1: \documentclass[preprint,prl,aps,showkeys,showpacs]{revtex4}
2: %\documentclass[prl,aps,twocolumn,showkeys,showpacs]{revtex4}
3: \begin{document}
4: \input{epsf.tex}
5: \epsfverbosetrue
6: \title{Bifurcations from stationary to pulsating solitons
7: in the cubic-quintic complex Ginzburg-Landau equation}
8: \author{Eduard N. Tsoy\footnote{Also at: Physical-Technical
9: Institute of the Uzbek Academy of Sciences,
10: Tashkent, Uzbekistan} and Nail Akhmediev}
11: \address{ Optical Sciences Group, Research School of Physical
12: Sciences and Engineering, \\
13: The Australian National University, Canberra, ACT 0200, Australia}
14: \begin{abstract}
15: Stationary to pulsating soliton bifurcation analysis
16: of the complex Ginzburg-Landau equation (CGLE) is presented.
17: The analysis is based
18: on a reduction from an infinite-dimensional dynamical dissipative system
19: to a finite-dimensional model. Stationary solitons, with constant amplitude
20: and width, are associated with fixed points in the model. For the first time,
21: {\em pulsating} solitons are shown to be stable {\em limit cycles} in the
22: finite-dimensional dynamical system. The boundaries between the two types
23: of solutions are obtained approximately from the reduced model.
24: These boundaries are reasonably close to those predicted by direct numerical
25: simulations of the CGLE.
26: \end{abstract}
27: \pacs{ 04.30.Nk Wave propagation and interactions; 05.45.Yv Solitons;
28: 42.65.Sf Dynamics of nonlinear optical systems; 42.65.Tg Optical solitons}
29: \maketitle
30: The complex Ginzburg-Landau equation (CGLE) is one of the basic
31: equations for modelling modulated amplitude waves \cite{Brusch2001},
32: spatio-temporal dynamics and spontaneous development of coherent
33: structures in a variety of nonlinear dissipative
34: systems~\cite{Reviews,Books}. Examples include pulse generation
35: by passively mode-locked soliton lasers \cite{Ippen94},
36: signal transmission in all-optical communication lines \cite{bakonyi_2},
37: travelling waves in binary fluid mixtures \cite{Kolodner},
38: and also pattern formation in many other physical systems
39: \cite{Gross}. Complicated patterns consist of simpler localized
40: solutions like fronts, pulses, sources and sinks \cite{saarlos92}.
41:
42: Pulsating soliton solutions of dissipative systems have attracted a
43: great deal of attention in recent years. They have been found
44: numerically \cite{Deiss94,AST01,SGGA04} and observed experimentally
45: \cite{SGGA04} in a fiber laser. Pulsating solitons form one set of
46: possible localized solutions of the CGLE, and they exist on an equal
47: basis with stationary solitons. Such localized waves exist, in
48: various forms in biology, chemistry and physics.
49:
50: A pulsating soliton can be be described as a limit cycle of an
51: infinite-dimensional dissipative dynamical system \cite{Springerbook}.
52: It is different from the higher-order solitons that are usually
53: connected with an integrable model \cite{Satsuma74}.
54: Although numerical simulations show clearly the existence of
55: pulsating solutions and their bifurcations from stationary solitons,
56: so far there has been no progress in finding analytic expressions for
57: pulsating solutions and bifurcation boundaries. The problem is not
58: simple as there are several parameters of the CGLE that define
59: the regions of existence for both stationary and pulsating solitons.
60: Hence, the bifurcation boundaries are surfaces in this multi-dimensional
61: space of the parameters.
62:
63: In this work we use a reduction from an infinite-dimensional to a
64: five-dimensional model, and we aim to find localized solutions of the
65: CGLE and the transformations that they are subjected to when the
66: system parameters are varied. Although exact solutions of the CGLE do
67: exist \cite{Books}, they can be presented explicitly only for certain
68: relations between the parameters of the equation. Furthermore, only
69: stationary solutions can be found. Hence, we are faced with the
70: necessity of finding an efficient approximation to tackle the problem. We have
71: found that the method of moments, originally developed by Maimistov
72: \cite{Maim93} for the perturbed nonlinear Schr\"{o}dinger equation
73: (NLSE) can be used for solving our problem. The moments are the integral
74: characteristics of the field under consideration. In principle, there
75: are an infinite number of equations for moments. One can obtain exact results
76: by using the complete set of these equations. However, in practice, one
77: uses a trial function with a finite number of parameters, and this is the
78: way to obtain a significant reduction in the number of variables used for the
79: description of the dynamics.
80:
81: The cubic-quintic complex Ginzburg-Landau equation, in dimensionless
82: form, is written as
83: %
84: \begin{eqnarray}
85: i \psi_t + {D \over 2} \psi_{xx} + |\psi|^2 \psi =
86: -\nu |\psi|^4 \psi +
87: \nonumber \\
88: i \delta \psi + i \epsilon |\psi|^2 \psi +
89: i \beta \psi_{xx} + i \mu |\psi|^4 \psi \equiv R[\psi] \ ,
90: \label{CGLE}
91: \end{eqnarray}
92: %
93: where $\psi(x,t)$ is the normalized envelope of the field, $t$ and $x$
94: are time and spatial variables, respectively, $D$ is the group
95: velocity dispersion coefficient, $\nu$ is the parameter of the quintic
96: nonlinearity, $\delta$ represents the linear loss, $\epsilon$ is the
97: nonlinear gain coefficient, $\beta$ stands for the spectral filtering,
98: and $\mu$ characterizes the saturation of the nonlinear gain. Stable soliton
99: solutions of the CGLE exist only for the following choices for the signs of
100: the coefficients: $\delta, \mu < 0$, $\beta, \epsilon > 0$, and any
101: sign for $\nu$ and $D$ (see e.g.~\cite{Books,AST01}). Hence, in this work,
102: we limit ourselves only to this range.
103:
104: The method of moments \cite{Maim93} is a reduction of the complete
105: problem of the evolution of a field that has an infinite number of degrees
106: of freedom to the evolution of a finite set of pulse
107: characteristics. For a localized solution with a single maximum, these
108: include the peak amplitude, pulse width and center-of-mass position.
109: For an arbitrary localized field, the two integrals, namely
110: the energy $Q$ and momentum $P$
111: %
112: \begin{eqnarray}
113: Q= \int_{-\infty}^{\infty}{ |\psi|^2 dx}\ , \quad
114: %\nonumber \\
115: P= {1 \over 2} \int_{-\infty}^{\infty}
116: { (\psi \psi_{x}^{\ast} - \psi^{\ast} \psi_{x}) dx}.
117: \label{EM}
118: \end{eqnarray}
119: are two basic variables evolving with $t$.
120: %
121: Three higher-order generalized moments, related to the pulse, are given
122: by the following expressions \cite{Maim93}:
123: %
124: \begin{eqnarray}
125: I_1 &=& \int_{-\infty}^{\infty}{x |\psi|^2 dx}\ , \quad
126: I_2 = \int_{-\infty}^{\infty}{(x-x_0)^2 |\psi|^2 dx}\ ,
127: \nonumber \\
128: I_3 &=& \int_{-\infty}^{\infty}{(x-x_0)
129: (\psi^{\ast} \psi_{x} - \psi \psi_{x}^{\ast}) dx}\ ,
130: \label{Integr}
131: \end{eqnarray}
132: %
133: where $x_0(t)= I_1/Q$.
134: Using Eq.~(\ref{CGLE}) one can derive the evolution equations
135: for the generalized moments. For the five integrals given above, one
136: can obtain the following \cite{Maim93}:
137: %
138: \begin{eqnarray}
139: && {d Q \over d t} = i \int_{-\infty}^{\infty}
140: {(\psi R^{\ast} - \psi^{\ast} R) dx}\ ,
141: \nonumber \\
142: && {d P \over d t} = -i \int_{-\infty}^{\infty}
143: {(\psi_x R^{\ast} + \psi_x^{\ast} R) dx}\ ,
144: \nonumber \\
145: && {d I_1 \over d t} = i D P + i \int_{-\infty}^{\infty}
146: {x (\psi R^{\ast} - \psi^{\ast} R) dx}\ ,
147: \label{ODEs} \\
148: && {d I_2 \over d t} = -i D I_3 + i \int_{-\infty}^{\infty}
149: {(x-x_0)^2 (\psi R^{\ast} - \psi^{\ast} R) dx}\ ,
150: \nonumber \\
151: && {d I_3 \over d t} = 2 P {d x_0 \over d t} +
152: i \int_{-\infty}^{\infty} { (2 D |\psi_x|^2 - |\psi|^4) dx } +
153: \nonumber \\
154: && 2 i \int_{-\infty}^{\infty} { (x-x_0)
155: (\psi_x R^{\ast} + \psi_x^{\ast} R)} +
156: i \int_{-\infty}^{\infty}
157: {(\psi R^{\ast} + \psi^{\ast} R) dx}\ .
158: \nonumber
159: \end{eqnarray}
160: %
161: Equations~(\ref{ODEs}) are general in the sense that they are
162: valid for a large class of NLSE-type evolution equations, including
163: Eq.~(\ref{CGLE}) with arbitrary coefficients as a particular case.
164: If we use an exact solution of Eq.~(\ref{CGLE}), $\psi$, then
165: equations~(\ref{ODEs}) are exact.
166:
167: For problems of a certain class, even the first two Eqs.~(\ref{ODEs})
168: may be sufficient when one deals with exact two-parameter reductions of the
169: CGLE solutions \cite{JOSA98}. In Ref.~\cite{JOSA98}, such an approach
170: was used to find the CGLE solutions in the form of stable soliton pairs
171: and trains. The method of moments has also been applied to find
172: stationary solutions of the CGLE~(\ref{CGLE}) in Ref.~\cite{Ostrov04}
173: although there was no attempt at finding pulsating solitons. Our
174: tests showed that five is the minimum number of moments
175: needed to describe pulsating solitons using the reduced system.
176: Having more of them may improve the accuracy, but the complexity of
177: the analysis then increases dramatically.
178:
179: The choice of the trial function is crucial for obtaining solutions with
180: the desired properties. Any reduction from an infinite-dimensional to
181: a finite-dimensional system will have deficiencies. In approaches like
182: this, the choice of the trial function can only be justified at the
183: last stage of analysis, when the approximate solutions are compared with
184: numerical simulations of the original equation. We use the fact that
185: soliton solutions remain localized even when they are pulsating.
186: Therefore, we take the sech-function:
187: \begin{equation}
188: \psi(x,t)= A\; \mbox{sech} \left( {x- x_0 \over w} \right)
189: e^{i[b (x-x_0) + c (x-x_0)^2] }\ ,
190: \label{Ansatz}
191: \end{equation}
192: where $A(t)$, $w(t)$ and $x_0(t)$ are the amplitude, width and maximum
193: position of the pulse, respectively,
194: $b(t)$ is the soliton velocity and $c(t)$ is the chirp
195: parameter. The chirp is highly important, as the numerical
196: simulations \cite{AST01,SGGA04} show. The number of parameters
197: in the trial function must correspond to the number of moments
198: used in the set of equations (\ref{ODEs}). More complicated trial functions
199: need more equations in (\ref{ODEs}).
200:
201: Now, the generalized moments can be expressed in terms of
202: the variable parameters of the trial function. Evaluation of the
203: integrals~(\ref{EM}) and (\ref{Integr}), with a help of
204: Eq.~(\ref{Ansatz}), gives the following expressions:
205: %
206: \begin{eqnarray}
207: Q= 2 A^2 w, \quad P= -2 i A^2 w b, \quad I_1= 2 A^2 w x_0,
208: \nonumber \\
209: I_2= (\pi^2/6) A^2 w^3, \quad I_3= i (2 \pi^2/3) A^2 w^3 c.
210: \label{Values}
211: \end{eqnarray}
212: %
213: Then, using Eqs.~(\ref{ODEs}), one can obtain a set of ordinary
214: differential equations for the soliton parameters
215: $Q,w,c,x_{0}$ and $b$:
216: %
217: \begin{eqnarray}
218: Q_t &=& F_1 \equiv 2 Q \nonumber \\
219: & & \times \left[ \delta - \beta b^2 + {\epsilon \over 3}
220: {Q \over w} + {2 \mu Q^2 - 5 \beta \over 15 w^2}
221: - {\pi^2 \over 3} \beta c^2 w^2 \right] ,
222: \nonumber \\
223: w_t &=& F_2 \equiv -{2 \epsilon \over \pi^2} Q + {8 \beta-\mu Q^2 \over
224: \pi^2 w} + 2 D c w -
225: {16 \pi^2 \over 15} \beta c^2 w^3 ,
226: \nonumber \\
227: c_t &=& F_3 \equiv -2 D c^2 - {1 \over \pi^2 w^2} \label{Model} \\
228: & & \times \left[
229: 4 \left( {\pi^2 \over 3} + 1 \right) \beta c +
230: {Q \over w}
231: + {8 \nu Q^2 -30 D \over 15 w^2} \right] , \nonumber\\
232: \nonumber \\
233: x_{0,t} &=& F_4 \equiv b (D - {2 \pi^2 \over 3} \beta c w^2) ,
234: \nonumber \\
235: b_t &=& F_5 \equiv - {4 \over 3} \beta ( {1 \over w^2} + \pi^2 c^2 w^2 )\,
236: b \ .
237: \nonumber
238: \end{eqnarray}
239:
240: Fixed points (FPs) of the dynamical system~(\ref{Model}) can be found from the
241: set of algebraic equations $F_j =0,\ j=1, \dots,5$. The stability of
242: the FPs is determined from the analysis of eigenvalues $\lambda_j$,
243: $j=1, \dots,5$, of the Jacobian matrix $ M_{ij} = \partial F_i /
244: \partial p_j $, where $\{p_1, \dots, p_5 \} \equiv \{ Q, w, c, x_0, b
245: \}$, and $i= 1, \dots, 5$. If the real part of at least one eigenvalue
246: is positive, then the corresponding fixed point is unstable. In
247: principle, the whole five-dimensional dynamical system (\ref{Model})
248: can be studied using the specialized software \cite{AUTO2000}. We analyse
249: the system based on general theory and using further simplifications.
250:
251: Firstly, we consider solutions with $b=0$ in Eqs. (\ref{Model}).
252: The real part of the eigenvalue which corresponds to $b$ is negative for
253: any set of the system parameters. The soliton center, $x_0(t)$, takes
254: a constant value for $t \to \infty$ and the real part of the
255: corresponding eigenvalue is zero (neutral stability). Therefore, we
256: can make a further reduction and consider a system with only three
257: variables, viz. $Q, w$, and $c$. We denote the three
258: corresponding eigenvalues by $\lambda_1$, $\lambda_2$ and
259: $\lambda_3$. Note that, since the characteristic equation for these
260: eigenvalues is cubic, then either $\lambda_1= \lambda_2^{\ast}$ and
261: $\lambda_3$ is real, or all three $\lambda_i$ are real.
262:
263: We find FPs numerically by solving the algebraic equations, and
264: calculate the characteristic eigenvalues by analyzing the Jacobian
265: matrix. In this
266: way, we have identified the regions where solutions have distinctive
267: features. The bifurcation diagram for the dynamical system ~(\ref{Model}) in
268: the ($\nu, \epsilon$)-plane is presented in Fig.~\ref{Fig:ne1}.
269: We take $D=1$ and $\delta= -0.1$ henceforth in the paper.
270: To facilitate comparison with exact results, the parameters of the model
271: are chosen to be the same as in the numerical simulations of
272: Ref.~\cite{AST01}, namely $\mu= - 0.1$ and $\beta=0.08$.
273:
274: When the value of the gain $\epsilon$ is small, there are no stable or
275: unstable FPs (i. e. no stationary solitons) in the system. This
276: region is located below $\epsilon=0.6$, and therefore it is not shown in
277: Fig.~\ref{Fig:ne1}. For moderate values of $\epsilon$,
278: there are two FPs. In the region below the curve $1$ in Fig.\ref{Fig:ne1}a,
279: one FP is stable while the other one is unstable.
280: The signs of the real and imaginary parts of the eigenvalues,
281: for the stable FP in this region are the following:
282: $\mbox{Re}[\lambda_{1,2}] <0$,
283: $\mbox{Im}[\lambda_{1}] = -\mbox{Im}[\lambda_{2}]$,
284: $\mbox{Re}[\lambda_{3}] <0$, and $\mbox{Im}[\lambda_{3}] =0$.
285: The stable FP corresponds to a stationary CGLE soliton with
286: constant soliton parameters $p_j$.
287:
288: % Figure 1
289: \begin{figure}[c]
290: \centering
291: \epsfxsize=8.5cm \mbox{\epsffile{fig1.eps}}
292: \caption{(a)
293: Regions of existence and stability of FPs and limit cycles (LCs) of the
294: reduced system in ($\nu, \epsilon$)-plane. The central region, between
295: the two solid lines $1$ and $2$, corresponds to stable LC. (b) Regions
296: of existence for various soliton solutions obtained from
297: numerical simulations of CGLE (\ref{CGLE}).
298: The region for pulsating solitons, found numerically in Ref.~\protect
299: \cite{AST01}, is copied from (b) to (a) for comparison.
300: The parameters of the dynamical system are shown in (b).} \label{Fig:ne1}
301: \end{figure}
302:
303: Curve $1$ in Fig.\ref{Fig:ne1}a is the bifurcation boundary
304: (threshold) where the stable fixed point turns into an unstable one.
305: Above curve $1$, $\mbox{Re}[\lambda_{1,2}] >0$,
306: $\mbox{Im}[\lambda_{1}] = -\mbox{Im}[\lambda_{2}]$,
307: $\mbox{Re}[\lambda_{3}] <0$ and $\mbox{Im}[\lambda_{3}] =0$.
308: On line $1$, a stable FP is transformed into a stable limit
309: cycle (LC) and an unstable FP. We confirmed, numerically, that the
310: {\em stable} limit cycle of model~(\ref{Model}) does indeed exist
311: between the solid curves $1$ and $2$. Two examples of LC in
312: ($Q, w, c$)-space are shown in Fig.~\ref{Fig:attr1}a and \ref{Fig:attr1}b.
313: The limit cycle has zero size on line $1$ in Fig.\ref{Fig:ne1}a. This corresponds
314: to a super-critcal Hopf bifurcation \cite{Guckenheimer}.
315: The limit cycle elongates in $Q$ axis direction when $\epsilon$
316: increases. It becomes infinitely large and breaks off on line $2$ in
317: Fig.\ref{Fig:ne1}a.
318:
319: The smaller region in Fig.\ref{Fig:ne1}a surrounded by the light solid curve, is
320: copied from Fig.\ref{Fig:ne1}b for comparison. This is the
321: boundary for the existence of pulsating solitons found by direct numerical
322: simulations of CGLE (\ref{CGLE}). One can see that the thresholds (solid
323: curves $1$ and $2$) of the existence of the stable LC provide reasonably good
324: estimates for the boundaries.
325:
326: The period of oscillations, $T$, depends on the parameters of the system.
327: A plot of $T$ versus $\epsilon$ at constant
328: $\nu= - 0.09$ is given in Fig.\ref{Fig:front}.
329: The upper curve represents the results of exact numerical simulations
330: of the CGLE, while the lower curve shows the period of oscillations
331: found from our finite-dimensional model.
332: There is an apparent difference in the values found for the period,
333: due to the drastic reduction in the number of degrees of freedom in the model.
334: However, the two curves have the same qualitative behaviour.
335: In particular, each curve starts with a finite value of the period $T$
336: at the lower boundary of the region where pulsating solitons
337: exist, and increases to infinity at the upper boundary.
338:
339: Above the curve $2$,
340: the soliton amplitude $A= [Q/(2w)]^{1/2}$ remains almost
341: constant, but the soliton energy $Q$ and the width $w$ increase
342: monotonically with $t$. This motion is tantamount to the localized
343: solution of CGLE with constant amplitude and the width
344: that increases indefinitely. The final stage of this motion is an
345: asymptotic transformation of the soliton into two separating fronts
346: with constant velocities.
347: Thus, the qualitative agreement above the line $2$ is also
348: fairly good and the main features of the
349: dynamics are captured correctly.
350:
351: % Figure 2
352: \begin{figure}[c]
353: \centering
354: \epsfxsize=8.5cm \mbox{\epsffile{fig2.eps}}
355: \caption{Two examples of stable limit cycle of the system (\ref{Model}) in
356: ($Q, w, c$)-space. The parameters are shown in (a) and (b)
357: respectively. Note the different scales for the axes in (a) and (b).}
358: \label{Fig:attr1}
359: \end{figure}
360:
361: Another slice of the parameter space, namely the ($\mu, \epsilon$)-plane,
362: is shown in Fig.~\ref{Fig:me1}a. As in Fig.~\ref{Fig:ne1}a, there are
363: two FPs in the region presented. One of them is always unstable. The
364: stable FP exists below the solid curve $1$. The bifurcation on the
365: curve $1$ is the same as that in Fig.~\ref{Fig:ne1}a. The solid curve
366: $2$ corresponds to the transition between limit cycles and the
367: solutions with increasing width. From Fig.~\ref{Fig:ne1} and
368: Fig.~\ref{Fig:me1}, one can see that the results obtained from the
369: model~(\ref{Model}) for pulsating solitons and the boundaries for their
370: existence are in reasonable agreement with direct numerical simulations
371: of the CGLE over a wide range of the system parameters.
372:
373: % Figure 3
374: \begin{figure}[c]
375: \centering
376: \epsfxsize=6cm \mbox{\epsffile{fig3.eps}}
377: \caption{Period of pulsations $T$ versus $\epsilon$.
378: Upper curve correspond to exact numerical results while the lower
379: curve describes the data obtained from the low-dimensional model.}
380: \label{Fig:front}
381: \end{figure}
382:
383: Thus, considering the evolution of the dissipative soliton profile, the
384: model~(\ref{Model}) is able to predict: (a) stable and unstable
385: stationary solitons; (b) periodic soliton pulsations; (c) the unlimited
386: increase in the soliton width at constant amplitude (which is equivalent to
387: splitting solitons into moving fronts); and (d) bifurcations between
388: these dynamical behaviours.
389:
390: % Figure 4
391: \begin{figure}[c]
392: \centering
393: \epsfxsize=8.5cm \mbox{\epsffile{fig4.eps}}
394: \caption{
395: (a) Diagram of existence and stability of FPs and LCs of the
396: reduced model in the ($\mu, \epsilon$)-plane. Stable LCs
397: exist between the curves $1$ and $2$.
398: (b) Existence diagram based on numerical simulations of CGLE
399: \protect \cite{AST01}. The solid curve bounding the region for
400: pulsating solitons in (b) is copied to (a) for comparison.
401: The parameters of the dynamical system are shown in (b).}
402: \label{Fig:me1}\end{figure}
403:
404: In summary, we have derived a finite-dimensional dynamical system
405: associated with the CGLE. We have demonstrated that the pulsating
406: solitons of the CGLE correspond to the limit cycles of this system,
407: while stationary solitons of the CGLE are related to the
408: fixed points. We have found the approximate bifurcation boundaries
409: between different types of solutions in the parameter space.
410:
411: \begin{acknowledgments}
412: This work was funded by the Australian Research Council. The authors
413: are grateful to Dr. A. Ankiewicz for a critical reading of the manuscript.
414: \end{acknowledgments}
415:
416: \begin{thebibliography}{99}
417: % 1
418: \bibitem{Brusch2001} L. Brusch, A. Torcini, M. van Hecke,
419: M. G. Zimmermannf, and M. B\"ar, Physica D, {\bf 160}, 127 (2001).
420:
421: \bibitem{Reviews} I. S.
422: Aranson and L. Kramer, Rev. Mod. Phys. {\bf 74}, 100 (2002).
423: % 2
424: \bibitem{Books} See e.g. N. Akhmediev and A. Ankiewicz, {\em Solitons:
425: Nonlinear Pulses and Beams}, (Capman \& Hall, London, 1997);
426: G. P. Agrawal, {\em Nonlinear Fiber Optics}, (Academic Press, London,
427: 1989).
428:
429: % 3
430: \bibitem{Ippen94} E. P. Ippen, Appl. Phys. B {\bf 58}, 159 (1994).
431: % 4
432: \bibitem{bakonyi_2} Z. Bakonyi, D. Michaelis, U. Peschel, G. Onishchukov,
433: and F. Lederer, J. Opt. Soc. Am. B {\bf 19}, 487 (2002).
434: % 5
435: \bibitem{Kolodner} P. Kolodner, Phys. Rev. Lett. {\bf 66}, 1165 (1991).
436: % 6
437: \bibitem{Gross} M. C. Gross and P. C. Hohenberg, Rev. Mod. Phys. {\bf 65},
438: 851 (1993).
439: % 7
440: \bibitem{saarlos92} W. van Saarlos and P. C. Hohenberg, Physica D
441: {\bf 56}, 303 (1992).
442: % 8
443: \bibitem{Deiss94} R. J. Deissler and H. R. Brand, Phys. Rev. Lett.
444: {\bf 72}, 478 (1994).
445:
446: \bibitem{AST01} N. Akhmediev, J. M. Soto-Crespo, and G. Town,
447: Phys. Rev. E {\bf 63}, 056602 (2001).
448:
449: \bibitem{SGGA04} J. M. Soto-Crespo, M. Grapinet, Ph. Grelu and
450: N. Akhmediev, Phys. Rev. E {\bf 70}, 066612 (2004).
451:
452: \bibitem{Springerbook} N. Akhmediev and A. Ankiewicz,
453: Dissipative Solitons in the Complex Ginzburg-Landau
454: and Swift-Hohenberg Equations, In: {\it Dissipative Solitons},
455: (Springer, Heidelberg, 2005).
456:
457: \bibitem{Satsuma74} J. Satsuma and N. Yajima,
458: Progr. Theor. Phys. Suppl. {\bf 55}, 284 (1974).
459:
460: \bibitem{Maim93} A. I. Maimistov,
461: J. Exp. Theor. Phys. {\bf 77}, 727 (1993)
462: [Zh. Eksp. Teor. Fiz. {\bf 104}, 3620 (1993), in Russian].
463:
464: \bibitem{JOSA98} N. Akhmediev, A. Ankiewicz, and J. M. Soto-Crespo,
465: J. Opt. Soc. Am. B {\bf 15}, 515 (1998).
466:
467: \bibitem{Ostrov04} M. N. Zhuravlev and N. V. Ostrovskaya,
468: J. Exp. Theor. Phys. {\bf 99}, 427 (2004).
469: [Zh. Eksp. Teor. Fiz. {\bf 126}, 483 (2004), in Russian.]
470:
471: \bibitem{AUTO2000} See e. g. the software ``AUTO2000''
472: on the web page http://indy.cs.concordia.ca/auto/
473:
474: \bibitem{Guckenheimer} J. Guckenheimer and P. Holmes,
475: ``Nonlinear oscillations, dynamical systems and bifurcations
476: of vector fields, (Springer-Verlag, New York, 1983).
477:
478: \end{thebibliography}
479: \end{document}
480:
481: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
482: