES

3.12 The whole molar phase diagram of the Fe-Cr-W system at 1 bar and 1673 K,
plotted on a Gibbs triangle. All the phase fields are here two-dimensional. . . . . 54

4.1 Thermodynamic properties of Cr in the liquid and bcc phases relative to the SER
reference. The horizontal lines represent the discontinuous changes at the melting
temperature. Each one of the phases is metastable on the “wrong” side of that
temperature. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

4.2 Same properties as in Fig. 4.1 but now given relative to the standard state, which
is here defined for bcc at the same temperature. . . . . . . . . . . . . . . . . . . 56

4.3 Standard Gibbs energy of formation of some oxides. They are all given per mole
of formula unit. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

4.4 Oxygen potential diagram for some oxides. The tendency to oxidize decreases
upwards in the diagram. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

4.5 Potential phase diagram for the Fe-O system. . . . . . . . . . . . . . . . . . . . . 58

xiii
LIST OF FIGURES

xiv
Chapter 1

Basic Thermodynamics

1.1 First law of thermodynamics

Energy can be transferred to a system from the surroundings as heat, Q, or work, W . The
first law of thermodynamics requires that energy cannot be generated nor destroyed and the
transferred energy must somehow be stored inside the system. Without analyzing in what way
it is stored, one would thus know that the energy of the system has changed by an amount equal
to what it has received. One has thus defined a special concept, the internal energy, U , and
in basic thermodynamics there is no need to discuss its physical nature.

dU = dQ + dW (1.1)

We will mainly be concerned with work done by an external pressure, P , by compressing the
system with an amount of −dV , yielding

dU = dQ − P dV (1.2)

Under constant volume, V , the heat transfer to the system will be

dQ = dU (1.3)

Under constant pressure, P , the heat transfer will be

dQ = dU + P dV ± V dP = d (U + P V ) − V dP = dH (1.4)

We have introduced the notation H for U + P V and that quantity is called enthalpy. Eq. 1.4
explains why it is often called heat content but it should be remembered that there is no ther-
modynamic way of distinguishing heat from work inside the system. In thermodynamics, heat
and work are just different mechanisms of transferring energy between two systems or between
a system and its surroundings. Enthalpy is a very useful property because it is more common to
operate under constant P than V . Using Eq. 1.4 one can determine experimentally how H varies
with T for various substances and such information is often collected in tables for properties of
various substances.

1
CHAPTER 1. BASIC THERMODYNAMICS

Exercise 1.1.1. Evaluate the heat of melting of 1 mole of pure fcc-Ni at 1 bar.
Hint 1) The melting point at 1 bar is 1728.25 K. 2) Integration of Eq. 1.4 yields
Qmelting = H liquid − H f cc because for a pure element the phases retain their prop-
erties during the whole melting process when P and T are kept constant.

1.2 Second law of thermodynamics

The second law of thermodynamics concerns another quantity, S, called entropy. A process
or reaction occurring spontaneously inside a system must increase the entropy of the system if
there is no heat exchange with the surroundings.

dip S > 0 for spontaneous internal processes (1.5)

An internal process will be infinitely slow if dip S = 0. This is sometimes regarded as a reversible
process. The symbol dip means that the change is caused by an internal process and dip S is
regarded as an internal entropy production. A heat exchange itself will change the entropy
by dQ/T and the total change of S in the system accompanying an internal process would be

dQ dQ
dS = + dip S > (1.6)
T T

since dip S > 0. No attempt will here be made to explain entropy or prove the second law.
Actually, both laws will be regarded as axioms.

Exercise 1.2.1. Evaluate ∆ip S for the solidification of 1 mole of liquid Ni in a

system at 1720 K and 1 bar.
Hint Solidification of the liquid can be regarded as an internal process of the system.
Liquid and fcc Ni are in equilibrium at 1728.25 K. At that temperature liquid Ni
will not solidify spontaneously, dip S = 0. At 1720 K the liquid is supercooled and
could solidify spontaneously, dip S > 0. The quantity dip S could be evaluated from
Eq. 1.6 in the integrated form, ∆ip S = ∆S
f cc liq f cc liq

− ∆Q/T . Eq. 1.4 then yields for one
mole ∆ip Sm = Sm − Sm − Hm − Hm /1720. The molar quantities should here
be taken from 1720 K.

1.3 The combined law

Combining Eq. 1.6, rewritten as T dS = dQ + T dip S, with the first law as given by Eq. 1.2 yields
the combined law,

T dS = dQ + T dip S = dU + P dV + T dip S > dU + P dV (1.7)

We can rearrange the terms in order to introduce P and T as the variables,

− T dip S = dU + P dV − T dS = d (U + P V − T S) − V dP + SdT (1.8)

2
 CHAPTER 1. BASIC THERMODYNAMICS

Figure 1.1: For an internal variable, the equilibrium is found by minimizing G if P and T are
constant.

Introducing the symbol G for U + P V − T S, a quantity called Gibbs energy, we obtain

dG = V dP − SdT − T dip S < V dP − SdT for spontaneous internal processes (1.9)

G is of more direct practical use than S because it depends on P and T and one usually considers
systems under constant P and T . A system kept under conditions of constant P and T may
undergo a change by a spontaneous internal process under exchange of heat and work with the
surroundings if it causes a negative change of G.

dG = −T dip S < 0 for spontaneous internal changes under constant P and T (1.10)

Exercise 1.3.1. Evaluate ∆ip S for the solidification of 1 mole of liquid Ni at 1720
K and 1 bar from information on ∆G for solidification. Hint ∆G may be obtained
from a table of Gm or from a computerized database.

1.4 Driving force and dissipation

Eq. 1.10 may be illustrated with a simple diagram, Fig. 1.1, where the x axis represents the
extent of an internal process, ξ, in a system kept under constant P and T . In contrast to
the external variables, which may be controlled from the outside, ξ is an internal variable,
which can change spontaneously even in a completely isolated system. Starting at ξ = 0, the
Gibbs energy here decreases as the process proceeds but the process will finally stop when it can
no longer decrease the Gibbs energy, i.e. at the point of minimum. If the system were initially on
the other side of the minimum, the process would spontaneously proceed in the other direction
and approach the same minimum. The minimum thus represents the equilibrium of the system
with respect to this internal process and the equilibrium condition under constant P and T is

(∂G/∂ξ)P,T = 0 (1.11)

One may thus determine the state of equilibrium with regard to this internal process if one
knows how G varies with ξ under constant P and T . It is evident that G must be a function
of ξ in addition to P and T , i.e. G = G (P, T, ξ). The rate of decrease of Gibbs energy, defined

3
CHAPTER 1. BASIC THERMODYNAMICS

from the slope of the G curve in Fig. 1.1, can be regarded as the driving force for increasing
the ξ variable, denoted D (or DF if there is a need to distinguish it from a diffusion coefficient).

D = − (∂G/∂ξ)P,T (1.12)

For a transformation between two states, α → β, one obtains by integration D = −∆Gm where
∆Gm = Gβm − Gα m . In view of Eq. 1.11, a system will be in internal equilibrium under constant
P and T if D = 0 for all possible internal processes. One often assumes that the rate of a process
is proportional to its driving force. A process could occur spontaneously only if D > 0.

If we consider only one process at a time, then Eq. 1.9 yields under constant P and T

D = − (∂G/∂ξ)P,T = T dip S/dξ (1.13)

We can identify D with T dip S/dξ and replace T dip S by Ddξ in all the previous equations,
e.g. Eq. 1.9,

dG = V dP − SdT − Ddξ (1.14)

R
In this connection, Ddξ may thus be regarded as the dissipation of Gibbs energy and Ddξ as
the dissipated Gibbs energy for a process. Evidently, an internal process occurring under D = 0
would not produce any entropy, nor dissipate any Gibbs energy. However, it is predicted to be
infinitely slow. As already mentioned, such a process is sometimes regarded as a reversible
process. Likewise, a process does not dissipate any Gibbs energy if dξ = 0. Of course, that
happens at equilibrium where D = 0 but also for a frozen process, which could happen if the
temperature is low enough to make the rate of the internal processes practically zero. In both
cases

dG = V dP − SdT (1.15)

It would thus be possible to vary P and T of a system without causing an internal process if all
possible internal processes are frozen. Eq. 1.15 will be our most useful form of the combined law
for a unary system.

Exercise 1.4.1. Find the driving force for the formation of solid Ni from liquid Ni
at 1720 K and 1 bar.
Hint The only quantity that changes during isobarothermal solidification of pure Ni
is the amounts of solid and liquid Ni. It may thus be convenient to express the extent
of solidification, ξ, by the amount of solid Ni, e.g. expressed as moles, N sol . The
driving force, D, will have the same value during the whole
solidification of asolpure
substance and we obtain D = −∆G/∆ξ = − Gsol − Gliq /∆N sol = Gliq m − Gm .

1.5 Variable Composition

Consider a system for which also the content may change by exchange of matter with the
surroundings. By generalizing Eq. 1.14 one obtains

4
 CHAPTER 1. BASIC THERMODYNAMICS

X
dG = V dP − SdT + µi dNi − Ddξ (1.16)

µi enters into this expression as one of several coefficients. They can all be expressed as partial
derivatives of G, e.g. V = (∂G/∂P )T,Ni ,ξ and µk = (∂G/∂Nk )T,P,Nj ,ξ . The subscript Nj indi-
cates that all Ni except Nk are kept constant. If also P and T are kept constant and there is
no internal entropy production, one obtains

µk = (∂G/∂Nk )T,P,Nj (1.17)

Due to its definition, µk is a partial derivative of G and is often called the partial Gibbs
energy of component k. P and T and all µi are regarded as external variables because they
can be controlled from the surroundings. V , S and Ni are also external variables but of all those
external variables only P , T and all the µi are regarded as potentials because at equilibrium they
must each have the same value in the whole system. They can be kept constant at fixed values
by requiring equilibrium with an infinite reservoir of matter with those values. In addition to be
regarded as a partial Gibbs energy, µk is also regarded as the chemical potential of component
k. It is a very important property of the system.

It is also possible to consider the system as a reservoir which can supply the surroundings with
matter of the potentials given by Eq. 1.17 because

   
−∂G ∂G
= = µk (1.18)
−∂Nk P,T,Nj ∂Nk P,T,Nj

P
The mole fraction xk ≡ Nk / Ni = Nk /N will now be introduced as a means of expressing
composition. From Eq. 1.16, we obtain

X 
dG = V dP − SdT + µi xi dN − Ddξ (1.19)

Consider a large amount of homogeneous matter with uniform P , T and composition. Define a
very small volume as the system and extend its limits gradually without really changing anything.
At any stage the Gibbs energy will be proportional to the amount of matter.

X  X
G= µi xi N = µi Ni (1.20)

It should be emphasized that one could often define a large number of internal processes and
their driving forces. They are also internal variables. In fact, all that can be used to describe
the internal state of a system can be regarded as internal variables.

The molar Gibbs energy is defined as

X
Gm ≡ G/N = µi xi (1.21)

The definition of chemical potential is illustrated for a binary solution phase by the molar Gibbs
energy diagram in Fig. 1.2.

In order to explain the diagram we shall soon need the following partial derivatives for xA =
NA /N and xB = NB /N

5
CHAPTER 1. BASIC THERMODYNAMICS

Figure 1.2: The chemical potentials of a binary alloy of composition xA , xB are found from
the intercepts made on the axes by the tangent.

 
∂xA N · 0 − NA · 1 −xA
= = (1.22)
∂NB NA N2 N

 
∂xB N · 1 − NB · 1 1 − xB
= 2
= (1.23)
∂NB NA N N

It is evident from Eq. 1.18 that the chemical potential of each component must have the same
value everywhere in a system at equilibrium. Otherwise, the Gibbs energy could decrease spon-
taneously by some material moving into the region with the lower value. The driving force for
such a reaction would be

D = µ0i − µ00i (1.24)

This is why the chemical potential is regarded as a potential. The same applies to P and T ,
which are also potentials. There is one exception. P will
 have different values on the two sides of
a curved interface with a specific interfacial energy, σ J/m2 . For a spherical interface a simple
mass balance shows that

∆P = 2σ/Vm (1.25)

One usually regards the elements A and B as a set of two independent components in a binary
system because one can vary the amount of one without changing the amount of the other.
However, one can define the set of two independent components differently. It may be convenient
to regard a combination of A and B as one of the independent components. When A is replaced
by Aa Bb but B is retained, Eq. 1.20 for a binary system can be rearranged

6
 CHAPTER 1. BASIC THERMODYNAMICS

X
G= µi Ni = NA µA + NB µB = NA µA + NB µB ± (b/a)NA µB
(1.26)
= NA (µA + (b/a)µB ) + (NB − (b/a)NA )µB = NA0 a Bb µAa Bb + NB0 µB

where

µAa Bb = aµA + bµB (1.27)

and NA0 a Bb = NA /a and NB0 = NB − (b/a)NA . The coefficients, a and b are usually small
numbers but may be replaced by the mole fractions, xA and xB . It should be emphasized that
µAa Bb may be a useful quantity even without defining a new independent component. When
Aa Bb molecules are actually present in a solution phase, then Eq. 1.27 expresses their chemical
potential at internal equilibrium between the three species. For a frozen state, all three chemical
potentials depend on their actual amounts. Eq. 1.27 can be applied only for a real or assumed
equilibrium.

Exercise 1.5.1. (Advanced) Use Eq. 1.17 to derive µA for a system composed of a
mixture of two stoichiometric A-B phases.
HintRemember that Gibbs energy is an additive quantity. How can one then keep
NB constant when one varies NA ?

1.6 Expressing chemical potentials through

the molar Gibbs energy

In order to apply the definition of µB , obtained from Eq. 1.17 for a binary system, to Fig. 1.2
we shall introduce the molar Gibbs energy Gm = G/N , which should really be described as a
function of xA or xB but is often given with an expression containing both mole fractions. It
is then implied that xA = 1 − xB for a binary system. As long as we obey Eqs. 1.22 and 1.23
we may treat Gm as a function of both xA and xB , regarded as independent variables, which is
often practical. We may thus derive expressions for the slope of the tangent, dGm /dxB , and
chemical potentials.

 
∂Gm ∂Gm ∂Gm ∂Gm
dGm = dxA + dxB = − dxB (1.28)
∂xA ∂xB ∂xB ∂xA

 
dGm ∂Gm ∂Gm
= − (1.29)
dxB ∂xB ∂xA

Using Eqs. 1.22 and 1.23 we obtain

7
CHAPTER 1. BASIC THERMODYNAMICS

   
∂G ∂ (N Gm )
GB ≡ µB ≡ =
∂NB NA ∂NB NA
     
∂N ∂Gm ∂xA ∂Gm ∂xB
= Gm +N · +N ·
∂NB NA ∂xA ∂NB NA ∂xB ∂NB NA
∂Gm ∂Gm (1.30)
= Gm − xA + (1 − xB )
∂xA ∂x
  B
∂Gm ∂Gm
= Gm + xA −
∂xB ∂xA
dGm
= Gm + xA
dxB

The chemical potentials can thus be obtained from the intersections of the tangent with the
component axes in Fig. 1.2. Taking the difference between Eq. 1.30 and a similar equation for
µA we obtain for the slope of the tangent

   
∂Gm ∂Gm ∂Gm ∂Gm
µB − µA = Gm + xA − − Gm − xB −
∂xB ∂xA ∂xA ∂xB
(1.31)
∂Gm ∂Gm dGm
= − =
∂xB ∂xA dxB

This relation is also illustrated by Fig. 1.2. Finally, it should be mentioned that one can in the
same way derive an expression for the chemical potentials in a multicomponent solution. It is
preferable to express it as follows.

∂Gm X ∂Gm
GB ≡ µB = Gm + − xi (1.32)
∂xB ∂xi

It is evident that this reduces to the third line of Eq. 1.30 for the binary case. In connection
to Eq. 1.18 it was mentioned that the chemical potential of a component could be regarded as
its partial molar Gibbs energy. Partial quantities of other molar properties can be defined in a
similar way according to the general relation,

 
∂A ∂Am X ∂Am
Ak ≡ = Am + − xi (1.33)
∂Nk P,T,Nj ∂xk ∂xi

As an example, by inserting H in place of A one obtains the partial enthalpy of component k in

the solution, usually called heat of solution of component k.

Since each potential must have the same value in all parts of a system at equilibrium, two
phases in equilibrium must have the same values for P , T and all the chemical potentials. In a
molar Gibbs energy diagram for which P and T are kept constant it is thus necessary that the
compositions of two phases in equilibrium are such that there is a common tangent and from
the intersections on the component axes one can read the chemical potentials of the two-phase
equilibrium. See Fig. 1.3.

A binary compound φ with constant composition is represented with a single point in the molar
Gibbs energy diagram but it is convenient to represent it with a very narrow parabola which is
consistent with the difficulty to vary its composition. Fig. 1.4 demonstrates that the chemical
potentials of the components are not uniquely defined in such a phase.

8
 CHAPTER 1. BASIC THERMODYNAMICS

Figure 1.3: The equilibrium between binary phases is found from the common tangent. Its
intercepts on the axes give the chemical potentials.

Figure 1.4: The chemical potentials are not defined by a phase of constant composition, a
so-called stoichiometric phase. It is here approximated by a very sharp parabola.

9
CHAPTER 1. BASIC THERMODYNAMICS

Exercise 1.6.1. Use Eq. 1.17 for evaluating the chemical potentials of Ag and Cu
at equilibrium in an Ag-Cu melt of 40 mass-% Cu at 1200 K and 1 bar.
Hint When you can define the conditions for a state of equilibrium in your software
then it should be possible to make your software evaluate most properties of the state.
Find out if it can evaluate µAg and µCu as partial derivatives, i.e. as (dG/dNi )P,T,Nj .

1.7 Gibbs–Duhem relation

In Fig. 1.4 one can give the tangent any slope through the point representing ϕ, a so-called
stoichiometric phase. This demonstrates that the chemical potentials are not defined for such a
phase alone. Two possibilities are shown in Fig. 1.4. They are identified by 0 and 00, respectively.
However, the µA , µB pair can be chosen in an infinite number of ways but according to Eq. 1.21
it must obey the relation

Gφm = xφA µA + xφB µB (1.34)

This relation can also be verified geometrically from Fig. 1.4. Since xφA and xφB are constant and
Gφm has a fixed value for this phase, we find

xφA dµA + xφB dµB = dGφm = 0 (1.35)

Actually, one can prove that this relation also holds for a solution phase where the composition
can vary, as long as the variation in composition is small. It is called the Gibbs–Duhem
relation and its general form is

X
xi dµi − Vm dP + Sm dT = 0 (1.36)

Multiplying by N we obtain an alternative form of the Gibbs–Duhem relation.

X
Ni dµi − V dP + SdT = 0 (1.37)

It should be emphasized that we have derived the Gibbs–Duhem relation only for a system
containing a single phase. In fact, it only applies to each phase separately if there are more than
one phase in the system.

Exercise 1.7.1. Consider the equilibrium between an A-rich phase and a stoichio-
metric A3 B2 compound. Use Eq. 1.34 to illustrate how µB can be obtained from a
molar Gibbs energy diagram.
Hint Remember that the molar Gibbs energy of all phases is expressed per mole of
atoms in the molar Gibbs energy diagram, not per formula unit.

10
 CHAPTER 1. BASIC THERMODYNAMICS

Exercise 1.7.2. The chemical potentials of A and B in an A-B solution phase can
be changed by adding some B or by exchanging some A for the same amount of B.
Check the validity of the Gibbs–Duhem relation at constant P and T for these two
cases numerically on a solid Ag-Cu alloy with 5 atom-% Cu at 1 bar and 1100 K.
Hint For these two cases the Gibbs-Duhem relation yields

NCu (∂µCu /∂NCu )NAg ,P,T + NAg (∂µAg /∂NCu )NAg ,P,T = 0
NCu (∂µCu /∂NCu )N,P,T + NAg (∂µAg /∂NCu )N,P,T = 0

An advanced thermodynamic program can give all these derivatives directly. If that
is not available to you, consider 1 mole of the alloy, first evaluate the chemical
potentials directly, then add a small amount of Cu, e.g. 0.01 mole, and evaluate the
new chemical potentials, and apply Eq. 1.37 to the differences obtained and, finally,
do the same after also removing the same amount of A.

Exercise 1.7.3. It is common to express the chemical potential of oxygen in a gas

through the partial pressure of the O2 molecules. Of course, the value of this quantity
is directly related to the chemical potential of O2 . Show that it is also related to
the chemical potential of O. The result may seem absurd if there are no O atoms
present. Some condition must be fulfilled. Which one?
Hint Apply Eq. 1.27.

1.8 Phases

A homogeneous system or portion of a system is regarded as a phase and is often denoted by

a Greek letter, e.g. α, β, γ, δ etc. Many phases can have a variable composition and can then
be regarded as solution phases. Phases are thus identified not by their composition but by
their structure, e.g. body-centered cubic (bcc) or liquid (L). Many phases have an almost or
practically constant composition, which can be represented by a chemical formula with small
integers as coefficients, e.g. Al2 O3 . They are regarded as stoichiometric phases.

V , U and S are all extensive properties and are additive in the sense that the value for a
composite system is the sum of the values of the parts, e.g.

V =Vα+Vβ (1.38)

For a homogeneous system, the value of an additive property is proportional to the size of the
system, N , and for all extensive properties one can define the molar quantity

Vm = V /N (1.39)

The subscript m indicates that the size is here measured as the total number of mole of atoms,
N . Tables with properties of substances always give molar properties. Molar properties can be
regarded as intensive because they have the same value everywhere in a homogeneous system.
However, they are not potentials like P , T and µi , which must have the same value everywhere
in the whole system at equilibrium.

11
CHAPTER 1. BASIC THERMODYNAMICS

G is also an additive property but only for systems with uniform P and T because they are
external variables. The molar Gibbs energy of a pure element A could be denoted ◦ Gα m (A) but
is usually abbreviated ◦ GαA . From Eq. 1.17 it is evident that µA for a pure element in the α
state is equal to ◦ Gα
A because N = N k and G = N · ◦ α
GA . For a stoichiometric phase like Al2 O3
one may define the molar Gibbs energy per mole of formula units, ◦ GAl2 O3 , i.e. for five moles of
atoms.

Exercise 1.8.1. Find the total Gibbs energy of a system containing 0.5 kg of Al2 O3
and 1 kg of CaO at 1 bar and 1500 K. Suppose the two substances have not reacted
with each other. Check the law of additivity.
Hint The system is not really in internal equilibrium since the two oxides have a
tendency to react with each other. To avoid that they react with each other, you
could simply omit all other phases from the set of data fetched from the database.
To check the law of additivity, evaluate the total G and G of the Al2 O3 phase and
G of the CaO phase separately and add them.

1.9 Gibbs phase rule

The value of a potential in a system can be changed to a new value by bringing the system in
contact with a large reservoir at another value. There are c + 2 such independent potentials,
basically P , T and the chemical potentials for c components. It may seem that one has the
freedom to vary all c + 2 potentials, i.e., one has c + 2 degrees of freedom. However, the Gibbs-
Duhem relation states that a phase cannot remain in a system under changed conditions unless
the variations of all the independent potentials are somehow related. The degrees of freedom to
vary the conditions are thus reduced to c + 2 − p where p is the number of phases one wants to
be present in the system because the Gibbs-Duhem relation is different for all the phases. One
also says that the variance for a system with p prescribed phases is

v =c+2−p (1.40)

When applied to a phase diagram v is the dimensionality of phase fields but it should be realized
that the rule was derived by considering potential variables. Consequently, it applies only to
phase diagrams with potential axes. Fig. 1.5 illustrates this difference with an example from the
Ag-Cu system, One degree of freedom has already been consumed by requiring that the pressure
is always 1 bar, yielding a remaining variance of v = c + 2 − p − 1 = 3 − p. The first diagram
has two potential axes and one-phase fields are two-dimensional, two-phase fields are linear and
the three-phase field is the point of intersection in agreement with the rule. The other phase
diagram has a two-phase field that is two-dimensional in conflict with the rule. The reason is
the use of a molar axis which is not a potential.

Exercise 1.9.1. Sometimes one plots the phase diagram for a system with different
sets of axes in order to illustrate different aspects. Fig. 1.5 shows the Ag-Cu phase
diagram at a pressure of 1 bar, plotted in two ways. Locate the fcc+L phase field in
both diagrams and try to apply the Gibbs phase rule. Explain the results.
Hint In this system there are two fcc phases. Choose one of them. Remember under
what conditions the rule was derived.

12
 CHAPTER 1. BASIC THERMODYNAMICS

Figure 1.5: The Ag-Cu phase diagram at 1 bar.

Exercise 1.9.2. Eq. 1.16 has some terms with a potential as variable and some with
an extensive quantity. Which one is intensive and which one is extensive in the term
Ddξ?
Hint You may for instance consider the solidification of a pure metal as the internal
process.

1.10 Introduction of new components

In Section 1.5 it was shown how one could define a new component for a particular purpose. A
common case is a solution phase containing associates between the basic components, e.g. a gas
phase with two components, H and O elements or H2 and O2 molecules, which may react to
form H2 O molecules. According to Eq. 1.27, equilibrium for the reaction requires that

µH2 O = 2µH + µO (1.41)

It could also be written

µH2 O = µH2 + 0.5µO2 (1.42)

because µH2 = 2µH and µO2 = 2µO at equilibrium, again according to Eq. 1.27 but now with
a = 2 and b = 0. The driving force for the reaction H2 + 0.5O2 → H2 O, before equilibrium has
been established, can be written as

D = µH2 + 0.5µO2 − µH2 O (1.43)

where all the chemical potentials are evaluated from the actual amounts of the species.

One could thus define a new set of components in any way provided that the Gibbs phase rule
is satisfied by the number of independent components being correct. As an example where a
new set of components may be convenient one can mention a system obtained by mixing several

13
CHAPTER 1. BASIC THERMODYNAMICS

stoichiometric oxides. Oxygen and the primary oxides may then be used as the set of independent
components instead of the set of elements. At equilibrium one can directly obtain the chemical
potentials for one set in terms of those in the other set using Eq. 1.27. However, if all the oxides
were stoichiometric and formed from different elements, there should be no way of varying the
oxygen content independently. Eq. 1.17 could not be applied to oxygen and the oxygen potential
could not be defined, nor the potentials of the other elements, only the potentials of the oxides.

Stoichiometric phases may be taken as an example. From a database one may obtain some ther-
modynamic properties of an oxide MO by tabulation, e.g. the molar Gibbs energy of a particular
crystalline structure, ◦ Gϕ
M O . Other pieces of information may be obtained be calculations from
the basic information. If a general kind of software for thermodynamic calculations is available
it probably requires that the degrees of freedom are reduced to zero by the specification of con-
ditions. For a system with only the MO phase one obtains v = c + 2 − p = 2 + 2 − 1 = 3.
Two further conditions could be prescribed values of P and T. It will be necessary to introduce
one more condition. It could be another phase formed by one or both of the components, e.g.
metallic M or gaseous O2 or the oxygen potential. One can use any such condition as long as
one is interested in only the properties of MO, e.g. its heat capacity or enthalpy. For a system
with oxides of two different metals, MO and LO, the situation will be the same and also the
remedy.

Exercise 1.10.1. For the Ca-O-Si system it may be natural to use the three elements
as components. However, suppose you are only interested in the reactions between
the two primary oxides, CaO and SiO2 , and you like to compute the CaO-SiO2 phase
diagram, which is really a quasibinary section of the ternary one. Should you expect
any problem? If so, try to solve it.
Hint A thermodynamic program normally treats the elements as components and
for a ternary system there are three. In the present case it may seem natural to
define CaO and SiO2 as the components but the program may still require a third
component. You have to introduce a third component that is not situated on the
CaO-SiO2 line. It may seem most logical to select O as the third component. To
avoid that other phases in the CaO-O-SiO2 system appear in the calculation it may
be convenient to use a very low O activity, e.g. 10−10 .

1.11 Gibbs energy of formation

Consider an A-B system consisting of pure A and pure B which are not soluble in each other.
◦ β
The values of µA and µB are thus established as ◦ Gα
A and GB . Then, suppose A and B atoms
in the correct amounts form one mole of a new stoichiometric phase, the φ phase. The change
of the molar Gibbs energy of the system will be

∆Gm = Gφm − xφA µA − xφB µB = Gφm − xφA ◦ Gα φ ◦ β

A − xB GB (1.44)

The formation of φ is illustrated with a molar Gibbs energy diagram in Fig. 1.6. In this case
the compound forms under a decrease of the Gibbs energy and the driving force is thus positive,
D = −∆Gm .

The formula of a compound is usually given with the coefficients a and b as small integers and
the Gibbs energy per mole of formula unit is GφAa Bb = (a + b) Gφm . The change per formula unit
of the new phase φ when formed from pure A and pure B will be,

14
 CHAPTER 1. BASIC THERMODYNAMICS

∆◦f GφAa Bb = ◦ GφAa Bb − a◦ Gα ◦ β

A − b GB = − (a + b) D (1.45)

The superscript ◦ indicates that a quantity refers to a pure component or stoichiometric com-
pound. The quantity ∆◦f Gφm is regarded as the standard Gibbs energy of formation of the
◦ β
φ phase if ◦ Gα
A and GB have been chosen as the molar Gibbs energies of pure A and B in their
most stable states at the same P and T , their standard states. See Section 1.14. This quantity
is thus negative for all compounds that can form spontaneously from the standard states of the
elements.

The magnitude of the driving force of a phase with variable composition depends on the exact
composition of the phase as it precipitates (in this specific case, the phase is stoichiometric, i.e.
always the same composition) but also on the composition of the parent phase if it is a solution
phase. The largest value is obtained with a parallel tangent construction in Fig. 1.7(a) which
illustrates the precipitation from a solution phase. The composition of that phase, (xα◦ α◦
A ; xB ),
controls the chemical potentials of both components and this phase acts as a reservoir of both
components. Instead of the molar Gibbs energies of the pure components one should here
introduce the chemical potentials of the parent solution phase in Eqs. 1.44 and 1.45 obtaining

D = xφA µα◦ φ α◦ φ
A + xB µB − Gm (1.46)

Gφm in Eq. 1.46 is defined for 1 mole of atoms. Multiplying Eq. 1.46 with a + b, rearranging and
inserting the value for ◦ GφAa Bb from Eq. 1.45 and knowing that xφA = a/(a+b) and xφB = b/(a+b),
we obtain

 
◦ α ◦ β ◦ φ
(a + b) D = a (µα◦
A − GA ) + b µα◦
B − GB − ∆f GAa Bb (1.47)

The composition of the solution phase will change gradually as the precipitation proceeds and the
driving force will thus decrease until it approaches zero as a final equilibrium is established when
the parallel tangents coincide. Fig. 1.7(a) thus shows the initial driving force and Fig. 1.7(b)
shows the total driving force for the complete process. It is expressed per mole of the precipitated
phase because it comes from a molar Gibbs energy diagram. It may be called integrated driving
force and may sometimes be denoted Dint since it is obtained by integration over the whole
process. It should be emphasized that an integrated driving force can be evaluated only between
two states of the same composition. In a molar Gibbs energy diagram they must fall on the
same vertical line.

Thermodynamic information of compounds is often tabulated in the form of standard Gibbs

energy of formation i.e. defined relative to the standard states at the current P and T . However,
in computerized databases it is preferable to store the properties of substances as functions of
temperature and it is then more convenient to have references at a fixed temperature. See
Section 1.14.

Exercise 1.11.1. Evaluate the standard Gibbs energy of formation of Cr23 C6 at 1

bar and 1000 K from your data bank system.
Hint Find the Gibbs energy for 1 mole of formula units of Cr23 C6 . Start by changing
the references of Cr and C to bcc-Cr and graphite at 1 bar and 1000 K.

15
CHAPTER 1. BASIC THERMODYNAMICS

Figure 1.6: Construction to find the driving force to form one mole of a binary phase from the
pure components.

Figure 1.7: (a) Construction to find the initial driving force D for precipitating a binary φ phase
from a solution phase of an initial composition xα◦
B . (b) Diagram showing the final composition,
α/φ
xB , of the parent phase at the α + φ equilibrium. Dint is the integrated driving force for the
complete precipitation process.

16
 CHAPTER 1. BASIC THERMODYNAMICS

1.12 Properties of the Gibbs energy function

As already discussed, it is possible to express all the thermodynamic properties of a substance

through its Gibbs energy. This is accomplished by the use of various partial derivatives of the
characteristic state function G (P, T, Ni ). Eq. 1.17 is just one example. In the same way one
obtains

V = (∂G/∂P )P,T,Ni (1.48)

S = − (∂G/∂T )P,Ni (1.49)

H ≡ U + P V = G + T S = G − T (∂G/∂T )P.Ni = (∂ (G/T ) /∂ (1/T ))P,Ni (1.50)

One defines the heat capacity as the capability of a system to receive heat under a given
increase of T , dQ/dT . It used to be called specific heat. Under constant P we obtain

CP = (∂H/∂T )P,Ni = −T ∂ 2 G/∂T 2

P,Ni
= T (∂S/∂T )P,Ni (1.51)

The molar heat capacity is defined as CP /N . Furthermore, one defines the isothermal com-
pressibility and thermal expansion as

κT = − (∂V /∂P )T,Ni /V = − ∂ 2 G/∂P 2

T,Ni
/ (∂G/∂P )T,Ni (1.52)

α = (∂V /∂T )P,Ni /V = ∂ 2 G/∂T ∂P

Ni
/ (∂G/∂P )T,Ni (1.53)

The definitions of kT and α can be applied to the molar volume without change of the numerical
value but the results of Eq. 1.51 is proportional to the size. It is usually applied to the molar
Gibbs energy and the result is called molar heat capacity.

These three quantities are particularly important properties and can be determined experimen-
tally for various substances. Such information is compiled into tables that have been the primary
source for the evaluation of other thermodynamic data, mainly for Gibbs energy. In recent years
one has started to collect evaluated functions of G (P, T, Ni ) from such information and store
them into computerized databases. One may use automatic procedures based on the above re-
lations and similar ones to directly obtain all the properties needed for a practical application.
With advanced software it may not even be necessary to inspect or handle the values of those
properties. They may be directly applied to the practical situation and the output could give
the answer to a technical question.

Exercise 1.12.1. Evaluate the enthalpy of pure Mo at 2000 K relative 25◦ C and 1
bar.
Hint You may not need to use Eq. 1.50 because most kinds of thermodynamic soft-
ware have special procedures for the evaluation of the properties based on the first
derivatives of G.

17
CHAPTER 1. BASIC THERMODYNAMICS

Exercise 1.12.2. Evaluate the thermal expansion of an alloy of Fe with 0.5 mass-%
C at 1300 K and 1 bar from stored information on G for the fcc phase, which is the
stable phase under those conditions.
Hint Again you need the kind of software that can evaluate partial derivatives.

1.13 Adiabatic changes

We can rearrange the second law from Eq. 1.6 using Eq. 1.13,

dS = dQ/T + dip S = dQ/T + (1/T ) Ddξ (1.54)

During a very rapid compression of a material one may imagine that there is no time for heat
conduction nor for an internal process. The first condition means that the compression is adia-
batic, dQ = 0, and the second condition means that the compression is reversible, because there
is not sufficient time for an internal process, dξ = 0. Eq. 1.54 shows that these conditions make
dS = 0. An adiabatic and reversible change is isentropic. The change of temperature during
this kind of compression can be expressed as (∂T /∂P )S .

Exercise 1.13.1. Consider a shock wave traveling through a plate of iron, which
initially is at 20◦ C. Evaluate dT /dP .
Hint A shock wave is very fast and there is very little time for heat conduction. It
may happen that there is time for some dislocation movements and the wave may
leave the material slightly deformed. Neglecting such effects one can approximate the
compression and the release of pressure on the back side of the wave as isentropic.
One could then obtain dT /dP when the wave is approaching as (∂T /∂P )S . It is
directly obtainable from an advanced data bank system for thermodynamics. If such
a system is not available one must express the partial derivative in terms of the
properties available in tables, i.e. CP , α and κT , which are second derivatives of
G with respect to P and T . In that case you should thus transform (∂T /∂P )S to
derivatives where P and T are the variables. Since S is also involved you should
consider a function S (T, P ).
   
∂S ∂S
dS = dT + dP
∂T P ∂P T
For dS = 0
     
∂S ∂T ∂S
0= +
∂T P ∂P S ∂P T
and using Eqs. 1.51 and 1.53
 , , 2 
−∂ 2 G
    
∂T ∂S ∂S −∂ G Vα
=− =− 2
=
∂P S ∂P T ∂T P ∂P ∂T ∂T CP /T
It should be emphasized that V , α and CP vary with P and T and detailed infor-
mation is required in order to integrate this result to higher pressures. That can be
obtained from a database but integration is still laborious. Much can be gained by
using an advanced data bank system.

18
 CHAPTER 1. BASIC THERMODYNAMICS

1.14 State of reference and standard states

Thermodynamic modeling of the properties of solution phases will be described in Chapter

2. The chemical potentials of the components vary with composition and different models use
different functions of the composition variables. The chemical potentials of the pure components
are usually given as the molar Gibbs energy of the component in the same crystallographic
structure as the solution phase, denoted by ◦ Gα A where the superscript
◦
indicates that the
◦ α
quantity concerns the component in pure form. GA is thus used as a reference for the chemical
potential of A in the α solution. It may be regarded as a model-based reference. When
comparing the chemical potential of a component in two different phases one may need a common
reference. In Fe-Cr solutions one may like to compare the chemical potential of chromium with
the value obtained for pure bcc-Cr at the same P and T , i.e., at the current P and T . This is
the most stable form of chromium and it is common practice to use the most stable form of the
element as a common reference. It is often regarded as the standard state of the element and
may be denoted by ◦ Gstd A . It is as in the definition of the standard Gibbs energy of formation
in Eq. 1.45. The relation between two references is found by applying the fact that the actual
chemical potential, µα A , is not affected by the choice of reference made when one wanted to
express it with a number.

◦ α ◦ std ◦ ◦ std
µα Gα


A = GA + f (comp) = GA + A − GA + f (comp) (1.55)

◦ ◦ std


One will thus have to add the quantity Gα
A − GA to the function f (comp).

There is also a need of references for the description of Gibbs energy as function of P and T .
It must be independent of P and T and it is then possible simply to use the most stable state
of the element at normalized values of P and T . This may be denoted by ◦ GN A
PT
where NPT
◦
stands for Normalized P and T . Usually, one uses 298.15 K (25 C) and 1 bar or possibly 1 atm.
If there is a need of yet another kind of reference, it may simply be denoted by ◦ Gref
A .

The kind of reference, represented by the Gibbs energy of a state, is not sufficient for all purposes
because there are two contributions to Gibbs energy, an enthalpy part and an entropy part and
each one requires a reference. The enthalpy has no natural reference and it can be chosen
arbitrarily but should follow international agreement. One usually uses the enthalpy of the most
stable state at 1 bar and 298.15 K. For the entropy there is a natural zero point for the elements
at 0 K according to the third law of thermodynamics. The problem is that it is difficult to
determine experimentally the difference of entropy between 298.15 K and 0 K. Nevertheless,
there is a system called the G-HSER system that uses this zero point. A dataset based on this
system should thus be able to return the absolute values of entropy. In practice this may be
best satisfied for the most stable phase of the elements. The considerable uncertainty for other
states of the elements has practical consequences only at very low temperatures. The enthalpy
SER SER
reference for one mole of an element will be denoted by HA and data are given as GA − HA
SER
for a unary (one component) system and as Gm − H for a solution or compound of more
than one component where H SER represents an average over the components.

1.15 Duhem’s theorem

In the state of equilibrium under given P , T and Ni all the internal variables have obtained
their optimum values, which minimize the Gibbs energy of the whole system. In principle, the
Gibbs energy of a system in equilibrium is thus uniquely defined by a function of variables and, in
principle, they may be controlled from the outside, so-called external variables, G = G (P, T, Ni ).
This is true independent of how many phases are involved, although the G function is more
complicated the more phases there are. For a closed system, i.e. a system with a given content

19
CHAPTER 1. BASIC THERMODYNAMICS

of matter, i.e. all the Ni , the equilibrium state can vary only with P and T . When computing
the state of equilibrium using a computer program, one could thus give two conditions, primarily
the values of P and T , in addition to all Ni . This is called Duhem’s theorem. Altogether
this makes 2 + c conditions and it is interesting to see how the number of necessary conditions
is related to the degrees of freedom according to the Gibbs phase rule.

Exercise 1.15.1. Using a thermodynamic data bank system, determine the state of
equilibrium of an alloy with 20 mole-% Cu in Ag at 800◦ C and 1 bar.
Hint The size is not prescribed and you can choose any size, e.g. N = 1. With that
choice NCu = 0.2 and NAg = 0.8. Or you could enter the condition as N = 1 and
xCu = 0.2 or even NCu = 0.2 and xAg = 0.8.

Exercise 1.15.2. From the previous exercise you know that there will be two phases,
fcc and liquid (L). You will also know the amounts and compositions of the phases.
Omit the condition on the value of the temperature and accept the value of the Cu
content in L as a new condition for an attempt to determine the temperature.
Hint Of course, you should expect to recreate the same state of equilibrium, i.e. to
find T = 1073 K.

1.16 Characteristic state function and

Gibbs energy model

The state of a system can be defined by the values of a sufficient set of variables that describe the
conditions for the system. The state of a homogeneous system of a given internal structure can
be completely defined with a set of independent state variables. Gibbs energy is one of several
functions of such state variables which describe thermodynamic properties of a system. It has
the very useful property that all other thermodynamic properties can be expressed through it
and its derivatives with respect to the state variables P , T and Ni . The function G(P, T, Ni )
may be regarded as a characteristic state function and the Helmholtz energy, F (V, T, Ni ), is
another example. The sets of variables, P, T, Ni and V, T, Ni , are regarded as their respective
set of natural variables.

By measurements and modeling one may hope to construct an equation yielding the value of
Gibbs energy as a function of its natural variables, i.e. G(P, T, Ni ), containing various numerical
parameters. It could be used for calculating values of all thermodynamic properties at fixed
values of P , T and Ni .

In the modeling of Gibbs energy of a system, P , T and Ni are all regarded as external variables
because their values can be controlled by actions from the surroundings. However, there are
also internal variables but their values will spontaneously be adjusted to changes in the external
variables until a state of equilibrium has been established. That state represents the Gibbs
energy minimum under the given external conditions. In principle, the Gibbs energy function
G(P, T, Ni ) is thus well-defined through the model under equilibrium conditions. On the other
hand, it is often impossible to express that function analytically. In modeling, Gibbs energy is
given as an equation with internal variables in addition to the external variables, G(P, T, Ni , ξk ),
and it can thus be evaluated for any set of values, including both kinds of variables. Gibbs
energy as a characteristic state function only applies under equilibrium conditions but in an
actual calculation one gives start values to the internal variables and gradually adjusts their

20
 CHAPTER 1. BASIC THERMODYNAMICS

values until a minimum of G has been found and the initial function has been reduced to the
characteristic state function G(P, T, Ni ). The modeled form of the Gibbs energy with internal
variables is very valuable also away from the states of equilibrium because it can give information
on the driving force for the processes that lead to equilibrium. The function G(P, T, Ni , ξk ) will
simply be regarded as a thermodynamic model.

For a system with more than one phase, the model will be composed of functions like
Gα α α α
m (P, T, xi , ξ1 , ξ2 , . . .) for all the possible phases, each one multiplied by the amount of that
α
phase, N , which is another internal variable. The equilibrium of the complex system could be
obtained by minimizing that complicated expression with respect to all the internal variables.
This is what the thermodynamic data bank system does when you ask for the equilibrium to be
computed.

It may be noted that phase equilibria and phase diagrams as well as properties of systems at
equilibrium can all be described by intensive quantities, i.e. potentials or molar quantities. In
principle, it should thus be possible to obtain all such information without involving the size of
the system. However, available methods of computation based on minimizing the Gibbs energy
normally require access to the size, which may be chosen arbitrarily.

Exercise 1.16.1. For the Ag-Cu system, try to use the condition that fcc and L
must be present in an equilibrium at 800◦ C and 1 bar.
Hint A way to define that both phases are present would be to require a certain
amount of each one, e.g. 1 mole. According to Duhem’s theorem it then remains to
define 2 + c − 2 = c = 2 conditions, which should be the values of P and T .

21
CHAPTER 1. BASIC THERMODYNAMICS

22
Chapter 2

Solution Models

2.1 Constitution and constituents

The basic features of thermodynamics were dealt with in Chapter 1 and the effect of variations
in the composition of phases was an important topic. Variations with the change of the content
of components were described, often the number of atoms of various elements expressed through
the numbers, Ni , or mole fractions, xi . Chapter 2 will now deal with the modeling of solution
phases. Many phases have more than one kind of sites for the atoms. The sites are then divided
into sublattices and a model may be based on the assumption that atoms of different elements
prefer to dissolve in different sublattices. Each atom will have a different effect on the properties
of the phase depending on in what sublattice it resides. An essential part of the model can then
be described with a formula and (A, B)1 (B, C, D)2 is an example with two sublattices of which
the second one has twice as many sites as the first one and it can dissolve atoms of the B, C and
D elements whereas the first one can dissolve A and B atoms. This situation will be regarded as
the constitution of the phase. The indices 1 and 2 are the stoichiometric coefficients of the
phase. A and B are the constituents of the first sublattice and B, C and D are the constituents
of the second sublattice. B can thus reside in both sublattices and will then be regarded as
two different constituents although they are atoms of the same element. The contents of the
constituents on each sublattice will be described with their fractions within the sublattice. They
are called site fractions and are denoted yjαk for constituent j in the k:th sublattice of phase α.
They are defined as yjαk = Njαk / Niαk . Naturally, for each sublattice,
P P αk
yi = 1. An ionized
atom may also occupy a site and is then regarded as a constituent different from the neutral
atom of the same element even if they reside in the same sublattice.

It may also happen that atoms combine to form molecules or other kinds of associates. When
a solution phase is modeled with an associate in the same sublattice as single atoms, they are
all regarded as constituents of that sublattice.

A model of a phase with two or more sublattices may be given as an analytical function of the
site fractions, not mole fractions. It will normally be possible to evaluate the mole fraction of a
phase from information on the site fractions and the stoichiometric coefficients. It should finally
be mentioned that anything that can occupy a site is regarded as a species irrespective of where
it is situated. A, B, C and D in the above example are thus four different species but an A+1 ion
would not be the same species as the neutral A atom. An A1 B2 associate (or molecule) would
also be regarded as a species if it could occupy a single site according to the model.

23
CHAPTER 2. SOLUTION MODELS

Exercise 2.1.1. The mole fractions of the components in a phase can be evaluated
from the site fractions if one considers the stoichiometric coefficients. Evaluate the
0 00
mole fraction of component B in the phase (A, B)1 (B, C, D)2 if yB = 0.2 and yB =
0.05. P
Hint Apply the definition xB = NB / Ni to one formula unit.

Exercise 2.1.2. Give the formula showing the constitution of an A1 B2 phase if one
has been able to decrease the mole fraction of B to 0.5 by some A atoms entering as
a second constituent of the second sublattice.
Hint PExpress the formula in a general way as A1 (A1−z Bz )2 and apply xB =
NB / Ni .

2.2 Ideal solutions and related non-ideal solutions

In a substitutional solution all the components can substitute for each other because they
occupy the same kind of lattice sites. With statistical methods one can show that already the
physical mixing of the
P components in a substitutional solution contributes to the molar entropy
by the amount −R P xi ln xi if the mixing is random. The effect on the molar Gibbs energy will
thus be −T S = +RT xi ln xi . If this is the only thermodynamic effect of the mixing, then the
chemical potential of each component in a phase α is given as

◦ α
µα α
i = Gi + RT ln xi (2.1)

◦ α
This is the ideal solution model, xα i is the mole fraction of component i and Gi is the
molar Gibbs energy of pure component i of the same structure (phase) as the solution and at the
current P and T. Pure i is often regarded as one of the end-members of the solution phase and
◦ α
Gi is a model-based reference for component i in the phase α. It may be selected in a different
way by applying another model. This will always happen if the solution cannot be extended all
the way to pure i. Inserting µα
i in Eq. 1.21 we find for the molar Gibbs energy of this solution,

X X X
◦ α
Gα
m = xα α
i µi = xα
i Gi + RT xα α
i ln xi (2.2)

The first summation of terms on the right hand side represents the average of the references
for the components and this summation may thus be regarded as the frame of reference for the
mixture. For a binary solution it may be regarded as a line of reference. That is illustrated in
Fig. 2.1, which shows the same case as Fig. 1.2 .

The last summation in Eq. 2.2 is recognizes as the effect of the ideal entropy of mixing in
the Gibbs energy. All the terms are negative because all mole fractions are less than unity.
This summation represents the distance of the Gm curve below the line of reference. Fig. 2.2
illustrates the same situation for a ternary
P ◦ solution phase. The triangle at the top represents
the model-based plane of reference, xi Gi .

In reality, one should always expect some deviation from ideality and it is often expressed with
an excess Gibbs energy, E Gαm , defined from

24
 CHAPTER 2. SOLUTION MODELS

Figure 2.1: Molar Gibbs energy diagram for an ideal binary α phase, illustrating the line of
reference and the construction to obtain the chemical potentials for a given composition.

Figure 2.2: Molar Gibbs energy diagram for a ternary phase, illustrating the plane of reference
and the construction to obtain the chemical potentials for a given composition.

25
CHAPTER 2. SOLUTION MODELS

X X X
◦ α
Gα
m = xα α
i µi = xα
i Gi + RT xα α
i ln xi +
E
Gα
m (2.3)

Of course, E Gα
m depends on the composition and several models have been proposed. By defini-
tion, E Gα
m ⇒ 0 as the composition approaches a pure component, e.g. one of the three corners in
Fig. 2.2. Deriving an expression for the chemical potential of a component i, we obtain instead
of Eq. 2.1,

◦ α
µα α
i = Gi + RT ln xi +
E
Gα
i (2.4)

E
Gα i is the partial excess Gibbs energy for component i in the α phase and is obtained by
applying Eq. 1.30 to Gα m from Eq. 2.3. If an analytical expression has been chosen for
E α
Gm
E α
as function of composition, then one can derive an expression for Gi by applying Eq. 1.33 to
E α
Gm . An example will be presented in Section 2.5.

There are many special models. When states of equilibrium or values of thermodynamic proper-
ties at equilibrium are calculated with computer, then the time and cost depend only marginally
on whether the data are expressed with the simple ideal solution model or more sophisticated
ones. However, it can make an enormous difference if one calculates by hand or even with a hand
calculator. The ideal solution model was used extensively before the advent of the computer.
It still predominates in teaching where exact results are not as important as the demonstration
of how various factors may affect the results. This is particularly true for reactions between
molecules in a solution, which will be considered in the next section.

◦ α
Numerical values can be obtained only relative to references, e.g. as the differece µα
i − Gi
α ◦ ref
from Eq. 2.4 or as µi − Gi by introducing another reference. One may then obtain numerical
values of

◦ ref ◦ α ◦ α
µα α
i − Gi = µi − Gi + ∆ref Gi (2.5)

where ∆ref◦ Gα ◦ α ◦ ref

i = Gi − Gi . This is identical to Eq. 1.45 when applied to a single component.
The molar Gibbs energy of the α phase can thus be stored as the numerical values of ∆ref◦ Gα
i in
addition to the excess parameter E Gα
m .

Exercise 2.2.1. In the regular solution model of a binary A-B system one assumes
that E Gm = LAB xA xB . Evaluate the partial excess Gibbs energy of component B.
Hint You can apply Eq. 1.32 directly by neglecting the other terms in Gm or Eq. 1.33
by identifying Am with E Gm .

2.3 Chemical activity and activity coefficient

Each kind of solution requires its own kind of model. Many models have similarities with the
ideal solution model in Eq. 2.1 but the mole fraction is replaced by a quantity called activity or
chemical activity, aB . Thermodynamically it is defined by its relation to the chemical potential.

µB = ◦ GB + RT ln aB (2.6)

26
 CHAPTER 2. SOLUTION MODELS

◦
GB can here be the Gibbs energy of any real or hypothetical state of pure B, selected for
expressing the activity. It defines the numerical values of the activity by making ln ab = 0, i.e
aB = 1 when µB = ◦ GB and it is thus a reference for the activity.

Let us now consider the case where ◦ GB is chosen as ◦ Gα B , the value for pure B of the same
phase. Then aB = xB if the solution is ideal as defined by Eq. 2.1. Real solutions deviate from
the ideal behaviour and one may write aB = fB xB where fB is the activity coefficient. It
may be regarded as a correction factor to xB and is normally a function of composition. Eq, 2.6
yield

µB = ◦ Gα α
B + RT ln aB
(2.7)
= ◦ Gα α α
B + RT ln fB + RT ln xB

Comparison with 2.4 yields the relation between the two approaches

RT ln fB = E Gα α E α


B i.e. fB ≡ exp GB /RT (2.8)

E
We will soon discuss the modelling of Gα
m.

When comparing with other phases it is essential to use common references for the chemical
potentials, e.g. by chosing standard states for the elements at the current P and T or global
references at a particular pressure and temperature, the SER reference being an example. See
Section 1.14. We will here use the notation Gref
B for any such reference. When combining pieces
of thermodynamic information on a system from different sources it is essential to check that they
are based on the same set of references. If that is not the case, it is necessary to transform
P α ◦ local the
data, which were given relative to a local set of references, e.g. as ∆Gα m ≡ G α
m − x i G i ,
into ∆ref Gαm , where the superscript ref now indicates that values refer to any set of agreed
references. It is obtained as

X X  
◦ ref
∆ref Gα α
m ≡ Gm − xα
i Gi = ∆Gα
m+ xα
i
◦
Glocal
i − ◦ Gref
i (2.9)

◦
Glocal
i − ◦ Gref
i is the difference in Gibbs energy of pure i in the two states of reference. When
comparing the activity of an element B in two different solution phases, one often finds that the
reference for pure B in each phase has been taken from the end-member B in that phase, i.e. ◦ Gα B
and ◦ GβB , respectively. It is then common to prefer the most stable state as the common reference,
◦ β
say α. See Section 1.14. The difference ◦ Gα B − GB is regarded as the lattice stability of pure
B in the less stable β state relative to the most stable state. It is thus negative by definition.

Exercise 2.3.1. One has studied solutions of Mn in fcc Fe-Ni and bcc Fe-Cr alloys
at 1200 K and has expressed the Mn activity using pure fcc-Mn and pure bcc-Mn,
respectively, as references. For both alloys, Fe-Mn-Ni and Fe-Mn-Cr, one obtained
aM n = 0.03. What alloy had the highest activity?
Hint To answer this question you must use a common reference and you may use any
state as the common reference, say fcc-Mn. With a data bank system you can simply
ask for the activity relative fcc-Mn instead of bcc-Mn as long as you have kept data
for the fcc phase. Of course, you could just as well solve thehproblem analytically
 byi
f cc−ref bcc−ref ◦ bcc ◦ f cc
changing to the fcc reference using aM n = aM n ·exp GM n − GM n /RT
but then you must first evaluate the difference for Mn in the two states at 1200 K.

27
CHAPTER 2. SOLUTION MODELS

2.4 Excess Gibbs energy

E
GαB in Eq. 2.7 was defined in Eq. 2.4. It is regarded as the partial excess Gibbs energy for
component B. For a binary A-B solution there would be partial excess Gibbs energies for both
components and they are not independent of each other. That will automatically be taken care
of by starting from an expression for the excess molar Gibbs energy, E Gα m , defined in Eq. 2.3
and deriving expressions for the various partial excess Gibbs energies by applying Eq. 1.33 to
E α
Gm . Naturally E Gα
m = 0 for all the pure elements if all the references have been chosen from
the pure elements in the phase under consideration, here to be denoted by ◦ Gα j . For this choice,
the simplest approach for a binary A-B solution would be

E
Gα α α
m = LxA xB (2.10)

With L = 0 this model degenerates to the ideal solution model, with L independent of com-
position but not of temperature. This is called the regular solution model. Some more
complicated solution models describe L as a power series in xα α
A − xB .

E
When applied to Gα
m from Eq. 2.10, Eq. 1.33 yields
E
Gα α 2
A = L(xB ) and

◦ α
µα α
A = GA + RT ln aA

= ◦ Gα α
A + RT ln xA +
E
Gα
A (2.11)
◦ 2
= Gα
A + RT ln xα
A + L (xα
B)

◦ α
µα α
B = GB + RT ln aB

= ◦ Gα α
B + RT ln xB +
E
Gα
B (2.12)
◦ 2
= Gα
B + RT ln xα
B + L (xα
A)

Eq. 2.3 yields

α 2
fBα = e[L(xA ) /RT ]
(2.13)

α 2
aα α α α [L(xA )
B = fB xB = xB e
/RT ]
(2.14)

It is easy to see that Gm = xA µA + xB µB will contain the excess term given by Eq. 2.10.

Fig. 2.3 a illustrates the effect of various values of the regular solution parameter on the molar
Gibbs energy in a binary solution and Figs. 2.3 b and 2.3 c present the corresponding variations
of the chemical activities when the pure elements in the same phase have been chosen as standard
states for the activities.

The high activity values in Figs. 2.3 b and c obtained with large L, cannot represent stable
states. There would even be a positive driving force for the formation of pure B from a solution
with aB > 1. To examine this problem we should go back to Fig. 2.3 a and draw a double
tangent to one of the curves for large L values. Fig. 2.4 illustrates that all states on the curve
between the tangent points could decrease their Gibbs energy by separating into two phases with
compositions on those points. The average Gibbs energy of such a “mechanical mixture” will

28
 CHAPTER 2. SOLUTION MODELS

Figure 2.3: a) The binary regular solution model for a series of values of the interaction energy.
b) The activity of A in the regular solutions illustrated in a). c) The activity of B in the regular
solutions illustrated in a).

Figure 2.4: Gibbs energy diagram illustrating the construction to find a miscibility gap. The
Gibbs energy curve is taken from Fig. 2.3

fall on the tangent. There is thus a miscibility gap between those points. For all alloys within
the miscibility gap there is a positive driving force for decomposition into two phases, one rich
in A and the other rich in B. Compare with Fig. 1.3.

For a ternary solution the regular solution model yields

X X
Gα
m = xi ◦ Gα
i + RT xi ln xi + LAB xA xB + LBC xB xC + LCA xC xA (2.15)

◦ α
µα α
A ≡ GA = GA + RT ln xA + LAB xB (1 − xA ) − LBC xB xC + LCA xC (1 − xA ) (2.16)

◦ α
µα α
B ≡ GB = GB + RT ln xB + LAB xA (1 − xB ) + LBC xC (1 − xB ) − LCA xC xA (2.17)

◦ α
µα α
C ≡ GC = GC + RT ln xC − LAB xA xB + LBC xB (1 − xC ) + LCA xA (1 − xC ) (2.18)

29
CHAPTER 2. SOLUTION MODELS

Figure 2.5: Comparison between the ideal and dilute solution models for a regular solution with
L = −2RT . The dilute solution model is based on the tangent at infinite dilution, i.e. Henry’s
law.

Exercise 2.4.1. An A-B solution phase can be approximated as a regular solution

with LAB = 20 kJ/mol at 800◦ C. Evaluate the activity coefficient for B at xB = 0.40.
Hint You must solve this exercise analytically if you cannot create your own database
for the regular solution model. Having evaluated the activity coefficient, fB , you
should also evaluate aB .

2.5 Dilute solutions

For a discussion of methods to treat information on activities at the two ends of a binary solution,
the curve for L = −2RT from Fig. 2.3 c is again shown in Fig. 2.5. It is interesting to note that
the modelled activity of B approaches the ideal solution model close to pure B. That is called
Raoult’s law. For small xα α
A , i.e., xB close to 1, it is demonstrated by Eqs. 2.12 and 2.14 which
yields

µα ∼◦ α α
B = GB + RT ln xB + 0 (2.19)

aα ∼ α 0 α
B = xB e = xB (2.20)

By extending the initial tangent to the activity curve in Fig. 2.5, one obtains Henry’s law for
dilute solutions. It is also demonstrated by Eqs. 2.12 and 2.14 for xα A close to 1, i.e. for small
xα
B,

µα ∼◦ α α
B = GB + RT ln xB + L (2.21)

aα ∼ α α α
B = xB exp(L/RT ) = fB xB (2.22)

where fBα = exp(L/RT ) (2.23)

30
 CHAPTER 2. SOLUTION MODELS

There are many cases where one has information only from a dilute range of a solution and
information may then be described with Henry’s law using ◦ Gα B as a reference for the chemical
potential of B. L can thus be evaluated from the initial slope of the B activity and be inserted
in Eq. 2.12 to yield a regular solution extrapolation to higher B contents. However, ◦ GαB would
not be known if there is no information on pure B in the α state. The combination ◦ Gα B +L
in Eq. 2.21 instead of L would then be the unknown parameter which could be evaluated from
Henry’s law. If one notices some deviation from Henry’s law at the highest B contents available,
then one can separate ◦ Gα B and L to obtain the best fit of the limited information and again
apply Eq. 2.12 for some extrapolation to higher B contents.

By tradition it is common to apply Henry’s law with any convenient measure of B content,
e.g. mole % or mass % or even molality. Denoting mass % with wB one writes Henry’s law as
aα α α
B = γB wB . In this case one usually uses the standard state of B as the reference.

◦ std α ∼ ◦ std
µα α α
B ≡ GB + RT ln aB = GB + RT ln wB + RT ln γB (2.24)

α
The activity coefficient has here been denoted by γB to make it evident that it differs from fBα .
eq
If wB is the B content in equilibrium with pure B in its standard state, i.e., at the activity 1,
α eq
then γB = 1/wB if Henry’s law applies.

Exercise 2.5.1. The experimental information on many terminal solutions is very

limited and often one only gives an ideal solution description based on the Gibbs
energy of the solute element in a hypothetical state with the same structure as the
solvent. On the other hand, there are methods to estimate that energy relative to the
stable state of the solute element in pure form. Study the description of the solution
of Si in fcc-Al in your data bank.
1. Plot the activity of Si in the fcc solution (suspend all other phases) at 500◦ C
all the way to pure Si using fcc-Si as reference.
2. Change the reference state to the stable state of Si at the current T and plot
the diagram again.
3. From the activity of pure Si obtained from the first diagram compared to the
second one, you may calculate the difference in Gibbs energy between the two
reference states.
Hint
2. To find the stable state of Si you may for instance calculate the binary phase
diagram.

3. Applying the definition of the chemical potential for the two cases you get
(1) (1) (2) (2)
µSi = ◦ GSi + RT ln aSi = ◦ GSi + RT ln aSi .

2.6 Phases with sublattices

There are many cases where all the lattice sites are not equivalent. Then one distinguishes
between two or more sublattices and various elements may prefer different sublattices. One may
define an ideal solution for such phases by requiring that there is random mixing within each
sublattice and no excess Gibbs energy term. For a solution phase (A, B)a (C, D)b there will be
four end-members, Aa Cb , Ba Cb , Aa Db and Ba Db . The ideal entropy of mixing is determined

31
CHAPTER 2. SOLUTION MODELS

by the fractions of the elements within each sublattice, the so-called site fractions yi , defined
for each sublattice separately. Quantities related to the sublattices can be identified by 0 and 00.
Since (NA + NB ) /a = (NC + ND ) /b we obtain (xA + xB ) /a = (xc + xD ) /b by dividing with
the total number of atoms, N . The relation to the ordinary mole fractions is

0 NB xB
yB = = (2.25)
NA + NB xA + xB

00 NC (a/b) NC (a/b) xC
yC = = = (2.26)
NC + ND NA + NB xA + xB

A simple sublattice solution model without any excess terms is defined for one mole of formula
units,

XX X X
Gm = yi0 yj00 ◦ Gia jb + aRT yi0 ln yi0 + bRT yj00 ln yj00 (2.27)

◦
Gia jb represents a model-based reference for the ia jb component compound and it is regarded
as a compound energy of an end-member. The double summation of terms is a weighted average
of all the compound energies and may be regarded as a model-based surface of reference. yi0 yj00
is the fraction of ij bonds between
P the
P two sublattices, assuming random mixing within each
0 0 00 00
sublattice and one can show that ia jb = 1 since yA + yB = 1 and yC + yD = 1. The other
two terms represent the entropy contributions caused by random mixing within each sublattice.

This model is regarded as the ideal compound energy model. It provides a basis for more com-
plicated and sophisticated models and in that connection it is called the Compound Energy
Formalism, CEF. The chemical potentials of the elements cannot be defined for such a phase
alone because Eq. 1.17 cannot be applied when there is a stoichiometric constraint, in this case
(NA + NB )/a = (NC + ND )/b. One component cannot be varied when all the others are kept
constant. Instead of regarding the constituents as the components of the system one may thus
regard the end-member compounds as components. For their chemical potentials one can derive
an equation with some resemblance to Eq. 1.32.

∂Gm ∂Gm 0 ∂Gm 0 ∂Gm 00 ∂Gm 00 ∂Gm

µAa Cb = Gm + 0 + 00 − yA 0 − yB 0 − yC 00 − yD 00 (2.28)
∂yA ∂yC ∂yA ∂yB ∂yC ∂yD

Without any excess terms, the result will be

µAa Cb = ◦ GAa Cb + yB
0 00
yD ∆◦ GAa Cb +Ba Db + aRT ln yA
0 00
+ bRT ln yC (2.29)

where

∆◦ GAa Cb +Ba Db = ◦ GAa Db +◦ GBa Cb −◦ GAa Cb −◦ GBa Db (2.30)

This quantity is often regarded as the Gibbs energy of the reciprocal reaction between pure
compounds, Aa Cb + Ba Db → Aa Db + BP a Cb . Fig. 2.6 illustrates the model-based surface of
reference for a reciprocal solution phase yi0 yj00 ◦ Gia jb and it should be emphasized that this
surface is not planar as the one in Fig. 2.2 unless ∆◦ GAa Cb +Ba Db = 0 in Eq. 2.30, i.e. ◦ GAa Dd +◦
GBa Cb = ◦ GAa Cb +◦ GBa Db . To save the properties of the reciprocal phase it is thus sufficient to

32
 CHAPTER 2. SOLUTION MODELS

Figure 2.6: Surface of reference for a reciprocal phase in a quaternary system.

store information on the four end-member compounds, which is done as described for ∆ref GφAa Bb
in Eq. 1.45.

Through Eq. 2.29 and three similar equations one can thus evaluate four chemical potentials,
µAa Cb , µAa Db , µBa Cb and µBa Db . However, by combining them, two and two, one can show that

µAa Cb + µBa Db = µAa Db + µBa Cb (2.31)

Actually, this relation can be derived directly by applying Eq. 1.27 since each side of Eq. 2.31 will
then be equal to aµA + aµB + bµC + bµD . This kind of system is regarded as a reciprocal system
and due to the Gibbs–Duhem relation, applied to the elements, it behaves as a ternary system
although it contains four elements. Only three of them can be regarded as independent of each
other. There is thus a compositional constraint and the number of components in the Gibbs
phase rule should be decreased by subtracting the number of constraints. This fact is illustrated
by the composition tetrahedron in Fig. 2.7 where the reciprocal solution phase is restricted to
the planar square. Only three of the chemical potentials in Eq. 2.31 can be independent. The
composition square in Figs. 2.6 and 2.7 can thus be compared with the composition triangle in
Fig. 2.2.

If one would like to consider the chemical potentials of the elements then one must fix one of
them to a constant value that may be chosen arbitrarily, say µD . Eq. 1.27 then yields

aµA = µAa Db − bµD (2.32)

aµB = µBa Db − bµD (2.33)

bµC = µAa Cb − µAa Db + bµD (2.34)

One can insert Eq. 2.29 and three similar equations in order to obtain analytical expressions for
the chemical potentials of the elements. One may regard µD as a reference because the other three

33
CHAPTER 2. SOLUTION MODELS

Figure 2.7: Composition tetrahedron for a quaternary system with the composition square for
a reciprocal solution phase, (A, B)1 (C, D)1 .

chemical potentials are given relative to µD and it is even possible that µD is fixed by interaction
with the surrounding. The component D may for instance be oxygen and the surrounding may
be a gas phase of a given oxygen pressure. All the other three chemical potentials are then
defined by the composition of the reciprocal phase.

Deviations from ideal behaviour can be described with excess Gibbs energy terms containing
regular solution parameters representing interactions between atoms on the same sublattice. In
0 00 00
the present case the terms would be of the following type yA yC yD LA:C,D . Higher order terms
can also be introduced.

If all elements can dissolve in both sublattices, at least to some degree, then one must include
many more compound energies e.g. ◦ GA:B . The model could thus turn very complicated. How-
ever, with a computer program there should be no practical problem once all the parameter
values have been evaluated.

It should be noted that the chemical potentials of all the elements can be evaluated from the
model if at least one element, e.g. A, dissolves in both sublattices because µA:A is equal to the
chemical potential of the end-member Aa Ab which is a + b atoms of A in the particular structure
of Aa Cb . One may then define a model-based reference ◦ Gref ◦
A = GA:A / (a + b)

When the compound energy formalism is applied to a solution between two compounds, which
differ on one sublattice only, e.g. Aa Cb and Aa Db , the result will be much simpler because
0
yA = 1. Without any excess terms Eqs. 2.27 and 2.29 will give

00 ◦ 00 ◦ 00 00 00 00
Gm = yC GAa Cb + yD GAa Db + bRT (yC ln yC + yD ln yD ) (2.35)

µAa Cb = ◦ GAa Cb + bRT ln yC

00
(2.36)

The compound energy formalism thus reduces to the ordinary ideal solution model by the site
fraction yi00 playing the role of the mole fraction xi . One may thus regard the compound energy

34
 CHAPTER 2. SOLUTION MODELS

formalism, CEF, as a generalization of the ordinary solution model. Numerical values of the
◦ γ
chemical potentials of component compounds can be defined as µAa Cb − a ◦ Gα
A − b GC or as
◦ ref ◦ ref
µAa Cb − a GA − b GC .

It should be emphasized that with the CEF, the site fractions are defined for the elements A,
B etc., which are the constituents, whereas the thermodynamic quantities are defined for the
end-member compounds.

Exercise 2.6.1. Suppose a = b = 1 and ◦ GAC = 10, ◦ GAD = 8, ◦ GBC = 2,

◦
GBD = 20 (kJ/mol) in Eq. 2.27. Derive an equation for the line of reference along
0
the AD-BC diagonal using yB as the variable.
0 00 0 0
Hint Start by expressing yA , yC and yD as functions of yB along the diagonal.

2.7 Interstitial solutions

Interstitial solutes reside in sites between the host atoms. One may define those sites as be-
longing to an interstitial sublattice, which is initially occupied by vacancies. Such solutions can
be described with the compound energy formalism and the vacancies may be regarded as an
additional species and constituent. We may identify the vacancies with the D species in section
2.3 and the interstitials with the C species. The Aa Vab compound will simply be a moles of A
with all interstitial sites vacant, i.e. a moles of pure A. From Eq. 1.27 we have

µAa V ab = aµA + bµV a (2.37)

and we find that µV a = 0. For a simple interstitial solution Aa (C, V a)b we get

µAa Cb = aµA + bµC (2.38)

µAa Cb − µAa V ab = aµA + bµC − aµA − bµV a = bµC (2.39)

µC = (1/b) (µAa Cb − µAa V ab )

= (1/b) (◦ GAa Cb + bRT ln yC
00
− ◦ GAa V ab − bRT ln yV00 a ) (2.40)
◦ ◦ 00 00
= (1/b) ( GAa Cb − a GA + bRT ln (yC / (1 − yC )))

For the solvent we obtain

µA = (1/a) µAa V ab = (1/a) [◦ GAa V ab + bRT ln (1 − yC

00
)] (2.41)

35
CHAPTER 2. SOLUTION MODELS

Exercise 2.7.1. The interstitial solution of C in fcc-Fe can be represented by the

formula F e1 (V a, C)1 . Using a database compute and plot µC and aC as functions
of the C content at 1200 K and 1 bar from 0 to 1.5 mass-% C, using graphite as
00
reference. Use the mole fraction of C, not yC .
Hint Of course, the computer can directly evaluate xC and plot the diagram.

2.8 The ideal gas model

For a monatomic species the ideal gas model predicts that

µJ = KJ + RT ln PJ (2.42)

PJ is regarded as the partial pressure of species J but is actually defined as yJ P where yJ is

the fraction of J relative to all other species in the gas and P is the total pressure. Eq. 2.42 can
thus be written as

µJ = KJ + RT ln P + RT ln yJ (2.43)

The ideal gas thus behaves as an ideal solution of the species but KJ + RT ln P contains a much
stronger pressure dependence through RT ln P than ◦ Gi in Eq. 2.1 does for condensed phases.
For a gas species one usually defines the pure species at a pressure of 1 bar as the reference at
the current temperature, i.e. the standard state. Then KJ is identical to ◦ GJ (1 bar, T ). With
that reference Eq. 2.42 can thus be written as

µJ = ◦ GJ (1 bar, T ) + RT ln PJ (2.44)

Of course, this requires that one expresses partial pressures in bar, not pascal, which is normally
required. For gaseous species we will let ◦ indicate that the species is in pure gaseous form, at
1 bar and at the current temperature. We will thus omit the information “(1 bar)” from that
notation. We get a similar expression for diatomic J2 species,

µJ2 = ◦ GJ2 + RT ln PJ2 (2.45)

where ◦ GJ2 is the Gibbs energy of one mole of the J2 species at 1 bar and the current temperature.
Between monatomic and diatomic species we have µJ2 = 2µJ at equilibrium according to Eq. 1.27
with a = 2 and b = 0. From information on PJ2 one can then evaluate

µJ = 0.5 (◦ GJ2 + RT ln PJ2 ) (2.46)

We could thus evaluate the equilibrium value of PJ from the current value of PJ2 by combining
Eqs. 2.44 and 2.46.

PJ2 = PJ2 exp [(◦ GJ2 − 2 ◦ GJ ) /RT ] (2.47)

36
 CHAPTER 2. SOLUTION MODELS

300

∆GO=HOSER-0.5oGO2(T) kJ
250

200

150

100

50

0
0 500 1000 1500 2000
Temperature [oC]

Figure 2.8: Difference between the SER reference and the reference for oxygen based on pure
O2 gas at the current temperature. The difference depends strongly on the temperature.

For the gaseous elements hydrogen, nitrogen and oxygen, the diatomic species is much more
abundant than the monatomic species. In these cases one often prefers to use 0.5 ◦ GJ2 as
standard state for the element J and use it also when it appears in other phases. In such cases
PJ2 is the partial pressure in an imagined atmosphere where J and J2 are in equilibrium. On
the other hand, in tables and computerized databases it is common to use the state of reference
defined at 1 bar and 25◦ C, e.g. the SER system for which Gref J = HJSER (1 bar, 298.15 K). It
◦
is often necessary to transform from the 0.5 GJ2 reference to another one and that is done
by adding ∆GJ = Gref J − 0.5 ◦ GJ2 (T ). This transformation term varies with temperature as
illustrated in Fig. 2.8 for changing the values of oxygen to the SER system.

When applying the law of mass action to a gas, it is advantageous to insert partial pressures
instead of yi fractions, which cannot be used without information on the total pressure. As an
example, for the gaseous reaction CO + 0.5O2
CO2 we obtain the equilibrium from µCO2 =
µCO + 0.5µO2 and inserting expressions for the chemical potentials according to Eq. 2.43 we
would get

yCO2 √
= P exp −∆◦ Greaction /RT = K y (P )


√ (2.48)
yCO yO2

where

∆◦ Greaction = ◦ GCO2 − ◦ GCO − 0.5◦ GO2 (2.49)

All the ◦ G values are here given for 1 bar and the current temperature. The equilibrium “con-
stant” would thus depend on the√total pressure, P . By instead using expressions according to
Eq. 2.44 we would eliminate the P factor and get

37
CHAPTER 2. SOLUTION MODELS

PCO2
= exp −∆◦ Greaction /RT = K P


p (2.50)
PCO PO2

Exercise 2.8.1. What is the partial pressure of monatomic H in a hydrogen gas

when in internal equilibrium at 1 bar and 2000 K?

SER
Exercise 2.8.2. Evaluate µO − HO and µO − 0.5◦ GO2 (T ) for air with 20% O2
at 1 bar and 1000 K.

Exercise 2.8.3. Find the value of the standard Gibbs energy of formation of
monatomic O from O2 at 1000 K.

2.9 Vapour pressure

Even a solid element M may have a vapour pressure, although usually very low. In that case,
the standard state used for the gas would be based on the solid state at 1 bar and the current T
and the chemical potential of the monatomic species in the gas, given by the ideal law for gases
through Eq. 2.44, would be

µM = ◦ Ggas ◦ std ◦ std ◦ gas

M + RT ln PM ± GM = GM + ∆f GM + RT ln PM (2.51)

◦ gas
µM − ◦ Gstd


PM = exp M − ∆f GM /RT (2.52)

where ∆◦f Ggas ◦ gas ◦ std

M = GM − GM is the standard Gibbs energy of formation of the monatomic M
species in a gas from solid M if ◦ Gref ◦ solid
m is chosen as GM at the current temperature. Similarly,
the partial pressure of the M2 constituent in the gas is

◦ gas
µM2 − 2 ◦ Gstd


PM2 = exp M − ∆f GM2 /RT (2.53)

where ∆◦f GgasM2 =

◦ gas
GM2 − 2 ◦ Gstd
M . There may even be more complex species. It is not self-
evident whether one should define the vapour pressure of an element M as the sum of the partial
pressures of all species or as a measure of the total content of M in the gas. As an example,
Fig. 2.9 illustrates the partial pressures of various gas species of sulphur over pure liquid sulphur.

Exercise 2.9.1. Find the partial pressures of Fe1 and Fe2 , respectively, in equilib-
rium with solid fcc-Fe and bcc-Fe, respectively, at 1750 K. Suppose the solid phase
is under a pressure of 1 bar caused by an inert atmosphere. From the results, decide
whether bcc or fcc Fe is most stable.

38
 CHAPTER 2. SOLUTION MODELS

Figure 2.9: The equilibrium partial pressures of various sulphur species over liquid sulphur. S8
is the most abundant species. Monatomic S is very rare and falls below the diagram.

2.10 Thermal vacancies

In a solid metal all the atoms are located in lattice sites but some sites are vacant. The number
of vacancies increases with increasing temperature and one talks about thermal vacancies. In
a simple thermodynamic model one considers the vacancies as a substitutional solute because
they can substitute for the atoms. The formula for one mole of lattice sites will thus be written
as (A, V a)1 and the following model for the Gibbs energy per mole of formula units as obtained
from the regular solution model will be

Gm = yA ◦ GA + yV a ◦ GV a + RT (yA ln yA + yV a ln yV a ) + yA yV a LA,V a (2.54)

yj is defined as mole j per mole of formula units, i.e. mole per mole of lattice sites. When
examining this model one must realize that it concerns a unary system with A as the only
component. The site fraction yV a is thus an internal variable. To find its value at equilibrium one
could study a partial derivative for which the content of A is kept constant. It is then necessary
to use an expression of Gibbs energy as function of NA and NV a . It is related to Eq. 2.54
through Gm = G (NA , NV a ) / (NA + NV a ). By further introducing yj = Nj / (NA + NV a ) we
obtain after omitting the excess term for simplification,

G (NA , NV a ) = (NA + NV a ) Gm
 
◦ ◦ NA NV a
= NA GA + NV a GV a + RT NA ln + NV a ln
NA + NV a NA + NV a (2.55)
◦ ◦
= NA GA + NV a GV a
+ RT [NA ln NA + NV a ln NV a − (NA + NV a ) ln (NA + NV a )]

39
CHAPTER 2. SOLUTION MODELS

Using a partial derivative we can now find the equilibrium.

 
∂G
= ◦ GV a + RT [ln NV a − ln (NA + NV a )]
∂NV a NA (2.56)
= ◦ GV a + RT ln yV a = 0

(∂G/∂NV a )NA looks as a definition of the chemical potential of vacancies and we will indeed
accept this. However, since NV a is an internal variable, it belongs to the kind of variables
represented by ξ in Section 1.4 for which D = −∂G/∂ξ = 0 at internal equilibrium. We thus
obtain the equilibrium content of vacancies as

yVeqa = exp (− ◦ GV a /RT ) (2.57)

It is not necessary to discuss the nature of ◦ GV a . It is sufficient to accept that it is just an

experimental parameter and its numerical value is obtained from information on yVeqa using
Eq. 2.57. On the other hand, it is interesting that one can derive the chemical potential of the
vacancies directly from the model defined by Eq. 2.54 by applying Eq. 1.32 which can be done
if we consider the vacancies as a real component which can be dissolved from a surrounding
vacuum. The site fractions will then become mole fractions. Eq. 1.32 will directly yield

µV a = ◦ GV a + RT ln yV a (2.58)

This would give the same result as Eq. 2.56 if it is accepted that equilibrium with the surroundings
implies that the chemical potential of vacancies has an absolute value of zero. That is a very
convenient procedure which is often applied. The present result supports this procedure.

For the regular solution model one obtains

µV a = ◦ GV a + RT ln yV a + yA
2
LAV a (2.59)

However, due to the low vacancy content it is not possible to distinguish experimentally between
the effects of ◦ GV a and LAV a . We can thus introduce a single parameter,

yVeqa = exp − ◦ 2


GV a + yA LAV a /RT ' exp [−EAV a /RT ] (2.60)

It will thus be EAV a that is evaluated from experimental information on yVeqa and it may again
be emphasized that it is not necessary to consider its physical interpretation.

For vacancies in a binary phase we obtain by applying the regular solution model to a ternary
solution,

Gm = yA ◦ GA + yB ◦ GB + yV a ◦ GV a
X (2.61)
+ RT yi ln yi + yA yB LAB + yA yV a LAV a + yB yV a LBV a

µV a = ◦ GV a + RT ln yV a − LAB yA yB + LAV a yA (1 − yV a ) + LBV a yB (1 − yV a )

(2.62)
' yA EAV a + yB EBV a + RT ln yV a

40
 CHAPTER 2. SOLUTION MODELS

The equilibrium content of vacancies in this approximation is again obtained with µV a = 0 and
it is thus predicted to vary with compositions as

yV a = exp [− (yA EAV a + yB EBV a − yA yB LAB ) /RT ] (2.63)

In a database one prefers to describe phases with the same mathematical model and Eq. 2.61
should thus be used as an expression of when there are thermal vacancies. To make it consistent
with Eqs. 2.60 and 2.63 one could simply put ◦ GV a to zero and interpret LAV a and LBV a as
EAV a and EBV a .

The vacancies considered in the discussion of interstitial solutions in Section 2.7 are regarded
as stoichiometric vacancies. Their number depends directly on the composition, not on the
temperature. Also for such vacancies it was shown that one may assume that their chemical
potential in equilibrium is equal to zero.

Exercise 2.10.1. Suppose the fraction of vacancies in a pure solid metal is 10−3 at
the melting point. What should be the fraction at half the melting point?
Hint Of course, half the melting point means half of the absolute melting tempera-
ture.

Exercise 2.10.2. Suppose a specimen of pure solid A is equilibrated at the melting

point and then quenched to half that temperature. What is the driving force for
decreasing the vacancy content at the new temperature? Use the value of ◦ GV a
obtained from the preceding exercise.

2.11 Solutions with associates

Non-ideal behaviour of an α solution may be described by assuming that the species have a
tendency to form more complex species, associates, which are sometimes identical to molecules
that appear in another phase. All the species are treated as constituents and Pthe composition
variables will be described with the fractions of the constituents, yk = Nk / Ni . This could
be called constituent fraction but itP looks as a special case of the definition of site fraction
introduced in Section 2.1, yjαk = Njαk / Niαk . It has thus been recommended to use the same
symbol, yi , for both and to apply the term site fraction to both although there are not separate
sublattices in the present case. The mole fraction of component d is obtained as

X XX
xα
d = aid yiα / aic yiα (2.64)
i i c

The first two summations include all constituents i and aid is the stoichiometric coefficient of
component d in the constituent i. The third summation includes all components, c, and aic is
the stoichiometric coefficient of component c in the constituent i.

In a simple model one assumes that all constituents mix randomly with each other. For a simple
A-B solution with free atoms and an A1 B1 associate one obtains per mole of constituents,

41
CHAPTER 2. SOLUTION MODELS

X
Gm (y1 , y2 , · · · ) = yA ◦ GA + yB ◦ GB + yAB ◦ GAB + RT yi ln yi (2.65)

◦
GAB is the standard Gibbs energy of the AB associate, a quantity that has to be evaluated
from the measured properties of the solution. This may be regarded as an ideal solution model
for a solution containing associates. Eq. 2.64 yields

yA + yAB
xA = (2.66)
yA + yB + yAB + yAB

It may seem rather complicated to evaluate the chemical potentials of the components for a state
of equilibrium from Gm (y1 , y2 , · · · ) since this quantity does not apply to a constant number of
atoms. This molar Gibbs energy holds for 1 mole of constituents, i.e. 1 mole of lattice sites if the
phase is solid. The situation was somewhat similar in the preceding section. It was then shown
how the problem can be solved by defining an internal variable and then reformulate the model
to an expression in a set of independent variables. However, it was then shown that a more
convenient method may be based on the application of Eq. 1.32 by treating all the site fractions
as independent mole fractions. That method is often applied to solutions with associates by
considering a frozen-in state where the free atoms cannot react to form the associates and their
numbers can be controlled by exchange of associates with the surroundings. All the constituents
can then be regarded as components. This is often a realistic assumption for aqueous solutions
at room temperature where reactions between molecules are very sluggish. One may thus apply
Eq. 1.32 to Gm (y1 , y2 , · · · ) by treating the site fractions, yi , as if they were mole fractions of the
components, xi , in Eq. 1.32. One may thus express their chemical potentials from Eq. 1.32,

∂Gm X ∂Gm
µj = Gm + − yi ' ◦ Gj + RT ln yj (2.67)
∂yj ∂yi

This procedure is allowed in two limiting cases, at low temperatures where internal reactions are
very sluggish and at high temperatures where internal equilibrium is almost satisfied.

At internal equilibrium, where the situation is not frozen-in, one can relate the chemical potential
of an associate to those of its components using Eq. 1.27. In the case of A1 B1 one can obtain
µAB directly from Eq. 2.67 but also by first evaluating µA and µB from Eq. 2.67 and then using
µAB = µA + µB .

eq eq eq
µAB = ◦ GAB + RT ln yAB = ◦ GA + RT ln yA + ◦ GB + RT ln yB (2.68)

We may thus calculate the equilibrium content of AB associates from

eq
yAB ◦
eq eq = exp (−∆ GAB /RT ) = K (2.69)
yA yB

where

∆◦ GAB = ◦ GAB − ◦ GA − ◦ GB (2.70)

∆◦ GAB may be regarded as the standard Gibbs energy of formation of AB associates within the
solution phase. It is determined as a fitting parameter to satisfy experimental information.

42
 CHAPTER 2. SOLUTION MODELS

Figure 2.10: Excess Gibbs energy accounted for with the associate solution model using a series
of values of the standard Gibbs energy of formation of AB associates.

Eq. 2.69 is an example of the law of mass action and K is a temperature dependent coeffi-
cient, often called equilibrium constant. In many cases such coefficients have been determined
experimentally and their values have often been presented in tables without any further ther-
modynamic analysis. It is common to compute equilibria from such constants using the law
of mass action and it should be remembered that it is based on an ideal solution model. The
applicability is often restricted to dilute solutions.

For more concentrated solutions the law of mass action cannot be applied and the calculation
turns more complicated. On the other hand, with an advanced software for thermodynamic
calculations, which can minimize the total Gibbs energy of a system under given conditions, the
result may be obtained directly.

The present model for solutions with associates is based on random mixing and could thus
be regarded as a kind of ideal solution model if no interactions between the constituents are
introduced through an excess Gibbs energy term. However, it should be emphasized that the
model is a method of describing a non-ideal behaviour as far as the real components, A and
B, are concerned. In some fields it is common to analyse a non-ideal behaviour in terms of
hypothesized associates and thus to predict the compositions and contents of associates that are
suspected to exist but may not. Fig. 2.10 illustrates the real deviation from ideal behaviour
in an A-B system, i.e. the excess Gibbs energy, E Gm , predicted with AB associates of various
standard Gibbs energy of formation. The curves represent the difference between the associate
model and the ideal solution model, i.e.

E
Gm = Gm (associate model, Eq. 2.65) − Gm (ideal solution model, Eq. 2.2)

Naturally, the excess Gibbs energy is zero for a pure component.

43
CHAPTER 2. SOLUTION MODELS

As already mentioned, it may take time for internal equilibrium in a solution to be established.
In such cases one sometimes regards the advance of sluggish internal reactions as extra degrees
of freedom but, at the same time, one treats the associates as additional components. The
Gibbs phase rule then takes the following form, (v + rea) = (c + ass) + 2 − p. Evidently, this
is correct only if the number of internal reactions that are considered is equal to the number of
associates that are considered. It is advisable to consider only the internal reactions that control
the amounts of the associates by formation from the real components and not reactions involving
more than one associate. Such reactions could be included if others are excluded but the total
number of internal reactions to be considered must be equal to the number of associates. One
talks about a set of “independent” reactions.

Before internal equilibrium has been established one could evaluate the chemical potential of
an associate from Eq. 2.67 under frozen-in conditions and the driving force for the formation of
more of it in a dilute solution is given by

D = µA + µB − µAB
= ◦ GA + RT ln yA + ◦ GB + RT ln yB − ◦ GAB − RT ln yAB (2.71)
◦
= −∆ GAB + RT ln (yA yB /yAB )

This can be reformulated using Eq. 2.69,

eq
yAB yA yB
D = RT ln eq eq (2.72)
yAB yA yB

It should be realized that this is actually the driving force for the reaction A + B → A1 B1 .
For thermal vacancies Eq. 2.56 concerned the creation of vacancies with no other constituents
involved. The relations derived for AB associates in this section can easily be generalized to
more complicated associates and to solutions with several kinds of associates.

Exercise 2.11.1. Analytical expressions for equilibrium constants are usually given
through Eq. 2.69 with ∆◦ GAB = α + βT but more exact descriptions could easily
be used in computerized databases. Use such a database to check the temperature
dependence of ∆◦f GH2 O . Make the examination from 0 to 1000◦ C.
Hint Plot ∆◦ GH2 O as function of T using the standard states of H2 and O2 as
references.

44