1:
2: \documentclass[aps,pre,superscriptaddress]{revtex4}
3: %twocolumn,superscriptaddress]{revtex4}
4: \usepackage[dvips]{graphics}
5: \usepackage{graphicx}
6: \usepackage{amsfonts}
7: \usepackage{amssymb}
8: \usepackage{amsmath}
9: \usepackage{subfigure}
10:
11: \setcounter{MaxMatrixCols}{10}
12:
13: \begin{document}
14:
15: \title{ Solitary Waves in Discrete Media with Four Wave Mixing }
16: \author{R.L. Horne}
17: \affiliation{Department of Mathematics, Florida State University, Tallahassee, FL 32306-4510}
18: \email{horne@math.fsu.edu}
19:
20: \author{P.G.\ Kevrekidis }
21: \affiliation{Department of Mathematics and Statistics, University of Massachusetts,
22: Amherst MA 01003-4515, USA}
23: \email{kevrekid,whitaker@math.umass.edu}
24: \author{N. Whitaker}
25: \affiliation{Department of Mathematics and Statistics, University of Massachusetts,
26: Amherst MA 01003-4515, USA}
27: %\email{whitaker@math.umass.edu}
28:
29: \begin{abstract}
30: In this paper, we examine in detail the principal branches of solutions
31: that arise in vector discrete models with nonlinear inter-component
32: coupling and four
33: wave mixing. The relevant four branches of solutions consist of two
34: single mode branches (transverse electric and transverse magnetic)
35: and two mixed mode branches, involving both components (linearly
36: polarized and elliptically polarized). These solutions are obtained
37: explicitly and their stability is analyzed completely in the anti-continuum
38: limit (where the nodes of the lattice are uncoupled), illustrating the
39: supercritical pitchfork nature of the bifurcations that give rise to
40: the latter two, respectively, from the former two. Then the branches
41: are continued for finite coupling constructing a full two-parameter
42: numerical bifurcation diagram of their existence. Relevant stability
43: ranges and instability regimes are highlighted and, whenever unstable,
44: the solutions are dynamically evolved through direct computations to
45: monitor the development of the corresponding instabilities. Direct
46: connections to the earlier experimental work of Meier {\it et al.}
47: [Phys. Rev. Lett. {\bf 91}, 143907 (2003)] that motivated the present
48: work are given.
49: \end{abstract}
50:
51: %\date{Submitted to {\em Physical Review E}, July 2005}
52:
53: \maketitle
54:
55: \section{Introduction}
56:
57: Recently, nonlinear Hamiltonian lattice dynamical systems with a large number of degrees of freedom
58: have become a focal point for a variety of application areas \cite{reviews}.
59: Several diverse physical contexts in which such models (typically discrete in space
60: and continuous in the evolution variable) arise are (i) spatial dynamics of optical beams in coupled waveguide
61: arrays arising in nonlinear optics \cite{reviews1}, (ii) temporal evolution of Bose-Einstein condensates (BECs)
62: in optical lattices in soft-condensed matter physics \cite{reviews2} and (iii) DNA double strand in biophysics
63: \cite{reviews3}.
64:
65: A prototypical model, applicable to different degrees of approximation,
66: in all of the above contexts is the so-called discrete nonlinear
67: Schr{\"o}dinger (DNLS) equation. This model was first proposed in
68: \cite{b3} and implemented for the first time experimentally in Refs. \cite{b7,b8}. A
69: systematic presentation of the experimental results
70: in optical systems is given in a recent
71: work \cite{JOSA}; for BEC-related experiments, see \cite{reviews2,morsch}.
72: In this model, self-localized excitations (discrete
73: solitons) are possible as a result of the interplay between the Kerr
74: nonlinearity and discrete linear coupling. Many properties of optical
75: discrete spatial solitons have been systematically explored in theory
76: and experiment, including generalizations to diffraction management \cite
77: {b9,Jena}, diffraction-managed solitons \cite{b10}, and soliton transport
78: and gating \cite{Eugenieva,dnc2,hector,bright}.
79:
80: On the other hand, a topic that has received considerably less
81: attention has been the study of vector analogs of the DNLS
82: equation. A number of studies have addressed the existence
83: and stability of diverse families of solitary waves/localized states
84: in the DNLS equation (see \cite{c7,c13,epjd}). These issues have also been
85: studied for both cubic and quadratic nonlinearities \cite{c13,c18} in
86: one and two dimensions \cite{c7,c19,hudock}.
87: However, the first experimental realization for
88: such a system occurred rather recently \cite{meier}
89: (see also the very recent paper \cite{meier2}). This work sheds new light
90: on the theoretical investigations of this topic, in that the relevant
91: model for the corresponding AlGaAs waveguide array experiment
92: included four wave mixing (FWM) terms that were rarely included
93: in previous theoretical studies. Subsequent studies (see
94: \cite{molina1,molina2}) addressed a number of specific properties of
95: the model system put forward in \cite{meier}. Such properties related to
96: switching, instability-induced amplification, modulational instability
97: as well as
98: an analysis of the energy barrier between stable discrete solitary waves centered on
99: a lattice site versus two lattice sites.
100:
101: %While these last studies have been quite useful in highlighting some
102: %of the properties of the vector system with FWM relevant to the experiments,
103: %a detailed stability analysis of the mode
104:
105: The focus of the present study is to extend the analysis of the vector NLS
106: lattice system incorporating FWM which is relevant to the experiments of \cite{meier}.
107: More specifically, we provide a systematic analysis of the
108: existence and stability
109: requirements of the principal solitary wave modes of the system.
110: In so doing, we have found that the concept of the anti-continuum limit
111: (in which the coupling between waveguides is absent) provides
112: a powerful tool for the study of the existence and stability concerning
113: these solutions. Subsequently, continuations in the coupling parameter
114: are used to construct a full two-parameter bifurcation diagram
115: that gives the complete picture of stability for both one-component as well
116: as ``mixed'' mode solutions (involving both components). We note that
117: all of the results obtained via our dimensionless model can be directly
118: placed in the context of
119: the experimental results obtained in Refs. \cite{meier,meier2}.
120:
121: The remainder of the paper is organized in the following way. In section II,
122: we present the theoretical model of interest as well as proposing a setup
123: that allows for analysis of the existence and stability of the model's solutions.
124: In section III, we fully analyze the relevant solutions and their stability
125: in the anti-continuum limit (i.e., zero coupling between adjacent waveguides).
126: In section IV, we proceed numerically to determine solutions to the model
127: for nonzero coupling and observe how the bifurcation points and stability
128: features are modified for this case. The instability
129: regions obtained in section IV are then monitored through direct
130: numerical experiments in section V. Finally, in section VI, we
131: summarize our findings and present our conclusions.
132:
133:
134: \section{Model and Setup}
135:
136: Following \cite{meier}, we will use the dimensionless model
137: \begin{eqnarray}
138: i \dot{a}_n &=&-a_n-\epsilon \left(a_{n+1}+a_{n-1}\right) -
139: \left(|a_n|^2 + A |b_n|^2\right)a_{n} - B b_n^2 a_n^{\star}
140: \label{req1}
141: \\
142: i \dot{b}_n &=& b_n-\epsilon \left(b_{n+1}+b_{n-1}\right) -
143: \left(|b_n|^2 + A |a_n|^2\right)b_{n} - B a_n^2 b_n^{\star}.
144: \label{req2}
145: \end{eqnarray}
146: In these equations, $a_n$ and $b_n$ are the appropriately
147: normalized, slowly-varying, complex field envelopes for the transverse
148: electric (TE) and transverse magnetic (TM) polarized waves respectively.
149: The constants $A$ and $B$ are respectively associated with
150: the cross-phase modulation (XPM) and four-wave mixing (FWM)
151: and were evaluated in \cite{meier} to be approximately equal
152: to $A \simeq 1$ and $B \simeq 1/2$. Wherever possible,
153: the results obtained in this analysis will be given for general $A$ and $B$,
154: even though the above values have been used in the specifics of
155: our numerical computations. The overdot in Eqs. (\ref{req1})-(\ref{req2})
156: denotes differentiation with respect to the evolution variable (which
157: is $z$ in this case) and the $\star$ denotes the complex conjugate.
158: For the properties of the waveguide array and incident light
159: used in the experiment (see \cite{meier} for details),
160: the dimensionless power
161: \begin{eqnarray}
162: P = \sum_n (|a_n|^2 + |b_n|^2)
163: \label{req3a}
164: \end{eqnarray}
165: is connected with its dimensional analog $P_d$ (measured in Watts)
166: according to $P_d \simeq 56.4 P$. Furthermore, the
167: dimensionless coupling $\epsilon$ used in \cite{meier} was
168: $\epsilon \simeq 0.921$. This is crucial since it serves to highlight
169: differences between the results obtained in the computational analysis
170: of \cite{meier} and the generalization of the corresponding results
171: presented herein. More generally, we take $\epsilon$ to be a free parameter
172: in our analysis. We note that $\epsilon$ is related to the dimensional coupling
173: constant, $\kappa$, given in \cite{meier}, via $\epsilon = 2.741\kappa$.
174: Here, $\kappa$ is measured in $\mbox{mm}^{-1}$.
175:
176:
177: We look for stationary solutions in the form: $a_n=\tilde{a}_n e^{i q z}$
178: and $b_n=\tilde{b}_n e^{i q z}$ and subsequently dropping the tildes,
179: we obtain the corresponding stationary equations:
180: \begin{eqnarray}
181: (q-1) a_n-\epsilon \left(a_{n+1}+a_{n-1}\right) -
182: \left(|a_n|^2 + A |b_n|^2\right)a_{n} - B b_n^2 a_n^{\star} &=& 0
183: \label{req3}
184: \\
185: (q+1) b_n-\epsilon \left(b_{n+1}+b_{n-1}\right) -
186: \left(|b_n|^2 + A |a_n|^2\right)b_{n} - B a_n^2 b_n^{\star} &=& 0.
187: \label{req4}
188: \end{eqnarray}
189: In this context, $q$ is an additional free parameter, namely the propagation
190: constant ($q$ is also used as a free parameter for the computations presented
191: in \cite{meier}). Our parameter $q$ is related to the corresponding dimensional
192: propagation constant, $\delta$, given in \cite{meier}, by $\delta = 0.3648q$.
193: Here, $\delta$ is measured in $\mbox{mm}^{-1}$.
194:
195:
196: It is worth pointing out that for Fig. 1 in Ref. \cite{meier}, the range
197: of parameters used in their computations appear to correspond to the case of
198: $q > 3$. This is verified by our analysis below. Again, this is highlighted
199: to account for the disparities of some of the conclusions obtained in
200: \cite{meier} with the ones their generalizations
201: presented in the following sections.
202:
203:
204: To perform linear stability computations, provided that an
205: exact solution $(a_n^0,b_n^0)$
206: of the stationary Eqs. (\ref{req3})-(\ref{req4})
207: has been obtained (either analytically in the anti-continuum
208: limit, or numerically for $\epsilon \neq 0$), we use the perturbation
209: ansatz:
210: \begin{eqnarray}
211: a_n &=& a_n^0 + \delta \left(c_n e^{-i \omega z} + d_n e^{i \omega^{\star} z}
212: \right)
213: \label{req5}
214: \\
215: b_n &=& b_n^0 + \delta \left(f_n e^{-i \omega z} + g_n e^{i \omega^{\star} z}
216: \right).
217: \label{req6}
218: \end{eqnarray}
219: In this ansatz, $\delta$ is a (small) formal expansion parameter and
220: %$\omega_{2}$ = -$\omega_{1} \equiv \omega$
221: $\omega$ is the eigenfrequency of the
222: perturbation eigenvector
223: $(c_n,d_n^{\star},f_n,g_n^{\star})^T$ (where the superscript denotes
224: the transpose). Then, upon lengthy but straightforward calculation,
225: one obtains the linearized matrix eigenvalue problem to O$(\delta)$.
226: These dynamical equations have the form:
227: \[
228: \omega \left(
229: \begin{array}{c}
230: c_{n} \\
231: d_{n}^{\star} \\
232: f_{n} \\
233: g_{n}^{\star}
234: \end{array}
235: \right) ={\bf L}\cdot \left(
236: \begin{array}{c}
237: c_{n} \\
238: d_{n}^{\star} \\
239: f_{n} \\
240: g_{n}^{\star}
241: \end{array}
242: \right) ,\newline
243: \]
244: where
245: \[
246: {\bf L}=\left(
247: \begin{array}{cccc}
248: L_{11} & L_{12} & L_{13} & L_{14} \\
249: L_{21} & L_{22} & L_{23} & L_{24} \\
250: L_{31} & L_{32} & L_{33} & L_{34} \\
251: L_{41} & L_{42} & L_{43} & L_{44}
252: \end{array}
253: \right) .\newline
254: \]
255: The $N \times N$ (where $N$ is the size of the lattice) blocks
256: of the linearization matrix are given by:
257: \begin{eqnarray}
258: L_{11} &=& (q-1)-\epsilon (\Delta_2+2) - 2 |a_n^0|^2 - A |b_n^0|^2
259: \label{req7}
260: \\
261: L_{12} &=& -(a_n^0)^2 - B (b_n^0)^2
262: \label{req8}
263: \\
264: L_{13} &=& -A a_n^0 (b_n^0)^{\star} - 2 B (a_n^0)^{\star} b_n^0
265: \label{req9}
266: \\
267: L_{14} &=& -A a_n^0 b_n^0
268: \label{req10}
269: \\
270: L_{21} &=& - L_{12}^{\star}
271: %((a_n^0)^{\star})^2 + B ((b_n^0)^{\star})^2
272: \label{req11}
273: \\
274: L_{22} &=& -L_{11}
275: \label{req12}
276: \\
277: L_{23} &=& - L_{14}^{\star}
278: \label{req13}
279: \\
280: L_{24} &=& - L_{13}^{\star}
281: \label{req14}
282: \\
283: L_{31} &=& L_{13}^{\star}
284: \label{req15}
285: \\
286: L_{32} &=& L_{14}
287: \label{req16}
288: \\
289: L_{33} &=& (q+1)-\epsilon (\Delta_2+2) - 2 |b_n^0|^2 - A |a_n^0|^2
290: \label{req17}
291: \\
292: L_{34} &=& - (b_n^0)^2 - B (a_n^0)^2
293: \label{req18}
294: \\
295: L_{41} &=& -L_{14}^{\star}
296: \label{req19}
297: \\
298: L_{42} &=& - L_{13}
299: \label{req20}
300: \\
301: L_{43} &=& - L_{34}^{\star}
302: \label{req21}
303: \\
304: L_{44} &=& - L_{33}
305: \label{req22}
306: \end{eqnarray}
307: In the above, the short-hand notation $(\Delta_2+2) z_n = z_{n+1} + z_{n-1}$,
308: invoking the concept of the discrete Laplacian $\Delta_2$. With this linearized
309: matrix eigenvalue problem, we can establish a
310: criterion for the existence and stability
311: of solutions to this system by examination of our
312: system at the anti-continuum limit
313: (i.e., when $\epsilon = 0$).
314:
315:
316: \section{The Anti-Continuum Limit}
317:
318: \subsection{Existence}
319:
320: We begin by examining eqs. $(\ref{req3})-(\ref{req4})$ for the case $\epsilon$ = 0.
321: Setting $a_n=r_n e^{i \theta_n}$ and $b_n=s_n e^{i \phi_n}$ in this case, one finds:
322: \begin{eqnarray}
323: (q-1)-(r_n^2+A s_n^2) - B s_n^2 e^{2 i (\phi_n-\theta_n)} &=& 0
324: \label{req23}
325: \\
326: (q+1)-(s_n^2+A r_n^2) - B r_n^2 e^{-2 i (\phi_n-\theta_n)} &=& 0.
327: \label{req24}
328: \end{eqnarray}
329: From the above, we infer that for solutions to exist in this limit
330: it is necessary for $2 (\theta_n-\phi_n)$ to be an integer multiple of
331: $\pi$, hence, we obtain that
332: \begin{eqnarray}
333: \theta_n-\phi_n= k \frac{\pi}{2},
334: \label{req25}
335: \end{eqnarray}
336: with $k \in {\cal Z}$.
337:
338: The simplest possible solutions are the ones
339: that involve only one of the two branches and were hence termed
340: TE and TM modes respectively in \cite{meier}. The TE solution
341: of Eqs. (\ref{req23})-(\ref{req24}) has the form (in the present limit)
342: \begin{eqnarray}
343: r_n= \pm \sqrt{q-1} \quad s_n =0
344: \label{req26}
345: \end{eqnarray}
346: and exists only for $q>1$.
347: On the other hand, the TM mode features
348: \begin{eqnarray}
349: r_n=0 \quad s_n= \pm \sqrt{q+1},
350: \label{req27}
351: \end{eqnarray}
352: and is only present for $q>-1$.
353:
354: Also, there are two possible mixed mode solutions permitted by the quantization
355: condition of eq. (\ref{req25}), allowing the exponential of Eqs.
356: (\ref{req23})-(\ref{req24}) to be $+1$ or $-1$ respectively. The first
357: one $(e^{2i(\theta_{n} - \phi_{n})} = 1)$ was characterized as a linearly polarized
358: (LP) branch in \cite{meier},
359: involving in-phase contributions from both the TE and TM components.
360: In this case, the linear system of Eqs. (\ref{req23})-(\ref{req24})
361: has the general solution:
362: \begin{eqnarray}
363: r_n &=& \pm \sqrt{\frac{(A+B)(q+1)-(q-1)}{(A+B)^2-1}}
364: \label{req28}
365: \\
366: s_n &=& \pm \sqrt{\frac{(A+B)(q-1)-(q+1)}{(A+B)^2-1}}.
367: \label{req29}
368: \end{eqnarray}
369: If $(A+B)^2>1$ (which is the case in the experimentally relevant
370: setting), this branch only exists for $(A+B) (q+1)>(q-1)$
371: and $(A+B) (q-1)>(q+1)$ [the sign of the two above
372: inequalities needs to be reversed for existence conditions
373: in the case of $(A+B)^2<1$]. Among the two conditions, in the
374: present setting, the second one is the most ``stringent''
375: for the case $A=2B=1$, which yields the constraint $q \geq 5$
376: (while the first condition requires for the same parameters $q \geq -5$).
377:
378: The second possibility for a mixed mode (stemming from
379: $e^{2 i (\theta_n-\phi_n)}=-1$ in Eqs. (\ref{req23})-(\ref{req24}))
380: yields the, so-called, elliptically polarized mode (EP) \cite{meier},
381: involving a $\pi/2$ phase shift between the TE and TM components.
382: The relevant amplitudes in this case are:
383: \begin{eqnarray}
384: r_n &=& \pm \sqrt{\frac{(A-B)(q+1)-(q-1)}{(A-B)^2-1}}
385: \label{req30}
386: \\
387: s_n &=& \pm \sqrt{\frac{(A-B)(q-1)-(q+1)}{(A-B)^2-1}}.
388: \label{req31}
389: \end{eqnarray}
390: The condition for this branch to exist if $(A-B)^2<1$
391: is $q-1 \geq (A-B) (q+1)$ and $q+1 \geq (A-B) (q-1)$
392: (once again the signs should be reversed if $(A-B)^2>1$).
393: In this case, the first condition is more constraining
394: than the second, imposing for our
395: special case of interest herein ($A=2B=1$) $q \geq 3$
396: (while the second condition only requires $q \geq -3$).
397:
398: \subsection{Stability}
399:
400: We can now turn to the study of the stability of the corresponding
401: branches via the anti-continuum limit ($\epsilon = 0$). The major advantage of
402: this limit is that all the blocks of the linearization
403: matrix $L$ now become diagonal. Furthermore, the structures
404: that we consider, per the branches presented above, involve the excitation
405: of a {\it single site} (all other sites are inert) with amplitudes
406: corresponding to the appropriate
407: TE, TM, LP or EP modes. The interesting byproduct of this formulation is
408: that all inert sites end up yielding purely diagonal entries
409: in the stability matrix equal to $(q-1)$ in the first (2N-1)$\times$(2N-1)
410: blocks and $(q+1)$ in the second set of (2N-1)$\times$(2N-1) blocks.
411: Hence the matrix will have $2N-1$ eigenfrequencies equal to $(q-1)$
412: and equally many eigenfrequencies of $(q+1)$. The remaining $4 \times 4$
413: matrix will correspond to the {\it single excited site} of the
414: lattice and its eigenvalues will yield the eigenfrequencies
415: pertaining to the corresponding branch of solutions. Hence, by solving
416: this considerably simpler eigenvalue problem, we can infer the full
417: stability picture for $\epsilon \neq 0$ from the anti-continuum limit.
418: This, we believe, is a major advantage in using this approach.
419:
420:
421: We note that due to the fact that the power is conserved for our system (see previous
422: section), two of the relevant eigenvalues of this $4 \times 4$ problem should be identically 0.
423: This is confirmed in the computations presented here.
424:
425:
426: Now, taking the case for the ``TE'' mode in eqs.$(\ref{req7})-(\ref{req22})$, we determine the
427: remaining pair of eigenvalues in this case:
428: \begin{eqnarray}
429: \omega_{TE}=\pm \sqrt{(q+1)^2+(A^2-B^2) (q-1)^2-2 A (q^2-1)}.
430: \label{req32}
431: \end{eqnarray}
432: Examining our model for the experimental case, $A$ = 1, $B$ = 1/2 and $q > 0$,
433: we find that this eigenfrequency is real for $q < 5$, while it is imaginary for $q>5$
434: (hence, implying the presence of an instability,
435: since the corresponding eigenvalue $\lambda=i \omega$ becomes real
436: and exponential growth occurs $\propto e^{\lambda z}$). Indeed, we can show that
437: for the TE mode, stability is ensured for $1 < q < 5$.
438:
439: For the TM mode, the corresponding expression can be obtained as:
440: \begin{eqnarray}
441: \omega_{TM}=\pm \sqrt{(q-1)^2+(A^2-B^2) (q+1)^2-2 A (q^2-1)},
442: \label{req33}
443: \end{eqnarray}
444: and for the case of interest, stability is ensured for $-1 < q < 3$, while
445: instability ensues (due to a real eigenvalue) for $q > 3$.
446:
447: For the LP mode, the expression for the corresponding eigenfrequency
448: is too cumbersome to provide in analytic form (even though such a form
449: has been obtained) for general $A$ and $B$. Hence, we will restrict
450: ourselves to the case of $A=2B=1$, where the relevant eigenfrequency
451: reads:
452: \begin{eqnarray}
453: \omega_{LP}= \pm \frac{2 \sqrt{2}}{5} \sqrt{q^2-25},
454: \label{req34}
455: \end{eqnarray}
456: implying stability for $q \geq 5$ (which coincides with the branch's
457: existence region).
458:
459: Finally, for the EP mode, the expression similarly reads:
460: \begin{eqnarray}
461: \omega_{EP} \pm \frac{2 \sqrt{2}}{3} \sqrt{q^2-9},
462: \label{req35}
463: \end{eqnarray}
464: implying, in principle, stability for $q \geq 3$ (again, coinciding with the
465: solutions domain of existence); see, however, the discussion below for
466: this branch.
467:
468: From the above observations, one can put together the complete
469: bifurcation picture of the four branches and their corresponding
470: stability in the anti-continuum limit (again, we present this here
471: for concreteness for $A=2B=1$). More specifically,
472: the TE branch exists for $q \geq 1$ and is stable for
473: $1 \leq q \leq 5$. For $q>5$, the branch is destabilized as
474: a new branch emerges, namely the LP branch, through a
475: pitchfork bifurcation; notice that the TM
476: component of this branch, per Eq. (\ref{req29}) is exactly zero
477: at $q=5$, hence it directly bifurcates from the TE branch.
478: The two branches of this super-critical pitchfork correspond
479: to the two signs of $s_n$ in Eq. (\ref{req29}). The bifurcating
480: branch ``inherits'' the stability of the TE branch for
481: all larger values of $q$, while
482: the latter branch remains unstable thereafter.
483:
484: In a similar development, the TM branch exists for $q \geq -1$
485: and is stable in the interval $-1 \leq q \leq 3$. However, at
486: $q=3$, a new branch (in fact, a pair of branches)
487: bifurcates acquiring non-zero $r_n$, beyond
488: the bifurcation point, as per Eq. (\ref{req30}). This is accompanied
489: by the destabilization of the TM branch (due to a real eigenvalue)
490: and the {\it apparent}
491: stability of the ensuing EP branch for all values of $q>3$.
492:
493: Fig. \ref{rfig1} shows how the two supercritical pitchfork bifurcations
494: are generated as a function of the coordinates $(r_{n}, s_{n}, q)$. We
495: can now turn on the coupling and see how the existence and stability of
496: each branch is affected by varying $q$ and $\epsilon$.
497:
498:
499: \begin{figure}[tb]
500: \medskip
501: \centerline{
502: \includegraphics[width=8cm,angle=0,clip]{rlh4.eps}
503: }
504: \caption{The full bifurcation diagram for $\epsilon=0$, illustrating the two
505: supercritical
506: pitchfork bifurcations at $q=3$ and $q=5$. These give rise to the formation
507: of the EP branch (the former) and the LP branch (the latter). The blue lines
508: illustrate stable branches, while the red lines show the unstable ones.}
509: \label{rfig1}
510: \end{figure}
511:
512: %\subsection{TE branch}
513:
514: %\subsection{TM branch}
515:
516: %\subsection{LP branch}
517:
518: %\subsection{EP branch}
519:
520: \section{Finite Coupling}
521:
522: One of the key features that will help us in our analysis of the
523: finite coupling case consists of determining where the bands of the continuous
524: spectrum of the linearized problem occurs. This is obtained analytically using
525: the plane wave ansatz $a_{n} \sim e^{i (kn - \omega z)}$ and $b_{n} \sim
526: e^{i (kn - \omega z)}$. Plugging our ansatz into the coupled linearized problem
527: (essentially, using eqs. $(\ref{req1})-(\ref{req2})$ and
528: ignoring the nonlinear terms), the
529: resulting dispersion relations are
530: \begin{eqnarray}
531: \omega &=& (q-1) - 2 \epsilon \cos(k)
532: \label{req36}
533: \\
534: \omega &=& (q+1) - 2 \epsilon \cos(k).
535: \label{req37}
536: \end{eqnarray}
537: Hence the continuous spectrum (in the infinite lattice case)
538: will consist of the frequency intervals $[q-1-2\epsilon,q-1+2\epsilon]$
539: and $[q+1-2\epsilon,q+1+2\epsilon]$. As may be expected, for
540: $\epsilon=0$, these intervals degenerate to two single points
541: ($\omega=q-1$ and $\omega=q+1$), in consonance with the discussion
542: of the previous section. But perhaps more importantly, for
543: $q-1 < 2 \epsilon$ (equivalently for $\epsilon > (q-1)/2$), the
544: continuous spectrum branch will be crossing the origin, leading
545: to the collision of the eigenvalues with their mirror symmetric
546: opposites, that will in turn lead to instabilities. Hence, we need
547: {\it only} consider couplings in the interval $\epsilon \in [0,(q-1)/2]$.
548:
549: We now numerically examine each of the main four branches, using
550: one-parameter continuation in the above-mentioned interval of
551: $\epsilon$'s for different values of $q$ and thus constructing
552: a two-parameter bifurcation diagram for each branch. Notice that
553: the solutions are numerically obtained for each pair of
554: $(q,\epsilon)$ by means of a fixed point iteration. Once
555: convergent to within a pre-set tolerance ($10^{-8}$ typically for
556: the computations presented herein), the procedure is followed by numerical
557: linear stability analysis, obtaining the eigenfrequencies for the linearization
558: matrix $L$.
559:
560: \subsection{TE branch}
561:
562: The continuation of the TE branch is detailed in Fig. \ref{rfig2}.
563: It turns out that especially in the case of this branch, solutions
564: cannot be obtained for $\epsilon > (q-1)/2$, i.e., the branch
565: terminates at that point with its amplitude going to zero at
566: this critical point. Within its region of existence, the branch
567: has a domain of stability and one of instability. The point of separation
568: between the two in the anti-continuum limit, studied
569: previously, was the critical point of $q=5$. For $\epsilon \neq 0$,
570: the separatrix curve is shown in Fig. \ref{rfig2} and can be well
571: approximated by the curve $\epsilon_{TE}^c \approx (4 \sqrt{2}/5) \sqrt{q-5}$.
572: Hence for $q\leq 5$, the solution is stable for all values of
573: $\epsilon$ in its range of existence ($0<\epsilon<(q-1)/2$), while
574: for $q \geq 5$, the solution is only stable for $\epsilon_{TE}^c < \epsilon<
575: (q-1)/2$ and unstable (due to a real eigenvalue pair) for $0< \epsilon <
576: \epsilon_{TE}^c$.
577:
578:
579: \begin{figure}[tbp]
580: \centerline{
581: \includegraphics[width=8.cm,height=5cm,angle=0,clip]{rlh5.eps}
582: \includegraphics[width=8.cm,height=5cm,angle=0,clip]{rlh5a.eps}
583: }
584: \medskip
585: \centerline{
586: \includegraphics[width=8cm,angle=0,clip]{rlh5b.eps}
587: }
588: \caption{TE branch: The top left set of panels shows the
589: details of the TE branch as a function of $\epsilon$ for
590: $q=3$. The topmost panel shows the norm of the branch as
591: a function of $\epsilon$, while the left and right panels
592: show the solution profile and its spectral plane for
593: $\epsilon=0.1$ and $\epsilon=0.9$ respectively. The top
594: right set of panels shows similar details, but for $q=5.5$.
595: The top right panel in the right case, which is unstable
596: for small $\epsilon$, shows the most unstable eigenvalue
597: real part. The middle and bottom panels show the (unstable) solution
598: and its stability for $\epsilon=0.5$ and the (stable) one for
599: $\epsilon=1.5$ respectively. The bottom panel shows the two
600: parameter bifurcation diagram of the coupling $\epsilon$ as a function
601: of $q$ (for details see text). All the relevant existence and stability
602: regimes have been accordingly labeled. The top left and top right
603: set of panels can be considered as two ``vertical cuts'' across
604: this bifurcation diagram for $q=3$ and $5.5$ respectively.}
605: \label{rfig2}
606: \end{figure}
607:
608: \subsection{TM branch}
609:
610: The TM branch is considerably more complicated, structurally, than
611: its TE counterpart. Firstly, it does not disappear beyond the critical
612: $\epsilon = (q-1)/2$, however, it does become unstable as predicted
613: previously, hence we will, once again, restrict ourselves to this
614: parameter range. Furthermore, similarly to the TE branch case, there is an
615: $\epsilon_{TM}^c$ below which the branch is {\it always} unstable,
616: whereas for $\epsilon>\epsilon_{TM}^c$, the branch {\it may}
617: be stable. At the anti-continuum limit, the critical point
618: for the instability is $q=3$, as discussed previously; for
619: $q>3$, the critical point is obtained numerically in Fig. \ref{rfig3}.
620: It can be well approximated (close to $q=3$) by
621: $\epsilon_{TM}^c \approx (9/10) \sqrt{q-3}$.
622:
623: However, within the range of {\it potential} stability
624: ($0\leq \epsilon \leq (q-1)/2$ for $q \leq 3$, and
625: $\epsilon_{TM}^c \leq \epsilon \leq (q-1)/2$ for
626: $q \geq 3$), we observe an additional large region of
627: instability in the two parameter bifurcation diagram of
628: Fig. \ref{rfig3}, due to a complex quartet of eigenvalues.
629: This instability appears to ``emanate'' from the point
630: with $(q,\epsilon)=(2.2,0)$ in the anti-continuum limit, and
631: to linearly expand its range as $\epsilon$ increases. Hence,
632: we will try to understand it at the level of $\epsilon=0$.
633: What happens at the value of $q=2.2$ in the $\epsilon=0$ limit
634: is that the point spectrum eigenfrequency of
635: Eq. (\ref{req33}) ``collides'' with the continuous spectrum
636: point of concentration, corresponding to $\omega=q-1$.
637: However, the eigenvector of this eigenvalue
638: \footnote{The eigenvector can be found explicitly for the eigenvalues of
639: Eq. (\ref{req33}) and in fact is $(-(4 \pm \sqrt{16-(q+1)^2})/(q+1),1,0,0)^T$}
640: has a Krein signature (see e.g. \cite{aubry,joh,kevkap} for relevant
641: definitions of this signature) opposite to that of the continuous
642: band at $\omega=q-1$. Hence, according to Krein theory \cite{mackay}, the
643: resulting collision leads to the formation of a quartet of eigenvalues
644: emerging in the complex plane and, in turn, implying the instability of
645: the TM configuration. As $\epsilon$ is varied from $0$, the continuous
646: spectrum band grows linearly in $\epsilon$, hence the corresponding
647: interval of $q$'s, where this instability is present
648: (due to the collision with the opposite Krein sign eigenvalue)
649: also grows at the
650: same rate. Along the same vein, it is worth pointing out that the line
651: of this instability threshold and that of $\epsilon=(q-1)/2$ are parallel.
652:
653: \begin{figure}[tbp]
654: \centerline{
655: \includegraphics[width=8.cm,height=5cm,angle=0,clip]{rlh6.eps}
656: \includegraphics[width=8.cm,height=5cm,angle=0,clip]{rlh6a.eps}
657: }
658: \medskip
659: \centerline{
660: \includegraphics[width=8cm,angle=0,clip]{rlh6b.eps}
661: }
662: \caption{TM branch: the figures are similar as in Fig. \ref{rfig3},
663: but now for the TM branch. The top left set of panels shows the
664: continuation in $\epsilon$ for $q=2.5$, along with the special
665: cases of $\epsilon=0.1$ (stable) and $\epsilon=0.5$
666: (unstable due to quartet of eigenvalues) in the left
667: and right middle and bottom panels. The top right set of panels
668: shows the $q=3.5$ (unstable) case with the middle/bottom panels
669: corresponding to $\epsilon=0.5$ (unstable due to real
670: eigenvalue pair)/$\epsilon=0.65$ (stable) respectively.} \label{rfig3}
671: \end{figure}
672:
673: \subsection{LP branch}
674:
675: The mixed LP branch, involving in-phase contributions of the
676: TE and TM has been continued for various values of $q>5$,
677: as a function of $\epsilon$. Recall that this
678: branch emerges through a supercritical pitchfork
679: as $q$ is varied for for fixed $\epsilon$ (see Fig. \ref{rfig2}).
680: In an exactly analogous way, if $\epsilon$ is varied
681: for a fixed $q>5$, the branch appears to terminate
682: at $\epsilon=\epsilon_{TE}^c$ through a subcritical pitchfork.
683: Hence, the LP branch only arises for $0<\epsilon<\epsilon_{TE}^c$
684: and $q>5$ and it is stable throughout its interval of existence.
685: For $\epsilon > \epsilon_{TE}^c$, this branch ``degenerates''
686: back into the stable (for such $\epsilon$'s) TE mode.
687: This is clearly illustrated in Fig. \ref{rfig4} that shows
688: the continuation of the LP branch as a function of $\epsilon$
689: for a fixed $q=5.5$. The top panels show the norms of the
690: two components illustrating that the second component is
691: absent for $\epsilon>\epsilon_{TE}^c=0.81$ in this case.
692: The middle and bottom left panels show a
693: case for $\epsilon=0.5<\epsilon_{TE}^c$
694: where a genuine (and stable) TE solution exists, while the corresponding
695: right panels are for $\epsilon=1>\epsilon_{TE}^c$, past the critical
696: point where the LP solution degenerates into a TE mode.
697:
698: \begin{figure}[tbp]
699: %\centerline{
700: %\includegraphics[width=8.cm,height=5cm,angle=0,clip]{rlh6.eps}
701: %\includegraphics[width=8.cm,height=5cm,angle=0,clip]{rlh6a.eps}
702: %}
703: %\medskip
704: \centerline{
705: \includegraphics[width=8cm,angle=0,clip]{rlh7.eps}
706: }
707: \caption{LP branch: The top left and top right panel shows the
708: dependence on $\epsilon$ of the respective powers
709: of the two components $P_1$ and $P_2$ ($P=P_1+P_2$), for
710: $q=5.5$. The middle and bottom panel
711: panel show the profile and the linear stability of the
712: LP mode while it is still present ($\epsilon=0.5$; left panels)
713: and of the TE mode that it degenerates to, beyond $\epsilon_{TE}^c$
714: ($\epsilon=1$; right panels).} \label{rfig4}
715: \end{figure}
716:
717:
718: \subsection{EP branch}
719:
720: The case of the EP branch, analyzed in Fig. \ref{rfig5},
721: is somewhat analogous to that
722: of the LP branch. For fixed $q$ close to (and larger than) $3$
723: and varying $\epsilon$,
724: the branch exists and is {\it stable} for $0<\epsilon<\epsilon_{TM}^c$.
725: For larger values of $\epsilon$, the EP mode degenerates into a
726: TM mode, as is shown in the top left panel of Fig. \ref{rfig5} for
727: $q=3.5$ (cf. with the analogous for the LP case of Fig. \ref{rfig4}).
728:
729: However, for $q>3.62$, this phenomenology appears to change and
730: an expanding (for increasing $q$) interval of oscillatory instability
731: within the range of existence of the EP branch
732: ($0<\epsilon<\epsilon_{TM}^c$) seems to arise. To rationalize
733: this feature, we again turn to the anti-continuum limit.
734: Just as in the case of the (TM) branch from which it originates,
735: the EP branch has a point spectrum eigenfrequency given by Eq. (\ref{req35})
736: which has a negative Krein signature and upon collision with the continuous
737: spectrum band of eigenfrequencies will give rise to an instability. Setting
738: the frequency of Eq. (\ref{req35}) equal to $q-1$, we obtain that
739: this collision occurs at $q=9$. For lower values of $q$, this ``collision''
740: will occur for a finite (nonzero) interval of values of $\epsilon$
741: --see e.g. the top right panels of Fig. \ref{rfig5} for $q=4.5$--, which,
742: in turn, yields the region of oscillatory instability of the EP mode,
743: clearly separated from its region of stability by a dashed line
744: in the bottom panel of Fig. \ref{rfig5}.
745:
746: In conclusion, the EP solutions exist in the regime where the TM
747: mode is unstable (cf. bottom panels of Figs. \ref{rfig3} and
748: \ref{rfig5}), illustrated by the curve originating from the point
749: $(q,\epsilon)=(3,0)$ in the anti-continuum limit. However, the
750: EP solution, as is shown in the top panels of the figure,
751: may (top left) or may not (top right) necessarily be stable
752: in its region of existence due to the appearance of an
753: oscillatory instability for sufficiently large $q$.
754:
755: \begin{figure}[tbp]
756: %\centerline{
757: %\includegraphics[width=8.cm,height=5cm,angle=0,clip]{rlh6.eps}
758: %\includegraphics[width=8.cm,height=5cm,angle=0,clip]{rlh6a.eps}
759: %}
760: %\medskip
761: \centerline{
762: \includegraphics[width=8cm,angle=0,clip]{rlh8.eps}
763: \includegraphics[width=8cm,angle=0,clip]{rlh9.eps}
764: }
765: \centerline{
766: \includegraphics[width=8cm,angle=0,clip]{rlh10.eps}
767: }
768: \caption{EP branch: The panels are analogous to
769: those of Fig. \ref{rfig3} but for the EP branch;
770: the top right panel, however, shows the non-vanishing
771: power of the second component.
772: Top left panels: $q=3.5$. The middle
773: and bottom panels correspond to $\epsilon=0.5<\epsilon_{TM}^c$
774: (left panels) and to $\epsilon=1>\epsilon_{TM}^c$ (right panels).
775: In the latter case, the solution has already degenerated into
776: a TM mode. Top right panels: q=4.5. The middle and bottom panels
777: correspond to the stable case of $\epsilon=0.1$ (left panels)
778: and the unstable case of $\epsilon=1$ (right panels). The bottom
779: panel shows the two-parameter diagram for this mode. The region
780: for $q>3$ and $0<\epsilon<\epsilon_{TM}^c$, where this modes exists
781: is separated by the dashed line into a stable and an unstable regime.
782: For comparison the extension of the regions of stability/instability
783: of the TM mode from Fig. \ref{rfig3} are also included. We note
784: that the choice of star symbols (instead of circles as e.g. in Figs.
785: \ref{rfig2}, \ref{rfig4}) in the spatial profile of $a_n$ is used
786: to indicate that the imaginary part of $a_n$ is shown.}
787: \label{rfig5}
788: \end{figure}
789:
790:
791: %****************************************************************
792:
793: % We need descriptions for figures 4 and 5! These concern some
794: % special case of our parameters. Do we need to include this or
795: % not?
796:
797: %****************************************************************
798:
799: \section{Dynamical Evolution of Unstable Branches}
800:
801: We now examine a series of different instability scenarios
802: pertaining to the branches of solutions discussed above. In so doing,
803: we solve Eqs. (\ref{req1})-(\ref{req2}) using a fourth-order
804: Runge-Kutta algorithm. To ensure the validity of
805: our numerical results, we have monitored the
806: conservation of
807: the power of Eq. (\ref{req3a}) and of the Hamiltonian (energy)
808: of the system:
809: \begin{eqnarray}
810: H= -\sum_n \left(\epsilon (a_n^{\star}a_{n+1}+a_n a_{n+1}^{\star}
811: +b_n^{\star}b_{n+1}+b_n b_{n+1}^{\star}) + |a_n|^2-|b_n|^2
812: + \frac{1}{2} (|a_n|^4+|b_n|^4+A |a_n|^2 |b_n|^2)
813: + \frac{B}{2} (a_n^2 (b_n^{\star})^2+b_n^2 (a_n^{\star})^2) \right).
814: \label{req38}
815: \end{eqnarray}
816: The power is conserved to numerical precision, while the Hamiltonian
817: is conserved to 1 part in $10^8$. In all cases, $150$ spatial sites
818: are used in our computations.
819:
820: Our first example, shown in Fig. \ref{rfig6},
821: illustrates the case of $q=5.5$ and $\epsilon=0.5$
822: for the TE mode. This is a case representative of
823: the region of parameter space
824: (cf. Fig. \ref{rfig2}), where the TE mode is unstable due to a real
825: eigenvalue pair (while the LP mode is stable).
826: The solution actually coincides with the steady TE branch until
827: approximately time $10$.
828: The configuration then deviates from the TE branch and
829: subsequently develops a breathing oscillation around a stable LP state.
830:
831: \begin{figure}[tbp]
832: \centerline{
833: \includegraphics[width=8.cm,height=5cm,angle=0,clip]{teq5p5ep5A.eps}
834: \includegraphics[width=8.cm,height=5cm,angle=0,clip]{teq5p5ep5B.eps}}
835: \caption{This demonstrates the instability in the TE mode when
836: $\epsilon =0.5$ and $q=5.5$. The left panel shows the space-time
837: contour plot of the
838: modulus square of the two fields ($|a_n(t)|^2$ in the top panel and
839: $|b_n(t)|^2$ in the bottom panel). The right panel shows the evolution
840: of the (square modulus of the) central site of the configuration as a
841: function of time.}
842: \label{rfig6}
843: \end{figure}
844:
845: The next three cases are concerned with the different scenarios
846: for the instability of the TM mode. Fig. \ref{rfig7},
847: with $q=3.5$ and $\epsilon=0.5$, is representative
848: of the region of parameter space
849: where the TM mode is unstable due to a real eigenvalue pair, while the
850: EP mode is stable. In this case, As in the previous case,
851: the solution coincides
852: with the steady TM mode up until approximately time $10$. It is then observed
853: that for
854: short times the configuration deviates from the TM branch and oscillates
855: around the (stable) EP configuration. However, for longer times,
856: the system appears to wander away from the latter configuration
857: and into a breathing state, where the two components have roughly
858: equal power and exchange a small fraction of it (almost) periodically.
859:
860:
861: \begin{figure}[tbp]
862: \centerline{
863: \includegraphics[width=8.cm,height=5cm,angle=0,clip]{tmq3p5ep5A.eps}
864: \includegraphics[width=8.cm,height=5cm,angle=0,clip]{tmq3p5ep5B.eps}
865: }
866: \caption{An instability of the TM mode is demonstrated here with
867: $\epsilon =0.5$ and $q=3.5$. The left panel again shows the
868: space-time contour plot of the
869: modulus square of the two fields ($|a_n(t)|^2$ in the top panel and
870: $|b_n|^2$ in the bottom panel). The right panel shows the evolution
871: of the (square modulus of the) central site of the configuration as a
872: function of time.}
873: \label{rfig7}
874: \end{figure}
875:
876: The second instability scenario of the TM branch, shown in
877: Fig. \ref{rfig8}, involves the case
878: of an eigenvalue quartet as e.g.
879: for $q=2.5$ and $\epsilon=0.5$ (cf. Fig. \ref{rfig3}).
880: In this case, we observe the configuration departing from the
881: unstable TM steady state and
882: resulting into a breathing solution involving both components. In this
883: and the remaining cases, the departure from the steady state occurs later
884: than in the first two cases.
885:
886: \begin{figure}[tbp]
887: \centerline{
888: \includegraphics[width=8.cm,height=5cm,angle=0,clip]{tmq2p5ep5A.eps}
889: \includegraphics[width=8.cm,height=5cm,angle=0,clip]{tmq2p5ep5B.eps}}
890: \caption{The instability of a TM mode is shown here for
891: $\epsilon =0.5$ and $q=2.5$, using exactly the same diagnostics
892: as in Fig. \ref{rfig7}.}
893: \label{rfig8}
894: \end{figure}
895:
896: Next, we examine the evolution of the instability for the TM
897: mode in the presence, also, of continuous spectrum instabilities
898: for $q=2.5$ and $\epsilon=1$ (cf. Fig. \ref{rfig3}). In this case
899: a different phenomenology occurs.
900: %We observe the breakup of
901: %the mode as it dissipates into smaller pieces.
902: The mode is completely destroyed in favor of small amplitude excitations.
903:
904: \begin{figure}[tbp]
905: \centerline{
906: \includegraphics[width=8.cm,height=5cm,angle=0,clip]{tmq2p5e1A.eps}
907: \includegraphics[width=8.cm,height=5cm,angle=0,clip]{tmq2p5e1B.eps}
908: }
909: \caption{Same as the previous 2 figures, but for the case of
910: $\epsilon =1$ and $q=2.5$. Here the nonlinear mode is completely destroyed
911: by the instability and results into small amplitude extended waves.}
912: \label{rfig9}
913: \end{figure}
914:
915: %Next, we examine the evolution of the instability for the TM
916: %mode in the presence, also, of continuous spectrum instabilities
917: %for $q=2.5$ and $\epsilon=.75$. This provides a validation of the
918: %phenomena that happens in figure \ref{rfig9}.
919: %we observe again the breakup of the mode into smaller pieces.
920: %\begin{figure}[tbp]
921: %\centerline{
922: %\includegraphics[width=8.cm,height=5cm,angle=0,clip]{tmq2p5e1A.eps}
923: %\includegraphics[width=8.cm,height=5cm,angle=0,clip]{tmq2p5e1B.eps}
924: %}
925: %\caption{An unstable TM mode with
926: %$\epsilon =.75$ and $q=2.5$.}
927: %\label{rfig10}
928: %\end{figure}
929:
930: Finally, we examine the evolution of the instability for an elliptically
931: polarized mixed mode case within its regime of oscillatory
932: instability (cf. Fig. \ref{rfig5}), for $\epsilon =0.5$ and $q=5$.
933: We observe the configuration departing the steady state and giving
934: rise to a persistent breathing state involving both components.
935:
936: \begin{figure}[tbp]
937: \centerline{
938: \includegraphics[width=8.cm,height=5cm,angle=0,clip]{mmq5ep5A.eps}
939: \includegraphics[width=8.cm,height=5cm,angle=0,clip]{mmq5ep5B.eps}
940: }
941: \caption{An unstable elliptically polarized mode with
942: $\epsilon =0.5$ and $q=5$ is monitored by the same diagnostics
943: as in the previous figures.}
944: \label{rfig11}
945: \end{figure}
946:
947:
948: \section{Conclusions}
949:
950: In this paper, we have examined in detail the principal modes arising in the
951: discrete vector lattices of AlGaAs waveguides, motivated by the recent
952: experimental investigations of \cite{meier,meier2}. We have constructed
953: the full two-parameter bifurcation diagrams of all four principal
954: branches of solutions. These include the transverse electric and
955: transverse magnetic modes (occupying respectively each of the two
956: components of the vector lattice), but also the linearly polarized
957: (two components in or out of phase) and elliptically polarized modes (two
958: components out of phase by odd multiples of $\pi/2$). The latter
959: two modes, emerge from the former two in pairs through a pitchfork
960: bifurcation that has been completely elucidated through the study
961: of the system in the anti-continuum limit. In fact, the anti-continuum
962: limit has provided with a very effective tool to
963: {\it analytically} highlight existence
964: and stability properties of the different branches as well as
965: potential instabilities. We have continued this
966: monoparametric picture (as a function of the propagation constant
967: of the solutions, but at zero coupling) to finite coupling fleshing
968: out the regimes of stability and instability, and of existence as
969: well as of disappearance of the relevant branches of solutions.
970: The relevant two-parameter diagrams (Fig. \ref{rfig2} for the TE
971: mode, Fig. \ref{rfig3} for the TM, Fig. \ref{rfig4} for the LP and
972: Fig. \ref{rfig5} for the EP mode) constitute the main result
973: of the present paper. We have, however, complemented our existence
974: and stability analysis with direct numerical ``experiments'', illustrating
975: the different instability scenario for the modes of interest and how
976: they can either lead to the excitation of other branches, or even to
977: the complete annihilation of the relevant localized waves.
978:
979: It is noteworthy that the process of obtaining such a multiparametric
980: bifurcation picture has allowed us to reveal a considerably deeper
981: understanding of the system's parameter space, in comparison to
982: earlier publications (such as e.g. the theoretical conclusions
983: of the works of \cite{meier,meier2}). In particular, for instance,
984: the limited parameter space examined in the latter setting (where
985: a monoparametric continuation was used for $\epsilon \simeq 0.921$
986: and apparently varying $q$ only for $q > 3$) led the authors to infer that the
987: TM mode is {\it always} unstable and similarly that the bifurcating
988: EP mode is {\it also} unstable. Here, we have illustrated explicitly
989: regions of stability of each of these modes in the parameter
990: space of the system (which definitely exist for lower values of the coupling).
991:
992: Finally, while the recent investigations of the various principal
993: branches have offered a rather substantial understanding of their
994: properties, there are multiple directions of future interest in this
995: topic that also appear within experimental reach. Such examples consist
996: e.g. of the study of two-dimensional waveguide arrays and
997: of discrete solitons \cite{solitons} and vortices \cite{vortices}
998: in them. Another such example is the study of more elaborate,
999: multi-site constructions, such as twisted modes \cite{twisted} even
1000: in one spatial dimension. While the above directions of examination
1001: of stationary states are of interest in their own right, an additional
1002: dimension of relevance revealed by our direct instability simulations of
1003: section V pertains to the
1004: systematic study of the existence and stability of breathing
1005: excitations due to their principal role in the system's dynamics.
1006: Such investigations are currently in progress
1007: and will be reported in future publications.
1008:
1009: \begin{thebibliography}{99}
1010: \bibitem{reviews} S. Aubry, Physica D {\bf 103},
1011: 201, (1997); S. Flach and C.R. Willis, Phys. Rep.
1012: {\bf 295} 181 (1998);
1013: D. Hennig and G. Tsironis,
1014: Phys. Rep. {\bf 307}, 333 (1999);
1015: P.G. Kevrekidis, K.O. Rasmussen, and A.R.
1016: Bishop, Int. J. Mod. Phys. B {\bf 15}, 2833 (2001).
1017:
1018: \bibitem{reviews1} D. N. Christodoulides, F. Lederer and Y. Silberberg,
1019: Nature \textbf{424}, 817 (2003); Yu. S. Kivshar and G. P. Agrawal,
1020: \textit{Optical Solitons: From Fibers to Photonic Crystals},
1021: Academic Press (San Diego,
1022: 2003).
1023:
1024:
1025: \bibitem{reviews2} P.G. Kevrekidis and D.J. Frantzeskakis,
1026: Mod. Phys. Lett. B {\bf 18}, 173 (2004).
1027: V.V. Konotop and V.A. Brazhnyi, Mod. Phys. Lett. B {\bf 18} 627, (2004);
1028: P.G. Kevrekidis {\it et al.}, Mod. Phys. Lett. B {\bf 18}, 1481 (2004).
1029:
1030: \bibitem{reviews3} M. Peyrard, Nonlinearity {\bf 17}, R1 (2004).
1031:
1032: \bibitem{b3} D.N. Christodoulides and R.I. Joseph, Opt. Lett. {\bf 13}, 794
1033: (1988).
1034:
1035:
1036: \bibitem{b7} H.S. Eisenberg, Y. Silberberg, R. Morandotti, A.R. Boyd, and
1037: J.S. Aitchison, Phys. Rev. Lett. {\bf 81}, 3383 (1998).
1038:
1039:
1040: \bibitem{b8} R. Morandotti, U. Peschel, J.S. Aitchison, H.S. Eisenberg, and
1041: Y. Silberberg, Phys. Rev. Lett. {\bf 83}, 2726 (1999).
1042:
1043:
1044: \bibitem{JOSA} H. S. Eisenberg, R. Morandotti, Y. Silberberg, J. M. Arnold,
1045: G. Pennelli, J. S. Aitchison, J. Opt. Soc. Am. B {\bf 19}, 2938 (2002).
1046:
1047: \bibitem{morsch} O. Morsch and M. Oberthaler,
1048: Bose-Einstein condensates in optical lattices,
1049: preprint (2004).
1050:
1051: \bibitem{b9} H.S. Eisenberg, Y. Silberberg, R. Morandotti, and J.S.
1052: Aitchison, Phys. Rev. Lett. {\bf 85}, 1863 (2000).
1053:
1054: \bibitem{Jena} T. Pertsch, T. Zentgraf, U. Peschel, A. Brauer
1055: and F. Lederer, Phys.
1056: Rev. Lett. {\bf 88}, 093901 (2002).
1057:
1058: \bibitem{b10} M.J. Ablowitz and Z.H. Musslimani, Phys. Rev. Lett. {\bf 87},
1059: 254102 (2001); U. Peschel and F. Lederer, J. Opt. Soc. Am. B {\bf 19}, 544
1060: (2002).
1061:
1062: \bibitem{Eugenieva} D.N. Christodoulides and E.D. Eugenieva, Phys. Rev.
1063: Lett. {\bf 87}, 233901 (2001)
1064:
1065: \bibitem{dnc2} J. Meier, G.I. Stegeman, D.N. Christodoulides,
1066: Y. Silberberg, R. Morandotti, H. Yang, G. Salamo, M. Sorel
1067: and J.S. Aitchison, Opt. Lett. {\bf 30}, 1027 (2005).
1068:
1069: \bibitem{hector} H.E. Nistazakis, P.G. Kevrekidis, B.A. Malomed,
1070: D.J. Frantzeskakis and A.R. Bishop,
1071: Phys. Rev. E {\bf 66}, 015601(R) (2002).
1072:
1073: \bibitem{bright} P.G. Kevrekidis, D.J. Frantzeskakis,
1074: R. Carretero-Gonz{\'a}lez, B.A. Malomed, G. Herring and A.R. Bishop,
1075: Phys. Rev. A {\bf 71}, 023614 (2005).
1076:
1077: \bibitem{c7} F. Lederer, S. Darmanyan and A. Kobyakov
1078: in {\it Spatial Optical Solitons}, edited by S. Trillo and
1079: W.E. Torruellas (Springer-Verlag, New York, 2001).
1080:
1081: \bibitem{c13} S. Darmanyan, A. Kobyakov, E. Schmidt and F. Lederer,
1082: Phys. Rev. E {\bf 57}, 3520 (1998).
1083:
1084: \bibitem{epjd} P.G. Kevrekidis, H.E. Nistazakis, D.J. Frantzeskakis,
1085: B.A. Malomed and R. Carretero-Gonz{\'a}lez,
1086: Eur. Phys. J. D {\bf 28}, 181 (2004).
1087:
1088: \bibitem{c18} T. Peschel, U. Peschel and F. Lederer, Phys. Rev.
1089: E {\bf 57}, 1127 (1998).
1090:
1091: \bibitem{c19} M.J. Ablowitz and Z.H. Musslimani, Phys. Rev. E
1092: {\bf 65}, 056618 (2002).
1093:
1094: \bibitem{hudock} J. Hudock, P. G. Kevrekidis, B. A. Malomed, and
1095: D. N. Christodoulides, Phys. Rev. E {\bf 67}, 056618 (2003).
1096:
1097: \bibitem{meier} J. Meier, J. Hudock, D. Christodoulides, G. Stegeman,
1098: Y. Silberberg, R. Morandotti and J.S. Aitchison,
1099: Phys. Rev. Lett. {\bf 91}, 143907 (2003).
1100:
1101: \bibitem{meier2} J. Meier, J. Hudock, D.N. Christodoulides, G.I. Stegeman,
1102: H.Y. Yang, G. Salamo, R. Morandotti, J.S. Aitchison and Y. Silberberg,
1103: J. Opt. Soc. Am. B {\bf 22}, 1432 (2005).
1104:
1105: \bibitem{molina1} R.A. Vicencio, M.I. Molina and Yu.S. Kivshar,
1106: Opt. Lett. {\bf 29}, 2905 (2004).
1107:
1108: \bibitem{molina2} R.A. Vicencio, M.I. Molina and Yu.S. Kivshar,
1109: Phys. Rev. E {\bf 71}, 056613 (2005).
1110:
1111: \bibitem{aubry} S. Aubry, Physica D {\bf 103}, 201 (1997)
1112:
1113: \bibitem{joh} M. Johansson and S. Aubry, Phys. Rev. E {\bf 61}, 5864
1114: (2000).
1115:
1116: \bibitem{kevkap} T. Kapitula, P.G. Kevrekidis and B. Sandstede,
1117: Physica D {\bf 195}, 263 (2004).
1118:
1119: \bibitem{mackay} R.S. MacKay,
1120: in: Hamiltonian Dynamical Systems,
1121: R.S. MacKay, J. Meiss (Eds.), Adam Hilger, 1987, p. 137.
1122:
1123: \bibitem{solitons} J.W. Fleischer, M. Segev, N.K. Efremidis
1124: and D.N. Christodoulides, Nature {\bf 422}, 147 (2003).
1125: J.W. Fleischer, T. Carmon, M. Segev, N.K. Efremidis and
1126: D.N. Christodoulides, Phys. Rev. Lett. {\bf 90}
1127: 023902 (2003); H. Martin, E.D. Eugenieva, Z. Chen
1128: and D.N. Christodoulides, Phys. Rev. Lett.
1129: {\bf 92} 123902 (2004).
1130:
1131: \bibitem{vortices} D.N. Neshev, T.J. Alexander, E.A. Ostrovskaya,
1132: Yu.S. Kivshar, H. Martin, I. Makasyuk and Z. Chen,
1133: Phys. Rev. Lett. {\bf 92}, 123903 (2004); J.W. Fleischer,
1134: G. Bartal, O. Cohen, O. Manela, M. Segev, J. Hudock and D.N. Christodoulides,
1135: Phys. Rev. Lett. {\bf 92} (2004) 123904.
1136:
1137: \bibitem{twisted} P.G. Kevrekidis, A.R. Bishop and K.{\O}. Rasmussen,
1138: Phys. Rev. E {\bf 63}, 036603 (2001); G.L. Alfimov, V.A. Brazhnyi and V.V.
1139: Konotop, Physica D {\bf 194}, 127 (2004);
1140: D.E. Pelinovsky, P.G. Kevrekidis and D.J. Frantzeskakis,
1141: nlin.PS/0410005.
1142:
1143:
1144: \end{thebibliography}
1145:
1146: \end{document}
1147: