1: \documentclass{amsart}

2: 

3: 

4: \usepackage{amsmath, amsthm, amsopn, amsfonts}

5: 

6: 

7: %\usepackage{showkeys}

8: 

9: \usepackage{color}

10: 

11: \usepackage{graphicx}

12: 

13: 

14: \usepackage{graphpap}

15: 

16: 

17: %\usepackage[english, francais]{babel}

18: 

19: 

20: 

21: 

22: 

23: %\textwidth= 14.5 cm

24: 

25: 

26: %\textheight= 20.5cm

27: 

28: 

29: %\hoffset = -1cm

30: 

31: 

32: % UNE MACRO POUR PLACER DU TEXTE SUR UNE BOITE. 

33: 

34: 

35: % SYNTAXE: \ecriture{boite}{

36: 

37: 

38: % \aat{coordhoriz}{coordvert}{texte}} la macro trace un quadrillage 

39: 

40: 

41: %de 50

42: 

43: 

44: % pour faciliter le placement du texte. On le supprime apres mise au

45: 

46: 

47: % point par la commande \pasdegrille en debut de fichier*

48: 

49: 

50: 

51: 

52: 

53: 

54: 

55: 

56: \def\finpreuve{\hfill $\Box$\par}

57: 

58: 

59: \def\pasdegrille{\let\grille = \pasgrille}

60: 

61: 

62: \def\ecriture#1#2{\setbox1=\hbox{#1}

63: 

64: 

65: \dimen1= \wd1

66: 

67: 

68: \dimen2=\ht1

69: 

70: 

71: \dimen3=\dp1

72: 

73: 

74: \grille #2 \box1 }

75: 

76: 

77: \def\aat#1#2#3{

78: 

79: 

80: \divide \dimen1 by 48

81: 

82: 

83: \dimen3=\dimen1

84: 

85: 

86: \multiply \dimen1 by #1

87: 

88: 

89: \advance \dimen1 by -\dimen3

90: 

91: 

92: \divide \dimen1 by 101

93: 

94: 

95: \multiply \dimen1 by 100

96: 

97: 

98: \divide \dimen2 by \count11

99: 

100: 

101: \multiply \dimen2 by #2 

102: 

103: 

104: \setbox0=\hbox{#3}\ht0=0pt\dp0=0pt

105: 

106: 

107:   \rlap{\kern\dimen1 \vbox to0pt{\kern-\dimen2\box0\vss}}\dimen1= \wd1

108: 

109: 

110: \dimen2=\ht1}

111: 

112: 

113: \def\pasgrille{

114: 

115: 

116: \count12= \dimen1 

117: 

118: 

119: \divide \count12 by 50

120: 

121: 

122: \divide \dimen2 by \count12

123: 

124: 

125: \count11 =\dimen2

126: 

127: 

128: \ 

129: 

130: 

131: \divide \dimen1 by 48

132: 

133: 

134: \setlength{\unitlength}{\dimen1}

135: 

136: 

137: \smash{\rlap{\ }}

138: 

139: 

140: \dimen1= \wd1

141: 

142: 

143: \dimen2=\ht1

144: 

145: 

146: }

147: 

148: 

149: \def\grille{

150: 

151: 

152: \count12= \dimen1 

153: 

154: 

155: \divide \count12 by 50

156: 

157: 

158: \divide \dimen2 by \count12

159: 

160: 

161: \count11 =\dimen2

162: 

163: 

164: \ 

165: 

166: 

167: \divide \dimen1 by 48

168: 

169: 

170: \setlength{\unitlength}{\dimen1}

171: 

172: 

173: \smash{\rlap{\graphpaper[1](0,0)(50, \count11)}}

174: 

175: 

176: \dimen1= \wd1

177: 

178: 

179: \dimen2=\ht1

180: 

181: 

182: }

183: 

184: 

185: 

186: 

187: 

188: \pasdegrille

189: 

190: 

191: %%%%

192: 

193: 

194: %%%

195: 

196: 

197: % FIN DE KLA

198: 

199: 

200: %MACRO POUR INCLURE DU TEXTE DANS UNE FIGURE

201: 

202: 

203: %

204: 

205: 

206: %

207: 

208: 

209: 

210: 

211: 

212: 

213: 

214: 

215: \usepackage{amssymb}

216: 

217: 

218: %\usepackage{draftcopy}

219: 

220: 

221: \ifx\pdfoutput\undefined

222: 

223: 

224: \usepackage{graphicx}  \DeclareGraphicsExtensions{.ps}

225: 

226: 

227: \def\Dessin#1{\begin{center}

228: 

229: 

230: \includegraphics{#1.mps}

231: 

232: 

233: \end{center}

234: 

235: 

236: }

237: 

238: 

239: \else

240: 

241: 

242: \usepackage{amsfonts}

243: 

244: 

245: \usepackage[pdftex]{graphicx}  \DeclareGraphicsExtensions{.pdf,.mps}

246: 

247: 

248: \def\Dessin#1{\begin{center}

249: 

250: 

251: \includegraphics{#1.mps}

252: 

253: 

254: \end{center}

255: 

256: 

257: }

258: 

259: 

260: \fi

261: 

262: 

263: 

264: 

265: 

266: \newcommand{\PSbox}[1]{\includegraphics[0in,0in][3in,3in]{#1}}

267: 

268: 

269: \usepackage[T1]{fontenc}

270: 

271: 

272: 

273: 

274: 

275: % Added by JW

276: 

277: 

278: \setlength{\marginparwidth}{.8in}

279: 

280: 

281: \newcommand{\mar}[1]{{\marginpar{\sffamily{\scriptsize #1}}}} 

282: 

283: 

284: \usepackage{epic, eepic}

285: 

286: 

287: % Have to temporarily widen margins to make marginal notes fit!

288: 

289: 

290: \setlength{\textheight}{8in} \setlength{\oddsidemargin}{0.35in}

291: 

292: 

293: \setlength{\evensidemargin}{0.35in} \setlength{\textwidth}{5.8in}

294: 

295: 

296: \setlength{\topmargin}{0.18in} \setlength{\headheight}{0.18in}

297: 

298: 

299: \setlength{\marginparwidth}{1.0in}

300: 

301: 

302: 

303: 

304: 

305: \newcommand{\abs}[1]{{\left\lvert{#1}\right\rvert}}

306: 

307: 

308: \newcommand{\norm}[1]{{\left\lVert{#1}\right\rVert}}

309: 

310: 

311: \newcommand{\bignorm}[1]{{\Bigl\lVert{#1}\Bigr\rVert}}

312: 

313: 

314: \newcommand{\pa}{{\partial}}

315: 

316: 

317: \newcommand{\h}{{\frac 12}}

318: 

319: 

320: \newcommand{\cinf}{{\mathcal{C}}^\infty}

321: 

322: 

323: 

324: 

325: 

326: \def\mathbbone{{\mathchoice {1\mskip-4mu \text{l}} {1\mskip-4mu \text{l}}

327: 

328: 

329: { 1\mskip-4.5mu \text{l}} { 1\mskip-5mu \text{l}}}}

330: 

331: 

332: 

333: 

334: 

335: \newcommand{\bc}{\mathbb C}

336: 

337: 

338: \newcommand{\br}{\mathbb R}

339: 

340: 

341: \newcommand{\bz}{\mathbb Z}

342: 

343: 

344: \newcommand{\im}{\operatorname{\rm Im}}

345: 

346: 

347: \newcommand{\mn}[1]{\Vert#1\Vert}

348: 

349: 

350: \newcommand{\op}{\operatorname{Op}}

351: 

352: 

353: \newcommand{\re}{\operatorname{\rm Re}}

354: 

355: 

356: \newcommand{\set}[1]{\left\{\,#1\,\right\}}

357: 

358: 

359: \newcommand{\w}[1]{\langle #1\rangle }

360: 

361: 

362: 

363: 

364: 

365: 

366: 

367: 

368: \def\squarebox#1{\hbox to #1{\hfill\vbox to #1{\vfill}}} 

369: 

370: 

371: \newcommand{\stopthm}{\hfill\hfill\vbox{\hrule\hbox{\vrule\squarebox 

372: 

373: 

374:                  {.667em}\vrule}\hrule}\smallskip} 

375: 

376: 

377: 

378: 

379: 

380: \pagestyle{headings}

381: 

382: 

383: 

384: 

385: 

386: 

387: 

388: 

389: 

390: 

391: 

392: 

393: 

394: 

395: 

396: 

397: 

398: %\pagestyle{plain}

399: 

400: 

401: 

402: 

403: 

404: \newcommand{\1}{{\bold 1}} 

405: 

406: 

407: \newcommand{\F}{{\mathcal F}} 

408: 

409: 

410: \newcommand{\CC}{{\mathbb C}}

411: 

412: 

413: \newcommand{\CI}{{\mathcal C}^\infty }

414: 

415: 

416: \newcommand{\HH}{{\mathcal H}}

417: 

418: 

419: \newcommand{\CIc}{{\mathcal C}^\infty_{\rm{c}} }

420: 

421: 

422: \newcommand{\CIb}{{\mathcal C}^\infty_{\rm{b}} }

423: 

424: 

425: \newcommand{\Op}{{\operatorname{Op}}}

426: 

427: 

428: %^{{w}}_h}}

429: 

430: 

431: \newcommand{\Oo}{{\mathcal O}} 

432: 

433: 

434: \newcommand{\Hh}{{\mathcal H}}

435: 

436: 

437: \newcommand{\pic}{{\mbox{Pic}}}

438: 

439: 

440: \newcommand{\Z}{{\mathbb Z}}

441: 

442: 

443: \newcommand{\Q}{{\mathbb Q}}

444: 

445: 

446: \newcommand{\RR}{{\mathbb R}}

447: 

448: 

449: \newcommand{\SP}{{\mathbb S}}

450: 

451: 

452: \newcommand{\ZZ}{{\mathbb Z}}

453: 

454: 

455: \newcommand{\NN}{{\mathbb N}}

456: 

457: 

458: \newcommand{\MO}{{\mathcal M}}

459: 

460: 

461: \newcommand{\U}{{\mathcal U}}

462: 

463: 

464: \newcommand{\N}{{\mathbb N}}

465: 

466: 

467: \newcommand{\vol}{\operatorname{vol}}

468: 

469: 

470: \newcommand{\rank}{\operatorname{rank}}

471: 

472: 

473: \newcommand{\itt}{\operatorname{it}}

474: 

475: 

476: \newcommand{\KKer}{\operatorname{ker}}

477: 

478: 

479: \newcommand{\supp}{\operatorname{supp}}

480: 

481: 

482: \newcommand{\Ran}{\operatorname{Ran}}

483: 

484: 

485: \newcommand{\UC}{U_{\circlearrowright}}

486: 

487: 

488: \newcommand{\QC}{Q_{\circlearrowright}}

489: 

490: 

491: \newcommand{\comp}{\operatorname{comp}}

492: 

493: 

494: \newcommand{\loc}{\operatorname{loc}}

495: 

496: 

497: \newcommand{\tr}{\operatorname{tr}}

498: 

499: 

500: \newcommand{\rest}{\!\!\restriction}

501: 

502: 

503: \newcommand{\ttt}{|\hspace{-0.25mm}|\hspace{-0.25mm}|}

504: 

505: 

506: \renewcommand{\Re}{\mathop{\rm Re}\nolimits}

507: 

508: 

509: \renewcommand{\Im}{\mathop{\rm Im}\nolimits}

510: 

511: 

512: 

513: 

514: 

515: % JW messed with the following:

516: 

517: 

518: \theoremstyle{plain}

519: 

520: 

521: \def\Rm#1{{\rm#1}}

522: 

523: 

524: \newtheorem{thm}{Theorem}

525: 

526: 

527: %   \renewcommand{\thethm}{\arabic{thm}}

528: 

529: 

530: \newtheorem{prop}{Proposition}[section]

531: 

532: 

533: %   \renewcommand{\theprop}{\arabic{prop}}

534: 

535: 

536: \newtheorem{cor}[prop]{Corollary}

537: 

538: 

539: %   \renewcommand{\thecor}{}

540: 

541: 

542: \newtheorem{lem}{Lemma}[section]

543: 

544: 

545: %   \renewcommand{\thelem}{\arabic{lem}}

546: 

547: 

548: %\numberwithin{lem}{section}

549: 

550: 

551: 

552: 

553: 

554: %\numberwithin{prop}{section}

555: 

556: 

557: \theoremstyle{definition}

558: 

559: 

560: \newtheorem{ex}{EXAMPLE}[section]

561: 

562: 

563: \newtheorem{exmple}{Example}[section]

564: 

565: 

566: \newtheorem{rem}{Remark}

567: 

568: 

569: \newtheorem{defn}[prop]{Definition} 

570: 

571: 

572: 

573: \numberwithin{equation}{section}

574: 

575: \newcommand{\thmref}[1]{Theorem~\ref{#1}}

576: \newcommand{\secref}[1]{Section~\ref{#1}}

577: \newcommand{\lemref}[1]{Lemma~\ref{#1}}

578: \newcommand{\exref}[1]{Example~\ref{#1}}

579: \newcommand{\corref}[1]{Corollary~\ref{#1}}

580: \newcommand{\propref}[1]{Proposition~\ref{#1}}

581: 

582: \def\bbbone{{\mathchoice {1\mskip-4mu \rm{l}} {1\mskip-4mu \rm{l}}

583: { 1\mskip-4.5mu \rm{l}} { 1\mskip-5mu \rm{l}}}}

584: 

585: 

586: \title[Eigenfunctions for partially rectangular billiards]

587: {Eigenfunctions for partially rectangular billiards}

588: \author[J. Marzuola]{Jeremy Marzuola}

589: \address{Mathematics Department, University of California \\

590: Evans Hall, Berkeley, CA 94720, USA}

591: \email{marzuola@math.berkeley.edu}

592: 

593: \usepackage{amssymb}

594: \usepackage{amsmath, amsthm, amsopn, amsfonts}

595: %\usepackage[dvips]{graphicx}

596: 

597: 

598: 

599: \def\11{{\rm 1~\hspace{-1.4ex}l} }

600: \def\R{\mathbb R}

601: \def\C{\mathbb C}

602: \def\Z{\mathbb Z}

603: \def\N{\mathbb N}

604: \def\E{\mathbb E}

605: \def\T{\mathbb T}

606: \def\O{\mathbb O}

607: \def\K{\mathbb K}

608: \def\cal{\mathcal}

609: 

610: 

611: \begin{document}    

612: 

613:    

614: \maketitle   

615:    

616: \section{Introduction}   

617: \label{in}

618: In this note, we further develop the methods of Burq-Zworski \cite{BZ3} 

619: to study eigenfunctions for billiards which have rectangular components: these include the Bunimovich 

620: billiard, the Sinai billiard, and the recently popular pseudointegrable billiards

621: \cite{Bog}.  The results are an application of a "black box" point of view as

622: presented in \cite{BZ2} by the same authors.  

623: 

624: 

625: 

626: \begin{figure}[ht]

627: \includegraphics[width=10cm]{sinai1.epsi}

628: \caption{Experimental images of eigenfunctions in a Sinai billiard

629: microwave cavity -- see {\tt http://sagar.physics.neu.edu}. We see

630: that there is always a non-vanishing presence near the boundary of the

631: obstacle as predicted by Theorem \ref{t:s} below.}

632: \label{fig:sin}

633: \end{figure}

634: 

635: 

636: By a partially rectangular billiard, we mean a connected planar domain,

637: $ \Omega $, with a piecewise smooth boundary, which contains a  rectangle,

638: $ R \subset \Omega $, such that if we decompose the boundary of 

639: $ R $, into pairs of parallel segments, $ \partial R = \Gamma_1 \cup \Gamma_2 $,

640: then $ \Gamma_i \subset \partial \Omega $, for at least one $ i$.

641: 

642: We show that for such billiards, 

643: the eigenfunctions of the Dirichlet, Neumann,

644: or periodic Laplacian cannot concentrate in closed sets in the interior of the rectangular part.

645: A combination of this elementary result with 

646: the now standard, but highly non-elementary, 

647: propagation results of Melrose-Sj\"ostrand \cite{MeSj} and Bardos-Lebeau-Rauch \cite{BLR},

648: can give further improvements -- see \cite{BZ2},\cite{BZ3}.

649: 

650: Here, we prove further non-concentration results, away from the obstacle in the Sinai billiard

651: (see Fig.1 and Theorem \ref{t:s}), and along certain trajectories

652: in pseudointegrable billiards, (see Fig.5 and Theorem \ref{t:j}).  

653: For recent motivation coming from the study of {\em quantum chaos}

654: we suggest \cite{Bog},\cite{BZ3},\cite{Do},\cite{Ze}, and references given there.

655: 

656: \medskip

657: \noindent

658: {\sc Acknowledgments.} This paper is a development of an unpublished work by N. Burq and M. Zworski, 

659: who treated the case of a square billiard rather than a general rectangular billiard.

660: The author is very grateful to Maciej Zworski and Nicolas Burq for allowing me to use results from their 

661: unpublished work, as well as many helpful conversations.  I would also like to thank Srinivas Sridhar for 

662: allowing the use the experimental images shown in Fig.\ref{fig:sin}.  

663: 

664: 

665: \section{Semiclassical Pseudodifferential Operators on a Torus}

666: \label{PDO}

667: 

668: In this section, we would like to discuss properties of Pseudodifferential Operators (PDO's) on a torus.  

669: To begin, we examine the nature of PDO's and their symbol classes.  In ${\mathbb R}^n$, we define a

670: Weyl quantization of an operator $a(x,hD)$ where $a \in {\mathcal S} ({\mathbb R}^{2n}), a=a(x, \xi)$ by:

671: $$a (x,hD) u(x) = \frac{1}{(2 \pi h)^n} \int_{{\mathbb R}^n} \int_{{\mathbb R}^n} e^{\frac{i}{h} \langle x-y, \xi \rangle}

672: a(x, \xi) u(y) dy d\xi ,$$

673: for $u \in {\mathcal S}$ and a symbol class by:  

674: $$S^k_\delta (m) = \{ a \in C^{\infty} ({\mathbb R}^{2n}) \| | \partial^\alpha a | \leq C_\alpha h^{-\delta |\alpha| -k} m 

675: \ \text{for all multi-indices} \ \alpha \},$$

676: where $m: {\mathbb R}^{2n} \to (0,\infty)$ is is an order function, i.e. there exist constants $C$, $N$ such that

677: $$ m(z) \leq C \langle z - w \rangle^N m(w).$$

678: We also have $$S^{-\infty}_\delta (m) = \bigcap_{k=-\infty}^{\infty} S^k_\delta (m).$$ 

679: For $k,\delta = 0$ we write simply $S (m)$. 

680: On a torus, however, $a$ and all its derivatives are bounded in the $x$ variable, thus for $h$ small and $k$ positive, we need not worry

681: about the derivatives in x, only those in $\xi$.  Also, for $k$ negative, provided

682: that we have the proper local regularity for our symbol $a$, this definition still works perfectly on a torus.  

683: 

684: 

685: Note also that we need only work with symbols that are periodic in the x-variable

686: with period determined by the dimensions of the torus.  In other words,

687: $a(x,\xi)=a(x+\gamma, \xi)$ for $\gamma \in (a{\mathbb Z}) \times (b{\mathbb Z})$, where $a, b \in {\mathbb R}$.  

688: Using this relation, we easily see the following propostion.

689: 

690: \begin{prop}

691: \label{p:3}  

692: If $a(x,\xi)$ is a periodic symbol in x with period $\gamma$, then $a(x,D) T_\gamma = T_\gamma a(x,D)$ where

693: $T_\gamma u(x) = u(x - \gamma)$.  

694: \end{prop}

695: \begin{proof}

696: We easily calculate:

697: 

698: \begin{eqnarray}

699: \label{eq:p3}

700: a(x,h D) T_\gamma (u) & = & \int_{{\mathbb R}^n} a(x,\xi) e^{\frac{i}{h} <x-y,\xi>} \hat{u} (y-\gamma) dy d\xi \\

701: &  = & \int_{{\mathbb R}^n} a(x-\gamma,\xi) e^{\frac{i}{h} <x-\gamma-{\tilde y},\xi>} \hat{u} (\tilde y) d{\tilde y} d\xi \\

702: &  = & T_\gamma(a(x,D) u)

703: \end{eqnarray}

704: \end{proof}

705: 

706: From the above proposition, it becomes clear that the properties of symbol classes in Euclidean space  

707: translate directly to properties of similarly defined symbol classes on a torus.  

708: For instance, we have the following result.

709: 

710: \begin{prop}

711: \label{p:4}

712: Given $u(x) = u(x+\gamma)$ where $\gamma$ is as above, and $u$ is $L^2$ on a torus, then $a(x,h D) u(x)$ is

713: $L^2$ on the torus, when $a \in {\mathcal S}_\delta (1)$, $0 \leq \delta \leq \frac{1}{2}$.

714: \end{prop}

715: \begin{proof}

716: Note that the condition on $a$ implies that it is $L^2$ bounded.

717: Given a function $u(x)$ which is periodic on a torus, we can write it as 

718: $\sum_\gamma T_\gamma u_0$ where $u_0 = \chi (x) u(x)$ and $\chi (x)$ is equal to 1 on a single copy

719: of the torus in the plane and 0 otherwise.  Note that no assumptions about the smoothness of $\chi(x)$ are made.  

720: Hence, $u_0 \in L^2$ and therefore, so is $a(x,h D) u_0$.  Then, 

721: $a(x,h D) u(x) = \sum_\gamma T_\gamma a(x,h D) u_0$.  The sum converges for each $x$ since $a(x,D) u_0$ will

722: have compact support and we have

723: $a(x,h D) u$ a periodic function that is $L^2$ on the torus. 

724: \end{proof}

725: 

726: Using similar techniques, we would like to develop the concept of a microlocal defect measure in this setting.

727: As shown in \cite{Zw}, we have the following theorem in Euclidean space:

728: \begin{thm}

729: \label{t:dm}

730: There exists a Radon measure $\mu$ on ${\mathbb R}^n$ and a sequence $h_j \rightarrow 0$ such that 

731: \begin{equation}

732: \langle a^w (x,h_j D) u(h_j), u(h_j) \rangle \rightarrow \int_{{\mathbb R}^{2n}} a(x,\xi) d\mu

733: \end{equation}

734: for all symbols $a \in S(1)$.

735: \end{thm}

736: 

737: We call $\mu$ a microlocal defect measure associated with the family $\{ u(h) \}_{0<h \leq h_0}$. 

738: Note that an $S(1)$ symbol on the torus corresponds to an $S(1)$ symbol on the plane, therefore this result proves

739: the existence of microlocal defect measures on a torus as well.  

740: 

741: \begin{proof}

742: 1.  Let $\{a_k \} \in C_c^\infty$ be dense in $C_0 ({\mathbb R}^{2n})$.  Select a sequence $h_j^1 \rightarrow 0$ such that

743: $$\langle a_1^w (x, h_j^1 D) u(h_j^1), u(h_j^1) \rangle \rightarrow \alpha_1.$$  Then, select a subsequence $\{ h_j^2 \} \subset \{ h_j^1 \}$

744: such that $$\langle a_2^w (x, h_j^2 D) u(h_j^2), u(h_j^2) \rangle \rightarrow \alpha_2.$$  Continue such that at the kth step, you take a subsequence

745: $\{ h_j^k \} \subset \{ h_j^{k-1} \}$ such that $$\langle a_k^w (x, h_j^k D) u(h_j^k), u(h_j^k) \rangle \rightarrow \alpha_k.$$  Then by

746: a diagonal argument, arrive at a sequence $h_j$ such that $$\langle a_k^w (x, h_j D) u(h_j), u(h_j) \rangle \rightarrow \alpha_k$$ for all $k=1,2,...$.

747: 

748: 2.  Define $\Phi(a_k) = \alpha_k$.  By a standard theorem on operator norms, we have for each $k$ that 

749: $$| \Phi (a_k) | = | \alpha_k | = \lim_{h_j \rightarrow \infty} | <a_k^w u(h_j), u(h_j) > | \leq

750: \limsup_{h_j \rightarrow \infty} C \| a_k^w \|_{L^2 \rightarrow L^2} \leq C \sup |a_k|.$$  The mapping $\Phi$ is bounded, linear

751: and densely defined, therefore uniquely extends to a bounded linear functional on $S(1)$, with the estimate $$ | \Phi (a) | \leq C \sup |a|$$

752: for all $a \in S(1)$.  The Riesz Representation Theorem therefore implies the existence of a (possibly complex valued) Radon measure on

753: ${\mathbb R}^{2n}$ such that $$\Phi (a) = \int_{{\mathbb R}^{2n}} a(x,\xi) d\mu.$$

754: \end{proof}

755: 

756: We now quote a general theorem about microlocal defect measures on Euclidean space which we can

757: then apply to a torus.  To state the propagation theorem in the form sufficient for our applications,

758: we follow \cite{BuIMRN}.

759: %\subsection{Semi-classical measures}

760: 

761: 

762: Let us consider a Riemannian manifold without boundary, $M$. By partitions of unity we can define semi-classical 

763: pseudo-differential operators $a(x, hD_x)$ associated to symbols $a(x, \xi)\in \CIc (T^* M)$. 

764: \par

765: Now we consider a sequence $(u_n)$ bounded in $L^2(M)$. satisfying

766: \begin{equation}\label{eq.equation}

767: (-h_n^2\Delta -1)u_n = 0.

768: \end{equation}  

769: Using~\eqref{eq.equation},

770: as in~\cite{GeLe93} (see also~\cite{BuIMRN}) we can prove the following result.

771: 

772: \begin{prop}

773: \label{prop2.1} There exist a subsequence $(n_{k})$ and a positive Radon  measure on $T^*M$, $\mu$ (a semi-classical measure for the sequence $(u_n)$), 

774: such that for any $a \in \CIc (T^* M)$

775: \begin{equation}

776: \label{eq2.8}\lim_{k\rightarrow + \infty}\left(a^w (x, h_{n_{k}}D_{x})u_{n_{k}}, u_{n_{k}}\right)_{ L^2(M)}= \langle \mu, a(x, \xi)\rangle.

777: \end{equation}

778: Furthermore this measure satisfies

779: \begin{enumerate}

780: \item The support of $\mu$ is included in the characteristic manifold:

781: \begin{equation}

782: \Sigma \stackrel{\rm{def}}{=} \{(x, \xi)\in T^*M; p(x,\xi)=\|\xi\|_x=1\}

783: \end{equation}

784: where  $\|\cdot \|_x$ is the norm for the metric at the point $x$,

785: \item The measure $\mu$ is invariant by the bicharacteristic flow (the flow of the Hamilton vector field of $p$):

786: \begin{equation}

787: \label{eq.inv}H_p \mu =0,

788: \end{equation}

789: \item For any $\varphi\in \CIc (T^*M)$, 

790: \begin{equation}

791: \label{eq.lim} \lim_{k \rightarrow + \infty} \|\varphi u_{n_k}\|^2 = \langle \mu, |\varphi|^2\rangle.

792: \end{equation}

793: \end{enumerate}

794: \end{prop}

795: The first two properties above are weak forms of the elliptic regularity and propagation of singularities results,  

796: whereas the last one states that there is no loss of $L^2$-mass at infinity in the $\xi$ variable.   

797: 

798: \begin{proof}

799: We will prove this proposition only for the case of a torus, but the methods are applicable to any manifold.

800: \bigskip

801: 

802: (0)  (Positivity)

803: We need to show that $a \geq 0$ implies $$\int_{{\mathbb T}^{2} \times {\mathbb R}^{2}} a(x, \xi) d\mu \geq 0.$$

804: Since $a \geq 0$, using the Garding inequality, we see that:  $$a^w (x,hD) \geq -Ch.$$

805: Let $h = h_j \rightarrow 0$, to see:

806: $$\int_{{\mathbb T}^{2} \times {\mathbb R}^{2}} a d\mu = \lim_{j \rightarrow \infty} \langle a^w (x,h_j D) u(h_j), u(h_j) \rangle \geq 0.$$

807: 

808: (1)  (Support of $\mu$)

809: Let $a$ be a smooth function such that $\text{supp} (a) \cap p^{-1} (1) = \emptyset$.  We must show 

810: $$\int_{{\mathbb T}^{2} \times {\mathbb R}^{2}} a d\mu = 0.$$

811: Select $\chi \in C^{\infty}_c ( {\mathbb T}^{2} \times {\mathbb R}^{2} )$ such that 

812: $$\text{supp} (a) \cap \text{supp} (\chi) = 0.$$

813: Then, $$a^w(x,hD) \left( ((p-1)^w+i \chi^w)^{-1} (p-1)^w \right) (x,hD) = a^w(x,hD) + {\mathcal O} (h^\infty)_{L^2 \rightarrow L^2}.$$

814: Apply $a^w (x,hD)$ to $u(h)$ to see that $a^w (x,hD) u(h) = o(1)$ and thus $\langle a^w (x,hD) u(h), u(h) \rangle \rightarrow 0$.

815: But, $$\langle a^w (x,hD) u(h_j), u(h_j) \rangle \rightarrow \int_{{\mathbb T}^{2} \times {\mathbb R}^{2}} a d\mu.$$

816: 

817: (2)  (Flow Invariance)

818: Select $a$ as above, then 

819: \begin{eqnarray} 

820: \langle [p^w,a^w] u(h), u(h) \rangle & = & \langle (p^w a^w -a^w p^w) u(h), u(h) \rangle \\

821:  & = & \langle a^w u(h), p^w u(h) \rangle - \langle p^w u(h), (a^w)^* u(h) \rangle \\

822:  & = & o(h) \ \text{as} \ h \rightarrow 0.

823: \end{eqnarray}

824: However, $[p^w,a^w] = \frac{h}{i} \{ p^w,a^w \} + O(h^2)$.  Hence, 

825: $$\langle [p^w,a^w] u(h), u(h) \rangle = \frac{h}{i} \langle \{ p,a \}^w u(h), u(h) \rangle + \langle o(h) u(h), u(h) \rangle.$$

826: As we let $h_j \rightarrow 0$, we get:  $$\int_{{\mathbb T}^{2} \times {\mathbb R}^{2}} \{p,a\} d\mu = 0.$$  

827: So, if $\Phi_t$ is the flow generated by the Hamiltonian vector field $H_p$, then 

828: $$\frac{d}{dt} \int_{{\mathbb T}^{2} \times {\mathbb R}^{2}} (\Phi_t* a) d\mu = 

829: \int_{{\mathbb T}^{2} \times {\mathbb R}^{2}} (H_p a)(\Phi_t) d\mu = 

830: \int_{{\mathbb T}^{2} \times {\mathbb R}^{2}} \{ p,a \} d\mu = 0.$$

831: Now, (3) follows easily by looking at the operator $|\varphi (x, \xi)|^2$ and applying the result about existence of a microlocal defect measure.

832: \end{proof}

833: 

834: 

835: \section{Partially rectangular billiards}

836: \label{prb}

837: 

838: In this section we will need to recall the basic control results \cite{Bu92},\cite{BZ2}

839: for rectagles, and the propagation results \cite{MeSj},\cite{BLR},\cite{BuIMRN},\cite{BuGe} for 

840: billiards. Since in the specific application presented in Section \ref{app} we only use propagation

841: away from the boundary, that is the only case we will review.

842: 

843: 

844: The following result from ~\cite{Bu92} is related to some earlier control results 

845: of Haraux~\cite{Ha} and  Jaffard~\cite{Ja}.\footnote{We remark that as noted in \cite{Bu92} 

846: the result holds for 

847: any product manifold $M= M_{x}\times M_{y}$, and the proof is essentially the same.}

848: \begin{prop}

849: \label{p:1} Let $\Delta$ be  the Dirichlet, Neumann, or periodic

850: Laplace operator on the rectangle $R= [0, 1]_{x} \times [0,a]_{y}$. 

851: Let $\omega_x$ be a non-empty open subset of $[0,1]$.  Then for any non-empty  $\omega \subset R$ of the form $

852: \omega= \omega_{x} \times [0,a]_{y}$, there 

853: exists $C$ such that for any solutions of

854: \begin{equation}

855: (\Delta -z) u =f \ \text{ on $R$}, \ u \rest_{\partial R}=0

856: \end{equation}

857: we have

858: \begin{equation}

859: \label{eq:6.12}\|u\|^2_{{L^2(R)}}\leq C \left(\|f \|^2_{H^{-1} (

860: [0,1]_{x}; L^2([0,a]_{y})) } +

861: \|u\rest _{\omega}\|^2_{{L^2(\omega)}} \right)

862: \end{equation}

863: \end{prop}

864: \begin{proof}

865: We will consider the Dirichlet case (the proof is the same in the other two 

866: cases) and 

867: decompose $u,f$ in terms of 

868: the basis of $L^2([0,a])$ formed by the Dirichlet eigenfunctions

869: $e_{k}(y)=  { \sqrt {{2}/a}}\sin(2k\pi y/a)$,

870: \begin{equation}

871: u(x,y)= \sum_{k}e_{k}(y) u_{k}(x), \qquad f(x,y)= \sum_{k}e_{k}(y) f_{k}(x).

872: \end{equation}

873: We get for $u_{k}, f_{k}$ the equation

874: \begin{equation}\label{estres.1}

875: \left(\Delta_{x}-\left(z+\left({2k\pi}/{a}\right)^2\right)\right)u_{k}= f_{k},\qquad u_{k}(0)=u_k(1)=0.

876: \end{equation}

877: We now claim that 

878: \begin{equation}

879: \label{eq:cont}

880: \|u_{k}\|^2_{{L^2([0,1]_{x})}}\leq C \left(\|f_k \|^2_{H^{-1}([0,1]_{x})} +

881: \|u_k \rest _{\omega_{x}}\|^2_{{L^2(\omega)}}\right), 

882: \end{equation}

883: from which, by summing the squares in $k$, we get~\eqref{eq:6.12}.

884: 

885: To see \eqref{eq:cont} we can use the propagation result above \ref{prop2.1} in dimension

886: one, but in this case an elementary calculation is easily available -- see

887: \cite{BZ3}.

888: \end{proof}

889: 

890: 

891: 

892: The following theorem is an easy consequence of Proposition \ref{p:1}:

893: 

894: \begin{thm}

895: \label{t:1}

896: Let $ \Omega $ be a partially rectangular billiard with the rectangular

897: part $ R \subset \Omega $, $ \partial R = \Gamma_1 \cup \Gamma_2 $, a decomposition 

898: into parallel components satisfying $ \Gamma_2 \subset \partial \Omega $. 

899: Let $ \Delta $ be the Dirichlet or Neumann Laplacian on $\Omega $. Then

900: for {\em any} neighbourhood of $ \Gamma_1 $ in $ \Omega $, $ V $, there exists $ C $

901: such that

902: \begin{equation}

903: \label{eq:t1}

904: - \Delta u = \lambda u \ \Longrightarrow \  \int_V | u ( x ) |^2 dx 

905: \geq \frac1C \int_R | u ( x ) |^2 dx \,, 

906: \end{equation}

907: that is, no eigenfuction can concentrate in $ R $ and away from $ \Gamma_1$.

908: \end{thm}

909: 

910: \begin{proof}

911: Let us take $x,y$ as 

912: the coordinates on the stadium, so that $x$ parametrizes $ \Gamma_2 \subset \partial 

913: \Omega $ and $ y$ parametrizes $ \Gamma_1 $, then  

914: \[ R = [0,1]_{x}\times [0,a]_{y} \,.\]

915: Let $\chi\in \CIc((0,1))$ be equal to $1$ on $[\varepsilon, 1- \varepsilon]$. Then $\chi(x) u(x,y)$ is a solution of

916: \begin{equation}

917: (\Delta-z)\chi u = [\Delta, \chi] u\text{ in $R$}

918: \end{equation} with the boundary conditions satisfied 

919: on $\partial R$. Applying Proposition~\ref{p:1}, we get 

920: \begin{equation}

921: \|\chi u\|_{L^2( R)}\leq C \left \| [ \Delta, \chi] u \|_{H^{-1}_{x}; L^2_{y}}+ 

922: \| u\rest _{\omega_{\varepsilon}}\|_{L^2( \omega_{\varepsilon})}\right)

923: \leq C' \| u\rest _{\omega_{\varepsilon}}\|_{L^2( \omega_{\varepsilon})} \,,

924: \end{equation}

925: where $\omega_{\varepsilon}$ is a neighbourhood of the support of $ \nabla\chi$.

926: Since a neighbourhood of $ \Gamma_1 $ in $ \Omega $ has to contain $ \omega_\varepsilon $

927: for some $ \varepsilon $, \eqref{eq:t1} follows.

928: \end{proof}

929: 

930: 

931: \section{Applications}

932: \label{app}

933: 

934: In \cite{BZ2} and \cite{BZ3}, Proposition \ref{p:1} is used to prove that in the

935: case of the Bunimovich billiard shown in Fig.\ref{fig:bath},
the states have nonvanishing 

936: density near the vertical boundaries of the rectangle. That follows from Theorem

937: \ref{t:1} which shows that we must have positive density in the wings of the

938: billiard, and the propagation result (in the boundary case) based on the fact that

939: any diagonal controls a disc geometrically (see  \cite[Section 6.1]{BZ2}; in fact 

940: we can use other control regions as shown in Fig.\ref{fig:bath}). 

941: Here we consider another case which accidentally generalizes a control theory 

942: result of Jaffard \cite{Ja}.

943: 

944: 

945: The Sinai billiard (see Fig.\ref{fig:sin}) 

946: is defined by removing a strictly convex open set, $ \mathcal O$, with 

947: a $ \CI $ boundary,

948:  from a flat torus, $ {\mathbb T}^2_{a,b} \stackrel{\rm{def}}{=}

949:  (a \SP^1 ) \times (b \SP^1 )$:

950: \[ S  \stackrel{\rm{def}}{=} {\mathbb T}^2_{a,b} \setminus {\mathcal O} \,.\]

951: The following theorem results by applying Theorem 1 to a torus with sides of 

952: arbitrary length.

953: 

954: \begin{figure}

955: \include{figure2}

956: \caption{Control regions in which eigenfunctions have positive mass and the

957: rectangular part for the Bunimovich stadium.}

958: \label{fig:bath}

959: \end{figure} 

960: 

961: 

962: \begin{thm}

963: \label{t:s}

964: Let $ V $ be any open neighbourhood of the convex boundary, $ \partial {\mathcal O} $, in 

965: a Sinai billiard, $ S $. If $ \Delta $ is the Dirichlet or Neumann

966: Laplace operator on $ S $ then there exists a constant, $ C = C ( V ) $, such that

967: \begin{equation}

968: \label{eq:ts}

969: - h^2 \Delta u = E(h) u \ \Longrightarrow \  \int_V | u ( x ) |^2 dx 

970: \geq \frac1C \int_S | u ( x ) |^2 dx \,, 

971: \end{equation}

972: for any $h$ and $|E(h) - 1| < \frac{1}{2}$.

973: \end{thm}

974: \begin{proof}

975: First note that we can easily limit ourselves to the case where our flat torus has

976: one side of length 1 and one side of length a.  Suppose that the result is not true, 

977: in other words, there exists a sequence of 

978: eigenfunctions $ u_n $, $ \| u _n \| = 1 $, with the corresponding 

979: eigenvalues $ \lambda_n \rightarrow \infty $,  such that $ \int_V | u_n ( x )|^2 dx

980: \rightarrow 0 $. 

981: 

982: We first observe that the only directions in the support of 

983: the corresponding semi-classical defect measure, $\mu$, have to be "rational", 

984: in other words, the trajectory must travel along a line of slope $\frac{m a}{n}$ where $m,n \in \NN$.  

985: The projection of a trajectory with an irrational direction is dense on the 

986: torus and hence must encounter the obstacle $ \partial {\mathcal O} $ (and 

987: consequently $V$). The propagation result recalled in Proposition \ref{prop2.1}, part (3),

988: gives a contradiction by choosing a proper test function $\phi$ which is nonzero on the support

989: of the measure $\mu$ resulting from our sequence of eigenfunctions (remark that we apply this result as long as the 

990: trajectory does not encounter the obstacle and consequently we need only the 

991: {\em interior} propagation).

992: 

993: Hence let us assume that there exists a rational direction in the support of 

994: the measure which then contains the periodic trajectory in that direction.  

995: As shown in Fig. \ref{f:3} we can find a maximal rectangular neighbourhood of 

996: the projection of that trajectory which avoids the obstacle.

997: 

998: \begin{figure}

999: \include{figure3}

1000: \caption{A maximal rectangle in a rational direction, avoiding the obstacle.

1001: Because the parrallelogram is certainly periodic and our region has uniform width,

1002: it is clear that the resulting rectangle is periodic.}

1003: \label{f:3}

1004: \end{figure} 

1005: 

1006: The rectangle can be described as $ R = [0,a_1]_{x_1} \times [0, b_1]_{y_1} $ with the the 

1007: $ y_1 $ coordinate parametrizing the trajectory. 

1008: Let $\epsilon$, $\delta >0$ be small.

1009: Let $ u $ be an eigenfunction in our sequence and define 

1010: $$ \chi = \{ \chi ( x_1 ) \in \CIc ({\mathbb T}^2) \ | \ \chi (x_1) = 1 \ \text{for all} \ x_1 \in (\epsilon,a_1-\epsilon), 

1011: \ \chi(x_1) = 0 \ \text{outside $(0,a_1)$} \}.$$  

1012: Note that we can then write $\chi = \chi(x,y)$ for $(x,y) \in {\mathbb T}^2_{1,a}$ as $x_1$ is simply a rotation and translation of 

1013: the standard coordinates.

1014: Then $ \chi ( x,y ) u ( x, y ) $ is a function on all of $ R $ satisfying the periodicity condition. 

1015: Let  $\Phi_\xi ( \nu ) = \Phi(\xi - \nu)$, where we define

1016: $$\Phi = \{ \Phi (\xi) \in \CIc ({\mathbb R}^2) \ | \ \Phi (\xi) = 1 \ \text{for} \ \xi \in B(0,\delta), 

1017: \ \Phi (\xi) = 0 \ \text{for} \ \xi \in {\mathbb R}^2 \setminus B(0,2 \delta) \}, $$ 

1018: where $B(0,\delta)$ is a ball centered at $0$ of radius $\delta$.

1019: Note, due to the compact support of this function in $\xi$,

1020: $\Phi_\xi (D)$ is in the symbol class $S(\langle \xi \rangle^{-N})$ for any $N$.

1021: Let $ \Delta_R $ is the (periodic) Laplacian on $ R $. Using Fourier decomposition

1022: we can arrange that $ [ \Delta_R , \Phi_\xi (D) ] = 0 $.  Since our eigenfunction is only defined on 

1023: ${\mathbb T}^2_{1,a} \setminus {\mathcal O}$, let us introduce a smooth function ${\chi}_0$ which is $0$ on a neighborhood $U$
of

1024: the obstacle and $1$ on ${\mathbb T}^2_{1,a} \setminus V$ where $U \subset V$.
 Choose $U$ and $V$ such that $\chi \chi_0 = \chi$.

1025: Hence, 

1026: \[  ( -h^2 \Delta_R - E(h) ) \Phi_\xi(D) \chi {\chi}_0 u = [ -h^2 \Delta_R , \Phi_\xi(D) \chi] u 

1027: =\Phi_\xi (D) [ -h^2 \Delta_R , \chi] \chi_0 u + {\mathcal O} ( h^{ \infty } ) 

1028: \,, \ \  \| u \| = 1 \,, 

1029: \]

1030: by the properties of $S(\langle \xi \rangle^{-N})$ operators acting on $L^2$ functions on a torus.

1031: As in the proof of Proposition \ref{p:1}, we now see that 

1032: \begin{equation}

1033: \label{eq:mic}

1034:  \| \Phi_\xi \chi {\chi}_0 u \|_{L^2} \leq C \int_\omega | \chi_0 u |^2 

1035: + {\mathcal O} ( h^{\infty } ) \,,

1036: \end{equation}

1037: where $ \omega $ is a neighbourhood of $ \nabla \chi $ (in

1038: the calculus of semi-classical pseudo-differential operators). 

1039: Since the semi-classical defect measure of $ \Phi_\xi \chi {\chi}_0 u $ (which is $|\Phi_\xi \chi {\chi}_0|^2\times \mu$) was assumed to be

1040: non-zero, \eqref{eq:mic} shows that the measure of $ {\chi}_0 u \rest_\omega $

1041: is non zero and consequently there is a point in the intersection of the supports of 

1042: $\mu$ and ${\chi}_0 u \rest_\omega$. But $\mu$ is invariant

1043: by the flow (as long as it does not intersect the obstacle) and hence, once we choose $\epsilon$, $\delta$ small enough 

1044: such that all the cut-offs above are very close to the

1045: boundary of $ R$, its support can be made intersect any neighbourhood of $ \partial {\mathcal O} $.

1046: \end{proof}

1047: 

1048: Now, from the above theorem, we see the following simple, but important consequences:

1049: 

1050: \begin{rem}

1051: Let $S= {\mathbb T}^2_{a,b} \setminus {\mathcal O}$ where $ {\mathcal O}$ is sufficiently smooth in the case

1052: of Neumann boundary conditions, but otherwise lacking restrictions.  Then, for $V$ any open neighborhood of 

1053: $\partial {\mathcal O}$, and $u$ a solution of $-h^2 \Delta u = E(h) u$ as above, then (4.1) is satisfied.  

1054: This follows from the above argument as the convexity of the obstacle was never used.

1055: Thus, the result holds for any obstacle (even connectedness is not assumed here)

1056: and is applicable to the special case of pseudointegrable billiards (see for instance

1057: \cite{Bog} for motivation and description).  In the next section, we use an argument similar to that above 

1058: in order to say even more about concentration along trajectories in specific pseudointegrable billiards. 

1059: By an elementary reflection principle, the result also holds for an obstacle inside a square with Dirichlet 

1060: or Neumann conditions on the boundary of the square.

1061: \end{rem}

1062: 

1063: \begin{rem}

1064: The proof above gives in fact the following estimate for any open neighbourhood, say $V$, of the obstacle:

1065: \begin{equation}

1066: \begin{gathered}

1067: \exists C; \forall u,\, f \in L^2(S) \text{ solutions of } (-\Delta + \lambda)u =f, \qquad u \rest_{\partial S} =0\\

1068: \|u\|_{L^2(S)}\leq C \left( \|f\|_{L^2(S)}+ \|u \11 _V\|_{L^2(V)}\right)

1069: \end{gathered} 

1070: \end{equation}

1071: and according to~\cite[Theorem 4]{BZ2}, this implies that the Schr{\"o}dinger equation in $S$ 

1072: is exactly controllable by $V$ in finite time. In fact, by working on the time evolution equation, 

1073: we could strengthen this result allowing an arbitrarily small time. 

1074: \end{rem}

1075: This latter result was previously known~\cite{Ja} 

1076: for the particular case $\Theta= \emptyset$ ($S= \mathbb{T}^2$) but the proof was 

1077: based on subtle results about Fourier series \cite{Ka}.

1078: 

1079: \def\cprime{$'$}

1080: 

1081: \begin{rem}

1082: As shown in \cite[Theorem 2$'$]{BZ2}, the results of Ikawa and G\'erard on 

1083: scattering by two convex obstacles (see \cite{BZ2} and references given there)

1084: give an estimate on the maximal concentration of an eigenfunction (or a 

1085: quasimode) on a closed orbit in a Sinai billiard. Let $ \chi \in \CI(S;[0,1])  $  be 

1086: supported in  a small neighbourhood of a closed transversally reflecting 

1087: orbit. Then  for any family $ ( - \Delta - \lambda ) u_\lambda = {\mathcal O}

1088: ( \lambda^{-\infty } ) $, $ \| u_\lambda \| = 1 $, 

1089: \[ C \int_S  | u ( x ) |^2 ( 1 - \chi ( x ) ) dx \geq \frac{1}{ \log \lambda } \,, \]

1090: that is a concentration on a closed trajectory, if at all possible, has to 

1091: be very weak.

1092: \end{rem}

1093: 

1094: \section{Pseudointegrable Billiards}

1095: \label{PIB}

1096: 

1097: We define a pseudointegrable billiard to be a plane polygonal billiard with corners whose angles are of the form $\frac{\pi}{n}$, 

1098: for any integer $n$ (see \cite{Bog1}).  

1099: In particular, we will be working with the billiard $P = {\mathbb T}^2_{a,b} \setminus { S}$ where

1100: $ S$ is a slit that is parrallel to a side of the torus but not a closed loop.  In Remark 1, 

1101: we point out that Theorem \ref{t:s} allows us to make statements about the $L^2$ mass

1102: of eigenfunctions in a neighborhood of the slit for pseudointegrable billiards.  For this particular type of billiard,

1103: it would be ideal to state that every eigenfunction must have non-zero mass in a small neighborhood of the edges of the slit (see Fig. \ref{f:4}).

1104: In this section, we prove a weaker result about non-concentration along certain classical

1105: trajectories in $P$ of semiclassical defect measures obtained from eigenfunctions $u$ such that $(-\lambda -\Delta)u = 0$ on $P$.  

1106:  

1107: 

1108: 

1109: \begin{figure}

1110: \include{figure4}

1111: \caption{A pseudointegrable billiard $P$ consisting of a torus with a slit, $S$ along which we have Dirichlet boundary conditions.

1112: We would like to show that eigenfunctions of the Laplacian on this torus must have concentration in the shaded regions $V_1$ and $V_2$.}

1113: \label{f:4}

1114: \end{figure} 

1115: 

1116: 

1117: As with the Sinai billiard, the classical behavior of trajectories must be taken into account in our treatment of this problem.   

1118: There cannot be concentration along trajectories that do not hit the slit as shown by Theorem \ref{t:s}.  If a trajectory 

1119: has irrational slope, it is dense in $P$, and thus has mass near the edges of the slit as in Section \ref{app}.

1120: Therefore, for our purpose, we concern ourselves only with rational trajectories which intersect the slit at some point.  As we are dealing with

1121: periodic boundary conditions, let us consider the

1122: plane tiled with copies of the billiard $P$.  

1123: 

1124: Assume that $P$ is oriented such that $S$ is parrallel to the $y$-axis.  

1125: Let $\gamma \in S^* (P)$ be a trajectory.  Given the natural projection

1126: $$ \pi_1: S^* (P) \rightarrow P,$$

1127: we take $\gamma' = \pi_1 (\gamma)$, or the physical path along which the trajectory travels. 

1128: Consider the projection

1129: $$\tilde{\pi} : \R^2 \rightarrow {\mathbb T}^2_{a,b}.$$

1130: We see that 

1131: $$\tilde{\pi} : \R^2 \setminus \tilde{S} \rightarrow P, \ \text{where} \ \tilde{S} = \tilde{\pi}^{-1} (S).$$

1132: Define $$\pi_2 : S^* (\R^2 \setminus \tilde{S} ) \rightarrow S^* (P)$$

1133: to be the obvious projection.  Let $\tilde{\gamma} = \pi_2^{-1} (\gamma)$.

1134: We can write $$\tilde{\gamma} = \displaystyle\cup_{j=1}^{\infty} \gamma_j,$$ where each $\gamma_j$ is a trajectory 

1135: in $ S^* (\R^2 \setminus \tilde{S})$.  We note that by construction, $\gamma_i \cap \gamma_j = \emptyset$ for $ i \neq j$.  To see this, assume that 

1136: $\gamma_i \cap \gamma_j = (x, \xi)$.  Then, $\gamma_i = \gamma_j$ as they would represent trajectories which travel 

1137: through the same point in the same direction.  Now, let $$\pi^*_1 : S^* (\R^2 \setminus \tilde{S}) \rightarrow \R^2 \setminus \tilde{S}.$$

1138: Select one trajectory from the above union, say $\gamma_1$.  Let $\gamma_1' = \pi^*_1 (\gamma_1)$. 

1139: We see that either $\gamma_1'$ is bounded in the $x$-direction

1140: or $\gamma_1'$ is unbounded in the $x$-direction.  Note that this property then holds for all $\gamma_j$, $j \in {\mathbb N}$.

1141: For a trajectory $\gamma$, if the resulting path $\gamma_1'$ is bounded in the $x$-direction,

1142: we say $\gamma$ is $x$-bounded.  We define $\gamma$ as $x$-unbounded if $\gamma_1'$ is unbounded in the $x$-direction.

1143: See Fig. \ref{f:7} for examples.  Now, we are prepared to state our theorem concerning the billiard $P$.

1144: 

1145: \begin{figure}

1146: \include{figure7}

1147: \caption{Some examples of $x$-bounded and $x$-unbounded trajectories.}

1148: \label{f:7}

1149: \end{figure}

1150: 

1151:   

1152: 

1153: \begin{thm}

1154: \label{t:j}

1155: Let $\gamma$ be an $x$-bounded trajectory on $P= \mathbb{T}^2 \setminus S$.

1156: If $ \Delta $ is the Dirichlet  Laplace operator on $ P $ then 

1157: there exists no microlocal defect measure obtained from the eigenfunctions  on $P$ 

1158: such that ${\rm{supp}} \  (d\mu) = \gamma$.

1159: \end{thm}

1160: 

1161: \begin{proof}

1162: 

1163: Let  $\gamma '$  be as above.  Let $V_\epsilon$ be an $\epsilon$ neighborhood of $\gamma'$.  

1164: If the theorem were false, we would have a sequence of eigenfunctions $u_n$, $\| u_n \|_{L^2} =1$ with the 

1165: property

1166: $$\displaystyle\int_{P \setminus V_\epsilon} | u_n |^2 dx \rightarrow 0,$$ for any $\epsilon.$ We show that

1167: this is impossible.

1168: 

1169: For each $u_n$, we have $(-\Delta - \lambda_n) u_n = 0$, $u_n|_S =0$, $u_n \in L^2(P)$.  

1170: Let $\tilde{\pi}$ be as above.

1171: We define the sequence

1172: $\tilde{u}_n = \tilde{\pi}^{-1} u_n$.  We have $(-\Delta - \lambda_n ) \tilde{u}_n = 0$,

1173: $\tilde{u}_n|_{\tilde{S}} = 0$, and $\tilde{u}_n \in L^2_{\text{per}} (\R^2 \setminus \tilde{S})$.

1174: 

1175: 

1176: If $\pi_2 : S^* (\R^2 \setminus \tilde{S}) \rightarrow S^* (P)$ is as above and 

1177: $\tilde{\gamma} = \pi_2^{-1} (\gamma)$, then $\tilde{u}_n \rightarrow d\tilde{\mu}$ with 

1178: $$\text{supp} \ (d\tilde{\mu}) = \tilde{\gamma} \subset S^* (\R^2 \setminus \tilde{S})\,. $$

1179: 

1180: Now, let $\pi_1^* : S^* (\R^2 \setminus \tilde{S}) \rightarrow \R^2 \setminus \tilde{S}$ be as above.

1181: Select one trajectory, say $\gamma_1$.  As 

1182: $\gamma_1$ is $x$-bounded, ${\gamma}_1' = \pi_1^* ( \gamma_1 )$ is contained in a strip in the plane which is 

1183: infinite in the $y$-direction and bounded in the $x$-direction.

1184: Thus, $\gamma_1'$ is contained in a strip, $C_0$, with minimal width in the $x$-direction.  Then,

1185: $\tilde{u}_n$ satisfies $(-\Delta -\lambda_n) \tilde{u}_n = 0$ on $C_0$, is periodic in the $y$-direction, and

1186: satisfies the following boundary conditions in the $x$-direction:  Dirichlet boundary conditions 

1187: along the slits that intersect the boundary of $C_0$ and periodic boundary conditions otherwise.

1188: 

1189: Without loss of generality, we can choose the $x$-coordinates such that the 

1190: boundaries of $C_0$ are

1191: $x=-R$ and $x=0$.  We can then reflect to a strip, say $\tilde{C}_1$, with boundaries $x=-R$ and $x=R$, by defining a new function on $\tilde{C}_1$ by:

1192: $$ \tilde{u}_n^{(1)} (x,y) =  \left\{ 

1193: \begin{array}{ll}

1194:  \ \ \tilde{u}_n (x,y)  &  x \in [-R,0], \\

1195:  -\tilde{u}_n (-x,y)  & x \in (0,R). 

1196: \end{array} \right.$$

1197: Note that $\tilde{u}_n$ is periodic with period $2R$.

1198: As a result, we have

1199: $$(-\Delta - \lambda_n) \tilde{u}_n^{(1)} = f_n^{(1)}$$ on $\tilde{C}_1$, where

1200: $$f_n^{(1)} = 2 u(0,y) {\delta}_{0} ' (x) - 2 u(R,y) {\delta}_{R} ' (x).$$

1201: We note that $ f_n^{(1)} $ is supported away from the slits, $ \tilde S $.

1202: 

1203: Define 

1204: $$\pi^{\sharp}_1 (x,y) = \left\{

1205:  \begin{array}{ll}

1206:  (x,y) & -R \leq x \leq 0, \\

1207:   ( -x,y) & 0 \leq x \leq R.

1208:   \end{array} \right.$$

1209: If $ \pi_1^\sharp \; : \; \tilde{C}_1 \rightarrow C_0 $, then $$ (\pi_1^\sharp)^{-1} ( \bigcup_j \gamma_j' ) $$ is again a  union 

1210: of paths resulting from disjoint trajectories.  

1211: Now, we iterate this procedure a finite number of times, 

1212: stopping the iteration when the disjoint trajectories in the lift intersect each slit only once.

1213: 

1214: 

1215: After each reflection, we restrict to a new minimal width strip, say $C_i$.  Let us call $\tilde{C}_i$ the strip resulting from the $i$th reflection.

1216: We define $\pi^{\sharp}_i :  \tilde{C}_i \rightarrow C_{i-1}$ for $1 \leq i < N$ such that

1217: $$\pi^{\sharp}_i (x,y) = \left\{

1218:  \begin{array}{ll}

1219:  (x,y) & (x,y) \in C_{i-1}, \\

1220:   ( 2 R_i -x,y) & (x,y) \in C_{i-1} '.

1221:   \end{array} \right.$$

1222: Here, $C_{i-1}'$ is defined as the reflected strip and $x=  R_{i-1}$ is the line of reflection for $\tilde{C}_i$.  We can subsequently

1223: define $f_n^{(i)}$ as a sum of delta functions resulting from jumps that occur after reflection, similar to $f_n^{(1)}$ above.

1224: We also have $\pi^N : \R^2 \rightarrow C_N$, the obvious projection that results after we tile the plane with copies of $C_N$.  

1225: So, we have:

1226: $$\R^2 \stackrel{\pi_N}{\rightarrow} C_N \subset \tilde{C}_N \stackrel{\pi^{\sharp}_{N}}{\rightarrow} C_{N-1} \subset \tilde{C}_{N-1}

1227: \stackrel{\pi^{\sharp}_{N-1}}{\rightarrow} ... 

1228: \stackrel{\pi^{\sharp}_{2}}{\rightarrow} C_1 \subset \tilde{C}_1 \stackrel{\pi^{\sharp}_{1}}{\rightarrow} C.$$

1229: Note that 

1230: \[ \pi_N^{-1} (\gamma_1') = \displaystyle\bigcup_j \gamma_{1,j}', \]

1231: where $\{ \gamma_{1,j}' \}$ is the set of all paths in $C_N$ generated by the trajectory $\gamma_1$ and the periodicity in $y$.

1232: 

1233: 

1234: \begin{figure}

1235: \include{figure6}

1236: \caption{This diagram describes how we "unfold" the eigenfunctions in order to derive a contradiction.}

1237: \label{fig:unf}

1238: \end{figure} 

1239: 

1240: After a finite number of reflections, we "unfolded" ${\gamma}_1 '$ to be a periodic line on a large strip, $C_N$, which does not 

1241: intersect a slit anywhere.  Now, let us choose $\Phi_\xi$, $\chi$, and $\chi_0$ as above in order to cut-off microlocally 

1242: on this strip around $\gamma_1$.  Again, recall that we can set $\chi \chi_0 = \chi$.  As $f_n^{(i)}$ is supported only in between the slits

1243: for each $i \in \mathbb{N}$, $1 \leq i \leq N$, by

1244: choosing $\Phi_\xi$ to commute with the periodic Laplacian, we have

1245: \[  ( -h^2 \Delta_R - E(h) ) \Phi_\xi \chi {\chi}_0 u_n = \Phi_\xi \chi f_n + [-h^2 \Delta_R , \Phi_\xi \chi ]  u 

1246: =\Phi_\xi  [ -h^2 \Delta_R , \chi ] {\chi}_0 u + {\mathcal O} ( h^{ \infty } ) 

1247: \,, \ \  \| u \| = 1 \,. 

1248: \]

1249: Thus, the result follows by contradiction from the proof of Theorem \ref{t:s}.  

1250: \end{proof}

1251: 

1252: \begin{rem}

1253: Though this result only shows non-concentration, the proof of Theorem \ref{t:s} can be used to show if $\gamma$ is an $x$-bounded trajectory and

1254: $u$ is an eigenfunction supported on $\gamma ' = \pi_1 (\gamma)$, then in fact there must be mass at the edges of the slits as desired.

1255: \end{rem}

1256: 

1257: \begin{rem}

1258: If instead of a torus, we had Dirichlet boundary conditions on the boundary of the rectangle as well as the slit, then 

1259: this non-concentration result can also be applied by an elementary reflection principle argument.  

1260: \end{rem}

1261: 

1262: \begin{thebibliography}{10}

1263: 

1264: %\bibitem{BS}

1265: %A. B\"acker, R. Schubert, and P. Stifter. 

1266: %\newblock On the number of bouncing ball modes in billiards.

1267: %\newblock {\em J. Phys. A: Math. Gen.} 30:6783-6795, 1997.

1268: 

1269: \bibitem{BLR} 

1270: C. Bardos, G. Lebeau and J. Rauch.

1271: \newblock 

1272: Sharp sufficient conditions for the observation, control, and stabilization of waves from the boundary.

1273: \newblock {\em SIAM J. Control Optim.} 30:1024--1065, 1992.

1274: 

1275: \bibitem{Bog}

1276: E. Bogomolny, U. Gerland, and C. Schmit. { Models of intermediate spectral statistics,} 

1277: {\em Phys. Rev. E} 59:1315-1318, 1999.

1278: 

1279: \bibitem{Bog1}

1280: E. Bogomolny and C. Schmit. { Structure of Wave Functions of Pseudointegrable Billiards,} 

1281: {\em Phys. Rev. Lett.} 92:244102, 2004.

1282: 

1283: \bibitem{Bu92}

1284: N.~Burq.

1285: \newblock

1286: Control for Schrodinger equations on product manifolds.

1287: \newblock{ \em Unpublished}, 1992

1288: 

1289: \bibitem{BuIMRN}

1290: N.~Burq.

1291: \newblock Semi-classical estimates for the resolvent in non trapping

1292:   geometries.

1293: \newblock {\em Int. Math. Res. Notices}, 5:221--241, 2002.

1294: 

1295: \bibitem{BuGe}

1296:          {N. Burq and P. G{\'e}rard},

1297: \newblock {Condition n{\'e}cessaire et suffisante pour la

1298:        contr{\^o}labilit{\'e} exacte des ondes}.

1299: \newblock{\em Comptes Rendus de L'Acad{{\'e}}mie des Sciences}, {749--752},{t.325, S{\'e}rie I}, 1996

1300: 

1301: 

1302: 

1303: \bibitem{BZ2}

1304: N.~Burq and M. Zworski.

1305: \newblock Geometric control in the presence of a black box.

1306: \newblock {\em JAMS}, to appear.

1307: 

1308: 

1309: \bibitem{BZ3}

1310: N.~Burq and M. Zworski.

1311: \newblock Bouncing ball modes and quantum chaos.

1312: \newblock {\em SIAM Review}, to appear.

1313: 

1314: 

1315: \bibitem{BuLe01}

1316: {N. Burq and G.Lebeau}.

1317: \newblock {Mesures de d\'efaut de compacit\'e, application au syst\`eme

1318:               de {L}am\'e}.

1319: \newblock{\em Ann. Sci. \'Ecole Norm. Sup. (4)}, No {34}, 817-870, 2001.

1320: 

1321: \bibitem{Do}

1322: H.~Donnelly.

1323: \newblock Quantum unique ergodicity.

1324: \newblock {\em Proc. Amer. Math. Soc.} 131:2945-2951, 2003.

1325: 

1326: \bibitem{Zw}

1327: L. C. Evans and M. Zworski.

1328: \newblock

1329: Lectures on Semiclassical Analysis.

1330: \newblock {\em Unpublished Lecture Series}, 2003.

1331: 

1332: \bibitem{GeLe93}

1333: {P. G{\'e}rard and E. Leichtnam}.

1334: \newblock {Ergodic Properties of Eigenfunctions for the {D}irichlet

1335:                   Problem}.

1336: \newblock{\em Duke Mathematical Journal,} No 71, 559--607, 1993

1337: 

1338: \bibitem{Ha}

1339: A. Haraux.

1340: \newblock S\'eries lacunaires et contr\^ole semi-interne des vibrations

1341:               d'une plaque rectangulaire.

1342: \newblock{\em J. Math. Pures Appl.} 68-4:457--465, 1989.

1343: 

1344: %\bibitem{Hor} L. H{\"o}rmander.

1345: %\newblock

1346: %The Analysis of Linear Partial Differential Operators. Vol. III, IV. 

1347: %\newblock Springer-Verlag, Berlin, 1985.

1348: 

1349: \bibitem{Ja}

1350: S. Jaffard.

1351: \newblock Contr\^ole interne exact des vibrations d'une plaque rectangulaire. 

1352: \newblock {\em Portugal. Math.} 47 (1990), no. 4, 423-429.

1353: 

1354: \bibitem{Ka} 

1355: J.P. Kahane.

1356: \newblock

1357: Pseudo-p\'eriodicit\'e et s\'eries de Fourier lacunaires.

1358: \newblock {\em Annales Sc. de l'Ecole Normale Sup\'erieure} 79, 1962.

1359: 

1360: 

1361: \bibitem{MeSj}

1362:  {R.B. Melrose and J. Sj{\"o}strand}.

1363: \newblock {Singularities of Boundary Value Problems {I \& II}},

1364: \newblock {\em Communications in Pure Applied Mathematics}, {31 \& 35}, {593- 617 \& 129-168}, {1978 \& 1982}.

1365: 

1366: 

1367: %\bibitem{Mi03}

1368: %L.~Miller. 

1369: %\newblock 

1370: %How violent are fast controls for {S}chr\"odinger

1371: %equation?

1372: %\newblock {\em preprint}, 2003.

1373: 

1374: \bibitem{Ze}

1375: S. Zelditch.

1376: \newblock 

1377: Quantum unique ergodicity.

1378: \newblock

1379: {\tt math-ph/0301035}

1380: 

1381: \end{thebibliography}

1382: \end{document}

1383: