Вы находитесь на странице: 1из 6

Materials Science and Engineering A 413414 (2005) 429434

About Kolmogorovs statistical theory of phase transformation


A.A. Burbelko

, E. Fra s, W. Kapturkiewicz
AGH University of Science and Technology, 23 Reymonta Str., 30-059 Krakow, Poland
Received in revised form 25 July 2005
Abstract
Many mathematical models of solidication use the following equation for the estimation of solid volume f
S
= 1 exp(n
1

n
2
), where
is time, n
1
and n
2
are constants. The fundamentals of the statistical theory of metal solidication were developed in classic papers written by
Kolmogorov [A.N. Kolmogorov, Bull. Acad. Sci. USSR. Ser. Math., 3 (1937) 355359 (in Russian)], Johnson and Mehl [W.A. Johnson, R.F.
Mehl, Trans. Metall. Soc. AIME 135 (1939) 416442], Avrami [M. Avrami, J. Chem. Phys. 7 (1939) 11031112; M. Avrami, J. Chem. Phys.
8 (1940) 212224; M. Avrami, J. Chem. Phys. 9 (1941) 177184]. The above equation in contemporary literature is named after these authors
(K-J-M-A), but this equation is only a particular solution of Kolmogorovs general solution f
S
=1exp(), where the function is the so-called
total extended volume (per unit volume) of the growing grains, if their overlapping is neglected. Kolmogorovs general solution for the -function
is an accurate development only in the case of geometric similarity of the growing grains, their uniform distribution and equal orientation in space.
Moreover, for an arbitrary moment and direction all grains must have the same absolute value of the growth velocity vector. If the conditions stated
above are neglected, the above equations can give overestimated results. The source of this overestimation is the, so-called, screening effect. The
presented solution takes into account the screening effect for the calculation of -function. This solution expands the scope of the Kolmogorovs
statistical theory of transformation. The proposed solution takes into consideration the growth of the grains of any shape, assuming their uniform
distribution in space.
2005 Elsevier B.V. All rights reserved.
Keywords: Transformation kinetics; Kolmogorov equation; Screening
1. Introduction
The rst empirical models of the solidication kinetics were
created in the rst half of the 20th century. Among todays empir-
ical models we can mention an Austin-Rickett equation. During
investigation of the isothermal decomposition of undercooled
austenite in steels, Austin and Rickett (acc. to [1]) proposed the
following kinetic equation:
f
S
()
(1 f
S
())
= B(T)
n
(1)
where f
S
is the volume content of decomposition products, B(T)
is a temperature-dependent value, is time, n is the constant.
For an arbitrary positive value of the coefcient n, the curve of
this function plotted in a time-related logarithmic scale assumes
the shape of the, so called, sigmoid curve. The above men-
tioned empirical equation is in many cases consistent with the

Corresponding author. Tel.: +48 12 617 27 34; fax: +48 12 633 63 48.
E-mail address: abur@agh.edu.pl (A.A. Burbelko).
experimental results [24]. As proved by Avrami [5], the Austin-
Rickett equation is a particular case of the general kinetic equa-
tion for structures composed of very ne grains still at the onset
stage of growth, before they start overlapping each other.
The rst model that allowed for an effect of crystal nucleation
and a linear motion of the crystal faces published in the 1937
by Kolmogorov [6]. However, the most often cited and most
popular in this eld are the works of Johnson and Mehl [7], and
Avrami [1,5,8].
Disregarding the fact that the results obtained by Johnson
Mehl (J-M) were a particular variation of Kolmogorovs general
solution (K), disclosed two years before the rst work by John-
son and Mehl appeared, most of the present day studies written
in the English language refer to the main equation in the statisti-
cal theory of solidication as to an A, J-M, J-M-A equation (the
rst letters of the names of its inventors), and occasionally only
they call it K or K-J-M-A solution [4,9,10].
The aim of the present study is an analysis of Kolmogorovs
equation related to Johnson and Mehl as well as Avramis
approach, and specically stating the conditions under which
these equations hold true.
0921-5093/$ see front matter 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.msea.2005.08.161
430 A.A. Burbelko et al. / Materials Science and Engineering A 413414 (2005) 429434
In the second part of the study a solution is presented to
cover the case of nucleation and the growth of particles, when
some of the conditions, for which Kolmogorovs or Johnson
MehlAvramis solution was developed are not satised.
2. An analysis of Kolmogorov, and Johnson, Mehl and
Avramis equations
The relationships between the volume content of transfor-
mation products and time, proposed by Kolmogorov, Johnson,
Mehl and Avrami, can be written in general form as:
f
S
= 1 exp(()) (2)
In this equation, function () has the following physical
meaning: it denes the, so called, extended volume of the
grains in the transformation product. This is the total relative
volume (that is the one related to a unit volume of the start-
ing parent phase) that all the crystals taken together might have
occupied, if they had had the chance to grow free, and if their
overlapping with the neighboring grains in the space and block-
ing of each other had been neglected:
() =

i
V
i
()
V
(3)
where V
i
is the extended volume i of grains in transformation
product (summed up it covers all the grains), V is the examined
volume.
Kolmogorov [6] allowed for random contact between the
grains and proposed a general kinetic equation. He gave exam-
ples of its application in the two simplest rules of the grain
nucleation and growth in the transformation products.
Let us now analyze the postulates on which Kolmogorovs
solution was based. In [6] the following assumptions were made:
P1 There is a volume V, which for =0 is totally occu-
pied by the parent phase. With a time lapse a part of
this volume V
1
() becomes occupied by the grains of
a transformation product.
P2 The volume of an arbitrary grain V
i
is much smaller
than V.
P3 In an arbitrary part of the volume of a source phase
V

<VV
1
() the solidication sites, i.e. the solidi-
cation nuclei, are formed. The probability of formation
within this volume of precisely one single grain during
the time lapse is equal to
p = ()V

+o() (4)
The probability of more nuclei appearing within this volume
is o(), and thus it has a value innitely small compared to .
This probability is independent of the distribution of nucleation
sites which have been previously formed. (It can be proved that
function is a volume-related nucleation rate.)
P4 The crystals are of a convex shape.
P5 Since the onset of nucleation, vector u of the lin-
ear velocity of motion of the surface of each grain
towards its parent phase will depend on time and
direction n:
u(, n) = k()c(n) (5)
where k() is the scalar function time-dependent only, and c(n)
is the vector whose modulus is direction-dependent only.
Function k() allows for changes in the growth rate irre-
spective of direction, while the vector function c(n) is time-
independent and allows for the anisotropy of the grain growth.
The last postulate means that the grains need not be of a
spherical shape and can assume various dimensions (which will
depend on the exact time when their growth has begun). A con-
nement is here the necessary condition that all the grains should
be of similar geometry, should have the same orientation in space
and a convex shape. This means that by analyzing an arbitrary
single grain, through application of mathematical operations of
a parallel transfer in space (without rotation), combined with
isotropic scaling, we should be able to picture all other single
grains. The linear growth velocity of all the grains in an arbitrary
direction at a given time will be the same, but it may depend on
the direction and may change in time (in a way similar for all
the grains).
During transformation, the grains are at random arrested in
their growth by the neighboring grains, which make themchange
their shape and symmetry. An extended volume of the grain
which starts growing at t
0
, is larger than its real volume, and at
the time instant t it amounts to:

i
=
4
3
c
3
__
t
t
0
k()d
_
3
(6)
where the value of coefcient c
3
is calculated by integration
done on the surface of a unit sphere S, with the centre placed at
the starting point of the system of coordinates:
c
3
=
1
4
_
S
c
3
(n)d (7)
If condition (5) is observed, the ratio between the functions
c(n) and k() (that is, their normalizing) can be selected in an
arbitrary way. Setting the k() allows for the maximum growth
speed on the surface of a non spherical grain, yields the integral
in Eq. (6) which determines the magnitude of the radius-vector
of a point on the grain surface in the direction of the most rapid
growth. Obviously, this choice in the case of isotropic growth
(that is for spherical grains) will make the integral in Eq. (6)
determine the radius of a sphere (c
3
=1). If the shape of the
grain is not spherical, then c
3
determines the volume ratio of
this grain and of its circumscribed sphere.
The common extended volume of all the grains per unit vol-
ume will be [6]:
=
4
3
c
3
_
t
0
(t

)
__
t
t

k()d
_
3
dt

(8)
where integration in brackets is performed since the moment t

of nucleation of the individual grains.


A.A. Burbelko et al. / Materials Science and Engineering A 413414 (2005) 429434 431
In this case Eq. (2) will take the following general form:
f
S
= 1 exp
_

4
3
c
3
_
t
0
(t

)
__
t
t

k()d
_
3
dt

_
(9)
Apart fromthis general solution, Kolmogorov proposed solu-
tions for two particular cases. In the rst case, when the nucle-
ation rate and the velocity of linear growth k() are constant,
the integration of Eq. (9) loads to the following formula for a
volume content of the transformation products:
f
S
= 1 exp
_

3
c
3
k
4
_
(10)
The second of those particular solutions is related with the
condition of an instantaneous nucleation of all the grains and
their growth at a constant rate. In this case, the value of function
will equal zero over the entire time interval; the only exception
will be an innitely small interval t at the very beginning of
the process. Within this interval , while product =
is conned and determines the number of grains in a unit
volume. Under such assumptions, integration of Eq. (9) loads to
the following relationship:
f
S
= 1 exp
_

4
3
c
3

3
_
(11)
Often, Kolmogorovs name (or K-J-M-A) is associated
not with a general solution (9), but with particular solutions
[4,10,12,13], having a common form similar to (10) and (11):
f
S
= 1 exp(n
1

n
2
) (12)
Taking the above into consideration, we propose to give the
name of Kolmogorov equation to a general solution (9), while
the name of K-J-M-A should rather be used for particular solu-
tions of the type given in (12).
Kolmogorovs general Eq. (9) and its particular forms (12)
(providing all the postulates mentioned in [6] are satised)
describe not only the solidication of metals but also the kinet-
ics of various phases and structural transformations taking place
in metallic and non-metallic materials in solid state; even the
growth of a colony of microorganisms. The possible ranges of
application of this equation and the relevant reference bibliog-
raphy are given in [11].
In the literature on the kinetics of phase transformations one
cancome across a formulationwhichsays: deviations fromK-J-
M-Aequation [1218]. Twocases shouldbe distinguishedhere.
The rst regards trials on the use of a particular solution (12)
to describe transformations accompanying the nucleation and/or
the linear grain growth when both proceed at a variable velocity.
The formulation which postulates the non-Kolmogorov equa-
tion is not valid for this situation, since the general solution (9)
still holds true.
The second case concerns a non-accomplishment of one of
the postulates P1P5. Under these circumstances, general solu-
tion (9) will no longer be valid. The most frequent cause of
deviations are:
- concave shape of the growing grains (e.g. dendrites growing
during solidication);
- random spatial orientation of the anisotropic convex grains;
- different velocities of the linear growth of grains in several
structural constituents;
- non-random spatial grain distribution.
3. Non-Kolmogorov growth
A detailed analysis of Kolmogorovs assumptions and of the
consequences resulting fromtheir non-observance was given by
Belenkiy [19]. Kolmogorovs proof was based on determinat-
ing of the probability of an arbitrary point in the liquid volume
being occupied by the extended volume of an arbitrary grain.
Belenkiy has proved that in the case when postulates P1P5
are satised, capture of this point by an extended volume of at
least one single grain will be equal to this point being occupied
by at least one of the growing grains (e.g. point O in Fig. 1a
is occupied by grain B and by the extended volume of grain A
and grain B). On the other hand, if the condition of a similar
orientation of the grains in space (Fig. 1b) is not satised, it
may happen that the point occupied by an extended volume of
the grain will physically remain in the untransformed phase. An
example like this is shown in Fig. 1b: the extended volume of
grain A occupies point O, but physically grain A has access to
this point blocked by grain B. The same situation may happen
also in the case when concave grains are present (Fig. 2).
The situation described here is known as a screening effect
(also shadowing or shielding). Belenkiy [19] has proved that
non-observance of the above mentioned Kolmogorovs postu-
late results in overestimation of the volume of the transformation
products according to Eq. (9). A fragment of the extended vol-
Fig. 1. Schematic representation of the screening effect with non-uniform
distribution of particles in space: (a) uniform distribution of particles in
spacephantoms are not formed; (b) phantom marked by gray.
432 A.A. Burbelko et al. / Materials Science and Engineering A 413414 (2005) 429434
Fig. 2. Schematic representation of the screening effect for concave grains:
phantoms P
A
and P
B
.
ume of the screened grain separated by the screen is usually
called phantom. Point O in Fig. 1b belongs to the phantom of
particle A. On this account, in the case of non-uniform distribu-
tion of the grains in space, or when their surfaces include some
concave parts as well, the classic Kolmogorov equation will give
an overestimated result. The value of the error will be equal to
a probability that a point selected at random in a fragment of
the parent space still not occupied by the growing particles will
move to a phantom. A similar result (overestimated appraisal)
will be obtained in the case of the grains growing at different
velocities in several structural constituents.
Starting with a monograph written by Belenkiy [19], the
effect of screening during particle growth was discussed many
times, but unfortunately an analytical solution offered until now
has been valid only for the task of a one-dimensional growth
in instantaneous and continuous nucleation and for the varied
growth rates of several types of particlesthe products of trans-
formation [20]. With regards to the modeling of the growth in a
two-dimensional space, with the growth velocities distributed in
an even manner, Andrienko et al. have used a numerical model
[20]. The study emphasizes the lack of any analytical solution
for two-dimensional tasks.
4. General solution for screening effect
A solution allowing for an arbitrary distribution of growth
velocities on the grain surface was proposed in [21]. To obtain
this solution, function S was introduced, where the value of S(u)
stands for an area of the unscreened extended surfaces of
grains per unit volume, on which the growth is proceeding at
a rate not higher than u. If there is a derivative to this function
S

(u) =
dS(u)
du
(13)
it may be used for calculating the surface size and the rate of
boundary migration on which it is located in a range from u to
u +du:
dS(u) = S

(u)du (14)
and, nally, by means of integration, for determinating of a value
of the function itself.
S(u) =
_
u
u
min
S

(u)du (15)
A local increase of the extended volume on the surface dS(u)
in time lapse is:
(u) = uS

(u)du (16)
By means of function S one can determine the extended
volume with respect to time :
=
_

0
__
u
max
u
min
uS

(u)du
_
dt (17)
As a result of the screening effect, some extended surfaces of
the growing grains are becoming the surfaces of the phantoms.
Based on the determined probability of this effect and the related
rate of changes inanarea of the growingextendedsurface, in[21]
an instantaneous screening velocity was calculated. Its measure
is the rate of changes in the natural logarithm of an area of
the unscreened extended surface of the growing grain. For the
growth taking place in a three- two-, and one-dimensional space,
the screening velocity amounts to, respectively [21]:
ln S

(u)

=
1
4u
_
u
u
min
(u u
1
)
2
S

(u
1
)du
1
(18)
ln S

(u)

=
u

_
u
u
min
_
_
1
_
u
1
u
_
2

u
1
u
arccos
u
1
u
_
S

(u
1
)du
1
(19)
ln S

(u)

=
_
u
u
min
(u u
1
)S

(u
1
)du
1
(20)
In the case of non-Kolmogorov growth, modeling of transfor-
mation kinetics requires the application of a system of Eqs. (2)
and (17) and of one of the Eqs. (18)(20). A simplest example
of this growth can be the solidication of eutectic grains (grains
of type 2) in hypoeutectic alloy, which is discussed later in this
text, when at the instant of the onset of eutectic solidication
a large volume of the liquid has already been occupied by the
grains of the primary phase (grains of type 1) [22]. The same
mathematical operation is applicable in the description of phase
transformations of porous materials, providing the walls of a
pore are assumed to act as a screening surface.
5. Partial solution for sequential transformation
Let us assume that the volume content of the grains of type
1 at the instant of the onset of eutectic solidication is f
1
, and
the common area of their surfaces per a unit volume of material
at the same instant equals S
1
. The linear growth velocity on the
facets of the grains of type 1, which will be responsible for the
screening effect exerted on the growth of the grains of type 2,
A.A. Burbelko et al. / Materials Science and Engineering A 413414 (2005) 429434 433
is in this particular case equal to zero. The extended volume of
these grains can be determined from Eq. (2):

1
= ln(1 f
1
) (21)
We shall assume that grains of type 2 appear with uniform
density in quantity N
2
in a unit volume simultaneously at the
time instant taken as equal to zero, and are growing next in a
spherical form at a constant rate u
2
. For such grains, growing
from a parent phase whose area has been partially occupied by
grains of type 1, Eq. (18) is transformed to the following form
ln S

(u
2
)

=
u
2
4
_
u
2
0
S

1
du
1
(22)
Using this equation, we can calculate the unscreened surface
of the extended volume of these grains
S
2
= 4N
2
u
2
2

2
exp
_

u
2
S
1

4
_
(23)
as well as a value of the extended volume itself:

2
= 4N
2
u
3
2
_

0
t
2
exp
_

u
2
S
1
t
4
_
dt (24)
Let us introduce further the following designations:
a = 4N
2
u
3
2
b =
u
2
S
1
4
(25)
By integrating (24) with respect to the above designations for
b >0 we obtain:

2
= ab
3
(2 (2 +2b +b
2

2
)e
b
) (26)
The growth velocity of the extended volume of the grains of
type 2 with their instantaneous nucleation under these conditions
is
d
2
d
= a
2
exp(b) (27)
Because of a form that this term assumes, and considering
the effect of coefcients a and b on the rate of changes in the
extended volume of the newgrains
2
, one can apply the follow-
ing designations: for a: growth coefcient, and for b: screening
coefcient.
For the grains of both types
f
1
+f
2
= 1 exp(
1

2
) (28)
allowing for (21) we obtain:
f
2
= (1 f
1
)(1 exp(
2
)) (29)
Having substituted in this formula the value of an extended
volume of the grains according to (26), we obtain an analytical
formula with solution for grains of type 2 in the case of their
instantaneous nucleation in presence of the grains of type 1:
f
2
= (1 f
1
)(1 exp(ab
3
(2 (2 +2b +b
2

2
)e
b
)))
(30)
The increase of volume of the type 2 grains due to the screen-
ing effect proceeds more slowly than it would result from the
Fig. 3. Screening coefcient value vs. change in volume content of the type 2
grains in the case of instantaneous nucleation: numbers in the diagram are the
values of the screening coefcient b, s
1
; bold line (K) is the growth without
the screening effect (classic Kolmogorov equation).
classic Kolmogorov equation, as shown in Fig. 3. In process-
ing the data shown in this drawing, it has been assumed that
the volume density of the spherical grains of type 1 amounts to
50%, while their quantity N
2
equals 10
6
cm
3
. The individual
experimental calculations were made for the growth coefcient
a =1 s
3
at different values of the screening coefcient. The par-
ticular values of the screeningcoefcient, of the blocking surface
and of the linear velocity of migration of the grain boundary for
grains of the second type, adopted in calculations, were deter-
mined from a denition given in (25)
u
2
=
3
_
a
4N
2
4.3 10
5
m/s (31)
S
1
=
4b
u
2
(32)
By denition (25), the growth coefcient value a is dependent
on properties of growing (type 2) grains only: the grain density
and the growth velocity. Screening coefcient b depends on the
velocity of growth of type 2 grains and on the screening grain
boundaries S
1
. An increase in the density of screening grain
boundaries S
1
while decreasing the average size of type 1 grains
increases the screening coefcient. The b value is not affected
however by the volume fraction of the screening grains. The
respective values are compiled in Table 1.
It follows from Fig. 3, in the case of an instantaneous nucle-
ation, that the screeningsurface of the grains of type 1maytotally
Table 1
Growth parameters
Variant of calculation Screening coef-
cient, b (s
1
)
Screening surface,
S
1
(cm
1
)
1 2.0 1860
2 1.5 1395
3 1.0 930
4 0.5 465
5 0.1 93
434 A.A. Burbelko et al. / Materials Science and Engineering A 413414 (2005) 429434
arrest the growth of the grains of type 2. The complete transfor-
mation and decay of the parent phase due to this mechanism of
growth is, under these conditions, impossible. For example, for
b =1 s
1
, about 6.5%of the untransformed parent phase remains
in the system.
The reason of the parent phase remaining is that the untrans-
formed volume was fully screened by type 1 grains from type 2
growing grains. The nucleation of new type 2 grains or growth
recommencement of the type 1 grains is necessary for the com-
pletion of solidication. The analytical solution (30) does not
take into account this transformation mechanism changing.
6. Conclusions
If Kolmogorovs postulates P1P5 are satised, the nucle-
ation rate is constant, and the linear velocity of the grain growth
is constant as well, the kinetics of phase transformations can
be described by the K-J-M-A Eq. (12), which is a particular
variation of Kolmogorovs general equation.
If the nucleation rate and/or the linear velocity of the grain
growth change during the transformation, the kinetics of this
phase transformation can be described by Kolmogorov general
Eq. (9).
In the case of non-Kolmogorov growth, modeling of trans-
formation kinetics require the application of a system of Eqs.
(2) and (17) and of one of the Eqs. (18)(20).
Acknowledgments
This paper was supported by the State Committee of Scien-
tic Research KBN (Project 4 T08A 041 25).
References
[1] M. Avrami, J. Chem. Phys. 7 (1939) 11031112.
[2] R.A. Clemente, A.M. Saleh, Phys. Rev. B, 65 (2002) art. no. 132102.
[3] S. Celotto, T.J. Bastow, Acta Mater. 49 (2001) 4151.
[4] M.J. Starink, J. Mater. Sci. 32 (1997) 40614070.
[5] M. Avrami, J. Chem. Phys. 8 (1940) 212224.
[6] A.N. Kolmogorov, Bull. Acad. Sci. USSR. Ser. Math. 3 (1937) 355359
(in Russian).
[7] W.A. Johnson, R.F. Mehl, Trans. Metall. Soc. AIME 135 (1939)
416442.
[8] M. Avrami, J. Chem. Phys., 9. (1941) 177184.
[9] M.C. Weinberg, D.P. Birnie, J. Non-Cryst. Solids 202 (3) (1996)
290296.
[10] M. Lusk, H.J. Jou, Metall. Mater. Trans. A 28 (2) (1997) 287291.
[11] R.A. Ramos, P.A. Rikvold, M.A. Novotny, Phys. Rev. B 59 (14) (1999)
90539069.
[12] M. Castro, F. Dominguez-Adame, A. Sanchez, T. Rodriguez, Appl. Phys.
Lett. 75 (15) (1999) 22052207.
[13] M. Castro, A. Sanchez, F. Dominguez-Adame, Phys. Rev. B 61 (10)
(2000) 65796586.
[14] A. Almansour, K. Matsugi, T. Hatayama, O. Yanagisawa, Mater. Trans.
Jim 37 (10) (1996) 15951601.
[15] E. Pineda, T. Pradell, D. Crespo, N. Clavaguera, M.T. Clavaguera-Mora,
J. Non-Cryst. Solids 287 (13) (2001) 9295.
[16] M.J. Starink, J. Mater. Sci. 36 (18) (2001) 44334441.
[17] E. Woldt, Metall. Mater. Trans. A 32 (10) (2001) 24652473.
[18] E. Pineda, T. Pradell, D. Crespo, Philos. Mag. A 82 (1) (2002) 107121.
[19] V.Z. Belenkiy, Geometrical Probabilistic Models of Crystallization. Phe-
nomenological Method. Nauka, Moscow, USSR, 1980, pp. 188 (in
Russian).
[20] Y.A. Andrienko, N.V. Brilliantov, P.L. Krapivsky, Phys. Rev. A 45 (1992)
22632269.
[21] A. Burbelko, Informatyka w technologii materia ow 2 (4) (2002)
106120 (in Polish).
[22] A. Burbelko, Geometrical-probabilistic model of grain growth (in prepa-
ration).

Вам также может понравиться