Acta metall, mater. Vol. 40, No. 9, pp. 2337 2342, 1992 0956-7151/92 $5.00 + 0.

00
Printed in Great Britain. All rights reserved Copyright t" 1992 Pergamon Press Ltd

GIBBS-DUHEM A N D CLAUSIUS-CLAPEYRON TYPE

EQUATIONS FOR ELASTICALLY STRESSED CRYSTALS
W. C. J O H N S O N t and H. S C H M A L Z R I E D z
tDepartment of Metallurgical Engineering and Materials Science, Carnegie Mellon University,
Pittsburgh, PA 15213-3890, U.S.A. and 2lnstitut f/ir physikalische Chemie und Elektrochemie,
Universit/it Hannover, W-3000 Hannover 1, Germany

(Received 13 December 1991)

Abstract--General thermodynamic expressions relating the intensive variables in homogeneously

deformed, nonhydrostatically stressed crystals are derived for both open and closed systems
(Gibbs-Duhem equation). These equations are then combined with the equilbrium conditions to write
Clausius-Clapeyron type equations for crystal-gas and multiphase crystalline systems. As the components
of the elastic deformation (strain) can be either thermodynamic intensive variables or densities, depending
upon the imposed boundary conditions and system geometry, various Gibbs Duhem type equations exist;
the equation to use depends on the problem under consideration.

Rrsum~-Des expressions thermodynamiques g~nrrales, qui relient les variables intensives dans des
cristaux d~formrs d'une manirre homogrne et contraints d'une manirre non hydrostatique, sont 6tablies
~i la lois pour les systrmes fermrs et ouverts (equation de Gibbs et Duhem). Ces +quations sont alors
combinres avec les conditions d'rquilibre pour 6crire des 6quations du type Clausius et Clapeyron pour
des systrmes cristal-gaz et des systrmes cristallisrs polyphasrs. Comme les composants de la d&ormation
61astique peuvent &re soit les variables, soit les densit~s thermodynamiques intensives selon les conditions
aux limites imposres et la g~om+trie du syst~me, on obtient diffrrents types d'~quations de Gibbs et
Duhem; l'~quation ~i utiliser drpend du probl~me consider&

Zusammenfassung--Allgemeine thermodynamische Ausdrficke, die die intensiven Variablen in homogen

verformten, nicht-hydrostatisch verspannten Kristallen verknfipfen, werden sowohl ffir offene als auch ffir
geschlossene Systeme abgeleitet (Gibbs Duhem-Gleichungen). Diese Gleichungen werden dann mit den
Gleichgewichtsbedingungen kombiniert und ffihren zu Gleichungen yon der Art Clausius-Clapeyron ffir
Kristall-Gas- und ffir vielphasige kristalline Systeme. Da die Komponenten der elastischen Verformung
(Verzerrung) entweder thermodynamische intensive Variabten oder Dichten sein krnnem je nach
eingesetzten Randbedingungen und Systemgeometrien, existieren verschiedene Gibbs-Duhem-Gleichun-
gem die Gleichung, die benutzt werden muB, hfingt von dem zu behandelnden Problem ab.

l. INTRODUCTION fields, like thermodynamic potentials, have the prop-

Thermodynamic variables are often classified as
either intensive or extensive. Extensive variables
depend upon the size of the system; the doubling of
the size of a single-phase homogeneous system leads
to a doubling of the extensive properties (e.g. energy,
entropy and mass). In contrast, an intensive variable
is independent of the size of the system. Those
intensive variables that are constant within the phase
and equal between phases in multiphase systems at
equilibrium are called thermodynamic potentials. If
the product of an intensive and extensive variable has
the dimensions of energy, then they are considered to
be conjugate variables.
Certain thermodynamic variables, such as the con-
jugate elastic stress and deformation tensors, do not
easily lend themselves to classification as either inten-
sive or extensive variables [1]. In such cases it is
advantageous to introduce a more general classifi-
cation scheme based on the concept of thermodyn-
amic field variables and densities [2]. Thermodynamic
erty that they assume identical values in and between
phases that are in equilibrium. Thermodynamic den-
sities remain constant upon the addition of two
identical systems, but are not equal between phases at
equilibrium. They are conjugate variables to the
thermodynamic field variables. Examples of thermo-
dynamic densities include the energy, entropy and
mass densities and certain components of the stress
and strain tensors.
The Gibbs D u h e m equation, as derived by
Gibbs for a fluid phase [3], is the differential of a
fundamental equation that connects the thermodyn-
amic potentials or field variables necessary for the
thermodynamic description of the phase. In conjunc-
tion with the equilbrium conditions, it forms the basis
of both the Clausius-Clapeyron type equations and
the phase rule and is often invoked in phenomeno-
logical transport equations to eliminate the interde-
pendency of the different driving forces [4]. Although
Gibbs and many subsequent investigators have
shown that the thermodynamics of crystalline solids

2337
2338 JOHNSON and SCHMALZRIED: EQUATIONS FOR ELASTICALLY STRESSED CRYSTALS
differs fundamentally from that of fluid systems, the
Gibbs-Duhem equation as derived for fluids is still
routinely applied to crystalline systems. Such appli-
cations are not valid in general, as the Gibbs-Duhem
equation for fluids accounts only for the chemical
components and does not consider the various struc-
ture elements (e.g. vacant lattice sites, interstitials of
various charge) arising from the crystalline nature of
the phase. In addition, crystals are almost never
hydrostatically stressed; i.e. it requires more than just
knowledge of the local pressure (or trace of the stress)
to describe the local thermodynamic state of a crystal.
In this work, we examine Gibbs-Duhem relations
for homogeneous, nonhydrostatically stressed, crys-
talline solids in both open and closed systems. The
Gibbs-Duhem equations are then combined with the
thermodynamic equilbrium conditions to illustrate
how Clausius-Clapeyron type equations for crystal-
gas and multiphase crystal systems can be obtained.
2. GIBBS--DUHEM EQUATION
We first briefly review two derivations of the
Gibbs-Duhem equation (GDE) for fluids that are
based on open and closed system approaches,
respectively. This provides a basis for contrasting the
similarities and differences of the GDE for crystalline
solids.
2.1. Fluids
The combined form of the first two laws of thermo-
dynamics (the Gibbs equation) for a simple bulk fluid
phase can be written for an open system as
dE=OdS-PdV+ ~ / t j dNj (1)
j=l
where E is the internal energy, S the entropy, 0 the
temperature, P the pressure, V the volume of the
fluid, and/~j and Nj are the chemical potential and
number of atoms, respectively, of thejth component.
The summation extends over the n different mass
components. The fundamental equation, equation
(1), connects the extensive variables. As E, S and Nj
are all proportional to V, equation (1) is a homo-
geneous equation of first order. It can be integrated
using Euler's theorem to obtain [5]
E=OS-PV + ~ #iNj. (2)
j=l
Differentiating equation (2) and using equation (1)
yields the Gibbs-Duhem equation
V d P = S dO + ~ Nyd#j. (3)
j=]
An equation connecting the thermodynamic field
variables can also be derived by expressing the Gibbs
equation on a per-unit-volume basis
de, = 0 ds, + ~/~j dpy (4)
)=l
where the subscript v denotes a density, measured per
unit volume. For example, P2 is the number of atoms
of component 2 per unit volume of fluid. Equation (4)
is the closed form of the Gibbs equation. It indicates
that only the number density of the components and
entropy can be independently varied in a given
volume element. Equivalently, an equal number of
thermodynamic fields will be independently variable.
Defining the thermodynamic potential, o~v, as
~o, = ev - Os~- ~ #j Ps (5)
j=l
gives
dco~= -s~dO - ~ pjd#s. (6)
j=l
0 and the/~j are clearly thermodynamic fields. That ~ov
is also a thermodynamic field can be proven by
considering the conditions for phase equilibria
between two fluids. In addition to the equality of
temperature and chemical potentials between phases,
there is an interfacial equilibrium condition obtained
from the virtual variation of allowing the transform-
ation of one phase (~) into the other phase (fl) [3]. At
equilibrium
co~,- o~. (7)
Since the fluid phases are homogeneous in thermo-
dynamic state at equilibrium, the co, potential is
clearly a thermodynamic field and equation (6) is the
GDE connecting the thermodynamic fields. If necess-
ary, the identification o~, = - P can be made using
equation (2). It is important to note that the
equations connecting the thermodynamic potentials
in the open-system formulation and the thermodyn-
amic field variables in the closed-system formulation
are identical for fluids.
2.2. Crystals
As noted by Schottky [6], the number of statisti-
cally distinguishable constituents of a crystal (the
structure elements) is usually greater than the number
of its chemical components. At the same time, how-
ever, the variations permitted in the structure
elements are constrained by the lattice itself. The
lattice constraints give rise to diffusion potentials, in
contrast to chemical potentials, which can be associ-
ated with the statistically distinguishable structure
elements. When local equilibrium obtains, the num-
ber of independent diffusion potentials again reduces
to the number of independent chemical components.
In the following discussion of GD equations for
closed and open systems, we assume that local equi-
librium is satisfied.
2.2.1. Closed systcm. To obtain an equation con-
necting the thermodynamic fields of a crystal, the
Gibbs equation is written in closed form for an n
component system as [7]
de,. = 0 ds,. + Tki d~k + ~ Mjt dp; (8)
j=l
 JOHNSON and SCHMALZRIED: EQUATIONS FOR ELASTICALLY STRESSED CRYSTALS
where the prime superscript indicates that all quan-
tities are measured on a per unit reference volume
basis, with the reference state assumed to be the
stress-free state. T is the first Piola-Kirchhoff stress
tensor, F is the deformation gradient, and Mjl is the
diffusion potential [7]. Here, vacancies (/) have been
arbitrarily chosen as the dependent structure element
in the diffusion potential. For a crystal in chemical
equilibrium with a fluid, Mjt = # j [7].
Equation (8) differs from the corresponding closed-
system expression for fluids, equation (4), in two
important ways. First, is the presence of the mechan-
ical work term indicating that the actual size and
shape of a unit volume of reference crystal can be
changed without accretion of mass. Secondly is the
introduction of the diffusion potential instead of a
chemical potential. The diffusion potential, which
also must be equal between and within phases at
equilibrium, emphasizes the crystallinity and struc-
ture of the solid. It represents the change in energy
associated with the exchange of two atomic species
(or structure elements) [7] holding the number of
lattice sites or unit cells fixed. So long as the lattice
is conserved, the addition of an atom to a unit volume
of crystal requires either the removal of another
atomic species or the annihilation of a vacant lattice
site, either in a regular or an interstitial sublattice.
The following results are entirely independent of the
choice of the dependent structure element.
The classification of the stress and strain com-
ponents as thermodynamic fields or densities depends
upon the system geometry and mechanical loading
conditions for a crystal [1]. As an example, consider
the two-phase system depicted in Fig. 1. The a-fl
interface is coherent and lies in the x~-x2 plane.
Traction boundary conditions are applied along the
x3 axis while either a displacement or traction bound-
ary condition can be applied along those surfaces
perpendicular to the x~ and x2 directions. In this
simple model, it is assumed that the shear com-
ponents of the stress and deformation vanish. The
coherency constraint requires continuity of defor-
mation across the interface. This gives
F~l = F,~ (9)
and
F~2 = F2~2. (lO)
In addition to the equality of diffusion potentials and
temperature between phases, equilibrium between the
phases requires [1]
T~3 = T~3 = T~3 3 (11)
and
q~, = ~v~, -- T~3 F]3 = ~ , = (.or0, - T3#3F~3 (12)
where T°3 is the mechanically applied stress in the x3
direction. Equation (12) is an interfacial equilibrium
condition obtained by allowing a virtual phase tran-
sition of one phase into the other. As such, the 4~o,
2339
%
I
I
I
I
I I
I I
I I
I I
I I
I
I
I
I
I I
I I
I I
I I
L -'
Tg3
a) Displacement Boundary Condition
r°3
-r,°, [
-7 li
rg3
b) Traction Boundary Condition
Fig. 1. A system geometry in which the phases are nonhy-
drostatically stressed and homogeneous in state. The mech-
anical boundary conditions can be given by specifying the
displacement (a) or the traction (b) on the edge of the planes.
potential in coherent crystalline systems is the equiv-
alent of the o),, potential in fluid systems.
Equations (9)-(11) show that the deformation
components Fix and F22 and the stress component T33
are thermodynamic fields. They are constant within
each phase and equal between phases at equilibrium.
In addition, the ~b,,,potential is a thermodynamic field
as the phases are homogeneous in the assumed
configuration. Differentiating q~,,.and using equation
(8) gives the fundamental equation
dq~, = -s~,dO + Tll dF u + T22dF22
- F 3 3 d T 3 3 - ~ p.;dM/,. (13)
j= 1
The fundamental equation, equation (13), connects
the thermodynamic fields for the system of parallel
plates depicted in Fig. 1. As such it can be considered
the Gibbs-Duhem equation for this nonhydrostati-
cally stressed crystal under closed-system conditions.
If the crystal experiences a hydrostatic stress and
one chooses to classify the stress components as the
thermodynamic field variables, the corresponding
fundamental equation, which could also then be
considered as the Gibbs-Duhem equation, is
dhv,= -so, dO + J dP - i p~ dMjt (14)
j= I
2340 JOHNSON and SCHMALZRIED: EQUATIONS FOR ELASTICALLY STRESSED CRYSTALS
where J, the determinant of the deformation tensor F,
is the ratio of the actual volume to the reference
volume and P = -Tkk/3.
In contrast to the "closed-system" GDE for a fluid,
equation (14) indicates that 0, P and the Mjt can be
independently varied in a given volume element of
hydrostatically stressed crystal. The potential, he,
where
h~,= -s~.O + J P - ~ p~Mjt (15)
j=]
is a dependent thermodynamic potential that is a
function of 0, P and Mjt. The independence of the
temperature, pressure and diffusion potentials in a
crystal element should not be surprising. Imagine that
the pressure (stress state) of a given volume element
in a large crystal that is initially in equilibrium and
under hydrostatic pressure is changed at constant
temperature. In a multicomponent system, the den-
sities of the various components (including vacancies)
would change in such a way that the diffusion
potential of every component remains constant and
equal to those of the surrounding crystal. (The sur-
rounding crystal acts as a chemical potential reser-
voir.) In a one component system, the vacancy
concentration in the volume element would change.
The final thermodynamic state of the element pos-
sesses the same temperature and diffusion potentials
as the original state, although the pressure (or state
of stress) has changed. This is precisely what occurs
in solute redistribution around dislocations and pre-
cipitates and can be treated using the open-system
elastic constants approach of Larch6 and Cahn [7].
2.2.2. Open system. Care must be taken in writing
the open-system Gibbs equation for a nonhydrostat-
ically stressed crystal in that the accretion terms,
corresponding to a change in the number of lattice
sites of the crystal, depend explicitly on how (and
where) mass is added or lattice is created [8]. This
problem was first treated by Gibbs for a solid [Gibbs'
equation (367)] and extended by Herring [9], among
others [7, 10], to stressed crystals containing
vacancies. If we imagine a hydrostatically stressed
crystal that does not exchange mass with its surround-
ings but is able to create vacancies at its surface,
thereby changing the number of unit cells comprising
the crystal, equilibrium is obtained when [7]
o9~,= --JP (16)
where P is the mechanically imposed pressure acting
on the crystal. Comparison of equation (16) with
equation (15) indicates that the potential h~, = 0 in a
system capable of changing the number of its cells or
vacancies, when the number of atoms of each mass
component is held fixed [10]. At equilibrium, dh,, = 0,
and the open-system Gibbs-Duhem equation for the
crystal, in view of equation (14), becomes
J dP = s., dO + ~ pj dMjt . (17)
j=l
Equation (17) also follows directly from the differen-
tiation of equation (16). Dividing equation (17) by
J yields the usual expression for the GDE
d e =sodO + ~ pjdMjt. (18)
j=1
Thus the open-system form of the GDE for a crystal
is different from the closed-system form.
3. CLAUSIUS-CLAPEYRON TYPE EQUATIONS
The generalized Clausius421apeyron equations
(CCE) express the interdependence of the thermodyn-
amic field variables for coexisting phases in equi-
librium. They give the "slope" of the coexistence
curve between two or more phases in a field diagram
or phase diagram of the first type (a phase diagram
depicting phase equilibria as a function of the ther-
modynamic field variables) [5]. In systems composed
entirely of fluid phases, each phase possesses its own
Gibbs-Duhem equation connecting the thermodyn-
amic field variables for the phase. The individual
GDE are connected through the conditions demand-
ing that the thermodynamic field variables be equal
between phases at equilibrium. This same procedure
is applicable to nonhydrostatically stressed systems as
shown below.
3.1. Crystalline systems
We consider equilibrium between two or more
phases which possess the geometry shown in Fig. 1.
The thermodynamic field variables have been
identified as 0, Mjt, FII , F22 , T33 and q~,.,. If we
designate the dependent thermodynamic field as
q~ = h0 and the independent thermodynamic fields by
ht (i = 1, 2 . . . . . k) with corresponding conjugate
thermodynamic density di defined by
d, = Oh__~o (19)
Ohi
then the GDE can be written as
k
E di dh, = o (20)
i=o
where we have defined do = - 1. If we consider a
system of two coexisting phases in which the exper-
imentally established thermodynamic fields, h4,
hs . . . . , hk are held fixed, then one obtains for the
generalized CCE [5]
dh I (dB2/d~) - - ( d ~ / d ~ )
dh--~2= (d~/d~) - (d~/d~)" (21)
If q~, is chosen as the dependent variable, one
obtains from equation (13)
(s~,-s~,) d0 + ( r ~ l - T~,) dF], + (r~2- T~2) dF22
+(F~3-F~3) aT33+ ~ (p~'-pf) dMjt = 0. (22)
j=l
 JOHNSON and SCHMALZRIED: EQUATIONS FOR ELASTICALLY STRESSED CRYSTALS
An expression similar to equation (22) was obtained
earlier examining phase equilibria between elastically
stressed heteroepitaxial thin films [11]. The equi-
librium conditions can also be applied to heteroge-
neously stressed systems, at least in a local sense at
the interface, to obtain a Gibbs-Duhem equation for
the interface [12].
3.2. Crystal-gas systems
Two different crystal-gas (or crystal-fluid) systems
can be imagined. The first is a system in which the
crystal is in a state of hydrostatic stress with the
mechanical state of the crystal being determined by
the pressure of the gas. This is the classic problem and
the corresponding CCE are obtained directly from
the open-system G D E for the crystal under hydro-
static pressure [equation (18)] and the G D E for the
gas [equation (3)]. If the pressure is chosen as the
dependent variable, then the corresponding CCE is
(s,, - s{) dO + ~ (pj - pf) d#j = 0 (23)
i= I
where the superscript f denotes the gas phase.
A second possible system consists of a nonhydro-
statically stressed crystal in equilibrium with a fluid.
One possible geometry, intended to model a thin film
in equilibrium with the melt from which it is grown,
is depicted in Fig. 2. The normal stress acting on
the crystal, T33, is assumed to be established by the
pressure of the fluid.
The interracial equilibrium condition valid at the
crystal-liquid interface is [7]
(o,,, = J~o{. (24)
Differentiating equation (24) and using the relation-
ship o9{= - P yields
(s .... Js{) dO - Tk, dF~k
+ ~ (pj - Jpf) d#j = P dJ. (25)
j-1
Equation (25) is further simplified by noting that
J=Fll[22/33 and 7 " 3 3 = - F uF22P. In the small
strain approximation, equation (25) becomes
X
T ~' Xl
Fluid
Crystal
Substrate
Fig. 2. Nonhydrostatically stressed planar film in equi-
librium with a fluid. The substrate imparts a displacement
field similar to that induced by the displacement boundary
conditions of Fig. l(a).
2341
(sv - s{) dO + ( p j - pf) d.j
y=l
--(all + P) den - (a22 + P) de22 = 0 (26)
where a~ and e,j are the Cauchy and small strain
tensors, respectively. The pressure or a33 component
of the stress tensor has been used as the dependent
variable in the CCE of equation (26).
If the pressure (a33) is to be established experimen-
tally, the appropriate CCE is determined as above; a
dependent thermodynamic field is chosen and the
equilibrium conditions are then employed. For
example, if #j is chosen as the independent variable,
it follows from the equilibrium condition d/l~= dM~
that
da),,, + s~,,dO - Tki dFik q- ~" p~ dM/~
j=2
where J* = p~/p{. Using the equilibrium condition,
equation (24), equation (27) simplifies in the small
strain limit to
(s, - J's{) dO + ~, ( p i - J*p~) dl~,
j-2
+(J*-J)dP-(an + P)-(a22+ P)=O. (28)
4. DISCUSSION
Unlike fluid systems, the open-system GDE for a
crystal differs from its closed-system form, even when
the elastic state of the crystal is a hydrostatic press-
ure. This difference reflects the two distinct ways in
which an atom can be added to a crystal, with or
without a change in the number of lattice sites. This
difference does not exist for a fluid system and the
open and closed-system G D E for a fluid are identical.
Which G D E should be used depends upon the
problem under consideration. Phenomena concerned
with the internal crystal would require use of the
closed-system GDE. This includes diffusion studies
and the determination of solute redistribution. The
open-system form under hydrostatic stress might be
more appropriate when examining properties of the
crystal as a whole.
The physical properties of the system may also
dictate which form of the G D E is most appropriate
to employ. Consider a simple oxide, AO, in contact
with oxygen gas, 02. The crystal will endeavor to
establish equilibrium with the gas by the transfer of
oxygen to the gas, if the chemical potential of oxygen
in the gas is less than that of the crystal, or by
accretion of oxygen to the crystal when the opposite
situation obtains. This process requires that charge
neutrality be maintained and, hence, that electronic
transport is involved in addition to ion transport
(1/202 + 2e' = 02-) [13].
If the diffusivity of the cation in the crystal, D A , is
much larger than the diffusivity of the oxygen ion,
2342 JOHNSON and SCHMALZRIED: EQUATIONS FOR ELASTICALLY STRESSED CRYSTALS
D O , the crystal will behave essentially as an open
system (for an appropriate time scale). The supply of
the cation to the interface or its removal from the
interface is sufficiently fast to maintain equilibrium
and to allow the new equilibrium concentration of
cation vacancies to form through the creation or
destruction of unit ceils at the interface. If, on the
other hand, D A ~ Do, the flow of cations to (or away
from) the interface will be insufficient to allow unit
cells of AO to be accreted (dissolved) at the interface
and the number of lattice sites of unit cells in the
crystal will remain approximately constant. In this
ease, the entire crystal might best be treated as a
closed system, Similar arguments would obtain for
the crystal in contact with a gas of the metallic
component.
A strain volume tensor, ~ij, is sometimes defined
as [14]
Qo = Ve~ (29)
where V and e are the volume and strain tensor,
respectively. ~ is assumed to be an extensive quantity
while the corresponding conjugate variable, the
stress, is the intensive variable. Although this desig-
nation might be useful in certain applications, it can
not be extended to multiphase crystalline systems.
Indeed, we have already seen that the stress com-
ponents can behave as either thermodynamic field
variables or thermodynamic densities and that their
proper identification as such is necessary for the
construction of phase diagrams and common-tangent
constructions in stressed systems [I].
5. SUMMARY
We have examined the Gibbs-Duhem equation,
coupling the thermodynamic potentials or field vari-
ables, for nonhydrostatically stressed crystalline
solids. The closed-system representation of this
equation, referred to a unit volume of crystal for
which the number of lattice sites or unit cells does not
change, demonstrates that the temperature, diffusion
(chemical) potentials, and elastic state are indepen-
dently variable. This is true even when the crystal is
hydrostatically stressed. The open-system represen-
tation of the GDE, in which the crystal is free to
change its number of lattice sites and is hydrostati-
cally stressed, is identical to that of its fluid counter-
part; the pressure, temperature and diffusion
potentials cannot be independently varied. The
appropriate form of the G D E to use, depends upon
the problem considered.
Clausius-Clapeyron type equations, coupling the
thermodynamic field variables between coexisting
phases in equilibrium, can be obtained from the GDE
of each phase for nonhydrostatically stressed crystals.
Acknowledgements--WCJ gratefully acknowledges the sup-
port of the Alexander von Humboldt Foundation through
a fellowship at the Institut fiir physikalische Chemic und
Elektrochemie, Universit/it Hannover, and the U.S.
National Science Foundation through grant I) MR-
8620026.
REFERENCES
1. W. C. Johnson and W. H. Mfiller, Acta metall, mater.
39, 89 (1991).
2. R. B. Griffiths and J. W. Wheeler, Phys. Rev. A 2, 1047
(1970).
3. J. W. Gibbs, The Scientific Papers--Vol. 1. Dover,
New York (1961).
4. S. R. de Groot and P. Mazur, Nonequilibrium Thermo-
dynamics. North Holland, Amsterdam (1963).
5. H. Schmalzried and A. Navrotsky, Festk6rperthermody-
namik. Verlag Chemic, Weinheim, Germany (1975).
6. W. Schottky, Halbleiter IV. Vieweg, Braunschweig
(1959).
7. F. C. Larch6 and J. W. Cahn, Acta metall. 21, 1051
(1973).
8. W. C. Johnson and J. W. Cahn, Scripta metall. 15, 1259
(1981).
9. C. Herring, The Physics of Powder Metallurgy (edited
by W. E. Kingston), Chap. 8. McGraw-Hill, New York
(1951).
I0. W. W. Mullins, in Proc. Int. Conf. on Solid-Solid Phase
Transformations (edited by H. I. Aaronson, D. E.
Laughlin, R. F. Sekerka and W. M. Wayman), p. 49.
TMS-AIME, Metals Park, Ohio (1982).
11. W. C. Johnson and C. S. Chiang, J. appl. Phys. 64, 1155
(1988).
12. P. W. Voorhess and W. C. Johnson, J. chem. Phys. 90,
2793 (1989).
13. G. Kutsche and H. Schmalzried, Solid St. Ionics 43, 43
(1990).
14. J. C. M. Li, Metall. Trans. 9A, 1353 (1978).