THEORETICAL CONSIDERATIONS*
A. J. ARDELLt
The theory of diffusion-controlled particle coarsening developed by Lifshitz and Slyozov and Wagner
(LSW) is modified to take into account the volume fraction, +, of precipitate. The modification is based
on the supposition that the diffusion of solute to a growing particle (or from a dissolving particle) will
depend on a distance characteristic of the spatial distribution of the particles in the polydisperse assemb-
ly. The characteristic distance chosen for computational purposes is related to the mean free path
between a particle and its nearest neighbor. While this is not the best choice, it serves to make the
problem tractable analytically. Steady-state diffusion under conditions of spherical symmetry is also
assumed.
The modified LSW (MLSW) theory predicts that while the basic t ijs kinetics of the LSW theory are
maintained, the coarsening rate increases with increasing volume fraction, even at very small values of
+. According to the MLSW theory, the theoretical distribution of particle sizes broadens rapidly at
small values of $. In the limit + = 1 the theoretical particle size distribution becomes identical to that
derived by Wagner for the case of interface controlled coarsening. All the results of the LSW theory are
recovered from the MLSW theory in the limit of zero volume fraction.
The results of the MLSW theory are in very good quantitative agreement with the available data on
the coarsening rates of Co precipitates in Cu-Co alloys. The apparent absence of volume fraction effects
on particle coarsening in a number of other alloy systems is discussed in light of the assumptions in-
volved in both the LSW and MLSW theories.
where c’ is now the concentration at R = r’, we see instantaneously growing or dissolving depends upon
that equation (1) is recovered in the limit r’ + co. the value of c,~ compared with the value of cl, irrespec-
Provided that steady-state diffusion to an individual tive of the precise diffusion geometry. The value of c’
ARDELL: ON PARTICLE COARSENING 63
(14)
- = 3crDsz(p - 1) (1 + BP).
dt
By setting equation (17) and its first derivative with discuss this point briefly later on). The equation that
respect to p equal to zero, we find describes the particle size distribution follows directly
3 from the expression presented by Sauthoff and Kahl-
3P7n2
(21) weit, who show that
* = @pm2 + (lp” fi)p, - 1 = 1 + 2BP, - B
where
(B” + p + 1P - (1 - #0
Pm = (22)
B * where A is a constant. The value of A can be obtained
The new parameter, pm, introduced in equations (21) by normalizing g(p) such that
and (22), is the theoretica relative maximum particle
size of the polydisperse assembly. The theoretical (24)
distribution of particle sizes must go to zero when
Letting
P 2 pm-
From equations (21) and (22), it is easy to show that
as ,4 -+ 0, v --+ 2714 and pm -+ 312. Thus, these results
of the LSW theory are recovered in the limit of zero
equation (24) can be rewritten as
volume fraction. As /3 + CD(#-+ 1)~ -+ 0 and pm + 2.
Consideration of this result in light of equations (19) A f(P¶n)
er(@-= A
and (20) shows that the coarsening rate should in-
(efh%d - gf(O)) = 1
(25)
3 s f(O) 3
crease without limit as the volume fraction increases
towards unity. Furthermore, the maximum particle The fun&ion j’(p) is zero at p = 0 and decreases,
size in the polydisperse assembly approaches twice approaching -CO as p approaches pm. This follows
the critical particle size, so that we can anticipate a from the fact that dp3/dr, from equation (17), is equal
broadening of the theoretical particle size distribution to zero when p = pmex. The value of A is then,
as CJincreases. The complete dependence of v and pm from equation (25), equal to -3. Thus,
upon Q is illustrated in Fig. 3, where it is observed
that v decreases and pm increases monotonically as + g(P) = $exp (S,‘$j$) (26)
varies from zero to unity.
Equation (26) can be inflated analytically only
when p = 0 (# = 0) and ,8 = co (# = 1). For the ease
In order to evaluate the variation of the distribution $I = 0, we obtain (apart from a no~alization constant)
of particle sizes as a function of 4, we can follow the the result of the LSW theory. When @ = CO the
procedures used by Lifshitz and Slyozov, or the particle size distribution is
particularly clear presentation of Hillert. Fortunate-
ly, this has already been done by Sauthoff and ; ps2
(27)
Kahlweit,(12) who arrived at equations (17) and (21)
by entirely different physical arguments (we shall t.0;p>2
Curiously, equation (27) is formally identical to the
‘r- F2.* distribution function derived by Wagner for the
case of interfa~-eontro~ed coarsening, and also to
the ~st~but~on function derived by Hillert for the
case of three-~mensiona1 grain growth, both of which
are governed by PI2 kinetics.
pm Figure 4(a) illustrates how g(p), computed numer-
i .6 ically, changes as a function of 4. The greatest change
in g(p) occurs for small values of 4, as illustrated by the
curves for $ = 0.005 and 0.05. When I?t,is greater
than 0.2, the shape of the distribution curve becomes
1.6
I
*o ,Y--Y
0. I 0.2 0.3 0.4
I a
0.5
”
/
0.6 0.7
I 0.6 *
0.9 1 1 .5
1.0 relatively insensitive to +.
In order to facilitate the comparison between
experimentally determined particle size distributions
Fra. 3. Illustrating the dependence of the parameters Y
and pm on the volume f&&ion, #. and the predictions of the NLSW theory, it is useful
ARDEL’L: ON PARTICLE COARSENING 67
0.90 -
0.88 I
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 3
d,
p’=r/i
W The rate fondant for coarsening, k(4) is thus
3%. 4. (a) Ill~tr~t~g the dependence of the theoretical
distribution of particle sizes on the volume fraction, 4;
(b) the particle size distributions redefined as functions
of the variable p’ = rjF_
ACTA METALLURGICA, VOL. 20, 1972
~(#) 27 IT3f#)
---= --
W) 4 44)
is shown in Fig. 6(a) for small volume fractions, and
in Fig. 6(b) for volume fractions up to 0.65. It is
seen in Fig. 6(a) that even at small values of 4, the
effect of 4 is appreciable. The coarsening rate is
twice that of the LSW theory when 4 is only about
0.008, a factor of three times as great when 4 is only
about 0.027, and five times as great when 4 is about
0.073. From Fig. 6(b), a ten-fold increase is predicted
01 I I I I 1 I for 4 ‘U 0.25, and a twenty-fold increase for (b N 0.55.
0 0.1 0.2 0.3 0.4 0.5 .0.6
To facilitate the comparison between the MLSW
4
theory and experimental data, the important para-
(b)
meters /l, Y, pm, j5 and ail are tabulated as a
FIG. 6. Illustrating how the rate constant, k($), for
diffusion-controlled coarsening varies with the volume function of 4 in Table 1.
fraction, r$; (a) at small volume fractions; (b) at volume
fractions up to + = 0.65. 3. APPLICATION TO COARSENING IN Cu-Go
ALLOYS
which, unlike k’($) in equation (20), can be compared
The only data in the literature that the author is
directly with experimental data. To emphasize the
aware of in which a volume fraction effect has defin-
effect of + on the coarsening rate, it is convenient to
itely been observed, are the data of Livingstono3) on
plot the ratio ~(~)/~(O), where
the coarsening of coherent Co-precipitates in Cu-Co
~GcDLR 8yc,D@ alloys. Ardell and Nicholson, in their analysis(14) of
k(0) = -$-- = ~
(31)
9RT Livingston’s data, showed that the interfacial free
)
ARDELL : ON PARTICLE COARSENING 69
k
fcms@3c x lO**f Y
@rgfcW
wt. XC0 #co k(#) k(O)
(measured) (calculated) - Old NC?&,
:::: 600
650 0.0168
0.0184 50.90
7.96 20.20
3.07 451
388 1.54
174
::: 600
650 0.0318
0.0195 13.00
0.83 5.62
0.31 IO20
521 318
199
Z-
energy, y, calculated from the experimental values of With respect to other available data on coarsening
k, increased with increasing $oo. These data, re- kinetics, the agreement between experiment &nd the
analyzed in light of the MLSW theory, are presented prediotions of the MLSW theory does not appear to be
in Table 2. Values of k(0) were obt&ned from the very good. Particle coarsening has beeninvestigated in
measured rate constants, k(#, from gr~ph~~~y a variety of alloy systems in which no effect of volume
i~~rpol~~d values of the theo~tio~l ratios in Table 1 fraction on the evening rates has been observed,
at the appropriate vaiues of 4. At T = 6OO”G, the In the N-Al system, for example, the y’(Xi&.l)
k(0) values so obtained are nearly constant as pre- precipitate coarsening kinetics have been investi-
dicted by theory, the only notable exception being gated(idJ6) in alloys containing from 9 to 20% y’*
the value for the 3.2% Co alloy (d, =: 0.0318). The There is no question thsGt the rate constants are
recalculated values of y(new), obtained from the entirely independent of t& over this range. In fact,
values of k(0) in Table 2 are seen to be independent in current investigations in this laboratory the y’
of &o- Furthermore, the new values ere in good coarsening kinetics have been studied in &Boys con-
agreement with the vctlue of approximately 196-200 taining up to 60% y’ and no effect on k has been
ergs/cm2 obtained by Servi and Turnbull(f5) in their observed.u7) This is puzzling, not only from the view-
study of precipitation kinetics in Cu-Co alloys. point of the MLSW theory, but also from that of the
original LSW theory, since the mean particle separ-
4. DISCUSSION ation in most of these alloys is much smaller ths,n the
Perhaps the most surprising result of the MLSW average particle size.
theory is the rapid variation of the coarsening param- Other systems in which no volume fraction
eters in TsbIe I at small volume fractions. This effects have been observed include Gu in ,-Fe@*) and
r&pid variation is due, physically, to the assumption NbC in liquid Fe .f6) The latter system wss studied
that only the mean free path between a particle and by Sttrian and Weart, who specifically looked for 8
its nearest lzeiqhbor affects the value of r’ in equation volume fraction effeot at very large values of
(1). The large effect on the coarsening rates predicted 4NbC (0.60-0.90). They found that t1/3 kinetics were
at small values of (F,is in good quantitative agreement obeyed over the entire range, but that the rate
with the data on the Cu-Go system, even though +oOis constsuts were independent of &,o. However, they
oonspjeuo~ly small. This is certainly an encoumg- also found that while the number of particles per unit
ing sign that the MLSW theory is basically correct, volume, Z, decreased with time as E-I, in agreement
because the assumptions used in the derivation of with W~gner’s(2~ predictions, 2 was also independent
equation (7) are increasingly accurate as I# approaches of 4, which is an impossible result. The source of
zero. Moreover, the applicability of the steady-state this inconsistency is not evident.
solution to the diffusion problem, equation (2), The results of Speioh and Oriani(ls) on the coorsen-
becomes more realistic as #J approaches zero. It is ing of Cu precipitates in &Fe are somewhat more
worth reiterating that P’~calculated from equation (11, disquieting. The range of $o,, in their study w&s
actually underest~~t~s the value of the “r&ius” of roughly equivalent to th& of #o. in Livingston’s Gu-
the Diriohlet region. Therefore, if we could oalmzfato Co alloys, and k was found to be clea.rly independent
this radius,we would find that the values of the param- of the volume fraction, The shape of the Cu pre-
eters in Table I wauld depend less strongly on +. cipitates in this alloy is not spherical, a factor which
This factor would probably improve the overall would influence the quantitative aspects of the diffu.
agreement between theory and experiment for the sion geometry, but on a semiquantitative basis a
Cu-Co system. volume fraction dependence of the coarsening rate
70 ACTA METALLURGICA, VOL. 20, 1972
would still be expected. The fact that none is ob- that owing to the difficulties usually encountered in
served is totally inconsistent with the MLSW theory. measuring particle size distributions to a high degree
However, the quantitative agreement with the LSW of accuracy, it is clearly difficult, if not impossible, to
theory is not too good in this case either, because the decide between diffusion-controlled and interface-
values of the interfacial free energy calculated from controlled coarsening on the basis of experimentally
the measured coarsening rates are too large by a measured particle size distributions alone.
factor of approximately eight. On examination of the available evidence, the
In addition to the available data on coarsening MLSW theory appears to fail miserably in explaining
kinetics, it is also possible to compare experimentally the data on the systems containing high volume
observed particle size distributions with the predic- fractions of the dispersed phase. As has already been
tionsof theMLSW theory. In almost every case known pointed out, this is puzzling because the original
to the author, the experimental distributions are LSW theory would not be expected to hold when the
broader than the theoretical distribution of the LSW average particle separation is smaller than the
theory. In the Ni-base y-y’ alloys there is good particle dimensions. According to equation (6), we
evidence that the magnitude of the y-y’ lattice misfit, would expect to find the condition i = r* in a mono-
which varies from one system to another, is at least disperse system when ,5 = 2, or when $ is approxim-
partly responsible for the observed broadening.og) ately 0.04. Even if we double this value to allow for
However, this factor cannot account for the other assorted errors introduced by various assumptions,
observation that in the Ni-AP and Ni-Sic12) systems the values of q$ in the y-y’ alloys studied to date all
the particle size distributions are evidently independ- exceed 0.08, and in the NbC-liquid Fe system studied
ent of 4. by Sarian and Weart the liquid Fe matrix was actually
In at least two systems, isoamyl alchol droplets in the minor phase. Nevertheless, for the y-y’ alloys
watert20) and u-Mn in Mg, fzl) the observed broadening in particular the quantitative agreement between
of the particle size distributions cannot be attributed experiment and the LSW theory is exceptionally good
to coherency strains. Unfortunately, the data on in all respects, save for the aforementioned breadth
these systems are not extensive enough for a quantita- of the particle size distributions. This is viewed
tive comparison with the MLSW theory. However, here as truly remarkable, since the particles are so
it is encouraging that the particle size distributions close together in most of the y-y’ alloys that neither
are broader than predicted by the LSW theory, since equations (1) nor (2) should be valid for describ-
the particles are spherical in both of them and the ing the concentration gradient at the interface of an
effect of volume fraction is the only factor of signifi- individual particle. It simply seems unreasonable to
cance, apart from expected experimental uncertain- expect steady-state diffusion to an individual particle,
ties. under conditions of spherical symmetry, to occur
It is interesting to note that the theoretical particle when 4 is so large. Viewed in this light, the failure
size distributions of the MLSW theory he somewhere of the MLSW theory at high volume fractions is
in between the LSW result for diffusion-controlled easier to accept and understand than the apparent
coarsening and the Wagner result for interface-con- success of the LSW theory.
trolled coarsening, which is identical to equation (27). Finally, it was pointed out earlier that Sauthoff and
This could account for some of the confusion which Kahlweit arrived at results similar to those presented
has arisen as to whether coarsening was actually inter- in this paper using an entirely different type of argu-
face or diffusion-controlled. For example, Smith,(si) ment. They assumed that the diffusion coefficient
who investigated the coarsening of cc-Mn particles in could be expressed (in the present notation) as D =
an Mg matrix, concluded that the u-Mn particle D,(l + @r/r*). In their development, p is left as a
size distributions agreed reasonably well with both disposable parameter, since there is no way to relate
the interface-controlled and diffusion-controlled (LSW) it to any of the processes which occur during particle
theoretical distributions. The values of #a_-Mn in coarsening. Just why the diffusion coefficient should
Smith’s alloy were very small (the order of O.OOl), but depend upon p is not made clear by Sauthoff and
even at these low values perceptible broadening is Kahlweit [one could assume y = yo(l + pp) and
predicted by the MLSW theory. The feeling here is obtain the same result]. Though their theoretical
that the u-Mn particles do grow by diffusion-controlled results are formally identical to those of the MLSW
coarsening, and that the observed breath of the cr-Mn theory, the physical basis for their assumption of a
particle size distribution is at least partly due to particle-size dependent diffusion-coefficient is some-
volume fraction effects. We may conclude from this what obscure.
ARDELL: ON PARTICLE COARSENING 71
5. SUMMARY AND CONCLUSIONS ent of ~4even though they are generally broader than
The MLSW theory developed in this paper predicts the distribution predicted by the LSW theory.
that particle coarsening rates should increase with Since the assumption of steady-state diffusion to an
increasing volume fraction of precipitate, whereas the individual spherical particle should not be valid when
basic t1i3 kinetics for diffusion-controlled coarsening 4 > 0.08, the breakdown of the MLSW theory is not
are unaffected. Even at very small volume fractions nearly as surprising as the apparent success of the
(4 < 0.01) the effect of 4 on the coarsening rate is LSW theory in describing the data on the y+y’
significant. The theoretical distribution of particle alloys. There is clearly a need for systematic experi-
sizes is predicted to broaden rapidly with increasing mental work on carefully selected alloys so that this
4, and approaches the particle size distribution for incongruity can be resolved.
interface-controlled coarsening in the limit 4 = 1.
6. ACKNOWLEDGEMENTS
The results of the MLSW theory agree with the
original LSW theory in the limit of zero volume frac- The author is grateful to Mr. Pramod Bansal for
tion. performing the bulk of the necessary computer work,
The predictions of the MLSW theory are in very to Professor Didier de Fontaine for valuable discus-
good quantitative agreement with the available data sions on several important points, and to the U.S.
on the coarsening of Co precipitates in Cu-Co alloys. Atomic Energy Commission for their financial support
However, the MLSW theory is inconsistent with the of this project.
observed absence of volume fraction effects on the REFERENCES
coarsening rates and particle size distributions for a 1. I. M.LIFSHITZ and V.V. SLYOZOV,J. phys.Ghem.Solids
10, 35 (1961).
number of other alloy systems, particularly those 2. C. WAGNER, 2. Elektrochem. 65, 581 (1961).
containing large volume fractions of precipitate 3. C. ZENER, J. appl. Phys. 20,950 (1949).
4. W. W. MULLINS and R.F. SEKERKA, J.appl. Phys. 34,
(4 > 0.08). In these systems, typified by the Ni-base 323 (1966).
y-y’ alloys, the particles are usually so close to- 5. S. S&IAN and H. W. WEART, J. appl. Phys. 37, 1675
(19fMl.
gether that individual particle growth under steady-
\----I-