CH 1

1 Thermodynamics and Phase Diagrams of Materials
Arthur D. Pelton
Centre de Recherche en Calcul Thermochimique, École Polytechnique,

Montréal, Québec, Canada
List of Symbols and Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.1.1 Notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2 Gibbs Energy and Equilibrium . . . . . . . . . . . . . . . . . . . . . . . 5
1.2.1 Gibbs Energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2.2 Chemical Equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3 Predominance Diagrams . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.3.1 Calculation of Predominance Diagrams . . . . . . . . . . . . . . . . . . . 7
1.3.2 Ellingham Diagrams as Predominance Diagrams . . . . . . . . . . . . . . 8
1.3.3 Discussion of Predominance Diagrams . . . . . . . . . . . . . . . . . . . 9
1.4 Thermodynamics of Solutions . . . . . . . . . . . . . . . . . . . . . . . 9
1.4.1 Gibbs Energy of Mixing . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.4.2 Chemical Potential . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.4.3 Tangent Construction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.4.4 Gibbs-Duhem Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.4.5 Relative Partial Properties . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.4.6 Activity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.4.7 Ideal Raoultian Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.4.8 Excess Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.4.9 Activity Coefficient . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.4.10 Multicomponent Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.5 Binary Phase Diagrams . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.5.1 Systems with Complete Solid and Liquid Miscibility . . . . . . . . . . . . 14
1.5.2 Thermodynamic Origin of Phase Diagrams . . . . . . . . . . . . . . . . . 16
1.5.3 Pressure-Composition Phase Diagrams . . . . . . . . . . . . . . . . . . . 19
1.5.4 Minima and Maxima in Two-Phase Regions . . . . . . . . . . . . . . . . . 20
1.5.5 Miscibility Gaps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.5.6 Simple Eutectic Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.5.7 Regular Solution Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
1.5.8 Thermodynamic Origin of Simple Phase Diagrams Illustrated by Regular
Solution Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
1.5.9 Immiscibility – Monotectics . . . . . . . . . . . . . . . . . . . . . . . . . 26
1.5.10 Intermediate Phases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
1.5.11 Limited Mutual Solubility – Ideal Henrian Solutions . . . . . . . . . . . . 29
1.5.12 Geometry of Binary Phase Diagrams . . . . . . . . . . . . . . . . . . . . . 31
Phase Transformations in Materials. Edited by Gernot Kostorz

Copyright © 2001 WILEY-VCH Verlag GmbH, Weinheim
ISBN: 3-527-30256-5
2 1 Thermodynamics and Phase Diagrams of Materials
1.6 Application of Thermodynamics to Phase Diagram Analysis . . . . . . 34

1.6.1 Thermodynamic/Phase Diagram Optimization . . . . . . . . . . . . . . . . 34
1.6.2 Polynomial Representation of Excess Properties . . . . . . . . . . . . . . . 34
1.6.3 Least-Squares Optimization . . . . . . . . . . . . . . . . . . . . . . . . . 35
1.6.4 Calculation of Metastable Phase Boundaries . . . . . . . . . . . . . . . . . 39
1.7 Ternary and Multicomponent Phase Diagrams . . . . . . . . . . . . . . 39
1.7.1 The Ternary Composition Triangle . . . . . . . . . . . . . . . . . . . . . . 39
1.7.2 Ternary Space Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
1.7.3 Polythermal Projections of Liquidus Surfaces . . . . . . . . . . . . . . . . 41
1.7.4 Ternary Isothermal Sections . . . . . . . . . . . . . . . . . . . . . . . . . 43
1.7.4.1 Topology of Ternary Isothermal Sections . . . . . . . . . . . . . . . . . . 45
1.7.5 Ternary Isopleths (Constant Composition Sections) . . . . . . . . . . . . . 46
1.7.5.1 Quasi-Binary Phase Diagrams . . . . . . . . . . . . . . . . . . . . . . . . 47
1.7.6 Multicomponent Phase Diagrams . . . . . . . . . . . . . . . . . . . . . . 47
1.7.7 Nomenclature for Invariant Reactions . . . . . . . . . . . . . . . . . . . . 49
1.7.8 Reciprocal Ternary Phase Diagrams . . . . . . . . . . . . . . . . . . . . . 49
1.8 Phase Diagrams with Potentials as Axes . . . . . . . . . . . . . . . . . . 51
1.9 General Phase Diagram Geometry . . . . . . . . . . . . . . . . . . . . . 56
1.9.1 General Geometrical Rules for All True Phase Diagram Sections . . . . . . 56
1.9.1.1 Zero Phase Fraction Lines . . . . . . . . . . . . . . . . . . . . . . . . . . 58
1.9.2 Choice of Axes and Constants of True Phase Diagrams . . . . . . . . . . . 58
1.9.2.1 Tie-lines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
1.9.2.2 Corresponding Phase Diagrams . . . . . . . . . . . . . . . . . . . . . . . 60
1.9.2.3 Theoretical Considerations . . . . . . . . . . . . . . . . . . . . . . . . . . 60
1.9.2.4 Other Sets of Conjugate Pairs . . . . . . . . . . . . . . . . . . . . . . . . 61
1.10 Solution Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
1.10.1 Sublattice Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
1.10.1.1 All Sublattices Except One Occupied by Only One Species . . . . . . . . . 62
1.10.1.2 Ionic Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
1.10.1.3 Interstitial Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
1.10.1.4 Ceramic Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
1.10.1.5 The Compound Energy Formalism . . . . . . . . . . . . . . . . . . . . . . 65
1.10.1.6 Non-Stoichiometric Compounds . . . . . . . . . . . . . . . . . . . . . . . 65
1.10.2 Polymer Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
1.10.3 Calculation of Limiting Slopes of Phase Boundaries . . . . . . . . . . . . 66
1.10.4 Short-Range Ordering . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
1.10.5 Long-Range Ordering . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
1.11 Calculation of Ternary Phase Diagrams From Binary Data . . . . . . . 72
1.12 Minimization of Gibbs Energy . . . . . . . . . . . . . . . . . . . . . . . 74
1.12.1 Phase Diagram Calculation . . . . . . . . . . . . . . . . . . . . . . . . . . 76
1.13 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
1.13.1 Phase Diagram Compilations . . . . . . . . . . . . . . . . . . . . . . . . . 76
1.13.2 Thermodynamic Compilations . . . . . . . . . . . . . . . . . . . . . . . . 77
1.13.3 General Reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
1.14 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
List of Symbols and Abbreviations 3
List of Symbols and Abbreviations
Symbol Designation
ai activity of component i
C number of components
cp molar heat capacity
E electrical potential of a galvanic cell
F degrees of freedom/variance
G Gibbs energy in J
g molar Gibbs energy in J/mol
gi partial molar Gibbs energy of i
Gi0 standard Gibbs energy of i
gi0 standard molar Gibbs energy of i
Dgi relative partial Gibbs energy i
gE excess molar Gibbs energy
giE excess partial Gibbs energy of i
DG Gibbs energy change
DG 0 standard Gibbs energy change
Dgm molar Gibbs energy of mixing
Dgf0 standard molar Gibbs energy of fusion
Dgv0 standard molar Gibbs energy of vaporization
H enthalpy in J
h molar enthalpy in J/mol
hi partial enthalpy of i
Hi0 standard enthalpy of i
hi0 standard molar enthalpy of i
Dhi relative partial enthalpy of i
hE excess molar enthalpy
hEi excess partial enthalpy of i
DH enthalpy change
DH 0 standard enthalpy change
Dhm molar enthalpy of mixing
Dh0f standard molar enthalpy of fusion
Dhv0 standard molar enthalpy of vaporization
K equilibrium constant
kB Boltzmann constant
n number of moles
ni number of moles of constituent i
Ni number of particles of i
N0 Avogadro’s number
pi partial pressure of i
P total pressure
P number of phases
qi general extensive variable
R gas constant
S entropy in J/K
s molar entropy in J/mol K
si partial entropy of i
Si0 standard entropy of i
si0 standard molar entropy of i
Dsi0 relative partial entropy of i
sE excess molar entropy
siE excess partial entropy of i
DS entropy change
DS 0 standard entropy change
Dsm molar entropy of mixing
Ds0f standard molar entropy of fusion
Dsv0 standard molar entropy of vaporization
T temperature
Tf temperature of fusion
Tc critical temperature
TE eutectic temperature
U internal energy
vi molar volume of i
vi0 standard molar volume of i
Xi mole fraction of i
Z coordination number
gi activity coefficient of i
e bond energy
h empirical entropy parameter
mi chemical potential of i
n number of moles of “foreign” particles contributed by a mole of solute
x molar metal ratio
s vibrational bond entropy
fi generalized thermodynamic potential
w empirical enthalpy parameter
b.c.c. body-centered cubic

f.c.c. face-centered cubic
h.c.p. hexagonal close-packed
1.2 Gibbs Energy and Equilibrium 5
1.1 Introduction ergy. General computer programs are avail-

able for the minimization of the Gibbs en-
An understanding of thermodynamics ergy in systems of any number of phases,
and phase diagrams is fundamental and es- components and species as outlined in Sec.
sential to the study of materials science. A 1.12. When coupled to extensive databases
knowledge of the equilibrium state under a of the thermodynamic properties of com-
given set of conditions is the starting point pounds and multicomponent solutions,
in the description of any phenomenon or these provide a powerful tool in the study
process. of materials science.
The theme of this chapter is the relation-
ship between phase diagrams and thermo- 1.1.1 Notation
dynamics. A phase diagram is a graphical
Extensive thermodynamic properties are
representation of the values of thermody-
represented by upper case symbols. For ex-
namic variables when equilibrium is estab-
ample, G = Gibbs energy in J. Molar prop-
lished among the phases of a system. Mate-
erties are represented by lower case sym-
rials scientists are used to thinking of phase
bols. For example, g = G/n = molar Gibbs
diagrams as plots of temperature versus com-
energy in J/mol where n is the total number
position. However, many other variables
of moles in the system.
such as total pressure and partial pressures
may be plotted on phase diagrams. In Sec.
1.3, for example, predominance diagrams 1.2 Gibbs Energy and Equilibrium
will be discussed, and in Sec. 1.8 chemical
potential – composition phase diagrams will 1.2.1 Gibbs Energy
be presented. General rules regarding phase
diagram geometry are given in Sec. 1.9. The Gibbs energy of a system is defined
In recent years, a quantitative coupling in terms of its enthalpy, H, entropy, S, and
of thermodynamics and phase diagrams temperature, T:
has become possible. With the use of com- G = H –TS (1-1)
puters, simultaneous optimizations of ther-
modynamic and phase equilibrium data can A system at constant temperature and pres-
be applied to the critical evaluation of bi- sure will approach an equilibrium state that
nary and ternary systems as shown in Sec. minimizes G.
1.6. This approach often enables good esti- As an example, consider the question of
mations to be made of the thermodynamic whether silica fibers in an aluminum ma-
properties and phase diagrams of multi- trix at 500 °C will react to form mullite,
component systems as discussed in Sec. Al6Si2O13
1.11. These estimates are based on structu- 13 9
ral models of solutions. Various models SiO 2 + 6 Al = Si + Al 6 Si 2 O13 (1-2)
2 2
such as the regular solution model, the sub-
lattice model, and models for interstitial If the reaction proceeds with the formation
solutions, polymeric solutions, solutions of of dn moles of mullite then, from the stoi-
defects, ordered solutions, etc. are dis- chiometry of the reaction, dnSi = (9/2) dn,
cussed in Secs. 1.5 and 1.10. dnAl = – 6 dn, and dnSiO2 = – 13/2 dn. Since
The equilibrium diagram is always cal- the four substances are essentially immis-
culated by minimization of the Gibbs en- cible at 500 °C, we need consider only the
standard molar Gibbs energies, Gi0. The action continues with the production of
Gibbs energy of the system then varies as: H2S, pH2S will increase while pH2 and pS2
will decrease. As a result, DG will become
9 0 13 0
0
dG / dn = g Al 6 Si 2 O13
+g − g progressively less negative. Eventually an
2 Si 2 SiO2 equilibrium state will be reached when
− 6 g Al
0
= DG 0 = − 830 kJ (1-3) dG/dn = DG = 0.
where DG 0 is the standard Gibbs energy For the equilibrium state, therefore:
change of reaction, Eq. (1-2), at 500 °C. DG 0 = – R T ln K (1-7)
Since DG 0 < 0, the formation of mullite 2 – 2 –1
= – R T ln ( pH2S pH2 pS2 )equilibrium
entails a decrease in G. Hence, the reaction
will proceed spontaneously so as to mini- where K, the “equilibrium constant” of the
mize G. reaction, is the one unique value of the ra-
tio ( p2H2S pH– 22 pS–12 ) for which the system will
1.2.2 Chemical Equilibrium be in equilibrium at the temperature T.
If the initial partial pressures are such
The partial molar Gibbs energy of an that DG > 0, then the reaction, Eq. (1-5),
ideal gas is given by: will proceed to the left in order to minimize
gi = gi0 + R T ln pi (1-4) G until the equilibrium condition of Eq.
(1-7) is attained.
where gi0 is the standard molar Gibbs en- As a further example, we consider the
ergy (at 1 bar), pi is the partial pressure in possible precipitation of graphite from a
bar, and R is the gas constant. The second gaseous mixture of CO and CO2 . The reac-
term in Eq. (1-4) is entropic. As a gas ex- tion is:
pands at constant T, its entropy increases.
Consider a gaseous mixture of H2 , S2 2 CO = C + CO2 (1-8)
and H2S with partial pressures pH2 , pS2 and Proceeding as above, we can write:
pH2S . The gases can react according to
dG/dn = gC + gCO2 – 2 gCO
2 H 2 + S 2 = 2 H 2S (1-5)
= ( gC0 + gCO
0
2
0
– 2 gCO 2
) + R T ln ( pCO2 /pCO )
If the reaction, Eq. (1-5), proceeds to the 0
= DG + R T ln ( pCO2 /pCO ) 2
(1-9)
right with the formation of 2 dn moles of = DG = – R T ln K + R T ln ( pCO2 /pCO 2
)
H2S, then the Gibbs energy of the system
2
varies as: If ( pCO2 /pCO ) is less than the equilibrium
constant K, then precipitation of graphite
dG/dn = 2 gH2S – 2 gH2 – gS2 will occur in order to decrease G.
= (2 gH0 2S – 2 gH0 2 – gS02 ) Real situations are, of course, generally
+ R T (2 ln pH2S – 2 ln pH2 – ln pS2 ) more complex. To treat the deposition of
= DG 0 + R T ln ( p2H2S pH– 22 pS–12 ) solid Si from a vapour of SiI4 , for example,
we must consider the formation of gaseous
= DG (1-6)
I2 , I and SiI2 so that three reaction equa-
DG, which is the Gibbs energy change of tions must be written:
the reaction, Eq. (1-5), is thus a function of SiI4 (g) = Si (sol) + 2 I2 (g) (1-10)
the partial pressures. If DG < 0, then the re-
action will proceed to the right so as to SiI4 (g) = SiI2 (g) + I2 (g) (1-11)
minimize G. In a closed system, as the re- I2 (g) = 2 I (g) (1-12)
1.3 Predominance Diagrams 7
The equilibrium state, however, is still that solid compounds of Cu, S and O. For ex-
which minimizes the total Gibbs energy of ample, at point Z, where pSO2 = 10– 2 and
the system. This is equivalent to satisfying pO2 = 10–7 bar, the stable phase is Cu2O.
simultaneously the equilibrium constants The conditions for coexistence of two and
of the reactions, Eqs. (1-10) to (1-12), as three solid phases are indicated respectively
will be shown in Section 1.12 where this by the lines and triple points on the diagram.
example is discussed further. For example, along the univariant line
(phase boundary) separating the Cu2O and
CuSO4 domains the equilibrium constant
1.3 Predominance Diagrams –2
K = pSO 2
pO– 23/2 of the following reaction is
satisfied:
1.3.1 Calculation of Predominance
Diagrams Cu2O + 2 SO2 + –23 O2 = 2 CuSO4 (1-13)
Predominance diagrams are a particu- Hence, along this line:

(1-14)
larly simple type of phase diagram which 3
log K = – 2 log pSO2 – – log pO2 = constant
2
have many applications in the fields of hot
corrosion, chemical vapor deposition, etc. This boundary is thus a straight line with a
Furthermore, their construction clearly il- slope of (– 3/2)/2 = – 3/4.
lustrates the principles of Gibbs energy In constructing predominance diagrams,
minimization and the Gibbs Phase Rule. we define a “base element”, in this case Cu,
A predominance diagram for the Cu – which must be present in all the condensed
S – O system at 1000 K is shown in Fig. phases. Let us further assume that there is
1-1. The axes are the logarithms of the no mutual solubility among the condensed
partial pressures of SO2 and O2 in the phases.
gas phase. The diagram is divided into Following the procedure of Bale et al.
areas or domains of stability of the various (1986), we formulate a reaction for the for-
Figure 1-1. Predomi-

nance diagram. log pSO2
versus log pO2 (bar) at
1000 K for the Cu – S – O
system (Bale et al., 1986).
mation of each solid phase, always from Fig. 1-3. For the formation reaction:
one mole of the base element Cu, and in-
4 Cu + O2 = 2 Cu2O (1-19)
volving the gaseous species whose pres-
sures are used as the axes (SO2 and O2 in we can write:
this example):
DG 0 = – R T ln K = R T ln ( pO2 )equilibrium
1
Cu + – O2 = CuO
2 = DH 0 – T DS 0 (1-20)
DG = DG 0 + R T ln pO–1/2 (1-15)
2 The diagonal line in Fig. 1-3 is thus a plot
1 1 of the standard Gibbs energy of formation
Cu + – O2 = – Cu2O
4 2
DG = DG 0 + R T ln pO–1/4 (1-16) of Cu2O versus T. The temperatures indi-
cated by the symbol M and 씲
2
M are the melt-
Cu + SO2 = CuS + O2 ing points of Cu and Cu2O respectively.
DG = DG 0 + R T ln ( pO2 pSO
–1
2
) (1-17) This line is thus simply a line taken from
the well-known Ellingham Diagram or
Cu + SO2 + O2 = CuSO4
DG 0 vs. T diagram for the formation of
DG = DG 0 + R T ln ( pSO
–1
2
pO–12 ) (1-18) oxides. However, by drawing vertical lines
and similarly for the formation of Cu2S, at the melting points of Cu and Cu2O as
Cu2SO4 and Cu2SO5 . shown in Fig. 1-3, we convert the plot to a
The values of DG 0 are obtained from ta- true phase diagram. Stability domains for
bles of thermodynamic properties. For any Cu (sol), Cu (l), Cu2O (sol), and Cu2O (l)
given values of pSO2 and pO2 , DG for each are shown as functions of T and of imposed
formation reaction can then be calculated. pO2 . The lines and triple points indicate
The stable compound is simply the one
with the most negative DG. If all the DG
values are positive, then pure Cu is the
stable compound.
By reformulating Eqs. (1-15) to (1-18) in
terms of, for example, S2 and O2 rather
than SO2 and O2 , a predominance diagram
with ln pS2 and ln pO2 as axes can be con-
structed. Logarithms of ratios or products
of partial pressures can also be used as
axes.
1.3.2 Ellingham Diagrams

as Predominance Diagrams
Rather than keeping the temperature
constant, we can use it as an axis. Figure
1-2 shows a diagram for the Fe – S – O sys-
tem in which R T ln pO2 is plotted versus T
at constant pSO2 = 1 bar. The diagram is of
the same topological type as Fig. 1-1.
A similar phase diagram of R T ln pO2 Figure 1-2. Predominance diagram. RT ln pO2 ver-
versus T for the Cu – O system is shown in sus T at pSO2 = 1.0 bar for the Fe – S – O system.
1.4 Thermodynamics of Solutions 9
Figure 1-3. Predominance diagram

(also known as a Gibbs energy-tem-
perature diagram or Ellingham dia-
gram) for the Cu – O system. Points
M and 앮M represent the melting
points of the metal and oxide re-
spectively.
conditions of two- and three-phase equilib- Predominance diagrams have found

rium. many applications in the fields of hot cor-
rosion, roasting of ores, chemical vapor
1.3.3 Discussion of Predominance deposition, etc. A partial bibliography on
Diagrams their construction and applications includes
In this section discussion is limited to Yokokowa (1999), Bale (1990), Bale et al.
the assumption that there is no mutual sol- (1986), Kellogg and Basu (1960), Ingra-
ubility among the condensed phases. The ham and Kellogg (1963), Pehlke (1973),
calculation of predominance phase dia- Garrels and Christ (1965), Ingraham and
grams in which mutual solubility is taken Kerby (1967), Pilgrim and Ingraham
into account is treated in Sec. 1.9, where (1967), Gulbransen and Jansson (1970),
the general geometrical rules governing Pelton and Thompson (1975), Shatynski
predominance diagrams and their relation- (1977), Stringer and Whittle (1975), Spen-
ship to other types of phase diagrams are cer and Barin (1979), Chu and Rahmel
discussed. (1979), and Harshe and Venkatachalam
We frequently encounter predominance (1984).
diagrams with domains for solid, liquid,
and even gaseous compounds which have
been calculated as if the compounds were
1.4 Thermodynamics of Solutions
immiscible, even though they may actually
1.4.1 Gibbs Energy of Mixing
be partially or even totally miscible. The
boundary lines are then no longer phase Liquid gold and copper are completely
boundaries, but are lines separating regions miscible at all compositions. The Gibbs en-
in which one species “predominates”. The ergy of one mole of liquid solution, g l, at
well known E – pH or Pourbaix diagrams of 1400 K is drawn in Fig. 1-4 as a function of
aqueous chemistry are examples of such composition expressed as mole fraction,
predominance diagrams. XCu , of copper. Note that XAu = 1 – XCu . The
Predominance diagrams may also be con- curve of g l varies between the standard mo-
structed when there are two or more base lar Gibbs energies of pure liquid Au and
0 0
elements, as discussed by Bale (1990). Cu, gAu and gCu .
nent i, and the derivative is taken with all

nj ( j9i) constant.
In the example of the Au – Cu binary liq-
uid solution, g Cu = (∂G l /∂nCu )T, P, nAu , where
G l = (nCu + nAu ) g l. That is, g Cu , which has
units of J/mol, is the rate of change of the
Gibbs energy of a solution as Cu is added.
It can be seen that g Cu is an intensive prop-
erty of the solution which depends upon
the composition and temperature but not
upon the total amount of solution. That is,
adding dnCu moles of copper to a solution
of given composition will (in the limit as
dnCu Æ 0) result in a change in Gibbs en-
ergy, dG, which is independent of the total
Figure 1-4. Molar Gibbs energy, g l, of liquid Au – mass of the solution.
Cu alloys at constant temperature (1400 K) illustrat- The reason that this property is called a
ing the tangent construction. chemical potential is illustrated by the fol-
lowing thought experiment. Imagine two
systems, I and II, at the same temperature
l
The function Dgm shown on Fig. 1-4 is and separated by a membrane that permits
called the molar Gibbs energy of mixing of only the passage of copper. The chemical
the liquid solution. It is defined as: potentials of copper in systems I and II
l
Dgm = g l – (XAu gAu
0 0
+ XCu gCu ) (1-21) are gCu I
= ∂ G I /∂nCuI
and gCu II
= ∂ G II /∂nCu
II
.
Copper is transferred across the membrane,
l
It can be seen that Dgm is the Gibbs energy with dn I = – dn II. The change in the total
change associated with the isothermal mix- Gibbs energy accompanying this transfer
ing of XAu moles of pure liquid Au and XCu is then: (1-24)
moles of pure liquid Cu to form one mole
of solution: dG = d (G I + G II ) = – (gCu
I II
– gCu II
) dnCu
I II
XAu Au (l) + XCu Cu (l) If gCu > gCu , then d (G I + G II ) is negative
II
when dnCu is positive. That is, the total
= 1 mole liquid solution (1-22)
Gibbs energy will be decreased by a trans-
Note that for the solution to be stable it is fer of Cu from system I to system II.
l
necessary that Dgm be negative. Hence, Cu will be transferred spontane-
ously from a system of higher g Cu to a sys-
1.4.2 Chemical Potential tem of lower g Cu . Therefore g Cu is called
the chemical potential of copper.
The partial molar Gibbs energy of com- An important principle of phase equilib-
ponent i, gi , also known as the chemical rium can now be stated. When two or more
potential, m i , is defined as: phases are in equilibrium, the chemical po-
gi = m i = (∂G/∂ni )T, P, n j (1-23) tential of any component is the same in all
phases.
where G is the Gibbs energy of the solu-
tion, ni is the number of moles of compo-
1.4.3 Tangent Construction 1.4.4 Gibbs–Duhem Equation

An important construction is illustrated Differentiation of Eq. (1-27) yields:
in Fig. 1-4. If a tangent is drawn to the
dg l = (XAu dgAu + XCu dgCu )
curve of g l at a certain composition
(XCu = 0.6 in Fig. 1-4), then the intercepts + (gAu dXAu + gCu dXCu) (1-30)
of this tangent on the axes at XAu = 1 and
Comparison with Eq. (1-28) then gives the
XCu = 1 are equal to g Au and g Cu respec-
Gibbs–Duhem equation at constant T and P:
tively at this composition.
To prove this, we first consider that the XAu dgAu + XCu dgCu = 0 (1-31)
Gibbs energy of the solution at constant T
and P is a function of nAu and nCu . Hence:
1.4.5 Relative Partial Properties
⎛ ∂G ⎞ l
⎛ ∂G ⎞ l
dGTl , P = ⎜ ⎟ dnAu + ⎜ ⎟ dnCu The difference between the partial Gibbs
⎝ ∂nAu ⎠ ⎝ ∂nCu ⎠ energy gi of a component in solution and
= g Au dnAu + gCu dnCu (1-25) the partial Gibbs energy gi0 of the same
component in a standard state is called the
Eq. (1-25) can be integrated as follows:
relative partial Gibbs energy (or relative
Gl nAu nCu chemical potential ), Dgi . It is most usual to
∫ dG l = ∫ g Au dnAu + ∫ gCu dnCu choose as standard state the pure compo-
0 0 0 nent in the same phase at the same temper-
G = g Au nAu + gCu nCu
l
(1-26 ) ature. The activity ai of the component rel-
ative to the chosen standard state is then
where the integration is performed at con- defined in terms of Dgi by the following
stant composition so that the intensive equation, as illustrated in Fig. 1-4.
properties g Au and g Cu are constant. This
integration can be thought of as describing Dgi = gi – gi0 = m i – m 0i = R T ln ai (1-32)
a process in which a pre-mixed solution of
constant composition is added to the sys- Note that gi and m i are equivalent symbols,
tem, which initially contains no material. as are gi0 and m 0i , see Eq. (1-23).
Dividing Eqs. (1-26) and (1-25) by From Fig. 1-4, it can be seen that:
(nAu + nCu ) we obtain expressions for the Dgm = XAu DgAu + XCu DgCu
molar Gibbs energy and its derivative:
= R T (XAu ln aAu + XCu ln aCu ) (1-33)
l
g = XAu gAu + XCu gCu (1-27)
The Gibbs energy of mixing can be di-
and vided into enthalpy and entropy terms, as
can the relative partial Gibbs energies:
dg l = gAu dXAu + gCu dXCu (1-28)
Since dXAu = – dXCu , it can be seen that Dgm = Dhm – T D sm (1-34)
Eqs. (1-27) and (1-28) are equivalent to the Dgi = Dhi – T D si (1-35)
tangent construction shown in Fig. 1-4.
These equations may also be rearranged Hence, the enthalpy and entropy of mixing
to give the following useful expression for may be expressed as:
a binary system: Dhm = XAu DhAu + XCu DhCu (1-36)
gi = g + (1 – Xi ) dg/dXi (1-29) D sm = XAu D sAu + XCu DsCu (1-37)
and tangent constructions similar to that of For the reaction, Eq. (1-38), at 1400 K,
Fig. 1-4 can be used to relate the relative DG 0 = – 68.35 kJ (Barin et al., 1977). The
partial enthalpies and entropies Dhi and activity of Cu in the liquid alloy at
Dsi to the integral molar enthalpy of mix- XCu = 0.6 is aCu = 0.43 (Hultgren et al.,
ing Dhm and integral molar entropy of mix- 1973). Substitution into Eq. (1-39) with
ing Dsm respectively. pO2 = 10– 4 bar gives:
dG/dn = DG = – 50.84 kJ
1.4.6 Activity Hence under these conditions the reaction
entails a decrease in the total Gibbs energy
The activity of a component in a solution
and so the copper will be oxidized.
was defined by Eq. (1-32).
Since ai varies monotonically with gi it
1.4.7 Ideal Raoultian Solutions
follows that when two or more phases are
in equilibrium the activity of any compo- An ideal solution or Raoultian solution
nent is the same in all phases, provided that is usually defined as one in which the ac-
the activity in every phase is expressed tivity of a component is equal to its mole
with respect to the same standard state. fraction:
The use of activities in calculations of
aiideal = Xi (1-40)
chemical equilibrium conditions is illus-
trated by the following example. A liquid (With a judicious choice of standard state,
solution of Au and Cu at 1400 K with this definition can also encompass ideal
XCu = 0.6 is exposed to an atmosphere in Henrian solutions, as discussed in Sec.
which the oxygen partial pressure is 1.5.11.)
pO2 = 10– 4 bar. Will Cu2O be formed? The However, this Raoultian definition of
reaction is: ideality is generally only useful for simple
substitutional solutions. There are more
2 Cu (l) + –12 O2 (g) = Cu2O (sol) (1-38) useful definitions for other types of solu-
where the Cu (l) is in solution. If the reactions such as interstitial solutions, ionic so-
tion proceeds with the formation of dn lutions, solutions of defects, polymer solu-
moles of Cu2O, then 2 dn moles of Cu are tions, etc. That is, the most convenient def-
consumed, and the Gibbs energy of the inition of ideality depends upon the solu-
Au – Cu solution changes by tion model. This subject will be discussed
in Sec. 1.10. In the present section, Eq.
– 2 (dG l /dnCu ) dn (1-40) for an ideal substitutional solution
will be developed with the Au – Cu solution
The total Gibbs energy then varies as: as example.
dG/dn = gCu2O – –12 gO2 – 2 (dG l /dnCu ) In the ideal substitutional solution model
it is assumed that Au and Cu atoms are
= gCu2O – –12 gO2 – 2 gCu nearly alike, with nearly identical radii and
0
= (gCu – –12 gO0 2 – 2 gCu
0
) electronic structures. This being the case,
2O
there will be no change in bonding energy
– –12 R T ln pO2 – 2 R T ln aCu or volume upon mixing, so that the en-
= DG 0 + R T ln ( pO–1/2 –2
aCu ) thalpy of mixing is zero:
2
ideal
= DG (1-39) Dhm =0 (1-41)
Furthermore, and for the same reason, the Hence Eq. (1-40) has been demonstrated
Au and Cu atoms will be randomly distrib- for an ideal substitutional solution.
uted over the lattice sites. (In the case of a
liquid solution we can think of the “lattice
1.4.8 Excess Properties
sites” as the instantaneous atomic positions.)
For a random distribution of NAu In reality, Au and Cu atoms are not iden-
gold atoms and NCu copper atoms over tical, and so Au – Cu solutions are not per-
(NAu + NCu ) sites, Boltzmann’s equation fectly ideal. The difference between a solu-
can be used to calculate the configurational tion property and its value in an ideal solu-
entropy of the solution. This is the entropy tion is called an excess property. The ex-
associated with the spatial distribution of cess Gibbs energy, for example, is defined
the particles: as:
(1-42)
S config = kB ln (NAu + NCu ) ! / NAu ! NCu ! gE = Dgm – Dgmideal (1-47)
where kB is Boltzmann’s constant. The con- Since the ideal enthalpy of mixing is zero,
figurational entropies of pure Au and Cu the excess enthalpy is equal to the enthalpy
are zero. Hence the configurational entropy of mixing:
of mixing, DS config, will be equal to S config.
hE = Dhm – Dhm
ideal
= Dhm (1-48)
Furthermore, because of the assumed close
similarity of Au and Cu, there will be no Hence:
non-configurational contribution to the en-
tropy of mixing. Hence, the entropy of gE = h E – T s E
mixing will be equal to S config. = Dhm – T s E (1-49)
Applying Stirling’s approximation, which Excess partial properties are defined simi-
states that ln N ! = [(N ln N) – N ] if N is larly:
large, yields:
giE = Dgi – Dgiideal
DSmideal = S config = − kB ( NAu + NCu ) (1- 43) = R T ln ai – R T ln Xi (1-50)
⎛ NAu NCu ⎞ siE = Dsi – Dsiideal = Dsi + R ln Xi (1-51)
× ⎜ NAu ln + NCu ln ⎟
⎝ NAu + NCu NAu + NCu ⎠
Also:
For one mole of solution, (NAu + NCu ) = N 0,
giE = hiE – T siE
where N 0 = Avogadro’s number. We also
note that (kB N 0 ) is equal to the ideal gas = Dhi – T siE (1-52)
constant R. Hence: Equations analogous to Eqs. (1-33),
(1-44)
DSmideal = – R (XAu ln XAu + XCu ln XCu ) (1-36) and (1-37) relate the integral and
partial excess properties. For example, in
Therefore, since the ideal enthalpy of mix- Au – Cu solutions:
ing is zero:
(1-45) gE = XAu g Au
E E
+ XCu g Cu (1-53)
Dgmideal = R T (XAu ln XAu + XCu ln XCu )
s E = XAu s Au
E E
+ XCu s Cu (1-54)
By comparing Eqs. (1-33) and (1-45) we
Tangent constructions similar to that of
obtain:
Fig. 1-4 can thus also be employed for ex-
Dgiideal = R T ln aiideal = R T ln Xi (1-46) cess properties, and an equation analogous
to Eq. (1-29) can be written: Eq. (1-50), Dgi < Dgiideal. That is, the com-
ponent i is more stable in the solution than
giE = gE + (1 – Xi ) dgE /dXi (1-55) it would be in an ideal solution of the same
The Gibbs – Duhem equation, Eq. (1-31), composition. If g i > 1, then giE > 0 and the
also applies to excess properties: driving force for the component to enter
into solution is less than in the case of an
E E
XAu dg Au + XCu dg Cu =0 (1-56) ideal solution.
In Au – Cu alloys, gE is negative. That is, 1.4.10 Multicomponent Solutions
Dgm is more negative than Dgmideal and so
the solution is thermodynamically more The equations of this section were de-
stable than an ideal solution. We say that rived with a binary solution as an example.
Au – Cu solutions exhibit negative devia- However, the equations apply equally to
tions from ideality. If gE > 0, then the solu- systems of any number of components. For
tion is less stable than an ideal solution and instance, in a solution of components
is said to exhibit positive deviations. A– B – C – D …, Eq. (1-33) becomes:
Dgm = XA DgA + XB DgB + XC DgC
1.4.9 Activity Coefficient + XD DgD + … (1-59)
The activity coefficient of a component
in a solution is defined as: 1.5 Binary Phase Diagrams
g i = ai /Xi (1-57)
1.5.1 Systems with Complete Solid
From Eq. (1-50): and Liquid Miscibility
giE = R T ln g i (1-58) The temperature – composition (T – X )

phase diagram of the CaO – MnO system is
In an ideal solution g i = 1 and giE = 0 for shown in Fig. 1-5 (Schenck et al., 1964;
all components. If g i < 1, then giE < 0 and by Wu, 1990). The abscissa is the composi-
Figure 1-5. Phase dia-

gram of the CaO – MnO
system at P = 1 bar
(after Schenck et al.,
1964, and Wu, 1990).
1.5 Binary Phase Diagrams 15
tion, expressed as mole fraction of MnO, (XMnO = 0.28). As the temperature is de-
XMnO . Note that XMnO = 1 – XCaO . Phase di- creased further, solid continues to precipi-
agrams are also often drawn with the com- tate with the compositions of the two
position axis expressed as weight percent. phases at any temperature being given by
At all compositions and temperatures in the liquidus and solidus compositions at
the area above the line labelled liquidus, a that temperature and with their relative
single-phase liquid solution will be ob- proportions being given by the lever rule.
served, while at all compositions and tem- Solidification is complete at 2030 °C, the
peratures below the line labelled solidus, last liquid to solidify having composition
there will be a single-phase solid solution. XMnO = 0.60 (point C).
A sample at equilibrium at a temperature The process just described is known as
and overall composition between these two equilibrium cooling. At any temperature
curves will consist of a mixture of solid during equilibrium cooling the solid phase
and liquid phases, the compositions of has a uniform (homogeneous) composition.
which are given by the liquidus and solidus In the preceding example, the composition
compositions at that temperature. For ex- of the solid phase during cooling varies
ample, a sample of overall composition along the line APC. Hence, in order for the
XMnO = 0.60 at T = 2200 °C (at point R in solid grains to have a uniform composition
Fig. 1-5) will consist, at equilibrium, of a at any temperature, diffusion of CaO from
mixture of liquid of composition XMnO = the center to the surface of the growing
0.70 (point Q) and solid of composition grains must occur. Since solid-state dif-
XMnO = 0.35 (point P). fusion is a relatively slow process, equi-
The line PQ is called a tie-line or co- librium cooling conditions are only ap-
node. As the overall composition is varied proached if the temperature is decreased
at 2200 °C between points P and Q, the very slowly. If a sample of composition
compositions of the solid and liquid phases XMnO = 0.60 is cooled very rapidly from the
remain fixed at P and Q, and only the rela- liquid, concentration gradients will be ob-
tive proportions of the two phases change. served in the solid grains, with the concen-
From a simple mass balance, we can derive tration of MnO increasing towards the sur-
the lever rule for binary systems: (moles of face from a minimum of XMnO = 0.28 (point
liquid) / (moles of solid) = PR /R Q. Hence, A) at the center. Furthermore, in this case
at 2200 °C a sample with overall composi- solidification will not be complete at
tion XMnO = 0.60 consists of liquid and solid 2030 °C since at 2030 °C the average con-
phases in the molar ratio (0.60 – 0.35) / centration of MnO in the solid particles
(0.70 – 0.60) = 2.5. Were the composition will be less than XMnO = 0.60. These con-
axis expressed as weight percent, then the siderations are discussed more fully in
lever rule would give the weight ratio of Chapter 2 of this volume (Müller-Krumb-
the two phases. haar et al., 2001).
Suppose that a liquid CaO – MnO solu- At XMnO = 0 and XMnO = 1 in Fig. 1-5 the
tion with composition XMnO = 0.60 is liquidus and solidus curves meet at the
cooled very slowly from an initial tempera- equilibrium melting points, or tempera-
ture of about 2500 °C. When the temperatures of fusion of CaO and MnO, which are
ture has decreased to the liquidus tempera- T f0(CaO) = 2572 °C, T f0(MnO) = 1842 °C.
ture 2270 °C (point B), the first solid The phase diagram is influenced by the
appears, with a composition at point A total pressure, P. Unless otherwise stated,
T – X diagrams are usually presented for Since, to a first approximation, Dhf0 and
P = const. = 1 bar. For equilibria involving Dsf0 are independent of T, Dgf0 is approxi-
only solid and liquid phases, the phase mately a linear function of T. If T > Tf0 ,
boundaries are typically shifted only by the then Dgf0 is negative. If T < Tf0 , then Dgf0 is
order of a few hundredths of a degree per positive. Hence, as seen in Fig. 1-6, as T
bar change in P. Hence, the effect of pres- decreases, the g s curve descends relative to
sure upon the phase diagram is generally g l. At 1500 °C, g l < g s at all compositions.
negligible unless the pressure is of the or- Therefore, by the principle that a system
der of hundreds of bars. On the other hand, always seeks the state of minimum Gibbs
if gaseous phases are involved then the ef- energy at constant T and P, the liquid phase
fect of pressure is very important. The ef- is stable at all compositions at 1500 °C.
fect of pressure will be discussed in Sec. At 1300 °C, the curves of g s and g l cross.
1.5.3. The line P1 Q1 , which is the common tan-
gent to the two curves, divides the compo-
sition range into three sections. For compo-
1.5.2 Thermodynamic Origin
sitions between pure Ge and P1 , a single-
of Phase Diagrams
phase liquid is the state of minimum Gibbs
In this section we first consider the ther- energy. For compositions between Q1 and
modynamic origin of simple “lens-shaped” pure Si, a single-phase solid solution is the
phase diagrams in binary systems with stable state. Between P1 and Q1 , a total
complete liquid and solid miscibility. Gibbs energy lying on the tangent line
An example of such a diagram was given P1 Q1 may be realized if the system adopts
in Fig. 1-5. Another example is the Ge – Si a state consisting of two phases with com-
phase diagram in the lowest panel of Fig. positions at P1 and Q1 and with relative
1-6 (Hansen, 1958). In the upper three pan- proportions given by the lever rule. Since
els of Fig. 1-6, the molar Gibbs energies of the tangent line P1 Q1 lies below both g s and
the solid and liquid phases, g s and g l, at g l, this two-phase state is more stable than
three temperatures are shown to scale. As either phase alone. Furthermore, no other
illustrated in the top panel, g s varies with line joining any point on g l to any point on
composition between the standard molar g s lies below the line P1 Q1 . Hence, this line
Gibbs energies of pure solid Ge and of pure represents the true equilibrium state of the
0 (s)
solid Si, g Ge and g 0Si(s), while g l varies system, and the compositions P1 and Q1 are
between the standard molar Gibbs energies the liquidus and solidus compositions at
0 (l)
of the pure liquid components g Ge and 1300 °C.
0 (l)
g Si . As T is decreased to 1100 °C, the points
0 (l)
The difference between g Ge and g 0Si(s) is of common tangency are displaced to
equal to the standard molar Gibbs energy higher concentrations of Ge. For T < 937 °C,
of fusion (melting) of pure Si, Dg f0(Si) = g s < g l at all compositions.
(g 0Si(l) – g 0Si(s) ). Similarly, for Ge, Dg f0(Ge) = It was shown in Fig. 1-4 that if a tangent
0 (l) 0 (s)
(g Ge – g Ge ). The Gibbs energy of fusion of is drawn to a Gibbs energy curve, then the
a pure component may be written as: intercept of this tangent on the axis at Xi = 1
is equal to the partial Gibbs energy or
Dgf0 = Dhf0 – T Dsf0 (1-60)
chemical potential g i of component i. The
where Dhf0 and Dsf0 are the standard molar common tangent construction of Fig. 1-6
enthalpy and entropy of fusion. thus ensures that the chemical potentials of
Figure 1-6. Ge – Si phase diagram at

P = 1 bar (after Hansen, 1958) and
Gibbs energy composition curves at
three temperatures, illustrating the com-
mon tangent construction (reprinted
from Pelton, 1983).
Ge and Si are equal in the solid and liquid If we rearrange Eq. (1-61), subtracting
phases at equilibrium. That is: the Gibbs energy of fusion of pure Ge,
l s Dg f0(Ge) = (g Ge
0 (l) 0 (s)
– g Ge ), from each side, we
g Ge = gGe (1-61)
get:
l s
g Si =g Si (1-62) l 0 (l) s 0 (s)
(g Ge – g Ge ) – (gGe – g Ge )
This equality of chemical potentials was 0 (l)
= – (g Ge 0 (s)
– g Ge ) (1-63)
shown in Sec. 1.4.2 to be the criterion for
phase equilibrium. That is, the common Using Eq. (1-32), we can write Eq. (1-63)
tangent construction simultaneously mini- as:
mizes the total Gibbs energy and ensures l
Dg Ge s
– DgGe = – Dg f0(Ge) (1-64)
the equality of the chemical potentials,
thereby showing that these are equivalent or
l s
criteria for equilibrium between phases. R T ln aGe – R T ln aGe = – Dg f0(Ge) (1-65)
l
where aGe is the activity of Ge (with re- Dgf0 = 0 when T = Tf0. Calculated phase dia-
spect to pure liquid Ge as standard state) in grams for Dsf0 = 3, 10 and 30 J/mol K are
the liquid solution on the liquidus, and aGes
shown in Fig. 1-7. A value of Dsf0 ≈ 10 is
is the activity of Ge (with respect to pure typical of most metals. However, when the
solid Ge as standard state) in the solid solu- components are ionic compounds such as
tion on the solidus. Starting with Eq. (1- ionic oxides, halides, etc., Dsf0 can be sig-
62), we can derive a similar expression for nificantly larger since there are several
the other component: ions per formula unit. Hence, two-phase
l s “lenses” in binary ionic salt or oxide phase
R T ln aSi – R T ln aSi = – Dg f0(Si) (1-66)
diagrams tend to be “fatter” than those
Eqs. (1-65) and (1-66) are equivalent to the encountered in alloy systems. If we are
common tangent construction. considering vapor – liquid equilibria rather
It should be noted that absolute values of than solid – liquid equilibria, then the
Gibbs energies cannot be defined. Hence, shape is determined by the entropy of
0 (l) 0 (l)
the relative positions of g Ge and g Si in vaporization, Dsv0 . Since Dsv0 is usually an
Fig. 1-6 are completely arbitrary. However, order of magnitude larger than Dsf0 , two-
this is immaterial for the preceding discus- phase (liquid + vapor) lenses tend to be
sion, since displacing both g 0Si(l) and g 0Si(s) by
0 (l)
the same arbitrary amount relative to g Ge
0 (s)
and g Ge will not alter the compositions of
the points of common tangency.
It should also be noted that in the present
discussion of equilibrium phase diagrams
we are assuming that the physical dimen-
sions of the single-phase regions in the
system are sufficiently large that surface
(interfacial) energy contributions to the
Gibbs energy can be neglected. For very
fine grain sizes in the sub-micron range,
however, surface energy effects can notice-
ably influence the phase boundaries.
The shape of the two-phase (solid + liq-
uid) “lens” on the phase diagram is deter-
mined by the Gibbs energies of fusion,
Dgf0 , of the components and by the mixing
terms, Dg s and Dg l. In order to observe
how the shape is influenced by varying
Dgf0 , let us consider a hypothetical system
A– B in which Dg s and Dg l are ideal Raoul-
tian (Eq. (1-45)). Let T f0(A) = 800 K and
T f0(B) = 1200 K. Furthermore, assume that
Figure 1-7. Phase diagram of a system A – B with
the entropies of fusion of A and B are equal
ideal solid and liquid solutions. Melting points of A
and temperature-independent. The enthalp- and B are 800 and 1200 K, respectively. Diagrams
ies of fusion are then given from Eq. (1-60) are calculated for entropies of fusion ∆Sf(A)
0
= ∆Sf(B)
0
=
by the expression Dhf0 = Tf0 Dsf0 since 3, 10 and 30 J/mol K.
very wide. For equilibria between two solid common tangent construction then gives
solutions of different crystal structure, the the equilibrium vapor and liquid composi-
shape is determined by the entropy of tions. The phase diagram depends upon the
solid – solid transformation, which is usu- Gibbs energies of vaporization of the com-
ally smaller than the entropy of fusion by ponents Dg v (Zn) and Dg v (Mg) as shown in
approximately an order of magnitude. Fig. 1-8.
Therefore two-phase (solid + solid) lenses To generate the isothermal pressure –
tend to be very narrow. composition (P – X ) phase diagram in the
lower panel of Fig. 1-8 we require the
1.5.3 Pressure – Composition Phase Gibbs energies of vaporization as functions
Diagrams of P. Assuming monatomic ideal vapors
and assuming that pressure has negligible
Let us consider liquid – vapor equilib-
effect upon the Gibbs energy of the liquid,
rium with complete miscibility, using as an
we can write:
example the Zn – Mg system. Curves of gv
and g l can be drawn at any given T and P, Dg v(i) = Dg0v(i) + R T ln P (1-67)
as in the upper panel of Fig. 1-8, and the
where Dg v(i) is the standard Gibbs energy
of vaporization (when P = 1 bar), which is
given by:
Dg0v(i) = Dh0v(i) – T Ds0v(i) (1-68)
For example, the enthalpy of vaporization

of Zn is Dh0v (Zn) = 115 300 J/mol at its nor-
mal boiling point of 1180 K (Barin et al.,
1977). Assuming that Dh0v is independent
of T, we calculate from Eq. (1-68) that
Ds0v (Zn) = 115 300 /1180 = 97.71 J/mol K.
From Eq. (1-67), Dg v (Zn) at any T and P is
thus given gy:
(1-69)
Dg v (Zn) = (115 300 – 97.71 T ) + R T ln P
A similar expression can be derived for the

other component Mg.
At constant temperature, then, the curve
of gv in Fig. 1-8 descends relative to gl as
the pressure is lowered, and the P – X phase
diagram is generated by the common tan-
gent construction. The diagram at 1250 K
in Fig. 1-8 was calculated under the as-
sumption of ideal liquid and vapor mixing
(gE (l) = 0, gE (v) = 0).
Figure 1-8. Pressure – composition phase diagram
of the Zn – Mg system at 1250 K calculated for ideal P – X phase diagrams involving liquid –
vapor and liquid solutions. Upper panel illustrates solid or solid – solid equilibria can be cal-
common tangent construction at a constant pressure. culated in a similar fashion through the fol-
lowing general equation, which gives the that the solid state is the stable state for
effect of pressure upon the Gibbs energy both pure components. However, in this
change for the transformation of one mole system gE (l) < gE (s), so that g s presents a
of pure component i from an a-phase to a flatter curve than does g l and there exists a
b-phase: central composition region in which gl < gs.
P Hence, there are two common tangent
b
Dga→ b = Dga0 → b + ∫ (vi − via ) dP (1- 70 ) lines, P1 Q1 and P2 Q2 . Such a situation
P =1 gives rise to a phase diagram with a mini-
where Dg 0aÆ b is the standard (P = 1 bar) mum in the two-phase region, as observed
Gibbs energy of transformation, and vib and in the Na2CO3 – K 2CO3 system (Dessu-
via are the molar volumes. reault et al., 1990) shown in Fig. 1-10. At a
composition and temperature correspond-
ing to the minimum point, liquid and solid
1.5.4 Minima and Maxima
of the same composition exist in equilib-
in Two-Phase Regions
rium.
As discussed in Sec. 1.4.8, the Gibbs en- A two-phase region with a minimum
ergy of mixing Dg m may be expressed as point as in Fig. 1-10 may be thought of as a
the sum of an ideal term Dgmideal and an ex- two-phase “lens” which has been “pushed
cess term gE. As has just been shown in down” by virtue of the fact that the liquid is
Sec. 1.5.2, if Dg ms and Dg ml
for the solid relatively more stable than the solid. Ther-
and liquid phases are both ideal, then a modynamically, this relative stability is ex-
“lens-shaped” two-phase region always re- pressed as gE (l) < gE (s).
sults. However in most systems even ap- Conversely, if gE (l) > gE (s) to a sufficient
proximately ideal behavior is the exception extent, then a two-phase region with a
rather than the rule. maximum will result. Such maxima in (liq-
Curves of g s and gl for a hypothetical uid + solid) or (solid + solid) two-phase re-
system A– B are shown schematically in gions are nearly always associated with the
Fig. 1-9 at a constant temperature (below existence of an intermediate phase, as will
the melting points of pure A and B) such be discussed in Sec. 1.5.10.
Figure 1-9. Isothermal Gibbs energy-composition Figure 1-10. Phase diagram of the K2CO3– Na2CO3
curves for solid and liquid phases in a system A – B in system at P = 1 bar (Dessureault et al., 1990).
which g E(l) > g E(s). A phase diagram of the type of
Fig. 1-10 results.
1.5.5 Miscibility Gaps
If gE > 0, then the solution is thermody-

namically less stable than an ideal solution.
This can result from a large difference in
size of the component atoms, ions or mole-
cules, which will lead to a (positive) lattice
strain energy, or from differences in elec-
tronic structure, or from other factors.
In the Au – Ni system, gE is positive in
the solid phase. In the top panel of Fig. 1-11,
gE (s) is plotted at 1200 K (Hultgren et al.,
1973) and the ideal Gibbs energy of
mixing, Dgmideal , is also plotted at 1200 K.
The sum of these two terms is the Gibbs
energy of mixing of the solid solution,
Dgms, which is plotted at 1200 K as well
as at other temperatures in the central panel
of Fig. 1-11. Now, from Eq. (1-45), Dgmideal
is always negative and varies directly
with T, whereas gE varies less rapidly with
temperature. As a result, the sum Dgms =
Dgmideal + gE becomes less negative as T de-
creases. However, the limiting slopes to the
Dgmideal curve at XAu = 1 and XNi = 1 are both
infinite, whereas the limiting slopes of gE
are always finite (Henry’s Law). Hence,
s
Dgm will always be negative as XAu Æ1
and XNi Æ1 no matter how low the temper-
ature. As a result, below a certain tempera-
s
ture the curve of Dgm will exhibit two neg-
ative “humps”. Common tangent lines Figure 1-11. Phase diagram (after Hultgren et al.,
1973) and Gibbs energy – composition curves of solid
P1 Q1 , P2 Q2 , P3 Q3 to the two humps at dif- solutions for the Au–Ni system at P = 1 bar. Letters
ferent temperatures define the ends of tie- “s” indicate spinodal points (Reprinted from Pelton,
lines of a two-phase solid – solid miscibility 1983).
gap in the Au – Ni phase diagram, which is
shown in the lower panel in Fig. 1-11
(Hultgren et al., 1973). The peak of the gap have less of an effect upon gE (l) in the liq-
occurs at the critical or consolute tempera- uid phase owing to the greater flexibility of
ture and composition, Tc and Xc . the liquid structure to accommodate differ-
When gE (s) is positive for the solid phase ent atomic sizes, valencies, etc. Hence, a
in a system it is usually also the case that solid – solid miscibility gap is often asso-
gE (l) < gE (s) since the unfavorable factors ciated with a minimum in the two-phase
(such as a difference in atomic dimensions) (solid + liquid) region, as is the case in the
which are causing gE (s) to be positive will Au – Ni system.
Below the critical temperature the curve gap at any temperature. Suppose that gE (s)
s
of Dgm exhibits two inflection points, indi- is so positve that Tc is higher than the min-
cated by the letter “s” in Fig. 1-11. These imum in the (solid + liquid) region. The re-
are known as the spinodal points. On the sult will be a phase diagram such as that of
phase diagram their locus traces out the the MgO – CaO system shown in Fig. 1-12
spinodal curve (Fig. 1-11). The spinodal (Doman et al., 1963; Wu, 1990).
curve is not part of the equilibrium phase The lower panel of Fig. 1-12 shows the
diagram, but it is important in the kinetics Gibbs energy curves at 2450 °C. The two
of phase separation, as discussed in Chap- common tangents define two two-phase re-
ter 6 (Binder and Fratzl, 2001). gions. As the temperature is decreased be-
low 2450 °C, the g s curve descends relative
to g l and the two points of tangency P1
1.5.6 Simple Eutectic Systems
and P2 approach each other until, at T =
The more positive gE is in a system, the 2374 °C, P1 and P2 become coincident at
higher is Tc and the wider is the miscibility the composition E. That is, at T = 2374 °C
Figure 1-12. Phase diagram

at P = 1 bar (after Doman et
al., 1963, and Wu, 1990) and
Gibbs energy – composition
curves at 2450 °C for the
MgO – CaO system. Solid
MgO and CaO have the
same crystal structure.
there is just one common tangent line con- the liquid has solidified, and during the re-
tacting the two portions of the g s curve at action the compositions of the three phases
compositions A and B and contacting the gl will remain fixed at A, B and E. For this
curve at E. This temperature is known as reason the eutectic reaction is called an in-
the eutectic temperature, TE , and the com- variant reaction. More details on eutectic
position E is the eutectic composition. For solidification may be found in Chapter 2
temperatures below TE , gl lies completely (Müller-Krumbhaar et al., 2001).
above the common tangent to the two por-
tions of the g s curve and so for T < TE a
1.5.7 Regular Solution Theory
solid – solid miscibility gap is observed.
The phase boundaries of this two-phase re- Many years ago Van Laar (1908) showed
gion are called the solvus lines. The word that the thermodynamic origin of a great
eutectic is from the Greek for “to melt many of the observed features of binary
well” since the system has its lowest melt- phase diagrams can be illustrated at least
ing point at the eutectic composition E. qualitatively by simple regular solution
This description of the thermodynamic theory. A simple regular solution is one for
origin of simple eutectic phase diagrams is which:
strictly correct only if the pure solid com-
gE = X A X B ( w – h T ) (1-72)
ponents A and B have the same crystal
structure. Otherwise, a curve for g s which where w and h are parameters independent
is continuous at all compositions cannot be of temperature and composition. Substitut-
drawn. ing Eq. (1-72) into Eq. (1-29) yields, for
Suppose a liquid MgO – CaO solution of the partial properties:
composition XCaO = 0.52 (composition P1) (1-73)
E 2 E 2
is cooled from the liquid state very slowly gA = XB (w – h T ) , gB = XA (w – h T )
under equilibrium conditions. At 2450 °C Several liquid and solid solutions con-
the first solid appears with composition Q1 . form approximately to regular solution be-
As T decreases further, solidification con- havior, particularly if gE is small. Examples
tinues with the liquid composition follow- may be found for alloys, molecular solu-
ing the liquidus curve from P1 to E and the tions, and ionic solutions such as molten
composition of the solid phase following salts and oxides, among others. (The very
the solidus curve from Q1 to A. The rela- low values of gE observed for gaseous solu-
tive proportions of the two phases at any T tions generally conform very closely to Eq.
are given by the lever rule. At a tempera- (1-72).)
ture just above TE , two phases are ob- To understand why this should be so, we
served: a solid of composition A and a liq- only need a very simple model. Suppose
uid of composition E. At a temperature just that the atoms or molecules of the compo-
below TE , two solids with compositions A nents A and B mix substitutionally. If the
and B are observed. Therefore, at TE , dur- atomic (or molecular) sizes and electronic
ing cooling, the following binary eutectic structures of A and B are similar, then the
reaction occurs: distribution will be nearly random, and the
liquid Æ solid1 + solid2 (1-71) configurational entropy will be nearly
ideal. That is:
Under equilibrium conditions the tempera-
ture will remain constant at T = TE until all gE ≈ Dhm – T S E (non-config) (1-74)
More will be said on this point in Sec. tively weak, then the solution is rendered
1.10.5. less stable, (w – h T ) > 0 and gE > 0.
We now assume that the bond energies Simple non-polar molecular solutions
eAA , e BB and eAB of nearest-neighbor pairs and ionic solutions such as molten salts of-
are independent of temperature and com- ten exhibit approximately regular behavior.
position and that the average nearest- The assumption of additivity of the energy
neighbor coordination number, Z, is also of pair bonds is probably reasonably realis-
constant. Finally, we assume that the en- tic for van der Waals or coulombic forces.
thalpy of mixing results mainly from the For alloys, the concept of a pair bond is, at
change in the total energy of nearest-neigh- best, vague, and metallic solutions tend to
bor pair bonds. exhibit larger deviations from regular be-
In one mole of solution there are (N 0 havior.
Z / 2) neareast-neighbor pair bonds, where In several solutions it is found that
N 0 is Avogadro’s number. Since the distri- | h T | < | w | in Eq. (1-72). That is, gE ≈ Dhm
bution is assumed random, the probability = w XA XB , and to a first approximation
that a given bond is an A–A bond is equal gE is independent of T. This is more often
to X A2 . The probabilities of B – B and A– B the case in non-metallic solutions than in
bonds are, respectively, X B2 and 2 XA XB . metallic solutions.
The molar enthalpy of mixing is then equal
to the sum of the energies of the nearest-
1.5.8 Thermodynamic Origin
neighbor bonds in one mole of solution,
of Simple Phase Diagrams Illustrated
minus the energy of the A–A bonds in XA
by Regular Solution Theory
moles of pure A and the energy of the B – B
bonds in XB moles of pure B: Figure 1-13 shows several phase dia-
grams, calculated for a hypothetical system
Dhm = (N 0 Z /2)
A– B containing a solid and a liquid phase
¥ (X A2 eAA + X B2 e BB + 2 XAB eAB ) with melting points of T f0(A) = 800 K and
– (N Z /2) (XA eAA) – (N 0 Z /2) (XB e BB )
0
T f0(B) = 1200 K and with entropies of fusion
= (N 0 Z ) [eAB – (eAA + e BB )/2] XA XB of both A and B set to 10 J/mol K, which is
= w XA XB (1-75) a typical value for metals. The solid and
liquid phases are both regular with temper-
We now define sAB , sAA and s BB as the ature-independent excess Gibbs energies
vibrational entropies of nearest-neighbor
pair bonds. Following an identical argu- gE (s) = w s XA XB and gE (l) = w l XA XB
ment to that just presented for the bond The parameters w s and w l have been varied
energies we obtain: systematically to generate the various pan-
s E (non-config) (1-76) els of Fig. 1-13.
0 In panel (n) both phases are ideal. Panels
= (N Z ) [sAB – (sAA + s BB )/2] = h XA XB
(l) to (r) exhibit minima or maxima de-
Eq. (1-72) has thus been derived. If A– B pending upon the sign and magnitude of
bonds are stronger than A–A and B – B (gE (l) – gE (s) ), as has been discussed in Sec.
bonds, then (eAB – h AB T ) < [(eAA – h AA T )/2 1.5.4. In panel (h) the liquid is ideal but
+ (e BB – h BB T ) / 2]. Hence, (w – h T ) < 0 positive deviations in the solid give rise to
and gE < 0. That is, the solution is rendered a solid – solid miscibility gap as discussed
more stable. If the A– B bonds are rela- in Sec. 1.5.6. On passing from panel (h) to
Figure 1-13. Topological changes in the phase diagram for a system A–B with regular solid and liquid phases,
brought about by systematic changes in the regular solution parameters ω s and ω l. Melting points of pure A and
B are 800 K and 1200 K. Entropies of fusion of both A and B are 10.0 J/mol K (Pelton and Thompson, 1975).
The dashed curve in panel (d) is the metastable liquid miscibility gap (Reprinted from Pelton, 1983).
panel (c), an increase in gE (s) results in a sultant lowering of the eutectic tempera-
widening of the miscibility gap so that the ture.
solubility of A in solid B and of B in solid Eutectic phase diagrams are often drawn
A decreases. Panels (a) to (c) illustrate that with the maximum solid solubility occur-
negative deviations in the liquid cause a ring at the eutectic temperature (as in Fig.
relative stabilization of the liquid with re- 1-12). However, panel (d) of Fig. 1-13, in
which the maximum solubility of A in the by sufficiently rapid cooling, then a meta-
B-rich solid solution occurs at approxi- stable liquid – liquid miscibility gap is ob-
mately T = 950 K, illustrates that this need served as shown in Fig. 1-13 (d). For exam-
not be the case even for simple regular so- ple, in the Na2O – SiO2 system the flattened
lutions. (or “S-shaped”) SiO2 liquidus heralds the
existence of a metastable miscibility gap of
importance in glass technology.
1.5.9 Immiscibility – Monotectics
In Fig. 1-13 (e), positive deviations in
1.5.10 Intermediate Phases
the liquid have given rise to a liquid – liquid
miscibility gap. The CaO – SiO2 system The phase diagram of the Ag – Mg
(Wu, 1990), shown in Fig. 1-14, exhibits system (Hultgren et al., 1973) is shown in
such a feature. Suppose that a liquid of Fig. 1-15 (d). An intermetallic phase, b¢, is
composition XSiO2 = 0.8 is cooled slowly seen centered approximately about the
from high temperatures. At T = 1815 °C the composition XMg = 0.5. The Gibbs energy
miscibility gap boundary is crossed and a curve at 1050 K for such an intermetallic
second liquid layer appears with a compo- phase has the form shown schematically in
sition of XSiO2 = 0.97. As the temperature is Fig. 1-15 (a). The curve g b¢ rises quite rap-
lowered further, the composition of each idly on either side of its minimum, which
liquid phase follows its respective phase occurs near XMg = 0.5. As a result, the b¢
boundary until, at 1692 °C, the SiO2-rich phase appears on the phase diagram only
liquid has a composition of XSiO2 = 0.99 over a limited composition range. This
(point B), and in the CaO-rich liquid form of the curve g b¢ results from the fact
XSiO2 = 0.74 (point A). At any temperature, that when XAg ≈ XMg a particularly stable
the relative amounts of the two phases are crystal structure exists in which Ag and Mg
given by the lever rule. atoms preferentially occupy different sites.
At 1692 °C the following invariant bi- The two common tangents P1 Q1 and P2 Q2
nary monotectic reaction occurs upon cool- give rise to a maximum in the two-phase
ing: (b¢ + liquid) region of the phase diagram.
(Although the maximum is observed very
Liquid B Æ Liquid A + SiO2 (solid) (1-77)
near XMg = 0.5, there is no thermodynamic
The temperature remains constant at reason for the maximum to occur exactly at
1692 °C and the compositions of the phases this composition.)
remain constant until all of liquid B is con- Another intermetallic phase, the e phase,
sumed. Cooling then continues with pre- is also observed in the Ag – Mg system,
cipitation of solid SiO2 with the equilib- Fig. 1-15. The phase is associated with a
rium liquid composition following the liq- peritectic invariant ABC at 744 K. The
uidus from point A to the eutectic E. Gibbs energy curves are shown schemati-
Returning to Fig. 1-13, we see in panel cally at the peritectic temperature in Fig.
(d) that the positive deviations in the liquid 1-15 (c). One common tangent line can be
in this case are not large enough to produce drawn to gl, g b¢ and g e.
immiscibility, but they do result in a flat- Suppose that a liquid alloy of composi-
tening of the liquidus, which indicates a tion XMg = 0.7 is cooled very slowly from
“tendency to immiscibility”. If the nuclea- the liquid state. At a temperature just above
tion of the solid phases can be suppressed 744 K a liquid phase of composition C and
Figure 1-14. CaO – SiO2 phase diagram at P = 1 bar (after Wu, 1990) and Gibbs energy curves at 1500 °C illus-
trating Gibbs energies of fusion and formation of the stoichiometric compound CaSiO3 .
This reaction occurs isothermally at 744 K

with all three phases at fixed compositions
(at points A, B and C). For an alloy with
overall composition between points A and
B the reaction proceeds until all the liquid
has been consumed. In the case of an alloy
with overall composition between B and C,
the b¢ phase will be the first to be com-
pletely consumed.
Peritectic reactions occur upon cooling
with formation of the product solid (e in
this example) on the surface of the reactant
solid (b¢), thereby forming a coating which
can prevent further contact between the re-
actant solid and liquid. Further reaction
may thus be greatly retarded so that equi-
librium conditions can only be achieved by
extremely slow cooling.
The Gibbs energy curve for the e phase,
g e, in Fig. 1-15 (c) rises more rapidly on ei-
ther side of its minimum than does the
Gibbs energy g b¢ for the b¢ phase in Fig. 1-
15 (a). As a result, the width of the single-
phase region over which the e phase exists
(sometimes called its range of stoichiome-
try or homogeneity range) is narrower than
for the b¢ phase.
In the upper panel of Fig. 1-14 for the
CaO – SiO2 system, Gibbs energy curves at
1500 °C for the liquid and CaSiO3 phases
are shown schematically. g 0.5 (CaSiO3 ) rises
extremely rapidly on either side of its min-
imum. (We write g 0.5 (CaSiO3 ) for 0.5 moles
of the compound in order to normalize to a
Figure 1-15. Ag – Mg phase diagram at P = 1 bar (af- basis of one mole of components CaO and
ter Hultgren at al., 1973) and Gibbs energy curves at
three temperatures.
SiO2 .) As a result, the points of tangency
Q1 and Q 2 of the common tangents P1 Q1
and P2 Q2 nearly (but not exactly) coincide.
a b¢ phase of composition A are observed Hence, the range of stoichiometry of the
at equilibrium. At a temperature just below CaSiO3 phase is very narrow (but never
744 K the two phases at equilibrium are b¢ zero). The two-phase regions labelled
of composition A and e of composition B. (CaSiO3 + liquid) in Fig. 1-14 are the two
The following invariant binary peritectic sides of a two-phase region that passes
reaction thus occurs upon cooling: through a maximum at 1540 °C just as the
Liquid + b¢ (solid) Æ e (solid) (1-78) (b¢ + liquid) region passes through a maxi-
mum in Fig. 1-15 (d). Because the CaSiO3 Eq. (1-78), to form a liquid of composition
single-phase region is so narrow, we refer C and another solid phase, b¢, of composi-
to CaSiO3 as a stoichiometric compound. tion A.
Any deviation in composition from the Another example of an incongruent
stoichiometric 1 : 1 ratio of CaO to SiO2 compound is Ca3Si2O7 in Fig. 1-14, which
results in a very large increase in Gibbs melts incongruently (or peritectically) to
energy. form liquid and Ca2SiO4 at the peritectic
The e phase in Fig. 1-15 is based on the temperature of 1469 °C.
stiochiometry AgMg3 . The Gibbs energy An incongruent compound is always as-
curve, Fig. 1-15 (c), rises extremely rapidly sociated with a peritectic. However, the
on the Ag side of the minimum, but some- converse is not necessarily true. A peritec-
what less steeply on the Mg side. As a re- tic is not always associated with an inter-
sult, Ag is virtually insoluble in AgMg3 , mediate phase. See, for example, Fig. 1-
while Mg is sparingly soluble. Such a 13 (i).
phase with a narrow range of homogeneity For purposes of phase diagram calcula-
is often called a non-stoichiometric com- tions involving stoichiometric compounds
pound. At low temperatures the b¢ phase such as CaSiO3 , we may, to a good approx-
exhibits a relatively narrow range of stoi- imation, consider the Gibbs energy curve,
chiometry about the 1 : 1 AgMg composi- g 0.5 (CaSiO3 ) , to have zero width. All that
tion and can properly be called a com- is then required is the value of g 0.5 (CaSiO3 )
pound. However, at higher temperatures it at the minimum. This value is usually
is debatable whether a phase with such a expressed in terms of the Gibbs energy
wide range of composition should be called of fusion of the compound, Dg 0f (0.5 CaSiO3 )
a “compound”. or the Gibbs energy of formation
From Fig. 1-14 it can be seen that if stoi- Dg 0form (0.5 CaSiO3 ) of the compound from
chiometric CaSiO3 is heated it will melt the pure solid components CaO and SiO2
isothermally at 1540 °C to form a liquid of according to the reaction: 0.5 CaO (sol) +
the same composition. Such a compound is 0.5 SiO2 (sol) = 0.5 CaSiO3 (sol). Both these
called congruently melting or simply a con- quantities are interpreted graphically in
gruent compound. The compound Ca2SiO4 Fig. 1-14.
in Fig. 1-14 is congruently melting. The b¢
phase in Fig. 1-15 is also congruently melt-
ing at the composition of the liquidus/sol- 1.5.11 Limited Mutual Solubility –
idus maximum. Ideal Henrian Solutions
It should be noted with regard to the con-
gruent melting of CaSiO3 in Fig. 1-14 that In Sec. 1.5.6, the region of two solids in
the limiting slopes dT /dX of both branches the MgO – CaO phase diagram of Fig. 1-12
of the liquidus at the congruent melting was described as a miscibility gap. That is,
point (1540 °C) are zero since we are really only one continuous g s curve was assumed.
dealing with a maximum in a two-phase re- If, somehow, the appearance of the liquid
gion. phase could be suppressed, then the two
The AgMg3 (e) compound in Fig. 1-15 is solvus lines in Fig. 1-12, when projected
said to melt incongruently. If solid AgMg3 upwards, would meet at a critical point
is heated it will melt isothermally at 744 K above which one continuous solid solution
by the reverse of the peritectic reaction, would exist at all compositions.
Such a description is justifiable only if g solvent ≈ 1

the pure solid components have the same g solute ≈ g solute
0
= constant (1-79)
crystal structure, as is the case for MgO
and CaO. However, consider the Ag – Mg (Care must be exercised for solutions other
system, Fig. 1-15, in which the terminal than simple substitutional solutions. Henry’s
(Ag) solid solution is face-centered-cubic Law applies only if the ideal activity is defined
and the terminal (Mg) solid solution is hex- correctly, as will be discussed in Sec. 1.10).
agonal-close-packed. In this case, one con- Treating, then, the h.c.p. (Mg) phase
tinuous curve for g s cannot be drawn. Each in the Ag – Mg system (Fig. 1-15 (b)) as a
solid phase must have its own separate Henrian solution we write:
Gibbs energy curve, as shown schemati-
g h.c.p. = (XAg g0Ag(f.c.c.) + XMg g0Mg
(h.c.p.)
)
cally in Fig. 1-15 (b) for the h.c.p. (Mg)
phase at 800 K. In this figure, g0Mg (h.c.p.)
and + R T (XAg ln aAg + XMg ln aMg )
0 (f.c.c.)
g Ag are the standard molar Gibbs ener- = (XAg g0Ag(f.c.c.) + XMg g0Mg
(h.c.p.)
) (1-80)
gies of pure h.c.p. Mg and pure f.c.c. Ag, 0
+ R T (XAg ln (g Ag XAg ) + XMg ln XMg )
while g0Ag(h.c.p.-Mg) is the standard molar
Gibbs energy of pure (hypothetical) h.c.p. 0
where aAg and g Ag are the activity and ac-
Ag in the h.c.p. (Mg) phase. tivity coefficient of silver with respect to
Since the solubility of Ag in the h.c.p. pure f.c.c. silver as standard state. Let us
(Mg) phase is limited we can, to a good ap- now combine terms as follows:
proximation, describe it as a Henrian ideal
solution. That is, when a solution is suffi- g h.c.p. = [XAg (g0Ag(f.c.c.) + R T ln g Ag
0
)
ciently dilute in one component, we can ap- + XMg g0Mg
(h.c.p.)
] (1-81)
proximate g Esolute = R T ln g solute by its value + R T (XAg ln XAg + XMg ln XMg )
in an infinitely dilute solution. That is, if
0
0
Xsolute is small we set g solute = g solute where Since g Ag is independent of composition,
0
g solute is the Henrian activity coefficient at let us define:
Xsolute = 0. Thus, for sufficiently dilute solu- 0
g0Ag(h.c.p.-Mg) = (g0Ag(f.c.c.) + R T ln g Ag ) (1-82)
tions we assume that g solute is independent
of composition. Physically, this means that From Eqs. (1-81) and (1-82) it can be seen
in a very dilute solution there is negligible that, relative to g0Mg (h.c.p.)
and to the hypothet-
0 (h.c.p.-Mg)
interaction among solute particles because ical standard state g Ag defined in
they are so far apart. Hence, each addi- this way, the h.c.p. solution is ideal. Eqs.
tional solute particle added to the solution (1-81) and (1-82) are illustrated in Fig. 1-
0
produces the same contribution to the ex- 15 (b). It can be seen that as g Ag becomes
cess Gibbs energy of the solution and so g E- larger, the point of tangency N moves
E
solute = dG /dnsolute = constant. to higher Mg concentrations. That is, as
From the Gibbs – Duhem equation, Eq. (g0Ag(h.c.p.-Mg) – g0Ag(f.c.c.) ) becomes more posi-
(1-56), if dg Esolute = 0, then dg Esolvent = 0. tive, the solubility of Ag in h.c.p. (Mg) de-
Hence, in a Henrian solution g solute is also creases.
constant and equal to its value in an infi- It must be stressed that g0Ag(h.c.p.-Mg) as de-
nitely dilute solution. That is, g solute = 1 and fined by Eq. (1-82) is solvent-dependent.
the solvent behaves ideally. In summary That is, g0Ag(h.c.p.-Mg) is not the same as, say,
then, for dilute solutions (Xsolvent ≈ 1) g0Ag(h.c.p.-Cd) for Ag in dilute h.c.p. (Cd) solid
Henry’s Law applies: solutions.
Henrian activity coefficients can usually This is the Gibbs Phase Rule. F is called
be expressed as functions of temperature: the number of degrees of freedom or vari-
ance of the system and is the number of pa-
R T ln g i0 = a – b T (1-83)
rameters which can and must be specified
where a and b are constants. If data are lim- in order to completely specify the state of
ited, it can further be assumed that b ≈ 0 so the system.
that R T ln g i0 ≈ constant. Binary temperature – composition phase
diagrams are plotted at a fixed pressure,
usually 1 bar. This then eliminates one de-
1.5.12 Geometry of Binary Phase gree of freedom. In a binary system, C = 2.
Diagrams Hence, for binary isobaric T – X diagrams
The geometry of all types of phase dia- the phase rule reduces to:
grams of any number of components is F=3–P (1-86)
governed by the Gibbs Phase Rule.
Consider a system with C components in Binary T – X diagrams contain single-
which P phases are in equilibrium. The phase areas and two-phase areas. In the sin-
system is described by the temperature, the gle-phase areas, F = 3 – 1 = 2. That is, tem-
total pressure and the composition of each perature and composition can be specified
phase. In a C-component system, (C – 1) in- independently. These regions are thus
dependent mole fractions are required to called bivariant. In two-phase regions,
describe the composition of each phase F = 3 – 2 = 1. If, say, T is specified, then the
(because S Xi = 1). Hence, the total number compositions of both phases are deter-
of variables required to describe the system mined by the ends of the tie-lines. Two-
is [P (C – 1) + 2]. However, as shown in Sec. phase regions are thus termed univariant.
1.4.2, the chemical potential of any compo- Note that the overall composition can be
nent is the same in all phases (a, b, g, …) varied within a two-phase region at con-
since the phases are in equilibrium. That is: stant T, but the overall composition is not a
parameter in the sense of the phase rule.
gia (T, P, X1a , X2a , X3a , …) Rather, it is the compositions of the indi-
= gib (T, P, X1b, X2b, X3b, …) vidual phases at equilibrium that are the
parameters to be considered in counting the
= gig(T, P, X1g, X2g, X3g, …) = … (1-84)
number of degrees of freedom.
where gia (T, P, X1a , X2a , X3a , …) is a func- When three phases are at equilibrium in
tion of temperature, of total pressure, and a binary system at constant pressure,
of the mole fractions X1a , X2a , X3a , … in F = 3 – 3 = 0. Hence, the compositions of
the a phase; and similarly for the other all three phases, as well as T, are fixed.
phases. Thus there are C (P – 1) indepen- There are two general types of three-phase
dent equations in Eq. (1-84) relating the invariants in binary phase diagrams. These
variables. are the eutectic-type and peritectic-type
Let F be the differences between the invariants as illustrated in Fig. 1-16. Let
number of variables and the number of the three phases concerned be called a, b
equations relating them: and g, with b as the central phase as shown
in Fig. 1-16. The phases a, b and g can be
F = P (C – 1) + 2 – C (P – 1) solid, liquid or gaseous. At the eutectic-
F=C–P+2 (1-85) type invariant, the following invariant re-
tectic reaction is l + s2 Æ s1 ; (ii) syntectics

(Fig. 1-13 (k)) in which a = liquid1 , b =
solid, g = liquid2 . The syntectic reaction is
l1 + l2 Æ s; (iii) peritectoids in which a =
solid1 , b = solid2 , g = solid3 . The peritec-
toid reaction is s1 + s3 Æ s2 .
An important rule of construction which
applies to invariants in binary phase dia-
grams is illustrated in Fig. 1-16. This ex-
tension rule states that at an invariant the
extension of a boundary of a two-phase re-
gion must pass into the adjacent two-phase
region and not into a single-phase region.
Examples of both correct and incorrect
constructions are given in Fig. 1-16. To
understand why the “incorrect extensions”
shown are not right consider that the (a + g)
phase boundary line indicates the composi-
tion of the g-phase in equilibrium with the
Figure 1-16. Some geometrical units of binary phase a-phase, as determined by the common
diagrams, illustrating rules of construction.
tangent to the Gibbs energy curves. Since
there is no reason for the Gibbs energy
curves or their derivatives to change dis-
action occurs isothermally as the system is continuously at the invariant temperature,
cooled: the extension of the (a + g) phase boundary
also represents the stable phase boundary
bÆa+g (1-87)
under equilibrium conditions. Hence, for
whereas at the peritectic-type invariant the this line to extend into a region labeled as
invariant reaction upon cooling is: single-phase g is incorrect.
Two-phase regions in binary phase dia-
a+gÆb (1-88)
grams can terminate: (i) on the pure com-
Some examples of eutectic-type invari- ponent axes (at XA = 1 or XB = 1) at a trans-
ants are: (i) eutectics (Fig. 1-12) in which formation point of pure A or B; (ii) at a
a = solid1 , b = liquid, g = solid2 ; the eutectic critical point of a miscibility gap; (iii) at an
reaction is l Æ s1 + s2 ; (ii) monotectics (Fig. invariant. Two-phase regions can also ex-
1-14) in which a = liquid1 , b = liquid2 , g = - hibit maxima or minima. In this case, both
solid; the monotectic reaction is l2 Æ l1 + s; phase boundaries must pass through their
(iii) eutectoids in which a= maximum or minimum at the same point as
solid1 , b = solid2 , g = solid3 ; the eutectoid shown in Fig. 1-16.
reaction is s2 Æ s1 + s3 ; (iv) catatectics in All the geometrical units of construction
which a = liquid, b = solid1 , g = solid2 ; the of binary phase diagrams have now been
catatectic reaction is s1 Æ l + s2 . discussed. The phase diagram of a binary
Some examples of peritectic-type invari- alloy system will usually exhibit several of
ants are: (i) peritectics (Fig. 1-15) in which these units. As an example, the Fe – Mo
a = liquid, b = solid1 , g = solid2 . The peri- phase diagram (Kubaschewski, 1982) is
Figure 1-17. Fe – Mo
phase diagram at P = 1 bar
(Kubaschewski, 1982).
shown in Fig. 1-17. The invariants in this and 2051 °C; a monotectic at 1692 °C; and
system are peritectics at 1540, 1488 and a peritectic at 1469 °C. The compound
1450 °C; eutectoids at 1235 and 1200 °C; Ca3SiO5 dissociates upon heating to CaO
peritectoids at 1370 and 950 °C. The two- and Ca2SiO4 by a peritectoid reaction at
phase (liquid + g) region passes through a 1789 °C and dissociates upon cooling to
minimum at XMo = 0.2. CaO and Ca2SiO4 by a eutectoid reaction at
Between 910 °C and 1390 °C is a two- 1250 °C. Maxima are observed at 2130 and
phase (a + g) g-loop. Pure Fe adopts the 1540 °C. At 1470 °C there is an invariant
f.c.c. g structure between these two temper- associated with the tridymite Æ cristobalite
atures but exists as the b.c.c. a phase at transition of SiO2 . This is either a peritec-
higher and lower temperatures. Mo, how- tic or a catatectic depending upon the rela-
ever, is more soluble in the b.c.c. than tive solubility of CaO in tridymite and cris-
0 (b.c.c.-Fe)
in the f.c.c. structure. That is, g Mo tobalite. However, these solubilities are
0 (f.c.c.-Fe)
< g Mo as discussed in Sec. 1.5.11. very small and unknown.
Therefore, small additions of Mo stabilize
the b.c.c. structure.
In the CaO – SiO2 phase diagram, Fig.
1-14, we observe eutectics at 1439, 1466
1.6 Application of important step in the development of data-

Thermodynamics to Phase bases for multicomponent systems.
Diagram Analysis
1.6.2 Polynomial Representation
1.6.1 Thermodynamic/Phase Diagram of Excess Properties
Optimization
Empirical equations are required to ex-
In recent years the development of press the excess thermodynamic properties
solution models, numerical methods and of the solution phases as functions of com-
computer software has permitted a quanti- position and temperature. For many simple
tative application of thermodynamics to binary substitutional solutions, a good rep-
phase diagram analysis. For a great many resentation is obtained by expanding the
systems it is now possible to perform a excess enthalpy and entropy as polynomi-
simultaneous critical evaluation of avail- als in the mole fractions XA and XB of the
able phase diagram measurements and of components:
available thermodynamic data (calorimet-
hE = XA XB [h0 + h1 (XB – XA ) (1-89)
ric data, measurements of activities, etc.) 2 3
with a view to obtaining optimized equa- + h2 (XB – XA) + h3 (XB – XA) + …]
tions for the Gibbs energies of each phase s E = XA XB [s0 + s1 (XB – XA ) (1-90)
which best represent all the data. These + s2 (XB – XA)2 + s3 (XB – XA)3 + …]
equations are consistent with thermody-
namic principles and with theories of solu- where the hi and si are empirical coeffi-
tion behavior. cients. As many coefficients are used as
The phase diagram can be calculated are required to represent the data in a
from these thermodynamic equations, and given system. For most systems it is a good
so one set of self-consistent equations de- approximation to assume that the coeffi-
scribes all the thermodynamic properties cients hi and si are independent of tempera-
and the phase diagram. This technique of ture.
analysis greatly reduces the amount of ex- If the series are truncated after the first
perimental data needed to fully character- term, then:
ize a system. All data can be tested for
g E = hE – T s E = XA XB (h0 – T s0 ) (1-91)
internal consistency. The data can be inter-
polated and extrapolated more accurately This is the equation for a regular solution
and metastable phase boundaries can be discussed in Sec. 1.5.7. Hence, the polyno-
calculated. All the thermodynamic proper- mial representation can be considered to be
ties and the phase diagram can be repre- an extension of regular solution theory.
sented and stored by means of a small set When the expansions are written in terms
of coefficients. of the composition variable (XB – XA ), as in
Finally, and most importantly, it is often Eqs. (1-89) and (1-90), they are said to be
possible to estimate the thermodynamic in Redlich – Kister form. Other equivalent
properties and phase diagrams of ternary polynomial expansions such as orthogonal
and higher-order systems from the assessed Legendre series have been discussed by
parameters for their binary sub-systems, as Pelton and Bale (1986).
will be discussed in Sec. 1.11. The analysis Differentiation of Eqs. (1-89) and (1-90)
of binary systems is thus the first and most and substitution into Eq. (1-55) yields the
1.6 Application of Thermodynamics to Phase Diagram Analysis 35
following expansions for the partial excess

enthalpies and entropies:
hAE = XB2 S
i=0
hi [(XB – XA )i
– 2 i XA (XB – XA )i – 1] (1-92)
hEB = XA2 S
i=0
hi [(XB – XA )i
+ 2 i XB (XB – XA )i – 1] (1-93)
sAE = XB2 S
i=0
si [(XB – XA )i
– 2 i XA (XB – XA )i – 1] (1-94)
sBE = XA2 S
i=0
si [(XB – XA )i
Figure 1-18. LiF – NaF phase diagram at P = 1 bar
+ 2 i XB (XB – XA )i – 1] (1-95) calculated from optimized thermodynamic parame-
ters (Sangster and Pelton, 1987). Points are experi-
Partial excess Gibbs energies, g Ei , are mental from Holm (1965). Dashed line is theoretical
then given by Eq. (1-52). limiting liquidus slope for negligible solid solubility.
Eqs. (1-89) and (1-90), being based upon
regular solution theory, give an adequate
representation for most simple substitu- each pure component at temperature T is
tional solutions in which deviations from given by:
ideal behavior are not too great. In other
cases, more sophisticated models are re- Dg f0 = Dhf0( Tf ) (1 − T / Tf )
quired, as discussed in Sec. 1.10. T
+ ∫ (cpl − cps ) (1 − 1/ T ) dT (1-96 )
Tf
1.6.3 Least-Squares Optimization
where Dh0f(Tf ) is the enthalpy of fusion at
Eqs. (1-89), (1-90) and (1-92) to (1-95) the melting point Tf , and cpl and c ps are the
are linear in terms of the coefficients. heat capacities of the pure liquid and solid.
Through the use of these equations, all The following values are taken from Barin
integral and partial excess properties (gE , et al. (1977):
hE , s E , g Ei , hEi , sEi ) can be expressed by
linear equations in terms of the one set of Dg f0(LiF) = 14.518 + 128.435 T
coefficients {hi , si }. It is thus possible to + 8.709 ¥ 10– 3 T 2 – 21.494 T ln T
include all available experimental data for – 2.65 ¥ 105 T –1 J/mol (1-97)
a binary phase in one simultaneous linear
Dg f0(NaF) = 10.847 + 156.584 T
least-squares optimization. Details have
been discussed by Bale and Pelton (1983), + 4.950 ¥ 10– 3 T 2 – 23.978 T ln T
Lukas et al. (1977) and Dörner et al. – 1.07 ¥ 105 T –1 J/mol (1-98)
(1980). Thermodynamic properties along the liq-
The technique of coupled thermody- uidus and solidus are related by equations
namic/phase diagram analysis is best illus- like Eqs. (1-64) and (1-65). Taking the
trated by examples. ideal activities to be equal to the mole frac-
The phase diagram of the LiF – NaF tions:
system is shown in Fig. 1-18. Data points
measured by Holm (1965) are shown on R T ln Xil – R T ln Xis + giE (l) – giE (s)
the diagram. The Gibbs energy of fusion of = – Dg f0(i) (1-99)
where i = LiF or NaF. Along the LiF-rich aLiF = 1 with respect to pure solid LiF as
liquidus, the liquid is in equilibrium with standard state. In the Henrian solution at
s
essentially pure solid LiF. Hence, X LiF =1 saturation,
E (s)
and gLif = 0. Eq. (1-99) then reduces to:
aLiF = g 0LiF XLiF = g 0LiF (1 – 0.915) = 1
l E (l) 0
R T ln X LiF +g Lif = – Dg f (LiF) (1-100)
Hence, the Henrian activity coefficient
l
From experimental values of X LiF on the in the NaF-rich solid solution at 649 °C
liquidus and with Eq. (1-97) for Dg f0(LiF) , is g 0LiF = 11.76. Since no solubilities have
E (l)
values of gLif at the measured liquidus been measured at other temperatures, we
points can be calculated from Eq. (1-100). assume that:
Along the NaF-rich solidus the solid so-
lution is sufficiently concentrated in NaF R T ln g 0LiF = R (922) ln (11.76) (1-104)
that Henrian behavior (Sec. 1.5.11) can be = 18 900 J/mol = constant
E (s)
assumed. That is, for the solvent, gNaF = 0. Using the notation of Eq. (1-82):
Hence, Eq. (1-99) becomes:
(l) (s) E (l)
g 0LiF
(s, NaF)
= g 0LiF
(s)
+ 18 900 J/mol (1-105)
R T ln X NaF – R T ln X NaF + gNaF
= – Dg f0(NaF) (1-101) where g 0LiF
(s)
is the standard Gibbs energy of
solid LiF, and g 0LiF (s, NaF)
is the hypothetical
Thus, from the experimental liquidus and standard Gibbs energy of LiF dissolved in
solidus compositions and with the Gibbs solid NaF.
energy of fusion from Eq. (1-98), values of The phase diagram drawn in Fig. 1-18
E (l)
gNaF can be calculated at the measured liq- was calculated from Eqs. (1-97) to (1-104).
uidus points from Eq. (1-101). Complete details of the analysis of the
Finally, enthalpies of mixing, hE , in the LiF – NaF system are given by Sangster and
liquid have been measured by calorimetry Pelton (1987).
by Hong and Kleppa (1976). As a second example of thermodynamic/
Combining all these data in a least- phase diagram optimization, consider the
squares optimization, the following expres- Cd – Na system. The phase diagram, with
sions for the liquid were obtained by Sang- points measured by several authors (Math-
ster and Pelton (1987): ewson, 1906; Kurnakow and Kusnetzow,
hE (l) = XLiF XNaF (1-102) 1907; Weeks and Davies, 1964) is shown in
Fig. 1-19.
¥ [– 7381 + 184 (XNaF – XLiF )] J/mol From electromotive force measurements
s E (l) = XLiF XNaF (1-103) on alloy concentration cells, several au-
¥ [– 2.169 – 0.562 (XNaF – XLiF )] J/mol thors have measured the activity coeffi-
cient of Na in liquid alloys. The data
Eqs. (1-102) and (1-103) then permit all are shown in Fig. 1-20 at 400 °C. From
other integral and partial properties of the the temperature dependence of g ENa =
liquid to be calculated. R T ln g Na , the partial enthalpy of Na in the
For the NaF-rich Henrian solid solution, liquid was obtained via Eq. (1-52). The re-
the solubility of LiF has been measured by sults are shown in Fig. 1-21. Also, hE of the
Holm (1965) at the eutectic temperature liquid has been measured by Kleinstuber
where the NaF-rich solid solution is in (1961) by direct calorimetry. These ther-
equilibrium with pure solid LiF. That is, modynamic data for g ENa , hENa and hE were
1.6 Application of Thermodynamics to Phase Diagram Analysis 37
Figure 1-19. Cd – Na phase diagram at P = 1 bar calculated from optimized thermodynamic parameters (Re-
printed from Pelton, 1988 a). 쑗 Kurnakow and Kusnetzow (1907), 왕 Mathewson (1906), × Weeks and Davies
(1964).
Figure 1-20. Sodium ac-

tivity coefficient in liquid
Cd – Na alloys at 400°C.
Line is calculated from
optimized thermodynamic
parameters (Reprinted
from Pelton, 1988a).
쏔 Hauffe (1940),
쎲 Lantratov and
Mikhailova (1971),
왕 Maiorova et al. (1976),
왔 Alabyshev and
Morachevskii (1957),
쑗 Bartlett et al. (1970).
Eq. (1-106) reproduces the calorimetric

data within 200 J/mol –1. Eqs. (1-52), (1-
58), (1-93) and (1-95) can be used to calcu-
late hENa and g Na . The calculated curves are
compared to the measured points in Figs.
1-20 and 1-21.
For the two compounds, Gibbs energies
of fusion were calculated (Pelton, 1988 a)
so as to best reproduce the measured phase
diagram:
Dg 0f (1/13 Cd11Na2 ) = 6816 – 10.724 T J/g-atom
(1-108)
Dg 0f (1/3 Cd2Na) = 8368 – 12.737 T J/g-atom
(1-109)
The optimized enthalpies of fusion of 6816
and 8368 J/g-atom agree within error lim-
its with the values of 6987 and 7878 J/g-
atom measured by Roos (1916). (See Fig.
1-14 for an illustration of the relation
between the Gibbs energy of fusion of a
compound and the phase diagram.)
The phase diagram shown in Fig. 1-19
was calculated from Eqs. (1-106) to (1-
109) along with the Gibbs energies of fu-
sion of Cd and Na taken from the literature
(Chase, 1983). Complete details of the
analysis of the Cd – Na system are given by
Pelton (1988 a).
Figure 1-21. Partial excess enthalpy of sodium in
It can thus be seen that one simple set of
liquid Cd –Na alloys. Line is calculated from opti-
mized thermodynamic parameters (Reprinted from equations can simultaneously and self-con-
Pelton, 1988a). 쎲 Lantratov and Mikhailova (1971), sistently describe all the thermodynamic
왕 Maiorova et al. (1976), 쑗 Bartlett et al. (1970). properties and the phase diagram of a bi-
nary system.
The exact optimization procedure will
optimized simultaneously (Pelton, 1988 a) vary from system to system depending
to obtain the following expressions for hE upon the type and accuracy of the avail-
and s E of the liquid: able data, the number of phases present, the
extent of solid solubility, etc. A large num-
hE (l) = XCd XNa [– 12 508 + 20 316 (1-106)
ber of optimizations have been published
¥ (XNa – XCd ) – 8714 (XNa – XCd )2 ] J/mol in the Calphad Journal (Pergamon) since
s E (l) = XCd XNa [– 15.452 + 15.186 (1-107) 1977.
¥ (XNa – XCd ) – 10.062 (XNa – XCd )2
– 1.122 (XNa – XCd )3 ] J/mol K
1.7 Ternary and Multicomponent Phase Diagrams 39
1.6.4 Calculation of Metastable Phase and Cd – Bi are found. Lines of constant

Boundaries mole fraction XBi are parallel to the Sn – Cd
edge, while lines of constant XSn and XCd
In the Cd – Na system just discussed, the
are parallel to the Cd – Bi and Bi – Sn edges
liquid exhibits positive deviations from
respectively. For example, at point a in Fig.
ideal mixing. That is, gE (l) > 0. This fact is
1-22, XBi = 0.05, XSn = 0.45 and XCd = 0.50.
reflected in the very flat liquidus in Fig.
Similar equilateral composition triangles
1-19 as was discussed in Sec. 1.5.9.
can be drawn with coordinates in terms of
By simply not including any solid phases
wt.% of the three components.
in the calculation, the metastable liquid
miscibility gap as well as the spinodal
curve (Sec. 1.5.5) can be calculated as 1.7.2 Ternary Space Model
shown in Fig. 1-19. These curves are im-
A ternary temperature – composition
portant in the formation of metallic glasses
“phase diagram” at constant total pressure
by rapid quenching.
may be plotted as a three-dimensional
Other metastable phase boundaries such
“space model” within a right triangular
as the extension of a liquidus curve below a
prism with the equilateral composition tri-
eutectic can also be calculated thermody-
angle as base and temperature as vertical
namically by simply excluding one or more
axis. Such a space model for a simple eu-
phases during the computations.
tectic ternary system A – B – C is illustrated
in Fig. 1-23. On the three vertical faces of
the prism we find the phase diagrams of the
1.7 Ternary and Multicomponent three binary subsystems, A – B, B – C and
Phase Diagrams C – A which, in this example, are all simple
eutectic binary systems. The binary eutec-
This section provides an introduction to tic points are e1 , e2 and e3 . Within the
ternary phase diagrams. For a more de- prism we see three liquidus surfaces de-
tailed treatment, see Prince (1966); Ricci scending from the melting points of pure
(1964); Findlay (1951); or West (1965). A, B and C. Compositions on these sur-
faces correspond to compositions of liquid
in equilibrium with A-, B- and C-rich solid
1.7.1 The Ternary Composition Triangle
phases.
In a ternary system with components In a ternary system at constant pressure,
A – B – C, the sum of the mole fractions is the Gibbs phase rule, Eq. (1-85), becomes:
unity, (XA + XB + XC ) = 1. Hence, there are
F=4–P (1-110)
two independent composition variables. A
representation of composition, symmetri- When the liquid and one solid phase are in
cal with respect to all three components, equilibrium P = 2. Hence F = 2 and the
may be obtained with the equilateral “com- system is bivariant. A ternary liquidus is
position triangle” as shown in Fig. 1-22 for thus a two-dimensional surface. We may
the Bi – Sn – Cd system. Compositions at choose two variables, say T and one
the corners of the triangle correspond to the composition coordinate of the liquid, but
pure components. Along the edges of the then the other liquid composition coordi-
triangle compositions corresponding to the nate and the composition of the solid are
three binary subsystems Bi – Sn, Sn – Cd fixed.
Figure 1-22. Projection of the liquidus surface of the Bi – Sn –Cd system onto the ternary composition triangle
(after Bray et al., 1961 – 1962). Small arrows show the crystallization path of an alloy of overall composition at
point a. (Reprinted from Pelton, 1996.)
Figure 1-23. Perspective view of ternary space

model of a simple eutectic ternary system. e1, e2 , e3
are the binary eutectics and E is the ternary eutectic.
The base of the prism is the equilateral composition
triangle. (Reprinted from Pelton, 1983.)
The A- and B-liquidus surfaces in Fig. 1- pure Cd. (Note that this fact cannot be de-
23 intersect along the line e1 E. Liquids duced from Fig. 1-22 alone.) Therefore, as
with compositions along this line are there- solidification proceeds, the liquid becomes
fore in equilibrium with A-rich and B-rich depleted in Cd, but the ratio XSn /XBi in the
solid phases simultaneously. That is, P = 3 liquid remains constant. Hence, the compo-
and so F – 1. Such “valleys” are thus called sition path followed by the liquid (its crys-
univariant lines. The three univariant lines tallization path) is a straight line passing
meet at the ternary eutectic point E at through point a and projecting to the Cd-
which P = 4 and F = 0. This is an invariant corner of the triangle. This crystallization
point since the temperature and the compo- path is shown on Fig. 1-22 as the line a b.
sitions of all four phases in equilibrium are In the general case in which a solid solu-
fixed. tion rather than a pure component or stoi-
chiometric compound is precipitating, the
crystallization path will not be a straight
1.7.3 Polythermal Projections
line. However, for equilibrium cooling, a
of Liquidus Surfaces
straight line joining a point on the crystal-
A two-dimensional representation of the lization path at any T to the overall compo-
ternary liquidus surface may be obtained as sition point a will extend through the com-
an orthogonal projection upon the base position, on the solidus surface, of the solid
composition triangle. Such a polythermal phase in equilibrium with the liquid at that
projection of the liquidus of the Bi – Sn – Cd temperature.
system (Bray et al., 1961 – 62) is shown in When the composition of the liquid has
Fig. 1-22. This is a simple eutectic ternary reached point b in Fig. 1-22 at T ≈ 435 K,
system with a space model like that shown the relative proportions of the solid Cd and
in Fig. 1-23. The constant temperature liquid phases at equilibrium are given by
lines on Fig. 1-22 are called liquidus iso- the lever rule applied to the tie-line d a b:
therms. The univariant valleys are shown (moles of liquid)/(moles of Cd) = d a /a b.
as heavier lines. By convention, the large Upon further cooling, the liquid composi-
arrows indicate the directions of decreas- tion follows the univariant valley from b to
ing temperature along these lines. E while Cd and Sn-rich solids coprecipitate
Let us consider the sequence of events as a binary eutectic mixture. When the
occurring during the equilibrium cooling liquidus composition attains the ternary eu-
from the liquid of an alloy of overall com- tectic composition E at T ≈ 380 K the invar-
position a in Fig. 1-22. Point a lies within iant ternary eutectic reaction occurs:
the field of primary crystallization of Cd.
liquid Æ s1 + s2 + s3 (1-111)
That is, it lies within the composition re-
gion in Fig. 1-22 in which Cd-rich solid where s1 , s2 and s3 are the three solid
will be the first solid to precipitate upon phases and where the compositions of all
cooling. As the liquid alloy is cooled, the four phases (as well as T ) remain fixed un-
Cd-liquidus surface is reached at T ≈ 465 K til all liquid is solidified.
(slightly below the 473 K isotherm). A In order to illustrate several of the fea-
solid Cd-rich phase begins to precipitate at tures of polythermal projections of liquidus
this temperature. Now, in this particular surfaces, a projection of the liquidus of a
system, Bi and Sn are nearly insoluble in hypothetical system A – B – C is shown in
solid Cd, so that the solid phase is virtually Fig. 1-24. For the sake of simplicity, iso-
Figure 1-24. Projection of

the liquidus surface of a
system A – B – C. The binary
subsystems A – B and C – A
are simple eutectic systems.
The binary phase diagram
B – C is shown in the insert.
All solid phases are assumed
pure stoichiometric compo-
nents or compounds. Small
arrows show crystallization
paths of alloys of composi-
tions at points a and b. (Re-
printed from Pelton, 1983.)
therms are not shown, only the univariant field, since e is a congruent compound,
lines with arrows to show the directions of while the composition of the d phase lies
decreasing temperature. The binary sub- outside of its field since d is incongruent.
systems A – B and C – A are simple eutectic Similarly for the ternary compounds, h is a
systems, while the binary subsystem B – C congruently melting compound while z is
contains one congruent binary phase, e, incongruent. For the congruent compound
and one incongruent binary phase, d, as h, the highest temperature on the h liquidus
shown in the insert in Fig. 1-24. The letters occurs at the composition of h.
e and p indicate binary eutectic and peritec- The univariant lines meet at a number of
tic points. The e and d phases are called bi- ternary eutectics E (three arrows converg-
nary compounds since they have composi- ing), a ternary peritectic P (one arrow en-
tions within a binary subsystem. Two ter- tering, two arrows leaving the point), and
nary compounds, h and z, with composi- several ternary quasi-peritectics P¢ (two
tions within the ternary triangle, as indi- arrows entering, one arrow leaving). Two
cated in Fig. 1-24, are also found in this saddle points s are also shown. These are
system. All compounds, as well as pure points of maximum T along the univariant
solid A, B and C (the “a, b and g” phases), line but of minimum T on the liquidus sur-
are assumed to be stoichiometric (i.e., there face along a section joining the composi-
is no solid solubility). The fields of pri- tions of the two solids. For example, s1 is at
mary crystallization of all the solids are in- a maximum along the univariant E1 P¢3 , but
dicated in parentheses in Fig. 1-24. The is a minimum point on the liquidus along
composition of the e phase lies within its the straight line z s1 h.
Let us consider the events occurring dur- univariant valley until the liquid reaches
ing cooling from the liquid of an alloy of the peritectic composition P. The invariant
overall composition a in Fig. 1-24. The pri- ternary peritectic reaction then occurs iso-
mary crystallization product will be the e thermally:
phase. Since this is a pure stoichiometric
liquid + e + b Æ z (1-113)
solid the crystallization path of the liquid
will be along a straight line passing Since there are three reactants, there are
through a and extending to the composition three possible outcomes: (i) the liquid is
of e as shown in the figure. consumed before either e or b and solidifi-
Solidification of e continues until the cation terminates at P; (ii) e is consumed
liquid attains a composition on the univari- first, solidification then continues along
ant valley. Thereafter the liquid composi- the path P P¢3 ; or (iii) b is consumed first
tion follows the valley towards the point P¢1 and solidification continues along the path
in co-existence with e and z. At point P¢1 P P¢1 . Which outcome occurs depends on
the invariant ternary quasi-peritectic reac- whether the overall composition b lies
tion occurs isothermally: within the tie-triangle (i) e b z; (ii) b z P, or
(iii) e z P. In the example shown, outcome
liquid + e Æ d + z (1-112)
(i) will occur.
Since there are two reactants in a quasi-
peritectic reaction, there are two possible
1.7.4 Ternary Isothermal Sections
outcomes: (i) the liquid is completely con-
sumed before the e phase; in this case, so- Isothermal projections of the liquidus
lidification will be complete at the point P¢1 ; surface do not provide information on the
(ii) e is completely consumed before the compositions of the solid phases at equilib-
liquid. In this case, solidification will con- rium. However, this information can be
tinue with decreasing T along the univari- presented at any one temperature on an iso-
ant line P¢1 E1 with co-precipitation of d and thermal section such as that shown for the
z until, at E, the liquid will solidify eutecti- Bi – Sn – Cd system at 423 K in Fig. 1-25.
cally (liquid Æ d + z + h). To determine This phase diagram is a constant tempera-
whether outcome (i) or (ii) occurs, we use ture slice through the space model of Fig.
the mass balance criterion that, for three- 1-23.
phase equilibrium, the overall composition The liquidus lines bordering the one-
a must always lie within the tie-triangle phase liquid region of Fig. 1-25 are identi-
formed by the compositions of the three cal to the 423 K isotherms of the projection
phases. Now, the triangle joining the com- in Fig. 1-22. Point c in Fig. 1-25 is point c
positions of d, e and z does not contain the on the univariant line in Fig. 1-22. An alloy
point a, but the triangle joining the compo- with overall composition in the one-phase
sitions of d, z and liquid at P¢1 does contain liquid region of Fig. 1-25 at 423 K will
the point a. Hence, outcome (ii) occurs. consist of a single liquid phase. If the over-
An alloy of overall composition b in Fig. all composition lies within one of the two-
1-24 solidifies with e as primary crystal- phase regions, then the compositions of the
lization product until the liquid composi- two phases are given by the ends of the tie-
tion contacts the univariant line. There- line which passes through the overall com-
after, co-precipitation of e and b occurs position. For example, a sample with over-
with the liquid composition following the all composition p in Fig. 1-25 will consist
Figure 1-25. Isothermal section of the Bi – Sn – Cd system at 423 K at P = 1 bar (after Bray et al., 1961–1962).
Extents of solid solubility in Bi and Sn have been exaggerated for clarity of presentation. (Reprinted from Pel-
ton, 1996.)
of a liquid of composition q on the liquidus is somewhat exaggerated in Fig. 1-25 for

and a solid Bi-rich alloy of composition r the sake of clarity of presentation.) Alloys
on the solidus. The relative proportions of with overall compositions rich enough in
the two phases are given by the lever rule: Bi or Sn to lie within the single-phase (Sn)
(moles of liquid)/(moles of solid) = p r /p q, or (Bi) regions of Fig. 1-25 will consist at
where p r and p q are the lengths of the tie- 423 K of single-phase solid solutions. Al-
line segments. loys with overall compositions at 423 K in
In the case of solid Cd, the solid phase is the two-phase (Cd + Sn) region will consist
nearly pure Cd, so all tie-lines of the (Cd + of two solid phases.
liquid) region converge nearly to the corner Alloys with overall compositions within
of the triangle. In the case of Bi- and Sn- the three-phase triangle d c f will, at 423 K,
rich solids, some solid solubility is ob- consist of three phases: solid Cd- and Sn-
served. (The actual extent of this solubility rich solids with compositions at d and f and
liquid of composition c. To understand this

better, consider an alloy of composition a
in Fig. 1-25, which is the same composi-
tion as the point a in Fig. 1-22. In Sec.
1.7.3 we saw that when an alloy of this
composition is cooled, the liquid follows
the path a b in Fig. 1-22 with primary pre-
cipitation of Cd and then follows the uni-
variant line with co-precipitation of Cd and
Sn so that at 423 K the liquid is at the com-
position point c, and two solid phases are in
equilibrium with the liquid.
Figure 1-26. A tie-triangle in a ternary isothermal
section illustrating the lever rule and the extension
1.7.4.1 Topology of Ternary Isothermal rule.
Sections
At constant temperature the Gibbs en-
ergy of each phase in a ternary system is single-phase region, and from each edge of
represented as a function of composition the triangle there extends a two-phase re-
by a surface plotted in a right triangular gion. The edge of the triangle is a limiting
prism with Gibbs energy as vertical axis tie-line of the two-phase region.
and the composition triangle as base. Just For overall compositions within the tie-
as the compositions of phases at equilib- triangle, the compositions of the three
rium in binary systems are determined phases at equilibrium are fixed at the cor-
by the points of contact of a common tan- ners of the triangle. The relative propor-
gent line to their isothermal Gibbs energy tions of the three phases are given by the
curves, so the compositions of phases at lever rule of tie-triangles, which can be de-
equilibrium in a ternary system are given rived from mass balance considerations. At
by the points of contact of a common tan- an overall composition q in Fig. 1-26 for
gent plane to their isothermal Gibbs energy example, the relative proportion of the g
surfaces. A common tangent plane can phase is given by projecting a straight line
contact two Gibbs energy surfaces at an in- from the g corner of the triangle (point c)
finite number of pairs of points, thereby through the overall composition q to the
generating an infinite number of tie-lines opposite side of the triangle, point p. Then:
within a two-phase region on an isothermal (moles of g)/(total moles) = q p /c p if com-
section. A common tangent plane to three positions are expressed in mole fractions,
Gibbs energy surfaces contacts each sur- or (weight of g)/(total weight) = q p /c p if
face at a unique point, thereby generating a compositions are in weight percent.
three-phase tie-triangle. Isothermal ternary phase diagrams are
Hence, the principal topological units of generally composed of a number of these
construction of an isothermal ternary phase topological units. An example for the Al –
diagram are three-phase (a + b + g) tie-tri- Zn – Mg system at 25 °C is shown in Fig.
angles as in Fig. 1-26 with their accompa- 1-27 (Köster and Dullenkopf, 1936). The
nying two-phase and single-phase areas. b, g, d, q, h and z phases are binary inter-
Each corner of the tie-triangle contacts a metallic compounds with small (~ 1 to 6%)
Another important rule of construction,

whose derivation is evident, is that within
any two-phase region tie-lines must never
cross one another.
1.7.5 Ternary Isopleths

(Constant Composition Sections)
A vertical isopleth, or constant composi-
tion section through the space model of the
Bi – Sn – Cd system, is shown in Fig. 1-28.
The section follows the line AB in Fig.
1-22.
The phase fields on Fig. 1-28 indicate
which phases are present when an alloy
Figure 1-27. Ternary isothermal section of the Al – with an overall composition on the line AB
Zn – Mg system at 25°C at P = 1 bar (after Köster and is equilibrated at any temperature. For ex-
Dullenkopf, 1936). (Reprinted from Pelton, 1983.) ample, consider the cooling, from the liq-
uid state, of an alloy of composition a
which is on the line AB (see Fig. 1-22). At
ranges of stoichiometry which can dissolve T ≈ 465 K, precipitation of the solid (Cd)
a limited amount (~ 1 to 6%) of the third phase begins at point a in Fig. 1-28. At
component. The t phase is a ternary phase T ≈ 435 K (point b in Figs. 1-22 and 1-28)
with a single-phase region existing over a the solid (Sn) phase begins to appear. Fi-
fairly extensive oval-shaped central com- nally, at the eutectic temperature TE , the
position range. Examination of Fig. 1-27 ternary reaction occurs, leaving solid (Cd)
shows that it consists of the topological + (Bi) + (Sn) at lower temperatures. The
units of Fig. 1-26. intersection of the isopleth with the univar-
An extension rule, a case of Schreine- iant lines on Fig. 1-22 occurs at points f and
makers’ Rule (Schreinemakers, 1915), see g which are also indicated on Fig. 1-28.
Sec. 1.7.5, for ternary tie-triangles is illus- The intersection of this isopleth with the
trated in Fig. 1-26. At each corner, the isothermal section at 423 K is shown in
extension of the boundaries of the single- Fig. 1-25. The points s, t, u and v of Fig.
phase regions, indicated by the broken 1-25 are also shown on Fig. 1-28.
lines, must either both project into the tri- It is important to note that on an isopleth
angle as at point a, or must both project the tie-lines do not, in general, lie in the
outside the triangle as at point b. Further- plane of the diagram. Therefore, the dia-
more, the angle between these extensions gram provides information only on which
must be less than 180°. For a proof, see phases are present, not on their composi-
Lipson and Wilson (1940) or Pelton tions. The boundary lines on an isopleth do
(1995). not in general indicate the phase composi-
Many published phase diagrams violate tions, only the temperature at which a
this rule. For example, it is violated in Fig. phase appears or disappears for a given
1-27 at the d-corner of the (e + d + t) tie-tri- overall composition. The lever rule cannot
angle. be applied on an isopleth.
Figure 1-28. Isopleth (constant composition section) of the Bi – Sn – Cd system at P = 1 bar following the line
AB at XSn = 0.45 of Fig. 1-22. (Reprinted from Pelton, 1996).
Certain geometrical rules apply to iso- 1.7.5.1 Quasi-Binary Phase Diagrams

pleths. As a phase boundary line is crossed,
one and only one phase either appears or Several of the binary phase diagrams
disappears. This Law of Adjoining Phase in the preceding sections (Figs. 1-5, 1-10,
Regions (Palatnik and Landau, 1964) is il- 1-12, 1-14, 1-18) are actually isopleths
lustrated by Fig. 1-28. The only apparent of ternary systems. For example, Fig. 1-12
exception occurs for the horizontal invari- is an isopleth at constant XO = nO /(nMg +
ant line at TE . However, if we consider this nCa + nO ) = 0.5 of the Mg – Ca – O system.
line to be a degenerate infinitely narrow However, all tie-lines lie within (or virtu-
four-phase region (L + (Cd) + (Bi) + (Sn)), ally within) the plane of the diagram be-
then the law is also obeyed here. cause XO = 0.5 in every phase. Therefore,
Three or four boundary lines meet at in- the diagram is called a quasi-binary phase
tersection points. At an intersection point, diagram.
such as point f or g, Schreinemakers’ Rule
applies. This is discussed in Sec. 1.9. 1.7.6 Multicomponent Phase Diagrams
Apparent exceptions to these rules (such
as, for example, five boundaries meeting at Only an introduction to multicomponent
an intersection point) can occur if the sec- phase diagrams will be presented here.
tion passes exactly through a node (such For more detailed treatments see Palatnik
as a ternary eutectic point) of the space and Landau (1964), Prince (1963), Prince
model. However, these apparent exceptions (1966) and Hillert (1998).
are really only limiting cases (see Prince, For systems of four or more components,
1963 or 1966). two-dimensional sections are usually plot-
ted with one or more compositional vari- at equilibrium for an overall composition
ables held constant. Hence these sections lying on the section. Tie-lines do not, in
are similar to the ternary isopleths dis- general, lie in the plane of the diagram, so
cussed in Sec. 1.7.5. In certain cases, sec- the diagram does not provide information
tions at constant chemical potential of one on the compositions or amounts of the
or more components (for example, at con- phases present. As a phase boundary is
stant oxygen partial pressure) can be use- crossed, one and only one phase appears or
ful. These are discussed in Sec. 1.8. disappears (Law of Adjoining Phase Re-
Two sections of the Fe – Cr –V– C system gions). If temperature is an axis, as in Fig.
(Lee and Lee, 1992) are shown in Figs. 1- 1-29, then horizontal invariants like the
29 and 1-30. The diagram in Fig. 1-29 is a line AB in Fig. 1-29 can appear. These can
T-composition section at constant Cr and V be considered as degenerate infinitely nar-
content, while Fig. 1-30 is a section at con- row phase fields of (C + 1) phases, where C
stant T = 850 oC and constant C content of is the number of components (for isobaric
0.3 wt.%. The interpretation and topologi- diagrams). For example in Fig. 1-29, on the
cal rules of construction of these sections line AB, five phases are present. Three or
are the same as those for ternary isopleths, four phase boundaries meet at intersection
as discussed in Sec. 1.7.5. In fact, the same points at which Schreinemakers’ Rule ap-
rules apply to a two-dimensional constant- plies. It is illustrated by the extrapolations
composition section for a system of any in Fig. 1-29 at points a, b and c and in Fig.
number of components. The phase fields 1-30 at points b, c, n, i and s (see discus-
on the diagram indicate the phases present sions in Sec. 1.9).
Figure 1-29. Section of the Fe – Cr – V – C system at 1.5 wt.% Cr and 0.1 wt.% V at P = 1 bar (Lee and Lee,
1992).
Figure 1-30. Section of

the Fe – Cr – V – C system
at 850 oC and 0.3 wt.% C
at P = 1 bar (Lee and Lee,
1992).
1.7.7 Nomenclature for Invariant + d) is a 1 Æ 3 reaction; and so on. The ter-

Reactions nary peritectic-type 3 Æ 1 reaction (a + b
+ g Æ d) is an invariant reaction in a ter-
As discussed in Sec. 1.5.12, in a binary nary system, a univariant reaction in a qua-
isobaric temperature – composition phase ternary system, a bivariant reaction in a
diagram there are two possible types of quinary system, etc.
invariant reactions: “eutectic-type” (b Æ a
+ g), and “peritectic type” (a + g Æ b). In
1.7.8 Reciprocal Ternary Phase
a ternary system, there are “eutectic-type”
Diagrams
(a Æ b + g + d), “peritectic-type” (a + b +
g Æ d), and “quasiperitectic-type” (a + b A reciprocal ternary salt system is one
Æ g + d) invariants (Sec. 1.7.3). In a consisting of two cations and two anions,
system of C components, the number of such as the Na+, K+/ F –, Cl – system of Fig.
types of invariant reaction is equal to C. A 1-31. The condition of charge neutrality
reaction with one reactant, such as (nNa+ + nK+ = nF – + nCl – ) removes one degree
(a Æ b + g + d + e) is clearly a “eutectic- of freedom. The system is thus quasiter-
type” invariant reaction but in general there nary and its composition can be repre-
is no standard terminology. These reactions sented by two variables, usually chosen
are conveniently described according to the as the cationic mole fraction XK = nK /
numbers of reactants and products (in the (nNa + nK ) and the anionic mole fraction
direction which occurs upon cooling). XCl = nCl /(nF + nCl ), where ni = number of
Hence the reaction (a + b Æ g + d + e) is a moles of ion i. Note that XNa = (1 – XK ) and
2 Æ 3 reaction; the reaction (a Æ b + g XF = (1 – XCl ).
ternary but is an isopleth of the four-com-

ponent Na – K – F – Cl system, and tie-lines
no longer necessarily lie in the plane of the
diagram.
In Fig. 1-31 the cationic and anionic
fractions are plotted as axes of a square.
Compositions corresponding to the four
neutral salts (KF, KCl, NaCl, NaF) are
found at the corners of the square. Edges of
the square correspond to the binary sub-
systems such as NaF – NaCl. A ternary
space model (analogous to Fig. 1-23) can
be constructed with temperature as vertical
axis. The phase diagram of Fig. 1-31 is a
polythermal projection of the liquidus sur-
face upon the composition square.
In this system, three of the binary edges
are simple eutectic systems, while the
NaCl – KCl binary system exhibits a sol-
idus/liquidus minimum. There is a ternary
eutectic at 570 °C in Fig. 1-31 (b). The
NaF – KCl diagonal contains a saddle point
at 648 °C in Fig. 1-31 (b). This saddle point
is a eutectic of the quasibinary system
NaF – KCl. That is, a binary phase diagram
NaF – KCl could be drawn with one simple
eutectic at 648 °C. However, the NaCl – KF
system, which forms the other diagonal, is
not a quasibinary system. If compositions
lying on this diagonal are cooled at equilib-
rium from the liquid, solid phases whose
compositions do not lie on this diagonal
can precipitate. Hence, a simple binary
phase diagram cannot be drawn for the
Figure 1-31. Projection of the liquidus surface of
the Na+, K+/F –, Cl – reciprocal ternary system.
NaCl – KF system.
a) Calculated from optimized binary thermodynamic For systems such as Ca2+, Na+/ F –, SO42–
parameters. in which the ions do not all have the same
b) As reported by Polyakov (1940). charge, composition axes are conveniently
expressed as equivalent ionic fractions
(e.g. YCa = 2 nCa /(2 nCa + nNa )), see Sec.
The assumption has, of course, been 1.9.2.1.
made that the condition (nNa + nK = nF + nCl ) The concept of reciprocal systems can be
holds exactly in every phase. If there is a generalized beyond simple salt systems
deviation from this exact stoichiometry, and is closely related to the sublattice
then the phase diagram is no longer quasi- model (Sec. 1.10.1).
1.8 Phase Diagrams with Potentials as Axes 51
For further discussion and references,

see Pelton (1988 b) and Blander (1964).
1.8 Phase Diagrams

with Potentials as Axes
So far we have considered mainly iso-
baric temperature – composition phase dia-
grams. However there are many other
kinds of phase diagrams of interest in ma-
terials science and technology with pres-
sure, chemical potentials, volume, etc. as
axes. These can be classified into geomet-
rical types according to their rules of con-
struction.
For instance, binary isothermal P – X di-
agrams as in Fig. 1-8 are members of the
same type as binary isobaric T – X diagrams
because they are both formed from the
same topological units of construction.
Other useful phase diagrams of this same
geometrical type are isothermal chemical Figure 1-32. Corresponding phase diagrams for the
potential – composition diagrams for ter- Co – Ni – O system at 1600 K (from Pelton and
nary systems. An example is shown in Thompson, 1973).
the lowest panel of Fig. 1-32 (Pelton and
Thompson, 1975) for the Co – Ni – O
system at T = 1600 K (and at a constant to- The usual isothermal section of the ter-
tal hydrostatic pressure of 1 bar). Here the nary Co – Ni – O system at 1600 K is shown
logarithm of the equilibrium partial pres- in the top panel of Fig. 1-32. There are two
sure of O2 is plotted versus the metal ratio single-phase regions with a two-phase re-
x = nNi /(nCo + nNi ), where ni = number of gion between them. The single-phase areas
moles of i. There are two phases in this are very narrow because oxygen is only
system under these conditions, a solid alloy very slightly soluble in the solid alloy
solution stable at lower pO2 , and a solid so- and CoO and NiO are very stoichiometric
lution of CoO and NiO stable at higher pO2 . oxides. In the central panel of Fig. 1-32
For instance, point a gives pO2 for the equi- this same diagram is shown with the com-
librium between pure Co and pure CoO at position triangle “opened up” by putting
1600 K. Between the two single-phase re- the oxygen corner at infinity. This can be
gions is a two-phase (alloy + oxide) region. done if the vertical axis becomes h = nO /
At any overall composition on the tie-line (nCo + nNi ) with the horizontal axis as
c d between points c and d, two phases will x = nNi /(nCo + nNi ). These are known as
be observed, an alloy of composition d and Jänecke coordinates. It can be seen in Fig.
an oxide of composition c. The lever rule 1-32 that each tie-line, e f, of the isothermal
applies just as for binary T – X diagrams. section corresponds to a tie-line c d of the
log pO2 – x diagram. This underscores the log pO2 – x diagram. For example, the (spi-
fact that every tie-line of a ternary isother- nel + (Fe, Cr) O + alloy) triangle with cor-
mal section corresponds to a constant chem- ners at points a, b and c corresponds to the
ical potential of each of the components. “eutectic-like” or eutecular invariant with
Another example of a log pO2 – x diagram the same phase compositions a, b and c at
is shown for the Fe – Cr – O system at log pO2 ≈ –10.7. We can see that within a
1573 K in the lower panel of Fig. 1-33 three-phase tie-triangle, pO2 is constant.
(Pelton and Schmalzried, 1973). The corre- An example of yet another kind of phase
sponding ternary isothermal section in diagram of this same geometrical type is
Jänecke coordinates is shown in the upper shown in Fig. 1-34. For the quaternary
panel. Each of the invariant three-phase Fe – Cr – O2 – SO2 system at T = 1273 K and
tie-triangles in the isothermal section at constant pSO2 = 10–7 bar, Fig. 1-34 is a
corresponds to an invariant line in the plot of log pO2 versus the molar metal ratio
Figure 1-33. Correspond-

ing phase diagrams for the
Fe–Cr–O system at 1573 K
(Pelton and Schmalzried,
1973). Experimental points
from Katsura and Muan
(1964).
Figure 1-34. Calculated phase diagram of log pO2 versus molar metal ratio at T = 1273.15 K and pSO2 = 10–7 bar
for the Fe – Cr – SO2 – O2 system.
x . Since log pO2 varies as –1/2 log pS2 when geometrical type. Often for ceramic
pSO2 and T are constant, Fig. 1-34 is also a systems, we encounter “binary” phase dia-
plot of log pS2 versus x . grams such as that for the “CaO – Fe2O3”
Plotting T versus x at constant pO2 in the system in Fig. 1-35, which has been taken
Fe – Cr – O system, or at constant pO2 and from Phillips and Muan (1958). How are
pSO2 in the Fe – Cr – SO2 – O2 system, will we to interpret such a diagram? How, for
also result in phase diagrams of this same instance, do we interpret the composition
Figure 1-35. Phase diagram for the

“CaO – Fe2O3” system in air ( pO2 =
0.21 bar) from Phillips and Muan
(1958) (Reprinted by permission of the
American Ceramic Society from Levin
et al., 1964).
axis when applied to the magnetite phase?

In light of the preceding discussion, it can
be seen that such diagrams are really T – x
plots at constant pO2 , where x is the metal
ratio in any phase. The diagram will be dif-
ferent at different oxygen partial pressures.
If pO2 is not fixed, the diagram cannot be
interpreted.
It can be seen that the diagrams dis-
cussed above are of the same geometrical
type as binary T – X diagrams because they
are all composed of the same geometrical
units of construction as in Fig. 1-16. Their
interpretation is thus immediately clear
to anyone familiar with binary T – X dia- Figure 1-36. Pressure-temperature phase diagram of
grams. Chemical potential – composition H2O.
Figure 1-37. Corresponding phase diagrams for the Fe – O system at PTOTAL = 1 bar (after Muan and Osborn,
1965).
diagrams (Figs. 1-32 to 1-34) are useful in of Figs. 1-1 to 1-3 are of this same geomet-
the study of hot corrosion, metallurgical rical type.
roasting processes, chemical vapor deposi- As yet another example of this geomet-
tion, and many aspects of materials pro- rical type of diagram, a plot of R T ln pO2
cessing. versus T for the Fe – O system is shown in
Another important geometrical type of Fig. 1-37 (b). Again, one-, two- and three-
phase diagram is exemplified by P – T phase regions are indicated by areas, lines
phase diagrams for one-component sys- and triple points respectively. Fig. 1-37 (a)
tems, as shown for H2O in Fig. 1-36. In is the binary T – composition phase diagram
such diagrams (see also Chapter 10 by for the Fe – O system. The correspondence
Kunz (2001)) bivariant single-phase re- between Figs. 1-37 (a) and 1-37 (b) is evi-
gions are indicated by areas, univariant dent. Each two-phase line of Fig. 1-37 (b)
two-phase regions by lines, and invariant “opens up” to a two-phase region of Fig.
three-phase regions by triple points. An 1-37 (a). Each tie-line of a two-phase re-
important rule of construction is the exten- gion in Fig. 1-37 (a) can thus be seen to
sion rule, which is illustrated by the broken correspond to a constant pO2 . Triple points
lines in Fig. 1-36. At a triple point, the in Fig. 1-37 (b) become horizontal invari-
extension of any two-phase line must pass ant lines in Fig. 1-37 (a).
into the single-phase region of the third Yet another type of phase diagram is
phase. Clearly, the predominance diagrams shown in Fig. 1-38. This is an isothermal
Figure 1-38. Phase diagram of log pS2 versus log pO2 at 1273 K and constant molar metal ratio nCr /(nFe + nCr) =
0.5 in the Fe – Cr – S2 – O2 system.
section at constant molar metal ratio Table 1-1. Corresponding pairs of potentials f i and
nCr /(nFe + nCr ) = 0.5 for the Fe – Cr – S2 – O2 extensive variables qi .
system. This diagram was calculated ther- fi : T P m1 m 2 … mC
modynamically from model parameters.
qi : S –V n1 n2 … nC
The axes are the equilibrium sulfur and
oxygen partial pressures. Three or four
boundary lines can meet at an intersection
point. Some of the boundary lines on Fig. where U is the internal energy of the
1-38 separate a two-phase region (a + b) system. The corresponding potentials and
from another two-phase region (a + g). extensive variables are listed in Table 1-1.
These lines thus represent the conditions It may also be noted that the corresponding
for three-phase (a + b + g) equilibrium. pairs are found together in the terms of the
general Gibbs – Duhem equation:
S dT – V dP + S ni dm i = 0 (1-115)
1.9 General Phase Diagram
Geometry
1.9.1 General Geometrical Rules
Although the various phase diagrams for All True Phase Diagram Sections
shown in the preceding sections may ap-
The Law of Adjoining Phase Regions ap-
pear to have quite different geometries, it
plies to all true sections. As a phase boun-
can be shown that, in fact, all true phase di-
dary line is crossed, one and only one
agram sections obey the same set of geo-
phase either appears or disappears.
metrical rules. Although these rules do not
If the vertical axis is a potential (T, P,
apply directly to phase diagram projections
m i ), then horizontal invariant lines like the
such as Figs. 1-22, 1-24 and 1-31, such di-
eutectic line in Fig. 1-12 or the line AB in
agrams can be considered to consist of por-
Fig. 1-29 will be seen when the maximum
tions of several phase diagram sections
number of phases permitted by the phase
projected onto a common plane.
rule are at equilibrium. However, if these
By “true” phase diagram we mean one in
are considered to be degenerate infinitely
which each point of the diagram represents
narrow phase fields, then the Law of Ad-
one unique equilibrium state. In the present
joining Phase Regions still applies. This is
section we give the general geometrical
illustrated schematically in Fig. 1-39 where
rules that apply to all true phase diagram
the three-phase eutectic line has been
sections, and we discuss the choices of
“opened up”. Similarly, if both axes are po-
axes and constants that ensure that the dia-
tentials, then many phase boundaries may
gram is a true diagram.
be degenerate infinitely narrow regions.
We must first make some definitions. In
For example, all phase boundaries on Figs.
a system of C components we can define
1-1 to 1-3, 1-36 and 1-37 (b) are degenerate
(C + 2) thermodynamic potentials f i . These
two-phase regions which are schematically
are T, P, m 1 , m 2 , …, m C , where m i is the
shown “opened up” on Fig. 1-40.
chemical potential defined in Eq. (1-23).
All phase boundary lines in a true phase
For each potential there is a corresponding
diagram meet at nodes where exactly four
extensive variable qi related by:
lines converge, as in Fig. 1-41. N phases
f i = (∂U/∂qi )q j ( j ≠ i ) (1-114) (a1 , a2 , …, aN ) where N ≥ 1 are common
1.9 General Phase Diagram Geometry 57
Figure 1-39. An isobaric binary T – X

phase diagram (like Fig. 1-12) with the
eutectic line “opened up” to illustrate that
this is a degenerate 3-phase region.
Figure 1-41. A node in a true phase diagram sec-

Figure 1-40. A potential – potential phase diagram tion.
(like Fig. 1-1 or Fig. 1-36) with the phase boundaries
“opened up” to illustrate that they are degenerate 2-
phase regions.
components was noted by Hillert (1985)
and proved by Pelton (1995). In the case of
to all four regions. Schreinemakers’ Rule degenerate phase regions, all nodes can
states that the extensions of the boundaries still be considered to involve exactly four
of the N-phase region must either both lie boundary lines if the degenerate boundar-
within the (N + 1)-phase regions as in Fig. ies are “opened up” as in Figs. 1-39 and
1-41 or they must both lie within the 1-40.
(N + 2)-phase region. This rule is illustrated An objection might be raised that a
by the extrapolations in Fig. 1-29 at points minimum or a maximum in a two-phase re-
a, b and c and in Fig. 1-30 at points b, c, n, gion in a binary temperature – composition
i and s. The applicability of Schreine- phase diagram, as in Fig. 1-10 or in the
makers’ Rule to systems of any number of lower panel of Fig. 1-16, represents an ex-
ception to Schreinemakers’ Rule. How- Fig. 1-1, all lines are actually two coinci-
ever, the extremum in such a case is not ac- dent ZPF lines.
tually a node where four phase boundaries The ZPF line concept is very useful in
converge, but rather a point where two the development of general algorithms for
boundaries touch. Such extrema in which the thermodynamic calculation of phase di-
two phase boundaries touch with zero agrams as discussed in Sec. 1-12.
slope may occur for a C-phase region in a
phase diagram of a C-component system 1.9.2 Choice of Axes and Constants
when one axis is a potential. For example, of True Phase Diagrams
in an isobaric temperature – composition
In a system of C components, a two-
phase diagram of a four-component sys-
dimensional diagram is obtained by choos-
tem, we may observe a maximum or a min-
ing two axis variables and holding (C – 1)
imum in a four-phase region separating two
other variables constant. However, not all
three-phase regions. A similar maximum
choices of variables will result in a true
or minimum in a (C – n)-phase region,
phase diagram. For example on the P – V
where n > 0, may also occur, but only for
diagram for H2O shown schematically in
a degenerate or special composition path.
Fig. 1-42, at any point in the area where the
For further discussion, see Hillert (1998).
(S + L) and (L + G) regions overlap there
are two possible equilibrium states of the
1.9.1.1 Zero Phase Fraction Lines system. Similarly, the diagram of carbon
activity versus XCr at constant T and P in
All phase boundaries on true phase dia-
the Fe – Cr – C system in Fig. 1-43 (Hillert,
gram sections are zero phase fraction
1997) exhibits a region in which there is no
(ZPF) lines, a very useful concept intro-
unique equilibrium state.
duced by Gupta et al. (1986). There are
In order to be sure that a diagram is a
ZPF lines associated with each phase. On
true phase diagram, we must choose one
one side of its ZPF line the phase occurs,
and only one variable (either fi or qi ) from
while on the other side it does not. For ex-
ample, in Fig. 1-30 the ZPF line for the a
phase is the line a b c d e f. The ZPF line for
the g phase is g h i j k l. For the MC phase the
ZPF line is m n c i o p q. The ZPF line for
M7C3 is r n b h s p k e t, and for M23C6 it is
u d j o s v. These five ZPF lines yield the en-
tire two-dimensional phase diagram. Phase
diagram sections plotted on triangular co-
ordinates as in Figs. 1-25 and 1-27 also
consist of ZPF lines.
In the case of phase diagrams with de-
generate regions, ZPF lines for two differ-
ent phases may be coincident over part
of their lengths. For example, in Fig. 1-12,
line CABD is the ZPF line of the liquid,
while CEBF and DEAG are the ZPF lines Figure 1-42. Schematic P – V diagram for H2O. This
for the a and b phases respectively. In is not a true phase diagram.
selected extensive variables we form the

ratio nFe /(nFe + nCr ). Fig. 1-38 is a plot
of m S2 versus m O2 at constant T, P and
nFe /(nFe + nCr ).
In Fig. 1-28 the selected variables are T,
P, nBi , nSn and nCd , and ratios are formed
from the selected extensive variables as
nCd /(nCd + nBi ) and nSn /(nCd + nBi + nSn )
= XSn . Fig. 1-28 is a plot of T versus
nCd /(nCd + nBi ) at constant P and XSn .
Fig. 1-42, the P – V diagram for H2O, is
not a true phase diagram because P and V
are members of the same conjugate pair.
For the diagram shown in Fig. 1-43, we can
choose one variable from each pair as fol-
Figure 1-43. Carbon activity versus mole fraction lows: T, P, m C , nFe , nCr . However the verti-
of Cr at constant T and P in the Fe – Cr – C system. cal axis is XCr = nCr /(nFe + nCr + nC ). This ra-
This is not a true phase diagram (from Hillert, 1997). tio is not allowed because it contains nC
which is not on the list of chosen variables.
That is, since we have chosen m C to be an
each of the (C + 2) conjugate pairs in Table axis variable, we cannot also choose nC .
1-1. (Also, at least one of these must be an Hence, Fig. 1-43 is not a true phase dia-
extensive variable qi .) From among the gram. A permissible choice for the vertical
n (1 ≤ n ≤ C + 2) selected extensive vari- axis would be nCr /(nFe + nCr ) (see Fig.
ables, (n – 1) independent ratios are then 1-33). Note that many regions of Figs. 1-42
formed. These (n – 1) ratios along with the and 1-43 do represent unique equilibrium
(C + 2 – n) selected potentials are the (C + 1) states. That is, the procedure given here is a
required variables. Two are chosen as axis sufficient, but not necessary, condition for
variables and the remainder are held con- constructing true phase diagrams.
stant. To apply this procedure simply, the com-
As a first example, consider a binary ponents of the system should be formally
system with components A – B. The conju- defined to correspond to the desired axis
gate pairs are (T, S), (P, – V ), (mA , nA ) and variables or constants. For example, in Fig.
(m B , nB ). Let us choose one variable from 1-1 we wish to plot pSO2 and log pO2 as
each pair as follows: T, P, nA , nB . From the axes. Hence we define the components as
selected extensive variables, nA and nB , we Cu – SO2 – O2 rather than Cu – S – O.
form a ratio such as nB /(nA + nB ) = XB . The In several of the phase diagrams in this
resultant phase diagram variables are T, P, chapter, log pi or R T ln pi has been substi-
XB . Choosing any two as axes and holding tuted for m i as axis variable or constant.
the third constant will give a true phase di- From Eq. (1-32), this substitution can
agram as in Fig. 1-6 or Fig. 1-8. clearly be made if T is constant. However,
As a second example, consider Fig. 1-38 even when T is an axis of the phase dia-
for the Fe – Cr – S2 – O2 system. We choose gram as in Fig. 1-37 (b), this substitution is
one variable from each conjugate pair as still permissible since m i0 is a monotonic
follows: T, P, m S2 , mO2 , nFe , nCr . From the function of T. The substitution of ln ai for
m i results in a progressive expansion and Na2O system we would choose as axes

displacement of the axis with increasing T the equivalent cationic and anionic frac-
that preserves the overall geometry of the tions, nNa /(nNa + 2 nCa ) and nCl /(nCl + 2 nO ),
diagram. whose denominators are equal because of
charge neutrality.
1.9.2.1 Tie-lines
1.9.2.2 Corresponding Phase Diagrams
If only potentials (T, P, m i ) are held con-
When only potentials are held constant
stant, then all tie-lines lie in the plane of
and when both axes are also potentials,
the phase diagram section. In this case, the
then the geometry exemplified by Figs. 1-1
compositions of the individual phases at
to 1-3, 1-26 and 1-37 (b) results. Such dia-
equilibrium can be read from the phase
grams were called “type-1 phase diagrams”
diagram, and the lever rule applies as, for
by Pelton and Schmalzried (1973). If only
example, in Figs. 1-6, 1-25, 1-33 or 1-34.
potentials are held constant and one axis is
However, if a ratio of extensive variables,
a potential while the other is a composition
such as a composition, is held constant as
variable, then the geometry exemplified
in the isopleths of Figs. 1-28 to 1-30, then
by Figs. 1-8, 1-12, 1-34, 1-37 (a), and the
in general, tie-lines do not lie in the plane.
lower panel of Fig. 1-33 results. These
If both axes are composition variables
were termed “type-2” diagrams. Finally, if
(ratios of ni ), and if only potentials are held
only potentials are held constant and both
constant, then it is desirable that the tie-
axes are compositions, then a “type-3” dia-
lines (which lie in the plane) be straight
gram as in the upper panels of Figs. 1-32
lines. It can be shown (Pelton and Thomp-
and 1-33 results.
son, 1975) that this will only be the case if
If the fi axis of a phase diagram is re-
the denominators of the two composition
placed by a composition variable that var-
variable ratios are the same. For example,
ies as its conjugate variable qi (ex: qi /q j ,
in the central panel of Fig. 1-32, which is in
qi /(qi + q j )), then the new diagram and the
Jänecke coordinates, the composition vari-
original diagram are said to form a pair of
ables, nCo /(nCo + nNi ) and nO /(nCo + nNi ),
corresponding phase diagrams. For in-
have the same denominator. This same dia-
stance, Figs. 1-37 (a) and 1-37 (b) are cor-
gram can be plotted on triangular coordi-
responding type-1 and type-2 phase dia-
nates as in the upper panel of Fig. 1-32 and
grams, while Fig. 1-33 shows a corre-
such a diagram can also be shown (Pelton
sponding pair of type-2 and type-3 dia-
and Thompson, 1975) to give straight tie-
grams. It is useful to draw corresponding
lines.
diagrams beside each other as in Figs. 1-37
Similarly, in the quasiternary reciprocal
or 1-33 because the information contained
phase diagram of Fig. 1-31 the vertical and
in the two diagrams is complementary.
horizontal axes are nNa /(nNa + nK ) and
nCl /(nCl + nF ). To preserve charge neutral-
1.9.2.3 Theoretical Considerations
ity, (nNa + nK ) = (nCl + nF ), and so the tie-
lines are straight. Generally, in quasiter- A complete rigorous proof that the pro-
nary reciprocal salt phase diagrams, cedure described in this section will always
straight tie-lines are obtained by basing the generate a true phase diagram is beyond
composition on one equivalent of charge. the scope of this chapter. As an outline of
For example, in the CaCl2 – NaCl – CaO – the proof, we start with the generalized
stability criterion: values. For example, in the predominance

(1-116)
diagram of Fig. 1-1, if the components are
(∂f1 /∂q1)f 2 , f 3 , …, qN , qN +1 , qN + 2 , …, qC + 2 ≥ 0
chosen as Cu – SO2 – O2 , then the com-
This equation states that a potential fi al- pound Cu2S is written as Cu2(SO2 )O– 2 ;
ways increases as its conjugate variable qi that is nO2 = –1. This is no problem in Fig.
increases when either f j or q j from every 1-1, since m O2 rather than nO2 was chosen
other conjugate pair is held constant. For from the conjugate pair and is plotted as an
instance, m i of a component always in- axis variable. However, suppose we wish
creases as that component is added to a to plot a diagram of m Cu versus m SO2 at con-
system (that is, as ni is increased) at constant T and P in this system. In this case,
stant T and P, when either the number of the chosen variables would be T, P, m SO2 ,
moles or the chemical potential of every nO2 . Since one of the selected extensive
other component is held constant. In a bi- variables must always be positive, and
nary system, for example, this means that since nO2 is the only selected extensive var-
the equilibrium Gibbs energy envelope is iable, it is necessary that nO2 be positive
always convex, as shown in Fig. 1-6. If the everywhere. For instance, a phase field for
envelope were concave, then the system Cu2S is not permitted. In other words, only
would be unstable and would separate into compositions in the Cu – SO2 – O2 subsys-
two phases, as shown in Fig. 1-11. tem are permitted. A different phase dia-
Consider first a phase diagram with axes gram would result if we plotted m Cu versus
f1 and f2 with f3 , f4 , …, fC +1 and qC + 2 m SO2 in the Cu – SO2 – S2 subsystem with nS2
constant. Such a diagram is always a true always positive. Cu2O would then not ap-
phase diagram. If the potential f1 is now re- pear, for example. That is, at a given m Cu
placed by q1 , the diagram still remains a and pSO2 we could have a low pO2 and a
true phase diagram because of Eq. (1-116). high pS2 in equilibrium with, for example,
The sequence of equilibrium states that oc- Cu2S, or we could have a high pO2 and a
curs as q1 is increased will be the same as low pS2 in equilibrium with, for example,
that which occurs as f1 is increased when Cu2O. Hence the diagram will not be a true
all the other variables (fi or qi ) are held diagram unless compositions are limited to
constant. the Cu – SO2 – O2 or Cu – SO2 – S2 triangles.
A true phase diagram is therefore ob- As a second example, if m SiO and m CO are
tained if the axis variables and constants chosen as variables in the SiO – CO– O
are chosen from the variables f1 , f2 , …, system, then the diagram must be limited
fN , qN +1 , qN +2 , …, qC +1 with qC + 2 held to nO > 0 (SiO – CO – O subsystem) or to
constant. The extensive variables can be nO < 0 (SiO – CO – Si – C subsystem).
normalized as (qi /qC + 2) or by any other in-
dependent and unique set of ratios. 1.9.2.4 Other Sets of Conjugate Pairs
It should be noted that at least one exten-
The set of conjugate pairs in Table 1-1 is
sive variable, qC + 2 , is considered to be con-
only one of many such sets. For example, if
stant across the entire diagram. In practice,
we make the substitution (H = T S + S ni m i )
this means that one of the extensive vari-
in Eq. (1-115), then we obtain another form
ables must be either positive or negative
of the general Gibbs – Duhem equation:
everywhere on the diagram. For certain
formal choices of components, extensive – H d (1/T ) – (V/T ) dP + S ni d ( m i /T ) = 0
composition variables can have negative (1-117)
This defines another set of pairs of conju- case the model is formally the same as that
gate potentials and extensive variables: of a simple substitutional solution, because
(1/T, – H ), (P, – V/T ), ( m i /T, ni ). Choosing the site fractions XMg and XCa of Mg2 + and
one and only one variable from each pair, Ca2 + cations on the cationic sublattice are
we can construct a true phase diagram by numerically equal to the overall component
the procedure described above. However, mole fractions XMgO and XCaO . Solid and liq-
these diagrams may be of limited practical uid MgO – CaO solutions have been shown
utility. This is discussed by Hillert (1997). (Wu et al., 1993) to be well represented by
simple polynomial equations for g E .
As a second example, the intermetallic e-
1.10 Solution Models FeSb phase exhibits non-stoichiometry to-
ward excess Fe. This phase was modeled
In Sec. 1.4.7, the thermodynamic expres- (Pei et al., 1995) as a solution of Fe and
sions for simple ideal substitutional solu- stoichiometric FeSb by assuming two sub-
tions were derived and in Secs. 1.5.7 and lattices: an “Fe sublattice” occupied only
1.6.2, the regular solution model and poly- by Fe atoms and an “Sb sublattice” occu-
nomial extensions thereof were discussed. pied by both Fe and Sb atoms such that, per
For other types of solutions such as ionic gram atom,
mixtures, interstitial solutions, polymeric Dgm = 0.5 R T ( yFe ln yFe + ySb ln ySb )
solutions, etc., the most convenient defini-
+ a yFe ySb (1-118)
tion of ideality may be different. In the
present section we examine some of these where ySb = (1 – yFe ) = 2 XSb is the site frac-
solutions. We also discuss structural order- tion of Sb atoms on the “Sb sites” and a is
ing and its effect on the phase diagram. For an empirical polynomial in ySb .
further discussion, see Pelton (1997).
1.10.1.2 Ionic Solutions
1.10.1 Sublattice Models
Let us take as an example a solution,
The sublattice concept has proved to be solid or liquid, of NaF, KF, NaCl and KCl
very useful in thermodynamic modeling. as introduced in Sec. 1.7.8. If the cations
Sublattice models, which were first devel- are assumed to mix randomly on a cationic
oped extensively for molten salt solutions, sublattice while the anions mix randomly
find application in ceramic, interstitial so- on an anionic sublattice, then the molar
lutions, intermetallic solutions, etc. Gibbs energy of the solution can be mod-
eled by the following equation which con-
1.10.1.1 All Sublattices Except One tains an ideal mixing term for each sublat-
Occupied by Only One Species tice:
In the simplest limiting case, only one g = (XNa XCl g 0NaCl + XK XF g 0KF
sublattice is occupied by more than one + XNa XF g 0NaF + XK XCl g 0KCl ) (1-119)
species. For example, liquid and solid + R T (XNa ln XNa + XK ln XK )
MgO – CaO solutions can be modeled by + R T (XF ln XF + XCl ln XCl ) + g E
assuming an anionic sublattice occupied
only by O2 – ions, while Mg2 + and Ca2 + where the factor (XNa XCl ), for example, is
ions mix on a cationic sublattice. In this the probability, in a random mixture, of
1.10 Solution Models 63
finding a Na ion and a Cl ion as nearest accounts for the fact that the number
neighbors. Differentiation of Eq. (1-119) of Na+ – F – and K+ – Cl – nearest-neighbor
gives the following expression for the ac- pairs will be higher than the number of
tivity of NaF: such pairs in a random mixture. This term
is usually not negligible.
R T ln aNaF = – XK XCl DG exchange (1-120)
The phase diagram in Fig. 1-31 (a) was
+ R T ln (XNa XF ) + g ENaF calculated by means of Eqs. (1-119) and
(1-122) solely from optimized excess
where DG exchange is the Gibbs energy
Gibbs energies of the binary systems and
change for the following exchange reaction
from compiled data for the pure salts.
among the pure salts:
Agreement with the measured diagram is
NaCl + KF = NaF + KCl; (1-121) very good.
DG exchange
=g 0
+g 0
–g 0
– g 0KF Eqs. (1-119) and (1-122) can easily be
NaF KCl NaCl
modified for solutions in which the num-
In this example, DG exchange < 0. The salts bers of sites on the two sublattices are not
NaF and KCl are thus said to form the equal, as in MgCl2 – MgF2 – CaCl2 – CaF2
stable pair. The first term on the right of solutions. Also, in liquid salt solutions the
Eq. (1-120) is positive. The members of the ratio of the number of lattice sites on one
stable pair thus exhibit positive deviations, sublattice to that on the other sublattice
and in Fig. 1-31 this is reflected by the flat can vary with concentration, as in molten
liquidus surfaces with widely spaced iso- NaCl – MgCl2 – NaF – MgF2 solutions. In
therms for NaF and KCl. That is, the mix- this case, it has been proposed (Saboungi
ing of pure NaF and KCl is unfavorable be- and Blander, 1975) that the molar ionic
cause it involves the formation of K+ – F – fractions in all but the random mixing
and Na+ – Cl – nearest-neighbor pairs at the terms of these equations should be replaced
expense of the energetically preferable by equivalent ionic fractions. Finally, the
Na+ – F – and K+ – Cl – pairs. If DG exchange is equations can be extended to multicompo-
sufficiently large, a miscibility gap will be nent solutions. These extensions are all dis-
formed, centered close to the stable diago- cussed by Pelton (1988 b).
nal joining the stable pair. For solutions such as liquid NaF – KF –
Blander (1964) proposed the following NaCl – KCl for which DG exchange is not too
expression for g E in Eq. (1-119): large, these equations are often sufficient.
For solutions with larger exchange Gibbs
g E = XNa g ENaCl–NaF + XK g EKCl–KF (1-122) energies, however, in which liquid immis-
E E
+ XF g NaF–KF + XCl g NaCl–KCl cibility appears, these equations are gener-
– XNa XK XF XCl (DG exchange )2/ZRT ally unsatisfactory. It was recognized by
Saboungi and Blander (1974) that this is
where, for example, g ENaCl–NaF is the excess due to the effect of non-randomness upon
Gibbs energy in the NaCl – NaF binary the second nearest-neighbor cation – cation
system at the same cationic fraction XNa as and anion – anion interactions. To take
in the ternary, and where Z is the first coor- account of this, Blander proposed addi-
dination number. That is, g E contains a tional terms in Eq.(1-122). Dessureault and
contribution from each binary system. The Pelton (1991) modified Eqs. (1-119) and
final term in Eq. (1-122) is a first-order (1-122) to account more rigorously for
correction for non-random mixing which non-random mixing effects, with good re-
sults for several molten salt systems with ions occupy both sublattices. The solu-
miscibility gaps. See also section 1.10.4. tion can be represented as (Fe21–+ x Ti3x + )A
4+ 3+
(Ti1– x Tix )B where x is the overall mole
fraction of Ti2O3 . The ions are assumed to
1.10.1.3 Interstitial Solutions
mix randomly on each sublattice so that:
As an example of the application of the
sublattice model to interstitial solutions we Ds ideal = – 2 R [(1 – x) ln (1 – x) + x ln x]
will take the f.c.c. phase of the Fe –V– C (1-124)
system. Lee and Lee (1991) have modeled Deviations from ideal mixing are as-
this solution using two sublattices: a metal- sumed to occur due to interlattice cation –
lic sublattice containing Fe and V atoms, cation interactions according to
and an interstitial lattice containing C
(FeA2+ – TiB4+ ) + (TiA3+ – TiB3+ )
atoms and vacancies, va. The numbers of
sites on each sublattice are equal. An equa- = (FeA2+ – TiB3+ ) + (TiA3+ – TiB4+ )
tion identical to Eq. (1-119) can be written DG = a + b T (1-125)
for the molar Gibbs energy:
where a and b are the adjustable parameters
g = (XFe Xva g 0Feva + XFe XC g 0FeC of the model. The probability that an A – B
+ XV Xva g 0Vva + XV XC g 0VC ) (1-123) pair is an (FeA2+ – TiB3+ ) or a (TiA3+ – TiB4+ )
+ R T (XFe ln XFe + XV ln XV) pair is equal to x (1 – x). Hence, g E =
x (1 – x) (a + b T ).
+ R T (XC ln XC + Xva ln Xva ) + g E
Similar models can be proposed for
where XFe = (1 – XV) and XC = (1 – Xva ) are other ceramic solutions such as spinels,
the site fractions on the two sublattices and pseudobrookites, etc. These models can
“Feva” and “Vva” are simply pure Fe and rapidly become very complex mathemati-
V, i.e., g 0Feva = g 0Fe . An expression for g E as cally as the number of possible species on
in Eq. (1-122), although without the final the lattices increases. For instance, in
non-random mixing term, was used by Lee Fe3O4 – Co3O4 spinel solutions, Fe2+, Fe3+,
and Lee with optimized binary g E parame- Co2+ and Co3+ ions are all distributed over
ters. Their calculated Fe –V– C phase dia- both the tetrahedral and octahedral sublat-
gram is in good agreement with experimen- tices. Four independent equilibrium con-
tal data. The sublattice model has been stants are required (Pelton et al., 1979) to
similarly applied to many interstitial solu- describe the cation distribution even for the
tions by several authors. ideal mixing approximation. This com-
plexity has been rendered much more tract-
able by the “compound energy model”
1.10.1.4 Ceramic Solutions (Sundman and Ågren, 1981; Hillert et al.,
Many ceramic solutions contain two or 1988). This is not actually a model, but is
more cationic sublattices. As an example, rather a mathematical formalism permit-
consider a solution of Ti2O3 in FeTiO3 (il- ting the formulation of various models in
menite) under reducing conditions. There terms of the Gibbs energies, g 0 , of “pseu-
are two cationic sublattices, the A and B docomponents” so that equations similar to
sublattices. In FeTiO3 , Fe2 + ions and Ti 4 + Eq. (1-119) can be used directly.
ions occupy the A and B sublattices, re-
spectively. With additions of Ti2O3 , Ti 3 +
1.10.1.5 The Compound Energy multicomponent solutions. It also provides

Formalism a conceptual framework for treating many
different phases with different structures.
As an example, the model for the
This facilitates the development of com-
FeTiO3 – Ti2O3 solution in Sec. 1.10.1.4
puter software and databases because many
will be reformulated. By taking all com-
different types of solutions can be treated
binations of an A-sublattice species and
as cases of one general formalism.
a B-sublattice species, we define two
real components, (Fe2+ )A (Ti 4 + )BO3 and
(Ti3+ )A (Ti 3+ )BO3 , as well as two “pseudo- 1.10.1.6 Non-Stoichiometric Compounds
components”, (Fe2+ )A (Ti 3+ )BO3– and
3+ 4+
(Ti )A (Ti )BO3 . + Non-stoichiometric compounds are gen-
Pseudocomponents, as in the present ex- erally treated by a sublattice model. Con-
ample, may be charged. Similarly to Eq. sider such a compound A1– d B1+ d . The sub-
(1-119) the molar Gibbs energy can be lattices normally occupied by A and B at-
written oms will be called, respectively, the A-sub-
lattice and the B-sublattice. Deviations
g = (1 – x)2 g 0FeTiO3 + x 2 g 0Ti2O3 (1-126) from stoichiometry (where d = 0) can occur
+ x (1 – x) g FeTiO3– + x (1 – x) g Ti2O3+ – T Ds ideal
0 0
by the formation of defects such as B atoms
on A sites, vacant sites, atoms occupying
Note that charge neutrality is maintained in
interstitial sites, etc. Generally, one type of
Eq. (1-126). The Gibbs energies of the two
defect will predominate for solutions with
pseudocomponents are calculated from the
excess A and another type will predomi-
equation
nate for solutions with excess B. These are
DG = a + b T called the majority defects.
= g 0FeTiO3– + g 0Ti2O3+ – g 0FeTiO3 – g 0Ti2O3 (1-127) Consider first a solution in which the
majority defects are A atoms on B sites and
where DG is the Gibbs energy change of B atoms on A sites: (A1– x Bx )A (A y B1– y )B . It
Eq. (1-125) and is a parameter of the follows that d = (x – y). In the compound
model. One of g 0FeTiO3– or g 0Ti2O3+ may be as- energy formalism we can write, for the mo-
signed an arbitrary value. The other is then lar Gibbs energy,
given by Eq. (1-127). By substitution of
0 0
Eq. (1-127) into Eq. (1-126) it may be g = (1 – x) (1 – y) g AB + (1 – x) y g AA
shown that this formulation is identical to 0
+ x (1 – y) g BB 0
+ x y g BA (1-128)
the regular solution formulation given in + R T [x ln x + (1 – x) ln (1 – x)
Sec. 1.10.1.4. Note that excess terms, g E,
could be added to Eq. (1-126), thereby giv- + y ln y + (1 – y) ln (1 – y)]
ing more flexibility to the model. In the 0
where g AB is the molar Gibbs energy of
present example, however, this was not re- (hypothetical) defect-free stoichiometric
quired. AB. Now the defect concentrations at equi-
The compound energy formalism is de- librium are those that minimize g. There-
scribed and developed by Barry et al. fore, setting (∂g /∂x) = (∂g /∂y) = 0 at con-
(1992), who give many more examples. An stant d, we obtain
advantage of formulating the sublattice (1-129)
model in terms of the compound energy xy ⎛ Dg + Dg 2 ⎞
= exp ⎜ − (1 − x − y ) 1 ⎟
formalism is that it is easily extended to (1 − x ) (1 − y ) ⎝ RT ⎠
0 0
where Dg1 = (g AA – g AB ) and Dg2 = 40)) are observed. This large discrepancy
0 0
(g BB – g AB ) are the Gibbs energies of for- can be attributed to the fact that the indi-
mation of the majority defects and where vidual segments of the polymer molecule
0 0
g BA has been set equal to g AB . At a given have considerable freedom of movement.
composition d = (x – y), and for given val- Flory (1941, 1942) and Huggins (1942)
ues of the parameters Dg1 and Dg2 , Eq. proposed a model in which the polymer
(1-129) can be solved to give x and y, segments are distributed on the solvent
which can then be substituted into Eq. sites. A large polymer molecule can thus be
(1-128) to give g. The more positive are oriented (i.e. bent) in many ways, thereby
Dg1 and Dg2 , the more steeply g rises on ei- greatly increasing the entropy. To a first ap-
ther side of its minimum, and the narrower proximation the model gives an ideal mix-
is the range of stoichiometry of the com- ing term with mole fractions replaced by
pound. volume fractions in Eq. (1-45):
Consider another model in which the
⎛ XA v A0
majority defects are vacancies on the B- Dg mideal = RT ⎜ XA ln
sublattice and B atoms on interstitial sites. ⎝ XA v A0 + XB v B0
We now have three sublattices with occu- XB v B0 ⎞
pancies (A)A (B1– y vay )B (Bx va1– x )I where + XB ln 0 ⎟
(1-131)
XA v A + XB v A ⎠
0
“I” indicates the interstitial sublattice. The
A-sublattice is occupied exclusively by A Lewis and Randall (1961) have com-
atoms. A vacancy is indicated by va. Stoi- pared the Flory – Huggins equation with
chiometric defect-free AB is represented experimental data in several solutions.
by (A) (B) (va) and (x – y) = 2 d /(1 – d ). In general, the measured activities lie be-
Per mole of A1– d B1+ d , the Gibbs energy is: tween those predicted by Eq. (1-131) and
0 by the Raoultian ideal equation, Eq. (1-45).
g = (1 – d ) {[(1 – x) (1 – y) gABva
0 0 0
A recent review of the thermodynamics
+ (1 – x) y gAvava + x (1 – y) gABB + x y gAvaB of polymer solutions is given by Trusler
+ R T [x ln x + (1 – x) ln (1 – x) (1999).
+ y ln y + (1 – y) ln (1 – y)]} (1-130)
1.10.3 Calculation of Limiting Slopes
Eq. (1-130) is identical to Eq. (1-128) apart
of Phase Boundaries
from the factor (1 – d ), and gives rise to an
equilibrium constant as in Eq. (1-129). From the measured limiting slope
Other choices of majority defects result in (dT/dX )XA =1 of the liquidus at the melting
very similar expressions. The model can point of a pure component A, much infor-
easily be modified to account for other mation about the extent of solid solubility,
stoichiometries Am Bn , for different num- as well as the structure of the liquid, can be
bers of available interstitial sites, etc., and inferred. Similar information can be ob-
its extension to multicomponent solutions tained from the limiting slopes of phase
is straightforward. boundaries at solid-state transformation
points of pure components.
1.10.2 Polymer Solutions Eq. (1-65) relates the activities along the
liquidus and the solidus to the Gibbs en-
For solutions of polymers in monomeric
ergy of fusion:
solvents, very large deviations from simple
l
Raoultian ideal behaviour (i.e. from Eq. (1- R T ln aA – R T ln a As = – Dg0f(A) (1-132)
In the limit XA Æ 1, the liquidus and solidus Agreement with the measured limiting liq-
converge at the melting point Tf (A) . Let us uidus slope is very good, thereby showing
assume that, in the limit, Raoult’s Law, Eq. that the solid solubility of NaF in LiF is not
(1-40), holds for both phases. That is, large.
l
aA = X Al and a As = X As . Furthermore, from In the general case, the solute B may dis-
Eq. (1-60), solve to form more than one “particle”. For
example, in dilute solutions of Na2SO4 in
Dg0f(A) Æ Dhf (A) (1 – T /Tf (A) )
MgSO4 , each mole of Na2SO4 yields two
Finally, we note that moles of Na+ ions which mix randomly
with the Mg2+ ions on the cationic sublat-
lim (ln XA ) = lim (ln (1 – XB )) = – XB tice. Hence, the real mole fraction of sol-
XA Æ 1 XA Æ 1
vent, XA , is (1 – n XB ) where n is the number
Substituting these various limiting values of moles of foreign “particles” contributed
into Eq. (1-132) yields: by one mole of solute. In the present exam-
lim (dX Al /dT – dX As /dT ) ple, n = 2.
XA Æ 1 Eq. (1-133) now becomes:
= Dh0f(A) /R (Tf (A) )2 (1-133)
lim (dX Al /dT – dX As /dT )
XA Æ 1
If the limiting slope of the liquidus,
lim (dX Al /dT ), is known, then the limiting = Dh0f(A) / n R (Tf (A) )2 (1-134)
XA Æ 1
slope of the solidus can be calculated, or The broken line in Fig. 1-44 is the limiting
vice versa, as long as the enthalpy of fusion liquidus slope calculated from Eq. (1-134)
is known. under the assumption of negligible solid
For the LiF – NaF system in Fig. 1-18, solubility.
the broken line is the limiting liquidus It can be shown (Blander, 1964) that Eq.
slope at XLiF = 1 calculated from Eq. (1- (1-134) applies very generally with the fac-
133) under the assumption that there is no tor n as defined above. For example, add-
solid solubility (that is, that dX As /dT = 0). ing LiF to NaF introduces only one foreign
Figure 1-44. Phase dia-

gram of the MgSO4 –
Na2SO4 system calculated
for an ideal ionic liquid so-
lution. Broken line is the
theoretical limiting liquidus
slope calculated for negli-
gible solid solubility taking
into account the ionic nature
of the liquid. Agreement
with the measured diagram
(Ginsberg, 1909) is good.
particle Li +. The F – ion is not “foreign”. or molecule is Zi Xi . Hence, mass balance

Hence, n = 1. For additions of Na2SO4 to equations can be written as
MgSO4 , n = 2 since two moles of Na+ ions
are supplied per mole of Na2SO4 . For Z1 X1 = 2 n11 + n12
CaCl2 dissolved in water, n = 3, and so on. Z 2 X2 = 2 n22 + n12 (1-136)
For C dissolving interstitially in solid Fe, where ni j is the number of i – j bonds in one
n = 1. The fact that the solution is intersti- mole of solution. “Coordination equivalent
tial has no influence on the validity of Eq. fractions” may be defined as:
(1-134). Eq. (1-134) is thus very general
and very useful. It is independent of the so- Y1 = 1 – Y2 = Z1 X1 /(Z1 X1 + Z 2 X2 ) (1-137)
lution model and of the excess properties,
where the total number of pairs in one mole
which become zero at infinite dilution.
of solution is (Z1 X1 + Z 2 X2 )/2. Letting Xi j
An equation identical to Eq. (1-134) but
be the fraction of i – j pairs in solution, Eq.
with the enthalpy of transition, Dh0tr , re-
(1-136) may be written as:
placing the enthalpy of fusion, relates the
limiting phase boundary slopes at a trans- 2 Y1 = 2 X11 + X12
formation temperature of a component. 2 Y2 = 2 X22 + X12 (1-138)
The molar enthalpy and excess entropy of
1.10.4 Short-Range Ordering
mixing are assumed to be directly related
The basic premise of the regular solution to the number of 1– 2 pairs:
model (Sec. 1.5.7) is that random mixing
Dhm – T s E (non-config)
occurs even when gE is not zero. To ac-
count for non-random mixing, the regular = (Z1 X1 + Z 2 X2 ) X12 (w – h T )/4 (1-139)
solution model has been extended though An approximate expression for the config-
the quasichemical model for short-range urational entropy of mixing is given by a
ordering developed by Guggenheim (1935) one-dimensional Ising model:
and Fowler and Guggenheim (1939) and
modified by Pelton and Blander (1984, Dsmconfig = − R ( X1 ln X1 + X2 ln X2 )
1986) and Blander and Pelton (1987). The R
model is outlined below. For a more com- − ( Z1 X1 + Z2 X2 )
2
plete development, see the last two papers
× [ X11 ln ( X11/ Y12 ) + X22 ln ( X22 / Y22 )
cited above, Degterov and Pelton (1996),
Pelton et al. (2000) and Pelton and Char- + X12 ln ( X12 / 2 Y1 Y2 )] (1-140)
trand (2000). The equilibrium distribution is calculated
For a binary system, consider the for- by minimizing Dgm with respect to X12 at
mation of two nearest-neighbor 1– 2 pairs constant composition. This results in a
from a 1–1 and a 2 – 2 pair: “quasichemical” equilibrium constant for
(1–1) + (2 – 2) = 2 (1– 2) (1-135) the reaction, Eq. (1-135):
Let the molar Gibbs energy change for 2
X12 ⎛ − (w − h T ) ⎞
this reaction be (w – h T ). Let the nearest- = 4 exp ⎜ ⎟ (1-141)
X11 X22 ⎝ RT ⎠
neighbor coordination numbers of 1 and 2
atoms or molecules be Z1 and Z 2 . The total When (w – h T ) = 0, the solution of Eqs. (1-
number of bonds emanating from an i atom 138) and (1-141) gives a random distribu-
tion with X11 = Y12, X22 = Y22 and 2 Y1 Y2 , and the following optimized (Kongoli et al.,
Eq. (1-140) reduces to the ideal Raoultian 1998) polynomial:
entropy of mixing. When (w – h T ) be- (1-142)
comes very negative, 1– 2 pairs predomi- (w – h T ) = – (70 017 + 9 T ) – 74 042 YS
nate. A plot of Dhm or s E (non-config) versus – (798 – 15 T ) YS3 + 40 791 YS7 J/mol –1
composition then becomes V-shaped and a
config
plot of Dsm becomes m-shaped, with Far fewer parameters are required than if a
minima at Y1 = Y2 = 1/2, which is the com- polynomial expansion of gE (as in Sec.
position of maximum ordering, as illus- 1.6.2) were used. Furthermore, and more
trated in Fig. 1-45. When (w – h T ) is quite importantly, the model permits successful
negative, the plot of gE also has a negative predictions of the properties of multicom-
V-shape. ponent systems as illustrated in Fig. 1-47,
For Fe – S liquid solutions, the activity where measured sulfur activities in quater-
coefficients of sulfur as measured by sev- nary liquid Fe – Ni – Cu – S solutions are
eral authors are plotted in Fig. 1-46. It is compared with activities calculated (Kon-
clear in this case that the model should be goli et al., 1998) solely from the optimized
applied with Z Fe = Z S . The curves shown in model parameters for the Fe – S, Ni – S
Fig. 1-46 were calculated from the quasi- and Cu – S binary systems. A pair exchange
chemical model with (w – h T ) expanded as reaction like Eq. (1-135) was assumed
for each M – S pair (M = Fe, Ni, Cu), and
an optimized polynomial expansion of
(wMS – h MS T ) as a function of YS , similar to
Eq. (1-142), was obtained for each binary.
Three equilibrium constant equations like
Eq. (1-141) were written, and it was as-
sumed that (wMS – h MS T ) in the quaternary
system was constant at constant YS . No ad-
justable ternary or quaternary parameters
were required to obtain the agreement
shown in Fig. (1-47), although the model
permits the inclusion of such terms if re-
quired.
Silicate slags are known to exhibit such
short-range ordering. In the CaO – SiO2
system, Dhm has a strong negative V-shape,
as in Fig. 1-45, but with the minimum at
XSiO2 = 1/3 which is the composition corre-
sponding to Ca2SiO4 . That is, the ordering
is associated with the formation of ortho-
silicate anions SiO44 –. In the phase diagram,
Fig. 1-14, the CaO-liquidus can be seen to
Figure 1-45. Molar enthalpy and entropy of mixing descend sharply near the composition
curves for a system AB calculated at 1000 °C with
Z1 =Z2 =2 from the quasichemical model for short- XSiO2 = 1/3. The quasichemical model has
range ordering with (ω – η T ) = 0, – 21, – 42, and been extended by Pelton and Blander
– 84 kJ. (1984) to treat silicate slags. The diagram
Figure 1-46. Activity coefficient of sulfur in liquid Fe – S solutions calculated from optimized quasichemical
model parameters and comparison with experimental data (Kongoli et al., 1998).
Figure 1-47. Equilibrium partial pressure of sulfur at 1200 oC over Fe – Ni – Cu – S mattes predicted by the
quasichemical model from binary data (Kongoli et al., 1998) and comparison with experimental data
(Bale and Toguri, 1996).
shown in Fig. 1-14 is thermodynamically model more flexible by permitting the Z i to

calculated (Wu, 1990). vary with composition and by expanding
Many liquid alloy solutions exhibit the (w – h T ) as polynomials in the bond
short-range ordering. The ordering is fractions Xi j rather than the overall compo-
strongest when one component is relatively nent fractions. A merger of the quasichem-
electropositive (on the left side of the peri- ical and sublattice models has also been
odic table) and the other is relatively elec- completed (Chartrand and Pelton, 2000),
tronegative. Liquid alloys such as Alk – Au permitting nearest-neighbor and second-
(Hensel, 1979), Alk – Pb (Saboungi et al., nearest neighbor short-range-ordering to
1985) and Alk – Bi (Petric et al., 1988 a), be treated simultaneously in molten salt so-
where Alk = (Na, K, Rb, Cs), exhibit curves lutions.
of Dhm and Dsm similar to those in Fig.
1-45 with one composition of maximum
1.10.5 Long-Range Ordering
ordering. For example, in the Au – Cs sys-
tem the minima occur near the composition In solid solutions, long-range ordering
AuCs; in Mg – Bi alloys the minima occur can occur as well as short-range ordering.
near the Mg3Bi2 composition, while in In Fig. 1-15 for the Ag – Mg system, a
K – Pb alloys the maximum ordering is at transformation from an a¢ to an a phase is
K 4Pb. shown occurring at approximately 665 K
It has also been observed that certain liq- at the composition Ag3Mg. This is an
uid alloys exhibit more than one composi- order – disorder transformation. Below the
tion of ordering. For example, in K – Te al- transformation temperature, long-range or-
loys, the “excess stability function”, which dering (superlattice formation) is observed.
is the second derivative of Dgm , exhibits An order parameter may be defined which
peaks near the compositions KTe8 , KTe decreases to zero at the transformation
and K 2Te (Petric et al., 1988 b) thereby temperature. This type of phase transfor-
providing evidence of ordering centred on mation is not a first-order transformation
these compositions. The liquid might be con- like those considered so far in this chapter.
sidered as consisting of a series of mutually Unlike first-order transformations which
soluble “liquid intermetallic compounds”. involve a change of state (solid, liquid, gas)
When (w – h T ) is expanded as a polyno- and also involve diffusion over distances
mial as in Eq. (1-142), the quasichemical large compared with atomic dimensions,
model and the polynomial model of Sec. order – disorder transformations, at least at
1.6.2 become identical as (w – h T ) ap- the stoichiometric composition (Ag3Mg in
proaches zero. That is, the polynomial this example), occur by atomic rearrange-
model is a limiting case of the quasichemi- ment over distances of the order of atomic
cal model when the assumption of ideal dimensions. The slope of the curve of
configurational entropy is made. Gibbs energy versus T is not discontinuous
When (w – h T ) is positive, (1–1) and at the transformation temperature. Order-
(2 – 2) pairs predominate. The quasichemi- ing and order – disorder transformations are
cal model can thus also treat such cluster- discussed in Chapter 8 (Inden, 2001).
ing, which accompanies positive devia- A type of order – disorder transformation
tions from ideality. of importance in ferrous metallurgy is the
Recent work (Pelton et al., 2000; Pelton magnetic transformation. Below its Curie
and Chartrand, 2000) has rendered the temperature of 769 °C, Fe is ferromagnetic.
Above this temperature it is not. The trans-

formation involves a change in ordering of
the atomic spins and is not first order. Ad-
ditions of alloying elements will change
the temperature of transformation. Mag-
netic transformations are treated in Chapter
4 (Binder, 2001). See also Miodownik
(1982) and Inden (1982).
1.11 Calculation of Ternary Phase

Diagrams From Binary Data
Among 70 metallic elements 70!/3! 67!
= 54 740 ternary systems and 916 895 qua-
ternary systems are formed. In view of the
amount of work involved in measuring
even one isothermal section of a relatively
simple ternary phase diagram, it is very im-
portant to have a means of estimating ter-
nary and higher-order phase diagrams.
The most fruitful approach to such pre-
dictions is via thermodynamic methods. In
recent years, great advances have been
made in this area by the international Cal-
phad group. Many key papers have been
published in the Calphad Journal.
As a first step in the thermodynamic ap-
proach, we critically analyze the experi-
mental phase diagrams and thermodynamic Figure 1-48. Projection of the liquidus surface of
data for the three binary subsystems of the the KCl – MgCl2 – CaCl2 system.
ternary system in order to obtain a set of a) Calculated from optimized binary thermodynamic
parameters (Lin et al., 1979).
mathematical expressions for the Gibbs b) As reported by Ivanov (1953).
energies of the binary phases, as was dis-
cussed in Sec. 1.6. Next, interpolation pro-
cedures based on solution models are used
to estimate the Gibbs energies of the ter-
nary phases from the Gibbs energies of the (Ivanov, 1953) and estimated (Lin et al.,
binary phases. Finally, the ternary phase di- 1979) liquidus projections of the KCl –
agram is calculated by computer from MgCl2 – CaCl2 system are shown in Fig.
these estimated ternary Gibbs energies by 1-48. The estimated phase diagram was
means of common tangent plane or total calculated from the thermodynamic prop-
Gibbs energy minimization algorithms. erties of the three binary subsystems with
As an example of such an estimation of the Gibbs energy of the ternary liquid ap-
a ternary phase diagram, the experimental proximated via the equation suggested by
1.11 Calculation of Ternary Phase Diagrams From Binary Data 73
Kohler (1960): tal effort involved in determining multi-

component phase diagrams.
g E = (1 – XA )2 g B/C
E
+ (1 – XB )2 g C/A
E
Reviews of various interpolation proce-
+ (1 – XC )2 g A/B
E
(1-143) dures and computer techniques for estimat-
ing and calculating ternary and higher-or-
In this equation, g E is the excess molar der phase diagrams are given by Ansara
Gibbs energy at a composition point in the (1979), Spencer and Barin (1979) and Pel-
E E E
ternary liquid phase and g B/C , g C/A and g A/B ton (1997).
are the excess Gibbs energies in the three Other equations, similar to the Kohler
binary systems at the same ratios XB /XC , Eq. (1-143) in that they are based on exten-
XC /XA and XA /XB as at the ternary point. If sion of regular solution theory, are used to
the ternary liquid phase as well as the three estimate the thermodynamic properties of
binary liquid phases are all regular solu- ternary solutions from the properties of the
tions, then Eq. (1-143) is exact. In the binary subsystems. For a discussion and
general case, a physical interpretation of references, see Hillert (1980). However,
Eq. (1-143) is that the contribution to g E for structurally more complex solutions
from, say, pair interactions between A and involving more than one sublattice or with
B particles is constant at a constant ratio significant structural ordering, other esti-
XA /XB apart from the dilutive effect of the mation techniques must be used. For a re-
C particles, which is accounted for by the view, see Pelton (1997).
term (1 – XC )2 taken from regular solution An example, the calculation of the
theory. phase diagram of the NaCl – KCl – NaF – KF
Ternary phase diagrams estimated in this system in Fig. 1-31, has already been pre-
way are quite acceptable for many pur- sented in Sec. 1.10.1.2.
poses. The agreement between the experi- The quasichemical model for systems
mental and calculated diagrams can be with short-range ordering was discussed
greatly improved by the inclusion of one for the case of binary systems in Sec.
or two “ternary terms” with adjustable 1.10.4. The model has been extended to
coefficients in the interpolation equations permit the estimation of ternary and multi-
for g E. For example, the ternary term component phase diagrams (Pelton and
a XKCl XMgCl2 XCaCl2 , which is zero in all Blander, 1986; Blander and Pelton, 1987;
three binaries, could be added to Eq. (1- Pelton and Chartrand, 2000). Very good re-
143) and the value of the parameter a sults have been obtained in the case of sili-
which gives the “best” fit to the measured cate systems. The liquidus surface of the
ternary liquidus could be determined. This, SiO2 – MgO – MnO system, estimated from
of course, requires that ternary measure- optimized binary data with the quasichem-
ments be made, but only a very few (even ical model for the liquid and under the
one or two in this example) experimental assumption of ideal mixing for the solid
liquidus points will usually suffice rather MgSiO3 – MnSiO3 and Mg2SiO4 – Mn2SiO4
than the larger number of measurements solutions, is shown in Fig. 1-49. Agree-
required for a fully experimental determi- ment with the measured phase diagram
nation. In this way, the coupling of the (Glasser and Osborn, 1960) is within ex-
thermodynamic approach with a few well perimental error limits.
chosen experimental measurements holds
promise of greatly reducing the experimen-
Figure 1-49. Projection of liquidus surface of the SiO2 – MgO – MnO system calculated from optimized binary
parameters with the quasichemical model for the liquid phase.
1.12 Minimization of Gibbs solution phases a, b, g, … . The total Gibbs

Energy energy of the system may be written as:
G = (nA g A0 + nB g B0 + …)
Throughout this chapter it has been
+ (na ga + nb gb + …) (1-144)
shown that phase equilibria are calculated
by Gibbs energy minimization. Computer where nA , nB , etc. are the number of moles
software has been developed in recent of the pure solid or liquidus; g A0 , g B0 , etc.
years to perform such calculations in sys- are the molar Gibbs energies of the pure
tems of any number of components, phases solids or liquids (which are functions of T
and species. and P); na , nb , etc. are the total number of
Consider a system in which several stoi- moles of the solution phases; g a , g b , etc.
chiometric solid or liquid compounds A, B, are the molar Gibbs energies of the solution
C, … could be present at equilibrium along phases (which are function of T, P and
with a number of gaseous, liquid or solid composition). For a given set of constraints
1.12 Minimization of Gibbs Energy 75
(such as fixed T, P and overall composi- has been asked to calculate the equilibrium
tion), the free energy minimization algo- state when 1 mol of SiI 4 is held at 1400 K
rithms find the set of mole numbers nA , nB , in a volume of 104 l. The thermodynamic
etc., na , nb , etc. (some may be zero) as well properties of the possible product species
as the compositions of all solution phases have been automatically retrieved from
which globally minimize G. This is the the database and the total Gibbs energy
equilibrium phase assemblage. Other con- has been minimized by the Solgasmix
straints such as constant volume or a fixed algorithm. At equilibrium there will be
chemical potential (such as constant pO2 ) 2.9254 mol of gas of the composition
may be applied. shown and 0.11182 mol of solid Si will
A discussion of the strategies of such al- precipitate. The total pressure will be
gorithms is beyond the scope of the present 0.0336 bar.
chapter. One of the best known general Although the calculation was performed
Gibbs energy minimization programs is by minimization of the total Gibbs energy,
Solgasmix written by Eriksson (1975) and substitution of the results into the equilib-
constantly updated. rium constants of Eqs. (1-10) to (1-12) will
When coupled to a large thermodynamic show that these equilibrium constants are
database, general Gibbs energy minimiza- satisfied.
tion programs provide a powerful tool for Another example is shown in Table 1-3
the calculation of phase equilibria. Several (Pelton et al., 1990). Here the program has
such expert database systems have been been asked to calculate the equilibrium
developed. They have been reviewed by
Bale and Eriksson (1990).
An example of a calculation performed
Table 1-3. Calculation of equilibrium state when re-
by the F * A * C * T (Facility for the Analy-
actants shown (masses in g) are held at 1873 K at a
sis of Chemical Thermodynamics) expert pressure of 1 atm. Calculations performed by mini-
system, which the author has helped to de- mization of the total Gibbs energy.
velop, is shown in Table 1-2. The program
100. Fe + 0.08 O + 0.4 Fe + 0.4 Mn + 0.3 Si + 0.08 Ar =
0.30793 litre ( 99.943 vol% Ar
+ 0.24987E – 01 vol% Mn
Table 1-2. Calculation of equilibrium state when + 0.24069E – 01 vol% SiO
1 mole SiI4 is held at 1400 K in a volume of 104 l. + 0.82057E – 02 vol% Fe
Calculations performed by minimization of the total + 0.79044E – 07 vol% O
Gibbs energy. + 0.60192E – 08 vol% Si
+ 0.11200E – 08 vol% O2
SiI4 =
+ 0.35385E – 15 vol% Si2 )
2.9254 ( 0.67156 I
(1873.0, 1.00 ,G)
+ 0.28415 SiI2
+ 0.24835E – 01 I2 + 0.18501 gram ( 49.948 wt.% SiO2
+ 0.19446E – 01 SiI4 + 42.104 wt.% MnO
+ 0.59083E – 05 SiI + 7.9478 wt.% FeO)
+ 0.23039E – 07 Si (1873.0, 1.00 ,SOLN 2)
+ 0.15226E – 10 Si2 + 100.99 gram ( 99.400 wt.% Fe
+ 0.21462EE – 11 Si3) + 0.33630 wt.% Mu
(1400.0,0.336E – 01,G) + 0.25426 wt.% Si
+ 0.11182 Si + 0.98375E – 02 wt.% O2)
(1400.0,0.336E – 01,S1, 1.0000) (1873.0, 1.00 ,SOLN 3)
state when 100 g Fe, 0.08 g oxygen, 0.4 g calculate all types of phase diagrams with
Fe, 0.4 g Mn, 0.3 g Si and 0.08 g Ar are any properly chosen variables as axes or
brought together at 1873 K at a total pres- constants.
sure of 1 bar. The database contains data
for a large number of solution phases as
well as for pure compounds. These data 1.13 Bibliography
have been automatically retrieved and the
total Gibbs energy has been minimized. At 1.13.1 Phase Diagram Compilations
equilibrium there are 0.30793 l of a gas
phase, 0.18501 g of a molten slag, and The classic compilation in the field of bi-
100.99 g of a molten steel of the composi- nary alloy phase diagrams is that of Hansen
tions shown. (1958). This work was continued by Elliott
The Gibbs energies of the solution (1965) and Shunk (1969). These compila-
phases are represented as functions of com- tions contain critical commentaries. A non-
position by various solution models (Sec. critical compilation of binary alloy phase
1.10). As discussed in Sec. 1.11, these diagrams is supplied in looseleaf form with
models can be used to predict the thermo- a continual up-dating service by W.G. Mof-
dynamic properties of N-component solu- fatt of the General Electric Co., Schenec-
tions from evaluated parameters for binary tady, N.Y. An extensive non-critical compi-
(and possibly ternary) subsystems stored in lation of binary and ternary phase diagrams
the database. For example, in the calcula- of metallic systems has been edited by
tion in Table 1-3, the Gibbs energy of the Ageev (1959 – 1978). An index to all com-
molten slag phase was estimated by the pilations of binary alloy phase diagrams up
quasichemical model from optimized pa- to 1979 was prepared by Moffatt (1979). A
rameters for the binary oxide solutions. critical compilation of binary phase dia-
grams involving Fe has been published by
Kubaschewski (1982). Ternary alloy phase
1.12.1 Phase Diagram Calculation
diagrams were compiled by Ageev (1959 –
Gibbs energy minimization is used to 1978).
calculate general phase diagram sections From 1979 to the early 1990s, the Amer-
thermodynamically using the zero phase ican Society for Metals undertook a project
fraction line concept (Sec. 1.9.1.1), with to evaluate critically all binary and ternary
data retrieved from databases of model co- alloy phase diagrams. All available litera-
efficients. For example, to calculate the di- ture on phase equilibria, crystal structures,
agram of Fig. 1-30, the program first scans and often thermodynamic properties were
the four edges of the diagram to find the critically evaluated in detail by interna-
ends of the ZPF lines. Each line is then fol- tional experts. Many evaluations have ap-
lowed from beginning to end, using Gibbs peared in the Journal of Phase Equilibria
energy minimization to determine the point (formerly Bulletin of Alloy Phase Dia-
at which a phase is just on the verge of be- grams), (ASM Int’l., Materials Park, OH),
ing present. When ZPF lines for all phases which continues to publish phase diagram
have been drawn, then the diagram is com- evaluations. Condensed critical evalua-
plete. Because, as shown in Sec. 1.9, all tions of 4700 binary alloy phase diagrams
true phase diagram sections obey the same have been published in three volumes
geometrical rules, one algorithm suffices to (Massalski et al., 1990). The ternary phase
1.13 Bibliography 77
diagrams of 7380 alloy systems have also Compilations of properties of solutions

been published in a 10-volume compilation (activities, enthalpies of mixing, etc.) are
(Villars et al., 1995). Both binary and ter- much more difficult to find. Hultgren et al.
nary compilations are available from ASM (1973) present the properties of a number
on CD-ROM. Many of the evaluations of binary alloy solutions. An extensive bib-
have also been published by ASM as liography of solution properties of all types
monographs on phase diagrams involving a of solutions was prepared by Wisniak and
particular metal as a component. Tamir (1978).
Each year, MSI Services (http://www. Thermodynamic/phase diagram optimi-
msiwp.com) publishes The Red Book, zation as discussed in Sec. 1.6.1 has been
which contains abstracts on alloy phase carried out for a large number of alloy, ce-
diagrams from all sources, notably from ramic and other systems. Many of these
the extensive Russian literature. MSI also evaluations have been published in the
provides a CD-ROM with extensive alloy international Calphad Journal, published
phase diagram compilations and reports. since 1977 by Pergamon Press. Several of
Phase diagrams for over 9000 binary, ter- the evaluations in the Journal of Phase
nary and multicomponent ceramic systems Equilibria discussed above include ther-
(including oxides, halides, carbonates, sul- modynamic/phase diagram optimizations,
fates, etc.) have been compiled in the 12- as do a number of the evaluations in Vol. 7
volume series, Phase Diagrams for Ceram- of Phase Diagrams for Ceramists.
ists (1964–96, Am. Ceramic Soc., Colum- Extensive computer databases of the
bus, OH). Earlier volumes were non-criti- thermodynamic properties of compounds
cal compilations. However, recent volumes and solutions (stored as coefficients of
have included critical commentaries. model equations) are available. These in-
Phase diagrams of anhydrous salt sys- clude F * A * C * T (http://www.crct.poly-
tems have been compiled by Voskresen- mtl.ca), Thermocalc (http://www.met.
skaya (1970) and Robertson (1966). kth.se), ChemSage (http://gttserv.lth.rwth-
An extensive bibliography of binary aachen.de), MTS-NPL (http://www.npl.
and multicomponent phase diagrams of all co.uk), Thermodata (http://www.grenet.fr),
types of systems (metallic, ceramic, aque- HSC (http://www.outokumpu.fi), and
ous, organic, etc.) has been compiled by MALT2 (http://www. kagaku.com). Gibbs
Wisniak (1981). energy minimization software permits the
calculation of complex equilibria from the
stored data as discussed in Sec. 1.12 as
1.13.2 Thermodynamic Compilations
well as the thermodynamic calculation of
Several extensive compilations of ther- phase diagram sections. A listing of these
modynamic data of pure substances of and other available databases is maintained
interest in materials science are available. at http://www.crct. polymtl.ca .
These include the JANAF Tables (Chase A bibliographic database known as
et al., 1985) and the compilations of Barin Thermdoc, on thermodynamic properties
et al. (1977), Barin (1989), Robie et al. and phase diagrams of systems of interest
(1978) and Mills (1974), as well as the to materials scientists, with updates, is
series of compilations of the National Insti- available through Thermodata (http://
tute of Standards and Technology (Wash- www.grenet.fr).
ington, D.C.).
1.13.3 General Reading Alper, A. M. (Ed.) (1970–1978), Phase Diagrams –

Materials Science and Technology, Vols. 1 – 5.
The theory, measurement and applica- New York: Academic.
Ansara, I. (1979), Internat. Met. Rev. 238, No. 1, 20.
tions of phase diagrams are discussed in a Bale, C. W. (1990), Canad. Metall. Quart. 29, 263.
great many texts. Only a few can be listed Bale, C. W., Eriksson, N. G. (1990), Canad. Metall.
here. A recent text by Hillert (1998) pro- Quart. 29, 105.
Bale, C. W., Pelton, A. D. (1983), Metall. Trans.
vides a complete thermodynamic treatment B 14, 77.
of phase equilibria as well as solution mod- Bale, C. W., Toguri, J. (1996), private communica-
eling and thermodynamic/phase diagram tion.
Bale, C. W., Pelton, A. D., Thompson W. T. (1986),
optimization. Canad. Metall. Quart. 25, 107.
A classical discussion of phase diagrams Barin, I. (1989), Thermochemical Data of Pure Sub-
in metallurgy was given by Rhines (1956). stances. 2. Weinheim: VCH Verlagsgesellschaft.
Barin, I., Knacke, O., Kubaschewski, O. (1977),
Prince (1966) presents a detailed treatment Thermochemical Properties of Inorganic Sub-
of the geometry of multicomponent phase stances. Berlin: Springer Verlag.
diagrams. A series of five volumes edited Barry, T. I., Dinsdale, A. T., Gisby, J. A., Hallstedt,
B., Hillert, M., Jansson, B., Jonsson, J., Sundman,
by Alper (1970 – 1978) discusses many as- B., Taylor, J. R. (1992), J. Phase Equilib. 13, 459.
pects of phase diagrams in materials sci- Bartlett, H. E. Neethling, A. J., Crowther, P. (1970),
ence. Bergeron and Risbud (1984) give an J. Chem. Thermo. 2, 523.
introduction of phase diagrams, with par- Bergeron, C. J., Risbud, S. H. (1984), Introduction to
Phase Equilibria in Ceramics. Columbus, Ohio:
ticular attention to applications in ceramic Amer. Ceramic Soc.
systems, see also Findlay (1951), Ricci Binder, K. (2001), in: Phase Transformations in Ma-
(1964) and West (1965). terials: Kostorz, G. (Ed.). Weinheim: Wiley-VCH,
Chap. 4.
In the Calphad Journal and in the Jour- Binder, K., Fratzl, P. (2001), in: Phase Transforma-
nal of Phase Equilibria are to be found tions in Materials: Kostorz, G. (Ed.). Weinheim:
many articles on the relationships between Wiley-VCH, Chap. 6.
Blander, M. (1964), Molten Salt Chemistry. New
thermodynamics and phase diagrams. York: Interscience, Chap. 3.
It has been beyond the scope of the Blander, M., Pelton, A. D. (1987), Geochim. et Cos-
present chapter to discuss experimental mochim. Acta 51, 85.
Bray, H. F., Bell, F. D., Harris, S. J. (1961 – 1962),
techniques of measuring thermodynamic J. Inst. Metals 90, 24.
properties and phase diagrams. For the Buckley, R. A. (1970), in: Techniques of Metals Re-
measurement of thermodynamic proper- search, Vol. IV, Part 1: Rapp, R. A. (Ed.). New
York: Interscience.
ties, including properties of solutions, the Chartrand, P., Pelton, A. D. (2000), Metall. Mater.
reader is referred to Kubaschewski and Trans., in press.
Alcock (1979). For techniques of measur- Chase, M. W. (1983), Bull. Alloy Phase Diag. 4, 124.
Chase, M. W., Davies, C. A., Downey, J. R., Frurip,
ing phase diagrams, see Pelton (1996), D. J., McDonald, R. A., Syverud, A. N. (1985),
Raynor (1970), MacChesney and Rosen- JANAF Thermochemical Tables, 3rd ed. New
berg (1970), Buckley (1970) and Hume- York: Amer. Inst. Physics.
Chu, W. F., Rahmel, A. (1979), Rev. High-Temp.
Rothery et al. (1952). Mater. IV, 139.
Degterov, S., Pelton, A. D. (1996), J. Phase Equilib.
17, 488.
Dessureault, Y., Pelton, A. D. (1991), J. Chim. Phys.
1.14 References 88, 1811.
Dessureault, Y., Sangster, J., Pelton, A. D. (1990),
Ageev, N. V. (Ed.) (1959–1978), Phase Diagrams of J. Electrochem. Soc. 137, 2941.
Metallic Systems, Vols. 1–22. Moscow: Acad. Sci. Doman, R. C., Barr, J. B., McNally, R. N., Alper, A.
USSR. M. (1963), J. Am. Ceram. Soc. 46, 313.
Alabyshev, A. F., Morachevskii, A. G. (1957), Z. Dörner, P., Henig, E.-Th., Krieg, H., Lucas, H. L.,
Neorg. Khim. 2, 669. Petzow, G. (1980), Calphad Journal 4, 241.
1.14 References 79
Elliott, R. P. (1965), Constitution of Binary Alloys. Katsura, T., Muan, A. (1964), TMS-AIME 230, 77.
First Supplement. New York: McGraw-Hill. Kellogg, H. H., Basu, S. K. (1960), TMS-AIME 218,
Eriksson, N. G. (1975), Chemical Scripta 8, 100. 70.
Findlay, A. (1951), The Phase Rule, 9th edn., revised Kleinstuber, T. (1961), Ph.D. Thesis, Univ. Munich,
by Campbell, A. N., Smith, N. A. New York: Germany.
Dover. Kohler, F. (1960), Monatsh. Chemie 91, 738.
Flory, P. J. (1941), J. Chem. Phys. 9, 660. Köster, W., Dullenkopf, W. (1936), Z. Metallk. 28,
Flory, P. J. (1942), J. Chem. Phys. 10, 5. 309.
Fowler, R. H., Guggenheim, E. A. (1939), Statistical Kongoli, F., Dessureault, Y., Pelton, A. D. (1998),
Thermodynamics. Cambridge: Cambridge Univ. Metall. and Mat. Trans. 29B, 591.
Press, p. 350. Kubaschewski, O. (1982), Iron-Binary Phase Dia-
Garrels, R. M., Christ, C. L. (1965), Solutions Miner- grams. New York: Springer Verlag.
als and Equilibria. New York: Harper and Row. Kubaschewski, O., Alcock, C. B., (1979), Metallur-
Ginsberg, A. S. (1909), Z. Anorg. Chem. 61, 130. gical Thermochemistry, 5th ed. New York: Per-
Glasser, E. P., Osborn, E. F. (1960), J. Amer. Ceram. gamon Press.
Soc. 43, 136. Kurnakow, N. S., Kusnetzow, A. N. (1907), Z. Anorg.
Guggenheim, E. A. (1935), Proc. Roy. Soc. A 148, Chem. 52, 173.
304. Kunz, M. (2001), in: Phase Transformations in Ma-
Gulbransen, E. A., Jansson, S. A. (1970), Proc. terials: G. Kostorz (Ed.). Weinheim: Wiley-VCH,
Symp. Electrochem. Soc. Fall meeting 1969. New Chap. 10.
York: The Electrochem. Soc., p. 3. Lantratov, M. F., Mikhailova, A. G. (1971), Zh. Prikl.
Gupta, H., Morral, J. E., Nowotny, H. (1986), Scripta Khimii 44, 1778.
Metall. 20, 889. Lee, B.-J., Lee, D. N. (1991), Calphad 15, 293.
Hansen, M. (1958), Constitution of Binary Alloys, Lee, B. J., Lee, D. N. (1992), J. Phase Equilib. 13, 349.
2nd edn. New York: McGraw-Hill. Levin, E. M., Robbins, C. R., McMurdie, H. F. (1964),
Harshe, G., Venkatachalam, S. (1984), J. Metals 36, and 11 supplements to (1996), Phase Diagrams for
34. Ceramists; Columbus, Ohio: Am. Ceramic Soc.
Hauffe, K. (1940), Z. Elektrochem. 46, 348. Lewis, G. N., Randall, M. (1961), Thermodynamics,
Hensel, F. (1979), Adv. Phys. 28, 555. revised by Pitzer, K. S., Brewer, L. New York:
Hillert, M. (1980), Calphad Journal 4, 1. McGraw-Hill.
Hillert, M. (1985), Int. Metall. Rev. 30, 45. Lin, P.-L., Pelton, A. D., Bale, C. W. (1979), J. Am.
Hillert, M. (1997), J. Phase Equilib. 18, 249. Ceram. Soc. 62, 414.
Hillert, M. (1998), Phase Equilibria, Phase Dia- Lipson, H., Wilson, A. J. C. (1940), J. Iron Steel Inst.
grams and Phase Transformations. Cambridge: 142, 122.
Cambridge University Press. Lukas, H. L., Henig, E.-Th., Zimmermann, B.
Hillert, M., Jansson, B., Sundman, B. (1988), Z. (1977), Calphad Journal 1, 225.
Metallk. 79, 81. MacChesney, J. B., Rosenberg, P. E. (1970), in Phase
Holm, J. L. (1965), Acta Chem. Scand. 19, 638. Diagrams – Materials Science and Technology,
Hong, K. C., Kleppa, O. J. (1976), J. Chem. Thermo- Vol. 1: Alper, A. M. (Ed.). New York: Academic,
dyn. 8, 31. Ch. 3.
Huggins, M. L. (1942), Ann. N. Y. Acad. Sci. 43, 1. Maiorova, E. A., Morachevskii, A. G., Kovalenko, S.
Hultgren, R., Desai, P. D., Hawkins, D. T., Gleiser, G. (1976), Elektrokhimiya 12, 313.
M., Kelly, K. K., Wagman, D. D. (1973), Selected Massalski, T. B., Okamoto, H., Subramanian, P. R.,
Values of the Thermodynamic Properties of the Kacprzak, L. (1990), Binary Alloy Phase Dia-
Elements and Binary Alloys. Metals Park, Ohio: grams, 2nd ed. Metals Park, OH: Am. Soc. Metals.
Am. Soc. of Metals. Mathewson, C. H. (1906), Z. Ang. Chem. 50, 180.
Hume-Rothery, W., Christian, J. W., Pearson, W. B. Mills, K. C. (1974), Thermodynamic Data for Inor-
(1952), Metallurgical Equilibrium Diagrams. ganic Sulphides, Selenides and Tellurides. Lon-
London: Inst. Phys. don: Butterworths.
Inden, G. (1982), Bull. Alloy Phase Diag. 2, 412. Miodownik, A. P. (1982), Bull. Alloy Phase Diag. 2,
Inden, G. (2001), in: Phase Transformations in Mate- 406.
rials: G. Kostorz (Ed.). Weinheim: Wiley-VCH, Moffatt, W. B. (1979), The Index to Binary Phase Col-
Chap. 8. lections, Schenectady. N.Y.: General Electric Corp.
Ingraham, T. R., Kellogg, H. H. (1963), TMS-AIME Muan, A., and Osborn, F. (1965), Phase Equilibria
227, 1419. Among Oxides in Steelmaking. Reading, MA.:
Ingraham, T. R., Kerby, R. (1967), Canad. Metall. Addison Wesley.
Quart. 6, 89. Müller-Krumbhaar, H., Kurz, W., Brener, E. (2001),
Ivanov, A. I. (1953), Izv. Sekt. Fiz.-Khim. Anal. Inst. in: Phase Transformations in Materials: G. Kostorz
Obshchei Neorg. Khim. Akad. Nauk SSSR 23, 197. (Ed.). Weinheim: Wiley-VCH, Chap. 2.
Palatnik, L. S., Landau, A. I. (1964), Phase Equilib- Robertson, W. D. (Ed.) (1966), Binary Phase Dia-
ria in Multicomponent Systems. NY: Holt, Rine- grams of Halide Salts, U. S. At. En. Comm. Con-
hart and Winston. tract AT (30-1) 2723, Washington: Clearinghouse
Pehlke, R. D. (1973), Unit Processes of Extractive for Federal Scientific and Techn. Info.
Metallurgy. New York: Elsevier. Robie, R. A., Hemingway, B. S., Fisher, J. R. (1978),
Pei, B., Björkman, B., Sundman, B., Jansson, B. Thermodynamic Properties of Minerals and Re-
(1995), Calphad 19, 1. lated Substances at 298.15 K and 1 Bar Pressure
Pelton, A. D. (1983), in: Physical Metallurgy: Cahn, and at Higher Temperatures. Washington: U.S.
R. W., Haasen, P. (Eds.). Amsterdam: Elsevier, Government Printing Office.
Chap. 7. Roos, G. D. (1916), Z. Anorg. Chem. 94, 329.
Pelton, A. D. (1988 a), Bull. Alloy Phase Diag. 9, 41. Saboungi, M.-L., Blander, M. (1974), High Temp.
Pelton, A. D. (1988 b), Calphad Journal 12, 127. Sci. 6, 37.
Pelton, A. D. (1995), J. Phase Equilib. 16, 501. Saboungi, M.-L., Blander, M. (1975), J. Am. Ceram.
Pelton, A. D. (1996), in: Physical Metallurgy: Cahn, Soc. 58, 1.
R. W., Haasen, P. (Eds.). Amsterdam, North-Hol- Saboungi, M.-L., Herron, S. J., Kumar, R. (1985),
land, Chap. 6. Ber. Bunsenges. Phys. Chem. 89, 375.
Pelton, A. D. (1997), in: Advanced Physical Chemis- Sangster, J., Pelton, A. D. (1987), J. Phys. Chem.
try for Process Metallurgy: Sano, N., Lu, W.-K, Ref. Data 16, 509.
Riboud P. (Eds.). New York, Academic, Chap. 3. Schenk, H., Frohberg, M. G., Nunninghof, R. (1964),
Pelton, A. D., Bale, C. W. (1986), Metall. Trans. Arch. Eisenhütten 35, 269.
A 17, 1057. Schreinemakers, F. A. H. (1915), Proc. K. Akad.
Pelton, A. D., Blander, M. (1984), Proc. AIME Symp. Wetenschappen, Amsterdam (Section of Sciences)
on Molten Salts and Slags. Warrendale, Pa.: The 18, 116.
Metall. Soc. AIME, p. 281. Shatynski, S. R. (1977), Oxidn. Metals 11, 307.
Pelton, A. D., Blander, M. (1986), Metall. Trans. Shunk, F. A. (1969), Constitution of Binary Alloys.
B 17, 805. Second Supplement. New York: McGraw-Hill.
Pelton, A. D., Chartrand, P. (2000), Metall. Mater. Spencer, P. J., Barin, I. (1979), Mater. Eng. Appl. 1,
Trans., in press. 167.
Pelton, A. D., Degterov, S. A., Eriksson, G., Robelin, Stringer, J., Whittle, D. P. (1977), Rev. Int. Htes.
C., Dessureault, Y. (2000), Metall. Mater. Trans., Temp. Réfract. 14, 6.
B 31, 651. Sundman, B., Ågren, J. (1981), J. Phys. Chem. Solids
Pelton, A. D., Schmalzried, H. (1973), Metall. Trans. 42, 297.
4, 1395. Trusler, J. P. M. (1999), in: Chemical Thermodynam-
Pelton, A. D., Schmalzried, H., Sticher, J. (1979), ics: Letcher, T. (Ed.). Abingdon, Oxon, UK:
J. Phys. Chem. Solids 40, 1103. Blackwell Science, Chap. 16.
Pelton, A. D., Thompson, W. T. (1975), Prog. Solid Van Laar, J. J. (1908), Z. Phys. Chem. 63, 216; 64,
State Chem. 10, part 3, 119. 257.
Pelton, A. D., Thompson, W. T., Bale, C. W., Erik- Villars, P., Prince, A., Okamoto, H. (1995), Hand-
son, N. G. (1990), High Temp. Sci. 26, 231. book of Ternary Alloy Phase Diagrams. Metals
Petric, A., Pelton, A. D., Saboungi, M.-L. (1988 a), Park, OH: Am. Soc. Metals.
J. Electrochem. Soc. 135, 2754. Voskresenskaya, N. K. (Ed.) (1970), Handbook of
Petric, A., Pelton, A. D., Saboungi, M.-L. (1988 b), Solid-Liquid Equilibria in Systems of Anhydrous
J. Chem. Phys. 85, 5070. Inorganic Salts, Vol. 1, TT69-55079/1, AEC-tr-
Phillips, B., Muan, A. (1958), J. Am. Ceram. Soc. 41, 69831/3, UC-4, Washington: U.S. At. En. Comm.
448. and Nat’l. Sci. Foundation.
Pilgrim, R. F., Ingraham, T. R. (1967), Canad. Weeks, J. R. Davies, H. A. (1964), AEC Report,
Metall. Quart. 6, 333. Conf. 660712-1, BNL-10372.
Polyakov, V. D. (1940), Izv. Sektora Fiz.-Khim Anal- West, D. R. F. (1965), Ternary Equilibrium Dia-
iza Inst. Obshchei Neorg. Khim. Akad. Nauk SSSR grams. New York: McMillan.
13, 308. Wisniak, J. (1981), Phase Diagrams: A Literature
Prince, A. (1963), Metall. Rev. 8, 213. Source Book, Vols. 1 and 2. New York: Elsevier.
Prince, A. (1966), Alloy Phase Equilibria. Amster- Wisniak, J., Tamir, A. (1978), Mixing and Excess
dam: Elsevier. Thermodynamic Properties: A Literature Source
Raynor, G. V. (1970), in: Physical Metallurgy, 2nd Book. New York: Elsevier.
edn: Cahn, R. W. (Ed.). Amsterdam: North Hol- Wu. P. (1990), Ph. D. thesis, Ecole Polytechnique,
land, Chap. 7. Montreal.
Rhines, F. N. (1956), Phase Diagrams in Metallurgy. Wu, P., Eriksson, G., Pelton, A. D. (1993), J. Am.
New York: McGraw-Hill. Ceram. Soc. 76, 2065.
Ricci, J. E. (1964), in: Molten Salt Chemistry: Yokokowa, H. (1999), J. Phase Equilib. 20, 258.
Blander, M. (Ed.). New York: Interscience, Chap. 4.

CH 1

Uploaded by

Copyright:

Available Formats

CH 1

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

CH 1

Uploaded by

Copyright:

Available Formats

1 Thermodynamics and Phase Diagrams of Materials

Centre de Recherche en Calcul Thermochimique, École Polytechnique,

List of Symbols and Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

Phase Transformations in Materials. Edited by Gernot Kostorz

1.6 Application of Thermodynamics to Phase Diagram Analysis . . . . . . 34

List of Symbols and Abbreviations

b.c.c. body-centered cubic

1.1 Introduction ergy. General computer programs are avail-

Predominance diagrams are a particu- Hence, along this line:

Figure 1-1. Predomi-

1.3.2 Ellingham Diagrams

Figure 1-3. Predominance diagram

conditions of two- and three-phase equilib- Predominance diagrams have found

nent i, and the derivative is taken with all

1.4.3 Tangent Construction 1.4.4 Gibbs–Duhem Equation

giE = R T ln g i (1-58) The temperature – composition (T – X )

Figure 1-5. Phase dia-

Figure 1-6. Ge – Si phase diagram at

For example, the enthalpy of vaporization

A similar expression can be derived for the

1.5.5 Miscibility Gaps

If gE > 0, then the solution is thermody-

Figure 1-12. Phase diagram

This reaction occurs isothermally at 744 K

Such a description is justifiable only if g solvent ≈ 1

tectic reaction is l + s2 Æ s1 ; (ii) syntectics

1.6 Application of important step in the development of data-

following expansions for the partial excess

Figure 1-20. Sodium ac-

Eq. (1-106) reproduces the calorimetric

1.6.4 Calculation of Metastable Phase and Cd – Bi are found. Lines of constant

Figure 1-23. Perspective view of ternary space

Figure 1-24. Projection of

of a liquid of composition q on the liquidus is somewhat exaggerated in Fig. 1-25 for

liquid of composition c. To understand this

Another important rule of construction,

1.7.5 Ternary Isopleths

Certain geometrical rules apply to iso- 1.7.5.1 Quasi-Binary Phase Diagrams

Figure 1-30. Section of

1.7.7 Nomenclature for Invariant + d) is a 1 Æ 3 reaction; and so on. The ter-

ternary but is an isopleth of the four-com-

For further discussion and references,

1.8 Phase Diagrams

Figure 1-33. Correspond-

Figure 1-35. Phase diagram for the

axis when applied to the magnetite phase?

Figure 1-39. An isobaric binary T – X

Figure 1-41. A node in a true phase diagram sec-

selected extensive variables we form the

m i results in a progressive expansion and Na2O system we would choose as axes

stability criterion: values. For example, in the predominance

1.10.1.5 The Compound Energy multicomponent solutions. It also provides

Figure 1-44. Phase dia-

particle Li +. The F – ion is not “foreign”. or molecule is Zi Xi . Hence, mass balance

shown in Fig. 1-14 is thermodynamically model more flexible by permitting the Z i to

Above this temperature it is not. The trans-

1.11 Calculation of Ternary Phase

Kohler (1960): tal effort involved in determining multi-

1.12 Minimization of Gibbs solution phases a, b, g, … . The total Gibbs