1: \documentclass[10pt]{article}

2: \usepackage{upgreek}

3: \usepackage{latexsym,epsfig,bm,amssymb}

4: \usepackage{color}

5: \usepackage{amsthm,mathrsfs}

6: %\usepackage{mathptmx}

7: \usepackage{times}

8: %\usepackage{showkeys}

9: 

10: \def\version{November 10, 2007}

11: 

12: %\pagestyle{myheadings}

13: %\markboth{}{{\rm Alexander Komech, Andrew Komech:}

14: %\ {\it\small Klein-Gordon field coupled to several nonlinear oscillators}

15: %\quad

16: %{\tt ukk.tex}}

17: 

18: \hyphenation{Shni-rel-man}

19: \hyphenation{quasi-mea-sure}

20: \hyphenation{Cuc-cagna}

21: \hyphenation{non-zero}

22: 

23: \textwidth 17cm

24: %\textheight 22cm

25: \textheight 23cm

26: \topmargin -1.4cm

27: %\topmargin 0mm

28: \oddsidemargin -3mm

29: \evensidemargin -3mm

30: 

31: \renewcommand{\arraystretch}{1.3}

32: 

33: \DeclareSymbolFont{AMSb}{U}{msb}{m}{n}

34: \DeclareSymbolFontAlphabet{\mathbb}{AMSb}

35: 

36: 

37: %%%Andrei !

38: %%eti moi makrosi pojalusita ne stirai! **********

39: 

40: \newcommand{\ov}{\overline}

41: \newcommand{\ci}{\cite}

42: 

43: \newcommand{\5}{\hspace{0.5mm}}

44: 

45: \newcommand{\beqn}{\begin{eqnarray}}

46: \newcommand{\eeqn}{\end{eqnarray}}

47: \newcommand{\be}{\begin{equation}}

48: \newcommand{\ee}{\end{equation}}

49: \newcommand{\ba}{\begin{array}}

50: \newcommand{\ea}{\end{array}}

51: \newcommand{\la}{\label}

52: 

53: \newcommand{\pa}{\partial}

54: \newcommand{\fr}{\frac}

55: 

56: \newcommand{\ds}{\displaystyle}

57: \newcommand{\E}{\mathscr{E}}

58: \newcommand{\I}{{\cal I}}

59: \newcommand{\bS}{{\bf S}}

60: \newcommand{\bGa}{\mbox{\boldmath $\Ga$}}

61: \newcommand{\bF}{{\bf F}}

62: \newcommand{\bA}{{\bf A}}

63: 

64: \newcommand{\de}{\delta}

65: \newcommand{\De}{\Delta}

66: 

67: \newcommand{\ve}{\varepsilon}

68: 

69: \newcommand{\Ga}{\Gamma}

70: \newcommand{\ga}{\gamma}

71: \newcommand{\gam}{\gamma}

72: \newcommand{\ka}{\kappa}

73: \newcommand{\lam}{\lambda}

74: \newcommand{\vp}{\varphi}

75: 

76: \newcommand{\om}{\omega}

77: \newcommand{\Om}{\Omega}

78: \newcommand{\Si}{\Sigma}

79: 

80: 

81: %\newcommand{\ft}[1]{\footnote[\thefootnote]{#1}

82: %{\bf\footnotemark[\value{footnote}]}

83: %$\square$

84: %\addtocounter{footnote}{1}

85: %}

86: 

87: \newcommand{\ft}[1]{

88: \addtocounter{footnote}{1}

89: {\bf\mbox{$\blacklozenge$}\footnote[\thefootnote]{{\bf AK:} #1}[AK: #1]}}

90: 

91: \newcommand{\bd}{\begin{definition}}

92:  \newcommand{\ed}{\end{definition}}

93: \newcommand{\bt}{\begin{theorem}}

94: \newcommand{\et}{\end{theorem}}

95: \newcommand{\bqt}{\begin{qtheorem}}

96: \newcommand{\eqt}{\end{qtheorem}}

97: 

98: 

99: \newcommand{\bp}{\begin{pro}}

100: \newcommand{\ep}{\end{pro}}

101: 

102: \newcommand{\bl}{\begin{lemma}}

103: \newcommand{\el}{\end{lemma}}

104: \newcommand{\bc}{\begin{cor}}

105: \newcommand{\ec}{\end{cor}}

106: 

107: \newcommand{\bex}{\begin{example}}

108: \newcommand{\eex}{\end{example}}

109: \newcommand{\bexs}{\begin{examples}}

110: \newcommand{\eexs}{\end{examples}}

111: 

112: 

113: \newcommand{\bexe}{\begin{exercice}}

114: \newcommand{\eexe}{\end{exercice}}

115: 

116: 

117: \newcommand{\br}{\begin{remark} }

118: \newcommand{\er}{\end{remark}}

119: 

120: 

121: 

122: %\newcommand{\toEF}{\stackrel{{\E}\sb{F}}\lr}

123: \newcommand{\lr}{\longrightarrow}

124: %\newcommand{\st}{\stackrel}

125: 

126: %%%%*****************************

127: 

128: \def\ot{\leftarrow}

129: 

130: \newcommand{\uupsilon}{{w}}

131: %\newcommand{\E}{{\cal E}}

132: \newcommand{\bmGamma}{\bm\Gamma}

133: %\newcommand{\bS}{{\bf S}}

134: %\newcommand{\bF}{{\bf F}}

135: %\newcommand{\bA}{{\bf A}}

136: %\newcommand{\toEF}{\stackrel{{\E}\sb{F}}\longrightarrow}

137: 

138: \def\xx{{m}}

139: \def\yy{{n}}

140: 

141: 

142: \newcommand\dist{\mathop{\rm dist}}

143: \newcommand\supp{\mathop{\rm supp}}

144: \newcommand\p{\partial}

145: \newcommand{\at}[1]{\vert\sb{\sb{#1}}}

146: \newcommand{\Spec}{\mathop{\rm Spec}}

147: \newcommand{\Bar}[1]{\mkern2mu\overline{\mkern-2mu#1\mkern-5mu}\mkern5mu}

148: \def\Re{{\rm Re\, }}

149: \def\Im{{\rm Im\,}}

150: \providecommand\C{{\mathbb C}}

151: \renewcommand\C{{\mathbb C}}

152: \newcommand{\R}{{\mathbb R}}

153: \newcommand{\N}{{\mathbb N}}

154: \newcommand\Z{{\mathbb Z}}

155: \newcommand{\Abs}[1]{\left\vert#1\right\vert}

156: \newcommand{\abs}[1]{\vert #1\vert}

157: \newcommand{\Norm}[1]{\left\Vert #1\right\Vert}

158: \newcommand{\norm}[1]{\Vert #1\Vert}

159: \newcommand{\const}{{\rm const}}

160: \newcommand\cnst{\mathop{C}}

161: \newcommand\sothat{{\rm :}\ }

162: \newcommand\sgn{\mathop{\rm sgn}}

163: \newcommand\Span{\mathop{\rm Span}}

164: \newcommand\mod{\mathop{\rm mod}}

165: 

166: \renewcommand{\labelenumi}{{\rm (}{\it\roman{enumi}}{\rm )}}

167: \providecommand{\ltor}[1]{

168: \ifnum #1=1{\it i}\else\ifnum #1=2{\it ii}\else\ifnum #1=3{\it iii}

169: \else\ifnum #1=4 {\it iv}\fi\fi\fi\fi

170: }

171: 

172: 

173: %\DeclareMathSymbol{\varGamma}{\mathord}{letters}{"00}

174: \DeclareMathSymbol{\varDelta}{\mathord}{letters}{"01}

175: \DeclareMathSymbol{\varTheta}{\mathord}{letters}{"02}

176: \DeclareMathSymbol{\varLambda}{\mathord}{letters}{"03}

177: %\DeclareMathSymbol{\varXi}{\mathord}{letters}{"04}

178: %\DeclareMathSymbol{\varPi}{\mathord}{letters}{"05}

179: %\DeclareMathSymbol{\varSigma}{\mathord}{letters}{"06}

180: %\DeclareMathSymbol{\varUpsilon}{\mathord}{letters}{"07}

181: \DeclareMathSymbol{\varPhi}{\mathord}{letters}{"08}

182: \DeclareMathSymbol{\varPsi}{\mathord}{letters}{"09}

183: \DeclareMathSymbol{\varOmega}{\mathord}{letters}{"0A}

184: 

185: 

186: % Italic ``theorems''

187: \theoremstyle{plain}

188: \newtheorem{theorem}{Theorem}[section]

189: \newtheorem{hypothesis}[theorem]{Hypothesis}

190: \newtheorem{lemma}[theorem]{Lemma}

191: \newtheorem{corollary}[theorem]{Corollary}

192: \newtheorem{proposition}[theorem]{Proposition}

193: \newtheorem{claim}{Claim}[section]

194: % Roman ``theorems''

195: \theoremstyle{definition}

196: \newtheorem{definition}[theorem]{Definition}

197: \newtheorem{assumption}{Assumption}%{assumption}

198: 

199: % Humble things: remarks and examples.

200: \theoremstyle{remark}

201: \newtheorem{remark}[theorem]{Remark}

202: \newtheorem{example}[theorem]{Example}

203: 

204: \makeatletter\@addtoreset{equation}{section}

205: \makeatletter\@addtoreset{theorem}{section}

206: %\makeatletter\@addtoreset{proposition}{section}

207: %\makeatletter\@addtoreset{lemma}{section}

208: %\makeatletter\@addtoreset{corollary}{section}

209: %\makeatletter\@addtoreset{remark}{section}

210: %\makeatletter\@addtoreset{assumption}{section}

211: %\makeatletter\@addtoreset{definition}{section}

212: \makeatother

213: 

214: \renewcommand{\theequation}{\thesection.\arabic{equation}}

215: \renewcommand{\thetheorem}{\thesection.\arabic{theorem}}

216: \renewcommand{\theproposition}{\thesection.\arabic{proposition}}

217: \renewcommand{\thelemma}{\thesection.\arabic{lemma}}

218: \renewcommand{\thecorollary}{\thesection.\arabic{corollary}}

219: \renewcommand{\theremark}{\thesection.\arabic{remark}}

220: %\renewcommand{\theassumption}{\Alph{assumption}}

221: \renewcommand{\theassumption}{\thesection.\arabic{assumption}}

222: \renewcommand{\thedefinition}{\thesection.\arabic{definition}}

223: 

224: 

225: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

226: 

227: \begin{document}

228: 

229: %MPI

230: %\title{

231: %On Global Attraction to Quantum Stationary States II.

232: %\\

233: %Several Nonlinear Oscillators Coupled to Massive Scalar Field

234: %}

235: 

236: %SIMA

237: \title{

238: On global attraction to solitary waves for the

239: Klein-Gordon field coupled to several nonlinear oscillators

240: }

241: %%short title:

242: %%Klein-Gordon field coupled to nonlinear oscillators

243: 

244: %%keywords:

245: %% Klein-Gordon equation, solitary waves, U(1)-invariance,

246: %% global attractor, solitary manifold,

247: %% Titchmarsh convolution theorem, nonlinear spectral analysis

248: %% spectral representation

249: 

250: \author{

251: {\sc Alexander Komech}

252: \footnote{On leave from 

253: Institute of the Information Transmission Problems, RAS.

254: Supported in part by Alexander von Humboldt Research Award,

255: by Max-Planck Institute for Mathematics in the Sciences (Leipzig),

256: and by grants FWF P19138-N13,

257: DFG 436 RUS 113/929/0-1, and RFBR 07-01-00018a.}

258: \\

259: {\it\small Faculty of Mathematics, University of Vienna, Wien A-1090, Austria}

260: \\

261: %{\rm\small and}

262: %\\

263: {\it\small

264: Institute for Information Transmission Problems,

265: Moscow 101447, Russia}

266: \\ \\

267: {\sc Andrew Komech}

268: \footnote{Supported in part

269: by Max-Planck Institute for Mathematics in the Sciences (Leipzig) and

270: by the National Science Foundation

271: under Grant DMS-0600863.

272: }

273: \\

274: {\it\small

275: Mathematics Department, Texas A\&M University,

276: College Station, TX 77843, USA}

277: \\

278: %{\rm\small and}

279: %\\

280: {\it\small

281: Institute for Information Transmission Problems,

282: Moscow 101447, Russia}

283: }

284: \date{\version}

285: 

286: \maketitle

287: %

288: \begin{abstract}

289: The global attraction is established for all finite energy solutions

290: to a model $\mathbf{U}(1)$-invariant nonlinear Klein-Gordon equation

291: in one dimension coupled to a finite number of nonlinear oscillators:

292: We prove that {\it each finite energy solution}

293: converges as $t\to\pm\infty$ to the set of all ``nonlinear eigenfunctions''

294: of the form $\phi(x)e\sp{-i\omega t}$

295: if  all oscillators are strictly nonlinear, and

296: the distances between neighboring oscillators are sufficiently small.

297: 

298: %This nonlinear energy transfer is justified

299: %by the Titchmarsh convolution theorem.

300: %This result for one oscillator was obtained in \cite{ubk-arma}.

301: 

302: Our approach is based on the analysis of {\it omega-limit trajectories}

303: which form the global attractor.

304: We show that their time spectrum is a priori compact.

305: Then

306: the nonlinear spectral analysis based on the Titchmarsh convolution theorem

307: allows to reduce the time-spectrum to one point.

308: This implies that each omega-limit trajectory is a solitary wave.

309: %shows that the spectrum of such limiting trajectories

310: %consists of at most one frequency.

311: Physically, the {\it global attraction} is caused by the nonlinear

312: energy transfer from lower harmonics to the continuous spectrum

313: and subsequent dispersive radiation.

314: The Titchmarsh theorem allows to prove that this energy transfer

315: and radiation are absent only for the solitary waves.

316: 

317: To check the sharpness of our conditions,

318: we construct counterexamples showing

319: the global attractor can contain

320: ``multifrequency solitary waves''

321: if the distance between oscillators is large 

322: % some of the neighboring oscillators is sufficiently large

323: or if some of them are linear.

324: %or linear combinations of distinct solitary waves.

325: \end{abstract}

326: 

327: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

328: 

329: \section{Introduction}

330: The long time asymptotics for

331: nonlinear

332: wave equations

333: have been the subject of intensive research,

334: starting with the pioneering papers by

335: Segal \cite{MR0153967,MR0152908},

336: Strauss \cite{MR0233062},

337: and Morawetz and Strauss \cite{MR0303097},

338: where the

339: nonlinear scattering and the local attraction to zero solution

340: were proved.

341: Local attraction to solitary waves,

342: or \emph{asymptotic stability},

343: in

344: $\mathbf{U}(1)$-invariant dispersive systems

345: was addressed in

346: \cite{MR1071238,MR1199635e,MR1170476,MR1334139}

347: and then developed in

348: \cite{MR1488355,MR1681113,MR1893394,MR1835384,MR1972870,MR2027616}.

349: Global attraction

350: to \emph{static},

351: stationary solutions in the dispersive systems

352: \emph{without $\mathbf{U}(1)$ symmetry}

353: was established in

354: \cite{MR1203302e,MR1359949,MR1412428,MR1434147,MR1726676,MR1748357}.

355: 

356: We would like to have

357: the dynamical description

358: of the Bohr transitions to quantum stationary states

359: in coupled nonlinear systems

360: of Quantum Physics.

361: This suggests investigation of the global attractors in

362: nonlinear Hamiltonian hyperbolic equations

363: with $\mathbf{U}(1)$-symmetry

364: (see \cite{ubk-arma} for the discussion).

365: The first result about

366: the global attraction to solitary waves in

367: a model with these properties

368: %$\mathbf{U}(1)$-invariant dispersive systems

369: was obtained in \cite{ubk-cr,ubk-arma},

370: where we considered the Klein-Gordon equation coupled to

371: one nonlinear oscillator.

372: 

373: We are aware of only one other recent advance \cite{MR2304091}

374: in the field

375: of nonzero global attractors for Hamiltonian PDEs.

376: In that paper, the global attraction for the nonlinear Schr\"odinger equation

377: in dimensions $n\ge 5$ was considered.

378: The dispersive wave was explicitly specified

379: using the rapid decay of local energy in higher dimensions.

380: The global attractor was proved to be compact, but it was

381: neither identified with the set of solitary waves nor was proved

382: to be of finite dimension \cite[Remark 1.18]{MR2304091}.

383: 

384: In the present paper, we prove the attraction

385: to the set of solitary waves

386: for all finite energy solutions to the Klein-Gordon equation

387: coupled to any finite number of nonlinear oscillators.

388: For the proof, we develop an approach of the spectral inflation

389: \cite{ubk-arma}

390: justified by the Titchmarsh Convolution Theorem.

391: This justification requires new arguments and appropriate conditions.

392: We demonstrate the sharpness of these conditions constructing

393: counterexamples.

394: 

395: Our model is based on the complex Klein-Gordon field $\psi(x,t)$, 

396: interacting with $N$ nonlinear oscillators

397: located at the points

398: $X\sb{1}<X\sb{2}<\dots<X\sb{N}$:

399: \begin{equation}\label{kg-no-0}

400: \ddot\psi

401: =\psi''-m^2\psi+\sum\sb{J}\delta(x-X\sb{J})F\sb{J}(\psi(X\sb{J},t)),

402: \qquad x\in\R,

403: \end{equation}

404: where

405: $m>0$

406: and

407: $F\sb{J}$ are nonlinear functions

408: describing nonlinear oscillators at the points $X\sb{J}$.

409: The dots stand for the derivatives in $t$,

410: and the primes for the derivatives in $x$.

411: All derivatives and the equation are understood in

412: the sense of distributions.

413: We assume that equation (\ref{kg-no-0}) is $\mathbf{U}(1)$-invariant;

414: that is,

415: \begin{equation}\label{inv-f}

416: F\sb{J}(e\sp{i\theta}\psi)=e\sp{i\theta} F\sb{J}(\psi),

417: \qquad\theta\in\R,\quad\psi\in\C,\quad 1\le J\le N.

418: \end{equation}

419: This symmetry leads to the charge conservation

420: and to the existence of the solitary wave solutions,

421: which are finite energy solutions of the following form:

422: \begin{equation}\label{soliton}

423: \psi\sb\omega(x,t)=\phi\sb\omega(x)e\sp{-i\omega t},\quad\omega\in\R,

424: \quad\phi\sb\omega\in H\sp{1}(\R).

425: \end{equation}

426: Above, $H\sp{1}(\R)$ denotes the Sobolev space.

427: 

428: \begin{definition}\label{dSS}

429: $S$

430: is the set of all functions

431: $\phi\sb\omega(x)\in H\sp{1}(\R)$ with $\omega\in\R$,

432: so that $\phi\sb\omega(x)e\sp{-i\omega t}$

433: is a solution to (\ref{kg-no-0}).

434: \end{definition}

435: Note that $S$ also contains the zero solution.

436: 

437: Generically,

438: the factor-space $S/\mathbf{U}(1)$

439: is isomorphic to a finite union of one-dimensional

440: intervals.

441: The set of all solitary waves for equation (\ref{kg-no-0})

442: is described in Proposition~\ref{prop-solitons}.

443: %Let us mention

444: %that there are numerous results on the existence of

445: %solitary wave solutions

446: %of the form $\phi(x)e\sp{-i\omega t}$

447: %to nonlinear Hamiltonian systems with $\mathbf{U}(1)$ symmetry

448: %\cite{MR0454365,MR695535,MR695536,MR765961,MR847126,MR1344729}.

449: Typically, such solutions exist for

450: $\omega$ from an interval or a collection of intervals

451: of the real line.

452: 

453: \bigskip

454: 

455: Our main result is the following long-time asymptotics:

456: In the case when

457: all oscillators are polynomial and strictly nonlinear

458: (see Assumptions~\ref{ass-a} and \ref{ass-nonl} below)

459: %is polynomial of order strictly greater than $1$

460: and all distances $\abs{X\sb{J+1}-X\sb{J}}$ are sufficiently small,

461: we prove

462: that any finite energy solution converges

463: to the set $S$ of all solitary waves:

464: \begin{equation}\label{attraction}

465: \psi(\cdot,t)\longrightarrow S,

466: \qquad t\to\pm\infty,

467: \end{equation}

468: where the convergence holds in local energy seminorms.

469: 

470: Let us give a brief sketch of our approach.

471: We introduce a concept of the omega-limit trajectories

472: $\beta(x,t)$ which 

473: play a crucial role in the proof.

474: We define omega-limit trajectories as the limits

475: \[

476: \psi(x,t+s\sb j)\rightarrow\beta(x,t),

477: \qquad

478: (x,t)\in\R^2,

479: \]

480: for some sequence of times

481: $s\sb j\to+\infty$.

482: We will prove that all omega-limit trajectories

483: are solitary waves, thus finishing the proof.

484: To complete this program,

485: we study the time spectrum of solutions,

486: that is, their Fourier-Laplace transform in time.

487: We need to prove that

488: $\beta(x,t)=\phi\sb\omega(x)e^{-i\omega t}$, that is,

489: that the time spectrum of $\beta$ consists of at most one frequency.

490: First, we show that the spectrum of the solution

491: at $x=X\sb{1}$ and $x=X\sb{N}$

492: is absolutely continuous for $\abs{\omega}>m$.

493: At the points $x\in(X\sb 1,X\sb N)$,

494: the nonlinearity may extend the singular part of the spectrum

495: to be at most $[-\Lambda,\Lambda]$,

496: for some bounded $\Lambda$.

497: Outside of this interval, the spectrum is absolutely continuous.

498: This allows to prove that the spectrum

499: of any omega-limit trajectory at $x=X\sb{1}$ and $x=X\sb{N}$

500: is contained in $[-m,m]$,

501: while at the points $x\in(X\sb 1,X\sb N)$

502: the spectrum is contained in $[-\Lambda,\Lambda]$.

503: %Assumption~\ref{ass-delta-small}

504: %guarantees that this spectrum

505: %does not contain frequencies which correspond to the

506: %``trapped modes'' that vanish at the endpoints

507: %of the intervals $[X\sb J,X\sb{J+1}]$.

508: The next important observation is

509: that each omega-limit trajectory

510: is also a solution to the original nonlinear Klein-Gordon equation.

511: This allows to apply the Titchmarsh theorem and prove

512: that the spectrum of any omega-limit trajectory

513: at all points $x\in\R$ consists of at most one frequency.

514: At this last step,

515: one needs

516: the assumptions that the oscillators are strictly nonlinear

517: and located sufficiently close to one another.

518: 

519: The requirement that the nonlinearities $F\sb{J}$ are polynomial

520: allows us to apply the Titchmarsh theorem

521: which is vital in the proof.

522: We construct counterexamples showing the sharpness of our assumptions

523: for the global attraction to the solitary waves. 

524: Namely, for $N=2$,

525: we construct multifrequency solitary waves

526: in the case when the distance $\abs{X\sb 2-X\sb 1}$ is sufficiently large

527: or one of the oscillators is linear.

528: %For $N=1$, a counterexample

529: %given by a superposition of two different solitary waves

530: %is constructed in \cite{ubk-arma}.

531: 

532: Let us mention that

533: in the case of $N$ oscillators,

534: considered in this paper,

535: the general plan of the proof is similar

536: to the case of one oscillator (see \cite{ubk-cr,ubk-arma}).

537: %Separation of dispersive components, absolute continuity of time spectrum 

538: %outside a bounded interval, compactness of time shifts

539: %of the bound component,

540: %and a nonlinear spectral analysis of omega-limit trajectories 

541: %by the Titchmarsh Convolution Theorem.

542: However, the justifications of all steps 

543: are based on new arguments.

544: In particular, the application of the Titchmarsh theorem

545: required a new construction.

546: 

547: \bigskip

548: 

549: Our paper is organized as follows.

550: In Section~\ref{sect-results}, we formulate our main results.

551: In Section~\ref{sect-splitting}, we separate the first dispersive component. 

552: In Sections~\ref{sect-spec} and ~\ref{sect-ac}, 

553: we construct spectral representation 

554: for the remaining component, 

555: and prove absolute continuity of its spectrum for high frequencies.

556: In Sections~\ref{sect-bound}, we  

557: separate the second dispersive component corresponding to the 

558: high frequencies

559: and establish compactness for the remaining \emph{bound} component

560: with the bounded spectrum.

561: In Section~\ref{sect-nonlinear-analysis},

562: we study omega-limit trajectories of the solution.

563: %({\it i}) first, we prove that any omega-limit trajectory also

564: %is a solution to  the nonlinear Klein-Gordon equation, 

565: %({\it ii}) second, we  reduce the spectrum 

566: %of the trajectory to a bounded set,

567: %({\it iii}) finally, we reduce the spectrum to a single point

568: %using the Titchmarsh Convolution Theorem. 

569: %This means that any  omega-limit trajectory is a solitary wave,

570: %and proves the global attraction to the set of all solitary waves.

571: In Section~\ref{sect-counterexamples} we collect counterexamples,

572: and in Appendix~\ref{sect-existence}

573: we establish global well-posedness.

574: 

575: 

576: \section{Main results}

577: \label{sect-results}

578: 

579: \subsection*{Model}

580: We consider the Cauchy problem for the Klein-Gordon equation

581: with the nonlinearity concentrated at the points

582: $X\sb{1}<X\sb{2}<\dots<X\sb{N}$:

583: \begin{equation}\label{kg-no}

584: \left\{

585: \begin{array}{l}

586: \ddot\psi(x,t)

587: =\psi''(x,t)-m^2\psi(x,t)

588: +\sum\sb{J}\delta(x-X\sb{J})F\sb{J}(\psi(X\sb{J},t)),

589: \qquad

590: x\in\R,

591: \quad t\in\R,

592: \\

593: \psi\at{t=0}=\psi\sb{0}(x),

594: \qquad

595: \dot\psi\at{t=0}=\pi\sb{0}(x).

596: \end{array}\right.

597: \end{equation}

598: If we identify a complex number $\psi=u+i v\in\C$

599: with the two-dimensional vector

600: $(u,v)\in\R\sp 2$,

601: then, physically, equation (\ref{kg-no}) describes small crosswise

602: oscillations of the infinite

603: string in three-dimensional space

604: $(x,u,v)$

605: stretched along the $x$-axis.

606: The string is subject to

607: the action of

608: an ``elastic force'' $-m^2\psi(x,t)$ and

609: coupled to nonlinear oscillators

610: of forces $F\sb{J}(\psi)$

611: attached at the points $X\sb{J}$.

612: We denote by $\mathcal{X}$

613: the set of all the locations of oscillators:

614: \begin{equation}\label{def-x}

615: \mathcal{X}=\{X\sb 1,\,X\sb 2,\,\dots,\,X\sb{N}\}.

616: \end{equation}

617: We will assume that

618: the oscillator forces $F\sb{J}$ admit real-valued potentials:

619: \begin{equation}\label{P}

620: F\sb{J}(\psi)=-\nabla U\sb{J}(\psi),\quad\psi\in\C,

621: \qquad

622: U\sb{J}\in C\sp 2(\C),

623: \end{equation}

624: where the gradient is taken with respect to $\Re\psi$ and $\Im\psi$.

625: We define

626: $\Psi(t)=

627: \left[\!\scriptsize{\begin{array}{c}

628: \psi(x,t)\\\pi(x,t)\end{array}}\!\right]$

629: and write the Cauchy problem

630: (\ref{kg-no})

631: in the vector form:

632: \begin{equation}\label{kg-no-cp}

633: \dot\Psi(t)

634: =

635: \left[\begin{array}{cc}0&1\\\p\sb x\sp 2-m^2&0\end{array}\right]

636: \Psi(t)

637: +

638: \sum\sb{J}

639: \delta(x-X\sb{J})\left[\begin{array}{c}0\\F\sb{J}(\psi)\end{array}\right],

640: \qquad

641: \Psi\at{t=0}

642: =\Psi\sb{0}

643: \equiv\left[\begin{array}{c}\psi\sb{0}\\\pi\sb{0}\end{array}\right].

644: \end{equation}

645: Equation (\ref{kg-no-cp})

646: formally can be written as a Hamiltonian system,

647: \begin{equation}\label{kg-no-h}

648: \dot\Psi(t)=\mathcal{J}\,D\mathcal{H}(\Psi),

649: \qquad

650: \mathcal{J}=\left[\begin{array}{cc}0&1\\-1&0\end{array}\right],

651: \end{equation}

652: where $D\mathcal{H}$ is the variational

653: derivative of the Hamilton functional

654: \begin{equation}\label{hamiltonian}

655: \mathcal{H}(\Psi)

656: =\frac 1 2

657: \int\limits\sb{\R}

658: \left(

659: \abs{\pi}\sp 2+\abs{\psi'}\sp 2+m^2\abs{\psi}\sp 2

660: \right)

661: dx

662: +\sum\sb{J} U\sb{J}(\psi(X\sb{J})),

663: \qquad

664: \Psi=\left[\begin{array}{c}\psi(x)\\\pi(x)\end{array}\right].

665: \end{equation}

666: We

667: assume that the potentials $U\sb{J}(\psi)$ are $\mathbf{U}(1)$-invariant,

668: where $\mathbf{U}(1)$ stands for the unitary group

669: $e\sp{i\theta}$, $\theta\in\R\mod 2\pi$.

670: Namely, we assume that

671: there exist $u\sb{J}\in C\sp 2(\R)$ such that

672: \begin{equation}\label{inv-u}

673: U\sb{J}(\psi)=u\sb{J}(\abs{\psi}\sp 2),

674: \qquad\psi\in\C,

675: \quad 1\le J\le N.

676: \end{equation}

677: 

678: \begin{remark}

679: In the context of the model of the infinite string in $\R\sp{3}$

680: that we described after (\ref{kg-no}),

681: the assumption (\ref{inv-u})

682: means that the potentials $U\sb{J}(\psi)$

683: are rotation-invariant with respect to the $x$-axis.

684: \end{remark}

685: 

686: Conditions (\ref{P}) and (\ref{inv-u})

687: imply that

688: \begin{equation}\label{def-a}

689: F\sb{J}(\psi)=

690: \alpha\sb{J}(\abs{\psi}^2)\psi,

691: \qquad\psi\in\C,

692: \end{equation}

693: where

694: $\alpha\sb{J}(\cdot)

695: =-2 u\sb{J}'(\cdot)\in C\sp 1(\R)$

696: are real-valued.

697: Therefore, (\ref{inv-f}) holds.

698: Since (\ref{kg-no-cp}) is $\mathbf{U}(1)$-invariant,

699: the N\"other theorem formally implies that the {\it charge functional}

700: \begin{equation}\label{cal-Q}

701: \mathcal{Q}(\Psi)

702: =\frac{i}{2}\int\sb{\R}

703: \left(\overline\psi\pi-\overline{\pi}\psi\right)\,dx,

704: \qquad

705: \Psi=\left[\begin{array}{c}\psi(x)\\\pi(x)\end{array}\right],

706: \end{equation}

707: is conserved for solutions $\Psi(t)$ to (\ref{kg-no-cp}).

708: 

709: Let us introduce

710: the phase space ${\E}$

711: of finite energy states for equation (\ref{kg-no}).

712: Denote by $L\sp 2$ the complex Hilbert space $L\sp 2(\R)$

713: with the norm $\norm{\cdot}\sb{L\sp 2}$,

714: and denote by $\norm{\cdot}\sb{L\sp 2\sb R}$ the norm in $L\sp 2(-R,R)$

715: for $R>0$.

716: 

717: \begin{definition}

718: \begin{enumerate}

719: \item

720: ${\E}$ is the Hilbert space of the states

721: $\Psi=(\psi,\pi)$,

722: with the norm

723: \begin{equation}\label{def-e}

724: \norm{\Psi}\sb{\E}^2

725: :=

726: \norm{\pi}\sb{L\sp 2}^2

727: +\norm{\psi'}\sb{L\sp 2}^2+m^2\norm{\psi}\sb{L\sp 2}^2.

728: \end{equation}

729: \item

730: ${\E}\sb{F}$ is the space $\E$

731: endowed with the Fr\'echet topology

732: defined by local energy seminorms

733: \begin{equation}\label{def-e-r}

734: \norm{\Psi}\sb{\E,R}^2

735: :=

736: \norm{\pi}\sb{L\sp 2(-R,R)}^2

737: +\norm{\psi'}\sb{L\sp 2(-R,R)}^2+m^2\norm{\psi}\sb{L\sp 2(-R,R)}^2,

738: \qquad

739: R>0.

740: \end{equation}

741: \end{enumerate}

742: \end{definition}

743: 

744: \begin{remark}

745: The space $\E\sb{F}$ is metrizable.

746: The metric could be introduced by

747: \begin{equation}\label{def-e-metric}

748: \dist(\Psi,\Phi)

749: =\sum\sb{R=1}\sp{\infty} 2^{-R}

750: \norm{\Psi-\Phi}\sb{\E,R}.

751: \end{equation}

752: \end{remark}

753: 

754: Equation (\ref{kg-no-cp})

755: is formally a Hamiltonian system with

756: the phase space ${\E}$

757: and the Hamilton functional $\mathcal{H}$.

758: Both

759: $\mathcal{H}$

760: and $\mathcal{Q}$ are continuous functionals on ${\E}$.

761: Let us note that

762: ${\E}={H\sp{1}}\oplus L\sp 2$,

763: where $H\sp{1}$ denotes the Sobolev space

764: \[

765: H\sp{1}=H\sp{1}(\R)

766: =\{\psi(x)\in L\sp 2(\R):\;\psi'(x)\in L\sp 2(\R)\}.

767: \]

768: We introduced into (\ref{def-e})

769: the factor $m^2>0$,

770: to have a convenient relation

771: $\mathcal{H}(\psi,\dot\psi)

772: =\frac 1 2\norm{(\psi,\dot\psi)}\sb{\E}^2+\sum\sb{J} U\sb{J}(\psi(X\sb{J}))$.

773: %%The space ${\E}\sb{F}$ is metrizable (but not complete).

774: 

775: \subsection*{Global well-posedness}

776: 

777: To have a priori estimates available for the proof of the global

778: well-posedness, we assume that

779: \begin{equation}\label{bound-below}

780: U\sb{J}(\psi)\ge {A}\sb{J}-{B}\sb{J}\abs{\psi}^2

781: \quad{\rm for}\ \psi\in\C,\quad

782: {\rm where}\quad {A}\sb{J}\in\R,\quad {B}\sb{J}\ge 0,

783: \quad

784: 1\le J\le N;

785: \quad

786: \sum\sb{J} {B}\sb{J}<m.

787: \end{equation}

788: 

789: \begin{theorem}\label{theorem-well-posedness}

790: Let $F\sb{J}(\psi)$ satisfy conditions (\ref{P})

791: and (\ref{inv-u}):

792: \[

793: F\sb{J}(\psi)=-\nabla U\sb{J}(\psi),\qquad

794: U\sb{J}(\psi)=u\sb{J}(\abs{\psi}^2),

795: \qquad u\sb{J}(\cdot)\in C\sp 2(\R).

796: \]

797: Additionally,

798: assume that (\ref{bound-below}) holds.

799: Then:

800: \begin{enumerate}

801: \item

802: For every $\Psi\sb{0}\in {\E}$ the Cauchy problem

803: (\ref{kg-no-cp}) has a unique solution

804: $\Psi(t)$ such that $\Psi\in C(\R,{\E})$.

805: \item

806: The map

807: $W(t):\;\Psi\sb{0}\mapsto\Psi(t)$

808: is continuous in ${\E}$

809: for each $t\in\R$.

810: \item

811: The energy and charge are conserved:

812: $\mathcal{H}(\Psi(t))=\const$,

813: $\mathcal{Q}(\Psi(t))=\const$,

814: $t\in\R$.

815: \item

816: The following \emph{a priori} bound holds:

817: $

818: \norm{\Psi(t)}\sb{\E}

819: \le C(\Psi\sb{0})$,

820: $t\in\R$.

821: \end{enumerate}

822: \end{theorem}

823: 

824: We prove this Theorem in Appendix~\ref{sect-existence}.

825: 

826: %%%%%%%%%%%%%%%%%%%%%%%%%%

827: \subsection*{Solitary waves and the main theorem}

828: 

829: \begin{definition}\label{def-solitary-waves}

830: \begin{enumerate}

831: \item

832: The solitary waves of equation (\ref{kg-no})

833: are solutions of the form

834: \begin{equation}\label{solitary-waves}

835: \psi(x,t)=\phi\sb\omega(x)e\sp{-i\omega t},

836: \qquad

837: {\rm where}

838: \quad

839: \omega\in\R,

840: \quad

841: \phi\sb\omega\in H\sp{1}(\R).

842: \end{equation}

843: \item

844: The solitary manifold

845: is the set

846: $

847: \bS

848: =

849: \left\{

850: (\phi\sb\omega,-i\omega\phi\sb\omega)

851: \sothat\omega\in\R,\ \phi\sb\omega\in H\sp 1(\R)

852: \right\}

853: \subset\E.

854: $

855: \end{enumerate}

856: \end{definition}

857: 

858: \begin{remark}

859: \begin{enumerate}

860: \item

861: Identity (\ref{inv-f}) implies that the set

862: $\bS$

863: is invariant under multiplication by $e\sp{i\theta}$,

864: $\theta\in\R$.

865: \item

866: Let us note that for any $\omega\in\R$

867: there is a zero solitary wave with

868: $\phi\sb\omega(x)\equiv 0$

869: since $F\sb{J}(0)=0$ by (\ref{def-a}).

870: \item

871: According to (\ref{def-a}),

872: $\alpha\sb{J}(\abs{C}^2)=F\sb{J}(C)/C\in\R$

873: for any $C\in\C\backslash 0$.

874: \end{enumerate}

875: \end{remark}

876: 

877: \begin{definition}\label{def-sn}

878: The function $F\sb{J}(\psi)$ is \emph{strictly nonlinear}

879: if the equation

880: $\alpha\sb{J}(C^2)=a$

881: has a discrete

882: (or empty)

883: set of positive roots

884: $C$ for each particular $a\in\R$.

885: \end{definition}

886: 

887: The following proposition provides a concise description

888: of all solitary waves.

889: Formally this proposition 

890: is not necessary for our exposition.

891: \begin{proposition}

892: \label{prop-solitons}

893: Assume that $F\sb{J}(\psi)$ satisfy (\ref{inv-f}) and that

894: $F\sb{J}(\psi)$, $1\le J\le N$, are strictly nonlinear

895: in the sense of Definition~\ref{def-sn}.

896: Then all solitary wave solutions

897: to {(\ref{kg-no})}

898: are given by {(\ref{solitary-waves})} with

899: \begin{equation}\label{solitary-wave-profile}

900: \phi\sb\omega(x)

901: =\sum\sb{J}

902: C\sb{J} e^{-\kappa(\omega)\abs{x-X\sb{J}}},

903: \qquad

904: \kappa(\omega)=\sqrt{m^2-\omega^2},

905: \end{equation}

906: where

907: $\omega\in[-m,m]$

908: and $C\sb{J}\in\C$,

909: $1\le J\le N$,

910: satisfy the following relations:

911: \begin{equation}\label{kaka}

912: %\kappa^2=m^2-\omega^2,

913: %\qquad

914: 2\kappa(\omega) C\sb{J}

915: =F\sb{J}\Big(\sum\sb{K}C\sb K e^{-\kappa(\omega)\abs{X\sb{J}-X\sb K}}\Big).

916: \end{equation}

917: \end{proposition}

918: 

919: \begin{remark}\label{remark-zero}

920: By (\ref{solitary-wave-profile}),

921: $\omega=\pm m$ can only correspond to zero solution.

922: \end{remark}

923: 

924: The proof of this Proposition repeats the proof

925: of a similar result for the case $N=1$

926: in \cite{ubk-arma}.

927: 

928: As we mentioned before,

929: we need to assume that the nonlinearities are

930: nonlinear polynomials.

931: This condition

932: is crucial in our argument:

933: It will allow to apply the Titchmarsh convolution theorem.

934: 

935: \bigskip

936: 

937: Let us formulate all the assumptions which we need

938: to formulate the main result.

939: 

940: \begin{assumption}\label{ass-a}

941: For all $1\le J\le N$,

942: \begin{equation}\label{f-is-such}

943: F\sb{J}(\psi)=-\nabla U\sb{J}(\psi),

944: \quad{\rm where}\quad

945: U\sb{J}(\psi)=\sum\limits\sb{n=0}\sp{p\sb{J}}u\sb{J,n}

946: \abs{\psi}\sp{2n}~,

947: %\end{equation}

948: %\begin{equation}

949: \qquad

950: \ \ u\sb{J,n}\in\R.

951: \end{equation}

952: \end{assumption}

953: 

954: \begin{assumption}\label{ass-nonl}

955: For all $1\le J\le N$, we have

956: \begin{equation}\label{nonli}

957: u\sb{J,p\sb{J}}>0

958: \quad

959: {\rm and}

960: \quad

961: p\sb{J}\ge 2.

962: \end{equation}

963: \end{assumption}

964: 

965: Assumptions~\ref{ass-a} and \ref{ass-nonl} guarantee that

966: all nonlinearities $F\sb{J}$ are strictly nonlinear

967: and satisfy (\ref{P}), (\ref{inv-u}),

968: and also that the bound (\ref{bound-below}) takes place.

969: %Moreover, by Proposition~\ref{prop-solitons},

970: %all nonzero solitary waves correspond to $\omega\in\R$.

971: %The convergence to the attractor

972: %will be understood as the uniform convergence

973: %of bounded continuous functions $\psi\in C\sb{b}(\R)$

974: %on compact sets.

975: %We denote by $C\sb{b,F}$

976: %the space $ C\sb{b}(\R)$ endowed with such convergence.

977: 

978: We introduce the following quantities:

979: \begin{equation}\label{def-mu}

980: \mu\sb{1}=m,

981: \quad

982: \mu\sb{J+1}=(2p\sb{J}-1)\mu\sb{J};

983: \qquad

984: \mu'\sb{N}=m,

985: \quad

986: %\mu'\sb{J}=(2p\sb{J}-1)\mu'\sb{J+1},

987: %%AC important correction ????

988: \mu'\sb{J}=(2p\sb{J+1}-1)\mu'\sb{J+1},

989: \qquad

990: 1\le J\le N-1,

991: \end{equation}

992: where $p\sb J$ are exponentials from (\ref{f-is-such}).

993: We also denote

994: \begin{equation}\label{def-m-lambda}

995: \Lambda=\max\limits\sb{1\le J\le N}(2p\sb{J}-1){M}\sb{J},

996: \qquad

997: {\rm where}

998: \quad

999: {M}\sb{J}=\min(\mu\sb{J},\mu'\sb{J}).

1000: \end{equation}

1001: 

1002: %We will show later that

1003: %the spectrum of any omega-limit trajectory

1004: %belongs to the intervals $[-\Lambda,\Lambda]$.

1005: 

1006: \begin{assumption}\label{ass-delta-small}

1007: The intervals

1008: $[X\sb{J},X\sb{J+1}]$, $\ 1\le J\le N-1$,

1009: are small enough so that

1010: \begin{equation}\label{delta-small}

1011: \Lambda<

1012: \sqrt{\frac{\pi^2}{\abs{X\sb{J+1}-X\sb{J}}^2}+m^2},

1013: \qquad

1014: 1\le J\le N-1.

1015: \end{equation}

1016: \end{assumption}

1017: 

1018: Our main result is the following theorem.

1019: 

1020: \begin{theorem}[Main Theorem]

1021: \label{main-theorem}

1022: Let Assumptions~\ref{ass-a}, ~\ref{ass-nonl}, and

1023: \ref{ass-delta-small} hold.

1024: Then for any $\Psi\sb{0}\in\E$

1025: the solution $\Psi(t)\in C(\R,\E)$

1026: to the Cauchy problem {(\ref{kg-no-cp})}

1027: converges to

1028: $\bS$:

1029: \begin{equation}\label{cal-A}

1030: \lim\sb{t\to\pm\infty}\dist(\Psi(t),\bS)=0,

1031: \end{equation}

1032: where

1033: $\dist(\Psi,\bS):=\inf\limits\sb{\bm s\in\bS}\dist(\Psi,\bm s)$,

1034: and $\dist$ is introduced in (\ref{def-e-metric}).

1035: \end{theorem}

1036: 

1037: \begin{remark}

1038: \begin{enumerate}

1039: \item

1040: The solution $\Psi(t)$ exists by Theorem~\ref{theorem-well-posedness}

1041: since Assumptions~\ref{ass-a} and \ref{ass-nonl}

1042: guarantee that conditions 

1043: (\ref{P}), (\ref{inv-u}), and (\ref{bound-below})

1044: hold.

1045: \item

1046: It suffices to prove Theorem~\ref{main-theorem}

1047: for $t\to +\infty$.

1048: \item

1049: In Sections~\ref{sect-example-w} and \ref{sect-example-li},

1050: we construct counterexamples to the

1051: convergence (\ref{cal-A})

1052: in the case when 

1053: Assumption~\ref{ass-nonl}

1054: or Assumption~\ref{ass-delta-small}

1055: are not satisfied.

1056: \item

1057: For the real initial data,

1058: we obtain a real-valued solution $\psi(t)$ to (\ref{kg-no}).

1059: Therefore, the convergence (\ref{cal-A})

1060: of $\Psi(t)=(\psi(t),\dot\psi(t))$

1061: to the set of pairs

1062: $(\phi\sb\omega,-i\omega\phi\sb\omega)$

1063: with $\omega\in\R$

1064: implies that $\Psi(t)$

1065: locally converges to zero:

1066: \[

1067: \lim\sb{t\to\infty}\dist(\Psi(t),0)=0.

1068: \]

1069: \end{enumerate}

1070: \end{remark}

1071: 

1072: 

1073: 

1074: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1075: \section{Separation of dispersive component}

1076: \label{sect-splitting}

1077: 

1078: Let us split the solution

1079: $\psi(x,t)$

1080: into two components,

1081: $\psi(x,t)=\chi(x,t)+\varphi(x,t)$,

1082: which are defined for all $t\in\R$

1083: as solutions to the following Cauchy problems:

1084: \begin{eqnarray}

1085: &&

1086: \ddot\chi(x,t)=\chi''(x,t)-m^2\chi(x,t),

1087: \qquad

1088: (\chi,\dot\chi)\at{t=0}=(\psi\sb{0}(x),\pi\sb{0}(x)),

1089: \label{kg-no-cp-1}

1090: \\

1091: \nonumber

1092: \\

1093: &&

1094: \ddot\varphi(x,t)=\varphi''(x,t)-m^2\varphi(x,t)

1095: +

1096: \sum\sb{J}

1097: \delta(x-X\sb{J})f\sb{J}(t),

1098: \qquad

1099: (\varphi,\dot\varphi)\at{t=0}=(0,0),

1100: \label{kg-no-cp-2-0}

1101: \end{eqnarray}

1102: where

1103: $(\psi\sb{0}(x),\pi\sb{0}(x))$ is the initial data from (\ref{kg-no}),

1104: and

1105: \begin{equation}\label{def-f}

1106: f\sb{J}(t):=F\sb{J}(\psi(X\sb{J},t)),

1107: \qquad

1108: t\in\R.

1109: \end{equation}

1110: The following lemma is proved in \cite[Lemma 3.1]{ubk-arma}.

1111: \begin{lemma}\label{lemma-decay-psi1}

1112: There is a local energy decay for $\chi$:

1113: %%in the space ${\E}$:

1114: \begin{equation}\label{dp0}

1115: \lim\sb{t\to\infty}

1116: \Norm{(\chi(\cdot,t),\dot\chi(\cdot,t))}\sb{{\E},R}=0,

1117: \qquad\forall R>0.

1118: \end{equation}

1119: \end{lemma}

1120: 

1121: Let $k(\omega)$ be the analytic function

1122: with the domain 

1123: $D:=\C\backslash((-\infty,-m]\cup[m,+\infty))$

1124: such that

1125: \begin{equation}\label{def-k}

1126: k(\omega)=\sqrt{\omega\sp 2-m^2},

1127: \qquad\Im k(\omega)>0,

1128: \qquad

1129: \omega

1130: \in D.

1131: \end{equation}

1132: Let us also denote its limit values for $\omega\in\R$ by

1133: \begin{equation}\label{def-k-plus}

1134: k\sb{\pm}(\omega):=k(\omega\pm i0),

1135: \qquad

1136: \omega\in\R.

1137: \end{equation}

1138: 

1139: \bigskip

1140: 

1141: \begin{figure}[htbp]

1142: \input ukk-domain.tex

1143: \caption{Domain $D$

1144: and the values of $k\sb{\pm}(\omega):=k(\omega\pm i0)$, $\omega\in\R$.}

1145: \label{fig-domain}

1146: \end{figure}

1147: As illustrated on Figure~\ref{fig-domain} 

1148: (where all square roots take positive values),

1149: we have

1150: \begin{equation}\label{k-plus-minus}

1151: k\sb{-}(\omega)=k\sb{+}(\omega)

1152: \quad{\rm for}\quad

1153: -m\le\omega\le m,

1154: \qquad

1155: k\sb{-}(\omega)=-k\sb{+}(\omega)

1156: \quad{\rm for}\quad

1157: \omega\in\R\backslash(-m,m),

1158: \end{equation}

1159: and also

1160: \begin{equation}\label{omega-k}

1161: \omega\,k\sb{+}(\omega)\ge 0

1162: \qquad

1163: {\rm for}

1164: \quad

1165: \omega\in\R\backslash(-m,m).

1166: \end{equation}

1167: 

1168: %%AC sentences swapped:

1169: We set $\mathcal{F}\sb{t\to\omega}[g(t)]

1170: =\displaystyle\int\sb{\R} e^{i\omega t}g(t)\,dt$

1171: for a function $g(t)$ from the Schwartz space $\mathscr{S}(\R)$.

1172: Let us study the Fourier transform 

1173: $\hat\chi(x,\omega):=\mathcal{F}\sb{t\to\omega}[\chi(x,t)]$,

1174: which is a continuous function of $x$

1175: valued in tempered distributions.

1176: 

1177: \begin{lemma}\label{lemma-chi-h1}

1178: \begin{itemize}

1179: \item

1180: $\hat\chi(x,\omega)$

1181: is a continuous function of 

1182: $x\in\R$ with values in 

1183: $L^1\sb{\rm loc}(\R)$, and 

1184: \begin{equation}\label{nul}

1185: \hat\chi(x,\omega)=0,

1186: \qquad

1187: \abs{\omega}<m.

1188: \end{equation}

1189: \item

1190: The following bound holds:

1191: \begin{equation}

1192: \sup\limits\sb{x\in\R}~\int\sb{\abs{\omega}>m}

1193: \abs{\hat\chi(x,\omega)}^2\omega\,k\sb{+}(\omega)\,d\omega<\infty.

1194: \end{equation}

1195: \end{itemize}

1196: \end{lemma}

1197: 

1198: \begin{proof}

1199: Set $\omega(k)=\sgn k\sqrt{m^2+k^2}$

1200: for $k\in\R$.

1201: Note that 

1202: the function $k\sb{+}(\omega)$

1203: for $\abs{\omega}>m$

1204: is inverse to the function $\omega(k)$, $k\ne 0$.

1205: We have:

1206: \begin{equation}

1207: \chi(x,t)=\frac{1}{2\pi}\int\sb{\R}

1208: e^{-i k x}\Big[

1209: \hat\psi\sb{0}(k)\cos(\omega(k)t)

1210: +\hat\pi\sb{0}(k)\frac{\sin(\omega(k)t)}{\omega(k)}

1211: \Big]\,dk.

1212: \end{equation}

1213: Hence, for the Fourier transform of $\chi(x,t)$,

1214: we obtain, for any $x\in\R$:

1215: \begin{eqnarray}

1216: &&

1217: \hat\chi(x,\omega)

1218: =\int\sb{\R}

1219: e^{-i k x}

1220: \Big[

1221: \hat\psi\sb{0}(k)\frac{\delta(\omega-\omega(k))+\delta(\omega+\omega(k))}{2}

1222: +\hat\pi\sb{0}(k)\frac{\delta(\omega-\omega(k))-\delta(\omega+\omega(k))}

1223: {2i\omega(k)}

1224: \Big]

1225: \,dk

1226: \nonumber

1227: \\

1228: &&

1229: \qquad

1230: =\int\sb{\abs{\omega'}>m}

1231: e^{-i k\sb{+}(\omega') x}

1232: \Big[

1233: \hat\psi\sb{0}(k\sb{+}(\omega'))\frac{\delta(\omega-\omega')+\delta(\omega+\omega')}{2}

1234: +\hat\pi\sb{0}(k\sb{+}(\omega'))\frac{\delta(\omega-\omega')-\delta(\omega+\omega')}

1235: {2i\omega'}

1236: \Big]

1237: \,\frac{\omega'\,d\omega'}{k\sb{+}(\omega')}.

1238: \nonumber

1239: \end{eqnarray}

1240: The above relation is understood in the sense of distributions

1241: of $\omega\in\R$.

1242: We used the substitution $k=k\sb{+}(\omega')$.

1243: Now (\ref{nul}) is obvious.

1244: Evaluating the last integral, we get:

1245: \[

1246: \hat\chi(x,\omega)

1247: =

1248: \frac{\omega}{2k\sb{+}(\omega)}

1249: \left\{

1250: e^{-i k\sb{+}(\omega) x}

1251: \hat\psi\sb{0}(k\sb{+}(\omega))

1252: +

1253: e^{i k\sb{+}(\omega) x}

1254: \hat\psi\sb{0}(-k\sb{+}(\omega))

1255: +

1256: e^{-i k\sb{+}(\omega) x}

1257: \frac{\hat\pi\sb{0}(k\sb{+}(\omega))}{i\omega}

1258: -

1259: e^{i k\sb{+}(\omega) x}

1260: \frac{\hat\pi\sb{0}(-k\sb{+}(\omega))}{i\omega}

1261: \right\},

1262: \qquad\abs{\omega}>m.\nonumber

1263: \]

1264: We took into account that

1265: $k\sb{+}(-\omega)=-k\sb{+}(\omega)$ for $\omega\in\R\backslash(-m,m)$

1266: %$\abs{\omega}\ge m$

1267: (see (\ref{k-plus-minus})).

1268: Thus, we have:

1269: \[

1270: \int\limits\sb{\abs{\omega}>m}

1271: \abs{\hat\chi(x,\omega)}^2

1272: \omega\,k\sb{+}(\omega)\,d\omega

1273: \le

1274: \int\limits\sb{\abs{\omega}>m}

1275: \Big[

1276: \frac{\omega^2

1277: \abs{\hat\psi\sb{0}(k\sb{+}(\omega))}^2}{k\sb{+}^2(\omega)}

1278: +

1279: \frac{\abs{\hat\pi\sb{0}(k\sb{+}(\omega))}^2}{k\sb{+}^2(\omega)}

1280: \Big]

1281: \omega\,k\sb{+}(\omega)\,d\omega

1282: =

1283: \int\limits\sb{\R}

1284: \Big[

1285: \abs{\hat\psi\sb{0}(k)}^2

1286: +

1287: \frac{\abs{\hat\pi\sb{0}(k)}^2}

1288: {\omega^2(k)}

1289: \Big]

1290: \omega^2(k)\,dk.

1291: \]

1292: The finiteness of the right-hand side

1293: follows from the finiteness of the energy of the

1294: initial data $(\psi\sb{0},\pi\sb{0})$:

1295: \[

1296: \norm{(\psi\sb{0},\pi\sb{0})}\sb{\E}^2

1297: =

1298: \frac{1}{2\pi}

1299: \int\sb{\R}

1300: \Big[

1301: \omega^2(k)

1302: \abs{\hat\psi\sb{0}(k)}^2

1303: +

1304: \abs{\hat\pi\sb{0}(k)}^2

1305: \Big]

1306: \,dk

1307: <\infty.

1308: \]

1309: \end{proof}

1310: 

1311: \section{Spectral representation}

1312: \label{sect-spec}

1313: The function $\varphi(x,t)=\psi(x,t)-\chi(x,t)$

1314: satisfies the following Cauchy problem:

1315: \begin{equation}

1316: \ddot\varphi(x,t)=\varphi''(x,t)-m^2\varphi(x,t)

1317: +

1318: \sum\sb{J}

1319: \delta(x-X\sb{J})f\sb{J}(t),

1320: \qquad

1321: (\varphi,\dot\varphi)\at{t=0}=(0,0),

1322: \label{kg-no-cp-2}

1323: \end{equation}

1324: with $f\sb{J}(t)$ defined in (\ref{def-f}).

1325: Note that

1326: $\psi(X\sb{J},\cdot)\in C\sb{b}(\R)$

1327: for $1\le J\le N$

1328: by the Sobolev embedding,

1329: since $(\psi(x,t),\dot\psi(x,t))\in C\sb{b}(\R,\E)$

1330: by Theorem~\ref{theorem-well-posedness}~({\it iv}).

1331: Hence, $f\sb J(t)\in C\sb{b}(\R)$.

1332: On the other hand,

1333: since $\chi(x,t)$

1334: is a finite energy solution to the free Klein-Gordon equation,

1335: we also have

1336: \begin{equation}\label{psi-1-bounds}

1337: (\chi(x,t),\dot\chi(x,t))\in C\sb{b}(\R,\E).

1338: \end{equation}

1339: Therefore, the function $\varphi(x,t)=\psi(x,t)-\chi(x,t)$ satisfies

1340: \begin{equation}\label{psi-2-bounds}

1341: (\varphi(x,t),\dot\varphi(x,t))\in C\sb{b}(\R,\E).

1342: \end{equation}

1343: The Fourier transform

1344: \begin{equation}

1345: \hat\varphi(x,\omega)

1346: =\mathcal{F}\sb{t\to\omega}[\varphi(x,t)],

1347: \qquad

1348: (x,\omega)\in\R^2,

1349: \end{equation}

1350: is a continuous function of $x\in\R$ 

1351: %\ft{prokommentirovat' vse mesta gde ispolzuetsa eta neporerivnost'!!}

1352: with 

1353: values in 

1354: tempered distributions of $\omega\in\R$.

1355: It satisfies the following equation (Cf. (\ref{kg-no-cp-2})):

1356: \begin{equation}\label{kg-no-cp-2-st}

1357: -\omega^2\hat\varphi(x,\omega)

1358: =

1359: \hat\varphi''(x,\omega)

1360: -m^2\hat\varphi(x,\omega)

1361: +\sum\sb{J}\delta(x-X\sb{J})\hat f\sb{J}(\omega),

1362: \qquad

1363: (x,\omega)\in\R^2.

1364: \end{equation}

1365: 

1366: We are going to construct a representation

1367: for the solution $\hat\varphi(x,\omega)$

1368: in a form suitable for our purposes.

1369: 

1370: \begin{lemma}\label{lemma-41}

1371: $\hat\varphi$ is a smooth function of $x\in\R\backslash\mathcal{X}$

1372: (where $\mathcal{X}=\{X\sb 1,\,X\sb 2,\,\dots,\,X\sb{N}\}$),

1373: with values in tempered distributions of $\omega\in\R$,

1374: and there exist quasimeasures

1375: $\hat\varPhi\sb{J}\sp{\pm}$,

1376: $1\le J\le N$,

1377: and

1378: $\hat\varTheta\sb J$,

1379: $1\le J\le N-1$,

1380: so that

1381: \begin{equation}\label{c-s}

1382: \hat\varphi(x,\omega)

1383: =

1384: \left\{\begin{array}{l}

1385: \hat\varPhi\sb{1}\sp{+}(\omega)e^{-ik\sb{+}(\omega)(x-X\sb{1})}

1386: +

1387: \hat\varPhi\sb{1}\sp{-}(\omega)e^{-ik\sb{-}(\omega)(x-X\sb{1})},

1388: \qquad

1389: x\le X\sb{1},

1390: \\

1391: \hat\varPhi\sb{J}(\omega)

1392: \cos(k\sb{+}(\omega)(x-X\sb{J}))

1393: +\hat\varTheta\sb{J}(\omega)\frac{\sin(k\sb{+}(\omega)(x-X\sb{J}))}{k\sb{+}(\omega)},

1394: \quad

1395: x\in [X\sb{J},X\sb{J+1}],\quad 1\le J\le N-1,

1396: \\

1397: \hat\varPhi\sb{N}\sp{+}(\omega)e^{ik\sb{+}(\omega)(x-X\sb{N})}

1398: +

1399: \hat\varPhi\sb{N}\sp{-}(\omega)e^{ik\sb{-}(\omega)(x-X\sb{N})},

1400: \qquad

1401: x\ge X\sb{N},

1402: \end{array}

1403: \right.

1404: \end{equation}

1405: where

1406: $

1407: \hat\varPhi\sb{J}(\omega)

1408: :=

1409: \hat\varPhi\sb{J}\sp{+}(\omega)+\hat\varPhi\sb{J}\sp{-}(\omega).

1410: $

1411: \end{lemma}

1412: 

1413: \begin{remark}\label{remark-quasimeasure}

1414: A tempered distribution $\mu(\omega)\in\mathscr{S}'(\R)$

1415: is called a {\it quasimeasure} if

1416: $\check\mu(t)=\mathscr{F}\sp{-1}\sb{\omega\to t}[\mu(\omega)]\in C\sb{b}(\R)$.

1417: For more details, see \cite[Appendix B]{ubk-arma}.

1418: \end{remark}

1419: 

1420: \begin{remark}\label{remark-relations}

1421: The representation (\ref{c-s})

1422: %Lemma~\ref{lemma-41}

1423: implies that

1424: \begin{equation}\label{a-is-varphi}

1425: \hat\varPhi\sb{J}(\omega)

1426: =

1427: \hat\varphi(X\sb J,\omega),

1428: \quad

1429: 1\le J\le N,

1430: \end{equation}

1431: \begin{equation}\label{varphi-is-varphi}

1432: \hat\varPhi\sb{1}\sp{+}(\omega)+\hat\varPhi\sb{1}\sp{-}(\omega)

1433: =\hat\varPhi\sb{1}(\omega)=\hat\varphi(X\sb{1},\omega),

1434: \qquad

1435: \hat\varPhi\sb{N}\sp{+}(\omega)+\hat\varPhi\sb{N}\sp{-}(\omega)

1436: %=\hat\varPhi\sb{N}(\omega)

1437: =\hat\varphi(X\sb{N},\omega),

1438: \end{equation}

1439: and also that

1440: \begin{equation}\label{a-is-varphi-prime}

1441: \hat\varphi'(X\sb J+0,\omega)=\hat\varTheta\sb{J}(\omega),

1442: \quad

1443: 1\le J\le N-1.

1444: \end{equation}

1445: \end{remark}

1446: 

1447: \begin{proof}

1448: {\bf Step 1: Complex Fourier-Laplace transform.}\quad

1449: We denote

1450: \begin{equation}\label{def-f-pm}

1451: f\sp{\pm}\sb{J}(t):=\theta(\pm t)f\sb{J}(t)

1452: =\theta(t)F\sb{J}(\psi(X\sb{J},t))

1453: \end{equation}

1454: and split $\varphi(x,t)$

1455: into

1456: \begin{equation}\label{def-varphi-pm}

1457: \varphi(x,t)=\varphi\sp{+}(x,t)+\varphi\sp{-}(x,t),

1458: \qquad

1459: {\rm where}

1460: \quad

1461: \varphi\sp\pm(x,t):=\theta(\pm t)\varphi(x,t).

1462: \end{equation}

1463: Then $\varphi\sp\pm(x,t)$ satisfy

1464: \begin{equation}

1465: \ddot\varphi\sp\pm(x,t)=\p\sb x^2\varphi\sp\pm(x,t)-m^2\varphi\sp\pm(x,t)

1466: +

1467: \sum\sb{J}

1468: \delta(x-X\sb{J})f\sb{J}\sp\pm(t),

1469: \qquad

1470: t\in\R,

1471: \label{kg-no-cp-2-pm}

1472: \end{equation}

1473: since $(\varphi\sp\pm,\dot\varphi\sp\pm)\at{t=0}=(0,0)$.

1474: Let us analyze the complex Fourier-Laplace transforms of

1475: $\varphi\sp\pm(x,t)$:

1476: \begin{equation}\label{FL}

1477: \displaystyle

1478: \tilde\varphi\sp\pm(x,\omega)

1479: =\mathcal{F}\sb{t\to\omega}[\theta(\pm t)\varphi(x,t)]

1480: :=

1481: \int\sb{0}\sp\infty e\sp{i\omega t}\theta(\pm t)\varphi(x,t)\,dt,

1482: \quad\omega\in\C\sp{\pm},

1483: \end{equation}

1484: where

1485: $\C\sp{\pm}:=\{z\in\C:\;\pm\Im z>0\}$.

1486: Due to (\ref{psi-2-bounds}),

1487: $\tilde\varphi\sp{\pm}(\cdot,\omega)$

1488: are $H\sp{1}$-valued analytic functions of $\omega\in\C\sp{\pm}$.

1489: In what follows, we will consider $\varphi\sp{+}$;

1490: the function $\varphi\sp{-}$ considered in the same way.

1491: 

1492: Equation (\ref{kg-no-cp-2-pm})

1493: implies that

1494: $\tilde\varphi\sp{+}$ satisfies

1495: \begin{equation}\label{FL-2}

1496: -\omega\sp 2\tilde\varphi\sp{+}(x,\omega)

1497: =

1498: \p\sb x^2\tilde\varphi\sp{+}(x,\omega)-m^2\tilde\varphi\sp{+}(x,\omega)

1499: +\sum\sb{J}\delta(x-X\sb{J})\tilde f\sb{J}\sp{+}(\omega),

1500: \quad\omega\in\C\sp{+}.

1501: \end{equation}

1502: The fundamental solutions

1503: $\displaystyle

1504: G\sb\pm(x,\omega)=\frac{e\sp{\pm i k(\omega)\abs{x}}}{\pm 2i k(\omega)}$

1505: satisfy

1506: \[

1507: G\sb\pm''(x,\omega)+(\omega^2-m^2)G\sb\pm(x,\omega)

1508: =\delta(x),

1509: \qquad\omega\in\C\sp{+}.

1510: \]

1511: The solution $\tilde\varphi\sp{+}(x,\omega)$

1512: could be written as a linear combination of these fundamental solutions.

1513: We use the standard ``limiting absorption principle''

1514: for the selection of the appropriate fundamental solution: Since

1515: $\tilde\varphi\sp{+}(\cdot,\omega)\in H\sp{1}$

1516: for $\omega\in\C\sp{+}$,

1517: only $G\sb{+}$ is acceptable,

1518: because for $\omega\in\C\sp{+}$ the function $G\sb{+}(\cdot,\omega)$

1519: is in $H\sp{1}$

1520: by definition (\ref{def-k}),

1521: while $G\sb{-}$ is not.

1522: This suggests the following representation:

1523: \begin{equation}\label{tilde-psi-tilde-f}

1524: \tilde\varphi\sp{+}(x,\omega)

1525: =-\sum\sb{J}

1526: \tilde f\sp{+}\sb{J}(\omega)G\sb{+}(x-X\sb{J},\omega)

1527: =-\sum\sb{J}

1528: \tilde f\sp{+}\sb{J}(\omega)\frac{e\sp{i k(\omega)\abs{x-X\sb{J}}}}{2i k(\omega)},

1529: \qquad\omega\in\C\sp{+}.

1530: \end{equation}

1531: The proof is straightforward since (\ref{tilde-psi-tilde-f})

1532: belongs to $H\sp{1}(\R)$ for $\omega\in\C\sp{+}$

1533: while the solution to (\ref{FL-2})

1534: which is an $H\sp{1}$-valued analytic function in $\omega$

1535: is unique.

1536: For $x\le X\sb{1}$, the relation (\ref{tilde-psi-tilde-f}) yields

1537: \begin{equation}\label{phi-1}

1538: \tilde\varphi\sp{+}(x,\omega)

1539: =-\sum\sb{J}

1540: \tilde f\sp{+}\sb{J}(\omega)\frac{e\sp{-i k(\omega)(x-X\sb{J})}}{2i k(\omega)}

1541: =e^{-i k(\omega)(x-X\sb{1})}\tilde\varphi\sp{+}(X\sb{1},\omega),

1542: \qquad

1543: x\le X\sb{1},

1544: \quad

1545: \omega\in\C\sp{+}.

1546: \end{equation}

1547: %Similarly, for $x\ge X\sb{N}$, the relation (\ref{tilde-psi-tilde-f}) yields

1548: %\begin{equation}\label{phi-N}

1549: %\tilde\varphi\sp{+}(x,\omega)

1550: %=e^{i k(\omega)(x-X\sb{N})}\tilde\varphi\sp{+}(X\sb{N},\omega),

1551: %\qquad

1552: %x\ge X\sb{N},

1553: %\quad

1554: %\omega\in\C\sp{+}.

1555: %\end{equation}

1556: For $x\in[X\sb{J},X\sb{J+1}]$, $1\le J\le N-1$,

1557: the relation (\ref{tilde-psi-tilde-f}) implies that

1558: \begin{equation}\label{ast0}

1559: \tilde\varphi\sp{+}(x,\omega)

1560: =\tilde\varPhi\sb{J}\sp{+}(\omega)\cos(k(\omega)(x-X\sb{J}))

1561: +\tilde\varTheta\sb{J}\sp{+}(\omega)

1562: \frac{\sin(k(\omega)(x-X\sb{J}))}{k(\omega)},

1563: \qquad

1564: x\in[X\sb{J},X\sb{J+1}],

1565: \quad

1566: \omega\in\C\sp{+},

1567: \end{equation}

1568: where $\tilde\varPhi\sb{J}\sp{+}$ and $\tilde\varTheta\sb{J}\sp{+}$,

1569: $1\le J\le N-1$,

1570: are analytic functions of $\omega\in\C\sp{+}$.

1571: We note that,

1572: by (\ref{tilde-psi-tilde-f}),

1573: \begin{equation}\label{a-b}

1574: \tilde\varPhi\sb{J}\sp{+}(\omega)

1575: =\tilde\varphi\sp{+}(X\sb{J},\omega),

1576: \qquad

1577: \tilde\varTheta\sb{J}\sp{+}(\omega)

1578: =\p\sb{x}\tilde\varphi\sp{+}(X\sb{J}+0,\omega)

1579: =-\sum\sb{J'}

1580: \sgn(X\sb{J}-X\sb{J'})

1581: \tilde f\sb{J'}\sp{+}(\omega)\frac{e^{ik(\omega)\abs{X\sb{J}-X\sb{J'}}}}{2}.

1582: \end{equation}

1583: %Since

1584: %$\tilde\varphi(x,\omega)=\tilde\varphi\sp{+}(x,\omega)+\tilde\varphi\sp{-}(x,\omega)$,

1585: %we proved the representation (\ref{c-s}) for $\omega\in\C\sp{+}$

1586: 

1587: \noindent

1588: {\bf Step 2: Traces on real line.}\quad

1589: Now we need to extend the relations (\ref{phi-1}) and (\ref{ast0})

1590: to $\omega\in\R$.

1591: The Fourier transform

1592: $\hat\varphi\sp{+}(x,\omega)

1593: :=\mathcal{F}\sb{t\to\omega}[\theta(t)\varphi(x,t)]$

1594: is a tempered $H\sp{1}$-valued distribution of $\omega\in\R$

1595: by (\ref{psi-2-bounds}).

1596: It is the boundary value

1597: of the analytic function $\tilde\varphi\sp{+}(x,\omega)$,

1598: in the following sense:

1599: \begin{equation}\label{bvp1}

1600: \hat\varphi\sp{+}(x,\omega)

1601: =\lim\limits\sb{\varepsilon\to 0+}\tilde\varphi\sp{+}(x,\omega+i\varepsilon),

1602: \qquad\omega\in\R,

1603: \end{equation}

1604: where the convergence

1605: is in the space

1606: of tempered distributions

1607: $\mathscr{S}'(\R,H\sp{1}(\R))$.

1608: Indeed,

1609: \[

1610: \tilde\varphi\sp{+}(x,\omega+i\varepsilon)

1611: =\mathcal{F}\sb{t\to\omega}[\theta(t)\varphi(x,t)e\sp{-\varepsilon t}],

1612: \qquad

1613: \theta(t)\varphi(x,t)e\sp{-\varepsilon t}

1614: \mathop{\longrightarrow}\limits\sb{\varepsilon\to 0+}

1615: \theta(t)\varphi(x,t),

1616: \]

1617: where the convergence holds in $\mathscr{S}'(\R,H\sp{1}(\R))$.

1618: Therefore, (\ref{bvp1}) holds by the continuity of the Fourier transform

1619: $\mathcal{F}\sb{t\to\omega}$ in $\mathscr{S}'(\R)$.

1620: 

1621: %%AC CHANGES!

1622: %Similarly to (\ref{bvp1}),

1623: The distributions

1624: $\hat\varPhi\sb{J}\sp{+}(\omega)$,

1625: $\hat\varTheta\sb{J}\sp{+}(\omega)\in\mathscr{S}'(\R)$,

1626: $\omega\in\R$,

1627: are

1628: %%AC ADDED line:

1629: defined as

1630: the boundary values of the

1631: functions

1632: $\tilde\varPhi\sb{J}\sp{+}(\omega)$ and $\tilde\varTheta\sb{J}\sp{+}(\omega)$

1633: analytic in $\omega\in\C\sp{+}$:

1634: \begin{eqnarray}\label{phi-to-phi}

1635: &&

1636: \hat\varPhi\sb{J}\sp{+}(\omega)=\lim\limits\sb{\varepsilon\to 0+}

1637: \tilde\varPhi\sb{J}\sp{+}(\omega+i\varepsilon),

1638: \qquad

1639: \omega\in\R,

1640: \quad

1641: 0\le J\le N,

1642: \\

1643: &&

1644: \label{theta-to-theta}

1645: \hat\varTheta\sb{J}\sp{+}(\omega)

1646: =\lim\limits\sb{\varepsilon\to 0+}

1647: \tilde\varTheta\sb{J}\sp{+}(\omega+i\varepsilon),

1648: \qquad

1649: \omega\in\R,

1650: \quad

1651: 1\le J\le N-1.

1652: \end{eqnarray}

1653: The above convergence holds in the space of quasimeasures

1654: by (\ref{a-b}),

1655: since

1656: $\tilde\varphi\sp{+}(X\sb{J},\omega)$

1657: and $\tilde f\sb{J}\sp{+}(\omega)$

1658: are quasimeasures (see Remark~\ref{remark-quasimeasure})

1659: while the exponential factors

1660: %%AC ADDED line:

1661: in (\ref{a-b})

1662: are multiplicators

1663: in the space of quasimeasures \cite[Appendix B]{ubk-arma}.

1664: Therefore, the

1665: formulas (\ref{ast0}) with $1\le J\le N-1$

1666: imply, in the limit $\Im\omega\to 0+$, that

1667: \begin{equation}\label{ast1}

1668: \hat\varphi\sp{+}(x,\omega)

1669: =\hat\varPhi\sb{J}\sp{+}(\omega)\cos(k(\omega+i0)(x-X\sb{J}))

1670: +\hat\varTheta\sb{J}\sp{+}(\omega)\frac{\sin(k(\omega+i0)(x-X\sb{J}))}{k(\omega+i0)},

1671: \qquad

1672: x\in [X\sb{J},X\sb{J+1}],

1673: \quad

1674: \omega\in\R,

1675: \end{equation}

1676: since

1677: $\cos(k(\omega+i0)(x-X\sb{J}))$

1678: and

1679: $\frac{\sin(k(\omega+i0)(x-X\sb{J}))}{k(\omega+i0)}$

1680: are smooth functions of $\omega\in\R$.

1681: %(By continuity, at $\omega=\pm m$ where $k(\omega)=0$,

1682: %the ratio $\frac{\sin(k(\omega)(x-X\sb{J}))}{k(\omega)}$

1683: %is to be understood as $x-X\sb{J}$.)

1684: Similar representation holds for $\hat\varphi\sp{-}(x,\omega)$.

1685: Therefore, the representation (\ref{c-s}) follows

1686: for $X\sb{1}\le x\le X\sb{N}$.

1687: 

1688: The formula (\ref{c-s}) for $x\le X\sb{1}$

1689: follows from taking the limit

1690: $\Im\omega\to 0+$

1691: in the expression (\ref{phi-1})

1692: for $\tilde\varphi\sp{+}(x,\omega)$

1693: and the limit

1694: $\Im\omega\to 0-$

1695: in a similar expression

1696: for $\tilde\varphi\sp{-}(x,\omega)$:

1697: \begin{equation}\label{phi-1-m}

1698: \tilde\varphi\sp{-}(x,\omega)

1699: =-\sum\sb{J}

1700: \tilde f\sp{-}\sb{J}(\omega)\frac{e\sp{-i k(\omega)(x-X\sb{J})}}{2i k(\omega)}

1701: =e^{-i k(\omega)(x-X\sb{1})}\tilde\varphi\sp{-}(X\sb{1},\omega),

1702: \qquad

1703: x\le X\sb{1},

1704: \quad

1705: \omega\in\C\sp{-},

1706: \end{equation}

1707: and then taking the sum of the resulting expressions.

1708: This justifies (\ref{c-s}) for $x\le X\sb{1}$.

1709: Similarly we justify (\ref{c-s}) for $x\ge X\sb{N}$.

1710: \end{proof}

1711: 

1712: \section{Absolute continuity of the spectrum}

1713: \label{sect-ac}

1714: 

1715: \begin{lemma}\label{lemma-continuity-0}

1716: The distributions

1717: $\hat\varPhi\sb{1}\sp\pm(\omega)$, $\hat\varPhi\sb{N}\sp\pm(\omega)$

1718: are absolutely continuous for $\abs{\omega}>m$,

1719: and moreover

1720: \begin{equation}\label{phi-0-41}

1721: \int\sb{\abs{\omega}>m}

1722: \left[

1723: \abs{\hat\varPhi\sb{1}\sp\pm(\omega)}\sp 2

1724: +

1725: \abs{\hat\varPhi\sb{N}\sp\pm(\omega)}\sp 2

1726: \right]

1727: \omega\,k\sb{+}(\omega)

1728: \,d\omega<\infty,

1729: \end{equation}

1730: where $\omega\,k\sb{+}(\omega)\ge 0$ by {(\ref{omega-k})}.

1731: \end{lemma}

1732: 

1733: The bound for each of

1734: $\hat\varPhi\sb{1}\sp\pm(\omega)$,

1735: $\hat\varPhi\sb{N}\sp\pm(\omega)$

1736: is obtained verbatim by applying the proof of

1737: \cite[Proposition 3.3]{ubk-arma}.

1738: 

1739: \begin{proposition}\label{prop-continuity}

1740: The distributions

1741: $\hat\varPhi\sb{J}(\omega)$, $1\le J\le N$,

1742: and

1743: $\hat\varTheta\sb{J}(\omega)$, $1\le J\le N-1$,

1744: are absolutely continuous for

1745: $\abs{\omega}>\mu\sb{J}$ and $\abs{\omega}>(2p\sb{J}-1)\mu\sb{J}$,

1746: respectively,

1747: with $\mu\sb{J}$ defined in (\ref{def-mu}).

1748: Moreover,

1749: for any $\epsilon>0$,

1750: \begin{equation}\label{41}

1751: \int\sb{\abs{\omega}

1752: >\mu\sb{J}+\epsilon}

1753: \abs{\hat\varPhi\sb{J}(\omega)}\sp 2

1754: \omega^2

1755: \,d\omega<\infty,

1756: \quad

1757: 1\le J\le N;

1758: \qquad

1759: \int\sb{\abs{\omega}>(2p\sb{J}-1)\mu\sb{J}+\epsilon}

1760: \abs{\hat\varTheta\sb{J}(\omega)}\sp 2

1761: \,d\omega<\infty,

1762: \quad

1763: 1\le J\le N-1.

1764: \end{equation}

1765: \end{proposition}

1766: 

1767: \begin{proof}

1768: We will use induction,

1769: proving the absolute continuity of

1770: $\hat\varphi(X\sb{J},\omega)$

1771: and $\p\sb x\hat\varphi(X\sb{J}\pm 0,\omega)$

1772: starting with $J=1$ and going to $J=N$.

1773: By Lemma~\ref{lemma-41},

1774: $\hat\varphi(X\sb 1,\omega)

1775: =\hat\varPhi\sb{1}(\omega)

1776: =\hat\varPhi\sb{1}\sp{+}(\omega)+\hat\varPhi\sb{1}\sp{-}(\omega)$

1777: and

1778: $

1779: \p\sb x\hat\varphi(X\sb 1-0,\omega)=

1780: -i k\sb{+}(\omega)\hat\varPhi\sb{1}\sp{+}(\omega)

1781: -i k\sb{-}(\omega)\hat\varPhi\sb{1}\sp{-}(\omega).

1782: $

1783: Hence,

1784: Lemma~\ref{lemma-continuity-0}

1785: implies that,

1786: for any $\epsilon>0$,

1787: \begin{equation}\label{phi-phi-1}

1788: \int\sb{\abs{\omega}>m+\epsilon}

1789: \abs{\hat\varphi(X\sb{1},\omega)}\sp 2

1790: \omega^2

1791: \,d\omega<\infty,

1792: \qquad

1793: \int\sb{\abs{\omega}>m+\epsilon}

1794: \abs{\hat\varphi'(X\sb{1}-0,\omega)}\sp 2

1795: \,d\omega<\infty.

1796: \end{equation}

1797: Now assume that for some $1\le J<N$ and for any $\epsilon>0$

1798: we have:

1799: \begin{equation}\label{phi-phi-j}

1800: \int\sb{\abs{\omega}>\mu\sb{J}+\epsilon}

1801: \abs{\hat\varphi(X\sb{J},\omega)}\sp 2

1802: \omega^2

1803: \,d\omega<\infty,

1804: \qquad

1805: \int\sb{\abs{\omega}>\mu\sb{J}+\epsilon}

1806: \abs{\hat\varphi'(X\sb{J}-0,\omega)}\sp 2

1807: \,d\omega<\infty.

1808: \end{equation}

1809: Lemma~\ref{lemma-41}

1810: and equation (\ref{kg-no-cp-2-st})

1811: yield the jump condition

1812: \begin{equation}\label{jump-j}

1813: \hat\varTheta\sb{J}(\omega)

1814: =

1815: \hat\varphi'(X\sb{J}+0,\omega)

1816: =\hat\varphi'(X\sb{J}-0,\omega)

1817: -\hat f\sb{J}(\omega),

1818: \qquad\omega\in\R,

1819: \end{equation}

1820: where $f\sb{J}(t)=F\sb{J}(\psi(X\sb{J},t))$ by (\ref{def-f}).

1821: 

1822: \begin{lemma}\label{lemma-fj-ac}

1823: For any $\epsilon>0$

1824: the following inequality holds:

1825: \begin{equation}\label{int-f2}

1826: \int\sb{\abs{\omega}>(2p\sb{J}-1)(\mu\sb{J}+2\epsilon)}

1827: \abs{\hat f\sb{J}(\omega)}^2

1828: \,d\omega<\infty.

1829: \end{equation}

1830: \end{lemma}

1831: 

1832: \begin{proof}

1833: Let

1834: $\zeta\sb{J}(\omega)\in C\sp\infty\sb{0}(\R)$

1835: be such that

1836: $\zeta\sb{J}(\omega)\equiv 1$ for $\abs{\omega}\le\mu\sb{J}+\epsilon$

1837: and

1838: $\zeta\sb{J}(\omega) \equiv 0$ for $\abs{\omega}\ge\mu\sb{J}+2\epsilon$.

1839: We denote $\psi(X\sb{J},t)$ by $\uppsi\sb{J}(t)$,

1840: and split it into

1841: \begin{equation}\label{psi-psi-psi}

1842: \uppsi\sb{J}(t)

1843: =\uppsi\sb{J,b}(t)+\uppsi\sb{J,d}(t),

1844: \end{equation}

1845: where the functions in the right-hand side are defined

1846: by their Fourier transforms:

1847: \begin{equation}

1848: \hat\uppsi\sb{J,b}(\omega)

1849: =\zeta\sb{J}(\omega)\hat\uppsi\sb{J}(\omega)

1850: =\zeta\sb{J}(\omega)\hat\psi(X\sb{J},\omega),

1851: \qquad

1852: \hat\uppsi\sb{J,d}(\omega)

1853: =(1-\zeta\sb{J}(\omega))\hat\uppsi\sb{J}(\omega)

1854: =(1-\zeta\sb{J}(\omega))\hat\psi(X\sb{J},\omega).

1855: \end{equation}

1856: By Lemma~\ref{lemma-chi-h1}

1857: and by (\ref{phi-phi-j}),

1858: we have

1859: \begin{equation}\label{chi-varphi-h1}

1860: \int\sb{\R}

1861: \Abs{(1-\zeta\sb{J}(\omega))

1862: \hat\chi(X\sb{J},\omega)}^2\omega^2\,d\omega<\infty,

1863: \qquad

1864: \int\sb{\R}

1865: \Abs{(1-\zeta\sb{J}(\omega))

1866: \hat\varphi(X\sb{J},\omega)}^2\omega^2\,d\omega

1867: <\infty.

1868: \end{equation}

1869: Since

1870: $

1871: \hat\uppsi\sb{J,d}(\omega)

1872: =(1-\zeta\sb{J}(\omega))

1873: (\hat\chi(X\sb{J},\omega)+\hat\varphi(X\sb{J},\omega)),

1874: $

1875: we also have

1876: \[

1877: \int\sb{\R}

1878: \Abs{(1-\zeta\sb{J}(\omega))

1879: \hat\uppsi\sb{J}(\omega)}^2\omega^2\,d\omega

1880: <\infty,

1881: \]

1882: proving that

1883: \begin{equation}\label{psi-jd-h1}

1884: \uppsi\sb{J,d}(t)\in H\sp 1(\R).

1885: \end{equation}

1886: For

1887: $

1888: \hat f\sb{J}(\omega)

1889: =

1890: \mathcal{F}\sb{t\to\omega}[F\sb{J}(\uppsi\sb{J}(t))]

1891: =

1892: \mathcal{F}\sb{t\to\omega}[F\sb{J}(\psi(X\sb{J},t))]

1893: $,

1894: taking into account (\ref{f-is-such}) and (\ref{psi-psi-psi}),

1895: we have:

1896: \begin{eqnarray}

1897: \hat f\sb{J}(\omega)

1898: &=&-\sum\sb{n=1}\sp{p\sb{J}}

1899: 2n\,u\sb{J,n}

1900: \underbrace

1901: {

1902: (\hat\uppsi\sb{J}\ast\hat{\overline\uppsi}\sb{J})

1903: \ast\dots\ast

1904: (\hat\uppsi\sb{J}\ast\hat{\overline\uppsi}\sb{J})

1905: }\sb{n-1}

1906: \ast

1907: \hat\uppsi\sb{J}

1908: \nonumber

1909: \\

1910: &=&

1911: .\ .\ .\ .\ .\ \ -\sum\sb{n=1}\sp{p\sb{J}}

1912: 2n\,u\sb{J,n}

1913: \underbrace

1914: {

1915: (\hat\uppsi\sb{J,b}\ast\hat{\overline\uppsi}\sb{J,b})

1916: \ast\dots\ast

1917: (\hat\uppsi\sb{J,b}\ast\hat{\overline\uppsi}\sb{J,b})

1918: }\sb{n-1}

1919: \ast

1920: \hat\uppsi\sb{J,b},

1921: \label{conv-conv}

1922: \end{eqnarray}

1923: where the dots in the right-hand side

1924: denote the convolutions of

1925: $\hat\uppsi\sb{J,b}$, $\hat{\overline\uppsi}\sb{J,b}$,

1926: $\hat\uppsi\sb{J,d}$, and $\hat{\overline\uppsi}\sb{J,d}$

1927: that contain at least one of

1928: $\hat\uppsi\sb{J,d}$, $\hat{\overline\uppsi}\sb{J,d}$.

1929: Since

1930: $\uppsi\sb{J,b}(t)$, $\uppsi\sb{J,d}(t)$ are bounded

1931: while $\uppsi\sb{J,d}(t)\in H\sp 1(\R)$

1932: by (\ref{psi-jd-h1}),

1933: all these terms belong to $L\sp 2(\R)$.

1934: Finally, since

1935: $\supp\hat\uppsi\sb{J,b}

1936: \subset[-\mu\sb{J}-2\epsilon,\mu\sb{J}+2\epsilon]$,

1937: the convolutions

1938: under the summation sign

1939: in the right-hand side of (\ref{conv-conv})

1940: are supported inside

1941: $[-(2p\sb{J}-1)(\mu\sb{J}+2\epsilon),(2p\sb{J}-1)(\mu\sb{J}+2\epsilon)]$

1942: and do not contribute into the integral (\ref{int-f2}).

1943: \end{proof}

1944: 

1945: Using

1946: (\ref{phi-phi-j}) and Lemma~\ref{lemma-fj-ac}

1947: to estimate the norms of

1948: $\p\sb x\hat\varphi(X\sb{J}-0,\omega)$ and $\hat f\sb{J}(\omega)$

1949: in the right-hand side in the relation (\ref{jump-j}),

1950: we conclude that

1951: \begin{equation}\label{bj-ac}

1952: \int\limits\sb{\abs{\omega}>(2p\sb{J}-1)(\mu\sb{J}+2\epsilon)}

1953: \Abs{

1954: \hat\varphi'(X\sb{J}+0,\omega)}^2\,d\omega

1955: <\infty.

1956: \end{equation}

1957: Now the inequalities

1958: \begin{equation}\label{phi-phi-jj}

1959: \int\limits\sb{\abs{\omega}>(2p\sb{J}-1)(\mu\sb{J}+2\epsilon)}

1960: \abs{\hat\varphi(X\sb{J+1},\omega)}^2

1961: \omega^2\,d\omega<\infty,

1962: \qquad

1963: \int\limits\sb{\abs{\omega}>(2p\sb{J}-1)(\mu\sb{J}+2\epsilon)}

1964: \abs{\hat\varphi'(X\sb{J+1}-0,\omega)}^2

1965: \,d\omega<\infty

1966: \end{equation}

1967: follow

1968: from the representation (\ref{c-s}) for $x\in[X\sb{J},X\sb{J+1}]$,

1969: where we apply the first inequality from (\ref{phi-phi-j})

1970: and the inequality (\ref{bj-ac}).

1971: Therefore, starting with (\ref{phi-phi-1}),

1972: one shows by induction that (\ref{phi-phi-j}) holds for all $1\le J\le N$.

1973: The estimates on

1974: $\hat\varPhi\sb{J}(\omega)=\hat\varphi(X\sb{J},\omega)$

1975: and 

1976: $\hat\varTheta\sb{J}(\omega)=\hat\varphi'(X\sb{J}+0,\omega)$

1977: stated in the Proposition

1978: follow from (\ref{phi-phi-j}) and (\ref{bj-ac}), respectively.

1979: This finishes the proof of Proposition~\ref{prop-continuity}.

1980: \end{proof}

1981: 

1982: \begin{corollary}\label{corollary-continuity}

1983: The distributions

1984: $\hat\varPhi\sb{J}(\omega)=\hat\varphi(X\sb J,\omega)$,

1985: $1\le J\le N$,

1986: are absolutely continuous for

1987: $\abs{\omega}>{M}\sb{J}$,

1988: while $\hat\varTheta\sb{J}(\omega)=\p\sb x\hat\varphi(X\sb J+0,\omega)$,

1989: $1\le J\le N-1$,

1990: are absolutely continuous for $\abs{\omega}>(2p\sb{J}-1){M}\sb{J}$,

1991: where ${M}\sb{J}:=\min(\mu\sb{J},\mu'\sb{J})$

1992: is defined in (\ref{def-m-lambda}).

1993: \end{corollary}

1994: 

1995: \begin{proof}

1996: In the proof of Proposition~\ref{prop-continuity},

1997: we could as well proceed from $J=N$ to $J=1$,

1998: proving the result stated in the Corollary.

1999: \end{proof}

2000: 

2001: \section{Compactness}

2002: \label{sect-bound}

2003: 

2004: \subsection*{Second dispersive component}

2005: Let $\zeta(\omega)\in C\sp\infty\sb{0}(\R)$ be such that

2006: $\zeta(\omega)\equiv 1$ for $\abs{\omega}<\Lambda$,

2007: where $\Lambda$ is from (\ref{def-m-lambda}).

2008: Define $\varphi\sb{d}(x,t)$

2009: by its Fourier transform:

2010: \begin{equation}\label{dd}

2011: \hat\varphi\sb{d}(x,\omega)

2012: :=

2013: (1-\zeta(\omega))\hat\varphi(x,\omega)

2014: \ \ x\in\R,\ \ \omega\in\R.

2015: \end{equation}

2016: 

2017: \begin{lemma}\label{lemma-phi-d-b}

2018: $\varphi\sb{d}(x,t)$ is a bounded continuous

2019: function

2020: of $t\in\R$

2021: with values in $H\sp 1(\R)$:

2022: \begin{equation}\label{phi-d-b}

2023: \varphi\sb{d}(x,t)\in C\sb{b}(\R,H\sp{1}(\R)).

2024: \end{equation}

2025: The local energy decay holds for $\varphi\sb{d}(x,t)$:

2026: \begin{equation}\label{led}

2027: \lim\sb{t\to\infty}

2028: \norm{(\varphi\sb{d},\dot\varphi\sb{d})}\sb{\E,R}=0,

2029: \qquad\forall R>0.

2030: \end{equation}

2031: \end{lemma}

2032: 

2033: \begin{proof}

2034: We generalize the proof of \cite[Proposition 3.6]{ubk-arma}.

2035: By Lemma~\ref{lemma-41},

2036: \begin{equation}\label{ftb00}

2037: \hat\varphi\sb{d}(x,\omega)

2038: =

2039: \left\{

2040: \begin{array}{l}

2041: (1-\zeta(\omega))

2042: \left[

2043: \hat\varPhi\sb{1}\sp{+}(\omega)e\sp{-ik\sb{+}(\omega)(x-X\sb{1})}

2044: +

2045: \hat\varPhi\sb{1}\sp{-}(\omega)e\sp{-ik\sb{-}(\omega)(x-X\sb{1})}

2046: \right]

2047: ,

2048: \qquad

2049: x\le X\sb{1},

2050: \\

2051: (1-\zeta(\omega))\hat\varPhi\sb{J}(\omega)\cos(k\sb{+}(\omega)(x-X\sb{J}))

2052: +

2053: (1-\zeta(\omega))

2054: \hat\varTheta\sb{J}(\omega)

2055: \frac{\sin(k\sb{+}(\omega)(x-X\sb{J}))}{k\sb{+}(\omega)(x-X\sb{J})},

2056: \qquad

2057: x\in[X\sb{J},X\sb{J+1}],

2058: \\

2059: (1-\zeta(\omega))

2060: \left[

2061: \hat\varPhi\sb{N}\sp{+}(\omega)e\sp{ik\sb{+}(\omega)(x-X\sb{N})}

2062: +

2063: \hat\varPhi\sb{N}\sp{-}(\omega)e\sp{ik\sb{-}(\omega)(x-X\sb{N})}

2064: \right],

2065: \qquad

2066: x\ge X\sb{N}.

2067: \end{array}

2068: \right.

2069: \end{equation}

2070: Each of the functions entering the above expression,

2071: considered on the whole real line,

2072: corresponds to a finite energy solution to a linear

2073: Klein-Gordon equation,

2074: satisfying the properties stated in the lemma.

2075: For example,

2076: define $u(x,t)$ by its Fourier transform:

2077: \[

2078: \hat u(x,\omega)

2079: :=

2080: (1-\zeta(\omega))\hat\varPhi\sb{1}(\omega)

2081: \cos(k\sb{+}(\omega)(x-X\sb{1})),

2082: \qquad

2083: x\in\R.

2084: \]

2085: Then $u(x,t)$ is a solution to a linear Klein-Gordon equation,

2086: and, by Proposition~\ref{prop-continuity},

2087: the corresponding initial data are of finite energy:

2088: \[

2089: (u(x,0),\dot u(x,0))\in\E.

2090: \]

2091: Hence

2092: $u(x,t)\in C\sb{b}(\R,H\sp 1(\R))$

2093: and satisfies the local energy decay of the form (\ref{led})

2094: (see \cite[Lemma 3.1]{ubk-arma}.

2095: This finishes the proof.

2096: \end{proof}

2097: 

2098: \subsection*{Compactness for the bound component}

2099: 

2100: We introduce the bound component of $\varphi(x,t)$

2101: by

2102: \begin{equation}\label{bb}

2103: \varphi\sb{b}(x,t)

2104: =\varphi(x,t)-\varphi\sb{d}(x,t)

2105: =\psi(x,t)-\chi(x,t)-\varphi\sb{d}(x,t),

2106: \qquad

2107: x\in\R,\ \ t\in\R.

2108: \end{equation}

2109: By Lemma~\ref{lemma-phi-d-b},

2110: \begin{equation}\label{ebb}

2111: \varphi\sb{b}(x,t)\in C\sb{b}(\R,H\sp{1}(\R)).

2112: \end{equation}

2113: Lemma~\ref{lemma-41} and (\ref{dd}), (\ref{bb})

2114: imply the multiplicative relation

2115: \begin{equation}\label{ftb0}

2116: \hat\varphi\sb{b}(x,\omega)

2117: =

2118: \left\{

2119: \begin{array}{l}

2120: \zeta(\omega)

2121: \big[

2122: \hat\varPhi\sb{1}\sp{+}(\omega)e\sp{-i k\sb{+}(\omega)(x-X\sb{1})}

2123: +

2124: \hat\varPhi\sb{1}\sp{-}(\omega)e\sp{-i k\sb{-}(\omega)(x-X\sb{1})}

2125: \big],

2126: \qquad

2127: x\le X\sb{1},

2128: \\

2129: \zeta(\omega)

2130: \big[\hat\varPhi\sb{J}(\omega)\cos(k\sb{+}(\omega)(x-X\sb{J}))

2131: +

2132: \hat\varTheta\sb{J}(\omega)

2133: \frac{\sin(k\sb{+}(\omega)(x-X\sb{J}))}{k\sb{+}(\omega)}

2134: \big],

2135: \qquad

2136: x\in[X\sb{J},X\sb{J+1}],

2137: \\

2138: \zeta(\omega)

2139: \big[

2140: \hat\varPhi\sb{N}\sp{+}(\omega)e\sp{i k\sb{+}(\omega)(x-X\sb{N})}

2141: +

2142: \hat\varPhi\sb{N}\sp{-}(\omega)e\sp{i k\sb{-}(\omega)(x-X\sb{N})}

2143: \big],

2144: \qquad

2145: x\ge X\sb{N}.

2146: \end{array}

2147: \right.

2148: \end{equation}

2149: By (\ref{ebb}), the functions

2150: \[

2151: \varphi\sb{b,J}(t)

2152: :=

2153: \varphi\sb{b}(X\sb{J},t)

2154: =\varphi(X\sb{J},t)-\varphi\sb{d}(X\sb{J},t)

2155: \]

2156: are bounded and continuous.

2157: Therefore,

2158: $

2159: \hat\varphi\sb{b}(X\sb{J},\cdot)\in\mathscr{S}'(\R)

2160: $

2161: are quasimeasures (see Remark~\ref{remark-quasimeasure}).

2162: 

2163: \begin{proposition}

2164: \begin{enumerate}

2165: \item

2166: The function $\varphi\sb{b}(x,t)$ is smooth

2167: for $x\in\R\backslash\mathcal{X}$

2168: (where $\mathcal{X}=\{X\sb 1,\,X\sb 2,\,\dots,\,X\sb{N}\}$)

2169: and $t\in\R$.

2170: \item

2171: For any $R>0$,

2172: \begin{equation}\label{bqda}

2173: \sup\limits\sb{

2174: \abs{x}\le R,\,x\notin\mathcal{X}}

2175: \,\,

2176: \sup\limits\sb{t\in\R}

2177: \abs{\p\sb x\sp\xx\p\sb t\sp\yy\varphi\sb{b}(x,t)}

2178: <\infty.

2179: \end{equation}

2180: \end{enumerate}

2181: \end{proposition}

2182: 

2183: The argument repeats

2184: the proof of Proposition~\cite[Proposition 4.1]{ubk-arma}.

2185: 

2186: \begin{remark}

2187: Let us note that

2188: the bounds (\ref{bqda}) are independent of $x$

2189: and remain valid for

2190: $x\notin\mathcal{X}$,

2191: although the derivatives

2192: $\p\sb x\sp\xx\p\sb t\sp\yy\varphi\sb{b}(x,t)$

2193: with $\xx\ne 0$

2194: may have jumps at $x=X\sb{J}$.

2195: (Note that

2196: this is the case for the solitary waves in (\ref{solitary-wave-profile}).)

2197: \end{remark}

2198: 

2199: We now may deduce the compactness of the set of

2200: translations of the bound component,

2201: $\{\varphi\sb{b}(x,s+t)\sothat s\ge 0\}$.

2202: 

2203: \begin{corollary}\label{coco}

2204: \begin{enumerate}

2205: \item

2206: By the Ascoli-Arzel\`a Theorem,

2207: for any sequence $s\sb{j}\to\infty$

2208: there exists a subsequence $s\sb{j'}\to\infty$

2209: such that

2210: \begin{equation}

2211: \varphi\sb{b}(x,s\sb{j'}+t)

2212: \to

2213: \beta(x,t),

2214: \qquad

2215: x\in\R,

2216: \quad

2217: t\in\R,

2218: \label{ol}

2219: \end{equation}

2220: and also for any nonnegative integers $\xx$ and $\yy$,

2221: \begin{equation}

2222: \p\sb x\sp\xx\p\sb t\sp\yy

2223: \varphi\sb{b}(x,s\sb{j'}+t)

2224: \to

2225: \p\sb x\sp\xx\p\sb t\sp\yy

2226: \beta(x,t),

2227: \qquad

2228: x\notin\mathcal{X},

2229: \quad

2230: t\in\R,

2231: \label{olpd}

2232: \end{equation}

2233: for some

2234: $\beta(x,t)\in C\sb{b}(\R,H\sp{1}(\R))$.

2235: The convergence in (\ref{ol}) and (\ref{olpd})

2236: is uniform in $x$ and $t$

2237: as long as $\abs{x}+\abs{t}\le R$, for any $R>0$.

2238: The convergence in (\ref{olpd})

2239: also holds for $x=X\sb{J}\pm 0$.

2240: \item

2241: By the Fatou Lemma,

2242: \begin{equation}\label{beta-beta}

2243: \sup\limits\sb{t\in\R}\norm{\beta(\cdot,t)}\sb{H\sp{1}}<\infty.

2244: \end{equation}

2245: \end{enumerate}

2246: \end{corollary}

2247: 

2248: We call {\it omega-limit trajectory}

2249: any function $\beta(x,t)$

2250: that can appear as a limit in (\ref{ol}), (\ref{olpd}).

2251: 

2252: \begin{remark}\label{remark-end}

2253: Previous analysis demonstrates that the long-time asymptotics

2254: of the solution $\psi(x,t)$ in $\E\sb{F}$

2255: depends only on the singular component

2256: $\varphi(x,t)$.

2257: Due to Corollary~\ref{coco},

2258: to conclude the proof of Theorem~\ref{main-theorem},

2259: it suffices to check that every omega-limit trajectory

2260: belongs to the set of solitary waves;

2261: that is,

2262: \begin{equation}\label{eidd}

2263: \beta(x,t)

2264: =

2265: \phi\sb{\omega\sb{+}}(x)e\sp{-i\omega\sb{+}t}

2266: \qquad

2267: {\rm for\ some\ }\omega\sb{+}\in[-m,m].

2268: \end{equation}

2269: \end{remark}

2270: 

2271: 

2272: \section{Nonlinear spectral analysis}

2273: \label{sect-nonlinear-analysis}

2274: \subsection*{Bounds for the spectrum}

2275: 

2276: By Lemmas~\ref{lemma-decay-psi1} and \ref{lemma-phi-d-b},

2277: the dispersive components $\chi(\cdot,t)$

2278: and

2279: $\varphi\sb{d}(\cdot,t)$ converge

2280: to zero in

2281: ${\E}\sb{F}$

2282: as $t\to\infty$.

2283: On the other hand,

2284: by Corollary~\ref{coco},

2285: the bound component

2286: $\varphi\sb{b}(x,t+s\sb{j'})$

2287: converges to $\beta(x,t)$

2288: as $j'\to\infty$,

2289: uniformly in every compact set of the plane $\R^2$.

2290: Hence,

2291: $\psi(x,t+s\sb{j'})=

2292: \varphi\sb{b}(x,t+s\sb{j'})+\chi(x,t+s\sb{j'})+\varphi\sb{d}(x,t+s\sb{j'})$

2293: also

2294: converges to $\beta(x,t)$

2295: uniformly in every compact set of the plane $\R^2$.

2296: Therefore, taking the limit in equation (\ref{kg-no}),

2297: we conclude that

2298: the omega-limit trajectory $\beta(x,t)$ also satisfies

2299: the same equation:

2300: \begin{equation}\label{kg-no-beta}

2301: \ddot\beta(x,t)

2302: =\beta''(x,t)-m^2\beta(x,t)+\sum\sb{J}\delta(x-X\sb{J})F\sb{J}(\beta).

2303: \end{equation}

2304: 

2305: \begin{remark}

2306: Note that the bound component

2307: $\varphi\sb{b}(x,t)$ itself generally does not satisfy equation (\ref{kg-no-beta}).

2308: \end{remark}

2309: 

2310: Taking the Fourier transform of $\beta$ in time,

2311: we see by (\ref{olpd})

2312: that $\hat\beta(x,\omega)$

2313: is a continuous function of $x\in\R$,

2314: smooth for $x\in\R\backslash\mathcal{X}$,

2315: with values in tempered distributions of $\omega\in\R$,

2316: and that it satisfies the corresponding stationary equation

2317: \begin{equation}\label{kg-no-beta-1}

2318: -\omega^2\hat\beta(x,\omega)

2319: =\hat\beta''(x,\omega)-m^2\hat\beta(x,\omega)

2320: +\sum\sb{J}\delta(x-X\sb{J})\hat g\sb{J}(\omega),

2321: \qquad

2322: (x,\omega)\in\R^2,

2323: \end{equation}

2324: valid in the sense of tempered distributions of $(x,\omega)\in\R^2$,

2325: where

2326: $\hat g\sb{J}(\omega)$

2327: are the Fourier transforms of the functions

2328: \begin{equation}

2329: g\sb{J}(t):=F\sb{J}(\beta(X\sb{J},t)),

2330: \qquad

2331: 1\le J\le N.

2332: \end{equation}

2333: We also denote

2334: \begin{equation}\label{def-upbeta}

2335: \upbeta\sb{J}(t):=\beta(X\sb{J},t),

2336: \qquad

2337: \Sigma\sb{J}:=\supp\hat\upbeta\sb{J},

2338: \qquad 1\le J\le N.

2339: \end{equation}

2340: 

2341: From (\ref{ftb0}),

2342: we know that the spectrum of $\varphi\sb{b}(x,t)$

2343: is bounded for all $x\in\R$.

2344: Hence,

2345: the convergence

2346: (\ref{olpd}) implies that

2347: the spectrum of $\beta(x,t)$ is also bounded.

2348: We will need more precise bounds on the size

2349: of the spectrum of $\beta$:

2350: 

2351: \begin{lemma}\label{lemma-m-m}

2352: \begin{enumerate}

2353: \item

2354: $\Sigma\sb{J}

2355: :=\supp\hat\upbeta\sb{J}

2356: \subset[-{M}\sb{J},{M}\sb{J}]$,

2357: \quad

2358: $1\le J\le N$;

2359: \item

2360: $\supp\hat\beta'(X\sb{J}+0,\omega)

2361: \subset[-(2p\sb{J}-1){M}\sb{J},(2p\sb{J}-1){M}\sb{J}]$,

2362: \quad

2363: $1\le J\le N-1$,

2364: with ${M}\sb{J}>0$ defined in (\ref{def-m-lambda}).

2365: \end{enumerate}

2366: 

2367: \end{lemma}

2368: 

2369: \begin{proof}

2370: We have the relation

2371: \[

2372: \varphi\sb{b}(x,s\sb{j}+t)

2373: =

2374: \frac{1}{2\pi}\int\sb{\R}

2375: e^{-i\omega t}

2376: e^{-i \omega s\sb{j}}

2377: \hat\varphi\sb{b}(x,\omega)\,d\omega,

2378: \qquad

2379: x\in\R,\quad t\in\R,

2380: \]

2381: where the integral is understood as the pairing

2382: of a smooth function (oscillating exponent) with a compactly supported

2383: distribution.

2384: Then the convergence (\ref{ol}) implies that

2385: \begin{equation}\label{phi-to-beta}

2386: e^{-i\omega s\sb{j'}}

2387: \hat\varphi\sb{b}(x,\omega)

2388: \to\hat\beta(x,\omega),

2389: \qquad

2390: x\in\R,

2391: \quad s\sb{j'}\to\infty,

2392: \end{equation}

2393: in the sense of quasimeasures.

2394: Since $\hat\varphi\sb{b}(X\sb J,\omega)$

2395: is locally $L\sp 2$ for $\abs{\omega}>{M}\sb{J}$

2396: by Corollary~\ref{corollary-continuity},

2397: the convergence (\ref{phi-to-beta})

2398: at $x=X\sb J$

2399: shows that

2400: $\hat\upbeta\sb{J}(\omega):=\hat\beta(X\sb{J},\omega)$

2401: vanishes for $\abs{\omega}>{M}\sb{J}$.

2402: This proves the first statement of the lemma.

2403: 

2404: The second statement is proved similarly.

2405: Namely, the convergence (\ref{olpd}) implies that

2406: \begin{equation}\label{phi-to-beta-pd}

2407: e^{-i\omega s\sb{j'}}

2408: \p\sb x\hat\varphi\sb{b}(X\sb{J}+0,\omega)

2409: \to\p\sb x\hat\beta(X\sb{J}+0,\omega),

2410: \qquad

2411: %x\in\R\backslash\mathcal{X}, \quad

2412: s\sb{j'}\to\infty,

2413: \end{equation}

2414: in the sense of quasimeasures.

2415: Since

2416: $\hat\varphi\sb{b}'(X\sb{J}+0,\omega)$

2417: is locally $L\sp 2$ for $\abs{\omega}>(2p\sb{J}-1){M}\sb{J}$

2418: by Corollary~\ref{corollary-continuity},

2419: the convergence (\ref{phi-to-beta-pd})

2420: shows that

2421: $\hat\beta'(X\sb{J}+0,\omega)$

2422: vanishes for $\abs{\omega}>(2p\sb{J}-1){M}\sb{J}$.

2423: \end{proof}

2424: 

2425: We denote

2426: \begin{equation}\label{def-kappa}

2427: \kappa(\omega):=-i k\sb{+}(\omega),

2428: \qquad

2429: \omega\in\R,

2430: \end{equation}

2431: where $k\sb{+}(\omega)$ was introduced in (\ref{def-k-plus}).

2432: We then have

2433: $\Re\kappa(\omega)\ge 0$,

2434: and also

2435: \[

2436: \kappa(\omega)=\sqrt{\omega^2-m^2}>0\quad{\rm for}\quad -m<\omega<m,

2437: \]

2438: in accordance  with

2439: (\ref{solitary-wave-profile}).

2440: 

2441: \begin{proposition}\label{prop-beta}

2442: The distribution

2443: $\hat\beta(x,\omega)$ admits the following representation:

2444: \begin{equation}\label{beta-beta-c}

2445: \hat\beta(x,\omega)

2446: =

2447: \left\{

2448: \begin{array}{l}

2449: \hat\upbeta\sb{1}(\omega)

2450: e^{\kappa(\omega)(x-X\sb{1})},

2451: \qquad

2452: x\le X\sb{1},

2453: \\

2454: \hat\upbeta\sb{J}(\omega){\cosh(\kappa(\omega)(x-X\sb{J}))}+

2455: \hat\beta'(X\sb{J}+0,\omega)

2456: \frac{\sinh(\kappa(\omega)(x-X\sb{J}))}{\kappa(\omega)},

2457: \qquad

2458: x\in[X\sb{J},X\sb{J+1}],

2459: \quad

2460: 1\le J\le N-1,

2461: \\

2462: \hat\upbeta\sb{N}(\omega)

2463: e^{-\kappa(\omega)(x-X\sb{N})},

2464: \qquad

2465: x\ge X\sb{N}.

2466: \end{array}

2467: \right.

2468: \end{equation}

2469: \end{proposition}

2470: 

2471: \begin{proof}

2472: By (\ref{phi-to-beta}),

2473: the middle line in (\ref{beta-beta-c})

2474: follows from the representation (\ref{c-s})

2475: since the multiplicators are smooth bounded functions of $\omega\in\R$.

2476: Taking the limit in the first line of (\ref{c-s}),

2477: we obtain the first line in (\ref{beta-beta-c})

2478: since

2479: $\Sigma\sb{1}\subset[-m,m]$

2480: by Lemma~\ref{lemma-m-m},

2481: while

2482: %%AC removed \pm subscript of \kappa(\omega)

2483: $k\sb{+}(\omega)=k\sb{-}(\omega)=i\kappa(\omega)$

2484: for $-m\le\omega\le m$ (Cf. (\ref{k-plus-minus}), (\ref{def-kappa})).

2485: Similarly we explain the last line in (\ref{beta-beta-c}).

2486: \end{proof}

2487: 

2488: \subsection*{Reduction to point spectrum}

2489: 

2490: 

2491: \begin{proposition}\label{prop-one-omega}

2492: Any omega-limit trajectory $\beta(x,t)$ 

2493: is a solitary wave:

2494: \[

2495: \beta(x,t)=\phi(x)e^{-i\omega\sb{+}t}

2496: \quad{\rm with}

2497: \quad

2498: \omega\sb{+}\in[-m,m]

2499: \quad

2500: {\rm and}

2501: \quad

2502: \phi(x)\in H^1(\R).

2503: \]

2504: \end{proposition}

2505: 

2506: \begin{proof}

2507: The proof is based on the following lemmas.

2508: 

2509: \begin{lemma}

2510: If $\Sigma\sb{1}=\emptyset$,

2511: then $\beta(x,t)\equiv 0$.

2512: \end{lemma}

2513: 

2514: \begin{proof}

2515: According to equation (\ref{kg-no-beta-1}),

2516: the function

2517: %$\hat\beta\in C(\R,\mathscr{S}'(\R))$

2518: %%AC C is not enough; we need smooth away from X.

2519: %%AC to be able to write \hat\beta'(X\sb{1}-0,...), etc.

2520: $\hat\beta$

2521: satisfies the following continuity and jump conditions

2522: at the point $X\sb{1}$:

2523: \begin{equation}\label{jump-condition}

2524: \hat\beta(X\sb{1}+0,\omega)=\hat\beta(X\sb{1}-0,\omega)

2525: =\hat\upbeta\sb{1}(\omega),

2526: \qquad

2527: \hat\beta'(X\sb{1}+0,\omega)=\hat\beta'(X\sb{1}-0,\omega)

2528: +

2529: \hat g\sb{1}(\omega),

2530: \qquad

2531: \omega\in\R.

2532: \end{equation}

2533: $\Sigma\sb{1}=\emptyset$

2534: means that $\hat\upbeta\sb{1}(\omega)\equiv 0$,

2535: that is, $\upbeta\sb{1}(t)\equiv 0$.

2536: Hence,

2537: $g\sb{1}(t)\equiv F\sb{1}(\upbeta\sb{1}(t))\equiv 0$,

2538: and $\hat g\sb{1}(\omega)\equiv 0$.

2539: On the other hand, the first line of (\ref{beta-beta-c})

2540: implies that

2541: $\hat\beta(x,\omega)\equiv 0$ for $x\le X\sb{1}$,

2542: and in particular

2543: $\hat\beta'(X\sb{1}-0,\omega)\equiv 0$.

2544: Therefore, the jump condition (\ref{jump-condition}) implies that

2545: $\hat\beta'(X\sb{1}+0,\omega)\equiv 0$.

2546: Hence,

2547: $\hat\beta(x,\omega)\equiv 0$ for $x\in[X\sb{1},X\sb{2}]$

2548: by the middle line of (\ref{beta-beta-c}).

2549: By induction, 

2550: $\hat\upbeta\sb{J}(x,\omega)\equiv 0$.

2551: \end{proof}

2552: 

2553: Now we consider the case $\Sigma\sb{1}\not=\emptyset$.

2554: 

2555: \begin{lemma}

2556: If $\Sigma\sb{1}\not=\emptyset$,

2557: then $\Sigma\sb{1}=\{\omega\sb{+}\}$

2558: for some $\omega\sb{+}\in[-m,m]$.

2559: \end{lemma}

2560: 

2561: \begin{proof}

2562: By Lemma~\ref{lemma-m-m},

2563: we know that $\Sigma\sb{1}\subset[-m,m]$.

2564: To show that $\Sigma\sb{1}$ consists of a single point,

2565: we assume that, on the contrary,

2566: $\inf\Sigma\sb{1}<\sup\Sigma\sb{1}$.

2567: By (\ref{f-is-such}),

2568: the Fourier transform

2569: $\hat g\sb{1}(\omega)$

2570: of $g\sb 1(t):=F\sb{1}(\beta(X\sb{1},t))$

2571: is given by

2572: \begin{equation}\label{conv}

2573: \hat g\sb{1}

2574: =

2575: -\sum\sb{n=1}\sp{p\sb{1}}

2576: 2n\,u\sb{1,n}

2577: \underbrace{

2578: (\hat\upbeta\sb{1}\ast\hat{\overline\upbeta}\sb{1})

2579: \ast\;\dots\;\ast

2580: (\hat\upbeta\sb{1}\ast\hat{\overline\upbeta}\sb{1})

2581: }\sb{n-1}

2582: \ast\hat\upbeta\sb{1}.

2583: \end{equation}

2584: Applying the Titchmarsh Convolution Theorem

2585: \cite{titchmarsh} (see also \cite[p.119]{MR1400006}

2586: and \cite[Theorem 4.3.3]{MR1065136})

2587: to the convolutions in (\ref{conv}),

2588: we obtain the following equalities:

2589: \begin{eqnarray}

2590: \label{supp-g}

2591: &&

2592: \inf\supp\hat g\sb{1}

2593: =\inf\supp\hat\upbeta\sb{1}+(p\sb{1}-1)

2594: \inf\supp(\hat\upbeta\sb{1}\ast\hat{\overline\upbeta}\sb{1})

2595: =\inf\Sigma\sb{1}+(p\sb{1}-1)(\inf\Sigma\sb{1}-\sup\Sigma\sb{1}),

2596: \\

2597: &&

2598: \sup\supp\hat g\sb{1}

2599: =\sup\supp\hat\upbeta\sb{1}+(p\sb{1}-1)

2600: \sup\supp(\hat\upbeta\sb{1}\ast\hat{\overline\upbeta}\sb{1})

2601: =\sup\Sigma\sb{1}+(p\sb{1}-1)(\sup\Sigma\sb{1}-\inf\Sigma\sb{1}),

2602: \label{supp-gg}

2603: \end{eqnarray}

2604: where we used the relations

2605: $

2606: \inf\supp\hat{\overline\upbeta}\sb 1=-\sup\supp\hat\upbeta\sb 1,

2607: $

2608: $

2609: \sup\supp\hat{\overline\upbeta}\sb 1=-\inf\supp\hat\upbeta\sb 1.

2610: $

2611: Note that the Titchmarsh theorem is applicable

2612: since $\supp\hat\upbeta\sb{1}$ is compact

2613: by Lemma~\ref{lemma-m-m}.

2614: Since we assumed that $\inf\Sigma\sb{1}<\sup\Sigma\sb{1}$,

2615: (\ref{supp-g}) and (\ref{supp-gg})

2616: imply that

2617: $

2618: \inf\supp \hat g\sb{1}

2619: <\inf\Sigma\sb{1},

2620: $

2621: $

2622: \sup\supp \hat g\sb{1}

2623: >\sup\Sigma\sb{1}.

2624: $

2625: Therefore, the jump condition (\ref{jump-condition})

2626: with $J=1$

2627: implies that

2628: \begin{equation}\label{imp}

2629: \inf\supp \hat\beta'(X\sb{1}+0,\cdot)

2630: =

2631: \inf\supp\hat g\sb{1}

2632: <\inf\Sigma\sb{1},

2633: \qquad

2634: \sup\supp \hat\beta'(X\sb{1}+0,\cdot)=

2635: \sup\supp\hat g\sb{1}

2636: >

2637: \sup\Sigma\sb{1}.

2638: \end{equation}

2639: The ratio

2640: $

2641: \sinh(\kappa(\omega)(X\sb{2}-X\sb{1}))/\kappa(\omega)

2642: $

2643: could only vanish at the points

2644: $\omega=\pm\omega\sb{1,n}$,

2645: where

2646: \[

2647: \omega\sb{J,n}

2648: :=\sqrt{\frac{\pi^2 n^2}

2649: {\abs{X\sb{J+1}-X\sb{J}}^2}+m^2},

2650: \qquad

2651: 1\le J\le N-1,

2652: \quad

2653: n\in\N.

2654: \]

2655: %Let us introduce the set of frequencies

2656: %which correspond to the ``trapped modes'',

2657: %vanishing at the endpoints of the interval $[X\sb{J},X\sb{J+1}]$:

2658: %\begin{equation}\label{def-W}

2659: %\mathcal{W}=\{\pm\omega\sb{J,n}

2660: %\sothat 1\le J\le N-1,\quad n\in\N\},

2661: %\qquad{\rm where}\quad

2662: %\omega\sb{J,n}

2663: %:=\sqrt{\frac{\pi^2 n^2}

2664: %{\abs{X\sb{J+1}-X\sb{J}}^2}+m^2}.

2665: %\end{equation}

2666: %Under Assumption~\ref{ass-delta-small},

2667: %all trapped modes have frequencies larger than $\Lambda$:

2668: %\begin{equation}\label{WW}

2669: %[-\Lambda,\Lambda]\cap\mathcal{W}=\emptyset.

2670: %\end{equation}

2671: Due to Assumption~\ref{ass-delta-small} and Lemma~\ref{lemma-m-m},

2672: $\supp\hat\beta'(X\sb{1}+0,\omega)\cap\{\pm\omega\sb{1,n}\sothat n\in\N\}

2673: =\emptyset$.

2674: Hence, the middle line of (\ref{beta-beta-c})

2675: at $x=X\sb{2}-0$

2676: and the inequalities (\ref{imp})

2677: imply that

2678: \begin{equation}\label{sigma2-sigma1}

2679: \inf\Sigma\sb{2}

2680: = \inf\supp\hat g\sb{1}

2681: <\inf\Sigma\sb{1},

2682: \qquad

2683: \sup\Sigma\sb{2}

2684: = \sup\supp\hat g\sb{1}>\sup\Sigma\sb{1}.

2685: \end{equation}

2686: We proceed by induction, proving that

2687: \begin{equation}\label{induction}

2688: \inf\Sigma\sb{1}>\inf\Sigma\sb{2}>\dots>\inf\Sigma\sb{N},

2689: \qquad

2690: \sup\Sigma\sb{1}<\sup\Sigma\sb{2}<\dots<\sup\Sigma\sb{N}.

2691: \end{equation}

2692: It then follows that $\inf\Sigma\sb{N}<\sup\Sigma\sb{N}$.

2693: Starting from $J=N$ and going to the left,

2694: we also prove the opposite inequalities:

2695: \begin{equation}\label{induction-2}

2696: \inf\Sigma\sb{1}<\inf\Sigma\sb{2}<\dots<\inf\Sigma\sb{N},

2697: \qquad

2698: \sup\Sigma\sb{1}>\sup\Sigma\sb{2}>\dots>\sup\Sigma\sb{N}.

2699: \end{equation}

2700: The contradiction of (\ref{induction}) and (\ref{induction-2})

2701: shows that

2702: our assumption that $\inf\Sigma\sb{1}<\sup\Sigma\sb{1}$

2703: was false,

2704: hence

2705: $\Sigma\sb{1}=\{\omega\sb{+}\}$

2706: for some $\omega\sb{+}\in[-m,m]$.

2707: \end{proof}

2708: 

2709: Thus, 

2710: $\supp\hat\upbeta\sb{1}(\omega)=\Sigma\sb{1}\subset\{\omega\sb{+}\}$,

2711: with $\omega\sb{+}\in[-m,m]$.

2712: Therefore,

2713: \begin{equation}\label{delom}

2714: \hat\upbeta\sb{1}(\omega)=a\sb 1\delta(\omega-\omega\sb{+}),

2715: \qquad {\rm with\ some}\ \ a\sb 1\in\C.

2716: \end{equation}

2717: Note that

2718: the derivatives $\delta\sp{(k)}(\omega-\omega\sb{+})$, $k\ge 1$

2719: do not enter the expression for

2720: $\hat\upbeta\sb{1}(\omega)=\mathcal{F}\sb{t\to\omega}[\beta(X\sb 1,t)]$

2721: since $\beta(x,t)$

2722: is a bounded continuous function of $(x,t)\in\R^2$

2723: due to the bound (\ref{beta-beta}).

2724: 

2725: \begin{lemma}\label{lem}

2726: $\hat\beta(x,\omega)=a(x)\delta(\omega-\omega\sb{+})$, 

2727: where $a(x)$ is a bounded continuous function.

2728: \end{lemma}

2729: 

2730: \begin{proof}

2731: For $x\le X\sb{1}$, the representation stated in the lemma 

2732: follows from the first line in (\ref{beta-beta-c})

2733: and from (\ref{delom}).

2734: Let us prove this representation for $X\sb{1}\le x\le X\sb{2}$.

2735: By (\ref{delom}), 

2736: we have $\upbeta\sb{1}(t):=\beta(X\sb 1,t)=a\sb{1}e^{-i\omega\sb{+}t}/2\pi$,

2737: hence

2738: $g\sb{1}(t):=F\sb{1}(\upbeta\sb{1}(t))=b\sb{1}e^{-i\omega\sb{+}t}$

2739: for some $b\sb{1}\in\C$

2740: due to the $U(1)$-invariance (\ref{inv-f}).

2741: Therefore, $\hat g\sb{1}(\omega)=2\pi b\sb{1}\delta(\omega-\omega\sb{+})$.

2742: Moreover, by (\ref{beta-beta-c}), we have

2743: $\hat\beta'(X\sb{1}-0,\omega)

2744: =\kappa(\omega\sb{+})a\sb 1\delta(\omega-\omega\sb{+})$.

2745: Hence,

2746: the jump condition (\ref{jump-condition}) implies that

2747: $\hat\beta'(X\sb{1}+0,\omega)=c\sb{1}\delta(\omega-\omega\sb{+})$,

2748: for some $c\sb 1\in\C$.

2749: Finally,  (\ref{beta-beta-c}) implies that

2750: $\hat\beta(x,\omega)=a(x)\delta(\omega-\omega\sb{+})$

2751: for $x\in [X\sb{1},X\sb{2}]$,

2752: with $a(x)$ a continuous complex-valued function of $x$.

2753: Proceeding by induction, we obtain 

2754: similar representation for $\hat\beta(x,\omega)$ for all $x\in\R$.

2755: \end{proof}

2756: 

2757: Now we can finish the proof of Proposition~\ref{prop-one-omega}.

2758: Lemma \ref{lem} implies that

2759: $\beta(x,t)=\phi(x)e^{-i\omega\sb{+}t}$,

2760: where

2761: $\phi(x)=a(x)/2\pi$.

2762: We conclude

2763: from (\ref{beta-beta})

2764: that $\phi\in H\sp{1}(\R)$,

2765: finishing the proof of Proposition~\ref{prop-one-omega}.

2766: Note that

2767: $\omega=\pm m$ could only correspond to the zero solution

2768: (see Remark~\ref{remark-zero}).

2769: \end{proof}

2770: 

2771: According to Remark~\ref{remark-end},

2772: Proposition~\ref{prop-one-omega}

2773: completes the proof of Theorem~\ref{main-theorem}.

2774: 

2775: \section{Multifrequency solitary waves}

2776: \label{sect-counterexamples}

2777: We will show that when the assumptions of Theorem~\ref{main-theorem}

2778: are not satisfied, then the attractor could be more complicated

2779: because the equation admits multifrequency solitary wave solutions.

2780: 

2781: \subsection{Wide gaps}

2782: \label{sect-example-w}

2783: 

2784: Let us consider equation (\ref{kg-no}) with $N=2$,

2785: under Assumptions~\ref{ass-a} and ~\ref{ass-nonl}.

2786: 

2787: \begin{proposition}

2788: If the Assumption~\ref{ass-delta-small}

2789: is violated, then the conclusion of Theorem~\ref{main-theorem}

2790: may no longer be correct.

2791: \end{proposition}

2792: 

2793: \begin{proof}

2794: We will show that

2795: if $L:=X\sb 2-X\sb 1$ is sufficiently large,

2796: then

2797: one can take $F\sb 1(\psi)$ and $F\sb 2(\psi)$

2798: satisfying

2799: Assumptions~\ref{ass-a} and~\ref{ass-nonl}

2800: such that

2801: the global attractor of the equation

2802: contains the multifrequency solutions

2803: which do not converge

2804: to solitary waves of the form (\ref{solitary-waves}).

2805: For our convenience, we assume that $X\sb{1}=0$, $X\sb{2}=L$.

2806: We consider the model (\ref{kg-no}) with the nonlinearity

2807: \begin{equation}

2808: F\sb{1}(\psi)=F\sb{2}(\psi)

2809: =F(\psi),

2810: \qquad

2811: {\rm where}

2812: \quad

2813: F(\psi)

2814: =\alpha\psi+\beta\abs{\psi}^2\psi,

2815: \qquad \alpha,\,\beta\in\R.

2816: \end{equation}

2817: In terms of the condition (\ref{f-is-such}), $p\sb{1}=p\sb{2}=2$.

2818: We take $L$ to be large enough:

2819: \begin{equation}\label{m-pi}

2820: L>\frac{\pi}{2\sp{3/2}m}.

2821: \end{equation}

2822: Consider the function

2823: \begin{equation}\label{psi-plus}

2824: \psi(x,t)=

2825: A (e^{-\kappa(\omega)\abs{x}}

2826: +

2827: e^{-\kappa(\omega)\abs{x-L}})

2828: \sin(\omega t)

2829: +B\chi\sb{[0,L]}(x)\sin(k(3\omega) x)\,\sin(3\omega t),

2830: \qquad

2831: A,\,B\in\C.

2832: \end{equation}

2833: Then

2834: $\psi(x,t)$ solves (\ref{kg-no}) for $x$

2835: away from the points $X\sb{J}$.

2836: We require that

2837: \begin{equation}\label{k-d}

2838: k(3\omega)=\frac{\pi }{L},

2839: \end{equation}

2840: so that $\psi(x,t)$ is continuous in $x\in\R$

2841: and symmetric with respect to $x=L/2$:

2842: \[

2843: \psi(x,t)

2844: =

2845: \psi(\frac{L}{2}-x,t),

2846: \qquad x\in\R.

2847: \]

2848: We need

2849: $\abs{\omega}<m$ to have $\kappa(\omega)>0$, and

2850: $3\abs{\omega}>m$ to have $k(3\omega)\in\R$.

2851: We take $\omega>0$, and thus

2852: $m<3\omega<3m$.

2853: By (\ref{k-d}), this means that we need

2854: \[

2855: m<\sqrt{\frac{\pi^2 }{L^2}+m^2}<3 m.

2856: \]

2857: The second inequality is satisfied

2858: by (\ref{m-pi}).

2859: 

2860: Due to the symmetry of $\psi(x,t)$

2861: with respect to $x=L/2$,

2862: the jump condition

2863: (\ref{jump-condition})

2864: both at $x=0$ and at $x=L$ takes

2865: the following identical form:

2866: \begin{equation}\label{j-at-0-l}

2867: 2A\kappa(\omega) \sin\omega t-B k(3\omega)\sin 3\omega t

2868: =

2869: F\big(A (1+e^{-\kappa(\omega)L})\sin(\omega t)\big).

2870: \end{equation}

2871: Using the identity

2872: \begin{equation}\label{trig}

2873: \sin^3\theta=\frac 3 4\sin\theta-\frac 1 4\sin 3\theta,

2874: \end{equation}

2875: we see that

2876: \begin{equation}

2877: F(A (1+e^{-\kappa(\omega)L})\sin\omega t)

2878: %\alpha\psi(L,t)+\beta\abs{\psi(L,t)}^2\psi(L,t)

2879: =

2880: \Big(\alpha A(1+e^{-\kappa(\omega)L})

2881: +

2882: \frac{3}{4}

2883: \beta \abs{A}^2 A(1+e^{-\kappa(\omega)L})^3

2884: \Big)

2885: \sin(\omega t)

2886: -\frac{1}{4}\beta \abs{A}^2 A(1+e^{-\kappa(\omega)L})^3

2887: \sin(3\omega t).

2888: \end{equation}

2889: Collecting in (\ref{j-at-0-l})

2890: the terms at $\sin\omega t$ and at $\sin 3\omega t$,

2891: we obtain the following system:

2892: \begin{equation}\label{c-e}

2893: \left\{

2894: \begin{array}{l}

2895: 2A\kappa(\omega)

2896: =\alpha A(1+e^{-\kappa(\omega)L})

2897: +\frac 3 4\beta \abs{A}^2 A(1+e^{-\kappa(\omega)L})^3,

2898: \\

2899: B k(3\omega)=\frac 1 4\beta \abs{A}^2 A(1+e^{-\kappa(\omega)L})^3.

2900: \end{array}

2901: \right.

2902: \end{equation}

2903: Assuming that $A\ne 0$,

2904: we divide the first equation by $A$:

2905: \begin{equation}

2906: 2\kappa(\omega)

2907: =\alpha (1+e^{-\kappa(\omega)L})

2908: +\frac 3 4\beta \abs{A}^2(1+e^{-\kappa(\omega)L})^3.

2909: \end{equation}

2910: The condition for the existence of a solution $A\ne 0$

2911: is

2912: \begin{equation}\label{2k-alpha}

2913: \Big(\frac{2\kappa(\omega)}{1+e^{-\kappa(\omega)L}}-\alpha\Big)\beta>0.

2914: \end{equation}

2915: Once we found $A$,

2916: the second equation

2917: in (\ref{c-e})

2918: can be used to express $B$ in terms of $A$.

2919: 

2920: \begin{remark}

2921: Condition (\ref{2k-alpha}) shows that we can choose

2922: $\beta<0$ taking large $\alpha>0$.

2923: The corresponding potential

2924: $U(\psi)=-\alpha\abs{\psi}^2/2-\beta\abs{\psi}^4/4$

2925: satisfies (\ref{bound-below})

2926: and Assumptions~\ref{ass-a} and~\ref{ass-nonl}.

2927: \end{remark}

2928: 

2929: %\begin{remark}

2930: %Many more solutions like (\ref{psi-plus}) are possible.

2931: %Consider the function

2932: %\begin{equation}

2933: %\psi(x,t)

2934: %=A\big(

2935: %e\sp{-\kappa(\omega)\abs{x}}-(-1)^n e\sp{-\kappa(\omega)\abs{x-\Delta}}

2936: %\big)\sin\omega t

2937: %+B\chi\sb{[0,\Delta]}\sin(k(n\omega)x)\sin(n\omega t),\qquad A,\,B\in\C.

2938: %\end{equation}

2939: %This function satisfies (\ref{kg-no}) if the values of $A$ and $B$

2940: %are such that the jump condition is satisfied:

2941: %\begin{equation}\label{eq-a-bn}

2942: %2A\kappa(\omega)\sin(\omega t)-B k(n\omega)\sin(n\omega t)

2943: %=F((1-(-1)^n e^{-\kappa(\omega)\Delta})\sin\omega t).

2944: %\end{equation}

2945: %This relation results in a polynomial equation on $A$.

2946: %The value of $B$ is then expressed in terms of $A$.

2947: %\end{remark}

2948: \end{proof}

2949: 

2950: \subsection{Linear degeneration}

2951: \label{sect-example-li}

2952: Let us consider equation (\ref{kg-no})

2953: with $N=2$,

2954: under Assumptions~\ref{ass-a} and ~\ref{ass-delta-small}.

2955: 

2956: \begin{proposition}

2957: If the Assumption~\ref{ass-nonl}

2958: is violated, then the conclusion of Theorem~\ref{main-theorem}

2959: may no longer be correct.

2960: \end{proposition}

2961: 

2962: \begin{proof}

2963: Again, we construct

2964: multifrequency solutions.

2965: Consider the equation

2966: \begin{equation}

2967: \ddot\psi

2968: =\psi''-m^2\psi

2969: +\delta(x)F\sb{1}(\psi)

2970: +\delta(x-L)F\sb{2}(\psi),

2971: \end{equation}

2972: where

2973: \begin{equation}

2974: F\sb{1}(\psi)=\alpha\psi+\beta\abs{\psi}^2\psi,

2975: \qquad

2976: F\sb{2}(\psi)=\gamma\psi,

2977: \qquad

2978: \alpha,\ \beta,\ \gamma\in\R.

2979: \end{equation}

2980: Note that the function $F\sb{2}$ is linear,

2981: failing to satisfy

2982: Assumption~\ref{ass-nonl}.

2983: The function

2984: \[

2985: \psi(x,t)

2986: =

2987: \left\{

2988: \begin{array}{l}

2989: (A+B) e^{\kappa(\omega)x}\sin(\omega t),

2990: \qquad x\le 0,

2991: \\

2992: \big(A e^{-\kappa(\omega)x}+B e^{\kappa(\omega)x}\big)

2993: \sin(\omega t)

2994: +C\sinh(\kappa(3\omega)x)\sin(3\omega t),

2995: \qquad x\in[0,L],

2996: \\

2997: (A e^{-\kappa(\omega)}+B e^{\kappa(\omega)(2L-x)})\sin(\omega t)

2998: +\frac{C}{\sinh(\kappa(3\omega)L)}

2999: e^{-\kappa(3\omega)(x-L)}\sin(3\omega t)

3000: ,

3001: \qquad

3002: x\ge L,

3003: \end{array}

3004: \right.

3005: \]

3006: where $\omega\in(0,m/3)$,

3007: will be a solution if

3008: the jump conditions are satisfied

3009: at $x=0$ and at $x=L$:

3010: \begin{equation}\label{j-at-0}

3011: -\psi'(0+,t)+\psi'(0-,t)=\alpha\psi(0,t)+\beta\psi^3(0,t),

3012: \end{equation}

3013: \begin{equation}\label{j-at-l}

3014: -\psi'(L+,t)+\psi'(L-,t)=\alpha\psi(L,t)+\beta\psi^3(L,t).

3015: \end{equation}

3016: We use the identity

3017: \[

3018: \alpha(A+B)\sin(\omega t)

3019: +\beta((A+B)\sin(\omega t))^3

3020: =\Big(\alpha(A+B)+\beta\frac{3(A+B)^3}{4}\Big)\sin(\omega t)

3021: -\beta\frac{(A+B)^3}{4}\sin(3\omega t)

3022: \]

3023: which follows from (\ref{trig}).

3024: Collecting the terms at $\sin(\omega t)$

3025: and at $\sin(3\omega t)$,

3026: we write the condition (\ref{j-at-0})

3027: as the following system of equations:

3028: \begin{eqnarray}

3029: &&

3030: 2\kappa(\omega)A=\Big(\alpha(A+B)+\beta\frac{3(A+B)^3}{4}\Big),

3031: \label{c01}

3032: \\

3033: &&

3034: -\kappa(3\omega)C=-\beta\frac{(A+B)^3}{4}.

3035: \label{c03}

3036: \end{eqnarray}

3037: Similarly, the condition (\ref{j-at-l}) is equivalent to the

3038: following two equations:

3039: \begin{eqnarray}

3040: &&

3041: 2B\kappa(\omega)e^{\kappa(\omega)L}

3042: =\gamma(A e^{-\kappa(\omega)L}+B e^{\kappa(\omega)L}),

3043: \label{cl1}

3044: \\

3045: &&

3046: \frac{\kappa(3\omega)C}{\sinh(\kappa(3\omega)L)}

3047: +

3048: \kappa(3\omega)C\cosh(\kappa(3\omega)L)

3049: =\gamma C\sinh(\kappa(3\omega)L).

3050: \label{cl3}

3051: \end{eqnarray}

3052: Equations (\ref{c01}), (\ref{c03}), (\ref{cl1}), and (\ref{cl3})

3053: could be satisfied for arbitrary $L>0$.

3054: Namely, for any $\omega\in(0,m/3)$,

3055: one uses (\ref{cl3}) to determine $\gamma$.

3056: For any $\beta\ne 0$,

3057: there is always a solution $A$, and $B$

3058: to the nonlinear system (\ref{c01}), (\ref{cl1}).

3059: Finally, $C$ is obtained from (\ref{c03}).

3060: 

3061: %\begin{remark}

3062: %Note that such an example for small $L$

3063: %requires very strong linear force at $x=L$.

3064: %Namely, as follows from (\ref{cl3}), $\lim\sb{L\to 0}\gamma=\infty$.

3065: %\end{remark}

3066: \end{proof}

3067: 

3068: 

3069: \appendix

3070: 

3071: \section{Global well-posedness}

3072: 

3073: 

3074: \label{sect-existence}

3075: 

3076: Here we prove Theorem~\ref{theorem-well-posedness}.

3077: We first need

3078: to adjust the nonlinearity $F$

3079: so that it becomes bounded, together with

3080: its derivatives.

3081: Define

3082: \begin{equation}\label{def-lambda-0}

3083: \lambda\sb{0}

3084: =\sqrt{\frac{\mathcal{H}(\psi\sb{0},\pi\sb{0})-\sum\sb{J}{A}\sb{J}}{m-\sum\sb{J}{B}\sb{J}}},

3085: \end{equation}

3086: where $(\psi\sb{0},\pi\sb{0})\in{\E}$ is the initial data

3087: from Theorem~\ref{theorem-well-posedness}

3088: and ${A}\sb{J}$, ${B}\sb{J}$ are constants from (\ref{bound-below}).

3089: Then we may pick a modified potential function

3090: $\widetilde{U}\sb{J}\in C\sp 2(\C,\R)$,

3091: $\widetilde{U}\sb{J}(\psi)=\widetilde{U}\sb{J}(\abs{\psi})$,

3092: $j=1,\,2$,

3093: so that

3094: \begin{equation}\label{new-U}

3095: \widetilde{U}\sb{J}(\psi)=U\sb{J}(\psi)

3096: \qquad{\rm for}\ \abs{\psi}\le\lambda\sb{0},

3097: \quad

3098: \psi\in\C,

3099: \end{equation}

3100: $\widetilde{U}\sb{J}(\psi)$ satisfy (\ref{bound-below})

3101: with the same constants ${A}\sb{J}$, ${B}\sb{J}$ as $U\sb{J}(\psi)$ do:

3102: \begin{equation}\label{new-U-2}

3103: \widetilde{U}\sb{J}(\psi)\ge {A}\sb{J}-{B}\sb{J}\abs{\psi}^2,

3104: \quad{\rm for}\ \psi\in\C,\quad

3105: {\rm where}

3106: \quad

3107: {A}\sb{J}\in\R,

3108: \quad

3109: {B}\sb{J}\ge 0,

3110: \quad

3111: 1\le J\le N,

3112: \quad

3113: \sum\sb{J}{B}\sb{J}<m,

3114: \end{equation}

3115: and so that

3116: $\abs{\widetilde{U}\sb{J}(\psi)}$,

3117: $\abs{\widetilde{U}\sb{J}'(\psi)}$,

3118: and $\abs{\widetilde{U}\sb{J}''(\psi)}$

3119: are bounded for $\psi\ge 0$.

3120: We define

3121: \begin{equation}\label{f-reg}

3122: \widetilde{F}\sb{J}(\psi)

3123: =-\nabla \widetilde{U}\sb{J}(\psi),

3124: \qquad\psi\in\C,

3125: \end{equation}

3126: where $\nabla$ denotes the gradient with respect to

3127: $\Re\psi$, $\Im\psi$;

3128: Then

3129: $\widetilde{F}\sb{J}(e\sp{is}\psi)

3130: =e\sp{is}\widetilde{F}\sb{J}(\psi)$ for any $\psi\in\C$, $s\in\R$.

3131: 

3132: We consider the Cauchy problem of type

3133: (\ref{kg-no}) with the modified

3134: nonlinearity,

3135: \begin{equation}\label{kg-no-a}

3136: \left\{

3137: \begin{array}{l}

3138: \ddot\psi(x,t)

3139: =\psi''(x,t)-m^2\psi(x,t)+\sum\sb{J}\delta(x-X\sb{J})

3140: \widetilde F\sb{J}(\psi(X\sb{J},t)),

3141: \qquad

3142: x\in\R,

3143: \quad t\in\R,

3144: \\

3145: \psi\at{t=0}=\psi\sb{0}(x),

3146: \qquad

3147: \dot\psi\at{t=0}=\pi\sb{0}(x).

3148: \end{array}\right.

3149: \end{equation}

3150: %which we rewrite in the vector form

3151: %in terms of

3152: %$\Psi=\left[\begin{array}{c}\psi(x,t)\\\pi(x,t)\end{array}\right]$:

3153: %\begin{equation}

3154: %\dot\Psi

3155: %=\left[\begin{array}{cc}0&1\\\p\sb x^2-m^2&0\end{array}\right]\Psi

3156: %+\sum\sb{J}\delta(x-X\sb{J})

3157: %\left[\begin{array}{c}0\\\widetilde{F}\sb{J}(\psi)\end{array}\right],

3158: %\qquad\Psi\at{t=0}

3159: %=\Psi\sb{0}

3160: %\equiv\left[\begin{array}{c}\psi\sb{0}(x)\\\pi\sb{0}(x)\end{array}\right].

3161: %\label{kg-no-ap}

3162: %\end{equation}

3163: Equation (\ref{kg-no-a})

3164: formally can be written as the following Hamiltonian system

3165: (Cf. (\ref{kg-no-h})):

3166: \begin{equation}\label{kg-no-h-t}

3167: \dot\Psi(t)=\mathcal{J}\,D\widetilde{\mathcal{H}}(\Psi),

3168: \qquad

3169: \mathcal{J}=\left[\begin{array}{cc}0&1\\-1&0\end{array}\right],

3170: \end{equation}

3171: where $D\widetilde{\mathcal{H}}$ is the variational

3172: derivative of the Hamilton functional

3173: \begin{equation}\label{kg-no-a-h}

3174: \widetilde{\mathcal{H}}(\Psi)

3175: =\int\limits\sb{\R}

3176: \left(

3177: \abs{\pi}\sp 2

3178: +\abs{\nabla\psi}\sp 2+m^2\abs{\psi}\sp 2

3179: \right)\,dx

3180: +\sum\sb{J}\widetilde{U}\sb{J}(\psi(X\sb{J},t)),

3181: \quad

3182: \Psi=\left[\begin{array}{c}\psi(x)\\

3183: \pi(x)\end{array}\right]\in\E,

3184: \end{equation}

3185: which is Fr\'echet differentiable in the space

3186: ${\E}=H\sp{1}\times L\sp 2$.

3187: By the Sobolev embedding theorem,

3188: $L\sp\infty(\R)\subset H\sp{1}(\R)$,

3189: and there is the following inequality:

3190: \begin{equation}\label{sobolev-embedding}

3191: \norm{\psi}\sb{L\sp\infty}^2

3192: \le\frac{1}{2m}

3193: (\norm{\psi'}\sb{L\sp 2}^2+m^2\norm{\psi}\sb{L\sp 2}^2)

3194: \le\frac{1}{2m}

3195: \norm{\Psi}\sb{\E}^2.

3196: \end{equation}

3197: Thus, (\ref{new-U-2}) leads to

3198: \begin{equation}\label{bound-on-u}

3199: \widetilde{U}\sb{J}(\psi(0))

3200: \ge {A}\sb{J}-{B}\sb{J}\norm{\psi}\sb{L\sp\infty}^2

3201: \ge {A}\sb{J}-\frac{{B}\sb{J}}{2m}\norm{\Psi}\sb{\E}^2.

3202: \end{equation}

3203: Taking into account (\ref{kg-no-a-h}),

3204: we obtain the inequality

3205: \begin{equation}

3206: \norm{\Psi}\sb{\E}\sp 2

3207: =2\widetilde{\mathcal{H}}(\Psi)-2\sum\sb{J}\widetilde{U}\sb{J}(\psi(X\sb{J}))

3208: \le 2\widetilde{\mathcal{H}}(\Psi)-2\sum\sb{J}{A}\sb{J}+\frac{\sum\sb{J}{B}\sb{J}}{m}

3209: \norm{\Psi}\sb{\E}^2,

3210: \qquad

3211: \Psi\in\E.

3212: \end{equation}

3213: It follows that

3214: \begin{equation}\label{t-bound-1}

3215: \norm{\Psi}\sb{\E}\sp 2

3216: \le\frac{2m}{m-\sum\sb{J}{B}\sb{J}}

3217: \Big(\widetilde{\mathcal{H}}(\Psi)-\sum\sb{J}{A}\sb{J}\Big),

3218: \qquad\Psi\in\E.

3219: \end{equation}

3220: 

3221: \begin{lemma}\label{lemma-same-u}

3222: \begin{enumerate}

3223: \item

3224: There is the identity

3225: $\widetilde{\mathcal{H}}(\Psi\sb{0})

3226: =\mathcal{H}(\Psi\sb{0})$.

3227: \item

3228: If $\Psi=\left[\begin{array}{c}\psi(x)\\\pi(x)\end{array}\right]\in{\E}$

3229: satisfies

3230: $\widetilde{\mathcal{H}}(\Psi)\le\widetilde{\mathcal{H}}(\Psi\sb{0})$,

3231: then

3232: $\ \widetilde{U}\sb{J}(\psi(x))=U\sb{J}(\psi(x))$

3233: for any $x\in\R$.

3234: \end{enumerate}

3235: \end{lemma}

3236: 

3237: \begin{proof}

3238: According to (\ref{t-bound-1}),

3239: the Sobolev embedding (\ref{sobolev-embedding}),

3240: and the choice of $\lambda\sb{0}$ in (\ref{def-lambda-0}),

3241: \begin{equation}

3242: \norm{\psi\sb{0}}\sb{L\sp\infty}^2

3243: \le\frac{1}{2m}\norm{\Psi\sb{0}}\sb{\E}^2

3244: \le\frac{\mathcal{H}(\Psi\sb{0})-\sum\sb{J}{A}\sb{J}}{m-\sum\sb{J}{B}\sb{J}}

3245: =\lambda\sb{0}^2.

3246: \end{equation}

3247: Thus, by (\ref{new-U}),

3248: $

3249: \widetilde{U}(\psi\sb{0}(x))=U(\psi\sb{0}(x))

3250: $

3251: for all $x\in\R$.

3252: This proves ({\it i}).

3253: 

3254: By (\ref{sobolev-embedding}),

3255: the relation (\ref{t-bound-1}),

3256: the condition

3257: $\widetilde{\mathcal{H}}(\Psi)\le\widetilde{\mathcal{H}}(\Psi\sb{0})$,

3258: and part ({\it i}) of the Lemma, we have:

3259: \[

3260: \norm{\psi}\sb{L\sp\infty}^2

3261: \le\frac{1}{2m}\norm{\Psi}\sb{\E}^2

3262: \le\frac{\widetilde{\mathcal{H}}(\Psi)-\sum\sb{J}{A}\sb{J}}{m-\sum\sb{J}{B}\sb{J}}

3263: \le\frac{\widetilde{\mathcal{H}}(\Psi\sb{0})-\sum\sb{J}{A}\sb{J}}{m-\sum{B}\sb{J}}

3264: =\frac{\mathcal{H}(\Psi\sb{0})-\sum\sb{J}{A}\sb{J}}{m-\sum\sb{J}{B}\sb{J}}

3265: =\lambda\sb{0}^2.

3266: \]

3267: Now the statement ({\it ii}) follows by (\ref{new-U}).

3268: \end{proof}

3269: 

3270: If $\Psi(t)$ solves (\ref{kg-no-h-t}),

3271: then

3272: $\widetilde{\mathcal{H}}(\Psi(t))=\widetilde{\mathcal{H}}(\Psi\sb{0})$,

3273: By Lemma~\ref{lemma-same-u}~({\it ii}),

3274: $\widetilde{U}\sb{J}(\psi(x,t))=U\sb{J}(\psi(x,t))$

3275: for all $x\in\R$, $t\in\R$.

3276: Hence, $\widetilde{F}\sb{J}(\psi(x,t))=F\sb{J}(\psi(x,t))$

3277: for all $x\in\R$, $t\ge 0$,

3278: allowing us to conclude that

3279: $\psi(t)$ solves (\ref{kg-no})

3280: as well as (\ref{kg-no-a}).

3281: The rest of the proof

3282: of Theorem~\ref{theorem-well-posedness}

3283: repeats the proof of a similar result for the case $N=1$

3284: \cite[Theorem 2.3]{ubk-arma}.

3285: 

3286: \bibliographystyle{sima}

3287: %\bibliographystyle{amsalpha}

3288: \bibliography{comech,ubk-mathsci,ubk-local,all}

3289: 

3290: \end{document}

3291: