1: %\documentstyle[12pt,aps,pra]{revtex}
2: \documentclass[twocolumn,aps,pre,showpacs,epsfig]{revtex4}
3: \usepackage[dvips]{graphics}
4: \usepackage{graphicx}
5: \usepackage{amsfonts}
6: \usepackage{amssymb}
7: \usepackage{amsmath}
8: \begin{document}
9: \title{Discrete breathers in BEC with two- and three-body
10: interactions in optical lattice}
11: \author{F.Kh. Abdullaev$^{1}$ , A. Bouketir$^{2}$, A. Messikh$^{3}$,
12: B.A. Umarov$^{1,3}$\footnote[7]{Corresponding author
13: (b\_umarov@yahoo.com)}} \affiliation{$^{1}$Physical-Technical
14: Institute, Uzbek Academy of Sciences, 700084, Tashkent-84,
15: G.Mavlyanov str, 2-b, Uzbekistan}
16: %\author{A. Bouketir}
17: \affiliation{$^{2}$Department of Science in
18: Engineering, Faculty of Engineering}
19: %\author{A.Messikh, B.A. Umarov\footnote[7]{Corresponding author(b\_umarov@yahoo.com)}}
20: \affiliation{$^{3}$Department of
21: Computational and Theoretical Sciences,
22: Faculty of Science, International Islamic University Malaysia, P.O.
23: Box 10, 50728 Kuala Lumpur, Malaysia}
24: \date{\today}
25: %\maketitle
26: %\vspace{1cm}
27: \begin{abstract}
28: We investigate the properties of discrete breathers in a
29: Bose-Einstein condensate with three-body interactions in optical
30: lattice. In the tight-binding approximation the Gross-Pitaevskii
31: equation with periodic potential for the condensate wavefunction
32: is reduced to the cubic-quintic discrete nonlinear Schr\"odinger
33: equation. We analyze the regions of modulational instabilities of
34: nonlinear plane waves. This result is important to obtain the
35: conditions for generation of discrete solitons(breathers) in
36: optical lattice. Also using the Page approach, we derive the
37: conditions for the existence and stability for bright discrete
38: breather solutions. It is shown that the quintic nonlinearity
39: brings qualitatively new conditions for stability of strongly
40: localized modes. The numerical simulations conform with the
41: analytical predictions.
42: \end{abstract}
43: \pacs{03.75.Lm;03.75.-b;05.30.Jp} \maketitle
44: %\begin{multicols}{2}
45: \section{Introduction}
46: Discrete breathers are important eigenmodes of the nonlinear
47: discrete and periodic systems and their existence has been
48: predicted for such systems as optical waveguide arrays, array of
49: Josephson junctions, magnetic chains etc \cite{rev,rev_f,Leder01}.
50: Recently discrete breathers in array of Bose-Einstein condensates
51: has attracted a great deal of attention
52: \cite{Tromb01,Abd01a,Kon04,Alfimov,Kevr04,AK,Morsh}. The localized
53: atomic matter wave in BEC in the optical lattice(a gap soliton)
54: has been observed recently in \cite{Ober04}. The theoretical
55: approaches mainly based on the analysis of the Gross-Pitaevskii
56: equation with periodic potential and cubic mean field
57: nonlinearity. It corresponds to taking into account elastic
58: two-body interactions and the potential of optical lattice. The
59: progress with Bose-Einstein condensates on the surface of atomic
60: chips and in atomic waveguides involves strong compression of BEC
61: and so to essential increasing the density of BECs. Then it pose
62: the problem of the taking into account three-body interactions
63: effects. This interaction represent interest also for the
64: understanding of the fundamental limits for the functioning of
65: devices using BEC \cite{Pu}. The existence of three-body
66: interactions can play important role in the condensate stability
67: \cite{Abd01b,Akhmediev}. Recently 3-body interaction arising due
68: to the Efimov resonance has been observed in an ultra-cold gas of
69: cesium atoms \cite{Kraemer}.
70:
71: Thus represent interest to investigate properties of discrete
72: breathers in BEC in optical lattices when two- and three-body
73: interactions are taken into account. One of the interesting
74: properties of this system is the existence of the gap-Townes
75: soliton\cite{AS05}. In the tight-binding approximation(see a
76: section below) we show that the GP equation with periodic
77: potential can be reduced to the cubic-quintic discrete nonlinear
78: Schr\"odinger(CQDNLS) equation. This regime occurs for the deep
79: optical lattice, which is realized in typical experiments in the
80: BEC.
81:
82: For the generation of discrete breathers the modulational
83: instability of nonlinear plane wave represent the possible
84: mechanism. Thus in this paper in the section 3 we discuss the
85: modulational instability in the CQDNLS equation, the stability
86: region of the wave in the lattice. In section 4 we discuss the
87: localized modes of the different symmetries and the conditions of
88: their stability.
89:
90:
91: \section{The model}
92:
93: Let consider BEC with two and three body interactions in optical
94: lattice and in highly anisotropic trap corresponding to quasi 1D
95: geometry. The Gross-Pitaevskii equation is
96: \begin{widetext}
97: %long equation goes here
98: \begin{equation}
99: i\hbar\psi_{t} = -\frac{\hbar^2}{2m}\nabla^{2}\psi + \frac{m}{2}
100: (\omega_{\perp}^2 \rho^2 + \omega_{x}^{2}x^{2}) + V_{opt}(\rho,x)
101: - g_{1}|\psi|^{2}\psi - g_{2}|\psi|^{4}\psi,
102: \end{equation}
103: \end{widetext}
104: where $g_1, \, g_2$ are coefficients proportional to two- and
105: three-body elastic interactions. For rubidium atoms it is expected
106: that the three-body interaction is attractive and $|g_3| \sim
107: 10^{-26}-10^{-27}$cm$^6$/s \cite{Pu}. Here $\omega_{\perp} \gg
108: \omega_{x}$, $V_{opt} = V_{0}\sin^{2}(kx)R(\rho).$ We look for the
109: solution of the form
110: \begin{equation}
111: \psi(r,t) =\phi_{0}(\rho) \phi (x,t),
112: \end{equation}
113: where $\phi_{0}$ is the solution of the radial linear equation
114: \begin{equation}
115: \frac{\hbar^2}{2m}\nabla_{\rho}^{2}\phi_{0} + \frac{m}{2}
116: \omega_{\perp}^{2}\rho^{2}\phi_{0} = \frac{1}{2}\hbar
117: \omega_{\perp}\phi_{0}.
118: \end{equation}
119: The solution for the ground state is
120: $$\phi_{0} = \sqrt{\frac{1}{\pi a_{\perp}^2}}
121: \exp(-\frac{\rho^{2}}{2a_{\perp}^2}).$$
122:
123: Multiplying both sides of the GP equation by $\phi_{0}$ and
124: integrating over transverse variable $\rho$ we obtain the quasi 1D
125: GP equation
126: \begin{widetext}
127: \begin{equation}
128: i\hbar \phi_{t} =-\frac{\hbar^2}{2m}\phi_{xx} + (\frac{m}{2}
129: \omega_{x}^{2}x^{2} + V_{0}\sin^{2}(kx))\phi - \frac{g_{1}}{2\pi
130: a_{\perp}^2}|\phi|^{2}\phi - \frac{g_{2}}{3\pi^2
131: a_{\perp}^4}|\phi|^{4}\phi.
132: \end{equation}
133: \end{widetext}
134: Introducing the dimensionless variables
135: $$t = t\nu, x = kx, \phi =
136: u\sqrt{\frac{2a_{s}\omega_{\perp}}{\nu}}, \nu =
137: \frac{E_{R}}{\hbar},$$ where $E_{R} = \hbar^2 k^2/2m$ is the
138: recoil energy,
139: we obtain the equation
140: %
141: \begin{equation}\label{GP}
142: iu_{t} + \frac{1}{2}u_{xx} + \lambda |u|^{2}u + \bar{\lambda}
143: |u|^{4}u - V_{0} \cos(2x)u =0.
144: \end{equation}
145: %
146: where $\lambda = \pm 1$ and
147: $$\bar{\lambda} = \frac{g_{2}}{3\pi^2 \hbar \nu a_{\perp}^4}
148: \left( \frac{\nu}{2a_{s}\omega_{\perp}} \right)^{5/2}.$$
149: %
150: In the case of deep optical lattice ($U \gg E_{R}$ i.e. $\nu \gg
151: 1$) the tight-binding approximation can be used
152: \cite{Tromb01,Abd01a,Kon04,Alfimov}. If the potential $U(x)$
153: periodic, i.e. $U(x + l) =U(x)$, the eigenvalue problem is:
154: \begin{equation}
155: \frac{d^{2}\phi_{\alpha,q}}{dx^2} + U(x)\phi_{\alpha,q} =
156: E_{\alpha,q} \phi_{\alpha,q},
157: \end{equation}
158: where for the optical lattice potential $U(x) = V_{0}\cos(2x)$.
159: Here $\phi_{\alpha,q}$ is the Bloch function, $\alpha$ is the
160: index labelling the energy bands $E_{\alpha}(q)$. Periodicity of
161: the lattice admits the expansion of the energy
162: $$E_{\alpha,q+q_{0}} = E_{\alpha,q} = E_{\alpha,-q},\
163: q_{0}=\frac{2\pi}{l}.$$ Thus we can expand the energy in the
164: Fourier series
165: \begin{equation}
166: E_{\alpha,q} = \sum_{n}\hat{\omega}_{n,\alpha}e^{iqnl}, \
167: \hat{\omega}_{n,\alpha} = \frac{l}{2\pi}\int_{-q_{0}/2}^{q_{0}/2}
168: E_{q,\alpha}e^{-iqnl}dq.
169: \end{equation}
170: For deep lattice case it is convenient to use the Wannier
171: functions\cite{Alfimov}:
172: \begin{equation}
173: w_{n,\alpha} = \sqrt{\frac{l}{2\pi}}\int_{-q_{0}/2}^{q_{0}/2}
174: \phi_{\alpha,q}(x)e^{-iqnl}dq,
175: \end{equation}
176: and
177: \begin{equation}
178: \phi_{\alpha,q} =
179: \sqrt{\frac{l}{2\pi}}\sum_{n}w_{n,\alpha}(x)e^{iqnl}.
180: \end{equation}
181: The main property of Wannier functions that they are strongly
182: localized in the bottoms of the potential and so are suitable for
183: the description of strongly localized modes in the periodic
184: potential. They form an orthonormal and a complet set
185: \begin{eqnarray}
186: \int_{-\infty}^{\infty}w_{n,\alpha}^{\ast}w_{n',\alpha'}(x)dx =
187: \delta_{n,n'}\delta_{\alpha,\alpha'},\nonumber\\
188: \sum_{n,\alpha}w_{n,\alpha}^{\ast}(x)w_{n,\alpha}(x') =
189: \delta(x-x').
190: \end{eqnarray}
191: Let look for the solution of CQ GP equation of the form of
192: expansion in Wannier functions
193: \begin{equation}
194: u(x,t) = \sum_{n,\alpha}c_{n,\alpha}(t)w_{n,\alpha}(x).
195: \end{equation}
196: Substituting this expansion into the GP equation (\ref{GP}) we
197: obtain the system of equation for coefficients $c_{n,\alpha}$
198:
199: \begin{widetext}
200: \begin{eqnarray}
201: ic_{n,\alpha;t} &=& \sum
202: c_{n_{1},\alpha}\hat{\omega}_{n-n_{1},\alpha}
203: %\nonumber\\&&
204: + \lambda
205: \sum_{\alpha_{1},\alpha_{2},\alpha_{3}}\sum_{n_{1},n_{2},n_{3}}
206: c_{1}^{\ast}c_{2}c_{3}W^{nnnn}_{\alpha\alpha\alpha\alpha}
207: %\nonumber\\&&
208: +
209: \bar{\lambda}\sum_{5\alpha}\sum_{5n}c^{\ast}_{1}c^{\ast}_{2}c_{3}c_{4}c_{5}
210: W_{6\alpha}^{6n},
211: \end{eqnarray}
212: \end{widetext}
213: where $$W_{nn_1n_2n_3}^{\alpha\alpha_1\alpha_2\alpha_3} =
214: \int_{-\infty}^{\infty}w_{n,\alpha}w_{n_1,\alpha_1}w_{n_2,\alpha_2}w_{n_3,\alpha_3}\
215: dx.$$ For deep optical lattice $V_0 > 5 E_R$ are important terms
216: with $n = n_{1}= n_{2} = n_{53}$ i.e. we are restricted by the
217: band $\alpha$. Then the system is reduced for equation
218:
219: \begin{widetext}
220: \begin{equation}
221: iC_{n,\alpha;t} = \hat{\omega_{1,\alpha}}(C_{n-1,\alpha} +
222: C_{n-1,\alpha}) + \lambda W_{4\alpha}^{4n}|C_{n}|^{2}C_{n} +
223: \bar{\lambda} W_{6\alpha}^{6n} |C_{n}|^{4}C_{n}.
224: \end{equation}
225: \end{widetext}
226: By the transformation $v_{n}
227: =C_{n,\alpha}\sqrt{W_{4\alpha}^{4n}}C_{n}$ this equation can be
228: transformed into
229: %
230: \begin{equation}\label{dcq}
231: iv_{n,t} + \kappa(v_{n+1} + v_{n-1}) + \lambda |v_{n}|^{2}v_{n} +
232: {\gamma} |v_{n}|^{4}v_{n}=0.
233: \end{equation}
234: %
235: Here $\kappa = \hat{\omega_{1,\alpha}}, \gamma = \bar{\lambda}
236: W_{6\alpha}^{6n}/(2(W_{4\alpha}^{4n})^{2}).$ It is the
237: cubic-quintic discrete nonlinear Schr\"odinger(CQDNLS) equation.
238: As showed the analysis performed in \cite{Alfimov}, the region of
239: parameters exists, when the tight-binding approximation describes
240: the GP equation with periodic potential with a high accuracy. It
241: is achieved for the amplitudes of the periodic potential $|V_0| >
242: 5 E_R, \,$ and $E_R$ is the recoil energy. That corresponds to
243: the experimentally realized values. When $\gamma =0$ the problem
244: is reduced to considered previously by
245: \cite{Tromb01,Abd01a,Alfimov}. In the recent work \cite{Malomed06}
246: the case when $\lambda >0, \gamma < 0$, motivated by the analysis
247: of localized states in array of waveguides with saturable
248: nonlinearity, has been considered. In the BEC case the parameter
249: $\gamma$ can has any sign. The conserved quantities are: The
250: norm(number of atoms) $N$ is:
251: \begin{equation}\label{norm}
252: N = \sum_{n}|v_{n}|^{2},
253: \end{equation}
254: and the Hamiltonian $H$ is:
255: \begin{equation}\label{ham}
256: H = -\sum_{n=-\infty}^{\infty}\left[ \kappa(v_{n}^{\ast}v_{n+1} +
257: v_{n}v_{n+1}^{\ast})
258: + \frac{\lambda}{2} |v_{n}|^{4} + \frac{\gamma}{3}
259: |v_{n}|^{6} \right].
260: \end{equation}
261: The equation of motion is:
262: \begin{equation}
263: iu_{n,t} = \frac{\delta H}{\delta u_{n}^{\ast}}.
264: \end{equation}
265: The momentum is not conserved due to the discretness of the
266: system. Due to the periodicity of the system the Hamiltonian is
267: periodic with period $n=1$. The different excitations induces
268: different effective periodic potential like the Peierls-Nabarro
269: potential. The height of the barrier depends on the localization
270: of the discrete soliton solution - grows with the decreasing of
271: the soliton width. The analysis of the Hamiltonian can help for
272: the calculation of the PN barrier between different discrete
273: solitons and so help for the investigation of the stability of
274: nonlinear modes \cite{Cai}.
275:
276: \section{Modulational instability of nonlinear plane
277: waves}
278:
279:
280: The nonlinear discrete equation has the plane wave solution
281: \begin{equation}\label{eq18}
282: v_n(t)=\nu_0 e^{i(qn+\omega t)},
283: \end{equation}
284: with the dispersion relation
285: \begin{equation}\label{eq19}
286: \omega = 2k\cos q +\lambda {\nu_{0}}^2 +\gamma {\nu_{0}}^{4}.
287: \end{equation}
288:
289: For special importance here are the staggered $q = \pi$ and
290: unstaggered $q = 0$ solutions\cite{ADG} where the dispersion
291: relation can be rewritten as
292: \begin{equation}\label{eq20}
293: \omega = \pm 2k +\lambda {\nu_{0}}^2 +\gamma {\nu_{0}}^4.
294: \end{equation}
295: The linear stability of the solution (\ref{eq18}) and (\ref{eq19})
296: can be investigated by looking for a solution in the form
297: \begin{equation}\label{eq21}
298: v_n(t)=(\nu_0+\delta v_n (t)) e^{i(qn+\omega t)},
299: \end{equation}
300:
301: where $\delta v_n (t)$ is a small perturbation on the
302: carrier wave, substituting (\ref{eq21}) in (\ref{dcq}) and keeping
303: only the linear terms on $\delta v_n (t)$ and $\delta v^\ast _n
304: (t)$ we obtain the following equation
305:
306: \begin{widetext}
307: \begin{equation}\label{eq22}
308: i\delta v_{n,t} +k\cos q (\delta v_{n+1}+\delta v_{n-1}-2 \delta
309: v_{n})+(\lambda+2\gamma {\nu_0}^2)(\delta v_{n}+\delta
310: v^\ast_{n})=0.
311: \end{equation}
312: \end{widetext}
313: Considering a modulation in the form
314: \begin{equation}
315: \delta v_{n} (t) = \left(
316: \begin{array}{cc}
317: \alpha \\
318: \bar{\beta}
319: \end{array}\right) e^{i(Qn+\Omega t)},
320: \end{equation}
321: where $Q$ and $\Omega$ are the wave number and the
322: frequency of the linear modulation waves respectively then
323: substituting it into (\ref{eq22}) we obtain the following system
324: of linear equations
325: \begin{eqnarray}
326: %\begin{array}
327: i \Omega \alpha +4k \cos q \sin^2 (Q/2)\bar{\beta}=0, \nonumber\\
328: -(4k\cos q \sin^2 (Q /2)+2(\lambda+2\gamma \nu_0 ^2)\nu_0
329: ^2)\alpha +i\Omega \bar{\beta}=0.
330: %\end{array}
331: \end{eqnarray}
332: In order to have a
333: nontrivial solution for the system the determinant of the
334: coefficient matrix should be zero that leads to the following
335: dispersion law
336:
337: \begin{widetext}
338: \begin{eqnarray}\label{eq25}
339: \Omega ^2 =8k \sin^2(Q/2)[2k\cos^2 (q)\sin^2 (Q/2)-\cos q
340: (\lambda+2\gamma \nu_0 ^2) \nu_0 ^2].
341: \end{eqnarray}
342: \end{widetext}
343: The plane wave solution (\ref{eq18}) is modulationally unstable
344: provided that $\Omega^2$ is negative and the gain is
345: \begin{widetext}
346: \begin{eqnarray}\label{eq26}
347: G=2|\sin(Q/2)|\sqrt{2k\cos q(\lambda+2\gamma \nu_0 ^2)\nu_0
348: ^2-2k\cos q \sin^2(Q/2))}.
349: \end{eqnarray}
350: \end{widetext}
351: From the gain equation we can see that the unstaggered solution
352: $(q=0)$ are unstable whenever $\lambda \nu_0 ^2 +2\gamma \nu_0^4
353: >0$ and the staggered solutions $(q=\pi)$ are unstable
354: whenever $\lambda \nu_0 ^2 +2\gamma \nu_0^4 < 0$ which leads to
355: the following cases:
356:
357: i) $q=0$. If $\lambda>0$, $\gamma<0$ then we have MI of the
358: unstaggered solution for $|\gamma|<\lambda/2\nu_0 ^2$ and if
359: $\lambda<0$, $\gamma>0$ then we have MI when
360: $|\gamma|>\lambda/2\nu_0 ^2$.
361:
362: ii) $q=\pi$. If $\lambda>0$, $\gamma<0$ then we have MI of the
363: staggered solution for $|\gamma|>\lambda/2\nu_0 ^2$ and if
364: $\lambda<0$, $\gamma>0$ then we have MI when
365: $|\gamma|<\lambda/2\nu_0 ^2$.
366:
367: iii) In contradiction with the continuum case which
368: is stable for the set of nonlinearities $\lambda<0$, $\gamma<0$
369: the staggered solution is modulationally instable regardless the
370: value of any parameter. However the unstaggered solution should
371: satisfy the inequality $\lambda \nu_0 ^2 +2\gamma \nu_0 ^4>2k$,
372: which is almost satisfied all the time for strongly localized
373: modes, in order to have MI.
374:
375: In the general case where $q$ can take any value we can
376: distinguish two regions of MI as it is shown in Fig. 1 for
377: positive $k \cos q$.
378:
379: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% FIG.1 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
380: \begin{figure}[here]
381: \includegraphics[height=8cm, width=8cm]{figure1.eps}
382: \caption{Modulational instability regions as function of the
383: parameters $\Lambda=\gamma/2k{\nu_0}^2 \cos q$ and
384: $B=\beta/4{\nu_0}^2 \cos q$. with $\cos q>0$, the labelled points
385: are the points discussed in the text. }
386: \end{figure}
387: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
388:
389: One is fully unstable and the other is conditional means it
390: depends on the wave number $q$ of the carrier wave and the wave
391: number $Q$ of modulation wave as it is shown in Fig. 2.
392:
393: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% FIG.2 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
394: \begin{figure}[here]
395: \includegraphics[height=8cm, width=8cm]{figure2.eps}
396: \caption{Regions of modulational instability in the plane $(q,Q)$
397: for the conditional stability region in Fig. 1, the labelled
398: points are the points discussed in the text. }
399: \end{figure}
400: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
401:
402: The conditional region represents the effect of the discreteness
403: on the modulational instability. Clearly we can see that if we fix
404: the cubic nonlinearity parameter and keep changing the quintic
405: nonlinearity parameter then we will have instability, conditional
406: instability and stability which shows that the quintic term can
407: lead to collapse of MI. The role of quintic term also is very
408: crucial on the wave numbers of the carrier and modulation waves
409: leading to MI, where increasing $\gamma$ in the negative direction
410: will shrink the MI region in $(q, Q)$ plane till a critical value
411: $\gamma_{cr}$. This eliminates all possible chances to get MI then
412: MI occurs again in the $(q, Q)$ plane in a mirror symmetry with
413: the one occured for $\gamma < \gamma_{cr}$. In the continuum
414: limit, when $Q<<1$ and $q<< 1$ equation (\ref{eq25}) reduced to
415: \begin{equation}\label{eq27}
416: \Omega^2=2kQ^2[kQ^2/2 -(\lambda+2\gamma \nu_0 ^2)\nu_0 ^2]
417: \end{equation}
418: and coincides with the one obtained in \cite{Darm98}. This MI gain
419: does not depend on the wave number q of the carrier wave contrary
420: to the discrete case.
421:
422: Based on the analytical results, the modulational instabilities of
423: carrier wave with wave number $q$ modulated by small oscillation
424: with wave number $Q$ occurs when the right hand side of dispersion
425: equation (\ref{eq25}) is negative. In order to check our
426: analytical results we have performed numerical simulations of
427: equation (\ref{dcq}) using fourth order Runge-Kutta scheme with a
428: time step chosen to preserve our conserved quantities which are
429: the total energy of the system and the number of atoms to accuracy
430: more than 10$^{-4}$. The numerical simulations have been performed
431: for a chain of $N=180$ units, with periodical boundary conditions
432: so that the wave numbers $q$ and $Q$ satisfy the relations $q=2\pi
433: m/N$ and $Q=2\pi M/N$, where $m$ and $M$ are integers. The other
434: parameters have been chosen to be $k=0.1$, $\lambda=1$ with
435: $\gamma$ varying, the amplitude of the modulation wave has been
436: chosen small compared to the amplitude of the carrier wave as
437: $\nu_0=0.5$, $\delta \nu_0=0.005$. The stability and instability
438: regions as it was predicted by the analytical results has been
439: checked for different points , starting with point labelled (1) in
440: Fig.1 that correspond to zero quintic term with $q=Q=\pi/3$ and
441: $q=\pi/3$, $Q=2\pi/3$. Our expectation of modulational instability
442: has been confirmed as it is shown in Fig. 3 and independently from
443: $q$ and $Q$.
444: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% FIG.3 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
445: \begin{figure}[here]
446: \includegraphics[height=8cm, width=8cm,angle=-90]{figure3.eps}
447: \caption{Modulational instability arises at different times at
448: point labelled (3) started immediately and at the point labelled
449: (4) the solution is stable for a long time before the occurrence
450: of MI. }
451: \end{figure}
452: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
453:
454:
455:
456: The role of
457: discreteness investigated within the conditional instability
458: region of Fig. 1 with quintic parameter $\gamma=-1.2$, where we
459: choose two points: the point labelled (3) with $q=\pi/9$,
460: $Q=2\pi/3$ and the point labelled (4) with $q=\pi/3$, $Q=2\pi/3$
461: from Fig.2 that correspond the same point labelled (2) in Fig. 1.
462: The numerical results demonstrate the dependence of stability on
463: $q$ and $Q$ as it is shown in Fig. 4.
464:
465: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% FIG.4 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
466: \begin{figure}[here]
467: \includegraphics[height=8cm, width=8cm,angle=-90]{figure4.eps}
468: \caption{ Modulational instability of the plane wave solution
469: occurs and do not occurs at the same point in plane $(q, Q)$ with
470: different $\gamma > \gamma_c$ $\gamma < \gamma_c$ }
471: \end{figure}
472: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
473:
474:
475: \section{Intrinsic localized modes(discrete
476: breathers)} In this section we study the strongly localized
477: solutions, which are standing and occupying few sites. In the case
478: when the coupling between sites are weak the analytical approach
479: can be used to find the discrete soliton (breather) solutions.
480:
481: We used Page approach \cite{Page} in order to examine the
482: existence of the most known localized modes: even, twisted and odd
483: modes(see Fig. 5). Twisted and even modes has been combined to one
484: mode where both can be described by one formula as far as the
485: evolution of one can be other after varying the quintic
486: nonlinearity term.
487:
488: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% FIG.5 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
489: \begin{figure}[here]
490: \includegraphics[height=10cm,width=5cm]{figure5a.eps}
491: \caption{Schematic representation of different kind of odd modes
492: (a unstaggered and b staggered), even modes (c unstaggered and d
493: staggered)and twisted modes (e unstaggered and f staggered).
494: }
495: \end{figure}
496: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
497:
498: By considering solution for equation (\ref{dcq}) as
499: $v_n(t)=\nu_n {\rm e}^{(i\omega t)}$, where $\nu_n$ represents
500: the respective amplitudes of a localized mode one can transform
501: equation (\ref{dcq}) to a system of algebraic equations depends on
502: the topology of the localized mode, hence we derive the conditions
503: of existence and then with linearization we derive the conditions
504: of stability for each mode as follows.\\
505:
506: {\bf Even mode}\\
507:
508: The even/twisted mode defined by its amplitudes as
509: $v_n=\nu_0\left(\cdots,0,\alpha,1,s,s\alpha,0,\cdots\right)$ ,
510: where , $n=\pm1,\pm2,\cdots$, as we are concerned with the
511: symmetric modes only $s=\pm1$. Varying the signs of $s$ and
512: $\alpha$ produces the different topologies of even/twisted modes
513: showed in Fig. 5. Taking in consideration the requirements for
514: strongly localized modes which are $|\alpha|\ll1$ and
515: $\alpha_n\approx0$ for $n>2$
516: we derive the dispersion relation and the formula of
517: the secondary
518: amplitude respectively as
519: \begin{eqnarray}
520: \omega&\equiv&
521: \omega_{ET}=k(\alpha+s)+\nu_0^2(\lambda+\gamma\nu_0^2)
522: \label{eq2},
523: \\
524: \alpha&=&\frac{k}{\nu_0^2(\lambda+\gamma\nu_0^2)} \label{eq3}.
525: \end{eqnarray}
526:
527: Hence a set of conditions for existence of strongly localized
528: modes can be derived as it is in table 1.
529: \begin{table}[h]
530: \begin{center}
531: \begin{tabular}{|l|c|}
532: \hline
533: Mode & Existence conditions\\
534: \hline
535: Odd (a) & $\lambda>-\gamma\,\nu_0^2$\\
536: \hline
537: Odd (b) & $\lambda<-\gamma\,\nu_0^2$\\
538: \hline
539: Even (c) and Twisted (e) & $\lambda>-\gamma\,\nu_0^2$\\
540: \hline
541: Even (d) and Twisted (f)& $\lambda<-\gamma\,\nu_0^2$\\
542: \hline
543: \end{tabular}
544: \caption{ Conditions for existences of different kinds of modes
545: shown in Fig. 5 and for positive coupling.}
546: \end{center}
547: \label{tab1}
548: \end{table}
549:
550:
551: To study the stability of these modes we followed the approach
552: developed in \cite{Darm98} by using a linear analysis, where we
553: impose a perturbation $\delta_n(t)$ on each non zero excitation
554: amplitude, hence the mode's amplitude can be written now as
555: $$\nu_n=\nu_0(\cdots,0,r\alpha+\delta_{-2},1+\delta_{-1},
556: s+\delta_1,\alpha+\delta_2,0,\cdots),$$ inserting it in
557: (\ref{dcq}) and with subsequent linearization we got eight-order
558: system of equations. The change of variables as
559: $\delta^{\pm}_j=\delta_{+j}\pm\delta_{-j}$ ($j$=1,2) reduces the
560: system to two independent four-order equations systems. Separating
561: the real and imaginary parts of the perturbation
562: $\delta^{\pm}_j=\delta^{\pm}_{jr}+i\,\delta^{\pm}_{ji}$
563: and introducing the scaled time $\tau=\omega_c\,t$,
564: we obtain
565:
566: \begin{widetext}
567: \begin{eqnarray}
568: \frac{d \delta^{\pm}}{d \tau}= \left(
569: \begin{array}{cccc}
570: 0& (s-p)\alpha & 0 &\alpha\\
571: 2-(3s-p)r\alpha+\displaystyle\frac{2\gamma\nu_0^2(1-r\alpha)}{(\lambda
572: +\gamma\nu_0^2)}
573: &0&\alpha&0\\
574: 0& -\alpha&0&1\\
575: \alpha&0&-1&0\\
576: \end{array}
577: \right)\delta^{\pm}, \label{eq4}
578: \end{eqnarray}
579: \end{widetext}
580: where
581: $\delta^{\pm}=(\delta^{\pm}_{1r},\delta^{\pm}_{1i},\delta^{\pm}_{2r},
582: \delta^{\pm}_{2i})$ and $p=\pm1$ stands for the symmetric
583: $\delta^+_j$ and antisymmetric $\delta^-_j$ perturbation
584: respectively. If we introduce $\delta^{\mp}\propto {\rm
585: exp}(g\tau)$ then the eigenvalues $g$ of (\ref{eq4}) are given by
586: the following equation
587:
588: \begin{widetext}
589: \begin{eqnarray}
590: &&g^4+\left[ 1+ \left(
591: \frac{2\gamma\nu_0^2(p-s)}{(\lambda+\gamma\nu_0^2)}+2(3-2ps)
592: \right)\alpha^2+
593: 2(p-s)\left(1+\frac{\gamma\nu_0^2(p-s)}{\lambda+\gamma\nu_0^2})
594: \right) \alpha \right]g^2
595: \nonumber\\
596: &&+ 2(p-s)\left( 1+\frac{\gamma\nu_0^2}{\lambda+\gamma\nu_0^2}
597: \right)\alpha+ \left(
598: 2(3-2ps)+\frac{2\gamma\nu_0^2(1-p+s)}{\lambda+\gamma\nu_0^2}
599: \right)\alpha^2
600: \nonumber\\
601: &&+\left(
602: 2(p-2s)-\frac{2\gamma\nu_0^2(1-p+s)}{\lambda+\gamma\nu_0^2}
603: \right)\alpha^3+\alpha^4=0. \label{eq5}
604: \end{eqnarray}
605: \end{widetext}
606: Hence similarly to the case without quintic nonlinearity
607: \cite{Darm98}, when the symmetry of the mode coincides with the
608: symmetry of perturbation that is $p=s$ the localized modes will
609: be stable without any conditions on the nonlinearity parameters.
610: In the case where the mode and the perturbation have different
611: symmetries that $p=-s$ then there is a possibility for the mode
612: to be unstable with the instability gain
613: \begin{widetext}
614: \begin{eqnarray}
615: g=-2\sqrt{1+\frac{\gamma \nu_0^2}{\gamma
616: \nu_0^2+\lambda}}\sqrt{s\alpha}\left(1 -
617: \left(\frac{5s}{4}+\frac{\gamma \nu_0^2(1-2s)}{2(2\gamma
618: \nu_0^2+\gamma)} \right)\alpha \right), \label{eq6}
619: \end{eqnarray}
620: \end{widetext}
621:
622: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% FIG.6 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
623: %\begin{figure}[here]
624: %\includegraphics[height=8cm, width=8cm]{figure6.eps}
625: %\caption{Schematic representation of regions of stability for even
626: %twisted modes defined in figure 1. It is clear that the staggered
627: %even mode (d) is not stable at all since the region of existence
628: %is different from the region of stability. }
629: %\end{figure}
630: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
631:
632: In contrast to the case of cubic nonlinearity only, where the mode
633: is stable only when $\alpha s<0$, which means only twisted modes
634: ( staggered and unstaggered ) are stable, it is clear in the
635: cubic-quintic case that the first coefficient of (\ref{eq6})
636: leads also to the possibility of stability of even modes
637: (staggered and unstaggered) $\alpha s>0$. However taking the
638: existence consideration only the unstaggered even mode is stable.
639: The relationships between nonlinearities that control the
640: stability is given in table 2.
641: \begin{table}[h]
642: \begin{center}
643: \begin{tabular}{|l|c|}
644: \hline
645: Mode & Stability Conditions\\
646: \hline
647: Even (c \& d) & $\lambda<-2\gamma\,\nu_0^2$\\
648: \hline
649: Twisted (e \& f) & $\lambda>-2\gamma\,\nu_0^2$\\
650: \hline
651: \end{tabular}
652: \caption{Conditions for stability of even/twisted modes showed in
653: Fig. 1}
654: \end{center}
655: \end{table}
656: In the numerical simulations, first we checked the stability of
657: the even modes when the symmetry of the perturbation is the same
658: as the symmetry of the mode. We found that the numerical results
659: confirm the analytical predictions and the modes are stable
660: always. For the case when the perturbation and the mode have
661: opposite symmetry we checked the validity of analytically derived
662: stability regions for the modes.
663:
664: The most important result here is that the predicted stable even
665: unstaggered mode, which is not possible in the case with cubic
666: nonlinearity only, see \cite{Darm98} and references therein, has
667: been demonstrated numerically see Fig. 6 for $\gamma = -0.6$ which
668: is less than $\gamma= \lambda/2{v_{0}}^2 =-0.5$. However when
669: $\gamma=-0.4$ which is greater than $\gamma= \lambda/2{v_{0}}^2
670: =-0.5$, the mode is unstable and it is transformed to the odd mode
671: after some time as it is shown in Fig. 7. The even twisted modes
672: are shown analytically to be stable within two different regions
673: and this result has been checked numerically, which is in
674: agreement with \cite{Darm98} for the case of cubic nonlinearity only.\\
675:
676:
677: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% FIG.6 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
678: \begin{figure}[here]
679: \includegraphics[height=8cm, width=6cm,angle=-90]{figure6.eps}
680: \caption{Evolution of perturbed unstaggered even mode (c) $\lambda
681: =1, v_{0}=1, \gamma=-0.6$. }
682: \end{figure}
683: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
684:
685:
686: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% FIG.7 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
687: \begin{figure}[here]
688: \includegraphics[height=8cm, width=6cm,angle=-90]{figure7.eps}
689: \caption{Evolution of perturbed unstaggered even mode (c) $\lambda
690: =1, v_{0}=1, \gamma=-0.4$. }
691: \end{figure}
692: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
693:
694:
695:
696:
697: {\bf Odd mode}\\
698:
699: The odd mode is defined by its amplitude as
700: $\nu_n=\nu_0(\cdots,0,\beta,1,s\beta,0,\cdots)$, where
701: $n=0,\pm1,\pm2,\cdots$ and $s=\pm1$ with the requirements for
702: strong localized modes are $\beta\ll1$ and $\beta_n\approx 0$ for
703: $n>1$ and after performing similar algebraic operations to those
704: done for even/twisted mode we obtain the dispersion relation and
705: the formula of the secondary amplitude respectively as .
706: \begin{eqnarray}
707: \omega&\equiv&
708: \omega_o=k^2\frac{1+s}{\nu_0^2(\lambda+\gamma\nu_0^2)}+\lambda\nu_0^2+\gamma\nu_0^4,
709: \label{eq7}
710: \\
711: \beta&=&\frac{k}{\nu_0^2(\lambda+\gamma\nu_0^2)}, \label{eq8}
712: \end{eqnarray}
713: which leads to a set of conditions for the existence of this mode
714: as it is shown previously in table 1. To study the stability of
715: this mode we insert perturbation $\delta_{n}(t)$ on each non zero
716: excitation amplitude as
717: $$\nu_n=\nu_0(\cdots,0,\beta+\delta_{-1},1+\delta_0,s\beta+\delta_1,0,\cdots),$$
718: then substitute it in equation (\ref{dcq}) and perform
719: linearization procedures, as we did for the even/twisted mode, we
720: got a six-order system of equations, again a change of variables
721: as
722: $\delta_1^+=\delta_{+1}+\delta_{-1}$ reduces the system
723: to four-order equations system.
724: Separating the real and imaginary parts of the
725: perturbation $\delta^+_1=\delta_{+1r}+i\,\delta_{1i}$ and
726: $\delta_0=\delta_{0r}+i\,\delta_{0i}$, then introducing the scaled
727: time $\tau=\omega_c t$ we obtain
728: \begin{eqnarray}\label{eq35}
729: \frac{d\delta^{\rm od}}{d\tau}= \left(
730: \begin{array}{cccc}
731: 0&0&0&-\beta\\
732: 2+\displaystyle\frac{\gamma\nu_0^2}{\lambda+\gamma\nu_0^2}&0&\beta&0\\
733: 0&-2\beta&0&1\\
734: 2\beta&0&-1&0
735: \end{array}
736: \right)\delta^{\rm od},
737: \end{eqnarray}
738:
739: where $\delta^{\rm od}=(\delta_{0r}^{\rm od},\delta_{0i}^{\rm
740: od}, \delta_{1r}^{\rm od},\delta_{1i}^{\rm od})$ . If we introduce
741: $\delta^{\mp}\propto {\rm exp}(g\tau)$ then the eigenvalues $g$ of
742: (\ref{eq35}) are given by the following equation
743: \begin{eqnarray}
744: g^4+\left[1+4s\beta\right]g^2+4\left[1+\frac{\gamma\nu_0^2}{\lambda+
745: \gamma\nu_0^2}+\beta^2\right]\beta^2=0, \label{eq36}
746: \end{eqnarray}
747: which have four roots and the one that control the stability is
748: \begin{equation}
749: g=\frac{1}{2}\sqrt{-2-8\beta^2+2\sqrt{1-24\beta^2+
750: \frac{16\lambda\beta^2}{\lambda\nu_0^2+\lambda}}}, \label{eq37}
751: \end{equation}
752: which shows that the instability gain Re($g$) will be a nonzero
753: only if $|\beta|>\beta_c$ with
754: $\beta_c=\displaystyle\pm\frac{1}{\sqrt{8}}\sqrt{\frac{\lambda+
755: \gamma\nu_0^2}{\lambda+3\gamma\nu_0^2}}$. For $\gamma =
756: \gamma_c=-\lambda/3\nu_0^2$, $\beta_c$ will be undefined, but we
757: can conclude from equation (\ref{eq36}) that the odd mode will be
758: stable whatever the value of $\beta$. It is important to mention
759: here that the value of $\beta_c$ is depend on the interplay
760: between the cubic and quintic nonlinearities and satisfies the
761: condition for strongly localized modes $\beta\ll1$ for a range of
762: values of these nonlinearities, in contrast to the case of the
763: CDNLSE discussed in \cite{Darm98} where $\beta_c = 1/\sqrt{8}$
764: which showed that the strongly localized mode is always stable
765: because $\beta < \beta_c$ always. Furthermore for the stability of
766: localized odd mode in CQDNLSE the condition
767: $\gamma>\gamma_c=-\lambda/3\nu_0^2$ is necessary for focusing
768: cubic nonlinearity and defocusing quintic nonlinearity. Finally we
769: should mention here that for both even/twisted modes and odd modes
770: cases, if we substitute $\gamma=0$ in our results, we get back to
771: the cubic case and hence we reproduce the results of
772: \cite{Darm98}. Numerical results for the case of odd modes
773: presented in Figs. 8-9, confirmed the new predicted result that is
774: the quintic nonlinearity term can lead to an unstable staggered
775: odd mode when $-2<\gamma<-1.5$, this range is chosen to keep the
776: mode strongly localized means the second amplitude $0.1\ \le \beta
777: \le 0.2$. However the unstaggered odd mode is stable always as it
778: is shown in Fig. 10 in agreement with the results for the case of
779: cubic nonlinearity only \cite{Darm98}.
780:
781: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% FIG.8 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
782: \begin{figure}[here]
783: \includegraphics[height=8cm, width=6cm,angle=-90]{figure8.eps}
784: \caption{Evolution perturbed of staggered odd mode (b) $\lambda
785: =1, v_{0}=1, \gamma=-1.7, \beta_c=0.140 < \beta=0.167$. }
786: \end{figure}
787: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
788:
789:
790: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% FIG.9 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
791: \begin{figure}[here]
792: \includegraphics[height=8cm, width=6cm,angle=-90]{figure9.eps}
793: \caption{Evolution of perturbed staggered odd mode (b) $\lambda
794: =1, v_{0}=1, \gamma=-1.6, \beta_c=0.146 > \beta=0.143$. }
795: \end{figure}
796: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
797:
798:
799: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% FIG.10 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
800: \begin{figure}[here]
801: \includegraphics[height=8cm, width=6cm,angle=-90]{figure10.eps}
802: \caption{Evolution of perturbed unstaggered odd mode (a) $\lambda
803: =1, v_0=1, \gamma=-0.3$.
804: }
805: \end{figure}
806: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
807:
808: \section{Conclusion}
809: In conclusion we have investigated the modulational instabilities
810: and discrete breathers in BEC with two- and three-body
811: interactions in the optical lattice. Following to the procedure
812: suggested in \cite{Alfimov} we derive the lattice nonlinear model
813: which describe the GP equation with periodic potential. This
814: approximation is reasonable for the atomic wave function is
815: approximated by a single Wannier function - so called a Wannier
816: soliton. The modulational instability in a lattice is the process
817: leading to the generation of discrete breathers in the optical
818: lattice. We find the regions of instabilities of nonlinear plane
819: wave and conform analytical predictions by the direct numerical
820: simulations of the CQDNLS equation. The lattice dispersion vary
821: sign in dependence of the quasi-momentum value in the Brilloin
822: zone. Thus we can control the regions of instability even for the
823: fixed signs of two and three body interactions. This circumstance
824: extend the possibility for creation of the localized modes in BEC
825: in optical potential. We find the expressions for the amplitudes
826: and frequencies of different strongly localized modes in optical
827: lattice such as even, odd, twisted stuggered and unstaggered
828: modes. We analyze the stability of these modes and found the
829: stability regions which are necessary for the search of these
830: modes in the experiments with BEC in optical lattice. For the
831: further investigations will be interesting to consider the regimes
832: beyond of the tight-binding approximation(the deep optical lattice
833: limit) by accounting the higher order terms in the Wannier
834: expansion. In particular it concerns the extensions of CQDNLSE
835: involving the discrete nonlinear models with nonlocal
836: nonlinearities\cite{TS2,Alfimov,KevPel,Khare}.
837:
838:
839:
840:
841:
842: \begin{thebibliography}{99}
843:
844: \bibitem{rev}
845: S. Flash and C.R. Willis, Phys.Rep. {\bf 295}, 181 (1998);
846:
847: \bibitem{rev_f}
848: The focus issue on "Nonlinear localized modes:physics and
849: applications," Chaos {\bf 13}, 586 (2003).
850:
851: \bibitem{Leder01}
852: F. Lederer, S. Darmanyan, and A. Kobyakov, in {\it Nonlinearity
853: and Disorder: Theory and Applications}, F. Kh. Abdullaev, O. Bang
854: and M. P. Soerensen (Eds.). Kluwer AP, pp.131-157 (2001).
855:
856: \bibitem{Tromb01}
857: A. Trombettoni and A. Smerzi, Phys.Rev.Lett. {\bf 86}, 2353
858: (2001).
859:
860: \bibitem{Abd01a}
861: F.Kh. Abdullaev, B.B. Baizakov, S.A. Darmanyan, V.V. Konotop, and
862: M. Salerno, Phys.Rev. A {\bf 64}, 043606 (2001).
863:
864: \bibitem{Kon04}
865: V.A. Brazhnyi and V.V. Konotop, Mod.Phys.Lett. B {\bf 18}, 627
866: (2004).
867:
868:
869: \bibitem{Alfimov}
870: G.Alfimov, P.G. Kevrekidis, V.V. Konotop, and M. Salerno,
871: Phys.Rev. E {\bf 66}, 046608 (2002).
872:
873: \bibitem{Kevr04}
874: P.G. Kevrekidis, D.J. Frantzeskakis, Mod.Phys.Lett. B {\bf 18},
875: 173 (2004).
876:
877: \bibitem{AK}
878: F.Kh. Abdullaev and V.V. Konotop, Phys.Rev. A {\bf 68}, (2003).
879:
880: \bibitem{Morsh}
881: Morsh and M. Oberthaler, Rev.Mod.Phys. {\bf 78}, 179 (2006).
882:
883: \bibitem{Ober04}
884: B.Eiermann, Th. Anker, M. Ablietz, M. Taglieber, P. Treutlein,
885: K.-P. Marzlin, and M.K. Oberthaler, Phys.Rev.Lett. {\bf 92},
886: 320401 (2004).
887:
888: \bibitem{Pu}
889: W. Zheng, E.M. Wright, H. Pu, and P. Meystre, Phys.Rev. A {\bf
890: 68}, 023605 (2003).
891:
892: \bibitem{Abd01b}
893: F.Kh. Abdullaev, A. Gammal, L. Tomio and T. Frederico, Phys.Rev. A
894: {\bf 63}, 043604 (2001).
895:
896: \bibitem{Akhmediev}
897: N. Akhmediev, M.P. Das, and A.V. Vagov, Int.J.Mod.Phys. B {\bf
898: 13}, 625 (1999).
899:
900: \bibitem{Kraemer}
901: T. Kraemer et al.,
902: %''Efimov resonance in an ultracold gas of
903: %cesium atoms",
904: arXive:cond-mat/0512394.
905:
906: \bibitem{AS05}
907: F.Kh. Abdullaev and M. Salerno, Phys.Rev. A {\bf 72}, 033617
908: (2005).
909:
910: \bibitem{Malomed06}
911: R. Carretero-Gonzales, J.D. Talley, C. Chong, and B.A. Malomed,
912: arXive:nlin/0512052.
913:
914: \bibitem{Cai}
915: D. Cai, A.R. Bishop, and N. Gronblech-Jensen, Phys.Rev.Lett. {\bf
916: 70}, 591 (1993).
917:
918: \bibitem{ADG}
919: F.Kh. Abdullaev, S.A. Darmanyan, and J. Garnier, Progress in
920: Optics, E.Wolf(Ed.), {\bf 42}, 301 (2002).
921:
922: \bibitem{Page}
923: J.B. Page, Phys.Rev. B {\bf 41}, 7835 (1990).
924:
925: \bibitem{Darm98}
926: S. Darmanyan, A. Kobyakov, and F. Lederer, JETP {\bf 86}, 682
927: (1998).
928:
929: %\bibitem{Akhmediev1}
930: %N.N. Akhmediev and A. Ankiewicz, {\it Solitons. Nonlinear pulses
931: %and beams}, Chapman $\&$ Hall, London (1997).
932:
933: \bibitem{TS2}
934: A. Smerzi and A. Trombettoni, Phys.Rev.A {\bf 68}, 023613 (2003).
935:
936: \bibitem{KevPel}
937: P.G. Kevrekidis, S.V. Dmitriev, and A.A. Sukhorukov,
938: arXive:nlin/0603046.
939:
940:
941: \bibitem{Khare}
942: A.Khare, K.Rasmussen, M. Salerno, M. Samuelsen, and A. Saxena,
943: arXive:clin/0603034.
944:
945: \end{thebibliography}
946:
947:
948: \end{document}
949: