1: \documentclass[12pt]{amsart}

2: \usepackage[T1]{fontenc}

3: \usepackage{amssymb}

4: \usepackage{a4wide}

5: \usepackage{graphicx}

6: \usepackage{verbatim}

7: \usepackage{amsmath}

8: \usepackage{version}

9: \usepackage{yhmath}

10: 

11: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Textclass specific LaTeX commands.

12:  \theoremstyle{plain}    

13:  \newtheorem{thm}{Theorem}%%[section]

14:  \numberwithin{equation}{section} %% Comment out for sequentially-numbered

15: %\numberwithin{figure}{section} %% Comment out for sequentially-numbered

16:  \theoremstyle{plain}

17:  \newtheorem{prop}{Proposition} %%Delete [thm] to re-start numbering

18:  \theoremstyle{plain}    

19:  \newtheorem{lem}{Lemma} %%Delete [thm] to re-start numbering

20:  \theoremstyle{plain}    

21:  \newtheorem{cor}{Corollary} %%Delete [thm] to re-start numbering

22:  \theoremstyle{plain}    

23:  \newtheorem{conj}{Conjecture} %%Delete [thm] to re-start numbering

24:  \theoremstyle{plain}    

25:  \newtheorem{question}{Question} %%Delete [thm] to re-start numbering

26: \theoremstyle{definition}

27: \newtheorem{rem}{Remark} %%Delete [thm] to re-start numbering

28: \theoremstyle{definition}

29: \newtheorem{Def}{Axioms} %%Delete [thm] to re-start numbering

30: 

31: 

32: 

33: \begin{document}

34: \input{def.tex}

35: 

36: \title%[Resonant eigenstates in quantum chaotic scattering ]

37: {Resonant eigenstates for a quantized chaotic system}

38: 

39: \author[S. Nonnenmacher]{St\'ephane Nonnenmacher}

40: \author[Mathieu Rubin]{Mathieu Rubin}

41: \address{Service de Physique Th\'eorique, 

42: CEA/DSM/PhT, Unit\'e de recherche associ\'e CNRS,

43: CEA/Saclay,

44: 91191 Gif-sur-Yvette, France}

45: \email{snonnenmacher@cea.fr}

46: 

47: \begin{abstract}

48: We study the spectrum of quantized open maps, 

49: as a model for the resonance spectrum of quantum

50: scattering systems. We are particularly interested in open

51: maps admitting a fractal repeller. Using the ``open baker's map'' 

52: as an example, we numerically

53: investigate the exponent appearing in the Fractal Weyl law

54: for the density of resonances; we show that this exponent is not 

55: related with the ``information dimension'', but rather the Hausdorff

56: dimension of the repeller. We then 

57: consider the {\em semiclassical measures} associated with 

58: the eigenstates: we prove that these measures are

59: conditionally invariant with respect to the classical dynamics.

60: We then address the problem

61: of classifying semiclassical measures among conditionally invariant ones. 

62: For a solvable model, the ``Walsh-quantized'' open baker's map, we manage to

63: exhibit a family of semiclassical measures with simple self-similar properties.

64: \end{abstract}

65: 

66: 

67: \maketitle

68: 

69: %%%%%%%%%%%%%%%%%%%%%%

70: %%%%%%%%%%%%%%%%%%%%%%

71: \section{Introduction}

72: %%%%%%%%%%%%%%%%%%%%%%

73: %%%%%%%%%%%%%%%%%%%%%%

74: 

75: %%%%%%%%%%%%%%%%%%%%%%

76: \subsection{Quantum scattering on $\IR^D$ and resonances} 

77: %%%%%%%%%%%%%%%%%%%%%%

78: 

79: In a typical scattering system, particles of positive energy

80: come from infinity, interact with a localized potential $V(q)$, and then leave

81: to infinity. The corresponding  quantum Hamiltonian 

82: $H_\hbar=-\hbar^2\Delta +V(q)$ has an

83: absolutely continuous spectrum on the positive axis.

84: Yet, the Green's function $G(z;q',q)=\la q'|(H_\hbar-z)^{-1}|q\ra$ admits

85: a meromorphic continuation from the upper half plane $\set{\Im z>0}$ to (some

86: part of) the lower

87: half-plane $\set{\Im z<0}$. This continuation generally has 

88: {\it poles} $z_j=E_j-\i\Gamma_j/2$, $\Gamma_j>0$,

89: which are called {\it resonances} of the scattering system.

90: 

91: The probability density of the corresponding ``eigenfunction'' $\varphi_j(q)$

92: decays in time like $\e^{-t \Gamma_j/\hbar}$, so physically $\varphi_j$ represents

93: a metastable state with decay rate $\Gamma_j/\hbar$, or lifetime $\tau_j=\hbar/\Gamma_j$.

94: In the semiclassical limit $\hbar\to 0$, we will call ``long-living''

95: the resonances $z_j$ such that $\Gamma_j=\cO(\hbar)$, equivalently with lifetimes 

96: bounded away from zero.

97: 

98: The eigenfunction $\varphi_j(q)$ is meaningful only

99: near the interaction region, while its behaviour outside that region

100: (exponentially increasing outgoing waves) is clearly unphysical.

101: As a result, one practical method to compute resonances (at least approximately)

102: consists in adding a {\it smooth absorbing potential} $-\i W(q)$ to the Hamiltonian

103: $\hat H$,

104: thereby obtaining a nonselfadjoint operator $H_{W,\hbar}=H_\hbar-\i W(q)$.

105: The potential $W(q)$ is supposed to vanish

106: in the interaction region, but is positive outside: its effect is to {\em absorb}

107: outgoing waves, as opposed to a real positive potential which would reflect

108: the waves back into the interaction region. Equivalently, the (nonunitary)

109: propagator $\e^{-\i H_{W,\hbar}/\hbar}$ kills wavepackets localized oustide the

110: interaction region. 

111: 

112: The spectrum of $H_{W,\hbar}$ in some neighbourhood of the

113: positive axis is then made of discrete eigenvalues $\tilde{z}_j$ associated with 

114: square-integrable eigenfunctions $\tilde\varphi_j$. 

115: Absorbing Hamiltonians of the type of $H_{W,\hbar}$ have been widely used in

116: quantum chemistry to study reaction or dissociation dynamics

117: \cite{LeforWyatt83,SeidMill92}; in those works it is implicitly assumed that eigenvalues

118: $\tilde{z}_j$ close to the real axis are small perturbations of the resonances

119: $z_j$, and that the corresponding eigenfunctions $\varphi_j(q)$, $\tilde\varphi_j(q)$ 

120: are close to one another in the interaction region.

121: Very close to the real axis (namely, for $|\Im\tilde{z}_j|=\cO(\hbar^n)$

122: with $n$ sufficiently large), one can prove that this is indeed the case \cite{Stef05}.

123: Such very long-living resonances are possible when the classical dynamics admits a

124: trapped region of positive Liouville volume.

125: In that case, resonances and the associated eigenfunctions can be approximated

126: by quasimodes of an associated closed system \cite{TangZw98}. 

127: 

128: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

129: \subsection{Resonances in chaotic scattering} 

130: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

131: 

132: We will be interested in a different situation, where the set of 

133: trapped trajectories has volume zero, and is a fractal hyperbolic repeller. 

134: This case encompasses the

135: famous 3-disk scatterer in 2 dimensions \cite{GaspRice89}, or its smoothing,

136: namely the 3-bump

137: potential introduced in \cite{Sjo90} and numerically studied in \cite{Lin02}.

138: Resonances

139: then lie deeper below the real line (typically, $\Gamma_j\gtrsim \hbar$),

140: and are not perturbations of an associated real spectrum. 

141: Previous studies have focussed in counting the number of resonances in small

142: disks around the energy $E$, in the semiclassical r\'egime. Based on the

143: seminal work of Sj\"ostrand \cite{Sjo90}, several authors have conjectured

144: the following Weyl-type law:

145: \bequ\label{e:Weyl-law}

146: \#\set{z_j\in \Res(H_\hbar)\,:\,|E-z_j|\leq \gamma\hbar}\sim C(\gamma)\hbar^{-d}\,.

147: \end{equation}

148: Here the exponent $d$ is related to the trapped set at energy $E$: the latter has

149: (Minkowski) dimension $2d+1$. 

150: This asymptotics was numerically checked by Lin and Zworski for

151: the 3-bump potential

152: \cite{Lin02,LSZ03}, and by 

153: Guillop\'e, Lin and Zworski for scattering on a hyperbolic surface \cite{GLZ04}. 

154: However, only the upper bounds for the number

155: of resonances could be rigorously proven \cite{Sjo90,Zw99,GLZ04,SjoZw05}. 

156: 

157: To avoid the

158: complexity of ``realistic'' scattering systems,

159: one can study simpler models, namely quantized open maps on a compact phase space, for instance

160: the quantized open baker's map studied in \cite{NonZw05,NonZw06,KNPS06} (see \S\ref{e:open-map}). 

161: Such a model is meant to mimick

162: the propagator of the nonselfadjoint Hamiltonian $H_{W,\hbar}$, in the case where the

163: classical flow at energy $E$ is chaotic in the interacting region. The above

164: fractal Weyl law has a direct counterpart in this setting; such a fractal scaling

165: was checked for the open kicked rotator in \cite{SchoTwo04}, 

166: and for the ``symmetric'' baker's map in \cite{NonZw05}.

167: In \S\ref{s:infodim?} we numerically check this fractal law for an 

168: asymmetric version of the open baker's map; apart from

169: extending the results of \cite{NonZw05}, this model allows to

170: specify more precisely the dimension $d$ appearing in the scaling law. 

171: To our knowledge, so far the only

172: system for which the asymptotics corresponding to \eqref{e:Weyl-law} 

173: could be rigorously proven is the ``Walsh quantization'' of 

174: the symmetric open baker's map \cite{NonZw06}, which will be described 

175: in \S\ref{s:walsh}.

176: 

177: After counting resonances, the next step consists in studying the 

178: long-living resonant eigenstates $\varphi_j$ or $\tilde\varphi_j$.

179: Some results on this matter 

180: have been announced by M.Zworski and the first author in \cite{NonZw-gap}.

181: Interesting numerics were performed by M.~Lebental and coworkers for a 

182: model of open stadium billiard, relevant to describe an experimental micro-laser cavity

183: \cite{Leben06}. In the framework of quantum open maps, eigenstates of the open Chirikov

184: map have been numerically studied by Casati {\it et al.} \cite{CasMasShep99}; the authors

185: showed that, in the semiclassical

186: limit, the long-living eigenstates concentrate on the classical hyperbolic repeller. They

187: also found that the phase space structure of the eigenstates are very correlated with their 

188: decay rate. In this paper we will consider general ``quantizable'' open maps,

189: %mainly focus on a different map, namely the open baker, 

190: and formalize the above observations into rigorous statements. Our main result is

191: Theorem~\ref{thm:main} (see \S\ref{s:semiclass}), which 

192: shows that the {\it semiclassical measure} associated

193: to a sequence of quantum eigenstates is (up to subtleties due to discontinuities) 

194: necessary an {\em eigenmeasure} of the classical open map.

195: Such eigenmeasures are necessarily supported on the

196: the {\em backward trapped set}, which, in the case of a chaotic dynamics, 

197: is a fractal subset of the phase space: this motivated

198: the denomination of ``quantum fractal eigenstates'' used in \cite{CasMasShep99}.

199: Let us mention that eigenstates of quantized open maps have been studied 

200: in parallel by J.~Keating and coworkers \cite{KNPS06}.

201: Our theorem~\ref{thm:main} provides a rigorous version of

202: statements contained in their work.

203: 

204: Inspired by our experience with closed

205: chaotic systems, in \S\ref{s:abundance} we attempt to

206: classify

207: semiclassical measures among all possible eigenmeasures, in particular

208: for the open baker's map.

209: In the case of

210: the ``standard'' quantized open baker, the classification remains open.

211: In \S\ref{s:walsh} we consider a solvable model,

212: the Walsh quantization of the open baker's map, introduced in \cite{NonZw05,NonZw06}. 

213: For that model, one can explicitly construct some semiclassical measures and partially

214: answer the above questions. A further study of semiclassical measures for

215: the Walsh model will appear in a joint publication with J.~Keating, M.~Novaes and M.~Sieber.

216: 

217: \subsection*{Acknowledgments}

218: M.~Rubin thanks the Service de Physique Th\'eorique for hospitality 

219: in the spring 2005, during which this work was initiated. S.~Nonnenmacher

220: has been partially supported from the grant ANR-05-JCJC-0107-01

221: of the Agence Nationale de la Recherche.

222: Both authors are grateful to J.~Keating, M.~Novaes and M.~Sieber for

223: communicating their results concerning the Walsh-quantized baker before 

224: publication, and for interesting discussions. We also thank M.~Lebental

225: for sharing with us her preliminary results on the open stadium billiard,

226: and the anonymous referees for their stimulating comments.

227: 

228: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

229: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

230: \section{The open baker's map}

231: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

232: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

233: \subsection{Closed and open symplectic maps}\label{e:open-map} 

234: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

235: 

236: Although many of the results we will present deal with a particular family of maps on the

237: 2-torus, namely the family of baker's maps, we start by some general considerations on

238: open maps defined on a compact metric space $\hM$, equipped with a probability measure $\mu_L$. 

239: We borrow some ideas and notations from the 

240: recent review of Demers and Young \cite{DemYou06}. We start with a invertible map $\hT:\hM\to\hM$,

241: which we assume to be piecewise smooth, and to preserve the measure $\mu_L$ (when

242: $\hM$ is a symplectic manifold, $\mu_L$ is the Liouville measure).

243: We then {\em dig a hole} in $\hM$, that is a certain subset $H\subset \hM$, 

244: and decide that points falling in the hole are no more iterated, but rather

245: ``disappear'' or ``go to infinity''. The hole is assumed to be a Borel subset of $\hM$.

246: 

247: Taking $M\defeq \hM\setminus H$, we are thus lead to consider 

248: the {\em open map} $T=\hT_{|M}:M\to \hM$, or equivalently, say that

249: $T$ sends points in the hole to infinity. By 

250: iterating $T$, we see that any point $x\in M$ has a certain time of escape $n(x)$, which is the

251: smallest integer such that $\hT^n(x)\in H$ (this time can be 

252: infinite). For each $n\in\IN^*$ we call 

253: \bequ

254: \label{e:M^n}

255: M^n=\set{x\in M,\ n(x)\geq n}=\bigcap_{j=0}^{n-1} \hT^{-j}(M)\,.

256: \eequ

257: This is the domain of definition of

258: the iterated map $T^n$. The {\em forward trapped set} for the open map $T$ is

259: made of the points which will never escape in the future:

260: \bequ\label{e:forward-TS}

261: \Gamma_- = \bigcap_{n\geq 1}M^n = \bigcap_{j= 0}^\infty \hT^{-j}(M)\,.

262: \eequ

263: These definitions allow us to split

264: the full phase space into a disjoint union

265: \bequ\label{e:split}

266: \hM=\big(\bigsqcup_{n=1}^\infty M_n\big)\sqcup \Gamma_-,

267: \eequ

268: where $M_1\defeq H$, and for each $n\geq 2$, $M_n\defeq M^{n-1}\setminus M^{n}$ 

269: is the set of points 

270: escaping exactly at time $n$.

271: 

272: We also consider the backward evolution given by the inverse map $\hT^{-1}$. 

273: The hole for this backwards 

274: map is the set $H^{-1}=\hT(H)$, and we call

275: $T^{-1}$ the restriction of $\hT^{-1}$ to $M^{-1}=\hT(M)$ 

276: (the ``backwards open map''). We also define $M^{-n}=\bigcap_{j=1}^n \hT^{j}(M)$, 

277: $M_{-1}=H^{-1}$, $M_{-n}=M^{-n+1}\setminus M^{-n}$ ($n\geq 2$). This leads to the

278: the {\em backward trapped set} 

279: \bequ\label{e:backward-TS}

280: \Gamma_+ = \bigcap_{j=1}^\infty \hT^{j}(M)\,,\quad\text{and the {\em trapped set} }

281: K=\Gamma_-\cap \Gamma_+\,.

282: \eequ

283: We also have a ``backward partition'' of the phase space:

284: \bequ\label{e:split2}

285: \hM=\big(\bigsqcup_{n=1}^\infty M_{-n}\big)\sqcup \Gamma_+.

286: \eequ

287: The dynamics of the open map $T$ is interesting if 

288: the trapped sets are not empty. This is be the case for the 

289: open baker's map we will study more explicitly (see \S\ref{s:baker-trapped}

290: and Fig.~\ref{f:Cantor}).

291: 

292: \medskip

293: 

294: In order to quantize the maps $\hT$ and $T$, one needs further assumptions.

295: In general, $\hM$ is

296: a {\em symplectic manifold}, $\mu_L$ its Liouville measure, and $\hT$ a canonical

297: transformation on $\hM$. Also, we will assume that $T$ is sufficiently regular (see

298: \S\ref{s:quantum}).

299: 

300: Yet, one may also consider ``quantizations'' on more general phase spaces, 

301: like in the axiomatic framework of Marklof and O'Keefe \cite{MarOK05}. 

302: As we will explain in \S\ref{s:Lipschitz}, 

303: the ``Walsh quantization'' of the baker's map is easier to analyze if we consider it

304: as the quantization of an open map on a certain symbolic space, which

305: is not a symplectic manifold. By extension, we also call ``Liouville'' the 

306: measure $\mu_L$ for this case.

307: 

308: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

309: \subsection{The open baker's map and its symbolic dynamics}\label{s:open-baker} 

310: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

311: We present the closed and open maps which will be our central examples: the baker's

312: maps and their associated symbolic shifts.

313: 

314: \subsubsection{The closed baker}

315: The phase space of the baker's map is the 2-dimensional torus

316: $\hM=\t2\simeq [0,1)\times [0,1)$. A point on $\t2$ is described with

317: the coordinates $x=(q,p)$, which we call respectively position (horizontal) 

318: and momentum (vertical), to insist on the symplectic structure $dq\wedge dp$.

319: %%%%%%%%%%%%%%%

320: \begin{figure}[htbp]

321: \begin{center}

322: \includegraphics[width=13cm]{bak32bis.eps}

323: \caption{\label{f:3-baker} Sketch of the closed baker's map $\hB_{\vr}$, and its

324: open counterpart $B_{\vr}$, for the case $\vr=\vr_{sym}=(1/3,1/3,1/3)$. 

325: The three rectangles form a Markov partition.}

326: \end{center}

327: \end{figure}

328: %%%%%%%%%%%%%%%

329: We split $\t2$ into three vertical rectangles $R_i$ with widths $(r_0,r_1,r_2)\defeq \vr$

330: (such that $r_0+r_1+r_2=1$, with all $r_\ep>0$), and first define a

331: closed baker's map on $\t2$ (see Fig.~\ref{f:3-baker}):

332: \bequ\label{e:closed-baker}

333: (q,p)\mapsto \hB_{\vr}(q,p)\defeq (q',p')= \left\{  \begin{array}{ll}

334: \big(\frac{q}{r_0},\,  p\,r_0\big) & \text{if}\  0 \leq q < r_0 \\

335: \big(\frac{q-r_0}{r_1}, \,  p\,r_1+r_0 \big)& \text{if}\  r_0 \leq q < r_0+r_1 \\

336: \big(\frac{q-r_0-r_1}{r_2}, \,  p\,r_2+r_1+r_0\big)& \text{if}\  1-r_2 \leq q < 1  \,.

337: \end{array}\right. 

338: \eequ

339: This map is invertible and symplectic on $\t2$. 

340: It is discontinuous on the boundaries of the rectangles $R_i$ but smooth (actually, affine)

341: inside them.

342: We now recall how $\hB_{\vr}$ can be conjugated with a symbolic dynamics. 

343: We introduce the right shift $\hat\sigma$ on the {\em symbolic space} 

344: $\Sigma=\set{0,1,2}^{\IZ}$:

345: $$

346: \bep=\ldots\ep_{-2}\ep_{-1}\cdot\ep_0\ep_1\ldots\in \Sigma\longmapsto 

347: \hat\sigma(\bep)=\ldots\ep_{-2}\ep_{-1}\ep_0\cdot\ep_1\ldots\,.

348: $$

349: The symbolic space $\Sigma$ can be mapped to the 2-torus as follows: every

350: bi-infinite sequence $\bep\in\Sigma$ is mapped to the point $x=J_{\vr}(\bep)$ with coordinates

351: \bequ\label{e:conjugation}

352: q(\bep)=\sum_{k=0}^\infty r_{\ep_0}r_{\ep_1}\cdots r_{\ep_{k-1}}\,\alpha_{\ep_k}\,,\qquad

353: p(\bep)=\sum_{k=1}^\infty r_{\ep_{-1}}r_{\ep_{-2}}\cdots r_{\ep_{-k+1}}\,\alpha_{\ep_{-k}}\,,

354: \eequ

355: where we have set $\alpha_0=0$, $\alpha_1=r_0$, $\alpha_2=r_0+r_1$. 

356: The position coordinate (unstable direction) depends on symbols 

357: on the right of the comma, while the momentum coordinate (stable direction)

358: depends on symbols on the left.

359: The map $J_{\vr}:\Sigma\to\t2$ is

360: surjective but not injective, for instance the sequences $\ldots \ep_{n-1}\ep_n 000\ldots$ 

361: (with $\ep_n\neq 0$) and $\ldots \ep_{n-1}(\ep_n-1) 222\ldots$ have the same image on $\t2$. 

362: For this reason, it is convenient to restrict $J_{\vr}$ to the subset $\Sigma'\subset\Sigma$

363: obtained by removing from $\Sigma$ the sequences ending by $\ldots 2222 \ldots$ on the

364: left or the right.

365: $\Sigma'$ is invariant through the shift $\hat\sigma$. 

366: The map $J_{\vr|\Sigma'}:\Sigma'\to\t2$ is now bijective, and it conjugates

367: $\hat\sigma_{|\Sigma'}$ with the baker's map $\hB_{\vr}$ on $\t2$:

368: \bequ\label{e:shift-A}

369: \hB_{\vr} = J_{\vr|\Sigma'}\circ \hat\sigma\circ (J_{\vr|\Sigma'})^{-1}\,.

370: \eequ

371: Any finite sequence 

372: $\bep=\ep_{-m}\ldots\ep_{-1}\cdot\ep_{0}\ldots\ep_{n-1}$ represents a {\em cylinder} 

373: $[\bep]\subset\Sigma$, which consists in the sequences sharing the same symbols 

374: between indices $-m$ and $n-1$. 

375: We call $J_{\vr}([\bep])$

376: a {\em rectangle} on the torus (sometimes we will note it $[\bep]$).

377: This rectangle has sides

378: parallel to the two axes; it has width $r_{\ep_0}r_{\ep_1}\cdots r_{\ep_{n-1}}$

379: and height $r_{\ep_{-1}}\cdots r_{\ep_{-m}}$.

380: 

381: For any triple, the Liouville measure on $\t2$ is the push-forward through $J_{\vr}$ 

382: of a certain Bernoulli measure on

383: $\Sigma$, namely $\mu^\Sigma_L= \nu^{\Sigma_-}_{\vr}\times \nu^{\Sigma_+}_{\vr}$

384: (see \S\ref{s:Bernoulli}).

385: 

386: %=======================================

387: \subsubsection{The open baker}\label{s:baker-trapped}

388: %=======================================

389: 

390: We choose to take for the hole the middle rectangle $H=R_1=\set{r_0\leq q<1-r_2}$.

391: We thus obtain an ``open baker's map'' $B_{\vr}$ defined on $M=\t2\setminus H$:

392: $$

393: (q,p)\mapsto B_{\vr}(q,p)\defeq (q',p')= \left\{  \begin{array}{ll}

394: \big(\frac{q}{r_0},\,  p\,r_0 \big) & \text{if}\  0 \leq q < r_0 \\

395: \quad\infty& \text{if}\  r_0 \leq q < 1-r_2 \\

396: \big(\frac{q-1+r_2}{r_2}, \,  p\,r_2+1-r_2\big)& \text{if}\  1-r_2 \leq q < 1 \,.

397: \end{array}\right. 

398: $$ 

399: The ``inverse map'' $B_{\vr}^{-1}$

400: is defined on the set $M^{-1}=B_{\vr}(M)$, that is outside the backwards hole 

401: $H^{-1}=B_{\vr}(H)=\set{r_0\leq p<1-r_2}$ (see Fig.~\ref{f:3-baker}, right).

402: 

403: Due to the choice of the hole, this open map is still easy to analyze 

404: through symbolic dynamics:

405: the hole $R_1$ is the image through $J_{\vr}$ of the set $\set{\ep_0=1}\cap \Sigma'$.

406: Let us define as follows the {\it open shift}

407: $\sigma$ on $\Sigma$:

408: \bequ\label{e:shift-B}

409: \bep\in \Sigma%=\ldots\ep_{-2}\ep_{-1}\cdot\ep_0\ep_1\ldots%\stackrel{B_{\vr}}{\longmapsto} 

410: \longmapsto \sigma(\bep)\defeq 

411: \begin{cases}\infty\quad\mbox{if}\quad \ep_0=1\\

412: \hat\sigma(\bep)%\ldots\ep_{-2}\ep_{-1}\ep_0\cdot\ep_1\ldots=\hat\sigma(\bep)

413: \quad \mbox{if}\quad\ep_0\in\set{0,2}\,.

414: \end{cases}

415: \eequ

416: $J_{\vr|\Sigma'}$ conjugates the open shift $\sigma_{|\Sigma'}$

417: with $B_{\vr}$ as in \eqref{e:shift-A}. Similarly, the backwards open shift 

418: $\sigma^{-1}$, which kills the sequences

419: s.t. $\ep_{-1}=1$ and otherwise moves the comma to the left, is conjugated with

420: $B_{\vr}^{-1}$.

421: 

422: \medskip

423: 

424: These conjugations allow to easily characterize the various trapped sets of $B_{\vr}$.

425: On the symbolic space $\Sigma$, the forward (resp. backward) trapped set of the

426: open shift $\sigma$ is given by the sequences $\bep$ such that 

427: $\ep_{n}\in\set{0,2}$ for all $n\geq 0$ (resp. for all $n<0$). To obtain the

428: trapped sets for the baker's map, we restrict ourselves on $\Sigma'$ and conjugate

429: by $J_{\vr|\Sigma'}$. The image sets are given by the direct products 

430: $\Gamma_-=\cC_{\vr}\times [0,1)$, $\Gamma_+=[0,1)\times \cC_{\vr}$ 

431: and $K=\cC_{\vr}\times\cC_{\vr}$,

432: where $\cC_{\vr}$ is (up to a countable set) 

433: the Cantor set on $[0,1)$ adapted to the partition $\vr$ 

434: (see Fig.~\ref{f:Cantor}).

435: %%%%%%%%%%%%%%%%%%%%

436: \begin{figure}[htbp]

437: \begin{center}

438: \includegraphics[width=16cm]{cantor3bis.eps}

439: \caption{\label{f:Cantor} In black we approximate the forward (left), 

440: backward (center)

441: and joint (right) trapped sets

442: for the symmetric open baker $B_{\vr_{sym}}$. 

443: On the left (resp. centre), red/gray scales (from white to

444: dark) correspond to points escaping 

445: at successive times in the future (resp. past), that is to sets $M_{n}$

446: (resp. $M_{-n}$), for $n=1,2,3,4$.}

447: \end{center}

448: \end{figure}

449: %%%%%%%%%%%%%%%%%

450: 

451: For future use, we define the subset $\Sigma''\subset\Sigma$ by

452: \bequ\label{e:Sigma''}

453: \Sigma''\defeq \Sigma\setminus

454: \big(\set{\ldots 222\cdot 0\ep_1\ep_2\ldots}\cup\set{\ldots 222\cdot 2\ep_1\ep_2\ldots}\cup 

455: \set{\ldots \ep_{-2}\ep_{-1}\cdot \ep_0 222\ldots}\big)\,.

456: \eequ

457: We notice that $\Sigma''\varsupsetneq\Sigma'$, and that $J_{\vr}(\Sigma\setminus \Sigma'')$ is

458: a subset of the discontinuity set of the map $B_{\vr}$ (see \S\ref{s:quantum-baker2}

459: and Fig.~\ref{f:gamma+}). 

460: The map $J_{\vr}$ realizes a kind of {\it semiconjugacy} between $\sigma_{|\Sigma''}$ and $B$:

461: \bequ\label{e:semiconj}

462: \forall \bep\in \Sigma'',\qquad  J_{\vr}\circ\sigma(\bep)=B_{\vr}\circ J_{\vr}(\bep).

463: \eequ

464: Above it is understood that both sides are ``sent to infinity'' if $\ep_{0}=1$. 

465: 

466: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

467: \subsection{Eigenmeasures of open maps}\label{s:eigenmeasures0} 

468: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

469: 

470: Before defining eigenmeasures of

471: open maps, we briefly recall how invariant measures emerge in the study of the

472: quantized {\it closed} maps.

473: 

474: \subsubsection{Quantum ergodicity for closed chaotic systems}

475: \label{s:eigenstates-IM} 

476: 

477: The quantum-classical correspondence between a closed symplectic map

478: $\hT$ and its quantization $\hT_{N}$ (see \S\ref{s:quantum}) 

479: has one important consequence:

480: in the semiclassical limit $N\to\infty$, stationary states of the quantum system (that is,

481: eigenstates of $\hT_{N}$) should reflect the stationary properties of the

482: classical map, namely its invariant measures. 

483: To be more precise, 

484: to any sequence of eigenstates $(\psi_N)_{N\to \infty}$ of the quantum map, 

485: one can associate at least one {\bf semiclassical measure} (see \S\ref{s:semicl-measure}).

486: Omitting problems due to the discontinuities of $\hT$, the quantum-classical

487: correspondence implies that a semiclassical measure $\mu$ must be 

488: invariant w.r.to $\hT$:

489: \bequ\label{e:invariant}

490: \hT^*\,\mu=\mu\quad\Longleftrightarrow 

491: \quad\text{for any Borel set } S\subset \t2,\quad \mu(\hT^{-1}(S))=\mu(S)\,.

492: \eequ

493: $\mu$ is then also invariant w.r.to the inverse map $\hT^{-1}$.

494: 

495: If the map $\hT$ is ergodic w.r.to the Liouville measure $\mu_{L}$ on $\hM$, 

496: the quantum ergodicity theorem (or Schnirelman's theorem \cite{Schni74})

497: states that, for ``almost any'' sequence $(\psi_N)_{N\to \infty}$,

498: there is a unique associated semiclassical measure, which is $\mu_{L}$ itself.

499: 

500: Such a theorem was first proven for eigenstates of the Laplacian on compact

501: Riemannian manifolds with ergodic geodesic flow \cite{Zel87,CdV85}, then for more general 

502: Hamiltonians \cite{HelMarRob87}, billiards \cite{GerLei93,ZelZwo96}

503: and maps \cite{BouzDB96,Zel96}. Quantum ergodicity for piecewise smooth

504: maps was proven in a general setting in \cite{MarOK05}, and the particular case of the

505: baker's map was treated in \cite{DENW06}. Finally, in \cite{AN06} it was shown that

506: the (closed) Walsh-baker's map, seen as the quantization of the right shift $\hat\sigma$

507: on $(\Sigma,\mu_L)$, also satisfies quantum ergodicity with respect to $\mu_L$.

508: 

509: It is generally unknown whether there exist ``exceptional sequences'' of eigenstates, 

510: converging to a different invariant measure. 

511: The absence of such sequences is expressed by the

512: quantum unique ergodicity conjecture \cite{RudSar94}, which has been proven only for

513: systems with arithmetic properties \cite{Linden06,KurRud00}. This conjecture has 

514: been disproved for some specific systems enjoying

515: large spectral degeneracies at the quantum level, allowing for

516: sufficient freedom to build up partially localized eigenstates \cite{FNdB03,AN06,Kelmer05}. 

517: Some special eigenstates of the standard quantum baker with interesting multifractal

518: properties have been numerically identified \cite{MeenakLaksh04}, 

519: but their persistence in the semiclassical limit remains unclear.

520: 

521: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

522: \subsubsection{Eigenmeasures of open maps}\label{s:eigenmeasures} 

523: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

524: 

525: We now dig the hole $H=\hM\setminus M$, and consider the open map $T=\hT_{|M}$. 

526: Eigenmeasures (also called conditionally invariant measures) of maps ``with holes'' have been less

527: studied than their invariant counterparts. The recent article of Demers and Young \cite{DemYou06}

528: summarizes most of the properties of these measures, for maps enjoying various dynamical

529: properties. Below we describe some of these properties for general maps, before 

530: being more specific in the

531: case of the baker's map.

532: 

533: A probability measure $\mu$ on $\hM$ which is invariant

534: through $T$ {\em up to a multiplicative factor} will be called

535: an eigenmeasure of $T$:

536: \bequ\label{e:eigenmeasure}

537: T^*\,\mu=\Lambda_\mu\,\mu

538: \quad\Longleftrightarrow\quad 

539: \text{for any Borel set } S\subset\hM,\quad \mu(T^{-1}(S))=\Lambda_\mu\,\mu(S)

540: \,. 

541: \eequ

542: Here $\Lambda_\mu\in [0,1]$ (or rather $\gamma_\mu=-\log\Lambda_\mu$) 

543: is called the ``escape rate'' or ``decay rate'' of the eigenmeasure $\mu$, and is

544: given by $\Lambda_{\mu}=\mu(\cS)$. It corresponds

545: to the fact that a fraction of the particles in the support of $\mu$ 

546: escape at each step. Our definition slightly differs from the one in \cite{DemYou06}:

547: their measures are supported on $M$ and normalized there, while we choose the normalization

548: $\mu(\hM)=1$.

549: 

550: Here are some simple properties:

551: %%%%%%%%%%%%%%%

552: \begin{prop}

553: Let $\mu$ be an eigenmeasure of $T$ with decay rate $\Lambda_\mu$. \\

554: If $\Lambda_\mu=0$, $\mu$ is supported in the hole $H$. \\

555: If $\Lambda_\mu=1$, $\mu$ is supported in the trapped set $K$ (invariant measure).\\

556: If $0<\Lambda_\mu<1$, $\mu$ is supported on the set $\Gamma_+\setminus K$.

557: \end{prop}

558: %%%%%%%%%%%%%%%

559: 

560: %%%%%%%%%%%%%%%

561: \begin{figure}[htbp]

562: \begin{center}

563: \includegraphics[width=8cm]{cantor6bis.eps}

564: \caption{\label{f:Cantor6} Various components $\Gamma_+^{(n)}$

565: of the forward trapped set for the open baker $B_{\vr_{sym}}$. 

566: Various red/grey scales (from light to dark) correspond to $n=1,2,3,4$.}

567: \end{center}

568: \end{figure}

569: %%%%%%%%%%%%%%%

570: Following the decomposition \eqref{e:split}, the set $\Gamma_+\setminus K$ can be split into a 

571: disjoint union (see Fig.~\ref{f:Cantor6}):

572: $$

573: \Gamma_+\setminus K=\bigsqcup_{n\geq 1} \Gamma_+^{(n)},\quad\text{where}\qquad

574: \Gamma_+^{(n)}\defeq \Gamma_+\cap M_{n}=

575: T^{-n+1}\Gamma_+^{(1)},

576: \quad n\geq 1\,.

577: $$

578: Following \cite[Thm. 3.1]{DemYou06}, any

579: eigenmeasure can be constructed as follows.

580: %&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&

581: \begin{prop}\label{p:eigen-decompo}

582: Take some $\Lambda\in [0,1)$ and $\nu$ an arbitrary Borel probability

583: measure on $\Gamma_+^{(1)}=\Gamma_+\cap H$.

584: Define the probability measure $\mu$ on $\t2$ as follows:

585: \bequ\label{e:eigen-decompo}

586: \mu= %(1-\Lambda) \sum_{n\geq 0}\Lambda^n\,(\hT^*)^n\,\nu=

587: (1-\Lambda) \sum_{n\geq 0}\Lambda^n\,(T^*)^n\,\nu\,.

588: \eequ

589: Then $T^*\,\mu=\Lambda\,\mu$.

590: All $\Lambda$-eigenmeasures of $T$ can be written this way.

591: \end{prop}

592: %&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&

593: This construction shows that, as long as $\Gamma_+^{(1)}$ is not empty 

594: (that is, there exists at least a point $x_0$ trapped in the past but escaping

595: in the future), there are plenty of eigenmeasures. 

596: 

597: The case where the map $\hT$ is uniformly hyperbolic has been studied in detail

598: by Chernov and Markarian \cite{Mark96,CherMar97}. The set $\Gamma_+^{(1)}$ is then 

599: uncoutable, and it is foliated by the unstable foliation. These authors 

600: focussed on eigenmeasures of the open map $T$ which are

601: absolutely continuous along the unstable direction. 

602: Even with this condition, one has plenty of eigenmeasures 

603: (we remind that

604: for a closed hyperbolic map, unstable absolute continuity is satisfied for a 

605: unique invariant measure, namely the SRB measure).

606: 

607: %=======================================

608: \subsubsection{Pure point eigenmeasures}\label{s:pp} 

609: %=======================================

610: 

611: Applying the recipe of Proposition~\ref{p:eigen-decompo} to the Dirac measure $\delta_{x_0}$ on

612: an arbitrary point $x_0\in\Gamma_+^{(1)}$, we obtain a simple, pure point $\Lambda$-eigenmeasure

613: supported on the backwards trajectory $\big(x_{-n}=T^{-n}(x_0)\big)_{n\geq 0}$. 

614: For any $\Lambda\in (0,1)$, we call this eigenmeasure

615: \bequ\label{e:mu-traject}

616: \mu_{x_0,\Lambda}\defeq (1-\Lambda)\sum_{n\geq 0} \Lambda^n\,\delta_{x_{-n}}\,.

617: \eequ

618: We recall that the pure point {\em invariant} 

619: measures of $T$ are localized on periodic orbits of $T$

620: on $K$ (which, for the case of a horseshoe map $T_{|K}$, form a countable family). 

621: On the opposite, 

622: for each $\Lambda\in [0,1)$ the family of pure point eigenmeasures 

623: $\mu_{x_0,\Lambda}$ is labelled by all $x_0\in \Gamma_+^{(1)}$ (an uncountable set).

624: 

625: %=======================================

626: \subsubsection{Natural eigenmeasure}\label{s:natural} 

627: %=======================================

628: 

629: As discussed in \cite{DemYou06}, one may search for the definition of a {\em natural} eigenmeasure of

630: $T$. The simplest (``ideal'') definition reads as follows: for any 

631: initial measure $\rho$ absolutely continuous w.r.to $\mu_L$, and

632: such that $\rho(\Gamma_-)> 0$,

633: \bequ\label{e:SRB-open}

634: \mu_{nat}=\lim_{n\to\infty} \frac{T^{*n}\,\rho}{\|T^{*n}\,\rho\|}\,,\qquad\text{where}\ 

635: \|\mu\|\defi \mu(\hM)\,.

636: \eequ

637: It was shown in \cite{Mark96,CherMar97}

638: that, for a hyperbolic open map, the limit exists and is independent of $\rho$. Besides,

639: the natural measure is then absolutely continuous along the unstable direction.

640: Yet, for a general open map the above limit does not necessarily exist, 

641: or it may depend on the initial distribution $\rho$ \cite{DemYou06}.

642: 

643: In case \eqref{e:SRB-open} holds, 

644: the eigenvalue $\Lambda_{nat}=\mu_{nat}(M)$ associated with this measure is generally called

645: ``the decay rate of the system'' by physicists.

646: For a {\em closed} chaotic map $\hT$, 

647: the natural measure is the Liouville 

648: measure $\mu_{L}$ (indeed, $\hT^{*n}\rho\to \mu_{L}$ is equivalent with

649: the fact that $\hT$ is {\em mixing} with respect to $\mu_{L}$). 

650: As explained in \S\ref{s:eigenstates-IM}, the quantum ergodicity theorem 

651: shows that this particular invariant measure is ``favored'' by quantum mechanics.

652: One interesting question we will address is the relevance of

653: $\mu_{nat}$ with respect to the quantized open map $T_N$.

654: 

655: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

656: \subsubsection{Bernoulli eigenmeasures of the open baker}\label{s:Bernoulli} 

657: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

658: 

659: We now focus on the open baker $B_{\vr}$ and the open shift $\sigma$ it is conjugated with,

660: and construct a family of eigenmeasures, called Bernoulli eigenmeasures. Some of

661: these measures will appear in \S\ref{s:walsh} as

662: semiclassical measures for the Walsh-quantized open baker.

663: 

664: 

665: The first equality in \eqref{e:conjugation} maps

666: the set $\Sigma_+$ of one-sided sequences $\cdot \ep_0\ep_1\ldots$ to

667: the position interval $[0,1]$. By a slight abuse, we also call this

668: mapping $J_{\vr}$. We will first construct Bernoulli measures on the

669: symbol spaces $\Sigma_+$ and $\Sigma$, and then push them on 

670: the interval or the torus using $J_{\vr}$. 

671: 

672: Let us recall the definition of Bernoulli measures on $\Sigma_+$. Choose

673: a {\em weight distribution} $\vP=(P_0,P_1,P_2)$, 

674: where $P_\ep\in[0,1]$ and $P_0+P_1+P_2=1$. We define the measure $\nu^{\Sigma_+}_{\vP}$

675: on $\Sigma_+$ as follows: 

676: for any $n$-sequence $\bep=\cdot\ep_0\ep_1\cdots\ep_{n-1}$, the weight of the cylinder $[\bep]$

677: is given by 

678: $$

679: \nu^{\Sigma_+}_{\vP}([\bep])=P_{\ep_{0}}\cdots P_{\ep_{n-1}}\,.

680: $$

681: The push-forward on $[0,1]$ of this measure will also be called a Bernoulli measure,

682: and called $\nu_{\vr,\vP}=J_{\vr}^*\,\nu^{\Sigma_+}_{\vP}$.

683: If we take $P_\ep=r_\ep$ for $\ep=0,1,2$, we recover $\nu_{\vr,\vr}=\nu_{Leb}$ the Lebesgue 

684: measure on $[0,1]$. 

685: If for some $\ep\in \set{0,1,2}$ we take $P_\ep=1$, we get for $\nu_{\vr,\vP}$ 

686: the Dirac measure at the point $q(\cdot \ep\ep\ep\ldots)$, which takes the respective

687: values $0$, $\frac{r_0}{1-r_1}$ and $1$. 

688: For any other distribution $\vP$,

689: the Bernoulli measure $\nu_{\vr,\vP}$ is purely singular

690: continuous w.r.to the Lebesgue measure. Fractal properties of the measures 

691: $\nu_{\vr,\vP}$ were studied in \cite{HJPPS86}. 

692:  If the weight $P_\ep$ vanishes for a single $\ep$ (e.g. $P_1=0$),

693: $\nu_{\vr,\vP}$ is supported on a Cantor set (e.g. $\cC_{\vr}$).

694: 

695: A Bernoulli measure on $\nu^{\Sigma_-}_{\vP}$ can be defined similarly.

696: By taking products of two Bernoulli measures, one easily constructs eigenmeasures of 

697: the open shift $\sigma$ or the open baker $B_{\vr}$.

698: %%%%%%%%%%%%%%%%%%%

699: \begin{prop}

700: Take any weight distribution $\vP$ such that $P_1<1$. Then the following hold.

701: 

702: $i)$ there is a unique auxiliary distribution,

703: namely $\vP^*=\set{\frac{P_0}{P_0+P_2},0,\frac{P_2}{P_0+P_2}}$,

704: such that the product measure

705: $$

706: \mu^\Sigma_{\vP}\defeq \nu^{\Sigma_+}_{\vP}\times\nu^{\Sigma_-}_{\vP^*}

707: $$

708: is an eigenmeasure of the open shift $\sigma$. The corresponding decay rate is

709: \bequ

710: \Lambda_{\vP}=1-P_1=P_0+P_2\,.

711: \eequ

712: $ii)$ for any $\vr$, the push-forward 

713: $\mu_{\vr,\vP}=J_{\vr}^*\,\mu^\Sigma_{\vP}$

714: is an eigenmeasure of the open baker $B_{\vr}$.

715: \end{prop}

716: %%%%%%%%%%%%%%%%%%%

717: By definition, the product measure has the following weight on a cylinder 

718: $[\bep]=[\ep_{-m}\ldots\ep_{n-1}]$:

719: $$

720: \mu^\Sigma_{\vP}([\bep])=P^*_{\ep_{-m}}\ldots P^*_{\ep_{-1}}\,P_{\ep_{0}}\cdots P_{\ep_{n-1}}\,.

721: $$

722: The proof of the first statement is straightforward.

723: There is a slight subtlety concerning push-forwards of $\sigma$-eigenmeasures, 

724: which is resolved in the following 

725: %============

726: \begin{prop}\label{l:eigenmeasure-push}

727: Let $\mu^\Sigma$ be an eigenmeasure of the open shift $\sigma:\Sigma\to\Sigma$. 

728: If $\mu^\Sigma$ does not charge the subset $\Sigma\setminus\Sigma''$ described in \eqref{e:Sigma''}, 

729: that is if $\mu^\Sigma(\Sigma\setminus\Sigma'')=0$, then its push-forward

730: $\mu=J_{\vr}^*\mu^{\Sigma}$ on $\t2$ is an eigenmeasure of $B_{\vr}$.

731: \end{prop}

732: %============

733: \begin{proof}

734: Take any Borel set $S\in\t2$.

735: Then the following identities hold:

736: $$

737: \mu(B_{\vr}^{-1}(S))\defeq \mu^\Sigma(J_{\vr}^{-1}\circ B_{\vr}^{-1}(S))

738: =\mu^\Sigma(\sigma^{-1}\circ J_{\vr}^{-1}(S))%\\

739: =\Lambda \, \mu^\Sigma(J_{\vr}^{-1}(S))

740: \defeq \Lambda \, \mu(S)\,.

741: $$

742: The second equality is a consequence of the semiconjugacy \eqref{e:semiconj}, which implies

743: the fact that the symmetric difference of the sets $J_{\vr}^{-1}\circ B_{\vr}^{-1}(S)$ and 

744: $\sigma^{-1}\circ J_{\vr}^{-1}(S)$

745: is necessarily a subset of $\Sigma\setminus\Sigma''$.

746: \end{proof}

747: The condition $\mu^\Sigma(\Sigma\setminus\Sigma'')=0$ in the proposition cannot be removed. 

748: Indeed, by applying the 

749: construction of proposition~\ref{p:eigen-decompo} to an initial ``seed'' 

750: supported on $\set{\ldots 222\cdot 1\ep_1\ldots}$, one obtains an eigenmeasure of $\sigma$ 

751: charging $\Sigma''$, and such that its push-forward is not an eigenmeasure of $B_{\vr}$.

752: 

753: To prove the second point of the Proposition, we remark that 

754: the only Bernoulli eigenmeasure $\mu^\Sigma_{\vP}$ charging $\Sigma\setminus\Sigma''$

755: is the Dirac measure on the sequence $\ldots 2222\ldots$, which is

756: pushed-forward to the delta measure at the origin of $\t2$. 

757: $\hfill\square$

758: 

759: If we take $\vP=\vr$, the push-forward is

760: the natural eigenmeasure $\mu_{\vr,\vr}=\mu_{nat}$ of the open baker $B_{\vr}$.

761: If $\vP\neq\vP'$, the Bernoulli eigenmeasures $\mu^\Sigma_{\vP}$ and

762: $\mu^\Sigma_{\vP'}$

763: are {\em mutually singular} (there exists disjoint Borel subsets $A$, $A'$ of $\Sigma$

764: such that $\nu_{\vr,\vP}(A)=\nu_{\vr,\vP'}(A')=1$),

765: eventhough they may share the same decay rate.

766: Except in the case $\vP=(1,0,0)$, $\vP'=(0,0,1)$, the push-forwards 

767: $\mu_{\vr,\vP}$ and $\mu_{\vr,\vP'}$ are also mutually singular.

768: 

769: 

770: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

771: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

772: \section{Quantized open maps}

773: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

774: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

775: 

776: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

777: \subsection{``Axioms'' of quantization}\label{s:quantum}

778: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

779: For appropriate $2D$-dimensional compact symplectic manifolds $\hM$, one may

780: define a sequence of ``quantum'' Hilbert spaces $(\hn)_{N\to\infty}$ of 

781: finite dimensions $N$, which are related with Planck's constant by

782: $N\sim \hbar^{-D}$. 

783: Quantum states are normalized vectors in $\hn$. One

784: also wants to quantize observables, that is functions $a\in C^\infty(\hM)$,

785: into operators $\Op_N(a)$ on $\hn$. An invertible (resp. open) map $\hT$ (resp. $T$)

786: is quantized into a family of

787: a {\em unitary} (resp. contracting) operators $\hT_N$ (resp. $T_N$), 

788: which satisfies certain properties when $N\to\infty$ (see below). 

789: 

790: For the example we will treat explicitly (the open baker's map), 

791: the propagators of the closed and open maps are related by

792: $T_N=\hT_N\circ\Pi_{M,N}$,

793: where $\Pi_{M,N}$ is a projector associated with the subset $M$: it kills 

794: the quantum states microlocalized in the hole, while keeping unchanged

795: the states microlocalized inside $M$ \cite{SaVa96}. 

796: Yet, in general the propagator

797: $T_N$ can be defined without having to construct $\hT_N$ beforehand.

798: 

799: The proof of our main result, Theorem~\ref{thm:main}, only uses some

800: ``minimal'' properties of the quantized observables and

801: maps. These properties were presented as

802: ``quantization axioms'' by Marklof and O'Keefe in the case of closed maps \cite{MarOK05}.

803: We adopt the same approach, namely define quantization through these ``minimal axioms'',

804: and state our result in this general framework.

805: Afterwards, we will check that these

806: axioms are satisfied for the quantized open baker.

807: 

808: %================================

809: \subsubsection{Axioms on observables}\label{s:axioms-obs}

810: %================================

811: 

812: All axioms will describe properties of the quantum operators in the semiclassical limit 

813: $N\to\infty$. With a slight abuse of notation, we 

814: write $A_N\sim B_N$ when two families of operators $(A_N)$, $(B_N)$ on $\hn$ satisfy

815: $$

816: \norm{A_N-B_N}_{\cL(\hn)}\Nto8 0\,.

817: $$

818: The axioms concerning the quantization of observables

819: read as follows \cite[Axiom~2.1]{MarOK05}. 

820: \begin{Def}\label{d:axiom-obs}

821: For any $a\in C^\infty(\hM)$, the operators

822: $\Op_N(a)$ on $\hn$ must satisfy, in the limit $N\to\infty$:

823: \bequ

824: \begin{split}\label{e:axiom-obs}

825: \Op_N(\bar{a})\sim \Op_N(a)^{\dagger}\,&\qquad \text{(asymptotic hermiticity)}\\

826: \Op_N(a)\,\Op_N(b)\sim \Op_N(ab)\,&\qquad \text{(0-th order symbolic calculus)}\\

827: \lim_{N\to\infty} N^{-1}\,\Tr \Op_N(a)=\int_{\hM}a\,d\mu_{L}&\qquad \text{(normalization)}\,.

828: \end{split}

829: \eequ

830: \end{Def}

831: These axioms are satisfied by all standard quantization recipes (e.g. geometric or

832: Toeplitz quantization on K\"ahler manifolds \cite{Zel96}). Notice that they do not

833: involve the symplectic structure on $\hM$, and can thus be extended to more general phase

834: spaces.

835: 

836: %================================

837: \subsubsection{Semiclassical mesures}\label{s:semicl-measure}

838: %================================

839: From there, we may define semiclassical measures associated with sequences of (normalized)

840: quantum states $(\psi_N\in\hn)_{N\to\infty}$. Such a sequence is said to converge to the

841: distribution $\mu$ on $\hM$ iff, for any observable $a\in C^\infty(\hM)$, 

842: \bequ\label{e:semiclass-measure}

843: \la \psi_N,\,\Op_N(a)\psi_N\ra\Nto8 \int_{\hM}a\,d\mu\,.

844: \eequ

845: From the above axioms, one can show that $\mu$ is necessarily 

846: a probability measure on $\hM$,

847: which is called the semiclassical measure associated with $(\psi_N\in\hn)$.

848: By weak compactness, from any sequence $(\psi_N\in\hn)$ one can always extract a

849: subsequence $(\psi_{N_k})$ converging in the above sense to a certain measure. The latter

850: is called {\it a} semiclassical measure of the sequence $(\psi_N)$.

851: 

852: %==============================

853: \subsubsection{Axioms on maps}\label{s:axiom-maps}

854: %==============================

855: 

856: In the axiomatic framework of \cite{MarOK05}, the conditions satisfied by 

857: the unitary propagators $(\hT_N)$ quantizing a closed map $\hT$ 

858: consist in some form of quantum-classical correspondence (Egorov's theorem), 

859: when evolving quantum observables $\Op_N(a)$ through these propagators. 

860: However, if $\hT$ is discontinuous (or nonsmooth) at some

861: points, its propagator $\hT$ will exhibit diffraction phenomena 

862: around these ``singular'' points,

863: which alter the propagation properties. At the classical level,

864: a smooth observable $a\in C^\infty(\hM)$ is transformed into 

865: a smooth observable $a\circ \hT\in C^\infty(\hM)$ only if $a$ vanishes

866: near the singular points of $\hT^{-1}$.

867: As a result, the quantum-classical correspondence

868: is a reasonable axiom only if the observable $a$ is

869: supported ``far away'' from the singular set of $\hT^{-1}$..

870: 

871: We now specify our assumptions on the open map $T:M\to\hM$. Firstly, the

872: hole $H$ will be a ``nice'' set, that is a set with nonempty interior and such 

873: that $\partial H$ has Minkowski content zero 

874: ($i.e.$ the volume of its $\eps$-neighbourhood vanishes when $\eps\to 0$).

875: We also assume that $M$, the domain of definition of $T$, can be decomposed using 

876: finitely or countably many open connected sets $O_i$:

877: $\overline{M}=\overline{\cup_{i\geq 1}O_i}$, with $O_i\cap O_j=\emptyset$, and for each $i$, 

878: $T_{|O_i}:O_i\to T(O_i)$ is a smooth canonical diffeomorphism, 

879: with all derivatives uniformly bounded. 

880: Let us split the hole into

881: $H=\ring{H}\sqcup DH$.

882: The {\it continuity set} (resp. {\it discontinuity set}) of the map $T$ is defined as

883: $$

884: C(T)=\sqcup_{i\geq 1}O_i\subset \ring{M}\,,

885: \quad\text{resp.}\quad D(T)\defi (M\setminus C(T))\sqcup DH\,.

886: $$

887: We assume that $D(T)$ has Minkowski content zero. We have the decomposition

888: $\hM=C(T)\sqcup \ring{H}\sqcup D(T)$. A similar decomposition holds for the inverse map:

889: \bequ\label{e:inverse-decompo}

890: \hM=C(T^{-1})\sqcup \ring{H}^{-1}\sqcup D(T^{-1}),\quad\text{with}\quad 

891: C(T^{-1})=T\big(C(T)\big)\,.

892: \eequ

893: Adapting the axioms of \cite{MarOK05} to the case of open maps, we set as follows

894: the characteristic

895: property of the operators $(T_N)_{N\to\infty}$ quantizing $T$.

896: \begin{Def}\label{d:axiom}

897: We say that the operators $\big(T_N\in\cL(\hn)\big)_{N\to\infty}$ quantize

898: the open map $T$ iff

899: \begin{itemize}

900: \item for $N$ large enough, $\norm{T_N}_{\cL(\hn)}\leq 1$ 

901: \item for any observable $a\in C^\infty_c\big(C(T^{-1})\sqcup \ring{H}^{-1}\big)$, we

902: have in the limit $N\to\infty$

903: \bequ\label{e:axiom2-map}

904: T_N^{\dagger}\,\Op_N(a)\,T_N \sim \Op_N\big(\bbbone_{M}\times(a\circ T)\big)\,.

905: \eequ

906: \end{itemize}

907: Here $C^\infty_c(S)$ indicates the smooth functions compactly supported inside $S$,

908: and $\bbbone_{M}$ is the characteristic function on $M$.

909: \end{Def}

910: Notice that, if $a$ is supported inside $C(T^{-1})\sqcup \ring{H}^{-1}$, 

911: the function $\bbbone_{M}\times(a\circ T)$ is well-defined,

912: smooth and supported inside $C(T)$. 

913: The factor $\bbbone_{M}$ ensures that

914: an observable supported inside $H^{-1}$ is ``semiclassically killed'' 

915: by the evolution through $T_N$.

916: The condition \eqref{e:axiom2-map} reminds of the definition of a

917: ``quantized weighted relation'' introduced in \cite{NonZw06}, but it is less precise

918: (it only describes the lowest order in $\hbar$).

919: 

920: \begin{rem}

921: Going back to the problem of potential scattering mentioned in the introduction,

922: we expect the operators $T_N$ to share some spectral

923: properties with the propagator $\exp\big(-\i H_{W,\hbar}/\hbar\big)$ of the

924: ``absorbing Hamiltonian'', in the semiclassical limit.

925: The eigenvalues $\set{\lambda_{j}}$ of $T_{N}$ 

926: should be compared with $\set{\e^{-\i \tilde{z}_j/\hbar}}$, or with 

927: $\set{\e^{-\i z_j/\hbar}}$, where $\set{\tilde{z}_j}$ (resp. $\set{z_j}$) are the

928: eigenvalues of $H_{W,\hbar}$ (resp. the resonances of $H_\hbar$). Similarly, 

929: the eigenstates of $T_N$ should share some microlocal properties with the eigenfunctions

930: $\tilde\varphi_j$ of $H_{W,\hbar}$ (resp. the resonant states $\varphi_j$ of $H_\hbar$)

931: inside the interaction region. Accordingly, we will sometimes 

932: call ``resonances'' and ``resonant

933: eigenstates'' the eigenvalues/states of quantized open maps.

934: \end{rem}

935: 

936: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

937: \subsection{The quantum open baker's map}\label{s:quantum-baker}

938: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

939: 

940: For any dimension $N=(2\pi\hbar)^{-1}$, the quantum Hilbert space $\hn$ 

941: adapted to the torus phase space is spanned by the

942: orthonormal {\em position basis} $\set{\bq_j,\ j=0,\ldots,N-1}$, localized at the 

943: discrete positions $q_j=\frac{j}N$. 

944: 

945: There are several standard ways to quantize observables on $\t2$: Weyl quantization,

946: Toeplitz (or anti-Wick) quantizations \cite{BouzDB96}, or Walsh quantization \cite{AN06}. 

947: Weyl and anti-Wick quantizations are

948: equivalent with each other in the semiclassical limit, in the sense that 

949: $\Op_N^{W}(a)\sim  \Op_N^{AW}(a)$ for any smooth observable. 

950: On the opposite, the

951: Walsh quantization (see \S\ref{s:walsh0}) is not equivalent 

952: to the previous ones.

953: 

954: After recalling the definition of the anti-Wick quantization on $\t2$ and

955: the ``standard'' quantization of the baker's map,

956: following the original approach of Balazs-Voros, Saraceno, 

957: Saraceno-Vallejos \cite{BV89,Sar90,SaVa96}, we check that the latter

958: satisfies Axioms~\ref{d:axiom}

959: with respect to the anti-Wick quantization.

960: 

961: %=====================================================

962: \subsubsection{Anti-Wick quantization of observables}\label{s:Husimi} 

963: %=====================================================

964: We recall the definition and properties of coherent states on $\t2$,

965: which we use to construct the anti-Wick quantization of observables,

966: and by duality the Husimi representation

967: of quantum states \cite{BouzDB96}.

968: The Gaussian coherent state in $L^2(\IR)$, localized at the phase space

969: point $\bx=(q_0,p_0)\in\IR^2$ and with squeezing parameter $\ssigma>0$, 

970: is defined by the normalized wavefunction

971: \begin{equation}

972:   \label{e:planeCS}

973:   \Psi_{\bx,\ssigma}(q)\defeq  \left( \frac{\ssigma}{\pi\hbar}\right)^{1/4}

974:    \,\e^{-\i \frac{p_0 q_0}{2\hbar}}\,

975: \e^{\i \frac{p_0 q}{\hbar}}\,

976: \e^{-\ssigma \frac{(q-q_0)^2}{2\hbar}}\,.

977: \end{equation}

978: When $\hbar=(2\pi N)^{-1}$, 

979: that state can be periodized on the torus, to yield the torus coherent state

980: $\psi_{\bx,\ssigma}\in\hn$ with following components in the basis $\set{\bq_j}$:

981: \bequ\label{e:torusCS}

982: \la\bq_j, \psi_{\bx,\ssigma}\ra=\frac{1}{\sqrt{N}}\,\sum_{\nu\in\IZ}

983: \Psi_{\bx,\ssigma}(j/N+\nu),\qquad

984: j=0,\ldots,N-1\,.

985: \eequ

986: For $s>0$ fixed, these states are asymptotically normalized when $N\to\infty$.

987: 

988: To any squeezing $\ssigma>0$ and inverse 

989: Planck's constant $N\in\IN$ we

990: associate the anti-Wick (or Toeplitz) quantization

991: \bequ\label{e:anti-Wick}

992: f\in C^\infty(\t2)\longmapsto

993: \OpAW(f)\defi \int_{\t2}  |\psi_{\bx,\ssigma}\ra\la\psi_{\bx,\ssigma}|\;f(\bx)\;N\,d\bx\,,

994: \eequ

995: which satisfies the Axioms~\eqref{d:axiom-obs} \cite{BouzDB96}.

996: By duality, this quantization defines, for any state $\psi\in\hn$, 

997: a Husimi distribution $H^\ssigma_\psi$:

998: $$

999: \forall f\in C^\infty(\t2),\qquad H^\ssigma_\psi(f)\defeq \la\psi,\OpAW(f)\,\psi\ra\,.

1000: $$

1001: For $\norm{\psi}=1$, this distribution is a probability measure, with density given by 

1002: the (smooth, nonnegative) Husimi function 

1003: \bequ\label{e:Husimi}

1004: H^\ssigma_{\psi}(\bx)= N\,|\la \psi_{\bx,\ssigma},\psi\ra|^2\,,\quad

1005: \bx\in\t2\,.

1006: \eequ

1007: Applying the definition \eqref{e:semiclass-measure} to the present framework, 

1008: a sequence of states $(\psi_N\in\hn)_{N\to\infty}$ converges

1009: to the measure $\mu$ on $\t2$ iff, for any given $\ssigma>0$, the Husimi measures

1010: $(H^{\ssigma}_{\psi_N})_{N\to\infty}$ weak-$*$ converge to the measure $\mu$.

1011: 

1012: Following Schubert's work \cite{Roman-PhD}, 

1013: one can also consider anti-Wick 

1014: quantizations  (and dual Husimi measures)

1015: in which the squeezing parameter $\ssigma$ in the integral \eqref{e:anti-Wick} 

1016: depends on the phase space point.

1017: Adapting the proofs of \cite{Roman-PhD} to the torus setting, one

1018: shows that all these quantizations are equivalent to one another:

1019: %+++++++++++++++++++++++++++++

1020: \begin{prop}\label{p:quant}

1021: Choose two functions  $\ssigma_1,\ssigma_2\in C^\infty(\t2,(0,\infty))$.

1022: Then the two associated anti-Wick quantizations become close

1023: to one another when $N\to\infty$:

1024: $$

1025: \forall f\in C^\infty(\t2),\qquad \norm{{\rm Op}^{{\rm AW}\!,\ssigma_1}_N(f)-

1026: {\rm Op}^{{\rm AW}\!,\ssigma_2}_N(f)}=\cO(N^{-1})\,.

1027: $$

1028: \end{prop}

1029: %+++++++++++++++++++++++++++++

1030: %As a result, the convergence of $(H_{\psi_N}^{\ssigma})$ to the semiclassical 

1031: %measure $\mu$ is independent of the choice of $\ssigma$.

1032: 

1033: %=====================================================

1034: \subsubsection{Standard quantization of the baker's map}\label{s:quantum-baker2}

1035: %=====================================================

1036: Strictly speaking, the quantization 

1037: of the closed baker's map $\hB_{\vr}$ is well-defined only 

1038: if the coefficients $\vr$ are such that

1039: \bequ\label{e:condition}

1040: N\,r_i=N_i\in\IN,\quad i=0,1,2\,.

1041: \eequ

1042: Yet, in the

1043: semiclassical limit $N\to\infty$ one can, if necessary,

1044: slightly modify the $r_i$ by amounts $\leq 1/N$ in order to satisfy this condition: such

1045: a modification is irrelevant for the classical dynamics. Assuming 

1046: \eqref{e:condition}, the quantization of $\hB_{\vr}$ on $\hn$ is given by the

1047: following unitary matrix in the position basis \cite{BV89}:

1048: \bequ\label{e:A_N}

1049: \hB_{\vr,N}=F_N^{-1}\begin{pmatrix}F_{N_0}&&\\&F_{N_1}&\\&&F_{N_2}\end{pmatrix}\,.

1050: \eequ

1051: Here $F_{N_i}$ denotes the $N_i$-dimensional discrete Fourier transform, 

1052: \bequ\label{e:DFT}

1053: (F_{N_i})_{jk}={N_i}^{-1/2}\,\e^{-2\i\pi jk/N_i},\quad j,k=0,\ldots N_i-1.

1054: \eequ

1055: Since we already have a quantization for the $\hB_{\vr}$ and the hole is the 

1056: rectangle $H=R_1=\set{r_0\leq q<1-r_2}$, a natural choice to quantize $B_{\vr}$ is to

1057: project on the positions $q\in [0,1)\setminus [r_0,1-r_2)$ and then apply $\hB_{\vr,N}$.

1058: One gets the following open propagator in the position basis \cite{SaVa96}:

1059: \bequ\label{e:B_N}

1060: B_{\vr,N}=F_N^{-1}\begin{pmatrix}F_{N_0}&&\\&0&\\&&F_{N_2}\end{pmatrix}\,.

1061: \eequ

1062: This is the ``standard'' quantization of the open baker's map $B_{\vr}$. These

1063: matrices are obviously contracting on $\hn$.

1064: The semiclassical

1065: connection between the matrices $\hB_{\vr,N}$ (resp. $B_{\vr,N}$) and the classical map 

1066: $\hB_{\vr}$ (resp. $B_{\vr}$) has

1067: been analyzed in detail in \cite{NonZw06}; this analysis implies the property

1068: \eqref{e:axiom2-map} with respect to the Weyl quantization.

1069: %%%%%%%%%%%%%%%%%%%%

1070: \begin{figure}[htbp]

1071: \begin{center}

1072: \includegraphics[width=.6\textwidth]{gamma+3.eps}

1073: \caption{\label{f:gamma+} On the left, we show the backward trapped set $\Gamma_+$ for the

1074: symmetric open baker $B_{\vr_{sym}}$ (black), and the discontinuity set $D(B_{\vr_{sym}}^{-1})$ 

1075: (pink/light gray). Their union gives $\Gamma_+\sqcup D_{-\infty}$, which contains

1076: the support of semiclassical measures (see Thm.~\ref{thm:main}, $i)$).

1077: The dotted lines indicate identical points on $\t2$. 

1078: On the right, we show the backwards image 

1079: $B_{\vr_{sym}}^{-1}(D(B_{\vr_{sym}}^{-1}))$ involved in 

1080: Thm~\ref{thm:main}, $iii)$ (pink/light gray), and the projection on $\t2$ of the set 

1081: $\Sigma\setminus\Sigma''$ defined in \eqref{e:Sigma''} (red/dark gray).} 

1082: \end{center}

1083: \end{figure}

1084: %%%%%%%%%%%%%%%%%%%%

1085: Below we give an alternative proof of the property \eqref{e:axiom2-map} for the open

1086: propagator $B_{\vr,N}$, with respect to the

1087: anti-Wick quantization. The proof uses the semiclassical propagation of 

1088: coherent states analyzed in \cite{DENW06}. 

1089: 

1090: To start with, we precisely give the continuity set

1091: of $B_{\vr}^{-1}$:

1092: $$

1093: C(B_{\vr}^{-1})=\tR_0\sqcup \tR_2\,,

1094: $$

1095: where $\tR_0=\set{q\in(0,1),\ p\in (0,r_0)}\subset B_{\vr}(R_0)$ 

1096: and similarly for $\tR_2$: in the case of the symmetric baker,

1097: these are the open grey rectangles on 

1098: the right plot of Fig.~\ref{f:3-baker}.

1099: On the other hand, the hole $H^{-1}=B_{\vr}(R_1)=\set{q\in [0,1),\ p\in [r_0,1-r_2)}$ (this is 

1100: the white strip, including the vertical side). 

1101: The discontinuity set $D(B_{\vr_{sym}}^{-1})$

1102: is shown in Fig.~\ref{f:gamma+} (pink/light gray lines in the left plot).

1103: 

1104: Any observable 

1105: $a\in C^\infty_c\big(C(B_{\vr}^{-1})\sqcup \ring{H}^{-1}\big)$ is supported 

1106: at a distance $\geq\delta>0$

1107: from the discontinuity set $D(B_{\vr}^{-1})$. 

1108: Let select some smooth $\ssigma\in C^\infty(\t2,(0,\infty))$, and consider

1109: the corresponding anti-Wick quantization (see \eqref{e:anti-Wick}).

1110: From the definition \eqref{e:anti-Wick}, the operator 

1111: ${\rm Op}^{{\rm AW}\!,\ssigma}(a)$ only involves coherent states

1112: $\psi_{\bx,\ssigma(\bx)}$ located at distance $\geq\delta>0$ from $D(B_{\vr}^{-1})$.

1113: Adapting the proof of \cite[Prop.5]{DENW06} to the general open baker $B_{\vr}$,

1114: one can show the following propagation for such states:

1115: \bequ\label{e:CS-evol}

1116: B_{\vr,N}^\dagger\; \psi_{\bx,\ssigma(\bx)}=\bbbone_{M^{-1}}(\bx)\,\e^{i \theta(\bx,N)} \,

1117: \psi_{\bx',\ssigma'(\bx')}

1118: +\cO_{\hn}(\e^{-N\,C(\ssigma,\del)})\,,\qquad N\to\infty\,.

1119: \eequ

1120: Here $\bx'=B_{\vr}^{-1}(\bx)$, $\ssigma'(\bx')=\ssigma(\bx)/r_\eps^2$ if $\bx'$ lies inside the

1121: rectangle $R_{\eps}$, and $\theta(\bx,N)$ is a phase which can be explicitly computed. 

1122: 

1123: Through the symplectic 

1124: change of variable $\by=B_{\vr}^{-1}(\bx)$ for $\by\in M$, one gets

1125: $$

1126: B_{\vr,N}^\dagger\,{\rm Op}^{{\rm AW}\!,\ssigma}(a)\,B_{\vr,N}= 

1127: {\rm Op}^{{\rm AW}\!,\ssigma'}(\bbbone_{M}\times (a\circ B_{\vr})) +

1128: \cO_{\cB(\hn)}(\e^{-N\,C'(\ssigma,\del)})\,.

1129: $$

1130: The function $\ssigma'$ is obtained by taking $\ssigma'(\by)=\ssigma(B_{\vr}(\by))/r_\eps^2$ for 

1131: $\by\in B_{\vr}^{-1}(\supp a\cap \tR_{\eps})$, and smoothly extending the function to 

1132: $\ssigma'\in C^\infty(\t2,(0,\infty))$.

1133: Since the quantizations with parameters $\ssigma$ and $\ssigma'$ are equivalent 

1134: (Prop.~\ref{p:quant}), we have proven

1135: the property \eqref{e:axiom2-map} for the family $(B_{\vr,N})$. $\hfill\square$

1136: 

1137: 

1138: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1139: \section{Fractal Weyl law for the quantized open baker} 

1140: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1141: In this section, which mainly presents numerical results, 

1142: we exclusively consider the open baker's maps $B_{\vr}$ and their 

1143: quantizations \eqref{e:B_N}. Our aim is to investigate the precise notion of dimension

1144: entering the fractal Weyl law conjectured below, through a numerical study which

1145: complements the one performed in \cite{NonZw05}. Still, we believe that most

1146: statements should hold as well if we replace $B_{\vr}$ by a map $T$ obtained by

1147: restricting an Anosov diffeomorphism $\hT$ outside some 

1148: ``small hole'', as described in \cite{CherMar97}.

1149: 

1150: From the explicit formula \eqref{e:B_N}, the subspace of $\hn$ spanned by 

1151: $\set{\bq_j,\ N_0\leq j<N-N_2}$ is in the kernel of $B_{\vr,N}$. We call

1152: ``nontrivial'' the spectrum of $B_{\vr,N}$ on the complementary subspace.

1153: That spectrum is situated inside the unit disk. In the semiclassical 

1154: limit $N\to\infty$, most of it accumulates 

1155: near the origin, which corresponds to ``short-living'' eigenvalues \cite{SchoTwo04}.

1156: We rather focus on ``long-living'' eigenvalues, situated in some annulus away from the origin. 

1157: By analogy with the case of potential scattering \eqref{e:Weyl-law}, a fractal Weyl law

1158: for the semiclassical density of ``long-living'' eigenvalues was conjectured in \cite{NonZw06}:

1159: %%%%%%%%%%%%%%%%%%%%%%%%%%%

1160: \begin{conj}[Fractal Weyl Law]\label{conj:FWL}

1161: Let $B_{\vr,N}$ be the quantized open baker's map described in \S\ref{s:quantum-baker}.  

1162: Then, for any radius $0<r<1$, there exists $C_r\geq 0$ such that

1163: \bequ\label{e:FWL}

1164: n(N,r)\defeq

1165: \#\set{ \lambda \in \Spec ( B_{\vr,N} ),\ :\  |\lambda|\geq r}=C_r\,N^{d}\,+o(N^{d }),

1166: \qquad N\to\infty\,.

1167: \eequ

1168: The eigenvalues

1169: are counted with multiplicities, and $2d$ is an appropriate 

1170: fractal dimension of the trapped set $K$. 

1171: \end{conj}

1172: %%%%%%%%%%%%%%%%%%%%%%%%%%%

1173: %So far, this conjecture could be proven only for the Walsh quantization $\tB_N$

1174: %of the symmetric baker $B_{\vr_{sym}}$, see \S\ref{s:Walsh-baker}.

1175: %This Weyl law was numerically checked for the open kicked rotator \cite{SchoTwo04}, and

1176: %for the ``standard'' quantum open baker $B_{\vr_{sym},N}$, especially 

1177: %for $N$ along geometric sequences $N=N_o\,3^k$, $k\geq 1$ \cite{NonZw05}. 

1178: 

1179: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1180: \subsection{Which dimension plays a role?}\label{s:infodim?} 

1181: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1182: %%%%%%%%%%%%%%%%%%%%

1183: \begin{figure}[htbp]

1184: \begin{center}

1185: \includegraphics[width=8cm]{ecart-r0=0.1-1-32.2-3.eps}

1186: \caption{\label{f:fit} Standard deviations when fitting the Weyl law \eqref{e:FWL} 

1187: to various dimensions, integrated on $0.1\leq r\leq 1$. The two marks on the horizontal

1188: axis indicate the theoretical values $d_I$ and $d_0$.}

1189: \end{center}

1190: \end{figure}

1191: %%%%%%%%%%%%%%%%%%%%

1192: %%%%%%%%%%%%%%%%%%%%

1193: \begin{figure}[htbp]

1194: \begin{center}

1195: \includegraphics[width=10cm]{spect-asymB-unscaled-BW.eps}

1196: \includegraphics[width=10cm]{spect-asymB-rescaled-BW2.eps}

1197: \caption{\label{f:asymB}Top: spectral counting function for the asymmetric baker $B_{\vr_{asym}}$, 

1198: for various values of Planck's constant $N$. Bottom: same curves vertically

1199: rescaled by $N^{-d_0}$. The thick tick mark indicates the radius $\sqrt{\Lambda_{nat}}$ 

1200: corresponding to the natural measure.}

1201: \end{center}

1202: \end{figure}

1203: %%%%%%%%%%%%%%%%%%%%

1204: In the proofs for upper bounds of the Weyl law \eqref{e:Weyl-law}, 

1205: the exponent $d$ is defined in terms of the 

1206: {\em  upper Minkowski dimension} of the trapped set $K$ \cite{Sjo90,Zw99,GLZ04,SjoZw05}. 

1207: In the case of the open baker $B_{\vr}$, we therefore expect that

1208: the exponent $d$ appearing in the conjecture (resp. $d+1$) is given by the

1209: Minkowski dimension of the Cantor set $\cC_{\vr}$ (resp. $\Gamma_+$), which is equal to

1210: its box, Hausdorff and packing dimensions. We call this theoretical

1211: value $d_0$.

1212: 

1213: For the symmetric baker $B_{\vr_{sym}}$, the Hausdorff dimension $d_H(\Gamma_+)=d_0+1$ 

1214: happens to be equal to the Hausdorff dimension of the

1215: natural measure $\mu_{nat}$, defined by

1216: $$

1217: d_H(\mu_{nat})=\inf_{A\subset\t2,\ \mu_{nat}(A)=1} d_H(A)\,.

1218: $$

1219: For a nonsymmetric baker's map $B_{\vr}$, those two Hausdorff dimensions 

1220: only satisfy the inequality

1221: $d_H(\mu_{nat})\leq d_H(\Gamma_+)$ (see the explicit expressions below).

1222: We want to investigate a possible ``role'' of the natural eigenmeasure $\mu_{nat}$

1223: regarding the structure of the quantum spectrum. It is therefore legitimate to

1224: ask the following 

1225: \begin{question}\label{qu0}

1226: Is the correct exponent in the Weyl law \eqref{e:FWL}

1227: given by $d_I=d_H(\mu_{nat})-1$ instead of $d_0=d_H(\Gamma_+)-1$?

1228: \end{question}

1229: Here the suffix $I$ indicates that $d_I$ is sometimes called

1230: the {\em information dimension} \cite{HJPPS86}.

1231: As mentioned above, both dimensions are equal for the symmetric baker $B_{\vr_{sym}}$,

1232: for which the Weyl law \eqref{conj:FWL} has been numerically tested in \cite{NonZw05}.

1233: They are also equal in the case of a closed map on $\t2$:

1234: in that case, the Weyl law has exponent $1$ (the whole spectrum

1235: lies on the unit circle), and we have $d_H(\t2)=d_H(\mu_{L})=2$. 

1236: 

1237: For a nonsymmetric baker $B_{\vr}$, the two dimensions

1238: take different values:

1239: $$

1240: d_0\quad \text{is the solution of}\quad r_0^d+r_2^d=1,\qquad\text{while}\quad

1241: d_I=\frac{r_0\log p_0+r_2\log p_2}{r_0\log r_0+r_2\log r_2}\,.

1242: $$

1243: To answer the above question, we

1244: considered a very asymmetric baker, taking $\vr_{asym}$ with $r_0=1/32$, $r_2=2/3$.

1245: The two dimensions then

1246: take the values $d_0\approx 0.493$, $d_I\approx 0.337$. We computed the counting function

1247: $n(N,r)$ for several radii $0.1\leq r\leq 1$ and several values of $N$. We then tried

1248: to fit the Weyl law \eqref{e:FWL} with an exponent $d$ varying in a certain range,

1249: and computed the standard deviations (see Fig.~\ref{f:fit}). 

1250: The numerical result is unambiguous: the best fit clearly occurs away from $d_I$, 

1251: but it is close to $d_0$. This numerical test rules out the possibility 

1252: that $d_I$ provides the correct exponent of the Weyl law, and suggests to 

1253: indeed take $d=d_0$.

1254: 

1255: To further illustrate the Weyl law for the asymmetric baker $B_{\vr_{asym}}$,  

1256: we plot in Fig.~\ref{f:asymB} (top) the counting functions

1257: $n(N,r)$ as a function of $r\in(0,1)$, for several values of $N$.

1258: On the bottom plot, we rescale $n(N,r)$ by the power $N^{-{d_0}}$: 

1259: the rescaled curves almost perfectly

1260: overlap, indicating that the scaling \eqref{e:FWL} is correct.

1261: \begin{rem}\label{r:fill}

1262: On figure~\ref{f:asymB} (right) the rescaled 

1263: counting function seems to converge to a function which is strictly decreasing

1264: on an interval $[\lambda_{\min},\lambda_{\max}]$, where $\lambda_{\min}\approx 0.1$, 

1265: $\lambda_{\max}\approx 0.9$. This implies that the spectrum of $B_{\vr_{asym},N}$ becomes

1266: dense in the whole annulus $\set{\lambda_{\min}\leq|\lambda|\leq\lambda_{\max}}$, 

1267: when $N\to\infty$. 

1268: Therefore, at this heuristic level, 

1269: for any $\lambda\in [\lambda_{\min},\lambda_{\max}]$

1270: one may consider sequences of 

1271: eigenvalues $(\lambda_N)_{N\geq 1}$ with the property $|\lambda_N|\Nto8 \lambda$.

1272: In particular, we may consider sequences converging to $\sqrt{\Lambda_{nat}}$.

1273: \end{rem}

1274: 

1275: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1276: \section{Localization of resonant eigenstates}\label{s:localization}

1277: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1278: 

1279: We now return to the general framework of an open map $T$

1280: on a subset $M$ of a compact phase space $\hM$, which is

1281: quantized into a sequence of operators $(T_N)_{N\to\infty}$ according to Axioms~\ref{d:axiom}.

1282: We want to study the semiclassical measures associated with the long-living eigenstates of $T_N$.

1283: 

1284: To start with, we fix some $\lambda_m\in(0,1)$, such that for $N$ large enough, 

1285: $\Spec T_N\cap \set{|\lambda|\geq \lambda_m}\neq \emptyset$. We

1286: can then consider sequences of eigenstates $(\psi_N)_{N\to\infty}$ such that, for $N$

1287: large enough, 

1288: \bequ\label{e:eigenstates}

1289: T_N\,\psi_N=\lambda_N\,\psi_N\,,\qquad \norm{\psi_N}=1\,,\qquad 

1290: |\lambda_N|\geq \lambda_{m}\,.

1291: \eequ

1292: The role of the (quite arbitrary) lower bound $\lambda_{m}>0$ is to ensure 

1293: that the eigenstates we consider are ``long-living''.

1294: Up to extracting a subsequence, we can assume that

1295: $(\psi_N)$ converges to a certain semiclassical measure $\mu$ 

1296: (see \S\ref{s:semicl-measure}). 

1297: 

1298: \medskip

1299: 

1300: To state our result, we first

1301: we need to analyze the continuity sets of the backward iterates of $T$. 

1302: For any $n\geq 1$, the map $T^{-n}$ is defined on the set $M^{-n}$. Its 

1303: continuity set $C(T^{-n})$ can be obtained iteratively through 

1304: $$

1305: C(T^{-n})=T\big(C(T^{-n+1})\cap C(T)\big)\,,\quad n\geq 2\,.

1306: $$

1307: The set $M_{-n}$ of points escaping exactly at 

1308: time $(-n)$ can also be split between its continuity subset

1309: its continuity subset

1310: $$

1311: CM_{-n}\defeq C(T^{-n+1})\cap \ring{M}_{-n}=T^{n-1}(C(T^{n-1})\cap \ring{H}^{-1})\,,

1312: $$

1313: which is a union of open connected sets, and its discontinuity subset

1314: $DM_{-n}=M_{-n}\setminus CM_{-n}$, which has Minkowski content zero (in the case $n=1$,

1315: we take $CM_{-1}=\ring{H}^{-1}$).

1316: Using this splitting of the $M_{-n}$, the decomposition \eqref{e:split2} can be recast into:

1317: \bequ\label{e:decompo-final}

1318: \hM=C_{-\infty}\sqcup D_{-\infty} \sqcup\Gamma_+,\quad\text{where}\quad

1319: C_{-\infty}=\big(\bigsqcup_{n=1}^\infty CM_{-n}\big),\quad

1320: D_{-\infty}=\big(\bigsqcup_{n=1}^\infty DM_{-n}\big)\,.

1321: \eequ

1322: The set $C_{-\infty}$ consists of points

1323: which eventually fall in the hole when evolved through $T^{-1}$, 

1324: and remain at finite distance from $D(T^{-1})$ all along their transient trajectory.

1325: 

1326: We can now state our main result concerning semiclassical measures.

1327: %%%%%%%%%%%%%

1328: \begin{thm}\label{thm:main}

1329: Assume that a sequence of eigenstates $(\psi_N)_{N\to\infty}$ of the 

1330: open quantum map $T_N$, with eigenvalues $|\lambda_N|\geq\lambda_m>0$,

1331: converges to the semiclassical measure $\mu$ on $\hM$. Then the following hold.

1332: 

1333: i) the support of $\mu$ is a subset of $\Gamma_+\sqcup D_{-\infty}$.

1334: 

1335: ii) If $\mu\big(C(T^{-1})\big)>0$, there exists $ \Lambda\in [\lambda_{m}^2,1]$ such

1336: that the eigenvalues $(\lambda_N)_{N\to\infty}$ satisfy

1337: $$

1338: |\lambda_N|^2\xrightarrow{N\to\infty} \Lambda\,.

1339: $$ 

1340: For any Borel subset $S$ not intersecting 

1341: $D(T^{-1})$, one has

1342: $\mu(T^{-1}(S))=\Lambda\,\mu(S)$. 

1343: 

1344: iii) If $\mu(D(T^{-1}))=\mu\big(T^{-1}(D(T^{-1}))\big)=0$, 

1345: then $\mu$ is an eigenmeasure of $T$, with decay rate $\Lambda$. 

1346: \end{thm}

1347: %%%%%%%%%%%%%

1348: \begin{proof}

1349: To prove the first statement, let us choose some $n\geq 1$, and take an observable

1350: $a\in C^\infty_c(CM_{-n})$. Every point $x\in\supp a$ has the property that

1351: for any $0\leq j <n-1$, $T^{-j}(x)\in C(T^{-1})$, while $T^{-n+1}(x)\in \ring{H}^{-1}$.

1352: Applying iteratively the property \eqref{e:axiom2-map}, one finds that in the 

1353: semiclassical limit,

1354: $$

1355: (T_N^{\dagger})^{n}\,\Op_N(a)\,(T_N)^{n}\sim 0\,.

1356: $$

1357: We take into account the fact that $\psi_N$ is a right eigenstate of $T_N$:

1358: $$

1359: \la \psi_N,\,\Op_N(a)\,\psi_N\ra=

1360: |\lambda_N|^{-2n}\,\la \psi_N,(T_N^{\dagger})^{n}\,\Op_N(a)\,(T_N)^{n}\,\psi_N\ra\,.

1361: $$

1362: Using $|\lambda_N|\geq \lambda_m$, these two expressions imply

1363: $\mu(a)\defeq \int a\,d\mu=0$.

1364: Since $n$ and $a$ were arbitrary, this shows $\mu(C_{-\infty})=0$, which is the first statement.

1365: 

1366: From the assumption in $ii)$, we may select $a\in C^\infty_c(C(T^{-1}))$ 

1367: such that $\int a\,d\mu>0$. The first iterate $a\circ T$ is supported in $C(T)\subset M$.

1368: Applying \eqref{e:axiom2-map}, we get

1369: \bequ\label{e:eigenstate}

1370: |\lambda_N|^{2}\,\la\psi_N,\,\Op_N(a)\,\psi_N\ra \sim 

1371: \la \psi_N,\,\Op_N(a\circ T)\,\psi_N\ra\,.

1372: \eequ

1373: For $N$ large enough, the matrix element

1374: $\la\psi_N,\,\Op_N(a)\,\psi_N\ra > \mu(a)/2$, so we may divide the

1375: above equation by this element, and obtain

1376: $$

1377: |\lambda_N|^2\Nto8 \frac{\mu(a\circ T)}{\mu(a)}\,.

1378: $$

1379: We call $\Lambda\geq |\lambda_m|^2$ this limit, which is obviously 

1380: independent of the choice of $a$. For any Borel set

1381: $S\subset C(T^{-1})$, one can approximate $\bbbone_{S}$ by smooth functions supported

1382: in $C(T^{-1})$, to prove that 

1383: $$

1384: \mu(T^{-1}(S))=\Lambda\,\mu(S)\,.

1385: $$

1386: If $S\subset \ring{H}^{-1}$, we know that $\mu(S)=0$ and $T^{-1}(S)=\emptyset$ ,

1387: so the above equality still makes sense. This proves $ii)$.

1388: 

1389: To obtain $iii)$, we

1390: split any Borel set $S$ into $S=(S\cap D(T^{-1}))\sqcup CS$.

1391: From $ii)$, we have $\mu(T^{-1}(CS))=\Lambda\,\mu(CS)$.

1392: By assumption, $\mu(S\cap D(T^{-1}))=\mu(T^{-1}(S\cap D(T^{-1})))=0$, so

1393: we get $\mu(T^{-1}(S))=\Lambda\,\mu(S)$.

1394: \end{proof}

1395: 

1396: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1397: \subsection{Semiclassical measures of the open baker's map} \label{s:semiclass}

1398: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1399: 

1400: In this section we apply the above theorem to the case of some open baker's map $B_{\vr}$,

1401: quantized as in \S\ref{s:quantum-baker2}. We also numerically compute the Husimi

1402: measures $H^1_{\psi_N}$ associated with some quantum eigenstates (we choose the isotropic

1403: squeezing $\ssigma=1$ by convenience): although we cannot really go to the semiclassical 

1404: limit, we hope that for $N\sim 1000$ these Husimi measures already give some idea of the

1405: semiclassical measures.

1406: 

1407: \subsubsection{Applying Theorem~\ref{thm:main}}

1408: 

1409: To simplify the presentation we will restrict the discussion to the symmetric baker 

1410: $B_{\vr_{sym}}$, which 

1411: will be denoted by $B$ in short. The discontinuity set $D(B^{-1})$, its backwards image 

1412: and the trapped sets were described in \S\ref{s:baker-trapped} and \S\ref{s:quantum-baker2}, and

1413: plotted in Fig.~\ref{f:gamma+}.

1414: The sets $M_{-n}$ and their continuity subsets are also simple to describe using symbolic dynamics.

1415: For any $n=1$, we know that $CM_{-1}=\ring{H}^{-1}$ is the open rectangle 

1416: $\set{q\in [0,1),\ p\in (1/3,2/3)}$. 

1417: For $n\geq 2$, 

1418: $CM_{-n}$ is the union of $2^{n-1}$ horizontal rectangles, indexed by the $n$-sequences 

1419: $\bep=\ep_{-1}\ldots\ep_{-n}$ such that $\ep_{-i}\in\set{0,2}$, $i=1,\ldots,n-1$, while

1420: $\ep_{-n}=1$. Each such rectangle is

1421: of the form $\set{q\in (0,1),\ p\in (p(\bep), p(\bep)+3^{-n})}$, where 

1422: $p(\bep)\equiv \cdot\,\ep_{-1}\ldots\ep_{n}$ in ternary decomposition. 

1423: Some of those  rectangles are shown in Fig.~\ref{f:Cantor} (center).

1424: The union of all these rectangles (for $n\geq 1$) makes up

1425: $C_{-\infty}$. Its complement $\Gamma_+\sqcup D_{-\infty}$ is given by

1426: $$

1427: \Gamma_+\sqcup D_{-\infty}=\big([0,1)\times \cC_{\vr_{sym}}\big)\cup D(B^{-1})\,.

1428: $$

1429: This set is shown in Fig.~\ref{f:gamma+} (left).

1430: %%%%%%%%%%%%%%%%%%%%

1431: \begin{figure}[htbp]

1432: \begin{center}

1433: \rotatebox{-90}{\includegraphics[width=.33\textwidth]{3states-nat.ps}}

1434: %\rotatebox{-90}{\includegraphics[width=6cm]{3states-sym.ps}}

1435: \rotatebox{-90}{\includegraphics[width=.33\textwidth]{3states-2maps.ps}}

1436: %\rotatebox{-90}{\includegraphics[width=6cm]{2states9-3.ps}}

1437: \caption{\label{f:eigenstates} Husimi densities of (right) 

1438: eigenstates of $B_{\vr,N}$ (black=large values, white=0).

1439: Top: 3 eigenstates of $B_{\vr_{sym},N}$ with $|\lambda_N|\approx \sqrt{\Lambda_{nat}}$.

1440: Bottom left, center: two eigenstates of $B_{\vr_{sym},N}$ with different $|\lambda_N|$. 

1441: Bottom right: one eigenstate of $B_{\vr_2,N}$.}

1442: \end{center}

1443: \end{figure}

1444: %%%%%%%%%%%%%%%%%%%%

1445: 

1446: In Figure~\ref{f:eigenstates}, we plot the Husimi densities $H^1_{\psi_N}$ of some

1447: (right) eigenstates of $B_{\vr,N}$, for the symmetric baker and 

1448: an asymmetric one, $\vr_{2}=(1/9,5/9,1/3)$.

1449: 

1450: \begin{rem}

1451: We notice that all Husimi functions are indeed

1452: very small in the horizontal rectangles $M_{-n}$ for $n=1,2$ (in case $N\leq 1500$),

1453: and also $n=3$ for $N=4200$. Using \eqref{e:CS-evol}, 

1454: one can refine the proof of Thm.~\ref{thm:main} to show that $H^1_{\psi_N}(x)=\cO(\e^{-cN})$ 

1455: for $x\in C_{-\infty}$, $N\to\infty$. 

1456: \end{rem}

1457: 

1458: \begin{rem}\label{r:rem2}

1459: Although this is not proven in our theorem, all the Husimi functions we have computed

1460: are very small on the set $D(B^{-1})\setminus \Gamma_+\subset \set{q=0}$

1461: (see the left plot in Fig.~\ref{f:gamma+}). 

1462: On the other hand, some of these Husimi functions are large on

1463: $D(B^{-1})\cap \Gamma_+$,

1464: so we cannot rule out the possibility that 

1465: some semiclassical measures $\mu$ charge this set. 

1466: For instance, in some of the Husimi plots ($e.g.$

1467: the bottom left in Fig.~\ref{f:eigenstates}), we clearly see

1468: a strong peak at the origin, and

1469: lower peaks on other points of $D(B^{-1})\cap \Gamma_+$.

1470: We call ``diffractive'' the components of an eigenstate localized on $D_{-\infty}$. 

1471: \end{rem}

1472: 

1473: From these observations, we state the following

1474: \begin{conj} \label{c:eigenm}

1475: Let $B_{\vr}$ be an open baker's map, quantized by \S\ref{s:quantum-baker2}. Then,

1476: \begin{itemize} 

1477: \item all long-living semiclassical measures are supported on 

1478: $\overline{\Gamma_+}$

1479: \item ``almost all'' long-living eigenstates are non-diffractive, 

1480: that is, their weight

1481: on $D(B^{-1})$ and $B^{-1}(D(B^{-1}))$ are negligible

1482: in the semiclassical limit. 

1483: The corresponding semiclassical measures are then eigenmeasures of $B_{\vr}$.

1484: \end{itemize}

1485: \end{conj}

1486: In the next section we further comment on the above Husimi plots.

1487: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1488: \subsection{Abundance of semiclassical measures}\label{s:abundance} 

1489: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1490: 

1491: Let us draw some consequences from Theorem~\ref{thm:main} and the following remarks.

1492: Statement $ii)$ of the theorem strongly constrains the 

1493: converging sequences of eigenstates: a sequence $(\psi_N)$ can

1494: converge to some measure $\mu$ (with $\mu(C(B^{-1})) > 0$) only if the

1495: corresponding eigenvalues $(\lambda_N)$ asymptotically approach the circle of radius 

1496: $\sqrt{\Lambda}$. 

1497: 

1498: We take for granted the density argument in Remark~\ref{r:fill} and assume that

1499: a ``dense interval'' $[\lambda_{\min},\lambda_{\max}]\subset (0,1]$

1500: exists for any open baker's map $B_{\vr}$. Hence,

1501: for any $\Lambda\in [\lambda_{\min}^2,\lambda_{\max}^2]$ there exist

1502: many sequences of eigenstates of $B_{\vr,N}$, 

1503: such that $|\lambda_N|^2\Nto8 \Lambda$. From our Conjecture~\ref{c:eigenm}, almost any

1504: semiclassical measure associated with such a sequence will be an eigenmeasure with

1505: decay rate $\Lambda$. Therefore, two converging sequences $(\psi_N)$, $(\psi_N')$

1506: associated with limiting decays $\Lambda\neq\Lambda'$ will necessarily converge

1507: to different eigenmeasures. This already shows that the semiclassical eigenmeasures

1508: generated by all possible sequences in the annulus 

1509: $\set{\lambda_{\min}\leq|\lambda|\leq\lambda_{\max}}$ form an uncountable family.

1510: 

1511: According to \S\ref{s:eigenmeasures}, for each decay rate $\Lambda\in (0,1)$,

1512: there exist uncountably many eigenmeasures. A natural question thus concerns the

1513: variety of semiclassical measures associated with a given 

1514: $\Lambda\in [\lambda_{\min}^2,\lambda_{\max}^2]$:

1515: 

1516: \begin{question}\label{qu1}$\ $

1517: 

1518: For a given $\Lambda\in [\lambda_{\min}^2,\lambda_{\max}^2]$, what are the

1519: semiclassical measures of $B_{\vr}$ of decay rate $\Lambda$?

1520: \begin{itemize}

1521: \item is there a unique such measure? 

1522: \item otherwise, is some limit measure  ``favored'', in the sense that ``almost all'' sequences

1523: $(\psi_N)$ with $|\lambda_N|^2\to \Lambda$ converge to $\mu$?

1524: \item can the {\em natural measure} $\mu_{nat}$ be obtained as a semiclassical measure?

1525: \end{itemize}

1526: \end{question}

1527: The same type of questions were asked by Keating and coworkers in ref.~\cite{KNPS06}. 

1528: At present we are unable to answer them rigorously for the quantum open baker $B_{\vr,N}$. 

1529: 

1530: From a heuristic point of view we notice the following features 

1531: in the plots of Fig.~\ref{f:eigenstates}. The three

1532: top plots correspond to eigenvalues $|\lambda_N|$ close to the value 

1533: $\sqrt{\Lambda_{nat}}\approx 0.8165$. However, the Husimi measures on the left and center

1534: seem very different from the natural measure (the latter is approximated by the black rectangles

1535: in Fig.~\ref{f:gamma+}). These two Husimi functions seem to charge different 

1536: parts of $\Gamma_+$. Only the rightmost state seems compatible with a convergence 

1537: to $\mu_{nat}$.

1538: 

1539: As commented in Remark~\ref{r:rem2}, the bottom-left state has strong concentrations on

1540: $D(B^{-1})\cap\Gamma_+$. For the various values of $N$ we have investigated, this 

1541: concentration seems characteristic of the eigenstate with the largest 

1542: $|\lambda_N|$. The discontinuities of $B_{\vr}$ manage to ``trap'' those quantum states 

1543: better than periodic orbits on $C(B^{-1})\cap\Gamma_+$. Such states seem ``very diffractive''.

1544: 

1545: Finally, the bottom-center state, with a smaller eigenvalue, clearly

1546: shows a selfsimilar structure along the horizontal direction, with a

1547: probability inside the hole higher than for the top eigenstates. 

1548: This feature had already been noticed in \cite{KNPS06} for averages over Husimi 

1549: functions with comparable $|\lambda_N|$.

1550: 

1551: In section~\ref{s:walsh} we will address Questions~\ref{qu1} for

1552: a different quantization of the symmetric open baker, namely 

1553: the Walsh-quantized open baker, where one can compute some semiclassical

1554: measures explicitly.

1555: 

1556: Before that, we explain how to construct approximate eigenstates

1557: ({\em pseudomodes}) for the quantum baker $B_{\vr_{sym},N}$.

1558: 

1559: %%%%%%%%%%%%%%%%%%%%%%%%%%

1560: \subsection{Pseusomodes and pseudospectrum} 

1561: %%%%%%%%%%%%%%%%%%%%%%%%%%

1562: 

1563: From the explicit representation of eigenmeasures given in Proposition~\ref{p:eigen-decompo},

1564: it is possible to construct approximate eigenstates of $T_N$ by 

1565: backward propagating wavepackets localized on the set $\Gamma_+^{(1)}=\Gamma_+\cap H$.

1566: 

1567: We call these approximate eigenstates 

1568: {\it pseudomodes}, by analogy with the recent literature on nonselfadjoint semiclassical

1569: operators (see e.g. \cite{DSZ04,BU03}). Those papers deal with pseudodifferential operators

1570: obtained by quantizing complex-valued observables, and they construct

1571: pseudomodes of error $\cO(\hbar^\infty)$, microlocalized at a single phase space point.

1572: 

1573: Our pseudomodes will be less precise and will be microlocalized on a countable set. 

1574: We will restrict ourselves to the case of the

1575: open symmetric baker $B=B_{\vr_{sym}}$, but our construction works for 

1576: any $B_{\vr}$, and can probably be extended to other maps or phase spaces.

1577: 

1578: Inspired by the pointwise eigenmeasures \eqref{e:mu-traject}, we construct 

1579: approximate eigenstates by backwards evolving coherent states localized in $\Gamma_+^{(1)}$.

1580: Precisely, for any $\bx_0\in \Gamma_+^{(1)}$, $\ssigma>0$ and $\lambda\in \IC$, $|\lambda|<1$, 

1581: $N\in\IN$, we define the following quantum state:

1582: \begin{equation}\label{e:quasim}

1583: \Psi^\ssigma_{\lambda,x_0}\defeq \sqrt{1-|\lambda|^2}\;

1584: \sum_{n\geq 0}\lambda^n\,B_{N}^{\dagger n}\,\psi_{\bx_0,\ssigma}\,,

1585: \end{equation}

1586: where $\psi_{\bx_0,\ssigma}$ is the coherent state defined in \S\ref{s:Husimi}, and $B_N$ is

1587: the standard quantization of $B$.

1588: %++++++++++++++++

1589: \begin{prop}

1590: Consider the symmetric open baker $B=B_{\vr_{sym}}$ and its quantization $(B_N)$.

1591: Fix $\lambda\in\IC$ with $|\lambda|<1$. 

1592: 

1593: $i)$ Choose $\vareps>0$ small and call $\alpha=(1-\vareps)\frac{\log 1/|\lambda|}{\log 3}$. 

1594: For any $N\in\IN^*$, 

1595: one can choose a point $x_0(N)\in \Gamma_+^{(1)}$ and a squeezing 

1596: parameter $\ssigma=\ssigma(N)>0$ such that the 

1597: states\\ $\big(\Psi_N\defi \Psi^{\ssigma(N)}_{\lambda,x_0(N)}\big)_{N\to\infty}$ defined

1598: in \eqref{e:quasim} satisfy

1599: $$

1600: \norm{\Psi_N}=1+\cO(N^{-\alpha}),\qquad 

1601: \norm{(B_N-\lambda)\Psi_N}=\cO(N^{-\alpha})\,,\qquad N\to\infty.

1602: $$

1603: $ii)$ for any $x_0\in \Gamma_+^{(1)}$, one can 

1604: select the points $\big(x_0(N)\big)$ such that

1605: $x_0(N)\Nto8 x_0$. In that case, the sequence $(\Psi_N)_{N\to\infty}$

1606: converges to the eigenmeasure $\mu_{x_0,\Lambda}$ described in \eqref{e:mu-traject},

1607: with $\Lambda=|\lambda|^2$.

1608: \end{prop}

1609: %++++++++++++++++

1610: This proposition shows that, for

1611: any $\alpha>0$ and $N$ large enough,

1612: the $N^{-\alpha}$-pseudospectrum of $B_N$ contains the disk 

1613: $\set{|\lambda|\leq 3^{-\alpha(1+\vareps)}}$.

1614: We notice that the errors are not very small, and increase when $|\lambda|\to 1$. We should

1615: emphasize that, in our numerical trials, we never found eigenstates of $B_N$ with

1616: Husimi measures looking like $\mu_{x_0,\Lambda}$.

1617: 

1618: \subsubsection*{Proof of the proposition}

1619: To control the series \eqref{e:quasim}, 

1620: we would like to ensure that the evolved

1621: coherent state $B_{N}^{\dagger j}\,\psi_{x_0(N),\ssigma(N)}$

1622: remains close to an approximate coherent state 

1623: $\psi_{x_{-j},\ssigma_{-j}}$ (as in \eqref{e:CS-evol})

1624: up to large times $j$. 

1625: For this we need the points $x_{-j}=B^{-j}(x_0(N))$ to stay ``far'' 

1626: from the discontinuity set

1627: $D(B^{-1})$, and we also need all $\psi_{x_{-j},\ssigma_{-j}}$ to be microlocalized

1628: in a small neighbourhood of $x_{-j}$. 

1629: Due to the hyperbolicity of $B^{-1}$, the second condition constrains

1630: the times $j$ to be smaller than the {\it Ehrenfest time} \cite{DENW06}

1631: \bequ\label{e:Ehrenf}

1632: n_E=(1-\vareps)\frac{\log N}{\log 3}\,.

1633: \eequ

1634: Here $\vareps>0$ is the small parameter in the statement of the proposition.

1635: 

1636: To identify a good ``starting point'' $x_0(N)$, we set $n=[n_E]$, 

1637: and consider the following subset of $\t2$, for $\delta,\gamma>0$:

1638: $$

1639: \cD_{n,\delta,\gamma}\defi\set{(q,p)\in\t2,\ q\in (1/3+\delta,2/3-\delta),\ 

1640: \forall k\in\IZ,\ |p-\frac{k}{3^n}|>\gamma}\cap\Gamma^{(1)}_+\,.

1641: $$

1642: We first show that this set is nonempty if $\delta<1/9$ and 

1643: $\gamma=3^{-n}\gamma'$, $\gamma'<1/9$.

1644: Take $q\in (1/3+\delta,2/3-\delta)$ arbitary, and select $p=p(\bep)$ with 

1645: the following properties: take all indices $\ep_{-j}\in\{0,2\}$, $j\geq 1$, 

1646: so that $p\in \cC_{\vr_{sym}}$,

1647: and require furthermore that the word $\ep_{-n-1}\ep_{-n-2}\in\set{02,20}$. 

1648: The point $(q,p)$ is then in $\cD_{n,\delta,\gamma}$. 

1649: 

1650: Let us take any point $x_0(N)$ in that set. It automatically lies in 

1651: $C(B^{-n})$, so its backwards iterates stay away from 

1652: $D(B^{-1})$ at least until the time $n$.

1653: Let us select the squeezing parameter

1654: $\ssigma(N)=s_0=N^{-1+\vareps}$. A simple adaptation

1655: of \cite[Prop.5]{DENW06} implies the following estimate:

1656: \bequ\label{e:semiclass1}

1657: \exists c,C>0,\ \ \forall j,\  0\leq j\leq n_E,\qquad

1658: \norm{B_{N}^{\dagger j}\,\psi_{\bx_0,\ssigma_0}-\e^{\i\theta_j}\,\psi_{x_{-j},\ssigma_{-j}}}\leq

1659: C\,\e^{-c\,N^\vareps}\,.

1660: \eequ

1661: Here we can take $c=\min(\delta^2,\gamma^{\prime 2})$, and $C$

1662: is uniform w.r.to the initial point $x_0$.

1663: From the values of $x_0$, $\ssigma_0$, one checks that 

1664: the components 

1665: $\la \bq_k,\psi_{x_0,\ssigma_0}\ra$ for $k/N\not\in [1/3,2/3]$ 

1666: are of order $\cO(\e^{-cN^\vareps})$: 

1667: that state is (very) localized in the hole $H=R_1$, so

1668: $$

1669: B_{N}\,\psi_{\bx_{0},\ssigma_{0}}= \cO(\e^{-c\,N^\vareps})\,.

1670: $$

1671: Similarly,

1672: for any $j\leq n_E$, the state $\psi_{x_{-j},\ssigma_{-j}}$ is localized inside a certain

1673: connected component of $M_{j+1}$ (one of the the pink/grey rectangles in Fig.~\ref{f:Cantor}, left). 

1674: In particular, the components of 

1675: $\la \bq_k,\psi_{x_{-j+1},\ssigma_{-j+1}}\ra$ are exponentially small in $H$. Therefore,

1676: $$

1677: \forall j,\  0\leq j\leq n_E,\qquad 

1678: B_{N}\,B^\dagger_{N}\psi_{\bx_{-j},\ssigma_{-j}}= 

1679: \psi_{\bx_{-j},\ssigma_{-j}}+\cO(\e^{-c\,N^\vareps})\,.

1680: $$

1681: Summing all terms $j\leq n_E$ in \eqref{e:quasim} 

1682: and estimating the remaining series using $\norm{B_N}\leq 1$, we obtain

1683: $$

1684: \norm{(B_{N}-\lambda)\,\Psi^\ssigma_{\lambda,x_0}}= \cO(|\lambda|^{n_E})\,.

1685: $$

1686: From the definition of $n_E$, we have $|\lambda|^{n_E}=N^{-\alpha}$.

1687: 

1688: The asymptotic normalization of $\Psi^\ssigma_{\lambda,x_0}$ is proven by estimating the

1689: overlaps between coherent states $\psi_{x_{-j},\ssigma_{-j}},\psi_{x_{-j'},\ssigma_{-j'}}$

1690: for $j,j'\leq n_E$. Because the sets $M_{j+1}$

1691: and $M_{j'+1}$ are disjoint for $j\neq j'$, the above mentioned localization

1692: properties imply that

1693: $$

1694: \forall j,j'\leq n_E,\qquad

1695: \la \psi_{x_{-j},\ssigma_{-j}},\psi_{x_{-j'},\ssigma_{-j'}}\ra=\delta_{j',j}+\cO(\e^{-c\,N^\vareps}),

1696: $$

1697: and the normalization estimate follows.

1698: This achieves the proof of $i)$.

1699: 

1700: To prove $ii)$, let us consider an arbitrary $x_0=(q_0,p_0)\in \Gamma^{(1)}_+$. 

1701: If $q_0\not\in\set{1/3,2/3}$,

1702: we take $\delta$ small enough such that $q_0\in (1/3+\delta,2/3-\delta)$ and set $q_0(N)=q_0$.

1703: Otherwise, we may let $\delta$ slowly decrease with $N$, and find 

1704: a sequence $\big(q_0(N)\big)$ such

1705: that $q_0(N)\in (1/3+\delta(N),2/3-\delta(N))$ and $q_0(N)\Nto8 q_0$.

1706: On the other hand, 

1707: $p_0\in\cC_{\vr_{sym}}=p(\bep)$ for a certain sequence $\bep$, with all 

1708: $\ep_{-j}\in\set{0,2}$. For each $N$, we

1709: inspect the word $\ep_{-n-1}\ep_{-n-2}$ of that sequence, where $n=[n_E]$ (see \eqref{e:Ehrenf}). 

1710: If this word is in the set $\set{02,20}$ we keep $p_0(N)=p_0$, otherwise we 

1711: replace this word by $02$ (keeping the other symbols unchanged) 

1712: to define $p_0(N)$. The point $x_0(N)=(q_0(N),p_0(N))$ is then in 

1713: $\cD_{n,\delta,\gamma}$, and 

1714: $x_0(N)\Nto8 x_0$.

1715: 

1716: The convergence to the measure $\mu_{x_0,\Lambda}$ is due to

1717: the localization of the coherent states $\psi_{x_{-j}, s_{-j}}$ for $j\leq n_E$.

1718: 

1719: $\hfill\square$

1720: 

1721: 

1722: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1723: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1724: \section{A solvable toy model: Walsh quantization of the open baker}\label{s:walsh}

1725: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1726: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1727: In this section we study an alternative quantization of the open symmetric

1728: baker $B_{\vr_{sym}}$, introduced in \cite{NonZw05,NonZw06}. 

1729: A similar quantization

1730: of the closed baker $\hB_{\vr_{sym}}$ was proposed and 

1731: studied in \cite{SchaCav00,TraSco02,ErmSara06},

1732: mainly motivated by research in quantum computation.

1733: From now on, we will drop the

1734: index $\vr_{sym}$ from our notations, and call $B=B_{\vr_{sym}}$.

1735: 

1736: A ``simplified'' quantization of $B$ was introduced in \cite{NonZw05,NonZw06}, 

1737: as a $N\times N$ matrix $\tB_N$, obtained by only keeping the ``skeleton''  

1738: of $B_N$ (see \eqref{e:B_N}). Although $\tB_N$ can be defined for any 

1739: $N$, its spectrum can be explicitly computed only 

1740: if $N=3^k$ for some $k\in\IN$.

1741: As explained in those references, $\tB_N$ can be interpreted as

1742: the ``Weyl'' quantization of a  

1743: multivalued map $\tB$ built upon $B$. In the present work we will stick to a different

1744: interpretation of $\tB_N$, valid in the case $N=3^k$: one can then introduce

1745: a (Walsh) quantization for observables on $\t2$, which is not equivalent with

1746: the anti-Wick quantization of \S\ref{s:Husimi}. We will check below that

1747: the matrices $\tB_N$ satisfies property \eqref{e:axiom2-map} with respect to that

1748: quantization and the open baker $B$. 

1749: Finally, this Walsh

1750: quantization is also suited to quantize observables on

1751: the symbol space $\Sigma$: the matrices $\tB_N$ are then quantizing the

1752: open shift $\sigma$. We will see that this interpretation is more ``convenient'', because

1753: it avoids problems due to discontinuities. 

1754: 

1755: We first recall the definition of the

1756: Walsh quantization of observables on $\t2$ (or $\Sigma$), and the associated

1757: Walsh-Husimi measures.

1758: 

1759: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1760: \subsection{Walsh transform and coherent states} \label{s:walsh0}

1761: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1762: 

1763: \subsubsection{Walsh coherent states}

1764: The Walsh quantization of observables on $\t2$ uses the decomposition of the quantum

1765: Hilbert space $\hn$ into a tensor product of ``qubits'', namely

1766: $\hn=(\IC^3)^{\otimes k}$ (clearly, this makes sense only for $N=3^k$). 

1767: Each discrete position $q_j=j/N$ can be represented by its ternary sequence

1768: $q_j= 0\cdot\, \ep_0\ep_2\cdots\ep_{k-1}$, with symbols $\ep_i\in\set{0,1,2}$. 

1769: Accordingly, each position eigenstate $\bq_j$ can be represented as a tensor product:

1770: $$

1771: \bq_j=e_{\ep_0}\otimes e_{\ep_2}\otimes\ldots\otimes e_{\ep_{k-1}}

1772: =|[\ep_0\ep_2\ldots\ep_{k-1}]_k\ra\,,

1773: $$

1774: where $\set{e_0,\,e_1,\,e_2}$ is the canonical (orthonormal) basis of $\IC^3$. The notation

1775: on the right hand side emphasizes the fact that this state is 

1776: associated with the cylinder $[\bep]=[\bep]_k$

1777: with $k$ symbols on the right of the comma, no symbol on the left. Its

1778: image on $\t2$ is a rectangle $[\bep]_k$ of height unity and width $3^{-k}=1/N$.

1779: 

1780: Walsh quantization consists in replacing the discrete Fourier transform

1781: \eqref{e:DFT} on $\hn$ by the Walsh(-Fourier) transform $W_{N}$, which is a unitary operator

1782: preserving the tensor product structure of $\hn$. 

1783: We define it through its inverse $W_{N}^*$, which maps the position basis to the orthonormal basis 

1784: of ``Walsh momentum states'': for any $j\equiv \ep_0\ldots\ep_{k-1}$, 

1785: $$

1786: \bp_j= W_{N}^*\,\bq_j\defeq F^*_3 e_{\ep_{k-1}}\otimes F^*_3 e_{\ep_{k-2}}\otimes\ldots

1787: \otimes F^*_3 e_{\ep_1}\otimes F^*_3 e_{\ep_0}

1788: $$

1789: (here $F_3^*$ is the inverse Fourier transform on $\IC^3$).

1790: To agree with our notations of \S\ref{s:open-baker}, we will index the symbols relative to the

1791: momentum coordinate by {\em negative} integers, so the Walsh momentum states will be denoted by

1792: $$

1793: \bp_j=|[\ep_{-k}\ldots\ep_{-1}]_0\ra\,,\quad\text{where}\quad 

1794: p_j=j/N=0 \cdot\,\ep_{-1}\ldots \ep_{-k}\,.

1795: $$

1796: This momentum state is associated with a rectangle of

1797: height $3^{-k}$ and width unity. 

1798: 

1799: More generally, for any $\ell\in\set{0,\ldots,k}$ and any sequence

1800: $\bep=\ep_{-k+\ell}\ldots\ep_{-1}\cdot\ep_0\ldots\ep_{\ell-1}$, one can construct

1801: a (Walsh-)coherent state

1802: \bequ\label{e:Walsh-CS}

1803: |[\bep]_{\ell}\ra  \defeq 

1804: e_{\ep_0}\otimes\ldots e_{\ep_{\ell-1}}\otimes 

1805: F_3^* e_{\ep_{-k+\ell}}\otimes\ldots F_3^* e_{\ep_{-1}} \,.

1806: \eequ

1807: This state is localized in the rectangle $[\bep]_\ell$ with height $3^{-k+\ell}$ and

1808: width $3^{-\ell}$, so it still has area $1/N$ 

1809: (all such rectangles are minimal-uncertainty, or ``quantum'' rectangles).

1810: 

1811: Like the squeezing parameter $\ssigma$ of Gaussian wavepackets (see \S\ref{s:Husimi}),

1812: the index $\ell$ describes the aspect ratio of the coherent state. When going

1813: to the semiclassical limit $k\to\infty$, we will

1814: always select $\ell(k)\sim k/2$, which corresponds to an ``isotropic'' squeezing. 

1815: One important

1816: difference between Gaussian and Walsh coherent states lies in the fact that the latter are

1817: {\em strictly} localized both in momentum and position. Another difference is that, for

1818: each $\ell$,  

1819: the $\ell$-coherent states make up a finite orthonormal basis of $\hn$, 

1820: instead of a continuous overcomplete family.

1821: 

1822: %==========================================

1823: \subsubsection{Walsh quantization for observables on $\Sigma$}\label{s:Lipschitz}

1824: %==========================================

1825: 

1826: As explained in \cite{AN06}, it is more natural to Walsh-quantize observables

1827: on the symbol space $\Sigma$ than on $\t2$.

1828: Indeed, if one equips $\Sigma$ with the metric structure

1829: $$

1830: d_{\Sigma}(\bep,\bep')=\max(3^{-n_+},3^{-n_-}),\quad n_+=\min\set{n\geq 0,\,\ep_n\neq \ep'_n},\ 

1831: n_-=\min\set{n\geq 0,\,\ep_{-n-1}\neq \ep'_{-n-1}}\,,

1832: $$

1833: the closed shift $\hat\sigma$ then acts as a {\it Lipschitz} map on $\Sigma$, and

1834: the hole $\set{\ep_0=1}$ is at finite distance from its complement in $\Sigma$.

1835: Indeed, the ``lift'' from $\t2$ to $\Sigma$ has the effect to

1836: ``blow up'' the lines of discontinuity of $B$.

1837: 

1838: The conjugacy $J:\Sigma\to\t2$ is also Lipschitz, so

1839: any Lipschitz function $F\in Lip(\t2)$ is pushed to 

1840: a function $f=F\circ J \in Lip(\Sigma)$. However, the converse is not true:

1841: if we use the inverse map $(J_{|\Sigma'})^{-1}:\t2\to\Sigma'$ and take an arbitrary

1842: $f\in Lip(\Sigma)$, the function $F= f\circ (J_{|\Sigma'})^{-1}$

1843: is generally discontinuous on $\t2$.

1844: 

1845: Let us select some $\ell\sim k/2$. The Walsh-quantization of a function $f\in Lip(\Sigma)$ is 

1846: defined as the following operator on $\hn$:

1847: \bequ\label{e:Walsh-quantiz}

1848: \Op_N^\ell(f)\defeq 3^k \sum_{[\bep]_{\ell}} 

1849: |[\bep]_{\ell}\ra \la[\bep]_{\ell}|\;\int_{[\bep]_{\ell}}f\,d\mu_L

1850: \eequ

1851: Here the sum goes over all ``quantum'' cylinders $[\bep]_{\ell}$, that is

1852: over all $3^k$ sequences $\bep=\ep_{-k+\ell}\ldots\ep_{-1}\cdot\ep_0\ldots\ep_{\ell-1}$.

1853: The integral over $f$ is performed using the uniform Bernoulli measure $\mu_L$, which

1854: is equivalent to the Liouville measure on $\t2$ (see the end of \S\ref{s:open-baker}).

1855: Notice the formal similarity of this quantization with the anti-Wick 

1856: quantization \eqref{e:anti-Wick}.

1857: 

1858: It is shown in \cite[Prop.3.1]{AN06} that this quantization satisfies the 

1859: Axioms~\ref{d:axiom-obs} (with Lipschitz observables), and that 

1860: two quantizations $\Op_N^{\ell_1}$, $\Op_N^{\ell_2}$ are semiclassically 

1861: equivalent if both $\ell_1,\,\ell_2\sim k/2$.

1862: 

1863: By duality, we define the Walsh-Husimi measure of a quantum state $\psi_N$. The corresponding 

1864: density is constant in each $\ell$-cylinder $[\bep]_{\ell}$:

1865: \bequ\label{e:WH}

1866: \forall \bal\in[\bep]_{\ell}\,,\qquad WH^\ell_{\psi_N}(\bal)=3^k\,|\la [\bep]_\ell , \psi_N\ra|^2\,.

1867: \eequ

1868: This density is originally defined on $\Sigma$, but can be 

1869: pushed-forward to $\t2$.

1870: Semiclassical measures

1871: are defined as the weak-$*$ limits of sequences $(WH^\ell_{\psi_N})$,

1872: where $N=3^k\to\infty$, and $\ell=\ell(k)\sim k/2$. One first obtains a measure $\mu^\Sigma$ 

1873: on $\Sigma$, which can be pushed on $\t2$ into $\mu=J^*\mu^\Sigma$ (equivalently,

1874: $\mu$ is the weak-$*$ limit of the Walsh-Husimi measures on $\t2$).

1875: 

1876: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1877: \subsection{Walsh quantization of the open baker}\label{s:Walsh-baker} 

1878: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1879: 

1880: We now recall the Walsh quantization of the open baker 

1881: $B=B_{\vr_{sym}}$, as defined in \cite{NonZw05,NonZw06}.

1882: Mimicking the standard quantization \eqref{e:B_N},

1883: we replace the Fourier transforms $F_{N}^*$, $F_{N/3}$ by their Walsh analogues

1884: $W_N$, $W_{N/3}$ (with $N=3^k$, $k\geq 0$), so that the Walsh-quantized open baker

1885: is given by the following matrix in the position basis:

1886: \bequ

1887: \widetilde B_{N}\defeq W_N^*\begin{pmatrix}W_{N/3}&&\\&0&\\&&W_{N/3}\end{pmatrix}\,.

1888: \eequ 

1889: For any set of vectors $v_0,\,\ldots\, v_{k-1}\in \IC^3$, this operator acts as follows

1890: on any tensor product state $v_0\otimes\ldots\otimes v_{k-1}$:

1891: \bequ\label{e:B_k}

1892: \widetilde B_{N}\big(v_0\otimes v_1\ldots\otimes v_{k-1}\big) =

1893: v_1\otimes\ldots\otimes v_{k-1}\otimes \widetilde F_3^* v_0\,.

1894: \eequ

1895: Here $\widetilde F_3^*=F_3^*\,\pi_{02}$, where $\pi_{02}$ is the orthogonal projector on 

1896: $\IC e_0\oplus \IC e_2$ in $\IC^3$.

1897: 

1898: One can generalize the quantum-correspondence

1899: of \cite[Prop.3.2]{AN06} to the open shift $\sigma$, and prove the Walsh version of

1900: Axioms~\ref{d:axiom}:

1901: %================

1902: \begin{prop}\label{p:Walsh-Egorov}

1903: 

1904: $i)$ Take any $f\in Lip(\Sigma)$. Then, in the limit $N=3^k\to\infty$, $\ell\sim k/2$, 

1905: $$

1906: \tB_N^\dagger\,\Op^\ell_N(f)\,\tB_N \sim  

1907: \Op^{\ell}_N\big(\bbbone_{\set{\ep_0\neq 1}}\times f\circ\sigma\big)\,.

1908: $$

1909: Notice that the function 

1910: $\bbbone_{\set{\ep_0\neq 1}}\times f\circ\sigma\in Lip(\Sigma)$.

1911: 

1912: $ii)$ If we take $F\in Lip_c(C(B^{-1})\sqcup \ring{H}^{-1})$ and $f=F\circ J$, we have

1913: $$

1914: \bbbone_{\set{\ep_0\neq 1}}\times f\circ\sigma=\big(\bbbone_{M}\times F\circ B\big)\circ J\,.

1915: $$

1916: \end{prop}

1917: %================

1918: The points $i)-ii)$ show that the family $(\tB_N)$

1919: satisfies the Axioms~\ref{d:axiom} with respect to the map $B$ on $\t2$ and

1920: the quantization \eqref{e:Walsh-quantiz} of observables in $Lip(\t2)$. 

1921: The point $i)$ alone shows that $(\tB_N)$ is a quantization 

1922: of the open shift $\sigma$ on $\Sigma$. As noticed above,

1923: the latter interpretation allows to get rid of problems of discontinuities.

1924: \begin{proof}

1925: Applying \eqref{e:B_k} to a coherent state $|[\bep]_\ell\ra$, we get the exact 

1926: evolution

1927: $$

1928: \tB_N^\dagger\,|[\bep]_\ell\ra=(1-\delta_{\ep_{-1},1})\,|[\sigma^{-1}(\bep)]_{\ell+1}\ra\,.

1929: $$

1930: That is, the coherent state $|[\bep]_\ell\ra$ is either killed if $\sigma^{-1}([\bep]_\ell)=\infty$, 

1931: or transformed into a coherent state associated with the cylinder 

1932: $[\sigma^{-1}(\bep)]_{\ell+1}=\sigma^{-1}([\bep]_{\ell})$.

1933: This exact expression, which is the quantum counterpart of the classical shift \eqref{e:shift-B},

1934: should be compared with the approximate expression \eqref{e:CS-evol}.

1935: From there, a straighforward computation shows that, for any $f\in Lip(\Sigma)$:

1936: $$

1937: \tB_N^\dagger\,\Op^\ell_N(f)\,\tB_N = 

1938: \Op^{\ell+1}_N\big((1-\delta_{\ep_{0},1})\times f\circ\sigma\big)\,.

1939: $$

1940: The semiclassical equivalence $\Op^{\ell+1}_N\sim \Op^{\ell}_N$

1941: finishes the proof of $i)$.

1942: 

1943: To prove $ii)$, we remark that, 

1944: if $F\in Lip_c(C(B^{-1})\sqcup\ring{H}^{-1})$ and $f=F\circ J$,

1945: the function $F\circ B$ is supported away from $D(B)$, and both functions

1946: $\bbbone_{\set{\ep_0\neq 1}}\times f\circ\sigma$ and 

1947: $\bbbone_{\set{\ep_0\neq 1}}\times F\circ B\circ J$ are supported inside $\Sigma''$. 

1948: The semiconjugacy \eqref{e:semiconj} shows that these two functions are equal.

1949: Finally, we notice that, for $\bep\in\Sigma''$, one has

1950: $\bbbone_{\set{\ep_0\neq 1}}(\bep)=\bbbone_{M}\circ J(\bep)$.

1951: \end{proof}

1952: 

1953: %==========================================

1954: \subsubsection{Spectrum of the Walsh open baker}

1955: %==========================================

1956: 

1957: The simple expression \eqref{e:B_k} allows to explicitly compute

1958: the spectrum of $\tB_{N}$ (see \cite[Prop. 5.5]{NonZw06}). 

1959: That spectrum is determined by the two nontrivial

1960: eigenvalues $\lambda_-,\ \lambda_+$ of the matrix 

1961: $\widetilde F_3^*$. These eigenvalues have moduli $|\lambda_+|\approx 0.8443$, 

1962: $|\lambda_-|\approx 0.6838$.

1963: The spectrum of $\widetilde B_{N}$ has a gap: the long-living eigenvalues are 

1964: contained in the annulus $\set{|\lambda_-|\leq |\lambda|\leq |\lambda_+|}$,

1965: while the rest of the spectrum lies at the origin.

1966: Most of the eigenvalues are degenerate. If we count multiplicities, the long-living

1967: ($\equiv$ nontrivial) spectrum satisfies the following asymptotics when $k\to\infty$:

1968: \bequ\begin{split}

1969: \forall r>0,\qquad

1970: &\#\set{\lambda_j\in\Spec(\tB_{N})\,,\ |\lambda_j|\geq r}=C_r\,2^k+o(2^k)\,,\\

1971: C_r&=\begin{cases} 1\,, & r < r_0\\

1972: 0\,, & r > r_0

1973: \end{cases}\,,\qquad r_0\defeq|\lambda_-\lambda_+|^{1/2}=3^{-1/4}\,.

1974: \end{split}\eequ

1975: The nontrivial

1976: spectrum is spanned by a subspace $\cH_{N,long}$ of dimension $2^k$.

1977: Since the trapped set for $B=B_{\vr_{sym}}$ has dimension $2d=2\,\frac{\log 2}{\log 3}$,

1978: the above asymptotics agrees with the Fractal Weyl law \eqref{e:FWL}.

1979: Although the density of resonances (counted with multiplicities) is 

1980: peaked near the circle $\set{|\lambda|=r_0}$, the spectrum (as a set) 

1981: densely fills the

1982: annulus $\set{|\lambda_-|\leq |\lambda|\leq |\lambda_+|}$ when $k\to\infty$ 

1983: (see Fig.~\ref{f:toy-log}). 

1984: %%%%%%%%%%%%%%%%%%%

1985: \begin{figure}[ht]

1986: \rotatebox{-90}{\includegraphics[width=.5\textwidth]{toy-log+.eps}}

1987: \caption{\label{f:toy-log} Nontrivial spectrum of the Walsh open baker 

1988: $\tB_N$ for $N=3^{10}$ (circles) and

1989: $3^{15}$ (crosses), using a logarithmic representation 

1990: (horizontal=$\arg\lambda_j$, vertical=$\log|\lambda_j|$). We plot horizontal lines

1991: at the extremal radii $|\lambda_{\pm}|$ of the spectrum (dashed),

1992: at the radius $r_0=|\lambda_-\lambda_+|^{1/2}$ of highest degeneracies (dotted)

1993: and at the radius corresponding to the natural measure (full).}

1994: \end{figure}

1995: %%%%%%%%%%%%%%%%%%%

1996: In the next section we construct some long-living eigenstates of $\tB_N$ and analyze

1997: their Walsh-Husimi measures. Due to the spectral degeneracies, there is

1998: generally a large freedom to select eigenstates $(\psi_N)$ associated with

1999: a sequence of eigenvalues $(\lambda_N)$. Intuitively, that freedom should 

2000: provide more possibilities for semiclassical measures.

2001: 

2002: We mention that 

2003: J.~Keating and coworkers have recently studied the eigenstates 

2004: of a slightly different version of the Walsh-baker, namely 

2005: a matrix $\tB_N'$ obtained by replacing $F_3$ by the ``half-integer Fourier transform'' 

2006: $(G_3)_{jj'}=3^{-1/2}\e^{-2\i\pi(j+1/2)(j'+1/2)/3}$ (see~\cite{KNNS07}). 

2007: 

2008: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2009: \subsection{Long-living eigenstates of the Walsh open baker} 

2010: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2011: 

2012: We first provide the analogue of 

2013: Theorem~\ref{thm:main} for the Walsh-baker. We remind that

2014: a semiclassical measure $\mu^\Sigma$ (or its push-forward $\mu$) 

2015: is now a weak-$*$ limit of some sequence of Walsh-Husimi measures

2016: $(WH^{\ell}_{\psi_N})_{N=3^k\to\infty}$,

2017: where $\psi_N$ are eigenstates of $\tB_N$, and the index $\ell\approx k/2$.

2018: 

2019: From the quantum-classical correspondence of Prop.~\ref{p:Walsh-Egorov} and

2020: using Prop.~\ref{l:eigenmeasure-push}, we deduce the following

2021: %%%%%%%%%%%%%%%%%%%%%%%%%

2022: \begin{cor}\label{thm:walsh}

2023: Let $\mu^\Sigma$ be a semiclassical measure for a sequence of long-living eigenstates 

2024: $(\psi_N,\lambda_N)$ of 

2025: the Walsh-baker $\tB_N$. Then:

2026: 

2027: $i)$ $\mu^\Sigma$ is an eigenmeasure for the open shift $\sigma$

2028: on $\Sigma$, and 

2029: the corresponding decay rate $\Lambda$ satisfies $\Lambda=\lim_{N\to\infty}|\lambda_N|^2$.

2030: 

2031: $ii)$ If $\mu^\Sigma(\Sigma\setminus\Sigma'')=0$ (where $\Sigma''$ is defined in \eqref{e:Sigma''}), 

2032: then $\mu=J^*\mu^{\Sigma}$ is an eigenmeasure

2033: of the open baker $B$, with decay rate $\Lambda$ .

2034: \end{cor}

2035: %%%%%%%%%%%%%%%%%%%%%%%%%

2036: From the structure of $\Spec(\tB_N)$ explained above, 

2037: there exist sequences of eigenvalues $(\lambda_N)_{N\to\infty}$ converging to any 

2038: circle of radius $\lambda\in [|\lambda_-|,|\lambda_+|]$. We also

2039: know that any semiclassical measure is an eigenmeasure of $\sigma$, so it is 

2040: meaningful to ask Questions~\ref{qu1} in the present framework 

2041: (setting $\lambda_{\max/\min}=|\lambda_{\pm}|$). We add the following question:

2042: are there semiclassical measures $\mu^\Sigma$ such that $\mu^\Sigma(\Sigma'')<1$?

2043: In the next section we give partial answers to these questions.

2044: 

2045: Concerning the last point in Question~\ref{qu1}, we notice that

2046: the ``physical'' decay rate for $B=B_{\vr_{sym}}$ is $\Lambda_{nat}=2/3$. The

2047: circle $\set{|\lambda|=\sqrt{\Lambda_{nat}}}$ is contained inside the annulus

2048: $\set{|\lambda_-|\leq |\lambda|\leq |\lambda_+|}$ where the nontrivial spectrum

2049: of $\tB_N$ is semiclassically dense, although it differs from

2050: the circle $\set{|\lambda|=r_0}$ where the spectral density is peaked, 

2051: see Fig.~\ref{f:toy-log}. 

2052: Still, there exist semiclassical measures of $\tB_N$ with

2053: eigenvalue $\Lambda_{nat}$, and it is relevant to ask whether $\mu_{nat}$ 

2054: can be one of these. 

2055: At present we are not able to answer that question. The next section

2056: shows that there are plenty of semiclassical measures with eigenvalue $\Lambda_{nat}$,

2057: so even if $\mu_{nat}$ is a semiclassical measure with $\Lambda=\Lambda_{nat}$, 

2058: it is certainly not the only one.

2059: 

2060: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2061: \subsection{Constructing the eigenstates of $\tB_N$}

2062: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2063: 

2064: In this section we construct one particular (right)

2065: eigenbasis of $\tB_N$ restricted to the subspace $\cH_{N,long}$ of long-living

2066: eigenstates.

2067: The construction starts from the (right) eigenvectors

2068: $v_\pm\in\IC^3$

2069: of $\widetilde F_3^*$ associated with $\lambda_{\pm}$. 

2070: Notice that these two vectors (which we take normalized)

2071: are not orthogonal to each other.

2072: For any sequence 

2073: $\bet=\eta_0\ldots\eta_{k-1}$, $\eta_i\in\set{\pm}$, we form the tensor product state 

2074: $$

2075: |\bet\ra\defeq v_{\eta_0}\otimes v_{\eta_1}\ldots\otimes v_{\eta_{k-1}}\,.

2076: $$

2077: The action \eqref{e:B_k} of $\tB_N$ implies that

2078: $$

2079: \tB_N |\bet\ra = \lambda_{\eta_0}\,|\tau(\bet)\ra\,,

2080: $$

2081: where $\tau$ acts as a cyclic shift on the sequence: 

2082: $\tau(\eta_0\ldots\eta_{k-1})=\eta_1\ldots\eta_{k-1}\eta_0$.

2083: The {\em orbit} $\set{\tau^j(\bet),\ j\in\IZ}$ contains 

2084: $\ell_{\bet}$ elements, where the {\em period} $\ell_{\bet}$ of the 

2085: sequence $\bet$ necessarily divides $k$.

2086: The states $\set{|\tau^j(\bet)\ra,\ j=0,\ldots,\ell_{\bet}-1}$ are not orthogonal 

2087: to each other, but form a linearly 

2088: independent family, which generates the $\tB_N$-invariant subspace 

2089: $\cH_{\bet}\subset \cH_{N,long}$. The eigenvalues

2090: of $\tB_N$ restricted to $\cH_{\bet}$ are of the form 

2091: $\lambda_{\bet,r}=\e^{2\i\pi r/\ell_{\bet}}\,

2092: \big(\prod_{j=0}^{\ell_{\bet}-1}\lambda_{\eta_j}\big)^{1/\ell_{\bet}}$ 

2093: (with indices $r=1,\ldots,\ell_{\bet}$), 

2094: and the corresponding eigenstates read

2095: \bequ\label{e:formula}

2096: |\psi_{\bet,r}\ra=\frac1{\sqrt{\cN_{\bet,r}}}

2097: \sum_{j=0}^{\ell_{\bet}-1} c_{\bet,r,j}\; |\tau^j(\bet)\ra\,,\qquad 

2098: c_{\bet,r,j}=\prod_{m=0}^{j-1}\frac{\lambda_{\eta_m}}

2099: {\lambda_{\bet,r}}\,,

2100: \eequ

2101: where $\cN_{\bet,r}>0$ is the factor which normalizes $|\psi_{\bet,r}\ra$.

2102: Up to a phase, this state is unchanged if $\bet$ is replaced by $\tau(\bet)$.

2103: In the next subsections 

2104: we explicitly compute the Walsh-Husimi measures of some of these eigenstates.

2105: %%%%%%%%%%%%%%%%%%%

2106: \begin{figure}[htbp]

2107: \begin{center}

2108: \rotatebox{-90}{\includegraphics[width=8.5cm]{max6.ps}}

2109: \rotatebox{-90}{\includegraphics[width=8.5cm]{min6.ps}}

2110: \caption{\label{f:maximal} Walsh-Husimi densities for the extremal eigenstates $\psi_{+,N}$,

2111: $\psi_{-,N}$ of $\tB_{N}$, with $N=3^6$, $\ell=3$. These are coarse-grained 

2112: versions of the Bernoulli measures $\mu_{\vP_{+}}$ and $\mu_{\vP_{-}}$.}

2113: \end{center}

2114: \end{figure}

2115: %%%%%%%%%%%%%%%%%%%

2116: 

2117: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2118: \subsubsection{Extremal eigenstates} \label{s:extremal}

2119: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2120: 

2121: The simplest case is provided by the sequence $\bet=++\cdots+$, which has period $1$, so that

2122: $|\psi_{+,N}\ra=|\bet\ra=v_+^{\otimes k}$ is the (unique) eigenstate 

2123: associated with the

2124: largest eigenvalue $\lambda_+$ (this is the longest-living eigenstate).

2125: For any choice of index 

2126: $0\leq \ell\leq k$, the Walsh-Husimi measure of $|\psi_{+,N}\ra$ factorizes:

2127: $$

2128: \text{for all $\ell$-rectangle }[\bep]_\ell,\qquad

2129: WH^\ell_{\psi_{+,N}}([\bep]_\ell)

2130: =\prod_{j=0}^{\ell-1} |\la v_{+},\,e_{\ep_j}\ra|^2\ 

2131: \prod_{j=-1}^{-k+\ell}|\la v_{+},\,F_3^*\, e_{\ep_j}\ra|^2\,.

2132: $$

2133: The second product involves the vector $w_+\defeq F_3 v_+$, with components

2134: $w_{+,\ep}=(1-\delta_{\ep,1})v_{+,\ep}/\lambda_+$.

2135: Following the notations of \S\ref{s:Bernoulli}, let $\mu^\Sigma_{\vP_+}$ 

2136: be the Bernoulli eigenmeasure of $\sigma$ with weights 

2137: $P_{+,\ep}=|v_{+,\ep}|^2$, $P^*_{+,\ep}=|w_{+,\ep}|^2$.

2138: The above expression shows that the Husimi measure $WH^\ell_{\psi_{+,N}}$ is equal

2139: to the measure  $\mu^\Sigma_{\vP_+}$, 

2140: {\em conditioned on the grid formed by the $\ell$-cylinders}. 

2141: Since the diameters of the cylinders decrease to

2142: zero as $k\to\infty$, $\ell(k)\sim k/2$, the Husimi measures 

2143: $(H^\ell_{\psi_{+,N}})$ converge to $\mu^\Sigma_{\vP_+}$.

2144: 

2145: One can similarly show that the Husimi functions of the eigenstates

2146: $\psi_-=v_-^{\otimes k}$, associated with the smallest nontrivial eigenvalue $\lambda_-$, 

2147: converge to the Bernoulli

2148: eigenmeasure $\mu^\Sigma_{\vP_-}$, with weights $P_{-,\ep}=|v_{-,\ep}|^2$, 

2149: $P^*_{-,\ep}=|w_{-,\ep}|^2$, where $w_{-}=F_3 v_{-}$.

2150: 

2151: In Fig.~\ref{f:maximal} we plot the Walsh-Husimi densities (pushed-forward on $\t2$)

2152: for $\psi_{+,N}$ and $\psi_{-,N}$,

2153: using the ``isotropic'' $\ell=k/2$. These give a clear idea of the

2154: selfsimilar structure of the respective semiclassical measures 

2155: $\mu_{\vP_+}$ and $\mu_{\vP_-}$. The weights have the approximate values

2156: $\vP_+\approx (0.579, 0.287, 0.134)$,

2157: $\vP_-\approx (0.088, 0.532, 0.380)$.

2158: 

2159: Considering the fact that the eigenvalues $\lambda_N$ close to the circles 

2160: of radii $|\lambda_+|$

2161: and $|\lambda_-|$ have small degeneracies, we propose the following

2162: %&&&&&&&&&&&&&

2163: \begin{conj}

2164: Any sequence of eigenstates $(\psi_N)_{N\to\infty}$ with eigenvalues

2165: $|\lambda_N|\to |\lambda_+|$ (resp.

2166: $|\lambda_N|\to |\lambda_-|$) converges

2167: to the semiclassical measure $\mu^\Sigma_{\vP_+}$ (resp. $\mu^\Sigma_{\vP_-}$).

2168: \end{conj}

2169: %&&&&&&&&&&&&&

2170: This conjecture can be proven for the version of the Walsh baker 

2171: $\tB_N'$ studied in \cite{KNNS07}: in 

2172: that case the two eigenvectors of $\tilde G^*_3$ replacing $v_{\pm}$

2173: are orthogonal to each other, which greatly simplifies the analysis.

2174: The limit measure $\mu^\Sigma_{\vP'_+}$ is then the ``uniform'' measure 

2175: on the trapped set $\set{\ep_n\neq 1,\ n\in\IZ}$, 

2176: with $\vP'_+={\vP'_+}^*=(1/2,0,1/2)$.

2177: 

2178: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2179: \subsubsection{Semiclassical measures in the ``bulk''} 

2180: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2181: 

2182: %%%%%%%%%%%%%%%%%%%

2183: \begin{figure}[htbp]

2184: \begin{center}

2185: \rotatebox{-90}{\includegraphics[width=8.5cm]{mix+-+-4-0.ps}}

2186: \rotatebox{-90}{\includegraphics[width=8.5cm]{mix++--4-0.ps}}

2187: \caption{\label{f:mix} Walsh-Husimi densities of two eigenstates $\psi_{\bet,0}$

2188: constructed from the sequences $\bet_{4}=+-+-$ (left) and $\bet'_{4}=++--$ (right).}

2189: \end{center}

2190: \end{figure}

2191: %%%%%%%%%%%%%%%%%%%

2192: 

2193: In this section we investigate some eigenstates of the form \eqref{e:formula},

2194: with eigenvalues $\lambda_N$ situated in the ``bulk'' of the nontrivial spectrum, 

2195: that is $|\lambda_N|\in (|\lambda_-|,|\lambda_+|)$. 

2196: 

2197: In the next proposition we show that there can be two different semiclassical

2198: measures with the same decay rate $\Lambda$. 

2199: This answers by the {\em negative} the first point in Question~\ref{qu1}.

2200: %%%%%%%%%%%%%

2201: \begin{prop}

2202: Choose a rational number $t=\frac{m}{n}$, with $n\geq 1$, $1\leq m\leq n-1$.

2203: 

2204: i) Select a sequence $\bet_n$ with $m$ $(+)$ and $n-m$ $(-)$. 

2205: For all $k'\geq 1$, form the repeated sequence $(\bet_n)^{k'}$, and

2206: choose $r=r(k')\in\IZ$

2207: arbitrarily. Then the sequence of eigenstates 

2208: $\big(\psi_{(\bet_n)^{k'},r(k')}\big)_{k=nk'\to\infty}$ 

2209: converges to the semiclassical measure $\mu^\Sigma_{\bet_n}$, which is

2210: a linear combination of Bernoulli measures for the iterated shift $\sigma^n$ (see \eqref{e:rho_bet_n}). 

2211: This measure is independent of the choice of $\big(r(k')\big)$, and has decay rate 

2212: $\Lambda_t\defeq\big|\lambda_-^{1-t}\,\lambda_+^{t}\big|^2$. Its push-forward is an

2213: eigenmeasure of $B$.

2214: 

2215: ii) If $\bet_n$

2216: and $\bet'_n$ are two sequences with $m$ $(+)$ and $n-m$ $(-)$, 

2217: which are not related by a cyclic permutation, then 

2218: the semiclassical measures $\mu^\Sigma_{\bet_n}$,

2219: $\mu^\Sigma_{\bet'_{n}}$ are mutually singular, and so are their

2220: push-forwards on $\t2$.

2221: \end{prop}

2222: %%%%%%%%%%%%%

2223: Note that the radius $\sqrt{\Lambda_t}$ lies in the ``bulk'' of the 

2224: nontrivial spectrum, but can be different from the radius $r_0$ where the spectral 

2225: density is peaked.

2226: In Fig.~\ref{f:mix} we plot the Husimi functions of two states $\psi_{\bet_4,0}$,

2227: $\psi_{\bet'_4,0}$

2228: constructed from two $4$-sequences $\bet_{4}$, $\bet'_{4}$ not cyclically related. 

2229: The two functions, which give a rough idea of the limit measures $\mu_{\bet_{4}}$, 

2230: $\mu_{\bet'_{4}}$, seemingly concentrate on different parts of the phase space.

2231: \begin{proof}

2232: For short we call $\bet=(\bet_n)^{k'}$ which has period $\ell_{\bet}=\ell_{\bet_n}$, and $r=r(k')$. 

2233: From \eqref{e:formula}, each state $|\psi_{\bet,r}\ra$ is a combination of 

2234: $\ell_{\bet}$ states $|\tau^j(\bet)\ra$, and its eigenvalue has exact modulus $\sqrt{\Lambda_t}$.

2235: When $k'\to\infty$, those states are asymptotically orthogonal to

2236: each other. Indeed, their overlaps can be decomposed as

2237: $$

2238: \la \bet|\tau^j(\bet)\ra=\big(\la \bet_n|\tau^j(\bet_n)\ra \big)^{k'}\,

2239: \quad j=0,\ldots,\ell_{\bet}-1\,,

2240: $$

2241: and for any $j\neq 0\bmod \ell_{\bet}$ we have $\bet_n\neq \tau^j(\bet_n)$, which implies

2242: $|\la \bet_n|\tau^j(\bet_n)\ra|\leq c$, with $c \defeq |\la v_+,v_-\ra|^2<1$.

2243: As a result, the normalization factor of $\psi_{\bet,r}$ satisfies

2244: $$

2245: \cN_{\bet,r}= \sum_{j=0}^{\ell_{\bet}-1} |c_{\bet,r,j}|^2+\cO(c^{k'})\,,\qquad k'\to\infty\,.

2246: $$

2247: To study the semiclassical measures of the sequence $(\psi_{\bet,r})_{k'\to\infty}$, 

2248: we fix some cylinder

2249: $[\bal]=[\alpha_{-l'}\ldots\alpha_{-1}\cdot \alpha_0\ldots\alpha_{l-1}]$

2250: and compute the weight of the measures $H^\ell_{\psi_{\bet,r}}$. If $k$ is large enough and

2251: $\ell\sim k/2$, the conditions $\ell>l$ and $k-\ell>l'$ are fulfilled, and this weight

2252: can be written

2253: \bequ\label{e:overlap}

2254: H^\ell_{\psi_{\bet,r}}([\bal])=

2255: \la\psi_{\bet,r}|\Pi_{[\bal]}|\psi_{\bet,r}\ra=

2256: \sum_{j,j'=0}^{\ell_{\bet}-1} c_{\bet,r,j}\,\bar{c}_{\bet,r,j'}\,

2257: \la\tau^{j'}(\bet)|\Pi_{[\bal]}|\tau^j(\bet)\ra\,,

2258: \eequ

2259: where the projector on $[\bal]$ is a tensor product operator:

2260: $$

2261: \Pi_{[\bal]}=\pi_{\alpha_0}\otimes\pi_{\alpha_1}\ldots\pi_{\alpha_{l-1}}\otimes 

2262: (I)^{\otimes k-l-l'}\otimes F_3^*\pi_{\alpha_{-l'}}F_3\otimes\ldots \otimes F_3^*\pi_{\alpha_{-1}}F_3\,.

2263: $$

2264: The tensor factor $(I)^{k-l-l'-1}$ implies that each matrix element 

2265: $\la\tau^{j'}(\bet)|\Pi_{[\bal]}|\tau^j(\bet)\ra$ contains

2266: a factor $\big(\la\tau^{j'}(\bet_n)|\tau^j(\bet_n)\ra\big)^{k'-\cO(1)}$; 

2267: for the same reasons as above, this element

2268: is $\cO(c^{k'})$ if $j\neq j'$. We are then lead to consider only

2269: the diagonal elements $j=j'$:

2270: \bequ\label{e:weight}

2271: \forall j=0,\ldots,\ell_{\bet}-1,\qquad

2272: \la\tau^{j}(\bet)|\Pi_{[\bal]}|\tau^j(\bet)\ra=

2273: \prod_{i=0}^{l-1} P_{\eta_{j+i},\alpha_i}\,

2274: \prod_{i'=1}^{l'} P^*_{\eta_{j-i'},\alpha_{-i'}}\,.

2275: \eequ

2276: As above the weights

2277: $P_{\pm,\ep}=|v_{\pm,\ep}|^2$, $P^*_{\pm,\ep}=|w_{\pm,\ep}|^2$,

2278: and the definition of $\eta_i$ was extended to $i\in\IZ$ by periodicity. We claim that

2279: the right hand side exactly corresponds to $\tilde\mu^\Sigma_{\tau^j(\bet_n)}([\bal])$,

2280: where $\tilde\mu^\Sigma_{\tau^j(\bet_n)}$ is a certain Bernoulli eigenmeasure for the iterated

2281: shift $\sigma^n$.

2282: The latter can be seen as a simple shift on the symbol space $\tilde\Sigma$ constructed from

2283: $3^n$ symbols $\teps\in\set{0,\ldots,3^n-1}$: 

2284: each $\teps$ is in one-to-one correspondence

2285: with a certain $n$-sequence $\ep_0\ldots\ep_{n-1}$.

2286: Adapting the formalism of \S\ref{s:Bernoulli}

2287: to this new symbol space,

2288: the Bernoulli measure $\tilde\mu^\Sigma_{\tau^j(\bet_n)}$ corresponds to the

2289: following weight distributions $\tilde\vP$, $\tilde\vP^*$:

2290: \bequ\label{e:weights}

2291: \text{for}\ \tilde\eps\equiv \ep_0\ldots\ep_{n-1},\qquad

2292: \tilde P_{\tilde\ep}=\prod_{i=0}^{n-1} P_{\eta_{n,j+i},\ep_i}\quad\text{and}\quad

2293: \tilde P^*_{\tilde\ep}=\prod_{i=1}^{n} P^*_{\eta_{n,j-i},\ep_{n-i}}\,.

2294: \eequ

2295: The measures $\tilde\mu^\Sigma_{\tau^j(\bet_n)}$, $j=0,\ldots,\ell_{\bet}-1$,

2296: are related to one another through $\sigma$:

2297: \bequ\label{e:class-cycle}

2298: \sigma^*\,\tilde\mu^\Sigma_{\tau^j(\bet_n)}= |\lambda_{\eta_{n,j}}|^2\,

2299: \tilde\mu^\Sigma_{\tau^{j+1}(\bet_n)}\,.

2300: \eequ

2301: Finally, the semiclassical measure associated with the sequence

2302: $\big(\psi_{\bet,r}\big)_{k=nk'\to\infty}$ is

2303: \bequ\label{e:rho_bet_n}

2304: \mu^\Sigma_{\bet_n}=

2305: \frac1{\cN_{\bet_n}}\sum_{j=0}^{\ell_{\bet}-1} C_{\bet_n,j}\,\tilde\mu^\Sigma_{\tau^j(\bet_n)}\,,\quad

2306: \text{where}\quad C_{\bet_n,j}=\prod_{m=0}^{j-1}\frac{|\lambda_{\eta_{n,m}}|^2}{\Lambda_t}\,,

2307: \quad \cN_{\bet_n}=\sum_{j=0}^{n-1}C_{\bet_n,j}

2308: \eequ

2309: This is

2310: a probability eigenmeasure of $\sigma$, with decay rate $\Lambda_t$. It only 

2311: depends on the orbit $\set{\tau^j\bet_n,\ j=0,\ldots,\ell_{\bet_n}-1}$, and not on the

2312: choice of $(r(k'))$. This measure does not charge $\Sigma\setminus\Sigma''$, so its push-forward is an

2313: eigenmeasure of $B$.

2314: 

2315: The proof of statement $ii)$ goes as follows: since $\bet_n$ and $\bet'_n$ are

2316: not cyclically related, for

2317: any $j,j'\in\IZ$, the weight distributions $\tilde\vP$ and $\tilde\vP'$ 

2318: defining respectively the 

2319: Bernoulli measures $\tilde\mu^\Sigma_{\tau^j(\bet_n)}$ and

2320: $\tilde\mu^\Sigma_{\tau^{j'}(\bet'_n)}$ (see \eqref{e:weights}) are different. 

2321: As a result, these two measures are mutually singular

2322: (see the end of \S\ref{s:Bernoulli}), and so are the 

2323: two linear combinations

2324: $\mu^\Sigma_{\bet_n}$, $\mu^\Sigma_{\bet'_n}$. Because the weights $\tilde\vP$, $\tilde\vP'$

2325: are different from $(1,0,\ldots,0)$ and $(0,\ldots,1)$, the push-forwards $\mu_{\bet_n}$,

2326: $\mu_{\bet'_n}$ are also mutually singular.

2327: \end{proof}

2328: By a standard density argument, we can exhibit semiclassical measures for arbitrary decay

2329: rates in $[|\lambda_-|^2,|\lambda_+|^2]$.

2330: \begin{cor}\label{c:eigenmeasures}

2331: Consider a real number $t\in [0,1]$, and a 

2332: sequence of rationals $(t_p = \frac{m_p}{n_p}\in [0,1]\big)_{p\to\infty}$ converging

2333: to $t$ when $p\to\infty$. 

2334: For each $p$, let $\bet_p$ be a sequence with $m_p$

2335: $(+)$ and $n_p-m_p$ $(-)$, and $\mu^\Sigma_{\bet_p}$ the 

2336: $\sigma$-eigenmeasure constructed above.

2337: Then any weak-$*$ limit of the sequence

2338: $\big(\mu^\Sigma_{\bet_p}\big)_{p\to\infty}$ is a semiclassical measure of $(\tB_N)$;

2339: it is a $\sigma$-eigenmeasure  of decay rate 

2340: $\Lambda_{t}=\big|\lambda_-^{1-t}\,\lambda_+^{t}\big|^2$, and

2341: its push-forward is an eigenmeasure of $B$. 

2342: \end{cor}

2343: Let us call $\mathfrak{M}^\Sigma(\Lambda_t)$ the family of semiclassical measures obtained

2344: this way, and 

2345: $\mathfrak{M}^\Sigma=\cup_{t\in[0,1]}\mathfrak{M}^\Sigma(\Lambda_t)$.

2346: We don't know whether the family $\mathfrak{M}^\Sigma$

2347: exhausts the full set of semiclassical measures for $(\tB_N)$. Still, 

2348: we can address the third point in Question~\ref{qu1} with respect to this family.

2349: \begin{prop}

2350: The family of eigenmeasures $\mathfrak{M}^\Sigma$ does not contain

2351: the natural measure $\mu^\Sigma_{nat}=\mu^\Sigma_{\vr_{sym}}$. The same statement

2352: holds after push-forward on $\t2$.

2353: \end{prop}

2354: \begin{proof}

2355: For any $\ep_0\in\set{0,1,2}$, let us compute the weight of a

2356: measure $\mu^\Sigma_{\bet_p}\in\mathfrak{M}^\Sigma$ on the cylinder

2357: $[\eps_0]$ (corresponding to the vertical rectangle $\overline{R_{\ep_0}}$). For the 

2358: natural measure we have

2359: $\mu^\Sigma_{nat}([\ep_0])=1/3$ for $\ep_0=0,1,2$.

2360: 

2361: From \eqref{e:weights}, for any $j\in\IZ$ one has

2362: $\tilde\mu^\Sigma_{\tau^j(\bet_p)}([\eps_0])= P_{\eta_{p,j},\ep_0}$. Combining this

2363: with \eqref{e:rho_bet_n}, we get

2364: $$

2365: \mu^\Sigma_{\bet_p}([\ep_0])=C_{+}\,P_{+,\ep_0}+ (1-C_+)\,P_{-,\ep_0},\quad\text{where}\quad

2366: C_{+}=\frac1{\cN_{\bet_p}}\sum_{j\,:\,\bet_{p,j}=(+)} C_{\bet_p,j}\,.

2367: $$

2368: For any $\mu^\Sigma\in\mathfrak{M}$, the weights $\mu^\Sigma([\ep_0])$  will take 

2369: the same form, for some $C_{+}\in [0,1]$. Using the approximate expressions for

2370: the weights given in \S\ref{s:extremal},

2371: one checks that the condition 

2372: $\mu^\Sigma([\ep_0])=1/3$ cannot be satisfied simultaneously for $\ep_0=0,1,2$.

2373: \end{proof}

2374: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2375: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2376: \section{Concluding remarks}

2377: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2378: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2379: 

2380: The main result of this paper is the semiclassical connection between, on the one hand,

2381: eigenfunctions of a quantum open map (which mimick ``resonance eigenfunctions''), on the

2382: other hand, eigenmeasures of the classical open map. 

2383: 

2384: We proved that, modulo some problems at the discontinuities of the classical map, 

2385: semiclassical measures 

2386: associated with long-living resonant:

2387: are eigenmeasures of

2388: the classical dynamics, and their decay rate is directly related with 

2389: those of the corresponding resonant eigenstates (see Thm~\ref{thm:main}).

2390: This result, which basically derives from Egorov's theorem, has been expressed

2391: in a quite general framework, and applied to the specific example of the open baker's map. 

2392: An analogue has been proven

2393: for the more realistic setting of Hamiltonian scattering \cite[Theorem~3]{NonZw-gap}.

2394: 

2395: Although the construction and classification of eigenmeasures with decay rates $\Lambda<1$

2396: is quite easy, the classification of semiclassical measures

2397: among all possible eigenmeasures remains largely open (see Question~\ref{qu1}).

2398: The solvable model provided by the

2399: Walsh-quantized baker provides some hints to these classification, in the form

2400: of an explicit family of semiclassical measures,

2401: but we have no idea whether these results

2402: apply to more general systems, not even the ``standard'' quantum open baker. Indeed,

2403: the high degeneracies of the Walsh-baker may be responsible for a nongeneric profusion 

2404: of semiclassical measures for that model.

2405: Interestingly, the natural eigenmeasure does not seem to play

2406: a particular role at the quantum level. 

2407: 

2408: A tempting way of constraining the set of semiclassical

2409: measures would be to adapt the ``entropic'' methods of \cite{AN06} to open chaotic maps. 

2410: A desirable output of these methods would be, for instance, to forbid

2411: semiclassical measures from being of the pure point type described in \S\ref{s:pp}.

2412: 

2413: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2414: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2415: \begin{thebibliography}{999999}

2416: 

2417: \bibitem{AN06}

2418: N.~Anantharaman and S.~Nonnenmacher, 

2419: {\em Entropy of semiclassical measures of the Walsh-quantized baker's map}, 

2420: Ann. Henri Poincar\'e {\bf 8} (2007) 37--74

2421: 

2422: \bibitem{BV89}

2423: N.~L.~Balazs and A.~Voros, {\em The quantized baker's transformation},

2424: Ann. Phys. (NY) {\bf 190} (1989) 1--31

2425: 

2426: \bibitem{BU03}

2427: D.~Borthwick and A.~Uribe, {\em On the pseudospectra of Berezin-Toeplitz operators}, 

2428: Meth. Appl. Anal. {\bf 10} (2003) 31--65

2429: 

2430: \bibitem{BouzDB96}

2431: A.~Bouzouina and S.~De~Bi\`evre, {\it Equipartition

2432: of the eigenfunctions of quantized ergodic maps on the torus}, 

2433: Commun. Math. Phys. {\bf 178} (1996)  83--105

2434: 

2435: \bibitem{CasMasShep99}

2436: G.~Casati, G.~Maspero and D.~Shepelyansky, {Quantum fractal eigenstates}, 

2437: Physica {\bf D 131} (1999) 311--316

2438: 

2439: \bibitem{CherMar97} N.~Chernov and R.~Markarian, {\it Ergodic properties of Anosov maps

2440: with rectangular holes}, Boletim Sociedade Brasileira Matematica {\bf 28} (1997) 271--314;

2441: N.~Chernov, R.~Markarian and S.~Troubetzkoy,

2442: {\it Invariant measures for Anosov maps with small holes},

2443: Ergod. Th. Dyn. Sys. {\bf 20} (2000) 1007--1044

2444: 

2445: \bibitem{CdV85} 

2446: Y.~Colin~de~Verdi\`ere, 

2447: {\it Ergodicit\'e et fonctions propres du laplacien}, 

2448: Commun. Math. Phys. {\bf 102} (1985) 497--502

2449: 

2450: \bibitem{DENW06}

2451: M.~Degli~Esposti, S.~Nonnenmacher and B.~Winn, 

2452: {\em Quantum Variance and Ergodicity for the baker's map},

2453: Commun. Math. Phys. {\bf 263} (2006) 325--352

2454: 

2455: \bibitem{DemYou06}

2456: M.~F.~Demers and L.-S.~Young, {\em Escape rates and conditionally invariant

2457: measures}, Nonlinearity {\bf 19} (2006) 377--397

2458: 

2459: \bibitem{DSZ04}

2460: N.~Dencker, J.~Sj\"ostrand and M.~Zworski, {\em Pseudospectra of semiclassical

2461: (pseudo-)differential operators},  Comm. Pure Appl. Math. {\bf 57} (2004) 384--415

2462: 

2463: \bibitem{ErmSara06}

2464: L.~Ermann and  M.~Saraceno, {\it Generalized Quantum Baker Maps as 

2465: perturbations of a simple kernel},

2466: Phys. Rev. {\bf E 74} (2006) 046205

2467: 

2468: \bibitem{FNdB03}

2469: F.~Faure, S.~Nonnenmacher and S.~De~Bi\`evre,

2470: {\it Scarred eigenstates for quantum cat maps of minimal periods},

2471: Commun. Math. Phys. {\bf 239} (2003) 449--492

2472: 

2473: \bibitem{GaspRice89}

2474: P. Gaspard and S.A. Rice, {\em Scattering from a classically chaotic repellor},

2475:  J. Chem. Phys. {\bf 90} (1989) 2225--2241; 

2476: ibid, {\em  Semiclassical quantization of the

2477: scattering from a classical chaotic repellor},

2478: J. Chem. Phys. {\bf 90} (1989) 2242--2254;

2479: A.~Wirzba, 

2480: {\it Quantum Mechanics and Semiclassics of Hyperbolic n-Disk Scattering Systems},

2481: Physics Reports {\bf 309} (1999) 1--116

2482: 

2483: \bibitem{GerLei93} 

2484: P.~G\'erard and E.~Leichtnam, 

2485: {\it Ergodic properties of eigenfunctions for the Dirichlet problem}, 

2486: Duke Math. J. {\bf 71} (1993) 559--607

2487: 

2488: \bibitem{GLZ04} L. Guillop\'e, K. Lin, and M. Zworski, 

2489: {\em The Selberg zeta function for convex co-compact Schottky groups,}

2490: Comm. Math. Phys, {\bf 245} (2004) 149--176

2491: 

2492: \bibitem{HJPPS86}

2493: T.C.~Halsey, M.H.~Jensen, L.P.~Kadanoff, I.~Procaccia and B.I.~Shraiman, 

2494: {\it Fractal measures and their singularities: The characterization of

2495: strange sets}, Phys. Rev. {\bf A 33} (1986) 1141--1151

2496: 

2497: \bibitem{HelMarRob87} 

2498: B.~Helffer, A.~Martinez and D.~Robert, {\it Ergodicit\'e et limite semi-classique}, 

2499: Commun. Math. Phys. {\bf 109} (1987) 313--326

2500: 

2501: \bibitem{Kelmer05}

2502: D.~Kelmer, {\it Arithmetic quantum unique ergodicity for symplectic linear

2503: maps of the multidimensional torus}, to appear in Ann. of Math., \texttt{math-ph/0510079};

2504: {\it ibid}, {\it Scarring on invariant manifolds for perturbed quantized hyperbolic toral automorphisms},

2505: preprint, \texttt{math-ph/0607033}

2506: 

2507: \bibitem{KNNS07}

2508: J.P.~Keating, S.~Nonnenmacher, M.~Novaes, and M.~Sieber,

2509: {\em On the resonance eigenstates of an open quantum baker map}, 

2510: in preparation.

2511: 

2512: \bibitem{KNPS06}

2513: J.P.~Keating, M.~Novaes, S.D.~Prado and M.~Sieber,

2514: {\it Semiclassical structure of chaotic resonance eigenfunctions}, 

2515: Phys. Rev. Lett. {\bf 97} (2006) 150406 

2516: 

2517: \bibitem{KurRud00} 

2518: P.~Kurlberg and Z.~Rudnick, {\it Hecke theory and

2519: equidistribution for the quantization of linear maps of the torus}, 

2520: Duke Math. J. {\bf 103} (2000) 47--77

2521: 

2522: \bibitem{Leben06}

2523: M.~Lebental, J.-S.~Lauret, J.~Zyss, C.~Schmit and E.~Bogomolny, 

2524: {\it Directional emission of stadium-shaped microlasers}, Phys. Rev. {\bf A 75}

2525: (2007) 033806

2526: 

2527: \bibitem{LeforWyatt83}

2528: C.~Leforestier and R.E.~Wyatt, {\it Optical potential for laser induced

2529: dissociation}, J. Chem. Phys. {\bf 78} (1983) 2334--2344

2530: 

2531: \bibitem{Lin02}

2532: K.~Lin, {\it Numerical study of quantum resonances in          

2533: chaotic scattering}, J. Comp. Phys. {\bf 176} (2002) 295--329;

2534: K.~Lin and M.~Zworski, {\it Quantum resonances in          

2535: chaotic scattering}, Chem. Phys. Lett. {\bf 355} (2002) 201--205

2536: 

2537: \bibitem{Linden06}

2538: E.~Lindenstrauss, {\it Invariant measures and arithmetic quantum unique ergodicity}, 

2539: Annals of Math. {\bf 163} (2006) 165-219

2540: 

2541: \bibitem{LSZ03}

2542: W.~Lu, S.~Sridhar, and M.~Zworski,

2543: {\em Fractal Weyl laws for chaotic open systems}, 

2544: Phys. Rev. Lett. {\bf 91} (2003) 154101

2545: 

2546: \bibitem{Mark96}

2547: R.~Markarian, A.~Lopes,

2548: {\it Open billiards: invariant and conditionally invariant probabilities on Cantor sets},

2549: SIAM J. Appl. Math. {\bf 56} (1996) 651--680

2550: 

2551: \bibitem{MarOK05}

2552: J.~Marklof and S.~O'Keefe, {\em Weyl's law and quantum ergodicity for maps with 

2553: divided phase space}, Nonlinearity {\bf 18} (2005) 277--304

2554: 

2555: \bibitem{MeenakLaksh04} N.~Meenakshisundaram and A.~Lakshminarayan,

2556: {\it Multifractal eigenstates of quantum chaos and the Thue-Morse sequence},

2557: Phys. Rev. {\bf E 71} (2005) 065303;

2558: ibid, {\it Using the Hadamard and related transforms for simplifying 

2559: the spectrum of the quantum baker's map}, J. Phys. {\bf A 39} (2006) 11205-11216

2560: 

2561: \bibitem{NonZw05}

2562: S.~Nonnenmacher and M.~Zworski, {\em Fractal Weyl laws in discrete models

2563: of chaotic scattering},

2564: J. Phys. {\bf A 38} (2005) 10683--10702 (special issue on ``Trends in quantum

2565: chaotic scattering''); S.~Nonnenmacher, {\it Fractal Weyl law for open chaotic maps}, in 

2566: {\it Mathematical physics of quantum mechanics}, J.~Asch and A.~Joye Eds.,

2567: Lect. Notes in Physics {\bf 690}, Springer, Berlin, 2006.

2568: 

2569: \bibitem{NonZw06}

2570: S.~Nonnenmacher and M.~Zworski, 

2571: {\em Distribution of resonances for open quantum maps}, 

2572: Commun. Math. Phys. {\bf 269} (2007) 311--365

2573: 

2574: \bibitem{NonZw-gap}

2575: S.~Nonnenmacher and M.~Zworski, 

2576: {\em Quantum decay rates in chaotic scattering}, talk given at 

2577: \'Ecole Polytechnique, Palaiseau, France (May 2006), 

2578: \texttt{http://math.berkeley.edu/$\sim$zworski/nzX.ps.gz}

2579: 

2580: \bibitem{RudSar94}

2581: Z.~Rudnick and P.~Sarnak, {\em The behaviour of eigenstates of arithmetic

2582: hyperbolic manifolds}, Commun. Math. Phys. {\bf 161} (1994) 195--213

2583: 

2584: \bibitem{Sar90}

2585: M.~Saraceno, {\em Classical structures in the quantized baker transformation}

2586: Ann. Phys. (NY) {\bf 199} (1990) 37--60

2587: 

2588: \bibitem{SaVa96} 

2589: M.~Saraceno and R.O.~Vallejos, {\it The quantized D-transformation}, 

2590: Chaos {\bf 6} (1996) 193--199

2591: 

2592: \bibitem{SchaCav00}

2593: R.~Schack and C.M.~Caves, {\it Shifts on a finite qubit string: 

2594: a class of quantum baker's maps},

2595: Appl. Algebra Engrg. Comm. Comput. {\bf 10}, 305--310 (2000)

2596: 

2597: \bibitem{Schni74} 

2598: A.~Schnirelman, {\it Ergodic properties of eigenfunctions}, 

2599: Usp. Math. Nauk. {\bf 29} (1974) 181--182

2600: 

2601: \bibitem{SchoTwo04}

2602: H.~Schomerus and J.~Tworzyd{\l}o, {\it Quantum-to-classical crossover of

2603: quasi-bound states in open quantum systems}, 

2604: Phys. Rev. Lett. {\bf 93} (2004) 154102

2605: 

2606: \bibitem{Roman-PhD}

2607: R.~Schubert, {\it Semiclassical localization in phase space},

2608: PhD thesis, University of Ulm, 2001.

2609: 

2610: \bibitem{SeidMill92}

2611: T.~Seideman and W.H.~Miller, {\it Calculation of the cumulative reaction

2612: probability via a discrete variable representation with absorbing boundary

2613: conditions}, J.~Chem.~Phys. {\bf 96} (1992) 4412--4422

2614: 

2615: \bibitem{Sjo90}

2616: J. Sj\"ostrand,  {\em

2617: Geometric bounds on the density of resonances for semiclassical problems},

2618: { Duke Math. J.}, {\bf 60} (1990) 1--57

2619: 

2620: \bibitem{SjoZw05} J. Sj\"ostrand and M. Zworski,

2621: {\em Fractal upper

2622: bounds on the density of semiclassical resonances},

2623: to appear in Duke Math. J., \texttt{math.SP/0506307}

2624: 

2625: \bibitem{Stef05}

2626: P.~Stefanov, {\it Approximating resonances with the Complex Absorbing 

2627: Potential Method},  Commun. PDE {\bf 30} (2005) 1843--1862

2628: 

2629: \bibitem{TangZw98}

2630: S.-H.~Tang and M.~Zworski, {\it From quasimodes to resonances},

2631: Math. Res. Lett. {\bf 5} (1998) 261--272;

2632: P.~Stefanov, {\it Quasimodes and resonances: sharp lower bounds},

2633: Duke Math. J. {\bf 99} (1999) 75--92

2634: 

2635: \bibitem{TraSco02}

2636: M.~M.~Tracy and A.~J.~Scott, {\it The classical limit for a class of 

2637: quantum baker's maps}, J. Phys. {\bf A 35} (2002) 8341--8360

2638: 

2639: \bibitem{Zel87}

2640: S.~Zelditch, 

2641: {\it Uniform distribution of the eigenfunctions on compact hyperbolic surfaces},

2642: Duke Math. J. {\bf 55} (1987) 919--941

2643: 

2644: \bibitem{Zel96}

2645: S.~Zelditch, {\em Quantum ergodicity of $C^*$ dynamical systems},

2646: Commun. Math. Phys {\bf 177} (1996) 507--528

2647: 

2648: \bibitem{ZelZwo96} S. Zelditch and M. Zworski, 

2649: {\it Ergodicity of eigenfunctions for Ergodic Billiards}, 

2650: Commun. Math. Phys {\bf 175} (1996) 673--682

2651: 

2652: \bibitem{Zw99} M. Zworski, {\em Dimension of the limit set and the density of

2653: resonances for convex co-compact Riemann surfaces}, Inv. Math. {\bf 136} (1999)

2654: 353--409

2655: \end{thebibliography}

2656: 

2657: \end{document}

2658: 

2659: 

2660: