Calculation of phase diagrams: Calphad

John Agren

The Calphad area is now maturing and recent advances The recent advances in the Calphad field fall into the
relate to Calphad as applied to engineering problems in the following categories: development of databases; develop-
materials industry. Such applications involve design of new ment of models; development of computational methods
alloys, process development and environmental interactions. and software; and applications

Most new Calphad evaluations are based on the regular

Address
Department of Mater& Science and Engineering, Royal Institute of
solution model and multisublattice model and the SGTE
Technology, Stockholm S- 10044, Sweden; e-mail: john@met.kth.se data for pure elements. References to the recent evalu-
ations are given; examples of the application of Calphad
Current Opinion in Solid State 6 Materials Science 1996,
l:355-360
evaluations to non-equilibrium processing conditions in-
volving diffusion are discussed in this review.
0 Current Science Ltd ISSN 1359-0286

Abbreviations
Calphad calculation of phase diagrams Development of databases
SGTE Scientific Group Thermodata Europe The assessment of experimental information and the
evaluation of model parameters to be entered into the
databases are quite tedious work which proceeds slowly.
Introduction Even though the procedures today are highly computer-
The calculation of phase diagrams (Calphad) method ized, an evaluation of a complex binary system may take
is becoming widely recognized as a powerful tool in up to half a year, even for an expert. It is encouraging,
industrial development of new materials and processes. however, that thenumber of groups presenting evaluations
The method was established in the seventies after the of satisfactory quality is now much higher than a decade
pioneering work by Kaufman and Bernstein [l] and Hillert ago. Tables 1 and 2 summarize the evaluated binary
[Z] and others. Its simple idea is based on the well and ternary systems of metallurgical interest reported
known thermodynamic law that a system with a given during 1995. It should be emphasized that all the listed
composition, temperature and pressure attains the state systems are based on the same description of the pure
of lowest Gibbs energy under these given conditions. elements, that is, the so-called SGTE substance database
Provided that the Gibbs energy (as function of pressure, [4]. As an example, a calculated liquidus projection of the
temperature and composition) is known for the individual CaO-MgOSiO system at atmospheric pressure has been
phases, it is possible to calculate the equilibrium state by produced by Huang et a/. [P] and is shown in Figure 1.
an energy minimization procedure. The diagram shows which phase would start forming upon
solidification for the various compositions. Their analysis
As new general computer software, like SOLGASMIX, was based on a two-sublattice model for the ionic melt.
Thermo-Calc, FACT and MTDATA, has become avail- This means that their database cannot be combined with
able over the past two decades, the demand for consistent databases based on other thermodynamic models of the
thermodynamic data sets has grown steadily. A consistent ionic melt.
thermodynamic data set contains only independent data,
usually the Gibbs energy as a function of compsition,
temperature and pressure for a given phase. All other Table 1
quantites are calculated from this Gibbs energy function Calphad evaluations of binary metallurgical systems in 1995.
whenever needed. The works on steel and refractory
System Authors References
metals were based on the regular solution model and the
two-sublattice model presented by Hillert and Staffansson AS-AI Lim et a/. 1171
Ba-o Zimmerman ef a/. I181
[3]. One of the most important problems when evaluating B-W Duschanek and Rogl [191
thermodynamic data for a general database is to agree upon Bi-0 Risold et a/. 1201
the description of the pure components and the models to Cd-Hg Yang et al. 1211
be used for the various phases. Before this problem was Cd-Pb Zakulski and Moser [221
Cd-Te, Hg-Te Yang et al. 1231
solved data from various groups were usually incompatible
Cr-Cu Zeng and H;irnUinen 1241
and could not be combined into a database. Fortunately, Cr-0, Fe-O, Ni-0 Kowalski and Spencer 1251
a general standard, based upon the work within SGTE Fe-Cu Chen and Jin 1261
(Scientific Group Thermodata Europe) and the annual Fe-Sb Pei et a/. I271
Li-Si Braga et al. 1281
Calphad meetings, has now been established among most
Mn-Se Chevalier et a/. I291
of the groups.
356 Metals and alloys

Table 2 bad approximation except when the system is close to

critical temperatures. Under these conditions there is a
Calphad evaluations of ternary metallurgical systems In 1995.
second-order transition to a state with long-range order,
System Authors References
and the stability of this state will be greatly overestimated
As-Ga-In Shen et al. 1301 using this approximation. This problem has been tackled
B-C-Ti Dushanek ef a/. 131’1 by many authors over the years. For example, Grujicic
Ba-Cu-0 Zimmerman et al. [321
and Owen [7] considered Fe-Cr-Ni alloys with N and
Cd-Hg-Te Yang et a/. f211
Fe-Nd-B Hallemans et al. (331
compared calculations based on the cluster-variation and
Fe-Ni-0 Lauoma I341 the Monte-Carlo methods, both of which account for
CaO-MgO-Si02 Huang and Hillert [5”1 short-range order. They concluded that at 1273 K, without
nitrogen the solution is almost random, whereas with
nitrogen there is an increased probability of finding
clusters with two or more Cr atoms around an N atom.
Bouchard and Bale [6”] evaluated a thermodynamic
description based on the dilute solution formalism for The authors suggested that this clustering is responsible
for the well known increase in strength of austentic steels
molten multicomponent iron alloys containing C, N, Ti, Si,
upon nitrogen addition.
Mn, S and I? Their evaluation is not based on the SGTE
data but the database can still be used to tackle a number
of important problems in steel metallurgy, for example Applications
formation of non-metallic inclusions. Such inclusions may Simulation of microsegregation in multicomponent alloys
form during solidification of the steel and have large The Calphad approach may not only be used to calculate
influence on the mechanical properties in general and the multi-dimensional phase diagrams of materials of interest
fatigue strength in particular. but may also be extended to non-equilibrium systems
if some additional assumptions are made. A relatively
New models simple extension, but very useful indeed, is the so-called
The regular solution and multisublattice models are Scheil simulation of solidification. The well known Scheil
based on the assumption of random mixtures of atoms equation, used to predict the amount of solid as a function
or ions within each sublattice. This assumption allows of temperature or the amount of microsegregation during
the use of a simple expression for the entropic effect solidification of binary alloys, is based on the assumption
of mixing but suffers from the drawback that short of homogeneous composition in the liquid, no diffusion at
range order is completely neglected. This is not a all in the solid and full thermodynamic equilibrium locally

Figure 1

Calculated liquidus projection of the

CaO-MgO-Si02 system. Published with SiO,
permission from [5”1.

0 0.2 0.4 0.6 0.8 1 .O

CaO
Mole-fraction MgO WO
 calculation of phase diagrams Agren 357

Figure 2 Figure 3

“‘-.Scheil model _

--- k3.0 K/set

.I.... R=0.05 Wsec

I I I

1740 1735 1730 1725 Comparison between predicted (lines) and experimental (symbols)
composition profiles in an Fe-Cr-Ni diffusion couple annealed for 21
Temperature /K hours at 12Oo’C. Filled, half-filled and empty symbols denote different
diffusion couples. ‘Distance’ refers to the distance from the initial
Changes of total fraction solid during solidiiication of an joint. Published with permission from [12*].
Fe-l 8%Cr-8.5%Ni alloy calculated from equilibrium and the Scheil
simulation and limited back diffusion model. The two dashed lines are
calculated after modifying the Scheil model to account for solid state
diffusion. R is the cooling rate. Published with permission from [WI.
Diffusion
Almost half a century ago Darken [9] realized that
diffusion is closely related to thermodynamics and he
wrote the interdiffusion coefficient of a binary system as:

at the solid/liquid phase interface (i.e. the compositions

D - (XBDA' + XADB*)O
on both sides of the phase interface may be read
directly from the phase diagram for the temperature under
consideration). These assumptions may also be applied where xA and ~1%are the mole fractions of B and A and
to the solidification of a multi-component liquid and DA* and DB* are the tracer diffusivity of A and B. The
result in a set of differential equations which, combined tracer diffusivities describe diffusion of low additions of
with an equilibrium calculation at each temperature step, radioactive isotopes of A and B. @ is the thermodynamic
will predict the solidification path of a multi-component factor and is related to the second derivative of the molar
alloy. The solidification path (i.e. the composition of the Gibbs energy with respect to concentration. Although
liquid as function of temperature) is quite important it may seem that the Darken relation is now firmly
from a practical viewpoint because it determines which established, its general validity is sometimes questioned.
solid phases actually form and what order they form in. It has been argued that it is only applicable to solutions
A slight change in the alloy composition may give a close to ideality, where Q, is unity. Such a view was
completely different solidification path and result in totally challenged by Karpe et al. [lo], who measured both
different properties of the final alloy. Over the past decade tracer diffusivities and interdiffusion in amorphous Ni-Zr
such calculations have been presented by several authors, alloys and evaluated a @ value as high as 33 at 623 I<
for example by Ohsasa et al. [8*]. In their calculation and 43at%Zr, that is, the interdiffusion would be 33
they were also able to account for solid state diffusion times more rapid than expected from the tracer diffusion
in a simplified manner by a numerical solution of the data and ideal solution behaviour. After comparing with
diffusion equation. Such solutions are usually based on a thermodynamic calculation, based on the Calphad
the finite-difference method in which the concentration approach and yielding 0~21, and taking into account all
profile is assumed to consist of a large number of linear uncertainties they concluded that their findings were in
segments. In the diffusion calculation the coarseness of satisfactory agreement with Darken’s relation. Thus, the
the structure, that is, the secondary dendrite arm spacing value of 0 need not be close to unity for the interdiffusion
(which in turn depends on the cooling rate), is necessary coefficient proposed by Darken to be applicable.
as input data. Figure 2 is reproduced from their report and
shows the calculated fraction of solid phase as a function The large value of the thermodynamic Factor for the
of temperature in an ordinary 18Cr-8.5Ni stainless steel above Ni-Zr alloys stems from the strong short-range
under various conditions. ordering tendencies of the amorphous phase. Since
358 Metals and alloys

Q, is related to the second derivative of the Gibbs and fee [13] Fe-Cr-Ni alloys by Jonsson. In its general
energy curve (i.e. its curvature), it is obvious from the multi-component version the diffusion coefficient matrix
Darken relation that ordering may have a very strong D (for a ternary system it is 2x2 matrix) is given by:
influence on diffusion. In fact, it has been known for
some time that, close to the ordering temperatures, the D=L@
interdiffusion coefficient in Fe-5 alloy has very unusual
deviations from the Arrhenius law. Rabkin et al. [ll’], where L is a matrix containing only kinetic coefficients
measured interdiffusion in Fe-Si alloys and compared (i.e. the atomic mobilities), and 0 is a matrix containing
the concentration dependence of interdiffusion with the basically the second derivatives of the Gibbs energy with
concentration dependence of the thermodynamic factor respect to alloy content. Jijnsson applied 0 as given
calculated using the Bragg-Williams-Gorsky model. They by the Calphad analyses and evaluated the elements of
concluded that the anomalous behaviour may be explained L in terms of the simple vacancy model for diffusion
by the thermodynamic factor. (diffusion takes place by atoms jumping to neighbouring
vacant sites and a random mixture between atoms and
vacancies is assumed). It was then possible for him to
Fiaure 4
calculate the concentration-dependent elements of the
multi-component diffusion coefficient matrix for a wide
30 range of compositions. Figure 3 shows an example of
concentration profiles calculated with Jonsson’s description
and measured experimentally. As can be seen the
agreement is excellent.
;A - 0.0001 s
Ii
------ 0.01 5 -
__--+.+---; . . ....t..... 0.1 s The data on Fe-Cr-Ni assessed by Jonsson was also
E ---
.1 ‘IIj , I
-----
10 5 used by Engstriim [14], who made numerical simulations
E 24 I:
Ii
6041.2 s _
I of ternary diffusion in a bee matrix with dispersed fee.
2 Ii
I: I
S
Ij
t In Engstrom’s calculation, diffusion in the matrix was
*i I
I:
22 III : , I calculated by means of a finite element method and the
,I , fraction and composition of the matrix and the dispersed
c LC-I
20
L____---_--_-_----______ fee phase were subsequently calculated by an equilibrium
I I 1 ’ , 8 I ’
calculation for each grid point.
0 1 2 3 4 5

distance (microns)
(a) Diffusion-controlled phase transformations
The combination of thermodynamic calculations and
I! “““‘I
13 Fe-21.9Cr-10.35Ni, 12OO’C - simulation of diffusion is no doubt a powerful tool
~-----__----_-~-------- to circumvent one of the most severe drawbacks with
; I’
\ I Calphad, that is, the result of Calphad is only expected
to hold as a limiting case when equilibrium is established.
The Scheil simulations, which have been modified to
include solid state diffusion 18.1, have already been
mentioned. Vitek et al. [15] have applied a similar pro-
cedure to simulate the diffusion-controlled transformation
- 0.0001 .s between ferrite and austenite in Fe-Cr-Ni alloys. In
..:/ /,’ .------
0.01 5 _
. . .._..0.1 s their calculation, the ternary Fe-Cr-Ni isothermal section
2’ --- 10s
,’ ----- 6041.2 s was calculated by means of the program Thermo-Calc
4 at various temperatures in the range 700-1300°C and
I 1 1 1 I I I I I
0 1 2 3 4 5 parameters were applied in order to have analytical
expressions for the phase boundaries of the phase diagram.
distance (microns)
The analytical expressions were then combined with a
finite-difference calculation of Cr and Ni diffusion in
Growth of ferrite in Fe-Cr-Ni alloy heat treated for various times ferrite and austenite. The migration rate of the moving
at 1200% (a) Cr and (b) Ni concentration profiles during growth. phase interface as well as the composition on both sides of
‘Distance’ denotes distance from initial center of ferrite region.
the interface were obtained. In their diffusion calculation
Published with permission from [15].
all off-diagonal elements of the diffusion coefficient
matrices were neglected, that is, ideal solution behaviour
was assumed. Figure 4 shows the calculated Cr and Ni
A systematic approach to analyzing diffusion data was concentration profiles after various annealing times at
suggested by our group in 1992 and applied to bee [12’] 1200°C.
 Caicutiion of phase diigramr Agren 359

Martensitic transformations 3. Hillert M, Staffansson L-i: The raguiar solution model for
stoichiometric phases and ionic m&a. Acta Chem Stand 1970,
As Fe-Mn alloys with more than 10% Mn are cooled 243610-3626.
rapidly from the austenitic state, which is face-centered 4. Dinsdale A: SGTE d8t8 for pure elem8nts Celphed - Computer
cubic, there is a martensitic transformation to the hexag- Coup Phase Dia 1991, 15:317-425.
onal close-packed structure. This transformation is the 5. Huang W, Hillert M: Thermodynamic assessment of the CaO-
basis for a shape-memory effect in Fe-Mn-Si alloys and MgO-SIC2 system, Mete/l Mathr Trans A 1995, 26:2293-2310.
Kis paper demonstrate8 that Calphad may be successfully applied to a
consequently it has attracted a lot of attention over the very complex system of large practical interest and invofves quite advanced
past decade. Cotes et al. [ 16.1 have studied experimentally Calphad modelling.

the M, and A, temperatures (i.e. the temperatures 6 Bouchard D, Bale C: Simult8n8ous optimization of
.. thermochemkal data for liquid iron alloys containing C, N, Ti,
where martensite starts forming upon cooling and the Si, Mn, S, and P. Mete/l Meter Trans 8 1995,26:467-404.
reverse transformation starts upon heating, respectively) This paper demonstrates very nicely how Caiphad and diffusion experiments
may be combined to give a definite answer to a somewhat controversial
in the Fe-Mn system and analyzed the information in question.
terms of thermodynamic models for the fee and hcp 7. Grujicic M, Owen WS: Models of short-range order in a face-
phases. In particular, they analyzed the effect of the centered cubk Fe-NI-Cr alloy with a high concentration of
nitrogen. Acta Metal/ Mater 1995.43:4201-4211.
magnetic transition at the Neel temperature, whereby
the paramagnetic austenite becomes antiferromagnetic. 8. Ohsaaa K, Nakaue S, Kudoh M, Narit8 T: Analysis of
. solidification path of Fe-Cr-NI ternary alloy. iron and Stee/
The Niel temperature increases with Mn content and Institute of Japan (/S/J) Intern8tionall995, 35:629-636.
the stabilization of the austenite caused by the magnetic This papar is a good example of how the Cafphad technique may be applied
to non-equilibrium conditions.
ordering suppresses the martensitic transformation and
9. Darken LS: Diffusion, mobiifty 8nd their interrelation through
causes the shape-memory effect to deteriorate, and thus free energy in binary metallic systems. Tmns A/ME 1946,
make the material unsuitable as a shape-memory metal. 175:164-l 94.
10. Karpe N, Kro9 JP, Bottiger Chechenin NG, Somekh RE, Greer AL:
The thermodynamk factor in interdlffusion: a strong effect in
Conclusions amoQhous Ni-Zr. Acta Mer8// Mater 1995,43:551-556.
The thermodynamic assessment of alloy systems proceeds 11. Rabkin E, Straumal B, Semenov V, Gust W, Predel B: The
steadily and consistent databases are gradually being com- . influence of an ordering transition on the interdlffusion in Fe-Si
alloys. Act8 MeraIl Merer 1995,43:3075-3063.
piled. Most assessments are based on the regular solution This paper demonstrates how thermodynamic modelling may be applied to
or multisublattice models, although some attention is represent deviations from the Arrhenius law for the diffusion coefficient in a
system with ordering.
being paid to short-range ordering effects. The SGTE data
of the pure elements seems generally accepted and is the 12. Jtinsson B: Assessment of the mobliities of Cr, Fe and Ni in
. bee Cr-Fe-Ni alloys. Iron 8ftd Steel institute of Japan f/S/J)
basis for almost all assessments. International 1995, 35:1415-l 421.
This paper summarizes very well how diiusional mobilities may be evaluated
by a Calphad type of approach.
The Calphad method is now more widely used to
13. Jijnsson 8: Assessment of the mobilities of Cr, Fe and Ni in fee
solve problems involving non-equilibrium conditions, such Cr-Fe and Cr-Ni alloys. Stand J Mets// 1995, 24:21-27.
as those occurring during solidification processing and 14. Engstrijm A: Interdiffusion in multiphase, Fe-Cr-Nl diffusion
heat treatment. The combination of thermodynamic and complex. Rand J Metal/ 1985, 24:12-20.
diffusion calculations is particularly attractive owing to its 15. Vitek JM. Vitek SA. David SA: Numerical modeiiina of dlffusion-
controlled phase iransformation in ternary syste& 8nd
conceptual simplicity, although the calculations may be application t the ferrlte/austenite transformation in the
quite complex. Thus, the future of the Calphad method Fe-Cr-Ni system. Metal/ Mater Tr8ns A 1995, 26:2007-2025.
is as an engineering tool (i.e. on the application side) 16. Cotes S, Sade M, Fern&de2 Guillermet A: Fcc/hcp martensitic
. transformation in the Fe-Mn svstem: exoerimental studv 8nd
and it should be evident by now that the method is
thermodynamic analysis of ph& stability. Metal/ Mater Tmns A
mature enough to be applied on a broad scale to industrial 1995,26:1957-l 969.
problems. On the scientific side, however, the bridging This paper demonstrates the application of Calphad analysis to bw temper-
atures and martensitic transformations.
between ab it&o energy calculations and Calphad is a
1 7. Lim SS, Rossiter P, Tibballs JE: Assessment of the Al-Ag binary
challenge. Despite large efforts there is still a huge cap phrse diagram. Calphad - Compur Coup Phase Dia 1995,
between the numbers evaluated from experimental data 19:131-141.

in terms of the Calphad models, and what is predicted ab 18. Zimmerman E. Hack K, Neuschiitz D: Thermochemical
8ssesrment of the binary system B8-0. C8/ph8d _Comput
initio by quantum mechanical calculations.
Coup Phase Dia 1995, 19:119-l 27.

19. Duschanek H, Rogl P: Critical assessment and thermodynamic

calculstion of the binary system boron-tungsten (B-W). J Phase
References and recommended reading Equrlrbria 1995, 16:150-l 61.
Papers of particular interest, published within the annual period of review,
20. Resold D, Hallstedt 8, Gauckler JJ, Lukas HL, Fries SG: The
have been highlighted as:
bismuth-oxygen system. J Phase Equilibrie 1995, 16:223-234.
. of special interest 21. Yang J, Silk NJ, Watson A, Bryant AW, Chart TG. Argent BB:
.. of outstanding interest The thermodyn8mic.s and ph8se diagmms of the Cd-Hg 8nd
Cd-Hg-Te systems. Calphad - Comput Coup Phase Dia 1995,
19:415-430.
1. Kaufman L, Bernstein l-f: Computer calculation of phase diagrams.
New York: Academic Press; 1970. 22. Zakulskl W, Moser Z: A calculation of the Cd-Pb (cadmlum-
lead) system. J Phase Equilibria 1995, 16:239-242.
2. Hillert M: Calcularions of phase equilibria. American Society for
Metals Semrnar on Phase Tlansformarions. Metals Park, Ohio: 23. Yang J. Silk NJ, Watson A, Bryant AW, Argent BB:
American Society for Metal; 1968:161-216. Thermodynsmic and phlse diagram 8ssessment of the Cd-Te
360 Metals and alloys

and fig-Te systems. Calphad - Comput Coup Phase Dia 1995, 30. Shen J-Y. Chantillon C. Ansara I. Watson A. Ruao B. Chart TG:
19:399-414. Optlmization of the thermodynamic and pha?1(5diagram data
in the ternary As-Ga-In system. Calphad - Comput Coup Phase
24. Zeng K, H&rn&iinen M: Thermodynamic analysis of stable and
Dia 1995, 19:215-228
metastable equilibria in the Cu-Cr system. Calphad - Comput
Coup Phase Dia 1995, 19:93-104.
31. Dushanek ii. Roal P. Lukas HL: A crttlcal assessment and
25. Kowalski M, Spencer PJ: Thermodynamic reevaluation of the . thermodynamic”~lculation of the boron-carbon-titanium (B-
Cr-0, Fe-O and Ni-0 systems: remodelling of the liquid, BCC C-T0 ternary system. J Phase Equilibria 1995, 16:46-60.
and FCC phases. Ca$had . Comput Coup Phase Dia 1995, This paper is an excellent example of a good Calphad analysis of a complex
19:229-243. ternary system.

26. Chen a, Jin Z: The Fe-Cu system: a thermodynamic evaluation.

32. Zimmerman E, Hack K, Mohammad A, Boudine A, Neuschiitz D:
Metal/ Mater Trans A 1995, 26:41?-428.
Experimental investigation and thermochemical assessment of
27. Pei B, Bjcrkman B, Sundman B, Jansson B: A thermodynamic the system Ba-Cu-0. Calphad - Comput Coup Phase Dia 1996,
assessment of the iron-antimony system. Calphad - Comput 19:179-188.
Coup Phase Dia 1995, 19:1-5.
33. Hallemans B, Wollants P, Roes JR: Thermodynamic assessment
28. Braga MH, Malhelros LF, Ansara J: Thermodynamic assessment
of the Fe-Nd-B phase diagram. J Phase Equilibria 1995,
of the Li-Si system. J Phase Equilrbna 1995, 16:324-329.
16:137-149.
29. Chevalier P-Y, Fisher E, Rivet A: A thermodynamic evaluation of
the Mn-Si system. Calphad _Comput Coup Phase Dia 1995. 34. Lauoma R: A thermodynamic analysis of the system Fi-Ni-0.
19:57-68. Calphad - Comput Coup Phase Dia 1995, 19:279-295.