A MICROSCOPIC THEORY FOR ANTIPHASE

BOUNDARY MOTION AND ITS APPLICATION

TO ANTIPHASE DOMAIN COARSENING
SASICEL i%l. ALLEN* and JOHN W. CAHNt
*Department of Materials Science and Engineering. Massachusetts Institute of Technology.
Cambridge. MA 02139. U.S.A.
tCenter for Materials Science. National Bureau of Standards. Washington. DC 20234, U.S.A.

(Received 12 October 1975)

Abstract-A microscopic diffusional theory for the motion of a curved antiphase boundary is presented.
The interfacial velocity is found to be linearly proportional lo the mean curvature of the boundary.
but unlike earlier theories the constant of proportionality does not include the specific surface free
energy. yet the diffusional dissipation of free energy is shown to be equal to the reduction in total
boundary free energy. The theory is incorporated into a model for antiphase domain coarsening. Experi-
mental measurements of domain coarsening kinetics in Fe-AI alloys were made over a temperature
range where the specific surface free energy was varied by more than two orders of magnitude. The
results are consistent with the theory: in particular. the domain coarsening kinetics do not have the
temperature dependence of the specific surface free energy.

R&sum&-Nous presentons une theorie microscopique de la diffusion appliquke au deplacement d’unc

paroi d’antiphase curviligne. Sous trouvons que la vitesse de I’interface est proportionnelle B la
courbure mo!enne du joint. mais. contrairement au cas des theories antkrieures. la constante de
proportionnaht6 n’inclut pas I’energie libre superficielle SpCcifique. bien que la dissipation par diffusion
de I’energie libre soit Cgale 6 la diminution de l’energie libre totale du joint. Nous incorporons cette
theorie dans un modtle de la croissance des domaines antiphases. Nous avons efTectui des mesures
expirimentales de cinetiques de croissance des domaines dans des alliagys Fe-Al. pour une gamme de
tempkratures dans laquelle I’energie libre superficielle specilique varlalt de plus de deux ordres de
grandeur. Les r&ultats sont en bon accord avec la theorie; en particulier. les cinetiques de grossissement
des domaines ne varient pas. en fonction de la temptrature. comme l’tnergie libre superficielle.

Zusammenfassung-Es wird eine mikroskopische, auf Diffusion aufbauende Theorie fir die Bewegung
einer gekriimmten Antiphasengrenze vorgelegt. Man findet, da0 die GrenzflXchengeschwindigkeit linear
mit der mittleren Kriimmung der. Grenze zusammenhtingt. die Proportionalit&konstante jedoch
anders als bei friiheren Theorien die freie OberflIchenenergie nicht enthllt. Es wird aber gezeigt, da13
die durch Diffusion dissipieite freie Energie der Verminderung der gesamten freien Energie der Grenze
entspricht. Die They2 wird in ein Model1 der Vergrijberung van Antiphasendomanen eingearbeitet.
Die Kinetik der DomznenvergrBberung wird an Fe-Al-Legierungen experimentell iiber einen Tem-
peraturbereich gemessen, in dem die spezifische freie OberflIchenenergie iiber mehr als zwei
GrBDenordnungen variiert. Die Ergebnisse stimmen mit der Theorie iiberein: insbesondere weist die
Kinetik der DomlnenvergrGberung die Temperaturabhgngigkeit der spezifischen ObertlIchenenergie
nicht auf.

1. INTRODUCTION in order that the evolution of the microstructure

Internal surfaces in solids-such as grain boundaries, could be understood.
coherent interphase interfaces and antiphase bound- Resistances to interfacial motion in solids are often
aries-are in general non-equilibrium features of real diffusional. Grain growth in pure metals requires re-
crystalline solids. They have a positive excess free arrangement of atoms in a grain boundary region,
energy. Thus, in polycrystalline materials, grain thus requiring diffusion’over distances of the order
boundaries migrate to reduce the total amount of of an atomic spacing. Long-range diffusion can also
grain boundary area. Local atomic arrangements near be required for interfacial motion, such as for the
rapidly moving interfaces can dialer significantly from growth of a precipitate from supersaturated solution.
arrangements near slowly moving ones. giving rise, or in cases where impurities segregate to the interface
for example. to impurity-drag phenomena. A com- and tend to be carried along with the interface as
plete theory of interfacial motion in solids would have it moves.
to account not only for the local structure, excess Theories to explain many of these phenomena have
free energy and geometry of the interface. but also been proposed [l-4]. A widely-used phenomenologi-
for the topological nature of the interfaces as a whole. cal theory of interfacial motion [l. 21 states that inter-

1085
1086 ALLEN AND CAHN: A THEORY FOR ANTIPHASE BOUNDARY MOTION

facial velocity V is proportional to thermodynamic region describes the surface free energy. In this paper
driving force, the proportionality constant being a we will show that the diffusion that results from the
positive quantity called the mobility. The driving gradients in the inhomogeneous region leads to a
force in this theory is the product of the mean of translation of the boundary proportional to its mean
the local principal curvatures (K, e K2) of the curvature but independent of irs surface fire energy.
boundary and its excess free energy per unit area 6. We do not explicitly introduce surface free energy,
resulting in the relationship nor for that matter curvature. into the diffusion equa-
tion, only the gradients in long-range order parameter
V= PG(K, -I- K,), (1) q. Within a curved ‘surface’. the quantity V2q contains
where p is the mobility which in some theories is quantities that originate from APB curvature and
inversely proportional to interface thickness [2]. One lead to interfacial motion. Because the surface free
of the motivations for the present study was to test energy is not a factor in the velocity. we then test
equation (1) experimentally for a problem in which the predictions of this theory against the phenomeno-
the surface free energy dependence could be investi- logical theory in systems which have second-order
gated. When an order-disorder transition is higher- transitions and thus where surface free energy can
order, the antiphase boundary (APB) energy is known be varied by several orders of magnitude.
to tend to zero continuously at the transition tem- The theoretical treatment that follows was pre-
perature [S], thus providing a means to vary the mag- sented earlier [9] in abbreviated form. The experi-
nitude of the surface free energy over a wide range. mental data presented here in support of the theory
Another approach to the theory of interfacial are much more extensive and conclusive than in the
motion. used by Langer and Sekerka [63, is based on earlier paper. We note that a recent paper by Chan
solving a diffusion equation that has been modified [lo] treating a more genera1 problem than that con-
to account for the thermodynamics of non-uniform sidered here obtains theoretical results similar to
systems [7]. The premise is that this modified free those we present.
energy is the basis of a diffusion potential whose
gradient leads to a flux [SJ. Langer and Seketka have 2. DIFFUSE-INTERFACE THEORY
used this diffusion theory to describe motion and de- AND ANTIPHASE 3OUNDARIES
viations from equilib~um in a plane interface between
two phases differing in composition and being forced Since we will treat interface motion in terms of
to move because of imposed fluxes through the bulk a diffusion equation which has been modified to
phases. The mathematical problem was quite cumber- account for gradients in the vicinity of coherent inter-
some in that there were three regions with differing faces, it is useful to recall some basic results of diffuse-
diffusion scales: the two phases and the interfacial interface theory [7].
region itself. Interfacial motion was controlled by We assume the existence of a function fo, the free
long-range diffusion. energy per unit volume of a homogeneous phase which
A much easier problem to study using the diffusion- is a function of the long-range order parameter q.
theory approach is the motion of a curved APB. For systems with second-order transitions, a11 scalar
These are coherent interfaces separating domains with properties includingf, must be even functions of the
identical properties in crystals with long-range order, order parameter, because states with order parameter
the domains differing by a relative displacement q’ and states with order parameter -rj are identical
which is not a superlattice translation. If the differ- (except for a relative displacement). Below the transi-
ences between the atomic species were ignored, the tion temperature,f, has the characteristic double-well
two domains would be part of the same crystal. form indicated in Fig. 1. For such a system the lowest
There are alternative ways of considering APBs and energy state is given by
their properties. It is customary to think of them as
surfaces, having geometrical properties such as area
and curvature, thermodynamic and chemical proper-
ties such as excess free energy per unit area and
adsorption, and kinetic properties such as their vel- Hence for a system with a second-order transition,
ocity in response to a driving force. But APBs also the equilibrium order parameters are of equal magni-
have a thickness which tends to infinity at the critical tude and opposite sign (+c and -q. in Fig. 1). The
point [5]. They are therefore also volumes in which APB is simply the interface between two regions, one
there ark composition and order parameter gradients. with order parameter +rl, and the other -qr
As long as the radii of curvature greatly exceed the Since the interfacial region near an APB comprises
thickness, both descriptions are feasible and it a volume in which the order parameter has values
becomes important to learn whether or not they are intermediate between +q, and -qc surface excess
equivalent. free energy depends on both the details of the func-
Near the critical point where the interface is thick tion f0 and the spatial variation of the order par-
and the gradients are small it has been shown [S, 73 ameter in the interfacial region. It is customary in
that the excess free energy due to this inhomogeneous discussing excess free energy due to interfaces to in-
 ALLEN AND CAHN: A THEORY FOR ANTIPHASE BOUSDARY MOTION 1087

Also, by using equation (7) to change the integration

variable in equation (8).
**‘l.

G= 2(~Afb)* ’ dq. (9)

i
* -‘Ill

For a coherent interface in an unstrained cubic lattice.

c is independent of crystallographic orientation of the
interface [7].
I I I
0
Interfacial thickness is easily estimated if the func-
-% ‘% 77 tion A& is known. An approximate relationship [7]
Fig. 1. The free energy per unit yolumefb is an even func- for the APB thickness l over which Sq’?g differs
tion of long-range order parameter q. Below the critical appreciably from zero is easily developed from equa-
temperature._& has minima at fq,.. && is the increase in tion (7) with the result
f, when q differs from v<.
i = ~~~Cti/~A~~~~~~~~. uot
traduce the function Af, which is simply the free
where (A&.X is the (local) maximum value of AfO
energy difference between a homogeneous state of ar-
which occurs at q = 0.
bitrary order parameter and that with q = +qr (see
For an equilibrium planar APB. the solution to
Fig. 1). For the entire system, the excess free energy
equation (7) for the profile ‘l&g) is invariant to trans-
due to interfaces is given by
lation. In the following section, we develop a kinetic
AF =
JJJ Afd I;: (3)
theory for diffusion in the interfaciai region contain-
ing a curved APB. We investigate the implications
of the kinetic theory when it is assumed that at some
where the function Ajis given by (7)
instant a gently curved moving boundary has the pro-
Af= AfO+ 2++)’ (4) file rip(g).
and R is calied the gradient energy coefficient. It can
also be shown [S, 1l] that the variational derivative 3. MICROSCOPIC THEORY FOR
bAF/61. which is the change in AF when q is varied ANTIPHASE BOUNDARY MOTION
at a point in a system in which 9 is varying spatially,
is given by The kinetic equations of continuous ordering reflect
the fact that the long-range order parameter n is not
SAF = SAf, a conserved quantity. If the free energy is not at a
-&- --.&- - 2KV”& (5)
minimum with respect to a local variation in ‘I, we
For planar interfaces at equilibrium, the variational postulate that there is an immediate change in q given
derivative aAF/sq must vanish everywhere, and AF by
will have its minimum value. That is,

where z is a positive kinetic coefficient. Substituting

equation (5) into (1 i). we obtain for the time-depen-
where g is distance normal to the interface (along the dent problem
gradient of 0). The first integral of equation (6) gives
the expression for an equilibrium planar interface [7]
(12)

(7)
where M = 2ra has dimensions of a diffusion coeffi-
Interfaces exist only if h’ is positive. cient (m”/s). This non-linear equation is similar to the
Equation (7) is readily integrated by separation of time-dependent GinsburgLandau equation in which
variables [7] to give the equilibrium profile of a planar there is no conservation. It has been discussed by
interface when the function Al* is known, subject to Kawasaki [ 123 and Metiu er a!. [13]. Because of the
the boundary conditions 4 = rfr at x = X, and ZAQSq term it is non-linear. Its solution will give
1 = -Q. at x = - X. Since the equation is invariant the evolution of the q field for any initial q distribu-
to translation we choose a particular solution q&g) tion. We now investigate the application of equation
such that n,(O) = 0. By combining equations (3), (4) (12) to the problem of the evolution of an 1 field
and (7) the excess free energy per unit area 6 of a in three dimensions which initially describes a curved
planar interface can be expressed APB.
It can be shown from a theorem by Wiener (14)
(8) that a curved APB obeying equation (12) cannot be
in equilibrium; that is, there are no time-independent
1085 ALLEN AND CAHN: A THEORY FOR ANTIPHASE BOUNDARY MOTION

solutions. For a spherical APB there is a simple proof where K1 and Kz are the principat curvatures of the
which is a coroIIary to one given previously [l [Link] iso-q surfaces.*
wish to demonstrate that there are no sphericalfy Equation (16) is a general equation governing the
symmetrical solutions to time-dependent readjustment of q near a curved APB.
We now investigate the motion of a gently curved
APB which at some particular time has the profile
q,(g) at all normal sections. For this case. equation
with the boundary conditions that at g = 0. (7) is valid, and we obtain for the kinetic equation
q = q,. f 0 and dq,/dg = 0: and at g = x, q = -q,
and dq/dg = 0. if we multiply this equation by dddg
and integrate term by term we obtain respectively for
each term
The velocity (&/Zc) of a constant q surface in the
boundary region is given by

In the limit in which the principal radii of curvature

and are much greater than the interfacial thickness. the
curvatures of the various iso-q surfaces can be con-
sidered to be independent of the vafue of the coor-
dinate g. Therefore a11 surfaces of constant q at a
‘point’ in the interface will move with the same vel-
Because the last integral is always positive, there is ocity K given by equations (17) and (1St as
no solution to equation (13) with these boundary con-
v= &l(KI + K,) (19)
ditions. The only spherically symmetric time indepen-
dent solution of equation (12) is p = constant, which and the assumed interfacial profile q&g) wilt be pre-
does not satisfy the boundary conditions below the served in the moving interface.
critical temperature. It is interesting to note at this point that there is
We now consider the time dependence of the order a significant difference between equation (,19) and the
parameter within a curved APB as given by equation early phenomenological theory which states that the
(12). We assume that the principal radii of curvature velocity is proportional to surface free energy multi-
of the APB are iarge compared to the equi~brium plied by mean curvature, equation (1). In the present
thickness and that initially the normal profile of the development, the surface free energy does not appear
APB is the same everywhere along the APB. Equation in the relationship.
(I 2) can be rewritten by expanding the term V2g. Thus
if t is a unit vector normal to the surfaces of constant 4. MACROSCOPIC THEORY OF
q, the vector Vq is given by DOMAIN GROWTH

Because instantaneous velocities and ~stantaneous

(14)
principal curvatures of a point on the interface are
difficult to measure, a direct experimental test of
where L”q/agis the rate of change of 4 in the direction
equation (19) is difficult to devise and execute. WC
oft. Then we have (see Appendix)
therefore choose to incorporate our result into a
theory of antiphase domain coarsening. in which
averages of curvature and velocity are r&ted.
Starting with the linear relation between local
Since the divergence of a unit normal vector to velocity and mean curvature. equation (19), and the
a surface is equal to the negative of the mean curva- geometrical relation between the change in area 6(dS)
ture (K, + K,)[16], we have for the kinetic equation when an element of area dS of a curved surface moves
a distance V&t in time 6t

we obtain after rearrangement and substitution of

l In the present treatment we adopt the convention that equation (19)
the mean curvature is positive when the surface is concave
on the side toward which g is directed. Thus a spherical dS
- = -M (K, + K2)2d.S. (21,
domain has its surface normal directed inward. An dr f
opposite convention was used in reference [Pf. with the
result that equations (IO) and (11) of that paper have the in which the time dependence of the experimentall?
opposite sign to the analogous equations in this paper. measurable total surface area S appears instead of
 ALLEN A?iD CAHN: A THEORY FOR ANTIPHASE [Link] XIOTiON 1089

the local velocities. We define the averaged squars with it and in so doing som2 elemrnts of volumz will
mean curvature KX lose and some will gain free energy.
. For the caSe of ‘almost planar’ APBs, whose radii
K,f,= ((K, 4 K,)‘) = f (K, + K?)‘dS. (22) of curvature greatly exceed the interfacial thickness.
i
equation (17) is valid. Making use of equations (5).
The surface area in a unit volume of specimen. S,. (15) and (17). equation (26) gives for the dissipation
then obeys the relation
-=-
?F
.*
dS,
-=
dt
-. CfK$S,. (13) ?t
2~
i>
C’I-+‘(KL+ K,)dz,.
zg
(27)

The volume of intsgration in equation (27) can be

We now consider superlattices having only two
separated into two parts: one for the antiphase
antiphase domains. The B2 (CsCl) structure is an
domain interiors. and on2 for volumes which com-
example. If dislocations in the crystal are ignored.
pletely contain the ~nhomogeneous regions where
each domain is multipiy connected [17]. As the
there are interfaces. The first part does not contribute
domain structure coarszns, interconnections are
to the dissipation. as Eq/c’g is zero everywhere within
broken. and both Ki and S, decrease. If we assume
the domain interiors. An elsment of volume in th2
that during coarsening the statistical measures of the
second part can be expressed as
structure change only as a result of the length scale
changes, then both .Sz and Ki will be proportional dc = [Link] (28)
to the reciprocal of the square of this length scale.
where dS is an-element of surface area and g, as
Hence
before, is the distance along the gradient (see Fig. 2).
I(: = 4s;. (23) Since the interface is thin relative to its radii of curva-
tures dS and K, and K, are independent of g. and
where statistical similarity of the structure during
the volume integral can be expressed
coarsening implies that Q is a constant. Substituting
equation (23) into (22) and integrating. we obtain
[S,(t)] -2 - [s,(o)]- L = 2#Mt, (24)
where S,(O) is the initial value of S, and S,(t) is the
valuz at time t. where the integration limit 6 is su~ciently large to
extend into regions on both sides of the interface
where &7/Q vanishes. From equation (8). the first in-
5. DISSIPATION IN THE
MICROSCOPIC THEORY tegral in equation (29) is seen to be equal to u, the
surface free energy. Comparing the second integrand
As has been noted in the introduction, it is custom- with equation (21). we have
ary to view surface 2x22s~ properties (like free energy)
on a per unit area basis, thus treating the interface
as a geometrical surface. In such a picture, inter-facial J7 V((K, + KJdS = - ‘$ (30)

motion lowers the free energy of the system by de- and hence equation (29) becomes
creasing the surface area of the interfaces. Thus the
dissipation, or rate of free energy decrease for the sys- SF dS
-_=fl-_. Q.E.D.
tem due to interfacial motion is given by St dt

SF dS The microscopic theory. in which interfacial motion

-;_ (25) as derived from a diffusion equation as not being pro-
CC =%.
portional to D gives the correct expression for dissipa-
In our treatment we did not implicitly define the tion of surface free energy.
surface tension 2 and the only irreversible process
was diffusional. Since the calculated velocity does not 6. SELECTION OF EXPERLMENTAL
depend on u it becomes important to check whether SYSTEM
the dissipation due to the diffusion that leads to inter-
face motion is equal to that given by equation (25). Equation (19) is thr central result of the theory for
The dissipation is given by a volume integral APB motion which we wish to test experimentally.
We have chosen to test the theory by studying the
ZF coarsening of antiphase domain structures in systems
-=
SK that have second-order transitions. Two critical
It is interesting to note that the integrand SAF/S~*ZI~,/Z~ aspects of the theory can be tested: first. that equation
which is I?AJ/& in a unit volume is not the dissipation (24) is obeyed for domain coarsening at a fixed tem-
because there can be a flux of free energy. A moving perature; and second. that over a range of tempera-
APB. for instance, will ‘carry’ excess free energy along tures where CTcan be varied appreciably, the quantity
1090 ALLEN AND CAHN: A TI-IEORY FOR ANTIPHASE BOUNDARY MOTION

Atomic arrangements and possible APB displace-

ment vectors in the B3 and DOS structures are shown
schematically in Fig. 4. In the 82 structure. there is
only one kind of APB. having a ja,(lll) displace-
ment vector. in the DOa structure two types of
displacement vectors produce APBr &(l Ii} and
jd,( 100). The two types in the DO3 structure can
be distinguished by dark-field electron microscopy
[23]. We chose to perform an electron microscope
study of domain coarsening in an Fe-23YJl alloy
with FeAl order. and in an Fe-247641 alloy we stud-
ied domain growth in both the FcAl and Fe,Al
Fig. 2. An element of ‘surface volume in a sheet of finite
phases. In the Fe,Al phase, we studied growth of the
thickness containing a diffuse interface. The volume ele-
ment dc is equal to dg-dS if the thickness of the inhomo- jab{ 100) antiphase domains only. This was accom-
geneous region is small relative to the radii of curvature. plished by producing very large j&l If} domains
in an FeAl ordering treatment prior to heat treatment
in the Fe,Al region of the phase diagram. A fine-scale
+M in equation (24) does not have the temperature Fe,Al domain structure consisting of only two
dependence of 6. The surface free energy can be domains, distinguished by a jab{ 100) displacement,
varied appreciably in any system in which the order- forms within each FeAl domain. This fmer domain
disorder transition is second order. Examples are structure is topologically very similar to that in the
Cu-Zn. Fe-Al and Fe-Si. FeAl phase.
A portion of the Fe-Al coherent phase dia- An extensive domain coarsening study in an Fe-Co
gram[t8. 19,20,21] is reproduced in Fig. 3. The alloy with smaI1 vanadium additions has already been
phases z:, FeAl, and Fe,Al possess the AZ(b.c.c.), 32 made by English [17]. Domain coarsening in an
and DO, structurest, respectively. The FeAl phase is Fe-49%Co-2Y;V alloy with B2 order was studied by
stable far from stoichiometry; the excess iron atoms X-ray diffraction. We have chosen to re-examine these
are accommodated substitutionally on the aluminum data in the light of our new theory. although there
sublattice. The x I* FeAl phase transition is second- are no data close to the disordering transition.
order along a tine in the phase diagram which ter-
minates at a tricritical point [22] at about 23% AIS
7. EXPER~~~ENTA~ TECHSIQUES
and 890K; below this temperature the phase change
is first-order. The FeAl--r FesAl transition is second- The Fe-Al alloys used in this study were the same
order until it terminates at a three-phase invariant. ones used in an earlier study [24]. Their nominal
compositions are Fe-23.O%Al and Fe-24.O%Al. The
following heat treatment sequence was used: homo-
genization treatment at 1050K or above in an inert
atmosphere, followed by a brine quench; and a single
ordering and antiphase domain growth ~eatment~ at
a temperature below the critical temperature for
ordering, followed by a brine quench. For the homo-
genization treatment the material was in rod form,
having 3 mm dia. For the ordering and domain
growth treatment, discs 3 mm in dia. and 0.2 mm
thick were wrapped in stainless steel foil and heat
treated in molten sait or lead. The temperature of
the molten bath was typically maintained constant
to within 1 K during heat treatments. Electron micro-
:
scope samples were prepared by electropolisbing the
700 I I I 3 mm discs using techniques described earlier [24].
20 22 24 26 28 30
Atom percent alumrum Heat treatment times used for the single ordering
and domain growth heat treatments were very much
Fig. 3. The coherent phase diagram for the Fe-Al system
greater than the time necessary for complete ordering:
c20.2 13.
in this way the term S;‘(O) in equation (24) was made
negligibly small. The maximum allowable heat treat-
t Strukturbericht notation. ment times are determined by the dislocation density
1 All compositions reported in this paper are atomic per- in the sample, because APBs terminate at dislocations
cent.
SAs noted earlier, FesAI specimens received an inter-
having a Burger’s vector which is not a translation
mediate FeAl ordering treatment which was carried out vector of the superlattice. Domain growth stopped
at about 900 K for at least 1 h. when domain size became of the order of the disloca-
 ALLEN AND CAHN: A THEORY FOR MJTIPHASE BOUNDARY MOTION 1091

FeAl Fe,AI

Fig. 4. Atomic arrangements and possible superlattice displacement vectors for the ordered phases
FeAl and Fe,AI. Both phases are derivatives of b.c.c.

tion spacing. Thus for measurements of the kinetics 8. EXPERIMENTAL RESULTS AKD
of antiphase domain coarsening, we endeavored to DISCUSSION
have the mean linear intercept for APBs be consider-
ably smaller than the mean distance between disloca- 8.1 Time dependence
tions, or other factors that interfere with the statistical
simiiarity assumed in equation (23). The Fe,Al The time dependence of antiphase domain coarsen-
domain coarsening heat treatments were terminated ing at constant temperature has been extensively in-
before the Fe,AI domain size reached the prior FeAl vestigated by others [27-331. We expect a linear rela-
domain size. tionship between time and the square of the domain
Measurements of APB surface area per unit volume size during the period where equations (1) or (19).
were made on transmission electron micrographs and equation (23) are valid. Implicit in (1) and (19)
using standard quantitative metallographic pro- is an absence of anisotropy, which obviously elimin-
cedures C25.261. For each data point a minimum of ates domain coarsening studies in LIZ structures such
38 intersections of APBs with the test circle were as Cu,Au [30,31-J and Ni,Fe[32] from consider-
counted, making the sampling error for .S, less than ation. Equation (23) eliminates the initial time period
10”;. Because micrographs were taken from areas of during which asymptotic domain shape implied by
finite specimen thickness, a first-order correction was equation (23) is being established. It also eliminates
applied in cases where a test line was tangent to (thus the long time period where the domain size is
overlapping the projected image width) an APB in approaching the grain or specimen size or the disloca-
a micrograph. In such a case one intersection of the tion spacing, whichever is smallest.
test line with the APB was counted. Since this paper is concerned with the distinction
An analysis of English’s data for domain growth between equations (1) and (19), the time dependence
in the Fe-Co-2V alloy requires that a value for S;*(O) is not a central issue. Nevertheless it is worthwhile
in equation (24) be taken into account because the to point out that the expected time dependence was
m~~imum domain sizes achieved in these experiments found both by English [I73 as shown by his Fig. 2.
were so small. The following method was used: Eng- and by us in a singie check (Table I, Fe-24D/4
lish reported domain sizes achieved after different 898 K). The time dependence implies that .Sft is a
heat treatment times for several temperatures, and constant in time, but not in temperature. It is equal
also the time for the samples to reach 90% of the to 2$M if equation (19) is valid, or 24~0 if equation
equilibrium value of the order parameter at each tem- (1) is valid. These should differ greatly in their tem-
perature. Heat treatment times which were longer perature dependence near the critical point where G
than this were considered to produce ‘fully’ ordered tends to zero. We note that several workers have
specimens, and these data were used for the present studied isothermal domain growth in the Fe,AI phase
domain growth analysis. The longest heat treatments experimentally at temperatures below 694 K.
at 898, 838, 823 and 798 K were discarded because [27,28,29] where the ordering times are rather long.
it appeared that there might have been systematic Careful scrutiny of the accompanying data concerning
errors in the domain size measurements when the the kinetics of the ordering reaction [27,29] show
domain sizes were large. A linear regession analysis that most of the reported data on domain size are
of D’ vs t (where D is the domain size and t is the for heat treatment times shorter than or approxi-
heat treatment time) for the remaining data at each mately equal to that required for complete ordering.
temperature was made. From the slope, a ‘best’ value For this reason, we do not make use of these results
of (.Szr)-’ at each temperature was computed. in testing our theory.
1092 ;\LLEN ASiD C;\HN: A THEORY FOR ANTIPHASE BOUNDARY MOTION

Table I. Antiphase domain growth in Fe-AI and Fe- The principal temperature dependence of M or p
G-X alloys should mirror that of diffusion within the APB, which
is a region of varying degrees of order. Even diffusion
T(K) r (s) 1S:t: - t (m’/s)
in homogeneously ordered alloys has an apparently
Fe-217,;A I, FeA 1order variable activation energy due to temperature de-
990 300 ;:; ; :;I” pendences of thermodynamic factors and correlation
987 300 factors [35.36-J. In addition it is known that the factor
983 300 I.5 x to-:: K in ICI should have a weak divergence close to
963 300 1.1 x 1o-*s
300 T proportional to [(T, - 7’+)/7J-“.o’ where the
953 1.1 x 10-1s
931 600 5.6 x IO-l6 exponent is equal to the product of the exponents
904 t200 2.2 x to-‘6 for the pair correlation function and the correlation
898 36X@ 1.1 x to-‘6 length [37]. Current best theoretical estimates from
898 360 1.t x to-‘6 a three-dimensional fsing model are 0.031 and 0.63,
871 3600 3.9 x to-”
respectively [38].
Fe-23xAI. FeAi order In addition. if p is inversely proportional to the
908 1200 disordered thickness of the APB as has been suggested for some
904 1100 1.3 x lo-l6 kinds of surfaces [2], an additional factor of [(7;. -
895 1440 6.9 x 10-l’ T)/T]“.63 would appear in equation (1) raising the
889 LZGO 5.4 x lo-” temperature exponent to almost 2.
884 1100 1.6 x LO-”
We now turn to the individual systems for which
883 2 1 6ooO 3.6 x to-‘*
we have listed domain coarsening data in Table 1.
Fe-24”i’AL
0 Fe&l order Domain gro~r~ in Fe-247&I with FeAl order.
768.5 [Link] 1.0 x lo-l9 These data span the temperature interval between 990
758 86,400 7.1 x to-?* and 871 K. The domain structure produced at 990 K
748 86.400 6.4 x IO-” is indicative of ordering during the quench; thus the
720 1.64 x lo6 5.8 x 1O-‘o
critical temperature for our 24% alloy is bracketed
Fe-26.5, Fe,At order [40] between 987 and 990 K. This alloy is two-phase below
823 1.01 X 10s 2 x to-‘8 about 860K (see Fig. 3). The semi-logarithmic plot
of the data in Fig. 5 shows that to an excellent
Fe-Co-ZV, FeAl order I:i7f
898 60 - 240 6.9 x to-”
approximation (Sit)-* is an Arrhenius function of
853 120 - 950 I.0 x lo-l9 temperature.
838 240 - 3200 43 x lo-= A regression analysis of these data was performed
823 880 - 5500 1:2 x t0-?O to determine 90% confidence intervals for the appar-
798 910 - 14,800 3.6 x to-”

S.2 Temperature dependence

Our data for domain coarsening in Fe-Al alloys
is listed in Table 1. In Fig. 5 we show Stt vs tempera-
ture for these data on an Arrhenius pIot. On such
a plot the deviations from a straight line demonstrate
the presence of temperature dependent factors or
terms other than a simple exp( - Q/RT).
It is important to establish whether the temperature
dependence of the data in Fig. 5 are consistent with
equations (1) or (19). The behavior of Q near the criti-
cal temperature is known to vary as [(T, - T)/T,I’:
where y is approximatefy 1.3 [34]. The other domi-
nant temperatue dependence in M or 1-1will be due
to the activation energy of atomic motion and will
be of the form exp( -Q/RT). Functions of the form
[(T, - T)/T,l’exp(-Q/R?“). when plotted on Fig. 5, .
Fe-Co- 2~. 82
will at low temperatures asymptotically approach \
Ref. 17 A
the slope of -Q/R but with 7 > 0 will reach a maxi-
10-f’; ’ t ’ 1 I 1 I I
mum and change siope at approximately I.10 1.20 t.30 1.40 I.50
(T, - Tl/T, = $TC/Q. Near T, the slope will tend to l/T~l03,K-~
X. If the exponent y were negative there would be
Fig. 5. Graph of experimental data for antiphase domain
no maximum and the slope would tend to -co, with growth in Fe-Al and Fe-Co-2V alloys. S, is the APB sur-
the marked deviation from the Arrhenius plot begin- face area per unit volume, t is the ordering time and T
ning at (T, - T)jl; = -yRTJQ. is the ordering temperature.
 .-!LLE?i 4ZD f.4HN: A THEQRY FOR .[Link] BOUSD.4RY MOTIO;u i!-Nj

ent sctiv-ation energy Q for the process. The resulting growth in the IZ”i; A! ahoy compares vvith that in
interval was the 24’0 alloy at the same temperature. The 22, alloy
is 80 K further from its critical temperature. It is more
[Link] mol I Q I 270,OGOf mol.
highly ordered and its value of D is at least an order
Ii :$I: -r was proportional to G multiplied by an of magnitude greater. In Fig 5 it is seen that in the
Arrhenius function having an activation energy of temperature range between 895 and 9gJK. domain
[Link] J!mol. then the plot of the data in Fig. 5 growth occurs nearly as rapidly in the 23°0.M alloy
would go through a maximum at about 45 K below as in the Z?q;\l alloy. Electron micrographs of the
the critical temperature. Clearly, then. the data do domain structures in the Fe-23”.,;.41 alloy reveal
not display the temperature dependence of o. quenched-in fluctuations in degree of long-range
Domain growrh in Fe-23U$l with FeAl ordrr. order that are absent in Fe-21?iAI alloys heat treated
These data span the temperature range from 908 to at similar temperatures. as shown in Fig. 6. In addi-
884 K. The critical temperature for this alloy is found tion. the APBs themselves are very ragged in the
experimentally to lie between 904 and 908 K. Below former alloy, and smoothly curved in the latter. It
about 890 K, electron micrographs show the micro- is interesting that even with these different structures
structure to be two-phase. The APBs in these tv.o- and 0-s domain growth in these two alloys occurs
phase specimens are coated with a distinct layer of with comparable speeds. as would be expected from
the disordered phase which is of the order of IOnm equation (19).
in thickness. Domain growth in Fe-23931 specimens which are
Antiphase domain growth in Fe.4 ordered two-phase 2 + FeAl is slowed considerably relative
Fe-23”&1 is of interest because this composition is to single-phase FeAl specimens. Since the z phase
close to the tricritical point in the phase diagram. coats the APBs and differs in composition from the
where the line of higher-order transitions for the ordered phase, this is not surprising. The apparent
x -+ FeAl reaction intersects rhe two-phase 3 + Fe.4 activation energy for domain coarsening in the two-
region. At the tricritical point, ?‘fWi&? vanishes in phase specimens. as indicated by the slope of the
the ordered phase [?I]. where c is the composition. curve for the 13”0Al alloy in Fig. 5. is much greater
Therefore. fluctuations in composition in FeAI- than for the single-phase Fe-X”/; allov.
ordered specimens will be large in the vicinity of the Donmin growrh in Fe-24P;AI ,vitil Fe,AI order.
tricritical point. Adsorption of excess iron atoms on These data span the temperature range between 768
APBs is also likely to occur near the tricritical point and 720 K. Only at the lowest temperature is ths
in the FeAl region of the phase diagram [39]. Since specimen clearly single-phase Fe,Al. At the other
our theory only accounts for variations in long-range temperatures studied. contrast in dark-field electron
order parameter. it is of interest to see how domain micrographs indicates that there is a layer of disor-

Fig. 6. Dark-field electron micrographs showing APB structures in (a) Fe-23”,Al and (b) Fs-2l”,,.Al
after identical ordering treatments at 904K for 1200s. The 2300 alloy is within 4K of its critical
temperature and shows quenched-in critical fluctuations.
dered phase at APBs At 755 and 768 K this layer As the beginning point for our theoreti&& treatment
is several nm thick. The data point at 770 K lies of the motion of a curved APB. we have chosen a
roughly along the extrapolation of the curve for FeAI- different phenomenological expression-a linear r&a-
ordered Fe-24431 alloys in Fig. 5. The value of tion between the rate or order change and the driving
:S:r: - ’ achieved in the two-phase x + Fe,Al speck- force for ordering. equation (11j. Our resuiting equa-
mens arrears to be relatively insensitive to the heat tion for interfasial velocity. equation 119). is not a
treatment temperature. relationship of velocity to driving force. since driving
Domain growth near T, in Fe-26.5?& with Fe,Al force must include (r. Nonetheless, it is gratifying to
or&. In a unique experiment, Swann et at. [JO] stud- note that the diffusional dissipation of free energy of
ied a specimen held for 28 h in a temperature gradient an APB obeying equation (19) is exactly equal to the
which included T, for their 26.5!& Al alloy. From an available driving force.
estimate of the domain size in their pub~shed micro- This paper concludes that equation (I) is not uni-
graphs. the data point in Table 1 was determined. versally valid. There are, however. clearly cases where
Their micrographs cover a temperature range of 7 K, it is valid. In the case of a curved soap film. the pres-
over which tr should k varied by more than a factor sure difference in the gases leads to gas transpiration
of twenty. Even to within a fraction of I K of the rates proportional to the driving force and, inciden-
critical temperature, there was no noticeable trend in tally, like Turnbull’s theory [2] inversely proportional
domain size with temperature. to the thickness. The soap film theory, like the theor?
Domain growth in Fe-Co-2V wirh FeAl order, The of Li for the motion of a cylindrical tilt boundary [41]
temperature dependence of English’s data for Fe- and the theory of the present paper, is derived from
Co-2V is evident from Fig. 5. The data display the more basic phenomenological theories which in turn
Arrhenius behavior found in the Fe-2-1:; Al alloy can each be tested in independent experiments. In Li‘s
with FeAl order. The reported activation energy [f7] case, he is invoking the force-velocity relation on the
in Fe-Co-2V, 290,ooOJ/mol, is somewhat higher than individual dislocations; in our case we invoke ordsr-
that for the Fe-Al afloy. The curve for Fe-Co-2V ing kinetics in a region that has not achieved equihb-
lies approximately two orders of magnitude below the rium order. At present we seem to have theories that
curve for Fe-24”/,AI. are linearly and non-linearly dependent in curvature,
On the basis of these individual experimental that are independent. linearly dependent and non-
results we conclude that APB velocity is not governed linearly dependent on o, and that are independent
by equation (1). Whether it is quantitativ~iy governed or linearly dependent on the reciprocal of the thick-
by equation (19) must await an experimentat measure- ness. If there is a universal law relating velocity to
ment of the geometrical factor (b and the kinetic curvature, it does not seem apparent at this time.
factor x which, as defined in equation (1 I), is a
phenomenological parameter which can be measured Acknowkt~ciyrmcnrs-We are grateful to the National
in diffusion or ordering experiments that are com- Science Foundation for support of the portion of the
pletely independent of any domain wall problems. research performed at MIT. Robert M. Allen of MIT per-
formed a statistical analysis of the data. Helpful discu&ous
with Professor James Litstrr of the Ph@cs Department
9. CRITIQUE at MIT concerning the critical behavior of the gradient
energy coefficient are gratefully acknowledged.
As our new theory now stands, its applicability to
antiphase domain coarsening in real systems is
REFERENCES
limited primarily by three constraints. First, the new
theory is strictly valid only for phases in which the 1. R. Smoluchowski, Phys. Rec. 83, 69 (f95t).
2. D. Turnbufi, Tinns. AfME 191, 661 fi95if.
orderdisorder reaction is higher than first-order. We 3. B. Sundquist. Met&. Trans. 4. 1919 fI??3).
note that Chan [lo] considered interfacia1 motion in 4. J. W. Cahn, Acra meroll. IO, 739 (1962).
systems which did not have symmetric (about r~= 0) 5. R. Kikuchi, J. Phys. Chem. Soiids 27. 1305 (19661
free energy functions, but the application of Chart’s 6. J. S. Langer and R. Sekerka. Acre metall. 23. 125
results to APB motion in such systems has [Link] (1975).
7. J. W. Cahn and 3. Hilliard. f. cheer Phys. 28, 258
made. Second, the new theory makes use of diffuse- (1958).
interface theory in a way that makes the results valid 8. J. W. Cahn, Acrn rnletolt. 9, 795 (19611
only for systems in which the APB energy is isotropic. 9. J. W. Cahn and S. M. Allen. J. de Physique, Colbqur
Third, when more than one kind of APB is possible, c-7, 54 (1977).
as in Cu,Au where both ‘conservative’ and non-con- 10. S.-K. Chan. J. them. Php. 67. 5755 (1977).
11. J. W. Cahn and J. E. Hilliard. Acre ttrerali. 19. 151
servative types can exist, equation (23) implies that (1971).
the proportion of each type be constant during It K. Kawasaki, Pray. rhror. Phys_ 57.4iO (1977).
domain coarsening. In Au,Cu there is evidence that t3. H. Metiu. H. Kitahari and 1. Ross. J. chern Pfr_vs.6%.
this latter constraint is not true [41]. 292 (1976).
14. R. Courant and D. Hilbert. .tfrthod~ r$ .Ifuthemmkzl
The phenomenological statement in equation (1) Physics. Vol. 2, p. 305. Interscience. sew York (1361t
linearly relates interfacial velocity to driving force. 15. J. W. Cahn and J. Hiiliard. J. them. Php. 31. &?s
Non-linear relations have also been proposed [3,42]. (1959).
 ALLEN AND CAHN: A THEORY FOR ANTIPHASE BOUNDARY MOTION 1095

16. C. Weatherbum. Difirenriaf Geometry o/ Three Dimen- 42. J. C. M. Li. Trans. A.l.M.E. US. 1591 (1969).
sions. p. 225. Cambridge Univ. Press. Cambridge 43. G. Arfken. Marhemarica/ Merhods for Physicists. pp.
(1927). 65-69. Academic Press. New York (1968).
17. A. English. Trans. AIME 236. 14 (1966).
18. P. R. Swami. W. R. Duff and R. M. Fisher. Trans.
.[Link] 245. 851 (1969).
19. H. Okamoto and P. A. Beck. Mernfi. Trans. 2. 569
(1971). APPENDIX
20. P. R. Swann. W. R. Duff and R. M. Fisher. Merall.
Trans. 3, 409 (I 972). Dicergence of a Gradient Vrcror in
21. S. M. Allen and J. W. Cahn. Acra merall. 23, 1017 Orrhoyonal Currilinear Coordinates
(1975).
22. R. B. Criffiths. J. them. Phys. 60, 195 (1974).
23. M. J. Marcinkowski and N. Brown. J. appl. Phys. 33, A general expression for the divergence of a gradient
537 (1962). vector to a scalar field t~(q,, qz. qJ) in orthogonal curvi-
24. S. M. Allen and J. W. Cahn. rtcta meroll. 24. 425 (1976). linear coordinates is given [43]. Consider the t7 field to
25. C. S. Smith and L. Guttman. Trans. A.I.M.E. 197. 81 be comprised of constant r~ surfaces. An orthogonal curvi-
(1953). linear coordinate system may be defined such that one
26. J. E. Hilliard. Applications of quantitative metallogra- coordinate 4, is everywhere normal to the iso7 surfaces.
phy in recrystallization studies, in Recrysralficarion. ,At any point the other two axes q2 and qJ are necessarily
Grain Growth and Textures. p. 267. Metals Park, Ameri- tangent to an iso-rl surface. Then we have
can Society for Metals (1965).
27. K. Oki, M. Hasaka and T. .Eguchi, Trans. Jap. Inst.
v’q = (A-1)
Metals 14. 8 (1973).
28. L. Rimlinger, C.r. Acad. Sci. Paris 272(C), 22 (1971).
29. H. Sagane. A. Yamamura. K. Oki and T. Eguchi, J.
Jap. Inst. %fetals 39. 1076 (1975). where (h,, h2, h,) is the metric which relates differential
30. M. Sakai and D. E. Mikkola. Metall. Trans. 2, 1635 distances in the coordinate system to differential incre-
(1971). ments of the coordinate axes. In particular, if the distance
31. C. L. Rase and D. E. Mikkola, lbferall. Trans. 6A. 2267 along the axis 4, is measured by a parameter 9. then
(1975).
32. F. Bley and M. Fayard. Acra metall. 24. 575 (1976). dy = h, dq, (A-2)
33. S. M. L. Sastry and H. Lipsitt. Mefall. Trans. 8(A). and equation (A-l) can be expressed
1543 (1977).
34. H. E. Stanley. Inrroduction to Phase Transitions and
Critical Phenomena, p. 47. Oxford University Press, (A-3)
Oxford (1971).
35. A. Kuper. D. Lazarus. J. R. Manning and C. T. Tomi- Expanding equation (A-3) and making use of (A-2) gives
zuka. Phys. Rev. 104. 1536 (1956). the result
36. J. E. Hilliard. in appendix to paper by J. E. Reynolds,
B. L. Averbach and M. Cohen, Acta metall. 5, 29
(1957).
37. J. Litster. private communication.
In the same notation the divergence of a unit normal g to
38. J. C. LeGuillou and J. Zinn-Justin, Phys. Rec. Lett.
39, 95 (1977). an iso-q surface is given by [43]

1
39. B. Widom. Phys. Rev. Lett. 34, 999 (1975). 1
40. P. R. Swarm, W. R. Duff and R. M. Fisher, Phys. Status v.g = - $h,h,) . (A-5)
Solidi 37, 577 (1970). hi&h, [ 24,
41. P. M. Bronsveld and S. Radelaar. J. phys. Sot. Japan
38. 1336 (1975). Equation (15) thus follows from equations (A-4) and (A-5).