nlin0604062/dw7.tex
1: \documentclass[aps,twocolumn,pre]{revtex4}
2: \usepackage{graphicx}
3: \usepackage{amsfonts}
4: \usepackage{amssymb}
5: \usepackage{amsmath}
6: 
7: \setcounter{MaxMatrixCols}{10}
8: 
9: \def\wfig{4.25cm}
10: \def\hfig{3.0cm}
11: \begin{document}
12: 
13: \title{\bf Symmetry Breaking in Symmetric and Asymmetric Double-Well Potentials}
14: 
15: \author{ 
16: G.\ Theocharis$^{1}$, 
17: P.G.\ Kevrekidis$^{2}$,  
18: D.J.\ Frantzeskakis$^{1}$, and 
19: P.\ Schmelcher$^{3,4}$ 
20: }
21: \affiliation{
22: $^{1}$ Department of Physics, University of Athens, Panepistimiopolis,Zografos, Athens 157 84, Greece \\
23: $^{2}$ Department of Mathematics and Statistics,University of Massachusetts, Amherst MA 01003-4515, USA \\
24: $^{3}$ Theoretische Chemie, Physikalisch-Chemisches Institut, INF 229, Universit\"at Heidelberg, 69120 Heidelberg, Germany \\
25: $^{4}$ Physikalisches Institut, Philosophenweg 12, Universit\"at Heidelberg, 69120 Heidelberg, Germany 
26: }
27: 
28: \begin{abstract}
29: 
30: Motivated by recent experimental studies of matter-waves and optical beams 
31: in double well potentials, we study the solutions of 
32: the nonlinear Schr\"{o}dinger equation in such a context.  
33: %examine analytically and numerically 
34: %the nonlinear waves that emerge in such potentials. 
35: Using a Galerkin-type approach, 
36: we obtain a detailed handle on the nonlinear solution branches of the problem, 
37: starting from the corresponding linear ones and predict the relevant bifurcations
38: of solutions for both attractive and repulsive nonlinearities. The results 
39: %These analytical predictions are found to be in excellent agreement with the full numerical results and serve to
40: illustrate the nontrivial differences that arise between the 
41: steady states/bifurcations emerging in symmetric and asymmetric double wells.
42: 
43: \end{abstract}
44: 
45: \maketitle 
46: 
47: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
48: 
49: %\section{Introduction}
50: 
51: {\it Introduction}. 
52: %
53: It is well known that the nonlinear Schr\"{o}dinger (NLS) equation is a 
54: fundamental 
55: model describing the evolution of a complex field envelope in nonlinear dispersive media \cite{NLS}. 
56: As such, it plays a key role in many different contexts, ranging from nonlinear and atom optics to 
57: plasma physics, fluid dynamics, and even biophysical models \cite{trubatch}.
58: The interest in the NLS equation has dramatically increased during the last 
59: few years, as it also describes the mean-field dynamics of Bose-Einstein condensates (BECs) \cite{gbec}. 
60: In this context, the NLS is also known as the 
61: Gross-Pitaevskii (GP) equation, and
62: typically incorporates external potentials that are used 
63: for the BEC confinement. Such potentials may be, e.g., harmonic (usually implemented 
64: by external magnetic fields) or periodic (implemented by the 
65: interference of laser beams), 
66: so-called optical lattices \cite{reviewsbec}. 
67: Importantly, NLS models with 
68: %such harmonic or periodic 
69: similar external potentials appear also in the 
70: context of optics, where they respectively describe the evolution of an optical beam in a 
71: graded-index waveguide or in periodic waveguide arrays \cite{kivshar,reviewsopt}. 
72: 
73: Another type of external potential, which has mainly been studied 
74: theoretically in the BEC context \cite{smerzi,kiv2,mahmud,bam,Bergeman_2mode,infeld} 
75: is the double well potential. 
76: Moreover, it has been demonstrated experimentally that a BEC 
77: either tunnels and performs Josephson oscillations between the wells, or is  
78: subject to macroscopic quantum self-trapping \cite{markus1}.
79: On the other hand, in the context of optics, a double well potential can be created 
80: by a two-hump self-guided laser beam in Kerr media \cite{HaeltermannPRL02}.
81: A different alternative was offered in \cite{zhigang}, wherein the first stages 
82: of the evolution of an optical beam, initially focused between the wells of a photorefractive 
83: crystal, were monitored. 
84:  
85: One of the particularly interesting features of either matter-waves or optical beams in 
86: double well potentials is the {\it spontaneous symmetry breaking}, i.e., the localization
87: of the respective order parameter in one of the wells of the potential. 
88: Symmetry breaking solutions of the NLS model have first been predicted 
89: in the context of molecular states \cite{ms} and, apart from the physical contexts 
90: of BECs \cite{smerzi,kiv2,mahmud,bam,Bergeman_2mode,infeld} (see also \cite{jackson}) 
91: and optics \cite{HaeltermannPRL02,zhigang} mentioned above, they 
92: have also been studied from a mathematical point of view in Refs. 
93: \cite{mathsy,mathsy1}. 
94: 
95: These works underscore the relevance and timeliness of a better 
96: understanding of the dynamics of nonlinear waves in double well potentials.  
97: In view of that, in the present work we offer a systematic methodology, 
98: based on a two-mode expansion, of how to tackle problems in double wells, 
99: as regards their stationary states and the bifurcations (and ensuing instabilities) 
100: that arise in them. This way, considering both cases of attractive and repulsive nonlinearities, 
101: we illustrate the ways in which a symmetric double well potential is different 
102: from an asymmetric one. In particular, we demonstrate that, 
103: contrary to the case of symmetric potentials where symmetry 
104: breaking follows a pitchfork bifurcation, 
105: in asymmetric double wells the bifurcation is of the saddle-node type.
106: 
107: The paper is structured as follows: in Section II, 
108: we present the model and set the analytical framework.
109: In Section III, we illustrate the value of the method by 
110: highlighting the significant differences of symmetric and asymmetric double wells. Finally,
111: in Section IV, we summarize our findings and discuss future directions.
112: 
113: {\it Model and Analytical Approach}.
114: %
115: In a quasi-1D setting, the evolution of the mean-field wavefunction of a BEC 
116: \cite{reviewsbec} (or the envelope of an optical beam \cite{kivshar})  
117: is described by the following normalized NLS (GP) equation, 
118: %
119: \begin{eqnarray}
120: i u_t=-\frac{1}{2} u_{xx} + s |u|^2 u + V(x) u - \mu u.
121: \label{beq1}
122: \end{eqnarray}
123: %
124: In the BEC (optics) context, $\mu$ denotes the chemical potential (propagation constant) and 
125: $s=\pm 1$ for attractive or repulsive interatomic interactions (focusing or defocusing Kerr nonlinearity) 
126: respectively; below, for simplicity, we will adopt the terms attractive and repulsive nonlinearity 
127: for $s=\pm 1$ respectively. Finally, in Eq. (\ref{beq1}), $V(x)$ is the double well potential, 
128: which is assumed to be composed by a parabolic trap 
129: (of strength $\Omega$) and a ${\rm sech}^2$-shaped barrier (of strength $V_0$, width $w$ and location $x_0$); 
130: in particular, $V(x)$ is of the form:
131: %
132: \begin{eqnarray}
133: V(x)=\frac{1}{2} \Omega^2 x^2 + V_0 {\rm sech}^2 \left(\frac{x-x_{0}}{w}\right),  
134: \label{beq2}
135: \end{eqnarray}
136: %
137: with the choice $x_{0}=0$ ($x_{0} \neq 0$) corresponding to a symmetric (asymmetric) double well. 
138: Note that such a double well can be implemented in BEC experiments 
139: upon, e.g., combining a magnetic trap with a sharply focused, blue-detuned laser beam \cite{expd}.  
140: Similar double wells can also be implemented e.g., in optical systems.
141: 
142: The spectrum of the underlying linear Schr{\"o}dinger equation ($s=0$) 
143: consists 
144: of a ground state, $u_{0}(x)$, and excited states, $u_{l}(x)$ ($l \ge 1$). In the 
145: nonlinear problem, using a Galerkin-type approach, we expand $u(x,t)$ as,
146: %
147: \begin{eqnarray}
148: u(x,t)=c_0(t) u_0(x) + c_1(t) u_1(x)+\cdots,
149: \label{beq3}
150: \end{eqnarray}
151: %
152: and truncate the expansion, keeping solely the first two modes; 
153: here $c_{0,1}(t)$ are unknown time-dependent complex prefactors. It is worth noticing that 
154: such an approximation (involving the truncation of higher order modes and 
155: the spatio-temporal factorization of the wavefunction), 
156: is expected to be quite useful for a weakly nonlinear analysis. In fact, as will be seen below, 
157: we will be able to identify the nonlinear states that stem from the linear ones, as well as their 
158: bifurcations.
159: 
160: Substituting Eq. (\ref{beq3}) into Eq. (\ref{beq1}), 
161: and projecting the result to the corresponding eigenmodes, we 
162: obtain the following ordinary differential equations (ODEs):
163: %
164: \begin{eqnarray}
165: i \dot{c}_0 &=& (\omega_0-\mu) c_0 - s A_0 |c_0|^2 c_0 - s B (2 |c_1|^2 c_0
166: + c_1^2 \bar{c}_0), 
167: \nonumber 
168: \\
169: &-& s \Gamma_1 |c_1|^2 c_1  - s \Gamma_0  (2 |c_0|^2 c_1 + c_0^2 \bar{c}_1)
170: \label{beq4}
171: \\
172: i \dot{c}_1 &=& (\omega_1-\mu) c_1 - s A_1 |c_1|^2 c_1 - s B (2 |c_0|^2 c_1
173: + c_0^2 \bar{c}_1), 
174: \nonumber 
175: \\
176: &-& s \Gamma_0 |c_0|^2 c_0  - s \Gamma_1  (2 |c_1|^2 c_0 + c_1^2 \bar{c}_0).
177: \label{beq5}
178: \end{eqnarray}
179: %
180: In Eqs. (\ref{beq4})-(\ref{beq5}), dots denote time derivatives, 
181: overbars denote complex conjugates, $\omega_{0,1}$ are the eigenvalues 
182: corresponding to the eigenstates $u_{0,1}$, while $A_0=\int u_0^4 dx$, $A_1=\int u_1^4 dx$, $B=\int u_0^2 u_1^2 dx$,
183: $\Gamma_0=\int u_0 u_1^3 dx$ and $\Gamma_1=\int u_1 u_0^3 dx$ are constants.
184: Recall that $u_{0}$ and $u_1$ are real (due to the Hermitian nature of
185: the underlying linear Schr{\"o}dinger problem) and are also orthonormal.
186: Notice also that in the symmetric case ($x_{0}=0$), due to the parity of the eigenfunctions, $\Gamma_0=\Gamma_1=0$.
187: 
188: We now use amplitude-phase (action-angle) variables, $c_j=\rho_j e^{i \phi_j}$, $j=0,1$
189: ($\rho_j$ and $\phi_j$ are assumed to be real), to 
190: derive from the ODEs (\ref{beq4})-(\ref{beq5}) a set of four equations. 
191: Introducing the function $\varphi \equiv \phi_1-\phi_0$, we find that the equations for 
192: $\rho_0$ and $\phi_0$ are,
193: %
194: \begin{eqnarray}
195: \dot{\rho}_0 &=& s [(\Gamma_0 \rho_0^2
196: + \Gamma_1 \rho_1^2) \sin(\varphi) - \rho_1^2 \rho_0 \sin(2 \varphi)], 
197: \label{beq6}
198: \\
199: \dot{\phi}_0 &=& (\omega_0-\mu) + s A_0 \rho_0^2 + 2 s B \rho_1^2
200: + s B \rho_1^2 \cos(2 \varphi) 
201: \nonumber
202: \\
203: &+& s \left(\frac{\Gamma_1 \rho_1^3}{\rho_0}
204: + 3 \rho_0 \rho_1 \Gamma_0 \right) \cos(\varphi), 
205: \label{beq7}
206: \end{eqnarray}
207: %
208: while the equations for $\rho_1, \phi_1$ are found by
209: interchanging indices $1$ and $0$ in the above equations. 
210: Next, taking into regard the conservation of the total norm, we obtain the equation 
211: $\rho_0^2+ \rho_1^2=N$, where $N=\int |u|^2 dx$ is the integral of motion of Eq. (\ref{beq1})
212: (the number of particles in BECs, or the power in optics). Finally, subtracting
213: Eq. (\ref{beq7}) for $\dot{\phi}_0$, and the corresponding one for $\dot{\phi}_1$, we obtain:
214: %
215: \begin{eqnarray}
216: \dot{\varphi} &=& -\Delta \omega + s (A_0 \rho_0^2 -  A_1 \rho_1^2) 
217: \nonumber 
218: \\
219: &-& s B (2 + \cos(2 \varphi)) (\rho_0^2-\rho_1^2)-s \frac{\cos(\varphi)}{\rho_0 \rho_1}
220: \nonumber
221: \\
222: &\times&  \left[\Gamma_0 \rho_0^2
223: (\rho_0^2-3 \rho_1^2) + \Gamma_1 \rho_1^2 (3 \rho_0^2 - \rho_1^2)\right].
224: \label{beq9}
225: \end{eqnarray}
226: %
227: Equations (\ref{beq6}), (\ref{beq9}) is a dynamical system, which, in principle, 
228: can be thoroughly 
229: investigated using phase-space analysis (such an approach has 
230: been presented in \cite{smerzi,kiv2} for similar systems that were derived using 
231: different expansion of the field $u$). Here, we will focus on the 
232: fixed points of 
233: the system [corresponding to the nonlinear 
234: eigenstates of Eq. (\ref{beq1})], and analyze 
235: their stability and bifurcations. 
236: 
237: {\it Results.} 
238: %and Comparison with Numerics}. 
239: Below we will analyze all possible cases ($s=\pm 1$, $x_0 = 0$, $x_0 \ne 0$)
240: %(attractive and repulsive nonlinearity, for symmetric and asymmetric potentials) 
241: for the double well of Eq. (\ref{beq2}) with $V_0=1$, 
242: $\Omega=0.1$, and $w=0.5$ (the results do not change qualitatively using different values). 
243: 
244: First we consider the case of attractive nonlinearity,  
245: i.e., $s=-1$, and a symmetric double well potential with $x_{0}=0$ 
246: (implying that $\Gamma_0=\Gamma_1=0$). In this case, 
247: the parameters involved in Eqs. (\ref{beq6}) and (\ref{beq9}) are found to be 
248: $A_0 = 0.09078$, $A_1=0.09502$, $B=0.08964$, $\omega_0=0.13282$ and $\omega_1=0.15571$. 
249: Then, it is readily observed that the possible real solutions of Eq. (\ref{beq6}) 
250: are $\rho_0=0$ and $\rho_1=0$, as well as $\varphi=0 \hspace{1.5mm} ({\rm mod} \hspace{1.5mm} \pi)$. 
251: The former two are continuations of the linear solutions in the nonlinear regime. 
252: However, the latter one is a non-trivial combination of
253: the two modes for $\varphi=\pi$ that results in an {\it asymmetric} pair of mirror-symmetric 
254: solutions \cite{jackson,zhigang}, emerging through a {\it pitchfork} bifurcation. 
255: From Eq. (\ref{beq9}), we obtain that this new branch of solutions
256: bifurcates from the symmetric branch $(\rho_0,\rho_1)=(\sqrt{N},0)$ for 
257: %
258: \begin{eqnarray}
259: N>N_c=\frac{\Delta \omega}{3 B-A_0},
260: \label{beq10}
261: \end{eqnarray}
262: %
263: and for $\mu<\mu_c=\omega_0-A_0 N=0.12115$.
264: These analytical predictions are in {\it excellent} agreement with 
265: the numerical results $\mu_c = 0.122 (\pm 0.001)$. It is also easy to see that the
266: anti-symmetric branch $(\rho_0,\rho_1)=(0,\sqrt{N})$ does not give
267: rise to such a bifurcation. The different branches of the full
268: numerical solutions (including the bifurcating ones) have been 
269: obtained through numerical fixed point algorithms solving the steady
270: state version of Eq. (\ref{beq1}), and using continuation of the
271: solutions over the parameter $\mu$. The results are shown in the top left panel of Fig.1, where 
272: the norm of the solutions $N=\int |u|^2 dx$ is shown as a function of the chemical potential 
273: (or propagation constant in optics) $\mu$. In addition, as expected from the nature of the bifurcation, 
274: the linear stability analysis has been used to illustrate the following: The emerging new asymmetric 
275: (i.e., ``symmetry breaking'') branch of solutions is stable, while the original symmetric branch 
276: is unstable beyond the bifurcation point due to a real eigenvalue $\lambda_r$ (see bottom left panel of Fig. 1). 
277: 
278: \begin{figure}[t]
279: %[h]
280: \includegraphics[width=4.2cm,height=5.2cm]{gt1}
281: \includegraphics[width=4.2cm,height=5.2cm]{gt2}
282: %\includegraphics[width=7cm]{fig1.eps}
283: \caption{(Color online) The top panels show the norm of the solutions of Eq.(\ref{beq1}) 
284: for attractive nonlinearity ($s=-1$) as a function of $\mu$ for symmetric 
285: (left panel, $x_{0}=0$) and asymmetric double well (right panel, $x_{0}=0.5$). 
286: The potential parameters are $\Omega=0.1$, $V_0=1$ and $w=0.5$. 
287: The 
288: %blue 
289: solid lines denote the symmetric solution, the 
290: %red 
291: dashed-dotted lines denote the antisymmetric one, while the 
292: %green 
293: dashed lines 
294: %(and cyan line in right panel) 
295: denote the asymmetric solutions that are generated from the bifurcation 
296: at $\mu_c \approx 0.122$ (pitchfork) and $\mu_c \approx 0.009$ (saddle-node) respectively. 
297: %As a result of that bifurcations the symmetric solutions destabilize as shown in the second row 
298: The bottom panels show the maximal real eigenvalue associated with the linear stability of the symmetric branches.
299: %around them. Moreover, due to the asymmetry of the double well, the two asymmetric branches 
300: %(dashed green and dashed cyan lines) are no longer equivalent and the 
301: %type of the bifurcation change from pitchfork to a saddle-node.
302: }
303: \label{fig1}
304: \end{figure}
305: 
306: Next, in the same case ($s=-1$), we consider 
307: a double well potenial with a {\it weak asymmetry} ($x_{0}=0.5$). 
308: In this case, the constants involved in Eqs. (\ref{beq6})-(\ref{beq9}) are found to be 
309: $A_0 = 0.14903$, $A_1=0.15618$, $B=0.02958$, $\omega_0=0.1249$ and $\omega_1=0.16535$, while 
310: $\Gamma_0=0.0407$ and $\Gamma_1=-0.04077$. Note that even such a weakly 
311: asymmetric case,  
312: renders the right well ``shallower'', in the following sense: the density, or power $N$ 
313: (regarding the ground state of the linear problem), in the right well is smaller than the one in the left well. 
314: Thus, in the nonlinear problem, the respective branches that bear the larger part 
315: of the density in the right or in the left well (i.e., the ones having, roughly speaking, 
316: the shape of a single pulse in each of the wells) are no longer equivalent. 
317: This results in a {\it significant} difference between the asymmetric and the symmetric case discussed above, 
318: namely there is no longer a pitchfork bifurcation, but instead, there is a {\it saddle-node} bifurcation. 
319: This result is shown in the top right panel of Fig. 1, where $N$ is shown as a function of $\mu$. 
320: It is readily observed that, due to the nature of the saddle-node bifurcation, two branches 
321: (one of which is stable and the other one is unstable) 
322: ``collide'' at some critical value of $\mu=\mu_c$ (see below) and disappear. These branches are 
323: the more ``symmetric'' one, that has support in both wells (see solid line in top left panel of Fig. 2), and the one 
324: pertaining to the state having the form of a single pulse in the shallower well (see dashed line in the 
325: rightmost top panel of Fig. 2). The instability of the former branch is depicted in the bottom right panel of Fig. 1, 
326: where the maximal real eigenvalue is shown as a function of $\mu$. 
327: On the other hand, there exists also another single-pulse branch 
328: supported over the deeper well (see top third panel of Fig. 2), which persists all the way to the linear limit. 
329: %This branch describes the state in which the pulse is in the 
330: %deeper well (see the bottom dashed line in top right panel of Fig. 1, as well as the dashed line 
331: %in the third top panel of Fig. 2), and persists all the way to the linear limit. 
332: Furthermore, the dash-dotted anti-symmetric branch of the top right of Fig. 1 is shown
333: in the second top panel of Fig. 2. 
334: 
335: The novel feature described above, namely the asymmetric breakdown of the pitchfork bifurcation into 
336: a saddle-node one, 
337: %and an ``isolated'' branch
338: is a {\it particular feature} of asymmetric double well potentials 
339: that, to the best of our knowledge, has not been previously appreciated.
340: %in this context. 
341: Notice that Eq. (\ref{beq9}) 
342: predicts that the saddle-node bifurcation occurs at $\mu_c=0.08748$, 
343: while the numerical result is $\mu_c=0.09 \pm 0.001$; apparently the two results are again in excellent agreement. 
344: 
345: \begin{figure}[t]
346: %[tbp]
347: \includegraphics[width=4.22cm,height=5.2cm]{dw_fig2b}
348: \includegraphics[width=4.22cm,height=5.1cm]{dw_fig2a}
349: \caption{(Color online) The steady state solutions of Eq.(\ref{beq1}), (see also Fig. \ref{fig1}) 
350: for the focusing, asymmetric case (top panels) and their linear stability (bottom panels) for $\mu=0.05$. 
351: The black-dashed line shows the double well potential.}
352: \label{fig2b}
353: \end{figure}
354: 
355: \begin{figure}[t]
356: %[h]
357: \includegraphics[width=4.25cm,height=5.2cm]{gt3.eps}
358: \includegraphics[width=4.25cm,height=5.2cm]{gt4.eps}
359: %\includegraphics[width=7cm]{fig3.eps}
360: \caption{(Color online) 
361: Same as in Fig.\ref{fig1} but for the repulsive nonlinearity ($s=+1$). 
362: The 
363: %red 
364: dashed-dotted lines denote the symmetric solution, the 
365: %blue 
366: solid the antisymmetric, while the 
367: %green 
368: dashed lines
369: %(and cyan line in right panel) 
370: denote the asymmetric solutions generated from the bifurcation at 
371: $\mu_c \approx 0.168$ (pitchfork) and $\mu_c \approx 0.207$ (saddle-node) 
372: respectively. Contrary to the case $s=-1$, the bifurcations originate from the anti-symmetric branch.}
373: \label{fig3}
374: \end{figure}
375: 
376: Let us now consider the repulsive nonlinearity ($s=+1$). 
377: In this case, for the symmetric potential ($x_{0}=0$), the pitchfork bifurcation still occurs; 
378: however, now it does not originate from the symmetric branch, but rather from the anti-symmetric one with 
379: $(\rho_0,\rho_1)=(0,\sqrt{N})$, giving again rise to symmetry breaking. 
380: %, i.e., to an asymmetric solution. 
381: Analyzing Eq. (\ref{beq9}), we find that this occurs when
382: %
383: \begin{eqnarray}
384: N> N_c \geq \frac{\Delta \omega}{3 B- A_1}, 
385: \label{beq11}
386: \end{eqnarray}
387: %
388: and for $\mu=\omega_0 + 3 B N=0.16822$, once again in excellent agreement
389: with the numerical result $\mu_c= 0.168 (\pm 0.001)$.
390: 
391: On the other hand, in the same case ($s=+1$) but for an asymmetric ($x_{0}=0.5$) 
392: double well, the bifurcation still originates from anti-phase (between the wells)
393: solutions, but again (as in the asymmetric case with $s=-1$), the bifurcation 
394: is of the saddle-node type. 
395: This  is theoretically predicted to occur at $\mu_c=0.21342$, 
396: once again in very close
397: agreement to the numerical result $\mu_c=0.207 \pm 0.001$. 
398: The details of the bifurcation diagrams, are illustrated in Fig. \ref{fig3} (analogously to Fig. 1), 
399: while the steady state solutions and their linear stability are shown in Fig. \ref{fig4b} (analogously to Fig. 2).
400: 
401: %Finally, in Fig. \ref{fig5}, one can observe the evolution of the unstable solutions for the asymmetric case. 
402: 
403: \begin{figure}[t]
404: %[tbp]
405: \includegraphics[width=4.25cm,height=5.2cm]{dw_fig4a.eps}
406: \includegraphics[width=4.25cm,height=5.2cm]{dw_fig4b.eps}
407: %\includegraphics[width=7cm]{unstabledwfig4.eps}
408: \caption{(Color online) The steady state solutions of Eq.(\ref{beq1}), 
409: (see also Fig. \ref{fig3}) for $s=+1$ in the asymmetric case (top panels) and their linear 
410: stability (bottom panels) for $\mu=0.22$. The black-dashed line shows the double well potential.}
411: \label{fig4b}
412: \end{figure}
413: 
414: {\it Conclusions.} In conclusion, we have presented a systematic
415: analysis based on a Galerkin, two-mode truncation of the 
416: stationary states of symmetric and asymmetric double well potentials.
417: The analysis has been carried out both for repulsive and attractive nonlinearities and, 
418: as such, can be relevant to a variety of physical contexts; these include matter-wave physics 
419: (most directly), nonlinear optics, as well as other contexts where 
420: it is relevant to consider double well potentials in the NLS model proper.
421: We have demonstrated that our analytical approach describes quite accurately, both 
422: {\it qualitatively} and {\it quantitatively} the features of the nonlinear solutions; 
423: numerical results were shown to be in excellent agreement with the analytical 
424: predictions. 
425: 
426: In the case of a symmetric double well potential, it has been shown 
427: that a symmetry-breaking (pitchfork) bifurcation of the ground state occurs 
428: for attractive nonlinearities, while it is absent for repulsive nonlinearities.
429: It has also been found that a similar bifurcation of the first excited state occurs 
430: in the relevant branches for repulsive nonlinearities, oppositely to the case of attractive ones,
431: where such bifurcation does not happen. 
432: Additionally, regarding the above feature, we have illustrated that 
433: symmetric potentials are very particular (degenerate) due to their 
434: special characteristic of mirror-equivalence of the emerging symmetry-breaking states.
435: We have shown that even weak asymmetries lift 
436: this degeneracy and lead to saddle-node bifurcations 
437: instead of pitchfork ones that were similarly quantified 
438: in both attractive and repulsive nonlinearity contexts. 
439: 
440: These results underscore the relevance of analyzing steady state features 
441: of nonlinear models (in the presence of external potentials) based 
442: on the states of the underlying linear equations. It would be 
443: particularly interesting to examine the extent to which dynamical 
444: features of such models can be captured by similar truncations. 
445: Such studies are currently in progress. 
446: %and will be reported in 
447: %future publications.
448: 
449: {\it Acknowledgements.} Constructive discussions with M.K. Oberthaler are kindly acknowledged. 
450: This work has been partially supported from ``A.S. Onasis'' Public Benefit Foundation (G.T.) and NSF (P.G.K.).
451: 
452: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
453: 
454: \begin{thebibliography}{99}
455: 
456: \bibitem{NLS} C. Sulem and P.L. Sulem, \newblock {\it The Nonlinear Schr{\"o}dinger Equation},
457: (Springer-Verlag, New York, 1999). 
458: 
459: \bibitem{trubatch} M.J. Ablowitz, B. Prinari and A.D. Trubatch,
460: {\it Discrete and Continuous Nonlinear Schr{\"o}dinger systems} 
461: (Cambridge University Press, Cambridge, 2003).
462: 
463: \bibitem{gbec} F. Dalfovo, S. Giorgini, L.P. Pitaevskii, and S. Stringari, 
464: %{\it et al.,} 
465: Rev. Mod. Phys. {\bf 71}, 463 (1999).
466: 
467: \bibitem{reviewsbec} P.G. Kevrekidis and D.J. Frantzeskakis,
468: Mod. Phys. Lett. B {\bf 18}, 173 (2004); 
469: V.V. Konotop and V.A. Brazhnyi, Mod. Phys. Lett. B {\bf 18} 627, (2004).
470: 
471: \bibitem{kivshar} Yu.S. Kivshar and G.P. Agrawal, 
472: \textit{Optical Solitons: From Fibers to Photonic Crystals}, 
473: Academic Press (San Diego, 2003).
474: 
475: \bibitem{reviewsopt} D. N. Christodoulides, F. Lederer and Y. Silberberg, 
476: Nature \textbf{424}, 817 (2003); J.W. Fleischer {\it et al.}, Opt. Expr. {\bf 13}, 1780 (2005).
477: 
478: \bibitem{smerzi} S. Raghavan {\it et al.}, Phys. Rev. A {\bf 59}, 620 (1999); 
479: S. Raghavan {\it et al.}, Phys. Rev. A {\bf 60}, R1787 (1999); 
480: A. Smerzi and S. Raghavan, Phys. Rev. A {\bf 61}, 063601 (2000).
481: 
482: \bibitem{kiv2} E.A. Ostrovskaya {\it et al.}, Phys. Rev. A \textbf{61}, 031601 (R) (2000).
483: 
484: \bibitem{mahmud} K.W. Mahmud, J. N. Kutz and W. P. Reinhardt, Phys. Rev. A \textbf{66}, 063607 (2002).
485: 
486: \bibitem{bam} V.S. Shchesnovich, B.A. Malomed, and R.A. Kraenkel, Physica D {\bf 188}, 213 (2004).
487: 
488: \bibitem{Bergeman_2mode} D. Ananikian and T. Bergeman, Phys. Rev. A 
489: \textbf{73}, 013604 (2006).
490: 
491: \bibitem{infeld} P. Zi\'{n} {\it et al.}, Phys. Rev. A
492: \textbf{73}, 022105 (2006).
493: 
494: \bibitem{markus1} M. Albiez {\it et al.}, Phys. Rev. Lett. {\bf 95},
495: 010402 (2005).
496: 
497: \bibitem{HaeltermannPRL02} C. Cambournac \textit{et al.}, Phys. Rev. Lett. 
498: \textbf{89}, 083901 (2002).
499: 
500: \bibitem{zhigang} P.G. Kevrekidis {\it et al.},
501: Phys. Lett. A {\bf 340}, 275 (2005).
502: 
503: \bibitem{ms} E.B. Davies, Commun. Math Phys. {\bf 64}, 191 (1979)
504: 
505: \bibitem{jackson}  R. K. Jackson and M. I. Weinstein, J. Stat. Phys. \textbf{116}, 881 (2004).
506: 
507: \bibitem{mathsy}  W. H. Aschenbacher {\it et al.}, 
508: %J. Fr\"{o}hlich, G. M. Graf, K. Schnee and M. Troyer, 
509: J. Math. Phys. \textbf{43}, 3879 (2002).
510: 
511: \bibitem{mathsy1} T. Kapitula and P.G. Kevrekidis, Nonlinearity {\bf 18}, 2491 (2005).
512: 
513: \bibitem{expd} 
514: C.\ Raman {\it et al.}, 
515: %M.\ K{\"{o}}hl, D. S.\ Durfee, C. E.\ Kuklewicz,
516: %Z.\ Hadzibabic and W.\ Ketterle, 
517: Phys.\ Rev.\ Lett.\ \textbf{83}, 2502
518: (1999); R.\ Onofrio {\it et al.}, 
519: %C.\ Raman, J.M.\ Vogels, J. R.\ Abo-Shaeer, A.P.\
520: %Chikkatur and W.\ Ketterle, 
521: Phys.\ Rev.\ Lett.\ \textbf{85}, 2228 (2000).
522: 
523: 
524: \end{thebibliography}
525: 
526: \end{document}
527: 
528: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
529: 
530: 
531: EXPANSIONS
532: %%%%%%%%%%
533: 
534: In the case $x_{0}=0$, the spectrum of the underlying linear Schr{\"o}dinger equation ($s=0$) 
535: typically consists of a symmetric ground state, $u_{0}(x)$, and 
536: an anti-symmetric first excited state $u_{1}(x)$ . In the case 
537: $x_{0} \neq 0$, 
538: 
539: (for the symmetric 
540: double well potential). Our approach, similarly to the spirit of
541: \cite{smerzi,mathsy1} consists of
542: %%%%%%%%%%%%%% 
543: In Eq. (\ref{beq3}), $u_{0,1}(x)$ are respectively the ground
544: and first excited state of the underlying linear problem and
545: $c_{0,1}(t)$ are their time-dependent prefactors.
546: Notice, however, that contrary to the spirit of \cite{smerzi}, we will
547: {\it not} expand to the basis of eigenfunctions localized in 
548: each well. 
549: 
550: It should be noticed that the expansion (\ref{beq3}) in the solutions of the respective linear problem 
551: is similar to the one used in \cite{zhigang,jackson,mathsy1}. On the other hand, it is different from the one used in 
552: \cite{smerzi} (where the expansion was to the basis of eigenfunctions localized in each well)
553: and the ones in \cite{kiv2} (where ``coupled mode theory'')
554: 
555: %%%%%%%%%%%%%%%%%%%%%%%%%%%%
556: 
557: TERMINOLOGY
558: 
559: Note that for 
560: $x_{0}=0$ the states $u_{0}(x)$ and $u_{1}(x)$ are symmetric and anti-symmetric respectively; 
561: below, we will use the same terminology also for $x_{0} \neq 0$. 
562: 
563: \begin{figure}[t]
564: %[tbp]
565: \includegraphics[width=4.2cm,height=4.2cm]{unstabledwfig2.eps}
566: \includegraphics[width=4.2cm,height=4.2cm]{unstabledwfig4.eps}
567: \caption{(Color online) Spatio-temporal countour plot of the density of the unstable 
568: solutions for the attractive (left panel)
569: and repulsive case (right panel) both for asymmetric potentials.}
570: \label{fig5}
571: \end{figure}
572: 
573: %\bibitem{markusref1} B.D. Josephson, Phys. Lett. {\bf 1}, 251 (1962).
574: