a

r
X
i
v
:
n
l
i
n
/
0
2
0
2
0
2
0
v
2


[
n
l
i
n
.
A
O
]


1
9

F
e
b

2
0
0
2
Quasi-stationary Stefan problem and computer simulation of
interface dynamics.
R. Andrushkiw, V. Gaychuk, and R. Zabrodsky
Abstract. The computer simulation of quasistationary Stefan problem has
been realized. Dierent representations of the Laplacian growth model are
considered. The main attention has been paid for the interface dynamics
represented by integro dierential equations. Numerical approach has been
realized by use of interpolating polynomials and exact quadrature formulae.
As a result system of ordinary dierential equations has been simulated.
There are various physical processes in nature, such as crystallization, combus-
tion, deposition etc., which make it is possible to single out moving boundary having
physical properties dierent on its either side. The location of the given boundary is
not known beforehand and is determined self-consistently by the distribution of its
physical parameters in the ambient space. Such problems as usually are singled out
as a separate section of modern mathematical physics, where the problems with the
moving boundaries, also known as Stefan problems, are considered. The number of
publications, devoted to this problem, is rather great. Therefore, we will mention
only some monographs [1, 2] including a lot of bibliographies on mathematical
models of various physical systems containing moving boundaries.
Such problems, as usual, have no analytical solution due to the essential non-
linearity stipulated by representation of boundary conditions for selfconsistent pa-
rameters allocated on the free boundary. Though it is sometimes possible to nd an
analytically self-similar solution to a Stefan problem in one-dimensional or quasi-
one-dimensional case, concerning the two-dimensional case such problems can be
solved using computer simulation only. Due to the fact that problems with the mov-
ing boundaries describe a wide class of important technological processes, such as
crystallization, melting, etching, oxidation, deposition, diusion etc., development
of numerical methods of their solution is especially important [1, 2].
1. Model
Let us consider crystallization of an undercooled melt located in innite con-
tainer, with initial temperature T < T
c
, where T
c
is the temperature of crystalliza-
tion [3]. It is supposed that growth of the interface of the solid and liquid phases
is controlled by the extract of heat generated in the process of crystallization inside
the melt (Fig. 1). Then the velocity v
n
of the motion of the crystallization front
along the normal vector n to the interface is given by the expression:
Key words and phrases. Laplacian growth, free boundary, dissipative structures, instability.
1
2 R. ANDRUSHKIW, V. GAFIYCHUK, AND R. ZABRODSKY
Figure 1. A schematic view of the interface during crystallization.
v
n
=

L
(T|

n). (1.1)
In the melt the distribution of temperature T is determined by the heat equation:
c

T
t
= T. (1.2)
Here c

and are the specic heat and density of the liquid phase; is thermal
conductivity; L is the latent heat of crystallization. The temperature T
s
at the
crystallization front usually satises the Gibbs-Thomson condition:
T
s
= T
c
(1 Kd
0
), (1.3)
where K is the local curvature of the surface; d
0
is the capillary length proportional
to the surface energy.
It is known that in systems described by equations (1.1) (1.3) regular dynam-
ics of the interface is unstable. The basic issues for understanding the processes
of dissipative structures formation during crystallization are presented in [4, 5],
clarifying the reason of formation of these structures for instance the morpholog-
ical instability. In case of the undercooled melt the mechanism of this instability
is stipulated by magnication of the motion velocity of convex segments of the
surface penetrating into the undercooled region. The latter stimulates their faster
propagating into depth of the melt.
The realization of instability of the homogeneous state of an interface for crys-
tallization of the undercooled melt, within a wide interval of wave numbers and the
essential instability of the whole process of shaping, reduces itself to large varieties
of possible structures and next to the reproduction of the self-similar forms of the
interface in smaller scales. It also indicates the fractal character of their formation.
In spite of apparent simplicity of the problem statement, even in two-dimensional
domains it is often senseless to perform direct numerical calculations for solving the
problem.
In same cases the microscopic temporal scales of the parameters specifying the
problem changes can appear to be considerably smaller than the temporal scales
of dissipation and distribution of an appropriate unlocal eld. As a result the
distribution of the unlocal eld can be considered quasistationary subject to the
QUASI-STATIONARY STEFAN PROBLEM AND COMPUTER SIMULATION OF INTERFACE DYNAMICS. 3
location of the boundary. In this case one can construct the equations of surface
motion containing only undened elds specied on the considered surface. It is
the basic methodology for this approach.
The essence of the method devised in [3, 6] implies replacement of a consistent
system of equations of motion by means of more simple ones, while keeping all
the properties of the origin problem. As the velocity in the obtained equations is
dened locally at each point, we obtain a set of equations of local dynamics. Some
simplications, however, restricted application of the approach proposed. We can
see that the equations obtained in [3] are satised for undercooling, when the
characteristic scale of an unlocal eld (for example temperature) l
t
= (cv)
1
is much smaller than the radius of surface curvature, r i.e.: l
t
r. Although in
this case one can not obtain solutions without numerical simulation making use of
appropriate approximations allows to nd new solutions, for example those in the
form similar to snowake.
If l
t
r , then in the eld of spatial changes of the temperature it is possible
to allocate two characteristic subregions. The rst one, with characteristic scale,
comparable to l
t
, corresponds to the change of the temperature, due to the eect of
thermal conduction, when the curvature of the boundary element can be neglected.
The second subregion lies in proximity to the boundary element, and its thickness
is equal to the average curvature radius r. The change of the temperature and its
gradient in this subregion depends on curvature of the boundary. The modulation
of the surface temperature and its spatial distribution, as the rst approximation
in the parameter r/l
t
, corresponds to the Laplace equation. Since harmonic func-
tions as the solution of Laplace equation describing the distribution of temperature,
methods of complex analysis appear to be very eective. This book presents meth-
ods of conformal mappings for analysis of motion of boundary surface at least in
two-dimensional case. Thus, if l
t
r the problem of crystallization of undercooled
melt (1.1) (1.3) and the problems considered above are formally equivalent and
are typical representatives of the quasistationary Stefan problem. Making use of
methods of complex variables allows us to reduce the problem discussed to a system
of ordinary dierential equations [8, 9, 10].
In this case problem (1.1) (1.3) becomes one of the simplest among the ones
considered above and is a kind of aquasistationary Stefan problem. However, even
in this simple case, determination of the boundary is a rather complicated mathe-
matical problem.
Reformulate now the problem (1.1) (1.3) in terms of Laplacian growth prob-
lem. We investigate the growth of the two-dimensional region Q and suppose that
its boundary is , which represents the physical interface (Fig. 1). The eld
outside the region Q, satises the Laplace equation
= 0. (1.4)
The boundary grows at a rate, that is proportional to the normal gradient of the
eld of the interface. Therefore the evolution of the interface is governed by the
following equation

n
=
n
|

. (1.5)
The potential eld satises the following boundary condition on the interface
|

= d
0
k|

, (1.6)
4 R. ANDRUSHKIW, V. GAFIYCHUK, AND R. ZABRODSKY
where k is the curvature of the interface , and d
0
is the dimensionless surface
tension parameter.
For this type of system the utilization of the theory of function of complex
variable is very fruitful. Theoretical methods of the functions of complex variable
appear to be rather useful in the construction of the eective surface dynamics of
the interface at least in the two-dimensional problem. In this case the central issue
of this problem belongs to the conformal mappings of the physical regions of the
space onto some special kind of regions. A large number of models for interfacial
dynamics had been proposed up to now. Reviews of these problems can be found in
[11, 9, 12]. The most famous equation was introduced by Shraiman and Bensimon
[14]. It is widely known as the conformal mapping equation and is still under
investigation.
In examples of the quasistationary Stefan problem, the main point of its solution
lies in construction of an appropriate potential specied by its boundary value.
Physically it is possible to do only when the boundary of the domain of the potential
is rather simple. The simplest examples of such region is a disk. If we know a
conformal mapping of a disk on some region, the solution to the posed boundary
problem reduces to change of variables in the solution for the disk. Thus, solution
of the quasistationary Stefan problem for the two-dimensional region is practically
reduced to determination of an appropriate conformal mapping or its equation
of motion. It is possible to compare the non-stationary Stefan problem with an
equivalent quasistationary Stefan problem. It allows us to solve a wider class of
problems in framework of the equations of motion of the appropriate conformal
mappings.
Let us consider the approach developed by Shraiman and Bensimon [9, 12,
14, 8, 16]. This approach is based on the Riemann mapping principle. This
principle states the existence of a conformal map from the exterior of domain Q
onto the standard domain, for example, the exterior of unit disk, with the boundary
corresponding to unit circle. In the outline under regard one can parametrize
evolution of the interface in z plane by means of time-dependent conformal map
F (w, t) which takes the exterior of the unit disk at each moment of time, |w| 1,
onto the exterior of Q. Therefore the evolution of the interface can be presented as
follows:
(, t) = lim
we
i
F (w, t) , 0 < 2. (1.7)
As it is shown by Shraiman and Bensimon [14], in the case of free surface tension
(d
0
= 0) the boundary map satises the following equation of motion
(, t)
t
= i
(, t)

S
_

(, t)

2
_

w=e
i
, (1.8)
where S [] means the Schwartz operator [13]. For the exterior of a unit circle with
some boundary condition f () the Schwartz operator can be presented as follows:
S [f ()] =
1
2
2
_
0
d

e
i

+ w
e
i

w
f
_

_
+ iC, (1.9)
where C is an arbitrary constant.
QUASI-STATIONARY STEFAN PROBLEM AND COMPUTER SIMULATION OF INTERFACE DYNAMICS. 5
It was shown[15] that equation (1.8) is equivalent to equation of motion of
certain eld (, t) =

(,t)

:
(, t)
t
=
1
R(t)
(, t)

1
2
2
_
0
d

v
n
(

, t)(

, t) cot

(, t)
R(t)
_
_
(, t)v
n
(, t)
1
2
2
_
0
(

, t)v
n
(

, t)d

_
_
+
+

2
(, t)v
n
(, t)
R(t)

1
2
2
_
0
d

ln (

, t) cot

(, t)
R(t)

1
2
2
_
0
d

v
n
(

, t)(

, t) cot

2
(1.10)
considered in [12, 16, 17]. Here the radius R and normal velocity v
n
satisfy to
the next equations:
dR
dt
=
1
2
2
_
0
d

, t)v
n
(

, t). (1.11)
v
n
(, , t) = ()
(, t)
R(t)
_
_
_

s
R(t)

1
2
2
_
0
d

cot

2
(

, t)d
0
(

, )

_
_
1
2
2
_
0
d

cot

2
ln (

, t)
_
_
+
+

s
R(t)

1
2
2
_
0
d

cot

2
(

, t)d
0
(

, )
_
_
_
. (1.12)
In local approximation last relationship has the form [12, 16, 17]:
v
n
(, , t) =
(, t)
R(t)
_
1
1
R(t)

2
_
. (1.13)
The equations (1.10),(1.11),(1.13) are one-dimensional equations determined
on the interval [0, 2].
Now let us consider the linear stage of stability of a cylindrical boundary.
In this case it is natural to put d
0
(, ) = d
0
= const, (, ) =
0
= const
. Linearizing equations (1.10), (1.11), (1.13) with respect to perturbations =

0
exp(
t
_
0
(t

) dt

+ ik) at k = 1, 2, ... and taking into account the relation

cot

2
e
ik

= i e
ik
signk
6 R. ANDRUSHKIW, V. GAFIYCHUK, AND R. ZABRODSKY
we obtain the following characteristic equation
=

R
_
k 2 +
d
0
R
k
2
(1 k)
_
. (1.14)
It can be seen from equation (1.14) that the instability is realized for rather
short-wave uctuations with k > 2 [5]. At d
0
= 0 the uctuations with wave
number
k
0
=
1
3
_
_

1 + 3
_
d
0

s
R
0
_
1
1
_
_
(1.15)
have the maximum increment of increase. The instability as d
0
0 is observed
within a wide spectrum of wave vectors, which causes complicated fractal structure
of dendritic formations. The surface energy damps the instability subject to chosen
wave vectors and gives rise to the minimum radius of curvature of a dendrite.
2. Numerical simulation of the interface dynamics
In the present section numerical simulation of the surface dynamics is given.
We consider the properties of the interface dynamics from the viewpoint of singular
integral equations.
2.1. Properties of the interface dynamics. Equations (1.10) and (1.11)
are not directly related to the physical mechanism determining motion of the in-
terface in real space. They determine the variation law governing the mapping of a
circle into the contour under the given normal velocity v
n
. In framework of this
approach we consider properties of equations (1.10), (1.11).
We will investigate the periodical structure formation, supposing that the struc-
ture will have period 2/k, where k is an integer. This make it possible to consider
the solution only on the interval [0, 2/k]. In fact, we shall consider some integrable
periodic function

f( + n2/k) =

f() (2.1)
In this case the any singular integrals J in the equation (1.10) will look like:
1
2
2
_
0
d

f(

) cot

2
=
k1

n=0
2/k
_
0
d

f(

)
_
cot

+ n2/k
2
_
= (2.2)
=
2/k
_
0
d

f(

)(

)du

(2.3)
where
(u

u) =
k1

n=0
cot

+ n2/k
2
.
So, we can use for computer simulation the interval [0,2/k].
QUASI-STATIONARY STEFAN PROBLEM AND COMPUTER SIMULATION OF INTERFACE DYNAMICS. 7
2.2. Numerical solution of the interface dynamics. The derived sys-
tems of nonlinear integro-dierential equations were investigated numerically. The
numeric simulation was based on approximation of unknown functions by trigono-
metrical polynomials. Expansions of this type make it possible to use well-known
quadrature formulae to evaluate singular integrals. As a result, the system of equa-
tions (1.10), (1.11) and (1.13) reduces to a system of ordinary non-linear dierential
equations with lled Jacobian.
During computing experiments the approach based on approximations of searched
functions by trigonometrical polynomials [18, 19]

f
n
(u) =
1
2
2n1

j=0

f(u
j
) sin[n(u u
j
)] cot
u u
j
2
(2.4)
with the given nodes set
u
j
=
j
n
, j = 0, 1, ..., 2n 1. (2.5)
was used. The representation of the function

f (u) in form (2.4) makes it possible
to apply the obtained analytical expressions practically for all derivatives of the
function

f during calculations.
Besides, expansion of the function

f (u) into a trigonometrical polynomial
makes it possible to use quadrature formulae [18, 19] for calculation of singular
integrals with Hilbert kernels, that is
J(u
0
m
) =
2n1

j=0
1
2n

f(u
j
) cot
u
j
u
0
m
2
(2.6)
being dened on the system of nodes
u
0
m
=
2m + 1
2
, m = 0, 1, ..., 2n 1. (2.7)
Taking now into account that all integrand functions on the right hand side of
equations (1.10), (1.12) are of the same kind it is possible to present them in the
form of (2.6) making the following assignments

m
(t, u) = (t, u
0
m
), m = 0, 1, ..., 2n 1. (2.8)
As a result from the equations of interface dynamics we obtain the system of ordi-
nary dierential equations:
d
m
dt
=

F
m
{
_
_
_

m
,
2n1

j=0
f
j
cot
u
j
u
0
m
2
_
_
_
, (2.9)
where

F
m
is a nonlinear operator being determined by the right hand side of equa-
tions (1.10), (1.11) (1.12), f
j
can take values v
n
(u
j
)(u
j
), ln (u
j
), (u
j
) and etc.
Numerical integration of specied systems was carried by means of discrete
calculation of the Jacobian when realizing implicit methods of Girr and Adams
8 R. ANDRUSHKIW, V. GAFIYCHUK, AND R. ZABRODSKY
Figure 2. Non-uniform distribution of the (, t), (, t) and
the corresponding interfaces for the following initial perturbation:
(, t
0
= 0) = 1.0 0.01 cos(4); d
0
= 0.05; interfaces on the gure
correspond to the time steps (t
step
= 0.510
4
) 18 times are plotted,
spaced between 0 < t 0.8 10
4
.
[20]. Choice of the value of discretization and the local error of integration was
made so that the necessary accuracy of the searched solution and the stability of
calculations were ensured. Making use of variable
_
u
0
m
_
changing in time, subject
to the transformations
dy =
R(t)d
(, t)
sin
_
+ () +

2
_
,
dx =
R(t)d
(, t)
cos
_
+ () +

2
_
(2.10)
for the system of equations (1.10), (1.11) and (1.13) it is possible to reproduce
geometrically dened interfaces in parametric form.
In this text we evidently do not claim the complete investigation of pattern
formation in free boundary problems. Our goal is to show that developed approach
can be applied to investigation of interface evolution. Results of computer simula-
tion presented below give rise only to some sketch of the structures realizing in the
systems under consideration.
During numerical simulation evolution of the interface was investigated. The
product term aect only on time scale of the process. Therefore, in the analysis
under regard = 1 is taken. It was found out that the surface energy inuences the
behavior of the eld (, t) and interfaces too much. At small values of the surface
energy rather small-sized structures can be formed. As d
0
increases the structures
take a more smooth form.
Figure 2 illustrates the characteristic form of non-uniform distributions of the
eld (plot a) and the corresponding interfaces (plot b).
It is possible to observe that relatively smooth variations of the interface cor-
respond to rather great variations in the eld as well as in the angle . Thus,
relatively small errors in determining do not cause great changes in the interface
geometry. In addition, as is seen from Figures 2 relatively even long sections of
the interface shrink when mapped onto the interval [0, 2], whereas sharp changes
in the contour , conversely, are stretched out. In this sense such mappings are
QUASI-STATIONARY STEFAN PROBLEM AND COMPUTER SIMULATION OF INTERFACE DYNAMICS. 9
Figure 3. Comparison of interface proles for dierent values of
surface tention d
0
: d
0
= 0.15a); d
0
= 0.12b); d
0
= 0.09c);
d
0
= 0.06d); d
0
= 0.04e); d
0
= 0.03f). Initial pertur-
bations for all the gures are the same: (, t
0
= 0) = 1.0
0.01 cos(4); (t
step
= 0.5 10
4
) 18 times are plotted, spaced be-
tween 0 < t 0.8 10
4
.
adaptive with respect to the information about the detailed geometric structure
of the interface.
The characteristic geometry of the structures obtained as a result of computer
simulation of a small perturbation is presented on Figures 3, and 4. The smaller
capillary length is the more diverse structure arises. Figures 3, and 4 are dierent
from each other due to periodicity of the initial perturbation.
If the system is inuenced by an uncorrelated noise with the normal law of
distribution unregular structures arise. During simulation it was taken that on
each step of integration the velocity at point i of the interval [0, 2] equal to
v
n
(
i
) +(/R)
1/2
v
n
(
i
), within v
n
(
i
) being chosen in the correspondence with
the normal law of distribution in limits
_
10
5
, 10
5
_
(Fig. 5).
For simulation of anisotropic growth of crystals the following expression (similar
as in [21]) was used:
d
0
() = (1 15
4
cos 4)d
0
=
=
_
_
1 15
4
cos 4
_
_

1
2
2
_
0
d

ln (

) cot

2
_
_
_
_
d
0
. (2.11)
The interfaces corresponding to the capillary length (2.11), subject to the
anisotropy of surface energy have the form of truly looking cross (Fig. 6, and
7), despite that at the initial moment the boundary surface presents a circle (up to
10 R. ANDRUSHKIW, V. GAFIYCHUK, AND R. ZABRODSKY
Figure 4. Comparison of interface proles for dierent values of
surface tention d
0
: d
0
= 0.15 a); d
0
= 0.12-b); d
0
= 0.09c);
d
0
= 0.06d); d
0
= 0.04e); d
0
= 0.03f). Initial perturbation for
all the gures is the same: (, t
0
= 0) = 1.00.01 cos(6); (t
step
=
0.5 10
4
) 18 times are plotted, spaced between 0 < t 0.8 10
4
.
Figure 5. Interface dynamics stimulated by external noise.
v
n
|
t=0

= 10
5
). According to the value of
4
such an anisotropy is exhibited at
small or large values of R and t. The realized distributions little depend on initial
perturbations and reduce frequently to rather complicated and exotic surfaces. A
more developed structure in the case of anisotropy is presented by Fig. 8. An
example of structures arising from nonhomogeneous initial conditions is illustrated
by Fig. 9, a). In the same time Fig. 9, b) illustrates an example of structures
developing from a at boundary.
It should be noted that at d
0
= 0 the results concerning smooth boundary
under stochastic excitation of short-wave perturbations, caused for example by
errors of calculations do not lead the developed structures due to instabilities of
computation.
References
QUASI-STATIONARY STEFAN PROBLEM AND COMPUTER SIMULATION OF INTERFACE DYNAMICS. 11
Figure 6. Non-uniform distribution of the (, t), (, t) and the
dendrite silhouette for surface tension anizotropy: (, t
0
= 0). =
1 0.01 cos(4); d
0
= 0.3;
4
= 0.05 interfaces on the gure cor-
respond to the time steps(t
step
= 0.25 10
4
) 15 times are plotted,
spaced between 0 < t 0.325 10
5
.
Figure 7. Comparison of interfacee proles for dierent values
of surface tension anizotropy: d
0
= 0.3; a)
4
=0.00; b)
4
=0.01 c);

4
=0.02; d)
4
=0.04. Initial perturbation for all the gures is the
same: (, t
0
= 0) = 1. 0.01 cos(4); (t
step
= 0.25 10
4
) 15 times
are plotted, spaced between 0 < t 0.325 10
5
.
[1] Moving Boundaries VI Computational Modelling of Free and Moving Boundary Problems,
Eds: B. Sarler, C.A. Brebbia, Series: Computational and Experimental Methods Vol : 4,
WIT Press, (2001).
[2] Free Boundary Problems: Theory and Applications Ed. by I. Athanasopoulos, (Chapman &
Hall /CRC Research Notes in Mathematics, 409) Hardcover (1999).
[3] E. Ben-Jacob , N. Goldenfeld , L. S.Langer, G. Schon, Dynamics of Interfacial Pattern for-
mation. Phys. Rev. Lett., 51, #21, p. 1930, (1983).
12 R. ANDRUSHKIW, V. GAFIYCHUK, AND R. ZABRODSKY
Figure 8. Dendrite silhoette for surface tension anizotropy: =
0.03, d
0
= 0.03, (, t
0
= 0) = 1 0.01 exp
_
/4
0.1
_
2
, (t
step
=
0.2 10
4
) 21 times are plotted, spaced between 0 < t 0.36 10
4
.
Figure 9. Dendrite silhoette for surface tension anizotropy: =
0.00, d
0
= 0.03, (, t
0
= 0) = 10.01 exp
_
/4
0.1
_
2
, (t
step
= 0.1
10
4
) 13 times are plotted, spaced between 0 < t 0.12 10
4
. a).
The evolution of the at boundary with small initial perturbation
(t = 0) = 1 0.01 cos u b).
[4] W. W. Mullins and R. F. Sekerka, Stability of a planar interface during solidication of a
dilute binary alloy. J. Appl. Phys., 35, 444, (1964).
[5] W. W. Mullins and R. F. Sekerka, Morphological stability of a Particle growing by diusion
or heat ow. J. Appl. Phys. v.34, 323, (1963).
[6] E. Ben-Jacob , N. Goldenfeld, J. S.Langer, G. Schon, String models of interfacial pattern
formation., Physica D., 12, #1/3, p. 245, (1984).
[7] E.Ben-Jacob, N.Goldenfeld, J. S.Langer, G.Schon, Pattern selection in Dendritic Solidica-
tion. Ibid. 53, #22. p. 2110, (1984).
[8] D. Bensimon Stability of viscous ngering, Phys. Rev. A, 33, #2, p.1302, (1986).
[9] D.Bensimon, L.P.Kadanojf, S.Liana, B. J.Shraiman, Chao Tang, Viscous ows in two dimen-
sions. Rev. of Modern Phys. 58, #4, p. 977, (1986).
[10] D.Bensimon, P.Peice, Tip-splitting solutions to a Stefan problem. Phys. Rev., 33, #6, p.
4471, (1986).
QUASI-STATIONARY STEFAN PROBLEM AND COMPUTER SIMULATION OF INTERFACE DYNAMICS. 13
[11] Dynamics of Curved Fronts, ed. by P.Pecle, Academic Press, San Diego (1988).
[12] V.V.Gaychuk, I.A.Lubashevskii, V.V.Osipov. Dynamics of surface structures in systems
with a free boundary. Naukova Dumka, Kiev, 1990. 214 pp. ISBN5-12-001274-4.
[13] H. Begehr, R. Gilbert, Transformations, transmutations, and kernel functions, 1, Longman
Publishers, Harlow, 399 pp., (1992).
[14] A. Shraiman, D. Bensiman, Singularities in nonlocal interface dynamics, Phys. Rev. A., 30,
# 5, 2840, (1984).
[15] V.Gaychuk, A.Shnyr, B.Datsko, Interface dynamics equations: their properties and com-
puter simulation, http://xxx.lanl.gov/nlin.AO/0011021 (2000).
[16] V.V.Gaychuk, I.A.Lubashevskii. Non-local dynamics of dendritic growth. - Zh. Khim. Fiz.
(USSR) v.7, No.9, p.852 - 859 (1988) (in Russian).
[17] V.V.Gaychuk, I.A.Lubashevskii. Analysis of the dynamics of phase interfaces in a quasi-
stationary Stefan problem. Zh. Vychisl. Mat. Mat. Fiz. (USSR) - v.29, No.9, p.1331-1345
(1989) (English translation in Compute. Math. Math. Phys. (UK)).
[18] S.M.Belozerkovsky, I.K.Lifanov, Numerical methods in the singular and integral equations.
Nauka, (1985). (in Russian).
[19] V.V.Panasyuk, M.P.Savruk, Z.T.Nazarchuk. Method of singular integral eguations in two-
dimensional diraction problem. Naukova Dumka, Kiev, 1984 (in Russian).
[20] Byrne J.,Hindmarch D.C. Sti ODE Solver: A review of Current and Coming Attractions //
J. Comput. Phys.- 1987.-70, N l.-P. 1-60.
[21] P.W. Vorhees, S.R. Coriell, G.A. McFadden, The eect of anisotropic crystal-melt surface
Tension on grain boundary groove morphology, J. Cryst. Growth. 67, 425, (1984).
New Jersey Institute of Technology University Heights Newark, NJ 07102-1982
E-mail address: roandr@megahertz.njit.edu
URL: http://eies.njit.edu/~andrushk/
Institute of Applied Problems of Mechanics and Mathematics, National Academy
of Sciences of Ukraine, Lviv
E-mail address: viva@iapmm.lviv.ua
Institute of Applied Problems of Mechanics and Mathematics, National Academy
of Sciences of Ukraine, Lviv
E-mail address: viva@iapmm.lviv.ua
This figure "Stochast.gif" is available in "gif"format from:
http://lanl.arxiv.org/ps/nlin/0202020v2
This figure "ad1.gif" is available in "gif"format from:
http://lanl.arxiv.org/ps/nlin/0202020v2
This figure "rus1.gif" is available in "gif"format from:
http://lanl.arxiv.org/ps/nlin/0202020v2
This figure "rus2.gif" is available in "gif"format from:
http://lanl.arxiv.org/ps/nlin/0202020v2
This figure "rus3.gif" is available in "gif"format from:
http://lanl.arxiv.org/ps/nlin/0202020v2
This figure "rus4an.gif" is available in "gif"format from:
http://lanl.arxiv.org/ps/nlin/0202020v2
This figure "izotrop_group_6.gif" is available in "gif"format from:
http://lanl.arxiv.org/ps/nlin/0202020v2
This figure "rus7.gif" is available in "gif"format from:
http://lanl.arxiv.org/ps/nlin/0202020v2
This figure "F11n.gif" is available in "gif"format from:
http://lanl.arxiv.org/ps/nlin/0202020v2