1: %\renewcommand{\theequation}{\thesection.\arabic{equation}}

2: %\usepackage{amsmath}

3: %\setcounter{MaxMatrixCols}{10}

4: 

5: 

6: \documentclass[pra,final,titlepage,showpacs]{revtex4}

7: \usepackage{amsfonts}

8: \usepackage{graphicx}

9: \usepackage{amssymb}

10: %%\usepackage{latexsym}

11: %%\usepackage{cite}

12: 

13: \newcommand{\eq}[1]{(\ref{#1})}

14: \newcommand{\eps}{\varepsilon}

15: \newcommand{\beq}{\begin{equation}}

16: \newcommand{\eeq}{\end{equation}}

17: \newcommand{\bea}{\begin{eqnarray}}

18: \newcommand{\eea}{\end{eqnarray}}

19: \newcommand{\bean}{\begin{eqnarray}}

20: \newcommand{\eean}{\end{eqnarray}}

21: %%\eqnobysec

22: \textheight 23cm

23: \textwidth 17cm

24: \topmargin=0pt

25: \headheight=0cm

26: \headsep=12pt

27: \footskip=1.5cm

28: \oddsidemargin=-0.2cm

29: \evensidemargin=-0.2cm

30: \pagestyle{plain}

31: \def\d{{\mathrm{d}}}

32: \def\Rset{\mathbb{R}}

33: \def\Zset{\mathbb{Z}}

34: \def\fix{\mathrm{Fix}}

35: \def\sgn{{\mathrm{sgn}}}

36: \def\eg{{e.g.}}

37: \def\ie{{i.e.}}

38: \renewcommand{\topfraction}{1}

39: \renewcommand{\bottomfraction}{1}

40: \renewcommand{\textfraction}{0}

41: 

42: \begin{document}

43: 

44: \title{Discrete Embedded Solitons}

45: \author{Kazuyuki~Yagasaki}

46: \affiliation{Department of Mechanical and Systems Engineering,

47:  Gifu University, Gifu 501-1193, Japan}

48:  \author{Alan R.~Champneys}

49: \affiliation{Department of Engineering Mathematics, University of

50: Bristol, Bristol BS8 1TR, UK}

51: \author{Boris A.~Malomed}

52:  \affiliation{

53: Department of Interdisciplinary Studies,

54:  School of Electrical Engineering,

55:  Faculty of Engineering, Tel Aviv University, Tel Aviv 69978, Israel}

56: 

57: 

58: \begin{abstract}

59: We address the existence and properties of discrete

60: \textit{embedded solitons} (ESs), that is localized waves existing

61: inside the phonon band in a nonlinear dynamical-lattice model. The

62: model describes a one-dimensional array of optical waveguides with

63: both $\chi ^{(2)}$ (second-harmonic generation) and $\chi ^{(3)}$

64: (Kerr) nonlinearities, for which a rich family of ESs are known to

65: occur in the continuum limit. First, a simple motivating problem

66: is considered, in which the $\chi ^{(3)}$ nonlinearity acts in a

67: single waveguide. An explicit solution is constructed

68: asymptotically in the large-wavenumber limit. The general problem

69: is then shown to be equivalent to the existence of a homoclinic

70: orbit in a four-dimensional reversible map. From properties of

71: such maps, it is shown that (unlike ordinary gap solitons),

72: discrete ESs have the same codimension as their continuum

73: counterparts. A specific numerical method is developed to compute

74: homoclinic solutions of the map, that are symmetric under a

75: specific reversing transformation. Existence is then studied in

76: the full parameter space of the problem. Numerical results agree

77: with the asymptotic results in the appropriate limit and suggest

78: that the discrete ESs may be semi-stable as in the continuous

79: case.

80: \end{abstract}

81: 

82: \pacs{05.45.Yv, 05.45.-a, 42.65.Tg, 42.65.Ky}

83: 

84: \maketitle

85: \section{Introduction}

86: 

87: An \emph{embedded soliton} (ES) is an isolated solitary wave in a

88: non-integrable system that resides insides the continuous spectrum

89: of linear waves. Unlike regular \emph{gap solitons}, the existence

90: of ESs in continuum models is generally of codimension-one in the

91: model's parameter space. That is, they are embedded into wider

92: families of \emph{generalized} solitary waves which have

93: nonvanishing periodic tails attached to them, such that the ES

94: occurs at isolated values of the frequency, or other intrinsic

95: parameter of the solution family at which the tail's amplitude

96: vanishes. The concept of ESs was introduced in \cite{YaMaKa:99}

97: (see also \cite{ChMaYaKa:01,YaMaKaCh:01} for more details),

98: although particular examples of waves of this kind were actually

99: known earlier (but not under this name)

100: \cite{ChGr:97,FuEs:97,ChMaFr:98}. In the last few years, ESs have

101: been shown to be of relevance in a diverse range of nonlinear-wave

102: models, see, e.g., ~\cite{KoChBuSa:02,AtMa:01,EsFuGo:03,Ya:03} and

103: references therein. One of their characteristic properties is that

104: they are typically, at best, \emph{semi-stable}. That is, the

105: linearization around them gives rise to no unstable eigenvalues,

106: but a non-trivial zero-eigenvalue mode can be found. The latter

107: leads to an algebraic-in-time instability for energy-decreasing

108: perturbations, and a similar algebraic-in-time relaxation back to

109: the original pulse for energy-increasing perturbations

110: \cite{PeYa:02}. Only in special cases of systems which have more

111: than one dynamical invariant, or for which the travelling-wave

112: system is reversible but not Hamiltonian, have examples been found

113: of ESs that are truly asymptotically stable \cite{Ya:03}. Also

114: known are physically relevant models with special symmetries that

115: support continuous (rather than discrete) families of ESs, and a

116: large part thereof may be stable \cite{Mak,Merhasin}.

117: 

118: All the above-cited examples pertain to continuum models. A

119: fundamental issue is whether solitons of the embedded type may

120: also occur in discrete nonlinear equations modelling dynamical

121: lattices. For example, in \cite{ChKi:99} it was argued that moving

122: topological defects in Frenkel-Kontorova (discrete sine-Gordon)

123: lattices can meaningfully be described as \emph{embedded kinks}.

124: Some of specific examples of moving discrete solitons constructed

125: by means of the \textit{inverse method} (a model is sought for to

126: which a given moving soliton is an exact solution \cite{Flach})

127: may also have the embedded character. Such waves, which connect a

128: zero state to itself, translated through a multiple of the period

129: of the lattice potential (the `topological charge'), were first

130: (indirectly) identified by Peyrard and Kruskal \cite{PeKr:84} and

131: have later received much attention in a wide variety of

132: applications \cite{KiBr:98}. In a co-moving frame, these waves

133: should be thought of as localized solutions to

134: infinite-dimensional \emph{advance and delay} (nonlocal)

135: equations, see works \cite{SaZoEi:00,AiChRo:03} where branches of

136: embedded kinks with the topological charge 1 and 2 were computed.

137: Another recent study \cite{ESdiscrete} reported the existence of

138: ESs in a discrete lattice of the Ablowitz-Ladik type, but with a

139: quintic nonlinearity added to the usual cubic terms, and also with

140: an additional next-nearest-neighbor linear coupling. By tuning the

141: nonlinear terms, an explicit analytical solution for a discrete ES

142: was found in that model (which somehow resembles the

143: above-mentioned inverse method). Apart from these, we are aware of

144: no other work that systematically considers the issue of ESs in

145: discrete models.

146: 

147: A general objective of this paper is to understand the existence

148: and multiplicity properties of discrete solitons in lattice

149: systems for which the corresponding continuum model is known to

150: support ESs. The first fundamental issue is whether the

151: codimension of their existence is preserved under discretization.

152: In dissipative systems, solitary pulses are supported due to the

153: balance between forcing and damping, hence the corresponding

154: homoclinic orbits are always of codimension one. Thus, formally

155: similar to ESs, dissipative solitons typically exist as solutions

156: to the corresponding stationary ordinary differential equation at

157: discrete values of a relevant parameter (temporal or spatial

158: frequency, depending on the physical realization). Then, it will

159: generically happen that, under the discretization of that

160: equation, the codimension of the solitons \emph{decreases}. This

161: is possible because discretization of homoclinic orbits to

162: hyperbolic equilibria leads to transverse homoclinic connections

163: that exist for a range of parameter values \cite{FiSc}. However,

164: for ESs we shall show that, in contrast, discrete ESs typically

165: keep \emph{the same} codimension as their continuum counterparts,

166: i.e., they also exist at discrete values of the corresponding

167: physical parameter.

168: 

169: In this work, we restrict ourselves to a discretization of a

170: single continuum model, for which the discrete counterpart finds a

171: direct physical interpretation. In fact, our results and

172: methodology will carry over to other discretized models that

173: support ESs (this expectation is supported by the general fact

174: that equations of the discrete nonlinear-Schr\"{o}dinger type,

175: that we consider below, may be derived as an asymptotic limit from

176: a much larger class of discrete models \cite{Morgante}).

177: Specifically, we shall start with the continuum model in which ESs

178: were identified \cite{YaMaKa:99}. It describes an optical medium

179: with competing quadratic ($\chi ^{(2)}$) and cubic ($\chi ^{(3)}$)

180: nonlinearities (see \cite{Bang1,Bang2,Bang3} and references

181: therein for physical applications). In this model, the evolution

182: variable $z$ is the propagation distance in the nonlinear optical

183: waveguide, and the variable $t$ is either the reduced time (in the

184: case of temporal solitons) or, which is more physically relevant,

185: a transverse spatial coordinate in a planar waveguide. In the

186: temporal case, such a model can be written in dimensionless form

187: as (see detailed explanations in reviews

188: \cite{Review,DiscrChi2soliton})

189: \begin{eqnarray}

190: &&i\frac{\partial u}{\partial z}+\frac{1}{2}\frac{\partial ^{2}u}{\partial

191: t^{2}}+u^{\ast }v+\gamma _{1}(|u|^{2}+2|v|^{2})u=0,  \nonumber \\[-1ex]

192: &&  \label{eqn:PDE} \\[-1ex]

193: &&i\frac{\partial v}{\partial z}+\frac{\delta }{2}\frac{\partial

194: ^{2}v}{\partial t^{2}}+qv+\frac{1}{2}u^{2}+2\gamma

195: _{2}(|v|^{2}+2|u|^{2})v=0, \nonumber

196: \end{eqnarray}where $u$ and $v$ are the local amplitudes of the fundamental-frequency (FF)

197: and second-harmonic (SH) fields, an asterisk denotes complex

198: conjugation, $\delta $ is the relative dispersion of the SH, $q$

199: is the wavenumber mismatch, and $\gamma _{1,2}$ measure the

200: relative strength of the cubic (Kerr) nonlinearity compared to its

201: quadratic (SH-generating) counterpart, whose strength is

202: normalized to be $1$. In the physical model, $\gamma _{1,2} $ must

203: have the same sign (typically, they are positive, but may, in

204: principle, be negative too). The spatial variant of the system

205: (\ref{eqn:PDE}) takes the same form, with a partly different

206: interpretation of the dimensionless parameters, and is easier to

207: implement experimentally \cite{Bang1,Bang2,Bang3}. In the spatial

208: domain, one always has $\delta \equiv 1$.

209: 

210: The existence of ESs in this model was established in

211: \cite{YaMaKa:99,YaMaKaCh:01} for both cases, $\gamma _{1,2}>0$ and

212: $\gamma _{1,2}<0$. In the case of relevance to spatial waveguides

213: ($\delta =1$), fundamental ESs are absent in the model, while

214: higher-order ones can be found.

215: 

216: In models such as (\ref{eqn:PDE}) which includes $\chi ^{(2)}$

217: interaction, ESs can be embedded only in the continuous spectrum

218: of the SH component, while the FF wavenumber can never be located

219: inside the continuous spectrum. This feature is stipulated by the

220: asymmetry between the two equations in system (\ref{eqn:PDE}). If

221: the SH supports linear waves, while the FF has the possibility of

222: exponential localization like $e^{-\lambda |t|}$, then the $u^{2}$

223: term, which drives the SH field in the second equation

224: (\ref{eqn:PDE}), allows it to have tails that decay exponentially

225: at a rate $e^{-2\lambda |t|}$. Such a solitary wave was said in

226: reference \cite{YaMaKa:99} to be \emph{tail-locked} and,

227: accordingly, the SH equation to be \emph{nonlinearisable} (because

228: the quadratic term can never be neglected in this equation). The

229: tail locking does not take place if the SH field supports its own

230: exponentially decaying tails that asymptotically (for

231: $|t|\rightarrow \infty $) dominate over the tails induced by the

232: term $u^{2}$.

233: 

234: In this paper we study the existence of discrete ESs in a model

235: arising from discretization of the (actually, spatial)

236: $t$-coordinate in (\ref{eqn:PDE}). One motivation for doing this

237: is to understand the effect of numerical approximation (which also

238: implies discretization, once a finite-difference scheme is

239: employed) on the computation of ESs in model systems. But this is

240: not our primary motivation. Lattice models play an increasingly

241: important role in describing physical phenomena in a number of

242: newly developed experimental settings. In particular, the discrete

243: version of the $\chi ^{(2)}:\chi ^{(3)}$ model describes an array

244: of the corresponding linearly-coupled waveguides. The creation of

245: discrete spatial solitons in a system of channel waveguides with

246: the $\chi ^{(2)}$ nonlinearity was recently reported

247: \cite{DiscrChi2soliton}. Our assumption concerning relevant

248: solutions is that, once the continuum version of the model readily

249: gives rise to ESs, then there is a chance to find them in the

250: corresponding lattice model too. In this paper, we show that this

251: is the case indeed and investigate how we may then pass to the

252: continuum limit. We report the corresponding discrete ES that can

253: be found, under special conditions, in an approximate analytical

254: form, and, in the general case, numerically.

255: 

256: The paper is organized as follows. In Section~2, we present the

257: lattice model to be considered in this work and discuss its

258: physical applicability. In Section~3, an asymptotic analysis is

259: undertaken for a reduced model where the cubic nonlinearity is

260: present only at a single site. Section~4 then goes on to discuss

261: the meaning of discrete ESs, as a solution to a system of

262: stationary finite-difference equations, in terms of a homoclinic

263: orbit to a fixed point of a discrete-time reversible map. This

264: makes it possible to understand that the solutions we seek must

265: have codimension one, similar to ESs in continuum models. In terms

266: of the same approach, in Section~5 we describe the numerical

267: procedure developed to search for discrete ES solutions. Note that

268: this requires special adaptation or other methods for finding

269: homoclinic orbits of maps, owing to the non-hyperbolic nature of

270: the fixed point. Our approach heavily relies on the reversibility

271: although it can be readily modified to treat non-reversible ESs.

272: Numerical results are reported in Section~6, in the form of the

273: corresponding codimension-one solution branches in the parameter

274: space. An array of actual shapes of the ESs is displayed too. Our

275: numerical results also show that the asymptotic analysis of

276: Section~3 can well predict the codimension-one set of ESs in the

277: parameter space in the model when the wavenumber is large. The

278: paper is concluded by Section~7 along with a preliminary stability

279: result for the fundamental ES.

280: 

281: \section{The models}

282: 

283: We now consider Eqs.~(\ref{eqn:PDE}) in the spatial domain. Direct

284: discretization of the spatial second derivative produces a lattice

285: model

286: \begin{eqnarray}

287: &&i\frac{du_{n}}{dz}+\frac{1}{2}D(u_{n+1}+u_{n-1}-2u_{n})+u_{n}^{\ast

288: }v_{n}+\gamma _{1}(|u_{n}|^{2}+2|v_{n}|^{2})u_{n}=0,  \nonumber \\[-1ex]

289: &&  \label{eqn:lattice} \\[-1ex]

290: &&i\frac{dv_{n}}{dz}+\frac{1}{2}\delta

291: D(v_{n+1}+v_{n-1}-2v_{n})+qv_{n}+\frac{1}{2}u_{n}^{2}+2\gamma

292: _{2}(|v_{n}|^{2}+2|u_{n}|^{2})v_{n}=0, \nonumber

293: \end{eqnarray}where $n$ is the discrete transverse coordinate which assumes integer

294: values, and $D=1/h^{2}$ with $h$ the stepsize of the

295: discretization. Effective coefficients of the lattice diffraction

296: for the FF and SH waves are $D_{1}\equiv D$ and $D_{2}=\delta D$;

297: %Note that straightforward discretization of the spatial-domain model,

298: with $\delta =1$, this yields $D_{2}=D_{1}$ in

299: Eqs.~(\ref{eqn:lattice}).

300: %However,

301: Unequal values of these coefficients, $D_{1}\neq D_{2}$ (i.e.,

302: $\delta \neq 1 $), including those with \emph{opposite signs}, can

303: in principle be realized experimentally (the latter possibility

304: was recently employed to predict the existence of discrete gap

305: solitons \cite{PanosZiad}). To do this one would use a

306: \textit{diffraction-management} technique, which is based on

307: oblique propagation of the beam across the waveguide array which

308: is modeled by the discrete system

309: \cite{DiffrManagement1,DiffrManagement2}. In that case, the

310: effective lattice-diffraction coefficients are

311: \begin{equation}

312: D_{m}=D_{m}^{(0)}\cos (mQ),\quad m=1,2,  \label{management}

313: \end{equation}where $Q\equiv k\sin \theta $ is the transverse wavenumber, $k$ is the

314: beam's wavenumber (normalization is such that the array spacing is

315: unity), $\theta $ is the angle between the Poynting vector and the

316: coordinate $z$ running along the waveguides, and $D_{1,2}^{(0)}$

317: are the original lattice-dispersion coefficients, corresponding to

318: $Q=0$. As seen in Eq.~(\ref{management}), one can efficiently

319: control the size and signs of $D_{1,2}$ by choosing an appropriate

320: angle $\theta $. In particular, the ratio $D_{2}/D_{1}$ can be

321: made large by choosing $Q=\pi /2+\varepsilon $, or $Q=3\pi

322: /2-\varepsilon $, for a small positive $\varepsilon $.

323: 

324: To cast the model in a normalized form, we note that $D_{1}$ can

325: be made positive, if it was originally negative, by means of the

326: complex conjugation (i.e., Eqs.~(\ref{eqn:lattice}) are replaced

327: by their counterparts for $u_{n}^{\ast }$ and $v_{n}^{\ast }$),

328: combined with the changes $v_{n}\rightarrow -v_{n}$ and $\gamma

329: _{1,2}\rightarrow -\gamma_{1,2}$. The size of $D_{1}$ may be set

330: equal to $1$ by means of the rescaling: $zD_{1}\rightarrow z$,

331: $\left( u_{n},v_{n}\right)/D_{1} \rightarrow \left(

332: u_{n},v_{n}\right)$, which leads to the final normalized form of

333: the discrete model,

334: \begin{eqnarray}

335: && i\frac{du_{n}}{dz}+\frac{1}{2}(u_{n+1}+u_{n-1}-2u_{n})+u_{n}^{\ast}v_{n}

336: +\tilde{\gamma}_{1}(|u_{n}|^{2}+2|v_{n}|^{2})u_{n}=0,  \nonumber \\[-1ex]

337: &&  \label{normalised} \\[-1ex]

338: && i\frac{dv_{n}}{dz}+\frac{1}{2}\delta(v_{n+1}+v_{n-1}-2v_{n})

339: +\tilde{q}v_{n}+\frac{1}{2}u_{n}^{2}

340: +2\tilde{\gamma}_{2}(|v_{n}|^{2}+2|u_{n}|^{2})v_{n}=0,  \nonumber

341: \end{eqnarray}

342: where

343: \begin{equation}

344: \tilde{q}=q/D_{1},\quad \tilde{\gamma}_{1,2}=D_{1}\gamma _{1,2}.

345: \label{delta}

346: \end{equation}

347: The dimensionless constants (\ref{delta}), including $\delta$, may, in

348: general, be positive, zero, or negative.

349: 

350: To obtain some analytical results, we shall also introduce a simplified

351: model where ESs can be found in approximate form. In the simplified model,

352: only the central site (the one corresponding to $n=0$) carries the Kerr

353: nonlinearity. Such a model is physically feasible too, if the central

354: waveguide in the array is doped to enhance its Kerr nonlinearity. Thus, the

355: simplified model is based on the equations

356: \begin{eqnarray}

357: && i\frac{d u_{n}}{d z} +\frac{1}{2}\left(u_{n+1}+u_{n-1}-2u_{n}\right)+u_{n}^{\ast }v_{n}

358: +\tilde{\gamma}_{1}\epsilon_{n}\left(|u_{0}|^{2}+2|v_{0}|^{2}\right)u_{0}=0,  \nonumber \\[-1ex]

359: &&  \label{eqn:mod} \\[-1ex]

360: && i\frac{d v_{n}}{d z} +\frac{1}{2}\delta

361: \left(v_{n+1}+v_{n-1}-2v_{n}\right)

362: +\tilde{q}v_{n}+\frac{1}{2}u_{n}^{2}

363: +2\tilde{\gamma}_{2}\epsilon_{n}\left(|v_{0}|^{2}+2|u_{0}|^{2}\right)v_{0}=0,

364: \nonumber

365: \end{eqnarray}

366: where $\epsilon _{n}=0$ for $n\neq 0$, and $\epsilon _{n}=1$ for $n=0$.

367: 

368: In either model, (\ref{eqn:lattice}) or (\ref{eqn:mod}), the ES

369: corresponds to a solution with FF component exponentially

370: localized:

371: \begin{equation}

372: u_{n}\sim A\exp (i\tilde{k}z-\lambda |n|)\qquad

373: \mbox{as

374: $|n|\rightarrow\infty$},  \label{uasympt}

375: \end{equation}where $A$ is a real amplitude and $\lambda $ is positive and real.

376: Simultaneously, the propagation constant $\tilde{k}$ must belong to the

377: \textit{phonon band} of the SH equation. That is, the SH component of the

378: solution generically will have a nonvanishing tail, of the form

379: \begin{equation}

380: v_{n}\sim C\exp (2i\tilde{k}z-ip|n|)\qquad \mbox{as $|n|\rightarrow \infty$},

381: \label{vasympt}

382: \end{equation}where $p$ is real. The existence of an ES corresponds to the vanishing of

383: the constant $C$ in Eq.~(\ref{vasympt}). Linearization of

384: Eqs.~(\ref{eqn:lattice}) and (\ref{eqn:mod}) and subsequent

385: substitution of expressions (\ref{uasympt}) and (\ref{vasympt})

386: yield the following relations:

387: \begin{equation}

388: \tilde{k}=2\sinh ^{2}(\lambda /2),  \label{k}

389: \end{equation}\begin{equation}

390: \tilde{q}-2\tilde{k}=2\delta \sin ^{2}(p/2).  \label{q}

391: \end{equation}From Eqs.~(\ref{k}) and (\ref{q}) we see that a discrete ES may exist in the

392: case of $\tilde{k}>0$, with $\tilde{q}$ belonging to the region

393: \begin{equation}

394: 0<\tilde{q}-2\tilde{k}<2\delta \quad \mathrm{or}\quad 2\delta

395: <\tilde{q}-2\tilde{k}<0,  \label{area}

396: \end{equation}depending on whether $\delta $ is positive or not. Note that we can only

397: choose $\tilde{k}$, $\tilde{q}$ and $\delta $ independently in the

398: set of $(\tilde{k},\tilde{q},\delta ,\lambda ,p)$ since $\lambda $

399: and $p$ are determined by the other three parameters via Eqs.

400: (\ref{k}) and (\ref{q}).

401: 

402: \section{Asymptotic analysis for the simplified model}

403: 

404: We first consider the simplified model (\ref{eqn:mod}), with the

405: $\chi ^{(3)} $ nonlinearity present only at the central site. An

406: ES solution is sought asymptotically under the assumption that

407: $|v_{n}|\ll |u_{n}|=O(1)$, the validity of which will be checked

408: \textit{a posteriori}. Accordingly, the quadratic term in the

409: first equation of the system (\ref{eqn:mod}) may be dropped to

410: first approximation, which makes the expressions (\ref{uasympt})

411: and (\ref{k}) an \emph{exact solution} for $n\neq 0$. At the point

412: $n=0$, the cross-phase-modulation term, $|v_{0}|^{2}u_{0}$, in the

413: first equation (\ref{eqn:mod}) may be dropped too to leading

414: order. Then, the equation at $n=0$ yields a final result for the

415: FF component of the soliton, $A^{2}=\tilde{\gamma}_{1}^{-1}\sinh

416: \lambda $, which implies that the discrete soliton in the FF

417: component is supported by itself, without coupling to the SH

418: component. Further, using Eq.~(\ref{k}) one can express $A^{2}$ in

419: terms of $\tilde{k}$,

420: \begin{equation}

421: A^{2}=\tilde{\gamma}_{1}^{-1}\sqrt{\tilde{k}(\tilde{k}+2)},  \label{Ak}

422: \end{equation}

423: which means that the solution exists only in the case of

424: $\tilde{\gamma}_{1}>0$.

425: 

426: Now, we tackle the second equation of (\ref{eqn:mod}) and

427: substitute the FF field in the form of Eqs.~(\ref{uasympt}) and

428: (\ref{k}). Obviously, at $n\neq 0$, an exact solution can be

429: found, which precisely corresponds to an ES, as it does not

430: contain the nonvanishing tail (\ref{vasympt}) and is `tail-locked'

431: to the square of the FF field, as explained in the Introduction.

432: We find explicitly that

433: \begin{equation}

434: v_{n}=B\exp (2i\tilde{k}z-2\lambda |n|),  \label{vsolution}

435: \end{equation}where

436: \begin{equation}

437: B=-\frac{A^{2}}{2[2\delta \sinh ^{2}\lambda

438: +(\tilde{q}-2\tilde{k})]}\equiv

439: -\frac{1}{2\tilde{\gamma}_{1}}\frac{\sqrt{\tilde{k}(\tilde{k}+2)}}{2\delta

440: \cdot \tilde{k}(\tilde{k}+2)+(\tilde{q}-2\tilde{k})}.  \label{B}

441: \end{equation}Substituting the expressions (\ref{uasympt}) and (\ref{vsolution}) into the

442: second equation of (\ref{eqn:mod}) at $n=0$, we have

443: \begin{equation}

444: \cosh \lambda =\frac{2\tilde{\gamma}_{2}}{\delta \tilde{\gamma}_{1}}.

445: \label{eqn:lambda}

446: \end{equation}Here, following the assumption $|v_{n}|\ll |u_{n}|$, we neglected the

447: self-phase-modulation term $|v_{0}|^{2}v_{0}$ at this point.

448: Equation~(\ref{eqn:lambda}) also implies that

449: \begin{equation}

450: \frac{2\tilde{\gamma}_{2}}{\delta \tilde{\gamma}_{1}}>1,  \label{eqn:12}

451: \end{equation}and, especially, that $\delta $ has the same sign as $\tilde{\gamma}_{2}$

452: since $\lambda ,\,\tilde{\gamma}_{1}>0$. Finally, using the

453: relations (\ref{k}) and (\ref{eqn:lambda}), we obtain

454: \begin{equation}

455: \tilde{k}=\frac{2}{\delta }\frac{\tilde{\gamma}_{1}}{\tilde{\gamma}_{2}}-1,

456: \label{final}

457: \end{equation}which is positive by virtue of (\ref{eqn:12}).

458: This result selects the \emph{single value} of $\tilde{k}$ at which the simplified model admits the

459: existence of an ES, with tail-locked SH component. It follows from

460: Eq.~(\ref{final}) that such a solution exists if the relative

461: lattice-diffraction coefficient (see Eq.~(\ref{delta})) belongs to

462: the interval

463: \begin{equation}

464: 0<|\delta |<2\tilde{\gamma}_{1}/|\tilde{\gamma}_{2}|,\quad

465: \tilde{\gamma}_{1}>0,  \label{12+}

466: \end{equation}where $\delta $ and $\tilde{\gamma}_{2}$ must be of the same sign.

467: 

468: It is now necessary to check compatibility of the solution with

469: the underlying assumption, $|v_{n}|\ll |u_{n}|$, which implies

470: $A^{2}\ll B^{2}$ (see Eqs.~(\ref{Ak}) and (\ref{B})).

471: Straightforward consideration shows that this condition amounts to

472: $|\delta|\ll|\tilde{\gamma}_{1,2}|$. We also note that, if the FF

473: component of the soliton is strongly localized, so that

474: $e^{-\lambda }\ll 1$ (see Eq.~(\ref{uasympt})), the simplified

475: model is actually equivalent to the full one, as the nonlinearity

476: in the first equation of (\ref{eqn:lattice}) is then negligible at

477: $n\neq 0$. From (\ref{k}) we see that this condition implies that

478: $\tilde{k}$ must be large and hence, from (\ref{final}), that

479: $|\delta|\ll 1$.

480: 

481: A noteworthy feature of Eqs.~(\ref{final}) and (\ref{12+}) is that

482: they do not involve the mismatch parameter $\tilde{q}$. However,

483: the condition that the soliton found above is embedded implies

484: that $\tilde{k}$ given by Eq.~(\ref{final}) must belong to the

485: interval (\ref{area}), which relates $\tilde{k}$ to $\tilde{q}$,

486: as $|\delta|\ll 1$. This approximate analysis, valid too for the

487: original model in the limit of small $|\delta|$, means that, for

488: an isolated selected $\tilde{k}$-value given by (\ref{final}),

489: there exists a curve of single-humped ESs approximately

490: parametrized by $\tilde{q}$ in the narrow interval

491: \begin{equation}

492: \tilde{q}\in (2\tilde{k},2\tilde{k}+2\delta) \quad\mbox{for

493: $\delta>0$}\quad\mbox{or}\quad (2\tilde{k}+2\delta,2\tilde{k})

494: \quad\mbox{for $\delta<0$} \label{qinterval}

495: \end{equation}

496: in the $(\tilde{q},\tilde{k})$-plane for $\delta$ fixed.

497: 

498: \section{Reversible maps}

499: 

500: Let us now consider the model (\ref{eqn:lattice}) in the general

501: case, and understand, from a dynamical systems point of view, why

502: discrete ESs should exist. In this context, one of essential

503: issues is the transition to the continuum limit. The scaling

504: leading to Eq. (\ref{normalised}), in which $D_{1}=1$, does not

505: allow this. Hence, we undo this scaling and set $D_{1}=D$,

506: $D_{2}=\delta D$. As above, we look for stationary solutions in

507: the form

508: \begin{equation}

509: u_{n}=U_{n}e^{ikz},\quad v_{n}=V_{n}e^{2ikz},  \label{eqn:UV}

510: \end{equation}where $U_{n}$ and $V_{n}$ are real. Equations~(\ref{eqn:lattice}) thus

511: reduce to

512: \begin{eqnarray}

513: &&\frac{1}{2}D(U_{n+1}+U_{n-1}-2U_{n})-kU_{n}+U_{n}V_{n}+\gamma

514: _{1}(U_{n}^{2}+2V_{n}^{2})U_{n}=0,  \nonumber \\[-1ex]

515: &&  \label{eqn:disc} \\[-1ex]

516: &&\frac{1}{2}\delta

517: D(V_{n+1}+V_{n-1}-2V_{n})+(q-2k)V_{n}+\frac{1}{2}U_{n}^{2}+2\gamma

518: _{2}(V_{n}^{2}+2U_{n}^{2})V_{n}=0,  \nonumber

519: \end{eqnarray}where parameters without the tildes are related to those with tildes by

520: \begin{equation}

521: q=\tilde{q}D,\quad k=\tilde{k}D,\quad \gamma _{1,2}=\tilde{\gamma}_{1,2}/D

522: \label{rescaledpars}

523: \end{equation}Note that the possible existence regions of ESs, given by expression (\ref{area}), now becomes

524: \begin{equation}

525: 0<q-2k<2\delta D\quad \mathrm{or}\quad 2\delta D<q-2k<0.  \label{area_scale}

526: \end{equation}

527: 

528: First and foremost, we want to establish the codimension of ES

529: solutions to (\ref{eqn:disc}). In order to do that, it is useful

530: to recast the system as a four-dimensional map. Specifically, upon

531: scaling the variables as $\xi _{n}\equiv \sqrt{2|\gamma _{1}|/D}\

532: U_{n}$ and $\eta _{n}\equiv \sqrt{2|\gamma _{1}|/D}\ V_{n}$, we

533: can indeed view Eqs.~(\ref{eqn:disc}) as defining a\emph{\

534: four-dimensional map},

535: \begin{equation}

536: \zeta _{n+1}=F_{\epsilon }(\zeta _{n}),  \label{eqn:F}

537: \end{equation}

538: where $\zeta _{n}\equiv (\chi _{n},\xi _{n},\mu _{n},\eta _{n})$ and

539: \[

540: F_{\epsilon }(\zeta _{n})=\left(

541: \begin{array}{c}

542: \xi _{n} \\

543: f_{\epsilon }^{(1)}(\zeta _{n}) \\

544: \eta _{n} \\

545: f_{\epsilon }^{(2)}(\zeta _{n})\end{array}\right) .

546: \]

547: Here we define

548: \begin{eqnarray}

549: f_{\epsilon }^{(1)}(\zeta _{n}) &=&-\chi _{n}+2\nu _{1}\xi _{n}-(\epsilon

550: \xi _{n}\eta _{n}+\kappa _{1}(\xi _{n}^{2}+2\eta _{n}^{2})\xi _{n}), \\

551: f_{\epsilon }^{(2)}(\zeta _{n}) &=&-\mu _{n}+2\nu _{2}\eta _{n}-\delta

552: ^{-1}\left( \frac{1}{2}\epsilon \xi _{n}^{2}+2\kappa _{2}(\eta _{n}^{2}+2\xi

553: _{n}^{2})\eta _{n}\right) ,

554: \end{eqnarray}

555: where

556: \begin{eqnarray}

557: \nu _{1} &=&1+\frac{k}{D},\quad \nu _{2}\equiv 1+\frac{2k-q}{\delta D},\quad

558: \epsilon \equiv \sqrt{\frac{2}{D|\gamma _{1}|}},  \nonumber \\[-1.5ex]

559: &&  \label{eqn:para} \\[-1.5ex]

560: \kappa _{1} &\equiv &{\mathrm{sgn}}(\gamma _{1}),\quad \kappa _{2}\equiv

561: \gamma _{2}/\left\vert \gamma _{1}\right\vert .  \nonumber

562: \end{eqnarray}

563: 

564: The map $F_{\epsilon }$ is \textit{reversible} \cite{De:76b,LaRo:98}, in the

565: sense that $F_{\epsilon }^{-1}(R(\zeta _{n}))=R(F_{\epsilon }(\zeta _{n}))$

566: holds, where $R:\mathbb{R}^{4}\rightarrow \mathbb{R}^{4}$ is the (linear)

567: involution given by

568: \begin{equation}

569: R:(\chi _{n},\xi _{n},\mu _{n},\eta _{n})\mapsto (\xi _{n},\chi _{n},\eta

570: _{n},\mu _{n}).  \label{R1}

571: \end{equation}

572: Note that the map $F_{\epsilon }$ is also reversible, under the second

573: reversibility

574: \begin{equation}

575: \hat{R}:(\chi _{n},\xi _{n},\mu _{n},\eta _{n})\mapsto (f_{\epsilon

576: }^{(1)}(\zeta _{n}),\xi _{n},f_{\epsilon }^{(2)}(\zeta _{n}),\eta _{n}).

577: \label{R2}

578: \end{equation}

579: In what follows, we consider orbits of the map that are reversible

580: with respect to $R$. These will correspond to ESs that have their

581: central peaks on a lattice site, which are the kind of lattice

582: solitons that are most frequently observed in physical

583: applications. In contrast, ESs that are symmetric under $\hat{R}$

584: would have their central peak between two lattice sites. Both $R$

585: and $\hat{R}$ arise from the fact that the ordinary differential

586: equations for steady solutions of the continuum model

587: (\ref{eqn:PDE}),

588: \begin{eqnarray}

589: &&\frac{1}{2}\frac{d^{2}U}{dt^{2}}-kU+UV+\gamma _{1}(U^{2}+2V^{2})U=0,

590: \nonumber \\[-1ex]

591: &&  \label{eqn:cont} \\[-1ex]

592: &&\frac{\delta }{2}\frac{d^{2}V}{dt^{2}}+(q-2k)V+\frac{1}{2}U^{2}+2\gamma

593: _{2}(V^{2}+2U^{2})V=0,  \nonumber

594: \end{eqnarray}

595: are reversible too, under an involution

596: $\tilde{R}:(U,dU/dt,V,dV/dt)\mapsto(U,-dU/dt,V,-dV/dt)$.

597: 

598: A fundamental characteristic of reversible maps is that if

599: $\{\zeta_{n}\}_{n=-\infty }^{\infty }$ is an orbit then

600: $\{R(\zeta_{-n})\}_{n=-\infty }^{\infty }$ is also an orbit. We

601: say that an orbit $\{\zeta _{n}\}_{n=-\infty}^{\infty}$ is

602: \textit{symmetric} (with respect to the reversibility) if

603: $\zeta_{n+1}=R(\zeta _{-n})$. Denote $\mathrm{Fix}(R)=\{\zeta \in

604: \mathbb{R}^{4}\,|\,F_{\epsilon }(\zeta )=R(\zeta )\}$. We easily

605: see that $\zeta =(\chi ,\xi ,\mu ,\eta )\in \mathrm{Fix}(R)$ if

606: and only if $\chi =f_{\epsilon }^{(1)}(\zeta )$ and $\mu

607: =f_{\epsilon }^{(2)}(\zeta )$. Thus, $\mathrm{Fix}(R)$ depends on

608: the particular form of the map $F_{\epsilon }$. An orbit $\{\zeta

609: _{n}\}_{n=-\infty }^{\infty }$ is symmetric if and only if $\zeta

610: _{0}\in \mathrm{Fix}(R)$, i.e., $\xi _{-1}=\xi _{1}$ and $\eta

611: _{-1}=\eta _{1}$.

612: 

613: The map $F_{\epsilon }$ has a fixed point at the origin $O$, whose Jacobian

614: matrix is

615: \[

616: J=\left(

617: \begin{array}{cccc}

618: 0 & 1 & 0 & 0 \\

619: -1 & 2\nu _{1} & 0 & 0 \\

620: 0 & 0 & 0 & 1 \\

621: 0 & 0 & -1 & 2\nu _{2}\end{array}\right) .

622: \]It follows from Eq.~(\ref{area_scale}) that $\nu _{1}>1$ and $|\nu _{2}|<1$,

623: hence the origin is a fixed point of $F_{\epsilon }$ of

624: saddle-center type. The saddle-center has one-dimensional stable

625: and unstable manifolds, $W^{s}(O)$ and $W^{u}(O)$, which are

626: tangent to the stable and unstable subspaces spanned by the

627: vectors $(1,\nu _{1}-\sqrt{\nu _{1}^{2}-1},0,0)$ and $(1,\nu

628: _{1}+\sqrt{\nu _{1}^{2}-1},0,0)$, respectively, and a

629: two-dimensional center manifold, $W^{c}(O)$, which is tangent to

630: the center subspace spanned by a set of two vectors, $(0,0,1,0)$

631: and $(0,0,0,1)$. By the fundamental property of reversible maps,

632: $W^{s}(O)=R(W^{u}(O))$ and $W^{u}(O)=R(W^{s}(O))$. Thus, if there

633: exists an orbit $\{\zeta _{n}\}_{n=-\infty }^{\infty }$ on

634: $W^{u}(O)$ such that $\zeta _{0}\in \mathrm{Fix}(R)$, then it is

635: also contained in $W^{s}(O)$ and is a symmetric homoclinic orbit

636: to $O$.

637: 

638: If $F_{\epsilon }$ were not reversible, then such intersections

639: between the one-dimensional stable and unstable manifolds in the

640: four-dimensional phase space would be of codimension two. However,

641: symmetric homoclinic orbits are of codimension one, since, for

642: their existence, we require an intersection between the

643: one-dimensional unstable manifold $W^{u}(O)$ and the

644: two-dimensional manifold $\mathrm{Fix}(R)$. Thus, since a

645: homoclinic orbit to $O$ for $F_{\epsilon }$ represents precisely

646: an ES in the lattice system (\ref{eqn:disc}), we have established

647: that:

648: 

649: \noindent \emph{Embedded solitons of the lattice model are of

650: \textbf{codimension-one} in the parameter space, provided they are

651: symmetric under a reversibility equivalent to $R$ (or $\hat{R}$)}.

652: 

653: \noindent Note that this is identical to the multiplicity result known in

654: the continuum version of the model \cite{ChMaYaKa:01}.

655: 

656: Finally, in what follows we shall also treat the case of pure

657: $\chi ^{(2)}$ nonlinearity, $\gamma _{1}=\gamma _{2}=0$. This case

658: is physically important, as experiments could be quite conceivably

659: be set up in a medium with negligible Kerr nonlinearity. However,

660: without the $\chi^{(3)}$ terms, no ES exists in the continuum

661: model \cite{ChMaYaKa:01}. It will therefore be important to find

662: out whether the same is true in the discrete model too.

663: 

664: With $\gamma _{1}=\gamma _{2}=0$, the scaling leading to $F_{\epsilon}$

665: becomes invalid, so we shall consider instead a family of four-dimensional

666: maps

667: \begin{equation}

668: \zeta _{n+1}=G_{\epsilon }(\zeta _{n}),  \label{eqn:G}

669: \end{equation}

670: where we define

671: \[

672: G_{\epsilon }(\zeta _{n})=\left(

673: \begin{array}{c}

674: \xi _{n} \\

675: g_{\epsilon }^{(1)}(\zeta _{n}) \\

676: \eta _{n}, \\

677: g_{\epsilon }^{(2)}(\zeta _{n})\end{array}\right) ,

678: \]

679: with

680: \begin{eqnarray}

681: g_{\epsilon }^{(1)}(\zeta _{n}) &=&-\chi _{n}+2\nu _{1}\xi _{n}-(\epsilon

682: \xi _{n}\eta _{n}+(1-\epsilon )(\xi _{n}^{2}+2\eta _{n}^{2})\xi _{n}), \\

683: g_{\epsilon }^{(2)}(\zeta _{n}) &=&-\mu _{n}+2\nu _{2}\eta _{n}-\delta

684: ^{-1}\left( \frac{1}{2}\epsilon \xi _{n}^{2}+2(1-\epsilon )(\eta

685: _{n}^{2}+2\xi _{n}^{2})\eta _{n}\right),

686: \end{eqnarray}

687: with $\nu _{1}$ and $\nu _{2}$ given by (\ref{eqn:para}). The map

688: $G_{\epsilon }$ is reversible under the same involutions $R$ and

689: $\hat{R}$. The map $G_{1}$ is equivalent to (\ref{eqn:disc}) with

690: $\gamma _{1}=\gamma _{2}=0$ if one sets $\xi _{n}=U_{n}$ and $\eta

691: _{n}=V_{n}$, and simultaneously $G_{0}$ coincides with $F_{0}$

692: with $\kappa_{1}=\kappa _{2}=1 $. The origin $O$ is also a

693: saddle-center of $G_{\epsilon }$ and has the same stable, unstable

694: and center subspaces as those of $F_{\epsilon }$. So we can apply

695: all the arguments presented above for $F_{\epsilon }$ to

696: $G_{\epsilon }$ if we replace $F_{\epsilon }$ with $G_{\epsilon }$

697: in the definition of $\mathrm{Fix}(R)$. In particular, a

698: homoclinic orbit to $O$ for $G_{\epsilon }$ represents an ES in

699: the lattice system (\ref{eqn:lattice}) with $\gamma _{1}=\gamma

700: _{2}=0$.

701: 

702: \section{Numerical procedure}

703: 

704: Several numerical procedures exist for finding homoclinic orbits to fixed

705: points in maps, see, e.g., \cite{K81,BK97a,BK97b,Y98a,BBV00}. Here we shall

706: use an adaptation of these methods that takes into regard both the

707: reversibility and the non-hyperbolic nature of the fixed point.

708: 

709: To compute symmetric homoclinic orbits for $F_{\epsilon }$, we consider the

710: three-dimensional algebraic problem

711: \begin{eqnarray}

712: &&\chi _{-N}-\left( \nu _{1}-\sqrt{\nu _{1}^{2}-1}\right) \xi _{-N}=0,

713: \label{eqn:con1} \\

714: &&\xi _{1}=\xi _{-1},\quad \eta _{1}=\eta _{-1},  \label{eqn:con2}

715: \end{eqnarray}

716: for $N>0$ sufficiently large, where $\zeta_{\pm 1}

717: \equiv(\chi_{\pm 1},\xi_{\pm 1},\mu_{\pm 1},\eta_{\pm 1})

718: =F_{\epsilon }^{N\pm 1}(\chi_{-N},\xi_{-N},0,0)$. The condition

719: (\ref{eqn:con1}) means that the point $(\chi _{-N},\xi _{-N},0,0)$

720: lies in the one-dimensional unstable subspace of the fixed point

721: at the origin, and condition (\ref{eqn:con2}) means that $\zeta

722: _{0}=(\chi _{0},\xi _{0},\mu_{0},\eta _{0})\in \mathrm{Fix}(R)$,

723: i.e., the finite orbit $\{\zeta_{n}\}_{n=-N}^{0}$ intersects

724: $\mathrm{Fix}(R)$ at $n=0$. Thus, adding a parameter as an extra

725: unknown variable, Eqs.~(\ref{eqn:con1}) and (\ref{eqn:con2})

726: represent a formally well-posed system of three equations for two

727: unknowns, $\chi _{-N}$ and $\xi _{-N}$, and the free parameter,

728: which can be chosen to be any of $\epsilon$, $\delta$, $k$, $D$,

729: $q$, $\kappa_{1}$ or $\kappa_{2}$. A non-trivial solution for $N$

730: sufficiently large gives an approximate homoclinic orbit $\{\zeta

731: _{n}\}_{n=-N}^{N+1}$ of $F_{\epsilon }$, that is symmetric under

732: $R$, which in turn corresponds to a discrete ES solution of the

733: lattice equation (\ref{eqn:lattice}) for $\gamma _{1,2}\neq 0$. In

734: the case $\gamma _{1,2}=0$, the same treatment can be used to find

735: symmetric homoclinic orbits for $G_{\epsilon }$. Note that one

736: test of the validity of this approximation is to measure the

737: distance of the point $(\chi _{-N},\xi _{-N},0,0)$ from the

738: origin. By virtue of the stable manifold theorem for maps

739: \cite{GH83}, we know that the error will be proportional to this

740: distance squared.

741: 

742: Branches of solutions to Eqs.~(\ref{eqn:con1}) and

743: (\ref{eqn:con2}) can be continued in a second parameter using

744: pseudo-arclength continuation. To this end, we use the code

745: \texttt{AUTO} \cite{Auto}. The problem now is finding a good

746: initial point along a branch of discrete ESs. Here we can use a

747: regular perturbation approach by first finding solutions to the

748: map $F_{0}$, making use of the fact that when $\epsilon =0$, the

749: $(\chi _{n},\xi _{n})$-plane is invariant under $F_{\epsilon }$.

750: The restriction of $F_{0}$ onto the invariant plane is given by

751: \begin{equation}

752: (\chi _{n+1},\xi _{n+1})=f(\chi _{n},\xi _{n}),  \label{eqn:2d}

753: \end{equation}

754: where

755: \[

756: f(\chi _{n},\xi _{n})\equiv (\xi _{n},-\chi _{n}+2\nu _{1}\xi _{n}-\kappa

757: _{1}\xi _{n}^{3}).

758: \]The two-dimensional map $f$ also has a hyperbolic saddle at the origin, and

759: is reversible under an involution,

760: \[

761: \bar{R}:(\chi _{n},\xi _{n})\mapsto (\xi _{n},\chi _{n}).

762: \]The stable (resp. unstable) manifold of the saddle, $\bar{W}^{s}(O)$

763: (resp. $\bar{W}^{u}(O)$), is tangent to its stable (resp. unstable) subspace spanned

764: by a vector $(1,\nu _{1}-\sqrt{\nu _{1}^{2}-1})$ (resp.

765: $(1,\nu_{1}+\sqrt{\nu _{1}^{2}-1}$)). By the reversibility,

766: $\bar{W}^{s}(O)=\bar{R}(\bar{W}^{u}(O))$ and vice versa. When $\nu

767: _{1}>0$ and $\kappa _{1}>0$, the stable and unstable manifolds

768: intersect transversely and form homoclinic tangles, as shown in

769: Fig.~\ref{fig:3a}.

770: 

771: \begin{figure}[t]

772: \begin{center}

773: \includegraphics[scale=1.2]{fig1.eps}

774: \end{center}

775: \caption{Homoclinic tangles for the two-dimensional map $f$ with

776: $\protect\nu_1=1.25$ and $\protect\kappa _{1}=1$, drawn using the

777: software \texttt{Dynamics} \protect\cite{NY97}. The circle

778: ``$\bullet$'' represents the saddle at the origin.} \label{fig:3a}

779: \end{figure}

780: 

781: Let $N>0$ be a sufficiently large integer, and let

782: $(\bar{\chi}_{-N},\bar{\xi}_{-N})$ be a point on the unstable

783: subspace such that

784: $(\bar{\xi}_{N+1},\bar{\chi}_{N+1})=(\bar{\chi}_{-N},\bar{\xi}_{-N})$,

785: where

786: $(\bar{\chi}_{N+1},\bar{\xi}_{N+1})=f^{2N+1}(\bar{\chi}_{-N},\bar{\xi}_{-N})$.

787: By the reversibility of $f$, the point

788: $(\bar{\chi}_{N+1},\bar{\xi}_{N+1})$ must be contained in the

789: stable subspace. The two points $(\bar{\chi}_{-N},\bar{\xi}_{-N})$

790: and $(\bar{\chi}_{N+1},\bar{\xi}_{N+1})$ are also close to the

791: saddle when $N>0$ is large. Hence, the orbit leaving at

792: $(\bar{\chi}_{-N},\bar{\xi}_{-N})$ on the unstable subspace and

793: arriving at $(\bar{\chi}_{N+1},\bar{\xi}_{N+1})$ on the stable

794: subspace is a good approximation to a symmetric homoclinic orbit

795: of $f$. Using an adaptation of the driver \texttt{HomMap}

796: \cite{Y98a,Y98b} to \texttt{AUTO97}, we can easily find such

797: approximate homoclinic points.

798: 

799: Now, in order to find non-trivial symmetric homoclinic solutions

800: of $F_{\epsilon }$ for $\epsilon >0$, we take $\epsilon $ as the

801: additional free parameter and choose $(\chi _{-N},\xi

802: _{-N},\epsilon )=(\bar{\chi}_{-N},\bar{\xi}_{-N},0)$ as the

803: starting solution to (\ref{eqn:con1}) and (\ref{eqn:con2}), where

804: $(\bar{\chi}_{-N},\bar{\xi}_{-N})$ denotes the homoclinic point on

805: the unstable subspace for $f$, obtained using the above procedure.

806: Fixing $\kappa _{1}=1$, we then performed continuation of these

807: algebraic equations in \texttt{AUTO}, using $\delta$ as the

808: continuation parameter. To obtain symmetric homoclinic orbits for

809: $\kappa _{1}=-1$, we also take $\delta $ and $\kappa _{1}$ as the

810: free and continuation parameters, respectively, and continue the

811: solution obtained above for $\kappa _{1}=1$ and $\epsilon

812: =\sqrt{2/(D|\gamma _{1}|)}$. The results are presented in

813: Figs.~\ref{fig:3b}-\ref{fig:3d}. As shown in Figs.~\ref{fig:3b}

814: and \ref{fig:3c}, new branches bifurcate from the one with

815: $\epsilon=0$ at discrete values of $\delta$ and can be continued

816: to $\epsilon=\sqrt{2/(D|\gamma_{1}|)} $ by varying $\delta $ and

817: $\epsilon $. As shown in Fig.~\ref{fig:3d}, the branches were also

818: continued from $\kappa _{1}=1$ to $\kappa _{1}=-1$ by varying

819: $\delta $ and $\kappa_{1}$. For $\kappa _{1}=-1$ and $\kappa

820: _{2}=1$, we could not find such symmetric homoclinic orbits of

821: $F_{\epsilon }$ with sufficient precision. In these computations,

822: we also chose the value of $N$ such that the distance between the

823: approximate homoclinic point and the origin was $1.5\times

824: 10^{-3}$ at most, and checked that the results did not change

825: significantly under increase of $N$. Our computations also

826: suggested that a possibly infinite number of branches of symmetric

827: homoclinic orbits could be obtained when $k$ or $q$ is large or

828: $\delta $ is small. These branches show oscillations in the

829: parameter plane, which are sensitive to the value of $N$ and do

830: not appear to converge as $N\rightarrow \infty$. These, probably

831: spurious branches, result from a large ratio between the imaginary

832: eigenvalue of the linearization and the real eigenvalue, which is

833: well known to be a singular limit for equations that bear ES

834: \cite{Ch:01,Lo:01}, and needs to be treated with great caution. In

835: the results that follow, we have stopped computation at points

836: where such oscillations first become evident, and have checked all

837: results for consistency in the limit of $N\rightarrow \infty $.

838: 

839: \begin{figure}[tbp]

840: \begin{center}

841: \includegraphics[scale=0.7]{fig2a.eps}\quad \includegraphics[scale=0.7]{fig2b.eps}

842: \end{center}

843: \caption{Numerical continuation of symmetric homoclinic orbits for

844: $F_{\protect\epsilon }$ when $D=100$ and $N=85$: (a) $k=0.3$,

845: $q=5$ and $\protect\kappa _{1}=\protect\kappa _{2}=1$; (b)

846: $k=0.5$, $q=1$, $\protect\kappa _{1}=1$ and $\protect\kappa

847: _{2}=-1$. Here $\protect\epsilon $ and $\protect\delta $ are

848: varied. The dotted line represents $\protect\epsilon

849: =\protect\sqrt{2/(D|\protect\gamma _{1}|)}$ with $\protect\gamma

850: _{1}=0.05$. } \label{fig:3b}

851: \end{figure}

852: 

853: \begin{figure}[tbp]

854: \begin{center}

855: \includegraphics[scale=0.7]{fig3.eps}

856: \end{center}

857: \caption{Numerical continuation of symmetric homoclinic orbits for

858: $G_{\protect\epsilon }$ when $D=10$, $k=3$, $q=1$ and $N=15$. Here

859: $\protect\epsilon $ and $\protect\delta $ are varied. The dotted

860: line represents $\protect\epsilon =1$. } \label{fig:3c}

861: \end{figure}

862: 

863: \begin{figure}[tbp]

864: \begin{center}

865: \includegraphics[scale=0.7]{fig4.eps}

866: \end{center}

867: \caption{Numerical continuation of symmetric homoclinic orbits for

868: $F_{\protect\epsilon }$ when $D=100$, $k=0.5$, $q=1$,

869: $\protect\kappa _{2}=-1 $, $N=85$ and $\protect\epsilon

870: =0.6324555\approx \protect\sqrt{2/\left( D|\protect\gamma

871: _{1}|\right) }$ with $\protect\gamma _{1}=0.05$. Here

872: $\protect\delta $ and $\protect\kappa _{1}$ are varied. The dotted

873: line represents $\protect\kappa_{1}=-1$. } \label{fig:3d}

874: \end{figure}

875: 

876: \section{Numerical results}

877: 

878: We now present continuation results obtained with \texttt{AUTO}

879: for symmetric homoclinic orbits of $F_{\epsilon }$ (or

880: $G_{\epsilon }$) under simultaneous variation of two relevant

881: parameters within the saddle-center parameter regions. By varying

882: the discreteness coefficient $D$ up to large values, we are also

883: able to compare the results to those of the continuum limit,

884: $D\rightarrow \infty $. We shall treat the cases $\delta >0$ and

885: $\delta <0$ separately and also discuss the possibility of

886: discrete ESs in the absence of cubic nonlinearity.

887: 

888: \subsection{The case of\/ $\protect\delta >0$}

889: 

890: Figure~\ref{fig:4a} depicts branches of discrete ESs in the

891: presence of $\chi ^{(3)}$ terms, with $\gamma

892: _{1}=\gamma_{2}=0.05$ and $k=0.3$. Two distinct branches of ESs

893: are displayed. Along each branch, the profile changes

894: continuously. Figure~\ref{fig:4b} displays an array of profiles of

895: these ESs for $D=100$ and $D=5$. Note that the second

896: (higher-$\delta$) branch can be considered to be a family of

897: \emph{fundamental}, i.e.\ single-humped, solitons throughout the

898: range of existence, whereas the double-humped structure of the SH

899: component of the first branch becomes evident for small $q$. This

900: property, and the location of the branches in the $(\delta ,q)$

901: plane, are fully consistent with results obtained in the continuum

902: limit \cite[Fig.~3]{ChMaYaKa:01}. Our numerical results also

903: suggest that there may exist further branches for lower $\delta

904: $-values, each subsequent one containing more oscillations in core

905: of the SH component. However, computation becomes unreliable

906: beyond the first two branches for the reasons given at the end of

907: the last section, and so we do not display those results here.

908: Also experience from the continuum model suggests that

909: non-fundamental ESs never have a chance to be stable.

910: 

911: \begin{figure}[tbp]

912: \begin{center}

913: \includegraphics[scale=0.7]{fig5a.eps}\quad \includegraphics[scale=0.7]{fig5b.eps}

914: \end{center}

915: \caption{Branches of the discrete ES solutions for $k=0.3$ and

916: $\protect\gamma _{1}=\protect\gamma _{2}=0.05$. (a) Solutions in

917: the $(\protect\delta ,q)$-plane for $(D,N)=(100,86)$ (solid line)

918: $(10,31)$ (dashed line) and $(5,23)$ (broken line). According to

919: (\protect\ref{area_scale}) ESs also exist only above the dotted

920: line $q=0.6$ for this $k$-value. Labeled points correspond to the

921: relevant subpanels of Fig.~\protect\ref{fig:4b} where solutions

922: profiles are displayed. (b) Solutions in the $(D,\protect\delta

923: )$-plane for $q=5$ and $N=22$. Note that, according to

924: (\protect\ref{area_scale}), we must have $\protect\delta D>2.2$

925: (the boundary is shown as a dotted line). } \label{fig:4a}

926: \end{figure}

927: 

928: \begin{figure}[tbp]

929: \begin{center}

930: \includegraphics[scale=0.55]{fig6a.eps}\qquad \includegraphics[scale=0.55]{fig6b.eps}\\[2ex]

931: \includegraphics[scale=0.55]{fig6c.eps}\qquad \includegraphics[scale=0.55]{fig6d.eps}\\[2ex]

932: \includegraphics[scale=0.55]{fig6e.eps}\qquad \includegraphics[scale=0.55]{fig6f.eps}\\[2ex]

933: \includegraphics[scale=0.55]{fig6g.eps}\qquad \includegraphics[scale=0.55]{fig6h.eps}

934: \end{center}

935: \caption{Profiles of the discrete ESs corresponding to the labeled

936: points in Fig.~\protect\ref{fig:4a}(a): (a) $\protect\delta

937: =0.58637 $ and $q=5$; (b) $\protect\delta =2.9894$ and $q=5$; (c)

938: $\protect\delta =0.65467$ and $q=0.6$; (d) $\protect\delta

939: =3.3693$ and $q=0.6$; (e) $\protect\delta =1.3496 $ and $q=5$; (f)

940: $\protect\delta =3.1432$ and $q=5$; (g) $\protect\delta =0.80213$

941: and $q=0.6$; (h) $\protect\delta =3.3824$ and $q=0.6$. In

942: panels~(a)-(d), $D=100$ and $N=85$, and in panels (e)-(h), $D=5$

943: and $N=22$. In this and all subsequent plots, the FF

944: ($u$-component) is interpolated by a solid line and the SH

945: ($v$-component) by a broken line. } \label{fig:4b}

946: \end{figure}

947: 

948: Figure \ref{fig:4a}(b) depicts continuation in $D$ of the solutions on these

949: two branches for $q=5$. In this case, we find that there is a lower limit on

950: $D$ beyond which no ESs exist. This corresponds to the right-hand inequality

951: in the first condition in (\ref{area_scale}).

952: %This is symptomatic of the fact that in the anticontinuum limit

953: %($D\rightarrow 0$) only a narrow band of $q$ values is allowed, as

954: %shown by the asymptotic analysis (\ref{qinterval}).

955: 

956: We will now check the validity of the analytical approximation

957: developed in Section~3 when $\tilde{k}$ and hence $k$ are large.

958: In Fig.~\ref{fig:4b1}, we display the two numerically computed ES

959: branches with the same values of $\gamma _{1}$ and $\gamma _{2}$

960: as those in Fig.~\ref{fig:4a} for $D=10$ when $k$ is rather large,

961: and compare them with the analytical prediction (\ref{final}),

962: taking into regard the relations (\ref{rescaledpars}). In

963: Fig.~\ref{fig:4b1}(a), the parameters $\delta $ and $k$ are varied

964: with $q=2k$, and in Fig.~\ref{fig:4b1}(b) the parameters $\delta $

965: and $q$ are varied for $k=50$. The analytical predictions are

966: plotted as dashed lines. We find good agreement between the

967: numerical and analytical results, especially for large $k$.

968: Figure~\ref{fig:4b2} displays the profiles of these ESs. The

969: fundamental soliton in Fig.~\ref{fig:4b2}(a) has a steep peak at

970: $n=0$, as assumed in the approximate analysis of Section~3.

971: 

972: \begin{figure}[tbp]

973: \begin{center}

974: \includegraphics[scale=0.7]{fig7a.eps}\quad \includegraphics[scale=1.23]{fig7b.eps}

975: \end{center}

976: \caption{Branches of discrete ESs for $\protect\gamma

977: _{1}=\protect\gamma_{2}=0.05$ and $D=10$; $N=7$ is chosen for

978: larger $\protect\delta$, and $N=5$ for smaller $\protect\delta $.

979: The analytical prediction given by Eqs.~(\protect\ref{final}) and

980: (\protect\ref{rescaledpars}) is plotted as a dashed line. (a)

981: Solutions in the $(\protect\delta,k)$-plane for $q=2k$. (b)

982: Solutions in the $(\protect\delta,q)$-plane for $k=50$. The ESs

983: exist in the region $100<q<100+10\protect\delta $, whose boundary

984: is shown as a dotted line.} \label{fig:4b1}

985: \end{figure}

986: 

987: \begin{figure}[tbp]

988: \begin{center}

989: \includegraphics[scale=0.55]{fig8a.eps}\qquad \includegraphics[scale=0.55]{fig8b.eps}

990: \end{center}

991: \caption{Examples of the ESs on solution branches in

992: Fig.~\protect\ref{fig:4b1}(b): (a) $\protect\delta =0.34718$,

993: $q=102$ and $N=7$; (b) $\protect\delta =0.0029014$, $q=100.02$ and

994: $N=5$. } \label{fig:4b2}

995: \end{figure}

996: 

997: It is well known that in the continuum case multi-pulse homoclinic

998: orbits exist (under a mild \emph{Birkhoff signature} condition

999: \cite{MiHoOR:92,KoChBuSa:02}) along families of curves in the

1000: parameter plane that accumulate on the curves of the fundamental

1001: ES solutions. Unlike the branches computed above, they do not

1002: feature a single-humped shaped in either component, but rather

1003: look like bound-states of two spatially separated fundamental

1004: solitons. We have found exactly the same solution families in the

1005: discrete model too. An example is presented in

1006: Fig.~\ref{bound-state}.

1007: 

1008: \begin{figure}[tbp]

1009: \begin{center}

1010: \includegraphics[scale=0.55]{fig9.eps}

1011: \end{center}

1012: \caption{A typical example of a two-pulse bound state ES in the

1013: lattice system (\protect\ref{eqn:disc}) with

1014: $\protect\delta=1.1261$, $\protect\gamma _{1}=\protect\gamma

1015: _{2}=0.05$, $D=5$, $k=0.3$, $q=7$ and $N=51$.} \label{bound-state}

1016: \end{figure}

1017: 

1018: \subsection{The case of\/ $\protect\delta <0$}

1019: 

1020: Figure~\ref{fig:4c} shows ES branches in the case of

1021: self-defocusing $\chi ^{(3)}$ terms, with $\gamma _{1}=\gamma

1022: _{2}=-0.05$ and $\delta <0$. Figures~\ref{fig:4d} and \ref{fig:4e}

1023: display the profiles of these ESs for $D=100$, $D=10$ and $D=6$.

1024: Note that these curves and the solutions on them for $D=100 $ are

1025: identical, to the accuracy depicted, to the corresponding ones

1026: found in the continuum counterpart of the model in Ref.

1027: \cite[Figs.~2,3]{YaMaKaCh:01}.

1028: \begin{figure}[tbp]

1029: \begin{center}

1030: \includegraphics[scale=0.7]{fig10a.eps}\quad \includegraphics[scale=0.7]{fig10b.eps}

1031: \end{center}

1032: \caption{Branches of ESs in the discrete system

1033: (\protect\ref{eqn:lattice}) with $q=1$ and $\protect\gamma

1034: _{1}=\protect\gamma _{2}=-0.05$: (a) In the $(\protect\delta

1035: ,k)$-plane for $D=100$ and $10$; (b) in the $(D,k)$-plane for

1036: $\protect\delta =-0.5$ and $N=30$. In panel~(a), the solid and

1037: dashed curves represent the results for $D=100$ and $10$,

1038: respectively. When $D=100$ (resp. $D=10$), $N=85$ (resp. $N=30$)

1039: were used for the most part but $N=93$ or $N=109$ (resp. $N=35$ or

1040: $N=40$) were used for the two outer curves with large

1041: $|\protect\delta |$. According to condition

1042: (\protect\ref{area_scale}), ESs exist only above $k=0.5$ and below

1043: $k=0.5-\protect\delta D$, when $q=1 $. } \label{fig:4c}

1044: \end{figure}

1045: Notice the variety of multi-humped shapes of the solitons

1046: belonging to these families. Like the continuum model, only the

1047: inner-most of these branches represents a fundamental soliton.

1048: Continuation towards the anti-continuum limit, $D\rightarrow 0$,

1049: becomes numerically problematic for these solitons. It was found

1050: difficult to compute the solutions with repeatable accuracy while

1051: varying $N$ below $D\approx 6$. Clearly, these branches become

1052: increasingly spiky as $D$ is decreased (see Fig.~\ref{fig:4e}),

1053: and it may happen that branches of ES solutions actually terminate

1054: before they reach the minimum value of $D$ at which they remain

1055: embedded, which would be $D_{\min}=2k-1$, for the values of $q$

1056: and $\delta $ used in Fig.~\ref{fig:4c}(b).

1057: 

1058: \begin{figure}[tbp]

1059: \begin{center}

1060: \includegraphics[scale=0.55]{fig11a.eps}\qquad \includegraphics[scale=0.55]{fig11b.eps}\\[2ex]

1061: \includegraphics[scale=0.55]{fig11c.eps}\qquad \includegraphics[scale=0.55]{fig11d.eps}\\[2ex]

1062: \includegraphics[scale=0.55]{fig11e.eps}\qquad \includegraphics[scale=0.55]{fig11f.eps}\\[2ex]

1063: \includegraphics[scale=0.55]{fig11g.eps}\qquad \includegraphics[scale=0.55]{fig11h.eps}

1064: \end{center}

1065: \caption{ESs on the branches in Fig.~\protect\ref{fig:4c}(a) for

1066: $\protect\delta =-1$: (a) $k=0.69564$; (b) $k=0.71312$; (c)

1067: $k=0.71383$; (d) $k=0.71387$; (e) $k=0.6895$; (f) $k=0.70811$; (g)

1068: $k=0.70898$; (h) $k=0.70902$. In panels (a)-(d), $D=100$ and

1069: $N=85$, and in panels~(e)-(h), $D=10$ and $N=30$. } \label{fig:4d}

1070: \end{figure}

1071: 

1072: \begin{figure}[tbp]

1073: \begin{center}

1074: \includegraphics[scale=0.55]{fig12a.eps}\quad \includegraphics[scale=0.55]{fig12b.eps}

1075: \end{center}

1076: \caption{ESs at the end of the above two branches in

1077: Fig.~\protect\ref{fig:4c}(b) for $D=6$: (a) $k=0.78427$; (b)

1078: $k=0.78426$. } \label{fig:4e}

1079: \end{figure}

1080: 

1081: Figure~\ref{fig:4f} shows the ES branches in a still more exotic

1082: case of opposite signs in front of the FF and SH $\chi ^{(3)}$ SPM

1083: terms, $\gamma _{1}=0.05$ and $\gamma _{2}=-0.05$. This case may,

1084: in principle, also be physically relevant -- not to ordinary

1085: optical materials, but rather to photonic crystals (see, e.g.,

1086: \cite{PhotCryst} and references therein). Figure~\ref{fig:4g}

1087: displays the profiles of these ESs for $D=100$ and $D=10$. Note

1088: similarity with the branches in Fig.~\ref{fig:4c} for small

1089: $\delta $. However, in this case it is found that the solution

1090: branches still exist for large $k$, rather than undergoing turning

1091: back with the increase of $k$. Also the internal oscillations on

1092: the non-fundamental branches become far less pronounced.

1093: 

1094: \begin{figure}[tbp]

1095: \begin{center}

1096: \includegraphics[scale=0.7]{fig13a.eps}\quad \includegraphics[scale=0.7]{fig13b.eps}

1097: \end{center}

1098: \caption{Branches of ESs with $q=1$, $\protect\gamma _{1}=0.05$

1099: and $\protect\gamma _{2}=-0.05$: (a) In the $(\protect\delta

1100: ,k)$-plane for $D=100$ and $10$; (b) in the $(D,k)$-plane for

1101: $k=1$ and $N=24$. In panel~(a), the solid and dashed curves

1102: represent the results for $D=100$ and $10$, respectively. When

1103: $D=100$ (resp. $D=10$), $N=85$ (resp. $N=30$) was used. In

1104: panel~(b), the forth solid line from the bottom still exists

1105: although it is very short and almost invisible. The ESs exist only

1106: in the region $0.5<k<0.5-\protect\delta D$ when $q=1$, whose

1107: boundary is indicated by a dotted line. } \label{fig:4f}

1108: \end{figure}

1109: 

1110: \begin{figure}[tbp]

1111: \begin{center}

1112: \includegraphics[scale=0.55]{fig14a.eps}\quad \includegraphics[scale=0.55]{fig14b.eps}\\[2ex]

1113: \includegraphics[scale=0.55]{fig14c.eps}\quad \includegraphics[scale=0.55]{fig14d.eps}\\[2ex]

1114: \includegraphics[scale=0.55]{fig14e.eps}\quad \includegraphics[scale=0.55]{fig14f.eps}\\[2ex]

1115: \includegraphics[scale=0.55]{fig14g.eps}\quad \includegraphics[scale=0.55]{fig14h.eps}

1116: \end{center}

1117: \caption{Examples of ESs on the branches in

1118: Fig.~\protect\ref{fig:4f}(a) for $k=1$: (a) $\protect\delta

1119: =-0.23601$; (b) $\protect\delta =-0.099245$; (c) $\protect\delta

1120: =-0.053237$; (d) $\protect\delta =-0.033776$; (e) $\protect\delta

1121: =-0.24838$; (f) $\protect\delta =-0.13304$; (g) $\protect\delta

1122: =-0.096003$; (h) $\protect\delta =-0.05109$. In panels (a)-(d),

1123: $D=100$ and $N=85$; in panels (e)-(h), $D=10$ and $N=30$. }

1124: \label{fig:4g}

1125: \end{figure}

1126: 

1127: The approximate analysis of Section~3 is also valid for $\gamma

1128: _{1}>0$ and $\gamma _{2},\,\delta <0$ when $k$ is large.

1129: Figure~\ref{fig:4h} shows the fundamental ES branch (the left one

1130: in Fig.~\ref{fig:4f}(a)) with the same values of $\gamma _{1}$ and

1131: $\gamma _{2}$ as those in Fig.~\ref{fig:4f} for $D=10$ when $k$ is

1132: rather large. In Fig.~\ref{fig:4h}(a), the parameters $\delta $

1133: and $k$ are varied for $q=2k$, and in Fig.~\ref{fig:4h}(b) the

1134: parameters $\delta $ and $q$ are varied for $k=50$, as in

1135: Figs.~\ref{fig:4b1}(a) and (b). The predictions based on Eqs.

1136: (\ref{final}) with (\ref{rescaledpars}) are plotted as dashed

1137: lines. A profile of the ES is also displayed in

1138: Fig.~\ref{fig:4h}(a). We see that Eq.~(\ref{final}) again

1139: approximates well the numerical result for the fundamental

1140: solitons when $k$ is large. The ES in Fig.~\ref{fig:4h}(a)

1141: features a steep peak at $n=0$, as assumed in the analytical

1142: approximation.

1143: 

1144: \begin{figure}[t]

1145: \begin{center}

1146: \includegraphics[scale=0.7]{fig15a.eps}\quad \includegraphics[scale=1.23]{fig15b.eps}

1147: \end{center}

1148: \caption{Branches of discrete ESs for $\protect\gamma _{1}=0.05$,

1149: $\protect\gamma _{2}=-0.05$, $D=10$ and $N=7$. The analytical

1150: prediction, given by Eqs.~(\protect\ref{final}) and

1151: (\protect\ref{rescaledpars}) is plotted by dashed lines. (a)

1152: Solutions in the $(\protect\delta ,k)$-plane for $q=2k$. (b)

1153: Solutions in the $(\protect\delta ,q)$-plane for $k=50$. The ESs

1154: exist in the region $100+10\protect\delta<q<100$, whose boundary

1155: is shown by the dotted line. An example of the ES, for

1156: $\protect\delta =-0.31873$ and $q=98$, is plotted in panel~(b). }

1157: \label{fig:4h}

1158: \end{figure}

1159: 

1160: Finally, we briefly discuss the case in which $\gamma _{1}=\gamma

1161: _{2}=0$, i.e., the $\chi ^{(3)}$ terms are absent. As shown in

1162: Fig.~\ref{fig:4i}, we could follow solutions of the

1163: three-dimensional algebraic problem for $G_{\epsilon}$. However,

1164: these solutions do not represent ESs since the origin is not a

1165: saddle-center but a hyperbolic saddle. Although the region

1166: $q/2<k<q/2-\delta D$ in which the origin is a saddle-center

1167: becomes wider when $D$ is larger, for the solution $k$ diverges to

1168: $\infty$ as $D\rightarrow \infty$. Thus, there seems to be no hope

1169: that an ES exists in this case. As mentioned above, this situation

1170: is very similar to that in the continuum model (\ref{eqn:PDE}).

1171: 

1172: \begin{figure}[t]

1173: \begin{center}

1174: \includegraphics[scale=1.23]{fig16a.eps}\quad \includegraphics[scale=0.7]{fig16b.eps}

1175: \end{center}

1176: \caption{Branches of solutions of the three-dimensional algebraic problem

1177: for $G_{\protect\epsilon }$ in the absence of the $\protect\chi ^{(3)}$

1178: nonlinearity, with $\protect\gamma _{1}=\protect\gamma _{2}=0$, and $q=1$.

1179: (a) In the $(\protect\delta ,k)$-plane for $D=10$; (b) in the $(D,k)$-plane

1180: for $\protect\delta =-0.1$ and $N=15$. In panel~(a), $N=15$ and $25$ were

1181: used for $k>1$ and $k<1$, respectively. The ESs would exist only in the

1182: region $0.5<k<0.5-\protect\delta D$, the boundaries of which are plotted as

1183: dotted lines, when $q=1$. An example of a \emph{non-embedded} (gap) soliton,

1184: found in this case for $\protect\delta =-0.1$ and $k=1.9027$, is plotted in

1185: panel~(a).}

1186: \label{fig:4i}

1187: \end{figure}

1188: 

1189: \section{Conclusions}

1190: 

1191: In this paper, we have established the existence of embedded

1192: solitons (ESs) in the discrete model of the second-harmonic

1193: generation in the presence of cubic nonlinearity. The model can be

1194: naturally realized as an array of channel waveguides in the

1195: spatial domain, therefore our results suggest possibilities for

1196: new experiments with discrete spatial solitons in nonlinear

1197: optics. We have also introduced a simplified model, with the $\chi

1198: ^{(3)}$ nonlinearity present solely at the central site, in which

1199: the existence of an ES was demonstrated in an asymptotic

1200: analytical form. In the general case, the existence region for the

1201: ESs in the discrete model was found numerically. Moreover, we have

1202: checked that the asymptotic analysis of the simplified model in

1203: the limit of large wavenumbers gives a good approximation of the

1204: codimension-one set in parameter space, on which the ESs exist.

1205: 

1206: More generally, we have established, that unlike the

1207: discretizations of other (dissipative) continuum models bearing

1208: localized solutions, the codimension of ESs does not change when

1209: one passes to a discrete version. Whereas continuum ESs may be

1210: regarded as homoclinic orbits to saddle-center equilibria in

1211: reversible flows (ordinary differential equations), discrete ESs

1212: should be thought of as homoclinic orbits to saddle-center fixed

1213: points of reversible maps. Both have codimension one in the

1214: parameter space. Understanding this property led us to the

1215: derivation of a numerical method for computing homoclinic orbits

1216: to nonhyperbolic equilibria of reversible maps.

1217: 

1218: Accurate investigation of stability of ESs in the lattice model is

1219: beyond the scope of the present investigation. Systematic results

1220: for the stability will be presented elsewhere; however, some

1221: preliminary results suggest that the fundamental discrete ESs

1222: inherit the \textit{semi-stability} property found in the

1223: corresponding continuous model \cite{YaMaKaCh:01,PeYa:02}. We

1224: expect that general arguments in favor of the semi-stable

1225: character of these solitons for the continuous case should apply

1226: in the discrete setting as well.

1227: 

1228: Figure~\ref{fig:7a} shows a preliminary computational result for

1229: $D=10$, $\delta =-1$, $k=0.6895$, $q=1$ and $\gamma _{1}=\gamma

1230: _{2}=-0.05$, corresponding to the fundamental ES of

1231: Fig.~\ref{fig:4d}(e). Here Eq.~(\ref{eqn:lattice}) was integrated

1232: numerically by the fourth-order Runge-Kutta method under the

1233: boundary condition

1234: \begin{equation}

1235: u_{-\bar{N}-1}(z)=u_{\bar{N}+1}(z)=v_{-\bar{N}-1}(z)=v_{\bar{N}+1}(z)=0

1236: \end{equation}for all $z\geq 0$ with $\bar{N}=93$ and the initial condition

1237: \begin{equation}

1238: u_{n}(0)=(1+c_{1})U_{n},\quad v_{n}(0)=(1+c_{2})V_{n}  \label{eqn:init}

1239: \end{equation}where $(U_{n},V_{n})$ represents an approximate ES given by the data of

1240: Fig.~\ref{fig:4d}(e) for $|n|\leq N=30$ and by $U_{n}=V_{n}=0$ for $|n|>N$, and $c_{1,2}$

1241: are small amplitudes of the initial disturbance, which were chosen

1242: to be $c_{1}=c_{2}=0.01$. The positive values of $c_{1,2}$ mean that the

1243: norm

1244: \begin{equation}

1245: E=\sum_{n=-\bar{N}}^{\bar{N}}(|u_{n}|^{2}+2|v_{n}|^{2})

1246: \end{equation}of the perturbed state  (in the temporal-domain optical model, it plays the

1247: role of energy) is greater than that of the unperturbed ES. In

1248: Fig.~\ref{fig:7a} we see the characteristic hallmark of

1249: semi-stability for the ES: The shapes of the perturbed wave are

1250: almost unchanged for a long period of $t $ in

1251: Figs.~\ref{fig:7a}(a) and(b), and the amplitudes $|u_{0}(t)|$ and

1252: $|v_{0}(t)|$ exhibit small oscillations near the unperturbed

1253: values in Figs.~\ref{fig:7a}(c) and(d). However, the perturbed ES

1254: was destroyed in the simulations when different signs of $c_{1}$

1255: and $c_{2}$ were chosen. Further investigation of stability and

1256: bifurcation will be the subject of future work.

1257: 

1258: \begin{figure}[t]

1259: \begin{center}

1260: \includegraphics[scale=0.8]{fig17a.eps}\quad \includegraphics[scale=0.8]{fig17b.eps}\\[3ex]

1261: \includegraphics[scale=0.6]{fig17c.eps}\quad \includegraphics[scale=0.6]{fig17d.eps}\\[0pt]

1262: \end{center}

1263: \caption{Evolution of the fundamental ES (corresponding to

1264: Fig.~\protect\ref{fig:4d}(e)) initiated by the nondestructive

1265: perturbation with $c_{1}=c_{2}=0.01$ in the initial conditions

1266: (\protect\ref{eqn:init}). The dashed lines in panels~(c) and (d)

1267: are the amplitudes of the $u_{0}$- and $v_{0}$-components of the

1268: unperturbed soliton.} \label{fig:7a}

1269: \end{figure}

1270: 

1271: Finally, our results so far pertain only to those discrete ESs

1272: that are symmetric with respect to the involution $R$ defined in

1273: Eq.~(\ref{R1}); solitons with this symmetry have an on-site

1274: central peak. It is also possible to apply techniques developed in

1275: this work to solutions symmetric with respect to involution

1276: $\hat{R}$, see Eq.~(\ref{R2}), that should feature an inter-site

1277: central peak. In other physical context, such waves are less

1278: likely to be stable than waves that are centered on a lattice site

1279: \cite{Review}. However, that understanding typically applies to

1280: regular (gap) discrete solitons and need not necessarily apply to

1281: ESs. We shall address this issue in future work.

1282: 

1283: \section*{Acknowledgements}

1284: 

1285:  K.Y. acknowledges support from the Japanese Society for

1286: the Promotion of Science, which enabled him to stay in Bristol and

1287: to perform this work. B.A.M. appreciates support of EPSRC

1288: ``critical mass" grant and mathematics programme and hospitality

1289: of the Department of Engineering Mathematics at the University of

1290: Bristol.

1291: 

1292: \section*{References}

1293: 

1294: %\bibliography{embed}

1295: 

1296: \begin{thebibliography}{99}

1297: \bibitem{YaMaKa:99} Yang J, Malomed B A and Kaup D J 1999 Embedded solitons

1298: in second-harmonic-generating systems \textit{Phys. Rev. Lett.} \textbf{83}

1299: 1958--1961

1300: 

1301: \bibitem{ChMaYaKa:01} Champneys A R, Malomed B A, Yang J and Kaup D J 2001

1302: Embedded solitons: Solitary waves in resonance with the linear spectrum

1303: \textit{Physica D} \textbf{152-153} 340--354

1304: 

1305: \bibitem{YaMaKaCh:01} Yang J, Malomed B A, Kaup D J and Champneys A R 2001

1306: Embedded solitons: A new type of solitary wave \textit{Math. Comp. Sim}

1307: \textbf{56} 585--600

1308: 

1309: \bibitem{ChGr:97} Champneys A R and Groves M D 1997 A global investigation

1310: of solitary waves solutions to a two-parameter model for water

1311: waves \textit{J. Fluid. Mech.} \textbf{342} 299--229

1312: 

1313: \bibitem{FuEs:97} Fujioka J and Espinosa A 1997 Soliton-like solution of an

1314: extended NLS equation existing in resonance with linear dispersive waves

1315: \textit{J. Phys. Soc. Jpn.} \textbf{66} 2601--2607

1316: 

1317: \bibitem{ChMaFr:98} Champneys A R, Malomed B A and M. J.~Friedman 1998

1318: Thirring solitons in the presence of dispersion \textit{Phys. Rev. Lett.}

1319: \textbf{80} 4168--4171

1320: 

1321: \bibitem{AtMa:01} Atai J and Malomed B A 2001 Solitary waves in systems with

1322: separated Bragg grating and nonlinearity \textit{Phys. Rev. E} \textbf{64}

1323: 066617

1324: 

1325: \bibitem{KoChBuSa:02} Kolossovski K, Champneys A R, Buryak A and R. A.

1326: Sammut 2002 Multi-pulse embedded solitons as bound states of quasi-solitons

1327: \textit{Physica D} \textbf{171} 153--177

1328: 

1329: \bibitem{EsFuGo:03} Espinosa-Ceron A, Fujioka J and Gomez-Rodriguez A 2003

1330: Embedded solitons: Four-frequency radiation, front propagation and radiation

1331: inhibition \textit{Physica Scripta} \textbf{67} 314--324

1332: 

1333: \bibitem{Ya:03} Yang J K 2003 Stable embedded solitons \textit{Phys. Rev.

1334: Lett.} \textbf{91} 143903

1335: 

1336: \bibitem{Mak} Mak W C K, Malomed B A and Chu P L 2004 Symmetric and

1337: asymmetric solitons in linearly coupled Bragg gratings \textit{Phys. Rev.} E

1338: 69 066610

1339: 

1340: \bibitem{Merhasin} Merhasin I M and Malomed B A 2004 Four-wave solitons in

1341: Bragg cross-gratings \textit{J. Optics B: Quant. Semiclass. Opt.} 6 S323-S332

1342: 

1343: \bibitem{PeYa:02} Pelinovsky D E and Yang J 2002 A normal form for nonlinear

1344: resonance of embedded solitons, \textit{Proc. Roy. Soc. Lond. A} \textbf{458}

1345: 1469--1497

1346: 

1347: \bibitem{ChKi:99} Champneys A R and Kivshar Yu S 2000 Origin of multikinks

1348: in dispersive nonlinear systems \textit{Phys. Rev. E} \textbf{61} 2551--2554

1349: 

1350: \bibitem{Flach} Flach S, Zolotaryuk Y and Kladko K 1999 Moving lattice kinks

1351: and pulses: An inverse method \textit{Phys. Rev E} \textbf{59} 6105-6115

1352: 

1353: \bibitem{Morgante} Morgante A M, Johansson M, Kopidakis G and Aubry S 2002

1354: Standing wave instabilities in a chain of nonlinear coupled oscillators

1355: \textit{Physica D} \textbf{162 }53-94

1356: 

1357: \bibitem{PeKr:84} Peyrard M and Kruskal M D 1984 Kink dynamics in the highly

1358: discrete {sine-Gordon} system \textit{Physica D} \textbf{14} 88--102

1359: 

1360: \bibitem{KiBr:98} Braun O M and Kivshar Yu S 1998 Nonlinear dynamics of the

1361: {Frenkel-Kontorova} model \textit{Phys. Rep.} \textbf{306} 1--108

1362: 

1363: \bibitem{SaZoEi:00} Savin A V. Zolotaryuk Y and Eilbeck J C 2000 Moving

1364: kinks and nanopterons in the nonlinear {Klein-Gordon} lattice

1365: \textit{Physica D} \textbf{138} 267--281

1366: 

1367: \bibitem{AiChRo:03} Aigner A A, Champneys A R and Rothos V R 2003 A new

1368: barrier to the existence of moving kinks in Frenkel-Kontorova lattices

1369: \textit{Physica D} \textbf{186} 148--170

1370: 

1371: \bibitem{ESdiscrete} Gonzalez-Perez-Sandi S, Fujioka J and Malomed B A 2004

1372: Embedded solitons in dynamical lattices \textit{Physica D} \textbf{197}

1373: 86--100

1374: 

1375: %%\bibitem{Scott} Scott, A 2003 \textit{Nonlinear Science: Emergence and

1376: %%Dynamics of Coherent Structures} 2nd ed (Oxford: Oxford University Press)

1377: 

1378: \bibitem{FiSc} Fiedler B and Scheurle J 1996 \textit{Discretization of

1379: Homoclinic Orbits, Rapid Forcing and ``Invisible'' Chaos} Memoirs of Amer.

1380: Math. Soc. 119 (Providence RI: American Mathematical Society)

1381: 

1382: \bibitem{Bang1} Bang O, Clausen C B, Christiansen P L and Torner L 1999

1383: Engineering competing nonlinearities \textit{Opt. Lett.} \textbf{24}

1384: 1413--1415

1385: 

1386: \bibitem{Bang2} Corney J F and Bang O 2001 Solitons in quadratic nonlinear

1387: photonic crystals \textit{Phys. Rev. E} \textbf{64} 047601

1388: 

1389: \bibitem{Bang3} Johansen S K, Carrasco S, Torner L and Bang O 2002

1390: Engineering of spatial solitons in two-period QPM structures \textit{Opt.

1391: Commun.} \textbf{203} 393--402

1392: 

1393: \bibitem{Review} Etrich C, Lederer F, Malomed B A, Peschel T, and Peschel U

1394: 2000 Optical solitons in media with a quadratic nonlinearity \textit{Progr.

1395: Opt.} \textbf{41} 483-568

1396: 

1397: \bibitem{DiscrChi2soliton} Iwanow R, Schiek R, Stegeman G I, Pertsch T,

1398: Lederer F, Min Y and Sohler W 2004 Observation of discrete quadratic

1399: solitons \textit{Phys. Rev. Lett.} \textbf{93} 113902

1400: 

1401: \bibitem{PanosZiad} Kevrekidis P G Malomed B A, and Musslimani Z 2003

1402: Discrete gap solitons in a diffraction-managed waveguide array \textit{Eur.

1403: Phys. J. D} \textbf{67} 013605

1404: 

1405: \bibitem{DiffrManagement1} Eisenberg H S, Silberberg Y, Morandotti R, Boyd A

1406: and Aitchison J S 2000 Diffraction management \textit{Phys. Rev. Lett.}

1407: \textbf{85} 1863--1866

1408: 

1409: \bibitem{DiffrManagement2} Pertsch T, Zentgraf T, Peschel U and Lederer F

1410: 2002 Anomalous refraction and diffraction in discrete optical systems

1411: \textit{Phys. Rev. Lett.} \textbf{88} 093901

1412: 

1413: \bibitem{De:76b} Devaney R L 1976 Reversible diffeomorphisms and flows

1414: \textit{Trans. Amer. Math. Soc.} \textbf{218} 89--113

1415: 

1416: \bibitem{LaRo:98} Lamb J S W and Roberts J A G 1998 Time-reversal symmetry

1417: in dynamical systems: A survey \textit{Physica D} \textbf{112} 1--39

1418: 

1419: \bibitem{GH83} Guckenheimer J and Holmes P 1983 \textit{Nonlinear

1420: Oscillations, Dynamical Systems, and Bifurcations of Vector Fields} (New

1421: York: Springer)

1422: 

1423: \bibitem{Auto} Doedel E, Champneys A R, Fairgrieve T F, Kuznetsov Y A,

1424: Sandstede B and Wang X 1997 \textit{AUTO97: Continuation and

1425: Bifurcation Software for Ordinary Differential Equations (with

1426: HomCont)} Available by anonymous ftp from

1427: \texttt{ftp.cs.concordia.ca}, directory \texttt{pub/doedel/auto}

1428: 

1429: \bibitem{NY97} Nusse E H and Yorke J A 1997 \textit{Dynamics: Numerical

1430: Explorations} 2nd ed. (New York: Springer)

1431: 

1432: \bibitem{K81} Kawakami H 1981 Algorithme num\'{e}rique d\'{e}finissant la

1433: bifurcation d'un point homocline \textit{C. R. Acad. Sc. Paris, S\'{e}rie I}

1434: \textbf{293} 401--403

1435: 

1436: \bibitem{BK97a} Beyn W-J and Kleinkauf J N 1997 The numerical computation of

1437: homoclinic orbits for maps \textit{SIAM J. Num. Anal.} \textbf{34} 1207--1236

1438: 

1439: \bibitem{BK97b} Beyn W-J and Kleinkauf J N 1997 Numerical approximation of

1440: homoclinic chaos \textit{Numer. Algorithms} \textbf{14} 25--53

1441: 

1442: \bibitem{Y98a} Yagasaki K 1998 Numerical detection and continuation of

1443: homoclinic points and their bifurcations for maps and periodically forced

1444: systems \textit{Int. J. Bifurcation Chaos} \textbf{7} 1617--1627

1445: 

1446: \bibitem{BBV00} Bergamin J M, Bountis T and Vrahatis M N 2002 Homoclinic

1447: orbits of invertible maps \textit{Nonlinearity} \textbf{15} 1603--1619

1448: 

1449: \bibitem{Y98b} Yagasaki K 1998 \textit{HomMap: An Auto driver for homoclinic

1450: bifurcation analysis of maps and periodically forced systems} (Gifu Japan:

1451: Gifu University)

1452: 

1453: \bibitem{Lo:01} Lombardi E 2000 \textit{Oscillatory Integrals and Phenomena

1454: beyond All Algebraic Orders: With Applications to Homoclinic Orbits in

1455: Reversible Systems} (Berlin: Springer)

1456: 

1457: \bibitem{Ch:01} Champneys A R 2001 Codimension-one persistence beyond all

1458: orders of homoclinic orbits to singular saddle centers in

1459: reversible systems \textit{Nonlinearity} \textbf{14} 87--112

1460: 

1461: \bibitem{MiHoOR:92} Mielke A, Holmes P and O'Reilly O 1992 Cascades of

1462: homoclinic orbits to, and chaos near, a Hamiltonian saddle-center \textit{J.

1463: Dyn. Diff. Eqns.} \textbf{4} 95--126

1464: 

1465: \bibitem{PhotCryst} Andreani L C, Agio M, Bajoni D, Belotti M, Galli M,

1466: Guizzetti G, Malvezzi A M, Marabelli F, Patrini M and Vecchi G 2003

1467: Morphology and optical properties of bare and polydiacetylenes-infiltrated

1468: opals \textit{Synthetic Metals} \textbf{139} 695--700

1469: \end{thebibliography}

1470: 

1471: \end{document}

1472: