Professional Documents
Culture Documents
JOHN W. CAHNt
Rate laws are derived for transformations in which the new phase nucleates on grain boundary surface,
grain edges, or grain corners, and in which the new phase grows with constant radial velocity. Time-
dependent and time-independent nucleation rates are treated. At high nucleation rates (relative to
growth rates and grain sizes) the rate law assumes a form independent of the nucleation rate, and the
half-time of the transformation becomes approximately 0.1 D/G, where D is the grain diameter and (2 the
radial growth velocity, independently of which sites are responsible for nucleation. St low nucleation
rates the rate law approaches that given by randomly distributed nucleation sites. The conditions which
determine which type of site will produce the most nuclei, and hence determine the form of the rate law,
are discussed.
It has recently been suggested by Clemm and (junctions of 4 grains), and will be progressively greater
Fisher(l) that the nucleation energy at a junction of as, one considers edges (junctions of 3 grains), grain
grain boundaries could be less than at a plane boundary surfaces, and the interior of grains. It is
boundary, and that these junctions might be favored the purpose of this paper to explore the suggestion
sites for nucleation. However, there are two factors of Clemm and Fisher, by deriving the rate laws associ-
opposing this purely energetic consideration. One ated with reactions nucleating at such sites, developing
factor results from the ease with which these sites can t.he limits under which a particular site is favored, and
transform, and might lead to a situation where these examining some of the characteristics of such a
sites lie wholly in transformed material at a time early reaction.
in the reaction and, hence, be no longer available to In a polycrystalline material, nucleation is assumed
nucleation. The other factor, affecting the nucleation to occur with characteristic frequencies on four t,ypes
rate, is the number of atoms associated with each type of sites. We shall designate the homogeneous nuclea-
of nucleation site. This will be least for corners tion frequency per unit volume as I,, the specific
grain boundary nucleation frequency per unit area as
*Received November 18, 1955. I,, the specific edge nucleation frequency per unit
eneral Electric Research Laboratory, Schenectady,
&w%ork. length as I,, the specific corner nucleation frequency
ACTA METALLURGICA, VOL. 4, SEPTEMBER 1956 449
1
450 ACTA METALLURGICA, VOL. 4, 1956
per corner as 1,. We should also like t*o define nuclea- for considerable distance from a growing particle,
tion frequencies per unit volume for each type of Johnson and Mehlc2) derived the rate law for grain
site, r;,, lir,, &,, SC. If S, L, C are respectively the boundary nucleation by assuming that the trans-
boundary area, edge length, grain corner number formation product could not grow across grain
(all per unit volume), then SV = I,; 8, = I$; boundaries. It will be necessary t.o examine this
8, = I,L; NC = ICC. The total rate of nucleation assumpt,ion in detail. At high nucleation rates many
per unit volume AT= Szl + gir, + Se -+- gC. nuclei per grain will form and the reaction will be
In order to compare expressions complaining 8, L, substantially eomplete when Gt is of t.he order of the
and C, it will be necessary to express them approxi- grain diameter. At low nucleation rates the reaction is
mately in terms of CL
grain diameter. If we assume that far from complete at that time. If one assumes that
all grains are equally large tetrakaidecahedra arranged the transformation product cannot grow across
so that they fill space [a body-centered-cubic array boundaries, then the fraction transformed will be
orient,ed so that the square faces are in (100) planes, equal to the fraction of the grains in which a nucleus
and hexagonal faces in (111) planes] and call the has formed, with a correction for that fraction of the
distance between
square faces D, then the edge untransfor~ned volume which is in grains which are in
length will be D/242. The number of grains per unit the process of transforming. This correctSion becomes
volume are 2/O, and a constant fraction of the untransformed volume at
5 = 3.350-l, time greater than the time it takes a nucleus starting
at time 0 to consume its grain, that is, at t > D/G.
I; = 8.5D-2, (1)
(This is true whether the nucleation rate is high or low,
C = 12.OD-3.
except that when it is high, at t = DIG the reaction is
While the act,ual numerical constants in equations close to completion.) Indeed, Johnson and Mehls
(1) are to be taken as approxiinate, the functional analysis at t > D/G reduces t,o
dependence on D is assumed to be exact,
1 - X = (1 + K)e-~B21~t,
DERIVATION OF THE RATE LAWS
where E is the constant* which denotes the fraction of
Nucleation and growth reactions fall into two the untransformed material which is in grains in the
broad classes. The first,
termed discontinuous, process of transforming. One predicts, therefore, that
will be treated in this report. In this class of reactions, where grain boundaries are not crossed by the growing
the transformation product which nucleates and grows transformed material, and where K is small, a plot of
has the same average composition as the untrans- log log (l/l - z) vs. log t would rapidly approach
formed material, and, except for a short distance slope unity whenever t > I)/@.
from the advancing interface, the untransformed An alternate assumption, and one which will be
material has the original composition. One can assume treated here, is that grain boundaries offer no resistance
that, after a time, steady-state conditions are achieved to a growing nodule. It is believed that this assump-
in which both the growth rate and the specific tion is a better approximation to the true situation.
nucleation rate are
independent of time. This Wherever grain boundaries stop a nodule, nucleation
assumption will be treated first. In the last section we on the other side appears to be greatly enhanced, and
shall trea,t the case where, for some reason, steady- this situation should be better approximated as
state conditions are not achieved in the nucleation
rate during a significant part of the reaction time.
The other class, to be discussed more fully in a
subsequent paper, is where the transformation
product is of a different composition and the untrans-
formed material is gradually changing composition.
Here the nucleation rate in the untransformed
material may be a strong function of time and distance where 6 = g ; z and y EUPC
dummy variables;
from a growing transformed region, because the (1 + ?/)2-a
and w = exp (4nb [ -(I -Y -z)
nucleation rate is a strong function of composition. 1 4Y 1
In the case of grain boundary nucleated reactions 1+?/+z
+(1+Y--~ogY+(l+Y)log~+
(and incidentally, for reactions which nucleate on
+ z lo_ (1 + z/P - x2
dislocations), the high diffusivity in a boundary could 0
4 II
be expected to affect furt,her nucleation drastically K is a function of b only. For small b, K is proportional to 6.
CARX: GHAIN BOUNDARY
contSinued growth and ncrt as the relatkely rare from a plane or edge, assuming that these are only
~~~~~~~n~~ of original ~~oleat,ion. impinging with nodules origin&ing from the same
fn the study of pea,rlite, for instance, where the plane OPedges. One next considers a random distri.
parameter b is small, a plot of Iog Iog (l/i - 2) vs, bution of planes or edges each giving rise tn a certain
lag the usudy results in a slope of 4 or more? which transformed v&me, and considers irn~i~~e~nent
is &e dope one WRUIdsxpe& from rand~mI~ d&k=i- a~~~~~~ L&se volnmes.
but.ed nodules growing with no res~rietian to their For the case of corner ~I~cIeat.iu~~,
it is just. assumed
%%~a1size except inlping~n~ent with one another, that tShe corners are randomly distributed, and
The occurrence of a,slope of more than 4 appears ta be limited in number.
due to t.ransienLeffects in nucleation.
The derivation of t,he rate laws center around tke
pmblem of ~~inge~~el~t:. The newly formed grain is
assumed to grow at cor~tsnt radiaI velocity until it
impinges on another growing grain. ft then st~3p3
growiq along the area of e#nta.i.&1%KiIl be e~i~~eR~e~~~
ta use a device used by Avrami,@) a.nd Johnson and
Mehl called extended volume fraction &X4.This is the
sum of the volumes of all grains, divided by the total
~olumc, ~s~~~i.~~gthat the grain5 never stop growing
and t*hatnew grains keep nucleating at the same rate
in transfornl~ as well us untransf~r~~ed material,
XT, can esmd unity, 1-f nucleators is randomly
dis~ri~~uted,then the a&al volume fraction X is
rigoronsly given by
xEZZ.2 -@--X6
For the case of grain boundary and grain edge The exknded volume fraction due to a b~und~y area
n~~cleat~ reactinns, two types of ~rn~~ngel~en~wiU be S per unit volume
treated separat,ely. Thie first type is impingement;
among nodules ~r~ginati~~~from the same pIma OF where
sdge. For example, if this ~oenrs early in the reaction,
the individual nodules would cease growing parallel to
the plane or edgee, further growth only occurring
~~r~en~~~~a~ $0 the edge or plane. Armeq3re8sion is
derived for bhe total volume of the nodules o~gina~ing
462 ACTA METALLURGICA, VOL. 4, 1956
and if this boundary area is assumed randomly time at a time between t and t + dt is e-c dt. The
distributed in the specimen extended volume fraction at time t is therefore
= 0 G(t - 7) ( T. It will be noticed that all of the above rate laws can
be plotted as log log (111 - XI) vs. log t to give
Integrating with respect to t and substituting x for
characteristic master curves, Figs. 1-3, in which two
r/Q& we get
parameters are undetermined as additive scale factors.
__- 1+2/l-9 Thus, instead of a family of curves obtained by
d1-~2-~210g--~-~- - x<l,
2 I
Johnson and Mehl, a single curve is obtained. The
parameters of Johnson and Mehl are related t,o the
zzz.z 0 x> 1. parameters a and 6 in the following way: The bs are
The volume occupied by nodules originating from a equivalent to Johnson and Mehls shape factor 1, in
unit length of E is that they can both be expressed as 8jD4/G times a
1 geometrical constant. The us are respectively
v,z m2?7rZ ar = 2rG2t2 ~(1 - eeze) ax. (4) proportional to bs13Z, b,l12Z, and b,Z. All curves have
10 s0 slope 4 when a is small and approach the equation
Randomly distributing an edge length L in zt unit X = 1_ p/3&GP
volnme will give
which is identical with that, obtained for randomly
X = 1- exp f-b,-lf,(u,)]
placed nucleation sites. In the vicinity of a = 1, the
where
curves change slope. This change of slope is due to a
process which we shall call site saturation. In the case
of corner nucleation it occurs because there are only a
limited number of corners. In the case of grain
boundary and edge nucleation it occurs because the
nucleation sites are not randomly distributed, but
$,(a8) was evaluated, using Simpsons rule. are in the vicinity of other nucleation sites. Hence,
they have a greater probability of being transformed.
Comer Nucleation Therefore, the fraction of boundary area or edge
Here there is no problem about impingement length transformed is larger than the volume fraction
between nodules starting from the same corner. transformed, and the nucleation rate which is pro-
The first nucleus is the only one which will determine portional to the untransformed area or length drops
how much volume will contribute to the extended off faster than the untransformed volume. In the
volume. If the specific nucleation rate per corner is I,, limit of a > 1, nucleation is so small conlpared to the
then the fraction of corners nucleating for the first untransformed volume fraction that the curves can be
CAHN: GRAIN BOUNDARY SUCLEATED REACTIOXS 463
O*(11-
O-0( )l-
0.1
10
t which is independent of grain size.
At very large values of b the sites would saturate
-, t: , very early in the reaction, and furt.her nucleation
0
would cease. For a grain boundary nucleated reaction
?
the t,ransformed region would occupy an ever-
2
y; 0.1 thickening slab of material with the original boundary
at its center. The rate law would then be of the form
2
e
001
which is wha,t equation (3) approaches. Sirnila.rl~,
one would expect that for edge-nucleated reactions
O.OOll
0.1 10 1 100
as -
FIG. 1. Master curve for t,ransformations which nucleat,e which is the limit of equation (5) for large values of b.
on grain boundary surfaces.
0.01
they can be generalized to the time for any fraction
0.1 1 10 100 1000 completion by substituting for ln 2, In (l/l - X),
where X is the desired fraction.
(2, M,,, W
Substituting 0.5 for X, and a = a(t,n), where tlls is
FIG. 4. Theoretical curves for half-times of transformations
which nucleate on grain boundary surfaces.
This Paper
Johnson and Mehl------
X = 1 _ e-- $CG3t3
6 l/3 =fs(G
S
In2
b _fe(ae)
e
In2
,I,3 =fc(o
c
In2 .
S, L, C into the bs and rearranging, we obtain should incrertse with the inverse diameter squared.*
At sufficiently la*rgegrain diameter, the grains have no
trouble nucleating, and the half.time is the time
required for the growing transformation product to
consume balf of its grain, which va.ries directly as the
d&meter.
The following conclusions can be drawn from t,he
curves: (1) If, and only if, site saturation occurs,
D = 4.18 [f&~~)]~~,
G4/2 is a8pproximately equal to O.lD. This corre-
since sponds to the upper portion of the curves, and is
independent of the type of nucleating site. (2) tr,, can
never be appreciably fess than 0.1 DIG for any reaction
where nucleat,ion occurs only on boundaries or
i~ltersect~ons of boundaries. (3) At su&iently tow
values of the parameter, b = ~~~4lG), the half-time is
given approximately by (&Gs)-rI, if the trsnsforma-
tion prodnct cnn cross boundaries. This is the same
result one would expect for homogeneous nucleation.
Since the fs are i~de~endeni~ of ,aa,in size, sub-
a plot of as against9.7 $,(a,); as against 8.8r~e.(~~)~~~3~
&it&ion into ($?Ga)--1~4 leads to the result tha,t t,,, is
of a, ngainst 4.18 ~C(uC)]1/3 is a plot of I) against Gi& proportional t.o the .ZW4 for boundary nucleation,
for the various types of nucleation, v&h D and G&, EPs for surface nucleat~ion,Df4 for corner nucleation,
plotted in the same units and Gfl to the appropriate as well as D1 for site saturat,ion and I>a for homo-
power as the scale factor. This is plotted in Figs. 4-6 geneous nueleiltion.
on a double log plot. The scale factor is now an
ndditive constant. Since it is the same additive
Rates if More Than One Ty-pe of&&? is Active
constant, the only permissible translation for the
Whenever more than one type of site i+ active,
master curves, if fG is unknown, is along a 45 line.
the impingement problem is considerably more
Also, points having a certain vahe of ~~~~!~~~~ are
oompficoted. Certa*inlimits can be discussed, however.
constrained to he on a certain 45 tine. A (~~~~~~~~)
Until one type of site begins t,o saturate, the rate law
axis is therefore indicated.
can be approximated by
l?igs. 4-6 can be used to obtain relationships
between Gtriz, D, and f/G, or any of their ~ornbin~t~o~s~
if the type of site responsible for nucleation is known.
A knowledge of the parameter It for example is where .& is the sum of all the $%.
sufficient to determine the value of i~~~~~~~ and the Because of t,he geometry, when the boundaries
parameter a. An ex~rimenta~~y determined relat,ion- saturate, the edges and corners saturate too, regardless
ship among these parameters oan be used to determine of their specific nucleation rate. When the edges
what type of sit,e is xesponsible for nucleation. For saturate, the corners saturate too, while the boundaries
comparison, the curve obtained from Johnson and can continue to nucleate. When corners saturate,
Mehls equations is also plotted (Fig. 4-dotted curve). both edges and boundaries can continue to nucleate.
The agreement is very good along the upper portion The sat.uration of a type of site Emits ail further
(rl_IP~$). However, one sees immediately that one nucleation on sites of lower dimensional~ty. t_
consequence of the ass~l~ption tha~t the trans-
formation product cannot cross a grain boundary is *Another way of looking at it is that. w-hen the grains
that, at small enough grain size, the reaction rate become sufficiLtntly small, the volume nucleation rate will
increase as tho surface area or as l/t>, while the amount of
decreases as the grain size decreases. This is apparent transform&ion per nucleus will decrease as P. The over-all
if one considers the consequences of that assumption. rate will decrease as P, and the half-time will incres,se I~Rl/D*.
For grain-edge and grain-corner nucleat,ion the half-times
At sufficiently small grain size the time for half- will similarly depend on D- end Do instead of 2P2 and Da.
reaction will be that required for half of the grains to $For convenience, we shall refer to the dimensionalitg of Q
site as heing 0 for corners, 1 for edges, 2 for boundaries, 3 for
obtain at Last one nucleus. However, t;he area hmnogeneous nucleation. It is intended only as a notation,
a.vailable for each grain t,o receive a nucleus decreases and what foXlows does plot hold for nucleation on other types of
sites to which a ~~e~siona~~y can be assigned, such as
as the square of its diameter, and the half-time vacancies, dislacation, etc.
456 ACTA METALLURGICA, VOL. 4, 1956
f @(l - 0)s
167~y3
and the maximum R
(AH)2 Teq log D/S
R *,tax- -___-_ at 8 = l/S.
36ay3
A, > &-;/! k-
FIG. 9. Typical example showing conditions for bar+
edge nucleation would never predominate, and the detectable nucleation 1; nucleation sufficiently rapid for
corner saturation 2; edge saturation 3, boundary surface
area marked edge in Fig. S would be absent. saturation 4.
455 ACTA METALLURGICA, VOL. 4, 1956
R rises initially as the undercoo~ng squared reaehes Similarly, one can derive an expression for homo-
GM, at e = 8, and has a very long plateau. If R,,, geneous nucleation, using J, for the number of nuclei
is small for certain values of K, only modes of lower present at time zero. This leads to
dimensionality will ever predominate. If R,,, is large,
R in the regions of interest in Figs. 8 and 9 can be X,=l.-eq-:;IG3t3[Jv
taken as proportional to undercooling squared.
for grain boundary nueleat~on; The determination of the n~~cleat,ion rate from a
rate law is therefore a very complicated problem.
X, = I - exp Even for equation (15), which has a simple solution
mathematically,
1 P log (1 - X)
x exp & = - s7Tc3-.-.-~-~~~
at4
x I,((1-g2- Y2]12dr))
dy]) (13) it is unlikely t,hat kinetic data can be obt,ained with
sufficient accuracy to make t,his solution useful.
for edge nucleation; and If the nucleation rate follows an expression of the
type
X,=1-exp(~CC$3~~-(l-J~)~
Ij = Njt
X 1,(1 --:I3 X exp (-llOdi) dr]) (14) it is possible to integrate equations (12-15). A plot of
log log (l/l - 2) against log t would be similar to
for corner nucleation. those of Figs. 13, except that the initial part of the
CAHN: GRAIN BOUNDARY NUCLEATED REACTIONS 459
curve would have slope 4 + n. The parameters a and where K = B,S, B,L, or B,C respectively for grain
b would become of the form boundary, edge, or corner nucleation. The dependence
a, = (BSG2)11(a+n). t of h/2 on grain size far from saturation would then be
B,
b,= ~
S:j+nGl+n
b, =
Be
Ll +nlZC;rl
+?I
REFERENCES
a = B l/(1+12) . t
c G
1. P. J. CLEMM and J. C. FISHER Acta Met. 3. 70-73 (1955).
b, =-
Bc 2. W. A. JOHNSONand R..F. MEHL Trans. Amer. Inst: Min.
(J(l+n)/3@+n Met. En+ 135, 416-442 (1939).
3. M. AVRAMIJ. Chem. Phys. 7, 1103-1112 (1939); 8,212-224
The half-time at saturation would be independent of (1940); 9, 177-184 (194-l).
4. J. E. BURKE and D. TURNBULL Progress in H&l Physics
the nucleat.ion rate, while the expected half-time at 3, 226 (1952).
low nucleation rates would be 5. E. SCHEIL and A. LANGEWEISE Arch. Eisen~hiittenw 11,
93 (1937).
6. F.C.HULL,R.A.COLTON,~~ R.F.MEHL Trans. Amer.
t,,, = Inst. Min. Net. Eng. 150, 185-207 (1942).
7. D. TURNBULL Trans. Amer. Inst. Min. Met. Eng. 175,
774 (1948).
8. J. C. FISHER, from ASM Symposium on Thermodynamics
in Physical Metallurgy, p. 201-241.