Vous êtes sur la page 1sur 21

Physica D 136 (2000) 245265

Dynamics of kinks in two-dimensional hyperbolic models


Horacio G. Rotstein
a,
, Alexander A. Nepomnyashchy
a,b
a
Department of Mathematics, Technion IIT, Haifa 32000, Israel
b
Minerva Center for Nonlinear Physics of Complex Systems, Technion IIT, Haifa 32000, Israel
Received 4 January 1999; received in revised form 7 May 1999; accepted 25 June 1999
Communicated by F.H. Busse
Abstract
We study the motion of fronts for an extended version of the nonlinear wave equation,
2

t t
+
2

t
=
2
+f () +
h +
4

t
with positive 1 in cartesian and polar coordinates and give a local description of the front in terms of its
normal velocity, acceleration and curvature. We study analytically and numerically the motion of planar and circular fronts
and perturbations on them. 2000 Elsevier Science B.V. All rights reserved.
Keywords: Kink dynamics; Hyperbolic models; Front motion; BornInfeld equation
1. Introduction
In this paper we consider the following equation:

t t
+
2

t
=
2
+f () +h +
4

t
(1)
in a bounded region R
2
with smooth boundary for Neumann boundary conditions on , with , h and
constants, and non-negative, and f () a real odd function with positive maximum equal to

, negative
minimum equal to

and precisely three zeros into the closed interval [a, a] located at 0 and a for some
positive constant a. With f () = sin , Eq. (1) is a two-dimensional generalization of the standard model of a large
area Josephson junction [1,2]. For = 0, this equation was investigated formerly in [2]. With f () = (
3
)/2,
h = = 0, Eq. (1) is a damped version of the nonlinear KleinGordon equation; the later equation has been
analyzed in [3].
At the same time, Eq. (1) can be considered as a hyperbolic generalization of the AllenCahn equation,

t
=
2
+f (), (2)

Corresponding author. Tel.: +972-4-8293205; fax: +972-4-8324654.


E-mail addresses: horacio@techunix.technion.ac.il (H.G. Rotstein), nepom@math.technion.ac.il (A.A. Nepomnyashchy)
0167-2789/00/$ see front matter 2000 Elsevier Science B.V. All rights reserved.
PII: S0167- 2789( 99) 00160- 8
246 H.G. Rotstein, A.A. Nepomnyashchy / Physica D 136 (2000) 245265
which describes a phase transition dynamics process, where is a non-conserved order parameter. Indeed, let us
consider a physical system in a bounded region R
2
which can be in either of two possible states (phases).
These two phases are separated by an interface of nite thickness which will be dened more precisely later. The
boundary is assumed to be smooth. The free energy of the system is assumed to be of the form
F

() =
_

2
2
()
2
+F()
_
dx, (3)
where is an order parameter and F is a double well potential. The functional derivative of (3) is given by
F

=
2
f (), (4)
where f () = F

(). Eq. (4) may be considered as a generalized force indicative of the tendency of the free
energy to decay towards a minimum. In the case of a non-conserved order parameter one can generalize Eq. (2) by
assuming that (x) decreases with a delayed response (rather than a rate proportional) to the generalized force (4),
as suggested by Amy Novick-Cohen [4,5],

t
=
1

_
t
0
a(t t

)
F

(x, t

) dt

. (5)
Integrodifferential equations describing the evolution of a conserved order parameter (concentration) have been
used in the past [6] as a generalization of the linearized CahnHilliard equation for spinodal decomposition of
binary mixtures in the case where there is a coupling to a slowly relaxing variable, e.g., phase separation of polymer
mixtures near the glass transition.
Here a(t ) is a differentiable, scalar-valued function on (0, ) with a() = 0 that may be considered as a
generalized force relaxation function. The prototype example that we are going to consider here is
a(t ) = e
t
, (6)
where is a non-negative constant. Substituting (4) into (5) with =
2
, we obtain

t
=
_
t
0
a(t t

)[
2
+f ()](t

) dt

. (7)
We now substitute (6) into (7), differentiate with respect to t , rearrange terms and rescale by means of the transfor-
mation t
1/2
t and
1/2
. Eq. (7) becomes

t t
+
2

t
=
2
+f (). (8)
Note that for a(t ) = (t ) Eq. (7) reduces to (2).
For (2) Allen and Cahn [7] and Rubinstein et al. [8] showed that curved fronts in the plane move with normal
velocity proportional to their curvature (see also [9]):
S
t
=
S
xx
1 +S
2
x
, (9)
where y = S(x, t ) is the cartesian description of interface in the plane. For a circular interface, the curvature being
the reciprocal of the radius R, (9) becomes R
t
= 1/R(t ) whose solution satisfying R(0) = R
0
is given by
R(t ) =
_
R
2
0
2t ; i.e., circles shrink to a point conserving their original shapes at t = R
2
0
/2. For more general
shapes such an expression for the distance of every point in the curve from the origin is difcult to obtain, but
there are some analytical results showing that the behavior is similar. Gage and Hamilton [10] proved that (9)
H.G. Rotstein, A.A. Nepomnyashchy / Physica D 136 (2000) 245265 247
shrinks convex curves embedded in the plane R
2
to a point. They showed that such curves remain convex and
become asymptotically circular as they shrink. Grayson [11] extended this result to a more general case showing
that embedded curves become convex without developing singularities; i.e., curve shortening shrinks embedded
plane curves smoothly to points with round limiting shape.
For (8) with = 0 Neu [3] proved that the motion of the interface is governed by the BornInfeld equation
[12,13]:
(1 S
2
t
)S
xx
+2S
x
S
t
S
xt
(1 +S
2
x
)S
t t
= 0. (10)
He also showed that circular interfaces moving according to (10) shrank to a point in nite time and established
that a closed kink could be stabilized against a collapse by slight perturbations of circular interfaces.
In Section 2 we show by means of a formal, although non-rigorous but self-consistent, asymptotic analysis that
for (1) curved fronts move according to a generalized version of (10). In Section 3 we study the stability of small
perturbations both analytically and numerically. As a consequence of dissipation ( > 0) and difference in potential
between the two phases = 0, non-planar fronts propagating steadily appear. This is analyzed in Section 4. When
the front is a closed curve it is appropriate to use the polar coordinate version of the extended BornInfeld equation.
This equation is presented in Section 5 and the formal derivation is shown in Appendix A. By a different approach,
which we explain later, it was shown [14] that in the framework of the sine-Gordon equation when there is no
dissipation and when both phases have equal potential ( = h = = 0 in Eq. (1)), the front equation (10) governs
the eccentric oscillation of annular kinks (pulsons [14]), although its equivalence with the BornInfeld equation
has not appeared to be clear up to now. By means of a linear perturbation analysis we obtain equations which describe
the evolution of slightly perturbed circular interfaces and study both analytically and numerically the inuence of
, h and as compared with the = h = = 0 case given in [14]. We also show analytically that for h 0
circular interfaces shrink to a point in nite time and that establishes the conditions under which this behavior takes
place for values of h < 0. In Section 6 we give a local and more compact description of the evolution of the interface
in terms of its normal velocity and curvature.
2. Equation for the front motion
In this section we show that the motion of the front for Eq. (1) is described in cartesian coordinates by
(1 S
2
t
)S
xx
+2S
x
S
t
S
xt
(1 +S
2
x
)S
t t
S
t
(1 +S
2
x
S
2
t
)

h(1 +S
2
x
S
2
t
)
3/2
S
t
(1 +S
2
x
) = 0, (11)
where the parameters

h and , which are linear functions of h and , respectively, are dened later. Note that Eq.
(11) is reduced to (9) when

h = = 0 in the limit of strong dissipation, , by rescaling, t t . For
=

h = = 0, i.e., when there is no dissipation and both phases have equal potential, Eq. (11) reduces to the
classical BornInfeld equation (10) [12] as predicted by Neu [3].
We assume for simplicity that for small 0 and all t [0, T ], the domain can be divided into two open
regions
+
(t ; ) and

(t, ) with a curve (t ; ), which does not intersect , separating between them. This
interface, dened by
(t ; ) := {x : (x, t ; ) = 0} , (12)
is assumed to be smooth, which implies that its curvature and its velocity are bounded independently of . We also
assume that there exists a solution (x, t ; ) of (1), dened for small , for all x and for all t [0, T ] with
internal layer. As 0 this solution, , is assumed to vary continuously through the interface being
+
when
248 H.G. Rotstein, A.A. Nepomnyashchy / Physica D 136 (2000) 245265
x
+
(t ; ),

when x

(t, ) and a smooth function with rapid spatial variation through the interface. By
carrying out a singular perturbation analysis for 1 we obtain the laws of motion of the interface treating it as
a moving internal layer of width O(), focusing on the dynamics of a fully developed layer and not on the process
by which it was generated.
In cartesian coordinates the interface is represented by y = S(x, t, ) for sufciently small. We assume that the
curvature of the front is small compared to its width and dene, on the interface, a new variable
z :=
y S(x, t, )

,
which is O(1) as 0. We call the asymptotic form of as 0 with z xed, i.e.,
= (z, x, t, ). (13)
The eld equation (8) in (z, x, t ) coordinates becomes (see Appendix B)
(1 +S
2
x
S
2
t
)
zz
+f () +[2S
t

zt
+S
t t

z
+ S
t

z
2S
x

zx
S
xx

z
+h S
2
x
S
t

zzz
S
t

zzz
]
+O(
2
) = 0. (14)
The asymptotic expansions of and S are assumed to have the form

0
+
1
, S S
0
as 0. (15)
Substituting into (14) and equating coefcients of the corresponding powers of we obtain the following problems
for O(1) and O(), respectively:
(1 +S
2
x
S
2
t
)
0
zz
+f (
0
) = 0, (16)
(1 +S
2
x
S
2
t
)
1
zz
+f

(
0
)
1
= (S
xx
S
t t
S
t
)
0
z
2S
t

0
zt
+2S
x

0
zx
h +S
t
(1 +S
2
x
)
0
zzz
. (17)
In order to solve (16) we dene a new variable
:=
z
(1 +S
2
x
S
2
t
)
1/2
. (18)
In terms of , Eq. (16) reads

+f (
0
) = 0, (19)
whose solution is
0
= (), the unique solution of

+f () = 0, () =

, (0) = 0. Thus

0
=
0
_
z
(1 +S
2
x
S
2
t
)
1/2
_
. (20)
In terms of , x and t , Eq. (17) reads (see Appendix B)

+f

(
0
)
1
=
S
xx
S
t t
S
t
(1 +S
2
x
S
2
t
)
1/2


2S
x
(S
x
S
xx
S
t
S
xt
)
(1 +S
2
x
S
2
t
)
3/2
(
0

+
0

)
+
2S
t
(S
x
S
xt
S
t
S
t t
)
(1 +S
2
x
S
2
t
)
3/2
(
0

+
0

) h +
S
t
(1 +S
2
x
)
(1 +S
2
x
S
2
t
)
3/2

. (21)
It is straightforward to check that

() satises the homogeneous equation

+f

(
0
)
1
= 0. (22)
H.G. Rotstein, A.A. Nepomnyashchy / Physica D 136 (2000) 245265 249
That means that the operator dened as follows:
:=

2

2
+f

(
0
) (23)
has a simple eigenvalue at the origin with

as the corresponding eigenfunction. Then the solvability condition for


Eq. (21) gives
S
xx
S
t t
S
t
(1 +S
2
x
S
2
t
)
1/2
_

)
2
d h
_

d +
S
t
(1 +S
2
x
)
(1 +S
2
x
S
2
t
)
3/2
_

2S
x
(S
x
S
xx
S
t
S
xt
) 2S
t
(S
x
S
xt
S
t
S
t t
)
(1 +S
2
x
S
2
t
)
3/2
_

d = 0. (24)
A simple calculation shows that
_

d =
1
2
_

)
2
d and
_

d =
_

)
2
d. (25)
We call
:=
_

)
2
d
_

)
2
d
and

h := h
(+) ()
_

)
2
d
. (26)
Replacing (25) and (26), multiplying Eq. (24) by (1 + S
2
x
S
2
t
)
3/2
and rearranging terms we obtain (11). Note
that for f () = (
3
)/2 (GinzburgLandau theory), () = tanh(/2), = /5 and

h := 3h, whereas for
f () = sin , () = 4 tan
1
e

, = /3 and

h := (/4)h.
3. Motion of planar fronts and their stability
In this section we analyze the evolution of a planar interface and the stability of linear perturbations. We obtain
that as long as the absolute value of the initial velocity is bounded from above by 1, the velocity of the planar
interface tends to a constant value as t and that the linear perturbations decay, either in a monotonic or
oscillatory way, to zero as t .
3.1. Propagation of the planar front
Consider (11) in the innite interval (, ) satisfying |S(x, t )| < as x . Assume that S(x, t )
depends weakly on x and let
S(x, t ) = S
0
(t ) +S
1
(x, t ), S(x, 0) = f
0
+f
1
(x) and S
t
(x, 0) = g
0
+g
1
(x), (27)
with f
0
and g
0
constants and |g
0
| < 1. Substituting (27) into (11) we obtain the following equations:
S
0,t t
+ S
0,t
(1 S
2
0,t
) +

h(1 S
2
0,t
)
3/2
+ S
0,t
= 0, (28)
and
S
1,t t
(1 S
2
0,t
)S
1,xx
+[ (1 S
2
0,t
) 2 S
2
0,t
3

hS
0,t
(1 S
2
0,t
)
1/2
+ ]S
1,t
= 0. (29)
Eq. (28) describes the evolution of a planar interface, whereas Eq. (29) is a linear equation for the small perturbation
of the planar interface R
1
(x, t ).
250 H.G. Rotstein, A.A. Nepomnyashchy / Physica D 136 (2000) 245265
From (28) and (27) we have the following second-order ODE:
S
0,t t
+ S
0,t
(1 S
2
0,t
) +

h(1 S
2
0,t
)
3/2
+ S
0,t
= 0, S
0
(0) = f
0
, S
0,t
(0) = g
0
. (30)
For the case

h = 0 (both phases having equal potential) we can give an analytical expression for the solution of
(30) for values of and satisfying the condition + (1 g
2
0
) > 0. Calling = + the solution of (30) is
given by
S
0
(t ) = f
0

ln(
_
g
2
0
+
_
)
_

t
1

ln(
_
g
2
0
+
_
g
2
0
+( g
2
0
)e
2t
), (31)
where the upper choice of signs correspond to g
0
> 0 and the lower choice of signs correspond to g
0
< 0. Note that
the direction of motion depends on the sign of g
0
and for t the front will cease to move and its coordinate S
0
(t )
asymptotically approaches the value f
0
(1/

) ln((g
2
0
)
1/2
+

). When, in addition, there is no damping


( = 0 and = 0), it can be seen from (30) that S
0
(t ) = f
0
+g
0
t .
In order to investigate this problem for values of

h = 0 (different potential in the two phases) we look at the xed
points of the differential equation and study their stability. First we consider the case = 0. We call v := S
0,t
and
the rst equation in (30) becomes
v
t
= v(1 v
2
)

h(1 v
2
)
3/2
. (32)
For =

h = 0, v(t ) = g
0
, i.e., the solution of (30) is S
0
(t ) = f
0
+ g
0
t . For = 0 (no dissipation) but

h = 0
(different potential in the two phases) the solution of (32) is
v(t ) =
g
0


ht (1 g
2
0
)
1/2
(1 g
2
0
+[g
0


ht (1 g
2
0
)
1/2
]
2
)
1/2
. (33)
For > 0 we cannot give an explicit solution for (32) but we see that there are three equilibrium points: v

= 1,
v
0
=

h/(
2
+

h
2
)
1/2
and v
+
= 1. We dene
F(v) := v(1 v
2
)

h(1 v
2
)
3/2
(34)
and calculate F

(v) at these equilibrium points obtaining F

(1) = F

(1) = 2 and F

h/(
2
+

h
2
)
1/2
) = .
Then 1 and 1 are unstable steady state solutions and

h/(
2
+

h
2
)
1/2
is an attractor corresponding to the balance
between the moving force caused by the difference of phases potential and dissipation. Note that depending on the
sign of

h the stable equilibrium is negative or positive and that lim

h
v
0
= 1, i.e., the front propagates in the
direction of the phase with a higher potential. Note also that for = 0 and

h > 0 (

h < 0), v
0
= 1 (v
0
= +1),
i.e., in absence of dissipation the velocity of the planar front tends to the limit value |v| = 1. Thus for g
0
= 1,
S
0
(t ) = f
0
t ; for g
0
=

h/(
2
+

h
2
)
1/2
, S
0
(t ) = (

h/(
2
+

h
2
)
1/2
)t + f
0
and for g
0
(1, 0) (0, 1),
v

h/(
2
+

h
2
)
1/2
as t .
The dynamics of the evolution of the position of a planar interface with time for various values of and

h is
illustrated in Fig. 1. The solutions were calculated numerically by means of the modied Euler method. We can
see that the response of the interface on increasing is to approach a constant value as t as predicted by
(31), whereas increasing or decreasing

h, for = 1 and = 0, produce a qualitative change in the evolution of the
interface no longer tending to a constant value as time increases but to a line with either positive, zero or negative
slope according to

h < 0,

h = 0 or

h > 0, respectively.
H.G. Rotstein, A.A. Nepomnyashchy / Physica D 136 (2000) 245265 251
Fig. 1. Position of a planar interface vs. time for S
0
(0) = 0, S
0,t
(0) = 0.5. (a)

h = 0, = 0 and = 1, 2, 5 from top to bottom. (b) = 1,
= 0 and

h = 0.2, 0.1, 0, 0.1, 0.2 from top to bottom.
3.2. Stability
In the previous section we already investigated the stability of the motion with a stationary velocity v
0
with
respect to planar disturbances and found that the stability condition is F

(v
0
) < 0. Now we shall consider spatially
inhomogeneous disturbances. If = 0, then
F(v) := v(1 v
2
)

h(1 v
2
)
3/2
v. (35)
252 H.G. Rotstein, A.A. Nepomnyashchy / Physica D 136 (2000) 245265
It is easy to show that for > 0 the equilibrium point v
0
, F(v
0
) is unique and stable; i.e., F

(v
0
) < 0.
We look for solutions of (29) for suitable initial and boundary conditions, i.e.,
S
1,t t
(1 v
2
0
)S
1,xx
F

(v
0
)S
1,t
= 0, S
1
(x, 0) = f
1
(x), S
1,t
(x, 0) = g
1
(x),
|S
1
(x, t )| < , as x , (36)
where F(v) is dened by (35). We look at solutions of the form S
1
(x, t ) = Re(T
k
(t ) e
ikx
) for k real, where T
k
(t )
satises
T

k
F

(v)T

k
+k
2

2
(1 v
2
0
)T
k
= 0, T
k
(0) = f
1,k
, T

k
(0) = g
1,k
,
where f
1,k
and g
1,k
are the Fourier transformof f
1
(x) and g
1
(x), respectively. We conclude that F

(v
0
) < 0, v
0
< 1
remains the necessary and sufcient condition for asymptotic stability of the planar front.
4. Steady propagation of non-planar fronts
Let us nd the solutions of (11) corresponding to propagation with a constant velocity: S(x, t ) =

S(x) + vt .
Substituting into Eq. (11) yields
(1 v
2
)

S
xx
v(1 +

S
2
x
v
2
)

h(1 +

S
2
x
v
2
)
3/2
v(1 +

S
2
x
) = 0. (37)
We rst analyze the case

h = 0 (equal potential of phases). The solution of (37) is

S(x) =
1
2a
ln [1 +tan
2
(a b x +c
1
)] +c
2
, (38)
where a = ( + )v/(1 v
2
), b
2
= ( (1 v
2
) + )/( + ) and c
1
and c
2
are integration constants which depend
on the initial conditions. This solution is a shortening nger of width /a b whose orientation depends on the
sign of v.
Next, we analyze the case = = 0 (no dissipation). The solution of (37) is

S(x) =
(1 v
2
)
1/2

h
[1 (

hx +c
1
)
2
]
1/2
+c
2
, (39)
where c
1
and c
2
are integration constants depending on the initial conditions. This solution corresponds to an
ellipse centered at (c
1
/

h, c
2
/

h) cutting the x-axis at ((1/

h)(c
1
(1 c
2
2

h
2
/(1 v
2
))
1/2
, 0)) and the y-axis at
(0, c
2
(1 c
2
2
)
1/2
(1 v
2
)
1/2
/

h). For v = 0 it becomes a circle, corresponding to an unstable critical nucleus


of the phase with a lower potential in the phase with the higher potential (see Section 3).
5. Motion of circular fronts and their stability
In this section we will use Eq. (11) written in polar coordinates:

2
(1
2
t
) +
2

t
_

2
(1
2
t
) +
2

(1
2
t
) +
t
_

2
(1
2
t
) +
2


h

t
(1 +
2

/
2
)
_
(
2
(1
2
t
) +
2

)
3
= 0, (40)
where = (, t ) represents the interface and

h and are dened as in the cartesian case. The derivation is shown
in Appendix A.
H.G. Rotstein, A.A. Nepomnyashchy / Physica D 136 (2000) 245265 253
5.1. Circular interfaces
Assume that (, t ) depends weakly on and let
(, t ) = R
0
(t ) +R
1
(, t ), (, 0) = f
0
+f
1
() and
t
(, 0) = g
0
+g
1
(), (41)
with f
0
and g
0
constants and |g
0
| < 1. Substituting (41) into (40) we obtain the following equations:
R
0,t t
=
_
1
R
0
+ R
0,t
_
(1 R
2
0,t
)

h(1 R
2
0,t
)
3/2
R
0,t
(42)
and
R
1,t t
= (R
1,
+R
1
)
1R
2
0,t
R
2
0
+
_
2
R
0
+ R
2
0
R
2
0,t
R
2
0
R
0,t
+ (1R
2
0,t
) +3

hR
0,t
(1R
2
0,t
)
1/2

_
R
1,t
. (43)
Eq. (42) describes the interface dynamics in the circular case, while Eq. (43) is a linear equation for the small
eccentric perturbation R
1
(, t ).
Let us consider (42) with suitable initial conditions for values of t such that R
0,t
(t ) > 0. We call t
c
the quenching
time, i.e., the time t such that R
0
(t
c
) = 0 and R
0
(t ) > 0 for all t < t
c
. The following second-order non-linear ODE,
where

h and are proportional to h and , respectively, has to be analyzed:
R
0,t t
= (1/R
0
+ R
0,t
)(1 R
2
0,t
)

h(1 R
2
0,t
)
3/2
R
0,t
, R
0
(0) = f
0
, R
0,t
(0) = g
0
, (44)
For =

h = = 0 (no dissipation and both phases having equal potentials) and g
0
= 0, a bounded solution
satisfying R
0,t
(0) = 0 is [15]:
R
0
(t ) = f
0
cos
_
t
f
0
_
. (45)
We see that circles shrink to a point when t = f
0
/2, and that |R
0,t
| < 1 for 0 < t < t
c
. For the more general case,

h = 0 and and both non-zero, we can prove that as long as 1 < g


0
0, R
0,t
(t ) < 0 for t (0, t
c
). To this
task, we multiply the rst equation in (44) by R
0,t
and rearrange terms obtaining
d
dt
ln
1 R
2
0,t
R
2
0
= 2 R
2
0,t
+2
R
2
0,t
1 R
2
0,t
.
Solving we get
R
2
0,t
+
1 g
2
0
f
2
0
R
2
0
exp
_
_
t
0
_
2 R
2
0,t
() +
2 R
2
0,t
()
(1 R
2
0,t
())
__
d = 1,
which shows that as long as |g
0
| < 1, |R
0,t
(t )| < 1 for 0 < t < t
c
. Nowassume that 1 < g
0
0. Then there exists
a t
0
> 0 such that R
0,t
(t ) < 0 and R
0
(t ) < f
0
for t < t
0
. (If g
0
= 0, then R
0,t t
(0) = 1/f
0
< 0.) Nowsuppose that
there exists a t
1
(0, t
c
) such that R
0,t
(t
1
) = 0 and R
0,t
(t ) < 0 for t (0, t
1
). Then R
0,t t
(t
1
) = 1/R
0
(t
1
) < 0,
i.e., t
1
is a maximum, a contradiction. We conclude that, like in the =

h = = 0 case (no dissipation and equal
potential in both phases), R
0,t
(t ) < 0 for t (0, t
c
).
In order to study the general case we make a phase plane analysis. For simplicity we will analyze the case = 0.
To this end we call
x := R
0
and y := R
0,t
. (46)
254 H.G. Rotstein, A.A. Nepomnyashchy / Physica D 136 (2000) 245265
Problem (44) becomes
x = y, x(0) = f
0
, y = (1/x +y)(1 y
2
)

h(1 y
2
)
3/2
, y(0) = g
0
. (47)
We can see that R
0
(t ) = t + f
0
is a solution for (44) with = 0 and g
0
= 1. By considerations of existence
and uniqueness of solutions of ordinary differential equations we conclude that in the phase plane there is a region
bounded by the lines y = +1 and y = 1 with the property that every curve starting in this region remains inside
it for all future time. We shall call this region D.
As a rst step we analyze the case =

h = 0 (no dissipation and no difference of potential between the two
phases) for arbitrary initial conditions in D. Dividing the second equation in (47) by the rst one and solving we
obtain that
c
2
x
2
+y
2
= 1, (48)
where c
2
= (1 g
2
0
)/f
2
0
> 0. This means that the orbits are ellipses (or a circle of radius 1 when the initial
conditions are in a circle of radius 1). As a matter of fact, since we are considering x > 0 the orbits are half ellipses
(or half a circle). Replacing (48) into the rst equation in (47) and solving we get
R(t ) =
1
c
cos(ct +), (49)
where = arccos(cf
0
), t < t
c
= (/2 )/c.
We now study the more general case ( = 0). A detailed analysis can be found in [5]. Recall that we are
considering positive values of x. In Figs. 2 and 3 we can see the phase portraits for = 1 (dissipation),

h = 0 (both
phases with equal potential) and for = 0 (no dissipation),

h = 1 (the potential of the phase inside the circle is
lower than that of the phase outside the circle). In both graphs the isoclines are drawn with dashed curves and the
trajectories are drawn with full curves. The arrowheads indicate the direction of motion of the trajectories. In Fig. 2
the isoclines are y 0 and y = 1/x which do not intersect each other, i.e., there is no equilibrium point. We can
see that that for every initial condition the corresponding trajectories reach the y-axis, i.e., circles shrink to points
in nite time. This is also true for all non-negative values of and

h and for every initial condition in . In Fig. 3
the isoclines are y 0 and 1/x (1 y
2
)
1/2
= 0. They intersect at the xed point (1,0), a saddle corresponding
to the unstable critical nucleus (see Section 2.4). The eigenvalues for the linearized matrix of (47) with = 0 are
1. We see that there exist initial conditions for which circles reach the y-axis, i.e., circles shrink to points in nite
time, and for other initial conditions circles grow unboundedly. This situation repeats for all negative values of

h
with = 0 where the xed point is (1/

h, 0) and the corresponding eigenvalues for the linearized matrix of (47)


with = 0 are

h. The case

h < 0, = 0 is not discussed here.
We can see that when circles shrink to a point (with = 0), near the singularity (R
0
0 or t t
c
), Eq. (44)
becomes R
2
0,t
= 1. Thus, dening t

= t t
c
, R
0
t

(by the way this function, which satises R


0
(0) = t
c
, is a
second solution of the rst equation in (44) passing through the singularity point). In fact, we can easily calculate
that R
0
(t

) = t

+O(t
3

) as t

0.
For large values of either of the parameters ,

h or , i.e., for large dissipation or a big difference of potential
between the two phases, the behavior, specially near the quenching point, is like the classical owby mean curvature
equation for a circle and the quenching points can be asymptotically determined. To see that let us look at Eq. (44) for

h = 0 (bothphases withequal potential) and = O(1). After rescaling, t t for becomes R


0,t
= 1/R
0
(classical owby mean curvature equation for a circle) whose solution is given by R
0
=
_
R
2
0
(0) 2t . This solution
is valid away from R
0
= 0 and (returning to the original time variable) it gets the value 1/ at t = /2 +1/2 .
Near R
0
= 0 we have seen in the last paragraph that R
0
= t +t
c
. Calculating at t one obtains t
c
= /2 +1/2 .
H.G. Rotstein, A.A. Nepomnyashchy / Physica D 136 (2000) 245265 255
Fig. 2. Phase portrait for = 1 and

h = = 0. The isoclines are drawn with dashed curves and the trajectories are drawn with full curves. The
arrowheads indicate the direction of motion of the trajectories.
Then t
c
/2 as . If = , then t
c
as . An analogous result would be obtained if we
change by in the above discussion. For

h Eq. (44) becomes R
2
0,t
= 1 whose decreasing solution is
R
0
= t
c
+f
0
. Thus t
c
f
0
as

h .
For small values of either ,

h or , i.e., for small dissipation or small difference of potential between the
two phases, when the remaining parameters are zero we can think of problem (44) as a small perturbation of the
=

h = = 0 case (no dissipation and both phases with equal potential) for which (49) is the solution. Setting
R
0
(t ) = r
0
(t ) +r
1
(t ) and solving for

h = = 0 and for = = 0 we obtain
r
1
() = c
2
(
11
6
sin()
1
16
sin(3)
1
2
cos()
1
4

2
sin())
and
r
1
(t ) = c
2
(
1
16
cos() +
1
16
cos(3)
1
4
sin()),
respectively.
5.2. Stability
We are interested in solutions to (43) for suitable initial and boundary conditions for values of t > 0 such that
R
0
(t ) > 0 and R
1
(, t ) > 0, i.e.,
256 H.G. Rotstein, A.A. Nepomnyashchy / Physica D 136 (2000) 245265
Fig. 3. Phase portrait for

h = 1 and = = 0. The isoclines are drawn with dashed curves and the trajectories are drawn with full curves.
The arrowheads indicate the direction of motion of the trajectories.
R
1,t t
q(t )R
1,t
= p(t )(R
1,
+R
1
), R
1
(, 0) = f
1
(), R
1,t
(, 0) = g
1
(),
R
1
(, t ) = R
1
(, t ), R
1,
(, t ) = R
1,
(, t ), (50)
where
p(t ) =
1 R
2
0,t
R
2
0
and q(t ) = 2R
0,t
R
0
+ R
2
0
R
0,t
R
2
0
(1 R
2
0,t
) +3

hR
0,t
(1 R
2
0,t
)
1/2
.
Problem (50) is a linear homogeneous PDE. Separation of variables gives solutions of the form
R
1
(, t ) = A
n
(t ) cos(n) and R
1
(, t ) = B
n
(t ) sin(n) (51)
for n = 1, 2, . . . , where A
n
and B
n
satisfy
A

n
q(t )A

n
= p(t )(1 n
2
)A
n
, A
n
(0) = f
c
1,n
, A

n
(0) = g
c
1,n
(52)
for n = 0, 1, . . . , and
B

n
q(t )B

n
= p(t )(1 n
2
)B
n
, B
n
(0) = f
s
1,n
, B

n
(0) = g
s
1,n
(53)
for n = 0, 1, . . . , where f
c
1,n
, g
c
1,n
, f
s
1,n
and g
s
1,n
are the coefcients of the sine and cosine Fourier expansions of
f
1
() and g
1
(), respectively. Let us look at the solutions (52). The analysis of (53) is similar. Let us analyze as
H.G. Rotstein, A.A. Nepomnyashchy / Physica D 136 (2000) 245265 257
a rst step, the case =

h = = 0 (no dissipation and equal potential of phases). Two linearly independent
solutions, which are just a generalization of those found in [14], are given by
A
n
(t ) =
_
n cos(ct +) sin [n(ct +)] sin(ct +) cos [n(ct +)],
n cos(ct +) cos [n(ct +)] +sin(ct +) sin [n(ct +)].
(54)
For f
0
= 1 and g
0
= 0 it reduces to [14]:
A
n
(t ) =
R
1,t
(0)
n
2
1
(n cos(t ) sin(nt ) sin(t ) cos(nt )) +
R
1
(0)
n
(n cos(t ) cos(nt ) sin(t ) sin(nt )) .
Consider an initial closed convex curve R(0) = 1 + A
2
(0) cos(2) with R
1,t
(0) = 0. Its evolution with time is
given by the former solution with n = 2. At t = t
c
= /2 (the quenching point for the circle) we have A
2
(t
c
) = 0;
i.e., the perturbation disappears at t = t
c
like for the classical parabolic case (9). This convex curve shrinks to a
point in nite time. This situation repeats either for even values of n with g
c
1,n
= 0 or for odd values of n with
f
c
1,n
= 0. For the remaining cases the perturbation is still present at the shrinking point for the circle t
c
. For those
cases where g
0
= 0 and f
0
= 1 the analysis and results are similar.
As we pointed out before, Neu [3] showed, for the case =

h = = 0 (no dissipation and equal potential of
phases), that closed kink can be stabilized against a collapse by the appearance of short-wavelength, small amplitude
waves.
In order to analyze the more general case ( = 0), for which we do not have an expression for R
0
(t ), we replace
the asymptotic approximation near the quenching point (obtained in Section 3.1.2) R
0
= t

+O(t
3

) as t

0,
where t

= t t
c
, into the rst equation in either (52) and/or (53) depending on the initial perturbation. Without
loss of generality we do it for (52) obtaining
A

_
2
t

+2 (4C +3

6C

h)t


_
A

n
+6C(n
2
1)A
n
= 0
for a constant C. As t

0 this equation is essentially that which would have been obtained for the case =

h =
= 0,
A

2
t

n
+6C(n
2
1)A
n
= 0.
Two linearly independent solutions are 1 +O(t
2

) and t
3

+O(t
5

) as t

0.
In order to have a clearer picture of the evolution of the perturbations, solutions of (44) were calculated numerically
by means of the modied Euler method adapting the step size by means of a RungeKutta method of order 4 in
order to control possible numerical instabilities due to the fact that near t
c
, 1/R
0
(details of the procedure can
be found in [16]). Eqs. (52) and (53), homogeneous linear second-order ODEs, were solved numerically for some
values of n by means of the modied Euler method until the instant t = t
c
. In Fig. 4 we can observe the shrinkage of
a perturbed circle initially with R() = 1+0.1 sin(2) and R
t
() = 0.1 for

h = = 0 and = 0, 5, respectively.
For = 0 we see that the interface rotates as it shrinks. This phenomenon is not observable for = 5. In Fig. 5
we can see the phenomenon of rotation of the interface as a perturbed circle initially with R() = 1 +0.1 sin(2)
and R
t
() = 0 shrinks for = = 0 and

h = 0.5, 0.9, respectively. In the latter case we can observe that
at the rst instant the curve shrinks in one direction, whereas grows in the other as it rotates. For

h = 1 this
effect is analytically clear since R(t ) = 1/

h is a solution of (44) which in turn makes the solution of (50) being


non-decreasing with time. As a result the interface rotates without shrinking.
258 H.G. Rotstein, A.A. Nepomnyashchy / Physica D 136 (2000) 245265
Fig. 4. Shrinkage of a perturbed circle for R(0) = 1 +0.1 sin(2), R
t
(0) = 0.1 =

h = 0. (a) = 0 and t = 0, 0.7, 1.2, 1.4. (b) = 5 and
t = 0, 1.55, 2.52, 2.95.
6. A local description of the evolution of the interface in terms of its normal velocity and curvature
In this section we present an equation for the evolution of the interface in terms of its kinematic and geometric
properties which reads
dv
dt
+ v(1 v
2
) (1 v
2
) +

h(1 v
2
)
3/2
+ v = 0, (55)
H.G. Rotstein, A.A. Nepomnyashchy / Physica D 136 (2000) 245265 259
Fig. 5. Shrinkage of a perturbed circle for R(0) = 1 +0.1 sin(2), R
t
(0) = 0, = = 0. (a)

h = 0.5 and t = 0, 0.9, 1.4, 1.65. (b)

h = 0.9
and t = 0, 1.5, 2.5, 2.8.
where dv/dt is the Lagrangian time derivative of v which is calculated along the trajectory of the interfacial point
moving with the normal velocity v:
dv
dt
= v
t
+v v
x
sin ,
where is the angle between the normal velocity and the y axis and sin = S
x
/(1 + S
2
x
)
1/2
. In order to obtain
this equation we express the normal velocity v and the curvature are given in cartesian coordinates by
260 H.G. Rotstein, A.A. Nepomnyashchy / Physica D 136 (2000) 245265
v =
S
t
(1 +S
2
x
)
1/2
, =
S
xx
(1 +S
2
x
)
3/2
. (56)
The Lagrangian derivative of v, dv/dt , is given by
dv
dt
= v
t
+v
x
v sin =
S
t t
(1 +S
2
x
)
1/2

2S
x
S
t
S
xt
(1 +S
2
x
)
3/2
+
S
2
x
S
2
t
S
xx
(1 +S
2
x
)
5/2
(57)
Dividing Eq. (11) by (1 +S
2
x
)
3/2
and rearranging terms we get
S
xx
(1 +S
2
x
)
3/2

S
2
t
S
2
x
S
xx
(1 +S
2
x
)
5/2
+
2S
x
S
t
S
xt
(1 +S
2
x
)
3/2

S
t t
(1 +S
2
x
)
1/2

S
2
t
S
xx
(1 +S
2
x
)
5/2

S
t
(1 +S
2
x
)
1/2
_
1
S
2
t
1 +S
2
x
_


h
_
1
S
2
t
1 +S
2
x
_
3/2

S
t
(1 +S
2
x
)
1/2
= 0. (58)
Replacing (56) and (57) we obtain (55).
For the two-dimensional driven damped sine-Gordon equation

t t
+
t
= sin f, (59)
where and f were assumed small, Malomed [2] has found, following physical considerations, an equation for the
shape of the moving front in the framework of the equilibrium shapes of driven quasi-one-dimensional kinks which
reads

4
f (1 v
2
)
1/2
+ = v. (60)
Eq. (59) can be obtained from (1) for = , f = h, = 0 and by rescaling t t , x x and y y. Note
that Eq. (60) is valid only if the velocity remains constant in time. Eq. (55), which is obtained in a self-consistent
way, is a generalization of Eq. (60) that includes the derivative of v in the normal direction to the interface and a
linear term for v.
7. Discussion
We have obtained Eqs. (11) and (A.12) as describing the motion of the interface for (1) in cartesian and polar
coordinates, respectively. For = h = = 0 (no dissipation and equal potential of phases) in (1), which implies

h = = 0 in the corresponding interface motion equation, the cartesian coordinate version (11) was known to be
the BornInfeld equation and it was obtained from (8) when f () = (
3
)/2 [3]. Our result extends the validity
of the BornInfeld equation to a more general class of equations including the sine-Gordon equation. The effect of
adding a linear dissipation term (1), specially that whose coefcient is , is a nonlinear contribution in (11) whose
exact form is not trivial. On the other hand, as we remarked before, for =

h = 0 and Eq. (11) is an
extension of the classical ow by mean curvature equation (curve shortening equation) or a memory version of it.
Thus, as we showed before, by increasing , for h = = 0, we pass from a hyperbolic behavior to a parabolic one.
As a consequence of the dissipation term ( > 0) the planar front, otherwise moving unboundedly with constant
velocity, tends asymptotically to a constant position which depends on the values of the parameters. A difference
of potential between the two phases (h = 0) produces an unbounded front motion, but the front evolves with a
velocity tending to 1 depending on the sign of h. In addition, when there is dissipation and when both phases have
different potential, non-planar front moving steadily appears.
H.G. Rotstein, A.A. Nepomnyashchy / Physica D 136 (2000) 245265 261
When dissipation is present and both phases have equal potential, a circular front still shrink to a point in nite
time. This behavior changes when the two phases have different potential. For each value of h and for some initial
conditions shrinkage does not occur anymore and circular fronts grow unboundedly.
In [14] it was shown that when = h = = 0, the polar coordinate version (A.12) governs the eccentric
oscillation of annular kinks (pulsons) in the framework of the sine-Gordon equation. This leads us immediately to
conclude that the oscillation of eccentric pulsons is in fact governed by the BornInfeld equation. The technique
used in [14] was different from that used here. Their starting point was the Lagrangian density for the sine-Gordon
equation under the assumption of a 2 sine-Gordon kink. Thus our approach permits the analysis of the more
general case.
In Section 4 we have obtained an equation for the interface motion in terms of geometrical and kinematic properties
of the interface, like in the owby mean curvature case. This enables us to look at the problemfroma different point
of view and generalize the work of Malomed [2], where kinks moving with constant velocities were considered,
and also to carry out, more easily, numerical calculations.
Acknowledgements
We acknowledge Amy Novick-Cohen for reading the manuscript, numerous discussions and constructive criti-
cism. We ackowledge Jacob Rubinstein and Chaim Charach for useful comments.
Appendix A. Derivation of the extended BornInfeld equation in polar coordinates
We make the same assumptions and follow essentially the same steps as in the cartesian coordinate case (see
Section 2). The interface in polar coordinates is represented by r = (, t, ) for sufciently small. In the interface
layer we dene a new variable
z :=
r (, t )

. (A.1)
As in the cartesian case, z is O(1) as 0. We call the asymptotic form of as 0 with z xed, i.e.,
= (z, , t, ), (A.2)
and we assume that the asymptotic expansions of and are

0
+
1
,
0
as 0. (A.3)
The eld equation (8) in terms of the (z, , t ) coordinate system becomes (see Appendix B)
_
1 +

2

2

2
t
_

zz
+f () +
_
2
t

zt
+
t t

z
+
t

z
+
1

z
2

z
2
z
2

3

zz

2

z
+h
t

zzz

2

zzz
_
+O(
2
) = 0. (A.4)
We now replace (A.3) into (A.4) and equate coefcients of the corresponding powers of obtaining the O(1) and
O() problems, respectively:
_
1 +

2

2

2
t
_

0
zz
+f (
0
) = 0, (A.5)
262 H.G. Rotstein, A.A. Nepomnyashchy / Physica D 136 (2000) 245265
_
1 +

2

2

2
t
_

1
zz
+f

(
0
)
1
=
_

2

t t

1

t
_

0
z
2
t

0
zt
+2

0
z
+2
z
2

3

0
zz
h +
t
_
1 +

2

2
_

0
zzz
. (A.6)
To solve (A.5) we dene a new variable
:=
z
(1 +
2

/
2

2
t
)
1/2
. (A.7)
In terms of (, , t ), Eq. (A.5) reads

+f (
0
) = 0, (A.8)
whose solution is
0
= (), the unique solution of

+f () = 0, () =

, (0) = 0. Thus

0
=
0
_
z
(1 +
2

/
2

2
t
)
1/2
_
. (A.9)
In the (, , t ) coordinate system, Eq. (A.6) reads (see Appendix B)

+f

(
0
)
1
=

/
2

t t
1/
t
(1 +
2

/
2

2
t
)
1/2

0

/
2

/
3

t
(1 +
2

/
2

2
t
)
3/2
(
0

+
0

)
+2
t
(

t
/
2

t
/
3

t t
)
(1 +
2

/
2

2
t
)
3/2
(
0

+
0

) +
2z
2

3
1
1 +
2

/
2

2
t

h +
(1 +
2

/
2
)
t
(1 +
2

/
2

2
t
)
3/2

. (A.10)
As in the cartesian case

() satises Eq. (22) which again means that the operator dened in (23) has a simple
eigenvalue at the origin with

as the corresponding eigenfunction. Then the solvability condition for Eq. (A.10)
gives

/
2

t t
1/
t
(1 +
2

/
2

2
t
)
1/2
_

)
2
d
+
2
t
(

t
/
2

t
/
3

t t
) 2(

/
2
)(

/
2

/
3

t
)
(1 +
2

/
2

2
t
)
3/2
_

d
+2

3
1
(1 +
2

/
2

2
t
)
1/2
_

d h
_

d +

t
+(
2

/
2
)
t
(1 +
2

/
2

2
t
)
3/2
_

d = 0.
(A.11)
Replacing (25) and (26), multiplying Eq. (A.11) by (1 +
2

/
2

2
t
)
3/2
and rearranging terms we obtain
(1
2
t
)

+2
t

t
(
2
+
2

)
t t

t
(
2
(1
2
t
) +
2

) (1
2
t
)
2


h
1

(
2
(1
2
t
) +
2

)
3/2

t
(
2
+
2

) = 0, (A.12)
which is a polar coordinate version of the extended BornInfeld equation (11). A simple calculation shows that this
equation is (40).
H.G. Rotstein, A.A. Nepomnyashchy / Physica D 136 (2000) 245265 263
Appendix B. Change of variables
B.1. From cartesian to (z, x, t ) coordinates
To go from cartesian to (z, x, t ) coordinates we transform derivatives as follows:

t
=
t

1

S
t

z
,

t t
=
t t

1

2S
t

zt
+
1

2
S
2
t

zz

S
t t

z
,

xx
=
xx

1

2S
x

zx
+
1

2
S
2
x

zz

S
xx

z
,

yy
=
1

zz
,

xxt
=
xxt

1

S
t

xxz

2S
xt

zx

1

2S
x

xzt
+
1

2
2S
x
S
t

xzz
+
1

2
2S
x
S
xt

zz
+
1

2
S
2
x

zzt

1

3
S
2
x
S
t

zzz

S
xxx

S
xx

zt
+
1

2
S
xx
S
t

zz
,

yyt
=
1

zzt

1

3
S
t

zzz
.
Replacing into (8) and rearranging terms we obtain
(1 +S
2
x
S
2
t
)
zz
+f () +[2S
t

zt
+S
t t

z
+ S
t

z
2S
x

zx
S
xx

z
+g S
2
x
S
t

zzz
S
t

zzz
]
+O(
2
) = 0. (B.1)
B.2. From (z, x, t ) to (, x, t ) coordinates
To go from (z, x, t ) to (, x, t ) coordinates, derivatives are transformed as follows:

z
= (1 +S
2
x
S
2
t
)
1/2
,

x
= z(1 +S
2
x
S
2
t
)
3/2
(S
x
S
xx
S
t
S
xt
),

t
= z(1 +S
2
x
S
2
t
)
3/2
(S
x
S
xt
S
t
S
t t
),

zx
= (1 +S
2
x
S
2
t
)
3/2
(S
x
S
xx
S
t
S
xt
),

zt
= (1 +S
2
x
S
2
t
)
3/2
(S
x
S
xt
S
t
S
t t
).

0
z
=
0

z
=
1
(1 +S
2
x
S
2
t
)
1/2

, (B.2)

0
zt
=
0

t
+
0

zt
=
S
x
S
xt
S
t
S
t t
(1 +S
2
x
S
2
t
)
3/2
(
0

+
0

), (B.3)

0
zx
=
0

x
+
0

zx
=
S
x
S
xx
S
t
S
xt
(1 +S
2
x
S
2
t
)
3/2
(
0

+
0

). (B.4)
264 H.G. Rotstein, A.A. Nepomnyashchy / Physica D 136 (2000) 245265
B.3. From polar to (z, , t ) coordinates
To go from polar to (z, , t ) coordinates we transform derivatives as follows:

t
=
t

1

z
,
t t
=
t t

1

2
t

zt
+
1

2
t

zz

t t

z
,
r
=
1

z
,
rr
=
1

zz
,

z
+
1

zz

z
,
rt
=
1

zt

1

zz
,
rrt
=
1

zzt

1

zzz
,

t
=
t

1

2
t

z

1

2
zt
+
1

2
2

zz
+
1

2
2

zz
+
1

zzt

zzz

zt
+
1

zz
.
In polar coordinates =
rr
+ (1/r)
r
+ (1/r
2
)

. From (A.1) we have r = + z. Developing in power


series we have 1/r = 1/ (z/
2
) + O(
2
) and 1/r
2
= 1/
2
(2z/
3
) + O(
2
). Replacing into (8) and
rearranging terms we obtain
_
1 +

2

2

2
t
_

zz
+f () +
_
2
t

zt
+
t t

z
+
t

z
+
1

z
2

z
2
z
2

3

zz

2

z
+g
t

zzz

2

zzz
_
+O(
2
) = 0. (B.5)
B.4. From (z, , t ) to (, , t ) coordinates
To go from (z, , t ) to (, , t ) coordinates, derivatives are transformed as follows:

z
=
_
1 +

2

2

2
t
_
1/2
,

= z
_
1 +

2

2

2
t
_
3/2
_

2


3

3

t

t
_
,

t
= z
_
1 +

2

2

2
t
_
3/2
_

2


3

3

t

t t
_
,

zx
=
_
1 +

2

2

2
t
_
3/2
_

2


3

3

t

t
_
,

zt
=
_
1 +

2

2

2
t
_
3/2
_

2


3

3

t

t t
_
,

0
z
=
0

z
=
1
(1 +
2

/
2

2
t
)
1/2

, (B.6)
H.G. Rotstein, A.A. Nepomnyashchy / Physica D 136 (2000) 245265 265

0
zt
=
0

t
+
0

zt
=

t
/
2

t
/
3

t t
(1 +
2

/
2

2
t
)
3/2
, (B.7)

0
zx
=
0

z
+
0

z
=

/
2

/
3

t
(1 +
2

/
2

2
t
)
3/2
. (B.8)
References
[1] D.W. Laughlin, A.C. Scott, Perturbation analysis of uxon dynamics, Phys. Rev. A 18 (1978) 16521680.
[2] B.A. Malomed, Dynamics of quasi-one-dimensional kinks in the two-dimensional sine-Gordon model, Physica D 52 (1991) 157170.
[3] J.C. Neu, Kinks and the minimal surface equation in Minkowski space, Physica D 43 (1990) 421434.
[4] H.G. Rotstein, A.A. Nepomnyashchy, A. Novick-Cohen, Hyperbolic phase transition dynamics in two dimensions, preprint, 1997.
[5] H.G. Rotstein, Phase transitions dynamics with memory, Ph.D. Thesis, Department of Mathematics, Technion IIT, 1998.
[6] K. Binder, H.L. Frisch, J. Jackle, Kinetics of phase separation in the presence of slowly relaxing structural variables, J. Chem. Phys. 85
(1986) 15051512.
[7] S.M. Allen, J.W. Cahn, A microscopic theory for antiphase boundary motion and its application to antiphase domain coarsening, Acta
Metallurgica et Materiallia 27 (1979) 10851095.
[8] J. Rubinstein, P. Sternberg, J.B. Keller, Fast reaction, slow diffusion and curve shortening, SIAM J. Appl. Math. 49(1) (1989) 116133.
[9] P.C. Fife, Dynamics of internal layers and diffusive interfaces, Regional Conference Series in Applied Mathematics, 1988.
[10] M. Gage, R.S. Hamilton, The heat equation shrinking convex plane curves, J. Differential Geometry 23 (1986) 6996.
[11] M.A. Grayson, The heat equation shrinks embedded plane curves to round points, J. Differential Geometry 26 (1987) 285314.
[12] G.B. Whitham, Linear and Nonlinear Waves, Wiley/Interscience, New York, 1974.
[13] M. Born, L. Infeld, Foundations of a new eld theory, Proc. Roy. Soc. A 144 (1934) 425.
[14] P.L. Christiansen, N. Gronbech-Jensen, P.S. Lomdahl, B.A. Malomed, Oscillations of eccentric pulsons, Physica Scripta 55 (1997) 131134.
[15] M.R. Samuelsen, Approximate rotationally symmetric solutions to the sine-Gordon equation, Phys. Lett. A 74 (1978) 2122.
[16] R.L. Burden, J.D. Faires, Numerical Analysis, PWS Publishing Company, Boston, 1980.

Vous aimerez peut-être aussi