1: \documentclass[12pt]{article}

2: \usepackage{amssymb,graphicx}

3: \setlength{\oddsidemargin}{0mm} \setlength{\evensidemargin}{0mm}

4: \setlength{\topmargin}{-15mm} \setlength{\textheight}{220mm}

5: \setlength{\textwidth}{155mm}

6: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

7: \def\no{\noindent}

8: \def\d#1#2{\frac{\displaystyle #1}{\displaystyle #2}}

9: \def\f#1#2{\frac{#1}{#2}}

10: \def\f{\frac}

11: %do not use \d : use \f since \d do not conform in font.

12: \def\di{\displaystyle\int}

13: \def\ds{\displaystyle\sum}

14: \def\dl{\displaystyle\lim}

15: \def\ld{\lambda}

16: \def\ov{\overline}

17: \def\wto{\rightharpoonup}

18: \def\tr{\triangledown}

19: \def\ii{\int\!\!\!\int}

20: \def\iint{\int\int}

21: \def\dint{\int\!\!\int}

22: \def\vap{\varepsilon}

23: \def\wt{\widetilde}

24: %\def\BR{{\mathbb R}} %Do not use \BR, use \R if possible

25: \def\R{{\mathbb R}}

26: \def\S{{\cal S}}

27: \def\D{{\cal D}}

28: \def\C{{\cal C}}

29: \def\E{{\cal E}}

30: \def\Leb{{\rm Leb}}

31: \def\ep{\epsilon}

32: \def\pa{\partial}

33: \def\p{\partial}

34: \def\vphi{\varphi}

35: \def\vp{\varphi}

36: \def\L{{\bf L}}

37: \def\Ll{L_{\rm loc}} %%This is a cute one: use {\rm }instead of\mbox{}

38: \def\H{{\cal H}}

39: \def\O{{\cal O}}

40: \def\N{{\cal N}}

41: \def\ga{\gamma}

42: \def\v{\vskip 1em}

43: \def\vs{\vskip 2em}

44: \def\vsk{\vskip 4em}

45: %\def\n{{\bf n}}

46: \def\th{\theta}

47: \def\de{\delta}

48: \def\be{\beta}

49: \def\ve{\varepsilon}

50: \def\vap{\varepsilon}

51: \def\na{\nabla}

52: \def\la{\lambda}

53: \def\bu{\bigtriangleup}

54: \def\rw{\rightarrow}

55: \def\al{\alpha}

56: \def\meas{\hbox{meas}\,}

57: \def\La{\Lambda}

58: \def\T{{\cal T}}

59: \def\tv{\hbox{Tot.Var.}}

60: \def\Tilde{\widetilde}

61: \def\Do{{\R^+\!\times\!\R}}

62: \def\Doo{{\R^+\times\R^+\times\R}}

63: \def\sqr#1#2{\vbox{\hrule height .#2pt

64: \hbox{\vrule width .#2pt height #1pt \kern #1pt

65: \vrule width .#2pt}\hrule height .#2pt }}

66: \def\square{\sqr74}

67: \def\endproof{\hphantom{MM}\hfill\llap{$\square$}\goodbreak}

68: %% \def\Do{{\R^+\times\R}}

69: \def\dv{\mbox{\rm div}}

70: \def\uu{\mbox{\bfseries\itshape u}}

71: \def\mm{\mbox{\bfseries\itshape m}}

72: \def\nn{\mbox{\bfseries\itshape n}}

73: \def\xx{\mbox{\bfseries\itshape x}}

74: \def\UU{\mbox{\bfseries\itshape U}}

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

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

77: %nonumber: use $$---$$

78: \newcommand{\ben}{\begin{eqnarray}}

79: \newcommand{\een}{\end{eqnarray}}

80: \newcommand{\beno}{\begin{eqnarray*}}

81: \newcommand{\eeno}{\end{eqnarray*}}

82: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

83: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

84: \newtheorem{Theorem}{Theorem}[section]

85: \newtheorem{Lemma}[Theorem]{Lemma}

86: \newtheorem{Def}{Definition}[section]

87: \newtheorem{Remark}{Remark}[section]

88: \newcommand{\QED}{$\square$}

89: %%%%%%%%%%%-------------------------------------------

90: \setlength{\baselineskip}{1.2\baselineskip}

91: \setlength{\arraycolsep}{0.5mm}

92: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

93: \begin{document}

94: \title{\bf  Global Conservative Solutions to a Nonlinear 

95: Variational Wave Equation}

96: \author{Alberto Bressan and Yuxi Zheng \\[2mm]

97: {\small Department of Mathematics, The Pennsylvania State University }\\[-2mm]

98: {\small E-mail: bressan@math.psu.edu; yzheng@math.psu.edu} }

99: \maketitle

100: 

101: 

102: \begin{abstract}

103: We establish the existence of a conservative weak solution to the

104: Cauchy problem for the nonlinear variational wave equation 

105: $u_{tt} - c(u)(c(u)u_x)_x=0$, for initial data of finite energy. 

106: Here  $c(\cdot)$ is any smooth function with uniformly positive

107: bounded values. 

108: \end{abstract}

109: 

110: \vspace*{3mm}

111: 

112: \noindent \underline{Mathematics Subject Classification (2000):}\quad 

113: 35Q35 \\

114: \noindent \underline{Keywords:}\quad

115: Existence, uniqueness, singularity, coordinate transformation

116: 

117: 

118: 

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

120: 

121: 

122: 

123: 

124: 

125: \section{Introduction}

126: 

127: \setcounter{equation}{0}

128: 

129: We are interested in the Cauchy problem

130: \beq

131: u_{tt} - c(u)\big(c(u) u_x\big)_x=0\,,\label{1.1}

132: \eeq

133: with initial data

134: \beq

135: u(0,x)=u_0(x)\,,\qquad

136: u_t(0,x)=u_1(x)\,.\label{1.2}

137: \eeq

138: Throughout the following, we assume that $c:\R\mapsto \R_+$ is

139: a smooth, bounded, uniformly positive function.

140: Even for smooth initial data, it is well known that the solution

141: can lose regularity in finite time (\cite{[GHZ]}).  It is thus of interest

142: to study whether the solution can be extended beyond the

143: time when a singularity appears.  This is indeed the main

144: concern of the present paper.  In (\cite{[BZZ]}) we considered the

145: related equation

146: \beq

147: u_t+f(u)_x={1\over 2} \int_0^x f''(u)u_x^2\,dx\label{1.3'}

148: \eeq

149: and constructed a semigroup of solutions, depending continuously

150: on the initial data.

151: Here we establish similar results for the nonlinear wave equation

152: (\ref{1.1}).  By introducing new sets of dependent and independent variables,

153: we show that the solution to the Cauchy problem can be

154: obtained as the fixed point of a contractive transformation.

155: Our main result can be stated as follows.

156: \v

157: \no{\bf Theorem 1.} {\it Let $c:\R\mapsto [\kappa^{-1},\,\kappa]$ be

158: a smooth function, for some $\kappa>1$.  Assume that the initial data

159: $u_0$ in (\ref{1.2}) is absolutely continuous, and that

160: $(u_0)_x\in\L^2\,$,   ~$u_1\in\L^2$.

161: Then the Cauchy problem (\ref{1.1})-(\ref{1.2}) admits a weak solution

162: $u=u(t,x)$, defined for all $(t,x)\in\R\times\R$.

163: In the $t$-$x$ plane, the function $u$ is locally H\"older continuous

164: with exponent $1/2$.   This solution $t\mapsto u(t,\cdot)$ is

165: continuously differentiable as a map with values in $\L^p_{\rm loc}$, for all

166: $1\leq p<2$. Moreover, it is Lipschitz continuous w.r.t.~the $\L^2$

167: distance, i.e.

168: \beq

169: \big\|u(t,\cdot)-u(s,\cdot)\big\|_{\L^2}\leq L\,|t-s|\label{1.lip}

170: \eeq

171: for all $t,s\in\R$.

172: The equation (\ref{1.1}) is satisfied in integral sense, i.e.

173: \beq

174: \dint \Big[\phi_t\, u_t - \big(c(u) \phi\big)_x c(u)\,u_x\Big]\,dxdt =0

175: \label{1.4'}

176: \eeq

177: for all test functions $\phi\in\C^1_c$.  Concerning the initial conditions, 

178: the first equality in (\ref{1.2})

179: is satisfied pointwise, while the second holds in  $\L^p_{\rm loc}\,$

180: for $p\in [1,2[\,$.

181: }

182: \v

183: Our constructive procedure yields solutions which depend continuously

184: on the initial data.  Moreover, the ``energy"

185: \beq

186: \E(t)\doteq \f12\int \Big[u_t^2(t,x) +c^2\big(u(t,x)\big) \,u_x^2(t,x)\Big]

187: \,dx\label{1.5'}

188: \eeq

189: remains uniformly bounded. More precisely, one has

190: \v

191: \no{\bf Theorem 2.} {\it A family of weak solutions to the

192: Cauchy problem (\ref{1.1})-(\ref{1.2}) can be constructed with the

193: following additional properties.

194: For every $t\in\R$ one has

195: \beq

196: \E(t)~\leq ~\E_0~\doteq~ \f12\int \Big[ u_1^2(x)+c^2\big(u_0(x)\big) 

197: (u_0)_x^2(x) \Big]\,dx\,.\label{1.7}

198: \eeq

199: Moreover, let a sequence of initial conditions satisfy

200: $$%\beq

201: \big\|(u_0^n)_x- (u_0)_x\big\|_{\L^2}\to 0\,,\qquad\qquad

202: \big\|u_1^n- u_1\big\|_{\L^2}\to 0\,,

203: $$%\eeq

204: and $u^n_0\to u_0$ uniformly on compact sets, as $n\to \infty$.

205: Then one has the convergence of the corresponding solutions $u^n\to u$,

206: uniformly on bounded subsets of the $t$-$x$ plane.}

207: \v

208: It appears in (\ref{1.7}) that the total energy of our solutions

209: may decrease in time.  Yet, we emphasize that

210: our solutions are {\it conservative}, in the following sense.

211: \v

212: \no{\bf Theorem 3.} {\it There exists a continuous family

213: $\{\mu_t\,;~~t\in\R\}$ of

214: positive Radon measures on the real line with the following properties.

215: \begin{description}

216: \item[](i) At every time $t$, one has $\mu_t(\R)=\E_0$.

217: 

218: \item[](ii) For each $t$, the absolutely continuous part of $\mu_t$

219: has density $\f12(u_t^2+c^2\,u_x^2)$ w.r.t.~the Lebesgue measure.

220: 

221: \item[](iii)  For almost every $t\in\R$, the singular part of $\mu_t$

222: is concentrated on the set where $c'(u)=0$.

223: \end{description}

224: }

225: In other words, the total energy represented by the measure $\mu$ is

226: conserved in time.  Occasionally, some of this energy is concentrated

227: on a set of measure zero. At the times $\tau$ when this happens,

228: $\mu_\tau$ has a non-trivial singular part and

229: $\E(\tau)<\E_0$.  The condition (iii) puts some restrictions

230: on the set of such times $\tau$. In particular, if $c'(u)\not= 0$

231: for all $u$, then this set has measure zero.

232: \v

233: The paper is organized as follows.  In the next two subsections

234: we briefly

235: discuss the physical motivations for

236: the equation and recall some known results on its solutions.

237: In Section 2 we introduce a new set of independent and dependent

238: variables, and derive some identities

239: valid for smooth solutions.  We formulate a set of equations in the 

240: new variables which is equivalent to (\ref{1.1}).  Remarkably,

241: in the new variables all singularities disappear: Smooth 

242: initial data lead to

243: globally smooth solutions.

244: In Section 3 we use a contractive transformation in a Banach space 

245: with a suitable weighted norm to show that there is a unique

246: solution to the set of equations in the new variables,

247: depending continuously on the data $u_0, u_1$.

248: Going back to the original variables $u,t,x$,

249: in Section 4 we establish the H\"older continuity

250: of these solutions $u=u(t,x)$, and show

251: that the integral equation (\ref{1.4'}) is satisfied.

252: Moreover, in Section 5, we study the conservativeness of the solutions,

253: establish the energy inequality and the Lipschitz continuity of the map

254: $t\mapsto u(t,\cdot)$.  This already yields a proof

255: of Theorem 2. In Section 6 we study

256: the continuity of the maps

257: $t\mapsto u_x(t,\cdot)$, $t\mapsto u_t(t,\cdot)$, completing the proof

258: of Theorem 1.  The proof of Theorem 3 is given in  Section 7.

259: 

260: \subsection{Physical background of the equation}

261: 

262: Equation (\ref{1.1}) has several physical origins. 

263: In the context of nematic liquid crystals, 

264: it comes as follows. 

265: The mean orientation of the long molecules in a nematic liquid crystal

266: is described by a director field of unit vectors,  

267: ${\mathbf n}\in{\mathbb S}^2$,

268: the unit sphere. Associated with the director field ${\mathbf n}$, there is

269: the well-known Oseen-Franck potential energy density $W$ given by

270: \beq

271: W\left({\mathbf n},\nabla{\mathbf n}\right) =

272: \alpha\left|{\mathbf n}\times(\nabla\times{\mathbf n})\right|^2

273: + \beta(\nabla\cdot{\mathbf n})^2 +

274: \gamma\left({\mathbf n}\cdot\nabla\times{\mathbf n}\right)^2.

275: \label{2b}

276: \eeq

277: The positive constants $\alpha$, $\beta$, and $\gamma$ are elastic

278: constants of the liquid crystal. For the special case 

279: $\alpha = \beta = \gamma$, the potential energy density reduces to

280: $$

281: W\left({\mathbf n},\nabla{\mathbf n}\right) = \alpha\left|

282: \nabla{\mathbf n}\right|^2,

283: $$

284: which is the potential energy density used in harmonic maps into the sphere 

285: ${\mathbb S}^2$.

286: There are many studies on the constrained elliptic system of equations for

287: ${\mathbf n}$ derived through variational principles from the potential 

288: (\ref{2b}), and on the parabolic flow associated

289: with it, see \cite{[3],[7],[10],[19],[26],[49]} and references therein.

290: In the regime in which inertia effects dominate

291: viscosity, however, the propagation of the orientation waves in the

292: director field may then be modeled by the least action principle (Saxton 

293: \cite{[37]})

294: \beq

295: \frac{\delta}{\delta u} \int \left\{ \p_t{\mathbf n}\cdot\p_t{\mathbf n} 

296: - W\left({\mathbf n},

297: \nabla{\mathbf n}\right)\right\}\,d{\mathbf x}dt  = 0,

298: \qquad {\mathbf n}\cdot {\mathbf n} = 1.

299: \label{2a}

300: \eeq

301: In the special case $\alpha = \beta = \gamma$, this variational principle 

302: (\ref{2a}) yields the equation for harmonic wave maps from $(1+3)$-dimensional

303:  Minkowski space into the two sphere, see \cite{[6],[40],[41]} for example.

304: For planar deformations depending on a single space variable $x$,

305: the director field has the special form

306: $$

307: {\mathbf n} =  \cos u(x,t){\mathbf e}_x + \sin u(x,t){\mathbf e}_y,

308: $$

309: where the dependent variable $u\in\R^1$ measures the angle

310: of the director field to the $x$-direction, and ${\mathbf e}_x$ and

311: ${\mathbf e}_y$ are the coordinate vectors in the $x$ and $y$ directions,

312: respectively.

313: In this case, the variational principle (\ref{2a}) reduces to (\ref{1.1})

314: with the wave speed $c$ given specifically by

315: \beq

316: c^2(u) = \alpha\cos^2u + \beta\sin^2u.

317: \label{2c}

318: \eeq

319: 

320: The equation (\ref{1.1}) has interesting connections with 

321: long waves on a dipole chain in the continuum limit (\cite{[16]},  

322: Zorski and Infeld \cite{[Zo]}, and Grundland and Infeld \cite{[17]}),

323: and in classical field theories and general relativity (\cite{[16]}).

324: We refer the interested reader to the article \cite{[16]} for these

325: connections.

326: 

327: This equation (\ref{1.1}) compares interestingly with other well-known

328: equations, e. g.

329: \beq

330: \p^2_tu -\p_x[p(\p_xu)] =0, \label{conlaw}

331: \eeq

332: where $p(\cdot)$ is a given function, considered by Lax \cite{[29]},

333: Klainerman and Majda \cite{[28]}, and Liu \cite{[34]}. Second related equation

334: is 

335: \beq

336: \p_t^2u- c^2(u) \Delta u = 0\label{lin}

337: \eeq

338: considered by  Lindblad \cite{[31]}, who established the global existence of 

339: smooth solutions of (\ref{lin}) with smooth, small, and spherically 

340: symmetric initial data in $\R^3$,

341: where the large-time decay of solutions in high space dimensions is crucial.

342: The multi-dimensional generalization of equation (\ref{1.1}),

343: \beq

344: \p_t^2u- c(u) \nabla\cdot \left(c(u) \nabla u\right) = 0,\label{multid}

345: \eeq

346: contains a lower order term proportional to $cc'|\nabla u|^2$,

347: which (\ref{lin}) lacks. This lower order term is responsible for

348: the blow-up in the derivatives of $u$. Finally, we note that equation 

349: (\ref{1.1}) also looks related to the perturbed wave equation

350: \beq

351: \p_t^2u - \Delta u+ f(u,\nabla u, \nabla \nabla u) = 0, \label{1.7cpde}

352: \eeq

353: where $f(u, \nabla u,\nabla\nabla u)$ satisfies an appropriate convexity

354: condition (for example, $f =u^p$ or $f=a (\p_tu)^2 + b|\nabla u|^2$) or

355: some nullity condition. Blow-up for (\ref{1.7cpde}) with a convexity condition

356: has been studied extensively, see 

357: \cite{[2],[14],[18],[23],[25],[30],[38],[43],[44]} and Strauss \cite{[45]}

358: for more reference. Global existence and uniqueness of solutions to 

359: (\ref{1.7cpde})

360: with a nullity condition depend on the nullity structure and large time decay

361: of solutions of the linear wave equation in higher dimensions

362: (see Klainerman and Machedon \cite{[27]} and references therein).

363: Therefore (\ref{1.1}) with the dependence of $c(u)$ on $u$ and the possibility

364: of sign changes in $c'(u)$ is familiar yet truly different.

365: 

366: 

367: Equation (\ref{1.1}) has interesting asymptotic uni-directional wave equations.

368: Hunter and Saxton (\cite{[21]}) derived the asymptotic equations

369: \beq

370: (u_t +u^nu_x)_x = \f12 nu^{n-1}(u_x)^2

371: \label{3cpde}

372: \eeq

373: for (\ref{1.1}) via weakly nonlinear geometric optics.

374: We mention that the $x$-derivative of equation

375: (\ref{3cpde}) appears in the high-frequency limit of the variational principle

376: for the Camassa-Holm equation (\cite{[1],[4],[5]}), which arises in the theory

377: of shallow water waves. A construction of global solutions to the

378: Camassa-Holm equations, based on a similar variable transformation 

379: as in the present paper, will appear in \cite{BrCo} 

380: 

381: 

382: \subsection{Known results}

383: In \cite{[22]}, Hunter and Zheng established the global existence of weak

384: solutions to (\ref{3cpde}) ($n=1$) with initial data of bounded variations.

385: It has also been shown that the dissipative solutions are limits of

386: vanishing viscosity. Equation  (\ref{3cpde}) ($n=1$) is also shown

387: to be completely integrable (\cite{[HZ]}). 

388: In \cite{[52]}--\cite{[59]}, Ping Zhang and Zheng

389: study the global existence, uniqueness, and regularity of the weak

390: solutions to (\ref{3cpde}) ($n=1, 2$) with $L^2$ initial data, and

391: special cases of (\ref{1.1}). The study of

392: the asymptotic equation has been very beneficial for both the blow-up

393: result \cite{[GHZ]} and the current global existence result for the wave 

394: equation (\ref{1.1}).

395: 

396: 

397: \section{Variable Transformations} %section 2

398: 

399: \setcounter{equation}{0}

400: 

401: We start by deriving some identities valid for smooth solutions.

402: Consider the variables

403: \beq

404: \left\{ 

405: \begin{array}{rcl}

406: R & \doteq  &u_t+c(u)u_x\,, \\

407: S & \doteq  &u_t-c(u)u_x\,,

408: \end{array} \right.\label{2.1}

409: \eeq

410: so that

411: \beq

412: u_t={R+S\over 2}\,,\qquad\qquad u_x={R-S\over 2c}\,.\label{2.2}

413: \eeq

414: By (\ref{1.1}), the variables $R,S$ satisfy

415: \beq

416: \left\{

417: \begin{array}{rcl}

418: R_t-cR_x &= & {c'\over 4c}(R^2-S^2), \\ [3mm]

419: S_t+cS_x &= & {c'\over 4c}(S^2-R^2).

420: \end{array} \right.

421: \label{2.3}

422: \eeq

423: Multiplying the first equation in (\ref{2.3}) by $R$ and the

424: second one by $S$, we obtain balance laws for $R^2$ and $S^2$, namely

425: \beq\left\{

426: \begin{array}{rcl}

427: (R^2)_t - (cR^2)_x & = & {c'\over 2c}(R^2S - RS^2)\, , \\ [3mm]

428: (S^2)_t + (cS^2)_x & = & - {c'\over 2c}(R^2S -RS^2)\,.

429: \end{array} 

430: \right.\label{2.4}

431: \eeq

432: As a consequence, the following quantities are conserved:

433: \beq

434: E\doteq {1\over 2}\big(u_t^2+c^2u_x^2\big)={R^2+S^2\over 4}\,,\qquad\qquad

435: M\doteq -u_tu_x={S^2-R^2\over 4c}\,.\label{2.5} 

436: \eeq

437: Indeed we have

438: \beq

439: \left\{ 

440: \begin{array}{rcl}

441: E_t+(c^2M)_x & = & 0\,, \\ 

442: M_t+E_x      & = & 0\,.

443: \end{array}\right.\label{2.6} %{2.7'}

444: \eeq

445: One can think of $R^2/4$ as the energy density of backward

446: moving waves, and $S^2/4$ as the energy density of forward moving waves.

447: 

448: We observe that, if $R,S$ satisfy (\ref{2.3}) and $u$ satisfies

449: (\ref{2.1}b), then the quantity

450: \beq

451: F \doteq  R - S -2cu_x \label{F}

452: \eeq

453: provides solutions to the linear homogeneous equation

454: \beq

455: F_t - c\,F_x = {c'\over 2c}(R+S+2cu_x)F. \label{FE}

456: \eeq

457: In particular, if $F\equiv 0$ at time $t=0$, the same

458: holds for all $t>0$.  Similarly, if $R,S$ satisfy (\ref{2.3}) and

459: $u$ satisfies (\ref{2.1}a), then the quantity

460: $$

461: G\doteq R+S-2u_t  

462: $$

463: provides solutions to the linear homogeneous equation

464: $$

465: G_t+c\,G_x={c'\over 2c}(R+S-2cu_x)G\,. 

466: $$

467: In particular, if $G\equiv 0$ at time $t=0$, the same

468: holds for all $t>0$.  We thus have

469: \v

470: \no{\bf Proposition 1.}

471: {\it Any smooth solution of (\ref{1.1}) provides a solution to

472: (\ref{2.1})--(\ref{2.3}). Conversely, any smooth solution of (\ref{2.1}b)

473:  and (\ref{2.3}) (or (\ref{2.1}a) and (\ref{2.3})) which satisfies

474: (\ref{2.2}b)  (or (\ref{2.2}a) ) at time $t=0$ provides

475: a solution to (\ref{1.1}). }

476: \vs

477: 

478: The main difficulty in the analysis of (\ref{1.1}) is the possible

479: breakdown of regularity of solutions.   Indeed, even for smooth

480: initial data, the quantities $u_x, u_t$ can blow up in finite time.

481: This is clear from the equations (\ref{2.3}), where the right

482: hand side grows quadratically, see (\cite{[GHZ]}) for handling change of 

483: signs of $c'$ and interaction between $R$ and $S$.

484: To deal with possibly unbounded values of $R,S$,

485: it is convenient to introduce a new set of dependent variables:

486: $$

487: w\doteq 2\arctan R\,,\qquad\qquad z\doteq 2\arctan S\,,

488: $$

489: so that

490: \beq

491: R=\tan{w\over 2}\,,\qquad S=\tan{z\over 2}\,.\label{2.9}

492: \eeq

493: Using (\ref{2.3}), we obtain the equations

494: \beq

495: w_t-c\,w_x={2\over 1+R^2}(R_t-c\,R_x)

496: ={c'\over 2c}{R^2-S^2\over 1+R^2}\,, \label{2.10}

497: \eeq

498: 

499: \beq

500: z_t+c\,z_x={2\over 1+S^2}(S_t+c\,S_x)

501: ={c'\over 2c}{S^2-R^2\over 1+S^2}\,. \label{2.11}

502: \eeq

503: 

504: \begin{figure}[htb]

505: \centering

506: \includegraphics[width=0.55\textwidth]{b77.eps} %Yuxi's

507: %\includegraphics[width=0.55\textwidth]{FIG/b77.eps} %AB's

508: \caption{\footnotesize Characteristic curves}

509: \label{fig1}

510: \end{figure}

511: 

512: To reduce the equation to a semilinear one, it is convenient to

513: perform a further change of independent variables (fig.~1).

514: Consider

515: the equations for the forward and backward characteristics:

516: \beq

517: \dot x^+=c(u)\,,\qquad\qquad \dot x^-=-c(u)\,.\label{2.12} %{1.5}

518: \eeq

519: The characteristics passing through the point $(t,x)$

520: will be denoted by

521: $$

522: s\mapsto x^+(s,t,x)\,,\qquad\qquad s\mapsto x^-(s,t,x)\,,

523: $$

524: respectively.

525: As coordinates $(X,Y)$ of a point $(t,x)$ we shall use the quantities

526: \beq

527: X\doteq \int_0^{x^-(0,t,x)} \big(1+R^2(0,x)\big)\,dx

528: \,,\qquad\qquad Y\doteq \int_{x^+(0,t,x)}^0

529: \big(1+S^2(0,x)\big)\,dx\,.\label{XY}

530: \eeq

531: Of course this implies

532: \beq

533: X_t- c(u)X_x =0\,,\qquad\qquad Y_t+ c(u)Y_x =0\,,\label{2.14} %eqnlbl{1.10}

534: \eeq

535: \beq

536: (X_x)_t- (c\,X_x)_x=0\,,\qquad\qquad (Y_x)_t+(c\,Y_x)_x=0\,.\label{2.15}%{1.11}

537: \eeq

538: Notice that

539: $$

540: X_x(t,x)= \lim_{h\to 0} {1\over h} \int_{x^-(0, t, x)}

541: ^{x^-(0, t, x+h)}\big(1+ R^2(0, x)\big)\,dx\,,

542: $$

543: $$

544: Y_x(t,x)= \lim_{h\to 0} {1\over h} \int_{x^+(0, t, x)}

545: ^{x^+(0, t, x+h)} \big(1+S^2(0, x)\big)\,dx\,.

546: $$

547: For any smooth function $f$, using (\ref{2.14}) one finds

548: \beq

549: \begin{array}{ccccccr}

550: f_t+cf_x &=& f_XX_t+f_Y Y_t+cf_X X_x+cf_Y Y_x & = & (X_t+cX_x)f_X

551: & = & 2cX_x f_X\,,\\

552: f_t-cf_x &=& f_XX_t+f_Y Y_t-cf_X X_x-cf_Y Y_x & = & (Y_t-cY_x)f_Y

553: & = & -2cY_x f_Y\,.

554: \end{array} \label{2.16} %{1.12}

555: \eeq

556: We now introduce the further variables

557: \beq

558: p\doteq {1+R^2\over X_x}\,,\qquad\qquad q\doteq {1+S^2\over -Y_x}\,.

559: \label{2.17}

560: \eeq

561: Notice that the above definitions imply

562: \beq

563: (X_x)^{-1}= {p\over 1+R^2}=p\,\cos^2{w\over 2}\,,\qquad\qquad

564: (-Y_x)^{-1}={q\over 1+S^2}=q\,\cos^2{z\over 2}\,.\label{2.18}%{1.13}

565: \eeq

566: {}From (\ref{2.10})-(\ref{2.11}), using (\ref{2.16})-(\ref{2.18}), 

567: we obtain

568: $$

569: 2c {(1+S^2)\over q}w_Y={c'\over 2c}

570: {R^2-S^2\over 1+R^2}\,,\qquad\qquad

571: 2c {(1+R^2)\over p}z_X={c'\over 2c}

572: {S^2-R^2\over 1+S^2}\,.

573: $$

574: Therefore

575: \beq

576: \left\{

577: \begin{array}{ccccc}

578: w_Y&=&{c'\over 4c^2}\,{R^2-S^2\over 1+R^2}\,{q\over 1+S^2}

579: &=&{c'\over 4c^2}\,\left( \sin^2{w\over 2}\cos^2{z\over 2}-

580: \sin^2{z\over 2}\cos^2{w\over 2}\right)\,q\,,\\[4mm]

581: z_X&=&{c'\over 4c^2}\,{S^2-R^2\over 1+S^2}\,{p\over 1+R^2}

582: &=&{c'\over 4c^2}\left( \sin^2{z\over 2}\cos^2{w\over 2}-

583: \sin^2{w\over 2}\cos^2{z\over 2}\right)\,p\,.

584: \end{array}\right.\label{2.19} %{2.19'}

585: \eeq

586: Using trigonometric formulas, the above

587: expressions can be further simplified as

588: $$

589: \left\{

590: \begin{array}{ccc}

591: w_Y &= &{c'\over 8c^2}\,( \cos z - \cos w) \,q\,,\\[4mm]

592: z_X &= &{c'\over 8c^2}\,( \cos w - \cos z) \,p\,.

593: \end{array}\right.

594: $$

595: 

596: 

597: Concerning the quantities $p,q$, we observe that

598: \beq

599: c_x~=~c'\, u_x~=~ c'\,{R-S\over 2c}\,. \label{2.20} %{1.7a}

600: \eeq

601: Using again (\ref{2.18}) and (\ref{2.15}) we compute

602: $$

603: \begin{array}{ccl}

604: p_t-c\,p_x &= & (X_x)^{-1}\,2R(R_t-cR_x)-(X_x)^{-2}

605: \big[(X_x)_t-c(X_x)_x\big](1+R^2)\\[4mm]

606: &=& (X_x)^{-1}\,2R\,{c'\over 4c}(R^2-S^2)-(X_x)^{-2}[c_xX_x](1+R^2)\\[4mm]

607: &=& {p\over 1+R^2}2R(R^2-S^2){c'\over 4c}-

608: {p\over 1+R^2} {c'\over 2c} (R-S)(1+R^2)\\[4mm]

609: &=& {c'\over 2c}\,{p\over 1+R^2} \big[ S(1+R^2)-R(1+S^2)\big]\,.

610: \end{array}

611: $$

612: $$

613: \begin{array}{ccl}

614: q_t+c\,q_x &= & (-Y_x)^{-1}\,2S(S_t-cS_x)-(-Y_x)^{-2}

615: \big[(-Y_x)_t+c(-Y_x)_x\big](1+S^2)\\[4mm]

616: &=& (-Y_x)^{-1}\,2S\,{c'\over 4c}(S^2-R^2)-(-Y_x)^{-2}[c_x(Y_x)](1+S^2)\\[4mm]

617: &=& {q\over 1+S^2}2S(S^2-R^2){c'\over 4c}-

618: {q\over 1+S^2} {c'\over 2c} (S-R)(1+S^2)\\[4mm]

619: &=& {c'\over 2c}\,{q\over 1+S^2} \big[ R(1+S^2)-S(1+R^2)\big]\,.

620: \end{array}

621: $$

622: In turn, this yields

623: \beq

624: \begin{array}{ccl}

625: p_Y &=& (p_t-c\,p_x){1\over 2c\,(-Y_x)}=

626: (p_t-c\,p_x){1\over 2c}\,{q\over 1+S^2}\\[4mm]

627: &=& {c'\over 4c^2}{S(1+R^2)-R(1+S^2)\over (1+R^2)(1+S^2)}\,pq

628: ={c'\over 4c^2}\,\left[ {S\over 1+S^2}-{R\over 1+R^2}\right]\,pq\\[4mm]

629: &=& {c'\over 4c^2}\,\left[ \tan{z\over 2}\,\cos^2{z\over 2}-

630: \tan{w\over 2}\,\cos^2{w\over 2}\right]\,pq={c'\over 4c^2}\,

631: {\sin z-\sin w\over 2}\,pq\,,

632: \end{array}   \label{2.21} %{2.21'}

633: \eeq

634: \beq

635: \begin{array}{ccl}

636: q_X &= & (q_t+c\,q_x){1\over 2c\,(X_x)}=

637: (q_t+c\,q_x){1\over 2c}\,{p\over 1+R^2}\\[4mm]

638: &=&{c'\over 4c^2}{R(1+S^2)-S(1+R^2)\over (1+S^2)(1+R^2)}\,pq

639: ={c'\over 4c^2}\,\left[ {R\over 1+R^2}-{S\over 1+S^2}\right]\,pq\\[4mm]

640: &=&{c'\over 4c^2}\,\left[ \tan{w\over 2}\,\cos^2{w\over 2}-

641: \tan{z\over 2}\,\cos^2{z\over 2}\right]\,pq={c'\over 4c^2}\,

642: {\sin w-\sin z\over 2}\,pq\,.

643: \end{array} \label{2.22}%{2.22'}

644: \eeq

645: Finally, by (\ref{2.16}) we have

646: \beq

647: \begin{array}{ccccccc}

648: u_X&=&(u_t+cu_x)\,{1\over 2c} {p\over 1+R^2}&=& {1\over 2c}\, {R\over 1+R^2}\,p

649: &=& {1\over 2c}\, \left(\tan{w\over 2}\,\cos^2{w\over 2}\right)\,p\,, \\[4mm]

650: u_Y& =&(u_t-cu_x)\,{1\over 2c} {q\over 1+S^2}&=&{1\over 2c}\, {S\over 1+S^2}\,q

651: &=& {1\over 2c}\, \left(\tan{z\over 2}\,\cos^2{z\over 2}\right)\,q\, \,.

652: \end{array} \label{2.23} %\eqnlbl{2.23'}$$

653: \eeq

654: 

655: 

656: \vs

657: Starting with the nonlinear equation (\ref{1.1}),

658: using $X,Y$ as independent variables we thus obtain a semilinear

659: hyperbolic system with smooth coefficients for the variables $u,w,z,p,q$.

660: Using some trigonometric identities, the set of equations (\ref{2.19}),

661: (\ref{2.21})-(\ref{2.22}) and (\ref{2.23}) can be rewritten as

662: \beq

663: \left\{

664: \begin{array}{ccc}

665: w_Y&=&{c'\over 8c^2}\,( \cos z - \cos w)\,q\,,\\[4mm]

666: z_X&=&{c'\over 8c^2}\,( \cos w - \cos z)\,p\,,

667: \end{array}\right. \label{2.24} %{1.16}

668: \eeq

669: \beq

670: \left\{

671: \begin{array}{ccc}

672: p_Y &= & {c'\over 8c^2}\, \big[\sin z-\sin w\big]\,pq\,,\\[4mm]

673: q_X &= & {c'\over 8c^2}\, \big[\sin w-\sin z\big]\,pq\,,

674: \end{array}\right.\label{2.25} %{1.17}

675: \eeq

676: \beq

677: \left\{ 

678: \begin{array}{ccc}

679: u_X &= & {\sin w\over 4c} \, p\,,\\[4mm]

680:  u_Y &= & {\sin z\over 4c} \, q\,.

681: \end{array}\right. \label{2.26} %{1.18}

682: \eeq

683: \v

684: \no{\bf Remark 1.}  The function $u$ can be determined by

685: using either one of the equations in (\ref{2.26}). One

686: can easily check that the

687: two equations are compatible, namely

688: \beq

689: \begin{array}{rcl}

690: u_{XY}

691: &= &-{\sin w\over 4c^2} \,c' u_Y\,p+{\cos w\over 4c}\, w_Y\,p+{\sin w\over

692: 4c}\,p_Y\\[4mm]

693: &=&  {c'\over 32\,c^3}\,\big[ -2\sin w\sin z+

694: \cos w\cos z -\cos^2 w-\sin^2 w + \sin w\sin z\big]\,pq\\[4mm]

695: &=&{c'\over 32\,c^3}\,\big[ \cos(w-z)-1\big]\,pq \\ [4mm] 

696: &= & u_{Y\!X} \,.

697: \end{array} \label{compat}

698: \eeq

699: \v

700: \no{\bf Remark 2.} We observe that the new system is

701: invariant under translation by $2\pi$

702: in $w$ and $z$.  Actually, it would be more precise to work with

703: the variables $w^\dagger\doteq e^{iw}$ and $z^\dagger\doteq e^{iz}$.

704: However, for simplicity we shall use the variables $w,z$,

705: keeping in mind that they range on the unit circle

706: $[-\pi,\pi]$ with endpoints identified.

707: \v

708: The system (\ref{2.24})-(\ref{2.26}) must now be supplemented

709: by non-characteristic boundary conditions,

710: corresponding to (\ref{1.2}). For this purpose, we observe

711: that $u_0,u_1$ determine the initial values of the functions

712: $R,S$ at time $t=0$.

713: The line $t=0$ corresponds

714: to a curve $\gamma$ in the $(X, Y)$ plane, say

715: $$

716: Y = \vp (X), \qquad X\in\R,

717: $$

718: where

719: $Y\doteq \vp(X)$ if and only if

720: $$

721: X=\int_0^x \big(1+R^2(0,x)\big)\,dx\,,\quad

722: Y=- \int_0^x \big(1+S^2(0,x)\big)\,dx\qquad\hbox{for some}~~x\in\R\,.

723: $$

724: We can use the variable $x$ as a parameter along the curve $\gamma$.

725: The assumptions $u_0\in H^1$, $u_1\in \L^2$ imply

726: $R,S\in \L^2$; to fix the ideas, let

727: \beq

728: \E_0\doteq  {1\over 4}\int \big[R^2(0,x)+S^2(0,x)\big]\,dx <\infty

729: \,.\label{E.0}

730: \eeq

731: The two functions

732: $$

733: X(x)\doteq \int_0^x \big(1+R^2(0,x)\big)\,dx \,,\qquad\qquad

734: Y(x)\doteq \int_x^0 \big(1+S^2(0,x)\big)\,dx

735: $$

736: are well defined and absolutely continuous.  Clearly,

737: $X$ is strictly increasing while $Y$

738: is strictly decreasing. Therefore, the map $X\mapsto \vp(X)$

739: is continuous and strictly decreasing.  From (\ref{E.0}) it

740: follows

741: \beq

742: \big|X+\vp(X)\big|\leq 4 \E_0\,.\label{xp}

743: \eeq

744: As $(t,x)$ ranges over the domain $[0,\infty[\,\times\R$, the

745: corresponding variables $(X, Y)$ range over the set

746: \beq

747: \Omega^+\doteq \big\{ (X,Y)\,; ~~Y\geq \vp(X)\big\}\,.\label{2.27}

748: \eeq

749: Along the curve

750: $$

751: \gamma\doteq\big\{ (X,Y)\,;~~Y=\vp(X)\big\}\subset\R^2

752: $$

753: parametrized by $x\mapsto \big(X(x), \,Y(x)\big)$,

754: we can thus assign the boundary data

755: $(\bar w, \bar z, \bar p, \bar q, \bar u)$$\in \L^\infty$

756: defined by

757: \beq

758: \left\{

759: \begin{array}{rcl}

760: {\bar w } &= & 2\arctan R(0,x)\,,\\ {\bar z} &= & 2\arctan S(0,x)\,,

761: \end{array}\right.\qquad \qquad

762: \left\{

763: \begin{array}{rcl}

764: {\bar p} &\equiv & 1\,,\\

765: {\bar q} &\equiv & 1\,,

766: \end{array}\right. \qquad\qquad  {\bar u } =u_0(x)\,. \label{2.28}

767: \eeq

768: We observe that the identity

769: \beq

770: F = \tan {{\bar w}\over 2}-\tan {{\bar z}\over 2} -2c({\bar u}){\bar u}_x = 0

771: \label{2.29}

772: \eeq

773: is identically satisfied

774: along $\gamma$. A similar identity holds for $G$.

775: 

776: 

777: 

778: \section{Construction of integral solutions} %3

779: 

780: \setcounter{equation}{0}

781: 

782: Aim of this section is to prove a global existence theorem for the

783: system (\ref{2.24})-(\ref{2.26}), describing the nonlinear wave equation

784: in our transformed variables.

785: \v

786: \no{\bf Theorem 4.} {\it Let the assumptions in Theorem 1 hold.

787: Then the corresponding

788: problem (\ref{2.24})-(\ref{2.26}) with boundary data (\ref{2.28})

789: has a unique solution, defined for all $(X,Y)\in\R^2$.}

790: \v

791: In the following, we shall construct the solution on the domain $\Omega^+$

792: where $Y\geq \vp(X)$. On the complementary set

793: $\Omega^-$ where $Y<\vp(X)$, the solution

794: can be constructed in an entirely similar way.

795: 

796: Observing that all equations (\ref{2.24})-(\ref{2.26})

797: have a locally Lipschitz continuous right hand side,

798: the construction of a local solution as fixed

799: point of a suitable integral transformation is straightforward.

800: To make sure that this solution is actually defined on the whole

801: domain $\Omega^+$, one must establish a priori bounds,

802: showing that $p,q$ remain bounded on bounded sets.

803: This is not immediately obvious from the equations (\ref{2.25}),

804: because the right hand sides have quadratic growth.

805: 

806: The basic estimate can be derived as follows. Assume

807: \beq

808: C_0\doteq \sup_{u\in \R} \left|{c'(u)\over 4c^2(u)}\right| <\infty\,.

809: \label{3.1}

810: \eeq

811: {}From (\ref{2.25}) it follows the identity

812: $$

813: q_X + p_Y = 0\,.

814: $$

815: In turn, this implies that the differential form

816: $p\,dX-q\,dY$ has zero integral

817: along every closed curve contained in $\Omega^+$.

818: In particular, for every $(X,Y)\in\Omega^+$,

819: consider the closed curve $\Sigma$ (see fig.~2) consisting of:

820: 

821: \begin{description}

822: \item[-] the vertical segment joining $\big( X,\,\vp(X)\big)$ with $(X,Y)$,

823: 

824: \item[-] the horizontal segment joining $(X,Y)$ with 

825: $\big( \vp^{-1}(Y),\,Y\big)$

826: 

827: \item[-] the portion of boundary $\gamma=\big\{Y=\vp(X)\big\}$ joining

828: $\big(\vp^{-1}(Y),\,Y\big)$ with $\big(X,\,\vp(X)\big)$.

829: \end{description}

830: 

831: 

832: \begin{figure}[htb]

833: \centering

834: \includegraphics[width=0.55\textwidth]{b78.eps} %Yuxi's

835: %\includegraphics[width=0.55\textwidth]{FIG/b78.eps} %AB's

836: \caption{\footnotesize The closed curve $\Sigma$ }

837: \label{fig2}

838: \end{figure}

839: 

840: \no Integrating along $\Sigma$, recalling that $p=q=1$ along

841: $\gamma$ and then using (\ref{xp}), we obtain

842: \beq

843: \begin{array}{rcl}

844: \int_{\vp^{-1}(Y)}^{X}p(X',Y)\,dX' + \int_{\vp(X)}^{Y}q(X, Y')\,dY'

845: &=& X - \vp^{-1}(Y) + Y -\vp(X) \\[4mm]

846: &\leq & 2(|X|+|Y|+4\E_0)\,.

847: \end{array} \label{3.2}

848: \eeq

849: Using (\ref{3.1})-(\ref{3.2}) in (\ref{2.25}), since $p,q>0$ we

850: obtain the a priori bounds

851: \beq

852: \begin{array}{rcl}

853: p(X,Y) &= &\exp\left\{

854: \int_{\vp(X)}^Y{c'(u)\over 8c^2(u)}\,

855: \big[\sin z-\sin w\big]\,q(X,Y')\,dY'\right\}\\[4mm]

856: &\leq&  \exp\left\{ C_0\,

857: \int_{\vp(X)}^Y\,q(X,Y')\,dY'\right\}\\[4mm]

858: & \leq & \exp\big\{ 2C_0(|X|+|Y|+ 4\E_0)\big\}\,.

859: \end{array} \label{3.3}

860: \eeq

861: Similarly,

862: \beq

863: q(X,Y)\leq \exp \big\{ 2C_0( |X|+|Y|+4\E_0)\big\} \,.\label{3.4}

864: \eeq

865: 

866: Relying on (\ref{3.3})-(\ref{3.4}), we now show that, on bounded sets

867: in the $X$-$Y$ plane, the

868: solution of (\ref{2.24})--(\ref{2.26}) with boundary conditions

869: (\ref{2.28}) can be obtained as the

870: fixed point of a contractive transformation.

871: For any given $r>0$, consider the bounded domain

872: $$

873: \Omega_r\doteq \big\{ (X,Y)\,; ~~Y\geq \vp(X)\,,\quad X\leq r\,,

874: \quad Y\leq r\big\}\,.

875: $$

876: Introduce the space of functions

877: $$

878: \Lambda_r\doteq\Big\{ f:\Omega_r\mapsto\R\,;~~\|f\|_*\doteq {\rm ess}\!\!

879: \sup_{(X,Y)\in\Omega_r}~

880: e^{-\kappa(X+Y)} \big|f(X,Y)\big|~<~\infty\Big\}

881: $$

882: where $\kappa$ is a suitably large constant, to be determined later.

883: For $w,z,p,q,u\in\Lambda_r$, consider the transformation

884: $\T(w,z,p,q,u)=(\tilde w,\tilde z,\tilde p,\tilde q,\tilde u)$

885: defined by

886: \beq

887: \left\{

888: \begin{array}{rcl}

889: \tilde w(X,Y)&=&\bar w(X,\vp(X))+\int_{\vp(X)}^Y

890: {c'(u)\over 8c^2(u)}\,( \cos z - \cos w )\,q\,dY,\\[4mm]

891: \tilde z(X,Y)&=&\bar z(\vp^{-1}(Y),Y)+  \int_{\vp^{-1}(Y)}^X

892: {c'(u)\over 8c^2(u)}\,( \cos w - \cos z)\,p\,dX\,,

893: \end{array}\right. \label{3.5}

894: \eeq

895: \beq

896: \left\{

897: \begin{array}{rcl}

898: \tilde p(X,Y) &=& 1+\int_{\vp(X)}^Y{c'(u)\over 8c^2(u)}\,

899: \big[\sin z-\sin w\big]\,\hat p\hat q\,dY\},\\[4mm]

900: \tilde q(X,Y)&=&1+\int_{\vp^{-1}(Y)}^X {c'(u)\over 8c^2(u)}\,

901: \big[\sin w-\sin z\big]\,\hat p\hat q\,dX\}\,,

902: \end{array}\right.\label{3.6}

903: \eeq

904: \beq

905: \tilde u(X,Y)=\bar u(X,\vp(X))+\int_{\vp(X)}^Y

906: {\sin z\over 4c} \, q\,dY.\label{3.7}

907: \eeq

908: In (\ref{3.6}), the quantities $\hat p,\hat q$ are defined as

909: \beq

910: \hat p\doteq \min\big\{ p\,,~2 e^{2C_0(|X|+|Y|+4\E_0)}\big\}\,,\qquad\qquad

911: \hat q\doteq \min\big\{ q\,,~2 e^{2C_0(|X|+|Y|+4\E_0)}\big\}\,.\label{3.8}

912: \eeq

913: Notice that $\hat p=p$, $\hat q=q$ as long as the a priori estimates

914: (\ref{3.3})-(\ref{3.4}) are satisfied.  Moreover, if in

915: the equations (\ref{2.24})-- (\ref{2.26}) the variables $p,q$ are replaced

916: with $\hat p,\hat q$, then the right hand sides

917: become uniformly Lipschitz continuous on

918: bounded sets in the $X$-$Y$ plane.

919: A straightforward computation now shows that the map $\T$

920: is a strict contraction on the space $\Lambda_r$, provided that

921: the constant $\kappa$ is chosen sufficiently big

922: (depending on the function $c$ and on $r$).

923: 

924: Obviously, if $r'>r$, then the solution of (\ref{3.5})--(\ref{3.7})

925: on $\Omega_{r'}$ also provides the solution

926: to the same equations on $\Omega_r$, when restricted to this smaller domain.

927: Letting $r\to\infty$, in the limit we thus obtain a unique

928: solution $(w,z,p,q,u)$ of (\ref{3.5})--(\ref{3.7}),

929: defined on the whole domain $\Omega^+$.

930: 

931: To prove that these functions satisfy the  (\ref{2.24})--(\ref{2.26}),

932: we claim that $\hat p=p$, $\hat q=q$ at every point $(X,Y)\in\Omega^+$.

933: The proof is by contradiction.

934: If our claim does not hold, since the maps $Y\mapsto p(X,Y)$,

935: $X\mapsto q(X,Y)$

936: are continuous, we can find some point

937: $(X^*, Y^*)\in\Omega^+$ such that

938: \beq

939: p(X,Y)\leq

940: 2 e^{2C_0(|X|+|Y|+4\E_0)}\,,\qquad \qquad q(X,Y)\leq 2 e^{2C_0(|X|+|Y|+4\E_0)}

941: \label{3.9}

942: \eeq

943: for all $(X,Y)\in\Omega^*\doteq \Omega^+\cap \{X\leq X^*\,,~~Y\leq Y^*\}$, but

944: either

945: $p(X^*,Y^*)\geq {3\over 2}  e^{2C_0(|X|+|Y|+4\E_0)}$ or

946: $q(X^*,Y^*)\geq {3\over 2} e^{2C_0(|X|+|Y|+4\E_0)}$.

947: By (\ref{3.9}), we still have $\hat p=p$, $\hat q=q$ restricted to $\Omega^*$,

948: hence the equations (\ref{2.24})--(\ref{2.26})  and the a priori bounds

949: (\ref{3.3})-(\ref{3.4}) remain valid.  In particular, these imply

950: $$

951: p(X^*,Y^*)\leq e^{2C_0(|X|+|Y|+4\E_0)}\,,\qquad\qquad q(X^*,Y^*)

952: \leq e^{2C_0(|X|+|Y|+4\E_0)}\,,

953: $$

954: reaching a contradiction.

955: \endproof

956: 

957: \v

958: \no{\bf Remark 3.} In the solution constructed above,

959: the variables $w, z$ may well grow outside the initial range $\,]-\pi, \pi[\,$.

960: This happens precisely when the quantities $R$, $S$

961: become unbounded, i.e.~when

962: singularities arise.

963: \v

964: For future reference, we state a useful

965: consequence of the above construction.

966: \v

967: \no{\bf Corollary 1.} {\it If the initial data $u_0, u_1$ are smooth,

968: then the solution $(u,p,q,w,z)$ of (\ref{2.24})--(\ref{2.26}), (\ref{2.28})

969: is a smooth function of the variables $(X,Y)$.

970: Moreover, assume that a sequence of smooth functions

971: $(u_0^m, u_1^m)_{m\geq 1}$ satisfies

972: $$

973: u_0^m\to u_0\,,\qquad (u_0^m)_x\to (u_0)_x\,,\qquad u_1^m\to u_1

974: $$

975: uniformly on compact subsets of $\R$. Then one

976: has the convergence of the corresponding

977: solutions:

978: $$(u^m,\,p^m,\, q^m,\, w^m,\, z^m)\to (u,p,q,w,z)$$

979: uniformly on bounded subsets of the $X$-$Y$ plane.}

980: \v

981: We also remark that the equations (\ref{2.24})--(\ref{2.26})

982: imply the conservation laws

983: \beq

984: q_X+p_Y=0\,,\qquad\qquad \left({q\over c}\right)_X

985: -\left({p\over c}\right)_Y=0\,.\label{3.10}

986: \eeq

987: 

988: 

989: \section{Weak solutions, in the original variables} %4

990: 

991: \setcounter{equation}{0}

992: 

993: By expressing the solution $u(X,Y)$ in terms of the original variables

994: $(t,x)$, we shall recover a solution of the Cauchy problem

995: (\ref{1.1})-(\ref{1.2}). This will provide a proof of Theorem 1.

996: 

997: As a preliminary, we examine the regularity of the solution

998: $(u, w, z, p, q)$ constructed in the previous section.

999: Since the initial data $(u_0)_x$ and $u_1$ are only assumed to

1000: be in $\L^2$, the functions $w,z,p,q$ may well be discontinuous.

1001: More precisely, on bounded subsets of the $X$-$Y$ plane,

1002: the equations (\ref{2.24})-(\ref{2.26}) imply

1003: the following:

1004: \v

1005: \begin{description}

1006: 

1007: \item[-] The functions $w,p$ are Lipschitz continuous w.r.t.~$Y$,

1008: measurable w.r.t.~$X$.

1009: 

1010: \item[-] The functions $z,q$ are Lipschitz continuous w.r.t.~$X$,

1011: measurable w.r.t.~$Y$.

1012: 

1013: \item[-] The function $u$ is Lipschitz continuous w.r.t.~both $X$ and $Y$.

1014: \end{description}

1015: \v

1016: The map $(X,Y)\mapsto (t,x)$ can be constructed as follows.

1017: Setting $f=x$, then $f=t$  in the two equations at

1018: (\ref{2.16}), we find

1019: $$

1020: \left\{

1021: \begin{array}{rcr}

1022: c &=& 2cX_x\,x_X\,, \\[3mm]

1023: -c&=& -2cY_x\,x_Y\,,

1024: \end{array}\right.\qquad\qquad

1025: \left\{

1026: \begin{array}{rcr}

1027: 1&= &2cX_x\,t_X\,,\\[3mm]

1028: 1&=& -2cY_x\,t_Y\,,

1029: \end{array}\right.

1030: $$

1031: respectively.

1032: Therefore, using (\ref{2.18}) we obtain

1033: \beq

1034: \left\{

1035: \begin{array}{ccccr} 

1036: x_X&~=~&{1\over 2X_x}&~=~&{(1+\cos w)\,p\over 4}\,,\\[4mm]

1037: x_Y&~=~&{1\over 2Y_x}&~=~&-{(1+\cos z)\,q\over 4}\,,

1038: \end{array}\right. \label{4.1}

1039: \eeq

1040: \beq

1041: \left\{

1042: \begin{array}{ccccr} 

1043:  t_X&=&{1\over 2cX_x}&=&{(1+\cos w)\,p\over 4c}\,,\\[4mm]

1044: t_Y&=&{1\over -2cY_x}&=&{(1+\cos z)\,q\over 4c}\,.

1045: \end{array}\right. \label{4.2}

1046: \eeq

1047: For future reference,

1048: we write here the partial derivatives of the inverse

1049: mapping, valid at points where $w,z\not= -\pi$.

1050: \beq

1051: \left\{

1052: \begin{array}{ccr}  

1053: X_x&=&{2\over (1+\cos w)\,p}\,,\\[4mm]

1054: Y_x&=&\,-{2\over(1+\cos z)\,q}\,,

1055: \end{array}\right. 

1056: \qquad\qquad

1057: \left\{

1058: \begin{array}{ccr} 

1059:  X_t&=&{2c\over (1+\cos w)\,p}\,,\\[4mm]

1060: Y_t&=&{2c\over(1+\cos z)\,q}\,.

1061: \end{array}\right. \label{4.3}

1062: \eeq

1063: We can now recover the functions $x=x(X,Y)$ by integrating

1064: one of the  equations in (\ref{4.1}). Moreover, we can

1065: compute $t=t(X,Y)$ by integrating one of the equations in (\ref{4.2}).

1066: A straightforward calculation shows that the two equations

1067: in (\ref{4.1}) are equivalent: differentiating the first w.r.t.~$Y$

1068: or the second w.r.t.~$X$ one obtains the same expression.

1069: $$

1070: \begin{array}{ccl} 

1071:  x_{XY} & = &{(1+\cos w)\,

1072: p_Y\over 4} -{p\sin w\, w_Y\over 4}\\[3mm]

1073:   & = & {c'\,pq\over 32 c^2}\,\big[ \sin z-\sin w+\sin(z-w)\big]

1074: =x_{Y\!X}\,.

1075: \end{array}

1076: $$

1077: Similarly, the equivalence of the two equations in (\ref{4.2})

1078: is checked by

1079: $$

1080: \begin{array}{cl} 

1081: &t_{XY}-t_{Y\!X}~=~

1082: \left({x_X\over c}\right)_Y+\left({x_Y\over c}\right)_X~=~{2\over c}\, x_{XY}

1083: -\left({x_X\over c^2}\,c' u_Y+{x_Y\over c^2}\,c'u_X\right)\\[3mm]

1084: &\quad={c'\,pq\over 16\, c^3}\,\big[ \sin z-\sin w+\sin(z-w)\big]\\[3mm]

1085: &\qquad\qquad

1086:  -{c'\,pq\over 16\, c^3}\big[ (1+\cos w)\sin z-(1+\cos z)\sin w\big]~=~0\,.

1087: \end{array}

1088: $$

1089: \v

1090: In order to define $u$ as a function of the original variables $t,x$,

1091: we should formally invert the map $(X,Y)\mapsto (t,x)$ and write

1092: $u(t,x)=u\big( X(t,x)\,,~Y(t,x)\big)$.

1093: The fact that

1094: the above map may  not be one-to-one does not

1095: cause any real difficulty. Indeed, given $(t^*,x^*)$, we can choose an

1096: arbitrary point $(X^*,Y^*)$ such that $t(X^*,Y^*)=t^*$, $x(X^*,Y^*)=x^*$,

1097: and define $u(t^*,x^*)=u(X^*,Y^*)$.

1098: To prove that the values of $u$ do not depend on the choice

1099: of $(X^*,Y^*)$, we

1100: proceed as follows.

1101: Assume that there are two distinct points such that

1102: $t(X_1, Y_1)=t(X_2,Y_2)=t^*$,

1103: $~x(X_1,Y_1)=x(X_2,Y_2)=x^*$.

1104: We consider two cases:

1105: \v

1106: \no Case 1: $X_1\leq X_2$, $Y_1\leq Y_2$. Consider the set

1107: $$

1108: \Gamma_{x^*}\doteq \Big\{ (X,Y)\,;~~x(X,Y)\leq x^*\Big\}

1109: $$

1110: and call $\partial \Gamma_{x^*}$ its boundary.

1111: By (\ref{4.1}), $x$ is increasing with $X$ and decreasing with $Y$.

1112: Hence, this boundary can be represented as

1113: the graph of a Lipschitz continuous function: $X-Y=\phi(X+Y)$.

1114: We now construct the Lipschitz continuous curve $\gamma$ (fig.~3a) consisting

1115: of

1116: \begin{description}

1117: 

1118: \item[-] a horizontal segment joining $(X_1,Y_1)$ with a point $A=(X_A,Y_A)$

1119: on $\partial \Gamma_{x^*}$, with $Y_A=Y_1$,

1120: 

1121: \item[-] a portion of the boundary $\partial \Gamma_{x^*}$,

1122: 

1123: \item[-] a vertical segment joining $(X_2,Y_2)$ to a point $B=(X_B,Y_B)$

1124: on $\partial \Gamma_{x^*}$, with $X_B=X_2$.

1125: \end{description}

1126: \no We can obtain a Lipschitz continuous parametrization

1127: of the curve $\gamma:[\xi_1,\xi_2]\mapsto \R^2$ in terms of

1128: the parameter $\xi=X+Y$.

1129: Observe that the map $(X,Y)\mapsto (t,x)$ is constant along $\gamma$.

1130: By (\ref{4.1})-(\ref{4.2}) this implies

1131: $(1+\cos w) X_\xi=(1+\cos z)Y_\xi = 0$,

1132: hence $\sin w\cdot X_\xi=\sin z\cdot Y_\xi=0$. We now compute

1133: $$

1134: \begin{array}{rl}

1135: u(X_2,Y_2)-&u(X_1,Y_1)=\int_\gamma \big(u_X\,dX+u_Y\,dY\big) \cr

1136: &=\int_{\xi_1}^{\xi_2} \left({p\,\sin w\over 4c}\,X_\xi-{q\,\sin z\over 4c}\,

1137: Y_\xi\right)\,d\xi=0\,,

1138: \end{array}

1139: $$

1140: proving our claim.

1141: \v

1142: \no Case 2: $X_1\leq X_2$, $Y_1\geq Y_2$.   In this case, we consider the

1143: set

1144: $$

1145: \Gamma_{t^*}\doteq \Big\{ (X,Y)\,;~~t(X,Y)\leq t^*\Big\}\,,

1146: $$

1147: and construct a curve $\gamma$ connecting $(X_1,Y_1)$ with

1148: $(X_2,Y_2)$ as in fig.~3b.  Details are entirely similar to Case 1.

1149: 

1150: 

1151: 

1152: \begin{figure}[htb]

1153: \centering

1154: \includegraphics[width=0.55\textwidth]{b84.eps} %Yuxi's

1155: %\includegraphics[width=0.55\textwidth]{FIG/b84.eps} %AB's

1156: \caption{\footnotesize Paths of integration }

1157: \label{fig3}

1158: \end{figure}

1159: 

1160: 

1161: 

1162: We now prove that the function $u(t,x)=u\big(X(t, x), Y(t, x)\big)$

1163: thus obtained is H\"older continuous on bounded sets.

1164: Toward this goal, consider any characteristic curve, say  $t\mapsto

1165: x^+(t)$, with $\dot x^+=c(u)$.   By construction, this

1166: is parametrized by the function $X\mapsto \big(t(X,\ov Y),\, x(X,\ov Y)\big)$,

1167: for some fixed $\ov Y$. Recalling (\ref{2.16}), (\ref{2.14}), (\ref{2.18}) 

1168: and (\ref{2.26}), we  compute

1169: \beq

1170: \begin{array}{rcl}

1171:  \int_0^\tau \big[ u_t+c(u)u_x\big]^2\,dt &= &\int_{X_0}

1172: ^{X_\tau} ( 2cX_x u_X)^2\, (2X_t)^{-1} dX\\[4mm]

1173: &=& \int_{X_0}

1174: ^{X_\tau}  2c\left( p\,\cos^2{w\over 2}\right)^{-1}\,

1175: \left({p\over 4c}\,2\sin{w\over 2}\cos {w\over 2}\right)^2\,dX \\[4mm]

1176: &\leq& \int_{X_0}

1177: ^{X_\tau}{p\over 2c}\,dX ~\leq ~C_\tau\,,

1178: \end{array}\label{4.4}

1179: \eeq

1180: for some constant $C_\tau$ depending only on $\tau$.

1181: Similarly, integrating along any backward characteristics $t\mapsto x^-(t)$

1182: we obtain

1183: \beq

1184: \int_0^\tau \big[ u_t-c(u)u_x\big]^2\,dt\leq C_\tau. \label{4.5}

1185: \eeq

1186: Since the speed of characteristics is $\pm c(u)$,

1187: and $c(u)$ is uniformly positive and

1188: bounded, the bounds (\ref{4.4})-(\ref{4.5})

1189: imply that the function $u=u(t,x)$ is H\"older

1190: continuous with exponent $1/2$.  In turn, this implies that

1191: all characteristic curves are $\C^1$ with H\"older continuous derivative.

1192: Still from (\ref{4.4})-(\ref{4.5}) it follows that the functions

1193: $R,S$ at (\ref{2.1}) are square integrable on bounded subsets of the

1194: $t$-$x$ plane.

1195: \v

1196: Finally, we prove that the function $u$ provides a weak solution

1197: to the nonlinear wave equation (\ref{1.1}). According to (\ref{1.4'}),

1198: we need to show that

1199: \beq

1200: \begin{array}{rcl}

1201: 0&= &\dint \phi_t\big[(u_t+cu_x)+(u_t-cu_x)\big]

1202: -\big(c(u)\phi\big)_x\big[(u_t+cu_x)-(u_t-cu_x)\big]

1203: \,dxdt\\[4mm]

1204: &=&\dint \big[\phi_t

1205: -(c\phi)_x\big]\,(u_t+cu_x)\,dxdt+\dint \big[\phi_t

1206: +(c\phi)_x\big]\,(u_t-cu_x)\,dxdt\\[4mm]

1207: &=&\dint \big[\phi_t

1208: -(c\phi)_x\big]\,R\,dxdt+\dint \big[\phi_t

1209: +(c\phi)_x\big]\,S\,dxdt\,.

1210: \end{array}\label{4.6}

1211: \eeq

1212: By (\ref{2.16}), this is equivalent to

1213: \beq

1214: \dint \Big\{-2cY_x \phi_Y \,R~+~2cX_x\phi_X \,S~+~

1215: c'(u_X\,X_x+u_Y\,Y_x) \,\phi\,(S-R)\Big\} \,dxdt =0\,.

1216: \label{4.7}

1217: \eeq

1218: It will be convenient to express the double integral in (\ref{4.7})

1219: in terms of the variables $X,Y$. We notice that, by (\ref{2.18}) and 

1220: (\ref{2.14}),

1221: $$

1222: dx\,dt={pq\over 2c\, (1+R^2)(1+S^2)}\,dXdY\,.

1223: $$

1224: Using (\ref{2.26}) and the identities

1225: \beq

1226: \left\{

1227: \begin{array}{ccc}

1228: {1\over 1+R^2}&=&\cos^2{w\over 2}={1+\cos w\over 2}\,, \\[4mm]

1229: {1\over 1+S^2}&=&\cos^2{z\over 2}={1+\cos z\over 2}\,,

1230: \end{array}\right.

1231: \qquad\qquad

1232: \left\{

1233: \begin{array}{ccc}

1234: {R\over 1+R^2} &=&{\sin w\over 2}\,,\\[4mm]

1235: {S\over 1+S^2} &=&{\sin z\over 2}\,,

1236: \end{array}\right.\label{4.8}

1237: \eeq

1238: the double integral in (\ref{4.6}) can thus be written as

1239: \beq

1240: \begin{array}{rl}

1241: \dint &\left\{ 2c\,{1+S^2\over q} \,\phi_Y\,R+

1242: 2c{1+R^2\over p}\,\phi_X\,S

1243: +c'\left( {\sin w\over 4c}\,p\,{1+R^2\over p}-

1244: {\sin z\over 4c}\,q\,{1+S^2\over q}\right)

1245: \,\phi\,(S-R)\right\}\\[4mm]

1246: &\qquad\qquad\cdot {pq\over 2c\,(1+R^2)\,(1+S^2)}\,dXdY\\[4mm]

1247: &=\dint \left\{ {R\over 1+R^2}\,p\,\phi_Y+{S\over 1+S^2}\,q\,\phi_X

1248: +{c'pq\over 8c^2}\left({\sin w\over 1+S^2}-{\sin z\over 1+R^2}

1249: \right)\phi\,(S-R)\right\}\,dXdY\\[4mm]

1250: &=\dint\bigg\{ {p\,\sin w\over 2}\,\phi_Y+{q\sin z\over 2}\,\phi_X\\[4mm]

1251: &\qquad +

1252: {c'pq\over 8c^2}\left( \sin w\sin z-\sin w\,\cos^2 {z\over 2}\tan{w\over 2}

1253: -\sin z\,

1254: \cos^2 {w\over 2}\,

1255: \tan{z\over 2}\right)\,\phi\bigg\}\,dXdY\\[4mm]

1256: &= \dint\left\{ {p\,\sin w\over 2}\,\phi_Y+{q\sin z\over 2}\,\phi_X +

1257: {c'pq\over 8c^2}\,\big[\cos (w+z)-1\big]\,\phi\right\} \,dXdY\,.

1258: \end{array} \label{4.9}

1259: \eeq

1260: Recalling (\ref{2.27}), one finds

1261: \beq

1262: \begin{array}{rcl}

1263: \left(p\,\sin w\over 2\right)_Y+

1264: \left(q\,\sin z\over 2\right)_X

1265: &=&(2c\,u_X)_Y+(2c\,u_Y)_X\\[4mm]

1266: &=&4c' u_Xu_Y+4c\,u_{XY}\\[4mm]

1267: &=&{c'\,pq\over 4c^2}\, \sin w\,\sin z+ {c'\,pq\over 8c^2}

1268: \,\big[\cos(w-z)-1\big]\\[4mm]

1269: &=& {c'\,pq\over 8c^2}\big[\cos(w+z)-1\big]\,.

1270: \end{array}\label{4.10}

1271: \eeq

1272: Together, (\ref{4.9}) and (\ref{4.10}) imply (\ref{4.7}) and hence (\ref{4.6}).

1273: This establishes the integral equation (\ref{1.4'}) for every

1274: test function $\phi\in\C^1_c$.

1275: %\v

1276: %To complete the proof of Theorem 1, we need to check the

1277: %initial conditions (1.2).   Since $u$ is H\"older continuous,

1278: %y construction the initial data $u_0$ is attained pointwise.

1279: %To prove that our function $u$ satisfies also

1280: %the second equality in (1.2), it suffices to check the

1281: %continuity of the maps

1282: %$$t\mapsto u_x(t,\cdot)\,,\qquad\qquad

1283: %t\mapsto u_t(t,\cdot)\,,\leqno(4.11)$$

1284: %as functions with values in $\L^p$, for $p\in [1,\,2[\,$.

1285: %\vfill\eject

1286: 

1287: 

1288: \section{Conserved quantities} %5

1289: 

1290: \setcounter{equation}{0}

1291: 

1292: From the conservation laws (\ref{3.10}) it follows that the 1-forms

1293: $p \,dX-q\,dY$ and $\f{p}{c}\,dX + \f{q}{c}\,dY$ are closed, hence 

1294: their integrals along any closed curve in th $X$-$Y$ plane vanish.

1295: From the conservation laws at (\ref{2.6}), it follows that

1296: the 1-forms

1297: \beq

1298: E\,dx-(c^2M)\,dt\,,\qquad\qquad M\,dx-E\,dt \label{5.1}

1299: \eeq

1300: are also closed. There is a simple correspondence. In fact

1301: $$

1302: E\,dx-(c^2M)\,dt = \f{p}{4}\,dX -\f{q}{4}\,dY -\f12\,dx

1303: \quad 

1304: -M\,dx+ E\,dt  = \f{p}{4c}\,dX + \f{q}{4c}\,dY -\f12\,dt.

1305: $$

1306: Recalling (\ref{4.1})-(\ref{4.2}), these can be written

1307: in terms of the $X$-$Y$ coordinates as

1308: \beq

1309: {(1-\cos w)\,p\over 8}\,dX-{(1-\cos z)\,q\over 8}\,dY\,,\label{5.2}

1310: \eeq

1311: \beq

1312: {(1-\cos w)\,p\over 8c}\,dX+{(1-\cos z)\,q\over 8c}\,dY\,,\label{5.3}

1313: \eeq

1314: respectively.

1315: Using (\ref{2.24})--(\ref{2.26}), one easily checks that these forms are indeed

1316: closed:

1317: \beq

1318: \left({(1-\cos w)\,p\over 8}\right)_Y={c'pq\over 64c^2}\Big[

1319: \sin z(1-\cos w)-\sin w(1-\cos z)\Big]

1320: =-\left({(1-\cos z)\,q\over 8}\right)_X\,,\label{sour}

1321: \eeq

1322: $$

1323: \left({(1-\cos w)\,p\over 8c}\right)_Y=

1324: {c'pq\over 64c^3}\Big[\sin(w+z)-(\sin w +\sin z)\Big]

1325: =\left({(1-\cos z)\,q\over 8c}\right)_X\,.

1326: $$

1327: In addition, we have the 1-forms

1328: \beq

1329: dx={(1+\cos w)\,p\over 4}\,dX-{(1+\cos z)\,q\over 4}\,dY\,,\label{5.4}

1330: \eeq

1331: \beq

1332: dt={(1+\cos w)\,p\over 4c}\,dX+{(1+\cos z)\,q\over 4c}\,dY\,,\label{5.5}

1333: \eeq

1334: which are obviously closed.

1335: 

1336: 

1337: 

1338: \begin{figure}[htb]

1339: \centering

1340: \includegraphics[width=0.55\textwidth]{b79.eps} %Yuxi's

1341: %\includegraphics[width=0.55\textwidth]{FIG/b79.eps} %AB's

1342: \caption{\footnotesize  }

1343: \label{fig4}

1344: \end{figure}

1345: 

1346: 

1347: The solutions $u=u(X,Y)$ constructed in Section 3 are

1348: {\it conservative}, in the sense that the integral of

1349: the form (\ref{5.2}) along every Lipschitz continuous, closed curve

1350: in the $X$-$Y$ plane is zero.     

1351: 

1352: To prove the inequality (\ref{1.7}), fix any $\tau>0$.  

1353: The case $\tau<0$ is identical. For a given $r>0$ arbitrarily large,

1354: define the set (fig.~4)

1355: \beq

1356: \Gamma\doteq \Big\{ (X,Y)\,;~~0\leq t(X,Y)\leq \tau\,,\quad X\leq r\,,

1357: ~~Y\leq r\Big\}\label{5.6}

1358: \eeq

1359: By construction, the the map $(X,Y)\mapsto (t,x)$ will act as follows:

1360: $$

1361: A\mapsto (\tau, a)\,,\qquad B\mapsto (\tau, b)\,,\qquad C\mapsto(0,c)\,,

1362: \qquad D\mapsto (0,d)\,,

1363: $$

1364: for some $a<b$ and $d<c$.

1365: Integrating the 1-form (\ref{5.2}) along the boundary of $\Gamma$ we

1366: obtain

1367: \beq

1368: \begin{array}{rl}

1369: \int_{AB}& {(1-\cos w)\,p\over 8}\,dX-{(1-\cos z)\,q\over 8}\,dY\, \\[4mm]

1370: &=\int_{DC} {(1-\cos w)\,p\over 8}\,dX-{(1-\cos z)\,q\over 8}\,dY\,

1371: -\int_{DA}{(1-\cos w)\,p\over 8}\,dX-\int_{CB}{(1-\cos z)\,q\over 8}\,dY\\[4mm]

1372: &\leq \int_{DC} {(1-\cos w)\,p\over 8}\,dX-{(1-\cos z)\,q\over 8}\,dY\,

1373: \\[4mm]

1374: &=\int_d^c \f12\Big[ u_t^2(0,x)+c^2\big(u(0,x)\big)\,u_x^2(0,x)\Big]\,dx

1375: \,.

1376: \end{array}

1377: \label{5.7}

1378: \eeq

1379: On the other hand, using (\ref{5.4}) we compute

1380: \beq

1381: \begin{array}{rl}

1382: \int_a^b \f12 \Big[ u_t^2(\tau,x)+& c^2\big(u(\tau,x)\big)\,u_x^2(\tau,x)

1383: \big]\,dx\\[4mm]

1384: &=\int_{AB\cap\{\cos w\not= -1\}}

1385: {(1-\cos w)\,p\over 8}\,dX+\int_{AB\cap \{\cos z\not= -1\}}

1386: {(1-\cos z)\,q\over 8}\,dY\\[4mm]

1387: &\leq \E_0\,.

1388: \end{array}\label{5.8}

1389: \eeq

1390: Notice that the last relation in (\ref{5.7}) is satisfied as an equality,

1391: because at time $t=0$, along the curve $\gamma_0$

1392: the variables  $w,z$ never assume the value $-\pi$.

1393: Letting $r\to +\infty$ in (\ref{5.6}), one has

1394: $a\to -\infty$, $b\to +\infty$.

1395: Therefore (\ref{5.7}) and (\ref{5.8}) together imply

1396: $\E(t)\leq \E_0$, proving (\ref{1.7}).

1397: \v

1398: 

1399: 

1400: \begin{figure}[htb]

1401: \centering

1402: \includegraphics[width=0.95\textwidth]{b80.eps} %Yuxi's

1403: %\includegraphics[width=0.95\textwidth]{FIG/b80.eps} %AB's

1404: \caption{\footnotesize Proving Lipschitz continuity }

1405: \label{fig5}

1406: \end{figure}

1407: 

1408: 

1409: 

1410: We now prove the Lipschitz continuity of the map $t\mapsto

1411: u(t,\cdot)$ in the $\L^2$ distance.

1412: For this purpose, for any fixed time $\tau$, we let

1413: $\mu_\tau=\mu_\tau^- + \mu_\tau^+$

1414: be the positive measure on the real line

1415: defined as follows. 

1416: In the smooth case,

1417: \beq

1418: \mu_\tau^-\big(\,]a,b[\,\big)

1419: =\frac{1}{4}\int_a^b R^2(\tau,x)\,dx\,,\qquad\qquad

1420: \mu_\tau^+\big(\,]a,b[\,\big)

1421: =\frac{1}{4}\int_a^b S^2(\tau,x)\,dx\,.

1422: \label{5.9'}

1423: \eeq

1424: To define $\mu_\tau^\pm$ in the general case,

1425: let $\gamma_\tau$ be the boundary of the

1426: set

1427: \beq

1428: \Gamma_\tau\doteq \big\{(X,Y)\,;~~t(X,Y)\leq\tau\big\}\,.\label{5.9}

1429: \eeq

1430: Given any open interval $]a,b[\,$,

1431: let $A=(X_A, Y_A)$ and $B=(X_B, Y_B)$ be the points on

1432: $\gamma_\tau$ such that

1433: $$

1434: x(A)=a\,,\qquad X_P-Y_P\leq X_A-Y_A\quad\hbox{for every

1435: point $P\in \gamma_\tau$ with $x(P)\leq a$}\,,

1436: $$

1437: $$

1438: x(B)=b\,,\qquad X_P-Y_P\geq X_B-Y_B\quad\hbox{for every

1439: point $P\in \gamma_\tau$ with $x(P)\geq b$}\,.

1440: $$

1441: Then

1442: \beq

1443: \mu_\tau\big(\,]a,b[\,\big)

1444: =\mu_\tau^-\big(\,]a,b[\,\big)+\mu_\tau^+\big(\,]a,b[\,\big)\,,

1445: \label{5.10}

1446: \eeq

1447: where

1448: \beq

1449: \mu_\tau^-\big(\,]a,b[\,\big)\doteq \int_{AB}

1450: {(1-\cos w)\,p\over 8}\,dX\,\qquad

1451: \mu_\tau^+\big(\,]a,b[\,\big)\doteq -\int_{AB}

1452: {(1-\cos z)\,q\over 8}\,dY\,.

1453: \label{5.10'}

1454: \eeq

1455: 

1456: Recalling the discussion at (\ref{5.1})--(\ref{5.2}), it is clear 

1457: that $\mu_\tau^-$, $\mu_\tau^+$

1458: are bounded, positive measures, and

1459: $\mu_\tau(\R)=\E_0$, for all $\tau$.

1460: Moreover, by (\ref{5.9'}) and (\ref{2.5}),

1461: $$

1462: \int_a^b c^2 u_x^2\,dx \leq \int_a^b\f12(R^2+S^2)dx \leq 2\mu(]a, b[).

1463: $$

1464: For any $a < b$, this yields the estimate

1465: \beq

1466: |u(\tau, b) -u(\tau, a)|^2 \leq |b-a|\int_a^b u_x^2(\tau, y)\,dy\leq 

1467: 2\kappa^2|b-a|\mu_\tau(]a, b[). 

1468: \label{5.11}

1469: \eeq

1470: Next, for a given $ h>0$, $y\in\R$, we seek an estimate on the distance

1471: $\big| u(\tau+ h, \,y)-u(\tau,y)|$. As in fig.~5,

1472: let $\gamma_{\tau+ h}$ be the boundary of the set

1473: $\Gamma_{\tau+ h}$, as in (\ref{5.9}).

1474: Let $P=(P_X,\,P_Y)$ be the point on $\gamma_\tau$

1475: such that

1476: $$

1477: x(P)=y\,,\qquad X_{P'}-Y_{P'}\leq X_P-Y_P\quad\hbox{for every

1478: point $P'\in \gamma_\tau$ with $x(P')\leq x$}\,,

1479: $$

1480: Similarly, let $Q=(Q_X,\,Q_Y)$ be the point on $\gamma_{\tau+ h}$

1481: such that

1482: $$

1483: x(Q)=y\,,\qquad X_{Q'}-Y_{Q'}\leq X_Q-Y_Q\quad\hbox{for every

1484: point $Q'\in \gamma_{\tau+ h}$ with $x(Q')\leq y$}\,.

1485: $$

1486: Notice that $X_P\leq X_Q$ and $Y_P\leq Y_Q$.

1487: Let $P^+=(X^+, Y^+)$ be a point on $\gamma_\tau$ with

1488: $X^+ = X_Q$, and let $P^-=(X^-, Y^-)$ be a point on $\gamma_\tau$ with

1489: $Y^- = Y_Q$.

1490: Notice that $x(P^+)\in \,]y,\, y+\kappa h[\,$, because the

1491: point $\big(\tau, x(Q)\big)$ lies on some characteristic

1492: curve with speed $-c(u)>-\kappa$, passing through the point $(\tau+ h,\,y)$.

1493: Similarly, $x(P^-)\in \,]y-\kappa h,\,y[\,$. Recalling that the forms 

1494: in (\ref{5.2}) and (\ref{5.5}) are closed, we obtain the estimate

1495: \beq

1496: \begin{array}{rcl}

1497: \big|u(Q)-u(P^+)\big|&\leq&\int_{Y^+}^{Y_Q}

1498: \big|u_Y(X_Q,Y)\big|\,dY\\[4mm]

1499: &=&\int_{Y^+}^{Y_Q}

1500: \left|{\sin z\over 4c}\,q\right|\,dY\\[4mm]

1501: &=&\int_{Y^+}^{Y_Q}

1502: \left({(1+\cos z)\,q\over 4c}\right)^{1/2}\left({(1-\cos z)\,q

1503: \over 4}\right)^{1/2}\,dY\\[4mm]

1504: &\leq &\left(\int_{Y^+}^{Y_Q}

1505: {(1+\cos z)\,q\over 4c}\,dY\right)^{1/2}\cdot

1506: \left(\int_{Y^+}^{Y_Q}{(1-\cos z)\,q

1507: \over 4}\,dY\right)^{1/2}\\[4mm]

1508: &\leq& \left\{ \int_{P^-P^+} \left[

1509: {(1-\cos w)\,p\over 4}\,dX-{(1-\cos z)\,q\over 4}\,dY\right]\right\}^{1/2}

1510: \cdot  h^{1/2}\,.

1511: \end{array}\label{5.12}

1512: \eeq

1513: The last term in (\ref{5.12}) contains the integral of the 1-form at 

1514: (\ref{5.2}),

1515: along the curve $\gamma_\tau$, between $P^-$ and $P^+$. Recalling the

1516: definition (\ref{5.10})--(\ref{5.10'}) and the estimate (\ref{5.11}),

1517:  we obtain the bound

1518: \beq

1519: \begin{array}{rcl}

1520: \big|u(\tau+ h , x)-u(\tau ,x)\big|^2&\leq &

1521: 2\big|u(Q)-u(P^+)\big|^2+2\big|u(P^+)-u(P)\big|^2\\[3mm]

1522: &\leq & 4 h\cdot\mu_\tau\big(]x-\kappa h,\,x+\kappa h[\big)

1523: + 4\kappa^2\cdot(\kappa h)\cdot\mu_\tau \big(]x\,,~x+ h[\big)\,.

1524: \end{array}\label{5.13}

1525: \eeq

1526: Therefore, for any $ h>0$,

1527: \beq

1528: \begin{array}{rcl}

1529: \big\|u(\tau+ h,\cdot)-u(\tau,\cdot)\big\|_{\L^2}

1530: &=&\left(\int\big|u(\tau+ h,x)-u(\tau,x)\big|^2\,dx\right)^{1/2}\\[4mm]

1531: &\leq&\left(\int

1532: 4(1+\kappa^3) h \cdot \mu_\tau\big(]x-\kappa h\,,~x+\kappa h[\big)

1533: \,dx\right)^{1/2}\\[4mm]

1534: &= &\Big( 4(\kappa^3+1) h^2\,\mu_\tau(\R)\Big)^{1/2}\\[4mm]

1535: &= & h\cdot\big[4(\kappa^3+1)\,\E_0\big]^{1/2}.

1536: \end{array} \label{5.14}

1537: \eeq

1538: This proves the uniform Lipschitz continuity of the map

1539: $t\mapsto u(t,\cdot)$, stated at (\ref{1.lip}).

1540: 

1541: 

1542: \section{Regularity of trajectories}  %6

1543: 

1544: \setcounter{equation}{0}

1545: 

1546: In this section we prove the continuity of the

1547: functions $t\mapsto u_t(t,\cdot)$ and $t\mapsto u_x(t,\cdot)$,

1548: as functions with values in $\L^p$.  This will complete the proof

1549: of Theorem 1.

1550: 

1551: We first consider the case where the initial data $(u_0)_x$, $u_1$

1552: are smooth with compact support.

1553: In this case, the solution $u=u(X,Y)$ remains smooth on the entire

1554: $X$-$Y$ plane.   Fix a time $\tau$ and let $\gamma_\tau$ be the

1555: boundary of the set $\Gamma_\tau$, as in (\ref{5.9}).

1556: We claim that

1557: \beq

1558: {d\over dt} u(t,\cdot)\bigg|_{t=\tau}= u_t(\tau,\cdot)\label{6.1}

1559: \eeq

1560: where, by (\ref{2.14}), (\ref{2.18}) and (\ref{2.26}),

1561: \beq

1562: \begin{array}{rcl}

1563: u_t (\tau,x) & \doteq&  u_X\,X_t+u_Y\,Y_t \\ [4mm]

1564:  & =&  {p\,\sin w\over 4c}\,

1565: {2c\over p(1+\cos w)}+{q\,\sin z\over 4c}\,

1566: {2c\over q(1+\cos z)}={\sin w\over 2(1+\cos w)}+{\sin z\over 2(1+\cos z)}\,.

1567: \end{array} \label{6.2}

1568: \eeq

1569: Notice that (\ref{6.2}) defines the values of $u_t(\tau,\cdot)$

1570: at almost every point $x\in\R$, i.e.~at all points outside the support of

1571: the singular part of the measure $\mu_\tau$ defined at (\ref{5.10}).

1572: By the inequality (\ref{1.7}), recalling that $c(u)\geq \kappa^{-1}$,

1573: we obtain

1574: \beq

1575: \int_\R \big|u_t(\tau,x)\big|^2\,dx\leq \kappa^2\,\E(\tau)

1576: \leq \kappa^2\E_0. \label{6.3}

1577: \eeq

1578: 

1579: To prove (\ref{6.1}), let any $\ve>0$ be given.

1580: There exist finitely many disjoint intervals

1581: $[a_i,\,b_i]\subset \R$, $i=1,\ldots,N$, with the following property.

1582: Call $A_i, B_i$ the points on $\gamma_\tau$ such that

1583: $x(A_i)=a_i$, $x(B_i)=b_i$.  Then one has

1584: \beq

1585: \min\big\{ 1+\cos w(P)\,,~1+\cos z(P)\big\}<2\ve\label{6.4}

1586: \eeq

1587: at every point $P$ on $\gamma_\tau$ contained in one of the arcs

1588: $A_iB_i$, while

1589: \beq

1590: 1+\cos w(P)>\ve\,,\qquad\qquad 1+\cos z(P)>\ve\,,\label{6.5}

1591: \eeq

1592: for every point $P$ along $\gamma_\tau$, not contained in any of the arcs

1593: $A_iB_i$.

1594: Call $J\doteq \cup_{1\leq i\leq N} [a_i, b_i]$, $J'=\R\setminus J$,

1595: and notice that, as a function of the original variables, $u=u(t,x)$

1596: is smooth in a neighborhood of the set $\{\tau\}\times J'$.

1597: Using Minkowski's inequality and the differentiability of $u$ on $J'$,

1598: we can write

1599: \beq

1600: \begin{array}{rl}

1601: \lim_{h\to 0}&

1602: {1\over h}\left(\int_{\R} \Big| u(\tau+h,x)-u(\tau,x)-h\,u_t(\tau, x)\Big|^p

1603: dx\right)^{1/p}\\[4mm]

1604: &\leq \lim_{h\to 0}

1605: {1\over h}\left(\int_J \Big| u(\tau+h,x)-u(\tau,x)\Big|^p

1606: dx\right)^{1/p}+\left(\int_J \big|u_t(\tau,x)\big|^p\,dx\right)^{1/p}

1607: \end{array}\label{6.6}

1608: \eeq

1609: We now provide an estimate on the measure of the ``bad" set $J$:

1610: \beq

1611: \begin{array}{rcl}

1612: \meas(J)&=&\int_J dx=\sum_i\int_{A_iB_i} {(1+\cos w)\,p\over 4}\,dX

1613: -{(1+\cos z)\,q\over 4}\,dY\\[4mm]

1614: &\leq &2\ve \sum_i \int_{A_iB_i}

1615: {(1-\cos w)\,p\over 4}\,dX

1616: -{(1-\cos z)\,q\over 4}\,dY\\[4mm]

1617: &\leq& 2\ve  \int_{\gamma_\tau}

1618: {(1-\cos w)\,p\over 4}\,dX

1619: -{(1-\cos z)\,q\over 4}\,dY

1620: \leq 2\ve\,\E_0\,.

1621: \end{array}\label{6.7}

1622: \eeq

1623: 

1624: Now choose $q=2/(2-p)$ so that

1625: ${p\over 2}+{1\over q}=1$. Using H\"older's inequality

1626: with conjugate exponents $2/p$ and $q$, and recalling (\ref{5.14}), we obtain

1627: $$

1628: \begin{array}{rcl}

1629: \int_J \Big| u(\tau+h,x)-u(\tau,x)\Big|^p

1630: dx&\leq &\meas(J)^{1/ q}\cdot

1631: \left(\int_J \Big| u(\tau+h,x)-u(\tau,x)\Big|^2

1632: dx\right)^{p/2}\\[4mm]

1633: &\leq &\big[ 2\ve\,\E_0\big]^{1/q}\cdot

1634: \Big(\big\|u(\tau+h,\cdot)-u(\tau,\cdot)\big\|^2_{\L^2}

1635: \Big)^{p/2}\\[4mm]

1636: &\leq& \big[ 2\ve\,\E_0\big]^{1/q}\cdot

1637: \Big(h^2 \big[ 4(\kappa^3+1)\,\E_0\big]

1638: \Big)^{p/2}.

1639: \end{array}

1640: $$

1641: Therefore,

1642: \beq

1643: \limsup_{h\to 0}

1644: {1\over h}\left(\int_J \Big| u(\tau+h,x)-u(\tau,x)-h\Big|^p

1645: dx\right)^{1/p}\leq [2\ve\,\E_0]^{1/pq}\cdot \big[ 4(\kappa^3+1)\,\E_0

1646: \big]^{1/2}.\label{6.8}

1647: \eeq

1648: In a similar way we estimate

1649: $$

1650: \int_J\big|u_t(\tau,x)\big|^p\,dx\leq

1651: \big[\meas(J)\big]^{1/ q}\cdot

1652: \left(\int_J \Big| u_t(\tau,x)\Big|^2

1653: dx\right)^{p/2},

1654: $$

1655: \beq

1656: \left(\int_J\big|u_t(\tau,x)\big|^p\,dx\right)^{1/p}\leq

1657: \meas(J)^{1/ pq}\cdot \big[\kappa^2\,\E_0\big]^{p/2}.\label{6.9}

1658: \eeq 

1659: Since $\ve>0$ is arbitrary, from (\ref{6.6}), (\ref{6.8}) and (\ref{6.9}) 

1660: we conclude

1661: \beq

1662: \lim_{h\to 0}~

1663: {1\over h}\left(\int_{\R} \Big| u(\tau+h,x)-u(\tau,x)-h\,u_t(\tau, x)\Big|^p

1664: dx\right)^{1/p} =0.

1665: \label{6.10}

1666: \eeq

1667: The proof of continuity of the map $t\mapsto u_t$ is similar.

1668: Fix $\ve>0$. Consider the intervals $[a_i,b_i]$ as before.

1669: Since $u$ is smooth on a neighborhood of $\{\tau\}\times J'$,

1670: it suffices to estimate

1671: $$

1672: \begin{array}{cl}

1673: & \limsup_{h\to 0}\,\int\big|u_t(\tau+h,x)-u_t(\tau,x)\big|^p\,dx \\[4mm]

1674: \leq &\limsup_{h\to 0}

1675: \int_J \big|u_t(\tau+h,x)-u_t(\tau,x)\big|^p\,dx\\[4mm]

1676: \leq & \limsup_{h\to 0} \big[\meas(J)\big]^{1/ q}\cdot

1677: \left(\int_J \Big| u_t(\tau+h,\,x)-u_t(\tau,x)\Big|^2 dx\right)^{p/2} \\[4mm]

1678: \leq & \limsup_{h\to 0}

1679: \big[2\ve\,\E_0\big]^{1/ q}\cdot \Big(\big\|u_t(\tau+h,\cdot)\big\|_{\L^2}

1680: +\big\|u_t(\tau,\cdot)\big\|_{\L^2}\Big)^p\\[4mm]

1681: \leq & \big[2\ve\E_0]^{1/q}\, \big[ 4\E_0\big]^p.

1682: \end{array}

1683: $$

1684: Since $\ve>0$ is arbitrary, this proves continuity.

1685: \v

1686: To extend the result to general initial data, such that 

1687: $(u_0)_x, u_1\in\L^2$, we consider a sequence of smooth initial

1688: data, with $(u^\nu_0)_x, u^\nu_1\in \C^\infty_c$,

1689: with

1690: $u_0^n\to u_0$ uniformly,

1691: $(u_0^n)_x\to (u_0)_x$ almost everywhere and in $\L^2$,

1692: $u_1^n\to u_1$ almost everywhere and in $\L^2$.

1693: 

1694: 

1695: The continuity of the function $t\mapsto u_x(t,\cdot)$

1696: as a map with values in $\L^p$, $1\leq p <2$, is proved in

1697: an entirely similar way.

1698: 

1699: 

1700: \section{Energy conservation}  %7

1701: 

1702: \setcounter{equation}{0}

1703: 

1704: This section is devoted to the proof of Theorem 3, stating

1705: that, in some sense, the total energy of the solution 

1706: remains constant in time. 

1707: 

1708: A key tool in our analysis is

1709: the {\it wave interaction potential}, defined as

1710: \beq

1711: \Lambda(t)\doteq (\mu_t^-\otimes\mu_t^+)\big\{ (x,y)\,; ~x>y\big\}\,.

1712: \label{7.1}

1713: \eeq

1714: We recall that $\mu_t^\pm$ are the positive measures defined at

1715: (\ref{5.10'}).

1716: Notice that, if $\mu_t^+, \mu_t^-$ are

1717: absolutely continuous w.r.t.~Lebesgue measure, so that

1718: (\ref{5.9'}) holds, then (\ref{7.1}) 

1719: is equivalent to

1720: $$

1721: \Lambda(t)\doteq \f14 \dint_{x>y} R^2(t,x)\,S^2(t,y)\,dxdy\,.$$

1722: \v

1723: {\bf Lemma 1.} {\it The map $t\mapsto \Lambda(t)$ has locally bounded

1724: variation. Indeed, there exists a one-sided

1725: Lipschitz constant $L_0$ such that

1726: \beq

1727: \Lambda(t)-\Lambda(s) ~\leq L_0\cdot (t-s)\qquad\qquad t>s>0\,.

1728: \eeq

1729: }

1730: 

1731: To prove the lemma, we first give a formal argument, valid

1732: when the solution $u=u(t,x)$ remains smooth.

1733: We first notice that (\ref{2.4}) 

1734: implies

1735: $${d\over dt} \,(4\Lambda(t))

1736: \leq -\int 2c\, R^2S^2\,dx +\int \big(R^2+S^2\big)\,dx\cdot

1737: \int{c'\over 2c} |R^2S-RS^2|\,dx

1738: $$

1739: $$\leq -2\kappa^{-1}\int  R^2S^2\,dx+ 

1740: 4\E_0 \left\|{c'\over 2c}\right\|_{\L^\infty}

1741: \int\big|R^2S-RS^2|\,dx \,,

1742: $$

1743: where $\kappa^{-1}$ is a lower bound for $c(u)$.

1744: For each $\ve>0$ we have

1745: $

1746: |R|\leq \ve^{-1/2}+\ve^{1/2}\,R^2

1747: $.

1748: Choosing $\ve>0$ such that

1749: $$

1750: \kappa^{-1}>4\E_0 \left\|{c'\over 2c}\right\|_{\L^\infty}\cdot 2\sqrt\ve\,,

1751: $$

1752: we thus obtain

1753: $$

1754: {d\over dt}(4 \Lambda(t))\leq -\kappa^{-1}

1755: \int R^2S^2\,dx +{16\,\E_0^2\over\sqrt\ve}

1756: \left\|{c'\over 2c}\right\|_{\L^\infty}\,.

1757: $$

1758: This yields the $\L^1$ estimate

1759: $$

1760: \int_0^\tau \int \big(|R^2S|+|RS^2|\big)\,dxdt =

1761: \O(1)\cdot \big[ \Lambda(0)+ \E_0^2\,\tau\big]=

1762: \O(1)\cdot (1+\tau)\E_0^2,

1763: $$

1764: where $\O(1)$ denotes a quantity whose absolute value

1765: admits a uniform bound, depending only on the function $c=c(u)$ and

1766: not on the particular solution under consideration.

1767: In particular, the map $t\mapsto \Lambda(t)$ has bounded variation

1768: on any bounded interval.

1769: It can be discontinuous, with downward jumps.

1770: 

1771: To achieve a rigorous proof of Lemma 1, we need to reproduce the 

1772: above argument in terms of the variables $X,Y$.

1773: As a preliminary, we observe that for every $\ve>0$

1774: there exists a constant $\kappa_\ve$

1775: such that

1776: \beq

1777: \begin{array}{rl}

1778:      & \left| \sin z(1-\cos w)-\sin w(1-\cos z)\right| \\[3mm]

1779: \leq & \kappa_\ve\cdot \left(\tan^2\frac{w}{2}+\tan^2\frac{z}{2}

1780: \right)(1+\cos w)(1+\cos z)+\ve (1-\cos w)(1-\cos z)

1781: \end{array}\label{7.3}

1782: \eeq

1783: for every pair of angles $w,z$.

1784: 

1785: Now fix $0\leq s<t$.  Consider the sets $\Gamma_s,\Gamma_t$ as in (\ref{5.9})

1786: and define

1787: $\Gamma_{st}\doteq \Gamma_t\setminus\Gamma_s$.  

1788: Observing that 

1789: $$dxdt = \frac{pq}{8c}(1+\cos w)(1+\cos z)dXdY\,,$$

1790: we can now write

1791: \beq

1792: \begin{array}{rl}

1793: & \int_s^t\int_{-\infty}^\infty \frac{R^2+S^2}{4}\,dxdt \ = \ (t-s)\E_0 \\[4mm]

1794: = & \dint_{\Gamma_{st}} \frac{1}{4}\left(\tan^2\frac{w}{2}+\tan^2\frac{z}{2}

1795: \right)\cdot \frac{pq}{8c}(1+\cos w)(1+\cos z)\,dXdY\,.

1796: \end{array}\label{7.4}

1797: \eeq

1798: The first identity holds only for smooth solutions, but the second

1799: one is always valid. Recalling (\ref{sour}) and (\ref{5.10'}), and then using 

1800: (\ref{7.3})-(\ref{7.4}), we obtain

1801: $$

1802: \begin{array}{l}

1803: \Lambda(t)-\Lambda(s) \leq -\dint_{\Gamma_{st}} 

1804: \frac{1-\cos w}{8} p\cdot \frac{1-\cos z}{8} q\,dXdY \\[4mm]

1805: \quad +\E_0\cdot \dint_{\Gamma_{st}} \frac{c'}{64 c^2} pq\left[ 

1806: \sin z(1-\cos w)-\sin w(1-\cos z)\right]\,dXdY\\[4mm]

1807: \leq -\frac{1}{64}\dint_{\Gamma_{st}} (1-\cos w)(1-\cos z)\,pq\,dXdY\\[4mm]

1808: \quad+\E_0\cdot \dint_{\Gamma_{st}} \frac{c'}{64 c^2} pq\left[ 

1809: \kappa_\ve\cdot \left(\tan^2\frac{w}{2}+\tan^2\frac{z}{2}

1810: \right)(1+\cos w)(1+\cos z)\right.\\[4mm]

1811: \qquad\left. +\ve (1-\cos w)(1-\cos z)\right]\,dXdY\\[4mm]

1812: \leq \kappa (t-s)\,,

1813: \end{array}

1814: $$

1815: for a suitable constant $\kappa$.  This proves the lemma.

1816: 

1817: 

1818: 

1819: \medskip

1820: 

1821: To prove Theorem 3, consider the three sets

1822: $$\Omega_1\doteq \Big\{(X,Y)\,;~~w(X,Y)=-\pi\,,\qquad z(X,Y)\not= -\pi

1823: \,,\qquad c'\big(u(X,Y)\big)\not= 0\Big\}\,,$$

1824: $$\Omega_2\doteq \Big\{(X,Y)\,;~~z(X,Y)=-\pi\,,\qquad w(X,Y)\not= -\pi

1825: \,,\qquad c'\big(u(X,Y)\big)\not= 0\Big\}\,,$$

1826: $$\Omega_3\doteq \Big\{(X,Y)\,;~~z(X,Y)=-\pi\,,\qquad w(X,Y)= -\pi

1827: \,,\qquad c'\big(u(X,Y)\big)\not= 0\Big\}\,.$$

1828: From the equations (\ref{2.24}), it follows 

1829: that 

1830: \beq

1831: \meas(\Omega_1)=\meas(\Omega_2)=0\,.

1832: \label{7.5}

1833: \eeq

1834: Indeed, $w_Y\not= 0$ on $\Omega_1$ and $z_X\not= 0$ on $\Omega_2$.

1835: 

1836: Let $\Omega_3^*$ be the set of Lebesgue points of $\Omega_3$.

1837: We now show that

1838: \beq

1839: \meas\Big(\big\{ t(X,Y)\,;~~(X,Y)\in \Omega_3^*\big\}\Big)=0\,.

1840: \label{7.6}

1841: \eeq

1842: To prove (\ref{7.4}), fix any $P^*=(X^*,Y^*)\in\Omega_3^*$ and let 

1843: $\tau=t(P^*)$. We claim that

1844: \beq

1845: \limsup_{h,k\to 0+} ~{\Lambda(\tau-h)-\Lambda(\tau+k)\over h+k}~=~+\infty

1846: \label{7.7}

1847: \eeq

1848: 

1849: 

1850: 

1851: \begin{figure}[htb]

1852: \centering

1853: \includegraphics[width=0.55\textwidth]{b85.eps} %Yuxi's

1854: %\includegraphics[width=0.55\textwidth]{FIG/b85.eps} %AB's

1855: \caption{\footnotesize  }

1856: \label{fig6}

1857: \end{figure}

1858: 

1859: By assumption, for any $\ve>0$ arbitrarily small we can find

1860: $\delta>0$ with the following property.

1861: For any square $Q$ centered at $P^*$ with side of length $\ell<\delta$, 

1862: there exists a

1863: vertical segment $\sigma$ and a horizontal segment $\sigma'$,

1864: as in fig.~6,

1865: such that

1866: \beq 

1867: \meas \big(\Omega_3\cap \sigma\big) \geq (1-\ve)\ell\,,

1868: \qquad\qquad  \meas \big(\Omega_3\cap \sigma'\big) \geq (1-\ve)\ell\,,

1869: \label{7.8}

1870: \eeq

1871: Call 

1872: $$t^+\doteq \max\Big\{ t(X,Y)\,;~~(X,Y)\in \sigma\cup\sigma'\Big\}\,,$$

1873: $$t^-\doteq \min\Big\{ t(X,Y)\,;~~(X,Y)\in \sigma\cup\sigma'\Big\}\,.$$

1874: Notice that, by (\ref{4.2}),

1875: \beq

1876: t^+-t^-\leq \int_\sigma \frac{(1+\cos w)p}{ 4c}\,dX

1877: +\int_{\sigma'} \frac{(1+\cos z)q}{ 4c}\,dY\leq c_0\cdot (\ve\ell)^2\,.

1878: \label{7.9}

1879: \eeq

1880: Indeed, the integrand functions are Lipschitz continuous. Moreover,

1881: they vanish oustide a set of measure $\ve\ell$.

1882: On the other hand, 

1883: \beq

1884: \Lambda(t^-)-\Lambda(t^+)\geq c_1 (1-\ve)^2\ell^2  -

1885: c_2 (t^+-t^-)

1886: \label{7.10}

1887: \eeq

1888: for some constant $c_1>0$. Since $\ve>0$ was arbitrary, 

1889: this implies (\ref{7.5}).

1890: 

1891: Recalling that the map $t\mapsto \Lambda$ 

1892: has bounded variation, from (\ref{7.5}) it follows  (\ref{7.4}). 

1893: 

1894: \smallskip

1895: We now observe that the singular part of 

1896: $\mu_\tau$ is nontrivial only if the set 

1897: $$\big\{ P\in\gamma_\tau\,;~~w(P)= -\pi ~~\hbox{or}~~z(P)=-\pi

1898: \big\}$$

1899: has positive 1-dimensional measure.

1900: By the previous analysis, restricted to the region

1901: where $c'\not= 0$, this can happen only for a 

1902: set of times having zero measure.

1903: 

1904: 

1905: 

1906: 

1907: 

1908: \vsk

1909: 

1910: {\bf Acknowledgment:}

1911: Alberto Bressan was supported by the Italian M.I.U.R.,

1912: within the research project \#2002017219, while  

1913: Yuxi Zheng has been partially supported by grants

1914: NSF DMS 0305497 and 0305114.

1915: 

1916: \vsk

1917: 

1918: \begin{thebibliography}{99}

1919: 

1920: 

1921: \bibitem{[1]} M.~Albers, R.~Camassa, D.~Holm, and J.~Marsden,

1922: The geometry of peaked solitons and billiard solutions of a class of 

1923: integrable PDE's, {\it Lett. Math. Phys.}, {\bf 32}(1994), pp.~137--151.

1924: 

1925: 

1926: \bibitem{[2]} M.~Balabane, Non--existence of global solutions for some

1927: nonlinear wave equations with small Cauchy data, {\it C. R. Acad. Sc. Paris},

1928: {\bf 301}(1985), pp.~569--572.

1929: 

1930: 

1931: \bibitem{[3]} H.~Berestycki, J.~M.~Coron and I.~Ekeland (eds.), 

1932: {\it Variational Methods},

1933: Progress in Nonlinear Differential Equations and Their Applications, Vol.~4,

1934: Birkh\"auser, Boston (1990).

1935: 

1936: \bibitem{BrCo} A.~Bressan and A.~Constantin, Global solutions to the

1937: Camassa-Holm equations, to appear.

1938: 

1939: \bibitem{[BZZ]}

1940: A.~Bressan, Ping Zhang, and Yuxi Zheng,

1941: On asymptotic variational wave equations,

1942: {\it Arch. Rat. Mech. Anal.}, submitted April 29, 2004.

1943: 

1944: 

1945: 

1946: \bibitem{[4]} R.~Camassa and D.~Holm, An integrable shallow water equation 

1947: with peaked solitons, {\it Phys.~Rev.~Lett.}, {\bf 71}(1993), pp.~1661--1664.

1948: 

1949: \bibitem{[5]} R.~Camassa, D.~Holm, J.~Hyman, A new integrable shallow water 

1950: equation, to appear in {\it Adv. in Appl. Mech.}

1951: 

1952: 

1953: \bibitem{[6]} D.~Christodoulou and A.~Tahvildar-Zadeh,  On the regularity of 

1954: spherically symmetric wave maps, {\it Comm. Pure Appl. Math.}, {\bf 46}(1993), 

1955: pp.~1041--1091.

1956: 

1957: 

1958: \bibitem{[7]} J.~Coron, J.~Ghidaglia, and F.~H\'elein (eds.), {\it Nematics},

1959: Kluwer Academic Publishers, 1991.

1960: 

1961: \bibitem{[10]} J.~L.~Ericksen and D.~Kinderlehrer (eds.), 

1962: {\it Theory and Application of

1963: Liquid Crystals}, IMA Volumes in Mathematics and its Applications, Vol.~5,

1964: Springer-Verlag, New York (1987).

1965: 

1966: \bibitem{[14]} R.~T.~Glassey, Finite--time blow--up for solutions of 

1967: nonlinear wave equations, {\it Math. Z.}, {\bf 177}(1981), pp.~323--340.

1968: 

1969: \bibitem{[GHZ]} R.~T.~Glassey, J.~K.~Hunter and Yuxi Zheng, Singularities in a

1970: nonlinear variational wave equation, {\it J. Differential Equations},

1971: {\bf 129}(1996), 49-78.

1972: 

1973: \bibitem{[16]} R.~T.~Glassey, J.~K.~Hunter and Yuxi Zheng, Singularities and

1974: oscillations in a nonlinear variational wave equation, {\it Singularities and

1975: Oscillations}, edited by J.~Rauch and M.~E.~Taylor,  IMA, Vol {\bf 91},

1976: Springer, 1997.

1977: 

1978: \bibitem{[17]} A.~Grundland and E.~Infeld, A family of nonlinear Klein-Gordon

1979: equations and their solutions, {\it J. Math. Phys.}, {\bf 33}(1992),

1980: pp.~2498--2503.

1981: 

1982: 

1983: \bibitem{[18]} B.~Hanouzet and J.~L.~Joly, Explosion pour des probl\`emes

1984: hyperboliques semi--lin\'eaires avec second membre non compatible,

1985: {\it C. R. Acad. Sc. Paris}, {\bf 301}(1985), pp.~581--584.

1986: 

1987: \bibitem{[19]} R.~Hardt, D.~Kinderlehrer, and Fanghua Lin,

1988: Existence and partial regularity of static liquid crystal configurations.

1989: {\it Comm. Math. Phys.}, {\bf 105}(1986),  pp.~547--570.

1990: 

1991: \bibitem{[21]} J.~K.~Hunter and R.~A.~Saxton, Dynamics of director fields,

1992: {\it SIAM J. Appl. Math.}, {\bf 51}(1991), pp.~1498-1521.

1993: 

1994: 

1995: \bibitem{[22]} J.~K.~Hunter and Yuxi Zheng, On a nonlinear hyperbolic

1996: variational equation I and II, {\it Arch. Rat. Mech. Anal.},

1997: {\bf 129}(1995), pp.~305-353 and 355-383.

1998: 

1999: \bibitem{[HZ]} J.~K.~Hunter and Yuxi Zheng, On a completely 

2000: integrable nonlinear hyperbolic variational equation,

2001: {\it Physica D.} {\bf 79}(1994), 361--386.

2002: 

2003: \bibitem{[23]} F.~John, Blow--up of solutions of nonlinear wave equations in

2004: three space dimensions, {\it Manuscripta Math.}, {\bf 28}(1979),

2005: pp.~235--268.

2006: 

2007: \bibitem{[25]} T.~Kato, Blow--up of solutions of some nonlinear hyperbolic

2008: equations, {\it Comm. Pure Appl. Math.}, {\bf 33}(1980), pp.~501--505.

2009: 

2010: 

2011: \bibitem{[26]} D.~Kinderlehrer,  Recent developments in liquid crystal theory,

2012: in {\it  Frontiers in pure and applied mathematics : a collection of papers

2013: dedicated to Jacques-Louis Lions on the occasion of his sixtieth birthday},

2014: ed. R.~Dautray, Elsevier, New York, pp.~151--178 (1991).

2015: 

2016: \bibitem{[27]} S.~Klainerman and M.~Machedon, Estimates for the null forms 

2017: and the spaces

2018: $H_{s,\delta}$, {\it Internat. Math. Res. Notices}, 1996, no.~17, pp.~853--865.

2019: 

2020: 

2021: \bibitem{[28]} S.~Klainerman and A.~Majda, Formation of singularities for wave

2022: equations including the nonlinear vibrating string, {\it Comm. Pure Appl. 

2023: Math.}, {\bf 33}(1980), pp.~241--263.

2024: 

2025: \bibitem{[29]} P.~Lax, Development of singularities of solutions of 

2026: nonlinear hyperbolic partial differential equations, {\it J. Math. Phys.}, 

2027: {\bf 5}(1964), pp.~611--613.

2028: 

2029: \bibitem{[30]} H.~Levine, Instability and non--existence of global solutions to

2030: nonlinear wave equations, {\it Trans. Amer. Math. Soc.}, {\bf 192}(1974),

2031: pp.~1--21.

2032: 

2033: \bibitem{[31]} H.~Lindblad, Global solutions of nonlinear wave equations,

2034: {\it  Comm. Pure Appl. Math.}, {\bf 45}(1992), pp.~1063--1096.

2035: 

2036: \bibitem{[34]} Tai-Ping Liu, Development of singularities in the nonlinear

2037: waves for quasi--linear hyperbolic partial differential equations,

2038: {\it J. Differential Equations}, {\bf 33}(1979), pp.~92--111.

2039: 

2040: 

2041: 

2042: \bibitem{[37]} R.~A.~Saxton, Dynamic instability of the liquid crystal

2043: director, in {\it Contemporary Mathematics} Vol.~100:

2044: Current Progress in Hyperbolic Systems,

2045: pp.~325--330, ed. W.~B.~Lindquist, AMS, Providence, 1989.

2046: 

2047: \bibitem{[38]} J.~Schaeffer, The equation $u_{tt}-\Delta u = |u|^p$ for 

2048: the critical value of $p$,

2049: {\it Proc. Roy. Soc. Edinburgh Sect. A}, {\bf 101A}(1985), pp.~31--44.

2050: 

2051: 

2052: \bibitem{[40]} J.~Shatah, Weak solutions and development of singularities

2053: in the $SU(2)$ $\sigma$-model, {\it Comm. Pure Appl. Math.}, {\bf 41}(1988),

2054: pp.~459--469.

2055: 

2056: \bibitem{[41]} J.~Shatah and A.~Tahvildar-Zadeh,  Regularity of harmonic 

2057: maps from Minkowski space into rotationally symmetric manifolds, 

2058: {\it Comm. Pure Appl. Math.}, {\bf 45}(1992), pp.~947--971.

2059: 

2060: \bibitem{[43]} T.~Sideris, Global behavior of solutions to nonlinear wave 

2061: equations in three

2062: dimensions, {\it Comm. in Partial Diff. Eq.}, {\bf 8}(1983), pp.~1291--1323.

2063: 

2064: \bibitem{[44]} T.~Sideris, Nonexistence of global solutions to semilinear

2065: wave equations in high dimensions, {\it J. Diff. Eq.}, {\bf 52}(1984), 

2066: pp.~378--406.

2067: 

2068: \bibitem{[45]} W.~Strauss, {\it Nonlinear wave equations}, CBMS Lectures 73, 

2069: AMS, Providence, 1989.

2070: 

2071: \bibitem{[49]} E.~Virga, {\it Variational Theories for Liquid Crystals},

2072: Chapman \& Hall, New York (1994).

2073: 

2074: 

2075: \bibitem{[52]} Ping Zhang and Yuxi Zheng, On oscillations of an

2076: asymptotic equation of a nonlinear variational wave equation,

2077: {\it Asymptotic Analysis}, {\bf 18}(1998), pp.~307--327 .

2078: 

2079: \bibitem{[53]} Ping Zhang and Yuxi Zheng, On the existence and

2080:  uniqueness of solutions to an asymptotic equation of a variational wave

2081: equation,  {\it Acta Mathematica Sinica}, {\bf 15}(1999), pp.~115--130.

2082: 

2083: \bibitem{[54]} Ping Zhang and Yuxi Zheng, On the existence and

2084: uniqueness to an asymptotic equation of a variational wave

2085: equation with general data, {\it Arch. Rat. Mech. Anal.} 

2086: {\bf 155}(2000), 49--83.

2087: 

2088: 

2089: \bibitem{[55]} Ping Zhang and Yuxi Zheng, Rarefactive solutions to a

2090: nonlinear variational wave equation,  {\it Comm. Partial Differential

2091: Equations}, {\bf 26}(2001), 381-419.

2092: 

2093: \bibitem{[56]} Ping Zhang and Yuxi Zheng, Singular and rarefactive solutions

2094: to a nonlinear variational wave equation, {\it Chinese Annals of

2095: Mathematics}, {\bf 22B}, 2(2001), 159-170.

2096: 

2097: \bibitem{[57]} Ping Zhang and Yuxi Zheng, Weak solutions to a nonlinear

2098: variational wave equation, {\it Arch. Rat. Mech. Anal.}, 

2099: {\bf 166} (2003), 303--319.

2100: 

2101: 

2102: \bibitem{58} Ping Zhang and Yuxi Zheng,  On the second-order 

2103: asymptotic equation of a variational wave equation, 

2104: {\it Proc A of the Royal Soc. Edinburgh, A. Mathematics}, 

2105: {\bf 132A}(2002), 483--509. 

2106: 

2107: \bibitem{[59]} Ping Zhang and Yuxi Zheng, Weak solutions to a

2108: nonlinear variational wave equation with general data,

2109: Annals of Inst. H. Poincar\'e, 2004 (in press).

2110: 

2111: 

2112: 

2113: \bibitem{[Zo]} H.~Zorski and E.~Infeld, New soliton equations for dipole 

2114: chains, {\it Phys. Rev. Lett.}, {\bf 68}(1992), pp.~1180--1183.

2115: 

2116: \end{thebibliography}

2117: \end{document}

2118: 

2119: 

2120: 

2121: %%%%%%%%%%%%%%%%%%%%%%%%%%

2122: %%%%%%%%%%%%%%%%%%%%%%%%%%

2123: %%%%%%%%%%%%%%%%%%%%%%%%%%

2124: 

2125: 

2126: {\bf 1. Comparison with the asypmtotic model}

2127: It is interesting to draw parallels with the asymptotic equations.

2128: For the asymptotic equation the term $\int u_x^2$ is conserved between

2129: characteristics. For \eqnref{1.1}, the term $\int _a^\infty u_X^2\,dX$

2130: is bounded for any $a\in \R$, where $X$ is the characteristic coordinate.

2131: The time direction is the $Y$ direction. The term ${1+R^2\over X_x}$ is

2132: an $O^\ast(1)$.

2133: 

2134: 

2135: 

2136: {\bf 2. Interaction potential estimate}

2137: 

2138: To obtain further bounds on the terms in \eqnref{1.10},

2139: define the interaction potential as

2140: $$

2141: \Lambda(t)\doteq \dint_{x<y} R^2(t,x)\,S^2(t,y)\,dxdy\,.$$

2142: By \eqnref{1.4} there holds

2143: $$

2144: \eqalign{{d\over dt} \Lambda(t)&

2145: \leq -\int 2c\, R^2S^2\,dx +\int \big(R^2+S^2\big)\,dx\cdot

2146: \int{c'\over 2c} |R^2S-RS^2|\,dx

2147: \cr

2148: &\leq -2\kappa^{-1}\int  R^2S^2\,dx+ 4E \left\|{c'\over 2c}\right\|_{\L^\infty}

2149: \int\big|R^2S-RS^2|\,dx \,,\cr}

2150: $$

2151: where $c_*>0$ is a lower bound for $c(u)$.

2152: Observe that, for each $\ve>0$ we have

2153: $$

2154: |R|\leq {1\over \sqrt \ve}+\sqrt\ve\,R^2\,.

2155: $$

2156: Choosing $\ve>0$ such that

2157: $$

2158: c_*>4E \left\|{c'\over 2c}\right\|_{\L^\infty}\cdot 2\sqrt\ve\,,

2159: $$

2160: we obtain

2161: $$

2162: {d\over dt} \Lambda(t)\leq -c_* \int R^2S^2\,dx +{16\,E^2\over\sqrt\ve}

2163: \left\|{c'\over 2c}\right\|_{\L^\infty}\,.

2164: $$

2165: This yields the $\L^1$ estimate

2166: $$

2167: \int_0^\tau \int \big(|R^2S|+|RS^2|\big)\,dxdt =

2168: \O(1)\cdot \big[ \Lambda(0)+ E^2\,\tau\big]=

2169: \O(1)\cdot (1+\tau)E^2,\eqnlbl{3.11}

2170: $$

2171: where $\O(1)$ denotes a quantity whose absolute value

2172: admits a uniform bound, depending only on the function $c=c(u)$ and

2173: not on the particular solution under consideration.

2174: 

2175: 

2176: 

2177: 

2178: 

2179: 

2180: 

2181: We discuss briefly the issue of uniqueness.

2182: As for the equations studied in (\cite{[BZZ]}), here uniqueness does not hold

2183: within the class of all

2184: H\"older continuous functions which satisfy (\ref{1.1}) in its

2185: distributional form (\ref{1.4'}).  The underlying reason is that,

2186: at a time $\tau$ when a singularity develops, the measure

2187: with density $1/2(u_t^2+c^2(u) u_x^2)$

2188: may concentrate as a Dirac mass on a single point. In this case, the

2189: corresponding energy functional

2190: (\ref{1.5'}) satisfies $\E(\tau)<\E(\tau-)$.

2191: For $t>\tau$, the solution can now be prolonged in two different ways.

2192: We can construct a ``conservative" extension, such that

2193: $$\lim_{t\to \tau+}  \E(t)=\lim_{t\to \tau-}  \E(t)\,.$$

2194: In addition, there exists a ``dissipative" extension, for which

2195: $$

2196: \lim_{t\to \tau+}  \E(t)=\E(\tau)<\lim_{t\to \tau-}  \E(t)\,.

2197: $$

2198: Since our solution is obtained as a fixed point of a

2199: contractive transformation, a natural way to achieve uniqueness

2200: is to impose conditions which imply that a given solution provides

2201: a fixed point to the same transformation. 

2202: 

2203: (?)

2204: Adopting this strategy, in Section 6 (?) we prove a uniqueness result

2205: valid for conservative solutions. (?)

2206: 

2207: (Theorem 4 has uniqueness. So it is a matter of interpretation only.)

2208: 

2209: 

2210: \section{Some examples}  %8

2211: 

2212: \setcounter{equation}{0}

2213: 

2214: We use some simple examples to illustrate the conservativeness of our

2215: solution. Suppose that the wave speed function $c(u)$ has a critical

2216: point at $u=u_0$. Consider the solution $(w, z, p, q, u)$ in the $(X, Y)$

2217: plane defined as follows: $(z, p, q, u) \equiv (0, 1, 1, u_0)$ for

2218: all $(X, Y)$; $w=0$ for $X< 1$ and $X>2$; $w=-\pi$ for $ 1 \leq X \leq 2$.

2219: This solution has constant energy $\E(t) = 1/4$. But this energy

2220: does not appear in the ordinary form of the energy through the energy density

2221: $E(t)$. When mapping back to the $(t, x)$ plane, we recover only the

2222: constant solution $u(t,x) = u_0$ with zero energy.  The lost energy travels

2223: along the backward characteristic $x+c(u_0)t = 1$ as a ``ghost particle''

2224: without interaction with the field $u(t,x)$.

2225: 

2226: Now if we change the field $u_0$ to a different one ahead of this

2227: ghost particle, then the state $w(X,Y)$

2228: will grow above $-\pi$, assuming that $u_0$ is a (local) maximum of $c$,

2229: upon examining the equation for $w$. Thus the ghost particle dissolves into

2230: the field and the energy manifests itself through the ordinay energy

2231: density function. See Figure 5. If we reverse the time, we obtain a

2232: scenario that some concentrated energy forms from ordinary energy

2233: and disappears forever. Thus our conservative solution may loose energy

2234: for a long time.

2235: 

2236: We can send two such ghost particles collide and create an open set in

2237: the $X-Y$ plane in which both space and time $x-t$ are frozen. 

2238: 

2239: %\begin{figure}[htb]

2240: %\centering

2241: %%\includegraphics[width=0.75\textwidth]{nwef5.eps} %Yuxi's

2242: %\includegraphics[width=0.75\textwidth]{FIG/b80.eps} %AB's

2243: %\caption{\footnotesize }

2244: %\label{fig5}

2245: %\end{figure}

2246: