1:
2: \documentclass{article}
3: \usepackage{graphicx}
4:
5: \begin{document}
6:
7: \title{Resonant nonlinearity management for nonlinear-Schr\"{o}dinger
8: solitons}
9: \author{Hidetsugu Sakaguchi$^{1}$ and Boris A. Malomed$^{2}$ \\
10: %EndAName
11: $^{1}$Department of Applied Science for Electronics and Materials,\\
12: Interdisciplinary Graduate School of Engineering Sciences,\\
13: Kyushu University, Kasuga, Fukuoka 816-8580, Japan\\
14: $^{2}$Department of Interdisciplinary Studies,\\
15: School of Electrical Engineering, Faculty of Engineering,\\
16: Tel Aviv University, Tel Aviv 69978, Israel}
17: \maketitle
18:
19: \begin{abstract}
20: We consider effects of a periodic modulation of the nonlinearity coefficient
21: on fundamental and higher-order solitons in the one-dimensional NLS
22: equation, which is an issue of direct interest to Bose-Einstein condensates
23: in the context of the Feshbach-resonance control, and fiber-optic
24: telecommunications as concerns periodic compensation of the nonlinearity.
25: We find from simulations, and explain by means of a straightforward
26: analysis, that the response of a fundamental soliton to the weak perturbation
27: is resonant, if the modulation frequency $\omega $ is close to the intrinsic frequency of the soliton. For higher-order $n$-solitons with $n=2$ and $3$, the response to
28: an extremely weak perturbation is also resonant, if $\omega $ is close to
29: the corresponding intrinsic frequency. More importantly, a slightly stronger
30: drive splits the $2$- or $3$-soliton, respectively, into a set of two or
31: three moving fundamental solitons. The dependence of the threshold
32: perturbation amplitude, necessary for the splitting, on $\omega $ has a
33: resonant character too. Amplitudes and velocities of the emerging
34: fundamental solitons are accurately predicted, using exact and approximate
35: conservation laws of the perturbed NLS equation.
36: \end{abstract}
37:
38: PACS numbers: 03.75.Lm, 05.45.Yv, 42.65.Tg
39:
40: \section{Introduction}
41:
42: The nonlinear Schr\"{o}dinger (NLS) equation is a universal model
43: of weakly nonlinear dispersive media \cite{Whitham,Zakharov}. The
44: existence and stability of solitons in the one-dimensional (1D)
45: version of the NLS equation with constant coefficients is a
46: well-established fact, which has important implications in various
47: areas of physics. In particular, solitons in fiber-optic
48: telecommunications \cite{Hasegawa} and quasi-1D Bose-Einstein
49: condensates (BECs) with attractive interactions between atoms
50: \cite{BECsoliton}, have drawn a great deal of interest.
51:
52: A new class of dynamical problems, which also have a vast potential for
53: physical applications, emerges in the investigation of soliton dynamics in
54: extended versions of the NLS equation, in which coefficients are periodic
55: functions of the evolutional variable. A well-known example is a nonlinear
56: fiber-optic link subjected to \textit{dispersion management} (DM), which
57: implies that the dispersion coefficient periodically alternates between
58: positive and negative values. The DM links support a family of stable
59: temporal solitons (see, e.g., Refs. \cite{DM}, and also Ref. \cite{Ming}).
60: Somewhat similar is a \textit{waveguide-antiwaveguide} system, which can be
61: realized in the spatial domain (planar optical waveguides). In the latter
62: case, a light beam is transmitted through a periodic concatenation of
63: nonlinear waveguiding and antiwaveguiding segments \cite{Kaplan}. A common
64: feature of the latter system with the DM is that the coefficient which
65: periodically jumps between positive and negative values also belongs to the
66: linear part of the equation.
67:
68: Another technique that may be useful for optical telecommunications is
69: \textit{nonlinearity management} (NLM), which assumes that the coefficient
70: in front of the nonlinear term periodically changes its sign. An advantage
71: offered by the NLM is a possibility to compensate the nonlinear phase shift
72: accumulating in pulses due to the Kerr nonlinearity of the optical fiber
73: \cite{Pare'}. In practical terms, the NLM can be implemented by dint of
74: elements with a strong quadratic ($\chi ^{(2)}$) nonlinearity, which are
75: periodically inserted into the fiber link. The $\chi ^{(2)}$ elements can
76: emulate a negative Kerr effect through the cascading mechanism \cite{Uri}.
77: Various other schemes of NLM in fiber-optic links were considered, including
78: its combination with the DM, amplifiers, etc. \cite{NMfiber-general}. A
79: related scheme makes use of the NLM in soliton-generating lasers based on
80: fiber rings \cite{NMfiber-laser}. The NLM for spatial solitons, which
81: assumes alternation of self-focusing and self-defocusing nonlinear layers in
82: planar \cite{Javid} or bulk \cite{Isaac} waveguides, was introduced too.
83:
84: All these systems may be regarded as examples of periodically
85: inhomogeneous optical waveguiding media. Other examples belonging
86: to the same general class are \textit{tandem waveguides} (see Ref.
87: \cite{tandem} and references therein) and the \textit{split-step}
88: \textit{model} \cite{SSM}. These are built as a juxtaposition of
89: linear segments alternating with ones featuring, respectively,
90: quadratic or cubic nonlinearity. A common feature of the models of
91: all these types is that they support \emph{robust} solitons,
92: despite a ``naive" expectation that solitons would quickly decay,
93: periodically hitting interfaces between strongly different
94: elements of which the system is composed.
95:
96: The above-mentioned optical media are described by the NLS equation, in
97: which the role of the evolutional variable belongs to the propagation
98: distance, while the remaining free variable is either the local time (for
99: temporal solitons), or the transverse coordinate(s), in the spatial-domain
100: models. Mathematically similar, but physically altogether different, models
101: describe BECs in the 1D geometry. In that case, the corresponding NLS
102: equation is usually called the Gross-Pitaevskii (GP) equation. It governs
103: the evolution of the mean-field wave function $\phi $ in time ($t$), the
104: other variable, $x$, being the coordinate along the quasi-1D trap. In the
105: normalized form, the GP equation is
106: \begin{equation}
107: i\phi _{t}=-\frac{1}{2}\phi _{xx}+U(x)\phi +g|\phi |^{2}\phi , \label{GPE}
108: \end{equation}where $U(x)$ is the potential which confines the condensate, and the
109: nonlinearity coefficient $g$ is proportional to the scattering
110: length of collisions between atoms. Two natural possibilities to
111: introduce a time-periodic (ac) ``management" in the BEC context
112: are either through a periodic modulation of the confining
113: potential, most typically in the form of $U(x,t)=\frac{1}{2}\left[
114: \kappa _{0}+\kappa _{1}\cos (\omega t)\right] x^{2}$, or by means
115: of time modulation of the scattering length, using the Feshbach
116: resonance (FR) \cite{FR}. In the latter case, the nonlinearity
117: coefficient in Eq. (\ref{GPE}) takes the form of
118: $g(t)=g_{0}+g_{1}\sin (\omega t)$. In either case, the modulation
119: is generated by a combination of dc and ac magnetic fields applied
120: to the BEC.
121:
122: The GP equation with the periodically modulated strength of the
123: trapping potential was considered for both $g>0$ (when solitons do
124: not exist, and the BEC as a whole is subjected to the
125: ``management", including the 2D and 3D cases) \cite{GarciaGarcia},
126: and $g<0$, when the soliton is the basic dynamical object
127: \cite{Tashkent}. In particular, a parametric resonance is possible
128: in the former case, and creation of an effectively trapping
129: potential, while the underlying one is anti-trapping, having
130: $\kappa _{0}<0$, by the high-frequency ac part of the potential
131: (with large $\omega $) was predicted in the latter case.
132:
133: The periodic modulation of the nonlinearity coefficient through
134: the ``ac FR management" is an especially interesting possibility,
135: as the FR is a highly efficient experimental tool, broadly used
136: for the study of various dynamical properties of the BECs
137: \cite{FR}. In particular, it has been predicted that the
138: modulation through the ac FR makes it possible to preclude
139: collapse and generate stable soliton-like structures in 2D (but
140: not 3D) condensates \cite{stabilization}; in fact, this prediction
141: is similar to the earlier considered possibility of the
142: stabilization of 2D spatial optical solitons in a bulk waveguide
143: subjected to the periodic NLM \cite{Isaac}. In the 1D model of the
144: GP type, subjected to the NLM, various stable dynamical states,
145: including Gaussian-shaped soliton-like objects, and ones of the
146: Thomas-Fermi type, were studied in detail \cite{Athens}. In
147: addition, analysis based on averaged equations was developed, for
148: this case, in Ref. \cite{DimDim} (similarly to the analysis
149: elaborated in Ref. \cite{Tashkent} for the case of the periodic
150: modulation of the trapping potential).
151:
152: The objective of this work is to study \emph{resonance effects}
153: produced by the ac FR management, i.e., harmonic modulation of the
154: nonlinearity coefficient, in the dynamics of fundamental and
155: higher-order 1D solitons in the NLS equation. We will focus on the
156: case when the ac part of the nonlinear coefficient is small in
157: comparison with its constant (dc) part $g_{0}$, which accounts for
158: the self-attraction in the BEC, and is normalized to be
159: $g_{0}=-1$. We also assume that the soliton's width is much
160: smaller than the effective size of the trap, hence the external
161: potential may be dropped. The consideration of the GP equation
162: without the trapping potential makes it possible to identify
163: fundamental dynamical effects for the solitons induced by the ac
164: FR management. In this connection, it is necessary to mention
165: that, in the 1D case, the trapping potential is not a crucial
166: factor, on the contrary to the 2D and 3D cases, where the external
167: potential plays a much more important role, in view of the
168: intrinsic instability of the multi-dimensional NLS solitons.
169:
170: Thus, we will be dealing with the normalized NLS\ equation in the form [cf.
171: Eq. (\ref{GPE})]
172: \begin{equation}
173: i\phi _{t}+\frac{1}{2}\phi _{xx}+\left[ 1+b\sin \left( \omega
174: t\right) \right] |\phi |^{2}\phi =0, \label{NLSE}
175: \end{equation}where the amplitude $b$ of the ac drive is small. Note that Eq. (\ref{NLSE})
176: conserves exactly two dynamical invariants: the norm, which is proportional
177: to the number of atoms in the BEC,
178: \begin{equation}
179: N=\int_{-\infty }^{+\infty }\left\vert \phi (x)\right\vert ^{2}dx, \label{N}
180: \end{equation}and the momentum,
181: \begin{equation}
182: P=i\int_{-\infty }^{+\infty }(\phi \phi _{x}^{\ast }-\phi ^{\ast }\phi
183: _{x})dx. \label{P}
184: \end{equation}
185:
186: We will demonstrate that the weak ac perturbation in Eq. (\ref{NLSE}) can
187: generate strong effects, if the driving frequency $\omega $ is close to
188: specific resonant values. These effects include intrinsic vibrations of the
189: fundamental soliton and splitting of the higher-order ones. We will also
190: propose analytical explanations to these effects. To the best of our
191: knowledge, these results have not been reported before for the present
192: simple model.
193:
194: The rest of the paper is organized in the following way. The resonant effect
195: of the periodic NLM on the fundamental soliton is reported in Section 2, and
196: the resonant splitting of $n$-solitons with $n=2$ and $3$ is investigated in
197: Section 3. Section 4 concludes the paper.
198:
199: \section{Resonant response of the fundamental soliton}
200:
201: \subsection{Numerical results}
202:
203: First, we consider the action of the ac perturbation in Eq. (\ref{NLSE}) on
204: the fundamental soliton, which, in the case $b=0$, is
205: \begin{equation}
206: \phi _{\mathrm{sol}}(x,t)=A~\mathrm{sech}(A\left( x-x_{0}\right) )\exp
207: \left( iA^{2}t/2\right) , \label{sol}
208: \end{equation}where $A$ is an arbitrary amplitude. Numerical simulations of Eq. (\ref{NLSE}
209: ) were performed in a sufficiently large domain, $0<x<L$, the
210: initial condition corresponding to the soliton (\ref{sol}) placed
211: at the center of the domain, $x_{0}=L/2$. We have performed
212: numerical simulations using the split-step Fourier method with
213: 1024 Fourier modes. The system size is $L=50$, and the time step
214: for the numerical simulation is $\Delta t=0.001$.
215:
216: Figure \ref{fig1}(a) displays a typical example of the time evolution of the
217: soliton's amplitude $|\phi (x=L/2)|$, under the action of the ac
218: perturbation with a very small amplitude, $b=0.0001$. The frequency of the
219: beatings observed in this figure can be clearly identified as $\omega
220: -\omega _{\mathrm{sol}}$, where $\omega _{\mathrm{sol}}\equiv A^{2}/2$ is
221: the intrinsic frequency of the unperturbed soliton (\ref{sol}). We have
222: checked that the the beating frequency is independent of the system's size
223: and the other details of the numerical scheme.
224:
225: \begin{figure}[tbph]
226: \begin{center}
227: \includegraphics[width=3.5in]{fig1.eps}
228: \end{center}
229: \caption{(a) A typical example of beatings in the evolution of the
230: fundamental-soliton's amplitude under the action of a very weak
231: perturbation in Eq. (\protect\ref{NLSE}), with $b=10^{-4}$ and
232: $\protect\omega =2.2$. The amplitude of the initial unperturbed
233: soliton is $A=2$ (the corresponding soliton's frequency is
234: $\protect\omega _{\mathrm{sol}}\equiv A^{2}/2=2$). (b) The
235: difference between the maximum and minimum values of the soliton's
236: amplitude vs. the perturbation frequency. The dashed line is a
237: fitting curve, $0.0003\cdot |\protect\omega -2|^{-1/2}$.}
238: \label{fig1}
239: \end{figure}
240:
241: \begin{figure}[tbph]
242: \begin{center}
243: \includegraphics[width=2.2in]{fig2.eps}
244: \end{center}
245: \caption{
246: Direct numerical solution of the linearized equation
247: (\protect\ref{lin}) for the perturbation around the fundamental
248: soliton, in the near-resonance case, with $\protect\omega =1.98$.}
249: \label{fig2}
250: \end{figure}
251:
252: The main resonant effect for the fundamental soliton is displayed
253: in Fig. \ref{fig1}(b), in the form of the difference between the
254: maximum and minimum (in time) values of its amplitude versus the
255: driving frequency $\omega $. The resonance at $\omega =\omega
256: _{\mathrm{sol}}=2$ (for $A=2$) is obvious. The simulations do not
257: reveal any noticeable subharmonic or higher-order resonance at
258: frequencies $\omega =1$, $3$ or $4$. Due to the scaling invariance
259: of Eq. (\ref{NLSE}), the plot shown in Fig. \ref{fig1}(b) does not
260: pertain solely to the particular value of the soliton's amplitude,
261: $A=2$, but is actually a universal one. It is easy to verify that
262: the ranges of the variables $t$ and $x$, which are shown in this
263: and other figures, correspond, in the normalized units, to
264: experimentally realistic configurations of the BECs in the
265: quasi-1D geometry.
266:
267: \subsection{Perturbative analysis}
268:
269: In order to explain the resonance shown above, we look for a
270: perturbed fundamental solution as $\phi (x,t)=\phi
271: _{\mathrm{sol}}(x,t)+\phi _{\mathrm{pert}}(x,t)$, where the first
272: term is the solution (\ref{sol}). Thus, we arrive at the driven
273: linearized equation for the perturbation,
274: \begin{eqnarray}
275: &&i\left( \phi _{\mathrm{pert}}\right) _{t}+\frac{1}{2}\left( \phi _{\mathrm{\ pert}}\right) _{xx}+A^{2}\mathrm{sech}^{2}(Ax)\left( 2\phi _{\mathrm{pert}}+e^{iA^{2}t}\phi _{\mathrm{pert}}^{\ast }\right) \nonumber \\
276: &=&\frac{i}{2}bA^{3}\mathrm{\ \ sech}^{3}(Ax)\left[ e^{i\left( \left(
277: A^{2}/2\right) +\omega \right) t}-e^{i\left( \left( A^{2}/2\right) -\omega
278: \right) t}\right] . \label{lin}
279: \end{eqnarray}The source of the resonant response is in the fact that the second term on
280: the right-hand side of Eq. (\ref{lin}) becomes time-independent
281: exactly at the resonance point, $\omega =\omega
282: _{\mathrm{sol}}\equiv A^{2}/2$. Figure \ref{fig2} displays the
283: evolution of the perturbation close to the resonance at $\omega
284: =1.98$, as found from direct numerical integration of the
285: linearized equation (\ref{lin}), with the initial condition $\phi
286: _{\mathrm{pert}}(x)=0$. As can be seen from the figure, the
287: perturbation grows in time at the center, and simultaneously
288: expands in space. Strictly speaking, the latter feature remains
289: valid as long as the size of the region occupied by the expanding
290: wave fields remains essentially smaller than the limit imposed by
291: the confining field.
292:
293: The linearized equation (\ref{lin}) is too difficult for an exact
294: analytical solution. However, the observation that the
295: characteristic spatial scale of the solution observed in Fig.
296: \ref{fig2} becomes much larger than the internal scale of the
297: function $\mathrm{sech}(Ax)$ suggests that principal features of
298: the solution can be understood from a simpler equation, in which
299: the term $\sim \mathrm{sech}^{2}(Ax)$ on the left-hand side of Eq.
300: (\ref{lin}) is neglected, and the source corresponding to the
301: second term on the right-hand side is approximated by a $\delta
302: $-function:
303: \begin{equation}
304: i\left( \tilde{\phi}_{\mathrm{pert}}\right) _{t}+\frac{1}{2}\left(
305: \tilde{\phi}_{\mathrm{pert}}\right) _{xx}=\mathrm{const}\cdot
306: \delta (x)e^{-i\Delta \omega \cdot t},~\mathrm{const}=\frac{i\pi
307: }{4}bA^{2}, \label{simple}
308: \end{equation}where $\Delta \omega \equiv \omega -\omega _{\mathrm{sol}}$, and $\mathrm{const}\equiv \left( ib/2\right) A^{3}\int_{-\infty }^{+\infty }\mathrm{sech}^{3}(Ax)dx$. Equation (\ref{simple}) can be solved by means of the Fourier
309: transform. After straightforward manipulations, this
310: yields\begin{equation}
311: \tilde{\phi}_{\mathrm{pert}}(x,t)=-\frac{\mathrm{const}}{2\sqrt{\pi
312: }}(1+i)e^{-i\Delta \omega \cdot t}\int_{0}^{t}\frac{dt^{\prime
313: }}{\sqrt{t^{\prime }}}\exp \left( \frac{ix^{2}}{2t^{\prime
314: }}+i\Delta \omega \cdot t^{\prime }\right) . \label{solution}
315: \end{equation}Further consideration shows that, for $\Delta \omega <0$, the asymptotic
316: form of the solution (\ref{solution}) at $t\rightarrow \infty $ amounts to
317: an exponentially localized stationary expression, which, by itself, is an
318: exact solution to Eq. (\ref{simple}):
319: \begin{equation}
320: \tilde{\phi}_{\mathrm{pert}}(x,t)=\frac{\mathrm{const}}{\sqrt{2\left\vert
321: \Delta \omega \right\vert }}\exp \left( i\left\vert \Delta \omega
322: \right\vert t-\sqrt{2\left\vert \Delta \omega \right\vert }|x|\right) .
323: \label{stationary}
324: \end{equation}In the case of $\Delta \omega >0$, the asymptotic form of the general
325: solution (\ref{solution}) corresponds to a symmetric region
326: occupied by plane waves emitted by the central source at
327: wavenumbers $k=\pm \sqrt{2\Delta \omega }$. The region expands in
328: time with the group velocities $v_{\mathrm{gr}}=k=\pm
329: \sqrt{2\Delta \omega }$, so that the asymptotic form of the
330: solution is
331: \begin{equation}
332: \tilde{\phi}_{\mathrm{pert}}(x,t)\approx
333: -i\frac{\mathrm{const}}{\sqrt{2\Delta \omega }}\cdot \left\{
334: \begin{array}{cc}
335: \exp \left( -i\Delta \omega \cdot t+i\sqrt{2\Delta \omega }|x|\right) , &
336: \mathrm{if}~|x|~<\sqrt{2\Delta \omega }t, \\
337: 0, & \mathrm{if}~|x|~>\sqrt{2\Delta \omega }t.\end{array}\right.
338: \label{expanding}
339: \end{equation}This asymptotic solution implies that the norm (\ref{N}) of the expanding
340: radiation field grows in time at the rate
341: $dN_{\mathrm{pert}}/dt=\sqrt{2/\Delta \omega }\left\vert
342: \mathrm{const}\right\vert ^{2}$, which, in fact, is the rate at
343: which the norm flows from the soliton to the radiation waves
344: emitted under the action of the ac perturbation.
345:
346: Both analytical expressions (\ref{stationary}) and
347: (\ref{expanding}) feature the $\left\vert \Delta \omega
348: \right\vert ^{-1/2}$ factor, that perfectly fits the numerical
349: data summarized in Fig. \ref{fig1}(b). Although these results,
350: obtained for the weak time-periodic FR management, seem very
351: simple, they have not been reported before, to the best of our
352: knowledge. We also notice that the usual variational approximation
353: (VA)\ for the NLS solitons, which is efficient in explaining a
354: number of other perturbative effects \cite{VA}, cannot account for
355: the occurrence of the resonance at $\omega =\omega
356: _{\mathrm{sol}}$, because the VA neglects radiation effects, while
357: the above consideration showed that it is exactly the radiation
358: field which is amenable for the manifestations of the resonance.
359:
360: The above results were obtained in the linear approximation, i.e.,
361: for a very small amplitude $b$ of the ac drive in Eq.
362: (\ref{NLSE}). At larger $b$, the perturbed soliton can either
363: survive or decay into radiation. In fact, a stability region for
364: the solitons in a similar model with a nonsmall perturbation,
365: which differs from that in Eq. (\ref{NLSE}) by the form of the
366: periodic modulation function, which is a piecewise-constant one,
367: rather than harmonic, was drawn in Ref. \cite{Javid} in the
368: context of a model for spatial optical solitons in a layered
369: waveguide. In the cases when a stable soliton established itself
370: in the strongly perturbed (``strongly nonlinearly-managed")
371: system, its formation from the initial configuration (\ref{sol})
372: went through emission of radiation and, sometimes, separation of a
373: small secondary pulse, while no pronounced resonance at $\omega
374: =\omega _{\mathrm{sol}}$ was observed. As we do not expect that
375: the replacement of the piecewise-constant modulation function by
376: the harmonic one should dramatically alter the stability region,
377: we do not consider this issue here in detail.
378:
379: \section{Resonant splitting of higher-order solitons}
380:
381: \subsection{Response to a very weak ac drive}
382:
383: As is well known, the unperturbed NLS equation gives rise to exact soliton
384: solutions of order $n$, in the form of periodically oscillating breathers,
385: which start from the initial conditions
386: \begin{equation}
387: \phi _{0}(x)=nA~\mathrm{sech}\left( A\left( x-L/2\right) \right) \label{n}
388: \end{equation}with an integer $n>1$ \cite{SatsumaYajima} [the expression (\ref{n}) assumes
389: that the initial configuration is placed at the center of the integration
390: domain]. The frequency of the shape oscillations (breathings) of the
391: higher-order soliton is
392: \begin{equation}
393: \omega _{\mathrm{br}}=4A^{2}, \label{omega_br}
394: \end{equation}irrespective of the value of $n$.
395:
396: Generally speaking, the higher-order solitons are unstable bound
397: complexes of fundamental solitons, as, in the absence of
398: perturbations, their binding energy is exactly zero, which is a
399: known consequence of the exact integrability of the unperturbed
400: NLS equation. Nevertheless, not any perturbation readily splits
401: the higher-order soliton into its fundamental constituents;
402: usually, the splitting is easily induced by specific
403: nonconservative terms added to the NLS equation, such as the one
404: accounting for the intra-pulse stimulated Raman scattering in
405: optical fibers \cite{Hasegawa}. The consideration of dynamics of
406: the higher-order solitons is also relevant, especially in the
407: context of BECs, as the corresponding initial configurations can
408: be created in the real experiment.
409:
410: We have studied in detail the $n$-solitons up to $n=5$. First, we
411: consider the case of a very small driving amplitude, $b=0.00005$.
412: Figure \ref{fig3} (a) displays oscillations of the amplitude
413: $|\phi (x=L/2)|$ of the $2$ -soliton, which corresponds to the
414: initial condition (\ref{n}) with $n=2$. The frequency of the basic
415: oscillations coincides with $\omega _{\mathrm{br}} $, as given by
416: the expression (\ref{omega_br}), while the frequency of the zoomed
417: beatings in Fig. \ref{fig3}(b) can be clearly identified with
418: $\omega -\omega _{\mathrm{br}}$. The resonant character of the
419: response of the $2$-soliton to the weak NLM is obvious from Fig.
420: \ref{fig3}(c).
421: \begin{figure}[tbph]
422: \begin{center}
423: \includegraphics[width=5in]{fig3.eps}
424: \end{center}
425: \caption{(a) Oscillations of the amplitude, $\left\vert
426: \protect\phi (x=L/2)\right\vert $, of the $2$-soliton created by
427: the initial condition (\protect\ref{n}) with $A=1$, in the case
428: of $b=0.5\times 10^{-4}$ and $\protect\omega =4.15$. (b) Zoom of
429: the previous panel around minimum values of the amplitude, which
430: reveals beatings at the frequency $\protect\omega -\protect\omega
431: _{\mathrm{br}}$. (c) The difference between the maximum and
432: minimum values of the soliton's amplitude vs. the driving
433: frequency $\protect\omega $. The dashed line is a fitting curve,
434: $0.0006/|\protect\omega -4|$; note the difference of the fitting
435: power, $-1$, from that, $-1/2$, in Fig. \protect\ref{fig1}(b).}
436: \label{fig3}
437: \end{figure}
438:
439: Note that the exact solution for the $n$-soliton features not only
440: the shape-oscillation frequency (\ref{omega_br}), but also an
441: overall frequency of the phase oscillations, which coincides with
442: the above-mentioned frequency $\omega _{\mathrm{sol}}=A^{2}/2$ for
443: the fundamental soliton, provided that the initial condition is
444: taken as in Eq. (\ref{n}). In the simulations, we also observed a
445: resonant response at $\omega =\omega _{\mathrm{sol}}$, but this
446: resonance was essentially weaker than the one at $\omega =\omega
447: _{\mathrm{br}}$. In particular, this is manifest in the fact that,
448: as well as in the case of the fundamental soliton, the fit to the
449: response around the former resonance is provided by the expression
450: $\left\vert \omega -\omega _{\mathrm{sol}}\right\vert ^{-1/2}$,
451: cf. Fig. \ref{fig1}(b), while the fit to the resonance at $\omega
452: =\omega _{\mathrm{sol}}$ demonstrates a more singular dependence,
453: $\sim \left\vert \omega -\omega _{\mathrm{br}}\right\vert ^{-1}$,
454: as seen in Fig. \ref{fig3}(c). Another qualitative difference
455: between the two resonances is that the one at $\omega =\omega
456: _{\mathrm{br}}$, with a larger (but still small) forcing parameter
457: $b $, leads to splitting of the higher-order solitons into
458: fundamental ones, as shown below, while, in the case of the
459: resonance at $\omega =\omega _{\mathrm{sol}}$, the increase of $b$
460: does not lead to the splitting.
461:
462: \subsection{Splitting of $2$- and $3$-solitons}
463:
464: Unlike the case of the fundamental soliton, the reaction of the
465: higher-order ones to larger values of the forcing parameter was
466: not studied before, therefore we have done it here. First, we aim
467: to demonstrate that the $2$-soliton readily splits into two moving
468: fundamental pulses, if the driving frequency is close to the
469: resonant value (\ref{omega_br}). The shape of each moving soliton
470: is very close to that given by the commonly known exact solution,
471: which can be obtained by application of the Galilean boost, with a
472: velocity $v$, to the zero-velocity fundamental soliton
473: (\ref{sol}),
474: \begin{equation}
475: \phi _{\mathrm{sol}}(x,t)=A~\mathrm{sech}(A(x-vt))\exp \left[ ivx+\left(
476: i/2\right) \left( A^{2}-v^{2}\right) t\right] . \label{boost}
477: \end{equation}
478:
479: Figure \ref{fig4} displays the evolution of the wave function for
480: the initial condition (\ref{n}) with $n=2$ in the resonant case
481: ($A=1$ and $\omega =4$), with the driving amplitude $b=0.0005$.
482: The latter value is still very small, but larger by a factor of
483: $10$ than in the case shown in Fig. \ref{fig3}. The amplitudes of
484: the two fundamental solitons, observed as a result of the
485: splitting, are close to $A_{1}=3$ and $A_{2}=1$ [note that they
486: exactly corresponds to the fundamental-soliton constituents of the
487: original $2$-soliton with $A=1$, in terms of the inverse
488: scattering transform (IST) \cite{SatsumaYajima}]. Velocities of
489: the splinters were measured to be $v_{1}=0.00197$ and
490: $v_{2}=0.0066$, respectively (with the ratio $v_{1}:v_{3}\approx
491: 1:3$). At the end of the simulation run ($t=1000$ ), the secondary
492: solitons are found at the distance, respectively, $4.5$ and $13.2$
493: from the central point, $x=L/2$.
494: \begin{figure}[tbph]
495: \begin{center}
496: \includegraphics[width=3.5in]{fig4.eps}
497: \end{center}
498: \caption{A typical example of the splitting of a $2$-soliton [generated by
499: the initial condition (\protect\ref{n}) with $n=2$ and $A=1$] into an
500: asymmetric pair of moving fundamental solitons, under the action of the weak
501: resonant drive, with $\protect\omega =4$ and $b=0.0005$. (a) The evolution
502: of $\left\vert u(x,t)\right\vert $. (b) The final configuration at $t=1000$.}
503: \label{fig4}
504: \end{figure}
505:
506: Similar near-resonant splittings were observed for $n$-solitons
507: with $n>2$. In particular, Fig. \ref{fig5} shows this outcome for
508: $n=3$, which corresponds to the initial configuration (\ref{n})
509: with $n=3$, $A=0.5$, $\omega =1$ and $b=0.0005$. This time, the
510: splitting gives rise to three moving fundamental solitons, whose
511: amplitudes are close to $A_{1}=2.5$, $A_{2}=1.5$, and $A_{3}=0.5$.
512: As well as in the case of $n=2$, these values correspond to the
513: constituents of the original $3$-soliton (with $A=0.5$), in terms
514: of the IST \cite{SatsumaYajima}. The velocities of the three
515: splinters are $v_{1}=-0.00146$, $v_{2}=0.0732$, and
516: $v_{3}=-0.0148$, so that the ratios between them are
517: $v_{1}/v_{2}\approx -1/5$ and $v_{3}/v_{2}\approx -2$.
518: \begin{figure}[tbph]
519: \begin{center}
520: \includegraphics[width=3.5in]{fig5.eps}
521: \end{center}
522: \caption{The same as in Fig. \protect\ref{fig4} for the
523: $3$-soliton, generated by the initial configuration
524: (\protect\ref{n}) with $n=3$ and $A=0.5$. In this case, the
525: forcing frequency and amplitude are $\protect\omega =1$ and
526: $b=0.0005$.} \label{fig5}
527: \end{figure}
528:
529: These results can be summarized in the form of diagrams which show
530: the minimum (threshold) value of the forcing amplitude $b$,
531: necessary for the splitting, versus the driving frequency $\omega
532: $. The splitting of the $2$- and $3$-solitons was registered if it
533: took place in the simulations of Eq. (\ref{NLSE}) that were run up
534: to the time, respectively, $t=600$ or $t=2000$ (still longer
535: simulations did not give rise to any essential difference in the
536: results). As is seen from Fig. \ref{fig6}, for both $2$- and
537: $3$-solitons these dependences clearly have a resonant shape, with
538: sharp minima at the frequency given by Eq. (\ref{omega_br}). It is
539: not quite clear why the forcing amplitude required for the
540: splitting is very small but finite even exactly at the resonance
541: point. This may be related to the accuracy of the numerical scheme
542: and/or the finite size of the integration domain. Similar
543: observations were also made in simulations of the $n$-solitons
544: with $n=4$ and $5$.
545: \begin{figure}[tbph]
546: \begin{center}
547: \includegraphics[width=3.5in]{fig6.eps}
548: \end{center}
549: \caption{The minimum values of the amplitude of the ac perturbation,
550: necessary for the splitting of the $2$-soliton (a) and $3$-soliton (b), as
551: functions of the driving frequency. The initial condition is taken in the
552: form of Eq. (\protect\ref{n}) with, respectively, $n=2$ and $A=1$, or $n=3$
553: and $A=0.5$. In both cases, the sharp minimum exactly corresponds to the
554: resonant frequency, as predicted by Eq. (\protect\ref{omega_br}).}
555: \label{fig6}
556: \end{figure}
557:
558: \subsection{Analytical results}
559:
560: The amplitudes and velocities of the fundamental solitons, into which the
561: higher-order ones split, can be predicted in an analytical from. As it was
562: already mentioned above, the amplitudes of the secondary solitons coincide
563: with those which correspond to the constituents (eigenvalues) of the
564: corresponding original $n$-soliton in terms of the IST. However, the
565: velocities of the emerging fundamental solitons cannot be forecast this way,
566: as, in terms of the IST, they are zero when the fundamental solitons are
567: bound into a higher-order one.
568:
569: Nevertheless, both the amplitudes and velocities of the final set of the
570: solitons can be predicted in a different way, using the exact and nearly
571: exact conservation laws of Eq. (\ref{NLSE}). Indeed, there are two exact
572: dynamical invariants, (\ref{N}) and (\ref{P}), and, in addition to that, the
573: unperturbed NLS equation has an infinite series of higher-order dynamical
574: invariants, starting from the Hamiltonian,
575: \begin{equation}
576: H=\frac{1}{2}\int_{-\infty }^{+\infty }(|\phi _{x}|^{2}-|\phi |^{4})dx.
577: \label{H}
578: \end{equation}Two next invariants, which do not have a straightforward physical
579: interpretation, are \cite{Zakharov}
580: \[
581: I_{4}=\frac{1}{2}\int_{-\infty }^{+\infty }\left( \phi \phi _{xxx}^{\ast
582: }+3\phi \phi _{x}^{\ast }|\phi |^{2}\right) dx,
583: \]\begin{equation}
584: I_{5}=\frac{1}{4}\int_{-\infty }^{+\infty }\left[ |\phi _{xx}|^{2}+2|\phi
585: |^{6}-\left( \left( |\phi |^{2}\right) _{x}\right) ^{2}-6|\phi
586: _{x}|^{2}|\phi |^{2}\right] dx. \label{I5}
587: \end{equation}
588:
589: In the case of the splitting of the $2$-soliton (\ref{n}) with the
590: amplitude $A$, the exact conservation of the norm (\ref{N}) and
591: approximate conservation of the Hamiltonian (\ref{H}) yield the
592: following relations between $A$ and the amplitudes $A_{1,2}$ of
593: the emerging fundamental solitons (splinters): $4A=A_{1}+A_{2}$,
594: and $28A^{3}\approx A_{1}^{3}+A_{2}^{3}$ (the latter relation
595: neglects small kinetic energy of the emerging solitons). These two
596: relations immediately yield $A_{1}=3A$ and $A_{2}=A$, which
597: coincides with the the above-mentioned numerical results, as well
598: as with the predictions based on the set of the $2$-soliton's IST\
599: eigenvalues. Furthermore, the exact momentum conservation yields a
600: relation involving the velocities $v_{1,2}$ of the secondary
601: solitons, $A_{1}v_{1}+A_{2}v_{2}=0$. With regard to the ratio
602: $A_{1}/A_{2}=3$, this implies $v_{1}/v_{2}=-A_{2}/A_{1}=-1/3$.
603: This relation is indeed consistent with the aforementioned
604: numerical results, although the absolute values of the velocities
605: cannot be predicted this way.
606:
607: Similarly, in the case of the splitting of the $3$-soliton, the
608: exact conservation of $N$ and approximate conservation of $H$ and
609: $I_{5}$ [see Eq. (\ref{I5})] yield the relations (which again
610: neglect small kinetic terms, in view of the smallness of the
611: observed velocities) $9A=A_{1}+A_{2}+A_{3}$, $153A^{3}\approx
612: A_{1}^{3}+A_{2}^{3}+A_{3}^{3}$, and $3369A^{5}\approx
613: A_{1}^{5}+A_{2}^{5}+A_{3}^{5}$. A solution to this system of
614: algebraic equations is $A_{1}=5A$, $A_{2}=3A$, $A_{3}=A$, which
615: are the same values that were found from the direct simulations,
616: and can be predicted as the IST eigenvalues. The conservation of
617: $P$ and $I_{4}$ gives rise to further relations,
618: $A_{1}v_{1}+A_{2}v_{2}+A_{3}v_{3}=0$ and
619: $(A_{1}v_{1}^{3}-A_{1}^{3}v_{1})+(A_{2}v_{2}^{3}-A_{2}^{3}v_{2})+(A_{3}v_{3}^{3}-A_{3}^{3}v_{3})=0
620: $. If the velocities $v_{1,2}$ are small, it follows from here
621: that
622: $v_{1}/v_{2}=-(A_{2}^{3}-A_{2}A_{3}^{2})/(A_{1}^{3}-A_{1}A_{3}^{2})=-1/5$,
623: and
624: $v_{3}/v_{2}=-(A_{2}^{3}-A_{2}A_{1}^{2})/(A_{3}^{3}-A_{3}A_{1}^{2})=-2$.
625: These results for the velocities are consistent with the numerical
626: situation observed in Fig.~5.
627:
628: \section{Conclusion}
629:
630: In this work, we have addressed a simple model, based on the NLS equation,
631: which describes an attractive Bose-Einstein condensate (BEC) in a quasi-1D
632: trap, with the nonlinearity strength subjected to a weak time-periodic (ac)
633: modulation (that can be imposed by means of the Feshbach-resonance
634: technique). The same model describes the nonlinearity management in
635: periodically inhomogeneous optical waveguides.
636:
637: It was found from direct simulations, and explained by means of a
638: straightforward perturbative expansion, that the response of a
639: fundamental soliton, in the form of temporal beatings of its
640: amplitude, to the weak ac perturbation is resonant when the
641: driving frequency $\omega $ is close to the soliton's intrinsic
642: frequency. For $n$-solitons (breathers), with $n=2$ and $3$, the
643: response to an extremely weak drive is also resonant, if $\omega $
644: is close to the breathing frequency. More interestingly, a
645: slightly stronger drive gives rise to splitting of the $2$- and
646: $3$-solitons into sets of two or three moving fundamental
647: solitons, respectively. The dependence of the minimum perturbation
648: amplitude, which is necessary for the splitting, on $\omega $ has
649: a clearly resonant character too. The amplitudes of the splinter
650: solitons, and the ratio of their velocities, can be easily
651: predicted on the basis of the exact and approximate conservation
652: laws of the perturbed NLS equation.
653:
654: \section*{Acknowledgment}
655:
656: The work of B.A.M. was partially supported by the grant No. 8006/03 from the
657: Israel Science Foundation.
658:
659: \begin{thebibliography}{99}
660: \bibitem{Whitham} G. B. Whitham. \textit{Linear and Nonlinear Waves} (John
661: Wiley \& Sons: New York, 1974).
662:
663: \bibitem{Zakharov} V. E. Zakharov, S. V. Manakov, S. P. Novikov, and L. P.
664: Pitaevskii. \textit{Theory of Solitons} (Consultants Bureau: New York, 1984)
665:
666: \bibitem{Hasegawa} A. Hasegawa and Y. Kodama. \textit{Solitons in Optical
667: Communications} (Oxford University Press: New York, 2004).
668:
669: \bibitem{BECsoliton} K. E. Strecker, G. B. Partridge, A. G. Truscott, and R.
670: G. Hulet, Nature \textbf{417}, 150 (2002); L. Khaykovich, F. Schreck, G.
671: Ferrari, T. Bourdel, J. Cubizolles, L. D. Carr, Y. Castin, and C. Salomon,
672: Science \textbf{296}, 1290 (2002); L. D. Carr and J. Brand, Phys. Rev. Lett.
673: \textbf{92}, 040401 (2004).
674:
675: \bibitem{DM} A. Berntson, N.J. Doran, W. Forysiak, and J. H. B. Nijhof, Opt.
676: Lett. \textbf{23}, 900 (1998); S. K. Turitsyn, E. G. Shapiro, S. B.
677: Medvedev, M. P. Fedoruk., and V. K. Mezentsev, Comp. Rend. Phys. \textbf{4},
678: 145 (2003).
679:
680: \bibitem{Ming} R. Grimshaw, J. He, and B. A. Malomed, Phys. Scripta \textbf{\ \ 53}, 385 (1996).
681:
682: \bibitem{Kaplan} A. Kaplan, B. V. Gisin, and B. A. Malomed, J. Opt. Soc. Am.
683: B \textbf{19}, 522 (2002).
684:
685: \bibitem{Pare'} C. Par\'{e}, A. Villeneuve, P. A. Belanger, N. J. Doran,
686: Opt. Lett. \textbf{21}, 459 (1996); C. Pare, A. Villeneuve, and S.
687: LaRochelle, Opt. Commun. \textbf{160}, 130 (1999).
688:
689: \bibitem{Uri} R. Driben, B. A. Malomed, M. Gutin, and U. Mahlab, Opt.
690: Commun. \textbf{218}, 93 (2003); R. Driben, B. A. Malomed, and U. Mahlab,
691: \textit{ibid}. \textbf{232}, 129 (2004).
692:
693: \bibitem{NMfiber-general} I. R. Gabitov and P. M. Lushnikov, Opt. Lett.
694: \textbf{27}, 113 (2002); K. Beckwitt, F. O. Ilday, and F. W. Wise, \textit{\
695: ibid}. \textbf{29}, 763 (2004); J. D. Ania-Castanon, I. O. Nasieva, N.
696: Kurukitkoson, S. K. Turitsyn, C. Borsier, and E. Pincemin, Opt. Commun.
697: \textbf{233}, 353 (2004).
698:
699: \bibitem{NMfiber-laser} F. O. Ilday and F. W. Wise, J. Opt. Soc. Am. B
700: \textbf{19}, 470 (2002).
701:
702: \bibitem{Javid} J. Atai and B. A. Malomed, Phys. Lett. A \textbf{298}, 140
703: (2002).
704:
705: \bibitem{Isaac} I. Towers and B. A. Malomed, J. Opt. Soc. Am. \textbf{19},
706: 537 (2002).
707:
708: \bibitem{tandem} S. Carrasco, D. V. Petrov, J. P. Torres, L. Torner, H. Kim,
709: G. Stegeman, and J. J. Zondy, Opt. Lett. \textbf{29}, 382 (2004).
710:
711: \bibitem{SSM} R. Driben and B. A. Malomed, Opt. Commun. \textbf{185}, 439
712: (2000).
713:
714: \bibitem{FR} S. Inouye, M. R. Andrews, J. Stenger, H. J. Miesner, D. M.
715: Stamper-Kurn, and W. Ketterle, Nature \textbf{392}, 151 (1998); J. L.
716: Roberts, N. R. Claussen, J. P. Burke, C. H. Greene, E. A. Cornell, and C. E.
717: Wieman, Phys. Rev. Lett. \textbf{81}, 5109 (1998); J. Stenger, S. Inouye, M.
718: R. Andrews, H. J. Miesner, D. M. Stamper-Kurn, and W. Ketterle, \textit{ibid}
719: . \textbf{82}, 2422 (1999); S. L. Cornish, N. R. Claussen, J. L. Roberts, E.
720: A. Cornell, and C. E. Wieman, \textit{ibid}. \textbf{85}, 1795 (2000); E. A.
721: Donley, N. R. Claussen, S. L. Cornish, J. L. Roberts, E. A. Cornell, and C.
722: E. Wieman, Nature \textbf{412}, 295 (2001).
723:
724: \bibitem{GarciaGarcia} J. J. Garc\'{\i}a-Ripoll, V. M. P\'{e}rez-Garc\'{\i}a
725: and P. Torres, Phys. Rev. Lett. \textbf{83}, 1715 (1999); J. J.
726: Garc\'{\i}a-Ripoll and V. M. P\'{e}rez-Garc\'{\i}a, Phys. Rev. A
727: \textbf{59}, 2220 (1999).
728:
729: \bibitem{Tashkent} F. Kh. Abdullaev and R. Galimzyanov, J. Phys. B \textbf{\
730: 36}, 1099 (2003).
731:
732: \bibitem{stabilization} H. Saito and M. Ueda, Phys. Rev. Lett. \textbf{90},
733: 040403 (2003); F. Kh. Abdullaev, J. G. Caputo, R. A. Kraenkel, and
734: B. A. Malomed, Phys. Rev. A \textbf{67}, 013605 (2003); G. D.
735: Montesinos, V. M. P\'{e}rez-Garc\'{\i}a, and H. Michinel, Phys.
736: Rev. Lett. \textbf{92}, 133901 (2004); G. D. Montesinos, V. M.
737: P\'{e}rez-Garc\'{\i}a, and P. J. Torres PJ, Physica D
738: \textbf{191}, 193 (2004).
739:
740: \bibitem{Athens} P. G. Kevrekidis, G. Theocharis, D. J. Frantzeskakis, and
741: B. A. Malomed, Phys. Rev. Lett. \textbf{90}, 230401 (2003).
742:
743: \bibitem{DimDim} D. E. Pelinovsky, P. G. Kevrekidis, and D. J.
744: Frantzeskakis, Phys. Rev. Lett. \textbf{91}, 240201 (2003).
745:
746: \bibitem{VA} D. Anderson, Phys. Rev. A \textbf{27}, 3135 (1983); B. A.
747: Malomed, Progr. Opt. \textbf{43}, 71 (2002).
748:
749: \bibitem{SatsumaYajima} J. Satsuma, and N. Yajima, Progr. Theor. Phys.
750: (Suppl.) \textbf{55}, 284 (1974).
751: \end{thebibliography}
752:
753: \end{document}
754: