1: \documentclass{article}

2: \usepackage{mathptmx} % use Times fonts if available on your TeX system

3: 

4: \pagestyle{myheadings}

5: \markboth{}{Alexander Komech, Andrew Komech:

6: \ {\rm Global attractor for nonlinear oscillator coupled to Klein-Gordon field}}

7: 

8: 

9: \hyphenation{Shni-rel-man}

10: \hyphenation{quasi-mea-sure}

11: \hyphenation{Cuc-cagna}

12: \hyphenation{non-zero}

13: 

14: \textwidth 17cm

15: \textheight 22cm

16: %\textheight 23cm

17: \topmargin -1.4cm

18: %\topmargin 0mm

19: \oddsidemargin -3mm

20: \evensidemargin -3mm

21: 

22: \usepackage{latexsym,epsfig,bm}

23: \usepackage{amssymb}

24: \usepackage{color}

25: \usepackage{amsthm}

26: \usepackage{mathrsfs}

27: 

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

29: \DeclareSymbolFontAlphabet{\mathbb}{AMSb}

30: 

31: 

32: \newcommand{\E}{{\cal E}}

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

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

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

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

37: \newcommand{\toEF}{\stackrel{{\E}\sb F}\longrightarrow}

38: 

39: \def\xx{{j}}

40: \def\yy{{k}}

41: 

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

43: \newcommand\p{\partial}

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

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

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

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

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

49: \providecommand\C{{\mathbb C}}

50: \renewcommand\C{{\mathbb C}}

51: \newcommand{\R}{{\mathbb R}}

52: \newcommand{\N}{{\mathbb N}}

53: \newcommand\Z{{\mathbb Z}}

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

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

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

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

58: \newcommand{\const}{{\rm const}}

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

60: \newcommand\sothat{{\rm :}\ }

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

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

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

64: 

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

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

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

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

69: }

70: 

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

72: %\DeclareMathSymbol{\varDelta}{\mathord}{letters}{"01}

73: %\DeclareMathSymbol{\varTheta}{\mathord}{letters}{"02}

74: %\DeclareMathSymbol{\varLambda}{\mathord}{letters}{"03}

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

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

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

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

79: %\DeclareMathSymbol{\varPhi}{\mathord}{letters}{"08}

80: %\DeclareMathSymbol{\varPsi}{\mathord}{letters}{"09}

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

82: 

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

84: %\spnewtheorem{assumption}{Assumption}{\bf}{\it}%{assumption}

85: %\spnewtheorem{lemma}{Lemma}[section]{\bf}{\it}%{assumption}

86: 

87: %\makeatletter\@addtoreset{equation}{section}

88: %\makeatletter\@addtoreset{theorem}{section}

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

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

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

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

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

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

95: %\makeatother

96: 

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

98: %\renewcommand{\thetheorem}{\thesection.\arabic{theorem}}

99: %\renewcommand{\theproposition}{\thesection.\arabic{proposition}}

100: %\renewcommand{\thelemma}{\thesection.\arabic{lemma}}

101: %\renewcommand{\thecorollary}{\thesection.\arabic{corollary}}

102: %\renewcommand{\theremark}{\thesection.\arabic{remark}}

103: %\renewcommand{\theassumption}{\thesection.\arabic{assumption}}

104: %\renewcommand{\thedefinition}{\thesection.\arabic{definition}}

105: 

106: 

107: % Italic ``theorems''

108: \font\thf cmssdc10 at 11pt

109: %%\font\mysmall cmr10 at 6pt

110: \theoremstyle{plain}

111: \newtheorem{theorem}{\thf Theorem}[section]

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

113: %\newtheorem{lemma}{\thf Lemma}[section]

114: \newtheorem{lemma}[theorem]{\thf Lemma}

115: \newtheorem{corollary}[theorem]{\thf Corollary}

116: \newtheorem{proposition}[theorem]{\thf Proposition}

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

118: % Roman ``theorems''

119: \theoremstyle{definition}

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

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

122: \renewcommand{\theassumption}{\Alph{assumption}}

123: 

124: % Humble things: remarks and examples.

125: \theoremstyle{remark}

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

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

128: 

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

130: \makeatother

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

132: 

133: \begin{document}

134: %

135: \title{

136: %On Global Attraction to Quantum Stationary States I.

137: %\\

138: %Nonlinear Oscillator Coupled to Massive Scalar Field

139: %

140: Global attractor for a nonlinear oscillator

141: coupled to the Klein-Gordon field

142: }

143: 

144: \author{

145: {\sc Alexander Komech}

146: \footnote{

147: On leave from Department of Mechanics and Mathematics,

148: Moscow State University, Moscow 119899, Russia.

149: Supported in part

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

151: the

152: Wolfgang Pauli Institute and the Faculty of Mathematics,

153: Vienna University,

154: by DFG Grant 436\,RUS\,113/615/0-1,

155: and by FWF Grant P19138-N13.}

156: \\

157: {\it Faculty of Mathematics, Wien A-1090, Austria}

158: \\ \\

159: {\sc Andrew Komech}

160: \footnote{

161: %Permanent address:

162: %Department of Mathematics, Texas A\&M University,

163: %College Station, TX 77843, USA.

164: Supported in part

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

166: by the NSF Grant DMS-0434698.

167: }

168: \\

169: {\it 

170: %Max Planck Institute, D-04103 Leipzig, Germany

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

172: College Station, TX, USA

173: }}

174: 

175: \maketitle

176: %

177: \begin{abstract}

178: The long-time asymptotics is analyzed for all finite energy

179: solutions to a model $\mathbf{U}(1)$-invariant

180: nonlinear Klein-Gordon equation in one dimension,

181: with the nonlinearity concentrated at a single point:

182: {\it each finite energy solution} converges

183: as $t\to\pm\infty$

184: to the set of all ``nonlinear eigenfunctions''

185: of the form $\psi(x)e\sp{-i\omega t}$.

186: The {\it global attraction}

187: is caused by the nonlinear

188: energy transfer from lower harmonics to the continuous spectrum

189: and subsequent dispersive radiation.

190: 

191: We justify this mechanism by the following novel

192: strategy

193: based on

194: \emph{inflation of spectrum

195: by the nonlinearity}.

196: We show that

197: any {\it omega-limit trajectory}

198: has the time-spectrum in the spectral gap $[-m,m]$ and

199: satisfies the original equation.

200: This equation implies the key

201: {\it spectral inclusion} for spectrum of the nonlinear term.

202: Then the application of the Titchmarsh Convolution Theorem

203: reduces

204: the spectrum of each omega-limit trajectory

205: to a single harmonic

206: $\omega\in[-m,m]$.

207: 

208: The research

209: is inspired by Bohr's postulate on

210: quantum transitions and

211: Schr\"o\-din\-ger's identification

212: of the quantum stationary states to the nonlinear eigenfunctions

213: of the coupled $\mathbf{U}(1)$-invariant

214: Maxwell-Schr\"o\-din\-ger and Maxwell-Dirac equations.

215: \end{abstract}

216: 

217: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

218: 

219: \section{Introduction}

220: \label{sect-introduction}

221: The long time asymptotics for

222: nonlinear

223: wave equations

224: have been the subject of intensive research,

225: starting with the pioneering papers by

226: Segal \cite{MR0153967,MR0152908},

227: Strauss \cite{MR0233062},

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

229: where the

230: nonlinear scattering and the local attraction to zero were considered.

231: The asymptotic stability of solitary waves

232: has been studied since the 1990s by

233: Soffer and Weinstein \cite{MR1071238,MR1170476},

234: Buslaev and Perelman \cite{MR1199635e,MR1334139},

235: and then by others.

236: The existing results

237: suggest that the set of

238: orbitally stable solitary waves

239: typically forms a local attractor,

240: that is, attracts

241: finite energy solutions

242: that were initially close to it.

243: 

244: In this paper, we consider the global attractor

245: for

246: all finite energy solutions.

247: For the first time, we prove that

248: in a particular $\mathbf{U}(1)$-invariant

249: dispersive Hamiltonian system

250: the global attractor is finite-dimensional

251: and is formed by solitary waves.

252: The investigation is inspired by Bohr's quantum transitions

253: (``quantum jumps").

254: Namely, according to Bohr's postulates \cite{bohr1913},

255: an unperturbed electron lives forever

256: in a \emph{quantum stationary state} $\vert E\rangle$

257: that has a definite value $E$ of the energy.

258: Under an external perturbation,

259: the electron

260: can

261: jump from one state to another:

262: \begin{equation}\label{transitions}

263: \vert E\sb{-}\rangle

264: \longmapsto

265: \vert E\sb{+}\rangle.

266: \end{equation}

267: The postulate

268: suggests the dynamical interpretation

269: of the transitions

270: as long-time attraction

271: \begin{equation}\label{ga}

272: \Psi(t)\longrightarrow\vert E\sb\pm\rangle,

273: \qquad

274: t\to\pm\infty

275: \end{equation}

276: for any trajectory $\Psi(t)$ of the corresponding dynamical system,

277: where the limiting states $\vert E\sb\pm\rangle$ generally

278: depend on the trajectory.

279: Then the quantum stationary states

280: should be viewed as the points of the \emph{global attractor}

281: $\mathcal{S}$ which is the set of all

282: limiting states

283: (see Figure~\ref{fig-gaa}).

284: \begin{figure}%[htbp]

285: {

286: \input f2.pstex_t

287: }

288: \vskip 0.5cm

289: \caption{

290: Attraction of any trajectory $\Psi(t)$ to the set

291: of solitary waves

292: as $t\to\pm\infty$.}

293: \label{fig-gaa}

294: \end{figure}

295: Following de Broglie's ideas,

296: Schr\"o\-din\-ger identified the stationary states

297: $\vert E\rangle$

298: as the solutions

299: of the wave equation that have the form

300: \begin{equation}\label{solitary-waves-0}

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

302: \qquad

303: \omega=E/\hbar,

304: \end{equation}

305: where $\hbar$ is Planck's constant.

306: Then the attraction (\ref{ga})

307: takes the form of the long-time asymptotics

308: \begin{equation}\label{asymptotics}

309: \psi(x,t)

310: \sim

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

312: \qquad

313: t\to\pm\infty,

314: \end{equation}

315: that hold for each finite energy solution.

316: However, because of the superposition

317: principle,

318: the asymptotics of type (\ref{asymptotics})

319: are generally impossible for

320: the linear autonomous Schr\"o\-din\-ger equation

321: of type

322: \begin{equation}

323: (i\p\sb t-V(x))\psi(x,t)

324: =(-i\nabla-{\bm A}(x))\sp 2\psi(x,t),

325: \end{equation}

326: where $V(x)$ and ${\bm A}(x)$ are

327: scalar and vector potentials of a static external Maxwell field.

328: An adequate description of this process

329: requires to consider the Schr\"o\-din\-ger

330: (or Dirac)

331: equation

332: coupled to the Maxwell system

333: which governs the time evolution of the Maxwell 4-potential

334: $A(x,t)=(V(x,t),{\bm A}(x,t))$.

335: This coupling is inevitable indeed,

336: because, again by Bohr's postulates,

337: the transitions (\ref{transitions})

338: are followed by electromagnetic radiation responsible for the

339: atomic spectra.

340: The coupled Maxwell-Schr\"o\-din\-ger system

341: was initially introduced in \cite{Sch81109}.

342: It is a $\mathbf{U}(1)$-invariant

343: nonlinear Hamiltonian system.

344: Its global well-posedness was considered in \cite{MR1331696}.

345: One might expect the following

346: generalization of asymptotics (\ref{asymptotics})

347: for solutions to the coupled Maxwell-Schr\"o\-din\-ger

348: (or Maxwell-Dirac)

349: equations:

350: \begin{equation}\label{asymptotics-a}

351: (\psi(x,t), A(x,t))

352: \sim

353: \left(

354: \phi\sb{\omega\sb\pm}(x)e\sp{-i\omega\sb\pm t},

355: A\sb{\omega\sb\pm}(x)

356: \right),

357: \qquad

358: t\to\pm\infty.

359: \end{equation}

360: The asymptotics of this form are not available yet

361: in the context of coupled systems.

362: Let us mention that

363: the existence of the solitary  waves for

364: the coupled Maxwell-Dirac equations

365: was established in \cite{MR1386737}.

366: 

367: The asymptotics (\ref{asymptotics-a}) would mean that

368: the set of all solitary waves

369: \[

370: \{\left(\phi\sb\omega(x),A\sb\omega(x)\right):\omega\in\C\}

371: \]

372: forms a global attractor for the coupled system.

373: Similar convergence to a global attractor

374: is well known for dissipative systems, like Navier-Stokes equations

375: (see \cite{MR1156492,He81,MR1441312}).

376: In this context,

377: the global attractor is formed by the \emph{static stationary states},

378: and the corresponding asymptotics (\ref{asymptotics})

379: only hold for $t\to+\infty$ (and with $\omega\sb{+}=0$).

380: 

381: Our main impetus for writing this paper

382: was the natural question

383: whether dispersive Hamiltonian systems

384: could, in the same spirit,

385: possess finite dimensional global attractors,

386: and whether such attractors

387: are formed by the solitary waves.

388: We prove such a global attraction

389: for a model nonlinear Klein-Gordon equation

390: \begin{equation}\label{KG-0}

391: \ddot\psi(x,t)

392: =\psi''(x,t)-m^2\psi(x,t)+\delta(x)F(\psi(0,t)),

393: \qquad

394: \quad x\in\R.

395: \end{equation}

396: Here $m>0$, $\psi(x,t)$ is a continuous complex-valued wave function,

397: and $F$ is a nonlinearity.

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

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

400: All derivatives and the equation are understood in

401: the sense of distributions.

402: Equation (\ref{KG-0}) describes the linear Klein-Gordon equation

403: coupled to the nonlinear oscillator.

404: We assume that equation (\ref{KG-0}) is $\mathbf{U}(1)$-invariant;

405: that is,

406: $$

407: F(e\sp{i\theta}\psi)=e\sp{i\theta}F(\psi),

408: \qquad

409: \psi\in\C,

410: \quad

411: \theta\in\R.

412: $$

413: Note that the group $\mathbf{U}(1)$ is also the (global) gauge group

414: of the coupled Maxwell-Schr\"o\-din\-ger and Maxwell-Dirac equations,

415: with the representation given by

416: \[

417: (\psi(x),A(x))\mapsto (e^{i\theta}\psi(x),A(x)).

418: \]

419: This gauge symmetry leads to the charge

420: conservation

421: and to the existence of the solitary wave solutions

422: of the form (\ref{asymptotics-a}) (see \cite{MR1386737}).

423: We clarify the special role of the

424: ``nonlinear eigenfunctions'',

425: or solitary waves, of equation (\ref{KG-0})

426: which are finite energy solutions of type

427: (\ref{solitary-waves-0}):

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

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

430: \end{equation}

431: We prove that indeed they form the global attractor for all finite

432: energy solutions to (\ref{KG-0}).

433: 

434: Equation (\ref{KG-0}) has the following key features

435: of the

436: coupled Maxwell-Schr\"o\-din\-ger and Maxwell-Dirac

437: equations:

438: ({\it i})

439: The linear part of this equation has a dispersive character;

440: ({\it iii})

441: It is a nonlinear Hamiltonian system;

442: ({\it iii})

443: It is $\mathbf{U}(1)$-invariant.

444: We suggest that just these features

445: are responsible for the global attraction,

446: such as

447: (\ref{asymptotics}), (\ref{asymptotics-a}),

448: to ``quantum stationary states''.

449: \bigskip

450: 

451: Let us introduce the set of all solitary waves.

452: 

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

454: Let $\mathcal{S}$ be the set of all functions

455: $\phi\sb\omega(x)\in H\sp 1(\R)$ with $\omega\in\C$,

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

457: is a solution to (\ref{KG-0}).

458: \end{definition}

459: Here $H\sp 1(\R)$ denotes the Sobolev space.

460: Generically,

461: the quotient $\mathcal{S}/\mathbf{U}(1)$

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

463: intervals.

464: We will give an explicit construction

465: of the set of all solitary waves for equation (\ref{KG-0});

466: See Proposition~\ref{prop-solitons}

467: and its proof in Appendix~\ref{sect-solitons}.

468: Let us mention

469: that there are numerous results on the existence of

470: solitary wave solutions

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

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

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

474: Typically, such solutions exist for

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

476: of the real line.

477: 

478: \bigskip

479: 

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

481: In the case when the nonlinearity $F$ is polynomial

482: of order strictly

483: greater than $1$,

484: we prove the attraction

485: of any finite energy solution

486: to the set $\mathcal{S}$ of all solitary waves:

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

488: \psi(\cdot,t)\longrightarrow\mathcal{S},

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

490: \end{equation}

491: where the convergence holds in local energy seminorms.

492: In the linear case, when $F(\psi)=a\psi$ with $a\in\R$,

493: there is generally no attraction

494: to $\mathcal{S}$; instead, we show that the global attractor

495: is

496: the linear span of all solitary waves, $ \langle\mathcal{S}\rangle$.

497: See Theorem~\ref{main-theorem-linear}.

498: 

499: 

500: \begin{remark}\label{remark-short}

501: Although we proved the attraction (\ref{attraction}) to $\mathcal{S}$,

502: we have not proved the attraction to a particular solitary wave,

503: falling short of proving (\ref{asymptotics}).

504: Hypothetically, a solution can be drifting along $\mathcal{S}$,

505: keeping asymptotically close to it,

506: but never approaching a particular solitary wave.

507: \end{remark}

508: 

509: \begin{remark}

510: The requirement that the nonlinearity $F$ is polynomial

511: allows us to apply the Titchmarsh convolution theorem

512: that is vital to the proof.

513: We do not know whether this requirement could be dropped.

514: \end{remark}

515: 

516: Let us mention related earlier results:

517: 

518: \begin{enumerate}

519: \item

520: The asymptotics of type (\ref{asymptotics})

521: were discovered first with $\psi\sb\pm=0$

522: in the scattering theory

523: \cite{MR0233062,MR0303097,MR498955,MR535231,MR654553,MR824083,MR1120284}.

524: In this case, the attractor $\mathcal{S}$ consists of the zero solution only,

525: and the asymptotics mean well-known \emph{local energy decay}.

526: 

527: \item

528: The \emph{global attraction}

529: of type (\ref{asymptotics})

530: with $\psi\sb\pm\ne 0$ and $\omega\sb{\pm}=0$

531: was established in

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

533: for a number of nonlinear

534: wave problems.

535: There the attractor $\mathcal{S}$

536: is the set of all \emph{static} stationary states.

537: Let us mention that this set could be infinite

538: and contain continuous components.

539: 

540: \item

541: First results on the asymptotics of type (\ref{asymptotics}),

542: with $\omega\sb\pm\ne 0$

543: were obtained for nonlinear $\mathbf{U}(1)$-invariant

544: Schr\"o\-din\-ger equations

545: in the context of asymptotic stability.

546: This establishes asymptotics of type (\ref{asymptotics})

547: but only for solutions close to the solitary waves,

548: proving the existence of a \emph{local attractor}.

549: This was first done in

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

551: and then developed in

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

553: and other papers.

554: \end{enumerate}

555: 

556: The \emph{global attraction} (\ref{attraction})

557: to the solitary waves with  $\omega\ne 0$

558: was announced

559: for the first time

560: in \cite{MR2032730}  for equation (\ref{KG-0}).

561: In the present paper we give the detailed proofs,

562: and also add the well-posedness result which is not trivial

563: since the Dirac delta-function $\delta(x)$ does not belong to $L^2(\R)$.

564: 

565: Let us mention that the attraction (\ref{asymptotics})

566: for equation (\ref{KG-0}) with $m=0$

567: was proved in

568: \cite{MR1203302e,MR1359949}; In that case $\omega\sb\pm=0$.

569: Our proofs for $m>0$ are quite different from

570: \cite{MR1203302e,MR1359949}, and

571: are based

572: on a nonlinear spectral analysis

573: of \emph{omega-limit trajectories} for $t\to +\infty$

574: (and similarly for $t\to -\infty$).

575: First, we prove that their time-spectrum

576: is contained in a finite interval $[-m,m]$,

577: since the

578: spectral density

579: is absolutely continuous

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

581: and the corresponding component of the solution disperses completely.

582: Second,

583: the nonlinear equation (\ref{KG-0}) implies

584: the crucial spectral inclusion:

585: The nonlinearity does not inflate the spectrum

586: of any omega-limit trajectory.

587: Finally, the Titchmarsh convolution theorem

588: allows us to reduce

589: the spectrum of the omega-limit trajectory

590: to a single harmonic

591: $\omega\sb{+}\in[-m,m]$.

592: This implies the attraction (\ref{attraction}).

593: 

594: 

595: \begin{remark}

596: The global attraction

597: (\ref{asymptotics}), (\ref{asymptotics-a})

598: for $\mathbf{U}(1)$-invariant equations

599: suggests the corresponding extension

600: to general $\mathbf{G}$-invariant equations

601: ($\mathbf{G}$ being the Lie group):

602: \begin{equation}\label{asymptotics-g}

603: \psi(x,t)

604: \sim

605: \psi\sb\pm(x,t)=e\sp{\bm\Omega\sb{\pm}t}\phi\sb\pm(x),

606: \qquad

607: t\to\pm\infty,

608: \end{equation}

609: where

610: $\bm\Omega\sb{\pm}$

611: belong to the corresponding Lie algebra

612: and

613: $e\sp{\bm\Omega\sb{\pm}t}$

614: are corresponding one-parameter subgroups.

615: Respectively,

616: the global attractor would consist of the solitary waves

617: (\ref{asymptotics-g}).

618: In particular, for the unitary group $\mathbf{G}=\mathbf{SU}(3)$,

619: the asymptotics (\ref{asymptotics-g})

620: relate the ``quantum stationary states'' to the structure of the

621: corresponding Lie algebra $\mathbf{su}(3)$.

622: On a seemingly related note,

623: let us mention that according to Gell-Mann -- Ne'eman theory \cite{GN64}

624: there is a correspondence

625: between the Lie algebras

626: and the classification of the elementary particles

627: which are the ``quantum stationary states''.

628: The correspondence

629: has been confirmed

630: experimentally by the discovery of the omega-minus Hyperon.

631: \end{remark}

632: 

633: \bigskip

634: The plan of the paper is as follows.

635: In Section~\ref{sect-results} we state the main assumptions and results.

636: Section~\ref{sect-splitting}

637: describes the exclusion of dispersive components

638: from the solution.

639: In Section~\ref{sect-spectral} we state the spectral properties of all

640: omega-limit trajectories

641: and apply the Titchmarsh Convolution Theorem.

642: For completeness,

643: we also give the exhaustive treatment of the linear case,

644: when

645: $F(\psi)=a\psi$ with $a\in\R$;

646: See Section~\ref{sect-linear-case}.

647: In Appendix~\ref{sect-solitons}, we collect the properties

648: of the solitary waves.

649: In Appendix~\ref{sect-quasimeasures}, we describe properties

650: of quasimeasures and corresponding multiplicators.

651: The global well-posedness of

652: equation (\ref{KG-0}) in $H\sp 1(\R)$

653: is proved in Appendix~\ref{sect-existence}.

654: 

655: \section{Main results}

656: \label{sect-results}

657: 

658: \subsection*{Model}

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

660: with the nonlinearity concentrated at a point:

661: \begin{equation}\label{KG}

662: \left\{

663: \begin{array}{l}

664: \ddot\psi(x,t)

665: =\psi''(x,t)-m^2\psi(x,t)+\delta(x)F(\psi(0,t)),

666: \qquad

667: x\in\R,

668: \quad t\in\R,

669: \\

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

671: \qquad

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

673: \end{array}\right.

674: \end{equation}

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

676: with the two-dimensional vector

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

678: then, physically, equation (\ref{KG}) describes small crosswise

679: oscillations of the infinite

680: string in three-dimensional space

681: $(x,u,v)$

682: stretched along the $x$-axis.

683: The string is subject to

684: the action of

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

686: coupled to a nonlinear oscillator

687: of force $F(\psi)$

688: attached at the point $x=0$.

689: 

690: 

691: We define

692: $\Psi(t)=

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

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

695: and write the Cauchy problem

696: (\ref{KG})

697: in the vector form:

698: \begin{equation}\label{KG-cp}

699: \dot\Psi(t)

700: =

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

702: \Psi(t)

703: +

704: \delta(x)\left[\begin{array}{c}0\\F(\psi)\end{array}\right],

705: \qquad

706: \Psi\at{t=0}

707: =\Psi\sb 0

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

709: \end{equation}

710: We will assume that

711: the oscillator force $F$ admits a real-valued potential:

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

713: F(\psi)=-\nabla U(\psi),\quad\psi\in\C,

714: \qquad

715: U\in C\sp 2(\C),

716: \end{equation}

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

718: Then equation (\ref{KG-cp})

719: formally can be written as a Hamiltonian system,

720: \[

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

722: \qquad

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

724: \]

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

726: derivative of the Hamilton functional

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

728: \mathcal{H}(\Psi)

729: =\frac 1 2

730: \int\limits\sb\R

731: \left(

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

733: \right)

734: dx

735: +U(\psi(0)),

736: \quad

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

738: \end{equation}

739: We

740: assume that the potential $U(\psi)$ is $\mathbf{U}(1)$-invariant,

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

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

743: Namely, we assume that

744: there exists $u\in C\sp 2(\R)$ such that

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

746: U(\psi)=u(\abs{\psi}\sp 2),

747: \qquad\psi\in\C.

748: \end{equation}

749: 

750: \begin{remark}

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

752: that we described after (\ref{KG}),

753: the potential $U(\psi)$

754: is rotation invariant with respect to the $x$-axis.

755: \end{remark}

756: 

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

758: imply that

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

760: F(\psi)=\alpha(\abs{\psi}^2)\psi,

761: \qquad\psi\in\C,

762: \end{equation}

763: where

764: $\alpha(\cdot)=-2 u'(\cdot)\in C\sp 1(\R)$

765: %%AC changed sign.

766: is real-valued.

767: Therefore,

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

769: F(e\sp{i\theta}\psi)=e\sp{i\theta} F(\psi),

770: \qquad\theta\in\R,\quad\psi\in\C.

771: \end{equation}

772: Then the N\"other theorem formally

773: implies that

774: the functional

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

776: \mathcal{Q}(\Psi)=\frac i2

777: \int\sb\R

778: \left(\overline\psi\pi-\overline\pi\psi\right)\,dx,

779: \qquad

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

781: \end{equation}

782: is

783: conserved for solutions $\Psi(t)$ to (\ref{KG-cp}).

784: 

785: We introduce

786: the phase space ${\E}$

787: of finite energy states for equation (\ref{KG-cp}).

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

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

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

791: for $R>0$.

792: 

793: \begin{definition}

794: \begin{enumerate}

795: \item

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

797: $\Psi

798: =

799: \left[\begin{array}{c}\psi(x)

800: \\

801: \pi(x)

802: \end{array}\right]$,

803: with the norm

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

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

806: :=

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

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

809: \end{equation}

810: \item

811: ${\E}\sb F$ is the space ${\E}$ endowed

812: with the Fr\'echet topology defined by the seminorms

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

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

815: :=

816: \norm{ \pi}\sb{L\sp 2\sb R}^2

817: +

818: \norm{\psi'}\sb{L\sp 2\sb R}^2+m^2\norm{\psi}\sb{L\sp 2\sb R}^2,

819: \qquad

820: R>0.

821: \end{equation}

822: \end{enumerate}

823: \end{definition}

824: 

825: The

826: equation (\ref{KG-cp})

827: is formally a Hamiltonian system with

828: the phase space ${\E}$

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

830: Both

831: $\mathcal{H}$

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

833: Let us note that

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

835: where $H\sp 1$ denotes the Sobolev space

836: \[

837: H\sp 1=H\sp 1(\R)

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

839: \]

840: We introduced into (\ref{def-e}), (\ref{def-e-r})

841: the factor $m^2>0$;

842: This provides the convenient relation

843: $\mathcal{H}(\Psi)=\frac 1 2\norm{\Psi}\sb{\E}^2+U(\psi(0))$.

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

845: 

846: \subsection*{Global well-posedness}

847: 

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

849: well-posedness, we assume that

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

851: U(\psi)\ge {A}-{B}\abs{\psi}^2

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

853: {\rm where}\ {A}\in\R\ {\rm and}\ 0\le {B}<m.

854: \end{equation}

855: 

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

857: Let $F(\psi)$ satisfy conditions (\ref{P}) and (\ref{inv-u}):

858: \[

859: F(\psi)=-\nabla U(\psi),\qquad

860: U(\psi)=u(\abs{\psi}^2),

861: \qquad u(\cdot)\in C\sp 2(\R).

862: \]

863: Additionally,

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

865: Then:

866: \begin{enumerate}

867: \item

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

869: (\ref{KG-cp}) has a unique solution

870: $\Psi(t)\in C(\R,{\E})$.

871: \item

872: The map

873: $W(t):\;\Psi\sb 0\mapsto \Psi(t)$

874: is continuous in ${\E}$ and ${\E}\sb F$

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

876: \item

877: The energy is conserved:

878: \begin{equation}\label{ec}

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

880: \quad

881: t\in\R.

882: \end{equation}

883: \item

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

885: \begin{equation}\label{eb}

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

887: \le C(\Psi\sb 0),

888: \qquad t\in\R.

889: \end{equation}

890: \end{enumerate}

891: \end{theorem}

892: 

893: We prove this theorem in Appendix~\ref{sect-existence}.

894: 

895: \begin{remark}

896: The value of the charge is also conserved:

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

898: $\ t\in\R$.

899: \end{remark}

900: 

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

902: 

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

904: \begin{enumerate}

905: \item

906: The solitary waves of equation (\ref{KG-cp})

907: are solutions of the form

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

909: \Psi(t)=\Phi\sb\omega e\sp{-i\omega t},

910: \ {\rm where}\

911: \omega\in\C,

912: \ \Phi\sb\omega

913: =

914: \left[\!\begin{array}{c}\phi\sb\omega

915: \\

916: -i\omega\phi\sb\omega

917: \end{array}

918: \!\right],

919: \ \phi\sb\omega\in H\sp 1(\R).

920: \end{equation}

921: \item

922: The solitary manifold

923: is the set

924: $

925: \bS=

926: \left\{

927: \Phi\sb\omega\sothat\omega\in\C

928: \right\}

929: $

930: of all amplitudes $\Phi\sb\omega$.

931: \end{enumerate}

932: \end{definition}

933: 

934: Identity (\ref{inv-f}) implies that the set $\bS$

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

936: $\theta\in\R$.

937: Let us note that since $F(0)=0$ by (\ref{def-a}),

938: for any $\omega\in\C$

939: there

940: is a zero solitary wave with

941: $\phi\sb\omega(x)\equiv 0$.

942: 

943: Note that, according to (\ref{def-a}),

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

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

946: We will need to distinguish the cases

947: when $F$ is linear and nonlinear; for this,

948: we introduce the following definition.

949: 

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

951: The function $F(\psi)$ is \emph{strictly nonlinear}

952: if the equation

953: $\alpha(C^2)=a$

954: has a discrete

955: (or empty)

956: set of positive roots

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

958: \end{definition}

959: 

960: \begin{lemma}\label{lemma-omega-real}

961: If $F(\psi)$ is strictly nonlinear

962: in the sense of Definition~\ref{def-sn},

963: then

964: nonzero solitary waves

965: exist only for

966: $\omega\in\R$.

967: \end{lemma}

968: 

969: We prove this Lemma

970: in Appendix~\ref{sect-solitons}.

971: 

972: \begin{proposition}[Existence of solitary waves]

973: \label{prop-solitons}

974: Assume that $F(\psi)$ satisfies (\ref{inv-f}) and

975: that one of two following conditions holds:

976: \begin{enumerate}

977: \item

978: $F(\psi)$  is strictly nonlinear

979: in the sense of Definition~\ref{def-sn};

980: \item

981: $F(\psi)=a\psi$ with $a\in\R$.

982: \end{enumerate}

983: Then all nonzero solitary wave solutions

984: to {\rm (\ref{KG-cp})}

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

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

987: \phi\sb\omega(x)=C e^{-\kappa\abs{x}},

988: \end{equation}

989: where $\kappa>0$, $\omega\in\C$, and $C\in\C\backslash 0$

990: satisfy the following relations:

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

992: \alpha(\abs{C}^2)=2\kappa,

993: \qquad

994: \kappa^2=m^2-\omega^2.

995: \end{equation}

996: Additionally, if $F(\psi)$ is strictly nonlinear,

997: then $\omega\in(-m,m)$.

998: \end{proposition}

999: We prove this Proposition

1000: in Appendix~\ref{sect-solitons}.

1001: 

1002: \begin{remark}

1003: Let us denote

1004: $\kappa\sb C={\alpha(\abs{C}^2)}/{2}$ and

1005: $\omega\sb C\sp\pm=\pm\sqrt{m^2-\kappa\sb C\sp 2}$ for

1006: $C \in\C$.

1007: Then the relation  (\ref{kaka})

1008: demonstrates that the set of all solitary waves

1009: can be parametrized as follows:

1010: \begin{enumerate}

1011: \item

1012: When $F(\psi)$ is strictly nonlinear,

1013: in the sense of Definition~\ref{def-sn},

1014: the profile function

1015: $\phi\sb C(x)=C e^{i\theta}e^{-\kappa\sb C\abs{x}}$,

1016: with $C\ge 0$ and

1017: $\theta\in [0,2\pi]$,

1018: corresponds to the solitary waves

1019: with $\omega=\omega\sb C\sp\pm$

1020: as long as $\kappa\sb C\in (0,m]$

1021: (so that $\phi\sb C\in H\sp 1$

1022: and $\omega$ is real in agreement with

1023: Lemma~\ref{lemma-omega-real}).

1024: 

1025: \item

1026: When $F(\psi)=a\psi$ with $a\in\R$,

1027: we see from (\ref{kaka})

1028: that $\kappa\sb C=a/2$

1029: is constant.

1030: If $a>0$,

1031: the profile function

1032: $\phi\sb C(x)=C e^{-a\abs{x}/2}$, with $C\in\C$,

1033: corresponds to the solitary waves

1034: with

1035: $\omega=\pm\sqrt{m^2-\frac{a^2}{4}}$.

1036: The restriction $\kappa\sb C\in (0,m]$

1037: no longer applies

1038: since the value of $\omega$ may be imaginary.

1039: (This is different from the case of strictly nonlinear $F$,

1040: when imaginary values of $\omega$

1041: are prohibited by Lemma~\ref{lemma-omega-real}.)

1042: If $a\le 0$,

1043: then there is only the zero solitary wave solution.

1044: \end{enumerate}

1045: 

1046: \end{remark}

1047: 

1048: As we mentioned before,

1049: we need to assume that the nonlinearity is polynomial.

1050: This assumption

1051: is crucial in our argument:

1052: It will allow to apply the Titchmarsh convolution theorem.

1053: Now all our assumptions on $F$

1054: can be summarized as follows.

1055: 

1056: \begin{assumption}\label{assumption-a}

1057: \begin{equation}\label{f-is-such}

1058: F(\psi)=-\nabla U(\psi),

1059: \qquad

1060: U(\psi)=\sum\limits\sb{n=0}\sp{N}u\sb n\abs{\psi}\sp{2n},

1061: \end{equation}

1062: where

1063: $\ u\sb n\in\R$,

1064: $\ u\sb N>0$,

1065: $\ N\ge 2$.

1066: \end{assumption}

1067: 

1068: This Assumption guarantees that the nonlinearity

1069: $F$ satisfies

1070: (\ref{P}) and (\ref{inv-u}),

1071: and also the bound (\ref{bound-below})

1072: from Theorem~\ref{theorem-well-posedness}.

1073: Moreover, Assumption~\ref{assumption-a}

1074: implies that $F$

1075: is strictly nonlinear in the sense of Definition~\ref{def-sn}.

1076: By Lemma~\ref{lemma-omega-real},

1077: this in turn implies that

1078: all nonzero solitary waves correspond to $\omega\in\R$.

1079: 

1080: Our main result is the following theorem.

1081: 

1082: \begin{theorem}[Main Theorem]

1083: \label{main-theorem}

1084: Let the nonlinearity $F(\psi)$

1085: satisfy Assumption~\ref{assumption-a}.

1086: Then for any $\Psi\sb 0\in \E$

1087: the solution $\Psi(t)\in C(\R,{\E})$

1088: to the Cauchy problem {\rm (\ref{KG-cp})}

1089: with $\Psi(0)=\Psi\sb 0$

1090: converges to $\bS$ in the space $\E\sb{F}$:

1091: \begin{equation}\label{cal-A}

1092: \Psi(t)\toEF \bS,

1093: \quad t\to \pm\infty.

1094: \end{equation}

1095: \end{theorem}

1096: 

1097: Let us note that

1098: the convergence to the set $\bS$

1099: in the space $\E\sb{F}$ is equivalent to

1100: \begin{equation}\label{metr}

1101: \lim\sb{t\to\pm\infty}\rho(\Psi(t),\bS)=0,

1102: \end{equation}

1103: where $\rho$ is a

1104: metric in the space $\E\sb{F}$

1105: and

1106: $\rho(\Psi(t),\bS):=\inf\limits\sb{\Phi\in \bS}\rho(\Psi(t),\Phi)$.

1107: 

1108: Let us also give the corresponding result for the linear case,

1109: when $F(\psi)=a\psi$ with $a\in\R$.

1110: We restrict our consideration to the case when

1111: $a<2m$.

1112: It is in this case that condition (\ref{bound-below}) is satisfied.

1113: We do not consider the case

1114: $a\ge 2m$, since in this case the solutions

1115: are generally not bounded in $\E$-norm

1116: (see Remark~\ref{remark-solitons-linear}),

1117: while our arguments rely significantly on the bounds (\ref{eb}).

1118: This case will be considered in more detail elsewhere.

1119: 

1120: \begin{theorem}[Linear case]

1121: \label{main-theorem-linear}

1122: Assume that $F(\psi)=a\psi$, where $a<2m$.

1123: Then for any $\Psi\sb 0\in \E$

1124: the solution $\Psi(t)\in C(\R,{\E})$

1125: to the Cauchy problem {\rm (\ref{KG-cp})}

1126: with $\Psi(0)=\Psi\sb 0$

1127: converges in the space $\E\sb{F}$

1128: to the linear span of $\bS$,

1129: which we denote by $\langle\bS\rangle$:

1130: \begin{equation}\label{cal-A-linear}

1131: \Psi(t)\toEF\langle\bS\rangle,

1132: \quad t\to \pm\infty.

1133: \end{equation}

1134: 

1135: \end{theorem}

1136: 

1137: 

1138: \begin{remark}

1139: 

1140: In Section~\ref{sect-linear-case}

1141: we will show that:

1142: \begin{enumerate}

1143: \item

1144: If $0<a<2m$,

1145: then

1146: $\langle\bS\rangle\ne\bS$.

1147: Particular solutions

1148: show that the attraction {\rm (\ref{cal-A})}

1149: does not hold in general

1150: (see Remark~\ref{noatt})

1151: and has to be substituted by {\rm (\ref{cal-A-linear})}.

1152: \item

1153: If $a\le 0$,

1154: $\langle\bS\rangle=\bS=\{0\}$.

1155: 

1156: \end{enumerate}

1157: \end{remark}

1158: 

1159: \subsection*{Strategy of the proof}

1160: For $m=0$

1161: the global attraction of type {\rm (\ref{cal-A})}

1162: is proved in \cite{MR1359949},

1163: where the proof was based

1164: on the direct calculation of the energy radiation

1165: for the wave equation.

1166: For the Klein-Gordon equation with $m>0$,

1167: the dispersive relation $\omega\sp 2=k\sp 2+m^2$

1168: results in the group velocities $v=\omega'(k)=k/\sqrt{k^2+m^2}$,

1169: so every velocity $0\le\abs{v}<1$ is possible.

1170: This complicates considerably the investigation of the energy propagation,

1171: so

1172: the approach \cite{MR1359949}

1173: built on the fact that the group velocity was $\abs{v}=1$

1174: no longer works.

1175: To overcome this difficulty,

1176: we introduce a new approach based on the

1177: nonlinear spectral analysis

1178: of the solution.

1179: 

1180: We prove the absolute continuity

1181: of the spectrum of the solution  for $\abs{\omega}>m$.

1182: This observation is similar to

1183: the well-known Kato Theorem.

1184: The proof is not obvious

1185: and relies on the complex Fourier-Laplace transform and

1186: the Wiener-Paley arguments.

1187: 

1188: We then split the solution into two components: Dispersive and bound,

1189: with the frequencies  $\abs{\omega}>m$ and $\omega\in [-m,m]$,

1190: respectively.

1191: The dispersive component is an oscillatory integral of plane waves,

1192: while the bound component

1193: is a superposition of exponentially decaying functions.

1194: The stationary phase argument leads to a local decay of

1195: the dispersive component,

1196: due to the absolute continuity of its spectrum.

1197: This reduces the long-time behavior of the solution to the

1198: behavior of the bound component.

1199: 

1200: Next, we establish the spectral representation

1201: for the bound component. For this, we need to know

1202: an optimal regularity of the corresponding spectral measure;

1203: We have found out

1204: that the spectral measure belongs to the space of

1205: \emph{quasimeasures}

1206: which are Fourier transforms of

1207: bounded continuous functions,

1208: \cite{Ga66}.

1209: The spectral representation

1210: implies compactness in the space of quasimeasures,

1211: which in turn leads to the existence

1212: of \emph{omega-limit trajectories}  for $t\to \infty$.

1213: 

1214: Further,  we prove that

1215: an omega-limit trajectory itself

1216: satisfies the nonlinear equation (\ref{KG-0}),

1217: and this implies

1218: the crucial spectral inclusion:

1219: The spectrum of the nonlinear term

1220: is included in the spectrum of the omega-limit trajectory.

1221: We then reduce the spectrum of this limiting trajectory

1222: to a single harmonic $\omega\sb{+}\in[-m,m]$

1223: using the Titchmarsh convolution theorem

1224: \cite{titchmarsh} (see also \cite[p.119]{MR1400006}

1225: and \cite[Theorem 4.3.3]{MR1065136}).

1226: In turn, this means that any omega-limit trajectory

1227: lies in the manifold $\bS$ of the solitary waves,

1228: which proves that $\bS$ is the global attractor.

1229: 

1230: Empirically,

1231: the last part of our argument is a contemplation of

1232: the radiative mechanism based on the

1233: \emph{inflation of spectrum}

1234: by the nonlinearity: A low-frequency perturbation

1235: of the stationary state does not radiate

1236: the energy until it generates (via a nonlinearity) ``a spectral line''

1237: embedded in

1238: the continuous spectrum outside $[-m,m]$.

1239: This embedded spectral line

1240: gives rise to the

1241: wave packets

1242: which bring the energy to infinity.

1243: This radiative mechanism has been originally observed

1244: in the numerical experiments with the nonlinear relativistic

1245: Ginzburg-Landau equation (see \cite{MR2133795}).

1246: The spectral inclusion for the  omega-limit trajectories

1247: expresses their {\it nonradiative nature}:

1248: The limiting trajectory cannot radiate

1249: since the initial energy was bounded.

1250: 

1251: \section{Separation of dispersive components}

1252: \label{sect-splitting}

1253: 

1254: It suffices to prove Theorem~\ref{main-theorem}

1255: for $t\to+\infty$;

1256: We will only consider the solution $\psi(x,t)$

1257: restricted to $t\ge 0$.

1258: In this section we

1259: eliminate two

1260: dispersive components from $\psi(x,t)$.

1261: 

1262: 

1263: \subsection*{First dispersive component}

1264: Let us split the solution

1265: $\Psi(t)

1266: =\left[\!\scriptsize{\begin{array}{c}\psi(x,t)\\\pi(x,t)\end{array}}\!\right]

1267: $

1268: into

1269: $\Psi(t)=\Psi\sb 1(t)+\Psi\sb 2(t)$,

1270: where

1271: $

1272: \Psi\sb 1(t)

1273: =\left[\!\scriptsize{

1274: \begin{array}{c}\psi\sb 1(x,t)\\\pi\sb 1(x,t)\end{array}}\!\right]

1275: $

1276: and

1277: $

1278: \Psi\sb 2(t)

1279: =\left[\!\scriptsize{

1280: \begin{array}{c}\psi\sb 2(x,t)\\\pi\sb 2(x,t)\end{array}}\!\right]

1281: $

1282: are defined for $t\ge 0$

1283: as solutions to the following Cauchy problems:

1284: \begin{eqnarray}

1285: &&

1286: \dot\Psi\sb 1(t)

1287: =

1288: \left[\begin{array}{cc}0&1\\\p\sb x\sp 2-m^2&0\end{array}\right]

1289: \Psi\sb 1(t),

1290: \qquad

1291: \Psi\sb 1\at{t=0}

1292: =\Psi\sb 0,

1293: \label{KG-cp-1}

1294: \\

1295: \\

1296: &&

1297: \dot\Psi\sb 2(t)

1298: =

1299: \left[\begin{array}{cc}0&1\\\p\sb x\sp 2-m^2&0\end{array}\right]

1300: \Psi\sb 2(t)

1301: +

1302: \delta(x)\left[\begin{array}{c}0\\f(t)\end{array}\right],

1303: \qquad

1304: \Psi\sb 2\at{t=0}=0,

1305: \label{KG-cp-2}

1306: \end{eqnarray}

1307: where

1308: $\Psi\sb 0

1309: =\left[\!\scriptsize{\begin{array}{c}\psi\sb 0\\\pi\sb 0\end{array}}\!\right]$

1310: is the initial data from

1311: (\ref{KG-cp}),

1312: and

1313: \begin{equation}\label{def-f}

1314: f(t):=F(\psi(0,t)),

1315: \qquad

1316: t\ge 0.

1317: \end{equation}

1318: Note that $\psi(0,\cdot)\in C\sb{b}(\Bar{\R\sp{+}})

1319: $

1320: by the Sobolev embedding

1321: since $\Psi\in C\sb{b}(\Bar{\R\sp{+}}, \E)$

1322: by Theorem~\ref{theorem-well-posedness}~({\it iv}).

1323: Hence,  $f(\cdot)\in C\sb{b}(\Bar{\R\sp{+}})$.

1324: On the other hand,

1325: since $\Psi\sb 1(t)$

1326: is a finite energy solution to the free Klein-Gordon equation,

1327: we also have

1328: \begin{equation}\label{psi-1-bounds}

1329: \Psi\sb 1

1330: \in C\sb{b}(\Bar{\R\sp{+}}, \E).

1331: \end{equation}

1332: Hence, the function $\Psi\sb 2(t)=\Psi(t)-\Psi\sb 1(t)$

1333: also satisfies

1334: \begin{equation}\label{psi-2-bounds}

1335: \Psi\sb 2

1336: \in C\sb{b}(\Bar{\R\sp{+}}, \E).

1337: \end{equation}

1338: 

1339: \begin{lemma}\label{lemma-decay-psi1}

1340: There is a local decay of $\Psi\sb 1$

1341:  in ${\E}\sb F$ seminorms.

1342: That is,

1343: $\forall R>0$,

1344: \begin{equation}\label{dp0}

1345: \Norm{

1346: \Psi\sb 1(t)

1347: }\sb{{\E},R}

1348: \to 0,

1349: \qquad t\to\infty.

1350: \end{equation}

1351: \end{lemma}

1352: \begin{proof}

1353: We have to prove that

1354: \begin{equation}

1355: \norm{\Psi\sb 1(t)}\sb{\E,R}^2

1356: =

1357: \int\limits\sb{\abs{x}<R}\Big( \abs{\pi\sb 1(x,t)}\sp 2

1358: +\abs{\psi\sb 1'(x,t)}\sp 2+m^2\abs{\psi\sb 1(x,t)}\sp 2\Big)\,dx

1359: \end{equation}

1360: goes to zero as $t$ tends to infinity.

1361: Fix a cutoff function $\zeta(x)\in C\sb 0\sp\infty(\R)$

1362: with $\zeta(x)=1$, $\abs{x}\le 1$ and $\zeta(x)=0$, $\abs{x}\ge 2$.

1363: For $r>0$,

1364: let $\zeta\sb r(x)=\zeta(x/r)$.

1365: Denote by $\Phi\sb{r}(t)$ and $\Theta\sb{r}(t)$

1366: the solutions to the free

1367: Klein-Gordon equation with the initial data

1368: $\zeta\sb r\Psi\sb 0$

1369: and

1370: $(1-\zeta\sb r)\Psi\sb 0$,

1371: respectively,

1372: so that $\Psi\sb 1(t)=\Phi\sb{r}(t)+\Theta\sb{r}(t)$.

1373: Then there exists $C\sb r>0$ that depends on $r$

1374: so that

1375: $\norm{\Phi\sb{r}(t)}\sb{\E,R}^2

1376: \le C\sb r (1+t)\sp{-1}$

1377: for $t>0$,

1378: since the solution

1379: $\Phi\sb{r}$ is represented by the integral with the Green function,

1380: which is the Bessel function decaying like

1381: $(1+t)\sp{-1/2}$ (see e.g. \cite[$(2.7')$, Chapter I]{MR1364201}).

1382: We then have:

1383: \begin{equation}\label{ert}

1384: \norm{\Psi\sb 1(t)}\sb{\E,R}^2

1385: \le C\sb r(1+t)^{-1}

1386: +

1387: C

1388: \norm{\Theta\sb{r}(t)}\sb{\E,R}^2

1389: \end{equation}

1390: where $r>0$ could be arbitrary.

1391: To conclude that the left-hand side of (\ref{ert})

1392: goes to zero,

1393: it remains to note that

1394: \begin{equation}\label{rec}

1395: \norm{\Theta\sb{r}(t)}\sb{\E,R}

1396: \le

1397: \norm{\Theta\sb{r}(t)}\sb{\E}

1398: =

1399: \norm{\Theta\sb{r}(0)}\sb{\E}

1400: \end{equation}

1401: where the last relation

1402: is

1403: due to the energy conservation for the free Klein-Gordon equation,

1404: and that the right-hand side of (\ref{rec})

1405: could be made arbitrarily small

1406: if $r>0$ is taken sufficiently large.

1407: \qed\end{proof}

1408: 

1409: \subsection*{Complex Fourier-Laplace transform}

1410: 

1411: Let us analyze the complex Fourier-Laplace transform of

1412: $\psi\sb 2(x,t)$:

1413: \begin{equation}\label{FL}

1414: \displaystyle

1415: \tilde\psi\sb 2(x,\omega)

1416: =\mathcal{F}\sp{+}\sb{t\to\omega}[\psi\sb 2(x,\cdot)]

1417: :=

1418: \int\sb 0\sp\infty e\sp{i\omega t}\psi\sb 2(x,t)\,dt,

1419: \quad\omega\in\C\sp{+},

1420: \end{equation}

1421: where

1422: $C\sp{+}:=\{z\in\C:\;\Im z>0\}$.

1423: Due to (\ref{psi-2-bounds}),

1424: $\tilde\psi\sb 2(\cdot,\omega)$

1425: is an $H\sp 1$-valued analytic function of $\omega\in\C\sp{+}$.

1426: Equation (\ref{KG-cp-2})

1427: for $\psi\sb 2$ implies that

1428: \[

1429: -\omega\sp 2\tilde\psi\sb 2(x,\omega)

1430: =

1431: \tilde\psi\sb 2''(x,\omega)-m^2\tilde\psi\sb 2(x,\omega)

1432: +\delta(x)\tilde f(\omega),\quad \omega\in\C\sp{+}.

1433: \]

1434: Hence,

1435: the solution $\psi\sb 2(x,\omega)$ is a linear combination

1436: of the fundamental solutions

1437: which satisfy

1438: \[

1439: G\sb\pm''(x,\omega)+(\omega^2-m^2)G\sb\pm(x,\omega)

1440: =\delta(x),

1441: \qquad \omega\in\C\sp{+}.

1442: \]

1443: These solutions are given by

1444: $\displaystyle

1445: G\sb\pm(x,\omega)=\frac{e\sp{\pm ik(\omega)\abs{x}}}{\pm 2ik(\omega)}$,

1446: where $k(\omega)$ stands for the analytic function

1447: \begin{equation}\label{def-k}

1448: k(\omega)=\sqrt{\omega\sp 2-m^2},

1449: \qquad\Im k(\omega)>0,

1450: \qquad

1451: \omega\in\C\sp{+},

1452: \end{equation}

1453: which we extend

1454: to $\omega\in\Bar{\C\sp{+}}$

1455: by continuity.

1456: We use the standard ``limiting absorption principle''

1457: for the selection of the fundamental solution: Since

1458: $\tilde\psi\sb 2(\cdot,\omega)\in H\sp 1$

1459: for $\omega\in\C\sp{+}$,

1460: only $G\sb{+}$ is appropriate,

1461: because for $\omega\in\C\sp{+}$ the function $G\sb{+}(\cdot,\omega)$

1462: is in $H\sp 1$ while $G\sb{-}$ is not.

1463: Thus,

1464: \begin{equation}\label{tilde-psi-tilde-f}

1465: \tilde\psi\sb 2(x,\omega)=-\tilde f(\omega)G\sb{+}(x,\omega)

1466: =-\tilde f(\omega)\frac{e\sp{i k(\omega)\abs{x}}}{2ik(\omega)},

1467: \qquad\omega\in\C\sp{+}.

1468: \end{equation}

1469: Define

1470: $\tilde z(\omega):=\mathcal{F}\sb{t\to\omega}\sp{+}[z(t)]$,

1471: with

1472: $z(t):=\psi\sb 2(0,t)$.

1473: Then

1474: $\tilde z(\omega)=-\tilde f(\omega)/(2ik(\omega))$, and

1475: (\ref{tilde-psi-tilde-f}) becomes

1476: \begin{equation}\label{p1}

1477: \tilde\psi\sb 2(x,\omega)=\tilde z(\omega)e\sp{ik(\omega)\abs{x}},

1478: \qquad\omega\in\C\sp{+}.

1479: \end{equation}

1480: Let us extend $\psi\sb 2(x,t)$ and $f(t)$ by zero for $t<0$:

1481: \begin{equation}\label{zn}

1482: \psi\sb 2(x,t)=0

1483: \quad

1484: {\rm and}

1485: \quad

1486: f(t)=0

1487: \quad

1488: {\rm for}\quad t<0.

1489: \end{equation}

1490: Then

1491: \begin{equation}\label{psi2}

1492: \psi\sb 2\in C\sb{b}(\R, H^1)

1493: \end{equation}

1494: by (\ref{psi-2-bounds}) since $\psi\sb 2(x,0+)=0$ by

1495: initial conditions in (\ref{KG-cp-2}).

1496: The Fourier transform

1497: $\hat\psi\sb 2(\cdot,\omega):=\mathcal{F}\sb{t\to\omega}[\psi\sb 2(\cdot,t)]$

1498: is a tempered $H\sp 1$-valued distribution of $\omega\in\R$

1499: by (\ref{psi-2-bounds}).

1500: The distribution $\hat\psi\sb 2(\cdot,\omega)$

1501: is the boundary value of

1502: the

1503: analytic function $\tilde \psi\sb 2(\cdot,\omega)$,

1504: in the following sense:

1505: \begin{equation}\label{bvp1}

1506: \hat\psi\sb 2(\cdot,\omega)

1507: =\lim\limits\sb{\varepsilon\to 0+}\tilde\psi\sb 2(\cdot,\omega+i\varepsilon),

1508: \qquad\omega\in\R,

1509: \end{equation}

1510: where the convergence

1511: is in the space of

1512: tempered distributions $\mathscr{S}'(\R\sb\omega,H\sp 1)$.

1513: Indeed,

1514: $\tilde\psi\sb 2(\cdot,\omega+i\varepsilon)

1515: =\mathcal{F}\sb{t\to\omega}[\psi\sb 2(\cdot,t)e\sp{-\varepsilon t}]$

1516: and

1517: $\psi\sb 2(\cdot,t)e\sp{-\varepsilon t}

1518: \mathop{\longrightarrow}\limits\sb{\varepsilon\to 0+}

1519: \psi\sb 2(\cdot,t)$

1520: where the convergence holds in $\mathscr{S}'(\R\sb t,H\sp 1)$ by (\ref{zn}).

1521: Therefore, (\ref{bvp1}) holds by the continuity of the Fourier transform

1522: $\mathcal{F}\sb{t\to\omega}$ in $\mathscr{S}'(\R)$.

1523: 

1524: Similarly to (\ref{bvp1}),

1525: the distributions   $\hat z(\omega)$ and $\hat f(\omega)$,

1526: $\omega\in\R$,

1527: are the boundary values of the analytic in $\C\sp{+}$

1528: functions $\tilde f(\omega)$ and $\tilde z(\omega)$,

1529: $\omega\in\C\sp{+}$, respectively:

1530: \begin{equation}\label{bv}

1531: \hat z(\omega)=\lim\limits\sb{\varepsilon\to 0+}

1532: \tilde z(\omega+i\varepsilon),

1533: \qquad

1534: \hat f(\omega)=\lim\limits\sb{\varepsilon\to 0+}

1535: \tilde f(\omega+i\varepsilon), \quad \omega\in\R,

1536: \end{equation}

1537: since the functions

1538: $z(t)$ and $f(t)$ are bounded for $t\ge 0$ and zeros for

1539: $t<0$.

1540: The convergence holds in the space of tempered distributions

1541: $\mathscr{S}'(\R)$.

1542: 

1543: Let us justify that the representation (\ref{p1})

1544: for $\hat\psi\sb 2(x,\omega)$

1545: is also valid when $\omega\in\R$

1546: if the multiplication in (\ref{p1})

1547: is understood in the sense of quasimeasures

1548: (see Appendix~\ref{sect-quasimeasures}).

1549: \begin{proposition}\label{prop-uniform}

1550: For any fixed $x\in\R$, the identity

1551: \begin{equation}\label{p1r}

1552: \hat\psi\sb 2(x,\omega)=

1553: \hat z(\omega)e\sp{ik(\omega)\abs{x}},\quad\omega\in\R,

1554: \end{equation}

1555: holds in the sense of tempered distributions.

1556: The

1557: right-hand side is defined as the product of quasimeasure

1558: $\hat z(\omega)$ by the multiplicator $e\sp{ik(\omega)\abs{x}}$.

1559: \end{proposition}

1560: 

1561: \begin{proof}

1562: The representation (\ref{p1r}) for $\omega\ne \pm m$

1563: follows from (\ref{p1}) -- (\ref{bv}) since $e\sp{ik(\omega)\abs{x}}$

1564: is a smooth function of $\omega\in\Bar{\C\sp{+}}$

1565: outside the points $\pm m$.

1566: 

1567: So, we only need to justify

1568: (\ref{p1}) in a neighborhood of each point $\omega=\pm m$.

1569: The main problem is

1570: the low regularity of $k(\omega)=\sqrt{\omega\sp 2-m^2}$

1571: at the points $\pm m$.

1572: Choose

1573: a cutoff function $\zeta(\omega)\in C\sb 0\sp\infty(\R)$

1574: such that

1575: \begin{equation}\label{cfu}

1576: \zeta\at{[-m-1,m+1]}\equiv 1.

1577: \end{equation}

1578: Let us note that $\psi\sb 2(x,\cdot)\in C\sb{b}(\R)$

1579: for each particular $x\in\R$

1580: by (\ref{psi2})

1581: and the Sobolev embedding theorem.

1582: Therefore,

1583: for each $x\in\R$,

1584: $\hat\psi\sb 2(x,\cdot)$

1585: belongs to the space $\mathcal{QM}(\R)$

1586: of quasimeasures

1587: which are defined as functions with bounded continuous

1588: Fourier transform

1589: (see Definition~\ref{def-quasimeasure}).

1590: In particular,

1591: $\hat z(\cdot)$ is a quasimeasure since it is the Fourier transform

1592: of the function $z(t):=\psi\sb 2(0,t)$.

1593: On the other hand,

1594: the function

1595: $e\sp{i k(\omega)\abs{x}}\zeta(\omega)$ is a

1596: multiplicator in $\mathcal{QM}(\R)$

1597: by Lemma~\ref{lemma-quasi-2}~({\it i})

1598: and  Lemma~\ref{lemma-quasi-1}~({\it i})

1599: (see Appendix~\ref{sect-quasimeasures}).

1600: Let us now prove that

1601: \begin{equation}\label{p1rz}

1602: \hat\psi\sb 2(x,\omega)\zeta(\omega)=

1603: \hat z(\omega)e\sp{ik(\omega)\abs{x}}\zeta(\omega),\quad\omega\in\R

1604: \end{equation}

1605: in the sense of quasimeasures.

1606: We define

1607: $\mu\sb\varepsilon(\omega):=\tilde z(\omega+i\varepsilon)

1608: =\mathcal{F}\sb{t\to\omega}[z(t)e\sp{-\varepsilon t}]$

1609: for $\varepsilon>0$.

1610: Then $\mu\sb\varepsilon(\omega)\in \mathcal{QM}(\R)$,

1611: and $\mu\sb\varepsilon(\omega)\stackrel{\mathcal{QM}}

1612: \longrightarrow \mu(\omega):= \hat z(\omega)$

1613: as $\varepsilon\to 0+$

1614: since

1615: $z(t)e\sp{-\varepsilon t}\stackrel{C\sb{b,F}}\longrightarrow z(t)$

1616: by (\ref{zn})

1617: (see Definition~\ref{dA}).

1618: 

1619: Let us denote

1620: $M\sb{x,\varepsilon}(\omega)

1621: =

1622: e\sp{i k(\omega+i\varepsilon)\abs{x}}\zeta(\omega)$

1623: for $\omega\in\R$ and $\varepsilon\ge 0$.

1624: By Lemmas~\ref{lemma-quasi-1}~({\it ii})

1625: and~\ref{lemma-quasi-2}~({\it ii}),

1626: $M\sb{x,\varepsilon}(\omega)$

1627: are multiplicators in the space of quasimeasures.

1628: This implies that

1629: \begin{equation}\label{mve'}

1630: \tilde z(\omega+i\varepsilon)

1631: e\sp{i k(\omega+i\varepsilon)\abs{x}}

1632: \zeta(\omega)

1633: \stackrel{\mathcal{QM}}\longrightarrow \hat z(\omega)

1634: e\sp{i k(\omega)\abs{x}}\zeta(\omega)

1635: \quad{\rm as}\quad \varepsilon\to 0+.

1636: \end{equation}

1637: On the other hand, the left-hand side converges to the left-hand side

1638: of (\ref{p1rz}) by

1639: (\ref{p1}) and (\ref{bvp1}).

1640: \qed\end{proof}

1641: 

1642: \subsection*{Absolutely continuous spectrum}

1643: We study the regularity of the spectral density $\hat z(\omega)$

1644: from (\ref{p1r}).

1645: Denote

1646: \begin{equation}\label{def-omega-delta}

1647: \varOmega\sb\delta

1648: :=(-\infty,-m-\delta)\cup(m+\delta,\infty),

1649: \qquad

1650: \delta\ge 0.

1651: \end{equation}

1652: Note that $\Bar{\varOmega\sb 0}=(-\infty,-m]\cup[m,\infty)$

1653: coincides with the

1654: continuous spectrum of the free Klein-Gordon equation,

1655: and the function $\omega k(\omega)$ is positive for

1656: $\omega\in\varOmega\sb{0}$.

1657: %%AC typo!! was: $\omega\in\Omega\sb{0}$.

1658: 

1659: \begin{proposition}\label{prop-continuity}

1660: The distribution $\hat z(\omega)$ is absolutely continuous

1661: for $\omega\in\varOmega\sb 0$,

1662: and

1663: $\hat z\in L\sp 1(\Bar{\varOmega\sb 0})$.

1664: Moreover,

1665: \begin{equation}\label{f2}

1666: \int\sb{\varOmega\sb 0}\abs{\hat z(\omega)}\sp 2

1667: \omega k(\omega)

1668: \,d\omega<\infty.

1669: \end{equation}

1670: \end{proposition}

1671: \begin{proof}

1672: Let us first explain the main idea of the proof.

1673: By (\ref{p1r}),

1674: the function $\psi\sb{2}(x,t)$ \emph{formally} is a

1675: {\it linear combination}

1676: of the functions $e\sp{ik\abs{x}}$ with the

1677: {\it amplitudes}

1678: $\hat z(\omega)$:

1679: \begin{equation}\label{osci}

1680: \psi\sb{2}(x,t)=\frac 1{2\pi}\int\sb\R

1681: \hat z(\omega)e\sp{ik(\omega)\abs{x}}e\sp{-i\omega t}\,d\omega,

1682: \qquad x\in\R.

1683: \end{equation}

1684: For $\omega\in\varOmega\sb 0$,

1685: the functions $e\sp{ik(\omega)\abs{x}}$ are of infinite $L^2$-norm,

1686: while $\psi\sb{2}(\cdot,t)$ is of finite $L^2$-norm.

1687: This is possible only if the amplitude is absolutely continuous

1688: in $\varOmega\sb 0$:

1689: For example, if we took $\hat z(\omega)=\delta(\omega-\omega\sb 0)$

1690: with $\omega\sb 0\in\varOmega\sb 0$,

1691: then $\psi\sb{2}(\cdot,t)$ would be of infinite $L^2$-norm.

1692: 

1693: For a rigorous proof, we use the Paley-Wiener arguments.

1694: Namely, the Parseval identity and (\ref{psi-2-bounds})

1695: imply that

1696: \begin{equation}\label{PW}

1697: \int\limits\sb\R

1698: \norm{\tilde\psi\sb 2(\cdot,\omega+i\varepsilon)}\sb{H\sp 1}\sp 2\,d\omega

1699: =2\pi\int\limits\sb 0\sp\infty e\sp{-2\varepsilon t}

1700: \norm{\psi\sb 2(\cdot,t)}\sb{H\sp 1}\sp 2\,dt

1701: \le\frac{\cnst}{\varepsilon},

1702: \quad \varepsilon>0.

1703: \end{equation}

1704: On the other hand,

1705: we can calculate the term in the left-hand side of (\ref{PW}) exactly.

1706: First, according to (\ref{p1}),

1707: \[

1708: \tilde\psi\sb 2(\cdot,\omega+i\varepsilon)

1709: =\tilde z(\omega+i\varepsilon)e^{ik(\omega+i\varepsilon)\abs{x}},

1710: \]

1711: hence (\ref{PW}) results in

1712: \begin{equation}\label{PW-1}

1713: \varepsilon

1714: \int\sb\R

1715: \abs{\tilde z(\omega+i\varepsilon)}^2

1716: \norm{

1717: e^{ik(\omega+i\varepsilon)\abs{x}}

1718: }\sb{H\sp 1}\sp 2

1719: \,d\omega

1720: \le\cnst,

1721: \qquad \varepsilon>0.

1722: \end{equation}

1723: 

1724: Here is a crucial observation about the

1725: norm of $e^{ik(\omega+i\varepsilon)\abs{x}}$.

1726: \begin{lemma}

1727: \begin{enumerate}

1728: \item

1729: For $\omega\in\R$,

1730: \begin{equation}\label{wei}

1731: \lim\sb{\varepsilon\to 0+}

1732: \varepsilon \norm{e^{ik(\omega+i\varepsilon)\abs{x}}}\sb{H\sp 1}\sp 2

1733: =n(\omega):=

1734: \left\{

1735: \begin{array}{ll}

1736: \omega k(\omega),&\abs{\omega}>m\\

1737: 0,&\abs{\omega}<m

1738: \end{array}

1739: \right.,

1740: \end{equation}

1741: where the norm in $H\sp 1$ is chosen to be

1742: $\norm{\psi}\sb{H\sp 1}

1743: =\left(\norm{\psi'}\sb{L\sp 2}^2

1744: +m^2\norm{\psi}\sb{L\sp 2}^2\right)\sp{1/2}.

1745: $

1746: \item

1747: For any $\delta>0$

1748: there exists $\varepsilon\sb\delta>0$ such that

1749: for $\abs{\omega}>m+\delta$

1750: and $\varepsilon\in(0,\varepsilon\sb\delta)$,

1751: \begin{equation}\label{n-half}

1752: \varepsilon \norm{e^{ik(\omega+i\varepsilon)\abs{x}}}\sb{H\sp 1}\sp 2

1753: \ge n(\omega)/2.

1754: \end{equation}

1755: \end{enumerate}

1756: \end{lemma}

1757: 

1758: \begin{proof}

1759: Let us compute the $H\sp 1$-norm

1760: using the Fourier space representation.

1761: Setting $k\sb\varepsilon=k(\omega+i\varepsilon)$,

1762: so that

1763: $\Im k\sb\varepsilon>0$,

1764: we get

1765: $\mathcal{F}\sb{x\to k}

1766: \left[e^{ik\sb \varepsilon\abs{x}}\right]

1767: =2ik\sb\varepsilon/(k\sb \varepsilon^2-k^2)

1768: $ for

1769: $k\in\R$.

1770: Hence,

1771: \begin{equation}\label{g-h1-norm}

1772: \norm{e^{ik\sb \varepsilon\abs{x}}}\sb{H\sp 1}\sp 2

1773: =

1774: \frac{2\abs{k\sb \varepsilon}^2}\pi

1775: \int\sb{\R}

1776: \frac{(k^2+m^2)dk}

1777: {\abs{k\sb \varepsilon^2-k^2}^2}

1778: =-4\Im\left[

1779: \frac{(k\sb \varepsilon^2+m^2)\Bar{k\sb \varepsilon}}

1780: {k\sb \varepsilon^2-\Bar{k\sb \varepsilon}^2}

1781: \right].

1782: \end{equation}

1783: The last integral

1784: is evaluated using the Cauchy theorem.

1785: Substituting the expression $k\sb\varepsilon^2=(\omega+i\varepsilon)^2-m^2$,

1786: we get:

1787: \begin{equation}\label{norma}

1788: \norm{e^{ik(\omega+i\varepsilon)\abs{x}}}\sb{H\sp 1}\sp 2

1789: =\frac{1}{\varepsilon}

1790: \Re

1791: \left[

1792: \frac{(\omega+i\varepsilon)^2\overline{k(\omega+i\varepsilon)}}{\omega}

1793: \right],

1794: \quad\varepsilon>0,

1795: \ \omega\in\R,\ \omega\ne 0.

1796: \end{equation}

1797: The relation (\ref{wei}) follows

1798: since the function $k(\omega)$ is real for $\abs{\omega}>m$,

1799: but

1800: is purely imaginary for  $\abs{\omega}<m$.

1801: 

1802: The second statement of the Lemma

1803: follows

1804: since $n(\omega)>0$ for $\abs{\omega}>m$, and

1805: $n(\omega)\sim\abs{\omega}^2$ for $\abs{\omega}\to\infty$.

1806: \qed\end{proof}

1807: 

1808: \begin{remark}

1809: Note that $n(\omega)$ in (\ref{wei})

1810: is zero for

1811: $\abs{\omega}<m$,

1812: since in that case the function

1813: $e^{ik(\omega)\abs{x}}$ decays exponentially in $x$

1814: and the $H\sp 1$-norm of

1815: $e^{ik(\omega+i\varepsilon)\abs{x}}$

1816: remains finite

1817: when $\varepsilon\to 0+$.

1818: \end{remark}

1819: 

1820: 

1821: Substituting

1822: (\ref{n-half})

1823: into (\ref{PW-1}), we get:

1824: \begin{equation}\label{fin}

1825: \int\sb{\varOmega\sb\delta}

1826: \abs{\tilde z(\omega+i\varepsilon)}\sp 2

1827: \omega k(\omega)

1828: \,d\omega

1829: \le 2C,

1830: \qquad

1831: 0<\varepsilon<\varepsilon\sb\delta,

1832: \end{equation}

1833: with the same $C$ as in (\ref{PW-1}),

1834: and the region $\varOmega\sb\delta$ defined in

1835: (\ref{def-omega-delta}).

1836: We conclude that for each $\delta>0$ the set of functions

1837: \[

1838: g\sb{\delta,\varepsilon}(\omega)

1839: =

1840: \tilde z(\omega+i\varepsilon)

1841: \abs{\omega k(\omega)}^{1/2},

1842: \qquad

1843: \varepsilon\in(0,\varepsilon\sb\delta),

1844: \]

1845: defined for $\omega\in\varOmega\sb\delta$,

1846: is bounded in the Hilbert space $L\sp 2(\varOmega\sb\delta)$,

1847: and, by the Banach Theorem, is weakly compact.

1848: The convergence of the distributions (\ref{bv})

1849: implies the following weak convergence in the Hilbert space

1850: $L\sp 2(\varOmega\sb\delta)$:

1851: \begin{equation}\label{wc}

1852: g\sb{\delta,\varepsilon}

1853: \rightharpoondown g\sb\delta,

1854: \qquad \varepsilon\to 0+,

1855: \end{equation}

1856: where the limit function

1857: $g\sb\delta(\omega)$ coincides with the distribution

1858: $

1859: \hat z(\omega)\abs{\omega k(\omega)}^{1/2}

1860: $

1861: restricted onto

1862: ${\varOmega\sb\delta}$.

1863: It remains to note that the norms

1864: of all

1865: functions $g\sb\delta$, $\delta>0$,

1866: are bounded in $L\sp 2(\varOmega\sb\delta)$ by (\ref{fin}),

1867: hence (\ref{f2}) follows.

1868: Finally, $\hat z(\omega)\in L\sp 1(\Bar{\varOmega\sb 0})$

1869: by (\ref{f2}) and the Cauchy-Schwarz inequality.

1870: \qed\end{proof}

1871: 

1872: Let us denote

1873: \begin{equation}\label{dpo}

1874: \hat z\sb{d}(\omega)

1875: :=

1876: \left\{

1877: \begin{array}{ll}

1878: \hat z(\omega),&\quad\omega\in\varOmega\sb 0,\\

1879: 0,&\quad\omega\in\R\setminus\varOmega\sb 0.

1880: \end{array}\right.

1881: \end{equation}

1882: Then, by Proposition~\ref{prop-continuity},

1883: $\hat z\sb{d}(\omega)\in L\sp 1(\R)$.

1884: 

1885: \subsection*{Second dispersive component}

1886: Proposition~\ref{prop-continuity} and

1887: the representation (\ref{osci})

1888: suggest that we introduce the function

1889: \begin{equation}\label{dd}

1890: \psi\sb{d}(x,t)

1891: =\frac 1{2\pi}\int\sb\R

1892: \hat z\sb{d}(\omega)e\sp{ik(\omega)\abs{x}}e\sp{-i\omega t}\,d\omega,

1893: \qquad x\in\R,\quad t\in\R,

1894: \end{equation}

1895: with $\hat z\sb{d}$ defined by (\ref{dpo}).

1896: The Fourier transform of

1897: $\psi\sb{d}(x,t)$

1898: is given by the formula similar to (\ref{p1r}):

1899: \begin{equation}

1900: \label{ftd}

1901: \hat\psi\sb{d}(x,\omega)

1902: =

1903: \displaystyle

1904: \hat z\sb{d}(\omega)e\sp{ik(\omega)\abs{x}},\qquad x\in\R,

1905: \qquad

1906: \omega\in\R.

1907: \end{equation}

1908: We will show that $\psi\sb d(x,t)$

1909: is a dispersive component of the solution $\psi(x,t)$,

1910: in the following sense.

1911: \begin{proposition}\label{prop-decay-psi-d}

1912: $\psi\sb{d}(\cdot,t)$

1913: is a bounded continuous $H\sp 1$-valued function:

1914: \begin{equation}\label{bound-psi-d}

1915: \psi\sb{d}(\cdot,t)\in C\sb{b}(\R,H\sp 1).

1916: \end{equation}

1917: The local energy decay holds

1918: for $\psi\sb{d}(\cdot,t)$:

1919: for any $R>0$,

1920: \begin{equation}\label{decay-psi-d}

1921: \Norm{

1922: \left[\!{\scriptsize{

1923: \begin{array}{c}

1924: \psi\sb{d}(\cdot,t)\\\dot\psi\sb{d}(\cdot,t)

1925: \end{array}

1926: }}\!\right]

1927: }\sb{{\E},R}

1928: \to 0,

1929: \qquad t\to\infty.

1930: \end{equation}

1931: \end{proposition}

1932: 

1933: \begin{proof}

1934: Changing the variable,

1935: we rewrite (\ref{dd}) as follows:

1936: \begin{equation}\label{ppmr}

1937: \psi\sb{d}(x,t)=\frac 1{2\pi}

1938: \int\sb\R

1939: \hat z(\omega(k))

1940: e\sp{i k\abs{x}}e\sp{-i\omega(k)t}

1941: \frac{k\,dk}{\omega(k)},

1942: \qquad x\in\R,

1943: \end{equation}

1944: where $\omega(k)=\sqrt{k^2+m^2}$

1945: is the branch analytic for $\Im k>0$ and continuous for  $\Im k\ge 0$.

1946: Note that the function $\omega(k)$, $k\in\R\backslash 0$,

1947: is the inverse function to

1948: $k(\omega)$ defined on $\Bar{\C\sp{+}}$

1949: (see (\ref{def-k}))

1950: and restricted onto $\varOmega\sb 0=(-\infty,-m)\cup(m,\infty)$.

1951: Let us introduce the functions

1952: \begin{equation}\label{ppm}

1953: \psi\sb\pm(x,t)\!:=\frac 1{2\pi}

1954: \int\sb\R

1955: \hat z(\omega(k))

1956: e\sp{\pm i k x}e\sp{-i\omega(k) t}

1957: \frac{kdk}{\omega(k)},

1958: \qquad

1959: x\in\R,

1960: \quad

1961: t\ge 0.

1962: \end{equation}

1963: Both functions $\psi\sb\pm(x,t)$

1964: are solutions to the free Klein-Gordon equation

1965: on the whole real line.

1966: The (free Klein-Gordon) energy of each solution

1967: is finite, since

1968: \begin{eqnarray}

1969: \Norm{

1970: \left[\!{\scriptsize{

1971: \begin{array}{c}\psi\sb\pm(\cdot,0)

1972: \\\dot \psi\sb\pm(\cdot,0)\end{array}

1973: }}\!\right]

1974: }\sb{{\E}}^2

1975: =

1976: \int\sb{\R}

1977: (\omega^2(k)+\abs{k}\sp 2+m^2)

1978: \abs{\hat z(\omega(k))}^2

1979: \frac{k^2}{\omega^2(k)}\,dk

1980: \nonumber

1981: \\

1982: =

1983: \int\sb{\R}

1984: 2\abs{\hat z(\omega(k))}^2

1985: k^2\,dk

1986: =

1987: 2\int\sb{\varOmega\sb 0}

1988: \abs{\hat z(\omega)}^2

1989: \omega k(\omega)\,d\omega

1990: <\infty.

1991: \nonumber

1992: \end{eqnarray}

1993: In the last inequality, we used (\ref{f2}).

1994: Hence, both $\psi\sb{-}$ and $\psi\sb{+}$

1995: are bounded continuous

1996: $H\sp 1$-valued functions:

1997: \begin{equation}\label{psipm}

1998: \psi\sb\pm\in C\sb{b}(\R, H^1),

1999: \end{equation}

2000: and for any $R>0$

2001: \begin{equation}\label{psi-pm-local-decay}

2002: \Norm{

2003: \left[\!{\scriptsize{

2004: \begin{array}{c}\psi\sb\pm(\cdot,t)\\\dot \psi\sb\pm(\cdot,t)\end{array}

2005: }}\!\right]

2006: }\sb{{\E},R}

2007: \to 0,

2008: \qquad

2009: t\to\infty

2010: \end{equation}

2011: by the same arguments as in

2012: the proof of

2013: Lemma~\ref{lemma-decay-psi1}.

2014: The function

2015: $\psi\sb{d}(x,t)$

2016: coincides with $\psi\sb{+}(x,t)$ for $x\ge 0$

2017: and with $\psi\sb{-}(x,t)$ for $x\le 0$:

2018: \[

2019: \psi\sb{d}(x,t)=\psi\sb{\pm}(x,t),

2020: \qquad \pm x\ge 0.

2021: \]

2022: Moreover,

2023: $\psi\sb{-}(0-,t)=\psi\sb{+}(0+,t)$,

2024: so

2025: $\psi\sb{d}(x,t)$ has no jump at $x=0$

2026: and therefore

2027: $\psi\sb{d}'(x,t)$ is square-integrable

2028: over the whole $x$-axis.

2029: Therefore,

2030:  (\ref{bound-psi-d}) follows from (\ref{psipm}), and

2031:  (\ref{decay-psi-d})

2032: follows from (\ref{psi-pm-local-decay}).

2033: \qed\end{proof}

2034: 

2035: \section{Bound component}

2036: \label{sect-bound}

2037: 

2038: \subsection*{Spectral representation}

2039: We introduce the bound component of the solution $\psi(x,t)$

2040: by

2041: \begin{equation}\label{bb}

2042: \psi\sb{b}(x,t)

2043: =\psi\sb{2}(x,t)-\psi\sb{d}(x,t)

2044: =\psi(x,t)-\psi\sb 1(x,t)-\psi\sb{d}(x,t),

2045: \ \ x\in\R,\ \ t\in\R.

2046: \end{equation}

2047: Then

2048: (\ref{psi2}) and

2049: Proposition~\ref{prop-decay-psi-d}

2050: imply that

2051: \begin{equation}\label{ebb}

2052: \psi\sb{b}\in C\sb{b}(\R,H\sp 1).

2053: \end{equation}

2054: By (\ref{ebb}),

2055: the function

2056: \[

2057: z\sb{b}(t):=\psi\sb{b}(0,t)

2058: =\psi\sb 2(0,t)-\psi\sb{d}(0,t)

2059: \]

2060: is bounded and continuous.

2061: Therefore, its Fourier transform

2062: $

2063: \hat z\sb{b}\in\mathscr{S}'(\R)

2064: $

2065: is a quasimeasure:

2066: \begin{equation}\label{ftbs}

2067: \hat z\sb{b}=\hat z-\hat z\sb{d}\in \mathcal{QM}(\R),

2068: \qquad\supp \hat z\sb{b}\subset [-m,m],

2069: \end{equation}

2070: where the last inclusion follows from (\ref{dpo}).

2071: Now (\ref{p1r}), (\ref{ftd}), and (\ref{bb})

2072: imply the multiplicative relation

2073: \begin{equation}\label{ftb0}

2074: \hat\psi\sb{b}(x,\omega)

2075: =

2076: \hat z\sb{b}(\omega)e\sp{ik(\omega)\abs{x}}.

2077: \end{equation}

2078: We denote

2079: \begin{equation}\label{def-kappa}

2080: \kappa(\omega):=-i k(\omega)=\sqrt{m^2-\omega\sp 2},

2081: \qquad

2082: \Re\kappa(\omega)\ge 0

2083: \quad{\rm for}\quad

2084: \Im\omega\ge 0,

2085: \end{equation}

2086: where $k(\omega)$ was introduced in (\ref{def-k}).

2087: Let us note that

2088: $\Re\kappa(\omega)\ge 0$

2089: and that $\kappa(\omega)>0$ for $\omega\in (-m,m)$.

2090: We rewrite (\ref{ftb0}) as

2091: \begin{equation}\label{ftb}

2092: \hat\psi\sb{b}(x,\omega)

2093: =

2094: \hat z\sb{b}(\omega)e\sp{-\kappa(\omega)\abs{x}},

2095: \qquad\omega\in\R.

2096: \end{equation}

2097: Therefore, (\ref{ftbs}) implies that

2098: $\hat\psi\sb{b}(x,\omega)$ for any fixed $x\in\R$

2099: is a quasimeasure with the support

2100: $\supp\hat\psi\sb{b}(x,\cdot)\subset[-m,m]$,

2101: and  finally,

2102: \begin{equation}\label{bdef}

2103: \psi\sb{b}(x,t)=\frac 1{2\pi}

2104: \langle

2105: \hat z\sb{b}(\omega)e\sp{-\kappa(\omega)\abs{x}},

2106: e\sp{-i\omega t}\rangle,\qquad x\in\R,\quad t\in\R.

2107: \end{equation}

2108: 

2109: \subsection*{Compactness}

2110: We are going to prove a compactness of the set of

2111: translations of the bound component,

2112: $\{\psi\sb{b}(x,s+t)\sothat s\ge 0\}$.

2113: We will derive the compactness from the following

2114: uniform estimates which we deduce from

2115: (\ref{ftbs}), (\ref{bdef})

2116: by Lemma~\ref{lemma-quasi-1}.

2117: 

2118: \begin{proposition}

2119: \begin{enumerate}

2120: \item

2121: The function $\psi\sb{b}(x,t)$ is smooth for $x\ne 0$

2122: and $t\in\R$,

2123: and

2124: the following representation holds

2125: for any fixed $\ x\ne 0$, $\ t\in\R$,

2126: and any nonnegative integers $\xx$, $\yy$:

2127: \begin{equation}\label{bqd}

2128: \p\sb x\sp\xx\p\sb t\sp\yy

2129: \psi\sb{b}(x,t)

2130: =

2131: \frac 1{2\pi}

2132: \Big\langle \hat z\sb{b}(\omega)(-\kappa(\omega)\sgn x)^{\xx}

2133: e\sp{-\kappa(\omega)\abs{x}},\,

2134: (-i\omega)^{\yy} e\sp{-i\omega t}

2135: \Big\rangle.

2136: \end{equation}

2137: \item

2138: For any $R>0$,

2139: there is a constant $C\sb{\xx,\yy,R}>0$ so that

2140: \begin{equation}\label{bqda}

2141: \sup\limits\sb{

2142: 0<\abs{x}\le R}

2143: \,\,

2144: \sup\limits\sb{t\in\R}

2145: \abs{\p\sb x\sp\xx\p\sb t\sp\yy\psi\sb{b}(x,t)}

2146: \le C\sb{\xx,\yy,R}.

2147: \end{equation}

2148: \end{enumerate}

2149: \end{proposition}

2150: 

2151: \begin{remark}

2152: Let us note that

2153: the bounds (\ref{bqda}) are independent of $x$

2154: and remain valid in the regions $x>0$ and $x<0$,

2155: although the derivatives

2156: $\p\sb x\sp\xx\p\sb t\sp\yy\psi\sb{b}(x,t)$

2157: with $\xx\ne 0$

2158: may have a jump at $x=0$.

2159: (This is the case for the solitary waves in (\ref{solitary-wave-profile}).)

2160: \end{remark}

2161: 

2162: \begin{proof}

2163: ({\it i})

2164: \ The representation (\ref{bqd}) with $\xx=0$

2165: and any $\yy$ follows

2166: directly

2167: from (\ref{ftbs}), (\ref{bdef}).

2168: Further, consider, for example,

2169: $\xx=1$ and $\yy=0$:

2170: \begin{equation}\label{bqdx}

2171: \p\sb x \psi\sb{b}(x,t)=\lim\limits\sb{\varepsilon\to 0}

2172: \frac 1{2\pi}

2173: \left\langle \hat z\sb{b}(\omega)

2174: \frac{e\sp{-\kappa(\omega)\abs{x+\varepsilon}}-

2175: e\sp{-\kappa(\omega)\abs{x}}}\varepsilon \zeta(\omega),

2176:  e\sp{-i\omega t}\right\rangle

2177: \end{equation}

2178: if the limit exists.

2179: Here $\zeta(\omega)\in C\sb{0}^\infty(\R)$ is any cutoff function

2180: which satisfies (\ref{cfu}).

2181: The relation (\ref{bqdx})

2182: follows by Lemma~\ref{lemma-quasi-1}~({\it ii}),

2183: if we verify that the following convergence

2184: holds in $L\sp 1(\R\sb t)$:

2185: \begin{equation}\label{cL1}

2186: \mathcal{F}\sb{\omega\to t}^{-1}

2187: \left[\frac{e\sp{-\kappa(\omega)\abs{x+\varepsilon}}

2188: -e\sp{-\kappa(\omega)\abs{x}}}\varepsilon \zeta(\omega)\right]\stackrel{L\sp 1}\longrightarrow

2189: \mathcal{F}\sb{\omega\to t}^{-1}

2190: [\p\sb x e\sp{-\kappa(\omega)\abs{x}}\zeta(\omega)].

2191: \end{equation}

2192: We rewrite the expression in the brackets in the left-hand side

2193: of (\ref{cL1}) as

2194: \begin{equation}\label{cL1r}

2195: \frac{e\sp{-\kappa(\omega)\abs{x+\varepsilon}}-

2196: e\sp{-\kappa(\omega)\abs{x}}}\varepsilon

2197: \zeta(\omega)

2198: =

2199: \int\sb 0\sp 1\p\sb x e\sp{-\kappa(\omega)\abs{x+\rho\varepsilon}}

2200: \zeta(\omega)

2201: \,d\rho.

2202: \end{equation}

2203: Now the convergence (\ref{cL1}) follows

2204: from the Puiseux expansion of type (\ref{MHfA})

2205: for $\p\sb x e\sp{-\kappa(\omega)\abs{x+\rho\varepsilon}}\zeta(\omega)$.

2206: 

2207: \noindent

2208: ({\it ii})

2209: \ For fixed nonnegative integers $\xx$ and $\yy$, denote

2210: \[

2211: N\sb x(\omega)

2212: =(-\kappa(\omega)\sgn x)\sp\xx

2213: e\sp{-\kappa(\omega)\abs{x}}(-i\omega)\sp\yy\zeta(\omega).

2214: \]

2215: Lemma~\ref{lemma-quasi-2}~({\it iii})

2216: implies that

2217: $N\sb x(\omega)=\hat K\sb x(\omega)$,

2218: with $K\sb x(\cdot)\in L\sp 1(\R)$.

2219: Then (\ref{bqd}) becomes

2220: \begin{equation}\label{bqdwb}

2221: \p\sb x\sp\xx\p\sb t\sp\yy\psi\sb{b}(x,t)

2222: =\frac 1{(2\pi)^2}

2223: \int\sb\R z\sb{b}(t-\tau)K\sb x(\tau)d\tau.

2224: \end{equation}

2225: The bounds (\ref{bqda}) follow

2226: by Lemma \ref{qmb} and (\ref{B*})

2227: since

2228: $z\sb{b}(t)=\psi\sb{b}(0,t)\in C\sb{b}(\R)$

2229: by (\ref{ebb}).

2230: \qed\end{proof}

2231: 

2232: \begin{corollary}\label{coco}

2233: By the Ascoli-Arzel\`a Theorem,

2234: for any sequence $s\sb{j}\to\infty$

2235: there exists a subsequence $s\sb{j'}\to\infty$

2236: such that

2237: for any nonnegative integers $\xx$ and $\yy$,

2238: \begin{equation}\label{olpd}

2239: \p\sb x\sp\xx\p\sb t\sp\yy

2240: \psi\sb{b}(x,s\sb{j'}+t)

2241: \to

2242: \p\sb x\sp\xx\p\sb t\sp\yy

2243: \beta(x,t),

2244: \qquad

2245: x\ne 0,

2246: \quad

2247: t\in\R,

2248: \end{equation}

2249: for some

2250: $\beta\in C\sb{b}(\R,H\sp 1)$.

2251: The convergence in {\rm (\ref{olpd})}

2252: is uniform in $x$ and $t$

2253: as long as $\abs{x}+\abs{t}\le R$, for any $R>0$.

2254: \end{corollary}

2255: 

2256: We call {\it omega-limit trajectory}

2257: any function $\beta(x,t)$

2258: that can appear as a limit in (\ref{olpd}).

2259: Previous analysis demonstrates that the long-time asymptotics

2260: of the solution $\psi(x,t)$ in $\E\sb F$

2261: depends only on the bound component

2262: $\psi\sb{b}(x,t)$.

2263: By Corollary \ref{coco},

2264: to conclude the proof of Theorem~\ref{main-theorem},

2265: it suffices to check that every omega-limit trajectory

2266: belongs to the set of solitary waves;

2267: that is,

2268: \begin{equation}\label{eidd}

2269: \beta(x,t)

2270: =

2271: \phi\sb{\omega\sb{+}}(x)e\sp{-i\omega\sb{+}t},

2272: \qquad

2273: x,\,t\in\R,

2274: \end{equation}

2275: with some $\omega\sb{+}\in[-m,m]$.

2276: 

2277: 

2278: \subsection*{Spectral identity for omega-limit trajectories}

2279: 

2280: Here we study the time spectrum of

2281: the omega-limit trajectories.

2282: 

2283: \begin{definition}

2284: Let $f$ be a tempered distribution.

2285: By $\Spec f$

2286: we denote the support of its Fourier transform:

2287: \[

2288: \Spec f:=\supp\tilde f.

2289: \]

2290: \end{definition}

2291: 

2292: \begin{proposition}\label{prop-beta}

2293: \begin{enumerate}

2294: \item

2295: For any omega-limit trajectory

2296: $\beta(x,t)$,

2297: the following spectral representation

2298: holds:

2299: \begin{equation}\label{ber}

2300: \beta(x,t)=\frac 1{2\pi}\langle \hat\gamma(\omega)e\sp{-\kappa(\omega)\abs{x}},e\sp{-i\omega t}\rangle,

2301: \qquad x\in\R,\qquad t\in\R,

2302: \end{equation}

2303: where $\hat\gamma\in \mathcal{QM}(\R)$, and

2304: \begin{equation}\label{spec-gamma}

2305: \supp \hat\gamma\subset [-m,m].

2306: \end{equation}

2307: \item

2308: The following bound holds:

2309: \begin{equation}\label{ebbe}

2310: \sup\limits\sb{t\in\R}

2311: \norm{\beta(\cdot,t)}\sb{H\sp 1}<\infty.

2312: \end{equation}

2313: \end{enumerate}

2314: \end{proposition}

2315: 

2316: Note that,

2317: according to (\ref{ber}),

2318: $\hat\gamma(\omega)$ is the Fourier

2319: transform of the function $\gamma(t):= \beta(0,t)$, $t\in\R$.

2320: 

2321: \begin{proof}

2322: The representation (\ref{bdef}) implies that

2323: \begin{equation}\label{bqdwi}

2324: \psi\sb{b}(x,s\sb{j}+t)

2325: =

2326: \frac{1}{2\pi}

2327: \langle \hat z\sb{b}(\omega)

2328: N\sb x(\omega)e\sp{-i\omega s\sb{j}},

2329: e\sp{-i\omega t}\rangle,

2330: \qquad x\ne 0,\qquad t\in\R,

2331: \end{equation}

2332: where $N\sb x$ corresponds to  $\xx=\yy=0$.

2333: The convergence  (\ref{olpd}) and the bounds (\ref{bqda})

2334: with  $\xx=\yy=0$ imply, by Definition~\ref{dA}, that

2335: \begin{equation}\label{dztA}

2336: z\sb{b}(s\sb{j'}+t)\stackrel{C\sb{b,F}}\longrightarrow \gamma(t),

2337: \qquad s\sb{j'}\to\infty,

2338: \end{equation}

2339: where $\gamma(t)$, $t\in\R$,

2340: is some continuous bounded function.

2341: Hence, by Definition~\ref{dQ},

2342: \begin{equation}\label{dztAd}

2343: \hat z\sb{b}(\omega)e\sp{-i\omega s\sb{j'}}

2344: \stackrel{\mathcal{QM}}\longrightarrow \hat\gamma(\omega),

2345: \qquad

2346: s\sb{j'}\to\infty.

2347: \end{equation}

2348: Now Lemma~\ref{lemma-quasi-1}~({\it ii})

2349: and

2350: Lemma~\ref{lemma-quasi-2}~({\it iii})

2351: imply that

2352: \begin{equation}\label{flA}

2353: \hat z\sb{b}(\omega)

2354: N\sb x(\omega)e\sp{-i\omega s\sb {k'}}

2355: \stackrel{\mathcal{QM}}\longrightarrow \hat\gamma(\omega)N\sb x(\omega),

2356: \qquad s\sb{j'}\to\infty.

2357: \end{equation}

2358: Hence, the representation (\ref{ber}) follows from

2359: (\ref{bqdwi}), and (\ref{spec-gamma}) follows from (\ref{ftbs}).

2360: Finally, the bound (\ref{ebbe})

2361: follows from (\ref{ebb}) and (\ref{olpd}).

2362: \qed\end{proof}

2363: 

2364: The relation (\ref{ber}) implies the

2365: basic spectral identity:

2366: 

2367: \begin{corollary}

2368: \label{corol}

2369: For any omega-limit trajectory

2370: $\beta(x,t)$,

2371: \begin{equation}\label{SiI}

2372: \Spec \beta(x,\cdot)=\Spec \gamma,

2373: \qquad

2374: x\in\R.

2375: \end{equation}

2376: \end{corollary}

2377: 

2378: \begin{remark}

2379: It is mainly for the proof of (\ref{SiI})

2380: that we develop the theory of the quasimeasures and multiplicators

2381: in Appendix~\ref{sect-quasimeasures}.

2382: This theory provides the compactness of the set of distributions

2383: $\{

2384: \hat z\sb{b}(\omega)e\sp{-i\omega s}\sothat s\ge 0\}$

2385: in the space of  quasimeasures

2386: (see (\ref{dztAd}))

2387: and the spectral representation (\ref{ber}).

2388: \end{remark}

2389: 

2390: \section{Nonlinear spectral analysis}

2391: \label{sect-spectral}

2392: Here we will derive (\ref{eidd})

2393: from the following identity:

2394: \begin{equation}\label{eide}

2395: \gamma(t)=Ce^{-i\omega\sb{+}t},

2396: \qquad

2397: t\in\R,

2398: \end{equation}

2399: which will be proved in three steps.

2400: 

2401: \subsection*{Step 1}

2402: The identity is equivalent to

2403: $\hat\gamma(\omega)\sim\delta(\omega-\omega\sb{+})$, so we start with

2404: an investigation of $\Spec \gamma:=\supp \hat \gamma$.

2405: \begin{lemma}

2406: For omega-limit trajectories

2407: the following spectral inclusion holds:

2408: \begin{equation}\label{Si}

2409: \Spec F(\gamma(\cdot))\subset \Spec \gamma.

2410: \end{equation}

2411: \end{lemma}

2412: 

2413: \begin{proof}

2414: The convergence  (\ref{olpd}) and equation (\ref{KG}),

2415: together with Lemma~\ref{lemma-decay-psi1}

2416: and Proposition~\ref{prop-decay-psi-d}~({\it ii}),

2417: imply that the limiting trajectory $\beta(x,t)$ is a solution to

2418: equation (\ref{KG}) (although $\psi\sb{b}(x,t)$ is not!):

2419: \begin{equation}\label{KG-beta}

2420: \ddot\beta(x,t)

2421: =\beta''(x,t)-m^2\beta(x,t)+\delta(x)F(\beta(0,t)),

2422: \qquad (x,t)\in\R\sp 2.

2423: \end{equation}

2424: Since $\beta(x,t)$ is smooth function for $x\le 0$ and $x\ge 0$,

2425: we get

2426: the following algebraic identity (cf. (\ref{ais})):

2427: \begin{equation}

2428: \label{AI}

2429: 0=\beta'(0+,t)-\beta'(0-,t)+F(\gamma(t)), \quad t\in\R.

2430: \end{equation}

2431: The identity implies the spectral inclusion

2432: \begin{equation}\label{spectral-inclusion}

2433: \Spec F(\gamma(\cdot))\subset

2434: \Spec \beta'(0+,\cdot)\cup\Spec \beta'(0-,\cdot).

2435: \end{equation}

2436: On the other hand,

2437: $\Spec \beta'(0+,\cdot)\cup\Spec \beta'(0-,\cdot)\subset\Spec \gamma$

2438: by (\ref{SiI}).

2439: Therefore, (\ref{spectral-inclusion})

2440: implies

2441:  (\ref{Si}).

2442: \qed\end{proof}

2443: 

2444: \begin{remark}

2445: The spectral inclusion (\ref{AI})

2446: follows from the algebraic identity (\ref{AI}),

2447: which in turn is a consequence of

2448: the fact that $\beta(x,t)$ solves (\ref{KG}).

2449: We cannot prove (\ref{spectral-inclusion})

2450: for the function $\psi\sb{b}(x,t)$ since

2451: generally it is \emph{not} a solution to (\ref{KG}).

2452: \end{remark}

2453: 

2454: \subsection*{Step 2}

2455: \begin{proposition}\label{pTN}

2456: For any omega-limit trajectory,

2457: the following identity holds:

2458: \begin{equation}\label{C}

2459: \abs{\gamma(t)}={\rm const},

2460: \qquad t\in\R.

2461: \end{equation}

2462: \end{proposition}

2463: 

2464: \begin{proof}

2465: We are going to show that (\ref{C})

2466: follows from the key spectral relations (\ref{spec-gamma}), (\ref{Si}).

2467: Our main assumption (\ref{f-is-such}) implies that the function

2468: $F(t):=F(\gamma(t))$ admits the representation (cf. (\ref{def-a}))

2469: \begin{equation}\label{frep}

2470: F(t)=\alpha(t)\gamma(t),

2471: \end{equation}

2472: where, according to (\ref{f-is-such}),

2473: \begin{equation}\label{arep}

2474: \alpha(t)=-\sum\limits\sb{n=1}\sp{N}  2n u\sb n\abs{\gamma(t)}\sp{2n-2},

2475: \qquad

2476: N\ge 2;

2477: \quad

2478: u\sb N > 0.

2479: \end{equation}

2480: Both functions $\gamma(t)$ and $\alpha(t)$ are bounded

2481: continuous functions in $\R$ by

2482: Proposition~\ref{prop-beta}~({\it iii}).

2483: Hence, $\gamma(t)$ and $\alpha(t)$ are

2484: tempered distributions.

2485: Furthermore, $\hat\gamma$ and $\hat{\overline\gamma}$ have the supports

2486: contained in $[-m,m]$

2487: by (\ref{spec-gamma}).

2488: Hence, $\hat\alpha$ also has a bounded support

2489: since it is

2490: a sum of convolutions of finitely many

2491: $\hat\gamma$ and $\hat{\overline\gamma}$ by (\ref{arep}).

2492: Then the relation (\ref{frep})

2493: translates into a convolution

2494: in the Fourier space,

2495: $

2496: \hat F=\hat\alpha\ast\hat\gamma/(2\pi),

2497: $

2498: and the spectral inclusion (\ref{Si}) takes the following form:

2499: \begin{equation}\label{Si-ast}

2500: \supp\hat F=

2501: \supp\,\hat\alpha\ast\hat\gamma\subset\supp\hat\gamma.

2502: \end{equation}

2503: Let us  denote $\bF=\supp \hat F$, $\bA=\supp \hat\alpha$, and

2504: $\bmGamma=\supp\hat\gamma$.

2505: Then the spectral inclusion (\ref{Si-ast}) reads as

2506: \begin{equation}\label{Si-ast-r}

2507: \bF\subset\bmGamma.

2508: \end{equation}

2509: On the other hand,

2510: it is well known that

2511: $\supp\hat\alpha\ast\hat\gamma\subset\supp\hat\alpha+\supp \hat\gamma$,

2512: or

2513: $

2514: \bF \subset\bA+\bmGamma.

2515: $

2516: Moreover, the Titchmarsh convolution theorem

2517: states that the last inclusion is exact for the ends

2518: of the supports:

2519: 

2520: \begin{theorem}[The Titchmarsh Convolution Theorem]

2521: {\it

2522: Let $\hat\alpha$,

2523: $\hat\gamma$ be two distributions in $\R$

2524: with compact supports $\bA$ and $\bmGamma$ respectively,

2525: and $\bF=\supp \hat\alpha*\hat\gamma$.

2526: Then}

2527: \begin{equation}\label{msc}

2528: \inf\bF=\inf\bA

2529: +\inf\bmGamma,

2530: \qquad

2531: \sup\bF=\sup\bA

2532: +\sup\bmGamma.

2533: \end{equation}

2534: \end{theorem}

2535: 

2536: This theorem was proved first in \cite{titchmarsh}

2537: for $\hat\alpha,\,\hat\gamma\in L\sp 1(\R)$ (see also \cite[p.119]{MR1400006}

2538: and \cite[Theorem 4.3.3]{MR1065136}).

2539: The Titchmarsh Convolution Theorem, together with (\ref{Si-ast-r}),

2540: allows us to conclude that

2541: $\inf\bA=\sup\bA=0$, hence

2542: $

2543: \bA\subset\{0\}.

2544: $

2545: Indeed,  (\ref{Si-ast-r}) and (\ref{msc})

2546: result in

2547: \begin{equation}\label{msci}

2548: \inf\bF=\inf\bA

2549: +\inf\bmGamma\ge\inf\bmGamma,

2550: \qquad

2551: \sup\bF=\sup\bA

2552: +\sup\bmGamma\le \sup\bmGamma,

2553: \end{equation}

2554: so that $\inf\bA\ge 0\ge \sup\bA$.

2555: Thus, we conclude that

2556: $\supp\hat\alpha=\bA\subset\{0\}$,

2557: therefore

2558: the distribution

2559: $\hat\alpha(\omega)$ is a finite

2560: linear combination of $\delta(\omega)$ and its derivatives.

2561: Then $\alpha(t)$ is a polynomial in $t$;

2562: since $\alpha(t)$ is bounded by

2563: Proposition~\ref{prop-beta}~({\it iii}),

2564: we conclude that $\alpha(t)$ is constant.

2565: Now the relation (\ref{C}) follows since

2566: $\alpha(t)$ is a polynomial  in

2567: $\abs{\gamma(t)}$, and its degree is strictly positive by

2568: (\ref{arep}).

2569: \qed\end{proof}

2570: \begin{remark}

2571: The boundedness of the spectrum of

2572: both $\gamma(t)$

2573: and $\alpha(t)$

2574: is critical for our argument,

2575: since otherwise the Titchmarsh convolution theorem

2576: does not apply.

2577: It is to ensure that the spectrum of

2578: $\alpha(t)$ is also bounded

2579: that we had to assume the polynomial

2580: character of the nonlinearity

2581: in Assumption~\ref{assumption-a}.

2582: \end{remark}

2583: 

2584: \subsection*{Step 3}

2585: Now the same Titchmarsh arguments imply that $\bmGamma:=\Spec \gamma$ is a point

2586: $\omega\sb{+}\in[-m,m]$.

2587: Indeed,

2588: (\ref{C}) means that $\gamma(t) \overline\gamma(t)\equiv C$, hence

2589: in the Fourier transform

2590: $\hat\gamma \ast\hat{\overline\gamma}=2\pi C\delta(\omega)$.

2591: Therefore,

2592: if $\gamma$ is not identically zero,

2593: the Titchmarsh Theorem implies that

2594: \[

2595: 0=\sup\bmGamma+\sup(-\bmGamma)

2596: =\sup\bmGamma-\inf\bmGamma.

2597: \]

2598: Hence

2599: $\inf\bmGamma=\sup\bmGamma$

2600: and therefore

2601: $\bmGamma=\{\omega\sb{+}\}\subset [-m,m]$,

2602: %%AC typo! Was: $\bmGamma=\omega\sb{+}\in [-m,m]$,

2603: so that $\hat\gamma(\omega)$

2604: is a finite linear combination

2605: of $\delta(\omega-\omega\sb{+})$ and its derivatives.

2606: As the matter of fact, the derivatives could not be present

2607: because of the boundedness of

2608: $\gamma(t)=\beta(0,t)$

2609: that follows from

2610: Proposition~\ref{prop-beta}~({\it iii}).

2611: Thus,

2612: $\hat\gamma\sim\delta(\omega-\omega\sb{+})$, which implies

2613: (\ref{eide}).

2614: 

2615: \subsubsection*{Conclusion of the proof of Theorem~\ref{main-theorem}}

2616: The representation

2617: (\ref{ber}) implies that $\beta(x,t)

2618: =C e\sp{-\kappa\sb{+}\abs{x}}e\sp{-i\omega\sb{+} t}$

2619: since $\hat\gamma\sim\delta(\omega-\omega\sb{+})$.

2620: Therefore, the equation

2621: (\ref{KG-beta}) and the bound (\ref{ebbe})

2622: imply

2623: that $\beta(x,t)$

2624: is a solitary wave.

2625: This completes the proof of Theorem~\ref{main-theorem}.

2626: 

2627: \section{Linear case}

2628: \label{sect-linear-case}

2629: Let us now give a complete treatment of the linear case

2630: and prove Theorem~\ref{main-theorem-linear}.

2631: We assume that

2632: $F(\psi)$ that enters (\ref{KG})

2633: is given by

2634: $F(\psi)=a\psi$,

2635: where $a\in\R$, and $a<2m$.

2636: Thus, the potential is given by

2637: $U(\psi)=-a\abs{\psi}^2/2$,

2638: and

2639: we consider

2640: the equation

2641: \begin{equation}

2642: \ddot\psi(x,t)

2643: =\psi''(x,t)-m^2\psi(x,t)+a\delta(x)\psi(0,t),

2644: \qquad

2645: x\in\R,

2646: \quad t\in\R.

2647: \label{KG-linear}

2648: \end{equation}

2649: All  conclusions of

2650: Theorem~\ref{theorem-well-posedness}

2651: on global well-posedness

2652: hold for equation (\ref{KG-linear})

2653: with $a<2m$ since in this case

2654: the condition (\ref{bound-below}) is satisfied.

2655: Let us note that if $a\ge 2m$,

2656: then the conclusions ({\it i}), ({\it ii}),

2657: and ({\it iii}) of Theorem~\ref{theorem-well-posedness},

2658: are still valid

2659: (their proofs in  Appendix~\ref{sect-existence}

2660: apply for bounded times,

2661: and then the conclusions

2662: follow

2663: for all times by the linearity of the equation).

2664: On the other hand,

2665: the a priori bound (\ref{eb}) is generally violated

2666: when $a/2\ge m$ (see below).

2667: 

2668: 

2669: \begin{remark}\label{remark-solitons-linear}

2670: Let us summarize the properties of the solitary

2671: waves for the linear case that follow from Proposition~\ref{prop-solitons}.

2672: Note that, according to (\ref{kaka}), $\kappa=a/2$.

2673: \begin{enumerate}

2674: \item

2675: For $a\le 0$ there are no nonzero solitary waves

2676: since we need $\kappa>0$ for (\ref{solitary-wave-profile})

2677: to be from $H\sp 1$.

2678: \item

2679: When $a>0$, $a\ne 2m$,

2680: all solitary waves are given by

2681: $

2682: \phi\sb\omega(x)=C e^{-a\abs{x}/2},

2683: $

2684: where $C\in\C$ and

2685: $\omega=\pm\omega\sb a$,

2686: where $\omega\sb a:=\sqrt{m^2-{a^2}/{4}}$.

2687: Note that

2688: if $a>2m$,

2689: then the values $\pm\omega\sb a$ are purely imaginary

2690: and the corresponding solitary waves are exponentially growing.

2691: \item

2692: If $a=2m$, then $\omega\sb 0=0$ and

2693: there is a nonzero static solitary wave solution

2694: $\phi\sb 0(x)=e^{-m\abs{x}}$.

2695: Besides,

2696: there is secular (linearly growing) solution

2697: $te^{-m\abs{x}}$.

2698: 

2699: \end{enumerate}

2700: 

2701: \end{remark}

2702: 

2703: \begin{remark}

2704: When $a>2m$,

2705: the values of $\omega$ are purely imaginary,

2706: and the $\E$-norm of solitary waves

2707: that correspond to $\pm\Im\omega>0$

2708: grows exponentially for $t\to\pm\infty$.

2709: When $a=2m$, we have $\omega=0$;

2710: The $\E$-norm

2711: of the secular solution

2712: grows linearly in time.

2713: In both cases ($a\ge 2m$),

2714: the a priori bound (\ref{eb}) fails.

2715: This illustrates that

2716: condition (\ref{bound-below}) is sharp,

2717: since this condition fails for

2718: the potential

2719: $U(\psi)=-\displaystyle\frac{a}{2}\abs{\psi}^2$

2720: with $a\ge 2m$.

2721: \end{remark}

2722: 

2723: \subsubsection*{Proof of Theorem~\ref{main-theorem-linear}}

2724: 

2725: 

2726: Let us prove the global attraction

2727: to the set $ \langle\bS\rangle$.

2728: We proceed as in the proof of Theorem~\ref{main-theorem}

2729: until we get to equation (\ref{KG-beta}).

2730: Since now $F(\psi)=a\psi$,

2731: (\ref{KG-beta})

2732: takes the following form:

2733: \begin{equation}\label{KG-beta-linear}

2734: \ddot\beta(x,t)

2735: =\beta''(x,t)-m^2\beta(x,t)

2736: +a\delta(x)\beta(0,t),

2737: \qquad (x,t)\in\R\sp 2.

2738: \end{equation}

2739: Now we cannot use  the Titchmarsh arguments,

2740: and we have to solve the equation  directly to prove

2741: that

2742: \begin{equation}\label{fol}

2743: \left[\begin{array}{c}

2744: \beta(\cdot,t)

2745: \\

2746: \dot\beta(\cdot,t)

2747: \end{array}

2748: \right]

2749: \in \langle\bS\rangle

2750: \qquad

2751: {\rm for}\quad t\in\R.

2752: \end{equation}

2753: In

2754: the Fourier transform

2755: $\hat\beta(x,\omega)=\mathcal{F}\sb{t\to\omega}[\beta(x,t)]$

2756: the equation (\ref{KG-beta-linear})  becomes

2757: \begin{equation}\label{KG-beta-linear-f}

2758: -\omega^2\hat\beta(x,\omega)

2759: =\hat\beta''(x,\omega)-m^2\hat\beta(x,\omega)

2760: +a\delta(x)\hat\beta(0,\omega),

2761: \qquad (x,\omega)\in\R\sp 2.

2762: \end{equation}

2763: On the other hand, the representation

2764: (\ref{ber}) implies that

2765: \begin{equation}\label{beri}

2766: \hat\beta(x,\omega)=\hat\gamma(\omega)e^{-\kappa(\omega)\abs{x}}

2767: \end{equation}

2768: Substituting this into  (\ref{KG-beta-linear-f}), we obtain

2769: \begin{equation}\label{berig}

2770: 2\kappa(\omega)\hat\gamma(\omega)\delta(x)=a\delta(x)\hat\gamma(\omega).

2771: \end{equation}

2772: Therefore,

2773: on the support of the distribution $\hat\gamma(\omega)$,

2774: the identity holds

2775: \begin{equation}\label{berigk}

2776: 2\kappa(\omega)=a,

2777: \end{equation}

2778: hence

2779: $\supp \hat\gamma\subset \bmGamma\sb a:=\{\omega\in [-m,m]: 2\kappa(\omega)=a\}$

2780: by (\ref{spec-gamma}).

2781: Now let us consider two cases.

2782: 

2783: \begin{enumerate}

2784: \item

2785: In the case $0<a<2m$,

2786: according to Remark~\ref{remark-solitons-linear}~({\it ii}),

2787: the set of finite energy solitary waves is given by

2788: \begin{equation}\label{ssw-1}

2789: \bS=\left\{C\sb 1

2790: \left[\begin{array}{c}e^{-a\abs{x}/2}\\

2791: i\omega\sb a e^{-a\abs{x}/2}

2792: \end{array}

2793: \right]

2794: +

2795: C\sb 2\left[\begin{array}{c}e^{-a\abs{x}/2}\\

2796: -i\omega\sb a e^{-a\abs{x}/2}

2797: \end{array}\right]

2798: \sothat\;

2799: C\sb 1,\,\,C\sb 2\in\C

2800: \right\}.

2801: \end{equation}

2802: 

2803: On the other hand, the set

2804: $\bmGamma\sb a$

2805: contains exactly two points $\pm\omega\sb a$

2806: since $0<a<2m$. Hence, $\hat\gamma$ is a linear combination of

2807: $\delta(\omega\pm\omega\sb a)$ and their derivatives.

2808: The derivatives are forbidden since

2809: $\gamma(t)$ is bounded, so finally

2810: \begin{equation}\label{fina}

2811: \beta(x,t)

2812: =

2813: \left(

2814: C\sb 1 e^{i\omega\sb a t}+C\sb 2 e^{-i\omega\sb a t}

2815: \right)

2816: e^{-a\abs{x}/2}.

2817: \end{equation}

2818: Now  (\ref{fol}) follows from  (\ref{fina}).

2819: \item

2820: In the case $a\le 0$,

2821: the set of finite energy solitary waves consists of

2822: the zero solution only by Remark~\ref{remark-solitons-linear}~({\it i}).

2823: For $a<0$, the set $\bmGamma\sb a$ is empty,

2824: hence $\beta(x,t)=0$ and (\ref{fol}) follows.

2825: When $a=0$, we have

2826: $\omega\sb a=m$

2827: and

2828: $\bmGamma\sb a=\{-m\}\cup\{m\}$.

2829: Any omega-limit point

2830: $\beta$ is given by

2831: (\ref{fina})

2832: with $a=0$.

2833: Since

2834: $\beta(\cdot,t)\in H\sp 1$,

2835: we conclude that $C\sb 1=C\sb 2=0$ in (\ref{fina}),

2836: so that $\beta(x,t)=0$ and the inclusion (\ref{fol}) follows.

2837: \end{enumerate}

2838: 

2839: This finishes the proof of Theorem~\ref{main-theorem-linear}.

2840: 

2841: \begin{remark}\label{noatt}

2842: For $0<a<2m$,

2843: a particular exact solution to (\ref{KG-linear}),

2844: e.g.  (\ref{fina}),

2845: with $C\sb 1\ne 0$ and $C\sb 2\ne 0$,

2846: shows that in general there could be no attraction to

2847: $\bS$.

2848: \end{remark}

2849: 

2850: \appendix

2851: 

2852: \section{Appendix: Solitary waves}

2853: \label{sect-solitons}

2854: 

2855: Here we prove Lemma~\ref{lemma-omega-real}

2856: and Proposition~\ref{prop-solitons}.

2857: 

2858: \subsubsection*{Proof of Lemma~\ref{lemma-omega-real}}

2859: Substituting $\phi\sb\omega(x) e^{-i\omega t}$

2860: into (\ref{KG}),

2861: we get the equation

2862: \begin{equation}\label{stat-eqn}

2863: -\omega^2\phi\sb\omega(x)e^{-i\omega t}

2864: =

2865: \phi\sb\omega''(x)e^{-i\omega t}

2866: -m^2\phi\sb\omega(x)e^{-i\omega t}

2867: +\delta(x)F(e^{-i\omega t}\phi\sb\omega(0)),

2868: \end{equation}

2869: where $(x,t)\in\R\times\R$.

2870: We can assume that $\phi\sb\omega(0)\ne 0$.

2871: Indeed, if $\phi\sb\omega(0)=0$,

2872: then (\ref{stat-eqn})

2873: turns into a homogeneous second-order linear differential equation,

2874: which together with the inclusion $\phi\sb\omega\in H\sp 1(\R)$

2875: results in $\phi\sb\omega(x)\equiv 0$.

2876: Equation (\ref{stat-eqn}) leads to the identity

2877: $

2878: e^{-i\omega t}\Delta

2879: =

2880: F(e^{-i\omega t}\phi\sb\omega(0)),

2881: $

2882: with

2883: $\Delta=\phi\sb\omega'(0-)-\phi\sb\omega'(0+)$.

2884: This results in

2885: \begin{equation}\label{delta-f-phi}

2886: \frac{\Delta}{\phi\sb\omega(0)}

2887: =\frac{F(e^{-i\omega t}\phi\sb\omega(0))}{e^{-i\omega t}\phi\sb\omega(0)}

2888: =\frac{F(e^{t\,\Im\omega}\phi\sb\omega(0))}{e^{t\,\Im\omega}\phi\sb\omega(0)},

2889: \qquad

2890: t\in\R.

2891: \end{equation}

2892: We used (\ref{inv-f}) in the last equality.

2893: The condition that $F(\psi)$ is strictly nonlinear

2894: (in the sense of Definition~\ref{def-sn})

2895: implies that (\ref{delta-f-phi})

2896: only holds at discrete values  of $t\,\Im\omega$;

2897: thus, $\Im\omega=0$, finishing the proof.

2898: 

2899: \subsection*{Proof of Proposition~\ref{prop-solitons}}

2900: The relation (\ref{stat-eqn}) turns into

2901: the following eigenvalue problem:

2902: \begin{equation}\label{NEP}

2903: -\omega\sp 2\phi\sb\omega(x)

2904: =

2905: \phi\sb\omega''(x)-m^2\phi\sb\omega(x)

2906: +\delta(x)F(\phi\sb\omega(x)),\quad x\in\R.

2907: \end{equation}

2908: The phase factor $e^{-i\omega t}$

2909: can be canceled out

2910: because either $F(\psi)=a\psi$ or,

2911: when $F$ is strictly nonlinear,

2912: we can use (\ref{inv-f})

2913: (since in this case $\omega\in\R$ by Lemma~\ref{lemma-omega-real}).

2914: Equation (\ref{NEP}) implies that away from the origin

2915: we have

2916: $$

2917: \phi\sb\omega''(x)=(m^2-\omega^2)\phi\sb\omega(x),

2918: \qquad

2919: x\ne 0,

2920: $$

2921: hence

2922: $\phi\sb\omega(x)=C\sb\pm e\sp{-\kappa\sb\pm\abs{x}}$ for $\pm x>0$,

2923: where $\kappa\sb\pm$

2924: satisfy

2925: $\kappa\sb\pm^2=m^2-\omega\sp 2$.

2926: Since $\phi\sb\omega(x)\in H\sp 1$,

2927: it is imperative that $\kappa\sb\pm>0$;

2928: we conclude that

2929: $\abs{\omega}<m$

2930: and that

2931: $\kappa\sb\pm=\sqrt{m^2-\omega^2}>0$.

2932: Moreover, since

2933: the function $\phi\sb\omega(x)$ is continuous,

2934: $C\sb{-}=C\sb{+}=C\ne 0$

2935: (since we are looking for nonzero solitary waves).

2936: We see that

2937: \begin{equation}\label{profile}

2938: \phi\sb\omega(x)=C e\sp{-\kappa\abs{x}},

2939: \qquad

2940: C\ne 0,

2941: \qquad

2942: \kappa\equiv\sqrt{m^2-\omega^2}>0.

2943: \end{equation}

2944: Equation (\ref{NEP}) implies the following gluing condition at $x=0$:

2945: \begin{equation}\label{ais}

2946: 0=\phi\sb\omega'(0+)-\phi\sb\omega'(0-)+F(\phi\sb\omega(0)).

2947: \end{equation}

2948: This condition and (\ref{profile})

2949: lead to the equation

2950: $

2951: 2\kappa C=F(C)

2952: $

2953: which is equivalent to (\ref{kaka}) for $C\ne 0$.

2954: 

2955: \section{Appendix: Quasimeasures and multiplicators}

2956: \label{sect-quasimeasures}

2957: 

2958: \subsection*{Quasimeasures}

2959: 

2960: Let us denote

2961: by $\check g$

2962: the inverse Fourier transform

2963: of a tempered distribution $g$:

2964: \[

2965: \check g(t)=\mathcal{F}\sb{\omega\to t}^{-1}[g(\omega)].

2966: \]

2967: 

2968: \begin{definition}\label{def-quasimeasure}

2969: 

2970: A tempered distribution

2971: $\mu(\omega)$ is a

2972: \emph{quasimeasure}

2973: if  $\check \mu\in C\sb{b}(\R)$.

2974: \end{definition}

2975: 

2976: For example,

2977: any function from $L\sp 1(\R)$ is a quasimeasure,

2978: and so is any finite Borel measure on $\R$.

2979: 

2980: \begin{lemma}\label{qmb}

2981: Let $\mu(\omega)$ be a quasimeasure and $\varphi(\omega)$ be a test function

2982: from the Schwartz space $\mathscr{S}(\R)$.

2983: Then

2984: \begin{equation}\label{mup}

2985: \abs{\langle \mu(\omega),\varphi(\omega)\rangle}\le

2986: C

2987: \Vert

2988: \check\varphi(t)

2989: \Vert\sb{L^1(\R)}.

2990: \end{equation}

2991: \end{lemma}

2992: 

2993: The lemma is a trivial consequence of the Parseval identity:

2994: \begin{equation}\label{mupp}

2995: \abs{\langle \mu(\omega),\varphi(\omega)\rangle}

2996: =

2997: {2\pi}\abs{\langle \check\mu(t),\check\varphi(t)\rangle}

2998: \le

2999: {2\pi}

3000: \Vert

3001: \check\mu(t)

3002: \Vert\sb{L^\infty(\R)}

3003: \Vert

3004: \check\varphi(t)

3005: \Vert\sb{L^1(\R)}.

3006: \end{equation}

3007: 

3008: \begin{definition}\label{dA}

3009: $C\sb{b,F}(\R)$

3010: is the vector space of bounded  functions

3011: $f(t)\in C\sb{b}(\R)$

3012: endowed with the following convergence:

3013: $f\sb\varepsilon(t)\stackrel{C\sb{b,F}}\longrightarrow f(t)$,

3014: $\varepsilon\to 0+\,$ if and only if

3015: \begin{enumerate}

3016: \item

3017: $\forall T>0$,

3018: \quad

3019: $\Vert f\sb\varepsilon(t)-f(t)\Vert\sb{C[-T,T]}

3020: \to 0$, $\varepsilon\to 0+$;

3021: \item

3022: $\sup\limits\sb{\varepsilon\in (0,1]}

3023: \Vert f\sb\varepsilon(t)\Vert\sb{C\sb{b}(\R)}

3024: <\infty$.

3025: 

3026: \end{enumerate}

3027: 

3028: \end{definition}

3029: 

3030: This type of convergence

3031: coincides with the convergence

3032: stated in the Ascoli-Arzel\`a theorem.

3033: Next we introduce the dual class of the

3034: ``Ascoli-Arzel\`a quasimeasures''.

3035: \begin{definition}\label{dQ}

3036: $\mathcal{QM}(\R)$

3037: is the

3038: linear space of all quasimeasures

3039: $\mu(\omega)$ endowed with the following convergence:

3040: \[

3041: \mu\sb\varepsilon(\omega)\stackrel{\mathcal{QM}}

3042: {\mathop{\longrightarrow}

3043: \limits\sb{\varepsilon\to 0+}}

3044: \mu(\omega)

3045: \quad{\rm if\ and\ only\ if}\quad

3046: \check\mu\sb\varepsilon(t)\stackrel{C\sb{b,F}}

3047: {\mathop{\longrightarrow}\limits\sb{\varepsilon\to 0+}}

3048: \check\mu(t).

3049: \]

3050: \end{definition}

3051: 

3052: 

3053: 

3054: 

3055: \subsection*{Multiplicators}

3056: 

3057: 

3058: Now let us give a simple characterization

3059: of multiplicators in $\mathcal{QM}(\R)$.

3060: Let us consider a continuous function $M(\omega)\in C(\R)$.

3061: We also denote by $M$ the

3062: corresponding operator of multiplication:

3063: \[

3064: M:\;\mu(\omega)\mapsto M(\omega)\mu(\omega),

3065: \qquad

3066: \mu(\omega)\in C\sb{0}^\infty(\R).

3067: \]

3068: 

3069: \begin{lemma}\label{lemma-quasi-1}

3070: \begin{enumerate}

3071: \item

3072: Let

3073: $\check M(t)\in L\sp 1(\R)$.

3074: Then the operator $M$

3075: extends

3076: to a linear continuous operator in the space of quasimeasures:

3077: \[

3078: M:\;\mathcal{QM}(\R)\to\mathcal{QM}(\R).

3079: \]

3080: \item

3081: Let $\mu\sb\varepsilon(\omega)\stackrel{\mathcal{QM}}\longrightarrow \mu(\omega)$  and

3082: $\check M\sb\varepsilon(t)\stackrel{L\sp 1}\longrightarrow\check M(t)$

3083: as $\varepsilon\to 0+$.

3084: Then

3085: \begin{equation}\label{mve}

3086: M\sb\varepsilon(\omega)\mu\sb\varepsilon(\omega)

3087: \stackrel{\mathcal{QM}}\longrightarrow M(\omega)\mu(\omega),

3088: \qquad\varepsilon\to 0+.

3089: \end{equation}

3090: \end{enumerate}

3091: \end{lemma}

3092: \begin{proof}

3093: First we define

3094: $M(\omega)\mu(\omega):=\mathcal{F}\sb{t\to\omega}

3095: [\big(\check M\ast\check\mu\big)(t)](\omega)$

3096: that agrees with the case

3097: $\mu\in C\sb{0}^\infty(\R)$.

3098: Then

3099: ({\it i}) follows from ({\it ii})

3100: with

3101: $M\sb\varepsilon=M$

3102: and $\mu\sb\varepsilon\in C\sb{0}^\infty(\R)$.

3103: To prove  ({\it ii}),

3104: we need to show that

3105: \begin{equation}\label{lemma-second-statement}

3106: \mathcal{F}\sb{\omega\to t}^{-1}[M\sb\varepsilon(\omega)\mu\sb\varepsilon(\omega)]

3107: =\big(\check M\sb\varepsilon\ast\check\mu\sb\varepsilon\big)(t)

3108: \stackrel{C\sb{b,F}}\longrightarrow

3109: \big(\check M\ast\check\mu\big)(t).

3110: \end{equation}

3111: We have to check both conditions ({\it i}) and ({\it ii})

3112: of Definition~\ref{dA} for the functions

3113: \begin{eqnarray}

3114: &&

3115: f\sb\varepsilon(t):=\mathcal{F}\sb{\omega\to t}^{-1}

3116: [M\sb\varepsilon(\omega)\mu\sb\varepsilon(\omega)]

3117: =\big(\check M\sb\varepsilon\ast\check\mu\sb\varepsilon\big)(t),

3118: \phantom{\int}

3119: \nonumber

3120: \\

3121: &&

3122: f(t):=\mathcal{F}\sb{\omega\to t}^{-1}

3123: [M(\omega)\mu(\omega)]=

3124: \big(\check M\ast\check\mu\big)(t).

3125: \nonumber

3126: \end{eqnarray}

3127: We have:

3128: \[

3129: f\sb\varepsilon(t)-f(t)=

3130: \big(\check M\sb\varepsilon\ast\check\mu\sb\varepsilon\big)(t)

3131: -\big(\check M\ast\check\mu\big)(t)

3132: =

3133: \big(\!(\check M\sb\varepsilon-\check M)\ast\check\mu\sb\varepsilon\big)(t)

3134: +\big(\check M\ast(\check\mu\sb\varepsilon-\check\mu)\!\big)(t).

3135: \]

3136: The first term in the right-hand side converges to zero

3137: uniformly in $t\in\R$

3138: since

3139:  $\check M\sb\varepsilon-\check M\to 0$ in $L\sp 1$ while

3140: $\check\mu\sb\varepsilon\in C\sb{b}(\R)$ and is bounded uniformly

3141: for $\varepsilon\in(0,1)$.

3142: Let us analyze the second term,

3143: \begin{equation}\label{lim-second-term}

3144: \int\sb{\R}\check M(\tau)(\check\mu\sb\varepsilon(t-\tau)-\check\mu(t-\tau))\,d\tau.

3145: \end{equation}

3146: Since $\check M\in L\sp 1$,

3147: for any $\delta>0$

3148: there exists a finite $R>0$ so that

3149: $

3150: \int\sb{\abs{\tau}>R}

3151: \abs{\check M(\tau)}\,d\tau

3152: \le\delta.

3153: $

3154: On the other hand, for any $T>0$,

3155: the difference $\check\mu\sb\varepsilon(t-\tau)-\check\mu(t-\tau)$

3156: is uniformly small for  $\abs{t}\le T$,

3157: $\abs{\tau}<R$ and small $\varepsilon$.

3158: Therefore,

3159: the integral  (\ref{lim-second-term})

3160: converges to zero uniformly in

3161: $\abs{t}\le T$ as $\varepsilon\to 0+$.

3162: Hence, the convergence  ({\it i}) of Definition~\ref{dA} follows.

3163: 

3164: Finally, the uniform bound ({\it ii}) of Definition~\ref{dA}

3165: for the functions $f\sb\varepsilon(t)$ is obvious.

3166: The convergence (\ref{lemma-second-statement}) is proved.

3167: \qed\end{proof}

3168: 

3169: \subsection*{Bounds for multiplicators}

3170: 

3171: Let us justify the properties of the multiplicators

3172: which we used in Section~\ref{sect-bound}.

3173: Recall that we use the notation

3174: \[

3175: M\sb{x,\varepsilon}(\omega)

3176: :=e\sp{ik(\omega+i\varepsilon)\abs{x}}\zeta(\omega),

3177: \qquad

3178: x\in\R,\quad \varepsilon\ge 0,

3179: \]

3180: where $\zeta(\omega)\in C\sb 0\sp\infty(\R)$

3181: is a fixed cutoff function,

3182: and also the notation

3183: \[

3184: N\sb x(\omega)

3185: :=(ik(\omega)\sgn x)\sp\xx e\sp{ik(\omega)\abs{x}}

3186: (-i\omega)\sp\yy \zeta(\omega),

3187: \qquad

3188: x\in\R,

3189: \]

3190: where

3191: $\xx$, $\yy$ are fixed nonnegative integers.

3192: 

3193: \begin{lemma}\label{lemma-quasi-2}

3194: For any fixed $x\in\R$ we have:

3195: \begin{enumerate}

3196: \item

3197: $\check M\sb{x,\varepsilon}(t)\in L\sp 1(\R)$ for any

3198: $\varepsilon\ge 0$.

3199: \item

3200: $\check M\sb{x,\varepsilon}(t)\stackrel{L\sp 1}

3201: \longrightarrow\check M\sb{x,0}(t)$,

3202: \quad$\varepsilon\to 0$.

3203: \item

3204: $\check N\sb x \in L\sp 1(\R)$, and for any $R>0$

3205: there exists $C\sb{\xx,\yy,R}>0$

3206: so that

3207: \begin{equation}\label{B*}

3208: \sup\limits\sb{\abs{x}\le R}

3209: \,

3210: \norm{

3211: \check N\sb x}\sb{L\sp 1(\R)}\le C\sb{\xx,\yy,R}.

3212: \end{equation}

3213: \end{enumerate}

3214: \end{lemma}

3215: 

3216: \begin{proof}

3217: For any fixed $x\in\R$, the Puiseux expansion holds:

3218: \begin{equation}\label{MHfA}

3219: e\sp{ik(\omega+i\varepsilon)\abs{x}}

3220: \sim 1+

3221: \sum\sb{\pm}\sum\limits\sb{j=1}^{\infty}

3222: C\sb j\sp\pm

3223: (x)(\omega+i\varepsilon\mp m)\sp{j/2},

3224: \qquad

3225: \omega+i\varepsilon\to \pm m,

3226: \qquad

3227: \varepsilon >0.

3228: \end{equation}

3229: Therefore,

3230: the function $\check M\sb{x,\varepsilon}(t)$

3231: is smooth and

3232: decays at least like $\abs{t}\sp{-3/2}$

3233: when $t\to\infty$.

3234: This finishes the proof of the first statement of the lemma.

3235: 

3236: The second statement of the lemma follows from (\ref{MHfA}).

3237: 

3238: The last statement of the lemma

3239: follows by the same arguments from

3240: the Puiseux expansion for $\check N\sb x(\omega)$

3241: similar to expansion (\ref{MHfA}) with $\varepsilon=0$.

3242: \qed\end{proof}

3243: 

3244: 

3245: 

3246: \section{Appendix: Global well-posedness}

3247: \label{sect-existence}

3248: 

3249: Here we prove Theorem~\ref{theorem-well-posedness}.

3250: We first need

3251: to adjust the nonlinearity $F$

3252: so that it becomes bounded, together with

3253: its derivatives.

3254: Define

3255: \begin{equation}\label{def-Lambda}

3256: \Lambda(\Psi\sb 0)

3257: =\sqrt{\frac{\mathcal{H}(\Psi\sb 0)-{A}}{m-{B}}},

3258: \end{equation}

3259: where $\Psi\sb 0\in{\E}$ is the initial data

3260: from Theorem~\ref{theorem-well-posedness}

3261: and ${A}$, ${B}$ are constants from (\ref{bound-below}).

3262: Then we may pick a modified potential function

3263: $\widetilde{U}\in C\sp 2(\C,\R)$,

3264: $\widetilde{U}(\psi)=\widetilde{U}(\abs{\psi})$,

3265: so that

3266: \begin{equation}\label{new-U}

3267: \widetilde{U}(\psi)=U(\psi)

3268: \qquad{\rm for}\ \abs{\psi}\le \Lambda(\Psi\sb 0),

3269: \quad

3270: \psi\in\C,

3271: \end{equation}

3272: $\widetilde{U}(\psi)$ satisfies (\ref{bound-below})

3273: with the same constants ${A}$, ${B}$ as $U(\psi)$ does:

3274: \begin{equation}\label{new-U-2}

3275: \widetilde{U}(\psi)\ge {A}-{B}\abs{\psi}^2,

3276: \quad{\rm for}\ \psi\in\C,\quad

3277: {\rm where}\ {A}\in\R\ {\rm and}\ 0\le {B}<m,

3278: \end{equation}

3279: and so that

3280: $\abs{\widetilde{U}(\psi)}$,

3281: $\abs{\widetilde{U}'(\psi)}$,

3282: and $\abs{\widetilde{U}''(\psi)}$

3283: are bounded for $\psi\ge 0$.

3284: We define

3285: \begin{equation}\label{f-reg}

3286: \widetilde{F}(\psi)

3287: =-\nabla \widetilde{U}(\psi),

3288: \qquad\psi\in\C,

3289: \end{equation}

3290: where $\nabla$ denotes the gradient with respect to

3291: $\Re\psi$, $\Im\psi$;

3292: Then $\widetilde{F}(e\sp{is}\psi)=e\sp{is}\widetilde{F}(\psi)$ for any $\psi\in\C$, $s\in\R$.

3293: 

3294: We consider the Cauchy problem of type

3295: (\ref{KG}) with the modified

3296: nonlinearity,

3297: \begin{equation}\label{KG-a}

3298: \left\{

3299: \begin{array}{l}

3300: \ddot\psi(x,t)

3301: =\psi''(x,t)-m^2\psi(x,t)+\delta(x)\widetilde F(\psi(0,t)),

3302: \qquad

3303: x\in\R,

3304: \quad t\in\R,

3305: \\

3306: \psi\at{t=0}=\psi\sb 0(x),

3307: \qquad

3308: \dot\psi\at{t=0}=\pi\sb 0(x),

3309: \end{array}\right.

3310: \end{equation}

3311: which we rewrite in the vector form

3312: in terms of

3313: $\Psi=\left[\begin{array}{c}\psi(x,t)\\\pi(x,t)\end{array}\right]$,

3314: similarly to (\ref{KG-cp}):

3315: \begin{equation}

3316: \dot\Psi

3317: =\left[\begin{array}{cc}0&1\\\p\sb x^2-m^2&0\end{array}\right]\Psi

3318: +\delta(x)\left[\begin{array}{c}0\\\widetilde{F}(\psi)\end{array}\right],

3319: \quad

3320: \quad\Psi\at{t=0}

3321: =\Psi\sb 0

3322: \equiv\left[\begin{array}{c}\psi\sb 0(x)\\\pi\sb 0(x)\end{array}\right].

3323: \label{KG-ap}

3324: \end{equation}

3325: This is a Hamiltonian system, with the Hamilton functional

3326: \begin{equation}\label{KG-a-h}

3327: \widetilde{\mathcal{H}}(\Psi)

3328: =\int\limits\sb\R

3329: \left(

3330: \abs{\pi}\sp 2

3331: +\abs{\nabla\psi}\sp 2+m^2\abs{\psi}\sp 2

3332: \right)\,dx

3333: +\widetilde{U}(\psi(0,t)),

3334: \quad

3335: \Psi=\left[\begin{array}{c}\psi(x)\\

3336: \pi(x)\end{array}\right]\in\E,

3337: \end{equation}

3338: which is Fr\'echet differentiable in the space

3339: ${\E}=H\sp 1\times L\sp 2$.

3340: By the Sobolev embedding theorem,

3341: $L\sp\infty(\R)\subset H\sp 1(\R)$,

3342: and there is the following inequality:

3343: \begin{equation}\label{sobolev-embedding}

3344: \norm{\psi}\sb{L\sp\infty}^2

3345: \le\frac{1}{2m}

3346: (\norm{\psi'}\sb{L\sp 2}^2+m^2\norm{\psi}\sb{L\sp 2}^2)

3347: \le\frac{1}{2m}

3348: \norm{\Psi}\sb{\E}^2.

3349: \end{equation}

3350: Thus, (\ref{new-U-2}) leads to

3351: \begin{equation}\label{bound-on-u}

3352: \widetilde{U}(\psi(0))

3353: \ge {A}-{B}\norm{\psi}\sb{L\sp\infty}^2

3354: \ge {A}-\frac{{B}}{2m}\norm{\Psi}\sb{\E}^2.

3355: \end{equation}

3356: Taking into account (\ref{KG-a-h}),

3357: we obtain the inequality

3358: \begin{equation}

3359: \norm{\Psi}\sb{\E}\sp 2

3360: =2\widetilde{\mathcal{H}}(\Psi)-2 \widetilde{U}(\psi(0))

3361: \le 2\widetilde{\mathcal{H}}(\Psi)-2{A}+\frac{{B}}{m}

3362: \norm{\Psi}\sb{\E}^2,

3363: \qquad

3364: \Psi\in\E,

3365: \end{equation}

3366: which implies

3367: \begin{equation}\label{t-bound-1}

3368: \norm{\Psi}\sb{\E}\sp 2

3369: \le\frac{2m}{m-{B}}\left(\widetilde{\mathcal{H}}(\Psi)-{A}\right),

3370: \qquad\Psi\in\E.

3371: \end{equation}

3372: 

3373: \begin{lemma}\label{lemma-same-u}

3374: \begin{enumerate}

3375: \item

3376: There is the identity

3377: $\widetilde{\mathcal{H}}(\Psi\sb 0)

3378: =\mathcal{H}(\Psi\sb 0)$.

3379: \item

3380: If $\Psi=\left[\begin{array}{c}\psi(x)\\\pi(x)\end{array}\right]\in{\E}$

3381: satisfies

3382: $\widetilde{\mathcal{H}}(\Psi)\le\widetilde{\mathcal{H}}(\Psi\sb 0)$,

3383: then

3384: $\ \widetilde{U}(\psi(0))=U(\psi(0))$.

3385: \end{enumerate}

3386: \end{lemma}

3387: 

3388: \begin{proof}

3389: \begin{enumerate}

3390: \item

3391: According to (\ref{t-bound-1}),

3392: the Sobolev embedding (\ref{sobolev-embedding}),

3393: and the choice of $\Lambda(\Psi\sb 0)$ in (\ref{def-Lambda}),

3394: \begin{equation}

3395: \norm{\psi\sb 0}\sb{L\sp\infty}^2

3396: \le\frac{1}{2m}\norm{\Psi\sb 0}\sb{\E}^2

3397: \le\frac{\mathcal{H}(\Psi\sb 0)-{A}}{m-{B}}

3398: =\Lambda(\Psi\sb 0)^2.

3399: \end{equation}

3400: Thus, according to the choice of $\widetilde{U}$

3401: (equality (\ref{new-U})),

3402: $

3403: \widetilde{U}(\psi\sb 0(0))=U(\psi\sb 0(0)),

3404: $

3405: proving ({\it i}).

3406: \item

3407: By (\ref{sobolev-embedding}), (\ref{t-bound-1}),

3408: the condition

3409: $\widetilde{\mathcal{H}}(\Psi)\le\widetilde{\mathcal{H}}(\Psi\sb 0)$,

3410: and part ({\it i}) of the Lemma, we have:

3411: \[

3412: \norm{\psi}\sb{L\sp\infty}^2

3413: \le\frac{1}{2m}\norm{\Psi}\sb{\E}^2

3414: \le\frac{\widetilde{\mathcal{H}}(\Psi)-{A}}{m-{B}}

3415: \le\frac{\widetilde{\mathcal{H}}(\Psi\sb 0)-{A}}{m-{B}}

3416: =\frac{\mathcal{H}(\Psi\sb 0)-{A}}{m-{B}}

3417: =\Lambda(\Psi\sb 0)^2.

3418: \]

3419: Hence, ({\it ii}) follows by (\ref{new-U}).

3420: \end{enumerate}

3421: \qed\end{proof}

3422: 

3423: \begin{remark}

3424: We will show that if $\Psi(t)$ solves (\ref{KG-ap}),

3425: then

3426: $\widetilde{\mathcal{H}}(\Psi(t))=\widetilde{\mathcal{H}}(\Psi\sb 0)$,

3427: and therefore

3428: $\widetilde{U}(\psi(0,t))=U(\psi(0,t))$

3429: by Lemma~\ref{lemma-same-u}~({\it ii}).

3430: Hence, $\widetilde{F}(\psi(0,t))=F(\psi(0,t))$

3431: for all $t\ge 0$,

3432: allowing us to conclude that

3433: $\Psi(t)$ solves (\ref{KG-cp})

3434: as well as (\ref{KG-ap}).

3435: \end{remark}

3436: 

3437: 

3438: \subsection*{Local well-posedness}

3439: The solution to the Cauchy problem

3440: \begin{equation}

3441: \dot\Xi

3442: =\left[\begin{array}{cc}0&1\\

3443: \p\sb x^2-m^2&0\end{array}\right]\Xi,

3444: \qquad

3445: \Xi(x,0)

3446: =\Xi\sb 0(x)

3447: =\left[\begin{array}{c}\xi\sb 0(x)\\\eta\sb 0(x)\end{array}\right]

3448: \end{equation}

3449: is represented by

3450: \begin{equation}\label{integral-representation-free}

3451: \Xi(x,t)

3452: =W\sb 0(t)

3453: \Xi\sb 0

3454: =

3455: \int\sb\R

3456: \left[\begin{array}{cc}\dot{G}(x-y,t)&{G}(x-y,t)\\

3457: \ddot{G}(x-y,t)&\dot{G}(x-y,t)\end{array}\right]

3458: \left[\begin{array}{c}\xi\sb 0(y)\\\eta\sb 0(y)\end{array}\right]

3459: \,dy,

3460: \end{equation}

3461: where ${G}(x,t)$

3462: is the forward fundamental solution to the Klein-Gordon equation,

3463: ${G}(x,t)

3464: =\theta(t-\abs{x})J\sb 0(m\sqrt{t\sp 2-x\sp 2})/2,

3465: $

3466: with $J\sb 0$ being the Bessel function

3467: (see e.g. \cite{MR1364201}).

3468: Then

3469: the solution to the Cauchy problem (\ref{KG-ap})

3470: can be represented by

3471: \begin{eqnarray}\label{integral-representation}

3472: &&

3473: \Psi(x,t)

3474: =W\sb 0(t)\Psi\sb 0

3475: +{Z}[\psi(0,\cdot)](t),

3476: \nonumber

3477: \\

3478: &&

3479: {\rm where}

3480: \quad

3481: {Z}[\psi(0,\cdot)](t):=

3482: \int\sb 0\sp{t}

3483: W\sb 0(t-s)

3484: \left[\begin{array}{c}0

3485: \\

3486: \delta(\cdot)\widetilde{F}(\psi(0,s))

3487: \end{array}\right]

3488: \,ds.

3489: \end{eqnarray}

3490: 

3491: \begin{lemma}\label{lemma-bessel}

3492: For any

3493: nonnegative integers $\xx$ and $\yy$

3494: there is a constant $C\sb{\xx,\yy}>0$ such that

3495: \begin{equation}\label{bessel-many}

3496: \abs{\p\sb x\sp\xx \p\sb t\sp\yy J\sb 0(m\sqrt{t^2-x^2})}

3497: \le C\sb{\xx,\yy}(1+t)^{\xx+\yy},

3498: \qquad

3499: \abs{x}<t.

3500: \end{equation}

3501: \end{lemma}

3502: 

3503: \begin{proof}

3504: The proof immediately follows from the observation

3505: that all the derivatives of the Bessel function  $J\sb 0(z)$

3506: are bounded for $z\in\R$, and that $J\sb 0(z)$

3507: is an absolutely converging Taylor series in even powers of $z$.

3508: Hence,

3509: all derivatives of the function $J\sb 0(\sqrt{r})$

3510: in $r$ are bounded for $r\ge 0$.

3511: \qed\end{proof}

3512: 

3513: The next lemma establishes

3514: the contraction principle

3515: for the integral equation

3516: (\ref{integral-representation}).

3517: 

3518: \begin{lemma}\label{lemma-bounds}

3519: There exists a constant $C>0$ so that

3520: for any two functions

3521: $

3522: \Psi\sb{k}(\cdot,t)

3523: =

3524: \left[\begin{array}{c}\psi\sb{k}(\cdot,t)\\

3525: \pi\sb{k}(\cdot,t)

3526: \end{array}\right]\in C([0,1],\E),

3527: $

3528: $k=1,\,2$,

3529: we have:

3530: \[

3531: \norm{

3532: {Z}[\psi\sb 1(0,\cdot)](t)-{Z}[\psi\sb 2(0,\cdot)](t)

3533: }\sb{\E}

3534: \le

3535: C t^{1/2}

3536: \sup\sb{s\in[0,t]}\norm{\Psi\sb 1(\cdot,s)-\Psi\sb 2(\cdot,s)}\sb{\E},

3537: \]

3538: for $0\le t\le 1$.

3539: \end{lemma}

3540: 

3541: \begin{proof}

3542: According to (\ref{integral-representation-free}) and

3543: (\ref{integral-representation}),

3544: \[

3545: {Z}[\psi\sb 1(0,\cdot)](t)-{Z}[\psi\sb 2(0,\cdot)](t)=

3546: \left[\begin{array}{c}

3547: I(x,t)\\

3548: \partial_t I(x,t)

3549: \end{array}\right],

3550: \]

3551: where

3552: \[

3553: I(x,t):=\displaystyle\int\sb 0\sp t

3554: {G}(x,t-s)

3555: \left(

3556: \widetilde{F}(\psi\sb 1(0,s))-\widetilde{F}(\psi\sb 2(0,s))

3557: \right)\,ds.

3558: \]

3559: First we prove the

3560: $L\sp 2$ estimate for $I(x,t)$.

3561: By the Sobolev embedding theorem,

3562: \begin{eqnarray}

3563: \Norm{

3564: I(\cdot,t)

3565: }\sb{L\sp 2}

3566: &\le&

3567: \cnst

3568: \Norm{

3569: \int\sb 0\sp t

3570: \theta(t-s-\abs{x})

3571: \abs{\widetilde{F}(\psi\sb 1(0,s))-\widetilde{F}(\psi\sb 2(0,s))}

3572: \,ds

3573: }\sb{L\sp 2}

3574: \nonumber\\

3575: \nonumber\\

3576: &

3577: \le&

3578: \cnst

3579: \sup\sb{z\in\C}\abs{\nabla\widetilde{F}(z)}

3580: \Norm{

3581: \int\sb 0\sp t

3582: \theta(t-s-\abs{x})

3583: \,ds

3584: }\sb{L\sp 2}

3585: \,

3586: \sup\sb{s\in[0,t]}

3587: \norm{\psi\sb 1(\cdot,s)-\psi\sb 2(\cdot,s)}\sb{H\sp 1}

3588: \nonumber\\

3589: \nonumber\\

3590: &

3591: \le&

3592: C'

3593: \,t\sp{3/2}

3594: \sup\sb{s\in[0,t]}

3595: \norm{\psi\sb 1(\cdot,s)-\psi\sb 2(\cdot,s)}\sb{H\sp 1},

3596: \label{e-l2}

3597: \end{eqnarray}

3598: where we took into account that

3599: $\abs{\nabla\widetilde{F}(z)}$

3600: is bounded

3601: due to the choice of $\widetilde{U}$.

3602: 

3603: Similarly, we derive

3604: the  $L\sp 2$ estimate

3605: for the derivative $\p\sb x I(x,t)$.

3606: We first analyze

3607: \[

3608: \p\sb x{G}(x,t-s)=

3609: \frac 1 2

3610: \theta(t-s-\abs{x})\p\sb x J\sb 0(m\sqrt{(t-s)\sp 2-x\sp 2})

3611: -\frac 1 2\sgn{x}\,\delta(t-s-\abs{x}).

3612: \]

3613: By Lemma~\ref{lemma-bessel} for $\abs{x}\le\abs{t-s}\le 1$,

3614: we have

3615: $

3616: \abs{

3617: \p\sb x J\sb 0(m\sqrt{(t-s)\sp 2-x\sp 2})}

3618: \le\cnst;

3619: $

3620: We conclude that

3621: $

3622: \Norm{

3623: \p\sb x I(\cdot,t)

3624: }\sb{L^2}

3625: $

3626: is bounded by

3627: \begin{eqnarray}\label{d-e-l2}

3628: &&

3629: \Norm{

3630: \int\sb 0\sp t

3631: \left[C

3632: \theta(t-s-\abs{x})

3633: +

3634: \frac{\delta(t-s-\abs{x})}{2}

3635: \right]

3636: \,ds

3637: }\sb{L\sp 2}

3638: \,\sup\sb{s\in[0,t]}

3639: \Abs{\widetilde{F}(\psi\sb 1(0,s))-\widetilde{F}(\psi\sb 2(0,s))}

3640: \nonumber\\

3641: \nonumber\\

3642: &&

3643: \qquad\qquad

3644: \le

3645: C

3646: \Norm{

3647: \theta(t-\abs{x})

3648: }\sb{L\sp 2}

3649: \,\sup\sb{s\in[0,t]}

3650: \norm{\psi\sb 1(\cdot,s)-\psi\sb 2(\cdot,s)}\sb{H\sp 1}

3651: \nonumber\\

3652: \nonumber\\

3653: &&

3654: \qquad\qquad

3655: \le

3656: C'

3657: \,t^{1/2}

3658: \,\sup\sb{s\in[0,t]}

3659: \norm{\psi\sb 1(\cdot,s)-\psi\sb 2(\cdot,s)}\sb{H\sp 1}.

3660: \end{eqnarray}

3661: The $L^2$-norm of $\p\sb t I(x,t)$

3662: is estimated similarly.

3663: \qed\end{proof}

3664: 

3665: For $E>0$,

3666: let us denote

3667: ${\E}\sb E=\{\Psi\sb 0\in \E

3668: \sothat

3669: \mathcal{H}(\Psi\sb 0)\le E\}$.

3670: 

3671: \begin{corollary}

3672: \label{cor-l8-existence}

3673: \begin{enumerate}

3674: \item

3675: For any $E>0$ there exists $\tau=\tau(E)>0$

3676: such that for any $\Psi\sb 0\in {\E}\sb E$

3677: there is a unique solution

3678: $\Psi(x,t)\in C([0,\tau],{\E})$

3679: to the Cauchy problem {\rm (\ref{KG-ap})}

3680: with the initial condition

3681:  $\Psi(0)=\Psi\sb 0$.

3682: \item

3683: The map $W(t):\;\Psi\sb 0\mapsto \Psi(t)$,

3684:  $t\in[0,\tau]$

3685: are continuous maps from ${\E}\sb E$ to $\E$.

3686: \end{enumerate}

3687: \end{corollary}

3688: 

3689: \subsection*{Smoothness of the solution}

3690: 

3691: In this section,

3692: we will study the smoothness of the solution

3693: \[

3694: \Psi(x,t)=(\psi(x,t),\pi(x,t))\in C([0,\tau],\E)

3695: \]

3696: constructed in Corollary~\ref{cor-l8-existence}~({\it i})

3697: assuming that $\psi\sb 0(x),\pi\sb 0(x)\in C\sb 0^\infty(\R)$.

3698: According to the integral representation (\ref{integral-representation}),

3699: $\psi(x,t)$, $t\in [0,\tau]$, can be represented as

3700: \begin{equation}\label{integral-representation-psi}

3701: \psi(x,t)

3702: =

3703: \int\limits\sb\R

3704: \left(

3705: \dot{G}(x-y,t)\psi\sb 0(y)+{G}(x-y,t)\pi\sb 0(y)

3706: \right)

3707: dy

3708: +

3709: \int\limits\sb 0\sp t {G}(x,t-s)\widetilde{F}(\psi(0,s))\,ds.

3710: \end{equation}

3711: 

3712: First, let us prove the smoothness

3713: of the function $\psi(0,t)$.

3714: 

3715: \begin{lemma}\label{lemma-smoothness-0}

3716: $\psi(0,t)\in C\sp\infty([0,\tau])$.

3717: \end{lemma}

3718: 

3719: \begin{proof}

3720: The integral representation (\ref{integral-representation-psi})

3721: implies  that,

3722: for $t\in [0,\tau]$,

3723: \begin{equation}\label{imt}

3724: \psi(0,t)

3725: =

3726: \int\limits\sb\R

3727: \left(

3728: \dot{G}(y,t)\psi\sb 0(y)+{G}(y,t)\pi\sb 0(y)

3729: \right)dy

3730: +

3731: \frac{1}{2}

3732: \int\limits\sb 0\sp t

3733: J\sb 0(m(t-s))

3734: \widetilde{F}(\psi(0,s))

3735: \,ds.

3736: \end{equation}

3737: The first integral is a smooth function.

3738: Further,

3739: from $\norm{\psi(\cdot,t)}\sb{H\sp 1}\le C<\infty$, $t\in [0,\tau]$,

3740: we conclude that $\abs{\psi(0,t)}$ is bounded.

3741: Hence, (\ref{imt}) implies

3742: that $\psi(0,\cdot)\in C([0,\tau])$,

3743: and then by induction that $\psi(0,\cdot)\in C\sp\infty([0,\tau])$

3744: since the Bessel function is smooth.

3745: \qed\end{proof}

3746: 

3747: Now, from (\ref{integral-representation-psi}),

3748: we conclude that $\psi(x,t)$ is smooth

3749: away from the singularities of ${G}(x,t)$.

3750: 

3751: \begin{proposition}\label{coro}

3752: The solution

3753: $\psi(x,t)$ is piecewise smooth

3754: inside each of the four regions of $[0,\tau]\times\R$

3755: cut off by the lines $x=0$ and $x=\pm t$.

3756: \end{proposition}

3757: 

3758: \begin{proof}

3759: The first integral

3760: in the right-hand side of (\ref{integral-representation-psi})

3761: is infinitely smooth in $x$ and $t$ for all $x\in\R$, $t\ge 0$.

3762: Now let us consider the second integral

3763: in the right-hand side of (\ref{integral-representation-psi}),

3764: which could be written as follows:

3765: \begin{equation}\label{ir-psi-two}

3766: \frac{\theta(t-\abs{x})}{2}

3767: \int\sb 0\sp{t-\abs{x}}

3768: J\sb 0(m\sqrt{(t-s)^2-x^2})

3769: \widetilde{F}(\psi(0,s))\,ds.

3770: \end{equation}

3771: Here the function

3772: $\widetilde{F}(\psi(0,s))$

3773: is smooth in $s\in[0,\tau]$ by Lemma~\ref{lemma-smoothness-0}.

3774: By Lemma~\ref{lemma-bessel},

3775: all the partial derivatives of

3776: $J\sb 0(m\sqrt{(t-s)^2-x^2})$

3777: in $x$ and $t$

3778: are continuous and uniformly bounded for $\abs{x}<t-s$,

3779: $t\le\tau$.

3780: Therefore, (\ref{ir-psi-two})

3781: is smooth,

3782: with all the derivatives uniformly bounded,

3783: in each of the regions

3784: $0\le x\le t$, $-t\le x\le 0$.

3785: In the regions $\abs{x}>t$,

3786: (\ref{ir-psi-two}) is identically equal to zero.

3787: \qed\end{proof}

3788: 

3789: \begin{lemma}\label{lemma-psi-dot}

3790: For

3791: $0<t\le\tau$,

3792: \begin{equation}\label{psi-dot-pm}

3793: \lim\sb{x\to 0-}\dot\psi(x,t)=\lim\sb{x\to 0+}\dot\psi(x,t).

3794: \end{equation}

3795: \end{lemma}

3796: 

3797: \begin{proof}

3798: We have to analyze only

3799: the contribution from the second term

3800: in the right-hand side of (\ref{integral-representation-psi}),

3801: that is,

3802: \[

3803: \p\sb t

3804: \int\limits\sb 0\sp t {G}(x,t-s)\widetilde{F}(\psi(0,s))\,ds

3805: =

3806: {G}(x,0+)\widetilde{F}(\psi(0,t))

3807: +

3808: \int\limits\sb 0\sp t

3809: \dot{G}(x,t-s)\widetilde{F}(\psi(0,s))\,ds.

3810: \]

3811: The first term in the right-hand side

3812: is equal to zero for $x\ne 0$.

3813: The second term

3814: is continuous since

3815: the Green function

3816: ${G}(x,t-s)$

3817: is smooth at $x=0$ for $t-s>0$.

3818: \qed\end{proof}

3819: 

3820: \begin{lemma}\label{lemma-psi-dot-prime}

3821: For

3822: $0<t\le\tau$,

3823: \begin{enumerate}

3824: \item

3825: $

3826: \dot\psi(x,t)+\psi'(x,t)

3827: $

3828: is continuous across

3829: the characteristic $x=t$.

3830: \item

3831: $

3832: \dot\psi(x,t)-\psi'(x,t)

3833: $

3834: is continuous across the characteristic $x=-t$.

3835: \end{enumerate}

3836: \end{lemma}

3837: 

3838: \begin{proof}

3839: The proofs for both statements of the Lemma are identical;

3840: we will only prove the first statement with $x>0$.

3841: We have to study only

3842: the contribution from the second term

3843: in the right-hand side of (\ref{integral-representation-psi}),

3844: i.e.

3845: \begin{equation}\label{psi-dot-rhs-2}

3846: (\p\sb t+\p\sb x)

3847: \int\limits\sb 0\sp t {G}(x,t-s)\widetilde{F}(\psi(0,s))\,ds

3848: =

3849: \int\limits\sb 0\sp t

3850: (\p\sb t+\p\sb x){G}(x,t-s)

3851: \widetilde{F}(\psi(0,s))\,ds.

3852: \end{equation}

3853: Here we took into account that, as above,

3854: ${G}(x,0+)\widetilde{F}(\psi(0,t))=0$ for $x\ne 0$.

3855: Next key observation

3856: is that,

3857: for $x>0$, the derivative $\p\sb t+\p\sb x$

3858: applied to ${G}(x,t)$,

3859: does not produce a delta-function:

3860: \[

3861: (\p\sb t+\p\sb x){G}(x,t)

3862: =\frac 1 2

3863: \left\{

3864: \theta(t-x)

3865: (\p\sb t+\p\sb x)J\sb 0(m\sqrt{t^2-x^2})

3866: \right\}.

3867: \]

3868: Hence, the integral (\ref{psi-dot-rhs-2})

3869: is continuous in $x$ and $t$ across the line $x=t$, $0<t\le\tau$

3870: by Lemma~\ref{lemma-bessel}

3871: \qed\end{proof}

3872: 

3873: 

3874: 

3875: 

3876: \subsection*{Energy conservation and global well-posedness}

3877: 

3878: \begin{lemma}\label{lemma-e-conserved}

3879: For the solution

3880: to the Cauchy problem (\ref{KG-ap})

3881: with the initial data

3882: $\Psi\sb 0\in\E$,

3883: the energy is conserved:

3884: $\widetilde{\mathcal{H}}(\Psi(t))=\const$, $t\in [0,\tau]$.

3885: \end{lemma}

3886: 

3887: \begin{proof}

3888: We follow \cite{MR1359949}.

3889: First, we prove that the energy is conserved

3890: for the smooth initial data with compact support:

3891: $\Psi\sb 0=\left[\begin{array}{c}\psi\sb 0\\\pi\sb 0\end{array}\right]$,

3892: with

3893: $\psi\sb 0$, $\pi\sb 0\in C\sp\infty\sb 0(\R)$.

3894: Consider the norm (\ref{def-e}),

3895: \begin{equation}

3896: \norm{\Psi(t)}\sb{\E}^2

3897: =

3898: \int\sb{-\infty}\sp\infty

3899: [

3900: \abs{\dot\psi}\sp 2+\abs{\psi'}\sp 2+m^2\abs{\psi}\sp 2

3901: ]\,dx,

3902: \qquad

3903: t\in[0,\tau].

3904: \end{equation}

3905: We split this integral into four pieces: The integration over

3906: $(-\infty,-t)$, $(-t,0)$, $(0,t)$, and $(t,\infty)$.

3907: By Proposition~\ref{coro},

3908: on the support of each of these integrals

3909: $\psi(x,t)$ for $t\in[0,\tau]$

3910: is a smooth function of $x$ and $t$.

3911: Then, differentiating,

3912: we may express $\p\sb t \norm{\Psi(t)}\sb{\E}^2$ as

3913: \begin{eqnarray}\label{p-t-e}

3914: &&

3915: \p\sb t \norm{\Psi(t)}\sb{\E}^2

3916: =

3917: \Big[

3918: \abs{\dot\psi}\sp 2+\abs{\psi'}\sp 2+m^2\abs{\psi}\sp 2

3919: \Big]\sb{x=-t-0}\sp{x=-t+0}

3920: \\

3921: &&

3922: -

3923: \Big[

3924: \abs{\dot\psi}\sp 2+\abs{\psi'}\sp 2+m^2\abs{\psi}\sp 2

3925: \Big]\sb{x=t-0}\sp{x=t+0}

3926: +2\int\sb{-\infty}\sp{\infty}

3927: [

3928: \dot\psi\ddot\psi+\psi'\dot\psi'+m^2\psi\dot\psi

3929: ]\,dx,

3930: \quad

3931: t\in[0,\tau].

3932: \nonumber

3933: \end{eqnarray}

3934: The terms $m^2\abs{\psi}\sp 2$ could be discarded

3935: due to continuity of $\psi$

3936: across the characteristics $x=\pm t$.

3937: Integrating by parts the terms $\psi'\dot\psi'$

3938: and using the cancellations of the integrals

3939: due to equation (\ref{KG-a}) away from $x=0$,

3940: we get:

3941: \begin{eqnarray}

3942: &&\p\sb t \norm{\Psi(t)}\sb{\E}^2

3943: =

3944: \Big[

3945: \abs{\dot\psi}\sp 2+\abs{\psi'}\sp 2

3946: -2\psi'\dot\psi

3947: \Big]\sb{x=-t-0}\sp{x=-t+0}

3948: \nonumber\\

3949: &&\qquad\qquad

3950: -

3951: \Big[

3952: \abs{\dot\psi}\sp 2+\abs{\psi'}\sp 2

3953: +2\psi'\dot\psi

3954: \Big]\sb{x=t-0}\sp{x=t+0}

3955: -2\Big[\psi'\dot\psi\Big]\sb{x=0-}\sp{x=0+}

3956: \nonumber\\

3957: \nonumber\\

3958: &&\qquad

3959: =

3960: \Big[

3961: (\dot\psi-\psi')\sp 2

3962: \Big]\sb{x=-t-0}\sp{x=-t+0}

3963: -\Big[

3964: (\dot\psi+\psi')\sp 2

3965: \Big]\sb{x=t-0}\sp{x=t+0}

3966: -2\Big[\psi'\dot\psi\Big]\sb{x=0-}\sp{x=0+}.

3967: \label{last-line}

3968: \end{eqnarray}

3969: According to Lemma~\ref{lemma-psi-dot-prime},

3970: the first two terms in (\ref{last-line})

3971:  do not give any contribution.

3972: Let us compute the contribution of the last term.

3973: According to Lemma~\ref{lemma-psi-dot},

3974: $\dot\psi(0\pm,t)=\dot\psi(0,t)$

3975: for $t\in [0,\tau]$, therefore

3976: \[

3977: \left[\psi'\dot\psi\right]\sb{x=0-}\sp{x=0+}

3978: =\left[\psi'(x,t)\right]\sb{x=0-}\sp{x=0+}\psi(0,t)

3979: =

3980: -\widetilde{F}(\psi(0,t))\dot\psi(0,t)

3981: =\frac{d}{dt}\widetilde{U}(\psi(0,t)).

3982: \]

3983: In the second equality,

3984: we computed the jump of $\psi'$

3985: using equation (\ref{KG-a}) and the piecewise smoothness of the solution.

3986: We conclude that

3987: \[

3988: \frac{d}{dt}

3989: \left\{\frac 1 2\norm{\Psi(t)}\sb{\E}^2+\widetilde{U}(\psi(0,t))\right\}=0,

3990: \]

3991: and hence the value of the functional

3992: $\widetilde{\mathcal{H}}$

3993: defined in (\ref{KG-a-h}) is conserved.

3994: 

3995: Since we proved the energy conservation

3996: for the initial data that constitute a dense set

3997: in ${\E}$

3998: and since the dynamical group

3999: is continuous

4000: in ${\E}$

4001: by

4002: Corollary~\ref{cor-l8-existence}~({\it ii}),

4003: we conclude that the energy is conserved

4004: for arbitrary initial data from ${\E}$.

4005: \qed\end{proof}

4006: 

4007: \begin{corollary}

4008: \label{coroll}

4009: \begin{enumerate}

4010: \item

4011: The solution $\Psi$ to the Cauchy problem (\ref{KG-ap})

4012: with the initial data

4013: $\Psi\at{t=0}=\Psi\sb 0\in {\E}$

4014: exists globally:

4015: $

4016: \Psi\in C\sb{b}(\R,{\E}).

4017: $

4018: \item

4019: The energy

4020: is conserved:

4021: $

4022: \widetilde{\mathcal{H}}(\Psi(t))=\widetilde{\mathcal{H}}(\Psi\sb 0),

4023: \qquad t\ge 0.

4024: $

4025: \end{enumerate}

4026: \end{corollary}

4027: 

4028: \begin{proof}

4029: Corollary~\ref{cor-l8-existence}~({\it i})

4030: yields a solution

4031: $\Psi\in L\sp\infty([0,\tau],{\E})$

4032: with a positive $\tau=\tau(E)$.

4033: However,

4034: the value of $\mathcal{H}(\Psi(t))$

4035: is conserved for $t\le \tau$

4036: by Lemma~\ref{lemma-e-conserved}.

4037: Corollary~\ref{cor-l8-existence}~({\it i})

4038: allows us to extend $\Psi$

4039: to the interval $[\tau,2 \tau]$,

4040: and eventually to all $t\ge 0$.

4041: In the same way we extend the solution $\Psi(t)$

4042: for all $t<0$.

4043: \qed\end{proof}

4044: 

4045: 

4046: \subsection*{Conclusion of the proof of Theorem~\ref{theorem-well-posedness}}

4047: 

4048: The trajectory

4049: $\Psi

4050: =\scriptsize{\left[\!\begin{array}{c}\psi(x,t)\\\pi(x,t)\end{array}\!\right]}

4051: \in C\sb{b}(\R,{\E})$

4052: is a solution to (\ref{KG-ap}),

4053: for which

4054: Corollary~\ref{coroll}~({\it ii})

4055: together with Lemma~\ref{lemma-same-u}~({\it i})

4056: imply the energy conservation  (\ref{ec}).

4057: By Lemma~\ref{lemma-same-u}~({\it ii}),

4058: $\widetilde{U}(\psi(0,t))=U(\psi(0,t))$,

4059: for all $t\in\R$.

4060: This tells us that

4061: $\psi(x,t)$ is a solution to (\ref{KG}).

4062: Finally, the a priori bound (\ref{eb}) follows from

4063: (\ref{t-bound-1}) and the conservation of $\mathcal{H}(\Psi(t))$.

4064: This finishes the proof of Theorem~\ref{theorem-well-posedness}.

4065: 

4066: \noindent

4067: {\bf Acknowledgements.}

4068: The authors thank H.~Brezis,

4069: V.S.~Buslaev, J.~Ginibre, P.-L. Lions, L.~Nirenberg, J.~Shatah,

4070: A.~Shnirelman, H.~Spohn,

4071: W.~Strauss, G.~Velo, and M.I.~Vishik for fruitful discussions,

4072: and

4073: A.~Merzon

4074: for helpful remarks on the Titchmarsh theorem.

4075: The authors are also indebted to P.~Joly, F.~Collino, and T.~Fouquet

4076: from Project ONDES (INRIA),

4077: and to A.~Vinnichenko

4078: for the help with numerical experiments

4079: for nonlinear wave equations.

4080: 

4081: %The first author was supported in part

4082: %by Max-Planck Institute for Mathematics in the Sciences (Leipzig),

4083: %the Wolfgang Pauli Institute and the Faculty of Mathematics,

4084: %Vienna University,

4085: %by DFG Grant (436 RUS 113/615/0-1),

4086: %and by FWF Grant P19138-N13.

4087: %

4088: %The second author was supported in part by

4089: %Max-Planck Institute for Mathematics in the Sciences (Leipzig)

4090: %and by the NSF Grants DMS-0434698 and DMS-0600863.

4091: 

4092: 

4093: \bibliographystyle{amsalpha}

4094: \bibliography{ubk-mathsci,ubk-local}

4095: 

4096: %

4097: \end{document}

4098: