The GibbsThomson Effect in Dilute

Binary Systems
MA QIAN
The solubility of a precipitate, which is commonly referred
to as the composition of the matrix in equilibrium with the
precipitate in the presence of a curved interface, has been
well documented. For a pure precipitate , which consists
of component B, in an matrix that is a dilute solution of
B in A, the GibbsThomson effect is well known as[1,2]
xB(r) xB() exp

Fig. 3The temperature dependence of the specific strength for the

tested alloys

2V m
RTr

[1]

where xB(r) and xB() are the equilibrium atomic fractions

of B in at a curved interface of a spherical precipitate ( )
of radius r and a planar interface, respectively. The term
is the specific interfacial free energy, which is assumed
to be isotropic and independent of precipitate size (r), and
V m is the molar volume of . When is not pure B, e.g.,
its composition is given by xB, a number of different expressions have been proposed or employed. These include[311]
2V m

xB(r) xB() exp

x RTr

[2]

xB(r) xB() exp

2V B
RTr

[3]

and
xB(r) xB() exp

Fig. 4Specific strength of (Ir,Rh)75Nb15Ni10 alloys compared with other

high-temperature materials.

alloys based on combining Ir with Rh, which has higher

specific strengths compared with Ir-, Rh-, and Ta-based
alloys and is comparable to W-based alloys at 1600 C.

REFERENCES
1. Y. Yamabe-Mitarai, Y. Koizumi, H. Murakami, Y. Ro, T. Maruko, and
H. Harada: Scripta Metall., 1997, vol. 36, pp. 393-98.
2. Y. Yamabe-Mitarai, Y. Ro, T. Maruko, and H. Harada: Metall. Mater.
Trans. A, 1998, vol. 29A, pp. 537-49.
3. Y.F. Gu, Y. Yamabe-Mitarai, and H. Harada: in Iridium, E.K. Ohriner,
R.D. Lanam, P. Panfilov, and H. Harada, eds., TMS, Warrendale, PA,
2000, pp. 73-84.
4. Y.F. Gu, Y. Yamabe-Mitarai, Y. Ro, T. Yokokawa, and H. Harada:
Metall. Mater. Trans. A, 1999, vol. 30A, pp. 2629-39.
5. Y.F. Gu, Y. Yamabe-Mitarai, X.H. Yu, and H. Harada: Mater. Lett.,
1999, vol. 41, pp. 45-51.
6. N.S. Stoloff and C.T. Sims: in Superalloy II, C.T. Sims, N.S. Stoloff,
and W.C. Hagel, eds., John Wiley & Sons, New York, NY, 1987, p. 519.
7. H. Tanaka, Y. Tan, A. Kasama, and R. Tanaka: Proc. Int. Conf. on
Advanced Technology in Experimental Mechanics 99, Japan Society
of Mechanical Engineers, Tokyo, Japan, 1999, p. 539.
METALLURGICAL AND MATERIALS TRANSACTIONS A

1 xB() 2V m
xB() xB() RTr

[4]

Equation [3], in which V B is the partial molar volume of

B in , is given by Gaskell in a well-known textbook.[6]
Unfortunately, this is an incorrect expression as pointed out
by Hillert.[12] To illustrate this, one may consider such a
case where the partial volume of an interstitial is so small
that it can be neglected. Equation [3] says that xB(r) should
then be independent of precipitate, but from a molar Gibbs
energy diagram, one can see that it is not independent.[12]
The reason Eq. [3] is incorrect may be elaborated as follows.
As can be seen from Reference 6, until Eq. [77], Gaskell is
only considering a single phase and Eq. [77] holds under
constant composition (i.e., a constant xB). However, it should
be noted that one cannot define the partial molar volume or
any other partial molar quantities of a component in a single
phase with a fixed composition unless the phase under consideration is in equilibrium with another phase.[12] Also,
when considering a two-phase equilibrium, Gaskells Eq.
[78], reproduced as Eq. [3] above, is derived under the
incorrect assumption that the two-phase equilibrium is determined by equal activity of one component, neglecting the
other one.[12] Thus, it is not by using the common tangent
construction. With a simple molar Gibbs energy diagram,

MA QIAN, Senior Research Fellow, is with the CRC for Cast Metals
Manufacturing (CAST), Department of Mining, Mineral and Materials
Engineering, The University of Queensland, Brisbane QLD 4072, Australia.
Contact e-mail: q.ma@minmen.uq.edu.au
Manuscript submitted September 27, 2001.
VOLUME 33A, APRIL 20021283

Fig. 1A molar Gibbs free energy diagram constructed to illustrate the equilibrium between an phase (the matrix) and a spherical phase (the precipitate)
with the radius of varying from to r. AH and DF are two common tangents (B, O, Q, and S are four points of tangency). P is the point of intersection
for AH and DF corresponding to xB on the composition axis. K is the orthogonal projection of point S on the Gibbs free energy curve Gm(). The three
dashed lines CE, MG, and NL are all constructed to be parallel to DF. As a result, CD EF, MN GL, and DN FL SK 2V m /r.

it is easy to demonstrate that this is wrong because the

composition of the precipitate phase must also change. If
the precipitate phase cannot change its composition, one
must still equilibrate both activities, i.e., use a common
tangent.[12] A generalized GibbsThomson equation has
been given by Hillert[9] for this case based on a molar Gibbs
energy diagram, i.e., Eq. [38] of Reference 9. Using Hillerts
Eq. [38],[9] and noting that when is a dilute solution of B
in A, the solute B obeys Henrys law and the solvent A obeys
Raoults law, one can obtain
(1 xB()) ln

1 xB(r)
2V m
xB(r)

x
()
ln

B
1 xB()
xB()
RTr

[5]
Equation [5] is an exact GibbsThomson expression valid
for a dilute binary two-phase system in which the composition of the precipitate phase, xB, is constant, i.e., fixed at
xB(). When xB is a function of r, xB() in Eq. [5] should
be taken as some value between xB() and xB(r).[12] Unfortunately, this value is not known explicitly when is a solution phase.
Equation [4] is derived for the limiting case when both
(xB(r) xB())/xB() 1 for the composition of and
(xB(r) xB())/xB() 1 (including xB constant) for
that of are satisfied.[711] Equation [2] can be regarded as
a variation of Eq. [4] when xB() xB(). Two other variations of Eq. [4] have also been given for the case when
and are both dilute solutions of B in A[13] and for the case
when is a dilute solution of B in A but is a dilute solution
of A in B.[14] The former is equivalent to assuming (1
xB())/(xB() xB()) 1/(xB() xB()) in Eq. [4][13]
1284VOLUME 33A, APRIL 2002

and the latter equivalent to assuming (1 xB())/(xB()

xB()) 1,[14] where Eq. [4] reduces to Eq. [1].
Equations [4] and [5], including their variations, have
probably reflected all of the representative expressions
reported to date for the GibbsThomson effect in a dilute
binary system. However, Eq. [5] is restricted to a constant
xB and Eq. [4] is limited to both (xB(r) xB())/xB()
1 and (xB(r) xB())/xB() 1. Thus, neither describes
the general case when xB and xB can freely change with the
size of . Hillerts Eqs. [45] and [49],[9] which are given in
terms of d 2Gm/dx2, and therefore valid for either a dilute or
a concentrated binary system, are also limited to small
changes in xB and xB.
In this article, using a constructed molar Gibbs energy
diagram, a general description of the GibbsThomson effect
in a dilute binary system is introduced for the case when
both xB and xB can freely change with the size of . As will
be shown, this description is the same form as Eq. [5] based
on Hillerts Eq. [38][9] for a constant xB. Then, it is shown
that Eq. [4] is a limit form of Eq. [5] when xB(r) xB().
Finally, the applicability of Eq. [4] is clarified by comparing
it with Eq. [5].
Similar to the molar Gibbs energy diagrams used by
Hillert,[9,13] Christian[10] and Martin et al.[11] for the formulation of Eq. [4], Figure 1 shows a molar Gibbs energy diagram
constructed for the equilibrium between an phase and a
spherical phase. This analysis assumes that (a) the second
phase is incompressible and of isotropic surface properties
and (b) there are no stresses within the solid matrix phase.
The latter assumption could happen either if all the molar
volumes are equal and there will be no local volume changes
or if the matrix phase is soft enough (e.g., by the temperature
being high) to relieve all stresses. Note that, in Figure 1,
METALLURGICAL AND MATERIALS TRANSACTIONS A

both xB and xB can freely change with the radius of .

Comparing triangles ABC and EBH yields
xB()
AC

1 xB() EH
Substituting AC A() A(r) CD
B(r) B() CD leads to

[6]
and

Apparently, if xB xA holds, i.e., if is a dilute solution

of B in A,[15,16] then Eq. [13] always holds and therefore MN
AN.
Equation [9] can thus be written with sufficient accuracy,
when is a dilute solution of B in A, in the form
xB( ) A() A(r) CD

xB( )
AN

EH

CD xA()( A(r) A()) xB()( B(r) B())

[7]

Substituting Eqs. [7] and [10] with some rearrangement

gives
xA()( A(r) A()) xB()( B(r) B())
2V m

Then, comparing triangles ABC and AOM gives

xB()
AC

xB() AM

[8]

xB() A() A(r) CD

xB()
AN MN

[9]

That is,

in which
AN AD SK A() A(r) 2V m /r

[10]

In the following, we shall prove that if is a dilute

solution of B in A, i.e., when xB 1 or xB xA,[15,16] then
MN AN always holds. Consequently, solving Eq. [9] can
result in a general description of the GibbsThomson effect.
The length of MN can be determined by comparing triangles AOM and GOH
xB()
AM

1 xB() GH

[11]

where AM A() A(r) SK MN and GH

B(r) B() SK MN (SK 2V m /r). Therefore,
MN xA()( A(r) A()) xB()( B(r)

B())

2 V m

[12]

[13]

[14]

1 xB(r)
xB(r)

x
()
ln
B
1 xB()
xB()

xB(r) xB() exp

2V m
(1 )xB()RTr

METALLURGICAL AND MATERIALS TRANSACTIONS A

[19]

in which

1 xB(r)

1 x ()
x (r)
x () ln
x ()

(1 xB()) ln
[15]

[18]

For easy comparison with Eq. [4], Eq. [5] can be written as

Namely,
xB A() A(r)

xA B(r) B()

2V m

RTr

On the other hand, comparing triangles APD and FPH gives

xB
AD

1 xB FH

[17]

which is identical to Hillerts Eq. [38],[9] on which Eq. [5]

is based, except that in Eq. [17], V m is the molar volume of
corresponding to xB(r) (Eq. [18] gives a description of
x(r) when is also a dilute solution). Equation [17] demonstrates that the expression of Eq. [5] derived for a constant
xB based on Hillerts Eq. [38][9] can be extended to the
general case where both xB and xB can freely change with
the size of . In other words, the expression of Eq. [5] can
be used as a general description of the GibbsThomson
effect in a dilute binary system.
It should be noted that Eq. [5] is not an exact expression
unless xB is constant. However, it is a sufficiently accurate
expression so long as xB xA or xB 1 holds for the
system under consideration as rigorously derived earlier (i.e.,
from Eqs. [10] to [17]). Note that to metallurgists, a dilute
solution is commonly considered to mean that in a small
concentration range the activity coefficient of the solute is
approximately constant.[17] It is therefore possible that the
constraint xB 1 or xB xA may not always be satisfied
in such a practically dilute system. Accordingly, the use of
Eq. [5] may become less accurate under these circumstances,
depending upon to what extent the approximation MN
AN is satisfied according to Figure 1.
A similar expression can be obtained for the dependency
of xB on r from Eq. [17] when is also a dilute solution
(either B in A or A in B), by noting that i (r)
i () i (r) i () at equilibrium, as follows:
(1 xB()) ln

Then, comparing Eqs. [12] and [10], the assumption MN

AN leads to

A(r) A() B(r) B()

[16]

[20]

VOLUME 33A, APRIL 20021285

Table I. Comparison between the Solubilities Calculated Based on Equations [4] and [5] (xB () 0.05)
(xB(r) xB(r))/xB(r) (Pct)
2V m /RT

10

2V m /RT 109 m

Precipitate Radii
(r) (nm)

xB() 0.2

xB() 0.5

xB() 0.9

xB() 0.2

xB() 0.5

xB() 0.9

15
20
50
100
200
500
1000
5000

*
*
*
9.72
1.77
0.25
0.059
0.0022

13.37
5.67
0.63
0.14
0.034
0.0053
0.0013
5.21 105

0.23
0.12
0.017
0.004
9.87 104
1.55 104
3.88 105
1.55 106

3.45
1.77
0.25
0.059
0.014
0.0022
5.65 104
2.25 105

0.062
0.034
0.0053
0.0013
3.27 104
5.21 105
1.31 105
5.21 107

1.77
9.87
1.55
3.88
9.69
1.55
3.88
1.55

103
104
104
105
106
106
107
108

*No solution to Eq. [5] for xB(r) due to the selected large capillary length.
xB(r): solubility calculated based on Eq. [4].
xB(r): solubility calculated based on Eq. [5].

Note that

m ln

x (r)
x () ln
x ()

(1 xB()) ln
(xB())

lim

(r)x()
xB
B

xB()(1 xB())

xB()(1 xB())
[18]

1 xB(r)
1 xB()

[21]

where LHopitals rule was used in the course of solving

Eq. [21]. It is interesting to note that on substituting Eq.
[21] into Eq. [19], one at once obtains Eq. [4], demonstrating
that Eq. [4] is actually a limit form of Eq. [5] when
xB(r) xB(). When 1, Eq. [19], i.e., Eq. [5], will
reduce to Eq. [2]. The condition 1 thus serves as an
appropriate constraint on Eq. [2]. This has been neglected
before.
Although Eq. [4] is a limit form of Eq. [5] when
xB(r) xB(), owing to its simpler form, it would be useful
to identify the circumstances where Eq. [4] can be used to
replace Eq. [5] in practice. Table I shows the results of
comparison between Eqs. [4] and [5] for different dilute
binary systems (the capillary length of a metallic system,
2V m /RT, is usually less than 108 m when typical values
are substituted[11]). The value of xB() is fixed at 0.05 in
these calculations because changing xB() only slightly
influences the results. As is evident from Table I, both equations lead to almost the same results in most cases of practical
interest, even when the constraint, (xB(r) xB())/xB()
1, on the composition of the matrix ( ) is not met for
Eq. [4]. The notable exceptions are when the system under
consideration is characterized by 2V m /RT 109 m and
xB() 0.5. Under these circumstances, applying Eq. [4]
to precipitates smaller than 200 nm in radius tends to result
in significant errors in the calculated values of xB(r). Otherwise, Eq. [4] can always be used as a good approximation
of Eq. [5].
For a compound precipitate of stoichiometric composition
such as AmBn , following the same methodology used for
deriving Eq. [17] or using Hillerts Eq. [38],[9] it can be
derived that
1286VOLUME 33A, APRIL 2002

1 xB(r)
xB(r)
2V m
n ln

1 xB()
xB()
RTr

[22]

and that when (xB(r) xB())/xB() 1,

xB(r) xB() exp

1 xB()
2V m

n/(m n) xB() (m n)RTr

[23]

in which V m is the volume of one mole of formula units

AmBn . Equation [22] was first given by Bhattacharyya and
Russell[19] (i.e., Eq. [8] of Reference 19). It may also be
derived by means of the GibbsDuhem relation.[20,21] Equation [23] was first derived by Trivedi[8] (Appendix III of
Reference 8) and has been used by the present author to
describe the coarsening of intermetallic or compound precipitates in binary systems.[22] Similarly, when 2V m /(m
n)RT 109 m, Eq. [23] can be used as a good approximation of Eq. [22] according to Table I. However, when
2V m /(m n)RT 109 m and r 200 nm, Eq. [22]
should be employed rather than Eq. [23]. A detailed comparison between Eq. [22] and the reported experimental measure
(r) for different sizes of Cu4Ti particles
ments of r and xTi
in a Cu-4Ti alloy has been reported previously.[20] It was
shown that Eq. [22] is in good agreement with the reported
experimental results.
Like most previous analyses, this analysis is also aimed
at solid-solid systems. However, it should be noted that the
thermodynamic equilibrium between a solid matrix and a
solid second-phase particle is very complicated.[23] The common tangent approach used in this analysis is strictly applicable to fluid-fluid systems, where surface stress and surface
energy are identical for a fluid-fluid interface.[8,23,24] A recent
detailed examination of the thermodynamic basis for the
common tangent approach has shown that this approach is
also applicable to fluid-solid systems if the solid secondphase particle under consideration is incompressible and of
isotropic surface properties.[25] With the assumptions made
earlier, i.e., (a) the second phase is incompressible and of
isotropic surface properties and (b) there are no stresses
within the solid matrix phase, this analysis has approximated
the solid-solid systems under consideration as if the matrix
METALLURGICAL AND MATERIALS TRANSACTIONS A

were liquid. In other words, this analysis has assumed that

the common tangent approach is applicable to the solidsolid systems analyzed.
In summary, the following points can be concluded from
this work.
1. The expression of Eq. [5] can be used as a general description of the GibbsThomson effect in a dilute binary system, wherein the compositions of both the matrix ( )
and the precipitate ( ) can freely change with the size
of , as long as the matrix ( ) remains to be a dilute
solution when xB varies.
2. Although Eq. [4] is a limit form of Eq. [5] when
xB(r) xB(), it can be used as a good approximation
of Eq. [5] in most cases of practical interest. The notable
exceptions are when the system under consideration is
characterized by 2V m /RT 109 m and xB() 0.5,
especially, when the precipitate particles are smaller than
200 nm in radius.
3. For a binary compound precipitate of stoichiometric composition such as AmBn , Eq. [23] can serve as a good
approximation of Eq. [22] when 2V m /(m n)RT
109. However, when 2V m /(m n)RT 109 m
and r 200 nm, Eq. [22] should be used rather than
Eq. [23].

The author is most grateful for many helpful discussions

with Professor Mats Hillert during the preparation of this
work and for his many helpful comments on the first and
second versions of this article. The key readers suggestion
of mentioning the difference in thermodynamic equilibrium
conditions between fluid-fluid and solid-solid systems is
acknowledged. This suggestion has led to the addition of
the entire paragraph immediately before summary.
REFERENCES
1. D.V. Ragone: Thermodynamics of Materials, John Wiley & Sons. Inc.,
New York, NY, 1995, vol. II, pp. 106-08.
2. D.A. Porter and K.E. Easterling: Phase Transformations in Metals
and Alloys, 2nd ed., Chapman & Hall, London, 1997, pp. 44-47.
3. H.B. Aaron and G.R. Kotler: Metall. Trans., 1971, vol. 2, pp. 393-408.
4. C.M. Sargent and G.R. Purdy: Scripta Metall., 1974, vol. 8, pp. 569-72.
5. G. Sauthoff : J. Phys. IV (Coll. CI), 1996, vol. 6, pp. 87-97.
6. D.R. Gaskell: Physical Metallurgy, 4th ed., R.W. Cahn and P. Haasen,
eds., North-Holland, Amsterdam, 1996, Part I, pp. 458-60.
7. G.R. Purdy: Met. Sci. J., 1971, vol. 5, pp. 81-85.
8. R.K. Trivedi: in Lectures on the Theory of Phase Transformations,
H.I. Aaronson, ed., TMS-AIME, New York, NY, 1975, pp. 51-81.
9. M. Hillert: in Lectures on the Theory of Phase Transformations, H.I.
Aaronson, ed., TMS-AIME, New York, NY, 1975, pp. 1-50.
10. J.W. Christian: The Theory of Transformations in Metals and Alloys,
2nd ed., Pergamon Press, Oxford, United Kingdom, 1975, pp. 179-82.
11. J.W. Martin, R.D. Doherty, and B. Cantor: Stability of Microstructures
in Metallic Systems, 2nd ed., Cambridge University Press, Cambridge,
United Kingdom, 1997, pp. 242-47.
12. M. Hillert: private communications, 2001.
13. M. Hillert: Phase Equilibria, Phase Diagrams and Phase Transformations, Cambridge University Press, Cambridge, United Kingdom,
1998, pp. 143-50.
14. R.T. Dehoff : Thermodynamics in Materials Science, McGraw-Hill,
Inc., New York, NY, 1993, pp. 377-82.
15. R.G. Mortimer: Physical Chemistry, 2nd ed., Academic Press, San
Diego, CA, 2000, pp. 209-11.
16. R.A. Alberty: Physical Chemistry, 7th ed., John Wiley & Sons, New
York, NY, 1995, pp. 217-18.
METALLURGICAL AND MATERIALS TRANSACTIONS A

17. C.H.P. Lupis: Chemical Thermodynamics of Materials, North-Holland,

New York, NY, 1983, p. 161.
18. C.E. Pearson: Handbook of Applied Mathematics, 2nd ed., Van Nostrand Reinhold Company Inc., New York, NY, 1983, pp. 96-97.
19. S.K. Bhattacharyya and K.C. Russell: Metall. Trans., 1972, vol. 3,
pp. 2195-99.
20. M. Qian and L.C. Lim: Scripta Mater., 1998, vol. 39, pp. 1451-55.
21. M. Qian and L.C. Lim: Metall. Mater. Trans. A, 2000, vol. 31A, pp.
2659-61.
22. M. Qian: Metall. Mater. Trans. A, 2000, vol. 31A, pp. 3195-97.
23. J.W. Cahn and F. Larche: Acta Metall., 1982, vol. 30, pp. 51-56.
24. J.W. Cahn: Acta Metall., 1980, vol. 28, pp. 1333-38.
25. M. Hillert and J. Agren: Acta Mater., in press.

Infrared Imaging during Low-Cycle

Fatigue of HR-120 Alloy
H. WANG, L. JIANG, Y.H. HE, L.J. CHEN, P.K. LIAW,
R.R. SEELEY, and D.L. KLARSTROM
Low-cycle fatigue behavior is an important engineering
parameter for superalloys under cyclic loading conditions.
While many researchers have conducted extensive studies
regarding low-cycle fatigue behavior of superalloys,[16]
very limited work was performed on HR-120 alloy. A typical
low-cycle fatigue study of high-temperature alloys includes
(1) fatigue tests at different strain levels as a function of
temperature, to generate a fatigue life vs total strain range
plot; (2) peak stress amplitude vs cycles to study cyclic
hardening or softening; and (3) microstructural studies of
specimens with different degrees of fatigue damage. Nondestructive evaluation tools, such as optical imaging and acoustic emission, are often used to monitor the fatigue process.
During strain-controlled fatigue tests, a significant increase
in specimen temperature occurred, especially at high strain
levels. This rise in specimen temperature is a well-known
fact, because almost all the energy of cyclic loading is expected to convert into heat. The specimen temperature can be
measured by attaching a thermocouple. However, contact
temperature measurements at a single point are not always
practical, and they provide limited useful information to
fatigue studies. Therefore, the temperature evolution during
the fatigue process has not been very well documented, and
it has not been using to interpret low-cycle fatigue data.
Infrared (IR) imaging, also called IR thermography, can
provide temperature maps of a specimen subjected to cyclic
loading. The IR images show the energy flow and heat
conduction in real time during the entire fatigue process.
Much more information can be derived from the temperature
maps as compared to thermocouple measurements. Only
limited studies using IR thermography during fatigue
tests,[7,8] with temperature resolution of 0.1 C can be found

H. WANG, Research Staff Member, is with the Oak Ridge National

Laboratory, Oak Ridge, TN 78931-6064. L. JIANG, Metallurgist, is with
GE CR&D, Schenectady, NY. Y.H. HE, Professor, is with National Key
Laboratory of Powder Metallurgy, South-Central China University. L.J.
CHEN, Visiting Scientist, and P.K. LIAW, Professor, are with the Department of Materials Science and Engineering, University of Tennessee at
Knoxville, Knoxville, TN 37996. R.R. SEELEY, manager, and D.L.
KLARSTROM, Director of Technology, are with Haynes International,
Inc., Kokomo, IN 46904.
Manuscript submitted January 17, 2001.
VOLUME 33A, APRIL 20021287