1: \documentclass[12pt]{iopart}

2: %\usepackage{iopams}

3: \usepackage{indentfirst}

4: \usepackage{graphicx,color}

5: \usepackage{subfigure}

6: \usepackage{psfrag}

7: \usepackage[hang, small, bf, margin=20pt, tableposition=top]{caption}

8: \setlength{\abovecaptionskip}{0pt}

9: \usepackage[T1]{fontenc}

10: \usepackage[square, comma, numbers, sort&compress]{natbib}

11: \def\newblock{\hskip .11em plus .33em minus .07em}

12: \newfont{\bbold}{msbm10 scaled\magstep1}

13: \newcommand{\bbf}[1]{\mbox{{\bbold #1}}}

14: 

15: % ABBREVIATIONS

16: \newcommand{\spc}{\quad}

17: \newcommand{\dt}{\dot}

18: \newcommand{\tbar}{\overline}

19: \newcommand{\bbar}{\underline}

20: \newcommand{\lft}{\left}

21: \newcommand{\rit}{\right}

22: \newcommand{\clyd}{\partial}

23: \newcommand{\mcal}{\mathcal}

24: \newcommand{\npg}{\newpage}

25: \newcommand{\eee}{\varepsilon}

26: \newcommand{\kkk}{\kappa}

27: \newcommand{\w}{\omega}

28: \newcommand{\om}{\Omega}

29: \newcommand{\ld}{\lambda}

30: \newcommand{\tht}{\theta}

31: \newcommand{\degrees}{\mbox{$^\circ$}}

32: \newcommand{\sech}{\mathrm{sech}}

33: \newcommand{\dd}{{\mathrm d}}

34: \newcommand{\ee}{{\mathrm e}}

35: \newcommand{\ii}{{\mathrm i}}

36: \newcommand{\sfr}[2]{\mbox{$\frac{#1}{#2}$}}

37: \newcommand{\shalf}{\sfr{1}{2}}

38: \newcommand{\elo}{E_0}

39: \newcommand{\wbr}{\mcal{W}_{\mathrm{br}}}

40: \newcommand{\ptwo}{\lft(\frac{p}{2}\rit)}

41: \newenvironment{eqn}{\begin{equation}}{\end{equation}}

42: \newenvironment{ary}{\begin{eqnarray}}{\end{eqnarray}}

43: \newenvironment{aln}{\begin{align}}{\end{align}}

44: \newenvironment{enum}{\begin{enumerate}}{\end{enumerate}}

45: \newenvironment{lst}{\begin{itemize}}{\end{itemize}}

46: \newcommand{\itc}{\emph}

47: \newcommand{\str}{\textrm}

48: \newcommand{\mb}{\mathbf}

49: \newcommand{\tb}{\textbf}

50: \newcommand{\fns}{\footnotesize}

51: \newcommand{\ssz}{\scriptsize}

52: 

53: %COMMANDS TO DO WITH NUMBERING

54: \newcommand{\lbl}{\label}

55: %\newcommand{\lbl}[1]{\framebox{\scriptsize #1}\qquad \label{#1}}

56: \makeatletter

57: \@addtoreset{equation}{section}

58: \makeatother

59: \renewcommand{\theequation}{\arabic{section}.\arabic{equation}}

60: 

61: %----------------------------------------------------------------------------

62: \begin{document}

63: 

64: \submitto{\JPA}

65: 

66: \title{Discrete breathers in a two-dimensional 

67: hexagonal Fermi-Pasta-Ulam lattice}

68: \author{Imran A Butt and Jonathan A D Wattis}

69: \address{Theoretical Mechanics,  School of Mathematical Sciences,

70: University of Nottingham, University Park, Nottingham, NG7 2RD,

71: UK} \eads{\mailto{imran.butt@maths.nott.ac.uk},

72: \mailto{jonathan.wattis@nottingham.ac.uk}}

73: 

74: \begin{abstract} \lbl{abs}

75: We consider a two-dimensional Fermi-Pasta-Ulam (FPU) lattice 

76: with hexagonal symmetry.  Using asymptotic methods based on 

77: small amplitude ansatz, at third order we obtain a reduction to a 

78: cubic nonlinear Schr{\"o}dinger equation (NLS) for the breather 

79: envelope.  However, this does not support stable soliton solutions, 

80: so we pursue a higher-order analysis yielding a generalised NLS, 

81: which includes known stabilising terms.   We present numerical 

82: results which suggest that long-lived stationary and moving breathers 

83: are supported by the lattice. We find breather solutions which move 

84: in an arbitrary direction, an ellipticity criterion for the wavenumbers 

85: of the carrier wave, asymptotic estimates for the breather energy, 

86: and a minimum threshold energy below which breathers cannot be 

87: found.  This energy threshold is maximised for stationary breathers, 

88: and becomes vanishingly small near the boundary of the elliptic domain 

89: where breathers attain a maximum speed.  Several of the results obtained 

90: are similar to those obtained for the square FPU lattice (Butt \& Wattis, 

91: {\em J Phys A}, {\bf 39}, 4955, (2006)),  though we find that the square 

92: and hexagonal lattices exhibit different properties in regard to the 

93: generation of harmonics, and the isotropy of the generalised NLS equation.  

94: \end{abstract}

95: 

96: \pacs{05.45.-a, 05.45.Yv \\[2ex]This version: \today}

97: 

98: 

99: %--------------------------------------------------------------

100: \section{Introduction} \lbl{hexint}

101: 

102: 

103: Discrete breathers are time-periodic and spatially localised exact

104: solutions of translationally invariant nonlinear lattices.  For a

105: brief review of some general properties of breathers in

106: higher-dimensional systems, see our earlier work \citep{buts06}.  

107: In particular, it is known that while some fundamental properties

108: such as the existence of breathers are not affected by lattice

109: dimension (see Flach \etal \citep{flat94} and Mackay and Aubry

110: \citep{mac94}), other properties are affected profoundly, for

111: instance, the energy properties of breathers (see Flach \etal

112: \citep{fla97}, Kastner \citep{kas04} and Weinstein \citep{wei99}).

113: In this paper we investigate how the symmetry of the lattice 

114: influences the properties of discrete breathers found therein.  

115: 

116: 

117: In Hamiltonian systems, stationary breathers occur in one-parameter 

118: families.  For a certain class of Hamiltonian systems, a critical spatial 

119: dimension $d_c$ exists such that for systems with $d \geq d_c$, 

120: there exists a positive lower bound on the energy of a breather 

121: family, and the breather energies do {\em not} approach zero even 

122: as the amplitude tends to zero.   For lattices with dimension 

123: $d < d_c$, there is no positive lower bound on the energy of 

124: breathers.  In other words, the energy of a family of breathers goes 

125: to zero with amplitude, and breathers of arbitrarily small energy can 

126: be found.    The critical dimension $d_c$ is typically two.  A small 

127: amplitude expansion yields the NLS reduction to $\ii F_T+\Delta F 

128: + \kappa |F|^{2\sigma}F=0$ which has a critical dimension of  

129: $d_c=2/\sigma$, with blow-up in NLS occurring when $\kappa>0$ 

130: and $d>d_c$.   For typical lattice potentials, $\sigma=1$ again 

131: confirming the critical dimension of $d_c=2$. 

132: 

133: 

134: Marin, Eilbeck and Russell have performed extensive numerical 

135: investigations of breather dynamics in two-dimensional lattices 

136: \citep{mar98,mar00,mar01}.  Their results suggest that moving 

137: breather modes exist (or are at least extremely long-lived), and 

138: that the lattice exhibits a strong directional preference whereby 

139: breathers can only move along symmetries of the lattice, and in 

140: no other direction.  Such quasi-one-dimensional behaviour is also 

141: observed in higher-dimensional lattices.

142: 

143: 

144: The work in this paper follows on from our earlier study 

145: \citep{buts06} of breathers in a two-dimensional lattice with square 

146: rotational symmetry;  hereupon referred to as the ``square'' lattice.  

147: This lattice has $C_4$ symmetry, by which we mean that a rotation 

148: through any multiple of $2\pi/4 = \pi/2$ about an axis perpendicular to 

149: the lattice plane through a lattice site maps the lattice onto itself.  

150: %

151: In \cite{buts06}, using the semi-discrete multiple-scale method 

152: (see Remoissenet \citep{rem85}), we determined an approximate 

153: form for (as well as the properties of) small 

154: amplitude breathers in a two-dimensional square Fermi-Pasta-Ulam 

155: (FPU) lattice \citep{fer40}.  We found a third-order analysis to be 

156: inadequate, since the partial differential equation obtained at this order 

157: describing the breather envelope exhibits \itc{blow-up}.  To

158: overcome this, we incorporated higher-order effects in the model,

159: thereby obtaining a modified partial differential equation which

160: includes known stabilising terms. From this, we determined regions

161: of parameter space where breather solutions are expected.

162: Numerical simulations supported the results of our analysis, and

163: suggested that, in contrast to the two-component lattices studied

164: by Marin \etal\ \citep{mar98,mar00,mar01} (that is, with two

165: degrees of freedom at each lattice site), there is no restriction upon 

166: the permitted directions of travel within the one-component square 

167: FPU lattice. We also found asymptotic estimates for the breather 

168: energy which confirmed the existence of a minimum threshold 

169: energy, in agreement with the work of Flach \etal\ \citep{fla97}.

170: 

171: 

172: \begin{figure}[htbp]

173: \begin{center}

174: \resizebox{5in}{!}{\includegraphics{hexlatsmen.eps}}

175: \caption{The 2D hexagonal electrical transmission lattice (HETL).}

176: \lbl{hexetl}

177: \end{center}

178: \end{figure}

179:  

180: 

181: In this paper, we consider a hexagonal electrical transmission lattice 

182: (HETL).  This two-dimensional network possesses $C_6$ (or hexagonal) 

183: rotational symmetry.  That is, rotation through any multiple of the angle 

184: $2\pi/6 = \pi/3$ about a lattice site maps the lattice onto itself.  The HETL 

185: is shown in \Fref{hexetl}, pictured from a point vertically above the plane 

186: of the lattice.  We note from \Fref{hexetl} that geometrically, the HETL is 

187: an arrangement of tessellating triangles (not hexagons).   Nevertheless, the 

188: arrangement in \Fref{hexetl} is referred to as ``hexagonal'' rather than 

189: ``triangular'' since these descriptions refer to its symmetry properties and 

190: not to the geometrical shapes which comprise the array (not all authors 

191: follow this convention).  One might expect the analysis for the hexagonal 

192: lattice to be more involved, since it is geometrically more complicated in 

193: having more links emanating from each node.  However, the hexagonal 

194: symmetry results in greater isotropy and hence simpler equations than 

195: those obtained for the square lattice.

196: 

197: 

198: We derive the equations of motion and demonstrate a Hamiltonian

199: formalism in \Sref{hexder}.  In Sections \ref{hexsym} and

200: \ref{hexasymm}, we present two cases for which the hexagonal FPU

201: lattice equations can be reduced to a two-dimensional nonlinear

202: Schr{\"o}dinger (NLS) equation with cubic nonlinearity.  We

203: consider lattices with a symmetric interaction potential, in which

204: case reduction to a cubic NLS equation can be performed for moving

205: breathers.  We find an ellipticity criterion for the wavenumbers

206: of the carrier wave in \Sref{hexdell}.  A reduction to the cubic NLS

207: equation can also be carried out for lattices with an asymmetric

208: potential, provided we consider only stationary breathers.  

209: 

210: 

211: As expected we find a minimum energy below which breathers cannot 

212: exist in the hexagonal FPU lattice.  We find that the

213: energy threshold is dependent upon the wavevector of the carrier

214: wave. It is maximised for stationary breathers, and becomes

215: arbitrarily small near the boundary of the elliptic domain.   

216: 

217: 

218: The cubic NLS equation admits only unstable Townes solitons.  Hence, 

219: in \Sref{hex5}, we extend our asymptotic analysis to higher order and 

220: find an isotropic generalised NLS equation which incorporates known 

221: stabilising terms.  Our analytic work is supplemented by numerical simulations 

222: presented in \Sref{hexnumerics}, which suggest that long-lived stationary

223: and moving breather modes are supported by the system.

224: In \Sref{hdisc}, we discuss the results obtained in this paper. 

225: 

226: 

227: %==================================================

228: \section{A two-dimensional hexagonal Fermi-Pasta-Ulam lattice}

229: \lbl{hexfpulat}

230: 

231: 

232: \subsection{Preliminaries} \lbl{hexpre}

233: 

234: 

235: Before we derive the equations governing the HETL, we describe our 

236: scheme for indexing the lattice nodes and our choice of basis vectors.  We 

237: introduce a rectangular lattice with basis vectors $\mcal{B}~=~\{ \mb{i'} \, , 

238: \, \mb{j'} \}$, where $\mb{i'} = \mb{i} = [1,0]^T$ and $\mb{j'} = [0,h]^T$ 

239: ($\mb{j}$ being $[0,1]^T$), illustrated in \Fref{sqbase}.  We use only half of 

240: the $(m,n)$ indices, namely, those for which the sum $m+n$ is even.  We 

241: choose an origin with coordinates $(0,0)$;  the position of the site $(m,n)$ 

242: is $m\mb{i'} + n\mb{j'}$.  In order for the hexagonal lattice to be regular, 

243: we specify $h=\sqrt{3}$. 

244: 

245:  

246: \begin{figure}[htbp]

247: \begin{center}

248: \vspace{-1.0cm}

249: \psfrag{0}{{\scriptsize $(m,n)$}}

250: \psfrag{1}{{\scriptsize $(m+2,n)$}}

251: \psfrag{2}[B][]{{\scriptsize $(m+1,n+1)$}}

252: \psfrag{3}[B][]{{\scriptsize $(m-1,n+1)$}}

253: \psfrag{4}[B][r]{{\scriptsize $(m-2,n)$}}

254: \psfrag{5}[B][b]{{\scriptsize $(m-1,n-1)$}}

255: \psfrag{6}[B][b]{{\scriptsize $(m+1,n-1)$}}

256: \psfrag{i}{{\footnotesize $\mb{i'}$}}

257: \psfrag{j}{{\footnotesize $\mb{j'}$}}

258: \resizebox{4in}{!}{\includegraphics{sqbasecorr2.eps}}

259: \caption{Labelling of nodes in the HETL with basis 

260: $ \mcal{B} = \{ \mb{i'} \, , \, \mb{j'} \}$.}

261: \lbl{sqbase}

262: \end{center}

263: \end{figure}

264: 

265: 

266: At every node of the HETL lies a nonlinear capacitor (not shown in 

267: \Fref{hexetl}), and between every node and each of its six nearest 

268: neighbours is a linear inductor.  An enlarged view of the area surrounding 

269: the capacitor at $(m,n)$ is shown in \Fref{hexunitdir}, where the capacitor 

270: is visible.  The variable $V_{m,n}$ denotes the voltage across the 

271: capacitor $(m,n)$ and $Q_{m,n}$ denotes the total charge stored on the 

272: capacitor at $(m,n)$.  Also, $I_{m,n}$, $J_{m,n}$ and $K_{m,n}$ are the 

273: currents through the inductors immediately to the right of site $(m,n)$ in 

274: the directions $\mb{e_i} = [2,0]^T$, $\mb{e_j} = [1,-\sqrt{3}]^T$ and 

275: $\mb{e_k}=[1,\sqrt{3}]^T$ respectively, as illustrated in \Fref{hexunitdir}. 

276: 

277: 

278: \begin{figure}[htbp]

279: \begin{center}

280: \vspace{1.3cm}

281: \psfrag{a}{{\large $V_{m,n}$}}

282: \psfrag{b}[B][]{{\large $V_{m+1,n-1}$}}

283: \psfrag{c}[B][]{{\large $V_{m+2,n}$}}

284: \psfrag{d}{{\large $V_{m+1,n+1}$}}

285: \psfrag{e}{{\large $I_{m,n}$}}

286: \psfrag{f}{{\large $J_{m,n}$}}

287: \psfrag{g}{{\large $K_{m,n}$}}

288: \psfrag{i}{{\normalsize $\mb{e_i}$}}

289: \psfrag{j}{{\normalsize $\mb{e_j}$}}

290: \psfrag{k}{{\normalsize $\mb{e_k}$}}

291: \resizebox{5.5in}{!}{\includegraphics{hexundir.eps}}

292: \caption{Enlarged view of the HETL at site $(m,n)$.}

293: \lbl{hexunitdir}

294: \end{center}

295: \end{figure}

296: 

297: 

298: %----------------------------------------------------------------

299: \subsection{Derivation of model equations} \lbl{hexder}

300: 

301: 

302: To derive the equations relating current, charge and voltage in the lattice

303: we apply Kirchoff's law 

304: \begin{eqnarray}

305: V_{m+2,n} - V_{m,n} & = -L \frac{\dd I_{m,n}}{\dd t}, \lbl{hki} \\[1ex]

306: V_{m+1,n-1} - V_{m,n} & = -L \frac{\dd J_{m,n}}{\dd t}, \lbl{hkj} \\[1ex]

307: V_{m+1,n+1} - V_{m,n} & = -L \frac{\dd K_{m,n}}{\dd t}, \lbl{hkk}

308: \end{eqnarray}

309: where the inductance $L$ is constant.  Conservation of charge gives

310: \begin{eqn} \lbl{hcc}

311: I_{m-2,n}-I_{m,n} + J_{m-1,n+1}-J_{m,n} + K_{m-1,n-1}-K_{m,n} =

312: \frac{\dd Q_{m,n}}{\dd t}.

313: \end{eqn}

314: Differentiating \eref{hcc} with respect to time, and then using 

315: \eref{hki}--\eref{hkk} to find $\dot{I}_{m-2,n}$, $\dot{J}_{m-1,n+1}$ 

316: and $\dot{K}_{m-1,n-1}$, we have  

317: \begin{eqnarray}

318: L \frac{\dd^2 Q_{m,n}}{\dd t^2} & = & \spc (V_{m+2,n} - 2 V_{m,n} + 

319: V_{m-2,n}) + (V_{m+1,n-1} - 2 V_{m,n} + V_{m-1,n+1}) \nonumber \\ & & 

320: + (V_{m+1,n+1} - 2 V_{m,n} + V_{m-1,n-1}). 

321: \lbl{heqvqfull} \end{eqnarray}

322: Equation \eref{heqvqfull} may be written in the abbreviated form

323: $L \ddot{Q}_{m,n} = (\delta^2_I + \delta^2_J + \delta^2_K)V_{m,n}$,

324: where the centred second-difference operators are defined by

325: \begin{eqnarray}

326: \lbl{hdi} \delta^2_I A_{m,n} & = A_{m+2,n}   - 2 A_{m,n} + A_{m-2,n},\\[1ex]

327: \lbl{hdj} \delta^2_J A_{m,n} & = A_{m+1,n-1} - 2 A_{m,n} + A_{m-1,n+1},\\[1ex]

328: \lbl{hdk} \delta^2_K A_{m,n} & = A_{m+1,n+1} - 2 A_{m,n} + A_{m-1,n-1} . 

329: \end{eqnarray}

330: Here, $A_{m,n}$ is an arbitrary quantity referenced by two indices; $\delta^2_I$, 

331: $\delta^2_J$ and $\delta^2_K$ are centred second-difference operators in 

332: the directions of $\mb{e_i}$, $\mb{e_j}$ and $\mb{e_k}$ respectively.   

333: Since the voltage $V_{m,n}$ is known in terms of the charge, we reformulate 

334: \eref{heqvqfull} in terms of the single quantity $Q_{m,n}$.  We invert the 

335: capacitor's nonlinear charge-voltage relationship $Q=VC(V)$ (see 

336: equations (2.9)--(2.14) of \cite{buts06} for details), to obtain 

337: \begin{equation} 

338: V(Q) = ( Q + a Q^2 + b Q^3 + c Q^4 + d Q^5 ) / C_0 , 

339: \lbl{hqvexp} \end{equation}

340: where $C_0=C(0)$.  Hence, the HETL equations \eref{heqvqfull} can 

341: be written as 

342: \begin{eqn} \lbl{heqq}

343: \frac{\dd^2 Q_{m,n}}{\dd t^2} = (\delta^2_I + \delta^2_J + \delta^2_K) 

344: \lft[Q_{m,n}+aQ_{m,n}^2+bQ_{m,n}^3+cQ_{m,n}^4+dQ_{m,n}^5\rit], 

345: \end{eqn}

346: where $m,n \in {\bbf{Z}}$ and, by rescaling the time variable, we set 

347: $LC_0=1$ without loss of generality.   Thus we have shown that the 

348: equation governing charge in the HETL \eref{heqq} is a two-dimensional 

349: analogue of the Fermi-Pasta-Ulam equation 

350: \begin{equation}

351: \frac{\dd^2 Q_j}{\dd t^2} = W'(Q_{j+1}) - 2 W'(Q_j) + W'(Q_{j-1}) , 

352: \end{equation}

353: which is a Hamiltonian system that can be derived from both 

354: \begin{equation}

355: H = \sum_j \shalf \pi_j^2 + W(\phi_{j+1}-\phi_j) , \;\;\; {\rm and} \;\;\; 

356: \widetilde H = \sum_j \shalf (P_{j+1}-P_j)^2 + W(Q_j) , 

357: \end{equation}

358: where $Q_j=\phi_{j+1}-\phi_j$.  

359: 

360: 

361: The lattice equations \eref{heqq} can be derived from the Hamiltonian 

362: \begin{eqnarray}

363: \widetilde{H} & =\sum_{m,n} & 

364: \mbox{$\frac{1}{2}$} (P_{m+2,n}-P_{m,n})^2 + 

365: \mbox{$\frac{1}{2}$} (P_{m+1,n-1} - P_{m,n})^2 \nonumber \\ & & + 

366: \mbox{$\frac{1}{2}$} (P_{m+1,n+1} - P_{m,n})^2 + \Upsilon (Q_{m,n}), 

367: \lbl{hham} \end{eqnarray}

368: where $\Upsilon (Q_{m,n})$ satisfies $\Upsilon\!\ '(Q_{m,n}) =

369: V(Q_{m,n})$ given in \eref{hqvexp}, hence   

370: \begin{eqn}

371: \Upsilon(Q) = \mbox{$\frac{1}{2}$} Q^2 + 

372: \mbox{$\frac{1}{3}$} a Q^3 + \mbox{$\frac{1}{4}$} b Q^4 + 

373: \mbox{$\frac{1}{5}$} c Q^5 + \mbox{$\frac{1}{6}$} d Q^6 .

374: \end{eqn} 

375: We describe potentials which satisfy $\Upsilon(-Q)=\Upsilon(Q)$ (that is, 

376: $a=c=0$) as `symmetric'.  The variables $P_{m,n}$ and $Q_{m,n}$ are 

377: conjugate momenta and displacement variables of the system; and 

378: eliminating $P_{m,n}$ from the equations 

379: \begin{eqn} \lbl{hhameq}

380: \frac{\dd Q_{m,n}}{\dd t} = -(\delta^2_I + \delta^2_J  +

381: \delta^2_K)P_{m,n}, \qquad

382: \frac{\dd P_{m,n}}{\dd t} = -\Upsilon\!\ '(Q_{m,n}) , 

383: \end{eqn}

384: yields (\ref{heqq}). In \Sref{heeh}, we derive expressions for 

385: the energy of breathers given the small amplitude solutions 

386: which are obtained in the next section. 

387: 

388: 

389: %--------------------------------------------------------------------

390: \subsection{Asymptotic analysis} \lbl{haa}

391: 

392: 

393: We apply the method of multiple-scales to determine an approximate 

394: analytic form for small amplitude breather solutions of \eref{heqq}, with 

395: slowly varying envelope.  We introduce new variables defined by

396:  \begin{eqn} \lbl{hexvars}

397:  X=\eee m,\spc Y=\eee h n,\spc\tau=\eee t\spc\str{and}\spc T=\eee^2 t;

398:  \end{eqn}

399: note the presence of the scaling factor $h$ in the definition of $Y$.  

400: We seek solutions of \eref{heqq} of the form

401: \begin{eqnarray} \fl 

402: Q_{m,n}(t) = \hspace*{-3mm} & \!\!\!\! \eee \ee^{\ii \psi} F(X,Y,\tau,T) + 

403: \eee^2 G_0(X,Y,\tau,T) +

404: \eee^2 \ee^{\ii \psi} G_1(X,Y,\tau,T) + \nonumber \\ & \!\!\!\! 

405: \eee^2 \ee^{2\ii \psi} G_2(X,Y,\tau,T) + \eee^3 H_0 (X,Y,\tau,T) +

406: \eee^3 \ee^{\ii \psi} H_1(X,Y,\tau,T) + \nonumber \\

407: & \!\!\!\!  \eee^3 \ee^{2\ii \psi} H_2(X,Y,\tau,T)

408: + \eee^3\ee^{3\ii \psi}H_3(X,Y,\tau,T) + \eee^4\ee^{\ii\psi}I_1(X,Y,\tau,T)

409: + \nonumber\\ & \!\!\!\!  

410: \eee^4 \ee^{2\ii \psi} I_2(X,Y,\tau,T) + \eee^4 \ee^{3\ii\psi} I_3(X,Y,\tau,T)

411: + \eee^4\ee^{4\ii \psi} I_4(X,Y,\tau,T) + \nonumber \\

412: & \!\!\!\! \eee^5\ee^{\ii \psi} J_1(X,Y,\tau,T)+ \cdots + \mathrm{c.c.},

413: \lbl{hexanz} \end{eqnarray}

414: where the phase $\psi$ of the carrier wave is given by $km+lhn+\w t$ (once 

415: again noting the extra factor $h$), and $\mb{k}=[k,l]^T$ and $\w(\mb{k})$ 

416: are its wavevector and temporal frequency respectively.  We substitute the 

417: ansatz \eref{hexanz} into the lattice equations \eref{heqq} and equate 

418: coefficients of each harmonic frequency at each order of $\eee$.  After 

419: much simplification, this yields the following system of equations: \\

420: \\

421: $\mathcal{O}(\eee \ee^{\ii\psi})$:

422: \begin{eqn} \lbl{hexdisp}

423: \w^2 F = 4\sin^2(k) F + 4\sin^2 \lft( \frac{k+lh}{2} \rit) F + 

424: 4\sin^2 \lft( \frac{k-lh}{2} \rit) F,

425: \end{eqn}

426: $\mathcal{O}(\eee^2 \ee^{\ii\psi})$:

427: \begin{eqn} \lbl{hexvel}

428: \w F_{\tau} = 2\sin k [2\cos k + \cos(lh)]F_X + 2h\cos k \sin(lh) F_Y,

429: \end{eqn}

430: $\mcal{O}(\eee^2 \ee^{2\ii\psi})$:

431: \begin{eqn} \lbl{hexg2}

432: \w^2 G_2 = [\sin^2(2k) + \sin^2(k+lh) + \sin^2(k-lh)](G_2 + a F^2),

433: \end{eqn}

434: $\mcal{O}(\eee^3 \ee^{\ii\psi})$:

435: \begin{eqnarray}

436: \fl 2\ii \w F_T + F_{\tau\tau}  & 

437: = & [4\cos(2k) + 2\cos k \cos(lh)] F_{XX} + 2h^2\cos k \cos(lh)

438: F_{YY} \nonumber \\

439: & & - 4h\sin k \sin(lh)F_{XY} \nonumber \\

440: & & - \ 8a \lft[ \sin^2(k) + \sin^2 \lft( \frac{k+lh}{2} \rit)

441: + \sin^2 \lft( \frac{k-lh}{2} \rit) \rit] [F(G_0+\tbar{G}_0)+\tbar{F}G_2] 

442: \nonumber\\

443: & & - \ 12 b \lft[ \sin^2(k) + \sin^2 \lft( \frac{k+lh}{2} \rit)

444: + \sin^2 \lft( \frac{k-lh}{2} \rit) \rit] |F|^2 F,  \lbl{hexnls}

445: \end{eqnarray}

446: $\mcal{O}(\eee^3 \ee^{3\ii\psi})$:

447: \begin{eqn} \fl 

448: 9\w^2H_3 =  4 \lft[ \sin^2(3k) + \sin^2 \lft( \frac{3k+3lh}{2} \rit) + 

449: \sin^2 \lft( \frac{3k-3lh}{2} \rit) \rit] \left( H_3 + 2aFG_2+bF^3 \right), 

450: \lbl{hexh3} \end{eqn}

451: $\mcal{O}(\eee^4 \ee^0)$:

452: \begin{eqn} \lbl{hexfo}

453: G_{0\tau\tau} = 6\, G_{0XX} + 2h^2 G_{0YY} + 

454: a \! \lft[ 6 \lft( |F|^2 \rit)_{XX} + 2h^2 \lft(|F|^2 \rit)_{YY} \rit] .

455: \end{eqn}

456: 

457: 

458: Though each equation plays a similar role to its counterpart in the square 

459: lattice, equations \eref{hexdisp}--\eref{hexfo} are more complicated. 

460: Equation \eref{hexdisp} is the dispersion relation for the system \eref{heqq}.  

461: Since we are interested only in solutions for which $F \neq 0$, \eref{hexdisp} yields 

462: \begin{eqn} \lbl{hexdisp2} 

463: \w^2  = 4\sin^2(k)  + 4\sin^2 \lft( \frac{k+lh}{2} \rit) + 

464: 4\sin^2 \lft( \frac{k-lh}{2} \rit),

465: \end{eqn}

466: which does not simplify significantly.  From equation \eref{hexvel}, 

467: we determine the velocity of the travelling wave $F$, finding 

468: \begin{eqn}  \lbl{hexvels}

469: F(X,Y,\tau,T) \equiv F(Z,W,T),

470: \end{eqn}

471: where $Z=X-u\tau$ and $W=Y-v\tau$, and the horizontal and vertical

472: velocity components ($u$ and $v$) are found to be

473: \begin{eqn}  \lbl{velsuv}

474: u = \frac{-2\sin (k)[2\cos(k)+\cos(lh)]}{\w} \spc\str{and}\spc

475: v = \frac{-2h\cos(k)\sin(lh)}{\w}.

476: \end{eqn}

477: Equation \eref{velsuv}, along with \eref{hexdisp2} enables the elimination 

478: of terms involving $G_1$ from \eref{hexnls} which are not shown.  We 

479: denote the angle at which the envelope $F$ propagates through the 

480: lattice by $\Psi$, which is measured from the direction of the basis vector 

481: $\mb{e_i}$ to the line of travel and hence is given by $\tan^{-1}(v/u)$, 

482: which in turn depends upon the wavevector  $\mb{k}$.   For both the cases 

483: that we consider (namely, symmetric and asymmetric interaction potentials), 

484: we find constraints upon the wavenumbers $k$ and $l$ which affect the 

485: velocity components $u$ and $v$.   By taking $k=\pi/3$ and  $l=\pi/h$ we 

486: find $u=v=0$, which corresponds to a static breather; and by choosing 

487: $k,l$ values which circle this point, we find breathers which can propagate 

488: in any direction ($0\leq\Psi<2\pi$); in other words, our analysis suggests 

489: that there is no restriction upon the direction of travel through the lattice.

490: 

491: 

492: Our aim is to reduce \eref{hexnls} to a nonlinear

493: Schr{\"o}dinger (NLS) equation for $F$.  Before this

494: can be done, the quantities $G_0$ and $G_2$ in \eref{hexnls} must

495: be found in terms of $F$.  As for the square lattice, it is straightforward

496: to determine $G_2$ from the algebraic equation \eref{hexg2}.

497: However, the partial differential equation \eref{hexfo} for $G_0$

498: can be solved for two special cases only; namely, 

499: symmetric potentials, in which case the reduction of \eref{hexnls}

500: to an NLS equation can be completed even for moving breathers, and

501: also asymmetric potentials, provided we confine our attention to

502: stationary breathers.  These two cases are considered in Sections 

503: \ref{hexsym} and \ref{hexasymm} respectively, where we also use 

504: our breather formulae to generate estimates for the breather energy.

505: 

506: 

507: %-------------------------------------------------------------------

508: \subsection{Asymptotic estimates for breather energy} \lbl{heeh}

509: 

510: 

511: The HETL is a lossless network, meaning that the total electrical energy 

512: is conserved.  This quantity is related to the Hamiltonian 

513: (\ref{hham}) by $E=\widetilde{H}/C_0$ and so is given by 

514: \begin{eqn} \lbl{hexentot}

515: E = \sum_{m,n} e_{m,n} =  \sum_{m,n} \frac{\Upsilon(Q_{m,n})}{C_0} 

516: + \mbox{$\frac{1}{2}$}L \lft( I_{m,n}^2 + J_{m,n}^2 + K_{m,n}^2 \rit) . 

517: \end{eqn}

518: The electrical energy $e_{m,n}$ associated with a unit of the lattice is 

519: (see \Fref{hexunitdir})

520: \begin{eqn} \lbl{hexenun} 

521: e_{m,n} = \frac{\Upsilon(Q_{m,n})}{C_0} + 

522: \mbox{$\frac{1}{2}$}L ( I_{m,n}^2 + J_{m,n}^2 + K_{m,n}^2 ) . 

523: \end{eqn}

524: To derive a leading-order estimate for the electrical energy, defined by 

525: \begin{eqn}  \lbl{hentotlo}

526: E_0 = \sum_{m,n} e_{m,n}^{(0)}=\sum_{m,n}\frac{Q_{m,n}^2}{2C_0}

527: + \frac{L}{2} \lft( I_{m,n}^2 + J_{m,n}^2 + K_{m,n}^2 \rit) , 

528: \end{eqn}

529: we use leading-order expressions for each of the terms in the terms in 

530: the summand of \eref{hexentot}.  The first term is $Q_{m,n}^2/(2C_0)$;  

531: from \eref{hqvexp}, it follows that $V_{m,n} \sim Q_{m,n}/C_0$, and 

532: so to leading order, (\ref{hexentot}) agrees with the linear approximation 

533: to the energy in the capacitor being $QV/2$.   To find leading-order 

534: expressions for the currents  $I_{m,n}$, $J_{m,n}$ and $K_{m,n}$, we 

535: use equations \eref{hki}--\eref{hkk}, which imply 

536: \begin{eqnarray}

537: Q_{m+2,n} - Q_{m,n} & = - \frac{\dd I_{m,n}}{\dd t}, \lbl{hkiq} \\[1ex]

538: Q_{m+1,n-1} - Q_{m,n} & = - \frac{\dd J_{m,n}}{\dd t}, \lbl{hkjq} \\[1ex]

539: Q_{m+1,n+1} - Q_{m,n} & = - \frac{\dd K_{m,n}}{\dd t}, \lbl{hkkq}

540: \end{eqnarray}

541: where $LC_0=1$.  The currents are determined by substituting the

542: expression for the breather $Q_{m,n}$ into \eref{hkiq}--\eref{hkkq} 

543: and then integrating with respect to time.  We obtain leading-order 

544: estimates for the energy of moving breathers in systems with symmetric 

545: potentials Section \ref{hemv}, and in Section \ref{hest}, the energies 

546: of stationary breathers with asymmetric potentials.

547: 

548: 

549: %----------------------------------------------------------------

550: \subsection{The dispersion relation for the HETL} \lbl{wplots}

551: 

552: 

553: In this section, we analyse the dispersion relation \eref{hexdisp2} 

554: for the system \eref{heqq}.  A contour plot of $\w$ against $k$ and 

555: $l$ is shown in \Fref{wcont}.  Since $w$ is periodic in both $k$ and 

556: $l$, with period $2\pi$ along the $k$-direction, and $2\pi/h$ in the 

557: $l$-direction, we consider only $k$ and $l$ such that 

558: $(k,l) \in \mcal{T}^2 = [0,2\pi] \times [0,2\pi/h]$.

559: 

560: 

561: The function $\w$ is minimised, and assumes the value zero, at the

562: centre of the circular patterns in \Fref{wcont}.  The minima are located 

563: at $(0,0)$, $(2\pi,0)$, $(2\pi,2\pi/h)$, $(0,2\pi/h)$ and $(\pi,\pi/h)$.  

564: Plus signs (`$+$') mark the points in $(k,l)$-space at which $\w$ is 

565: maximised and takes the value $\w=3$.  The maxima lie at the centres 

566: of the equilateral triangles, the wavevectors corresponding to these 

567: points are denoted $\mb{k_1}, \ldots, \mb{k_6}$, where

568: \begin{equation} \begin{array}{ lclcl}

569: \mb{k_1}=[\pi/3,\pi/h]^T,  && 

570: \mb{k_2}=[2\pi/3,0]^T,  && 

571: \mb{k_3}=[4\pi/3,0]^T,    \\ 

572: \mb{k_4}=[5\pi/3,\pi/h]^T, && 

573: \mb{k_5}=[4\pi/3,2\pi/h]^T, &&

574: \mb{k_6}=[2\pi/3,2\pi/h]^T.

575: \end{array} \end{equation} 

576: The arrangement of the points corresponding to these wavevectors

577: in $(k,l)$-space reflects the hexagonal symmetry properties of the

578: lattice \eref{heqq}.  It may be verified using \eref{velsuv} that

579: the velocity components $u$ and $v$ are both zero for each of the

580: wavevectors $\{ \mb{k_1},\ldots, \mb{k_6} \}$.

581: 

582: 

583: \begin{figure}[htbp]

584: \begin{center}

585: \resizebox{4.5in}{!}{\includegraphics{wcont.ps}}

586: \put(-250,103){\footnotesize $\mb{k_1}$}

587: \put(-200,44){\footnotesize $\mb{k_2}$}

588: \put(-100,44){\footnotesize $\mb{k_3}$}

589: \put(-48,103){\footnotesize $\mb{k_4}$}

590: \put(-100,183){\footnotesize $\mb{k_5}$}

591: \put(-200,183){\footnotesize $\mb{k_6}$}

592: \caption{Contour plot of $\w(\mb{k})$; $\w$ attains its maximum 

593: value of 3 at the points $\mb{k_1},\ldots,\mb{k_6}$, and 

594: is minimised at the points marked `o', where $\w=0$. }

595: \lbl{wcont}

596: \end{center}

597: \end{figure}

598: 

599: 

600: %------------------------------------------------------------------

601: \subsection{Lattices with a symmetric potential} \lbl{hexsym}

602: 

603: 

604: In this section, we consider lattices with a symmetric interaction potential, 

605: where $\Upsilon (Q)$ is even and $\Upsilon\!\ '(Q)$  has odd symmetry. 

606: This corresponds to $a=c=0$ in \eref{heqq}.  Since there are no even 

607: harmonics for vibrations controlled by symmetric potentials, it follows 

608: that $G_0$ and $G_2$ are both zero.  In \eref{hexnls}, the term 

609: $F_{\tau\tau}$ is eliminated using $F_{\tau\tau} = u^2F_{ZZ} + 2uvF_{ZW} 

610: + v^2F_{WW}$ which is derived from \eref{hexvels}.  This 

611: leads to (\ref{hexnls}) being reduced to an NLS equation 

612: \begin{eqnarray}

613: \fl 2\ii\w F_T + \lft[ u^2 - 4\cos(2k) - 2\cos k\cos(lh) \rit] F_{ZZ}

614: + \lft[ v^2 - 2h^2\cos k \cos(lh) \rit] F_{WW} \nonumber \\

615: + \lft[ 2uv + 4h \sin k \sin(lh) \rit] F_{ZW} + 3b\w^2|F|^2F = 0, 

616: \lbl{hexrnls} \end{eqnarray}

617: where the velocities $u$ and $v$ are given by \eref{velsuv}.  By

618: applying an appropriate change of variables, we remove 

619: the mixed derivative term from  \eref{hexrnls}, and reduce the

620: equation to a standard form.  To simplify the appearance of

621: subsequent expressions, we denote the coefficients of $F_{ZZ}$,

622: $F_{WW}$ and $F_{ZW}$ by $D_1 = u^2 - 4\cos(2k) - 2\cos

623: k\cos(lh)$, $D_2 =  v^2 - 2h^2\cos k \cos(lh)$ and $D_3 = 2uv + 4h

624: \sin k \sin(lh)$ respectively.  A suitable transformation is thus

625: \begin{eqn} \lbl{hextr}

626: \xi = \frac{hZ}{\sqrt{D_1}} \spc \str{and} \spc

627: \eta = \frac{h(2D_1 W - D_3 Z)}{\sqrt{D_1 (4D_1 D_2 - D_3^2)}},

628: \end{eqn}

629: which implies \eref{hexrnls} becomes

630: \begin{eqn} \lbl{hexcan}

631: 2\ii\w F_T + 3\nabla^2 F + 3b\w^2|F|^2F = 0,

632: \end{eqn}

633: where the differential operator $\nabla^2 F \equiv F_{\xi\xi} +

634: F_{\eta\eta}$ is isotropic in the $(\xi,\eta)$ variables.   An approximation 

635: to the Townes soliton solution of (\ref{hexcan}) is given by equation 

636: (\ref{vartown}) in the appendix.  Substituting the resulting expression for $F$ 

637: into \eref{hexanz} yields a leading-order expression for the breather

638: \begin{eqn} \lbl{hexlo}

639: Q_{m,n}(t) = 2 \eee \alpha \cos[km+lhn+(\w+\eee^2\lambda)t] 

640: \ \sech(\beta r) + \mcal{O}(\eee^3),

641: \end{eqn}

642: where $\alpha$ and $\beta$ are determined as described by equation 

643: (\ref{alpbet}) in the appendix, with $D=3/2\w$, $B=3b\w/2$.  Further,

644: $r=\sqrt{\xi^2+\eta^2}$ is found in terms of the physical discrete

645: variables $m$ and $n$ by inverting the transformations

646: \eref{hextr} and reverting back to the variables $Z$ and $W$, using

647: \begin{eqnarray} 

648: r^2 &=& \xi^2 + \eta^2 = \frac{ 4 h^2 \eee^2 

649: ( D_2(m\!-\!ut)^2 + D_1(hn\!-\!vt)^2 - D_3(m\!-\!ut)(hn\!-\!vt) )}

650: {4D_1D_2 -D_3^2}.  \nonumber \\ && \lbl{hexr}

651: \end{eqnarray}

652: The terms $D_1$, $D_2$ and $D_3$ are known from \eref{hexrnls}, the 

653: velocities $u$ and $v$ are given by \eref{velsuv} and $\w$ is given in 

654: \eref{hexdisp2}.

655: 

656: 

657: %------------------

658: \subsubsection{Determining the domain of ellipticity. } \lbl{hexdell}

659: 

660: 

661: We confine our attention to elliptic NLS equations.  We seek to determine 

662: the region $\mcal{D}$ of $(k,l)$-parameter space (that is, the two-torus 

663: $\mcal{T}^2 = [(0,2\pi)]\times [0,2\pi/h]$) where the NLS equation 

664: \eref{hexrnls} is elliptic. By definition, this equation is elliptic when 

665: $D_3^2 < 4 D_1 D_2$, where $D_1$, $D_2$ and $D_3$ are given in 

666: \eref{hexrnls}.  Whilst the region $\mcal{D}$ cannot be specified 

667: explicitly,  it is simple to find numerically, and is illustrated in  \Fref{econt}.  

668: Defining the function $e(k,l) = 4D_1(k,l) \! \cdot \! D_2(k,l) - D_3(k,l)^2$, 

669: we are concerned with the region where $e(k,l)>0$.  The subdomains 

670: have been labelled $\{ \mcal{D}_1 ,\ldots, \mcal{D}_6\}$ in \Fref{econt}, 

671: where $\mcal{D} = \mcal{D}_1 \cup \ldots \cup \mcal{D}_6$.  

672: 

673: 

674: \addtolength{\captionmargin}{-15pt}

675: \begin{figure}[ht]

676: \begin{center}

677: \resizebox{4.5in}{!}{\includegraphics{econt.ps}}

678: \put(-250,130){\footnotesize $\mcal{D}_1$}

679: \put(-200,44){\footnotesize $\mcal{D}_2$}

680: \put(-100,44){\footnotesize $\mcal{D}_3$}

681: \put(-48,130){\footnotesize $\mcal{D}_4$}

682: \put(-100,220){\footnotesize $\mcal{D}_5$}

683: \put(-200,220){\footnotesize $\mcal{D}_6$}

684: \caption{The domain $\mcal{D} = \mcal{D}_1 \cup \ldots \cup 

685: \mcal{D}_6$ in which the NLS equation \eref{hexrnls} is elliptic.}

686: \lbl{econt}

687: \end{center}

688: \end{figure}

689: \addtolength{\captionmargin}{+15pt}

690: 

691: 

692: Again, the hexagonal symmetry properties of the HETL are reflected clearly 

693: in the function $e(k,l)$, which has six maxima (at which $e(k,l)=36$), each

694: lying at the centre of one of the closed curves in \Fref{econt}.  The maxima 

695: of $e(k,l)$ coincide with the six maxima of $\w(k,l)$ shown in \Fref{wcont}, 

696: namely, at the wavevectors $\{\mb{k_1},\ldots,\mb{k_6}\}$ in $\mcal{T}^2$; 

697: $e(k,l)$ is minimised ($e(k,l)=-48$) at the six midpoints of the line 

698: segments which connect adjacent maxima. 

699: 

700: 

701: \begin{figure}[ht]

702: \begin{center}

703: \resizebox{2.5in}{!}{\includegraphics{WD1.eps}} \quad 

704: \resizebox{2.5in}{!}{\includegraphics{SPD1.eps}}

705: \end{center}

706: \caption{Contour plots of $\w(k,l)$ (left) and speed 

707: ($\sqrt{u^2+v^2}$, right) for wavevectors $\mb{k}\in\mcal{D}_1$. 

708: Contours are for $2.82\leq\w\leq3$ in steps of $0.02$ and speeds from 

709: zero to 0.6 in steps of 0.1.  The central point corresponds to $\w=3$ and 

710: zero speed, as one considers wave vectors nearer to the edge of 

711: $\mcal{D}_1$, the frequency $\w$ decreases and the speed increases. }

712: \lbl{wspcont}

713: \end{figure}

714: 

715: 

716: Figure \ref{wspcont} shows a contour plot for the frequency $\w$ in 

717: $\mcal{D}_1$: the ellipticity constraint only permits breathers with a 

718: relatively high frequency, that is, with $\w>2.82$.  This constraint in turn 

719: implies that not all breather envelope velocities are attainable, only breathers 

720: with speeds upto about $0.7$ lattice sites per second are permitted; the plot 

721: on the right of Figure \ref{wspcont} shows that only breathers with speeds 

722: upto about $0.3$ sites per second can move in any direction. There is then an 

723: intermediate range of speeds, between $0.3$ and $0.7$ sites per second where 

724: breathers can only move in certain directions, these correspond to the lattice 

725: directions. The larger the speed, the more restricted is the direction of motion. 

726: 

727: 

728: %-------------------

729: \subsubsection{Breather energy. } \lbl{hemv}

730: 

731: 

732: To calculate the leading-order energy $\elo$ of moving breathers in 

733: lattices with a symmetric potential we use \eref{hexlo} which we write as

734: \begin{eqn} \lbl{hxs2}

735: Q_{m,n}(t) \sim 2 \eee \alpha \cos \Phi \ \sech(\beta r),

736: \end{eqn}

737: where $\Phi = km + lhn + \om t$ is the phase of the carrier wave,

738: $(k,l) \in \mcal{D}$, $\om = \w + \eee^2\lambda$ is the breather

739: frequency including the first correction term,  ($\psi=km+lhn+\w t$ 

740: is only the leading-order expression).  $\ld$ parameterises the 

741: breather amplitude, $\alpha=\alpha(\lambda)$, $\beta=\beta(\lambda)$ 

742: and $r^2$ are as described in equations \eref{hexlo} and \eref{hexr}.  

743: 

744: 

745: We now find expressions for the currents $I_{m,n}$, $J_{m,n}$ and

746: $K_{m,n}$: the current $I_{m,n}$ is obtained by substituting the 

747: expression for $Q_{m,n}$ \eref{hxs2} into \eref{hkiq} and integrating 

748: with respect to time.   Owing to the complexity of the expression for $r$ 

749: given by \eref{hexr}, the left-hand side of \eref{hkiq} can not be integrated 

750: with respect to time. However, the variable $r$ varies more slowly in time 

751: than $\Phi$. Hence,  integration by parts (using $\int f'(t) g(\eee t) \dd t = 

752: [f(t) g(\eee t)] - \eee \int f(t) g'(\eee t)\dd t$) gives, to leading-order, 

753: \begin{eqn} \lbl{himov} \fl

754: I_{m,n}\sim\frac{2\eee\alpha}{\w}\lft[1-\cos(2k)\rit]\sin\Phi\,\sech(\beta r)

755: - \frac{2\eee\alpha}{\w} \sin(2k) \cos \Phi \,\sech(\beta r),

756: \end{eqn}

757: where we have taken the constant of integration to be zero, and

758: $\om \sim \w$ to leading order.  Similarly, substituting for $Q_{m,n}$ in 

759: equations \eref{hkjq} and \eref{hkkq} and integrating, we find 

760: \begin{eqnarray} \fl

761: J_{m,n} & \sim & \frac{2\eee\alpha}{\w} \lft[ 1-\cos(k-lh) \rit] 

762: \sin \Phi\, \sech(\beta r) - 

763: \frac{2\eee\alpha}{\w}\sin(k-lh)\cos\Phi\,\sech(\beta r), \lbl{hjmov}\\[1ex] 

764: \fl  K_{m,n} & \sim & \frac{2\eee\alpha}{\w} \lft[ 1-\cos(k+lh) \rit] 

765: \sin \Phi\, \sech(\beta r)

766: - \frac{2\eee\alpha}{\w} \sin(k+lh) \cos \Phi \,\sech(\beta r). \lbl{hkmov}

767: \end{eqnarray}

768: Substituting these expressions into \eref{hentotlo}, we obtain 

769: \begin{eqnarray}

770: E_0 & \sim \sum_{m,n} & \frac{2\eee^2\alpha^2}{C_0} 

771: \cos^2 \Phi \ \sech^2(\beta r) \nonumber \\ & & + 

772: \frac{2L\eee^2\alpha^2}{\w^2}\sech^2(\beta r) \lft\{ \lft[ 

773: (1-\cos(2k))\sin\Phi - \sin(2k)\cos\Phi  \rit]^2 \right. \nonumber \\

774: & & + \lft[ (1-\cos(k-lh))\sin\Phi - \sin(k-lh)\cos\Phi \rit]^2 \nonumber \\ 

775: & & + \left. \lft[ (1-\cos(k+lh))\sin\Phi - \sin(k+lh)\cos\Phi \rit]^2 \rit\} .  

776: \lbl{hmvan} \end{eqnarray}

777: We replace the sum by an integral since the variables $Z=\eee (m-ut)$ 

778: and $W=\eee (hn-v t)$  vary slowly with $m$ and $n$.  To simplify the 

779: resulting integral, we approximate the terms $\cos^2\Phi$, $\sin^2\Phi$ 

780: and $\sin\Phi\cos\Phi$ by their average values of $\sfr{1}{2}$, $\sfr{1}{2}$ 

781: and $0$ respectively.  Hence \eref{hmvan} becomes 

782: \begin{eqn} \lbl{henmvint}

783: E_0 \sim \sum_{m,n} \frac{2\eee^2\alpha^2}{C_0} \sech^2(\beta r).

784: \end{eqn}

785: 

786: 

787: {}From the definition of $r$ (\ref{hexr}) we note the function 

788: $\sech(\beta r)$ is not in general radially symmetric in $m$, $n$.  Hence 

789: we work in $(\xi,\eta)$-space to facilitate evaluation of the double integral 

790: which approximates the double sum in (\ref{henmvint}).   Evaluating the 

791: Jacobian associated with the transformation from $(m,n)$ to $(\xi,\eta)$ 

792: coordinates, we find 

793: \begin{eqn}

794: E_0 \sim \frac{\alpha^2}{h^3C_0}\sqrt{4D_1D_2 - D_3^2} \int \!\!\! \int

795: \sech(\beta\sqrt{\xi^2+\eta^2}) \ \dd \xi \dd \eta. 

796: \lbl{hexemtri} \end{eqn}

797: Evaluating this and substituting for $\alpha$ and $\beta$ in terms of 

798: $D=3/2\w(k,l)$ and $B=3b\w(k,l)/2$ (see the appendix) we find 

799: \begin{eqn} \lbl{henar}

800: E_0 \sim \frac{4\pi\log 2(2\log 2 + 1)}

801: {3\, h^3 \, C_0 \, b \, \w^2(4\log 2 - 1)} \sqrt{4D_1D_2 - D_3^2}.

802: \end{eqn}

803: It is evident from \eref{henar} that the leading-order energy $E_0$ is 

804: independent of the breather amplitude, again confirming the existence 

805: of a minimum energy of moving breathers in the two-dimensional HETL 

806: with symmetric potential.  However, the threshold energy does depend 

807: upon the wavenumbers $k$ and $l$ hence moving breathers have a 

808: different threshold energy.   A plot of the expression \eref{henar} is 

809: shown in \Fref{henmov}.  We see that $\elo$, given by \eref{henar}, is 

810: strictly positive in the region of ellipticity $\mcal{D}$, and is maximised 

811: (attaining the same value) at each of the points corresponding to 

812: wavevectors $\{ \mb{k_1},\ldots, \mb{k_6} \}$, that is, at the points which

813: correspond to stationary breathers.  The threshold energy (\ref{henar}) 

814: decays to zero towards the boundary of the elliptic domain $\mcal{D}$ (see 

815: \Fref{econt}).   Hence, as for breathers in the square lattice, the energy 

816: threshold for moving breathers is {\em lower} than that for stationary 

817: breathers.  The threshold becomes arbitrarily small at the boundary of the 

818: domain of ellipticity.

819: 

820: 

821: \begin{figure}[htbp]

822: \begin{center}

823: \resizebox{3.5in}{!}{\includegraphics{henmov.ps}}

824: \caption{Plot of $E_0(k,l)$ for lattices with a symmetric potential.}

825: \lbl{henmov}

826: \end{center}

827: \end{figure}

828: 

829: 

830: %-------------------------------------------------------------------

831: \subsection{Lattices with an asymmetric potential} \lbl{hexasymm}

832: 

833: 

834: In this section, we consider the more general scenario where the potential 

835: $\Upsilon\!\ '(Q)$ is asymmetric.  In this case, the terms $a$ and $c$ in 

836: \eref{hqvexp} and \eref{heqq} are not both zero.   If \eref{hexnls} is to 

837: be reduced to an NLS equation in  $F$, both $G_0$ and $G_2$ must be 

838: found in terms of $F$.  Whilst $G_2$ is given by the simple algebraic 

839: equation \eref{hexg2}; in order to find $G_0$, the partial differential 

840: equation \eref{hexfo} must be solved.  We  assume that $G_0$ travels 

841: at the same velocity as $F$, that is, $G_0(X,Y,\tau,T) \equiv G_0(Z,W,T)$.  

842: Eliminating $G_{0\tau\tau}$,  \eref{hexfo} becomes

843: \begin{eqn} \lbl{hexg0}

844: (u^2 - 6)G_{0ZZ} + (v^2 - 2h^2)G_{0WW} + 2uvG_{0ZW} =

845: 6a|F|_{ZZ}^2 + 2ah^2|F|_{WW}^2.

846: \end{eqn}

847: For general $k$ and $l$, it is difficult to solve for $G_0$ explicitly.  

848: However, for any of the wavevectors $\{\mb{k_1},\ldots,\mb{k_6}\}$, 

849: the velocities $u$ and $v$ become zero, and so \eref{hexg0} becomes 

850: $\nabla^2 G_0 = -a\nabla^2|F|^2$, where the operator $\nabla^2$, 

851: defined by $\nabla_{(Z,W)}^2 \equiv \clyd_Z^2 + \clyd_W^2$, is 

852: equivalent to $\nabla_{(X,Y)}^2 \equiv \clyd_X^2 + \clyd_Y^2$. It 

853: follows that $G_0 = -a|F|^2$ for each of these wavevectors.

854: 

855: 

856: Equations \eref{hexdisp}--\eref{hexfo} are the same whichever 

857: of the wavevectors $\{\mb{k_1},\ldots,\mb{k_6}\}$ is used.  Thus, 

858: \eref{hexdisp} gives $\w=3$, and  \eref{hexg2} implies $G_2 = aF^2/3$.  

859: Substituting these expressions for $G_0$ and $G_2$ in  \eref{hexnls} 

860: gives the following NLS equation for asymmetric potentials for any 

861: of the wavevectors $\{\mb{k_1},\ldots,\mb{k_6}\}$

862: \begin{eqn} \lbl{hnlsasy}

863: 2\ii\w F_T + 3 \nabla^2 F + \w^2(3b-\mbox{$\frac{10}{3}$}a^2)|F|^2F=0,

864: \end{eqn}

865: The anomalous dispersion regime thus corresponds to $b>10a^2/9$.  

866: Soliton solutions are derived in the appendix;   \eref{hnlsasy} 

867: corresponds to $D = 1/2$ and $B =(9b-10a^2)/2$, and also 

868: $r^2 = X^2 + Y^2$.   Substituting the solution for $F$ into the lattice 

869: ansatz \eref{hexanz}, along with the known expressions for $G_0$ and 

870: $G_2$ gives the following second-order formula for stationary 

871: breathers in lattices with an asymmetric potential 

872: \begin{eqnarray} 

873: \fl Q_{m,n}(t) = 2\ \! \eee  \ \! \alpha\cos[km+lhn+(\w+\eee^2\lambda)t]\ 

874: \sech(\beta r) \nonumber \\ 

875: + \mbox{$\frac{2}{3}$}\ \! a \ \! \eee^2 \ \! \alpha^2 \sech^2(\beta r)\!\lft\{ 

876: \cos[2km+2lhn+(2\w+2\eee^2\lambda)t] - 3 

877: \rit\} + \mcal{O}(\eee^3), \lbl{hexqasym} 

878: \end{eqnarray} 

879: where $\alpha$ and $\beta$ are determined in the appendix, 

880: $r=\sqrt{X^2+Y^2}$, and $X=\eee m$ and $Y=\eee h n$ in terms of the 

881: original discrete variables $m$ and $n$.

882: 

883: 

884: %------------------

885: \subsubsection{Breather energy. } \lbl{hest}

886: 

887: 

888: We calculate the leading-order energy $\elo$ of stationary breathers in 

889: lattices with an asymmetric potential.  In this case, stationary breathers 

890: are given by \eref{hexqasym}, from which we take only the leading order term 

891: \begin{eqn} \lbl{hxa2}

892: Q_{m,n}(t) \sim 2 \eee \alpha \cos \Theta \ \sech(\beta r),

893: \end{eqn}

894: where $\Theta = km + lhn + \om t$ is the phase of the carrier wave 

895: with $(k,l)$ corresponding to one of $\{ \mb{k_1},\ldots, \mb{k_6}\}$, 

896: $\om = \w+\eee^2\lambda$ is the breather frequency, and $w=3$. 

897: 

898: 

899: The currents $I_{m,n}$, $J_{m,n}$ and $K_{m,n}$ are obtained 

900: by substituting \eref{hxa2} into equations \eref{hkiq}--\eref{hkkq} 

901: and integrating with respect to time, taking the constant of 

902: integration to be zero.   Thus, to leading order, we find

903: \begin{eqnarray}

904: I_{m,n} & \sim & \frac{3\eee\alpha}{\w} \sin \Theta\, \sech(\beta r)

905: - \frac{\sqrt{3}\eee\alpha}{\w} \cos \Theta \,\sech(\beta r),\lbl{heni} \\[1ex]

906: J_{m,n} & \sim & \frac{3\eee\alpha}{\w} \sin \Theta\, \sech(\beta r)

907: + \frac{\sqrt{3}\eee\alpha}{\w} \cos \Theta \,\sech(\beta r), \lbl{henj} \\[1ex]

908: K_{m,n} & \sim & \frac{3\eee\alpha}{\w} \sin \Theta\, \sech(\beta r)

909: + \frac{\sqrt{3}\eee\alpha}{\w} \cos \Theta \,\sech(\beta r).  \lbl{henk}

910: \end{eqnarray}

911: Substituting these expressions into  \eref{hentotlo} 

912: gives the leading-order energy 

913: \begin{eqnarray} \fl

914: E_0 \sim \sum_{m,n} \frac{2\eee^2\alpha^2}{C_0} 

915: \cos^2 \Theta \ \sech^2(\beta r) \nonumber \\

916: +  \frac{Lh^2}{2\w^2}\,\eee^2\alpha^2  \lft[  3h^2\sin^2\Theta + 

917: 2h\sin\Theta\cos\Theta + 3\cos^2\Theta  \rit]  \sech^2(\beta r)  . 

918: \lbl{henas}

919: \end{eqnarray}

920: We approximate the term in square brackets by taking the average

921: values of $\cos^2 \Theta$, $\sin^2\Theta$, $\sin\Theta\cos\Theta$ 

922: as \mbox{$\frac{1}{2}$}, \mbox{$\frac{1}{2}$} and $0$ respectively. 

923: 

924: 

925: We replace the double sum by an integral giving

926: \begin{eqn} \lbl{henint}

927: E_0 \sim  \frac{2\alpha^2}{hC_0} \int \!\!\! \int

928: \sech^2(\beta \sqrt{X^2+Y^2}) \ \dd X \dd Y.

929: \end{eqn}

930: This can be evaluated to 

931: \begin{eqn} \lbl{henex}

932: E_0 \sim \frac{4\pi\log 2}{h C_0} \frac{\alpha^2}{\beta^2}

933: = \frac{8 \pi \log 2(2\log 2+1)}{C_0h(9b-10a^2)(4\log 2-1)} . 

934: \end{eqn}

935: Again, this estimate for energy is independent of the breather

936: amplitude $\lambda$, demonstrating the energy threshold properties

937: of the two-dimensional HETL; namely, the activation energy

938: required to create a breather in the HETL is an $\mcal{O}(1)$

939: quantity, irrespective of its amplitude ($2\eee\alpha\ll 1$).

940: 

941: 

942: %==============================================

943: \section{Higher-order asymptotic analysis}  \lbl{hex5}

944: 

945: 

946: The cubic NLS equation exhibits blow-up, which cannot occur in a discrete 

947: system since energy is conserved, and even if all the energy were localised 

948: at a single site, the amplitude would still be finite.   The two-dimensional 

949: cubic NLS equation description of the breather envelope is lacking.   

950: Higher-order dispersive and nonlinear effects play an important role in the 

951: dynamics of such discrete systems and therefore must be incorporated.  In 

952: this section we extend our analysis of the lattice equations \eref{heqq} to 

953: fifth-order, and derive a generalised NLS equation which includes higher-order 

954: dispersive and nonlinear terms.  It then remains to determine whether this 

955: generalised NLS equation supports stable soliton solutions.  

956: 

957: 

958: Due to the complexity of a fifth-order analysis of  a general asymmetric 

959: potential, we consider only lattices with a symmetric potential, that is, 

960: those for which $a=c=0$ in \eref{heqq}. Since no second or fourth harmonic

961: terms are generated by the nonlinearity, we use a much simpler ansatz, namely 

962: \begin{eqn} \lbl{h5anzgen}

963: Q_{m,n}(t) = \eee \ee^{\ii \psi} F(X,Y,\tau,T)

964: + \eee^3 \ee^{3\ii \psi}H_3(X,Y,\tau,T) + \cdots + \mathrm{c.c.,}

965: \end{eqn}

966: where the phase $\psi = km+lhn+\w t$.  In this case, in addition to the 

967: equations \eref{hexdisp2}--\eref{velsuv}, \eref{hexrnls} and \eref{hexh3}, 

968: we also have \\ 

969: \\

970: $\mcal{O}(\eee^5 \ee^{\ii\psi})$:

971: \begin{eqnarray} \fl 

972: F_{TT} =&& \hspace*{-12mm}

973: \mbox{$\frac{1}{6}$} \lft[ 8\cos(2k) + \cos k \cos(lh) \rit] F_{XXXX} 

974: + h^2\cos k \cos(lh)F_{XXYY}  \nonumber \\ &&\hspace*{-16mm}

975: + \mbox{$\frac{1}{6}$} h^4 \cos k \cos(lh)F_{YYYY} 

976: - \mbox{$\frac{2}{3}$} h \sin k \sin(lh) F_{XXXY} 

977: - \mbox{$\frac{2}{3}$} h^3 \sin k \sin(lh)F_{XYYY} 

978: \nonumber\\ &&\hspace*{-16mm}

979: + 6b \lft[ 2\cos(2k) + \cos k \cos(lh) \rit](|F|^2F)_{XX}

980: + 6bh^2\cos k\cos(lh)(|F|^2F)_{YY} \nonumber \\ && \hspace*{-16mm}

981: -12b\sin k \sin(lh) \lft[  2hFF_X\tbar{F}_Y  + 2hFF_Y \tbar{F}_X

982: + 2h\tbar{F}F_XF_Y + hF^2\tbar{F}_{XY} + 2hF\tbar{F}F_{XY} \rit] 

983: \nonumber \\  && \hspace*{-16mm} 

984: - 3b\w^2\tbar{F}^2H_3 - 10\w^2 d |F|^4F . 

985: \lbl{hfogen} \end{eqnarray}

986: We simplify this by restricting attention to stationary breathers.

987: Accordingly, only one extra timescale, $T=\eee^2 t$, is required, and we 

988: fix the wavenumbers $k$ and $l$ to correspond to one of the wavevectors 

989: $\{ \mb{k_1},\ldots, \mb{k_6} \}$.  Hence we obtain the equations: \\

990: \\

991: $\mcal{O}(\eee^3 \ee^{\ii\psi})$: \\

992: \begin{eqn} \lbl{h5nls}

993: 2\ii \w F_T + 3\nabla^2 F + 3b\w^2 |F|^2F = 0,

994: \end{eqn}

995: $\mcal{O}(\eee^5 \ee^{\ii\psi})$: \\

996: \begin{eqn} \lbl{h550}

997: F_{TT} = -\mbox{$\frac{3}{4}$} \nabla^4F - 9b\nabla^2(|F|^2F) -

998: 10\w^2d|F|^4F , 

999: \end{eqn} 

1000: in addition to $\w=3$ (from the $\mcal{O}(\eee \ee^{\ii\psi})$

1001: equation) and $H_3=0$ (from $\mcal{O}(\eee^3 \ee^{3\ii\psi})$). 

1002: A consequence of the hexagonal symmetry of the HETL is that all 

1003: differentials on the right-hand side of \eref{h550} are isotropic.  

1004: 

1005: 

1006: In order to obtain a generalised NLS equation, we combine the

1007: higher-order equation \eref{h550} with the cubic two-dimensional

1008: NLS equation \eref{h5nls}.  The $F_{TT}$ term on the

1009: left-hand side of \eref{h550} is eliminated  by

1010: differentiating \eref{h5nls} with respect to $T$ and substituting

1011: for $F_{TT}$ into \eref{h550}.  The resulting expression is 

1012: \begin{eqnarray}

1013: 6 \ii F_T + 3\nabla^2 F + 27b|F|^2F + \frac{\eee^2}{2} \nabla^4 F

1014: + \frac{9\eee^2}{4}(40d-27b^2)|F|^4F & \nonumber \\ \qquad \qquad 

1015: + \frac{27b\eee^2}{4} \nabla^2(|F|^2F) - \frac{9b\eee^2}{2}|F|^2\nabla^2F

1016: - \frac{9b\eee^2}{4} F^2\nabla^2\tbar{F} = 0 , & \lbl{hexgennls}

1017: \end{eqnarray}

1018: which is isotropic. 

1019: 

1020: 

1021: To the best of our knowledge, this perturbed form of the NLS equation 

1022: has not been studied in the literature before; although it is similar to 

1023: and slightly simpler than the corresponding equation derived for the 

1024: square lattice \cite{buts06}.  It is also similar to the perturbed NLS 

1025: equation considered by Davydova \etal \citep{dav03}, namely 

1026: \begin{equation}

1027: \ii F_T + D \nabla^2 F + B |F|^2 F + P \nabla^4 F + K |F|^4 F =0 , 

1028: \lbl{davydova} 

1029: \end{equation} 

1030: which is known to have stable soliton solutions.  The anomalous 

1031: dispersion case is $BD>0$, which in our equation (\ref{hexgennls}) 

1032: corresponds to $b>0$;   Davydova's criteria for soliton existence is 

1033: $PK>0$ which implies $40d > 27b^2$.  Hence, it is in this parameter 

1034: regime that we seek breathers solutions of the HETL.  The numerical 

1035: results presented in  \Sref{hexnumerics} show that long-lived breather 

1036: solutions are supported by the two-dimensional hexagonal FPU lattice, 

1037: suggesting that the additional perturbing terms in \eref{hexgennls}

1038: do not destabilise the Townes soliton.

1039: In fact, numerical simulations (presented in \cite{thesis} but not here) of the 

1040: case $40d<27b^2$ show the breather mode to be long-lived, suggesting 

1041: that the terms on the second line of (\ref{hexgennls}) are stabilising. 

1042: 

1043: 

1044: We have been unable to find a variational formulation of

1045: \eref{hexgennls}, and are therefore unable to use the methods of 

1046: Davydova {\em et al.}\ \cite{dav03} and Kuznetsov {\em et al.}\ 

1047: \citep{kuz}.  Alternative possible methods include the modulation 

1048: theory of Fibich and Papanicolaou \citep{fib98,fib99}.

1049: 

1050: 

1051: %=============================================

1052: \section{Numerical results}  \lbl{hexnumerics}

1053: 

1054: \subsection{Preliminaries} \lbl{hnumpre}

1055: 

1056: 

1057: Using a fourth-order Runge-Kutta scheme, we solve the equations 

1058: \begin{eqnarray}

1059: \frac{\dd Q_{m,n}}{\dd t} & = & R_{m,n}, \nonumber \\[1ex]

1060: \frac{\dd R_{m,n}}{\dd t} & = & (\delta^2_I + \delta^2_J +

1061: \delta^2_K)

1062: \lft[ Q_{m,n} + a Q_{m,n}^2 + b Q_{m,n}^3 + c Q_{m,n}^4 + d

1063: Q_{m,n}^5 \rit], \lbl{heqqfo}

1064: \end{eqnarray}

1065: numerically.  Introducing the variable $R_{m,n}$ converts the 

1066: system of second-order ordinary differential equations \eref{heqq} 

1067: to an equivalent system of first-order differential equations.  

1068: 

1069: 

1070: We present the results of simulations for a range of parameter

1071: values.  From \eref{velsuv}, the velocity of the envelope $(u,v)$ 

1072: depends upon the wavevector $\mb{k}=[k,l]^T$, and hence we 

1073: obtain moving breathers by choosing $(k,l) \in \mcal{D}$.  

1074: In \Fref{ellcontnum}, we show the points in  $\mcal{D}_1 \subset  

1075: \mcal{D}$ for which we solve the lattice equations \eref{heqqfo} 

1076: numerically.  These points correspond to the wavevectors $\mb{k_1} 

1077: = [\pi/3,\pi/h]^T$, $\mb{k_a} = [1.4,\pi/h]^T$, $\mb{k_b} = 

1078: [0.79,1.7324]^T$ and $\mb{k_c} = [0.8,1.9987]^T$.  

1079: Selecting $(k,l)$ too near to the boundary of $\mcal{D}$ results 

1080: in a sharply elongated breather, which are difficult to simulate as 

1081: they require a large domain.  Wavevectors close to any of 

1082: $\{ \mb{k_1},\ldots, \mb{k_6} \}$, lead to breather modes with small 

1083: speeds, a check of breather velocity would then require a long-time 

1084: simulation.  In practice we have chosen wavenumbers which do not 

1085: lead to a severely elongated breather envelope, and yet have velocities 

1086: which result in observable displacements over reasonable times.   We 

1087: present simulations of stationary breathers in systems with asymmetric 

1088: potentials ($a,c \neq 0$) in \Sref{hnra} and, in Sections 

1089: \ref{hnstsy}--\ref{hpsi130}, simulations of stationary and moving breathers 

1090: in lattices with symmetric potentials (that is, $a=c=0$ in \eref{heqqfo}). 

1091: 

1092: 

1093: \begin{figure}[htbp]

1094: \begin{center}

1095: \resizebox{4in}{!}{\includegraphics{ellcontnum.eps}}

1096: \put(-136,114){\scriptsize $\mb{k_1}$}

1097: \put(-89,114){\scriptsize $\mb{k_a}$}

1098: \put(-172,91){\scriptsize $\mb{k_b}$}

1099: \put(-170,140){\scriptsize $\mb{k_c}$}

1100: \put(-87,164){\fns $\mcal{D}_1$}

1101: \caption{Wavevectors in $\mcal{D}_1 \subset \mcal{D}$ for 

1102: which breathers are simulated.}

1103: \lbl{ellcontnum}

1104: \end{center}

1105: \end{figure}

1106: 

1107: 

1108: %-----------------------------------------------------------------

1109: \subsection{Initial data and boundary conditions} \lbl{hbcics}

1110: 

1111: 

1112: We generate initial data by using the analytic expressions for

1113: breather solutions derived in \Sref{haa}. The formulae for $Q_{m,n}$ 

1114: and $R_{m,n}$ are found in terms of the original discrete variables 

1115: $m$ and $n$, and then shifted horizontally and vertically so that initially 

1116: the breather lies at the centre of the lattice.   We impose periodic 

1117: boundary conditions for the lattice in both horizontal and vertical 

1118: directions, converting the two-dimensional arrangement illustrated 

1119: in \Fref{hexetl} into a two-torus.  In long-time simulations, moving 

1120: breathers which approach an edge of the lattice reappear from the 

1121: opposite edge.  We select the site $(1,1)$ to lie at the bottom left-hand 

1122: corner of the arrangement as illustrated in \Fref{hexbcs}.  Sites along 

1123: the boundaries and corners are missing between one and four neighbours.  

1124: We introduce fictitious sites along the boundaries and corners 

1125: where necessary to effect periodic boundary conditions.  

1126: 

1127: 

1128: An illustration of a small finite lattice is shown in \Fref{hexbcs}.  

1129: The dots represent capacitors located at lattice sites, and the 

1130: lines represent some of the inter-connecting inductors.  From 

1131: equations \eref{heqqfo},  the charge $Q_{m,n}$ stored on 

1132: each capacitor depends upon the charge stored on the capacitors 

1133: located at its six neighbouring sites,  two in each of the directions 

1134: $\mb{e_i}$, $\mb{e_j}$ and $\mb{e_k}$.  

1135: For the sake of clarity, the inductors connecting the capacitors at

1136: the center of each hexagon to its six nearest neighbours are not

1137: shown (see \Fref{hexetl}).  It is not necessary that the lattice

1138: should be ``square,'' meaning that the lattice could comprise $M

1139: \times N$ lattice sites, with $M \neq N$.  However, typically, we

1140: consider $M=N$ as in \Fref{hexbcs}, since the numerical routines

1141: are simpler to encode when the lattice \itc{is} square, and in  

1142: the examples below, we consider lattices with $N \leq 50$.  

1143: 

1144: 

1145: \addtolength{\abovecaptionskip}{10pt}

1146: \begin{figure}[htbp]

1147: \begin{center}

1148: \vspace{1.0cm}

1149: \resizebox{2.5in}{!}{\includegraphics{hexbcs.eps}}

1150: \put(-204,-10){${Q_{1,1}}$}

1151: \put(-10,-10){${Q_{2N-1,1}}$}

1152: \put(-173,153){${Q_{2,N}}$}

1153: \put(-18,153){${Q_{2N,N}}$}

1154: \caption{Periodic boundary conditions for the two-dimensional HETL.}

1155: \lbl{hexbcs}

1156: \end{center}

1157: \end{figure}

1158: \addtolength{\abovecaptionskip}{-10pt}

1159: 

1160: 

1161: %--------------------------------------------------------------------

1162: \subsection{Numerical computation of breather energy} \lbl{hnumen}

1163: 

1164: 

1165: Since the HETL is lossless the total energy is conserved and 

1166: can be used as a check of the accuracy of our numerical scheme.   

1167: We compute the leading-order energy by expressing the summand 

1168: $e_{m,n}^{(0)}$ \eref{hentotlo} in terms of the output variables of 

1169: the numerical routine, namely $Q_{m,n}$ and $R_{m,n}$.  The first 

1170: term of $e_{m,n}^{(0)}$ is simply dependent on $Q_{m,n}$.  It

1171: remains to find the currents $I_{m,n}$, $J_{m,n}$ and $K_{m,n}$ in

1172: terms of $Q_{m,n}$ and $R_{m,n}$.  The details of this calculation 

1173: depend upon whether the interaction potential is symmetric or asymmetric.

1174: As a check of the numerical scheme we verify that the sum is 

1175: conserved and that it agrees with the asymptotic estimate \eref{henar}.

1176: 

1177: 

1178: %-------------------------------------------------------------

1179: \subsubsection{Lattices with a symmetric potential. }

1180: 

1181: 

1182: First, we find the current $I_{m,n}$ in terms of $Q_{m,n}$ and

1183: $R_{m,n}$.  Differentiating the leading-order expression for the breather 

1184: given by \eref{hxs2}, we have (retaining leading-order terms only)

1185: \begin{eqn} \lbl{hnem}

1186: \dot{Q}_{m,n} = R_{m,n} \sim -2\eee\alpha \w \sin \Phi \ \sech (\beta r),

1187: \end{eqn}

1188: where $\alpha$, $\beta$ and $r^2$ are as defined in \Sref{hexsym}.

1189: Comparing the analytic expression \eref{himov} for the current

1190: $I_{m,n}$ with the expressions for $Q_{m,n}$ \eref{hxs2} and

1191: $R_{m,n}$ \eref{hnem}, we find

1192: \begin{eqn} \lbl{hniqr}

1193: I_{m,n} = \frac{\cos(2k)-1}{\w^2} R_{m,n} - \frac{\sin(2k)}{\w}Q_{m,n}.

1194: \end{eqn}

1195: Similarly, comparing the expressions \eref{hjmov} and \eref{hkmov}

1196: for the currents $J_{m,n}$ and $K_{m,n}$ respectively with

1197: equations \eref{hxs2} and \eref{hnem}, we find 

1198: \begin{eqnarray}

1199: J_{m,n} & = & \frac{\cos(k-lh)-1}{\w^2} R_{m,n} -

1200: \frac{\sin(k-lh)}{\w}Q_{m,n},

1201: \lbl{hnjqr} \\[1ex]

1202: K_{m,n} & = & \frac{\cos(k+lh)-1}{\w^2} R_{m,n} -

1203: \frac{\sin(k+lh)}{\w}Q_{m,n}. \lbl{hnkqr}

1204: \end{eqnarray}

1205: Substituting these expressions into \eref{hentotlo}, gives an expression 

1206: for $E_0$ which does not simplify, so we do not reproduce it here.  

1207: 

1208: 

1209: %------------------- 

1210: \subsubsection{Lattices with an asymmetric potential. }

1211: 

1212: 

1213: Differentiating the leading-order expression \eref{hxa2} for the

1214: charge $Q_{m,n}$ in lattices with an asymmetric potential gives

1215: \begin{eqn} \lbl{hxa2dot}

1216: \dot{Q}_{m,n} = R_{m,n} \sim 

1217: -2\eee\alpha \w \sin \Theta \ \sech (\beta r) , 

1218: \end{eqn}

1219: where $\Theta=\pi m/3 +\pi n + \om t$ with $\om=3$.  Using \eref{hxa2} 

1220: and \eref{hxa2dot}, the current $I_{m,n}$  given by \eref{heni} can be 

1221: expressed in terms of $Q_{m,n}$ and $R_{m,n}$ as

1222: \begin{eqn} \lbl{hiqqd}

1223: I_{m,n} \sim -\frac{\sqrt{3}}{2\w} \lft[ \frac{\sqrt{3}R_{m,n}}{\w} + 

1224: Q_{m,n} \rit]\!.

1225: \end{eqn}

1226: Similarly, from  \eref{henj}--\eref{henk}, we have

1227: \begin{eqn} \lbl{hjkqqd}

1228: J_{m,n}  = K_{m,n} \sim -\frac{\sqrt{3}}{2\w} 

1229: \lft[ \frac{\sqrt{3}R_{m,n}}{\w} - Q_{m,n} \rit]\!.

1230: \end{eqn}

1231: These are equivalent to substituting $k=\pi/3$ and $l=\pi/h$ into 

1232: (\ref{hniqr})--(\ref{hnkqr}).  Inserting the expressions for $I_{m,n}$, 

1233: $J_{m,n}$ and $K_{m,n}$ (\eref{hiqqd}--\eref{hjkqqd}) into \eref{hentotlo} yields 

1234: \begin{eqn} \lbl{hentotqqd}

1235: E_0 = \sum_{m,n} \frac{1}{72 C_0} \lft[ 

1236: 45Q_{m,n}^2 + 3R_{m,n}^2 - 2\sqrt{3}Q_{m,n}R_{m,n} \rit]\! . 

1237: \end{eqn}

1238: 

1239: 

1240: %------------------------

1241: \subsubsection{Effective breather width. }

1242: 

1243: 

1244: Numerical computation of the energy as described above allows us to 

1245: check that the total lattice energy is conserved, but does not indicate 

1246: whether a breather changes shape over time.  To remedy this, we define 

1247: breather widths in the $m$ and $n$ directions, the sum of which we 

1248: denote $\mcal{W}_{\mathrm{br}}$, where

1249: \begin{eqn} \lbl{hbrwdth}

1250: \mcal{W}_{\mathrm{br}}^2 = \frac{r_{20}}{\elo} + \frac{r_{02}}{\elo}

1251: - \lft( \frac{r_{10}}{\elo} \rit)^2 -  \lft( \frac{r_{01}}{\elo} \rit)^2,

1252: \end{eqn}

1253: and $r_{10}$, $r_{01}$, $r_{20}$ and $r_{02}$ are defined by

1254: \addtolength{\arraycolsep}{0.5cm}

1255: \begin{displaymath}

1256: \begin{array}{c c}

1257: r_{10}=\sum_{m,n} m e_{m,n},\quad&r_{20}=\sum_{m,n}m^2 e_{m,n},\\[8pt]

1258: r_{01}=\sum_{m,n} hn e_{m,n},&r_{02} = \sum_{m,n} h^2n^2 e_{m,n}. 

1259: \end{array}

1260: \end{displaymath}

1261: \addtolength{\arraycolsep}{-0.5cm}

1262: The variation in $\wbr$ over time gives a measure of the distortion 

1263: suffered by a breather.

1264: 

1265: 

1266: %------------------------------------------------------------------

1267:  \subsection{Stationary breather in a lattice with symmetric potential}

1268: \lbl{hnstsy}

1269: 

1270: 

1271: We investigate stationary breather solutions in the anomalous dispersion

1272: regime which corresponds to $b>0$ (see \eref{hexgennls}).  We set 

1273: $k=\pi/3$ and $l=\pi/h$, corresponding to $\mb{k_1}$ in \Fref{ellcontnum}, 

1274: so that the velocities $u$ and $v$ are zero.   Davydova's result 

1275: implies that we expect to find stable soliton solutions when $PK>0$.  

1276: {}From \eref{davydova}, $P=\eee^2/2$, hence this inequality implies 

1277: $40d > 27b^2$.  Hence we choose $b=d=1$ as well as $N=30$, $\eee=0.2$ 

1278: and $\lambda=1$.  Using the technique outlined in the appendix we 

1279: calculate $\alpha=1.0212$, $\beta=1.8670$.  The breather frequency is 

1280: $\w+\eee^2\lambda=3.040$, and therefore the period of oscillation is $T=2.0668$. 

1281: 

1282: 

1283: As is the case for all our simulations, the initial profile of the breather 

1284: is located at the centre of the lattice, as illustrated in  \Fref{simh1}\,(a). 

1285: At $t=0$, the breather energy is $\elo=0.7606$, and $\wbr=3.74$.

1286: The asymptotic estimate for $E_0$ given by \eref{henar} is 0.7523, 

1287: which is only 1\% different from the numerically obtained value.

1288: In \Fref{simh1}\,(c) we show the breather after 30 oscillations.  

1289: Plots of the local energy $e_{m,n}$ at $t=0$

1290: and $t=30T$ are presented in Figures \ref{simh1}\,(b) and

1291: \ref{simh1}\,(d).  After thirty oscillations the breather has shed 

1292: a small amount of energy, which is manifested as small amplitude 

1293: radiation throughout the lattice.  Accordingly, the breather 

1294: appears a little distorted in shape compared to its initial profile.  

1295: In particular, at $t=30T$, we find $\wbr=4.21$. The energy of the 

1296: breather at $t=30T$ is 0.7087, giving $\Delta E_0/E_0=-0.0682$, 

1297: 

1298: 

1299: \begin{figure}[htbp]

1300: \centering % PICTURE 1

1301: \subfigure[Profile at $t=0$.]{

1302: \includegraphics[width=.45\textwidth]{hs1a.eps}}

1303: \hspace{.3in} % PICTURE 2

1304: \subfigure[Plot of $e_{m,n}$, $E_0=0.7606$.]{

1305: \includegraphics[width=.45\textwidth]{hs1aen.eps}}\\

1306: \vspace{.3in} % PICTURE 3

1307: \subfigure[Profile at $t=30T=62.82$.]{

1308: \includegraphics[width=.45\textwidth]{hs1b.eps}}

1309: \hspace{.3in} % PICTURE 4

1310: \subfigure[Plot of $e_{m,n}$, $E_0=0.7087$.]{

1311: \includegraphics[width=.45\textwidth]{hs1ben.eps}}

1312: \caption{Stationary breather in a lattice with symmetric potential, 

1313: see Section \ref{hnstsy} for details.}

1314: \label{simh1}

1315: \end{figure}

1316: 

1317: 

1318: %---------------------------------------------------------------------

1319: \subsection{Breather moving along a lattice direction ($\Psi=0\degrees$)}

1320: \lbl{hpsi0}

1321: 

1322: 

1323: In \Fref{simh2}, we show a simulation of a breather moving along a

1324: lattice direction parallel to the $m$-axis, that is, $\Psi=0\degrees$.  We 

1325: have chosen $\mb{k} = \mb{k_a} = [1.4,\pi/h]^T$ so that $u=0.4445$ 

1326: and $v=0$ and thus $\Psi=\tan^{-1}(v/u) = 0\degrees$.  The breather 

1327: frequency is $\w=2.9265$, and hence the period $T=2.1470$. Following 

1328: the calculation outlined in the appendix, the amplitude and width 

1329: parameters are $\alpha=1.0339$ and $\beta=1.8440$; the remaining 

1330: parameters being $b=1$, $d=1$, $N=30$, $\eee=0.1$ and $\lambda=1$.

1331: 

1332: 

1333: The initial profile of the breather is shown in \Fref{simh2}\,(a), and at 

1334: this time, the calculated energy is $E_0=0.5537$, whilst the asymptotic 

1335: estimate \eref{henar} is $E_0=0.5522$.  It may be observed that the 

1336: breather is not radially symmetric. In fact, it is slightly elongated in the 

1337: direction parallel to the $m$-axis, that is, parallel to the direction of 

1338: motion. This is because the point corresponding to the wavevector 

1339: $\mb{k_a}$ is near the boundary of the region of ellipticity $\mcal{D}_1$ 

1340: as illustrated in \Fref{econt}. 

1341: 

1342: 

1343: \Fref{simh2}\,(b) shows the breather at the later time of  $t=21.58T=46.3513$, 

1344: at which time we find $\elo=0.5562$.  By this time, the breather has reached 

1345: the right-hand edge of the lattice and, owing to periodic boundary conditions, 

1346: it reemerges from the left-hand side as shown in \Fref{simh2}\,(c). The 

1347: breather remains localised, without spreading, though it leaves behind a 

1348: small amount of energy in its path (visible in Figures \ref{simh2}\,(b) and 

1349: \ref{simh2}\,(c)).   The computed energy changes much less than for the 

1350: previous simulation (here, $\Delta E_0/E_0=0.0287$). 

1351: 

1352: 

1353: The velocity of the breather can be measured from a plot of the energy 

1354: $e_{m,n}$, which is shown in \Fref{simh2}\,(d), viewed from directly 

1355: above the plane of the lattice.  From this plot, we note that  the 

1356: breather has travelled 42 units at an average speed of 0.42 units per 

1357: second, which is 5.5\% lower than our predicted speed of 0.4445 units 

1358: per second.  This is consistent with a small amount of energy being 

1359: shed as the system transforms from our approximated initial conditions 

1360: into the precise shape of the breather.

1361: 

1362: 

1363: \begin{figure}[t]

1364: \centering % PICTURE 1

1365: \subfigure[Profile at $t=0$, $E_0=0.5537$.]{

1366: \includegraphics[width=.45\textwidth]{hs2a.eps}}

1367: \hspace{.3in} % PICTURE 2

1368: \subfigure[Profile at $21.58T$, $E_0=0.5562$.]{

1369: \includegraphics[width=.45\textwidth]{hs2b.eps}}\\

1370: \vspace{.3in} % PICTURE 3

1371: \subfigure[Profile at $t=46.59T$, $E_0=0.5696$.]{

1372: \includegraphics[width=.45\textwidth]{hs2c.eps}}

1373: \hspace{.3in} % PICTURE 4

1374: \subfigure[Plot of $e_{m,n}$.]{

1375: \includegraphics[width=.45\textwidth]{hs2cen.eps}}

1376: \caption{Breather moving along a lattice direction, $\Psi=0\degrees$, 

1377: see Section \ref{hpsi0} for details.}

1378: \label{simh2}

1379: \end{figure}

1380: 

1381: 

1382: %--------------------------------------------------------------------

1383: \subsection{Breather moving at $\Psi=210\degrees$} \lbl{hpsi210}

1384: 

1385: 

1386: We now show that it is possible to simulate breathers moving in

1387: directions other than a lattice direction (that is, $\Psi \neq 0\degrees$).  

1388: We set $k=0.79$ and $l=1.7324$, which corresponds to $\mb{k_b}$ in

1389: \Fref{ellcontnum}.  It may be verified that $u=-0.1999$ and

1390: $v=-0.1154$ units per second, leading to $\Psi = 210\degrees$ as 

1391: required.  Although not a lattice vector, this direction is an axis of 

1392: symmetry of the lattice. For the wavevector $\mb{k_b}$, we have 

1393: $\w=2.9675$ and

1394: hence $T=2.1174$,  the remaining parameters being $b=1$, $d=1$,

1395: $N=30$, $\eee=0.1$ and $\lambda=1$.  The variational parameters

1396: $\alpha$ and $\beta$ are 1.0267 and 1.8568 respectively.  The

1397: breather is shown at times $t=25$, 50, 75 and 100 seconds in

1398: \Fref{simh6}.  Clearly, the breather does not deform significantly

1399: as it travels, nor does it radiate much energy.  The initial energy is 

1400: computed to be $E_0=0.6306$, and at $t=100$, the energy is 0.6275, 

1401: a loss of 0.5\%.   The asymptotic estimate for the energy is 

1402: $E_0=0.6184$ --- 2\% different from the numerically computed value.

1403: The motion of the breather is charted in \Tref{tsimh6}, the final 

1404: measured values for the velocities $u$ and $v$ give an average 

1405: speed of 0.2152 units per second, 6.8\% below the predicted 

1406: speed of 0.2308 units per second. The angle of travel is 

1407: almost identical to the expected value of $\Psi=210\degrees$.

1408: 

1409: 

1410: \begin{figure}[t]

1411: \centering % PICTURE 1

1412: \subfigure[Profile at $t=25$, $\elo = 0.6295$.]{

1413: \includegraphics[width=.45\textwidth]{hs6a.eps}}

1414: \hspace{.3in} % PICTURE 2

1415: \subfigure[Profile at $t=50$, $\elo = 0.6292$.]{

1416: \includegraphics[width=.45\textwidth]{hs6b.eps}}\\

1417: \vspace{.3in} % PICTURE 3

1418: \subfigure[Profile at $t=75$, $\elo = 0.6235$.]{

1419: \includegraphics[width=.45\textwidth]{hs6c.eps}}

1420: \hspace{.3in} % PICTURE 4

1421: \subfigure[Profile at $t=100$, $\elo = 0.6275$.]{

1422: \includegraphics[width=.45\textwidth]{hs6d.eps}}

1423: \caption{Breather moving at $\Psi=210\degrees$, 

1424: see Section \ref{hpsi210} for details.}

1425: \label{simh6}

1426: \end{figure}

1427: 

1428: 

1429: \setlength{\arrayrulewidth}{1pt}

1430: \begin{table}

1431: \begin{center}

1432: \begin{tabular}[htbp]{| c | c | c | c | c | c | c |} \hline 

1433: {\ssz{Time}} & {\ssz{Horizontal}} & {\ssz{Vertical}} & 

1434: {\ssz{Average horizontal}} & {\ssz{Average vertical}} & {} & {} \\[-8pt]  

1435: {\ssz{(s)}} & {\ssz{displacement}} & {\ssz{displacement}} & 

1436: {\ssz{velocity (units s$^{-1}$)}} & {\ssz{velocity (units s$^{-1}$)}} & 

1437: {\ssz{$\tan\Psi$}} & {\ssz{$\Psi$}} \\  \hline 

1438: {\ssz{$25$}} & {\ssz{$-3.5$}}  & {\ssz{$-2$}}  & {\ssz{$-0.14$}}  & {\ssz{$-0.08$}}  & {\ssz{$0.5714$}} & {\ssz{$209.74\degrees$}} \\ \hline 

1439: {\ssz{$50$}} & {\ssz{$-8$}}  & {\ssz{$-4.5$}}  & {\ssz{$-0.16$}}  & {\ssz{$-0.09$}}  & {\ssz{$0.5625$}} & {\ssz{$209.36\degrees$}} \\ \hline 

1440: {\ssz{$75$}} & {\ssz{$-13$}}  & {\ssz{$-7.5$}}  & {\ssz{$-0.1733$}}  & {\ssz{$-0.1$}}  & {\ssz{$0.5770$}} & {\ssz{$209.99\degrees$}} \\ \hline 

1441: {\ssz{$100$}} & {\ssz{$-18.5$}}  & {\ssz{$-11$}}  & {\ssz{$-0.185$}}  & {\ssz{$-0.11$}}  & {\ssz{$0.5946$}} & {\ssz{$210.74\degrees$}} \\ \hline

1442: \end{tabular}

1443: \end{center}

1444: \caption{Summary of breather motion ($\Psi=210\degrees$).}

1445: \label{tsimh6}

1446: \end{table}

1447: 

1448: 

1449: %-------------------------------------------------------------------

1450: \subsection{Breather moving at $\Psi=130\degrees$} \lbl{hpsi130}

1451: 

1452: 

1453: We have presented simulations of breathers which move along axes of 

1454: symmetry of the lattice.  In \Sref{hexdell}, from the results of our asymptotic 

1455: analysis, we found that breather solutions could be constructed for any 

1456: direction of travel.  In this section, we test the mobility of breathers along 

1457: directions which do not correspond to axes of symmetry of the lattice.

1458: We have successfully propagated breathers in a range of directions, 

1459: and here we show only one such breather, moving at an angle

1460: $\Psi=130\degrees$.  We set $k=0.8$ and $l=1.9987$, from which we 

1461: find  $u=-0.2160$, $v=0.2575$, $\Psi = 130\degrees$, $\w=2.9502$ 

1462: and the period is $T=2.1298$.  The remaining parameters are $b=1$, 

1463: $d=1$, $N=30$, $\eee=0.1$, $\lambda=1$, $\alpha=1.0297$ and 

1464: $\beta=1.8514$.  From Figure \ref{simh7}, 

1465: Even at 120 seconds, the breather remains localised without suffering

1466: appreciable degradation nor is much radiation left behind in its wake.   

1467: Initially the energy $E_0$ is computed as 0.5977, only 0.2\% different 

1468: from the asymptotic estimate of $E_0=0.5965$.   

1469: After 120 seconds, $E_0$ is 0.5897, a change of 1.3\%.  

1470: To find the velocity of the breather, we have recorded its motion

1471: and presented relevant data in \Tref{tsimh7}.  From (\ref{velsuv}), 

1472: the average speed is predicted to be 0.3254 units per second; 

1473: whilst our numerical simulation shows an average speed of 0.3361 

1474: units per second,  3.2\% lower than the expected speed.  The 

1475: direction of motion of the breather is predicted accurately.

1476: 

1477: 

1478: \begin{figure}[t]

1479: \centering % PICTURE 1

1480: \subfigure[Profile at $t=30$, $\elo = 0.5961$.]{

1481: \includegraphics[width=.45\textwidth]{hs7a.eps}}

1482: \hspace{.3in} % PICTURE 2

1483: \subfigure[Profile at $t=60$, $\elo = 0.5966$.]{

1484: \includegraphics[width=.45\textwidth]{hs7b.eps}}\\

1485: \vspace{.3in} % PICTURE 3

1486: \subfigure[Profile at $t=90$, $\elo = 0.5939$.]{

1487: \includegraphics[width=.45\textwidth]{hs7c.eps}}

1488: \hspace{.3in} % PICTURE 4

1489: \subfigure[Profile at $t=120$, $\elo = 0.5897$.]{

1490: \includegraphics[width=.45\textwidth]{hs7d.eps}}

1491: \caption{Breather moving at $\Psi=130\degrees$, see Section 

1492: \ref{hpsi130} for details.}

1493: \label{simh7}

1494: \end{figure}

1495: 

1496: 

1497: \begin{table}

1498: \begin{center}

1499: \begin{tabular}[htbp]{| c | c | c | c | c | c | c |} \hline 

1500: {\ssz{Time}} & {\ssz{Horizontal}} & {\ssz{Vertical}} & 

1501: {\ssz{Average horizontal}} & {\ssz{Average vertical}} & {} & {} \\[-8pt] 

1502: {\ssz{(s)}} & {\ssz{displacement}} & {\ssz{displacement}} & 

1503: {\ssz{velocity (units s$^{-1}$)}} & {\ssz{velocity (units s$^{-1}$)}} & 

1504: {\ssz{$\tan\Psi$}} & {\ssz{$\Psi$}} \\ \hline 

1505: {\ssz{$30$}} & {\ssz{$-6$}}  & {\ssz{$7$}}  & {\ssz{$-0.2$}}  & {\ssz{$0.2333$}}  & {\ssz{$-1.1667$}} & {\ssz{$130.60\degrees$}} \\ \hline 

1506: {\ssz{$60$}} & {\ssz{$-11$}}  & {\ssz{$13.5$}}  & {\ssz{$-0.1833$}}  & {\ssz{$0.225$}}  & {\ssz{$-1.2273$}} & {\ssz{$129.17\degrees$}} \\ \hline 

1507: {\ssz{$90$}} & {\ssz{$-17.5$}}  & {\ssz{$21.5$}}  & {\ssz{$-0.1944$}}  & {\ssz{$0.2389$}}  & {\ssz{$-1.2286$}} & {\ssz{$129.14\degrees$}} \\ \hline  

1508: {\ssz{$120$}} & {\ssz{$-25$}}  & {\ssz{$30$}}  & {\ssz{$-0.2083$}}  & {\ssz{$0.25$}}  & {\ssz{$-1.2$}} & {\ssz{$129.81\degrees$}} \\ \hline

1509: \end{tabular}

1510: \end{center}

1511: \caption{Summary of breather motion ($\Psi=130\degrees$) 

1512: see Section \ref{hpsi130} for details.}

1513: \label{tsimh7}

1514: \end{table}

1515: 

1516: 

1517: %----------------------------------------------------------------------

1518: \subsection{Stationary breather in a lattice with asymmetric potential} 

1519: \lbl{hnra}

1520: 

1521: 

1522: In Sections \ref{hnstsy}--\ref{hpsi130}, we have shown simulations

1523: of lattices with a symmetric potential, that is, with $a$ and $c$ not 

1524: necessarily zero in \eref{heqqfo}.   We now consider the more general 

1525: case for which the interaction potential is asymmetric ($a\neq0 \neq c$).  

1526: To illustrate a stationary breather, we generate initial data using 

1527: \eref{hexqasym},  with the wavevector $\mb{k} = \mb{k_1} = 

1528: [\pi/3,\pi/h]^T$.  As discussed in \Sref{hexasymm}, anomalous dispersion 

1529: corresponds to $b>10a^2/9$, hence we choose $a=1$, $b=2.5$, $c=0$ 

1530: and $d=1$.  The remaining parameter values are $N=30$, $\eee=0.1$, 

1531: $\lambda=1$, hence $\w=3$, $T=2.0668$, $\alpha=0.8665$ and $\beta=1.8670$. 

1532: 

1533: 

1534: The breather is shown in \Fref{simh5} after $10$, $30$ and $40$ full 

1535: oscillations.  Initially, we find the breather's width and energy are  

1536: $\wbr=7.46$ and $\elo=0.5429$, the latter being only 0.26\% different 

1537: from the asymptotic estimate \eref{henex} of 0.5416.   The accompanying 

1538: plots of $e_{m,n}$ demonstrate clearly that the breather preserves its 

1539: form and remains localised, even after 80 seconds.  At $t=40T$, we find 

1540: that $\wbr=6.83$, a narrowing of the breather by 8\%.  Also, at $t=40T$, 

1541: the numerically computed value of the energy $E_0=0.5371$, showing 

1542: that the energy does not fluctuate significantly.  

1543: 

1544: 

1545: \begin{figure}[htbp]

1546: \centering % PICTURE 1

1547: \subfigure[$t=10T=20.69$.]{

1548: \includegraphics[width=.45\textwidth]{hs5b.eps}}

1549: \hspace{.3in} % PICTURE 2

1550: \subfigure[Plot of $e_{m,n}$, $E_0=0.5453$.]{

1551: \includegraphics[width=.45\textwidth]{hs5ben.eps}} \\

1552: \vspace{.3in} % PICTURE 3

1553: \subfigure[$t=30T=41.34$.]{

1554: \includegraphics[width=.45\textwidth]{hs5c.eps}}

1555: \hspace{.3in} % PICTURE 4

1556: \subfigure[Plot of $e_{m,n}$, $E_0=0.5425$.]{

1557: \includegraphics[width=.45\textwidth]{hs5cen.eps}} \\

1558: \vspace{.3in} % PICTURE 5

1559: \subfigure[$t=40T=82.67$.]{

1560: \includegraphics[width=.45\textwidth]{hs5d.eps}}

1561: \hspace{.3in} % PICTURE 4

1562: \subfigure[Plot of $e_{m,n}$, $E_0=0.5371$.]{

1563: \includegraphics[width=.45\textwidth]{hs5den.eps}}

1564: \caption{Stationary breather in a lattice with an asymmetric potential, 

1565: see Section \ref{hnra} for details.}

1566: \label{simh5}

1567: \end{figure}

1568: 

1569: 

1570: %=================================================

1571: \section{Discussion} \lbl{hdisc}

1572: 

1573: 

1574: In this paper, we have found approximations to discrete breathers

1575: in a two-dimensional hexagonal FPU lattice with a scalar-valued 

1576: function at each node.  We have shown that

1577: the lattice equations can be reduced to a cubic NLS equation for

1578: two special cases.  In \Sref{hexsym}, we obtained moving

1579: breathers when the interaction potential is symmetric; in

1580: this case an ellipticity criterion for the wavevector is found.

1581: In the appendix we summarise our technique for approximating 

1582: soliton solutions of the two-dimensional NLS equation. 

1583: 

1584: 

1585: In \Sref{hexasymm}, we considered lattices with an asymmetric 

1586: potential, in which case a reduction to NLS could only be performed

1587: for stationary breathers.   

1588: The theoretical methods employed here are similar to those used on 

1589: the square lattice of \citep{buts06}; however,  several important 

1590: differences emerge in the course of the analysis.  

1591: We found that the anomalous dispersion regime corresponds to 

1592: $b>4a^2/3$ in the square lattice and $b>10a^2/9$ in the hexagonal.  

1593: Furthermore we find that stationary breathers involve the 

1594: generation of second harmonics in the hexagonal lattice 

1595: ($G_2=aF^2/3$), whereas they are suppressed in the square 

1596: lattice ($G_2=0$).  In both lattices, the quadratic nonlinearity 

1597: generates a small amplitude slowly-varying mode ($G_0=-a|F|^2$). 

1598: 

1599: 

1600: For symmetric interactions we found an associated ellipticity constraint: 

1601: moving breathers only occur for certain wavevectors, this means that 

1602: all breathers have (i) a relatively high frequency, and (ii) a maximum 

1603: speed, which depends on the wave vector and hence on the 

1604: direction of travel, (iii) a threshold energy which also depends on 

1605: wavevector and hence is related to speed and direction of travel.  

1606: We find that the threshold energy is lower for breathers at the 

1607: edge of the domain of ellipticity in wavevector space. These breathers 

1608: have lower frequencies and faster speeds.  The fastest-moving 

1609: breathers are restricted to moving along lattice directions. 

1610: 

1611: 

1612: We also presented asymptotic estimates for the breather energy in

1613: Sections \ref{hexsym} and \ref{hexasymm}.  As expected, we found a

1614: minimum energy required to create breathers in the hexagonal

1615: lattice.  The threshold energy for moving breathers is smaller

1616: than that required for stationary breathers and becomes vanishingly

1617: small at the boundary of the domain of ellipticity.

1618: 

1619: 

1620: In \Sref{hex5}, we extended the small amplitude expansion to fifth-order 

1621: and derived a higher-order equation which more correctly describes 

1622: the shape and stability properties of the breather envelope.  We 

1623: obtained a generalised NLS equation \eref{hexgennls} with a variety 

1624: of perturbation terms, some of which are known to be stabilising.  This 

1625: equation is slightly simpler than the corresponding equation obtained 

1626: for the square lattice, namely (3.11) of \citep{buts06}.  In particular, 

1627: the higher-order dispersive terms in \eref{hexgennls} are isotropic, 

1628: reflecting the hexagonal rotational symmetry of the lattice.  For stationary 

1629: breathers in the case of a symmetric potential we find that the cubic 

1630: nonlinearity does not give rise to third harmonics, that is, $H_3=0$ in 

1631: the hexagonal lattice, in contrast to $H_3=bF^3/8$ in the square lattice.

1632: 

1633: 

1634: In \Sref{hexnumerics}, we illustrated  these breather modes, showing 

1635: that both stationary and moving breathers are long-lived.  The 

1636: breather profiles change little over time, a small amount of 

1637: energy is shed due to the initial conditions being only approximate.  

1638: We have successfully propagated long-lived breathers moving in 

1639: directions which are not axes of symmetry of the lattice (for 

1640: instance, $\Psi=130\degrees$),  suggesting that there is no absolute 

1641: restriction upon the direction of travel; this is in contrast to the 

1642: observations reported by Marin \etal \citep{mar00, mar98} for 

1643: mechanical (two-component) two-dimensional lattices, who could 

1644: only find breathers which travelled along axes of symmetry of the lattice. 

1645: 

1646: 

1647: %--------------------------------

1648: \ack

1649: 

1650: IAB would like to thank both the UK Engineering and Physical Sciences 

1651: Research Council for financial assistance, and also Qamran Yaqoob for his 

1652: assistance in the preparation of several diagrams that appear in this paper. 

1653: 

1654: 

1655: %------------------------------------------------------------

1656: \appendix 

1657: \renewcommand{\theequation}{\Alph{section}\arabic{equation}}

1658: \section{Approximation to Townes soliton} \lbl{app}

1659: 

1660: 

1661: Since analytic formulae for Townes solitons are unavailable, we use the 

1662: Rayleigh-Ritz method to find time-harmonic radially symmetric solutions of

1663: \begin{equation}

1664: \ii F_T + D \nabla^2 F + B |F|^2 F= 0 ,  

1665: \lbl{app-pde} 

1666: \end{equation}

1667: of the form $F({\bf x},T)=e^{\ii\lambda T}\phi(r)$ where 

1668: $r=|{\bf x}|=\sqrt{\xi^2 + \eta^2}$.   The function $\phi$ satisfies 

1669: \begin{equation} \lbl{nlsrl}

1670: -\lambda \phi + D\nabla^2 \phi + B\phi^3=0 ,

1671: \end{equation}

1672: where $\nabla^2=\partial_\xi^2 + \partial_\eta^2$.

1673: Equation (\ref{nlsrl}) arises from a variational derivative of  

1674: \begin{equation} \lbl{hvar}

1675: \mcal{E}(\phi) = \int \int \shalf \lambda |\phi|^2 +

1676: \shalf D |\nabla \phi|^2 - \sfr{1}{4} B |\phi|^4 d^2{\bf r}.

1677: \end{equation}

1678: Using a trial solution of the form $\phi=\alpha\,\sech(\beta r)$, 

1679: where $\alpha$ and $\beta$ are undetermined parameters, we find 

1680: \begin{equation} \lbl{hab}

1681: \mcal{E}(\alpha,\beta) = \frac{D(1+2\str{ln}2)}{12} \alpha^2 -

1682: \frac{B(4\str{ln}2-1)}{24}\frac{\alpha^4}{\beta^2} + \frac{\lambda

1683: \str{ln}2}{2}\frac{\alpha^2}{\beta^2} ; 

1684: \end{equation}

1685: $\alpha$ and $\beta$ are determined by seeking stationary 

1686: points of the action $\mcal{E}$, namely  ${\partial \mcal{E}}/

1687: { \partial \alpha} = {\partial \mcal{E}}/{ \partial \beta} = 0$.

1688: Hence we find 

1689: \begin{equation} \lbl{alpbet}

1690: \alpha = \sqrt{\frac{12\lambda\str{ln}2}{B(4\str{ln}2 -1)}} , \qquad 

1691: \beta = \sqrt{\frac{6\lambda\str{ln}2}{D(2\str{ln}2 + 1)}} , 

1692: \end{equation}

1693: and so our approximation to the Townes soliton solution of (\ref{app-pde}) is

1694: \begin{equation} \lbl{vartown}

1695: F = \sqrt{\frac{12\lambda\log2}{B(4\log2 -1)}} \ \exp (\ii \lambda T) \

1696: \sech \left( \sqrt{\frac{6\lambda\log 2}{D(2\log2 + 1)}}

1697: \sqrt{\xi^2 + \eta^2}\right) .

1698: \end{equation}

1699: 

1700: 

1701: \footnotesize

1702: %--------------------------------------------------------------

1703: \begin{thebibliography}{99}

1704: 

1705: \renewcommand{\itemsep}{0pt}

1706: \renewcommand{\parsep}{0pt}

1707: 

1708: \bibitem{thesis} IA~Butt. PhD Thesis. University of Nottingham, 

1709: to be published at {\tt http://etheses.nottingham.ac.uk/} (2006). 

1710: 

1711: \bibitem{buts06} IA~Butt and JAD Wattis.  Discrete breathers 

1712: in a two-dimensional {F}ermi-{P}asta-{U}lam lattice.

1713: \emph{J. Phys. A: Math. Gen.}, {\bf 39}, 4955, (2006).

1714: 

1715: \bibitem{but1d} IA~Butt and JAD Wattis.   Asymptotic analysis of 

1716: combined breather-kink modes in a Fermi-Pasta-Ulam chain. 

1717: \emph{submitted}, (2006). 

1718: 

1719: \bibitem{dav03}  TA~Davydova, AI~Yakimenko, and YA~Zaliznyak.

1720: Two-dimensional solitons and vortices in normal and anomalous 

1721: dispersive media.  \emph{Phys. Rev. E}, {\bf 67}, 026402, (2003).

1722: 

1723: \bibitem{fer40} E~Fermi, J~Pasta, and S~Ulam.  Studies of nonlinear 

1724: problems.   Los Alamos Scientific Report, 1940, originally unpublished. 

1725: Later published in {\em Lectures in Applied Mathematics}, {\bf 15}, 143, (1974).

1726: 

1727: \bibitem{fib98} G~Fibich and G~Papanicolaou.   A modulation 

1728: method for self-focusing in the perturbed critical nonlinear 

1729: {S}chr{\"o}dinger equation. \emph{Phys. Lett. A}, {\bf 239}, 167, (1998).

1730: 

1731: \bibitem{fib99}  G~Fibich and G~Papanicolaou.   Self-focusing in the 

1732: perturbed and unperturbed nonlinear {S}chr{\"o}dinger equation in 

1733: critical dimension.   \emph{SIAM J. Appl. Math.}, {\bf 60}, 183, (1999).

1734: 

1735: \bibitem{fla97}  S~Flach, K~Kladko, and RS~MacKay.  Energy thresholds 

1736: for discrete breathers in one-, two-, and  three-dimensional 

1737: lattices.  \emph{Phys. Rev. Lett.}, {\bf 78}, 1207, (1997).

1738: 

1739: \bibitem{flat94} S~Flach, K~Kladko, and CR~Willis.  Localized excitations 

1740: in two-dimensional lattices. \emph{Phys. Rev. E}, {\bf 50}, 2293, (1994).

1741: 

1742: \bibitem{gia04} J~Giannoulis and A~Mielke.   The nonlinear 

1743: {S}chr{\"o}dinger equation as a macroscopic limit for an oscillator chain 

1744: with cubic nonlinearities.  \emph{Nonlinearity}, {\bf 17}, 551, (2004).

1745: 

1746: \bibitem{gia06} J~Giannoulis and A~Mielke.  Dispersive evolution 

1747: of pulses in oscillator chains with general interaction potentials.

1748: \emph{Disc. Cont. Dyn. Sys. B}, {\bf 6}, 493, (2006).

1749: 

1750: \bibitem{kas04}  M~Kastner.  Energy thresholds for discrete 

1751: breathers.  \emph{Phys. Rev. Lett.}, {\bf 92}, 104301, (2004).

1752: 

1753: \bibitem{kuz} EA Kuznetsov, AM Rubenchik and VE Zakharov. Soliton 

1754: stability in plasmas and hydrodynamics. \emph{Phys Rep}, {\bf 142}, 103, (1986).

1755: 

1756: \bibitem{mac94} RS~MacKay and S~Aubry.  Proof of existence of 

1757: breathers for time-reversible or {H}amiltonian networks of weakly 

1758: coupled oscillators.  \emph{Nonlinearity}, {\bf 7}, 1623, (1984).

1759: 

1760: \bibitem{mar98} JL~Marin, JC~Eilbeck, and FM~Russell.  

1761: Localised moving breathers in a 2{D} hexagonal lattice. 

1762: \emph{Phys. Lett. A}, {\bf 248}, 225, (1998).

1763: 

1764: \bibitem{mar00} JL~Marin, JC~Eilbeck, and FM~Russell.

1765: \emph{2D breathers and applications (Nonlinear Science 

1766: at the Dawn of the 21st Century)}.  Eds: PL Christiansen 

1767: and MP Soerensen.   Springer, Berlin, (2000).

1768: 

1769: \bibitem{mar01}  JL~Marin, FM~Russell, , and JC~Eilbeck.  Breathers 

1770: in cuprate-like lattices.  \emph{Phys. Lett. A}, {\bf 281}, 21, (2001).

1771: 

1772: \bibitem{rem85} M~Remoissenet.  Low-amplitude breather 

1773: and envelope solitons in quasi-one-dimensional physical models.

1774: \emph{Phys. Rev. B}, {\bf 33}, 2386, (1985).

1775: 

1776: \bibitem{wei99} MI~Weinstein.  Excitation thresholds for nonlinear 

1777: localized modes on lattices. \emph{Nonlinearity}, {\bf 12}, 673, (1999).

1778: 

1779: \end{thebibliography}

1780: 

1781: 

1782: \end{document}

1783: