SURFACE SCIENCE 3 (1965) 429-444 o North-Holland Publishing Co.

, Amsterdam

GENERALIZATION OF THE GIBBS-THOMSON EQUATION

CHARLES A. JOHNSON
Edgar C. Bain Laboratory for Fundamental Research, United States Steel Corporation
Research Center, MonroeviNe, Pennsylvania, U.S.A.

Received 24 May 1965

The Gibbs-Thomson equation relates the chemical potential of the vapor in equilibrium
with a spherical drop to the radius and isotropic surface free energy of the drop. It is
shown that this equation has a simple generalization to the case of arbitrary anisotropic
surface free energy. This general form of the Gibbs-Thomson equation provides a simple
and direct connection between the size and shape of a crystal in equilibrium with its vapor,
the chemical potential of the vapor in equilibrium with the crystal and the detailed behavior
of the specific surface free energy, y, as a function of surface orientation.
HerringI) has given two equations which relate the chemical potential of the vapor in
equilibrium with a smoothly curved surface element or a facet on a given body of arbitrary
(i.e., not necessarily equilibrium) shape to the detailed behavior of y as a function of
surface orientation and to the size and geometry of the given body. It is shown that
Herring’s equation for a smoothly curved surface element can be made to take a consider-
ably simpler form by comparing the geometry of the given body to that of an equilibrium
body; if the given body does in fact have the equilibrium shape the still simpler general
Gibbs-Thomson equation is obtained. Herring’s equation for a facet is also shown to yield
the generalized Gibbs-Thomson equation when the given body is an equilibrium body.

1. Introduction

For a one component system the Gibbs-Thomson equation relates the

chemical potential, p”(r), of the vapor in equilibrium with a spherical
droplet of radius r to the chemical potential, &co), of the vapor in equi-
librium with a flat surface of the same substance at the same temperature
through

where y is the (isotropic) specific surface free energy of the droplet and LI,,
is the molecular volume in the condensed phase. This equation is readily
derived by considering the transport of material from the flat surface to the
droplet so as to increase the radius of the droplet.2)
The chemical potential of the vapor in equilibrium with a small body will
be uniform if and only if the body is in equilibrium with respect to changes

429
430 CHARLES A. JOHNSON

of shape. Equation (1) applies only to the case of isotropic surface free energy
and spherical drops. It is clear, for example, that one cannot apply equation
(1) to the rounded edge of a plate-like precipitate: for, if the surface free
energy is assumed to be isotropic then the particle is not in shape equilibrium,
while if it is supposed that the plate-like shape is the equilibrium shape, then
the surface free energy is certainly not isotropic.
A crystalline solid will, in general, have a specific surface free energy
which varies with the orientation of the surface relative to the crystal axes:
if surface orientation is denoted by the unit normal to the surface, A, and y
is the specific surface free energy, then y=y (A). The question of the equi-
librium shape of a small crystal was posed by Gibbss) and solved under
some restrictions by Wulff4); Wulff’s solution was first proved rigorously
by Dinghas5), and both solution and proof were shown to apply to the most
general variation of y(A) by Herring.6) One can expect that a generalization
of the Gibbs-Thomson equation will hold for equilibrium-shaped particles
no matter what the behavior of y(A). Such a generalization, valid for equi-
librium shapes which are (ideal) polyhedra, has been given by Volmer.7)
As Herring6) has pointed out, polyhedra from only a rather special subclass
of the possible equilibrium shapes of crystallites, and recent experimental
work has clearly shown that a variety of other forms 0ccur.s)
The principal result to be presented here is a generalization of the Gibbs-
Thomson equation which is valid for any variation of y(ii) whatever. A simple
derivation, based on Wulff’s construction, will be given first. It will also be
shown that the equation can be derived from the equations which Herringl)
has given for the chemical potential of the vapor in equilibrium with a body
of arbitrary shape and arbitrary y(h). In this connection Herring’s general
equations will be put into forms which depend only on the geometry of the
given body and that of an equilibrium-shape body of the same substance at
the same temperature.
The term vapor, as used here, is intended to denote the phase surrounding
the small (condensed) body of interest, whether this is a vapor phase or a
condensed phase. In applying the Gibbs-Thomson equation, as derived here,
to solid state problems, it is necessary that: (1) the precipitate particle be in
shape equilibrium ;* and (2), there be no significant strains in the system.
(The further generalization to multi-component systems follows directly
along the lines set down by Volmer.)T) With these conditions satisfied, the
generalized Gibbs-Thomson equation relates the enhanced solubility of a

* For a crystalline particle contained in a crystalline matrix, different precipitate-matrix

orientation relationships will generally lead to differing metastable equilibrium shapes
(and, hence, to differing y-plots). The results derived in this paper are valid for any par-
ticular orientation relationship which is stationary.
 GENERALIZATION OF THE GIBBS-THOMSON EQUATION 431

small particle and, equivalently, the boundary values of the chemical po-
tential for the diffusion problem involved in the coarsening of precipitates
(Ostwald ripening)s), directly to y(A). Recent work on the coarsening of small
precipitate particles has shown that, for suitable systems, values of y(A) for
precipitate-matrix interfaces can be obtained in this way is).
In the following two sections the connection between y(A) and the equi-
librium shape (Wulff’s theorem) and the derivation of the generalized Gibbs-
Thomson equation are given for crystals possessing a center of symmetry.
In section 4 the same program is carried out without the restriction to
crystals having central symmetry. In section 5, Herring’s equations are
introduced and interpreted in terms of the geometry of an equilibrium body;
it is shown that both these equations reduce to the generalized Gibbs-
Thomson equation when applied to an equilibrium body. The results
are summarized in section 4.

2. Wulff’s construction

The construction given by Wulff4) determines the shape which, for given
y(A) and fixed volume, has the least surface free energy; this shape can be
referred to as the equilibrium shape (or body) corresponding to y(A).

Fig. 1. The y-plot and WullT’s construction of the equilibriumshape

432 CHARLES A. JOHNSON

The behavior of y as a function of surface orientation, fi, can be described

as follows: from a fixed origin, construct a vector p(A) along the direction
I? such that the length of p(A) is proportional to y(A) (fig. 1). These vectors
map out a closed surface which will be referred to as a “y-plot”. Following
Wulff*~e), construct the plane which passes through the end of, and is normal
to, p(a), for every A. The body which is contained within this set of planes
is then geometrically similar to the equilibrium shape (fig. 1).
From a practical point of view one ordinarily has the inverse of Wulff’s
problem: given the equilibrium shape, what is y(h)? It is convenient to split
the process of determining y(A) into two stages. The first is the determi-

Fig. 2. Construction of the relative y-plot from the equilibrium shape

(full inversion symmetry).

nation of relative values of y(A); the second step is to place these values on
an absolute scale. The Gibbs-Thomson equation, together with py from, say,
coarsening experiments, can provide this absolute scale. But the first step,
that of going from the equilibrium body to a relative y-plot, is not unique
unless the crystal structure under consideration has full (3-dimensional)
inversion symmetry. For the sake of simplicity the general case is treated
later; in this and the next section it is assumed that the crystal (and, hence,
the y-plot) has full inversion symmetry.
From an origin at the center of inversion of the equilibrium body, we
erect vectors r(A) which trace out the surface of the body (fig. 2). Wulff’s
theorem states that the normal distance, measured from a fixed origin, of a
 GENERALIZATION OF THE GIBBS-THOMSON EQUATION 433

surface element on an equilibrium body is proportional to the specific

surface free energy of that element. (For the present case, symmetry de-
termines the origin unambiguously.) Thus, defining

1(R) = r(A)*A (2)

we have r(h)= const. *L(A). To fix this constant, suppose that r(fi,)= yO is
known. Then, with I(&,)=&,

y(A) = FL(r) = +)+ (3)

0 0

for all orientations A which occur on the equilibrium body.

3. Derivation of the generalized Gibbs-Thomson equation

Consider a crystallite of a pure substance which is in equilibrium with its

vapor, and therefore has the equilibrium shape. The size of the crystallite is
specified by the value of A, relevant in equation (3). If the chemical potentials
of a molecule in the solid and vapor are respectively pLs(Ao)and &A,), then
equilibrium implies ,u,(n,)=&n,).* Similarly, for a flat surface (that is, the
surface of an infinitely large body) of the same substance at the same
temperature, p,( CO)=~~( co).
Now, following the treatment given by Lewis and Randallz), we consider
the changes in free energy consequent upon evaporating dn molecules from
the flat surface and condensing them on the equilibrium body in such a way
as to increase the size of the equilibrium body without changing its shape.
The volume free energy change in this process is

dFV = [cl,@,) - ~~(a)] dn T (4)

which must be equal to the change in surface free energy of the crystallite,
dF,. Because of the equality of solid and vapor chemical potentials, equation
(4) is equivalent to
dF, = [I*&~) - ~&o)]dn. (5)

The volume of the crystal may be written as

v = Kn; (6)

where K is a shape factor. In the process under consideration the shape is

invariant, and thus
dV = 3K$d(1,),

* As Herring I) has shown, the possibility of point defects in the crystal requires a generali-
zation of this equation. This has no effect on the equation for the chemical potential of
the vapor as a function of crystal size.
434 CHARLES A. JOHNSON

whence
dn = z = 3KG d &)
(7)
QO QO ’
where Q, is the molecular volume in the crystal.
In order to evaluate the change in surface free energy, we shall use
equation (3) to determine the total surface free energy of an equilibrium
body. Consider an equilibrium body, as in the previous section, whose
surface is mapped out by Y(A), r(d) being taken from the center of inversion
of the body. Equation (3) holds, with A,, being the normal distance of the
surface element of normal ii, from our chosen origin, and y(ii,)=y,.
The total surface free energy of the body is

F, = y(ri)dS (8)
s

where the integral is taken over the surface of the body. Using equation (3),
equation (8) becomes

F, = f r(A)*I?dS.
0 s
s

The divergence theorem, which states that

sS
f(x,y,z)-AdS =
s
”
VfdI’*

transforms the surface integral of equation (9) into the integral of V*r(fi)
over the volume, I’, of the equilibrium body. Since V *r(A) = 3, we obtain

FSzz3+ d&$ (II)

A0 s 0
V

Using equation (6) in (1 l),

F, = 3Ky,i;, (12)
and
dF, = 6Ky,A, d (A,).

Hence, from (7) and (12), with dF,=dF,,

2Y(A)
P,(I,)-P”(oo)=~R,=
0
I(A (13)

* The divergence theorem holds provided that (i) f is differentiable and has continuous
partial derivatives, and (ii) j f. S dS exists. Both conditions are fulfilled in the present
problem.
 GENERALIZATION OF THE GIBBS-THOMSON EQUATION 435

which is the generalization of the Gibbs-Thomson equation for crystals in

equilibrium with their vapor (for crystals having full inversion symmetry).
According to equation (13), the chemical potential of the vapor in equi-
librium with a small crystal is directly proportional to the specific surface
free energy of any given surface orientation, I?, and inversely proportional
to the width of the crystal (the distance between the parallel planes tangent
to the body at opposite sides) for the same orientation fi. (According to
Wulff’s theorem this ratio is the same for all surface orientations on an
equilibrium body.)
The proof of equation (13) contains no restrictive assumptions about the
“smoothness” of the equilibrium body: the proof is valid for polyhedral
equilibrium bodies as well as for smoothly curved equilibrium bodies. It does
not, of course, apply to orientations (such as those between two facets of a
polyhedron) which do not occur on the equilibrium body. It is apparent that
for a spherical equilibrium body (isotropic surface free energy) equation (13)
reduces to equation (1).

4. The Gibbs-Thomson equation for crystals of less than

full inversion symmetry

For crystals having less than full inversion symmetry the process of
constructing a y-plot from the equilibrium body is not uniquely determined;
this is because we no longer have physical reasons which uniquely determine
the origin for the construction. It is none the less possible to obtain a Gibbs-
Thomson equation which, in this case, relates pV to pairs of values of the
specific surface free energy for opposed surface orientations.
We choose an origin anywhere within the equilibrium body and erect
vectors r(li) which trace out the surface of the body (fig. 3). Let us suppose
that we know the absolute values of the specific surface free energy, yi = y(A,)
and y; = y( - A,), for a pair of opposed surface orientations + A,. Then,
defining the normal distances of the surface elements f A, from the chosen
origin to be
12; = r(A,)*A,
I, = r( - A,)-( - A,)
(14)

we have a y-plot, based on the chosen origin,

r(ri)*ii. (15)

This is not, in general, the “physical” y-plot, but it does have the two in-
variance properties required for the derivation of the Gibbs-Thomson
equation :
436 CHARLES A. JOHNSON

(1) The sum y(A) + y( - A) is the same no matter what the choice of origin.
(2) The total surface free energy is also independent of the choice of origin.
Equation (15) is the expression of Wulff’s theorem in the form: the sum of
the surface free energies for opposed orientations rtr A is proportional to the
width of the equilibrium body in the direction A. Using (15) the total surface
free energy is

F, = 1 y(R)dS = (~)jr(A)*AdS. (16)

S s

;r Equilibrium shape

Fig. 3. Construction of the relative y-plot from the equilibrium shape (general case).

The divergence theorem, when applied to (16), yields

T,=(~)SV.r(A)dV=3(~)V (17)
V

independent of the choice of origin.

On writing the volume of the body as

V = K (A,+ + A,)” (18)

and considering changes of size d(Ai +A,), the remainder of the derivation
is the same as in the preceding section, leading to

[A@,’ + A,> - A (41 = 2 (19)

which reduces to equation (13) for any orientation ri for which + A and - i?
are crystallographically equivalent.
 GENERALIZATION OF THE GIBBS-THOMSON EQUATION 437

5. Herring’s equations

Herringi) has given two equations which determine the chemical potential
of the vapor in equilibrium with any surface element of a crystal of arbitrary
shape for which y(A) is known. These equations determine, respectively, the
value of pV above a smoothly curved surface element and the average value
of pu, above a flat facet. The equations are only valid when equilibrium with
respect to transfer of matter between the vapor and solid exists; their appli-
cation to non-equilibrium shapes is therefore restricted to situations in which
transport through the vapor phase is much slower than transport across the
smface (e.g., solid-state reactions). For a non-equilibrium shape pu, will vary
over the surface, so that p”=p,(A).
These two equations will, in turn, be shown to reduce to equation (19)
(the Gibbo-Thomson equation) when the crystal under consideration has the
equilibrium shape. The derivations are carried out along the lines followed
in the previous section: no assumption as to crystal symmetry is made.This
involves an arbitrary choice of origin for constructing y(R) from the equi-
librium shape. The invariance properties mentioned in the previous section
are sufficient to make all final results independent of the choice of origin.

5.1. HERRING'S EQUATION FOR A SMOOTHLY CURVED SURFACE

Consider a surface element of orientation A, on a given body of arbitrary

shape. It is assumed that the principle radii of curvature, R,(R,) and R,(ri,),
of the body at A, are both non-zero and finite (the body is smoothly curved
and not “flat” in any direction), and it is also assumed that the second
derivatives of y with respect to changes of surface orientation are finite at
A, (so that the equilibrium shape is not “flat” at A,). Then, according to
Herringi), the difference between the chemical potential of the vapor in
equilibrium with the surface element A, on the given body, &R,), and the
chemical potential of the vapor in equilibrium with a very large body, c(“(co),
is given by

where R,(A,) and &(A,) are the principal radii of curvature of the body
at R,; these are the radii of the circles of curvature in the principal planes
Pr and PZ. The angle a1 is defined as follows : let A, be a variable unit normal
lying in the plane P1; then cos 8r E 4, -A,. Similarly, for A, lying in Pz,
cos 8, =A, .A,. Q, is again the molecular volume in the condensed phase.
The size of the body enters equation (20) through the values of R, and R,.
438 CHARLES A. JOHNSON

The quantities in the square brackets of equation (20) can be expressed

in terms of the geometry of an equilibrium body of the same substance.
To this end we introduce an equilibrium body, of size determined by
I(&-,)+ A( - &,)=A: + A,, with its crystal axes aligned parallel with those
of the given body. At A, on the equilibrium body we construct planes parallel
to P, and Pz ; these will not, in general, be the principal planes of the equi-
librium body at fi,.
To evaluate the quantity

1
a2y
[ y+ig fi,
we need to consider the behavior of am as ii varies, with li constrained to

(b)

Fig. 4. a) Projection of an equilibrium body onto one of the principal planes of the
given body.
b) Geometry of the curve Cl.

lie in PI. The surface elements, fi, of interest are those which are tangent to
the cylinder, circumscribed about the equilibrium body, whose generators
are normal to P, (fig. 4a)*. The intersection of this cylinder with the plane
P, is the shadow of the equilibrium body when illuminated normal to P,.
The curve, C,, which bounds this shadow determines the behavior of y(A)
for li lying in PI.
We choose an arbitrary origin, 0, within the equilibrium body and erect
the vectors r(R) which trace out the surface of the equilibrium body. Then,

* G. D. Chakerian, private communication.

 GENERALIZATION OF THE GIBBS-THOMSON EQUATION 439

according to equation (15),

yo+ _
P(A).A=J(A)
YP)=
( >
A,++ n,
r(fi).A.

is the support function, based on the origin 0, of the equi-

librium body. And, if we project 0 onto the plane Pi, thus defining an origin
0, for the curve C,, then, for orientations fi which lie in PI,A(h)is also the
support function of the curve C,. Thus

(21)

where A and d2A/de2, are to be measured along the curve C,.

Let p1 (B,) be the radius of curvature of C, at A,, and 6 be the vector from
0, to the center of this circle of curvature (fig. 4b). Then, for orientations A
sufficiently close to A,, r(A)=6+pi(A,,)ri, whence

n(h) = A.6 + p,(A,). (22)

On expressing the difference between neighboring surface unit normals ii
and A, in terms of 8, =cos-‘(A.&,) and evaluating the appropriate limit,
it can readily be shown that

(23)

so that, from equations (21), (22) and (23)

Y(h,) +(&g&=
(gi)Pl (fro). (24)

In like manner, if C, is the curve bounding the shadow of the equilibrium

body when illuminated normal to plane P2, and p,(ri,) is the radius of
curvature of C, at A,, then

y(A,) + (gi)P2@o).
(gg&= (25)

Using (24) and (25) we can now rewrite equation (20) in a form which
depends on the geometry of an equilibrium body, a pair of values of r(li)
and the geometry of the given body, and does not require any particular
crystal symmetry :

P”(fiO)- P”(W)= J-20 ($!)[$$ +$$] (26)

where R,(R,)and R,(A,)are the principal radii of curvature of the given
440 CHARLES A. JOHNSON

bodyat4. ~,(fid and ~~(4,)are the radii of curvature, at R,, of the shadows
of an equilibrium body on the principal planes, P, and Pz, of the given body.
These “shadow” curvatures are connected with the principal radii of curva-
ture of the equilibrium body at A, in the following way 11): let CIbe the angle
between PI and one of the principal planes of the equilibrium body at A, ;
then, if the principal radii of the equilibrium body are R’l”(fi,) and Rsq(R,),

p1 (ii,) = cos’c( Relq(i2,,)+ sin’cl Riq (A,)

p2 (A,) = sin2a R”,‘(A,) + cos2a Rzq(f?,) . (27)

For the special case a=0 (that is, the principal planes PI and P2 of the given
body at A, are also principal planes of the equilibrium body at R,) Equations
(27) reduce to
~1 (fi,) = R;‘(&,)
(28)
~2 (fir,> = R’,4(&,).

Suppose that the given body does in fact have the equilibrium shape; the
given body differs from the equilibrium body which has been used to evaluate
y(Ei) only by a scale factor. We adjust the size* of the equilibrium body to
make this scale factor unity: the given body and the equilibrium body are
now identical, so that RFq(li)=Ri(fi), i= 1, 2, and x=0 for all A. From equa-
tions (26) and (28)

(29)
which is equation (19), the generalized Gibbs-Thomson equation.

5.2. HERRING’S EQUATIONFOR A FLAT FACET

Equations (20) or (26) apply only to the surface of a body at a point where
it is continuously curved. Neither form of this equation can be applied to
surface elements of a given body which are flat, i.e., facets, Herringl) has
given an equation by which the average chemical potential of the vapor in
equilibrium with a flat facet can be determined. In this section it will be
shown that this equation reduces to the Gibbs-Thomson equation if the
given body does in fact have the equilibrium shape.
Consider a body having a flat facet, of area A, and surface normal vector
A
n,, which is bounded by an edge, so that the surface normal changes dis-
continuously on crossing this edge. Let s be distance measured along the
edge (from some arbitrary starting point); further, let ri(s) be the normal to

* The dimensions of the equilibrium body enter equation (26) only through the ratios
pl (&)/(A,+ + LO-), i= 1, 2. These ratios are independent of the size of the equilibrium
body.
 GENERALIZATION OF THE GIBBS-THOMSON EQUATION 441

the element of surface adjacent to the facet at s, so that 19(s)~cos-~(A(s).A,)

is the dihedral angle across the edge at s (see fig. 5). Finally, let yl=y(A,)
and r(s)=y(A(s)) be the specific surface free energies of the facet and of the
orientations bordering the facet, respectively.
Herringl) has shown that, using these definitions, the difference between
the average chemical potential of the vapor in equilibrium with this facet,
R(A,), and the chemical potential of the vapor in equilibrium with a very
large body, &(co), is given by

,E,(PI~)- P,(CO)=: - y,ctne(s>]ds

[Y(S)CSC~(S) (30)
1
where the integral is to be taken around the edge. Q,, is the molecular volume
in the condensed phase. We introduce an equilibrium body, of size (At + A,),

Fig. 5. Geometry of a facet and its adjacent surface elements.

oriented so that its crystal axes are parallel to those of the given body. Then
according to equation (15)
Yo+
+ YO
r(fi>=
( 1A@)
n,’

where IZ(A)=r(A). R, r(A) being the tracing vector of the surface, measured
from an arbitrarily chosen origin, 0. Defining A,=A(A,) and A(s)=A(A(s)),
(30) becomes

=520d +y0
jj,(fi,) - p,(a)
AI kf=i-)+$
csce [A(s) (s) - 1, ctn0 (s)] ds .

We suppose that the given body is, in fact, identical with the equilibrium
(31)

body used to define A(A); the reduction of (31) to equation (19) consists in
showing that
*
A, = + [q+sce(s) - 1, ctne(s)]ds.
P
442 CHARSES A. JOHNSON

Referring to fig. 5, the vector equation

&ii, + T,(s) = IL(s)A(s) + T(s)

obviously holds. Taking the scalar product of this equation with h(s), noting
that ii(s)*T(s)=O, A(s ) .li,=cosB(s) and ii(s). T,(s)=IT,(s)J sin B(s),

i(s) = A1 costi + ITI (s)l sine(s)

or
ITI (s)l = L(s)cs&(s) - 1, ctno(s).
But

A, = 3 ITI (s)/ds
whence

A, = f [A(s)cscB(s) - &ctn0(s)]ds. (32)

Using (32) in (31) we have

the Gibbs-Thomson equation. (The bar, denoting an average value, has been
removed from &fir) since a body which is everywhere in equilibrium with
its vapor must have ,LL”
the same everywhere.)

6. Summary

It has been shown that, for a crystal that is everywhere in equilibrium with
its vapor (and thus is in equilibrium with respect to changes of shape), the
chemical potential of the vapor is related to the size and shape of the crystal
and to its (anisotropic) specific surface free energy in a very simple way
(equations (13) or (19)). This relationship, which may be called the general-
ized Gibbs-Thomson equation, bears a very strong resemblance to the classic
Gibbs-Thomson equation for the vapor pressure of a fluid drop. This result
is, on the one hand, an extension of Volmer’s7) result for crystals which are
polyhedra and, on the other hand, it can also be obtained by applying
Herring’s general equations to an equilibrium shape. The particular ad-
vantage of the generalized Gibbs-Thomson equation is that of simplicity,
not only in form but in application.
As an illustration of the application of the generalized Gibbs-Thomson
equation, let us suppose that we have a small crystal possessing central
symmetry, which is known to be in equilibrium (as far as changes of shape
are concerned) with its environment. (This environment, which has been
 GENERALIZATION OF THE GIBBS-THOMSON EQUATION 443

referred to as a vapor phase throughout this paper, can equally well be a

crystalline matrix in which the substance of the crystal is soluble.) Further,
suppose that the chemical potential of the vapor, relative to that for an
infinitely large body of the same substance at the same temperature has been
determined, either directly by a measurement of vapor pressure or enhanced
solubility (concentration gradient) or indirectly from data on the coarsening
of a distribution of such crystals in a solid matrixs*ra). Then, from equation
(13), the specific surface free energy of any particular surface orientation is
given by
P”(iO>
- AC4
Y(fi)=
( 2520
where~~(2,)and~4co > are the chemical potentials
) n(a)

of the vapor in equi-

(33)

librium with the crystal under consideration and with an infinitely large
crystal, Q, is the molecular volume in the crystal, and ,4(A) is the normal
distance from the center of symmetry of the crystal to the surface element
of orientation A (A(b)) is half the width of the crystal in the direction R;
see figs. 2 and 3).
In situations where equilibrium obtains across every element of the surface,
but not between widely separated elements of the surface (e.g., the approach
to the equilibrium shape of a particle in a solid matrix or the advance of a
curved interface in a diffusion-controlled solid state transformation reaction)
then one or the other of Herring’s general equations must be used. Both of
these equations require detailed knowledge of y(A). It has been shown that,
in Herring’s equation for a smoothly curved surface element, this detailed
knowledge of y(A) can equally well be replaced by a knowledge of the
geometry of the equilibrium shape (equation 26).

Acknowledgements

The comments and suggestions of Professor G. D. Chakerian of the

Mathematics Department of the University of California (Davis) were
essential to the development of much of the subject matter of this paper,
and are most gratefully acknowledged.
Thanks are also due to Drs. J. C. M. Li, N. Louat and R. A. Oriani,
all of this Laboratory, for very helpful discussions.

References

1) C. Herring, The Physics of Powder Metallurgy (McGraw-Hill

Book Company, New
York, 1951) p. 143f.
2) G. N. Lewis and M. Randall, Thermodynamics (2nd edition) (McGraw-Hill Book
Company, New York, 1961) p. 482.
444 CHARLES A. JOHNSON

3) J. W. Gibbs, Collected Works, Vol. I (Longmans, Green and Co., New York, 1928)
p. 321.
4) G. Wulff, Zeit. f. Krist. 34 (1901) 449.
5) A. Dinghas, Zeit. f. Krist. 105 (1944) 304.
6) C. Herring, Structure and Properties of Solid Surfaces (Univ. of Chicago Press,
Chicago, 1952) p. 5.
7) M. Volmer, Kinetik der Phsenbildung (T. Steinkopf Verlag, Leipzig, 1939) p. 87f.
8) B. E. Sundquist, Acta Met. 12 (1964) 67, 585.
9) C. Wagner, Zeit. f. Elektrochemie 65 (1961) 581.
10) G. R. Speich and R. A. Oriani, to be published in Trans. Met. Sot. AIME.
11) W. Blaschke, Kreis und Kugel (2nd edition) (Walter de Gruyter and Co., Berlin, 1956)
p. 115f.