PHASE EQUILIBRIA IN CONDENSED SYSTEMS

0. KUBASCREWSKI and I. BARIN

Lehrstuhl für Metallurgie der Kernbrennstoffe und Theoretische Hüttenkunde

der Rheinisch- Westjiilischen Technischen Hochschule Aachen, GFR

ABSTRACT
The evaluation of metallurgical equilibrium diagrams from thermochemical
data in contrast to the use of conventional methods is <iemonstrated with a
nurober of examples: first with hinary alloy systems, i.e. eutectic systems
(Ga-Zn, Ga-In), solidus-liquidus gaps (Mg-Cd, Ni-Co, Ni-Cu, Fe-Cr, Ntr.-Ta),
miscihility gaps (Cr-Mo, Ni-Cu, Bi-Zn) and systems exhibiting intermediary
phases (Fe-U, Fe-V, Fe-Cr), and then with ternary and quaternary systems,
i.e. solution-forming systems (Fe-Co-Ni, Fe-Cr-Ni, Fe-Cr- V, Au- Pd- Pt,
Pb-Sn-Cd, Pb-Cd·-Zn - Sn) and compound-forming systems (Mg-0-S,
Fe-0-Cl). The practical significance of this new development is discussed.

Although much of this paper is devoted to metallurgical equilibria, the

underlying principles are equally applicable to other substances, such as
oxide, salt or even organic systems. The emphasis is on relatively high
temperatures at which kinetic checks are less predominant and the knowledge
of chemical equilibrium is thus more important.
Metallurgical equilibria are commonly described by temperature versus
concentration plots at normal pressure, known as phase diagrams or equili-
brium diagrams*, which provide fundamental information for the metallur-
gist, whether he be chemically, physically or mechanicalJy oriented. Good
equilibrium diagrams are equally irnportant in other fields, such as those of
oxide-oxide systems for the ceramist.
The standard of conventional phase diagram work in metallurgy was set
in the years 1897-1905 by Heycock and Neville, whose degree of accuracy
has seldom been surpassed in the temperature range 20-1000°C. In Germany,
the evolution of this field is inextricably connected with the name of Gustav
Tammann.
The principles of the derivation of phase boundaries from Gibbs energies
were clearly indicated in papers published by van Laar around 1908. On the
other band, since phase boundaries in chemical systems represent equilibria.
chemical activities may be evaluated from phase diagrams. One example of
many is Rey's calculation 1 ofthe activity of silica and metal oxides in silicate
*The author makes a distinction hetween 'phase diagrams' obtained by conventional experi-
mental methods and ·cquilibrium diagrams· calculated from thermochemical data.

469
 0. KUBASCREWSKI AND I. BARIN
melts. The use of independently determined thermochemical data for the
evaluation of equilibrium diagrams is more recent. Using estimated entropies
of fusion, Hauffe and Wagner 2 calculated liquidus curves in metallic systems
exhibiting compound formation as early as 1940, but truly the first evaluation
of phase boundaries based on the experimental thermochemical data is
Wagner's calculation 3 of the solidus-liquidus gap in the system silver-gold.
Much of the recent work of the present author and his associates has been
devoted to the systematic development of this field.

AG r

"-V
\dj~
___________ \ ______ ; ____ ___ ___ _
0

-_
...
E~' Ö

N_.
Figure ]. Gibbs energy versus concentration for a hypothetical binary system at a given
temperature and pressure.

There are two major aspects of the evaluation of phase boundaries from
thermochemical data in cantrast to the conventional methods. Thermo-
cbemical measurements may be carried out at kinetically favourable
temperature8 and the results used to calculate phase boundaries at any
temperature including those at which equilibration may take months or
years. The lmowledge of t he exact position of such phase boundaries is, of
course, of the greatest practical significance because the use of certain alloys
in plant may extend over years at these temperatures and success or peernature
failure depends on whether the alloy was originally in a stable or metastable
state.
The second advantageofthe thermochemical over the conventional method
is only now becoming apparent, namely the possibility of extrapolating the
470
 PHASE EQUILIBRIA IN CONDENSED SYSTEMS

thermochemical properties of binary systems into multi~component ranges

with sufficient accuracy to permit calculation of the phase boundaries in
ternary, quaternary and more involved systems. This does not yet eliminatethe
need for the application of conventional methods but substantially reduces
the number of such tests.
Systematic experimental sturlies of the thermochemical properties of
alloys were first carried out for the sheer pleasure of breaking new ground, as
so often in fundamental science. This attitudestill prevails in some laboratories
concerned with the measurement of such data. All too often is the most impor-
tant aspect, the exact relationship ofthe results with the equilibrium diagram,
neglected. It may be repeated that good phase boundaries are vital thermo-
chemical information. lnvestigators of thermochemical data are to be
requested to assess their results critically together with alJ the available
information on pbase diagrams 4 and earlier thermochemical work 5 . If this
is done conscientiously, much unnecessary dupJication of work may be
avoided.
The pririciple of the evaluation of phase boundaries from Gibbs energies
may be seen from Figure 1, in which the Gibbs energies of formation of

Points calculated by fixing

3 points on the liquidus (N•2)
900 shown by 0

Mean experimentolly
determined boundories

800
:::.!.

1:11

...
L..
::)

0
1,.
bl
Q.
e '
~ '' '
700 ' '\
\
'' ''
............. '~

600

Mg Cd

Figure 2. Calculated and experimental solidus and liquidus curves in the system magnesium-
cadmium.
471
 0. KUBASCREWSKI AND LBARIN
various phases are plotted against concentration at a given temperature and
pressure in a hypothetical binary system. Since it is always the phase or
mixture of phases with the lowest Gibbs energy that is stable, tangents to the
lowest AG-T curves have been drawn in Figure 1, the points of contact giving
the phase boundaries at that temperature and pressure. Thus, Figure 1
illustrates the graphical method of evaluation. Alternatively, the Gibbs energy
curves may be expre~sed by polynomials and the tangents obtained by calcu-
lation invoking the help of a computer. The latter method is not necessarily
superior, as it may be difficult to describe the refinements in the thermo-
chemical data by polynomials, as has been demonstrated with the solidus-
liquidus gap in the system magnesium-cadmium by Gale and Davis 6 , as
shown in Figure 2.
By now, quite a large number of equilibrium diagrams of binary systems
have been evaluated and consistent sets of thermochemical data produced.

EUTECTIC SYSTEMS
If, in a simple eutectic system, the mutual solid solubilities are sufficiently
small to be disregarded, the liquidus temperatures (~iq) may be calcu1ated
from the partial heats (Al() and entropies (AS;) of solution of the component
i at mole fraction Nh
~iq = (Afl; + L;)/(AS; + ai) (1)

Ga-Zn
200
0 Pushin ef al.
0 Valenti and Florani
!::. Heumann and Predel
Calculated
X Present work
ISO ~ Zn crystals found

ou
~100
+J
0
'-
C'ia1
a..
E
~
so
t
~0

0 tO 20 30
At. 0
/o Zr'\
Figure 3. Calculated liquidus curve in the system galliurn- zinc and experimentalliquidus points.
472
 PHASE EQUILIBRIA IN CONDENSED SYSTEMS
where Li and ai are respectively the heat and entropy of fusion of component
i. Experience has shown that liquidus curves thus calculated from good
thermochemical data mostly agree very weil with those measured by thermal
analysis. This observation is important because it means that experimental
liquidus curves may be incorporated in the thermodynamic evaluation of a
metallic system including other than just eutectic systems.
However, there are exceptions, namely systems with semipmetals having
a pronounced tendency to supercooling and therefore to diffusional delays.

-E
65 In

0·5
X
~ 0-4
E
0
.... _
,..J
0 0·3
I
----x...a o
~0·2~
O·IL t.
6. Ä
--x--- x..___~
0

--..Jl_OY
0
01Ga 0·1
I
0·2
I I I I
0 ·3 0·4 0·5 0·6 0·7
-0 X
0·8 0·9 In
Nyn-
Figure 4. Partial excess entropies of solution of liquid indium
in liquid indium- gallium alloys
calculated from the heats of mixing and liquidus points obtained by thermal analysis.

In such cases the thermochemical method has proved tobe superior to the
conventional methods. An example is the system gallium-zinc where the
calculated liquidus temperatures at high gallium contents had been found
by Ku baschewski and Chart 7 to be up to 20 oc higher than those obtained by
normal thermal analysis, as may be seen from Figure 3. Only after prolonged
annealing (7 days) above the liquidus and the application of particularly
slow cooling rateswas the calculated curve confirmed (crosses in Figure 3).
In similar cases it is advisable to obtain the eutectic composition by
combining partial heats of mixing and thermal liquidus points and extra-
polating the resulting excess entropies of solution into the concentration
range rich in semi-metal, as has been demonstrated by Hayes and
Kubascbewski 8 with gallium- indium aiJoys. Figure 4 shows how faulty
thermal liquidus points lead to increasing deviations of the partial excess
entropies from the mean curve as the gallium side is approached.

SOLIDUS-UQUIDUS GAPS
H two metals are mutually completely soluble in the solid and liquid states.
the solutions are separated by a solidus-liquidus gap. Mak.ing a nurober of
simplifying assumptions, namely that both the solid and liquid solutions are
regu1ar and have the same heat of mixiog and that the width of tbe gap in
473
 0. KUBASCHEWSKI AND I. BARIN
terms of temperature (L\fJ) is required only for approximately equiatomic
compositions, Wagner 3 derived the following equation :

L\0 NiLl + N~L2 ( bT )z if Ntz ~ Ns2 (2)

. = RT 2 /N 11 N~ + TL\H~ bNh
where the superscripts, s and 1, refer to the solid and liquid phases, res~
pectively, and the subscripts, 1 and 2, to the two component metals. L\Hh is
the second derivative of the excess Gibbs energy with respect to the mole
fraction of the liquid phase and (bT /bNi} the slope of the liquidus curve.
Where the solidus-Iiquidus gap is small, equation (2) despite its limitations
is quite useful. In the system cobalt-nickel, all the thermodynamic excess
functions were found tobe zero within tbe experimental errors 9 . In equation
(2), the T L\Hh term is therefore zero and (bT jbN~) obtained from the difference
in melting points ofthe component metals. Near-ideal behaviour in the system
Co-Ni is compatible with a calculated solidus-liquidus gap of L\fJ = 0.26 oc
for the equiatomic alloy. Another metallic system likely to exhibit near-ideality
is the system Nb-Ta. Using equation (2), tbe known melting points and esti -
mated heats of fusion (ar = 1.95 cal g-atom - t K), one finds L\() ~ 21 oc for
the equiatomic alloy. Naturally, this result will be difficult to verify experi-
mentally because of the very high melting points.
Where the solidus-liquidus gapsarewider and are required for the whole
range of compositions, Wagner's equation is obviously inadequate. An
adequate equation to account for deviations from regularity, differences in

-500

-1000

llGsol

-1500

Cu 0·2 0·4 0·6 0·8 Ni

NNi
Figure 5. Gibbs energies of mixing of, respectively, liquid and solid copper-nickel alloys from
liquid copper and solid nicket.
474
 PHASE EQUILIBRIA IN CONDENSED SYSTEMS

excess functions of solid and liquid and compositions otber than equiatomic
would be very complex. Tbe mode of evaluation frequently employed in such
a case is to assess the thermochemical properties of the solid and liquid
so1utions across the whole range of compositions, to calculate the integral
Gibbs energies at intervals of, say, 5 at.% for a series of temperatures with
due regard to the standard states of the components, to plot the Gibbs ener-
gies against composition and to draw the tangent to the two curves at temp-
erature T, as exemplified by Figure 5 for the system copper-nickcP 0 at a
selected temperature of 1250 ac. The tangent touches the two curves at the
solidus and Iiquidus compositions. In the interval, the solid-liquid mixture
has a lower Gibbs energy than either of the homogeneaus solutions.

1400

1300

1200
u
o .. 1100
~
:;,
-e1:11 1ooo - Calculated
0. - - - Experimental
E
~ 400

300

200 lmmiscibility region

IO~u 10 20 30 40 50 60 70 80 90 Ni
At. 0/o Ni-
Figure 6. Equilibrium diagram of the system copper-nickel.

In Figure 6, the calculated solidus and liquidus curves are represented by

the fulllines whereas the dotted solidus curve represents the results ofthermal
analyses. Meanwhile the calculated solidus temperatures have been con-
firmed by differential analyses of thoroughly annealed alloys by Predel and
Mobs 11 . This example illustrates the point made earlier: calculated solidus
curves often lie substantially higher than the experimental ones owing mostly
to the shortage of time allowed for equilibration during thermal analysis. This
shortcoming of conventional methods is likely to be found to apply to many
more 'established' solidus curves than is generally appreciated. An earlier
example where this was found to be true is the chromium-iron system 1 oa.
In view of the greater reliability of the experimental liquidus in contrast
to the solidus temperatures, it is recommended to adjust the thermochemical
475
 0. KUBASCREWSKI AND I. BARIN
data for the solid and liquid solutions within the experimental tolerance
so as to represent the liquidus curves obtained by conventional methods
exactly rather than to rely entirely on the thermochemical results which are
often so scattered as to prohibit an accurate assessment. The chromium-iron
system is a case in point. For illustration, the numerous measurements of the
partial Gibbs energy of chromium in liquid alloys by different investigators
have been converted with an estimated heat of solution into excess entropies.
3·0 , -·-------.-----<0·--~~----.------.---- ~-~ ---,,-----------..-
+
t
I· 0
I
~2·or :
E I 0
•
0 I 8
ö 1'0 0

:." r·
::::-
0 '
o X
~
V

w6.
IV)
1·0 tG

<l I
l

Figure 7. Partial excess entropies of solution of liquid chromium in liquid chromium-iron

alloys. (The fullline represents the results of a calculation from other pertinent thermochemical
data, which is compared with the experimental results of various authors after combination
with the accepted heat of mixing.)

Figure 7 shows the scatter of the experimental thermochemical results from

five different sources whereas the ful11ine is compatible with the Jiquidus curve
and the thermochemical properties ofthe solid solutions. Actua11y, none ofthe
results of the Gibbs energy measurements has been used in the evaluation of
~S2r by Müller and Kubaschewski
100
•

MISCIBILITY GAPS
Jf two metals form solid solutions with a positive heat of mixing, the occur-
rence of a miscibility gap at some lower temperature is to be expected, the
mutual solubilities approaching zero as 0 K is approached. The reason for
this behaviour follows from the Gibbs-Helmholtz equation and the charac-
teristic shapes of the ~H versus N and TI!!S versus N curves. Whereas the
~H versus N curves tend towards a parabolic shape, the T ~S versus N
curves approach N = 0 and N = 1 asymptotically. Figure 8 shows the
resulting AG versus N curve for the chromium-molybdenum system at
800°C taking the shape of an inverted camel-back 6 • The tangent touches the
11G versus N curve at the compositions of the phase boundary at 800 oc.
This example also demonstrates the superiority of the thermochemical
method of phase boundary determination because a 60% Cr alloy bad to be
476
 PHASE EQUILIBIRA IN CONDENSED SYSTEMS

annealed at 800 °C for over a year to prove incontrovertibly that it de-

composed into aMo-rich and a Cr-rich solution 12, whereas technical alloys
may have to withstand high temperatures for Ionger periods than that.
At Iower temperatures, the progress of alloy deterioration may become so
slow that a calculated miscibility gap would be meaningless from the techno-
logical point of view. The miscibility gap in the system copper-nickel which
100 ~----------------------------------~400

,_
II '\

I
I _,,\
50 I 200
''\
'\
\\
0 \
-
\
E
0u \ Mo
0
>-.
0'1
L.
C)l
c:
C)l

C)l
C)l
'-
u.
...._., -so -200
0
(!)
<l
\
\
\
-100
2 = - 400
3 =
4 =

-ISO ~----------~------~------------~~ -~0

Figure 8. Calculated Gibbs energies of chromium-molybdenum alloys at 1073 K using different
analytical approximations (2, 3, 4 : upper limits of the summation index in the relationsbips for
the calculation of the Gibbs energy versus atomic fraction).

closes at 322 oc (Figure 5) is such a hypothetical example because the period

after which decomposition would become noticeable may be estimated to be
very far in excess of the life~time of any technical Cu-Ni alloy.
The investigation by thermal analysis ofthe phase boundary of a miscibility
gap even in the liquid state seems tobe subject~ on occasio~ to considerable
errors owing to the small differences in Gibbs energy, that is, small 'driving
forces', as exemplifie-d by the system Bi-Zn (Figure 9) where admittedly one
of the components is a semi-metal. In this case again, the thermochemical
properties obtained at higher. i.e. kinetically more favourable, temperatures
allow a definite decision to be made on which of the controversial results to
accept.
477
 0. KUBASCREWSKI AND I. BARlN
Bi -Zn

A Spring ond Romanoff !i_ _

X Hass and Jellinek -x Ax
0 Seith and Johnen
700
Calculated

X
600
Ä
u X
0

..,~ 500
2~
0.
E
~
r;. 400

300

0·2 0·4 0·6 0·8

Nzn-
Figure 9. Calculated equilibrium diagram of the system bismuth-zinc and experimental points.

SYSTEMS EXHIBITING INTERMEDIARY PHASES

Systems forming intermetallic compounds of narrow composition ranges
can be fairly easily treated by the thermochemical method because it is
mostly the liquidus temperatures and the eutectic or peritectic temperatures
and compositions that are required. The shape ofthe liquidus curves depends
mainly on the heats and entropies of fusion of the compounds and the
component metals. If necessary, entropies of fusion can be fairly well
estimated and if the meJting points are known, the heats of fusion are also
obtained. An early approach to the problern by Hauffe and Wagner 2 has
already been mentioned.
Figure 10 illustrates, for a selected temperature, the graphical evaluation
of the phase boundaries in the uranium-iron system. The entropies of fusion
of UFe 2 and U 6 Fe were estimated by Kubaschewski 13 and their melting
points adopted from the work of Grogan and Clews 14. The resulting equi1i-
brium diagram (Figure 11) is in good agreement with the otherwise indepen-
dent in vestigation of Grogan and Clews.
As an example ~ta system of a more involved nature and requiring more
experimental thermochemical effort, the ferrovanadium aJloys have been
478
 PHASE EQUILIBRIA IN CONDENSED SYSTEMS
+1000 ~-------------------------,

I
I
I
I
I

-2000
e
0
.....
0
I
CTI
AGiiq.

0
-3000
0

-4000

-5000 /
I
_ __!

u 20 40 60 80 Fe
At. 0 /o Fe
Figure 10. Gibbs energies of formation of liquid uranium-iron alloys and of the phase UFe 2
from the solid metals at 1328 K.

G Grogan
- · - Calculate:d

1400

u
0
QJ
I-
:;)

e
QJ
Q.

~ 1000
I-

u 20 40 60 80 Fe
Figure 11. Calculated equilibrium diagram and experimentalpoints for the system uranium-iron.
479
 0 . KUBASCHEWSKI AND I. BARIN

selected. It has been mentioned earlier that systems of complete mutual

solubility below the melting points but a positive heat of mixing are likely to
form a miscibility gap at lower temperature. If on the other band the heat of
mixing is negative, one would expect formation of some type of superstructure
at lower temperature. This is the case in the system Fe-V which forms a (J

1500

- ... - Experiment 15
- Calculated
1300

1100
u0
~
....;::,0
'- 900 Magnetic
~ trans.
,.,..,,
E ,, ...

~ ;"
t1
700

300

Fe 0·2 0·4 0·6 0·8 V

xv -· -
Figure 12. Calculated and experimental phase boundaries for the cx and cr phases in the system
iron-vanadium.

phase of fairly wide extension below about 1160 °C. Spencer and Putland 15
measured the heat of mixing of disordered rt alloys and heats and tempera-
tures of the transformation (J--+ r:x and adopted Gibbs energies of formation
for the rt alloys from torsion-effusion measurements of Myles and Aldred 16 .
With this information they calculated the r:~./rx + (Jj(J boundaries which are
shown as full lines in Figure 12. Agreement with the experimental phase
boundaries (dotted lines) is quite good in this case. This is by no means always
so: in particular does one often find solid solubilities at the lower temperatures
480
 PHASE EQUlLIBRIA IN CONDENSED SYSTEMS

smal1er than with conventional methods and some indication of this may be
seen in Figure 12 on the vanadium-rieb side. The discrepancies are again due
to diffusional delays in equilibration to which the thermochemical method is
much less prone.
The system chromium-iron presents a somewhat special case in that it
exhibits positive heats of mixing in the diserdered state but forms a (J phase
1900 ..--- or--- .---.---.----..--·

1800

1700
Liquid

1600

J
- - - Calculated
...... . Metastoble equilibrium d/ot'
820°

700
0(+0' oC'+O'

600

500

400

3oo ~~~--~~~-~-~~~--~~---~~
Fe 0·2 0·4 0·6 0·8 Cr
X Cr

Figure 13. Calculated equilibrium diagram ofthe system chromium-iron.

which is also endothermic. In such a case one must expect the ordered
structure to exist only over a limited temperature range below which a
miscibility gap appears. This was indeed found to be the case in the Cr-Fe
system 100 with practical consequences that will not be elaborated here
(Figure 13).

KAUFMAN'S APPROACH
Kaufman in Cambridge, Mass., has also concerned hirnself with the
derivation of equilibrium diagrams by thermochemical reasoning for many
481
 0. K UBASCHEWSKI AND I. BAR IN

years. An earlier approach is particularly interesting. He quite rightly argued

that it would take a long time to provide the experimental thermochemical
data needed for the full and accurate elaboration ofthe important metallurgi-
cal equilibrium diagrams and that short-cuts accepting a loss of accuracy were
desirable. Kaufmanfirst showed (1959) that the difference in Gibbs energy
between crystallographic forms of a metal made an important contribution
in controlling a11oying behaviour, phase transformations and stacking fault

3400

L
3000
~

~ 2600
b ß )--~------~:::;::='~~,.._
'-
"'
0..
: 2200
.....

1800

1400 .

I0 0 0 '-....l--_....__;.__...._----'-_....___-- L.....---'-...J..-._J

w so Rh
F'igure 14. Experimental phasediagram of the system tungsten-rhenium (thin lines) and tentative
calculation of the phase boundaries by a simplified method after Kaufman (thick lines).

energy of the metal over a wide range of temperature and pressure. He then
set out to obtain relative lattice stability information conceming the body-
centred cubic, hexagonal close-packed and face-centred cubic configurations
of metals in various ways.
In certain cases the energy differences are known because the metals are
stable in different modifications at some temperature andjor pressure and
can be determined directly. In other cases, the energy differences can be
obtained by extrapolation from binary systems. The missing data were
estimated. In a review paper by Kaufman 17, the relative phase stabilities of
the transition metals were surveyed and numerous binary equilibrium
482
 PHASE EQUILIBRIA IN CONDENSED SYSTEMS
diagrams computed, considering only the b.c.c., f.c.c., h.c.p. and liquid phases.
This was done by assuming that similar modifications formed ideal solutions
tbermochemically. As one example of many, the experimental and computed
phase diagram of the system tungsten- rhodium is shown in Figure 14.
Naturally, deviations are substantial since the approach deliberately dis-
regards other contributions to the energy of mixing, such as the strain energy
(size factor), vibrational, magnetic and electronic terms, but then (as Dr.
Kaufman pointed out to the present authors) 'a start has to be made
somewhere '.
Meanwhile, as anticipated, Kaufman has been using more sophisticated
n1ethods of approach including experimental thermochemical evidence. His
review paper 18 contributed to the Metallurgical Chemistry Symposium at
Brunel University is recommended for perusal.

TERNARY AND QUATERNARY SOLUTIONS

The second advantage of the thermochemical method of phase boundary
evaluation pointed out earlier is the potential application to the extrapola~
tion of the binary data into multi-component ranges. When the thermo-
chemical properties of ternary systems are available, the mode of evaluation
of phase boundaries is in principle the same as that indicated by Figure 1,
only that the tangents become tangent planes. For quaternary systems, the
evaluation can no Ionger be envisaged geometrically but offers no great
problern to a mathematical treatment provided a computer is available.
A nurober of equations have been proposed for deriving temary properties
from binary data for sing1e-phase systems. Since the bond mechanism of
metallic solutions is complicated, these equations are essentially empirical,
based on the regular solution model or the quasi-chemical approximation.
They have been surveyed by Ansara 19 whose group under Professor Bonnier
at the University of Grenoble has also made substantial contributions in the
field under review. Spencer and Hayes 20 have examined four of these equa-
tions and compared the heats and excess Gibbs energies, respectively,
calculated by these methods with experimental temary data.
Kohler 21 suggested that excess Gibbs energies in a ternary solution
A-B-C, in which the atomic fractions are N A• NB and Ne, may be expressed
in binary terms by the equation:
~GE= (1 - N.J 2 [L\GLcJNu/Nc + (1 ~ Ns) 2 [ÖGÄ-c]NA/Nc
+ (1 - N c) [ßG~-B]NA/Na
2
(3)
the binary compositions, at wbich the binary ~GE terms are taken, being
shown in Figure 15.
Toop 22 proposed the following equation:
E [ NA E . Ne E ] · 2[
~G = 1 -NB ~GA-B+ 1 -NB ~GB-CJNa
E ]
+ (1- Nß} ßGA-C NA/Ne (4)

The binary terms are shown in Figure 16. Both the Kohlerand Toop equations
may be derived from a rigorous equation given by Darken 23 provided the
ternary system and its constituent binaries exhibit regular solution behaviour.
483
 0. KUBASCHEWSKI AND I. BARIN

Composition NA N 8 N c

A c

Figure 15. Illustration of the binary terms used in the application of the Kohler equation.

Composition NA Ne Nc

~Gi-e] N 13

A c

Figure ·16. Illustration of the binary terrns used in the application of the Toop and Bonnier
equations.
484
 PHASE EQUILIBRIA IN CONDENSED SYSTEMS

Bonnierand Caboz's equation 24 takes the form

~G =
E [ NA ~GA-B+
j _Na E . Ne
1 _Na ~GB-c No+ (1 - Ns) ~GA-C NA/Ne
E E ] J ·[ (5)

where the binary Gibbs energy terms are the same as those in the Toop
equation (Figure 16).
The Margu1es equation 25 differs from the previous three in that it contains
a ternary term whereas the others contain only binary terms. There are
seven constants: two for each constituent binary plus one for the ternary:
~GE= NAN 8 [NAKBA + N 8 KA 8 ] + NANc[NAKcA + NcKAc]
+ NaNc[NaKca + NcKac] + NANaNcKABc (6)
For low-affinity solutions, these methods mostly give similar values and
agree quite we11 with the experimental ones, often within the experimental
accuracy. For high-affinity systems deviations are larger. Figure 17 shows

Co

0·2 -155 0·8

X
,........ -92
'):":,
1 0·4 - 260 . , ,
.,,,
0·6 \
+
-\4. 61
-340 .,,, ~
X . ·,,
X /~"'-- .......... , X ~'
0·6 -2ao ,~ . ..-·-."'-.'-~sa·~ o·4
,(' ""'·\._ ~
/ ., - s2ox /-fh2_0'
. -. ' ~ "\.. \
/ .I
I.

y·
/ .,.,... . '\.
.........'·~, ,, '\· ' \ \
0·8 ·~-200 I I /'X . ''- ~ \ 0·2
X~'X ,.. / . ~~ -679 . -634 ~ ,.~ \\
186ll I 'I ·.· -800 '\ \"
- I - 332 ,. 1 /1 -774 X,_~,:::--:.:::..~ '\\ ~\
x. .
96 ~ , .'1 ~.';.t' ._.,..'- • \'
130 ., (/ ~ '· . \
-X I ':>,..\"\ \,
Ftl 0·2 0·4 0·6 0·8 Ni
X Ni ----........
x Exptlrimtlntal
- - .... - Kohltlr tlquation
-·-·- Bonnter (const. Ni ) equation

Figure 17. Iso-enthalpies of formation of solid iron-cobalt-nickel alloys at 1473 K.

485
 0 . KUBASCREWSKI AND I. BARIN
as an example the experimental heats of formation (in place of the excess
Gibbs energies) of iron-cobalt-nickel alloys in comparison with the values
calculated by means of the Kohl er and Bonnier equations 26 . However, it is
difficult at the present stage to decide about the relative merits of the various
modes of extrapolation and further experimental values must be obtained
for a systematic survey. Ansara 19 who has made a thorough study of the
subject seems to give precedence to the Kohler equation. The calculated and
measured excess Gibbs energies of the system Pb-Cd-Sn are shown in
Fiqure 18. Cd

I
I I

I
I
I \
\
\
I
\+
('()-

0·5

I
I
I
I
I
I

Pb

Figure 18. Calculated and experimental excess Gibbs energies for liquid Iead-tin-cadmium
alloys.

As mentioned earlier, some thermochemical information is required for

the temary alloys in addition to complete sets of values for the binary
border systems, when the Margules equation is applied. An example of its
use is the calculation of the thermochemical properties of, and the extent of
the miscibility gap in, the system gold-palladium-platinum by Kubaschewsk.i
and Counsell 27 • From measured values, the excess entropy and heat of "'
mixing were expressed by the following equations:
~HEjcal (g-atom) - 1 = N AuNPd( - 11141 N Au - . 3065NPd)

+ NPdNPt( -5234Npd - 3021Nrt) + NPtNAu(5691NPt - 634N Au)

+ 3036N AuNPdNPt (7)
486
 PHASE EQUILIBRIA IN CONDENSED SYSTEMS

ßSEjcal(Kg-atom) - 1 = -3NAuNPd - 2.9 N~dNPt- 1.7 NpdN~t

- 0.9 N~tN Au - 3.5 NPtNÄu + 9 N AuNPdNPt (8)
The resulting boundaries of the ternary miscibility gap are compared at two
temperatures with the experimental boundaries after Raub and Wörwag 28
in Figure /9.

.... 40 ~. 0
Q 60
0
~ ..0
......
"';(
"' o-
'
I '
ao '/ ' \
\
\
- 20
\
\
\
I
I
I

Pt
At. O/o -... Au

Figure /9. Calculated (dashed) and experimental (full) phase boundaries of the miscibility gap
in the system gold-palladium- platinum at two temperatures.

From the practical point of view perhaps the most important ternary
system in metallurgy is that of the iron---chromium-nickel alloys. Figure 20
shows the calculated and experimental austenite-ferrite equilibria 29 at
1277 oc, that is, above the stability range of the a phase. Phase boundaries
in the latter range may be computed if estimates of the thermochemical
properties in the Fe-Ni and Ni- Cr systems can be provided, those for the
Fe--Cr system being well known. Kubaschewsld has indeed made an earlier
attempt to estimate the phase boundaries at 1277 oc by assuming that the
u phase in the Ni-Cr system is just stable below 700 oc at 75 at.% Ni.
Kaufman, in some other context, assumed cr-Ni-er tobe just metastable, but
the effect on phase boundary calculations in the temary system should not be
great and further experimental evidence is to be awaited.
However, Spencer and Counsel1 30 have calculated the r:t/u relations in a
similar system, Fe-Cr-V, at temperatures of 427, 700 and 900 ~C from
assessed thermochemical data for the binaries. Only the values for cr-Cr-V
bad to be estimated since this phase has never been observed and must
therefore be metastable. First, the heats and entropies of the r:J.-(J trans-
formation in pure chromium and vanadium were estimated from data in the
Fe-V and Fe-Cr systems. (This is of course plain formalism because the pure
metals do not and cannot exist in the (J form.) Tbe phase boundary cal-
culations were then carried out using the method described by Counsell,
Lees and Spencer 29 . As an example the phase relationshlps at 700 oc are
reproduced in Figure 21. There are certain discrepancies compared with the
experimental results of Martens and Duwez 3 \ in particular with regard to
487
 0 . KUBASCREWSKI AND I. BARIN

the relative positions of the tie-Iines. Errors in the calculated equilibria could
arise from the method used to fit the experimental data and to sum the
three binary expressions for calculations of temary Gibbs energies. However,
Martens and Duwez experienced difficulties in locating the phase boundaries
and tie-lines, which is not surprising considering the relatively low tempera-
tures; moreover, in the binary Fe-Cr system their r:t + a region is not in good

Cr

1550 K ----- Jenkins et al.

0·8 - - Calculated

0·2

F'c 0·2 0 ·6 0·8 Ni

XNi----e>
Figure 20. Calculated and experimental phase boundaries of the austenite- ferrite equilibrium
in the system iron-<:hromium- nickel at 1550 K.

agreement with the assessed boundaries. Considering all this. the agreement
between the two sets of results is fair and provides a further i11ustration of the
potential of thermochemical calculations of phase boundaries in ternary
systems using only binary alloy data.
Next the method should be extended to quatemary systems. An industrially
important problern of this kind would be the effect of fourth elements on the
austenite-ferrite equilibrium in iron- nickel-chromium alloys mentioned
above, but only one attempt to calculate phase boundaries in a fourth order
system has been made, namely those of the miscibility gap in liquid Pb--Cd-
Zn- Sn alloys by Counsell, Lees and Spencer 29 . The method has been
described by the authors as 'the minimization of the Gibbs energy', the
computational method chosen being that described by Neiderand Mead 32 .
488
 PHASE EQUILIBRIA IN CONDENSED SYSTEMS

Cr
Colculated and Ouwez 700°C
Mortens

0 ·2 - 0·8

0·6

\ X
0·4 Cr

0·8 0·2

Fe '------...'.1..--""-'-_.__~__"_- ·--=--"-~==c:IL-_y__· _ _ _____, V

0·2 0·4 0·6 0·8
Xv--
Fi~111re 21. Calculated and experimental phase boundaries of the cx-<r equilibrium in the system
iron-chromium-vanadium at 973 K.

N sn = O·t

T = 714 K

Pb Cd
Figure 22. Section of the quaternary system lead-cadmium-zinc- tin at 10 at.% Sn. (The full
curve and the tie-lines ~how the calculated. and the dashed curve the experimental two..phase
region at 714 K.)
489
 0. KUBASCREWSKI AND I. BARIN

In Figure 22, the computed and experimental phase boundary of the misci-
bility gap at 441 oc and 10 at.% Sn are shown for the sake of illustration.
This also raises the question of data representation in bigher order
systems, which is actually much older than the present approach. No satis-
factory answer has as yet been found, but the problern is under consideration,
as may be seen from the discussion about it at the Metallurgical Chemistry
Symposium 1971 33 .
From the same source, information may be gathered on the activities of
other laboratories in the field under review, in particular at tne Carnegie-
Mellon University, Pittsburgb (C. H. P. Lupis) and the Royal Institute of
Technology, Stockholm (M. Hillert, G. Kirchner and others), which cannot
all be e1aborated here. Actually, there are few fields in metallurgy where
scientific interest has been growing recently at so fast a rate as in the appli-
cation of chemical thermodynamics to the calculation of equilibrium
diagrams.
A severe restriction on the accuracy of the calculated phase boundaries is
imposed by their great sensitivity to small errors in the thermochernical
data, the required accuracy for which is as yet rarely achieved in experi-
mental thermochemistry. Use should therefore be made of existing. reliable
phase boundary information or additional phase diagram work performed
by means of conventional methods, and the basic data thus gradually
adjusted. Alloys melting below 1000°C, it is true, are quite we11 covered,
but these are of less practical interest. The present author is directing bis
attention to metallurgical systems of greater industria1 significance (say. to
alloyed steels and Nimonic alloys whjch incorporate quite a proportion of
the Periodic Table) and experimental thermochemists are needed to join
this effort because the task is huge. \

COMPOUND-FORMING SYSTEMS
The previous sections have dealt with the equilibrium diagrams of system.s
involving soJid and liquid solutions. However, in process metallurgy
numerous problems are encountered where the formation of solutions is
on1y of minor significance whereas equilibrium with a gaseous pbase must
be taken into account. The question then arises which solid and liquid
compounds are stable under given conditions of temperature and composi..
tion of the gas. Quite a nurober of such equilibria have been studied experi-
mentally and discussed on the basis of available thermochemical data at the
Institut für Theoretische Hüttenkunde at Aachen, where calculation and
plotting of the diagrams is performed by a computer 34 .
For the graphical representation of the stability rangeseither the tempera-
ture and one of the partial pressures of tbe gaseous components or two of the
partial pressures at constant temperature are cbosen. The systems studied
include Pb-S- 0 35 , Cu- S-0 36, Fe-S-0 37 , U-O-F 38, U-O-Cl 39 and
Mg-{J--s4o_
The systems alkaHne earth metal-oxygen- sulphur are of considerable
practical interest because of the potential use of the oxides for the removal
of sulpbur dio.xide from industrial gases by sulphate formation. Figure 23
shows the equilibrium diagram of the system magnesium-oxygen- sulphur
490
 PHASE EQU!LIBRIA IN CONDENSED SYSTEMS

in terms of temperature and oxygen pressure at a sulphur dioxide pressure

of 10- 4 atm which is the tolerable maximum with regard to air pollution.
F or the construction of the diagram, the thermochemical data of the con-
densed phases MgS, MgS0 3 , MgS0 4 and MgO and the gaseous species 80 2
and 0 2 after Barin and Knacke 45 have been perused. The phase boundaries
intersect at a quadruple point MgO~MgS-MgS0 4 -S0 2 which shifts with
increasing so2 pressure to higher temperatures.

Temperoture, °C
200 300 t. 00 500 600 700 800 900 lOOO
0 ·- ·- ·-· -·-·:_, _ ·---~-=-- -·- · - ·-·- · - ·- · -· ---·----· - · -· -·- .
!
I
784"C
p =1Q-4atm
502

Mg()

tooot r K
Figurt> 23 . Stahility areas in terms of oxygen pressure and temperaturein the system magnesium-
oxygen-sulphur at constant sulphur dioxide pressure oo- 4
atm).

Sulphur-containing exhaust gases generally have a content of, say. 0.1 atm
of free oxygen. Thus, the matter of interest is the intersection of the
( - 2 log Pso 2 ) line with the indicated (]og Po 1 = - 1) horizontal representing
the maximum temperature at which the S0 2 content of the exhaust gas is
reduced to the required Iimit. More revealing for practical purposes would be
a plot of log Pso1. versus 1jT .at p0 2 = 1 atm including similar plots for
other MO/MS0 4 equilibria. This shows the increasing stability of the
alkaline earth sulphates with atomic number 40 ; so that the temperature of
reaction may be increased to kinetically more favourable temperatures.
However, this aspect is not quite relevant in the contex:t of the present paper.
Another example has been selected from the fie]d of process metallurgy;
namely the system iron-oxygen-chlorine. important for the chlorination of
ores containing ferric oxides which aretobe either concentrated or selectively
vaporized. In this case a graphical representation is chosen in the form of
isothermal sections at 427 and 927oC (Figures 24 and 25) based on the
491
 0. KUBASCREWSKI AND I. BARIN
700 K (427 °C)
Or---~--~------------------------ ----------~
Gas phase(FeCl3 ,f"e2Cls)P t-s:: 1 atm
r-----~~----~~~~~~------~~--~0 0
~e

-5 -

-10 FeCl2(sol)
ö
Q

~
-15

Fe203
-20

l
-35 -30 -25 -20 -15 -10 -5 0
togpo2
Figure 24. Isothermal section of the system iron--oxygen-chlorine at 700 K. (P IJC • = PF~I)
+ PFe c is the total pressure of volatile chlorides.)
2 16

thermochemical properties of the gaseous compounds FeC1 2 , FeCI 3 and

FeCl 6 , and those of the condensed phases: iron, wüstite, magnetite, hematite,
FeOCl and FeC1 2 .
It may be seen from Figure 24 that chlorination of the iron oxides at lower
temperatures produces surface coatings consisting of FeC1 2 or at higher
oxygen pressures, FeOCl. Such surface coatings are of course undesirable

0 Gas phase(FeCl3+FeCl2)
p 5 =1 atm

0.1
-5 FeClz(ti )
N
0
Cl.
01
.9

Fe FeO

-20 -15 ~ 10 -5 0
log p 02
Figure 2.5. lsothermal section of the system iron·-oxygen-chlorine at 1200 K for Precl, + P r ctClt,
= l, 0.1 and. 0.01 atm.
492
 PHASE EQUILIBRIA IN CONDENSED SYSTEMS

because they represent kinetic obstacles. At tbe higher temperature (Figure

25), FeOO is no longer stable and chlorination ofbematite proceeds directly,
forming gaseous iron cblorides. A reduction of the oxygen pressure would be
technologically sensible.
Figure 25 also shows that gaseous iron chlorides may easily be trans-
formed with oxygen to Fe 2 0 3 producing chlorine. The conversion of the
chlorides at 700 K takes place via the formation of FeOCI, as has been
confirmed experimentally41 .
It has been mentioned earlier that thermochemical considerations alone
rarely describe technical processes fu11y. Kinetic conditions may play an
important role. During the chlorination of wüstite and magnetite under the
appropriate conditions, for instance, it has been observed by Mann 42 that it is
not liquid dichloride that is formed but gaseous trichloride whilst the lower
oxides are being oxidized to Fe 2 0 3 • This kinetic effect was first observed by
Gat 43 in the chlorination ofU0 2 and interpreted as beingdue to apreferential
chemisorption of chlorine from a Cl 2 -CO mixture leading to an assimilation
of the liberated oxygen ions by the oxide lattice44 .
Whereas the thermochemical conditions can normally be evaluated from
existing data.. the kinetic interactions have to be investigated experimentally
from case to case. Nevertheless. their thermochemical background, i.e. the
equilibrium diagrams. should be known first and it is tobe expected that the
thermochemical approach will increasingly find its practical application.

ACKNOWLEDGEMENTS
The authors are grateful to Dr P. J. Spencer of the National Physical
Laboratory, who assisted with the preparation of this paper, and to Professor
0. Knacke of the Technische Hochschule. Aachen. who discussed and
criticized its contents.

1
M. Rey, Faraday Soc. Disc. 4, 257 ( 1948).
2
K. Hauffe and C. Wagner. Z. Elektrocltemte. 46. 160 (1940).
3 C. Wagner, Acta Met. 2. 242 (1954).
4
M. Hansen, Constitucion ofBinary Alloys, McGraw-Hill, New York (1958): 1st Spplmt. {1965).
R. P. Elliott: 2nd Spplmt. ( 1969). F. A. Shunk.
5
R Hultgren et al .• Selected Values o.f the Therm.odynamic Properties o.f Metals and Alloys.
Wiley, New York ( 1963).
6
B. Gale and J. M. Davis. Metal Sei. J. S, 25 0971).
7
0. Kubaschewski and T. G. Chart, J. /nst. Metals 93. 329 (1965).
8
F. H. Hayes and 0. Kubaschewski. J. lnst. Metals, 97. 381 (1969).
9
F. H. Hayes. F. Müller and 0. Kubaschewski, J. lnst. Metals. 98, 20 (197Q).
10
L. Elford, F. Müller and 0. Kuba.schewski. Ber. Buusenges. Phys. Chem. 73, 601 (1969).
10
· " F. Müller and 0. KubaschewskL High Temp.-Higlt Press. 1. 543 (1969).
11 B. Predel and R. Mohs, Arclt. Eisenhilttenwes. 42, 575 (1971).
l l 0. Kubaschewskl, Refractory Merals. AGARD Conf., Oslo, Pergamon, Oxford, p. 191 (1964).
1 3 0. Kubaschewski Tlzermodynamics of Nuclear Materials, Proc. l.A.E.A. Sympos .. Vienna.

p. 219 ( 1962).
14
J. D. Gtogan and C. J. G. Clews. J. l'lst. Mnals. 77, 571 0950).
15 P. J. Spencer and F. H. Putland, J. lron an.d Steel Jnsr. 211, 293 ( 1973).
16
K. M. Myles and A. T. Aldred, J. Phys. Chem. 68,64 (1964).
17
L. Kaufman. Pha.~e Stahility in Metals &: Alloys. Proc. Batteile Colloquium, Geneva &
Villars, p. 125 (1967).
493
 0. KUBASCHEWSKI AND I. BARIN

ts L. ~aufman, Metallurgical Chemistry, Proc. Sympos., Brunel Univ. & NPL, HMSO, p. 373
(1972).
19 I. Ansara, Metallurgical Chemistry, Proc. Sympos., Brunel Univ. & NPL. HMSO, p. 403
(1972).
20
P. J. Spencer, F. H. Hayes and 0. Kubaschewski. Rev. Chim. M iner. 9. 13 ( 1972).
21
F. Kohler, Monatslt. Chemie, 91, 738 (1960).
22
G. W. Toop, 1rans. Met. Soc. AlME. 233, 850 (1965).
23 L. S. Darken. J. Amer. Chem. Soc. 72. 2909 (1950).
24
E. Bonnier and R. Caboz, Compt. rend. 250. 527 ( 1960).
25
K. Wohl, Trans. A.I.C.E. 42, 215 (1946).
26 F. H. Hayes, P. J. Spencer and L Elford, Chemical Metallurgy of Iron & Stee/, lron & Steel
Institute: London (1973).
27 0. Kubaschewski and J. F. Counsell. M onatsh. Chemie, 102, 1724 (1971 ).
28 E. Rauband G. Wörwag. Z. Metallkunde, 46, 513 (1955).
29 J. F. Counsel1, E. B. Lees and P. J. Spencer, M etal Sei. J. 5. 210 (1971 ).
30 P. J. Spencer and J. F. Counsell, z. Metallkunde, 64,662 (1973).
31 H. Martens and P. Duwez, Trar.s. Amer. Soc. Met. 44. 484 ( 1952).
32 J. A. Nelder and R. Mead Computer J. 7, 308 ( 1965).
33 0. Kubaschewski (Ed.), Metallurgical Chemistry, Proc. Sympos., Brunel Univ. & NPL,
HMSO: London, pp. 483 et seq. (1972).
34
U. Gat and R. Hörbe, Berichte der Kernforschungsan./age Jülich Nr. 596, Zentralbibliothek
KFA Jülich (1969).
35
0. Knacke and W. Neumann. Erzmetall. 9, 261 (1956).
36
0. Knacke, W. Hirschwald and P. Reinitzer. Erzmetall. 10. 123 (1957).
37
0. Knacke. J. Gaube. H. Lydtin and H. Pohl, Arch. Eisenlzüttenwes. 28, 109 ( 1963).
3 !! 0. Knacke, G. Lossmann and F. Müller, Z. anorg. u. allg. Chem. 370,91 (1969).
39
0. Knacke. F. Müller and E. van Rensen. Z . Physik. Chem. Neue Folge. 80,91 (1972).
40
H. A. Friedricbs. 0. Knacke and W. Schultze, Arch. Eisenhüttenwes. 44. 125 (1973).
41
A. W. Henderson, T. T. Campbelland F. E. Block, Met. Trans. 3, 2579 ( 1972).
42
E. Mann, Dissertation. Technische Hochschule Aacben ( 1972).
43
U. Gat, Dissertation. Technische Hochschule Aachen ( 1968}.
44
0. Kllacke, Metallurgical Chemistry, Proc. Sympos .. Brunel Univ. and NPL, HMSO, p. 549
(1971).
45
I. Bariu and 0. Knacke, Thermochemical Properries of lnorganic Compotmds, Springer :
Berlin ( 197J).

494