(This is a sample cover image for this issue. The actual cover is not yet available at this time.

This article appeared in a journal published by Elsevier. The attached

copy is furnished to the author for internal non-commercial research
and education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elseviers archiving and manuscript policies are
encouraged to visit:
http://www.elsevier.com/copyright

Author's personal copy

Materials Letters 89 (2012) 191194

Contents lists available at SciVerse ScienceDirect

Materials Letters
journal homepage: www.elsevier.com/locate/matlet

An analytical model for complete solute trapping during rapid solidication

of binary alloys
S.L. Sobolev
Institute of Problems of Chemical Physics, Academy of Sciences of Russia, Chernogolovka, Moscow Region 142432, Russia

a r t i c l e i n f o

abstract

Article history:
Received 14 June 2012
Accepted 15 August 2012
Available online 28 August 2012

An analytical model has been developed to describe solute partitioning during rapid solidication of
binary alloys under local nonequilibrium conditions. The model takes into account the deviations from
equilibrium both at the solidliquid interface according to the kinetic approach of Jackson et al. based
on Monte Carlo simulations and in the bulk liquid using local nonequilibrium diffusion model (LNDM).
The dimensionless growth parameter b found in the Monte Carlo simulations as the important
parameter for solute trapping has been modied for the local nonequilibrium diffusion case. An
analytical expression has been developed for the velocity-dependent partition coefcient KLNDM(V)
which predicts complete solute trapping and an abrupt transition from diffusion-controlled to
diffusionless solidication when the interface velocity V passes through the critical point V VD. The
predictions are in good agreement with the experimental results, molecular dynamic studies, and
Monte Carlo computer simulations.
& 2012 Elsevier B.V. All rights reserved.

Keywords:
Crystal growth
Diffusion
Metals and alloys
Segregation
Solidication

1. Introduction
Much of the current detailed understanding of rapid solidication of binary alloys under nonequilibrium conditions has been
developed based on Monte Carlo computer simulations, which
employ a detailed picture of the solidliquid interface [13], and
on local nonequilibrium diffusion model (LNDM) [47], which
takes into account deviation from local equilibrium of solute
diffusion in the bulk liquid. The most obvious manifestation of the
deviations from equilibrium during rapid solidication is solute
trapping which reduces the segregations at the solidliquid
interface. The degree of solute trapping is usually quantied by
the partition coefcient K, dened as the ratio of the concentration of solute in the solid to that in the liquid at the interface. The
effect of solute trapping has been investigated both theoretically
and experimentally [13,515]. The most outstanding question in
all these studies is transition to complete solute trapping K(V)1
at high interface velocity V. Conventional modes of solute trapping result in K(V)-1 only at V-N. LNDM predicts complete
solute trapping K 1 at a nite interface velocity [57]. The
molecular dynamics simulations [810] and the experimental
data [1115] conrm the possibility of reaching the complete
solute trapping at a nite interface velocity. The exact knowledge
of the K vs. V relation is of considerable practical importance
to model transition to diffusionless solidication, solute and
disorder trapping, grain renement, morphological stability and

E-mail address: SSL55@yandex.ru

0167-577X/$ - see front matter & 2012 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.matlet.2012.08.065

microstructure formation of commercial materials. Moreover,

rapid solidication of binary alloys is of fundamental interest to
physicists in the general context of the evolution of dynamic
systems far from local equilibrium. To address this issue, we will
incorporate the LNDM approach [47] into the Jackson et al.
model [13] to take into account deviations from equilibrium
both at the solidliquid interface and in the bulk liquid.

2. Jackson et al. model

The probability that an atom will be trapped into a crystal
depends on the ratio of the distance that the interface advances
during this time to the distance that the atom can diffuse during this
time [13]. The distance L which a solute atom can move by
diffusion during time t is L(Dt)1/2. The parameter L(Dt)1/2 is
usually dened as diffusion length and provides a measure of how
far the concentration has propagated in the X-direction by diffusion
in time t in the laboratory reference frame. It is a consequence of a
solution to parabolic diffusion equation (see Fig. 1 dashed line) that
is valid at a relatively low interface velocity. The ratio of these two
distances is the dimensionless parameter b [13]:

b Vt=Dt1=2 aVuK =D1=2

where uK is the normalized difference between the rates at which

atoms join and leave the crystal, and a is the cube root of the atomic
volume. The partition coefcient for small solute concentration
takes the form [13]
K J V K E 1=1 Ab

Author's personal copy

192

S.L. Sobolev / Materials Letters 89 (2012) 191194

3. Solute diffusion under local nonequilibrium conditions

1.0
At high interface velocity the solute diffusion in the bulk liquid
occurs under local nonequilibrium conditions and the solute
concentration C(X) in the moving reference frame is governed
by the hyperbolic diffusion equation [6]

concentration

0.8
Vd
0.6

@C
@2 C
@C
@2 C
@W
t 2 V
D 1V 2 =V 2D 2 W t
@t
@X
@t
@t
@X

0.4

where t is the relaxation time to local equilibrium, and W is the

kinetic rate (source) function. In the laboratory reference frame
the Fourier transformation (@/@t-io, @/@X-ih) of Eq. (3) gives the
dispersion law

0.2
0.0

ioto2 Dh

X distance

Fig. 1. Schematic concentration proles in the laboratory reference frame illustrating the diffusive distance. Solid lineconcentration prole calculated form
hyperbolic diffusion equation. Diffusive front moves with constant velocity VD and
diffusive distance is H VDt. Dashed lineconcentration prole calculated from
classical parabolic diffusion equation. Diffusive distance is L (Dt)1/2. At tct the
concentration proles coincide.

1.2

diffusion controlled
solidification V<Vd

diffusionless
solidification V>Vd

1.0

0.8
0.6

0.4

0.2

3
0.0

Vd
0.2

0.4

0.6

0.8

1.0

1.2

1.4

Nondimensional interface velocity

Fig. 2. Partition coefcient KJ, Eq. (2a), (curve 1); nondimensional effective
diffusion coefcient DLNDM/D, Eq. (6), (curve 2); local nonequilibrium boundary
layer dLNDM, Eq. (7), (curve 3); local equilibrium boundary layer (curve 4); and
partition coefcient KLNDM, Eq. (9), (curve 5) as functions of dimensionless
interface velocity j V/VD.

where KE is equilibrium partition coefcient and A is a constant. For

near equilibrium conditions when b-0 the partition coefcient
KJ(V)-KE. At the other extreme b-N the partition coefcient
KJ(V)-1. For a rough interface, the growth rate V is proportional
to uK [1,2], i.e. VVcuK, so that b is proportional to V:

b Va=V c D1=2
where Vc is the atomic diffusion speed VD in the WilsonFrenkel
expression for the diffusion controlled growth, or the thermal
velocity of the atoms for growth which is not thermally activated
[1,2]. Thus, KJ(V) takes the form
n

K J V K E 1=1 A

2a

where A*A(a/Vc D)1/2. Taking into account that VD D/a, we have

A*A(VcVD )  1/2, so that for diffusion controlled growth A*A/VD.
Eq. (2a) predicts complete solute trapping KJ(V)-1 as V-N (curve 1
in Fig. 2).

At the high-frequency limit o-N Eq. (4) predicts a nite

speed of the diffusive wave VD (D/t)1/2 i.e. maximum speed with
which the diffusion perturbations can propagate in the bulk liquid
[47]. For steady-state regimes, solution to Eq. (3) can be
expressed as
Z 1
1
1iujexpiuxZu
Cn x
du
5
2p 1
1j2 u2 iuj
R1
where Zu 1 oxexpiuxdx, j V/VD is nondimensional
interface velocity, Cn(x)(C(X) C0)/C0, and x XV/D. Eq. (5)
results in the solute concentration in the bulk liquid
C LNDM X Ci C 0 expVX=DLNDM V C0
(
D1V 2 =V 2D , V oV D
LNDM
D
V
0,
V 4V D

0.0

where DLNDM(V) is the effective diffusion coefcient [57]; C0 and

Ci are the solute concentration in the liquid far from (X-  N)
and at the interface (X 0), respectively. The effective diffusion
coefcient DLNDM is shown in Fig. 2 (curve 2) as a function of
nondimensional interface velocity j V/VD. For V 4VD the solute
concentration ahead of the interface C(X) C0 as DLNDM(V) 0. It
implies that when the interface velocity V passes through the
critical point V VD, an abrupt transition from diffusion controlled
to diffusionless solidication occurs. The result arises due to the
wave nature of solute diffusion under local nonequilibrium
conditions and has a clear physical meaning: a source of perturbations (i.e. the interface) moving with a velocity greater than the
maximum speed of perturbations cannot disturb the medium
ahead of itself. Expression (6) yields the solute boundary layer
(
D1V 2 =V 2D =V, V o V D
dLNDM V=V D
7
0,
V 4VD
The boundary layer dLNDM (curve 3 in Fig. 2) shrinks more
rapidly with increasing interface velocity, than expected from the
classical mass transport theory (curve 4 in Fig. 2) and equals zero
for V4VD.
Thus at high interface velocity V  VD, when solute diffusion
occurs under local nonequilibrium diffusion conditions, the solute
concentration has propagated in the X-direction of the laboratory
reference frame a distance HVDt (see Fig. 1 solid line). From the
microscopic point of view the diffusive speed VD can be treated as
the maximum speed with which an individual atom moves in
X-direction in the bulk liquid. In this situation the distance that a
solute atom can diffuse away from the interface in the moving
reference frame is (VD V)t. Hence, the probability that an atom
will be trapped into the crystal under local nonequilibrium
conditions, i.e. the ratio of the distance that the interface
advances during time t to the distance that the atom can diffuse

Author's personal copy

S.L. Sobolev / Materials Letters 89 (2012) 191194

193

during this time, takes the form

LNDM

V
V D V

-1,

V oV D

1.0
8

V 4VD

Substituting Eq. (8) in Eq. (2) gives the modied partition

coefcient
(
K E 1=1 AV=V D V , V oV D
9
K LNDM V
1,
V 4VD
Eq. (9) conrms the reasonable expectation that KLNDM-1 for
V-VD and KLNDM-KE as V-0 (see Fig. 2 curve 5).

KE = 0.4
VD = 4.8m/s
A = 2.9

0.9
Partition coefficient, K

bLNDM

A* = 0.8

0.8

Ge Si

0.7

4. Discussion and comparison with experiment

LNDM

Under local nonequilibrium diffusion conditions the modied

trapping probability bLNDM-N as V-VD. It implies complete
solute trapping KLNDM(V)1 at the nite interface velocity V VD
(see Eq. (9) and Fig. 2 curve 5). Molecular dynamic studies [810]
and the experimental data [11,12] have also shown the complete
solute trapping at a nite interface velocity. The velocity dependence of the partition coefcient was measured for rapid solidication of polycrystalline SiAs alloys induced by pulsed laser
melting [13,14]. The experimental results are compared with
predictions of the Jackson et al. model, Eq. (2a), and LNDM,
Eq. (9), on Fig. 3. The best t is obtained for VD 2.8 m/s, A 5.5,
and An 2.2 s/m. The K vs. V experimental data obtained by pulsed
laser melting of GeSi alloys [15] are shown in Fig. 4 together
with the predictions of both Jackson et al. model and LNDM with
VD 4.8 m/s, A2.9, and A* 0.8 s/m. The values of A and A* are in
good agreement with those of Monte Carlo simulations [2,3] and
theoretical prediction A* A/VD for diffusion controlled growth.
This implies that a best t is obtained with approximately the
same values of A for both models. The values of VD are also
consistent with those of other investigations [79,13,14].

Jackson et al.

0.6

Vd
1

Interface velocity, m/s

Fig. 4. Partition coefcient K as a function of interface velocity V for GeSi alloys
with KE 0.4, VD 4.8 m/s, A 2.9, A* 0.8 s/m.

The results presented in Figs. 3 and 4 show that at a relatively

low interface velocity the Jackson et al. model ts the data well
but the nonequilibrium partition coefcient KLNDM(V) ts experimental data and molecular dynamic results better both at high
and low interface velocity and predicts complete solute trapping
KLNDM(V)1 for V4VD.

5. Conclusion
The work extends the most recent theory for interface kinetics
due to Jackson et al. [13] by introducing a dimensionless
parameter (trapping probability) bLNDM that takes into account
the local nonequilibrium solute diffusion effects. It can be used for
Monte Carlo simulations of alloy solidication under local nonequilibrium conditions. The underlying physical picture for the
extension, based on insights gained from LNDM, is that the solute
concentration perturbations move with a nite velocity which
may be on the order of or even less than the interface velocity.
The model successfully describes the transition from diffusion
controlled to diffusionless solidication with complete solute
trapping for VZVD. The model can be used to study phase
transformations and microstructure evolution during the processing of commercial materials under nonequilibrium conditions.

1.0
KE = 0.3
VD = 2.8m/s

0.9

A = 5.5

Partition coefficient

A* = 2.2

0.8

0.7
Si As

References

LNDM
Jackson et al.

0.6

0.5
Vd
0.5

1.0

1.5

2.0

2.5

3.0

Interface velocity, m/s

Fig. 3. Partition coefcient K as a function of interface velocity V for SiAs alloys
with KE 0.3, VD 2.8 m/s, A 5.5, A* 2.2 s/m.

[1] Jackson KA. The interface kinetics of crystal growth processes. Interface Sci
2002;10:15969.
[2] Jackson KA, Beatty KM, Gudgel KA. An analytical model for non-equilibrium
segregation during crystallization. J Cryst Growth 2004;271:48194.
[3] Beatty KM, Jackson KA. Monte Carlo modeling of dopant segregation. J Cryst
Growth 2004;271:495512.
[4] Sobolev SL. Transport processes and travelling waves in systems with local
nonequilibrium. Sov Phys Usp 1991;34:2179.
[5] Sobolev SL. Local-nonequilibrium model for rapid solidication of undercooled melts. Phys Lett A 1995;199:3836.
[6] Sobolev SL. Rapid solidication under local nonequilibrium condition. Phys
Rev E 1997;55:684554.

Author's personal copy

194

S.L. Sobolev / Materials Letters 89 (2012) 191194

[7] Sobolev SL. Local non-equilibrium diffusion model for solute trapping during
rapid solidication. Acta Mater 2012;60:27118.
[8] Cook SJ, Clancy P. Impurity segregation in Lennard-Jones A/AB heterostructures. J Chem Phys 1993;99:217581.
[9] Yu Q, Clancy P. Molecular dynamic simulation of crystal growth in Si1  xGex/
Si(100) heterostructures. J Cryst Growth 1995;149:458.
[10] Yang Y, Humadi H, Buta D, Laird BB, Sun D, Hoyt JJ, et al. Atomistic
simulations of nonequilibrium crystal-growth kinetics from alloy melts. Phys
Rev Lett 2011;107:02550514.
[11] Eckler K, Cochrane RF, Herlach DM, Feuerbacher B, Jurisch M. Evidence for a
transition from diffusion-controlled to thermally controlled solidication in
metallic alloys. Phys Rev B 1992;45:501922.

[12] Willnecker R, Herlach DM, Feuerbacher B. Evidence of nonequilibrium

processes in rapid solidication of undercooled melts. Phys Rev Lett 1989;
62:270710.
[13] Kittl JA, Aziz MJ, Brunco DP, Thompson MO. Nonequilibrium partitioning
during rapid solidication of SiAs alloys. J Cryst Growth 1995;148:17282.
[14] Kittl JA, Sanders PG, Aziz MJ, Brunco DP, Thompson MO. Complete experimental test of kinetic model for rapid alloy solidication. Acta Mater
2000;48:4797811.
[15] Brunco DP, Thompson MO, Hoglung DE, Aziz MJ, Gossmann HJ. Germanium
partitioning in silicon during rapid solidication. J Appl Phys 1995;78:
157582.