nlin0507047/sol.tex
1: %\documentstyle[aps,pra,epsfig]{revtex}
2: %\documentstyle[aps,multicol,pra,epsfig]{revtex}
3: \documentstyle[aps,preprint,pra,epsfig]{revtex}
4: 
5: \begin{document}
6: 
7: \draft
8: 
9: \title{Propagation of Discrete Solitons in Inhomogeneous Networks}
10: 
11: \author{R. Burioni$^1$, D. Cassi$^1$, P. Sodano$^2$, 
12: A. Trombettoni$^1$, and A. Vezzani$^1$}
13: \address{$^1$ I.N.F.M. and Dipartimento di Fisica, Universit\`a
14: di Parma, parco Area delle Scienze 7A Parma, I-43100, Italy}
15: \address{$^2$ Dipartimento di Fisica and Sezione I.N.F.N.,
16: Universit\`a di Perugia, Via A. Pascoli Perugia, I-06123, Italy}
17: 
18: \date{\today}
19: \maketitle
20: 
21: \begin{abstract} 
22: In many physical applications solitons propagate on supports
23: whose topological properties may induce new and interesting effects. In this
24: paper, we investigate the propagation of solitons on chains with a topological 
25: inhomogeneity generated by the insertion of a 
26: finite discrete network on the chain. 
27: For networks connected by a link to a single site of the
28: chain, we derive a general criterion 
29: yielding the momenta for perfect reflection and transmission 
30: of traveling solitons and we discuss solitonic motion on chains 
31: with topological inhomogeneities. 
32: \end{abstract}
33: 
34: %\pacs{PACS: 74.25.Dw, 05.30.Jp, 74.50.+r, 85.25.Cp}
35: 
36: %\begin{multicols}{2}
37:  
38: {\bf In the last decades, a huge amount of work has been devoted 
39: to the study of the propagation of discrete solitons in regular, 
40: translational invariant lattices. 
41: However, in several systems, like networks of 
42: nonlinear waveguide arrays, 
43: Bose-Einstein condensates in optical lattices, 
44: arrays of superconducting Josephson junctions and silicon-based 
45: photonic crystals, one can engineer the shape 
46: (i.e., the topology) of the network. Correspondingly, 
47: an interesting task is
48: the study of the propagation of solitons in inhomogeneous networks. 
49: The general idea of this work 
50: is that network topology strongly affects
51: the soliton propagation. We provide a 
52: general argument giving the momenta of perfect transmission and 
53: reflection for a soliton scattering through a finite general network  
54: attached to a site of a chain: the momenta of perfect transmission and 
55: reflection are related in a simple way to the energy levels of the 
56: attached network. This criterion directly links the transmission coefficients 
57: with the network adjacency matrix, which encodes all the
58: relevant informations on its topology. Such relation puts into evidence 
59: the topological effects on the soliton propagation. 
60: The situations where finite linear chains, Cayley trees and other 
61: simple structures are attached to a site of an unbranched chain are 
62: investigated in detail.}
63: 
64: \section{Introduction}  
65: 
66: The analysis of nonlinear models on regular lattices 
67: \cite{flach98,scott99,hennig99}, 
68: as well as the investigation of linear models on 
69: inhomogeneous and fractal networks \cite{nakayama94} 
70: has attracted a great deal of attention in the last decades: 
71: while nonlinearity dramatically modifies the dynamics, allowing for 
72: soliton propagation, energy localization, and the existence of discrete
73: breathers \cite{scott99}, topology mainly affects the
74: energy spectrum giving rise to interesting phenomena such as anomalous
75: diffusion, localized states, and fracton dynamics \cite{nakayama94}. 
76: It is now both timely and highly desirable to begin a thorough 
77: investigation of nonlinear models on general inhomogeneous networks, 
78: since one expects not only interesting 
79: new phenomena arising from the interplay 
80: between nonlinearity and topology, but also an high potential impact 
81: for applications to biology \cite{peyrard89,special} and 
82: to signal propagation in optical waveguides \cite{kivshar03}. 
83: Recently, the effects of uniformity break on soliton propagation 
84: \cite{christodoulides01,kevrekidis03} 
85: and localized modes \cite{mcgurn02} has been investigated 
86: by considering $Y$-junctions  
87: \cite{christodoulides01,kevrekidis03} (consisting of a  
88: long chain inserted on a site of a chain yielding 
89: a star-like geometry) or geometries like  
90: junctions of two infinite waveguides or the waveguide coupler 
91: \cite{mcgurn02}. Here, we consider general finite 
92: networks inserted on a chain.
93: 
94: The discrete nonlinear Schr\"odinger equation (DNLSE) is a paradigmatic 
95: example of a nonlinear equation on a lattice which has been 
96: successfully applied to several contexts \cite{kevrekidis01,ablowitz04}: 
97: in particular, it 
98: has been used to describe the physics of 
99: arrays of coupled optical waveguides \cite{eisenberg98,morandotti99} 
100: and arrays of Bose-Einstein condensates \cite{trombettoni01}. 
101: It is well known that, on a homogeneous chain, 
102: the DNLSE is not integrable \cite{ablowitz04}; nevertheless,
103: soliton-like wavepackets can propagate for a long time and 
104: the stability conditions of soliton-like solutions can be derived 
105: within variational approaches \cite{malomed96}. 
106: Furthermore, the dynamics of traveling pulses has been investigated 
107: in detail in literature \cite{duncan93,flach99,gomez04} (more references 
108: are in \cite{kevrekidis01}). 
109: The simplest example of an inhomogeneous 
110: chain is provided by an external potential localized on a site of the chain 
111: \cite{forinash94,konotop96,krolikowski96,aceves96,kevrekidis01}:  
112: an experimental set up with a single defect has been recently 
113: realized with coupled optical waveguides \cite{mandelik03}. 
114: Another relevant example of an inhomogeneous chain is obtained 
115: adding an additional Fano degree of freedom coupled to 
116: the a site of the chain, which gives the so-called Fano-Anderson 
117: model (see \cite{miroshnichenko03,flach03,miroshnichenko05} 
118: and references therein).
119: 
120: As a first step in the investigation of the properties of nonlinear models 
121: on general inhomogeneous networks, we shall analyze the 
122: propagation of DNLSE solitons on a class 
123: of inhomogeneous networks built by suitably adding a 
124: finite topological inhomogeneity to an unbranched chain. 
125: The general framework where this analysis can 
126: be carried out is provided by graph theory \cite{harary69}; 
127: in particular, we shall consider networks where a finite 
128: discrete graph $G^0$ is attached by a link to a site of the 
129: homogeneous, unbranched chain (see Fig.1) 
130: while all the sites potentials $\epsilon_i$ are set to 
131: a constant. Such systems may be experimentally realized 
132: by placing the nonlinear waveguides in a suitable 
133: inhomogeneous arrangement, like the one depicted in Fig.1. We mention 
134: that in arrays of Bose-Einstein condensates one can build up geometries, 
135: which differ from the unbranched chain, by properly superimposing 
136: the laser beams creating the optical lattices \cite{oberthaler03}. 
137: In superconducting Josephson arrays \cite{fazio01}, 
138: present-day technologies allow for to prepare the insulating support for 
139: the junctions in order to create structures 
140: of the form "chain + a topological defect".
141: In the context of coupled nonlinear waveguides \cite{hennig99}, 
142: one should couple the waveguides according the geometry 
143: of the graph $G^0$, and couple this network of waveguides to a single 
144: waveguide of the unbranched chain; a similar engineering should be 
145: requested to realize photonic crystal circuits \cite{birner01} 
146: obtained merging the circuit $G^0$ to the unbranched chain.
147: 
148: As we shall discuss in Section IV, 
149: the shape of the attached graph $G^0$ affects the transmission
150: and reflection coefficients as a function of the soliton momentum: 
151: as an example of this general phenomenon, we consider 
152: unbranched chains to which simple graphs, like finite chains and Cayley trees, 
153: are added. Our analysis points to the fact that 
154: the topology of the network (i.e., how its sites are connected) controls 
155: the transmission properties, and that one can modify soliton propagation 
156: by varying the topology of the inserted network. In particular, we shall 
157: show that the momenta of perfect transmission are determined by the energy 
158: levels of the inserted graph, i.e., by the
159: eigenfrequencies of the $G^0$'s oscillation modes in the linear case 
160: (see Section IV).   
161:     
162: On a chain, stable solitonic wavepackets can propagate for long times 
163: \cite{kevrekidis01}. When a graph $G^0$ is inserted, one can study 
164: how the presence of this topological inhomogeneity modifies the soliton 
165: propagation. We numerically evaluate the transmission 
166: coefficients and we compare the numerical results 
167: with analytical findings obtained for a relevant soliton class, 
168: to which we refer as {\em large-fast solitons}. 
169: For large fast solitons the transmission coefficients
170: can be evaluated within a linear approximation \cite{miroshnichenko03}. 
171: Indeed the characteristic time for the soliton-topological defect collision
172: are very small with respect to the soliton dispersion 
173: time; therefore, the soliton scattering can be approximated by
174: the scattering of a plane-wave. 
175: However, as we numerically checked in the Figs.2-6, 
176: the nonlinearity still plays a role, giving long-lived solitons, especially 
177: near the momenta of perfect reflection or perfect transmission: 
178: it keeps the soliton shape during its propagation. 
179: Since in many experimental settings 
180: one can easily check if the reflected wavepacket 
181: is vanishing, this work could provide a basis for a topological 
182: engineering of solitonic propagation on inhomogeneous networks. 
183: 
184: The plan of the paper is the following. In Section II we review the
185: properties of the DNLSE on a chain and we introduce the variational
186: approach to investigate the soliton dynamics. 
187: In Section III, by using graph theory \cite{harary69}, we
188: define the DNLSE on inhomogeneous networks built by adding a 
189: topological perturbation to an unbranched chain; furthermore, we explain the
190: numerical techniques used in the paper for the study of the soliton
191: scattering, and we discuss the range of validity of the linear approximation
192: used in the analytical computations. In Section IV we present 
193: our analysis yielding
194: the conditions on the spectrum of the finite graph $G^0$ in 
195: order to obtain total reflection and transmission of solitons. 
196: In Section V we study the relevant case when $G^0$
197: is a finite, linear chain and we show that the Fano-Anderson model 
198: \cite{miroshnichenko03,flach03} can be realized within our approach 
199: by considering a single link attached to the unbranched chain. 
200: In Section VI we show that, for self-similar graphs $G^0$, the values 
201: of momenta for which perfect reflection occurs becomes perfect 
202: transmission momenta when the next generation of the graph is considered and 
203: we study in detail the case of Cayley trees as an 
204: example of self-similar structures. In Section VII we study the transmission 
205: coefficients for three different inserted finite graphs: 
206: loops, stars and complete graphs. Finally,  
207: Section VIII is devoted to our concluding remarks. 
208: 
209: \section{DNLSE on a chain}
210: 
211: Besides its theoretical interest, 
212: the DNLSE describes the properties of interesting systems, such as 
213: arrays of coupled optical waveguides and arrays
214: of Bose-Einstein condensates. On a chain the DNLSE reads
215: \begin{equation}
216: \label{DNLS-retta}
217: i  \frac{\partial \psi_n}{\partial \tau} = - \frac{1}{2} (\psi_{n+1} 
218: + \psi_{n-1}) + \Lambda \mid \psi_n \mid ^2 \psi_n + \epsilon_{n} \psi_n
219: \end{equation}
220: where $n$ is an integer index denoting the site position and
221: the normalization condition is $\sum_n \mid \psi_n \mid ^2=1$. 
222: In Eq.(\ref{DNLS-retta}) one has a kinetic coupling term only between 
223: nearest-neighbour sites, but the effect of next- nearest-neighbour 
224: coupling and long-term coupling has been also often considered 
225: (see the reviews \cite{hennig99,kevrekidis01} for more references): 
226: in the present paper we consider only a constant 
227: nearest-neighbour interaction, but from the next Section 
228: we allow for that the number of nearest-neighbours of a site 
229: is not constant across the network (like for the simple chain), but it can 
230: vary according the topology of the graph.
231:  
232: In condensate arrays, $\psi_n(\tau)$ is the wavefunction of the 
233: condensate  in the $n$th well. Time $\tau$ is in 
234: units of $\hbar/2K$, 
235: where $K$ is the tunneling rate between neighbouring condensates; 
236: $\epsilon_n=E_n/2K$ where $E_n$ 
237: is an external on-site field superimposed to the optical lattice and 
238: the nonlinear coefficient is $\Lambda=U/2K$, where $U$ 
239: is due to the interatomic interaction 
240: and it is proportional to the scattering length ($U$ 
241: is positive for $^{87}Rb$ atoms and is negative for $^{7}Li$ atoms). 
242: 
243: In arrays of one-dimensional coupled optical waveguides
244: \cite{eisenberg98} $\psi_n(\tau)$ is the electric 
245: field in the $n$th-waveguide at the position $\tau$ and the DNLSE
246: describes the spatial evolution of the field.
247: The parameter $\Lambda$ is proportional to the Kerr nonlinearity 
248: and the on-site potentials $\epsilon_n$ are
249: the effective refraction indices of the individual waveguides.
250: As the light propagates along the array, the coupling
251: induces an exchange of power among the single waveguides.
252: In the low power limit (i.e. when the nonlinearity is negligible),
253: the optical field spreads over the whole array.
254: Upon increasing the power, the output field narrows
255: until it is localized in a few waveguides, and discrete solitons can 
256: finally be observed \cite{eisenberg98,morandotti99}.
257: Experiments with defects (i.e., with particular 
258: waveguides different from the others) have been already reported 
259: \cite{mandelik03}.  
260: 
261: On a chain DNLSE soliton-like wavepackets 
262: can propagate for a long time even if the equation 
263: is not integrable \cite{hennig99}. 
264: Let us consider, at $\tau=0$, a gaussian wavepacket centered 
265: in $\xi(\tau=0) \equiv \xi_0$, 
266: with initial momentum $k$ and width $\gamma(\tau=0) \equiv \gamma_0$: 
267: its time  
268: dynamics are studied resorting to the Dirac time-dependent 
269: variational approach \cite{dirac30} which well 
270: reproduces the exact results in the continuum theory \cite{cooper93}. 
271: In its discrete version the wavefunction can be written 
272: as a generalized gaussian
273: \begin{equation}
274: \label{gaussol}
275: \psi_{n}(\tau)=\sqrt{{\cal K}} \, \cdot \, 
276: e^{ -\frac{(n-\xi)^2}{\gamma^2} + ik(n-\xi) +
277: i \frac{\delta}{2}(n-\xi)^2}
278: \end{equation}
279: where $\xi(\tau)$ and $\gamma(\tau)$ are, respectively,
280: the center and the width of the density $\rho_n = \mid \psi_n \mid^2$,
281: and $k(\tau)$ and $\delta(\tau)$  are the momenta conjugate to 
282: $\xi(\tau)$ and $\gamma(\tau)$ respectively; 
283: ${\cal K}$ is just a normalization factor.
284: The wave packet dynamical evolution is obtained from the Lagrangian 
285: ${\cal L}= \sum_n  i \dot{\psi}_n \psi_n^\ast - \cal{H}$, 
286: with the equations of motion 
287: for the variational parameters $\xi,\gamma,k,\delta$. 
288: In the absence of external potential ($\epsilon_n=0$), 
289: one obtains the Lagrangian \cite{trombettoni01a}
290: \begin{displaymath}
291: {\cal L}={\cal K} \sum\limits_{n=-\infty}^{\infty} 
292: e^{-(2n^2+2n-4n\xi+2\xi^2-2\xi+1)/\gamma^2} 
293: \cos{[\delta(n+1/2-\xi)+k]} - 
294: \frac{\Lambda {\cal K}^2}{2} \sum\limits_{n=-\infty}^{\infty} 
295: e^{ - 4 (n-\xi)^2 / \gamma^2}  
296: \end{displaymath}
297: \begin{equation}
298: \label{LAG-N-finito}
299: +{\cal K} \sum\limits_{n=-\infty}^{\infty} \left\{
300: -\frac{\dot{\delta}}{2} (n-\xi)^2 + \delta \dot{\xi}(n-\xi) -
301: \dot{k}(n-\xi)+k \dot{\xi} \right\} \,  
302: e^{-2(n-\xi)^2/\gamma^2}. 
303: \end{equation}
304: With $\gamma$ not too small ($\gamma \gg 1$), we can replace 
305: the sums over $n$ with integrals: to evaluate the error committed, 
306: we recall that \cite{knuth89} 
307: \begin{equation}
308: \frac{\sum\limits_{n=-\infty}^{\infty} e^{-\frac{(n-\xi)^2}
309: {\gamma^2}}}{\int\limits_{-\infty}^{\infty} dn \, \, \, 
310: e^{-\frac{(n-\xi)^2}{\gamma^2}} } 
311: = 1+ O(e^{- \pi^2 \gamma^2}).
312: \label{knuth}
313: \end{equation}
314: In this limit the normalization factor becomes 
315: ${\cal K}=\sqrt{2/\pi\gamma^2}$. 
316: We finally get \cite{trombettoni01} 
317: \begin{equation} 
318: \label{LAG}
319: {\cal L}=k \dot{\xi} - \frac{\gamma^2 \dot{\delta}}{8} - 
320: \frac{\Lambda}{2 \sqrt{\pi \gamma^2} } + \cos{k} \, \cdot \, 
321: e^{-\eta}, 
322: \end{equation}
323: where $\eta = 1 / 2 \gamma^2 + \gamma^2 \delta^2 / 8$.
324: The equations of motion are
325: \begin{eqnarray}
326: \dot{k} & = & 0 \label{var1} \\
327: \dot{\xi} & = & \sin{k} \, \cdot \, e^{-\eta} \label{var2} \\ 
328: \dot{\delta} & = & \cos{k} \Big(4 / \gamma^4-\delta^2 \Big) 
329: e^{-\eta} + 2 \Lambda / \sqrt{\pi} \gamma^3 \label{var3} \\ 
330: \dot{\gamma} & = & \gamma \delta \cos{k} \, \cdot \, 
331: e^{-\eta}: \label{var4}
332: \end{eqnarray} 
333: $k(\tau)=k$ is conserved. 
334: Notice that, due to the discreteness, the group velocity 
335: cannot be arbitrarily large ($\dot{\xi} \approx \sin{k} \le 1$). 
336: As Eq.(\ref{knuth}) clearly shows, the variational equations 
337: of motions (\ref{var4}) are meaningful only for large solitons: 
338: the Peierls-Nabarro potential does not appear in (\ref{var4}), 
339: and the equations feature momentum conservation. We mention 
340: that in uniform DNLSE chains a threshold condition for 
341: the soliton propagation appears: only if the soliton is
342: sufficiently broad solitons may freely move \cite{papa03}. In the following 
343: we shall consider only large-fast solitons, so that the variational 
344: equations of motions (\ref{var4}) are appropriate; however, to study 
345: the propagation of localized discrete breathers in inhomogeneous networks 
346: one should study the Lagrangian (\ref{LAG-N-finito}).
347: 
348: When $\dot{\gamma}=0$ and $\dot{\delta}=0$, the shape of the wavefunction
349: does not vary and one has a variational soliton-like solution 
350: where the center of mass move with a constant velocity $\dot{\xi}= constant$. 
351: If $\Lambda>0$ 
352: the conditions $\dot{\gamma}=0$ and $\dot{\delta}=0$ can be satisfied
353: only if $\cos{k}<0$ (i.e., only when the effective 
354: mass is negative): for this reason in the following 
355: we take only momenta $\pi/2 \le k \le \pi$ (positive velocities) 
356: or $-\pi \le k \le - \pi/2$ (negative velocities). In particular, 
357: for $\delta(\tau=0) \equiv \delta_0=0$ and large enough solitons 
358: ($\gamma_0 \gg 1$), the condition on
359: $\Lambda$ allowing for a soliton solution is \cite{trombettoni01}
360: \begin{equation}
361: \label{lambda_sol}
362: \Lambda_{sol} \approx  2 \sqrt{\pi} \frac{\mid \cos{k} \mid} 
363: {\gamma_0}.
364: \end{equation}
365: 
366: The stability of variational solutions has been 
367: numerically checked showing that the shape of the solitons
368: is preserved for long times.
369: In the following, we use the term ``solitons'' to name the solutions of the 
370: variational equations (\ref{var1})-(\ref{var4}). 
371: One can have a similar criterion using other 
372: variational approaches \cite{malomed96}. One also expects that 
373: the integrable version of the DNLSE, the so-called Ablowitz-Ladik 
374: equation \cite{ablowitz76,ablowitz04}, 
375: provides results very similar to those obtained in this paper. 
376: %In the following, for a soliton solution, we will omit 
377: %the subscript $_0$ in $k$ and $\gamma_0$. 
378: 
379: \section{DNLSE on graphs}
380: 
381: The DNLSE (\ref{DNLS-retta}) can be generalized to a general discrete network
382: by means of graph theory. A graph $G$ is given by a set of sites $i$ 
383: connected pairwise by set of unoriented links $(i,j)$
384: defining a neighbouring relation between the sites. The topology 
385: of a graph is 
386: described by its adjacency matrix $A_{i,j}$
387: which is defined to be $1$ if $i$ and $j$ are nearest-neighbours, and $0$
388: otherwise. The DNLSE on a graph reads as 
389: \begin{equation} 
390: \label{DNLS-gen} 
391: i \frac{\partial \psi_i}{\partial \tau} = - \frac{1}{2} \sum_{j}  A_{i,j}
392: \psi_{j}+ \Lambda \mid \psi_i \mid ^2 \psi_i + \epsilon_{i} \psi_i
393: \end{equation}   
394: 
395: Equation (\ref{DNLS-gen}) describes the wavefunction dynamics in a wide range 
396: of discrete physical systems, and it can be applied to regular
397: lattices as well as to inhomogeneous networks such as fractals, 
398: complex biological structures and glasses. The properties of 
399: Eq.(\ref{DNLS-gen}) on small graphs has been investigated in 
400: \cite{eilbeck85}. 
401: The first term in the right-hand side of Eq.(\ref{DNLS-gen}) represents 
402: the hopping between nearest-neighbours (with tunneling rate proportional 
403: to $A_{i,j}$), the second is the nonlinear term, and the third one 
404: describes superimposed external potentials. 
405: We remark that in Eq.(\ref{DNLS-gen}) the numbers of nearest neighbours 
406: is site-dependent. Furthermore, since $A_{i,j}=1$ if $i$ and $j$ 
407: are nearest-neighbours, one is assuming that the tunneling rate 
408: between neighbouring sites entering Eq.(\ref{DNLS-gen}) 
409: is constant across the array and
410: it is (in the chosen units) equal to $1$. Below, we shall consider also 
411: the case of a tunneling rate which is not constant across the array. 
412: 
413: We shall focus on the situation where the graph $G$ is obtained 
414: by attaching a finite graph $G^0$ to a single site of the 
415: unbranched chain (see Fig.1) and setting
416: $\epsilon_i=0$ for all the sites. We denote the sites of the unbranched 
417: chain and of the
418: graph $G^0$ with latin indices $m,n,\dots$ and greek indices 
419: $\alpha, \beta, \cdots$ respectively. 
420: A single link connects the site $n=0$ of the chain with
421: the site  $\alpha$  of the graph $G^0$.
422: 
423: The scattering of a soliton through the topological perturbation 
424: can be numerically studied as follows. 
425: At $\tau=0$ (hereafter, we refer to $\tau$ as a time
426: even if for the optical waveguides it represents a spatial variable) one 
427: prepares a gaussian soliton (\ref{gaussol}) centered well to the left of $0$ 
428: (i.e., $\xi_0<0$) moving
429: towards $n=0$ ($\sin(k)>0$) with a width related to the  nonlinear 
430: coefficient according to Eq.(\ref{lambda_sol}). From Eq.(\ref{DNLS-gen}) 
431: the time evolution of the wavefunction may be numerically evaluated: 
432: when $\tau_s \approx \xi_0 /\sin(k)$ the soliton scatters through the 
433: finite graph $G^0$
434: ($\sin(k)$ being the group velocity of the soliton). 
435: At a time $\tau$ well after
436: the soliton scattering (i.e. $\tau \gg \tau_s$), 
437: the reflection and transmission coefficients ${\cal R}$
438: and ${\cal T}$ are given by
439: \begin{equation} 
440: {\cal R}=\sum_{n<0} \mid \psi_n(\tau)\mid^2 \label{R}
441: \end{equation} 
442: \begin{equation} 
443: {\cal T}=\sum_{n>0} \mid \psi_n(\tau)\mid^2.
444: \label{T} 
445: \end{equation} 
446: 
447: Note that, while in the linear case ($\Lambda=0$) one has  ${\cal
448: R}+{\cal T}=1$, in general, nonlinearity violates unitarity by allowing for 
449: phenomena such as soliton trapping; nevertheless, there are regimes 
450: where soliton trapping is negligible and ${\cal R}+{\cal T} \approx 1$.
451: We numerically checked that in the time dynamics reported 
452: in the paper this condition is well satisfied.
453: Situations corresponding to resonant scattering (i.e. 
454: ${\cal R}=0$ or ${\cal T}=0$) have a particular relevance: in fact, 
455: these situations can be easily experimentally detected, and the
456: soliton-like solution is stable also well after the scattering, 
457: as it is numerically verified 
458: in different examples of resonant reflection and transmission.
459: 
460: For an important class of soliton solutions (to which we 
461: refer as large-fast solitons) the scattering through a 
462: topological inhomogeneity can be analytically studied using a 
463: linear approximation.
464: The interaction between the soliton and the topological 
465: inhomogeneity is characterized by two time-scales: 
466: the time of the soliton-defect interaction  
467: $\tau_{int}=\gamma/\sin{k}$ and the soliton dispersion 
468: time (i.e. the time scale in which 
469: the wavepacket will spread in absence of interaction) $\tau_{disp}= 
470: \gamma/ (4 \sin{(1/2\gamma) \cos{k})}$ \cite{miroshnichenko03}. 
471: For {\em large} ($\gamma \gg 1$, as in many 
472: relevant experimental settings) and {\em fast} solitons, i.e.
473: \begin{equation}
474: \label{linearaprox}
475: v=\sin{k} \gg (2/ \gamma) \cos{k},
476: \end{equation}
477: one has that the soliton 
478: may be considered as a set of non interacting plane waves while 
479: experiencing scattering on the graph; thus the 
480: soliton transmission may be studied by considering, in the linear regime 
481: (i.e., $\Lambda=0$), the transport coefficients of a plane wave across the 
482: topological defect. The use of the linear 
483: approximation for the analysis of the interaction of a 
484: fast soliton with a local defect in the continuous 
485: nonlinear Schr\"odinger equation is reported in \cite{cao95}. 
486: Later, we shall compare the analytical findings with a numerical solution 
487: of Eq.(\ref{DNLS-gen}), namely with the reflection and 
488: transmission coefficients ${\cal R}$ and ${\cal T}$ given by 
489: Eqs.(\ref{R})-(\ref{T}).
490: 
491: \section{A general argument for resonant transmission}
492: 
493: In this Section we show that, if a large-fast soliton scatters through 
494: a topological perturbation of an unbranched chain, the soliton momenta 
495: for perfect reflection and transmission are completely determined 
496: by the spectral properties of the attached graph $G^0$: 
497: in particular one has ${\cal R}=1$ if 
498: $2 \cdot \cos{k}$ coincides with an energy level of $G^0$, while
499: ${\cal T}=1$ if  $2 \cdot \cos{k}$ is an energy level 
500: of the reduced graph $G^r$, i.e., of the graph obtained from $G^0$ by 
501: cutting the site $\alpha$ from $G^0$ (see Fig.1).
502: In algebraic graph theory (see e.g. \cite{harary69,biggs74})
503: the energy level of a graph is simply defined as an eigenvalue of its
504: adjacency matrix. We will call $A^0$ and $A^r$ the adjacency matrices of
505: $G^0$ and $G^r$, respectively.
506: 
507: %More formally, the adjacency matrix ${ A }^0_{\eta,\eta'}$ 
508: %of the attached graph $G^0$ is 
509: %${ A }^0_{\eta,\eta'}=1$ when the sites $\eta$ and $\eta'$ 
510: %are nearest-neighbours of $G^0$ 
511: %and $0$ otherwise, and similarly for 
512: %${ A }_{\eta,\eta'}^r$ 
513: %(we recall that instead $A_{i,j}$ is the adjacency matrix of the whole 
514: %network). 
515: 
516: For large-fast solitons, 
517: the pertinent eigenvalue equation to investigate is   
518: \begin{equation}
519: \label{DLS}
520: -\frac{1}{2} \sum_{j} A_{i,j} \psi_j=\mu \psi_i
521: \end{equation}
522: (here $i$ is a generic site of the network). The solution 
523: corresponding to a plane wave 
524: coming from the left of the chain is $\psi_n=a e^{ikn}+be^{-ikn}$ 
525: for $n<0$ and $\psi_n=c e^{ikn}$ for $n>0$, so that 
526: $\mu=-\cos{k}$. The reflection coefficient is given by 
527: ${\cal R}=\mid b / a \mid ^2$ and the transmission coefficient 
528: by ${\cal T}=\mid c / a \mid ^2$. 
529: The continuity at $0$ requires $a+b=c$. The equation in $0$ is
530: \begin{equation}
531: -\frac{1}{2} (a e^{-ik}+b e^{ik}+c e^{ik}+\psi_{\alpha})=-\cos{k} 
532: \cdot (a+b)
533: \label{eq_0}
534: \end{equation}
535: while in $\alpha$ one has
536: \begin{equation}
537: -\frac{1}{2} (a+b+\sum_{\eta \in G^0} A^0_{\alpha,\eta} \psi_\eta)=
538: -\cos{k} \cdot \psi_{\alpha}. 
539: \label{eq_alpha}
540: \end{equation}
541: At the sites $\eta$ of $G^r$ one obtains 
542: \begin{equation}
543: -\frac{1}{2} \sum_{\eta' \in G^0} A^0_{\eta,\eta'} \psi_{\eta'}=-\cos{k} 
544: \cdot \psi_\eta.
545: \label{eq_altri}
546: \end{equation}
547:  
548: For perfect reflection, 
549: i.e. for the momenta $k$'s such that ${\cal R}(k)=0$, 
550: one has $c=0$, $a=-b$ and from Eq.(\ref{eq_0}) $\psi_\alpha=-2a \sin{k}$.
551: Therefore  Eqs.(\ref{eq_alpha}) and (\ref{eq_altri}) reduce to
552: the eigenvalue equation for the adjacency matrix $A^0$ and,
553: apart from the trivial case $\cos(k)=0$, they are 
554: satisfied only if $2\cos{k}$ coincides with an eigenvalue of ${ A^0}$. 
555: 
556: At variance, in order to find the momenta $k$'s such that ${\cal T}(k)=0$ 
557: (perfect transmission), one has $b=0$, $a=c$ and 
558: from Eq.(\ref{eq_0}) $\psi_\alpha=0$. Therefore, Eqs.(\ref{eq_altri}) 
559: reduces to the eigenvalue equation for $A^r$ and it is satisfied only if 
560: $2\cos{k}$ coincides with an eigenvalue of $A^r$.
561: 
562: This general argument can be easily extended to the situation 
563: where $p$ identical
564: graphs $G^0$ are attached to $n=0$: indeed, now one has only to replace 
565: in Eq. (\ref{eq_0}) $\psi_{\alpha}$ with $p\psi_{\alpha}$ and the conditions
566: for ${\cal T}(k)=0$ and ${\cal R}(k)=0$ do not change.
567: 
568: The stated result holds for the case where the tunneling rates 
569: between the neighbour sites of $G^0$ are constant (and equal to 1 in the 
570: chosen units): 
571: however one can also consider in $G^0$ non-uniform tunneling rates  
572: $t^0_{\eta,\eta'}>0$, where 
573: $\eta$ and $\eta'$ are nearest-neighbour sites belonging to $G^0$ 
574: ($t_{\eta,\eta'}^0=0$ if $A^0_{\eta,\eta'}=0$). The DNLSE  
575: at a site $\eta$ of $G^0$ becomes 
576: \begin{equation}
577: i \frac{\partial \psi_{\eta}}{\partial \tau}= - 
578: \frac{1}{2} \sum_{\eta'} t_{\eta,\eta'}^0 \psi_{\eta'}+\Lambda 
579: \mid \psi_\eta \mid^2 \psi_\eta
580: \end{equation} 
581: while the DNLSE at the sites $n$ of the of the chain remain unchanged. 
582: Now, the criterion states that ${\cal R}=1$ if 
583: $2\cos{k}$ coincides with an eigenvalue of the matrix $t_{\eta,\eta'}^0$, 
584: while ${\cal T}=1$ if $2\cos{k}$ is an eigenvalue of 
585: the matrix $t_{\eta,\eta'}^r$ defined as $t_{\eta,\eta'}^r=t_{\eta,\eta'}^0$ 
586: if $\eta$ and $\eta'$ belong to the reduced graph $G^r$.   
587: 
588: \section{Finite Linear Chains}
589: 
590: As a first simple application of the argument given in Section IV,  
591: we consider a single site $\alpha$ attached via a single link 
592: to the site $0$ of the unbranched chain. For Bose-Einstein 
593: condensates in optical lattices \cite{oberthaler03} 
594: the setup 
595: "infinite chain + single link" or "infinite chain + finite chain" 
596: may be realized by using two pairs of 
597: counterpropagating laser beams to create a star-shaped geometry 
598: in the $x$-$y$ plan \cite{brunelli04} and manipulating the frequencies of 
599: the superimposed harmonic magnetic potential 
600: so that in the $y$ direction only few sites 
601: can be occupied. In an optic context, 
602: chains of coupled waveguides are routinely built and studied 
603: \cite{kivshar03}: one
604: can obtain the configuration "infinite chain + single link" by
605: coupling a further waveguide to a waveguide of the chain. 
606: 
607: The DNLSE in the sites $0$ and $\alpha$ 
608: reads 
609: \begin{equation}
610: i \frac{\partial \psi_0}{\partial \tau}=-\frac{1}{2} (\psi_1+\psi_{-1}+
611: \psi_\alpha)+\Lambda \mid \psi_0 \mid^2 \psi_0 
612: \label{singolo_link_0}
613: \end{equation}
614: and 
615: \begin{equation}
616: i \frac{\partial \psi_\alpha}{\partial \tau}=-\frac{1}{2} \psi_0+
617: \Lambda \mid \psi_\alpha \mid^2 \psi_\alpha.
618: \label{singolo_link_alpha}
619: \end{equation}
620: It is transparent from Eqs.(\ref{singolo_link_0})-(\ref{singolo_link_alpha}) 
621: that the wavefunction $\psi_\alpha(\tau)$ may be interpreted as an 
622: additional local Fano degree of freedom, 
623: yielding the so-called Fano-Anderson 
624: model \cite{miroshnichenko03,flach03,miroshnichenko05}.  
625: In our approach, such degree of freedom is interpreted 
626: as a single link attached to the unbranched chain. As it is well known, 
627: the Fano-Anderson model describes interesting scattering properties: 
628: adding a generic finite graph (instead of a single link) 
629: gives rise to a yet richer variety of behaviors. We mention that 
630: in \cite{miroshnichenko05} the Fano degree of freedom is coupled 
631: to several sites of the unbranched chain: this would correspond in our 
632: description to a site linked to several sites of the chain, and, 
633: in general, to graphs attached to several sites of the chain. 
634: For simplicity, in the following we limit ourself to graphs inserted 
635: in a single site of the unbranched chain.
636:  
637: 
638: For large-fast solitons, when a single link is added to the unbranched 
639: chain, the 
640: reflection coefficient ${\cal R}$ from 
641: Eqs.(\ref{singolo_link_0})-(\ref{singolo_link_alpha}) 
642: is found to be \cite{miroshnichenko03}
643: \begin{equation}
644: {\cal R}=\frac{1}{1+4 \sin^2{(2k)}}:
645: \label{R_singolo_link}
646: \end{equation}
647: in the regime where $\tau_{disp} \gg \tau_{int}$, 
648: we verified that the numerical results of the soliton scattering against 
649: the link are in agreement with Eq.(\ref{R_singolo_link}) 
650: (see Fig.2). One sees that, when $k$ is approaching $\pi$ 
651: (solitons becoming slower), the agreement becomes worse.
652: From the general results, 
653: there are no fully transmitted momenta and ${\cal R}=1$ only 
654: for $k=\pi/2$ and $k=\pi$, as one can also see by a direct inspection of 
655: (\ref{R_singolo_link}). 
656: 
657: One can also attach a finite chain of length $L$ at the site $0$. 
658: In the linear approximation
659: for large-fast solitons (i.e., Eq.(\ref{DLS})), the solution
660: corresponding to a plane wave
661: coming from the left of the unbranched chain is $\psi_n=a e^{ikn}+be^{-ikn}$
662: for $n<0$ and $\psi_n=c e^{ikn}$ for $n>0$, while in the attached chain 
663: $\psi_{\alpha}=f e^{ik \alpha}+ge^{-ik \alpha}$ 
664: ($\alpha=1,\cdots,L$ denotes the sites of the attached chain, 
665: and $\alpha=1$ is the site linked to $n=0$). 
666: Eq.(\ref{DLS}) for $n=0$, $\alpha=1$, and $\alpha=L-1$ yields 
667: respectively 
668: \begin{equation}
669: a+b=c=f+g
670: \label{links_1}
671: \end{equation}
672: \begin{equation}
673: -\frac{1}{2} (a e^{-i k} + b e^{-i k} + c e^{i k} + f e^{i k} + g^{-i k})=
674: \mu (a+b)
675: \label{links_2}
676: \end{equation}
677: and
678: \begin{equation}
679: -\frac{1}{2} (f e^{i k (L-1)} + g e^{-i k (L-1)})=
680: \mu (f e^{i k L} + g e^{-i k L})
681: \label{links_3}
682: \end{equation}
683: where $\mu=-\cos{k}$. 
684: 
685: We have five unknowns ($a$, $b$, $c$, $f$, and $g$) 
686: and four equations (\ref{links_1})-(\ref{links_2})
687: (the remaining condition being provided by the normalization). 
688: One may easily determine $b/a$, $c/a$, $f/a$, and $g/a$, getting  
689: \begin{equation}
690: \frac{b}{a}=\frac{e^{2 i k}(e^{2ikL}-1)}{1-2e^{2ik}+e^{2ik(L+2)}},
691: \end{equation}
692: which leads to
693: \begin{equation}
694: {\cal R}=\mid b/a \mid^2 = 
695: \frac{\sin^2{(kL)}}{[\cos{(kL)}-\cos{(k(L+2))}]^2+\sin^2{(kL)}}.
696: \label{R_L_links}
697: \end{equation}
698: and ${\cal T}=\mid c/a \mid^2 = 1 - {\cal R}$. 
699: For $L=1$, Eq.(\ref{R_L_links}) reduces to Eq.(\ref{R_singolo_link}). 
700: In agreement with the general argument of Section IV,
701: the number of minima and maxima increases with $L$.
702: Eq.(\ref{R_L_links}) for $L=2$ is compared in Fig.2 with the numerical 
703: results. 
704: 
705: We notice that in the limit $L \to \infty$ the considered problem 
706: corresponds to the propagation of a soliton in a the so-called 
707: {\em star graph}, which has been 
708: recently investigated in the context of two-dimensional networks of nonlinear 
709: waveguide arrays \cite{christodoulides01} and 
710: $Y$-junctions for matter waves \cite{kevrekidis03}.
711: 
712: \section{Cayley trees}
713:  
714: Eq.(\ref{R_L_links}) yields 
715: that the values of $k$ allowing for perfect 
716: reflection (${\cal R}=1$) for the length $L$ coincide with the  momenta of full
717: transmission (${\cal T}=1$) when $G^0$ is a chain of length $L+1$. This
718: property is  readily understood: in fact, if $G^0$ is
719: a chain of length $L$ the $k$'s for which ${\cal R}=1$ correspond to
720: the energy levels of $G^0$. If $G^0$ is a chain of length $L+1$, the values of
721: $k$'s for which ${\cal T}=1$ correspond to energy levels  of the reduced graph
722: $G^r$, which, in this case, is again given by a chain of length $L$.  This is
723: clearly  a general property of any self-similar graph. As an example, we study
724: in this section the situation  in which $G^0$ is a Cayley tree of branching
725: rate $p$ and generation $L$ (see Fig.3). 
726: 
727: Let us consider the linear 
728: approximation  for large-fast solitons  Eq.(\ref{DLS}). 
729: The plane wave coming from the left of the unbranched chain is 
730: $\psi_n=a e^{ikn}+be^{-ikn}$ 
731: for $n<0$ and $\psi_n=c e^{ikn}$ for $n>0$. This fixes
732: $\mu=-\cos{k}$. Furthermore, from the continuity in $0$ it follows $a+b=c$.
733: 
734: For a Cayley tree, the eigenfunction must have, 
735: by symmetry, the same value at all the sites 
736: belonging to the same generation. If we denote by $\psi_\beta$ 
737: the eigenfunction at the sites at distance 
738: $\beta=1,\cdots,L$ from $n=0$, 
739: the eigenvalue equation (\ref{DLS}) at the site $\alpha=2,\cdots,L-1$ reads
740: \begin{equation}
741: -\frac{1}{2}(\psi_{\alpha-1}+p\psi_{\alpha+1})=\mu \psi_{\alpha}.
742: \label{eig_eq_Cayley}
743: \end{equation}
744: The plane wave solutions of Eq.(\ref{eig_eq_Cayley}) can be written as 
745: \begin{equation}
746: \psi_{\alpha}=\frac{1}{p^{\alpha/2}}(f e^{ik' \alpha}+ge^{-ik' \alpha}),
747: \label{cayley}
748: \end{equation}
749: where $\alpha=1,\cdots,L$: in this way 
750: one gets from Eq.(\ref{eig_eq_Cayley}) 
751: $\mu=-\sqrt{p} \cos{k'}$, so that 
752: $k'=\arccos{(p^{-1/2} \cos{k})}$. 
753: Eq.(\ref{DLS}) for $n=0$, $\alpha=1$, and $\alpha=L-1$ gives respectively  
754: \begin{equation}
755: -\frac{1}{2} (a e^{-i k} + b e^{i k} + c e^{i k})-\frac{1}{2 \sqrt{p}}
756: (f e^{i k'} + g e^{-i k'})=\mu(a+b)
757: \label{cayley_1}
758: \end{equation}
759: \begin{equation}
760: -\frac{1}{2} (a + b)-\frac{1}{2}
761: (f e^{2 i k'} + g e^{-2 i k'})=\frac{\mu}{\sqrt{p}} 
762: (f e^{i k'} + g e^{-i k'})
763: \label{cayley_2}
764: \end{equation}
765: and 
766: \begin{equation}
767: -\frac{1}{2} (f e^{i k' (L-1)} + g e^{-i k' (L-1)})=
768: \frac{\mu}{\sqrt{p}} (f e^{i k' L} + g e^{-i k' L}).
769: \label{cayley_3}
770: \end{equation}
771: Using Eqs.(\ref{cayley_1})-(\ref{cayley_3}) and the condition $a+b=c$, 
772: one can determine $b/a$, $c/a$, $f/a$, and $g/a$: the resulting 
773: expressions is rather involved and here we will not explicitly write them. 
774: In Fig.3 we plot the numerical and analytical 
775: results for the reflection coefficient ${\cal R}$ when two Cayley trees, 
776: respectively with $L=5$ and $L=6$, are attached 
777: to the unbranched chain. One sees that when one pass to the next 
778: generation, the momenta for which full reflection occurs 
779: becomes momenta of full transmission.
780: 
781: \section{Further Examples of inserted graphs}
782: 
783: In general, it is possible to consider the scattering of a large-fast soliton 
784: through a large variety of inhomogeneous networks. Here,
785: we analyze three further examples of network topologies: 
786: stars, loops and complete graphs. In each case, 
787: the reflection coefficients for large-fast solitons are
788: derived with a procedure analogous to the one adopted
789: in the previous section for Cayley trees.
790: The analytical findings are compared with numerical results.
791: 
792: {\em Loops:} Let us consider a loop graph, i.e., a finite chain  
793: of $L$ sites $\alpha_1, \cdots, \alpha_L$ such that 
794: the sites $\alpha_1$ and $\alpha_L$ are linked to the site 
795: $n=0$ of the unbranched chain. For large-fast solitons,
796: the reflection coefficient ${\cal R}$ in the linear approximated regime is
797: \begin{equation}
798: {\cal R}=2 \, \, \, \frac{\left[ 1+\cos{(k(L-1))}\right]^2 +
799: \sin^2{(k(L-1))}}{6-2 \cos{(2k)}+5 \cos{k} \cos{(kL)}-\cos{(k(L+3))} 
800: +\sin{k} \sin{(kL)}}.
801: \label{R_L_loop}
802: \end{equation}
803: We note that the number of momenta of perfect reflection 
804: and perfect transmission increases with $L$. 
805: Furthermore, at large $L$, the transmission properties 
806: of the loops become similar to those of a finite (not closed) 
807: chain [see Eq.(\ref{R_L_links})]. 
808: In Fig.4 the numerical and analytical results are compared
809: and a figure of the loop graph with $L=3$ is provided.  
810: 
811: {\em Stars:} The $p$-star 
812: is the graph composed by a central site linked to $p$ 
813: sites, which are in turn connected only to the 
814: central site. Let us consider the case where $G^0$ is a $p$-star
815: graph and $\alpha$ is the center, linked to $n=0$. 
816: In the linear approximation we have
817: \begin{equation}
818: {\cal R}=\frac{1}{1+\left( \frac{\sin{(3k)}}{\cos{k}}-(p-1) \tan{k}\right)^2} 
819: \label{R_L_star}.
820: \end{equation}
821: In Fig.5 the numerical and analytical results are compared. 
822: Eq.(\ref{R_L_star}) shows that perfect transmission (${\cal R}=0$) 
823: is obtained only for for the momentum $k=\pi/2$. 
824: This can be directly proved
825: applying the criterion of Section IV: indeed 
826: $G^0$ is a $p$-star, while $G^r$ 
827: consists of $p$ disconnected sites. 
828: Therefore, all the eigenvalues of the adjacency matrix $A^r$ equal
829: zero, and the only momentum of perfect transmission is $k=\pi / 2$. 
830: 
831: {\em Complete graphs:} The complete graph $K_M$ of $M$ sites 
832: is the graph where every pair of sites in linked \cite{harary69}: e.g, 
833: $K_3$ is a triangle.
834: Inserting $K_{L+1}$ at the site $n=0$ of the unbranched chain 
835: (so that $n=0$ is one of the sites of $K_{L+1}$), one gets:
836: \begin{equation}
837: {\cal R}=\frac{1}{1+4 \frac{(L-1-2 \cos{k})^2}{L^2} \sin^2{k}}. 
838: \label{R_L_simpl}
839: \end{equation}
840: For $L >> 1$, ${\cal R} \approx 
841: 1/ (1+4 \sin^2{k})$, therefore 
842: the complete graph behaves as a single effective 
843: defect [compare with Eq.(\ref{R_singolo_link})]. The comparison between the 
844: numerical and analytical results is presented in Fig.6.   
845: 
846: \section{Conclusions} 
847: 
848: As a first step in addressing the issue of the interplay 
849: between nonlinearity and topology, we studied the discrete nonlinear 
850: Schr\"odinger equation on a network 
851: built by attaching to a site of an unbranched chain a topological
852: perturbation $G^0$. 
853: The relevant situation corresponding to the Fano-Anderson model 
854: is obtained when one considers a single link attached to the linear chain. 
855: We showed that, by properly selecting the attached graph, 
856: one is able to control the perfect 
857: reflection and transmission of traveling solitons. 
858: We derived a general criterion 
859: yielding - once the energy levels of the graph $G^0$ is known - 
860: the momenta at which the soliton is fully reflected or fully transmitted. 
861: For self-similar graphs $G^0$, we found that the values 
862: of momenta for which perfect reflection occurs become perfect 
863: transmission momenta when the next generation of the graph is considered. 
864: For finite linear chains, loops, stars and complete graphs,  
865: we studied the transmission coefficients and we compared 
866: numerical results form the discrete nonlinear 
867: Schr\"odinger equation with analytical estimates.  
868: Our results evidence the remarkable influence of topology 
869: on nonlinear dynamics and are amenable to interesting applications 
870: in optics since one may think of engineering inhomogeneous chains acting 
871: as a filter for the motion of soliton \cite{burioni05}.
872: 
873: {\em Acknowledgments:} We thank M. J. Ablowitz, 
874: P. G. Kevrekidis and B. A. Malomed for discussions. 
875: 
876: \begin{thebibliography}{99}
877: 
878: %\bibitem{davydov82} A. S. Davydov, {\em Biology and Quantum Mechanics}, 
879: %Oxford, Pergamon Press, 1982.
880: 
881: \bibitem{flach98} 
882: S. Flach and C.R. Willis, Phys. Rep. {\bf 295}, 181 (1998). 
883: 
884: \bibitem{scott99} 
885: A. C. Scott, {\em Nonlinear Science: Emergence and
886: Dynamics of Coherent Structures}, Oxford University Press (1999).
887: 
888: \bibitem{hennig99} 
889: D. Hennig and G. P. Tsironis, Phys. Rep. {\bf 307}, 333 (1999). 
890: 
891: %\bibitem{alexander82} S. Alexander and R. Orbach, J. Phys. (Paris) 
892: %Lett. {\bf 43}, L625 (1982).
893: 
894: \bibitem{nakayama94} 
895: T. Nakayama, K. Yakubo, and R. L. Orbach, 
896: Rev. Mod. Phys. {\bf 66}, 381 (1994).
897: 
898: \bibitem{peyrard89} M. Peyrard and A. R. Bishop, 
899: Phys. Rev. Lett. {\bf 62}, 2755 (1989).
900: 
901: \bibitem{special} 
902: {\em Proceedings of the Conference on 
903: Future Directions of Nonlinear Dynamics in Physical and 
904: Biological Systems, Denmark} 
905: edited by P. L. Christiansen, J. C. Eilbeck, and R. D. Parmentier 
906: [Physica D {\bf 68}, 1-186 (1993)].
907: 
908: \bibitem{kivshar03} Y. S. Kivshar and G. P. Agrawal, {\em Optical Solitons: 
909: from Fibers to Photonic Crystals}, San Diego, Academic Press (2003).
910: 
911: \bibitem{christodoulides01} D. N. Christodoulides and E. D. Eugenieva, 
912: Phys. Rev. Lett. {\bf 87}, 233901 (2001); Opt. Lett. {\bf 26}, 
913: 1876 (2001). 
914:  
915: \bibitem{kevrekidis03} P. G. Kevrekidis, D. J. Frantzeskakis, 
916: G. Theocharis, and I. G. Kevrekidis, Phys. Lett. A {\bf 317}, 
917: 513 (2003).
918: 
919: \bibitem{mcgurn02} A. R. McGurn, Phys. Rev. B {\bf 61}, 13235 (2000); 
920: {\em ibid.} {\bf 65}, 075406 (2002).
921: 
922: \bibitem{kevrekidis01} 
923: P. G. Kevrekidis, K. \O. Rasmussen, and A. R. Bishop, Int. J. Mod. B {\bf 15}, 
924: 2833 (2001). 
925: 
926: \bibitem{ablowitz04} 
927: M. J. Ablowitz, B. Prinari, and A. D. Trubatch, 
928: {\em Discrete and Continuous Nonlinear Schr\"odinger Systems}, 
929: University Press (2004).
930: 
931: \bibitem{eisenberg98} 
932: H. S. Eisenberg, Y. Silberberg, R. Morandotti, A. R. Boyd, and 
933: J. S. Aitchison, Phys. Rev. Lett. {\bf 81}, 3383 (1998).
934: 
935: \bibitem{morandotti99} 
936: R. Morandotti, U. Peschel, J. S. Aitchison, H. S. Eisenberg, and 
937: Y. Silberberg, Phys. Rev. Lett. {\bf 83}, 2726 (1999).
938: 
939: \bibitem{trombettoni01}
940: A. Trombettoni and A. Smerzi, Phys. Rev. Lett. {\bf 86}, 2353 (2001).
941: 
942: \bibitem{malomed96} B. A. Malomed and M. I. Weinstein. Phys. Lett. A 
943: {\bf 220}, 91 (1996).
944: 
945: \bibitem{duncan93} D. B. Duncan, J. C. Eilbeck, H. Feddersen, and 
946: J. A. D. Wattis, Physica D {\bf 68}, 1 (1993). 
947: 
948: \bibitem{flach99} 
949: S. Flach and K. Kladko, Physica D {\bf 127} 61 (1999).
950: 
951: \bibitem{gomez04}
952: J. Gomez-Gardenes, L. M. Floria, M. Peyrard, and A. R. Bishop, 
953: Chaos {\bf 14}, 1130 (2004). 
954:  
955: \bibitem{forinash94} 
956: K. Forinash, M. Peyrard and B. A. Malomed, Phys. Rev. E {\bf 49}, 3400 (1994).
957: 
958: \bibitem{konotop96}
959: V. V.  Konotop, D. Cai, M. Salerno, A. R. Bishop and 
960: N. Gr\o nbech-Jensen, 
961: Phys. Rev. E {\bf 53}, 6476 (1996).
962: 
963: \bibitem{krolikowski96}
964: W. Krolikowski, and Yu. S. Kivshar, J. Opt. Soc. Am. B {\bf 13}, 876 (1996).
965: 
966: \bibitem{aceves96}
967: A. B. Aceves, C. De Angelis, T. Peschel, R. Muschall, F. Lederer, S. Trillo,   
968: and S. Wabnitz, Phys. Rev. E {\bf 53}, 1172 (1996).
969: 
970: \bibitem{mandelik03} R. Morandotti, H. S. Eisenberg, D. Mandelik, 
971: Y. Silberberg, D. Modotto, M. Sorel, C. R. Stanley, J. S. Aitchison, 
972: Opt. Lett. {\bf 28}, 834 (2003).
973: 
974: \bibitem{miroshnichenko03} A. E. Miroshnichenko, S. Flach, and 
975: B. A. Malomed, Chaos {\bf 13}, 874 (2003). 
976: 
977: \bibitem{flach03} S. Flach, A. E. Miroshnichenko, V. Fleurov, 
978: and M. V. Fistul, Phys. Rev. Lett. {\bf 90}, 084101 (2003).
979: 
980: \bibitem{miroshnichenko05} A. E. Miroshnichenko, 
981: S. F. Mingaleev, S. Flach, and Y. S. Kivshar, 
982: Phys. Rev. E {\bf 71}, 036626 (2005). 
983: 
984: \bibitem{harary69} F. Harary, {\em Graph Theory}, Addison-Wesley, 
985: Reading (1969). 
986: 
987: \bibitem{oberthaler03} M. K. Oberthaler and T. Pfau, 
988: J. Phys.: Condens. Matter {\bf 15}, R233 (2003).
989: 
990: \bibitem{fazio01} R. Fazio and H. van der Zant,  
991: Phys. Rep. {\bf 355}, 235 (2001). 
992: 
993: \bibitem{birner01} A. Birner, R. B. Wehrspohn, 
994: U. M. G\"osele, and K. Busch, Adv. Mater. {\bf 13}, 377 (2001).
995: 
996: \bibitem{dirac30} P. A. M. Dirac, 
997: Proc. Camb. Phil. Soc. {\bf 26}, 376 (1930).
998: 
999: \bibitem{cooper93} F. Cooper, H. Shepard, C. Lucheroni, 
1000: and P. Sodano, Physica D {\bf 68}, 344 (1993).
1001: 
1002: \bibitem{trombettoni01a} A. Trombettoni and A. Smerzi, 
1003: J. Phys. B {\bf 34}, 4711 (2001).
1004: 
1005: \bibitem{knuth89} R. E. Graham, D. E. Knuth, and O. Patashnik, 
1006: {\em Concrete Mathematics: A Foundation for Computer Science}, 
1007: Addison-Wesley, 1989.
1008: 
1009: \bibitem{papa03} I. E. Papacharalampous, P. G. Kevrekidis, 
1010: B. A. Malomed, and D. J. Frantzeskakis, 
1011: Phys. Rev. E {\bf 68}, 046604 (2003).
1012: 
1013: \bibitem{ablowitz76} 
1014: M. J. Ablowitz and J. F. Ladik, J. Math. Phys. {\bf 17}, 1011 (1976).
1015: 
1016: \bibitem{eilbeck85} J. C. Eilbeck, P. S. Lomdahl, and A. C. Scott, 
1017: Physica D {\bf 16}, 318 (1985). 
1018: 
1019: \bibitem{cao95} X. D. Cao and B. A. Malomed, 
1020: Phys. Lett. A {\bf 206}, 177 (1995). 
1021: 
1022: \bibitem{biggs74} N. L. Biggs, 
1023: {\em Algebraic Graph Theory}, Cambridge University Press (1974).
1024: 
1025: \bibitem{brunelli04} I. Brunelli, G. Giusiano, F. P. Mancini, P. Sodano, and 
1026: A. Trombettoni, J. Phys. B {\bf 37}, S275 (2004). 
1027: 
1028: \bibitem{burioni05} R. Burioni, D. Cassi, 
1029: P. Sodano, A. Trombettoni, and A. Vezzani, cond-mat/0502280.
1030: 
1031: \end{thebibliography}
1032: 
1033: \begin{figure}[h]
1034: \centerline{\psfig{%bbllx=5mm,bblly=6mm,bburx=184mm,bbury=137mm,%
1035: figure=fig1.eps,width=82mm,angle=0}}
1036: \caption{Inserting a graph $G^0$ on a site of a linear chain: 
1037: the points of the chain are denoted with integers $n$ and 
1038: the point in which the graph is attached is $n=0$; 
1039: $\alpha$ is the 
1040: point of $G^0$ connected to $0$. 
1041: $G^r$ is obtained subtracting $\alpha$ from $G^0$.}
1042: \end{figure}
1043: 
1044: \begin{figure}[h]
1045: \centerline{\psfig{%bbllx=5mm,bblly=6mm,bburx=184mm,bbury=137mm,%
1046: figure=fig2.eps,width=82mm,angle=0}}
1047: \caption{Reflection coefficient $\cal R$ as a function of $k$ 
1048: (with $k$ between $\pi/2$ and $\pi$) when a chain with length $1$ 
1049: (i.e., a single link) and $2$ are attached. 
1050: Empty circles ($L=1$) and stars ($L=2$) 
1051: correspond to the numerical solution of Eq.(\ref{DNLS-gen}): 
1052: in this figure, as well in the followings, 
1053: as initial condition we choose a Gaussian with initial 
1054: width $\gamma_0=40$ and momentum $k$. 
1055: Solid lines correspond to the analytical prediction (\ref{R_L_links}).}
1056: \end{figure}
1057: 
1058: \begin{figure}[h]
1059: \centerline{\psfig{%bbllx=5mm,bblly=6mm,bburx=184mm,bbury=137mm,%
1060: figure=fig3.eps,width=82mm,angle=0}}
1061: \caption{Reflection coefficient $\cal R$ as a function of $k$ 
1062: when a Cayley tree with length $5$ 
1063: and $6$ are attached. 
1064: Empty circles ($L=6$) and stars ($L=5$) 
1065: correspond to the numerical solution of Eq.(\ref{DNLS-gen}). 
1066: Solid lines correspond to the analytical prediction (see text). 
1067: As required from the general argument in Section III, the values of $k$ 
1068: for which one has perfect reflection (${\cal R}(k)=1$) for $L=5$ correspond 
1069: to perfect transmission (${\cal R}(k)=0$) for $L=6$.}
1070: \end{figure}
1071: 
1072: \begin{figure}[h]
1073: \centerline{\psfig{%bbllx=5mm,bblly=6mm,bburx=184mm,bbury=137mm,%
1074: figure=fig4.eps,width=82mm}}
1075: \caption{Reflection coefficient $\cal R$ as a function of $k$ 
1076: when loops with $L=4$, $7$ and $10$ are inserted at a site 
1077: of the unbranched chain.  
1078: Empty circles ($L=4$), squares ($L=7$) and diamonds ($L=10$) are obtained from 
1079: the numerical solution of Eq.(\ref{DNLS-gen}). 
1080: Solid lines correspond to the analytical prediction (\ref{R_L_loop}).
1081: The small figure represents a loop with 
1082: $L=3$ inserted in the unbranched chain.}
1083: \end{figure}
1084: 
1085: \begin{figure}[h]
1086: \centerline{\psfig{%bbllx=5mm,bblly=6mm,bburx=184mm,bbury=137mm,%
1087: figure=fig5.eps,width=82mm}}
1088: \caption{Reflection coefficient $\cal R$ as a function of $k$ 
1089: when stars with $p=3$, $4$ and $6$ are inserted.
1090: Empty circles ($p=2$), squares ($p=3$) and diamonds ($p=5$) are obtained from 
1091: the numerical solution of Eq.(\ref{DNLS-gen}). 
1092: Solid lines correspond to the analytical prediction (\ref{R_L_star}).
1093: The inset represents the situation where the attached graph $G^0$ is
1094: a star with $p=3$.}
1095: \end{figure}
1096: 
1097: \begin{figure}[h]
1098: \centerline{\psfig{%bbllx=5mm,bblly=6mm,bburx=184mm,bbury=137mm,%
1099: figure=fig6.eps,width=82mm}}
1100: \caption{Reflection coefficient $\cal R$ as a function of $k$ 
1101: when the complete graphs $K_3$ and $K_{11}$ are inserted at a site 
1102: of an unbranched chain. Empty circles ($K_3$) and 
1103: squares ($K_{11}$) are obtained from 
1104: the numerical solution of Eq.(\ref{DNLS-gen}). 
1105: Solid lines correspond to the analytical prediction (\ref{R_L_simpl}).  }
1106: \end{figure}
1107: 
1108: %\end{multicols}
1109: 
1110: \end{document}
1111: 
1112: 
1113: