Supersonic Discrete Kink-Solitons and Sinusoidal Patterns with "Magic" wavenumber in Anharmonic Lattices
Extracted source text ยท arXiv cond-mat/0303463
\documentclass{epl}
\title{Supersonic Discrete Kink-Solitons and Sinusoidal Patterns with
``Magic" Wavenumber in Anharmonic Lattices}
\author{Yuriy A. Kosevich \inst{1} Ramaz Khomeriki \inst{2} \and Stefano Ruffo \inst{3,4}}
\institute{
\inst{1} Instituto de Investigacion en Comunicacion Optica, Universidad Autonoma de San
\\ Luis Potosi, Alvaro Obregon 64, 78000 San Luis Potosi, S.L.P., Mexico \\
\inst{2} Department of Physics, Tbilisi State University,
3 Cha\-vchava\-dze avenue \\ Tbilisi 380028, Republic of Georgia \\
\inst{3} Dipartimento di Energetica ``Sergio Stecco" and CSDC Universit\`a di Firenze,
\\Via S. Marta, 3, I-50139 Firenze, Italy \\
\inst{4} INFM and INFN Firenze, Italy
}
\pacs{05.45.-a}{Nonlinear dynamics and nonlinear dynamical systems}
\pacs{63.20.Pw}{Localized modes}
\begin{document}
\maketitle
\begin{abstract}
The sharp pulse method is applied to Fermi-Pasta-Ulam (FPU) and Lennard-Jones
(LJ) anharmonic lattices. Numerical simulations reveal the presence of high
energy strongly localized ``discrete'' kink-solitons (DK), which move with
supersonic velocities that are proportional to kink amplitudes. For small
amplitudes, the DK's of the FPU lattice reduce to the well-known ``continuous''
kink-soliton solutions of the modified Korteweg-de Vries equation. For high
amplitudes, we obtain a consistent description of these DK's in terms of
approximate solutions of the lattice equations that are obtained by restricting
to a bounded support in space exact solutions with sinusoidal pattern
characterized by the ``magic'' wavenumber $k=2\pi/3$. Relative displacement
patterns, velocity versus amplitude, dispersion relation and exponential tails
found in numerical simulations are shown to agree very well with analytical
predictions, for both FPU and LJ lattices.
\end{abstract}
\vspace{-0.6cm}
\section{Introduction}
The interest in studying moving nonlinear localized excitations (solitons,
kink-solitons, etc.) is mainly motivated by the fact that they can be
related to transport properties. In this respect, one
dimensional lattices of anharmonic oscillators can serve as a testing ground for
nonlinear transport.
In this Letter we present numerical and analytical studies of the dynamical
properties of strongly localized excitations of anharmonic lattices
with two-body interparticle potentials~\cite{fermi}. Invariance under the
symmetry transformation
that shifts the positions of all the oscillators by the same amount
relates anharmonic lattices with interparticle couplings to a
wide class of systems which share
such continuous symmetry, e.g. quasi-one dimensional easy plane ferromagnets
and antiferromagnets, ferrimagnetic systems with spiral structures
and even quantum Hall double layer pseudo-ferromagnets~\cite{magnets}.
The spontaneous breakdown of this translational symmetry leads to the apperance of
gapless Goldstone modes, which itself generates kink-soliton solutions in the
presence of nonlinearity.
Anomalous transport properties~\cite{anomalous} could appear for all systems in
this class. It is extremely important to assess whether such properties
could be related to the existence of moving localized excitations.
In the weakly nonlinear continuum limit and for gapless dispersion relations one
can prove~\cite{japan} that all localized solutions can be described by the
Korteweg-de Vries (KdV) or modified KdV (mKdV)
equations~\cite{zabusky,yuriy1,yuriy2,poggi0,ramaz2}.
On the lattice, small-amplitude solutions can be represented as KdV or mKdV solitons
in terms of relative displacements, while they are kinks for absolute displacements.
The energy of such objects is proportional to the kink amplitude and thus
lattice discreteness has to be taken into account when considering high
energy (i.e. strongly localized) objects. Since in this limit one cannot
use the ordinary continuum approximation, Rosenau \cite{rosenau1} has
recently proposed a theoretical approach that takes explicitly into account
lattice discreteness. Applying this approach to the
Fermi-Pasta-Ulam (FPU) lattice~\cite{fermi}, he derived, in the limit
of strong nonlinearity, a different
continuum equation, which supports a new kind of localized solution that
he called {\it compacton}~\cite{rosenau}.
On the other hand, the existence of approximate kink-like solutions of
the FPU lattice with characteristic sinusoidal displacement pattern with
the ``magic" wavenumber $k=2\pi/3$ has been predicted on the basis of a direct
analysis of the discrete lattice equations \cite{kosevich}.
This ansatz relies on the existence of exact extended propagating
nonlinear sinusoidal waves having this particular wavenumber~\cite{kosevich,poggi}.
Such patterns are also extremely successful in modelling stationary and
slowly-moving discrete breathers in the FPU lattice~\cite{kosevich,lepri1,yuriy3}.
The numerical study of moving strongly localized excitations of anharmonic
lattices has been so far limited to discrete breathers (DB)~\cite{flach} that
move with small velocity.
Two types of numerical methods have been commonly used to study the time
evolution of DB's: {\it i)} exact~\cite{flach,aubry} or approximate
\cite{sandusky,takeno} DB's are put initially onto the lattice and {\it ii)}
modulational
instability of zone-boundary and large-wavenumber modes is
induced~\cite{lepri1,yuriy3,burlakov,peyrard,ruffo,ullmann,mirnov}.
The first method allows one to investigate single
moving localized objects, but it assumes the knowledge of exact or approximate
solutions. By using the second method, a number of
localized modes is produced: they move, collide and merge, and it is
therefore hard to investigate each single moving object.
On the other hand, no mathematical
proof of existence of moving DB's has been obtained (see \cite{sepulchre} for
a discussion of this issue and a possible way out).
In this Letter we show a new numerical method to produce strongly localized
objects. This method is inspired by the one used in real experiments
to generate optical solitons in fibers as well as
magnetization envelope solitons in Yttrium Iron Garnet film
waveguides~\cite{yttrium}. A sharp pulse
is applied at one end of an unperturbed sample. Then, the system itself chooses
the stable
propagating localized mode and the only thing that is left is
to follow its motion.
We have first performed these numerical experiments
for an FPU chain with quadratic and quartic
interactions between neighbors, which describes purely
transverse vibrational excitations.
For large amplitudes, we observe supersonic strongly
localized discrete kink-solitons (DK) in which just three oscillators are
excited.
They show very low energy loss. We have repeated these experiments for more
realistic Lennard-Jones (LJ) interparticle potentials, also revealing the existence
of DK's.
Fast moving kinks have been observed in the Frenkel-Kontorova (FK)
model~\cite{braun,savin}. Moving kinks appear also in translationally invariant
lattices with Hertz potentials (see e.g. Ref.~\cite{rosas} and references therein),
but in this case there is no attractive force and the localized wave dynamics is
different from that of the FPU lattice.
\section{Sharp Pulse Method}
The equations of motion of the FPU-$\beta$ lattice~\cite{fermi} are
\begin{equation}
\ddot u_n=u_{n+1}+u_{n-1}-2u_n+(u_{n+1}-u_n)^3+(u_{n-1}-u_n)^3,
\label{2}
\end{equation}
where $u_n$ is the transverse displacement of the $n$-th oscillator from its equilibrium
position in dimensionless units.
In the numerical experiment we oscillate the first particle of the chain in the
following manner:
\begin{eqnarray}
u_1&=&A_0\cdot sin\bigl[\omega_0(t_0-t)\bigr] \qquad \mbox{if} \qquad 0<t\leq
t_0 \qquad \mbox{and} \nonumber \\ &&
u_1=0 \qquad \mbox{if} \qquad t>t_0,
\label{1}
\end{eqnarray}
while the right end $u_{N}$ is pinned. After applying the above pulse,
we pin both ends of the lattice.
The numerical experiment shows the existence of kinks that propagate with
large and slightly different velocities. Throughout this paper we will use
two different representations of the spatial profile: either we use absolute
lattice displacements $u_n$ or we show relative displacements
$v_n=u_{n+1}-u_n$. Hence the kinks are shown as true steps in the first
representation (see Fig. 1a) or as sharp peaks in the second representation
(see Fig. 1b).
As time evolves, the plateaus between the DK's widen but the shape of the
kinks does not change. During the collisions, the kink-solitons cross each
other without any appreciable energy loss (this is not shown here for FPU
lattices but later on for LJ lattices, see the right graph in Fig. 4).
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{figure}[b]
%h=here, t=top, b=bottom, p=separate figure page
\begin{center}\leavevmode
\includegraphics[height=5cm]{fpu1.eps}
\includegraphics[height=5cm]{fpu2.eps}
\end{center}
\caption{\label{fig1} a) Absolute $u_n$ and b) relative $v_n=u_{n+1}-u_{n}$
displacements versus the lattice site $n$ at time $t=20$ in dimensionless units.
In the $u$-space we observe a sequence of kinks with different amplitudes,
while in the $v$-space the localized objects have solitonic form of different
height. This simulation is performed with $A_0=11$, $\omega_0=\sqrt{2}$, $t_0=1$
and $N=200$.}
\end{figure}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
The existence of these localized objects is possible because of the peculiar
symmetry $u_n\rightarrow u_n+const$ of the interparticle potential. Systems with
such a symmetry have an infinitely degenerate ground state and its spontaneous
breakdown leads to the appearance of gapless Goldstone modes that determine the
formation of kinks.
Since the potential energy of the FPU chain is only a function of the relative
displacements $v_n=u_{n+1}-u_{n}$, these are the variables to be considered for
the appropriate physical picture. In these new variables the equations of motion
(\ref{2}) can be exactly rewritten as
\begin{equation}
\ddot v_n=v_{n+1}+v_{n-1}-2v_n+v_{n+1}^3+v_{n-1}^3-2v_n^3. \label{4}
\end{equation}
Eq. (\ref{4}) leads, in the continuum approximation and selecting one direction
for the wave motion, to the modified Korteweg-de Vries
(mKdV) equation if one keeps only the first terms in the two small parameters
$1/N$ and
amplitude~\cite{zabusky}. The mKdV equation supports exact soliton
solutions. Indeed, for small pulse amplitudes we could observe mKdV solitons.
However, as the driving amplitude $A_0$ of the pulse is increased we observe the
DK's of Fig. 1. Therefore, in terms of relative displacements, the objects that
we
observe in our simulations can be considered as discrete extension of the
solutions
of the mKdV equation.
The existence of fast moving objects with amplitude
independent short localization length was first considered in
Ref.~\cite{kosevich}. In the remaining of this Letter, we develop a
theoretical approach that, making use of this result allows us to describe
the main physical properties of the DK's we observe in our simulations.
\section{Approximate Analytical Description}
Exact extended sinusoidal wave solutions of the FPU (and other) lattices have been
recently found by several authors~\cite{kosevich,poggi,chechin}. One of such solutions
has the ``magic" $k=2\pi/3$ wavenumber and can be written in terms of relative
displacements
as $v_n=A\cos(kn-\omega t)$. It has been proposed that such exact solution can acquire
a compact support and maintain its validity as approximate solution~\cite{kosevich}.
For instance, in the case of stationary and slowly-moving discrete
breathers in the FPU-$\beta$
lattice, half a period truncations of the sinusoidal envelope of the displacement
patterns with $k_{e}=\pi - k=\pi/3$ wavenumber
well reproduce large amplitude localized oscillations~\cite{kosevich,lepri1,yuriy3}.
Following the same strategy, we here propose a similar truncation that turns out
to describe very well our supersonic DK's:
\begin{eqnarray}
v_n&=&\pm\frac{A}{2}\left[1+\cos\right(\frac{2\pi}{3}n-\omega t\left)\right]
\qquad \mbox{if} \qquad -\pi<\frac{2\pi}{3}n- \omega t<\pi;
\label{111}
\end{eqnarray}
and $v_n=0$ otherwise. In the large amplitude limit we can neglect
short exponential tails in the sinusoidal pattern (\ref{111}) (see, however, Eq. (6)
below). Substituting this ansatz into Eq~(\ref{4})
and performing the rotating wave approximation \cite{rwa} for the
second harmonic, we obtain the following relations for the frequency
$\omega$ and velocity $V$ as the functions of the amplitude $A$ of the DK's.
\begin{equation}
\omega=\sqrt{3+(45/16)A^2}; \qquad V=\omega/(2\pi/3)
=3\sqrt{3+(45/16)A^2}/(2\pi). \label{112}
\end{equation}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{figure}[t]
%h=here, t=top, b=bottom, p=separate figure page
\begin{center}\leavevmode
\includegraphics[height=5cm]{fpu31.eps}
\includegraphics[height=5cm]{fpu41.eps}
\end{center}
\caption{\label{fig2} Left graph: Velocity of the localized discrete
kink-solitons versus amplitude of relative displacements. Open circles represent
the result of numerical experiments. The solid line corresponds to
the analytical formula (\ref{112}), which follows from the ``magic" wavenumber
solution. The dotted horizontal line
indicates the maximal group velocity of linear waves, which equals 1.
Right graph: Relative displacements of a given oscillator versus time. Open
circles are the results of numerical measurements, the
solid line corresponds to formula (\ref{111}).}
\end{figure}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
First of all, we can observe that the analytical solution of formula (\ref{111})
is consistent with the numerical observation that three
lattice sites are always excited (see Fig.1). Besides that, we have
numerically checked the dependence of velocity on amplitude
(see left graph in Fig. 2) and the time dependence of the relative
displacement of a given oscillator (see right graph in Fig. 2).
Our theoretical expression gives a very good agreement with all numerical data.
Furthermore, the right graph of Fig. 2 shows that the relative displacement
patterns of both the measured and the analytical solution has everywhere a
smooth envelope. Therefore, we can conclude that the compacton with sharp
cosine-like support and with
non-smooth relative displacement envelope, described in Refs.
~\cite{rosenau1,rosenau}, doesn't compare well with our numerical results.
We'll come back to this and related issues in a forthcoming
publication~\cite{newus}.
In relation to this latter comment, it is important to remark that our DK's have
a localized bulk sided by exponential tails (see Fig. 3a). The decay length
$\Lambda$ is a function of the velocity, as shown in Fig. 3b. This dependence
can be derived analytically representing the relative displacements in the tails as
$v_n\sim \exp{\left[\pm(n-Vt)/\Lambda\right]}$ and substituting this ansatz into
equations (\ref{4}). Considering the limit $n-Vt\gg \Lambda$, one gets the
relation~\cite{kosevich}:
\begin{equation}
V^2=4\Lambda^2\sinh^2(1/2\Lambda). \label{tail}
\end{equation}
As follows from Eq.~(6), the decay length $\Lambda$ is real positive only for
supersonic excitations with $V>1$ and it diverges for $V=1$. Again, the agreement with
numerical data, shown in Fig. 3b, is very good.
To check the stability of the approximate solution (\ref{111}), we directly put it on
the unperturbed lattice, imposing to the kink-soliton the initial velocity given by
formula (\ref{112}). In several simulations we observe the long-living motion
of a slightly modified kink-soliton. Since in our
experiments the boundaries are pinned, we observe edge reflection. Both at edge
reflections and during collisions, the kink-solitons do not significantly
change their amplitudes and shapes.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{figure}[t]
%h=here, t=top, b=bottom, p=separate figure page
\begin{center}\leavevmode
\includegraphics[height=4.5cm]{fputailsfit.eps}
\includegraphics[height=4.5cm]{fputailsfit1.eps}
\end{center}
\caption{\label{fig3} a) Logarithmic plot of the relative displacement
pattern of a single DK which moves from the left to the right. We use
the right exponential decay, which is not affected by radiation, to
fit the decay length $\Lambda$. b) Dependence of the decay length, $\Lambda$,
upon the velocity of the DK. The solid line is obtained from formula (\ref{tail}) and
the points are results of numerical experiments. The dashed line shows the
velocity, $V=1$, at which the decay length diverges.}
\end{figure}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\section{Realistic Potentials}
In order to check the generality of our results we have performed a similar
analytical and numerical study of the Lennard-Jones (LJ) interparticle potential
$V(r)=4e\left[\left(\frac{s}{r}\right)^{12}-\left(\frac{s}{r}\right)^6\right]$,
where $r$ is a distance between neighboring particles, $e$ and $s$ are constants.
After the appropriate rescalings of space and time:
$r\rightarrow (2)^{1/6}s\cdot r$ and ~~$t\rightarrow t\sqrt{m/12e}$,
we obtain the following equations of motion for the relative displacements:
\begin{equation}
\frac{d^2v_n}{dt^2}=\frac{2}{(1+v_n)^{13}}- \frac{1}{(1+v_{n-1})^{13}} -\frac{1}{(1+v_{n+1})^{13}}-
\frac{2}{(1+v_n)^{7}}+ \frac{1}{(1+v_{n-1})^{7}} +\frac{1}{(1+v_{n+1})^{7}}. \label{lennard}
\end{equation}
The dispersion relation for linear waves can be derived by substituting the pattern
$v_n=Acos(kn-\omega t)$ in Eq.~(\ref{lennard}). In the $A\rightarrow 0$ limit
we get the value of maximal group velocity of linear waves
$v_{max}=\sqrt{6}$
(see dashed line in the left panel of Fig. 4).
Seeking for strongly localized solutions, we substitute again the ``magic"
wavenumber sinusoidal pattern (\ref{111}) into the equations of motion (\ref{lennard}).
We choose the minus sign in the relative displacement pattern (\ref{111}),
which corresponds to the conditions under which we observe supersonic
kinks in LJ lattices. Then, we expand the r.h.s. of the equations of motion in
powers of $(A/2)\cos(2\pi n/3-\omega t)$, keeping terms
up to fifth order. To treat the powers of cosines we again perform
the rotating wave approximation~ \cite{page}.
We don't report here the formula that we obtain for the velocity-amplitude
relation to save space, but its graph is plotted in the left panel in Fig. 4
as a solid line.
The agreement between analytics and numerics is quite satisfactory,
confirming that the ``magic" wavenumber ansatz works also for LJ
potentials. We have also investigated numerically the DK's interaction process.
As seen from the right panel of Fig. 4, kink-solitons retain their shapes and
velocities after interaction. The effect of the interaction manifests itself
only in
tiny trajectory shifts, as it is expected from the weakly nonlinear continuum
approximation (see e.g. \cite{ramaz2}).
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{figure}[t]
%h=here, t=top, b=bottom, p=separate figure page
\begin{center}\leavevmode
\includegraphics[height=4.5cm]{lennard1.eps}
\includegraphics[height=12cm]{lennard2.eps}
\end{center}
\caption{\label{fig33} Left Graph: ``Magic" wavenumber ansatz for the
discrete kink-soliton velocity
versus amplitude for a LJ interparticle potential. The analytical result
(solid line) compares extremely well with numerics (open circles) .
The dashed horizontal line indicates the maximal group velocity of linear waves
(equal to $\sqrt{6}$).
Right graph: Interaction of the DK's created at the opposite ends of the
chain.
The strongly localized DK's retain their shapes after interaction.}
\end{figure}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\section{Conclusions}
Summarizing, we have used the sharp pulse method to generate large amplitude,
supersonic discrete kink-solitons with amplitude independent short localization
length.
These objects reduce to the well known kink-soliton solutions of the mKdV
equation for small
amplitudes. We have shown that large amplitude kink-solitons
are characterized by a sinusoidal pattern with the ``magic" wavenumber $k=2\pi/3$,
and propagate with supersonic velocities increasing with the amplitude.
Both the FPU and the more realistic LJ interparticle potential have
been considered. The agreement between the lattice kink theory and the numerical
experiments for both types of anharmonic lattices is so good that it encourages
us to draw
the general conclusion that strongly localized discrete kink-solitons
are common for systems with two-body interparticle potentials.
\acknowledgments We thank O.M. Braun for sending us his papers on the
Frenkel-Kontorova model
and P. Rosenau for useful correspondence and for sending us
his paper before publication.
This work was supported by the NATO expert visit award No PST.EV.979337 and the
LOCNET European Commission RTN HPRN-CT-1999-00163.R.Kh. is supported by USA CRDF award
No GP2-2311-TB-02 and NATO reintegration grant No FEL.RIG.980767.
\begin{thebibliography}{0}
\bibitem{fermi} E. Fermi, J. Pasta, S. Ulam, M. Tsingou, in {\it The Many-Body
Problems}, edited by D.C. Mattis (World Scientific, Singapore, 1993 reprinted).
\bibitem{magnets} I.B. Spielman, J.P. Eisenshtein, L.N. Pfeiffer, K.W. West, Phys.
Rev. Lett., {\bf 87}, 036803 (2001).
\bibitem{anomalous} S. Lepri, R. Livi and A.Politi, Phys. Rep., {\bf 377 }, 1 (2003).
\bibitem{japan} T. Taniuti, Prog. Theor. Phys. Suppl., {\bf 55}, 1, (1974).
\bibitem{zabusky} N.J. Zabusky, M.D. Kruskal, Phys. Rev. Lett., {\bf 15}, 240,
(1965).
\bibitem{yuriy1}Yu. A. Kosevich, Phys. Rev. B, {\bf 47}, 3138, (1993); Phys. Rev. B,
{\bf 48}, 3580, (1993).
\bibitem{yuriy2}Yu. A. Kosevich, Phys. Lett. A, {\bf 173}, 257, (1993).
\bibitem{poggi0} P. Poggi, S. Ruffo, and H. Kantz, Phys. Rev. E, {\bf 52}, 307, (1995).
\bibitem{ramaz2} R. Khomeriki, Phys. Rev. E, {\bf 65}, 026605, (2002).
\bibitem{rosenau1} P. Rosenau, Phys. Lett. A, {\bf 311}, 39, (2003).
\bibitem{rosenau} P. Rosenau, J.M. Hyman, Phys. Rev. Lett., {\bf 70},564,(1993);
P. Rosenau, Phys. Rev. Lett., {\bf 73}, 1737, (1994).
\bibitem{kosevich} Yu. A. Kosevich, Phys. Rev. Lett., {\bf 71}, 2058, (1993).
\bibitem{poggi} P. Poggi, S. Ruffo, Physica D, {\bf 103}, 251, (1997).
\bibitem{lepri1} Yu. A. Kosevich, S. Lepri, Phys. Rev. B, {\bf 61}, 299,
(2000).
\bibitem{yuriy3} Yu. A. Kosevich, G. Corso, Physica D, {\bf 170}, 1, (2002).
\bibitem{flach} S. Flach, C.R. Willis, Phys. Rep., {\bf 295}, 181, (1998).
\bibitem{aubry} S. Aubry, Physica D, {\bf 103}, 201, (1997).
\bibitem{sandusky} K.W. Sandusky, J.B. Page, K.E. Schmidt, Phys. Rev. B., {\bf
46}, 6161, (1992);
\bibitem{takeno} S. Takeno, K. Hori, J. Phys. Soc. Japan, {\bf 59}, 3037,
(1990).
\bibitem{burlakov} V.M. Burlakov, S.A. Kiselev, V.I. Rupasov, Phys. Lett. A.,
{\bf 147}, 130, (1990).
\bibitem{peyrard} I. Daumont, T. Dauxois, M. Peyrard, Nonlinearity, {\bf 10},
617, (1997).
\bibitem{ruffo} T. Cretegny, T. Dauxois, S. Ruffo, A. Torcini, Physica D, {\bf
121}, 109, (1998).
\bibitem{ullmann} K. Ullmann, A.J. Lichtenberg, G. Corso, Phys. Rev. E., {\bf
61}, 2471, (2000).
\bibitem{mirnov} V.V. Mirnov, A.J. Lichtenberg and H. Guclu, Physica D, {\bf
157}, 251, (2001).
\bibitem{sepulchre} R.S. Mackay and J.-A. Sepulchre, J.Phys A: Math. and Gen.,
{\bf 35}, 3985 (2002).
\bibitem{yttrium} P. De Gasperis, R. Marcelli and G. Miccoli, Phys. Rev. Lett.
{\bf 59}, 481, (1987).
\bibitem{braun} O.M. Braun, Yu.A. Kivshar, Phys. Rep, {\bf 306}, 1, (1998).
\bibitem{savin} A.V. Savin, Zh. Eksp. Teor. Fiz., {\bf 108}, 1105, (1995).
\bibitem{rosas} A. Rosas, K. Lindenberg, Phys. Rev. E., {\bf 68}, 041304,
(2003) and cond-mat/0308296.
\bibitem{chechin} G.M. Chechin, N.V. Novikova and A.A. Abramenko, Physica D,
{\bf 166}, 208 (2002); B. Rink, Physica D, {\bf 175}, 31 (2003), S. Shinohara,
J. Phys. Soc. Jpn., {\bf 71}, 1802 (2002).
\bibitem{rwa} S.Takeno, K. Kisoda, A.J. Sievers, Prog. Theor. Phys. Suppl., {\bf
94}, 242, (1988).
\bibitem{page} J.B. Page, Phys. Rev. B,{\bf 41}, 7835, (1990).
\bibitem{newus} Y.A. Kosevich, R. Khomeriki and S. Ruffo, in preparation.
%\bibitem{flach7} S. Flach, Phys. Rev. E, {\bf 50}, 3134, (1994).
\end{thebibliography}
\end{document}