1: \documentclass[aps,onecolumn]{revtex4}
2: \usepackage{caption2}
3: \usepackage{float}
4: \usepackage{amsmath}
5: \usepackage{amssymb}
6: \usepackage{graphicx}
7:
8: \begin{document}
9:
10: \title{Multidimensional \textit{semi-gap} solitons in a periodic potential}
11: \author{Bakhtiyor B. Baizakov${1}$, Boris A. Malomed${2}$ and Mario
12: Salerno${1}$\\
13: ${1}$ Dipartimento di Fisica "E. R. Caianiello", Universit{\'a} di
14: Salerno, via S. Allende, I-84081 Baronissi (SA), Italy \\
15: ${2}$ Department of Interdisciplinary Studies, School of
16: Electrical
17: Engineering, Faculty of Engineering, \\
18: Tel Aviv University, Tel Aviv 69978, Israel}
19:
20: \begin{abstract} The existence, stability and other dynamical
21: properties of a new type of multi-dimensional (2D or 3D) solitons
22: supported by a transverse low-dimensional (1D or 2D, respectively)
23: periodic potential in the nonlinear Schr\"{o}dinger equation with
24: the self-defocusing cubic nonlinearity are studied. The equation
25: describes propagation of light in a medium with normal
26: group-velocity dispersion (GVD). Strictly speaking, solitons
27: cannot exist in the model, as its spectrum does not support a true
28: bandgap. Nevertheless, the variational approximation (VA) and
29: numerical computations reveal stable solutions that seem as
30: completely localized ones, an explanation to which is given. The
31: solutions are of the gap-soliton type in the transverse
32: direction(s), in which the periodic potential acts in combination
33: with the diffraction and self-defocusing nonlinearity.
34: Simultaneously, in the longitudinal (temporal) direction these are
35: ordinary solitons, supported by the balance of the normal GVD and
36: defocusing nonlinearity. Stability of the solitons is predicted by
37: the VA, and corroborated by direct simulations.
38: \end{abstract}
39:
40: \maketitle
41:
42: \section{Introduction}
43:
44: Recently, a variety of two- and three-dimensional (2D and 3D)
45: solitons have been investigated in models based on the nonlinear
46: Schr\"{o}dinger (NLS) or Gross-Pitaevskii (GP) equations with a
47: spatially periodic potential and cubic nonlinearity, see a review
48: \cite{review}. The physical models of this type emerge in the
49: context of Bose-Einstein condensation (BEC)
50: \cite{multi-d,bmsEPL,Yang,Estoril,bms,mih}, where the periodic
51: potential is created as an optical lattice (OL), i.e.,
52: interference pattern formed by coherent beams illuminating the
53: condensate, and in nonlinear optics, where similar models apply to
54: photonic crystals \cite{PhotCryst}. A different but allied setting
55: is provided by a cylindrical OL (``Bessel lattice"), which can
56: also support stable 2D \cite{Kartashov} and 3D \cite{mihalache}
57: solitons. Additionally, models combining a periodic lattice
58: potential and saturable nonlinearity give rise to 2D solitons,
59: that were predicted in Ref. \cite{PhotorefrPrediction} and
60: observed in several experiments in photorefractive media,
61: including fundamental solitons \cite{Photorefr2Dsolitons} and
62: vortices \cite{PhotorefrVortices}. It is also relevant to mention
63: that experimental observation of spatiotemporal self-focusing of
64: light in silica waveguide arrays, in the region of anomalous
65: group-velocity dispersion (GVD), was reported in Ref.
66: \cite{Cheskis}.
67:
68: In models with the cubic nonlinearity, these solutions were
69: investigated in a quasi-analytical form, which combines the
70: variational approximation (VA) \cite{va} to predict the shape of
71: the solitons, and the Vakhitov-Kolokolov (VK) criterion \cite{vk}
72: to examine their stability. Final results were provided by
73: numerical methods, relying upon direct simulations of the
74: underlying NLS/GP equations. A conclusion obtained by means of
75: these methods is that, unlike their 1D counterparts,
76: multi-dimensional solitons in periodic potentials can exist only
77: in a limited domain of the $\left( N,\varepsilon \right) $ plane,
78: where $N$ and $\varepsilon $ are the norm of the solution and
79: strength of the OL potential, respectively. The most essential
80: limitation on the existence domain of 2D solitons is that $N$
81: cannot be too small (in a general form, a minimum value of the
82: norm, as a necessary condition for the existence of 2D solitons
83: supported by lattice potentials, was discussed in Ref.
84: \cite{efremidis}). Unlike it, $\varepsilon $ may be arbitrarily
85: small, as even at $\varepsilon =0$ the 2D\ NLS equation has a
86: commonly known weakly unstable solution in the form of the
87: \textit{Townes soliton}, at a single value of the norm,
88: $N=N_{\mathrm{T}}$ \cite{Townes} ($N_{\mathrm{T}}\approx 11.7$ for
89: the NLS equation in the usual 2D form, $iu_{t}+\nabla
90: ^{2}u+|u|^{2}u=0$). Small finite $\varepsilon $ gives rise to a
91: narrow stability region,
92: \begin{equation}
93: 0<N_{\mathrm{T}}-N<\left( \Delta N\right) _{\max }\sim \varepsilon
94: \label{nearTownes}
95: \end{equation}for the 2D solitons \cite{bmsEPL}. Crossing the lower border of the
96: existence domain (\ref{nearTownes}) leads to disintegration of the localized
97: state into linear Bloch waves (radiation) \cite{bs}.
98:
99: In the case of the attractive cubic nonlinearity (which
100: corresponds to BEC where atomic collisions are characterized by a
101: negative scattering length, while this is the case of the normal,
102: self-focusing Kerr effect), 2D and 3D solitons can be stabilized
103: not only by the potential lattice whose dimension is equal to that
104: of the ambient space, but also by \textit{low-dimensional}
105: periodic potentials, whose dimension is smaller by one, i.e., 2D
106: and 3D solitons can be stabilized by a quasi-1D \cite{Estoril,bms}
107: or quasi-2D \cite{Estoril,bms,mih} OL, respectively [in the former
108: case, the qualitative estimate (\ref{nearTownes}) for the width of
109: the stability region at small $\varepsilon $ is correct too];
110: however, 3D solitons cannot be stabilized by a quasi-1D lattice
111: potential \cite{Estoril,bms} [this is possible if the 1D potential
112: is applied in combination with the \textit{Feshbach-resonance
113: management}, i.e., periodic reversal of the sign of the
114: nonlinearity coefficient \cite{Warsaw}, or in combination with
115: \textit{dispersion management}, i.e., periodically alternating
116: sign of the local GVD coefficient \cite{Michal}]. Solitons can
117: exist in such settings because the attractive nonlinearity
118: provides for stable self-localization of the wave function in the
119: free direction (one in which the low-dimensional potential does
120: not act), essentially the same way as in the 1D NLS equation, and,
121: simultaneously, the lattice stabilizes the soliton in the other
122: directions (in the 3D model with the quasi-1D OL potential, the
123: self-localization in the transverse 2D subspace, where the
124: potential does not act, is possible too, but the resulting soliton
125: is unstable, the same way as the above-mentioned Townes soliton).
126: An important aspect of settings based on the low-dimensional OL
127: potentials is mobility of the solitons along the free direction,
128: which opens the way to study collisions between solitons and
129: related dynamical effects \cite{bms}.
130:
131: In the case of defocusing nonlinearity, which corresponds to a positive
132: scattering length in the BEC, or self-defocusing nonlinearity in optics
133: (negative Kerr effect), the soliton cannot support itself in the free
134: direction. Localization in that direction may be provided by an additional
135: external confining potential; however, the resulting pulse is not a true
136: multidimensional soliton, but rather a combination of a \textit{gap soliton}
137: (a weakly localized state created by the interplay of the repulsive
138: nonlinearity and periodic potential \cite{multi-d}),which was recently
139: created experimentally in a 1D BEC \cite{Oberthaler}) in the direction(s)
140: affected by the OL, and of a \textit{Thomas-Fermi state}, directly confined
141: by the external potential in the remaining direction \cite{bms}.
142:
143: Thus, no soliton can be supported by a low-dimensional lattice in
144: the BEC model (GP equation) with self-repulsion (the latter
145: corresponds to the most common situation in the experiment
146: \cite{Oberthaler}). On the other hand, a new possibility may be
147: considered in terms of nonlinear optics. Indeed, one may combine
148: three physically relevant ingredients, viz., (i) an effective
149: periodic potential in the transverse direction(s), while the
150: medium is uniform in the propagation direction, (ii)
151: self-defocusing nonlinearity, and (iii) normal GVD. The latter is
152: readily available, as most optical materials feature normal GVD,
153: in compliance with its name. As concerns the negative cubic
154: nonlinearity, it is possible in semiconductor waveguides, or may
155: be engineered artificially, through the cascading mechanism, in a
156: quadratically nonlinear medium with a proper longitudinal
157: quasi-phase-matching \cite{QPM}. Also quite encouraging for the
158: study of multidimensional solitons proposed in this work are
159: recent observations of 1D \cite{jason1} and 2D
160: \cite{Photorefr2Dsolitons} solitons in optically induced waveguide
161: arrays (photonic lattices) with \emph{self-defocusing}
162: nonlinearity.
163:
164: The setting outlined above can be realized in both 2D and 3D
165: geometry, where the necessary transverse modulation of the
166: refractive index is provided, respectively, by the transverse
167: structure in a planar photonic-crystal waveguide, or in a
168: photonic-crystal fiber. To the best of our knowledge, in either
169: case the model is a novel one. A soliton in this medium, if it
170: exists, will be of a \textit{mixed type}: in the transverse
171: direction(s), it is, essentially, a 1D or 2D spatial gap soliton,
172: supported by the combination of the effective periodic potential
173: and self-defocusing nonlinearity, while in the longitudinal
174: direction it is a temporal soliton of the ordinary type, which is
175: easily sustained by the joint action of the self-defocusing
176: nonlinearity and normal GVD. Thus, one may anticipate stable
177: spatiotemporal\textit{\ solitons}, alias ``light bullets", in this
178: model. Due to their mixed character, they may be called
179: \textit{semi-gap solitons}. The issue is of considerable interest
180: in view of the lack of success in experiments aimed at the
181: creation of ``bullets" in more traditional nonlinear-optical
182: settings \cite{review}. The only earlier proposed scheme for the
183: stabilization of 2D spatiotemporal optical solitons in periodic
184: structures, that we are aware of, assumed the use of a planar
185: waveguide with constant self-focusing nonlinearity and
186: longitudinal dispersion management \cite{WarsawDM}.
187:
188: On the other hand, it is necessary to stress that, rigorously
189: speaking, completely localized solutions cannot exist in the
190: present model: its linear spectrum cannot give rise to any true
191: bandgap, in which genuine solitons could be found (see below);
192: instead, one may expect the existence of quasi-solitons,
193: consisting of a well-localized ``body" and small nonvanishing
194: ``tails" attached to it. Nevertheless, we will produce families of
195: solutions which seem as stable perfectly localized objects. This
196: is possible because the ``tails" may readily turn out to be so
197: tiny that they remain completely invisible in numerical results
198: (possibly being smaller than the error of the numerical scheme),
199: and, of course, they will be invisible in any real experiment. An
200: explanation to this feature is provided by the fact that bandgaps,
201: which ``almost exist" in the system's spectrum, do not exist in
202: the strict sense because they are covered by linear modes with
203: very large wavenumbers. As shown in Ref. \cite{Isaac}, in this
204: case the amplitude of the above-mentioned tails (which are
205: composed of the linear modes with very large wavenumbers) is
206: exponentially small. In fact, families of \emph{stable}
207: ``practically existing" solitons in a second-harmonic-generating
208: system with opposite signs of the GVD at the fundamental-frequency
209: and second harmonics, where solitons cannot exist in the rigorous
210: mathematical sense, were explicitly found in that system, in both
211: multi- \cite{Isaac} and one- \cite{Kale} dimensional settings.
212: Implicitly (without discussion of this issue), ``practically
213: existing" solitons (although, in this case, they were unstable
214: against small perturbations) were also found in a recent work
215: \cite{Alejandro}, which was dealing with a 2D model of a planar
216: nonlinear waveguide with the cubic nonlinearity, that features a
217: Bragg grating in the longitudinal direction, and is uniform along
218: the transverse coordinate. In the latter model, true solitons
219: cannot exist, as the spectrum of the system does not support a
220: full bandgap.
221:
222: The objective of the present paper is to explore 2D and 3D
223: spatiotemporal solitons (which may be, strictly speaking,
224: ``quasi-solitons", in the above sense, but feature completely
225: localized shapes) and their stability in the proposed medium. In
226: Section 2 we fix the mathematical form of the 2D version of the
227: model, analyze its spectrum, and apply the VA to the study of
228: solitons. In Section 3, direct numerical results demonstrating the
229: existence of very robust 2D solitons, and their delocalization
230: when the lattice strength $\varepsilon $ becomes too small, are
231: reported (the comparison with the VA prediction shows that the VA
232: provides for a crude approximation in the present model). In
233: Section 4, we additionally consider the effect of variation of the
234: nonlinearity coefficient along the propagation distance on 2D
235: solitons. Finally, numerical results for 3D solitons are collected
236: in Section 5, and the paper is concluded by Section 6.
237:
238: \section{Formulation and variational analysis of the two-dimensional model}
239:
240: The model of a planar waveguide corresponding to the outline given above is
241: based on the following variant of the 2D NLS equation for the local
242: amplitude $u(z,x,t)$ of the electromagnetic field:
243: \begin{equation}
244: iu_{z}-\frac{1}{2}u_{tt}+\frac{1}{2}u_{xx}+\varepsilon \cos (2x)u-|u|^{2}u=0,
245: \label{2dnlse}
246: \end{equation}
247: where $z$ is the propagation distance, $x$ is the transverse coordinate, and
248: $t$ is the reduced time, defined the same way as in fiber optics.
249: The signs in front of the GVD ($u_{tt}$) and cubic terms
250: correspond, as said above, to the normal GVD and self-defocusing
251: nonlinearity, $\varepsilon $ is the amplitude of the transverse
252: modulation of the refractive index (which is assumed sinusoidal,
253: but the results will be nearly the same for more realistic forms
254: of the modulation which correspond to the actual photonic-crystal
255: structure), and the period of the modulation is scaled to be $\pi
256: $. Dynamical invariants of Eq. (\ref{2dnlse}) are the norm of the
257: solution (in optics, it is the total energy),
258: \begin{equation}
259: N\equiv \int_{-\infty }^{+\infty }dx\int_{-\infty }^{+\infty
260: }dt|u(z,x,t)|^{2}, \label{N}
261: \end{equation}
262: together with the longitudinal momentum and Hamiltonian,
263: \begin{eqnarray}
264: P &=&i\int_{-\infty }^{+\infty }dx\int_{-\infty }^{+\infty }dt~u_{t}^{\ast
265: }u, \\
266: H &=&\int_{-\infty }^{+\infty }dx\int_{-\infty }^{+\infty
267: }dt\left[ \frac{1}{2}\left\vert u_{x}\right\vert
268: ^{2}-\frac{1}{2}\left\vert u_{t}\right\vert
269: ^{2}+\right. \nonumber \\
270: &&\left. \frac{1}{2}|u|^{4}-\varepsilon \cos (2x)|u|^{2}\right] .
271: \end{eqnarray}
272:
273: To find the linear spectrum of the model, one looks for solutions to the
274: linearized version of Eq. (\ref{2dnlse}) as
275: \begin{equation}
276: u(z,t,x)=\exp \left( ikz-i\omega t\right) F_{E}(x), \label{linearized}
277: \end{equation}
278: where $k$ is a real propagation constant, $\omega $ is an arbitrary real
279: eigenvalue,\begin{equation} E\equiv \omega ^{2}-2k, \label{E}
280: \end{equation}and $F_{E}(x)$ is a solution of the Mathieu equation,\begin{equation}
281: F^{\prime \prime }+2\varepsilon \cos \left( 2x\right) F+EF=0,
282: \label{Mathieu}
283: \end{equation}
284: corresponding to the eigenvalue $E$. It is commonly known that the Mathieu
285: equation gives rise to bandgaps in its own spectrum, i.e., to
286: forbidden intervals of the values of $E$, within which no regular
287: quasi-periodic solutions of Eq. (\ref{Mathieu}) can be found.
288: However, since all large values of $E$ ($E\gg \varepsilon $)
289: belong to the allowed band, where such solutions exist, it is
290: obvious that \emph{no} value of $k$ may fall in a forbidden
291: bandgap. Indeed, using Eq. (\ref{E}), one can construct \emph{any}
292: real value of the propagation constant as $k=\left( \omega
293: ^{2}-E\right) /2$, taking very large $E$ and, accordingly, very
294: large $\omega $.
295:
296: On the other hand, the same consideration suggests that, in some
297: cases, the necessary values of $E$ and $\omega $ may be very large
298: indeed. It was shown, in a general form, in Ref. \cite{Isaac} that
299: short-period waves corresponding to such large parameters, which
300: build up into a possible ``tail" attached to the soliton's ``body"
301: (that makes it a quasi-soliton), will have an exponentially small
302: amplitude, rendering the tail totally negligible (in particular,
303: it may be completely invisible in numerical solutions). Therefore,
304: it makes sense to look for ``practically existing" solitons in the
305: present model.
306:
307: We start searching for stationary solutions of the ``mixed" type, which, as
308: explained above, are expected to feature a gap-soliton (weakly localized)
309: shape along $x$ and strong ordinary localization in $t$, by adopting the
310: following variational ansatz,
311: \begin{equation}
312: u(z,x,t)=Ae^{ikz}\left[ x^{-1}\sin (ax)\right] \mathrm{sech}(\alpha t),
313: \label{2dansatz}
314: \end{equation}
315: where $a$ and $\alpha $ are, respectively, the transverse and longitudinal
316: inverse widths of the soliton, and $A$ is its amplitude. Using the obvious
317: Lagrangian representation of Eq. (\ref{2dnlse}) and well-known VA formalism
318: \cite{va}, one can readily derive the following equations for the parameters
319: of the ansatz,
320: \begin{equation}
321: a=\left( \frac{3\varepsilon }{1+(N/3\pi )^{2}}\right) ^{1/3}, \label{a}
322: \end{equation}
323: \begin{equation}
324: \alpha =\frac{aN}{3\pi },~A=\frac{N}{\sqrt{6}\pi },~k=\varepsilon
325: -\frac{\varepsilon ^{2/3}}{2\left( 27\pi ^{2}\right)
326: ^{1/3}}\frac{27\pi ^{2}+5N^{2}}{\left( 9\pi ^{2}+5N^{2}\right)
327: ^{2/3}}, \label{2dva}
328: \end{equation}
329: where $N$ is the norm defined by Eq. (\ref{N}). It follows from these
330: equations that the condition $dk/dN<0$, i.e., the necessary stability
331: condition, according to the above - mentioned VK criterion \cite{vk}, always
332: holds, as shown in Fig. \ref{fig1}.
333:
334: \begin{figure}[tbh]
335: \centerline{\includegraphics[width=8cm,height=6cm,clip]{figure1.eps}}
336: \caption{The propagation constant vs. norm for the 2D solitons, as predicted
337: by the variational approximation, Eq. (\protect\ref{2dva}), for different
338: values of the strength of the periodic potential $\protect\varepsilon $. The
339: negative slope, $dk/dN<0$, implies stability of the solitons according to
340: the Vakhitov-Kolokolov criterion.}
341: \label{fig1}
342: \end{figure}
343:
344: The stability of the soliton families, predicted by the VK criterion as per
345: Fig. \ref{fig1}, is, generally, corroborated by direct numerical
346: simulations, see the next section (with a caveat that the VA predicts the
347: shape of the solitons in only a qualitatively correct form, as explained
348: below). However, it should also be mentioned that the applicability of the
349: VK criterion to gap solitons in lattice models has never been proven,
350: therefore one should apply this method with due care. In particular, it is
351: known that the gap solitons may be unstable in some cases when they are
352: expected to be VK-stable \cite{VK-}, which is not very surprising, as the VK
353: criterion ignores complex stability eigenvalues, that may give rise to
354: oscillatory instabilities. More perplexing is the fact that some gap-soliton
355: families in a lattice model with the cubic-quintic nonlinearity, that are
356: formally predicted to be VK-unstable, are in reality \emph{completely stable}
357: \cite{VK+}.
358:
359: To complete the discussion of the VA, it is necessary to notice
360: that ansatz (\ref{2dansatz}) is irrelevant if it predicts $a\ll
361: 1$, as it would mean that the soliton is very broad in the
362: $x$-direction, and it does not feel the underlying lattice
363: structure, $\cos (2x)$ in Eq. (\ref{2dnlse}). We therefore limit
364: the applicability of the VA by a (roughly defined) condition,
365: $a>1$. According to Eq. (\ref{a}), this implies that the potential
366: must be strong enough,
367: \begin{equation}
368: \varepsilon >\frac{1}{3}\left( 1+\frac{N^{2}}{9\pi ^{2}}\right) .
369: \label{crit}
370: \end{equation}
371:
372: \section{Numerical results for two-dimensional solitons}
373:
374: Direct simulations of Eq. (\ref{2dnlse}) (propagation in $z$) started with
375: an initial localized waveform, which we took as ansatz (\ref{2dansatz}) with
376: the parameters predicted by Eqs. (\ref{2dva}), or just a Gaussian with
377: rather arbitrary parameters -- for instance,
378: \begin{equation}
379: u(x,t,0)=A\exp [-\frac{a}{2}(x^{2}+t^{2})]. \label{Gauss}
380: \end{equation}
381: It was observed that the initial waveform undergoes intense evolution,
382: shedding off some radiation waves that were absorbed at edges of the
383: integration domain. The domain was large enough -- in most cases, $\left(
384: -8\pi ,+8\pi \right) $ in both directions ($x$ and $t$) -- so that the
385: solitons, which are typically well localized within a region $\left(
386: -5, +5 \right) $, see Figs. 2, 3, 7, and 9 below, are not affected by the
387: absorbers.
388: \begin{figure}[tbh]
389: \centerline{\includegraphics[width=8cm,height=6cm,clip]{figure2.eps}}
390: \caption{Cross sections of a typical established 2D soliton of the
391: semi-gap type. It has self-trapped from the initial configuration
392: (\protect\ref{Gauss}) with $A=1$ and $a=1$. The solid and dashed
393: lines represent, respectively, the sections along the transverse
394: directions $x$ at $t=0$, and temporal direction $t$ at $x=0$. The
395: shapes of the cross sections are displayed on the logarithmic
396: scale, to illustrate the fundamental observation of the practical
397: vanishing of the tails attached to the soliton's ``body".}
398: \label{fig2}
399: \end{figure}
400:
401: The evolution of the pulse ends with the establishment of a 2D
402: stationary soliton, \emph{without any visible tails}, as shown on
403: the logarithmic scale in Fig. 2. It is relevant to mention that,
404: as seen from Figs. \ref{fig2} and \ref{fig3}, the characteristic
405: sizes of the soliton in the $x$ and $t$ directions are on the same
406: order of magnitude, hence the values of the GVD coefficient and
407: its effective lattice-diffraction counterpart are, roughly, equal.
408:
409: In fact, the solitons self-trap even from the initial
410: configurations that are quite different from their final shape,
411: which attests to strong robustness of the solitons. The soliton's
412: stability was then additionally tested against small random
413: perturbations, by simulating the evolution of the initial
414: configuration $U(x,t)=U_{0}(x,t)[1+\sigma u_{p}(x,t)]$, where
415: $U_{0}(x,t)$ is the numerically found soliton, $\sigma $ is a
416: small amplitude of the perturbation, and $u_{p}(x,t)$ is a random
417: function. An example of a stable localized state self-trapped from
418: the initial Gaussian (\ref{Gauss}) with $A=1$ and $a=1$, with the
419: norm $N_{0}=\pi $, is displayed in Fig. \ref{fig3}.
420:
421: It should be said that direct comparison of the VA predictions
422: with the numerical results shows only a qualitative agreement: for
423: example, the VA predicts, with the same value of the norm, a
424: soliton whose width in the temporal direction exceeds the actual
425: width of the soliton in Fig. \ref{fig3}, by \ factor in excess of
426: $2$, and, accordingly, the amplitude of the numerically found
427: soliton significantly exceeds that predicted by the VA. Therefore,
428: the VA provides for only a crude approximation in this model;
429: nevertheless, its qualitative predictions, such as the stability
430: of the solitons predicted by the VK criterion, are correct. In
431: this connection, it is relevant to mention that no good version of
432: VA has been thus far proposed for gap solitons (this technical
433: issue was considered, in some detail, in Ref. \cite{Arik}), and
434: the solitons in the present model are still more complex objects.
435: \begin{figure}[tbh]
436: \centerline{\includegraphics[width=8cm,height=4cm,clip]{figure3.eps}}.
437: \caption{A stable localized solution of Eq. (\protect\ref{2dnlse})
438: with $\protect\varepsilon =2.0$, produced from the initial
439: Gaussian pulse (\protect \ref{Gauss}) with the initial norm
440: $N_{0}=\protect\pi $. The amplitude, inverse width along the
441: temporal direction, and norm of this localized state are $A=0.74$,
442: $\protect\alpha =0.56$, and $N=2.29$, respectively. Note that
443: $27\%$ of the initial norm was lost with emitted radiation in the
444: course of the evolution. Notice that the soliton features no
445: extended tail.} \label{fig3}
446: \end{figure}
447:
448: Further numerical analysis of the 2D model has demonstrated that,
449: following the known pattern of the delocalization transition of 2D
450: solitons in lattice potentials \cite{bs}, the soliton solutions
451: cease to exist when the strength of the periodic potential
452: $\varepsilon $ or the norm of the localized state $N$ fall below
453: some critical values. In this case, the localized waveform
454: undergoes disintegration, transforming into a quasi-linear
455: nonstationary extended state. This state keeps expanding until it
456: eventually hits edge absorbers, thus completely disappearing.
457: Naturally, the expansion occurs faster in the temporal (alias
458: longitudinal) direction, where it is not impeded by any potential
459: structure. Figure \ref{fig4} illustrates the disintegration of the
460: localized state (the one from Figure \ref{fig3}), following
461: gradual decrease of $\varepsilon $ along the propagation
462: direction. It is relevant to stress that the disintegration of the
463: soliton at small $\varepsilon $ is inevitable, as Eq.
464: (\ref{2dnlse}) with $\varepsilon =0$ has no 2D soliton solution,
465: unlike the Townes soliton, which would be a solution for the
466: equation with $\varepsilon =0$ and reverse sign in front of the
467: $u_{xx}$ term.
468:
469: \begin{figure}[tbh]
470: \centerline{\includegraphics[width=8cm,height=4cm,clip]{figure4a.eps}}
471: \centerline{\includegraphics[width=8cm,height=4cm,clip]{figure4b.eps}}
472: \caption{Dynamical disintegration of the localized state shown in
473: Fig. \protect\ref{fig3} as a result of a gradual decrease of the
474: strength of the periodic potential along the propagation distance,
475: so as $\protect\varepsilon (z)=\protect\varepsilon
476: _{0}(1-z/z_{\mathrm{end}})$, with $\protect\varepsilon _{0}=2$ and
477: $z_{\mathrm{end}}=1800$. Cross sections of the wave profile are
478: displayed: (a) along the transverse coordinate $x$; (b) in the
479: temporal direction $t$. The delocalization occurs around $z=650$,
480: at $\protect\varepsilon $ close to a critical value,
481: $\protect\varepsilon _{\mathrm{cr}}\simeq 1.3$.} \label{fig4}
482: \end{figure}
483:
484: Detecting the delocalization transition of the semi-gap solitons at critical
485: values of the nonlinear coefficient (or rescaled norm) and/or the strength
486: of the periodic potential can be used to locate the lower border of their
487: existence region in the parameter space $\left( N,\varepsilon \right) $.
488: Actually, the transition from the localized state to the extended one is
489: quite steep (see Fig. \ref{fig5}), which allows quite accurate determination
490: of the critical values of the parameters. However, delineating the full
491: existence region of the semi-gap solitons is a harder problem, as the
492: limiting effect at large values of the norm, which determines the upper
493: border, is \emph{splitting} of the soliton (see below), rather than collapse
494: (singularity formation) in the case of lattice gap solitons with the
495: self-focusing nonlinearity \cite{bmsEPL}. Precise shapes of the existence
496: and stability domains of lattice gap solitons can be rather complex, as
497: shown in Ref. \cite{Kartashov} for the case of saturable nonlinearity.
498:
499: Accurate determination of the full stability borders for the solitons in the
500: present model, going beyond the use of the VK criterion and collection of
501: typical examples of direct simulations, will be a subject of separate work.
502: \begin{figure}[tbh]
503: \centerline{\includegraphics[width=8cm,height=6cm,clip]{figure5.eps}}
504: \caption{The amplitude of the 2D semi-gap soliton abruptly decays at the
505: point of the delocalization transition, which was displayed in Fig. \protect
506: \ref{fig4}. The present figure illustrates the steepness of the transition
507: at $\protect\varepsilon \simeq 1.3$, with the slowly decreasing strength of
508: the periodic potential, starting from $\protect\varepsilon =2.0$. }
509: \label{fig5}
510: \end{figure}
511:
512: Finally, Eq. (\ref{2dnlse}) features obvious Galilean invariance
513: in the longitudinal ($t$) direction, which makes it possible to
514: generate a \textit{boosted} soliton $u_{c}$, with an arbitrary
515: inverse-velocity shift $c$, from a given soliton $u$, as
516: \[
517: u_{c}(z,t)=u(z,t-cz)e^{-i\left( c^{2}/2\right) z-ict}.
518: \]
519: The use of such two pulses with the $c_{1}\neq c_{2}$ makes it possible to
520: study collisions between the moving solitons, which, however, should be a
521: subject of a separate work.
522:
523: \section{Effects of nonlinearity modulation}
524:
525: As the nonlinearity is a key factor necessary for the existence of solitons,
526: in this section we address response of the 2D solitons to variation of the
527: nonlinearity strength along the propagation coordinate, $z$. In nonlinear
528: optics, a variable nonlinearity coefficient can be created by dint of
529: different physical mechanisms, such as variation of a dopant concentration,
530: optically controlled photorefraction, or simply by using a variable
531: thickness of the planar waveguide.
532:
533: Thus, we replace Eq. (\ref{2dnlse}) by its variant which includes
534: a variable nonlinear coefficient $\chi (z)$,
535: \begin{equation}
536: iu_{z}-\frac{1}{2}u_{tt}+\frac{1}{2}u_{xx}+\varepsilon \cos
537: (2x)u-\chi (z)|u|^{2}u=0. \label{chi}
538: \end{equation}
539: Then, we follow the transformation of the soliton with slow increase or
540: decrease of $\chi (z)$. Figure \ref{fig6} displays the evolution of the
541: soliton's parameters as the nonlinearity coefficient gradually increases
542: ten-fold.
543:
544: \begin{figure}[tbh]
545: \centerline{\includegraphics[width=8cm,height=6cm,clip]{figure6.eps}}
546: \caption{The evolution of the norm ($N$), amplitude ($A$), and
547: inverse temporal width ($\protect\alpha $) resulting from the slow
548: linear increase of the nonlinear coefficient $\protect\chi $ in
549: Eq. (\protect\ref{2dnlse}), $\protect\chi (z)=1+\protect\gamma
550: z/z_{\mathrm{end}}$,\textrm{\ }with $\protect\gamma =9$ and
551: $0<z<z_{\mathrm{end}}=2000$. Note significant decrease of the norm
552: $N$, opposed by little variation of $A$ and $\protect\alpha $. The
553: initial localized state is the same as in Fig.
554: \protect\ref{fig3}.} \label{fig6}
555: \end{figure}
556: In this case, Fig. \ref{fig7} shows that the final waveform,
557: corresponding to $\chi =10$, has not changed notably compared to
558: initial one (see Fig. \ref{fig3}), which implies that the increase
559: of the nonlinearity is countered by the loss of the norm. The
560: excess norm is shed off with linear waves which are absorbed on
561: the domain boundaries. It appears that the shape of the localized
562: wave, being weakly sensitive to the value of the norm (hence, to
563: the strength of the nonlinearity too), is actually fixed by the
564: strength $\varepsilon $ of the periodic potential. This
565: shape-invariance property of the mixed-type solitons is very
566: different from what is manifested by both ordinary solitons and
567: gap solitons per se, whose shapes are particularly sensitive to
568: the strength of the nonlinearity, at a fixed amplitude of the
569: periodic potential.
570:
571: \begin{figure}[tbh]
572: \centerline{\includegraphics[width=8cm,height=4cm,clip]{figure7.eps}}
573: \caption{The final shape of the soliton from Fig. \protect\ref{fig3},
574: established after by slow increase of the nonlinearity by a factor of $10$,
575: as illustrated in the previous figure. Note that an extended tail attached
576: to the soliton does not appear in this case either.}
577: \label{fig7}
578: \end{figure}
579:
580: If the coefficient of nonlinearity slowly \emph{decreases}, as $\chi
581: (z)=1-z/z_{\mathrm{end}}$ with $z_{\mathrm{end}}=2000$, the disintegration
582: of the localized state is observed when falls to the level of $\chi \simeq
583: 0.5$, resembling the picture in Fig. \ref{fig4}. Thus, we conclude that the
584: strengths of both the periodic potential and nonlinearity must exceed some
585: critical values in order to sustain the localized states. In this respect,
586: the multidimensional semi-gap solitons resemble regular gap solitons in
587: lattice potentials \cite{efremidis,bs}.
588:
589: One of characteristic features of ordinary solitons in 1D nonintegrable
590: systems is splitting of the soliton when the GVD coefficient \cite{grimshaw}
591: or nonlinearity \cite{Radik} is abruptly changed. In the present model, one
592: can observe a similar effect for 2D solitons. Figure \ref{fig8} displays an
593: example of the splitting of a soliton into two fragments, which then
594: separate along the free direction $t$, after the\ nonlinearity coefficient
595: was suddenly increased by an order of magnitude.
596:
597: \begin{figure}[tbh]
598: \centerline{\includegraphics[width=8cm,height=6cm,clip]{figure8.eps}}
599: \caption{Splitting of the soliton from Fig. \protect\ref{fig2}
600: along the free direction $t$, following the rapid increase of the
601: nonlinear coefficient by a factor of $10$, which is performed by
602: setting $\protect\chi (z)=1+\protect\gamma \tanh
603: (4z/z_{\mathrm{end}})$ in Eq. (\protect\ref{chi}), with
604: $\protect\gamma =9$ and $z_{\mathrm{end}}=10$.} \label{fig8}
605: \end{figure}
606:
607: \section{The three-dimensional case}
608:
609: The 3D version of Eq. (\ref{2dnlse}) has a straightforward form,
610: \begin{equation}
611: iu_{z}-\frac{1}{2}u_{tt}+\frac{1}{2}(u_{xx}+u_{yy})+\varepsilon \lbrack \cos
612: (2x)+\cos (2y)]u-|u|^{2}u=0. \label{3dnlse}
613: \end{equation}
614: Similarly to the 2D case, stationary solutions to Eq.
615: (\ref{3dnlse}) can be numerically found by using a Gaussian pulse
616: as the initial condition and propagating in $z$. Stability of the
617: 3D solitons was verified by simulating the evolution of a soliton
618: with a random perturbation added to it. As well as in the 2D case,
619: the simulations were performed with the absorbers placed at
620: borders of the integration domain (under the condition that the
621: length of the domain was much larger than a characteristic size of
622: the soliton). As a result, it was concluded that robust 3D
623: solitons exist as generic solutions, without any visible tails
624: attached to them. An example of a stationary 3D soliton, which was
625: found to be quite robust in stability simulations, is displayed in
626: Fig. \ref{fig9}. In particular, Fig. \ref{fig9}(b) clearly
627: demonstrates that, as well as in the 2D case, the soliton has,
628: roughly, equal sizes in the spatial and temporal directions, i.e.,
629: the GVD and diffraction modified by the periodic potential play
630: equally important roles in supporting the solitons.
631:
632: \begin{figure}[tbh]
633: \centerline{\includegraphics[width=8cm,height=4cm,clip]{figure9a.eps}}
634: \centerline{\includegraphics[width=8cm,height=4cm,clip]{figure9b.eps}}
635: \caption{A three-dimensional stationary soliton solution to Eq.
636: (\protect\ref{3dnlse}) with $\protect\varepsilon =2.0$ is shown
637: through its two cross sections: (a) perpendicular to the free
638: direction, in the $t=0$ plane, and (b) parallel to the free
639: direction, in the $y=0$ plane. The soliton was obtained by direct
640: propagation in $z$ of an initial Gaussian with the norm
641: $N_{0}=2\protect\pi $. The norm of the established soliton is
642: $N=4.16$, i.e., a third of the initial norm was lost in the course
643: of the adjustment of the initial pulse to the stationary shape of
644: the 3D soliton.} \label{fig9}
645: \end{figure}
646:
647: Besides the fundamental 3D pulses, such as the one displayed in
648: Fig. \ref{fig9}, the model can also support 3D solitons with
649: embedded \textit{vorticity}. Analogy with known results for
650: gap-soliton vortices in the 2D lattice models \cite{GapVortex}
651: suggests that the vortex solitons too may easily be stable in the
652: present model. However, detailed investigation of the vortex
653: solutions, as well as collecting systematic data about the family
654: of the fundamental 3D solitons, requires numerous runs of lengthy
655: simulations of the 3D equation, which is beyond the scope of the
656: present paper.
657:
658: \section{Conclusion}
659:
660: In this work, we have proposed a new type of the multidimensional
661: model in nonlinear optics. It combines self-defocusing
662: nonlinearity and normal group-velocity dispersion with periodic
663: modulation of the local refractive index in the one or two
664: transverse directions (in the 2D and 3D models, respectively).
665: Strictly speaking, multidimensional (spatiotemporal) solitons
666: cannot exist in media of this type, as the system's spectrum
667: contains no true bandgap. Nevertheless, solitons which seem as
668: completely localized ones are predicted by the variational
669: approximation, and found in direct simulations. These solitons are
670: solutions of a mixed type, as in the free (longitudinal, alias
671: temporal) direction they are regular solitons, while in the
672: transverse direction(s) they are objects of the gap-soliton type
673: (hence the solution as a whole was called a \textit{semi-gap
674: soliton}). The existence of the solitons requires that both the
675: norm of the solution (in other words, the nonlinearity strength
676: $\chi $) and the strength $\varepsilon $ of the spatially periodic
677: transverse potential exceed certain minimum values, otherwise the
678: pulses decay into linear waves. Actually, the solitons are much
679: more sensitive to $\varepsilon $ than to $\chi $.
680:
681: The results reported in this paper call for further work, that should be
682: aimed at accurate identification of borders of the solitons' stability
683: regions, especially in the 3D model (which requires running very massive
684: simulations), investigation of collisions between solitons, that may move
685: freely in the longitudinal direction, and the study of vortex solitons in
686: the 3D case.
687:
688: \section*{Acknowledgements}
689:
690: B.B.B. thanks the Department of Physics at the University of
691: Salerno (Italy) for a two-year research grant. B.A.M. appreciates
692: hospitality of the same Department. The work of this author was
693: partially supported by the grant No. 8006/03 from the Israel
694: Science Foundation. M.S. acknowledges a partial financial support
695: from the MIUR, through the inter-university project PRIN-2003.
696:
697: \begin{thebibliography}{99}
698: \bibitem{review} B. A. Malomed, D. Mihalache, F. Wise, and L. Torner, J.
699: Opt. B: Quant. Semicl. Opt. \textbf{7}, R53 (2005).
700:
701: \bibitem{multi-d} B. B. Baizakov, V. V. Konotop, M. Salerno, J. Phys. B
702: \textbf{35}, 5105 (2002); E. A. Ostrovskaya and Y. S. Kivshar, Phys. Rev.
703: Lett. \textbf{90}, 160407 (2003); H. Sakaguchi and B. A. Malomed, J. Phys. B
704: \textbf{37}, 1443 (2004).
705:
706: \bibitem{bmsEPL} B. B. Baizakov, B. A. Malomed and M. Salerno, Europhys.
707: Lett. \textbf{63}, 642 (2003).
708:
709: \bibitem{Yang} J. Yang and Z. Musslimani, Opt. Lett. \textbf{23}, 2094
710: (2003); Y. V. Kartashov, A. A. Egorov, L. Torner, D. N. Christodoulides,
711: Opt. Lett. \textbf{29}, 1918 (2004).
712:
713: \bibitem{Estoril} B. B. Baizakov, M. Salerno, and B. A. Malomed, in
714: \textit{Nonlinear Waves: Classical and Quantum Aspects}, (Kluwer Academic
715: Publishers, Dordrecht, 2004) edited by F. Kh. Abdullaev and V. V. Konotop,
716: p. 61; also available at \newline
717: http://rsphy2.anu.edu.au/\symbol{126}asd124/Baizakov\_2004\_61\_Nonlinear
718: Waves.pdf
719:
720: \bibitem{bms} B. B. Baizakov, B. A. Malomed and M. Salerno, Phys. Rev. A
721: \textbf{70}, 053613 (2004).
722:
723: \bibitem{mih} D. Mihalache, D. Mazilu, F. Lederer, Y. V. Kartashov, L.-C.
724: Crasovan, and L. Torner, Phys. Rev. E \textbf{70}, 055603 (2004).
725:
726: \bibitem{PhotCryst} P. Xie, Z.-Q. Zhang, and X. Zhang, Phys. Rev. E
727: \textbf{\ 67}, 026607 (2003); A. Ferrando, M. Zacar\'{e}s,
728: P. Fern\'{a}ndez de C\'{o}rdoba, D. Binosi, and J. A. Monsoriu,
729: Opt. Exp. \textbf{11}, 452 (2003);
730: \textit{ibid}. \textbf{12}, 817 (2004).
731:
732: \bibitem{Kartashov} Y. V. Kartashov, V. A. Vysloukh, and L. Torner, Phys.
733: Rev. Lett. \textbf{93}, 093904 (2004); \textbf{94}, 043902 (2005).
734:
735: \bibitem{mihalache} D. Mihalache, D. Mazilu, F. Lederer, B. A. Malomed, Y.
736: V. Kartashov, L.-C. Crasovan, and L. Torner, Phys. Rev. Lett. \textbf{95},
737: 023902 (2005).
738:
739: \bibitem{PhotorefrPrediction} N. K. Efremidis, S. Sears, D. N.
740: Christodoulides, J. W. Fleischer, and M. Segev\textit{, }Phys. Rev. E
741: \textbf{66}, 046602 (2002).
742:
743: \bibitem{Photorefr2Dsolitons} J.~W. Fleischer, M. Segev, N.~K. Efremidis,
744: and D.~N. Christodoulides, Nature \textbf{422}, 147 (2003).
745:
746: \bibitem{PhotorefrVortices} D. N. Neshev, T. J. Alexander, E. A.
747: Ostrovskaya, and Y. S. Kivshar, H. Martin, I. Makasyuk, and Z. Chen, Phys.
748: Rev. Lett. \textbf{92}, 123903 (2004); J. W. Fleischer, G. Bartal, O. Cohen,
749: O. Manela, M. Segev, J. Hudock, and D. N. Christodoulides, \textit{ibid}.
750: \textbf{92}, 123904 (2004); G. Bartal, O. Manela, O. Cohen, J. W. Fleischer,
751: and M. Segev, \textit{ibid}. \textbf{95}, 053904 (2005).
752:
753: \bibitem{Cheskis} D. Cheskis, S. Bar-Ad, R. Morandotti, J. S. Aitchison, H.
754: S. Eisenberg, Y. Silberberg, and D. Ross, Phys. Rev. Lett. \textbf{91},
755: 223901 (2003).
756:
757: \bibitem{va} D. Anderson, Phys. Rev. \textbf{A27}, 1393 (1983); B. A.
758: Malomed, Progr. Optics \textbf{43}, 69 (2002).
759:
760: \bibitem{vk} M. G. Vakhitov and A. A. Kolokolov, Radiophys. Quantum
761: Electron. \textbf{16}, 783 (1973).
762:
763: \bibitem{efremidis} N. K. Efremidis, J. Hudock, D. N. Christodoulides, J. W.
764: Fleischer, O. Cohen, and M. Segev, Phys. Rev. Lett. \textbf{91}, 213906
765: (2003).
766:
767: \bibitem{Townes} L. Berg\'{e}, Phys. Rep. \textbf{303}, 259 (1998).
768:
769: \bibitem{bs} B. B. Baizakov and M. Salerno, Phys. Rev. A \textbf{69}, 013602
770: (2004).
771:
772: \bibitem{Warsaw} M. Trippenbach, M. Matuszewski, and B.A. Malomed, Europhys.
773: Lett. \textbf{70}, 8 (2005); M. Matuszewski, E. Infeld, B. A. Malomed, and
774: M. Trippenbach, Phys. Rev. Lett. \textbf{95}, 050403 (2005).
775:
776: \bibitem{Michal} M. Matuszewski, E. Infeld, B. A. Malomed, and M.
777: Trippenbach, \textit{Stabilization of three-dimensional light bullets by a
778: transverse lattice in a Kerr medium with dispersion management}, Opt.
779: Commun., in press (2005).
780:
781: \bibitem{Oberthaler} B. Eiermann, Th. Anker, M. Albiez, M. Taglieber, P.
782: Treutlein, K.-P. Marzlin, and M. K. Oberthaler, Phys. Rev. Lett.
783: \textbf{92}, 230401 (2004).
784:
785: \bibitem{QPM} O. Bang, C. B. Clausen, P. L. Christiansen, and L. Torner,
786: Opt. Lett. \textbf{24}, 1413 (1999).
787:
788: \bibitem{jason1} J. W. Fleischer, T. Carmon, M. Segev, N. K. Efremidis and
789: D. N. Christodoulides, Phys. Rev. Lett. \textbf{90}, 023902 (2003).
790:
791: \bibitem{WarsawDM} M. Matuszewski, M. Trippenbach, B. A. Malomed, E. Infeld,
792: and M. Skorupski, Phys. Rev. E \textbf{70}, 016603 (2004).
793:
794: \bibitem{Isaac} I. N. Towers, B. A. Malomed, and F. W. Wise, Phys. Rev.
795: Lett. \textbf{90}, 123902 (2003).
796:
797: \bibitem{Kale} K. Beckwitt, Y.-F. Chen, F. W. Wise and B. A. Malomed, Phys.
798: Rev. E \textbf{68}, 057601 (2003).
799:
800: \bibitem{Alejandro} T. Dohnal and A. B. Aceves, Stud. Appl. Math., \textbf{115}, 209 (2005).
801:
802: \bibitem{Arik} A. Gubeskys, B. A. Malomed, and I. M. Merhasin, Stud. Appl.
803: Math. \textbf{115}, 255 (2005).
804:
805: \bibitem{VK-} P. J. Y. Louis, E. A. Ostrovskaya, C. M. Savage, and Y. S.
806: Kivshar, Phys. Rev. A \textbf{67}, 013602 (2003); Y. V. Kartashov, V. A.
807: Vysloukh, L. Torner, Optics Express, \textbf{12}, 2831 (2004).
808:
809: \bibitem{VK+} I. M. Merhasin, B. V. Gisin, R. Driben, and B. A. Malomed,
810: Phys. Rev. E \textbf{71}, 016613 (2005).
811:
812: \bibitem{grimshaw} R. Grimshaw, J. He, and B.A. Malomed, Phys. Scripta
813: \textbf{53}, 385 (1996).
814:
815: \bibitem{Radik} R. Driben and B. A. Malomed, Opt. Commun. \textbf{185}, 439
816: (2000).
817:
818: \bibitem{GapVortex} E. A. Ostrovskaya and Y. S. Kivshar, Phys. Rev. Lett.
819: \textbf{93}, 160405 (2004); H. Sakaguchi and B. A. Malomed, J. Phys. B
820: \textbf{37}, 2225 (2004).
821: \end{thebibliography}
822:
823: \end{document}
824: