1: \documentstyle[graphicx,psfig]{oldelsart}

2: \newcommand{\leqsi}{\; {\scriptstyle {< \atop \sim}} \;}

3: \newcommand{\geqsi}{\; {\scriptstyle {> \atop \sim}} \;}

4: \newcommand{\vs}[1]{\vspace*{#1}}

5: \newcommand{\prL}{Phys.\ Rev.\ Lett.\ }

6: \newcommand{\pr}{Phys.\ Rev.\ }

7: \newcommand{\jpb}{J.\ Phys.\ B: Atom.\ Mol.\ Opt.\ Phys.}

8: \newcommand{\epl}{Europhys.\ Lett.\ }

9: \newcommand{\epj}{Eur.\ Phys. J.\ }

10: \newcommand{\ch}{CHAOS\ }

11: \newcommand{\ap}{Adv.\ Phys.\ }

12: \newcommand{\phr}{Phys.\ Rep.\ }

13: \newcommand{\amop}{Adv.\ At.\ Mol.\ Opt.\ Phys.\ }

14: \newcommand{\natw}{Die Naturwissenschaften\ }

15: \newcommand{\pla}{Physics Letters A\ }

16: \newcommand{\anp}{Ann.\ Phys.\ }

17: \newcommand{\spu}{Sov.\ Phys.\ Usp.\ }

18: \newcommand{\josab}{J.\ Opt.\ Soc.\ Am.\ B\ }

19: \newcommand{\zpd}{Z.\ Phys.\ D\ }

20: \newcommand{\apl}{Appl.\ Phys.\ }

21: \newcommand{\csf}{Chaos, Solitons and Fractals\ }

22: \newcommand{\psc}{Physica Scripta\ }

23: \newcommand{\pha}{Physica A\ }

24: \newcommand{\jpa}{J.\ Phys.\ A\ }

25: \newcommand{\sptp}{Sov.\ Phys.\ Tech.\ Phys.\ }

26: \newcommand{\jpf}{J.\ Phys.\ France\ }

27: \newcommand{\phd}{Physica D\ }

28: \newcommand{\nc}{Il Nuovo Cimento\ }

29: \newcommand{\rmp}{Rev.\ Mod.\ Phys.\ }

30: \newcommand{\jpc}{J.\ Phys.\ C\ }

31: \newcommand{\ajp}{Am.\ J.\ Phys.\ }

32: \newcommand{\jcp}{J.\ Chem.\ Phys.\ }

33: \newcommand{\zp}{Z.\ Phys.\ }

34: \def\bea{\begin{eqnarray}}

35: \def\eea{\end{eqnarray}}

36: \def\be{\begin{equation}}

37: \def\ee{\end{equation}}

38: % harmless do not remove - needed for polish names with french

39: %version of latex

40: \def\c#1{\setbox0=\hbox{#1}\ifdim\ht0=1ex\accent24 #1%

41:   \else{\ooalign{\hidewidth\char24\hidewidth\crcr\unhbox0}}\fi}

42: %\tightenlines

43: %\psdraft

44: \begin{document}

45: \date{}

46: 

47: \begin{frontmatter}

48: 

49: \title{Non-dispersive wave packets in periodically driven quantum systems

50: \vskip -1truecm}

51: 

52: \author{Andreas Buchleitner$^1$, Dominique Delande $^{1,2}$,

53:  and Jakub Zakrzewski$^3$}

54: 

55: \address{$^1$Max-Planck-Institut f\"ur Physik komplexer Systeme,

56:         Dresden, Germany }

57: \address{$^2$Laboratoire Kastler-Brossel, Tour 12, \'Etage 1, Universit\'e Pierre et

58: Marie Curie,\\

59: 4 Place Jussieu, 75005 Paris, France

60: }

61:  \address{$^3$Instytut Fizyki imienia Mariana Smoluchowskiego, Uniwersytet

62: Jagiello\'nski,

63:  Reymonta 4, PL-30-059 Krak\'ow, Poland

64: }

65: 

66: \tableofcontents

67: 

68: \begin{abstract}

69: With the exception of the harmonic oscillator, quantum wave-packets

70: usually spread as time evolves. This is due to the

71: non-linear character of the classical equations of motion which

72: makes the various components of the wave-packet evolve at various

73: frequencies. We show here that, using the nonlinear 

74: resonance between an internal frequency of a system and an external 

75: periodic driving, it is possible to overcome this spreading

76: and build non-dispersive (or non-spreading)

77: wave-packets which are well localized and follow a classical

78: periodic orbit without spreading. From the quantum mechanical

79: point of view, the non-dispersive wave-packets are time periodic

80: eigenstates of the Floquet Hamiltonian, localized in the nonlinear

81: resonance island.

82: 

83: We discuss the general mechanism which produces the non-dispersive wave-packets,

84: with emphasis on simple realization in the electronic motion

85: of a Rydberg electron driven by a microwave field.

86: We show the robustness of such wavepackets  for a model 

87: one-dimensional as well as for realistic three dimensional atoms.

88: We consider their  essential properties such as

89: the stability versus ionization, the characteristic

90: energy spectrum and long lifetimes. The requirements

91: for experiments aimed at observing such non-dispersive wave-packets are

92: also considered.

93: 

94: 

95: The analysis is extended to situations in which the driving frequency is

96: a multiple of the internal atomic frequency. Such a case allows us to discuss

97: non-dispersive states composed of several, macroscopically separated

98: wave-packets communicating among themselves by tunneling. Similarly we

99: briefly discuss other closely related phenomena in atomic and molecular

100: physics as well as possible further extensions of the theory.    

101: \end{abstract}

102: 

103: \begin{keyword}

104: wave-packet, dispersion, spreading, coherent states, Rydberg atoms,

105: non-linear resonance, atom-field interaction

106: \PACS{

107: 05.45.Mt, 03.65.Sq, 32.80.Qk, 32.80.Rm, 42.50.Hz}

108: \end{keyword}

109: 

110: \end{frontmatter}

111: 

112: 

113: \section{Introduction}

114: \subsection{What is a wave packet?}

115: \label{INTRO1}

116: It is commonly accepted that, for macroscopic systems like comets, cars, 

117: cats, and dogs \cite{englert95}, quantum objects

118: behave like classical ones. 

119: Throughout this report, we will understand by

120: ``quantum objects'' 

121: physical systems governed by the Schr\"odinger

122: equation: the system can then be entirely described by its state 

123: $|\psi\rangle$, which, mathematically speaking,

124: is just a vector in Hilbert space.

125: We will furthermore restrict ourselves to the dynamics 

126: of a single, spin-less particle, 

127: such as to have an immediate representation of $|\psi\rangle$ in configuration 

128: and momentum space by the wave functions 

129: $\langle\vec{r}|\psi\rangle =\psi(\vec{r})$ 

130: and $\langle\vec{p}|\psi\rangle =\psi(\vec{p})$, respectively. 

131: The same object is in classical mechanics described by its phase space coordinates 

132: $\vec{r}$ and $\vec{p}$ (or variants thereof), 

133: and our central concern will be to understand

134: how faithfully we can mimic the classical time evolution of $\vec{r}$ and $\vec{p}$ by a 

135: {\em single} quantum state $|\psi\rangle$, in the {\em microscopic} realm.

136: 

137: Whereas classical dynamics are described by Hamilton's equations of motion, which determine

138: the values 

139: of $\vec{r}$ and $\vec{p}$ at any time, given some initial condition $(\vec{r_0},\vec{p_0})$, 

140: the quantum evolution is described by the Schr\"odinger equation,

141: which propagates the wave function. Hence, it is suggestive to

142: associate a classical particle with a quantum state $|\psi\rangle$

143: which is optimally localized around the classical particle's phase

144: space position, at any time $t$.              

145: However, quantum mechanics imposes a fundamental limit

146: on 

147: localization, expressed by 

148: Heisenberg's uncertainty relation

149: \begin{equation}

150: \Delta z\ .\ \Delta p \ge \frac{\hbar}{2},

151: \label{heisenberg}

152: \end{equation}

153: where $\Delta z$ and $\Delta p$ are the uncertainties 

154: (i.e., square roots of the variances) of the 

155: probability distributions of $z$ and its conjugate momentum $p$ in

156: state $|\psi\rangle$, 

157: respectively 

158: (similar relations hold for 

159: other choices of canonically conjugate coordinates). Consequently, the

160: best we can hope for is a quantum state localized with a finite width

161: $(\Delta z,\Delta p)$ around the particle's classical position

162: $(z,p)$, with $\Delta z$ and $\Delta p$ much smaller than the typical

163: scales of the classical trajectory. This, however, would satisfy our

164: aim of constructing a quantum state that mimics the classical motion,

165: provided $|\psi\rangle$ keeps track of the classical time evolution

166: of $z$ and $p$, and $\Delta z$ and $\Delta p$ remain small as time

167: proceeds. After all, also classical bodies follow their center of

168: mass trajectory even if they have a finite volume. Quantum states

169: which exhibit these properties at least on a finite time scale are

170: called ``wave-packets'', simply due to their localization properties

171: in phase space \cite{schroe26}.

172: 

173: More formally, a localized solution of a wave equation like the

174: Schr\"odinger equation can be conceived as a linear superposition of

175: plane waves (eigenstates of the momentum operator) or of any other

176: suitable basis states. From a purely technical point of view, such a

177: superposition may be seen as a {\em packet of waves}, hence, a

178: wave-packet. Note, however, that any strongly localized object is a

179: wave-packet in this formal sense, though not all superpositions of

180: plane waves qualify as localized objects. In addition, this formal

181: definition quite obviously depends on the basis used for the

182: decomposition. Therefore, the only sensible definition of a

183: wave-packet can be through its localization properties in phase space,

184: as outlined above. 

185: 

186: What can we say about the localization properties of a quantum state

187: $|\psi\rangle$ as time evolves? For simplicity, let us assume that

188: the Hamiltonian describing the dynamics has the time-independent form 

189: \begin{equation}

190: H = \frac{{\vec p}^2}{2m} + V({\vec r}),

191: \end{equation} 

192: with $V({\vec r})$ 

193: some potential.

194: The 

195: time evolution of $|\psi\rangle$ is then described by the 

196: Schr\"odinger equation 

197: \begin{equation}

198: \left[ -\frac{{\hbar}^2}{2m} \Delta + V({\vec r})  \right]\psi({\vec r},t) 

199: = i \hbar \frac{\partial \psi({\vec r},t)}{\partial t}.

200: \end{equation}

201: The expectation values of position and momentum in this state are

202: given by 

203: \begin{eqnarray}

204: <{\vec r}(t)> = \langle \psi(t)|\ {\vec r}\ |\psi(t) \rangle , \\ 

205: <{\vec p}(t)> = \langle \psi(t)|\ {\vec p}\ |\psi(t) \rangle ,

206: \end{eqnarray}

207: with time evolution 

208: \begin{eqnarray}

209: &&\frac{d<{\vec r}>}{dt} = \frac{1}{i\hbar}\ <[{\vec r},H]> = 

210: \frac{<{\vec p}>}{m},\\

211: &&\frac{d<{\vec p}>}{dt} = \frac{1}{i\hbar}\ <[{\vec p},H]> = 

212: - <\nabla V({\vec r})>,

213: \label{ehrenfest1}

214: \end{eqnarray}

215: and $[.,.]$ 

216: the commutator. 

217: These 

218: are almost the classical equations of motion generated by $H$, apart

219: from the right hand side of eq.~(\ref{ehrenfest1}), and apply for {\em

220: any} $|\psi\rangle$, irrespective of its localization

221: properties. If we additionally assume $|\psi\rangle$ to be

222: localized within a spatial region where $\nabla V(\vec{r})$ is

223: essentially constant, we have $<\nabla V({\vec r})> \simeq \nabla

224: V(<{\vec r}>)$, and therefore

225: \begin{eqnarray}

226: \frac{d<{\vec r}>}{dt} & = & \frac{<{\vec p}>}{m},\\

227: \frac{d<{\vec p}>}{dt} & \simeq & - \nabla V(<{\vec r}>),

228: \label{ehrenfest2}

229: \end{eqnarray}

230: precisely identical to the classical equations of motion. This is

231: nothing but 

232: Ehrenfest's theorem and tells us that the quantum expectation values

233: of $\vec{r}$ and $\vec{p}$ of an initially localized wave-packet

234: evolve according to the classical dynamics, as long as

235: $|\psi\rangle$ remains localized within a range where $\nabla

236: V(\vec{r})$ is approximately constant. However, these equations do not

237: yet give us any clue on the time evolution of the uncertainties

238: $(\Delta z,\Delta p)$ (and of those in the remaining degrees of

239: freedom), and, consequently, neither on the time scales on which they

240: are reliable. 

241: 

242: On the other hand, given a localized wave-packet at time $t=0$, a

243: decomposition 

244: \begin{equation}

245: |\psi (t=0)\rangle = \sum_n c_n \ |\phi_n\rangle

246: \label{decompo}

247: \end{equation}

248: with coefficients 

249: \begin{equation}

250: c_n = \langle \phi_n | \psi(t=0) \rangle

251: \label{coeffs}

252: \end{equation}

253: in unperturbed energy eigenstates 

254: \begin{equation}

255: H |\phi_n\rangle = E_n |\phi_n\rangle,\ n=1,2,\ldots,

256: \end{equation}

257: tells us immediately that 

258: \begin{equation}

259: |\psi (t)\rangle =  

260: \sum_n c_n \exp \left( -i \frac{E_n t}{\hbar}\right) \ |\phi_n\rangle 

261: \label{psit}

262: \end{equation}

263: cannot be stationary, {\em except} for 

264: \begin{equation}

265: |\psi (t=0)\rangle = |\phi_j\rangle ,

266: \label{single}

267: \end{equation}

268: for some suitable $j$. 

269: 

270: The eigenstates $|\phi_j\rangle$ are typically delocalized over a 

271: large part of phase space (for example, over

272: a classical trajectory, see section~\ref{SQ}), and thus

273: are not wave-packets. There is however, an exception:

274: in the vicinity of a (stable) fixed point of the classical

275: dynamics (defined \cite{lichtenberg83} as a point in phase space 

276: where the 

277: time derivatives 

278: of positions and momenta vanish simultaneously), there exist localized 

279: eigenstates, see section~\ref{GM}.

280: 

281: For a particle moving in a one-dimensional, binding potential bounded

282: from below, there is a stable fixed

283: point at any potential minimum. 

284: The quantum mechanical ground state of this system is 

285: localized near

286: the fixed point at the global minimum of the potential and is a

287: wave-packet, though a very special one: 

288: it does

289: not evolve in time. Note that there is no need for the potential to be 

290: harmonic, any potential minimum will do.

291: The same argument can be used for a one-dimensional binding potential 

292: whose origin 

293: moves with uniform velocity.

294: The problem can be 

295: reduced to the previous one  by transforming 

296: to the moving frame where 

297: the potential is

298: stationary. 

299: Back in the laboratory frame, 

300: the ground state of the particle in the 

301: moving frame will appear

302: as a wave-packet which moves 

303: at uniform velocity. Obviously, expanding the wave-packet in a

304: stationary 

305: basis

306: in the laboratory 

307: frame will result in an awfully complicated decomposition, with

308: time-dependent coefficients, and this example clearly 

309: illustrates

310: the importance of the proper choice of the referential.\footnote{In passing, 

311: note

312: that such a situation is actually realized in particle accelerators: 

313: electromagnetic

314: fields are applied to the particles, such 

315: that these are trapped at some fixed 

316: point 

317: (preferably stable) in

318: an accelerated frame \cite{lichtenberg83}.}

319: 

320: If eq.~(\ref{single}) is not fulfilled, the initial localization of

321: $|\psi(t=0)\rangle$ (which is equivalent to an appropriate choice

322: of the $c_n$ in eq.~(\ref{coeffs})) will progressively deteriorate as

323: time evolves, the wave-packet will {\em spread}, 

324: due to the accumulation of relative phases of the

325: different contributions to the sum in eq.~(\ref{psit}). Whereas the

326: classical dynamics in a one-dimensional binding potential $V(\vec{r})$ are described

327: by periodic orbits (at any energy), the quantum dynamics are in

328: general {\em not} periodic. A return to the initial state is {\em only}

329: possible if {\em all} the phases $\exp(-i E_n t/\hbar)$ simultaneously take

330: the same value. This implies that all the energy levels $E_n$ 

331: (with $c_n\neq 0$) are

332: equally spaced, or that all the 

333: level spacings are integer multiples of some quantity.

334: In practice, this is realized only for the harmonic oscillator 

335: (in any dimension), and for tops or rotors where the Hamiltonian

336: is proportional to some component of an angular 

337: momentum variable. Another possibility is to use the linear Stark effect

338: in the hydrogen atom which produces manifolds of equally spaced energy levels.

339: However, experimental imperfections (higher order Stark effect and

340: effect of the ionic core on non-hydrogenic Rydberg atoms) break the

341: equality of the spacings and consequently lead to dispersion \cite{raman97}.

342:  

343: The equality of consecutive spacings has a simple classical interpretation: 

344: since all classical trajectories are

345: periodic with the {\em same} period, the system is exactly back in its initial state

346: after an integer number of periods. In other words, in those special cases, 

347: there is no wave-packet

348: spreading at long times.

349: 

350: However, for more generic systems, the energy levels are not equally spaced,

351: neither are the spacings simply related, and a wave-packet {\em will} spread. 

352: For a 

353: one-dimensional,

354: time-independent system, it is even possible to estimate the time after which 

355: the wave-packet

356: has significantly spread (this phenomenon is also known as the ``collapse'' 

357: of the

358: wave-packet \cite{yeazell90,alber91}). 

359: This is done by expanding the various energies $E_n$ around

360: the ``central'' energy $E_{n_0}$ of the wave-packet:

361: \begin{equation}

362: E_n \simeq E_{n_0} + (n-n_0) \frac{dE_n}{dn}(n_0) + \frac{(n-n_0)^2}{2} 

363: \frac{d^2E_n}{dn^2}(n_0).

364: \label{esec}

365: \end{equation}

366: 

367: The wave-packet being initially localized, its energy is more or less

368: well defined and only a relatively small number $\Delta n \ll n_0$

369: of the coefficients $c_n$ have significant values.

370: At short times, the contribution of the second order term in eq.~(\ref{esec})

371: to the evolution can be neglected. Within this approximation, the important energy levels

372: can be considered as equally spaced, and one obtains a periodic motion of the

373: wave-packet, with period:

374: \begin{equation}

375: T_{\rm recurrence} = \frac{2\pi \hbar}{\frac{dE_n}{dn}(n_0)}.

376: \label{trec}

377: \end{equation}

378: In the standard semiclassical 

379: WKB approximation (discussed in section \ref{wkb}) \cite{landau2}, this is

380: nothing but the classical period of the motion at energy $E_{n_0}$, 

381: and one recovers the similarity

382: between the quantum motion of the wave-packet and the classical motion of a

383: particle.

384: 

385: At longer times, the contributions of the various eigenstates to the dynamics 

386: of the

387: wave-packet will come out of phase because of the second order term in 

388: eq.~(\ref{esec}),

389: resulting in spreading and collapse of the wave-packet. A rough estimate of 

390: the collapse time

391: is thus when the relevant phases have changed by $2\pi.$

392: One obtains:

393: \begin{equation}

394: T_{\rm collapse} \simeq \frac{1}{(\Delta n)^2}\ \frac{2 \hbar}{\frac{d^2E_n}{dn^2}(n_0)}.

395: \label{tcol}

396: \end{equation}

397: Using the standard WKB approximation, one can show that this expression 

398: actually

399: corresponds to the time needed

400: for the corresponding classical phase space density to 

401: significantly

402: spread under the influence of the {\em classical} evolution. 

403: 

404: At still longer times, a pure quantum phenomenon appears, due to the 

405: discrete

406: nature of the energy spectrum. Since the $(n-n_0)^2$ factors in 

407: eq.~(\ref{esec}) are all

408: integers, the second order contributions to the

409: phase are all integer multiples of the phase of the $n-n_0=1$ term. If the 

410: latter is an

411: integer multiple of $2\pi$, {\em all} the second order contributions 

412: will rephase,

413: inducing a revival of the wave-packet in its original shape. A refined 

414: estimation of the revival time actually shows that this analysis 

415: overestimates the

416: revival time by a factor two.\footnote{It must also be noted that, at simple rational multiples 

417: (such as 1/3, 1/2, 2/3) of the revival time, one observes ``fractional revivals'' 

418: \cite{parker86,alber86,averbukh89,yeazell91},

419: where only part of the

420: various amplitudes which contribute to eq.~(\ref{psit}) rephase. This

421: generates a wave-function split

422: into several individual wave-packets, localized at different positions along 

423: the

424: classical orbit.}

425:  The correct result is

426: \cite{parker86,alber86,averbukh89,yeazell91}:

427: \begin{equation}

428: T_{\rm revival} = \frac{2\pi \hbar}{\frac{d^2E_n}{dn^2}(n_0)}.

429: \label{trev}

430: \end{equation}

431: 

432: Based on the very elementary considerations above, we can so far

433: draw the following conclusions:

434: \begin{itemize}

435: \item An initially localized wave-packet will follow the classical

436: equations of motion for a finite time $t\sim T_{\rm recurrence}$;

437: \item its localization properties {\em cannot} be stationary as time

438: evolves;

439: \item in general, the initial quasi-classical motion is followed by collapse

440: and revival, with the corresponding time scales $T_{\rm

441: recurrence}<T_{\rm collapse}<T_{\rm revival}$.

442: \end{itemize}

443: 

444: In the sequel of this report, we will show how under very general

445: conditions it is indeed possible to create wave-packets as single

446: eigenstates of quantum systems, i.e., as localized ground states in an

447: appropriately defined reference frame.

448: The most suitable framework is to consider

449: quantum evolution in classical phase space, that provides a picture

450: which is {\em independent} of the choice of the basis and allows for

451: an immediate comparison with  the classical Hamiltonian flow. In

452: addition, such a picture motivates a semiclassical interpretation,

453: which we will expand upon in sec.~\ref{SQ}. The appropriate technical

454: tool for a phase space description are quasiprobability distributions

455: \cite{hillery84} 

456: as the Wigner representation

457: of the state $|\psi(t)\rangle$,

458: \begin{equation}

459: W({\vec r},{\vec p}) = \frac{1}{(2\pi \hbar)^f}

460: \ \int \psi^*\left({\vec r}+ \frac{\vec x}{2}\right)  

461: \psi \left({\vec r}- \frac{\vec x}{2}\right)

462: \exp{\left( i \frac{{\vec x}.{\vec p}}{\hbar}\right)}

463: \ {\rm d}^f{\vec x},

464: \end{equation}

465: where $f$ is the number of degrees of freedom. 

466: 

467: The Wigner density $W({\vec r},{\vec p})$ is real, but not necessarily

468: positive \cite{hillery84,moyal49}.

469: Its  

470: time-evolution follows from the Schr\"odinger equation \cite{hillery84}:

471: \begin{equation}

472: \hbar \frac{\partial W({\vec r},{\vec p},t)}{\partial t} =

473: -2 H({\vec r},{\vec p},t) \sin \left(\frac{ \hbar \Lambda}{2}\right )\ 

474: W({\vec r},{\vec p},t),

475: \label{dwsdt}

476: \end{equation}

477: where 

478: \begin{equation}

479: \Lambda = \overleftarrow{\nabla_{\vec p}} \overrightarrow{\nabla_{\vec r}} 

480: - \overleftarrow{\nabla_{\vec r}} \overrightarrow{\nabla_{\vec p}}, 

481: \end{equation}

482: and the arrows indicate in which direction the derivatives act.

483: Eq.~(\protect\ref{dwsdt}) can serve to

484: motivate the semiclassical approach. Indeed,

485: the $\sin$ function can be expanded in a Taylor series, i.e. a power 

486: expansion in $\hbar.$ 

487: At lowest non--vanishing order, only terms linear in $\Lambda$ 

488: contribute and

489: one obtains:

490: \begin{equation}

491: \frac{\partial W({\vec r},{\vec p},t)}{\partial t} = \{H, W({\vec r},{\vec p},t)\},

492: \label{liouville}

493: \end{equation}

494: where $\{.,.\}$ denotes the classical Poisson bracket \cite{lichtenberg83}:

495: \footnote{We choose here the most common definition of the Poisson bracket.

496: Note, however, that 

497: some

498: authors \protect{\cite{landau1,haake90}} use the opposite sign!}

499: \begin{equation}

500: \{f,g\} = \sum_{i=1...f}{\frac{\partial f}{\partial r_i} 

501: \frac{\partial g}{\partial p_i} - \frac{\partial f}{\partial p_i} 

502: \frac{\partial g}{\partial r_i} }.

503: \label{poisson_brackets}

504: \end{equation}

505: Eq.~(\ref{liouville}) is nothing but 

506: the classical Liouville equation \cite{lichtenberg83}

507: which describes the  classical evolution of a phase 

508: space density.

509: Hence, in the ``semiclassical limit'' $\hbar \rightarrow 0,$ the

510: Wigner density evolves classically.

511: Corrections of higher power in $\hbar$ can be calculated systematically.

512: For example, the next order is $\hbar^3 H\Lambda^3W/24$ in eq.~(\ref{dwsdt}), 

513: and

514: generates terms which contain third order 

515: derivatives (in either position and/or

516: momentum) of the

517: Hamiltonian. Therefore, for a Hamiltonian of maximal degree two

518: in 

519: position and/or momentum, all higher order terms in eq.~(\ref{dwsdt}) vanish 

520: and the

521: Wigner distribution follows the classical evolution for an arbitrary 

522: initial

523: phase space density, 

524: and for arbitrarily long times. The harmonic oscillator 

525: is an example

526: of such a 

527: system \cite{schroe26,glauber63}, in agreement with our discussion of

528: eq.~(\ref{esec}) above.

529: 

530: 

531: Now, once again, why does a wave-packet spread? At first sight, it could be thought that

532: this is due to the higher order terms in eq.~(\ref{dwsdt}), and thus of

533: quantum origin. This is not true and spreading of a wave-packet

534: has a purely classical origin, as illustrated 

535: by the following example. Let us consider

536: a one-dimensional, free particle (i.e. no potential), initially described by a

537: Gaussian wave-function with average position $z_0$,

538: average momentum $p_0>0$, and spatial width $\sigma$:

539: \begin{equation}

540: \psi(z,t=0) = \frac{1}{\pi^{1/4} \sqrt{\sigma}}

541: \exp \left( i \frac {p_0 z}{\hbar}

542: - \frac{(z-z_0)^2}{2\sigma^2}\right ).

543: \label{wp_gaussian}

544: \end{equation}

545: The corresponding Wigner distribution is a Gaussian in phase space:

546: \begin{equation}

547: W(z,p,t=0)=\frac{1}{\pi \hbar} \ \exp \left( - \frac {(z-z_0)^2}{\sigma^2} 

548: -\frac {\sigma^2 (p-p_0)^2}{\hbar^2} \right).

549: \label{wigner_gaussian}

550: \end{equation}

551: As the Hamiltonian is quadratic in the momentum, without potential, 

552: this distribution

553: evolves precisely alike the equivalent 

554: classical phase space density. Hence, the part of the 

555: wave-packet

556: with $p>p_0$ will evolve 

557: with a 

558: larger velocity than the part with $p<p_0.$ 

559: Even if both parts are initially localized close to $z_0$, 

560: the  

561: contribution of different velocity classes 

562: implies that their distance will increase

563: without bound at long times. The wave-packet will therefore 

564: {\em spread}, because

565: the various {\em classical} trajectories have different velocities. Spreading is thus

566: a completely classical phenomenon. 

567: 

568: This can be seen quantitatively by calculating the exact quantum evolution.

569: One obtains

570: \begin{equation}

571: W(z,p,t)=\frac{1}{\pi \hbar} \exp \left( - \frac {(z-z_0-pt/m)^2}{\sigma^2} 

572: -\frac {\sigma^2 (p-p_0)^2}{\hbar^2} \right)

573: \end{equation}

574: for the Wigner distribution, and 

575: \begin{equation}

576: \psi(z,t) = \frac{1}{\pi^{1/4}}\ 

577: \frac{{\rm e}^{i\phi}}{\left(\sigma^2+\frac{\hbar^2t^2}{m^2\sigma^2}\right)^{1/4}}

578: \ \exp \left\{ i \frac{p_0 z}{\hbar}- \frac{\left(z-z_0-\frac{p_0t}{m}\right)^2}{2\sigma^2+\frac{2i\hbar t}{m}}\right\}

579: \end{equation}

580: for the wave-function 

581: (${\rm e}^{i\phi}$ is an irrelevant, complicated

582: phase factor).

583: The former is 

584: represented in fig.~\ref{classical_spreading}, together with the

585: evolution of a swarm of classical particles with an initial

586: phase space density identical to the one of the initial quantum

587: wave-packet. Since the quantum evolution follows exactly the classical one,

588: the phase space volume of the wave-packet is preserved. However, the

589: Wigner distribution is progressively stretched along the $z$ axis.

590: This results in a less and less localized wave-packet, with

591: \begin{equation}

592: \left\{

593: \begin{array}{l}

594: \displaystyle

595: \Delta z (t) = \frac{\sigma}{\sqrt{2}} \sqrt{1+\frac{\hbar^2t^2}{m^2\sigma^4}},\\

596: \displaystyle

597: \Delta p (t) = \frac{\hbar}{\sigma \sqrt{2}}.

598: \end{array}

599: \right.

600: \end{equation}

601: The product $\Delta z\Delta p,$ initially minimum ($\hbar/2)$, continuously increases

602: and localization is eventually lost.

603: 

604: \psfull

605: \begin{figure}

606: \centerline{\psfig{figure=bdzf01.ps,width=8cm}}

607: \caption{Evolution of the Wigner density of a wave-packet for a free

608: particle moving in a one-dimensional configuration space, compared 

609: to the classical 

610: evolution of a swarm

611: of classical particles with the same initial probability density. 

612: While the uncertainty in momentum does not vary

613: with time, the Wigner density stretches along the position coordinate

614: and loses its initial minimum uncertainty character. This implies

615: spreading of the wave-packet. This has a purely classical origin, as shown

616: by the classical evolution of the swarm of particles, which closely follow the

617: quantum evolution. The contour of the Wigner density is chosen to contain

618: 86\% of the probability.}

619: \label{classical_spreading}

620: \end{figure}

621: 

622: \subsection{Gaussian wave-packets -- Coherent states}

623: \label{coherent_states}

624: 

625: We have already realized above that,

626: for the harmonic oscillator, the second derivative $d^2E_n/dn^2$ in 

627: eq.~(\ref{esec}) 

628: vanishes

629: identically, and

630: a wave-packet does not spread, undergoing periodic motion. 

631: For this specific 

632: system,

633: one can define a restricted class of wave-packets, which are minimum

634: uncertainty states

635: (i.e., $\Delta z\Delta p = \hbar/2)$, and remain minimal under

636: time-evolution \cite{schroe26}. 

637: Nowadays, these states are known as ``coherent'' states of the harmonic oscillator

638: \cite{glauber63}, and are frequently employed in the analysis of simple quantum

639: systems such as the quantized electromagnetic field \cite{mandel90,cct92}.

640: They have Gaussian wave-functions, see fig.~\ref{coherent_state}(a), 

641: given by eq.~(\ref{wp_gaussian}), and characterized by an average position

642: $z_0$, an average momentum $p_0$, and a spatial width

643: \begin{equation}

644: \sigma = \sqrt{\frac{\hbar}{m\omega}},

645: \label{cohsigma}

646: \end{equation}

647: where $\omega$ is the classical eigenfrequency of the harmonic oscillator. 

648: The corresponding

649: Wigner distribution, 

650: eq.~(\ref{wigner_gaussian}), also

651: has 

652: Gaussian shape. The properties

653: of coherent states are widely discussed in the litterature, see

654: \cite{cct73,liboff80}.

655: 

656: In the ``naturally scaled'', dimensionless coordinates 

657: $z\sqrt{m\omega/\hbar}$ and $p/\sqrt{m\omega \hbar}$,

658: the classical trajectories of the harmonic oscillator 

659: are circles, and the Wigner distribution is an

660: isotropic Gaussian centered at $(z_0,p_0),$ see 

661: fig.~\ref{coherent_state}(b). 

662: Under time evolution, which follows precisely the classical

663: dynamics, its isotropic Gaussian shape is preserved.

664: 

665: \psfull

666: \begin{figure}

667: \centerline{\psfig{figure=bdzf02a.eps,width=7cm,angle=-90}

668: \psfig{figure=bdzf02b.eps,width=7cm,angle=-90}}

669: \caption{(a): Coherent state of the harmonic oscillator 

670: \protect\cite{schroe26}. The

671: probability density and the modulus of the wave-function

672: (solid line) have a Gaussian distribution. The real (dashed line)

673: and the imaginary (dotted line) part of the wave-function show, in addition, 

674: oscillations which reflect

675: the non-vanishing momentum. Whereas the envelope of the wave-function

676: preserves its shape under time evolution, the frequency of the oscillations

677: exhibits the same 

678: time dependence as the momentum of the corresponding classical particle. 

679: (b): Contour of the corresponding Wigner distribution which shows

680: localization in both position and momentum. The isovalue for the contour

681: is chosen to enclose 86\% of the total probability.} 

682: \label{coherent_state} 

683: \end{figure}

684: 

685: An important point when discussing wave-packets is to avoid the

686: confusion between localized wave-packets (as defined above) and minimum 

687: uncertainty

688: (coherent or squeezed \cite{cct92}) states. The latter are

689: just a very restricted class of localized states. They are the best ones in 

690: the sense

691: that they have optimum localization. On the other hand, as soon as dynamics 

692: is considered,

693: they have nice properties only for harmonic oscillators. In generic systems, 

694: they

695: spread exactly like other wave-packets.

696: Considering only coherent states as good semiclassical analogs of classical

697: particles is in our opinion 

698: a too formal point of view. Whether

699: the product $\Delta z\Delta p$ is exactly $\hbar/2$ or slightly

700: larger is certainly of secondary relevance for the semiclassical 

701: character of the wave-packet. What counts is that, in the semiclassical limit

702: $\hbar \rightarrow 0,$ the wave-packet is asymptotically perfectly

703: localized in all directions of phase space. This was Schr\"odinger's original 

704: concern, without reference to the actual value of $\Delta z\Delta p$ 

705: \cite{schroe26}.

706: 

707: Finally, for future applications, let us define the so called Husimi representation

708: of the quantum wave-function~\cite{husimi40}. It is the squared projection of a given

709: quantum state over a set of coherent states. 

710: Let us denote the gaussian wavefunction of eq.~(\ref{wp_gaussian}) 

711: (with $\sigma$ given by eq.~(\ref{cohsigma})) as 

712: $|{\mathrm Coh}(z_0,p_0)\rangle$. Then the

713: Husimi representation of $\psi(z)$ is defined as

714: \begin{equation}

715: {\mathrm Hus}(z,p)=\frac{1}{\pi}|\langle {\mathrm Coh}(z,p)|\psi\rangle|^2,

716: \label{husimi_def}

717: \end{equation} 

718: where the factor $1/\pi$ is due to the resolution of unity in the coherent states basis

719: \cite{cct73} and is often omitted (to confuse the reader). Alternatively,

720: the Husimi function may be looked upon as a Wigner function convoluted with

721: a Gaussian~\cite{hillery84}.

722: 

723: \subsection{A simple example: the one-dimensional hydrogen atom}

724: \label{INTRO2}

725: 

726: We now illustrate the ideas discussed in the preceding sections, using the

727: specific example of a one-dimensional hydrogen atom.

728: This object is both, representative of generic systems, and useful for

729: atomic systems to be discussed later in this paper.

730: We choose the simplest hydrogen atom: we neglect all relativistic, spin and QED

731: effects, and assume that the nucleus is infinitely massive. The Hamiltonian 

732: reads:

733: \begin{equation}

734: H = \frac{p^2}{2m} - \frac{e^2}{z},

735: \label{ham_h_1d}

736: \end{equation}

737: where $m$ is the mass of the electron, $e^2=q^2/4\pi\epsilon_0$, with $q$

738: the elementary charge, and

739: $z$ is restricted to the positive real axis. The validity of this

740: model as compared to the real 3D atom will be discussed in sec.~\ref{LIN}.

741: 

742: Here and in the rest of this paper, we will use atomic units, defined by 

743: $m$, $e^2$

744: and $\hbar$. The unit of length is

745: the Bohr radius $a_0=\hbar^2/me^2=5.2917\times

746:  10^{-11}\ {\rm m}$, the unit of time

747: is 

748: $\hbar^3/me^4 = 2.4189\times 10^{-17}\ {\rm s}$, the unit of 

749: energy is the 

750: Hartree

751: $me^4/\hbar^2 = 27.2\ {\rm eV},$ {\em twice} the ionization energy of the

752: hydrogen atom, and the unit of frequency is 

753: $me^4/2\pi\hbar^3 = 6.5796\times 10^{16}\ {\rm Hz}$ \cite{bethe77}.

754: 

755: With these premises, the energy levels are:

756: \footnote{The present analysis

757: is restricted to bound states of the atom. Continuum (i.e., scattering) 

758: states also

759: exist  but usually do not significantly contribute to the

760: wave-packet dynamics. If needed, they can be incorporated without any

761: fundamental difficulty~\cite{goldberger50,faisal87}.}

762: \begin{equation}

763: E_n= - \frac{1}{2n^2},\ \ \ \ \ \ {\rm for}\ n\geq 1.

764: \label{energy_levels}

765: \end{equation}

766: Clearly, the levels are not equally spaced, and therefore (see eq.~(\ref{esec})) 

767: any wave-packet

768: will spread. 

769: 

770: Fig.~\ref{wp_h1d} shows the evolution of a wave-packet

771: built from a linear combination of eigenstates of $H$, using a Gaussian

772: distribution of the coefficients $c_n$ in eq.~(\ref{psit}). The distribution

773: is  

774: centered at $n_0=60$, with

775: a width $\Delta n = 1.8$ for the $|c_n|^2.$ 

776: The calculation is done numerically, but is simple 

777: in the hydrogen atom since

778: all ingredients -- energy levels and eigenstates -- are known analytically. 

779: At time $t=0$, the wave-packet is localized at the outer turning point

780: (roughly at a distance $2n_0^2$ from the origin), and has 

781: zero initial momentum; its shape is roughly Gaussian.

782: After a quarter of a classical Kepler 

783: period $T_{\rm recurrence}$, it is significantly closer to the nucleus,

784: with negative velocity, following the classical trajectory. 

785: After half a period, it has reached the nucleus (it is essentially 

786: localized near

787: the origin). However, interference fringes are clearly visible: they 

788: originate from

789: the interference between the head of the wave-packet, which has already 

790: been reflected

791: off the nucleus, and its tail, which has not yet reached the nucleus.

792: After 3/4 of a period, the interference 

793: fringes have disappeared, and the wave-packet propagates

794: to the right. It has already spread significantly. After one period,

795: it is close to its initial position, but no more as well localized as 

796: initially.

797: This recurrence time is given by

798: eqs.~(\ref{trec}) and (\ref{energy_levels}):

799: \begin{equation}

800: T_{\rm recurrence} =2\pi n_0^3.

801: \label{hrecur}

802: \end{equation}

803: 

804: \psfull

805: \begin{figure}

806: \centerline{\psfig{figure=bdzf03.ps,width=14cm,angle=-90}}

807: \caption{Time evolution of an initially localized

808: wave-packet in the one-dimensional hydrogen atom,

809: eq.~(\protect\ref{ham_h_1d}). The wave-packet is constructed 

810: as a linear superposition of energy eigenstates of $H$, with a Gaussian 

811: distribution (centered at $n_0=60$, with a 

812: width $\Delta n=1.8$ of the 

813: $|c_n|^2$) of the coefficients $c_n$ in eq.~(\protect\ref{decompo}). Time $t$ is

814: measured in units of the classical Kepler period $T_{\rm recurrence}$,

815: eq.~(\protect\ref{hrecur}). Note the quasiclassical approach of the wave-packet 

816: to the nucleus, during the first half period (top left), with the

817: appearance of interference fringes as the particle is accelerated towards the

818: Coulomb center. After one period (top right) the wave-packet almost resumes

819: its initial shape at the outer turning point of the classical motion, but

820: exhibits considerable dispersion (collapse) 

821: after few  Kepler cycles (bottom

822: left). Leaving a little more time to the quantum evolution, we observe a

823: non-classical revival after approx. $20$ Kepler cycles (bottom right).

824: Recurrence, collapse and revival times are very well predicted by

825: eqs.~(\protect\ref{hrecur}), (\protect\ref{hcol}) and (\protect\ref{hreviv}).}

826: \label{wp_h1d} 

827: \end{figure}

828: 

829: After few periods, the wave-packet has considerably spread and is now

830: completely delocalized along

831: the classical trajectory. The time for the collapse of the

832: wave-packet is well predicted by eq.~(\ref{tcol}):

833: \begin{eqnarray}

834: T_{\rm collapse} & \simeq & \frac{2 n_0^4}{3(\Delta n)^2} = 

835: \frac{n_0}{3 \pi (\Delta n)^2}\times T_{\rm recurrence}\nonumber \\

836: & = & 1.96\times T_{\rm recurrence},\ 

837: {\rm for}\

838: n_0=60\ {\rm and}\ \Delta n=1.8.

839: \label{hcol}

840: \end{eqnarray}

841: Finally, after 20 periods, the

842: wave-packet revives with a shape similar to its initial state. Again, this 

843: revival

844: time is in good agreement with the theoretical prediction, 

845: eq.~(\ref{trev}):

846: \begin{eqnarray}

847: T_{\rm revival} & = & \frac{2 \pi n_0^4}{3} = \frac{n_0}{3}\times T_{\rm

848: recurrence} \nonumber \\ 

849: & = & 20\times T_{\rm recurrence},\ {\rm for}\

850: n_0=60\ {\rm and}\ \Delta n=1.8.

851: \label{hreviv}

852: \end{eqnarray}

853: At longer times, the wave-packet continues to alternate between collapses 

854: and revivals.

855: In fig.~\ref{dpdx}, we show the temporal evolution 

856: of the product $\Delta z \Delta p.$

857: It is initially close to the Heisenberg limit (minimum value) $\hbar/2$, 

858: and oscillates at the

859: frequency of the classical motion with a global increase. 

860: When the wave-packet has completely

861: spread, the uncertainty product is roughly constant, with apparently 

862: erratic fluctuations. 

863: At the revival time, the uncertainty undergoes again rather orderly

864: oscillations of a relatively large magnitute reaching, at minima,

865: values close to $\hbar$. That is a manifestation of

866:  its partial relocalization.

867: 

868: \psfull

869: \begin{figure}

870: \centerline{\psfig{figure=bdzf04.eps,width=8cm,angle=-90}}

871: \caption{Time evolution of the uncertainty product $\Delta z\Delta p$ 

872: (in units of $\hbar$) of the wave-packet shown in fig.~\protect\ref{wp_h1d}. 

873: Starting out from minimum uncertainty, $\Delta z\Delta p\simeq

874: \hbar/2$ (the Heisenberg limit, eq.~(\protect\ref{heisenberg})), the

875: wave-packet exhibits some transient spreading on the time scale of a 

876: Kepler period $T_{\rm recurrence}$, thus reflecting the classical motion

877: (compare top left of fig.~\protect\ref{wp_h1d}), collapses on a time scale of

878: few Kepler cycles (manifest in the damping of the oscillations of $\Delta

879: z\Delta p$ during the first five classical periods), shows a fractional

880: revival around $t\simeq 10\times T_{\rm recurrence}$, and a full revival at 

881: $t\simeq 20\times T_{\rm recurrence}$. Note that, nontheless, even at the full

882: revival the contrast of the oscillations of the uncertainty product is reduced

883: as compared to the initial stage of the evolution, as a consequence of 

884: higher-order corrections which are neglected in eq.~(\protect\ref{esec}).}

885: \label{dpdx} 

886: \end{figure}

887:  

888: For the three-dimensional hydrogen atom, the energy spectrum is exactly the same as

889: in one dimension. This implies that the 

890: temporal dynamics is built from exactly the same

891: frequencies; thus, the 3D dynamics is essentially the same 

892: as the 1D dynamics.\footnote{In a generic,

893: multidimensional, integrable

894: system, there are several different classical 

895: frequencies along the various degrees

896: of freedom. Hence, only partial revivals of the 

897: wave-packet at various times

898: are observed. The 3D hydrogen atom is {\em not} generic, 

899: because the three frequencies

900: are degenerate, which opens the possibility of 

901: a {\em complete} revival, {\em simultaneously}

902: along all three coordinates.}

903: Indeed, collapses and revivals of the wave-packet were also observed, under

904: various experimental conditions, in the laboratory

905: \cite{yeazell90,yeazell91,yeazell88,marmet94}.

906: Fig.~\ref{wp_3d} shows the evolution of a 

907: minimum uncertainty 

908: wave-packet of the 3D atom, initially localized

909: on a circular Kepler orbit of the electron. It is built as a linear

910: combination of circular hydrogenic states (i.e., states with

911: maximum angular and magnetic quantum numbers $L=M=n-1$), using the

912: same Gaussian distribution of the coefficients as in fig.~\ref{wp_h1d}.

913: As expected, the wave-packet 

914: spreads along the

915: circular trajectory (but not transversally to it) and eventually

916: re-establishes its

917: initial shape after $T_{\rm revival}.$ Figure~\ref{swarm_3d_disp} 

918: shows the corresponding 

919: evolution of a swarm of classical particles, 

920: for the same initial phase space density.

921: The spreading 

922: of the classical distribution and of the quantum wave-packet

923: proceeds very similarly, 

924: whereas the

925: revival is completely absent in the classical evolution, which once more

926: illustrates 

927: its purely quantum origin.

928: 

929: \psfull

930: \begin{figure}

931: \centerline{\psfig{figure=bdzf05an.ps,width=6cm,angle=0}

932: \psfig{figure=bdzf05bn.ps,width=6cm,angle=0}}

933: 

934: \smallskip

935: 

936: \centerline{\psfig{figure=bdzf05cn.ps,width=6cm,angle=0}

937: \psfig{figure=bdzf05dn.ps,width=6cm,angle=0}}

938: 

939: \smallskip

940: 

941: \centerline{\psfig{figure=bdzf05en.ps,width=5cm,angle=0}}

942: \caption{Time evolution of a wave-packet launched along a

943: circular Kepler trajectory, with the same Gaussian weights $c_n$,

944: eq.~(\protect\ref{coeffs}), as employed for the one-dimensional 

945: example displayed

946: in fig.~\protect\ref{wp_h1d}, i.e., centered around the principal

947: quantum number $n_0=60$.

948: Since the relative

949: phases accumulated during the time evolution only depend on $n_0$ -- see

950: eq.~(\protect\ref{psit}) -- we observe precisely the same behaviour as in the

951: one-dimensional case: classical propagation 

952: at short times (top), followed by spreading and collapse

953: (middle), and revival (bottom). The snapshots of the

954: wave function are taken at times (in units of $T_{\rm recurrence}$) 

955: $t=0$ (top left), $t=0.5$ (top right), 

956: $t=1$ (middle left), $t=12.5$ (middle right), and $t=19.45$ (bottom).

957: The cube size is 10000 Bohr radii, centered on the nucleus (marked with 

958: a cross). The radius of the circular wave-packet trajectory 

959: equals approx.

960: 3600 Bohr radii.}

961: \label{wp_3d} 

962: \end{figure}

963: 

964: \psfull

965: \begin{figure}

966: \centerline{\psfig{figure=bdzf06a.eps,width=6cm,angle=-90}

967: \psfig{figure=bdzf06b.eps,width=6cm,angle=-90}}

968: \caption{Classical time evolution of a Gaussian (in spherical coordinates) 

969: phase space density fitted to the minimum uncertainty wave-packet of

970: fig.~\protect\ref{wp_3d} at time $t=0$ (left). As time evolves, the classical 

971: phase space density spreads along the circular Kepler orbit ($t=12.5$, right),

972: but exhibits no revival. Hence, wave-packet spreading is of purely classical

973: origin, 

974: only the revival is 

975: a quantum feature.  

976: The cube size is 10000 Bohr radii, centered on the nucleus (marked with 

977: a cross).  The radius of the circular wave-packet trajectory 

978: equals approx.

979: 3600 Bohr radii.

980: }

981: \label{swarm_3d_disp} 

982: \end{figure}

983: 

984: Finally, let us notice that collapse and revival of a 3D wave-packet 

985: depend

986: on the principal quantum number $n_0$ only -- see eqs.~(\ref{hcol}) and 

987: (\ref{hreviv}) -- and

988: are independent

989: of other parameters which characterize 

990: the classical motion, such as the eccentricity and the orientation

991: of the classical elliptical trajectory. This establishes that a 3D 

992: wave-packet 

993: with low average

994: angular momentum (and, a fortiori, a 1D wave-packet as shown in

995: fig.~\ref{wp_h1d}) -- which deeply explores the non-linearity of the 

996: Coulomb force -- does not

997: disperse faster than a circular wave-packet which essentially feels

998: a constant force. Hence, arguments on the non-linear character

999: of the interaction should be used 

1000: with some caution.

1001: 

1002: There have been several experimental realizations of electronic wave-packets

1003: in atoms \cite{raman97,yeazell90,yeazell91,yeazell88,mallalieu94,weinacht99},

1004: either along the pure radial coordinate or even along angular coordinates too.

1005: However, all these wave-packets dispersed rather quickly. 

1006: 

1007: \subsection{How to overcome dispersion}

1008: \label{INTRO3}

1009: 

1010: 

1011: Soon after the discovery of quantum mechanics, the spreading of 

1012: wave-packets was realized and attempts were made to overcome 

1013: it~\cite{schroe26}.

1014: From eq.~(\ref{psit}), it is however clear that this is only possible

1015: if the populated energy levels are equally spaced. In practice, this condition

1016: is 

1017: only met for the harmonic oscillator (or simple tops and rotors). In 

1018: any other system, the anharmonicity of the

1019: energy ladder will 

1020: induce dispersion. Hence, the situation

1021: seems hopeless.

1022:  

1023: Surprisingly, it is

1024: classical 

1025: mechanics which provides us with a possible solution. Indeed, as discussed

1026: above, a quantum wave-packet spreads exactly as the corresponding

1027: swarm of classical particles. Hence, dispersion can be overcome if all 

1028: classical trajectories

1029: behave similarly in the long time limit. In other words, if an initial 

1030: volume of phase

1031: space remains well localized under time evolution, it is reasonable to expect

1032: that a wave-packet built on this initial volume will not spread either. 

1033: The simplest

1034: example is to consider a stable fixed point: 

1035: by definition \cite{lichtenberg83}, 

1036: every initial condition in its vicinity

1037: will forever remain close to it. The corresponding wave-packet indeed does 

1038: not spread

1039: at long times \ldots though

1040: this is of limited interest, as it is 

1041: simply at rest!

1042: 

1043: Another possibility is to use a set of classical trajectories which all

1044: exhibit the

1045: same periodic motion, {\it with the same period for all trajectories}. This

1046: condition, however, is too restrictive, since it leads us back 

1047: to the harmonic oscillator.

1048: Though, we 

1049: can slightly relax this constraint by allowing classical trajectories

1050: which are not strictly periodic but quasi-periodic and staying forever

1051: in the vicinity of a well defined periodic orbit: A wave-packet built

1052: on such orbits should evolve along the classical periodic orbit

1053: while keeping a finite dispersion around it.

1054: 

1055: It happens that there is a simple possibility 

1056: to generate such 

1057: classical trajectories {\em locked} on a periodic orbit, which is to

1058: drive the system by an external periodic driving. The general 

1059: theory of nonlinear dynamical systems (described in section \ref{CD}) 

1060: \cite{lichtenberg83,ottb} shows 

1061: that when a nonlinear system (the internal frequency of which depends on 

1062: the initial conditions) is subject to an external

1063: periodic driving, a {\em phase locking} phenomenon -- known as

1064: a nonlinear resonance -- takes place. For initial conditions where the

1065: internal frequency is close to the driving frequency (quasi-resonant

1066: trajectories), the effect of the coupling is to force the motion towards the

1067: external frequency. In other words, trajectories which, in the absence

1068: of the coupling, would oscillate at a frequency slightly lower than

1069: the driving are pushed forward by the nonlinear coupling, and 

1070: trajectories

1071: with slightly larger frequency are pulled backward. In a certain region

1072: of phase space -- 

1073: termed 

1074: ``nonlinear resonance island'' -- all trajectories

1075: are trapped, and locked on the external driving. At the center of the resonance

1076: island, there is a stable periodic orbit 

1077: which precisely evolves with the driving frequency.

1078: If the driving is a small perturbation, this periodic orbit

1079: is just the periodic orbit which, in the absence of driving, has exactly

1080: the driving frequency. All the trajectories in the resonance island are

1081: winding around the central orbit with their phases locked on the

1082: external driving. The crucial point for our purposes 

1083: is that the resonance island occupies

1084: a finite volume of phase space, i.e., it traps all trajectories in a window

1085: of internal frequencies centered around the driving frequency. The size of this

1086: frequency window {\em increases} 

1087: with the amplitude of the system-driving coupling, and, as we shall see in

1088: section~\ref{GM}, can be made large enough to support wave-packet

1089: eigenstates of the corresponding quantum system.

1090: 

1091: The classical trapping mechanism is illustrated

1092: in fig.~\ref{wp_clas} which shows a swarm of classical particles launched

1093: along a circular Kepler orbit of a

1094: three-dimensional hydrogen atom exposed to a resonant, circularly polarized

1095: microwave field: the effect of the microwave field is to lock the particles

1096: in the vicinity of a circular trajectory. Note that also the phase along

1097: the classical circular trajectory is locked: the particles

1098: are grouped in the direction of the microwave field and follow

1099: its circular motion without any drift. There is a striking difference

1100: with the situation shown previously in fig.~\ref{swarm_3d_disp}, where the

1101: cloud of particles rapidly spreads in the absence of the microwave field

1102: (the same swarm of initial conditions is used in the two figures). In

1103: figure \ref{wp_clas}, there

1104: are few particles (about 10\%) in the swarm which are not phase 

1105: locked with the microwave field. This is due to the finite 

1106: subvolume of phase space which is effectively phase locked. 

1107: Particles in the tail

1108: of the initial Gaussian distribution may not be trapped 

1109: \cite{lee97}. 

1110: 

1111: 

1112: \psfull

1113: \begin{figure}

1114: \centerline{\psfig{figure=bdzf07.eps,width=6cm,angle=-90}}

1115: \caption{The initially Gaussian distributed swarm of classical particles,

1116: shown in fig.~\protect\ref{swarm_3d_disp}, after evolution

1117: in the presence of the Coulomb field of the nucleus, with a

1118: resonant, circularly polarized microwave

1119: field in the plane of the circular Kepler trajectory added. 

1120: The nonlinear resonance between the unperturbed Kepler motion and

1121: the driving field locks the phase of the particles on the phase of the driving

1122: field. As opposed to the free (classical) evolution depicted in 

1123: fig.~\protect\ref{swarm_3d_disp}, 

1124: the classical distribution does {\em not} exhibit

1125: dispersion along the orbit, except for few particles launched from the

1126: tail of the initial Gaussian distribution, 

1127: which are not trapped by the principal resonance island.}

1128: \label{wp_clas} 

1129: \end{figure}

1130: 

1131: Although

1132: the microwave field applied in fig.~\ref{wp_clas} 

1133: amounts to less than 5\% of the Coulomb field

1134: along the classical trajectory, it is sufficient to

1135: synchronize the classical motion. The same phenomenon 

1136: carries over to quantum mechanics, and 

1137: allows the creation of non-dispersive wave-packets, as will

1138: be explained in detail in section \ref{QD}.

1139: 

1140: \subsection{The interest of non-dispersive wave-packets}

1141: \label{INTRO4}

1142: 

1143: Schr\"odinger dreamt of the possibility of building quantum wave-packets 

1144: following classical trajectories \cite{schroe26}. He succeded for the harmonic 

1145: oscillator, but failed for other systems \cite{schroe68}. 

1146: It was then believed that

1147: wave-packets must spread if the system is nonlinear, and this is

1148: correct for time-independent systems.

1149: However, this is not true {\em in general}, and we have seen in the

1150: previous section that clever use of the non-linearity may, 

1151: on the contrary, {\em stabilize} a wave-packet and preserve it

1152: from spreading. Such non-dispersive wave-packets are thus

1153: a realization of Schr\"odinger's dream. 

1154: 

1155: One has to emphasize strongly that they are {\em not} some variant

1156: of the coherent states of the harmonic oscillator. They are

1157: of intrinsically completely different origin. Paradoxically,

1158: they exist only if there is some non-linearity, i.e. some

1159: unharmonicity,  in the classical system. They have

1160: some resemblance with classical solitons which are localized

1161: solutions of a non-linear equation that propagate without

1162: spreading. However, they are {\em not} solitons, as they are

1163: solutions of the {\em linear} Schr\"odinger equation. 

1164: They are simply new objects. 

1165: 

1166: Non-dispersive wave-packets in atomic systems were identified

1167: for the hydrogen atom exposed

1168: to a linearly polarized \cite{delande94,abu95d} and 

1169: circularly polarized \cite{ibb94}  microwave fields quite independently

1170: and using different physical pictures. The former approach associated

1171: the wave-packets with single Floquet states localized in the vicinity of

1172: the periodic orbit corresponding to atom-microwave nonlinear resonance.

1173: The latter treatment relied on the fact  that a transformation

1174: to a frame corotating with the microwave field removes

1175: the explicit time-dependence of the Hamiltonian for the

1176: circular polarization (see section~\ref{CP}). The states localized 

1177: near the equilibria of the rotating system

1178: were baptized ``Trojan wave-packets" to stress the analogy

1179: of the stability mechanism with Trojan asteroids. Such an approach

1180: is, however, 

1181: restricted to a narrow class of systems where the time-dependence

1182: can be removed and lacks the identification of the non-linear

1183: resonance as the relevant mechanism. We thus prefer to use

1184: in this review the more general term ``non-dispersive wave-packets"

1185: noting also that in several other papers ``non-spreading wave-packets''

1186: appear equally often.  

1187: 

1188: 

1189: Apart from their possible practical applications (for example,

1190: for the purpose of quantum control of atomic or molecular fragmentation 

1191: processes

1192: \cite{weinacht99}, or for information 

1193: storage \cite{ahn00,meyer00,kwiat00,bucksbaum00} in a confined volume 

1194: of (phase) space

1195: for long times), they show the fruitful character of classical

1196: nonlinear dynamics. Indeed, here the nonlinearity is not 

1197: a nuisance to be minimized, but rather 

1198: the essential ingredient. From complex nonlinear dynamics,

1199: a simple object is born. The existence of such non-dispersive wave-packets

1200: it extremely difficult to understand (let alone to predict) 

1201: from quantum mechanics

1202: and the Schr\"odinger equation alone. The classical nonlinear dynamics

1203: point of view is by far more illuminating and predictive. It is 

1204: the classical mechanics inside which led Berman and Zaslavsky \cite{berman77}

1205: to the pioneering discussion of states associated with the classical

1206: resonance island, the subsequent studies \cite{henkel92,holthaus94,holthaus95}

1207: further identified such states for driven one-dimensional systems using the

1208: Mathieu approach without, however, discussing the wave-packet aspects of

1209: the states. 

1210: The best proof of the importance of the classical mechanics inside

1211: is that the non-dispersive wave-packets  

1212: could have been discovered for a very long

1213: time (immediately after the formulation of the Schr\"odinger equation),

1214: but were actually identified only during the last ten years 

1215: \cite{abuth,delande94,abu95d,ibb94,farrelly95a,ibb95,farrelly95,kalinski95a,kalinski95b,delande95,kuba95a,ibb96,eberly96,kalinski96a,brunello96,kalinski96b,kalinski97,kalinski98,ibb97a,ibb97b,delande97,kuba97a,cerjan97,kuba97b,abu96,kuba98,kuba98a,abu98a,delande98,sachath,hornbergerda,hornberger98,sacha98a,sacha98b,sacha99a,yeazell00}, after

1216: the recent major developments of nonlinear dynamics. 

1217: 

1218: 

1219: 

1220: \section{Semiclassical quantization}

1221: \label{SQ}

1222: 

1223: In this section, we briefly recall the basic results

1224: on the semiclassical quantization of Hamiltonian systems, 

1225: which we will 

1226: need

1227: for

1228: the construction of non-spreading wave-packets in classical phase space, as well as

1229: to 

1230: understand

1231: their properties. This section does not contain

1232: any original material.

1233: 

1234: \subsection{WKB quantization}

1235: 

1236: \label{wkb}

1237: 

1238: For a one-dimensional, bounded, time-independent system, the Hamilton function

1239: (equivalent to the total energy)

1240: is a classical constant of motion, 

1241: and the  

1242: dynamics are periodic.

1243: It is possible to define canonically conjugate action-angle variables

1244: $(I,\theta)$, such that the Hamilton function depends 

1245: on the action 

1246: alone. The usual definition of the action along a periodic orbit (p.o.)

1247: writes:

1248: \begin{equation}

1249: I=\frac{1}{2\pi}\ \oint_{\mathrm p.o.}{p\ dz},

1250: \label{action_def}

1251: \end{equation}

1252: where $p$ is the momentum along the trajectory. 

1253: 

1254: The WKB (for Wentzel, Kramers, and Brillouin) method \cite{Berry_WKB}

1255: allows to construct 

1256: an approximate solution of the Schr\"odinger

1257: equation, in terms of the classical action-angle variables and of

1258: Planck's constant $\hbar:$

1259: \begin{equation}

1260: \psi(z) = \frac{1}{\sqrt{p}}\ \exp{\left(\frac{i}{\hbar}

1261: \int{p\ dz}\right)}

1262: \label{psi_wkb}

1263: \end{equation}

1264: as an integral along the classical trajectory.

1265: This construction is possible

1266: if and only if the phase accumulated along a period of the orbit is an integer

1267: multiple of $2\pi$. This means that  the quantized states are those where

1268: the action variable $I$ is an integer multiple of $\hbar.$ This simple

1269: picture has to be slightly amended because the semiclassical

1270: WKB approximation for the wave-function breaks down at the turning points

1271: of the classical motion,

1272: where the velocity of the classical particle 

1273: vanishes and, consequently, the

1274: expression~(\ref{psi_wkb}) diverges. This failure can be repaired \cite{Berry_WKB}

1275: by adding an additional phase $\pi/2$ for each turning point. This

1276: leads to the final quantization condition

1277: \begin{equation}

1278: I=\frac{1}{2\pi}\ \oint_{\mathrm p.o.}{p\ dz} = 

1279: \left(n+\frac{\mu}{4}\right) \hbar,

1280: \label{WKB}

1281: \end{equation}

1282: where $n$ is a non-negative integer and $\mu$ -- the ``Maslov index'' --

1283: counts the number of turning points along the periodic orbit

1284: ($\mu =2$ for a simple 1D periodic orbit).

1285: 

1286: Thus, the WKB recipe is extremely simple: when the classical Hamilton

1287: function $H(I)$

1288: is expressed in terms of the action $I$, the semiclassical energy

1289: levels are obtained by calculating $H(I)$ for the quantized values of 

1290: $I$:

1291: \begin{equation}

1292: E_n = H\left(I=\left(n+\frac{\mu}{4}\right)\hbar\right).

1293: \label{semen}

1294: \end{equation}

1295: 

1296: Finally, as a consequence of eqs.~(\ref{WKB}),(\ref{semen}), 

1297: the spacing between two

1298: consecutive semiclassical energy 

1299: levels is simply related to the classical frequency $\Omega$ of the motion,

1300: \begin{equation}

1301: E_{n+1}-E_n \simeq \frac{d E_n}{d n} = \hbar \frac{d H}{d I} = \hbar

1302: \Omega(I),

1303: \label{cprinc}

1304: \end{equation}

1305: a result which 

1306: establishes the immediate correspondence between a resonant 

1307: transition between two quantum mechanical eigenstates in the semiclassical

1308: regime, and resonant driving of the associated classical trajectory.

1309: 

1310: In the vicinity of a fixed (equilibrium) point, the Hamilton function 

1311: can be expanded

1312: at second order (the first order terms

1313: are zero, by definition of the fixed point), 

1314: leading to the ``harmonic approximation''. If the fixed point is

1315: stable, the semiclassical WKB quantization of the harmonic approximation gives 

1316: exactly the quantum result, although the semiclassical wave-function,

1317: eq.~(\ref{psi_wkb}), is incorrect. This remarkable feature is not true

1318: for an unstable fixed point (where classical trajectories escape far

1319: from the fixed point), and the WKB approximation fails in this case.    

1320: 

1321: \subsection{EBK quantization}

1322: \label{EBKsec}

1323: 

1324: For a multi-dimensional system, 

1325: it is a much more complicated task to extract the quantum mechanical 

1326: eigenenergies 

1327: from the classical dynamics of a Hamiltonian system.

1328: The problem can be solved for {\em integrable\/} systems,

1329: where there are as many constants of motion as degrees of 

1330: freedom~\cite{ozorio88}.

1331: This is known as EBK (for Einstein, Brillouin, and Keller) quantization

1332: \cite{einstein17}, and is a simple extension

1333: of the WKB quantization 

1334: scheme.

1335: Let us choose two degrees of freedom for simplicity, the extension

1336: to higher dimensions being straightforward. If the system

1337: is integrable, the Liouville-Arnold theorem~\cite{lichtenberg83}

1338: assures the existence of two pairs of canonically conjugate

1339: action-angle variables, $(I_1,\theta_1)$ and $(I_2,\theta_2)$, 

1340: such that the classical Hamilton function depends only on the actions:

1341: \begin{equation}

1342: H = H(I_1,I_2).

1343: \end{equation}

1344: The classical motion is periodic along each angle (the actions being constants

1345: of the motion)

1346: with frequencies

1347: \begin{eqnarray}

1348: \Omega_1 & = & \frac{\partial H}{\partial I_1},\\

1349: \Omega_2 & = & \frac{\partial H}{\partial I_2}.

1350: \end{eqnarray}

1351: In the generic case, these two frequencies are incommensurate, such that the

1352: full motion in the four-dimensional phase space 

1353: is quasi-periodic, and densely fills the

1354: so-called ``invariant torus" defined by the constant values $I_1$ and $I_2$.

1355: The semiclassical wave-function is constructed similarly to the WKB 

1356: wave-function.

1357: Turning points are now replaced by caustics \cite{ottb,einstein17,bornwolf} 

1358: of the classical

1359: motion (where the projection of the invariant torus on configuration

1360: space is singular), but the conclusions are essentially

1361: identical. The single-valued character of the wave-function requires the

1362: following quantization of the actions:

1363: \begin{eqnarray}

1364: I_1 & = & \left(n_1 + \frac{\mu_1}{4}\right) \hbar,\\

1365: I_2 & = & \left(n_2 + \frac{\mu_2}{4}\right) \hbar,

1366: \label{EBK}

1367: \end{eqnarray}

1368: where $n_1,n_2$ are two non-negative integers, and

1369: $\mu_1,\mu_2$ the Maslov indices (counting the number of caustics encountered

1370: on the torus)

1371: along the $\theta_1,\theta_2$ directions.

1372: Once again, 

1373: the semiclassical energy levels (which now depend on two quantum

1374: numbers) are obtained by substitution of these quantized values into the classical Hamilton

1375: function.

1376: 

1377: An alternative formulation of the EBK criterium 

1378: is possible using the original position-momentum 

1379: coordinates. Indeed, eq.~(\ref{EBK}) just expresses that, along any closed

1380: loop on the invariant torus, the phase accumulated 

1381: by the wave-function is an integer

1382: multiple of $2\pi$ (modulo the Maslov contribution). 

1383: Using the canonical invariance of the total action

1384: \cite{ozorio88}, 

1385: the EBK quantization conditions can be written as:

1386: \begin{equation}

1387: \frac{1}{2\pi} \oint_{{\mathrm closed\ path}\ \gamma_i}{\vec{p}. \d \vec{r}} = 

1388: \left(n_i + \frac{\mu_i}{4}\right) \hbar

1389: \label{ebkgen}

1390: \end{equation}

1391: where the integral has to be taken along two topologically 

1392: independent closed paths 

1393: $(\gamma_1,\gamma_2)$ on the

1394: invariant torus.   

1395: 

1396: Note that, as opposed to the WKB procedure in a 1D situation, 

1397: the EBK quantization uses the invariant tori of the classical

1398: dynamics, {\em not\/} the trajectories themselves.

1399: When there is a stable periodic orbit, it is 

1400: surrounded by invariant tori. 

1401: The smallest quantized torus around the stable orbit is associated with

1402: a quantum number equal to zero for the motion transverse to the orbit: 

1403: it defines 

1404: a narrow tube around the orbit, whose projection on configuration

1405: space will be 

1406: localized in the immediate vicinity

1407: of the orbit. Thus, the corresponding wave-function

1408: will also be localized close to this narrow tube, i.e.,

1409: along the stable periodic orbit in configuration space.

1410: Transversely to the orbit, the wave-function (or the Wigner function)

1411: will essentially look like the ground state of an harmonic oscillator,

1412: i.e. like a Gaussian. 

1413: 

1414: Finally, let us note that it is also possible to develop an analogous EBK

1415: scheme for periodically time dependent Hamiltonians \cite{breuer91}

1416: using the notion of an extended phase space \cite{lichtenberg83}.

1417: Such an approach will be extensively used in the next section, so it is 

1418: discussed in detail there.

1419: 

1420: 

1421: \subsection{Scars}

1422: 

1423: \label{scars}

1424: 

1425: When a periodic orbit is unstable, there is no torus closely surrounding it.

1426: However, it often happens that quantum eigenstates exhibit an increased

1427: probability density in the vicinity of unstable periodic orbits. This scarring

1428: phenomenon  is nowadays relatively well understood, and the interested

1429: reader may 

1430: consult references \cite{heller89,stoeckmann99}.

1431: 

1432: Similarly, some quantum states have

1433: an increased probability density 

1434: in the vicinity of an unstable equilibrium point

1435: \cite{abu95d,jensen89b,leopold94}.

1436: This localization is only partial. Indeed, since a quantum eigenstate

1437: is a stationary structure, some probability density

1438: {\em must} localize along the unstable directions of the classical Hamiltonian

1439: flow \cite{ozorio88}, and the localization cannot be perfect.

1440: This is in sharp contrast with stable equilibrium points and stable

1441: periodic orbits which -- see above -- optimally support

1442: localized eigenstates.

1443: 

1444: Note that there is, however, a big difference between scarring and localization

1445: in the vicinity of an unstable fixed point. The latter phenomenon

1446: is of purely classical origin. Indeed, close to an 

1447: equilibrium point, the velocity goes to zero and the particle consequently 

1448: spends

1449: more time close to the equilibrium point than further away from it. The quantum

1450: eigenfunctions have the same property: the probability density is large

1451: near the equilibrium point. This trivial enhancement of the

1452: probability density is already well known 

1453: for a one-dimensional

1454: system where the WKB wave-function, eq.~(\ref{psi_wkb}),

1455: diverges when the momentum

1456: tends to zero. 

1457: The localization effect

1458: near an unstable point is just the quantum manifestation of the

1459: critical slowing down of the classical particle~\cite{delande97}.

1460: 

1461: \section{Non-dispersive wave-packets and their realization in various 

1462: atomic systems}

1463: 

1464: \subsection{General model -- nonlinear resonances}

1465: \label{GM}

1466: 

1467: In this section, we present the general theory of non-dispersive wave-packets.

1468: As explained in section \ref{INTRO3}, the basic ingredients

1469: for building a non-dispersive wave-packet are a non-linear

1470: dynamical system and an external periodic driving which is

1471: resonant with an internal frequency of the dynamical system.

1472: We present here a very general theory starting out from classical

1473: mechanics which provides us with the most suggestive 

1474: approach to non-linear resonances. In a second step, we choose a pure

1475: quantum approach 

1476: giving essentially the same physics. 

1477: 

1478: We use a one-dimensional model, which displays all the interesting

1479: features of non-linear resonances. 

1480: While the direct link between classical nonlinear resonances, the corresponding

1481: Floquet states, and non-dispersive wave-packets has been identified only 

1482: recently \cite{abuth,delande94,abu95d,delande95,abu96} some 

1483: aspects of the developments presented below

1484: may be found in earlier studies  

1485: \cite{berman77,zaslavsky81,henkel92,holthaus95}.

1486: 

1487: Several complications not included in the simple one-dimensional

1488: model are important features of ``real systems". 

1489: They are discussed at a later stage in this paper:

1490: \begin{itemize}

1491: \item the effect of additional degrees

1492: of freedom, in sections \ref{LIN3D}-\ref{EP};

1493: \item higher 

1494: nonlinear resonances (where the driving frequency is a multiple

1495: of the internal frequency), in section \ref{HOR};

1496: \item an unbounded phase space, leading to the decay of 

1497: non-dispersive wave-packets (as ``open quantum systems''), 

1498: in section \ref{ION};

1499: \item sources of  ``decoherence'', such as

1500: spontaneous emission of atomic wave-packets, in section \ref{SPO};

1501: \item deviations from 

1502: temporal periodicity, in section \ref{PTP}.

1503: \end{itemize}

1504: 

1505: 

1506: In 

1507: particular cases, an apparently simpler approach is also possible

1508: (such as the use of the rotating frame for a Rydberg atom exposed to a

1509: circularly polarized electromagnetic field, see section \ref{CP}).

1510: Despite all its advantages, it may be quite specific and 

1511: too restricted to reveal nonlinear resonances 

1512: as the actual cause of the

1513: phenomenon. Here, we seek the most general description.

1514: 

1515: \subsubsection{Classical dynamics}

1516: \label{CD}

1517: 

1518: Let us start from a time-independent, bounded, one-dimensional system

1519: described by the Hamilton function

1520: $H_0(p,z)$.

1521: Since energy is conserved, the motion is confined to a

1522: one-dimensional manifold in two-dimensional phase space.

1523: Except for energies which define a fixed point of the

1524: Hamiltonian dynamics (such that $\partial H_0/\partial z = 0$

1525: and $\partial H_0/\partial p = 0;$ these fixed points generically

1526: only exist at some 

1527: isolated values of energy, for example at $E=0$

1528: for the harmonic oscillator), the motion is periodic

1529: in time, and the phase space trajectory is a simple closed loop.

1530: 

1531: It is always possible to find a set of canonically

1532: conjugate phase space coordinates adapted to the dynamics of the system.

1533: These are the action-angle coordinates $(I,\theta)$,

1534: whose existence is guaranteed

1535: by the Liouville-Arnold theorem \cite{lichtenberg83}, with:

1536: \begin{eqnarray}

1537: &&0 \leq I, \\

1538: &&0 \leq \theta \leq 2\pi, \label{thetaint} \\

1539: &&\left\{ \theta , I \right\} = 1,

1540: \end{eqnarray}

1541: and $\{ .,.\}$ the usual Poisson brackets, eq.~(\ref{poisson_brackets}).

1542: 

1543: A fundamental property is that the Hamilton function in these coordinates

1544: depends on $I$ alone, not on $\theta:$

1545: \begin{equation}

1546: H_0 = H_0(I).

1547: \end{equation}

1548: As a consequence of 

1549: Hamilton's

1550: equations of motion, $I$ is a

1551: constant of motion, and 

1552: \begin{equation}

1553: \theta =\Omega t +\theta_0

1554: \label{thetatime}

1555: \end{equation}

1556: evolves linearly in time,

1557: with the angular velocity

1558: \begin{equation}

1559: \Omega(I) = \frac{\partial H_0}{\partial I}(I),

1560: \label{internal_omega}

1561: \end{equation}

1562: which depends on the 

1563: action $I$.

1564: The period of the motion at a given value of $I$ reads

1565: \begin{equation}

1566: T = \frac{2\pi}{\Omega(I)}.

1567: \end{equation}

1568: 

1569: In simple words, the action $I$ is nothing but the properly

1570: ``rescaled" total energy, and the angle $\theta$ just measures how time

1571: evolves

1572: along the (periodic) orbits. 

1573: In a one-dimensional system, the action variable

1574: can be expressed as an

1575: integral along the orbit, see eq.~(\ref{action_def}).

1576: 

1577: 

1578: Suppose now that the system is exposed to a periodic driving force, such

1579: that the Hamilton function, in the original coordinates, writes

1580: \begin{equation}

1581: \label{ham}

1582: H = H_0(p,z) + \lambda V(p,z) \cos \omega t,

1583: \label{h_gen}

1584: \end{equation}

1585: with $\omega$ the frequency of the periodic drive and

1586: $\lambda$ some small parameter which determines

1587: the strength of the perturbation.

1588: For simplicity, we choose a single cosine function to define the

1589: periodic driving.  For a more

1590: complicated dependence on time \cite{ringot00}, 

1591: it is enough to expand it in a Fourier

1592: series, see section \ref{EP}.

1593: The equations become slightly more complicated, but the physics

1594: is essentially identical.

1595: 

1596: We now express the perturbation $V(p,z)$ in action-angle coordinates.

1597: Since $\theta$ is $2\pi-$periodic, eq.~(\ref{thetaint}), we obtain a Fourier series:

1598: \begin{equation}

1599: V(I,\theta) = \sum_{m=-\infty}^{+\infty}

1600: {V_m(I)\ \exp (im\theta)}.

1601: \label{fourier_components}

1602: \end{equation}

1603: Note that, as $\theta$ evolves linearly with time $t$ for the unperturbed

1604: motion (and therefore parametrizes an unperturbed periodic orbit),

1605: the $V_m$ can also be seen as the Fourier

1606: components of $V(t)$ evaluated along the classical, unperturbed trajectory.

1607: Furthermore, since the Hamilton function is real, $V_{-m}=V_m^*.$

1608: Again for the sake of simplicity, we will assume that both are

1609: real and thus equal.

1610: The general case can be studied as well, at the price of slightly

1611: more complicated formulas.

1612: 

1613: Plugging eq.~(\ref{fourier_components}) in eq.~(\ref{h_gen}) results in the

1614: following Hamilton function in action-angle coordinates,

1615: \begin{equation}

1616: H = H_0(I) + \lambda  \sum_{m=-\infty}^{+\infty}

1617: {V_m(I)\ \exp (im\theta)} \cos \omega t,

1618: \end{equation}

1619: which (assuming $V_{-m}=V_m$ -- see above) can be rewritten as

1620: \begin{equation}

1621: H = H_0(I) + \lambda \sum_{m=-\infty}^{+\infty}

1622: {V_m(I)\ \cos (m\theta-\omega t)}.

1623: \label{hamfou}

1624: \end{equation}

1625: For $\lambda$ sufficiently small, the phase space trajectories of the perturbed dynamics 

1626: will remain close to the unperturbed ones (for short times).

1627: This means

1628: that $m\theta -\omega t$ evolves approximately linearly in time

1629: as 

1630: $(m\Omega - \omega)t$ (see eq.~(\ref{thetatime})),

1631: while $I$ is slowly varying.

1632: It is therefore reasonable to expect that all the terms

1633: $ V_m(I)\ \cos (m\theta-\omega t)$ will oscillate rapidly and average out

1634: to zero,

1635: leading to an effective approximate Hamiltonian identical to the unperturbed

1636: one.

1637: Of course, this approach is too simple. Indeed, close to a  

1638: ``resonance'', 

1639: where $(s\Omega - \omega)$ is small, the various terms  

1640: $V_m\ \cos (m\theta-\omega t)$ oscillate, except for the $m=s$ term which

1641: may evolve very {\em slowly} and affect the dynamics considerably.

1642: For simplicity, we restrict the present analysis to the principal

1643: resonance such that $\Omega \simeq \omega$.  The extension

1644: to higher  resonances (with $s\Omega\simeq\omega$) is discussed

1645: in section \ref{HOR}.

1646: 

1647: 

1648: Our preceding remark is the basis of the ``secular approximation''

1649: \cite{lichtenberg83,cct92}.

1650:  The

1651: guiding idea

1652: is to perform a canonical change of coordinates involving

1653: the slowly varying variable $\theta -\omega t.$ Because of the explicit

1654: time dependence, this requires first 

1655: the passage to an extended phase space,

1656: which comprises time as an additional coordinate. The Hamilton function

1657: in extended phase space is defined by

1658: \begin{equation}

1659: {\cal H} = P_t + H,

1660: \label{hamext}

1661: \end{equation}

1662: with $P_t$ the momentum canonically conjugate to the new coordinate

1663:  - time $t$. The physical time $t$ is now parametrized by

1664: some

1665: fictitious time, say $\xi$. However,

1666: \begin{equation}

1667: \frac{\partial{\cal H}}{\partial P_t}=\frac{dt}{d\xi}=1,

1668: \end{equation}

1669:  i.e., $t$ and

1670: $\xi$ are essentially identical.

1671: ${\cal H}$, being independent of $\xi$,

1672:  is conserved as  $\xi$ evolves.

1673: The requested transformation to slowly varying variables $\theta -\omega t$ 

1674: reads:\footnote{This

1675: {\em canonical}

1676: change of coordinates is often refered to as

1677: ``passing to the rotating frame". It should however be emphasized that

1678: this

1679: suggests the correct picture {\em only}

1680: in phase space

1681: spanned by the action-angle coordinates $(I,\theta)$. In the original

1682: coordinates $(p,z)$, the transformation is usually very complicated, and

1683: only rarely

1684: a standard rotation in configuration space (see also section \ref{CP}).}

1685: \begin{eqnarray}

1686: &&\hat{\theta} = \theta - \omega t, \label{rotframe_a} \\

1687: &&\hat{I} = I,\label{rotframe_b} \\

1688: &&\hat{P}_t = P_t + \omega I,

1689: \label{rotframe_c}

1690: \end{eqnarray}

1691:  which transforms ${\cal H}$ into

1692: \begin{equation}

1693:  \hat{\cal H} = \hat{P}_t + H_0(\hat{I}) -\omega \hat{I}+ \lambda

1694: \sum_{m=-\infty}^{+\infty}

1695: {V_m(\hat{I})\ \cos (m\hat{\theta}+(m-1)\omega t)}.

1696: \label{floq_sc_gen}

1697: \end{equation}

1698: This Hamilton function does not

1699: involve any approximation yet. Only in the next step

1700: we average $\hat{\cal H}$ over the fast variable $t,$ i.e., over one period

1701: of the external driving. This has the effect of canceling all oscillating

1702: terms in the sum, except

1703: the resonant one, defined by $m=1$.

1704: Consequently, we are left with the approximate, 

1705: ``secular'' Hamilton

1706: function:

1707: \begin{equation}

1708: {\cal H}_{\mathrm sec} = \hat{P}_t + H_0(\hat{I}) -\omega \hat{I}+

1709: \lambda

1710: V_1(\hat{I}) \cos \hat{\theta}.

1711: \label{hsec_ap}

1712: \end{equation}

1713: 

1714: The secular Hamilton function no longer depends on time. Hence,

1715: $\hat{P}_t$ is a constant

1716: of motion and we are left with an integrable Hamiltonian system living in a

1717: two-dimensional phase space, spanned by $(\hat{I},\hat{\theta})$.

1718: The above averaging procedure is valid at first order in $\lambda$.

1719: Higher order expansions, using, e.g., the Lie algebraic

1720: transformation method~\cite{lichtenberg83}, 

1721: are possible.\footnote{An example is given

1722: in \cite{abu97}, in a slightly different situation,

1723: where the perturbation is not resonant with the

1724: internal frequency.} Basically, the interesting physical phenomena

1725: are already present at lowest non-vanishing order, to which we will

1726: restrain in the following.

1727: 

1728: The dynamics

1729: generated by the secular Hamilton function

1730: is rather simple. At order zero in $\lambda,$

1731: $\hat{I}$ is constant and $\hat{\theta}$ evolves linearly with time.

1732: As we can read from eq.~(\ref{hsec_ap}),  a

1733: continuous family (parametrized by the value of

1734: $0\leq \hat{\theta}<2\pi$)

1735: of fixed points exists

1736: if $d\hat{\theta}/dt=\partial{\cal H}_{\mathrm

1737: sec}/\partial\hat{I}$ vanishes,

1738: i.e., at actions $\hat{I}_1$ such that

1739: \begin{equation}

1740: \Omega(\hat{I}_1) = \frac{\partial H_0}{\partial I}(\hat{I}_1) = \omega.

1741: \label{resonance_condition}

1742: \end{equation}

1743: Thus, unperturbed trajectories that are {\em resonant} with the

1744: external drive are fixed points

1745: of the unperturbed secular

1746: dynamics. This is precisely why slowly

1747: varying variables

1748: are introduced.

1749: 

1750: Typically, eq.~(\ref{resonance_condition}) has only isolated solutions --

1751: we will

1752: assume that in the following. Such is the case when

1753: $\partial^2 H_0/\partial I^2$ does not vanish -- excluding

1754: the

1755: pathological situation of the harmonic oscillator, where all trajectories

1756: are simultaneously resonant.

1757: Hence,

1758: if $\partial^2 H_0/\partial I^2$ is positive, the line 

1759: ($\hat{I}=\hat{I}_1,0\leq \hat{\theta}<2\pi$, parametrized by $\hat\theta$)

1760: is a minimum of the unperturbed secular Hamilton function ${\cal H}_{\mathrm

1761: sec}$;

1762: if $\partial^2 H_0/\partial I^2$

1763: is negative, it is a maximum.

1764: 

1765: At first order in $\lambda$, the fixed points of the secular

1766: Hamiltonian should have an action close to $\hat{I_1}.$ Hence, it is

1767: reasonable to perform

1768: a power expansion of the unperturbed Hamiltonian in the vicinity of

1769: $\hat{I}=\hat{I_1}.$

1770: We obtain the following approximate  Hamiltonian:

1771: \begin{equation}

1772: {\cal H}_{\mathrm pend} = \hat{P}_t + H_0(\hat{I_1}) -\omega\hat{I}_1 + \frac{1}{2}

1773: H^{''}_0(\hat{I_1})\ (\hat{I}-\hat{I_1})^2 + \lambda

1774: V_1(\hat{I_1}) \cos \hat{\theta},

1775: \label{eqpend}

1776: \end{equation}

1777: with:

1778: \begin{equation}

1779: H^{''}_0 = \frac{\partial^2 H_0}{\partial I^2}.

1780: \label{second_derivative}

1781: \end{equation}

1782: Consistently at lowest order in $\lambda,$ it is not necessary 

1783: to take into

1784: account the dependence of $V_1$ on $\hat{I}.$

1785: 

1786: As already anticipated by the label, ${\cal H}_{\mathrm pend}$ defined

1787: in eq.~(\ref{eqpend}) describes a usual, one-dimensional

1788: pendulum: $\hat\theta$ represents the angle of the pendulum 

1789: with the vertical axis,

1790: $\hat{I}-\hat{I_1}$ its angular velocity, $1/H^{''}_0(\hat{I_1})$

1791: its momentum of inertia

1792: and $\lambda V_1(\hat{I_1})$ the gravitational field.

1793: This equivalence of the secular Hamilton function with that of a pendulum,

1794:  in the

1795: vicinity of the resonant action $\hat{I}_1$, is extremely useful to gain

1796: some physical insight

1797: in the dynamics of any Hamiltonian system close to a resonance. In particular,

1798: it will render our

1799: analysis of non-dispersive wave-packets rather simple.

1800: 

1801: Figure~\ref{pendulum} shows the isovalue lines of

1802: ${\cal H}_{\mathrm pend}$ in the $(\hat{I},\hat{\theta})$ plane, i.e. the

1803: classical phase space trajectories in the presence of the

1804: resonant perturbation. In

1805: the absence of the resonant perturbation, these should be horizontal 

1806: straight lines at

1807: constant  $\hat{I}.$

1808: We observe that the effect of the resonant perturbation is mainly

1809: to create a new structure, called the ``resonance island", located around

1810: the resonant action $\hat{I}_1.$ 

1811: 

1812: \psfull

1813: \begin{figure}

1814: \centerline{\psfig{figure=bdzf08.eps,width=10cm,angle=-90}}

1815: \caption{Isovalues of the Hamiltonian ${\cal H}_{\mathrm pend}$ of a

1816: pendulum in a gravitational field. This Hamiltonian is a good

1817: approximation for the motion of a periodically driven system when

1818: the driving frequency is resonant with the internal frequency

1819: of the system. The stable equilibrium point of the pendulum is surrounded

1820: by an island of librational motion (shaded region). 

1821: This defines the resonance island of the periodically driven system, where

1822: the internal motion is {\em locked} on the external driving. This 

1823: nonlinear phase-locking

1824: phenomenon is essential for the existence of non-dispersive wave-packets.}

1825: \label{pendulum}

1826: \end{figure}

1827: 

1828: To characterize this structure, let us examine the

1829: fixed points of the Hamiltonian~(\ref{eqpend}). They are easily calculated

1830: (imposing $\partial{\cal H}_{\mathrm pend}/\partial\hat{I}=\partial{\cal

1831: H}_{\mathrm pend}/\partial\hat\theta=0$), and

1832: located at

1833: \begin{equation}

1834: \hat{I}=\hat{I_1},\ \ \ \hat{\theta}=0,\ \ {\mathrm with\ energy}\

1835: H_0(\hat{I_1}) -\omega\hat{I}_1 + \lambda V_1(\hat{I_1}),

1836: \label{stablep}

1837: \end{equation}

1838: and

1839: \begin{equation}

1840: \hat{I}=\hat{I_1},\ \ \ \hat{\theta}=\pi,\ \ {\mathrm with\ energy}\

1841: H_0(\hat{I_1}) -\omega\hat{I}_1- \lambda V_1(\hat{I_1}),

1842: \label{resoact}

1843: \end{equation}

1844: respectively.

1845: 

1846: If $H^{''}_0(\hat{I_1})$ (and thus the ``kinetic energy'' part

1847: in ${\cal H}_{\mathrm pend}$)

1848: is positive, the minimum of the potential

1849:  $\lambda V_1(\hat{I_1})\cos\hat{\theta}$ corresponds to a global

1850:  minimum of ${\cal H}_{\mathrm pend}$, and thus to a stable

1851: equilibrium point. The maximum of $\lambda V_1(\hat{I_1})\cos\hat{\theta}$

1852: is a saddle point of ${\cal H}_{\mathrm pend}$, 

1853: and thus 

1854: represents an unstable equilibrium point,

1855: as the standard intuition suggests. For $H^{''}_0(\hat{I}_1)<0$

1856: the situation is reversed -- and less intuitive 

1857: since the

1858: ``kinetic energy'' is negative -- and the {\em maximum}

1859: of ${\cal H}_{\mathrm pend}$ is now a stable equilibrium point,

1860: as the reader may easily check by

1861: standard linear stability analysis in the vicinity of the

1862: fixed point. 

1863: Thus, in compact form, if

1864: $\lambda V_1(\hat{I_1}) H^{''}_0(\hat{I_1})$ is positive,

1865: $\hat{\theta}=\pi$ is

1866: a stable equilibrium point, while $\hat{\theta}=0$ is

1867: unstable. 

1868: If $\lambda V_1(\hat{I_1}) H^{''}_0(\hat{I_1})$ is negative, the stable and unstable

1869: points are interchanged.

1870: 

1871: There are two qualitatively different types of motion:

1872: \begin{itemize}

1873: \item Close to the stable equilibrium point of the pendulum, $\hat{\theta}$

1874: oscillates

1875: periodically, with an amplitude smaller than $\pi$.

1876: This is the ``librational motion'' of the pendulum inside the

1877: resonance

1878: island. Any trajectory started within 

1879: this region of phase space (the shaded area in

1880: fig.~\ref{pendulum})

1881: exhibits 

1882: librational motion. It should be realized

1883: that the resonance island confines the motion to finite intervals in $\hat{I}$

1884: and $\hat{\theta}$, and thereby strongly affects all trajectories with action

1885: close to the resonant action. According to eqs.~(\ref{stablep},\ref{resoact}),

1886: the resonance island is associated with the energy range

1887: $[H_0(\hat{I_1})-\omega\hat{I}_1- |\lambda V_1(\hat{I_1})|,

1888: H_0(\hat{I_1})-\omega\hat{I}_1+ |\lambda

1889: V_1(\hat{I_1})|].$

1890: \item For any initial energy outside that energy range

1891:  the pendulum has sufficient

1892: kinetic energy to rotate. This is the ``rotational

1893: motion'' of the pendulum outside the resonance island, where $\hat{\theta}$ is

1894: an unbounded and

1895: monotonous function of time. Far from the center of the island, the motion

1896: occurs at almost constant unperturbed action $\hat{I}$, with an almost

1897: constant

1898: angular velocity in $\hat{\theta},$ 

1899: tending to the unperturbed

1900: motion. This illustrates that the effect of the perturbation is 

1901: important

1902: for initial conditions close to the resonance island, but negligible for 

1903: non-resonant trajectories.

1904: \end{itemize}

1905: 

1906: The size of the resonance island can be simply estimated from

1907: eq.~(\ref{eqpend}) and

1908: fig.~\ref{pendulum}. The extension in $\hat{\theta}$ is $2\pi$, its width in

1909: $\hat{I}$ (which depends on $\hat{\theta})$ is:

1910: \begin{equation}

1911: \Delta \hat{I} = 4 \sqrt{\left|

1912: \frac{\lambda V_1(\hat{I_1})}{ H^{''}_0(\hat{I_1})}\right|},

1913: \label{width-island}

1914: \end{equation}

1915: and the total area \cite{lichtenberg83}:

1916: \begin{equation}

1917: A(\lambda) = 16 \sqrt{\left|

1918: \frac{\lambda V_1(\hat{I_1})}{ H^{''}_0(\hat{I_1})}\right|}.

1919: \label{area}

1920: \end{equation}

1921: 

1922: The dependence of $A(\lambda)$ on $\sqrt{|\lambda|}$ implies that even

1923: a small perturbation may

1924: induce

1925: significant changes in the phase space structure, provided the pertubation

1926: is resonant.

1927: 

1928: The above picture is valid in the rotating frame defined by

1929: eqs.~(\ref{rotframe_a}-\ref{rotframe_c}). If we go back to the original

1930: action-angle coordinates $(I,\theta),$ the stable (resp. unstable) fixed

1931: point of the secular Hamiltonian is mapped on a

1932: stable (resp. unstable) periodic orbit

1933: whose period is {\em exactly} equal to the period of the driving perturbation,

1934: as a consequence of eq.~(\ref{resonance_condition}).

1935: Any trajectory started in the vicinity of the stable periodic orbit will

1936: correspond to an initial point close to the fixed

1937: point in the rotating frame,

1938: and thus will remain trapped within the resonance island. In the original

1939: coordinate frame, it will appear

1940: as a trajectory evolving close to the stable periodic orbit forever.

1941: In particular, 

1942: the difference in $\theta$ between the stable periodic orbit and any orbit 

1943: trapped in the resonance island remains bounded within $(-\pi,+\pi)$, for arbitrarily 

1944: long times. This means that the phase of any trapped trajectory cannot drift with 

1945: respect to the phase of the periodic orbit.

1946:  As the latter evolves at the driving frequency, we

1947: reach the conclusion that the phase of any trajectory started in the

1948: resonance island

1949: will be {\em locked} on the phase of the driving field. This is the very origin

1950: of the phase locking phenomenon discussed in section~\ref{INTRO3}

1951: above. A crucial point herein

1952: is that

1953: the resonance island

1954: covers a significant part of phase space with finite volume:

1955: it is the

1956: whole structure, not few trajectories, which is phase locked. This is why, 

1957: further down, we will be able to build quantum wave-packets on this structure, 

1958: which will be phase locked to the classical orbit 

1959: and will not spread. The classical version of a non-spreading

1960: wave-packet

1961: thus consists of a family of trajectories, trapped within the resonance island, such

1962: that

1963: this family is invariant under the evolution generated by the pendulum

1964: Hamiltonian. The simplest

1965: possibility

1966: is to sample all trajectories within the ``energy" range

1967: $[H_0(\hat{I_1}) -\omega \hat{I}_1 - |\lambda V_1(\hat{I_1})|,

1968: H_0(\hat{I_1})-\omega \hat{I}_1+ |\lambda

1969: V_1(\hat{I_1})|]$ of the pendulum.

1970: In real space, this will appear as a localized probability density

1971: following the classical stable periodic orbit, reproducing 

1972: its shape exactly after each period of the drive.

1973: 

1974: So far, to derive the characteristics

1975: of the resonance island, we have consistently used

1976: first order

1977: perturbation theory, which is valid for small $\lambda.$ At higher values of

1978: $\lambda,$ higher order terms come into play and modify the shape and the

1979: precise location

1980: of the resonance island. However,

1981: it is crucial to note that the island itself considered as a structure is

1982: robust,

1983: and will survive up to rather high values of $\lambda$ (as a consequence of

1984: the KAM theorem \cite{lichtenberg83}).

1985: Since the size of the resonance island grows with

1986: $|\lambda|$, 

1987: eq.~(\ref{area}), the island may occupy

1988:  a significant

1989: area in phase space and eventually interact with

1990: islands associated

1991: with other resonances, for sufficiently large $|\lambda|$. The mechanism of this ``resonance

1992: overlap'' is rather well understood \cite{chirikov59,chirikov79}: in general,

1993: the motion close to the separatrix (where the period of the classical motion

1994: of the pendulum tends to infinity) is 

1995: most sensitive to higher order 

1996: corrections.

1997: The general scenario

1998: is thus the 

1999: nonintegrable perturbation of the separatrix and the 

2000: emergence of a

2001: ``stochastic" layer of

2002: chaotic motion in phase space, as 

2003: $|\lambda|$ is increased. At

2004: still larger

2005: values of $|\lambda|$, 

2006: chaos may invade large parts of phase space, 

2007: and the resonance island may shrink

2008: and finally

2009: disappear.

2010: While considering  

2011: realistic examples later-on, we shall 

2012: enter the

2013: non-perturbative regime. Let us, however, consider first the quantum

2014: perturbative picture.

2015: 

2016: \subsubsection{Quantum dynamics}

2017: \label{QD}

2018: 

2019: As shown in the previous section, the dynamics of

2020: a one-dimensional system exposed to a weak, resonant, periodic driving

2021: is essentially regular and analogous to the one of a pendulum,

2022: eq.~(\ref{second_derivative})

2023: (in the rotating frame, eqs.~(\ref{rotframe_a}-\ref{rotframe_c})).

2024: In the present section, we will

2025: show that the same

2026: physical picture can be employed in quantum mechanics, to construct

2027: non-dispersive wave-packets. They will follow the stable

2028: classical trajectory locked on the external drive,

2029: and 

2030: exactly reproduce their initial shape 

2031: after 

2032: each period.

2033: 

2034: Our starting point is the time-dependent Schr\"odinger equation

2035: associated with the Hamiltonian (\ref{ham}):\footnote{For simplicity,

2036: we use the same notation for

2037: classical and quantum quantities, the distinction between them

2038:  will

2039: become clear from the context.}

2040: \begin{equation}

2041: i\hbar \frac{d|\psi(t)\rangle}{dt} = (H_0 + \lambda V  \cos \omega t)

2042: |\psi(t)\rangle.

2043: \label{se}

2044: \end{equation}

2045: Since the Hamiltonian (\ref{ham}) is periodic in time, the Floquet

2046: theorem\footnote{The Floquet theorem \protect\cite{floquet1883} in the time

2047: domain is strictly equivalent to the Bloch theorem for potentials

2048: periodic in space \protect\cite{mermin76}.}

2049: guarantees that the general solution of eq.~(\ref{se}) is given by

2050: a linear combination

2051: of elementary, time-periodic states -- the so-called ``Floquet eigenstates'' of

2052: the system -- multiplied by oscillatory functions:

2053: \begin{equation}

2054: |\psi(t)\rangle = \sum_j c_j \exp\left( - i \frac{{\cal E}_jt}

2055: {\hbar}\right) |{\cal E}_j(t)\rangle,

2056: \label{solse}

2057: \end{equation}

2058: with

2059: \begin{equation}

2060: |{\cal E}_j(t+T)\rangle =|{\cal E}_j(t)\rangle .

2061: \label{flstate}

2062: \end{equation}

2063: The ${\cal E}_j$ are the 

2064: ``quasi-energies" of the system.

2065: Floquet states  and quasi-energies are eigenstates

2066: and eigenenergies

2067: of the Floquet Hamiltonian

2068: \begin{equation}

2069:  {\cal H} = H_0 + \lambda V  \cos \omega t - i \hbar \frac{\partial}{\partial

2070:  t}.

2071: \label{calhq}

2072: \end{equation}

2073: Note that, because

2074: of the time-periodicity with period $T=2\pi/\omega$, the quasi-energies are

2075: defined modulo $\hbar \omega$ \cite{shirley65}.

2076: 

2077: The Floquet Hamiltonian (\ref{calhq}) is nothing but the quantum analog of

2078: the classical Hamiltonian (\ref{hamext})

2079: in 

2080: extended phase space.

2081: Indeed, $- i \hbar \partial /\partial t$ is the quantum version of the

2082: canonical momentum $P_t$ conjugate to time $t.$

2083: In 

2084: strict analogy with the classical discussion of the previous section,

2085: it is the Floquet Hamiltonian in 

2086: extended phase space which will be the

2087: central object of our discussion. It contains all the

2088: relevant information on the system, encoded in its eigenstates.

2089: 

2090: In a quantum optics or atomic physics context -- 

2091: with the external perturbation 

2092: given by quantized modes of the electromagnetic field -- the

2093: concept of ``dressed atom" is widely used \cite{cct92}. There, 

2094: a given field mode 

2095: and the atom are treated on an equal footing, as a composite quantum system, 

2096: leading to a time-independent Hamiltonian (energy is conserved for the entire system

2097: comprising atom {\em and} field).

2098: This picture

2099: is 

2100: indeed very close to the Floquet picture. If the field mode is in a 

2101: coherent state \cite{cct92} with a large average number of

2102: photons, 

2103: the electromagnetic field can be treated (semi)classically -- i.e., replaced 

2104: by a $\cos$ time dependence and a fixed amplitude $F$ -- and the

2105: energy spectrum of the dressed atom exactly coincides with the

2106: spectrum of the Floquet Hamiltonian \cite{shirley65}.

2107: 

2108: By its mere definition, eq.~(\ref{flstate}),

2109: each Floquet eigenstate is

2110: associated with a strictly time-periodic probability density in configuration

2111: space.

2112: Due to this periodicity with the period of the driving field,

2113: the probability density of a Floquet eigenstate in general changes its shape

2114: as time evolves, but recovers its initial shape after 

2115: each period.

2116: Hence, the Floquet picture provides clearly the simplest approach to

2117: non-dispersive wave-packets. Given the ability to build a Floquet

2118: state which is well localized at a given phase of the driving field, it will

2119: {\em automatically} represent a non-dispersive wave-packet.

2120: In our opinion, this is a much simpler approach than the attempt to

2121: build an {\it a priori} localized wave-packet and try to minimize

2122: its spreading during the subsequent evolution

2123: \cite{farrelly95a,ibb95,ibb94,kalinski95a}.

2124: 

2125: Note that also

2126: the reverse property holds true. Any state with $T$-periodic probability

2127: density (and, in particular, any localized wave-packet propagating along a

2128: $T$-periodic classical orbit)

2129: has to be a single Floquet eigenstate:

2130: Such a state can be expanded into the Floquet eigenbasis, and during one

2131: period, the various components of the expansion accumulate phase factors

2132: $\exp(-i{\mathcal E}_iT/\hbar).$ Hence, 

2133: the only solution which allows for a $T$-periodic density is

2134: a one-component expansion, i.e., a single Floquet eigenstate

2135: \footnote{One

2136: might argue that Floquet states differing in energy by an

2137: integer multiple of $\hbar\omega$ could be used. However, 

2138: the Floquet spectrum is $\hbar\omega$-periodic by construction

2139: \cite{shirley65}, and two

2140: such states

2141: represent the same physical state.}.

2142: 

2143: To summarize, the construction of non-dispersive

2144: wave-packets in a time-periodic system is equivalent to finding

2145: localized Floquet eigenstates. The existence of such states is far

2146: from obvious, as the Floquet spectrum is usually very complex, composed of

2147: quasi-bound states, resonances and continua. This is why

2148: a semiclassical analysis can be very helpful in finding these objects.

2149: 

2150: \subsubsection{Semiclassical approximation}

2151: \label{sapp}

2152: 

2153: 

2154: Dealing with highly excited states, a semiclassical

2155: approximation can be used to determine quasi-energies and

2156: Floquet eigenstates~\cite{breuer91}. If the driving perturbation is sufficiently weak,

2157: we have shown in section~\ref{CD} that the classical dynamics close to a

2158: nonlinear resonance is

2159: essentially regular and accurately described by the pendulum

2160: Hamiltonian~(\ref{eqpend}). It describes a system with two

2161: degrees of freedom (along $t$ and $\hat\theta$, with their

2162: conjugate momenta $\hat{P}_t$ and $\hat{I}$, respectively)

2163: which is essentially regular. For semiclassical quantization, 

2164: we may then

2165: use the standard EBK rules, eq.~(\ref{EBK}), introduced in section~\ref{SQ}.

2166: 

2167: The momentum $\hat{P}_t$ is a constant of the motion,

2168: and the isovalue curves of $\hat{P}_t,\hat{I},\hat{\theta}$

2169: lying on the

2170: invariant tori

2171: can be used for the EBK quantization scheme. 

2172: Along such a curve, $t$

2173: evolves from $0$ to $2\pi/\omega$, with $\hat{\theta}=\theta-\omega t$

2174: kept constant. Thus, $\theta$ itself is changed by $2\pi$, what

2175: implies that  the Maslov index $\mu$ of the unperturbed $(I,\theta)$ motion

2176: has to be included,

2177: leading to the following quantization condition for $\hat{P}_t$,

2178: \begin{equation}

2179: \frac{1}{2\pi}\int_0^{T}\hat{P}_tdt=\frac{\hat{P}_tT}{2\pi}=

2180: \left(k+\frac{\mu}{4}\right)\hbar ,

2181: \label{ebkpt}

2182: \end{equation}

2183: with integer $k$.

2184: Since $T=2\pi/\omega$ is just the period of the resonant driving,

2185: we get the quantized values of $\hat{P}_t:$ 

2186: \begin{equation}

2187: \hat{P}_t = \left(k+\frac{\mu}{4}\right)\ \hbar\omega

2188: \label{ebkpt2}

2189: \end{equation}

2190: which are equally spaced by

2191: $\hbar \omega.$ Thus, we recover semiclassically 

2192: the $\omega$-periodicity of the Floquet spectrum.

2193: 

2194: For the motion in the $(\hat{I},\hat{\theta})$ plane, we can use

2195: the isocontour lines

2196: of the pendulum Hamiltonian $H_{\mathrm pend}$, eq.~(\ref{eqpend}), 

2197: as closed paths,

2198: keeping $\hat{P}_t$ and $t$ constant.

2199: Depending on the nature of the pendulum motion (librational or

2200: rotational), the topology of the closed paths is different,

2201: leading to distinct expressions:

2202: \begin{itemize}

2203: \item For trapped librational motion, inside the resonance island,

2204: the path is isomorphic to a circle in the $(\hat{I},\hat{\theta})$ plane,

2205: with a Maslov index equal to two. The quantization condition

2206: is:

2207: \begin{equation}

2208: \frac{1}{2\pi}\oint \hat{I}d\hat{\theta}=\left(N+\frac{1}{2}\right)\hbar,\

2209: ({\rm librational\ motion})

2210: \label{ebkires}

2211: \end{equation}

2212: with $N$ a non-negative integer.

2213: Of special interest is the ``fundamental" state, $N=0$, which exhibits

2214: maximum localization within the resonance island and

2215: is therefore

2216: expected to represent the optimal non-dispersive wave-packet.

2217: \item For unbounded rotational motion, outside the

2218: resonance island, the path includes a $2\pi$ phase change for $\theta$

2219: and acquires the Maslov index of the unperturbed motion:

2220: \begin{equation}

2221: \frac{1}{2\pi}\int_{0}^{2\pi} \hat{I}d\hat{\theta}=\left(

2222: N+\frac{\mu}{4}\right)\hbar,\ ({\rm rotational \ motion}).

2223: \label{ebkinres}

2224: \end{equation}

2225: \end{itemize}

2226: 

2227: This semiclassical quantization scheme is expected to work provided

2228: the classical phase space velocity is sufficiently large, see section~\ref{SQ}.

2229: This may fail close to the stable and unstable fixed points where

2230: the velocity vanishes. Near the stable equilibrium

2231: point, the expansion of the Hamiltonian at second order leads to an

2232: approximate harmonic Hamiltonian with frequency:

2233: \begin{equation}

2234: \omega_{\mathrm harm} = \sqrt{|\lambda V_1(\hat{I_1})H^{''}_0(\hat{I_1})|}.

2235: \label{omega_harmonic}

2236: \end{equation}

2237: 

2238: In this harmonic approximation, the semiclassical quantization

2239: is known to be exact~\cite{landau2}.

2240:  Thus, close to the stable

2241: equilibrium, the quasi-energy

2242: levels, labeled by the non-negative integer $N$, are

2243: given by the harmonic approximation. There are various cases which

2244: depend on the signs

2245: of $H^{''}_0(\hat{I_1})$, $V_1(\hat{I_1})$, and $\lambda$,

2246:  with the general

2247: result

2248: given by

2249: \begin{equation}

2250: {\mathcal E}_{N,k} = H_0(\hat{I_1}) - \omega \hat{I_1} +

2251: \left(k+\frac{\mu}{4}\right)\hbar\omega

2252: -\ {\mathrm sign}(H^{''}_0(\hat{I_1}))

2253: \left[ |\lambda V_1(\hat{I_1})| -

2254: \left(N+\frac{1}{2}\right) \hbar \omega_{\mathrm harm} \right].

2255: \label{spectrum_harmonic}

2256: \end{equation}

2257: 

2258: For $N=0,k=0,$ this gives

2259:  a fairly accurate estimate of the

2260: energy of the non-dispersive wave-packet with optimum localization.

2261: The EBK semiclassical scheme provides us also with some

2262: interesting information on the eigenstate. Indeed, the invariant

2263: tori considered here are tubes surrounding the resonant

2264: stable periodic orbit. They cover the $[0,2\pi]$ range

2265: of the $t$ variable but are well localized in the

2266: transverse $(\hat{I}=I,\hat{\theta}=\theta-\omega t)$ plane,

2267: with an approximately Gaussian phase space distribution.

2268: Hence, at any fixed time $t$, the Floquet

2269: eigenstate will appear as a Gaussian distribution

2270: localized around the point $(I=\hat{I_1},\theta=\omega t).$ 

2271: As this point precisely defines the resonant, stable

2272: periodic orbit, one expects the $N=0$ state to be

2273: a Gaussian wave-packet following the classical orbit.

2274: In the original $(p,z)$ coordinates, the width of the

2275: wave-packet will depend on the system under consideration

2276: through the change of variables $(p,z)\to (I,\theta),$

2277: but the Gaussian character is expected to be approximately valid

2278: for both the phase space density and the configuration space

2279: wave-function, as long as the change of variables is smooth.

2280: 

2281: Let us note that low-$N$ states may be considered as excitations

2282: of the $N=0$ ``ground'' state. Such states has been termed

2283: ``flotons'' in \cite{holthaus95} where

2284: their wave-packet character was, however, not considered.

2285: 

2286: The number of eigenstates trapped within the resonance island -- i.e.

2287: the number of non-dispersive wave-packets -- is easily evaluated

2288: in the semiclassical limit, as it is the maximum $N$ with librational

2289: motion.  It is roughly the area of the

2290: resonance island, eq.~(\ref{area}), divided by $2\pi\hbar$:

2291: \begin{equation}

2292: {\mathrm Number\ of\ trapped\ states} \simeq \frac{8}{\pi \hbar}

2293: \sqrt{\left|\frac{\lambda V_1(\hat{I_1})}{ H^{''}_0(\hat{I_1})}\right|}.

2294: \label{number_of_trapped_states}

2295: \end{equation}

2296: 

2297: Near the unstable fixed point -- that is at the energy which separates

2298: librational and rotational motion -- the semiclassical quantization fails 

2299: because of the critical slowing down in its  vicinity~\cite{marion}.

2300: The corresponding

2301: quantum states -- known as separatrix states~\cite{leopold94} -- are

2302: expected to be dominantly

2303: localized near the unstable fixed point, simply because the classical

2304: motion there slows down, and the pendulum spends more time close to its

2305: upright position. This localization is once again of purely classical origin,

2306: but not perfect: some part of the wave-function

2307: must be also localized 

2308: along the separatrix, which autointersects

2309: at the hyperbolic fixed point. Hence, the Floquet eigenstates associated with

2310: the unstable fixed points

2311: are not expected to form non-dispersive wave-packets with optimum

2312: localization (see also sec.~\ref{scars}).

2313: 

2314: \subsubsection{The Mathieu approach}

2315: \label{section_mathieu}

2316: The pendulum approximation, eq.~(\ref{eqpend}), for a resonantly driven system can also be found

2317: by a pure quantum description~\cite{berman77,holthaus95}. Let us consider

2318: a Floquet state of the system.

2319: Its spatial part can

2320: be expanded in the eigenbasis of the unperturbed Hamiltonian $H_0$,

2321: \begin{equation}

2322: H_0 |\phi_n\rangle = E_n|\phi_n\rangle ,

2323: \end{equation}

2324: while the time-periodic wave-function can be

2325: expanded in a Fourier series. One obtains:

2326: \begin{equation}

2327: |\psi(t)\rangle =\ \sum_{n,k} c_{n,k}

2328: \ \exp (-ik\omega t)\ \ |\phi_n\rangle,

2329: \label{wfmat}

2330: \end{equation}

2331: where the coefficients $c_{n,k}$ are to be determined.

2332: The Schr\"odinger equation for the Floquet states (eigenstates of ${\cal H}$),

2333: eq.~(\ref{calhq}), with quasi-energies ${\cal E}$ and 

2334: time dependence (\ref{solse}),

2335: reads:

2336: \begin{equation}

2337: c_{n,k}({\cal E}+k\hbar \omega - E_n) = \frac{\lambda}{2} \ \sum_{p}

2338: {\langle \phi_n|V|\phi_{n+p}\rangle\ (c_{n+p,k+1} + c_{n+p,k-1})}.

2339: \label{coupled_equations}

2340: \end{equation}

2341: 

2342: For $\lambda=0,$ the solutions of eq.~(\ref{coupled_equations})

2343: are trivial: ${\cal E}=E_n-k\hbar\omega$, which is nothing

2344: but the unperturbed energy spectrum modulo $\hbar\omega.$

2345: \footnote{Note that, as a consequence of the negative sign of the argument 

2346: of the exponential factor in eq.~(\ref{wfmat}), the energy shift $k\hbar\omega$ 

2347: appears here with a negative sign in the expression for $\cal E$ -- in contrast 

2348: to semiclassical expressions alike eq.~(\ref{spectrum_harmonic}), where we chose

2349: the more suggestive positive sign. Since $k=-\infty\ldots +\infty$, both conventions 

2350: are strictly equivalent.}

2351: In the presence of a small perturbation, only quasi-degenerate states

2352: with

2353: values close to

2354: $E_n-k\hbar\omega$ will be efficiently coupled. In the semiclassical limit,

2355: see section \ref{SQ}, eq.~(\ref{cprinc}),

2356: the unperturbed eigenenergies $E_n,$ labeled by a non-negative

2357: integer,

2358: are locally approximately spaced by $\hbar \Omega,$

2359: where $\Omega$ is the frequency of the unperturbed classical motion.

2360: Close to 

2361: resonance, $\Omega \simeq \omega$ (eq.~(\ref{resonance_condition})), and 

2362: thus: 

2363: \begin{equation}

2364: E_{n}-k \hbar \omega \simeq E_{n+1} - (k+1) \hbar \omega \simeq  E_{n+2} -

2365: (k+2)\hbar \omega\simeq  ...,

2366: \label{selrule}

2367: \end{equation}

2368: so that only states with the same value of $n-k$ will be efficiently coupled.

2369: 

2370: The first approximation is thus to neglect the couplings which do not

2371: preserve $n-k.$ This is just the quantum version of the

2372: secular approximation for the classical dynamics.

2373: Then, the set of equations~(\ref{coupled_equations})

2374: can be rearranged in independent blocks, each subset being characterized by $n-k$.

2375: The various subsets are in fact identical, except for a shift in energy

2376: by an integer multiple of $\hbar\omega.$ This is nothing but the

2377: $\hbar\omega$-periodicity  

2378: of the Floquet spectrum already encountered in secs.~\ref{QD} and \ref{sapp}.

2379: As a consequence, we can

2380: consider

2381: the

2382: $n-k=0$ block alone.

2383: 

2384: Consistently, since eq.~(\ref{selrule}) is valid close to the center

2385: of the resonance only, one can expand the quantities of interest in the

2386: vicinity of the center of the resonance, and use semiclassical 

2387: approximations for 

2388: matrix elements of $V.$ Let $n_0$ denote the effective, resonant quantum number

2389: such that, with eqs.~(\ref{cprinc}) and (\ref{resonance_condition}),

2390: \begin{equation}

2391: \left. \frac{dE_n}{dn} \right|_{n=n_0}= \hbar \omega.

2392: \label{space}

2393: \end{equation}

2394: Note that, by this definition, $n_0$ is {\em not} necessarily an integer.

2395:  In the semiclassical

2396: limit, where $n_0$ is large, the WKB approximation connects $n_0$ to the center

2397: of the classical resonance island, see eq.~(\ref{semen}),

2398: \begin{eqnarray}

2399: &&n_0 + \frac{\mu}{4} = \frac{\hat{I_1}}{\hbar},\\

2400: &&E_{n_0} = H_0(\hat{I_1}),

2401: \end{eqnarray}

2402: with $\mu$ being the Maslov index along the resonant trajectory.

2403: Furthermore, for $n$ close to $n_0,$ we can expand

2404: the unperturbed energy at second order in $(n-n_0)$,

2405: \begin{equation}

2406: E_n \simeq E_{n_0} + (n-n_0) \hbar \omega + \frac{1}{2}

2407: \ \left.\frac{d^2E_n}{dn^2}\right|_

2408: {n_0} \ (n-n_0)^2,

2409: \end{equation}

2410: where

2411: the second derivative $d^2E_n/dn^2$

2412: is directly related to the classical quantity $H^{''}_0$, see

2413: eq.~(\ref{second_derivative}), within the semiclassical WKB

2414: approximation, eq.~(\ref{semen}). Similarly, the matrix elements of $V$

2415: are related to the classical Fourier components

2416: of the potential~\cite{landau2}, eq.~(\ref{fourier_components}),

2417: \begin{equation}

2418: \langle \phi_n|V|\phi_{n+1} \rangle \simeq 

2419: \langle \phi_{n+1}|V|\phi_{n+2} \rangle \simeq V_1(\hat{I_1}),

2420: \label{matel_mat}

2421: \end{equation}

2422: evaluated at the center $\hat{I}=\hat{I}_1$ -- 

2423: see eq.~(\ref{resonance_condition}) --

2424: of the resonance zone.

2425: 

2426: With these ingredients and $r=n-n_0$, eq.~(\ref{coupled_equations}) 

2427: is transformed

2428: in the following set of approximate equations:

2429: \begin{equation}

2430: \left[ {\cal E} - H_0(\hat{I_1}) + 

2431: \omega\left(\hat{I}_1 -\frac{\mu\hbar}{4}\right) - 

2432: \frac{\hbar^2}{2} H^{''}_0(\hat{I_1}) r^2

2433: \right] d_r =

2434: \lambda V_1 (d_{r+1} + d_{r-1}),

2435: \label{tridiag}

2436: \end{equation}

2437: where

2438: \begin{equation}

2439: d_r \equiv c_{n_0+r,n_0+r}.

2440: \label{diag_mat}

2441: \end{equation}

2442: 

2443: Note that, because of eq.~(\ref{space}),

2444: the $r$ values are not necessarily integers, but all have the same

2445: fractional part.

2446: 

2447: The tridiagonal set of coupled equations (\ref{tridiag})

2448: can be rewritten as a differential

2449: equation. Indeed, if one introduces the following function associated with

2450: the Fourier

2451: components $d_r$,

2452: \begin{equation}

2453: f(\phi) = \sum_r{\exp (ir\phi) d_r},

2454: \label{dual}

2455: \end{equation}

2456: eq.~(\ref{tridiag}) can be written as

2457: \begin{equation}

2458: \left[- \frac{\hbar^2}{2} H^{''}_0(\hat{I_1}) \frac{d^2 }{d\phi^2} +

2459: H_0(\hat{I_1}) -

2460: \omega\left(\hat{I}_1 -\frac{\mu\hbar}{4}\right)

2461: + \lambda V_1 \cos \phi \right] f(\phi) = {\cal E} f(\phi),

2462: \label{mathieu}

2463: \end{equation}

2464: which is nothing but the quantum version of the pendulum

2465: Hamiltonian,

2466: eq.~(\ref{eqpend}).

2467: Thus, the present calculation is just the purely quantum 

2468: description of the non-linear resonance phenomenon.

2469: The dummy variable

2470: $\phi$

2471: introduced for convenience coincides with the classical angle

2472: variable  $\hat{\theta}$. 

2473: In general, $r$ is not an integer, so that the various

2474: $\exp (ir\phi)$ in eq.~(\ref{dual}) are not periodic functions

2475: of $\phi.$ However, as all $r$ values have the same fractional part,

2476: if follows that $f(\phi)$ must satisfy ``modified'' 

2477: periodic boundary conditions of the form

2478: \begin{equation}

2479: f(\phi + 2\pi) = \exp (-2i\pi n_0) f(\phi).

2480: \label{boundary}

2481: \end{equation}

2482: The reason for this surprising boundary condition is clear:

2483: $r=n-n_0$ is the quantum analog of $\hat{I}-\hat{I}_1.$

2484: In general, the resonant action $\hat{I}_1$ 

2485: is {\em not} an integer or half-integer multiple

2486: of $\hbar$ -- exactly as $n_0$ is not an integer. The semiclassical

2487: quantization, eq.~(\ref{ebkinres}), which expresses

2488: the $\hat{\theta}$ periodicity of the eigenstate,

2489: applies for the $\hat{I}$ variable. When expressed 

2490: in terms of the variable $\hat{I}-\hat{I}_1$,

2491: it contains the additional phase shift 

2492: present in eq.~(\ref{boundary}).

2493:   

2494: Few words of caution are in order: the equivalence

2495: of the semiclassical quantization with the pure

2496: quantum approach holds in the semiclassical limit only,

2497: when the quantum problem

2498: can be mapped on a pendulum problem. In the general case, it is not possible to

2499: define a quantum angle variable \cite{loudon}.

2500: Hence, the quantum treatment presented here

2501: is no

2502: more general or more powerful than the semiclassical treatment. They both rely

2503: on the same approximations and have the same limitations: perturbative

2504: regime (no overlap of resonances) and semiclassical approximation.

2505: 

2506: Finally, eq.~(\ref{mathieu}) can be written in its standard form, known as

2507: the ``Mathieu equation" \cite{abramowitz72}:

2508: \begin{equation}

2509: \frac{d^2y}{dv^2} + (a-2q\cos 2v) y = 0.

2510: \label{mat_eq}

2511: \end{equation}

2512: The correspondence with eq.~(\ref{mathieu}) is established via:

2513: \begin{eqnarray}

2514: &&\phi = 2v,

2515: \label{map_mathieu1}\\

2516: &&a=\frac{8\left[{\cal E}-H_0(\hat{I_1})+

2517: \omega\left(\hat{I}_1 -\frac{\mu\hbar}{4}\right)

2518: \right]}{\hbar^2 H^{''}_0(\hat{I_1})},

2519: \label{map_mathieu2}\\

2520: &&q=\frac{4\lambda

2521: V_1}{\hbar^2 H^{''}_0(\hat{I_1})}.

2522: \label{map_mathieu3}

2523: \end{eqnarray}

2524: 

2525: The boundary condition, eq.~(\ref{boundary}),

2526: is fixed by the so-called ``characteristic exponent"

2527: in the Mathieu equation,

2528: \begin{equation}

2529: \nu = -2 n_0\ \ \ ({\mathrm mod}\ 2).

2530: \label{cexp}

2531: \end{equation}

2532: 

2533: The Mathieu equation has solutions (for a given characteristic exponent)

2534: for a discrete set of values of $a$ only. That implies quantization of

2535: the quasi-energy levels, according to eq.~(\ref{map_mathieu2}).

2536: The quantized values $a_{\kappa}(\nu,q)$

2537: depend on $q$ and $\nu$, and are labeled\footnote{In 

2538: the standard text books as \protect\cite{abramowitz72},

2539: the various solutions of the Mathieu equation are divided in 

2540: odd and even solutions, and furthermore in $\pi$- and $2\pi$-periodic 

2541: functions. In our case, only the

2542: ``$a_{2p}$" and ``$b_{2p}$" (in the language of \protect\cite{abramowitz72})

2543: are to be considered.} 

2544: by a non-negative

2545: integer $\kappa.$

2546: They are well known

2547: -- especially asymptotic expansions are available both in the

2548: small and in the large $q$ regime -- and

2549: can be found in standard handbooks

2550: \cite{abramowitz72}. For example, fig.~\ref{fig_mathieu}(a) shows the

2551: first $a_{\kappa}(q)\equiv a_{\kappa}(\nu=0,q)$ curves for the case of 

2552: ``optimal" resonance (see below), where

2553: $n_0$ is an integer and thus

2554: the characteristic exponent $\nu$ vanishes.

2555: Equivalently, the figure can be interpreted as the evolution of the

2556: energy levels of a pendulum

2557: with the gravitational field.

2558: 

2559: 

2560: \psfull

2561: \begin{figure}

2562: \centerline{\psfig{figure=bdzf09a.ps,width=7cm,angle=-90}

2563: \psfig{figure=bdzf09b.eps,width=7cm,angle=-90}}

2564: \caption{Eigenvalues $a_{\kappa}(q)$ of the Mathieu equation, 

2565: for a characteristic

2566: exponent $\nu=0.$ These represent the energy levels of a pendulum as a function

2567: of the gravitational field, see

2568: eqs.~(\protect\ref{eqpend},\protect\ref{mathieu}). 

2569: (a): Eigenvalues in the range $[-2q,2q]$ are

2570: associated with the librational bounded motion of the pendulum, while

2571: eigenvalues above $2q$ are associated with rotational modes.

2572: The dotted lines represent the energies of the stable equilibrium

2573: point (lower line) and of the unstable equilibrium point (separatrix, upper

2574: line). Near the separatrix, the classical motion slows down and the

2575: quantal energy levels get closer. (b): Details for the first excited states

2576: together with the semiclassical WKB prediction for the energy levels

2577: (dashed lines, eqs.~(\protect\ref{semen})). 

2578: The semiclassical prediction is very accurate, except

2579: in the vicinity of the separatrix.}

2580: \label{fig_mathieu}

2581: \end{figure}

2582: 

2583: The quasi-energy levels of the driven

2584: system can now be expressed as a function of $a_{\kappa}(\nu,q):$

2585: \begin{equation}

2586: {\cal E}_{\kappa} = H_0(\hat{I_1}) - 

2587: \omega\left(\hat{I}_1 -\frac{\mu\hbar}{4}\right)

2588: + \frac{\hbar^2}{8} H^{''}_0(\hat{I_1})\ a_{\kappa}(\nu,q).

2589: \label{prediction_mathieu}

2590: \end{equation}

2591: Together with eqs.~(\ref{map_mathieu1})-(\ref{map_mathieu3}),

2592: this equation gives the quasienergy

2593: levels of a periodically driven system in the vicinity of the resonance zone.

2594: The full, quasi-resonant part of the Floquet spectrum of the driven system is built

2595: from these quantized values through shifts $k\hbar\omega,$ with

2596: arbitrary integer values of $k$.

2597: 

2598: A visual inspection of fig.~\ref{fig_mathieu} immediately shows the existence

2599: of two regions in the energy diagram: within the ``inner region",

2600: $|a_{\kappa}(\nu,q)| \leq 2q,$ the energy levels

2601: form a regular fan of curves 

2602: and tend to decrease with $q.$ On the contrary,

2603: for $a_{\kappa}(\nu,q) > 2q,$ the energy levels increase with $q$. 

2604: Around $a_{\kappa}(\nu,q)=2q,$

2605: a transition region is visible with a series of apparent avoided crossings

2606: between the levels. This has a simple semiclassical explanation.

2607: The stable fixed point of the pendulum described by the Mathieu

2608: equation (\ref{mat_eq}) lies at $v=0$ with energy $-2q$,

2609: what explains why

2610: the $a_{\kappa}(\nu,q)$ values are always larger. The unstable

2611: fixed point has an energy $+2q.$ Thus, in the range 

2612: $a_{\kappa}(\nu,q) \in [-2q,2q],$

2613: the pendulum

2614: is trapped in a region of

2615: librational motion. The energy levels can be approximated

2616: using the standard WKB quantization in the $(\hat{I},\hat{\theta)}$

2617: plane, as described in section~\ref{CD}. 

2618: The number of such states is given by eq.~(\ref{number_of_trapped_states})

2619: which can be rewritten, using eq.~(\ref{map_mathieu3}), as :

2620: \begin{equation}

2621: {\mathrm Number\ of\ trapped\ states} \simeq \frac{4\sqrt{|q|}}{\pi}.

2622: \label{number_of_trapped_states_2}

2623: \end{equation}

2624: At the center of the island (states with small $\kappa$

2625: and/or large $|q|$), an

2626: approximate expression for $a_{\kappa}(\nu,q)$ reads \cite{abramowitz72}:

2627: \begin{equation}

2628: a_{\kappa}(\nu,q) \approx -2|q|+4\left(\kappa+\frac{1}{2}\right) \sqrt{|q|}.

2629: \label{math_asy}

2630: \end{equation}

2631: It does not depend on $\nu$ (what physically means in our case 

2632: that the states deeply inside the resonance island are insensitive

2633: to the boundary condition).

2634: When inserted in eq.~(\ref{prediction_mathieu}), it

2635: yields exactly the energy levels of eq.~(\ref{ebkires}), with

2636: $\kappa=N.$ Hence, the

2637: Mathieu approach agrees with the harmonic approximation within the

2638: resonance island, for sufficiently large islands.

2639:  

2640: For $a_{\kappa}(\nu,q) > 2q,$ the

2641: pendulum undergoes

2642: a rotational, unbounded motion, which again can be quantized using WKB.

2643: For small $|q|$, 

2644: $a_{\kappa}(\nu,q)\approx 4([(\kappa+1)/2]-\nu/2)^2$ (where

2645: $[\ ]$ stands for the integer part), see \cite{abramowitz72}. 

2646: With this  in

2647: eq.~(\ref{prediction_mathieu}), 

2648: one recovers the known Floquet spectrum, in the limit of a

2649: vanishingly  weak perturbation. 

2650: Note, however, that in this rotational mode, the eigenstates

2651: {\em are} sensitive to the boundary conditions (and the $a_{\kappa}(\nu,q)$

2652: values depend on $\nu$). This is essential for

2653: the correct $\lambda \rightarrow 0$ limit.

2654: 

2655:  

2656: Around $a_{\kappa}(\nu,q) = 2q,$ 

2657: the pendulum is close to the separatrix between

2658: librational and rotational motion: the period of the classical motion

2659: tends to infinity (critical slowing down). That explains the locally enhanced

2660: density of states apparent in fig.~\ref{fig_mathieu}(a).

2661: 

2662: In fig.~\ref{fig_mathieu}(b), we also plot the semiclassical WKB

2663: prediction for the

2664: quantized $a_{\kappa}(q)$ values. 

2665: Obviously, the  agreement with the exact ``quantum"

2666: Mathieu result is very good, even for weakly excited states, except

2667: in the vicinity of the separatrix. This is not unexpected because the

2668: semiclassical

2669: approximation is known

2670: to break down near the unstable fixed point, see

2671: sections~\ref{SQ} and \ref{sapp} above.

2672: 

2673: The Mathieu equation

2674: yields accurate predictions for

2675: properties of

2676: non-dispersive wave-packets in  periodically driven

2677: quantum systems. Indeed, in the range  $a_{\kappa}(\nu,q) \in [-2q,2q],$ 

2678: the classical

2679: motion is trapped inside the resonance island, and the corresponding

2680: quantum eigenstates are expected to be non-dispersive wave-packets.

2681: In particular,

2682: the lowest state in the resonance island, associated

2683: with the ground state of the pendulum $\kappa=0$, 

2684: corresponds to the semiclassical eigenstate $N=0$, see eq.~(\ref{ebkires}), 

2685: and represents 

2686: the non-dispersive wave-packet

2687: with the best localization properties.

2688: As can be seen in fig.~\ref{fig_mathieu}(b), the

2689: semiclassical quantization for this state is in excellent agreement with the

2690: exact Mathieu result. This signifies that eq.~(\ref{spectrum_harmonic})

2691: can be used for {\em quantitative} predictions of the quasi-energy of this

2692: eigenstate.

2693: 

2694: As already mentioned,

2695: the characteristic exponent $\nu$ does not play a major role inside the

2696: resonance island, as the eigenvalues  $a_{\kappa}(\nu,q)$ there depends

2697: very little on $\nu.$ However, it is an important parameter outside the

2698: resonance, close to the separatrix especially at small $q.$ Indeed,

2699: at $q=0$, the minimum eigenvalue is obtained

2700: for $\nu=0: a_{0}(\nu=0,q=0)=0.$ 

2701: This implies that, even for very small $q,$ the ground state

2702: enters most rapidly the resonance island. On the opposite, the

2703: worst case is $\nu=1$ where the lowest eigenvalue is doubly degenerate:

2704: $a_{0}(\nu=1,q=0)=a_{1}(\nu=1,q=0)=1.$ 

2705: As we are interested in the ground state of the

2706: pendulum (the one with maximum localization), 

2707: the situation for $\nu=0$ is preferable: not only the state enters rapidly the

2708: resonance island, but it is also separated from the other states by an

2709: energy gap and is thus more robust versus any perturbation.

2710: We will call this situation ``optimal resonance". From eq.~(\ref{cexp}),

2711: it is associated with an integer value of $n_0.$ On the contrary, a

2712:  half-integer value of $n_0$ corresponds to $\nu=1$ and the

2713:  least optimal case. 

2714: 

2715: 

2716: \subsection{Rydberg states in external fields}

2717: \label{RSEF}

2718: 

2719: \subsubsection{Rydberg atoms}

2720: 

2721: In order to construct non-dispersive wave-packets, a quantum

2722: system subject to periodic driving with classically non-linear dynamics is

2723: needed. The latter requirement rules out the harmonic oscillator,

2724: and all its variants. The simplest periodic driving is certainly

2725: provided by

2726: an externally applied, monochromatic electromagnetic field.

2727: Extremely stable, tunable and well controlled sources

2728: exist over a wide range of frequencies.

2729: 

2730: Furthermore, incoherent

2731: processes which destroy the phase coherence of the quantum wave-function have

2732: to be minimized. Otherwise, they will spoil

2733: the localization properties

2734: of the non-dispersive wave-packets and -- in the worst case -- destroy

2735: them completely. Therefore, the characteristic time scales of the

2736: incoherent processes should be at least much longer than the period of the

2737: driving field.

2738: In this respect, atomic electrons appear as

2739: very good candidates,

2740: since -- given suitable experimental conditions -- atoms can be

2741: considered as practically isolated from the external world, with

2742: spontaneous emission of photons as the only incoherent process.

2743: Spontaneous emission is usually a very slow 

2744: mechanism, especially for highly

2745: excited states: the spontaneous life-time of typical atomic states

2746: is at least four or five orders

2747: of magnitude longer than the classical Kepler period (typically

2748: nanoseconds vs. femtoseconds for weakly excited states~\cite{bethe77}).

2749: 

2750: The Coulomb interaction between  

2751: the nucleus and the

2752: electrons is highly

2753: non-linear, which is very favourable. The efficiency

2754: of the coupling with an external electromagnetic

2755: field is known to increase rapidly with the degree of excitation of the atom

2756: \cite{bethe77}.

2757: As we have seen in the preceeding sections~\ref{CD} and \ref{QD},

2758: non-dispersive wave-packets are the quantum mechanical counterparts

2759: of nonlinear resonances in periodically driven Hamiltonian systems, where

2760: the period of the drive matches some intrinsic time scale of the unperturbed

2761: Hamiltonian dynamics.

2762: Due to the immediate correspondence between the classical Kepler problem

2763: and the hydrogen atom, the relevant time scale in this simplest atomic system

2764: is the

2765: unperturbed classical Kepler period, which 

2766: -- compare eq.~(\ref{cprinc}) --

2767: coincides with the inverse level spacing between neighbouring

2768: eigenstates of the unperturbed atom, for large quantum numbers.

2769: Hence, the driving field frequency has to be chosen resonant

2770: with an atomic transition

2771: in the Rydberg regime, typically around the

2772: principal quantum number $n_0=60$. This is the microwave domain,

2773:  where excellent sources exist.

2774: Thus, we believe that atomic Rydberg states are very well suited

2775: for the experimental preparation of non-dispersive wave-packets\footnote{Note,

2776: however, that this is a specific choice. Any driven quantum system 

2777: with a sufficiently high denisty of states and mixed regular-chaotic 

2778: classical dynamics will exhibit nondispersive wave-packets. Since a mixed 

2779: phase space structure is the generic scenario for dynamical systems,

2780: nondispersive wave-packets are expected to be a completely general and 

2781: ubiquitous phenomenon.}.

2782: 

2783: In most cases, the energy scale involved in the dynamics of Rydberg

2784: electrons is so small that the inner electrons of the ionic core

2785: can be considered as frozen and ignored. Thus, we will consider

2786: mainly the hydrogen atom as the simplest prototype. Multi-electron

2787: effects are discussed in section~\ref{HE}. Alternative systems

2788: for observing non-dispersive wave-packets are considered in section

2789: \ref{OGB}.

2790: 

2791: Compared to our simple one-dimensional model introduced in section~\ref{GM} 

2792: above,

2793: an {\em atom}

2794: displays a couple of additional features:

2795: \begin{itemize}

2796: \item A real hydrogen atom is a three-dimensional (3D) system. 

2797: However, it is a degenerate system, 

2798: because the energy depends on the principal quantum number

2799: $n$ only, but not on the angular or magnetic quantum numbers $L$ and $M$,

2800: respectively.

2801: Thus, the structure of the energy levels, which is crucial for the

2802: properties of the non-dispersive wave-packet, see section \ref{QD},

2803: is identical in the 1D and 3D cases. Before discussing the properties

2804: of 3D wave-packets in sections~\ref{LIN3D} to \ref{MA}, we will

2805: consider a simplified 1D model of the hydrogen atom in section~\ref{LIN1D}.

2806: \item Although spontaneous emission is a weak incoherent process,

2807: it 

2808: nontheless limits the life time of the non-dispersive wave-packets which may

2809: decay to lower lying states, losing the phase coherence

2810: of the electronic wave-function. Non-dispersive wave-packets

2811: exhibit specific spontaneous decay properties which are studied in

2812: section~\ref{SPO};

2813: \item The electron in a hydrogen atom is not necessarily bound.

2814: It may ionize, especially when the atom is exposed to

2815: a microwave field. This is a {\em coherent} decay process 

2816: where the ionized electron keeps its phase coherence.

2817: There are no exact bound states in the system, but rather resonances.

2818: From a quantum point of view, the Floquet spectrum is no longer

2819: discrete but continuous and we actually deal with an {\em open system}. 

2820: In a more elementary language,

2821: the atom can successively absorb several photons so that its energy

2822: exceeds the ionization threshold. 

2823: If initially prepared in a wave-packet eigenstate, this is a pure quantum

2824: phenomenon, since the classical dynamics remain trapped within the

2825: resonance island forever. The multi-photon ionization 

2826: may then be 

2827: considered as a tunneling process from 

2828: inside the resonance island

2829: to the non-resonant part of phase space, 

2830: where the Rydberg electron eventually escapes

2831: to infinity. 

2832: This picture is elaborated in  section~\ref{ION}.

2833: \end{itemize}

2834: 

2835: \subsubsection{Hamiltonian, basis sets and selection rules}

2836: 

2837: In the presence of a microwave field, the dipole 

2838: approximation~\cite{cct92,loudon}

2839: can be used to describe  the atom-field interaction. Different gauges

2840: can be used, the physics being of course independent of the choice

2841: of gauge. The most common choices are the length and the velocity gauges.

2842: For simplicity, in our discussion,

2843: we shall use the length gauge, although actual

2844: quantum calculations are usually a bit easier in the velocity gauge

2845: \cite{shakeshaft88,abu95c,cormier96}.

2846: The Hamiltonian reads:

2847: \begin{equation}

2848: H = \frac{\vec{p}^2}{2m} - \frac{q^2}{4\pi \epsilon_0 r}

2849: -q \vec{r}\cdot\vec{F}(t)

2850: \label{ham_3d_si}

2851: \end{equation}

2852: where $q$ is the (negative) charge of the electron, $m$ its mass, and

2853: $\vec{F}(t)$ the microwave electric field acting on the atom.

2854: We neglect here all relativistic, spin, QED effects, etc, and assume

2855: an infinitely massive nucleus.

2856: Later, unless specified otherwise, we shall use atomic units, where

2857: $|q|\equiv 4\pi\epsilon_0\equiv m\equiv \hbar\equiv 1$, and the

2858: Hamiltonian reduces to

2859: \begin{equation}

2860: H = \frac{\vec{p}^2}{2} - \frac{1}{r}

2861: +\vec{r}\cdot\vec{F}(t).

2862: \label{h}

2863: \end{equation}

2864: Here, 

2865: \begin{equation}

2866: H_0= \frac{\vec{p}^2}{2} - \frac{1}{r}

2867: \label{h0}

2868: \end{equation}

2869: describes the unperturbed atomic part, with 

2870: bound energy spectrum: 

2871: \begin{equation}

2872: E_n= - \frac{1}{2n^2},\ n\geq 1.

2873: \label{spectrum}

2874: \end{equation}

2875: The external driving

2876: \begin{equation}

2877: V = \vec{r}\cdot \vec{F}(t)

2878: \label{v}

2879: \end{equation}

2880: is characterized by

2881: the amplitude $F$ of the microwave field, and by its  

2882: frequency

2883: $\omega.$ 

2884: We shall consider in detail the case of linear polarization, 

2885: \begin{equation}

2886: V=F z \cos \omega t,

2887: \label{vlinear} \end{equation}

2888: in section \ref{LIN}, and that of circular polarization

2889: \begin{equation}

2890: V=F( x \cos \omega t +  y \sin \omega t),

2891: \label{vcirc} 

2892: \end{equation}

2893: in section~\ref{CP}. The general case of elliptic polarization

2894: will be 

2895: studied in section~\ref{EP}.

2896: 

2897: In the treatment of the perturbed Coulomb problem, eq.~(\ref{h}),

2898: there is an apparent difficulty with the singularity of the Coulomb

2899: potential at the origin, especially for the restricted one-dimensional

2900: model of the atom.

2901: This led several authors,

2902: see, e.g.,

2903: \cite{javanainen88,su91,reed91,burnett92,grobe93,grobe94}, to consider

2904: unphysical potentials without singularity, for example

2905: of the type $(r^2+a^2)^{-1/2}.$ This is completely unnecessary

2906: and potentially dangerous. Indeed, such a potential breaks

2907: the Coulomb degeneracy which is responsible for the closed character

2908: of elliptical Kepler orbits (in the classical world), and

2909: for the degeneracy of the energy levels (in the quantum world).

2910: Such an unphysical symmetry breaking strongly modifies the structure

2911: of the non-linear resonance island, and affects the existence

2912: and properties of non-dispersive wave-packets outlined in section~\ref{LIN3D}.

2913: 

2914: 

2915: The Coulomb singularity can be rigorously regularized (in any

2916: dimension), both in classical and in quantum mechanics. In classical

2917: mechanics, this is made possible through the well-known

2918: Kustaanheimo-Stiefel transformation \cite{kusta65}, used by

2919: various authors for perturbed Coulomb problems, see, e.g.,

2920: \cite{delande84,rath88,griffiths92,gebarowski95,schlagheck99}.

2921:  In quantum mechanics, one may  use a basis set

2922: of non-orthogonal functions, known as the Sturmian functions.

2923: Ultimately, the whole analysis relies on the dynamical symmetry

2924: properties of the Coulomb interaction and the associated SO(4,2) group

2925: \cite{barut79,chen80,chen81,delandeth}.

2926: It not only allows to treat the Coulomb singularity properly, but also

2927: to define a basis set of Sturmian functions 

2928:  extremely efficient for

2929: numerical calculations. The most common set 

2930: are ``spherical" Sturmian functions characterized by three quantum numbers,

2931: $L$ and $M$ for the angular structure (the associated wave-functions are

2932: the usual spherical harmonics), and the positive integer $n$, for the radial

2933: part. 

2934: 

2935: As discussed in section~\ref{QD}, the quantum properties

2936: of the non-dispersive wave-packets are 

2937: encoded in the

2938: spectrum of the Floquet Hamiltonian (\ref{calhq}), which acts

2939: in configuration space extended by the time axis. The temporal

2940: properties are completely independent of the spatial dimension,

2941: and any Floquet eigenstate can be expanded in a Fourier series

2942: indexed by the integer $k$, 

2943: as in eq.~(\ref{wfmat}). The Schr\"odinger equation

2944: for the Floquet Hamiltonian is tridiagonal in $k,$ as in

2945: eq.~(\ref{coupled_equations}).

2946: When the spatial part of the wave-function is 

2947: expressed in a Sturmian basis,

2948: one finally obtains

2949: a generalized eigenvalue problem $(A-{\cal E} B)|\psi> = 0$, where both,

2950: $A$ and $B$, are sparse matrices, the elements of which are known analytically

2951: and obey the following selection rules:

2952: \begin{equation}

2953: |\Delta M|\leq 1,\ \ |\Delta L| \leq 1,\ \ |\Delta n| \leq 2,\ \ |\Delta k| \leq 1.

2954: \end{equation}

2955: When a static electric or magnetic field is added, see section~\ref{MA},

2956: some additional non-zero matrix elements exist, but sparsity is preserved.

2957: The eigenvalues can then be calculated using an efficient diagonalization

2958: routine such as the Lanczos 

2959: algorithm \cite{lanczos,delande91,ericsson80,grimes94}.

2960: 

2961: Because -- in the presence of a microwave field -- the system is unbounded,

2962: there are in general no exact bound states but rather resonances.

2963: Using Sturmian functions, the

2964: properties of the resonances can be calculated {\em directly} 

2965: using the complex

2966: rotation technique 

2967: \cite{balslev71,graffi85,yajima82,nicolaides78,reinhardt83,ho83,moiseyev98}. 

2968: The price

2969: to pay is to diagonalize complex symmetric matrices, instead of

2970: real symmetric ones. The advantage is that the resonances are obtained

2971: as complex eigenvalues $E_n-i\Gamma_n/2$ of the complex rotated Hamiltonian,

2972: $E_n$ being the position of the resonance, and $\Gamma_n$ its width.

2973: All

2974: essential properties of resonances can be obtained from complex

2975: eigenstates \cite{abu94}.

2976: 

2977: \subsubsection{Simplified 1D and 2D models}

2978: 

2979: Because explicit calculations for the real 3D

2980:  hydrogen atom may be

2981: rather complicated, it is fruitful to study also simplified 

2982: 1D and 2D approximations of the real world.

2983: Let us first consider the simplified

2984: restriction of the atomic motion to one single dimension 

2985: of configuration space,

2986: \begin{equation}

2987: H_0=\frac{p^2}{2} - \frac{1}{z}, \ \ \ \ {\mathrm with}\ z>0,

2988: \label{h0_1d}

2989: \end{equation}

2990: with the external driving along $z$,

2991: \begin{equation}

2992: V = F z \cos \omega t.

2993: \label{v_1d}

2994: \end{equation}

2995: The energy spectrum of $H_0$ is identical to the spectrum

2996: in 3D~\cite{englefield72}:

2997: \begin{equation}

2998: E_n^{(1D)} = - \frac{1}{2n^2},\ \ \ \ {\mathrm with}\ n\geq 1.

2999: \label{spectrum_1d}

3000: \end{equation}

3001: Such  a one-dimensional model allows to grasp

3002: essential features of the driven atomic dynamics, and provides the simplest

3003: example for the creation of non-dispersive wave-packets by a near-resonant

3004: microwave field. The classical dynamics live on a three-dimensional

3005: phase space, spanned by the single dimension of configuration space, the canonically conjugate momentum, and

3006: by time. This is the lowest dimensionality for a Hamiltonian system to display

3007: mixed regular-chaotic character \cite{lichtenberg83}.

3008: 

3009: For a circularly (or elliptically) polarized microwave, a 1D model

3010: is of course inadequate. One can use a two-dimensional model

3011: where the motion of the electron is restricted to the polarization

3012: plane. The energy spectrum in two dimensions is:

3013: \begin{equation}

3014: E_n^{(2D)}= - \frac{1}{2\left(n+\half\right)^2},\ \ \ \ {\mathrm with}\ n\geq 0.

3015: \label{spectrum_2d}

3016: \end{equation}

3017: It differs from the 3D (and 1D) energy spectrum by the additional

3018: 1/2 in the denominator, due to the specific

3019: Maslov indices induced by  the Coulomb 

3020: singularity.

3021: 

3022: \subsubsection{Action-angle coordinates}

3023: 

3024: \label{action-angle_hydrogen}

3025: 

3026: In order to apply the general theory of nonlinear resonances and

3027: non-dispersive wave-packets derived in section~\ref{GM}, we need

3028: the action-angle coordinates for the hydrogen atom.

3029: For the simplified 1D model, the result is simple:

3030:  the principal action $I$ and the canonically conjugate

3031: angle $\theta$ are defined by \cite{goldstein80,jensen84}

3032: \begin{eqnarray}

3033: I & = & \sqrt{\frac{a}{2}}, \nonumber \\

3034: \theta & = & \left\{

3035: \begin{array}{lc}

3036: \eta-\sin\eta, & \ \ p\geq 0, \\ 2\pi-\eta+\sin\eta, & \ \ p<0,

3037: \end{array} \right. \nonumber \\

3038: \eta & = & 2\sin^{-1}\sqrt{\frac{z}{a}},\ \ \  a=-E^{-1},

3039: \label{aa_lin1d}

3040: \end{eqnarray}

3041: where $a$

3042: is the maximum distance.

3043: In celestial mechanics, $\theta$ and $\eta$ are known as

3044: the {\em mean} and the {\em eccentric anomaly}, respectively \cite{marion}.

3045: The Hamilton function depends on the action through:

3046: \begin{equation}

3047: H_0=-\frac{1}{2I^2},

3048: \label{h0_I}

3049: \end{equation}

3050: and the classical Kepler frequency reads:

3051: \begin{equation}

3052: \Omega = \frac{dH_0}{dI} = \frac{1}{I^3}.

3053: \label{omk}

3054: \end{equation}

3055: Due to the  Coulomb singularity at $z=0$,

3056:  the Maslov index

3057: of this system is $\mu=0$ instead of $\mu=2$, and the semiclassical 

3058: energy spectrum,

3059:  eq.~(\ref{semen})\footnote{As we are using atomic units, $\hbar$ is 

3060: unity, and

3061: the principal quantum number just coincides with the action.},

3062: matches the exact quantum spectrum, eq.~(\ref{spectrum_1d}).

3063: 

3064: The classical equations of motion can be solved exactly and it is easy

3065: to obtain the Fourier components of the dipole operator \cite{meerson82}:

3066: \begin{equation}

3067: z(\theta) = I^2 \left(\frac{3}{2} - 2\sum_{m=1}^{\infty}{ \frac{J'_m(m)}{m}

3068: \cos (m\theta)} \right),

3069: \label{v_1dfou}

3070: \end{equation}

3071: where $J'_n(x)$ denotes the  derivative of the usual Bessel function.

3072: The strongly non-linear character of the Coulomb interaction is responsible

3073: for the slow decrease of the Fourier components at high $m$.

3074: 

3075: For the 2D and 3D hydrogen atom, the action-angle variables

3076: are similar, but more complicated because

3077: of the existence of angular degrees of freedom.

3078: The classical trajectories are ellipses with focus at the nucleus.

3079: The fact that all bounded trajectories are periodic manifests the

3080: degeneracy of the classical dynamics. As a consequence, although

3081: phase space is six-dimensional with three angle and three action 

3082: variables in 3D

3083: -- four-dimensional with two angle and two action variables in 2D

3084: -- the Hamilton function depends only on the total action $I$, precisely

3085: like the 1D hydrogen atom, i.e. through eq.~(\ref{h0_I}).

3086: In 3D, the Maslov index is zero, so that the energy spectrum is

3087: again given by eq.~(\ref{spectrum}), and the semiclassical

3088: approximation is exact. However, a different

3089: result holds for the 2D hydrogen atom, where the

3090: Maslov index is $\mu=2$ (still yielding exact agreement between the

3091: semiclassical and   the quantum spectrum, cf.

3092: eqs.~(\ref{ebkgen},\ref{spectrum_2d})).

3093: 

3094: The action-angle variables which 

3095: parametrize a general Kepler ellipse  are well known

3096: \cite{goldstein80}.

3097: In addition to the action-angle variables $(I,\theta)$ which determine

3098: the total action and the angular position

3099: of the electron along the Kepler ellipse,

3100: respectively, the orientation of the ellipse in space is defined by 

3101: two angles: $\psi$, canonically conjugate to the total angular

3102: momentum $L$, and the polar angle $\phi$, canonically conjugate to $M$, the

3103: $z$-component of the angular momentum.

3104: The angle $\psi$ conjugate to $L$ has a

3105: direct physical meaning for $M=0$: it represents

3106: the angle between the Runge-Lenz vector $\vec{A}$ (oriented along the major

3107: axis) of the Kepler ellipse, and the

3108: $z$-axis.

3109: For the 2D hydrogen atom (in the $(x,y)$ plane), 

3110: the orientation 

3111: of the ellipse is defined by the angle $\psi$

3112: (canonically conjugate to the total angular

3113: momentum $L$)

3114:  between the Runge-Lenz vector

3115: and the $x$-axis.

3116: 

3117: Also the Fourier components of the unperturbed

3118: classical position operator $\vec{r}(t)$

3119: are well known \cite{casati88}. In the local coordinate

3120: system of the Kepler ellipse (motion in the

3121: $(x',y')$ plane, with major axis along $x'$), one gets:

3122: \begin{eqnarray}

3123: x'&=&-\frac{3e}{2}I^2+2I^2\sum_{m=1}^\infty

3124: \frac{J_{m}^{'}(me)}{m}\cos m\theta \label{v_3d_x} \\

3125: y'&=&2I^2\frac{\sqrt{1-e^2}}{e}\sum_{m=1}^\infty

3126: \frac{J_{m}(me)}{m}\sin m\theta \label{v_3d_y} \\

3127: z'&=&0

3128: \label{v_3d_z}

3129: \end{eqnarray}

3130: where 

3131: \begin{equation}

3132: e = \sqrt{1-\frac{L^2}{I^2}}

3133: \label{eccentricity}

3134: \end{equation}

3135: denotes the eccentricity of the ellipse.

3136: $J_{m}(x)$ and $J^{'}_{m}(x)$ are the ordinary

3137: Bessel function and

3138: its derivative, respectively.

3139: 

3140: In the laboratory frame, the various components can be found

3141: by combining these expressions with the usual Euler

3142: rotations~\cite{marion,goldstein80}. The set of three Euler angles

3143: describes the successive rotations required for the

3144: transformation between the

3145: laboratory frame and the frame $(x',y',z')$ linked to the classical

3146: Kepler ellipse. We choose to rotate successively by an angle $\phi$ 

3147: around the $z$ laboratory axis, an angle $\beta$ around the 

3148: $y$-axis\footnote{Some authors define the second Euler rotation with

3149: respect to the $x$-axis. The existence of the two definitions 

3150: makes a cautious physicist's life much harder,

3151:  but the physics does not -- or at least

3152: should not -- depend

3153: on such ugly details.}, and an angle $\psi$ around the $z'$ axis.

3154: The physical interpretation of $\phi$ and $\psi$ is simple: 

3155: $\phi$ corresponds to a rotation around the $z$-axis, and is thus 

3156: canonically conjugate

3157: to the $z$-component of the angular momentum, noted $M$. Similarly,

3158: $\psi$ corresponds to a rotation around the axis of the total angular

3159: momentum $\vec{L}$, and is thus canonically conjugate

3160: to $L.$ 

3161: By construction, the third angle $\beta$ is 

3162: precisely the angle

3163: between the angular momentum $\vec{L}$ and the $z$-axis. Thus:

3164: \begin{equation}

3165: \cos \beta = \frac{M}{L},

3166: \label{beta}

3167: \end{equation}

3168: and $\beta =\pi/2$ for $M=0$.

3169: Altogether, the coordinates in the laboratory frame are related

3170: to the local coordinates through

3171: \begin{eqnarray}

3172: x&=&(\cos \psi \cos \beta \cos \phi - \sin \psi \sin \phi)\ x' +

3173: (-\sin\psi \cos\beta \cos\phi -\cos\psi \sin\phi)\ y' + \sin \beta \cos\phi\ z',

3174: \nonumber \\

3175: y&=&(\cos \psi \cos \beta \sin \phi + \sin \psi \cos \phi)\ x' +

3176: (-\sin\psi \cos\beta \sin\phi +\cos\psi \cos\phi)\ y' + \sin \beta \sin\phi \ z',

3177: \nonumber \\

3178: z&=& -\cos\psi \sin\beta \ x'  +  \sin\psi \sin\beta \ y' +  \cos\beta \ z',

3179: \label{rot_euler}

3180: \end{eqnarray}

3181: which, combined with eqs.~(\ref{v_3d_x}-\ref{v_3d_z}), allows for a complete

3182: expansion of the classical trajectories in terms of action-angle\

3183: coordinates.

3184: 

3185: The situation is somewhat simpler for the 2D model of the hydrogen atom.

3186: There, the angular momentum $\vec{L}$ is aligned along the $z$-axis,

3187: which means that $L=M$ and $\beta=0.$ Also, the rotation around

3188: $\vec{L}$ by an angle $\psi$ 

3189: can be absorbed in a rotation by an angle $\phi$ around the $z$-axis.

3190: Therefore, one is left with

3191: two pairs $(I,\theta)$ and $(M,\phi)$ of action-angle variables.

3192: The relation between the laboratory and the local coordinates

3193: reads:

3194: \begin{eqnarray}

3195: x&=& \cos \phi \ x' - \sin\phi \ y' 

3196: \nonumber \\

3197: y&=&\sin \phi \ x' + \cos\phi\ y',

3198: \label{twodrotfr} 

3199: \end{eqnarray} 

3200: which is nothing but a rotation of angle $\phi$ in the plane of the

3201: trajectory.

3202: Formally, the 2D result, eq.~(\ref{twodrotfr}), can be obtained from the 3D one,

3203: eq.~(\ref{rot_euler}), by

3204: specializing to $\beta=\psi=0$.

3205: The eccentricity, eq.~(\ref{eccentricity}), of

3206: the trajectory now reads:

3207: \begin{equation}

3208: e = \sqrt{1-\frac{M^2}{I^2}}.

3209: \end{equation}

3210: 

3211: \subsubsection{Scaling laws}

3212: 

3213: \label{scaling_laws}

3214: 

3215: It is well known that the Coulomb interaction 

3216: exhibits particular

3217: scaling properties: for example, all 

3218: bounded trajectories are similar

3219: (ellipses), whatever the (negative) energy. Also, the

3220: classical period 

3221: scales in a well-defined way with the

3222: size of the orbit (third Kepler law). This originates from

3223: the fact that the Coulomb potential is a homogeneous function

3224: -- of degree $-1$ -- of the radial distance $r$. Similarly,

3225: the dipole operator responsible for the coupling between the Kepler 

3226: electron and the 

3227: external driving field is a homogenous function

3228: -- of degree $1$ -- of $r.$ It follows that the

3229: classical equations of motion of the hydrogen

3230: atom exposed to an electromagnetic field

3231: are invariant under the following

3232: scaling transformation:

3233: \begin{eqnarray}

3234: \left \{

3235: \begin{array}{l}

3236: {\vec r}\rightarrow\alpha^{-1} {\vec r}, \\

3237: {\vec p}\rightarrow\alpha^{1/2} {\vec p}, \\

3238: H_0\rightarrow \alpha H_0, \\

3239: t\rightarrow\alpha^{-3/2} t, \\

3240: F\rightarrow\alpha^2 F,  \\

3241: \omega\rightarrow\alpha^{3/2}\omega , \\

3242: V \rightarrow \alpha V.

3243: \end{array}

3244: \right.

3245: \label{scaling_law}

3246: \end{eqnarray}

3247: where $\alpha $ is an arbitrary, positive real number.

3248: Accordingly, the action-angle variables

3249: transform as

3250: \begin{eqnarray}

3251: \left \{

3252: \begin{array}{l}

3253: I \rightarrow \alpha^{-1/2} I,\\

3254: L \rightarrow \alpha^{-1/2} L,\\

3255: M \rightarrow \alpha^{-1/2} M,\\

3256: \theta \rightarrow \theta,\\

3257: \psi \rightarrow \psi,\\

3258: \phi \rightarrow \phi.

3259: \end{array}

3260: \right.

3261: \label{scaling_aa}

3262: \end{eqnarray}

3263: It is therefore useful to introduce the ``scaled" total angular momentum

3264: and its component 

3265: along the $z$-axis, by chosing 

3266: \begin{equation}

3267: \alpha =I^2,

3268: \label{choosescale}

3269: \end{equation}

3270: what leads to

3271: \begin{eqnarray}

3272: \left \{

3273: \begin{array}{l}

3274: L_0 = \frac{L}{I},\\

3275: M_0 = \frac{M}{I}.

3276: \label{scaling_lm}

3277: \end{array}

3278: \right.

3279: \end{eqnarray}

3280: The eccentricity of the classical ellipse then reads:

3281: \begin{equation}

3282: e=\sqrt{1-L_0^2},

3283: \end{equation}

3284: and only depends -- as it should -- on scaled quantities. Similarly,

3285: the Euler angles describing the orientation of the ellipse are

3286: scaled quantities, by virtue of eq.~(\ref{scaling_aa}).

3287: 

3288: When dealing with non-dispersive wave-packets, it will be 

3289: useful

3290: to scale the amplitude and the frequency of the external field

3291: with respect of the action $\hat{I}_1$ of the resonant orbit.

3292: With the above choice of $\alpha$, eq.~(\ref{choosescale}), the scaling 

3293: relation (\ref{scaling_law}) for $\omega$ defines the scaled frequency 

3294: \begin{equation}

3295: \omega_0=\omega I^3\,

3296: \label{omega0}

3297: \end{equation}

3298: which turns into $\omega_0=\Omega \hat{I}_1^3$ with the resonance condition,

3299: eq.~(\ref{resonance_condition}), and enforces

3300: \begin{equation}

3301: \hat{I}_1 = \omega^{-1/3},

3302: \label{scaling_action}

3303: \end{equation}

3304: by virtue of eq.~(\ref{omk}).

3305: Correspondingly, the scaled external field is defined as 

3306: \begin{equation}

3307: F_0=FI^4,

3308: \label{scaling_F}

3309: \end{equation}

3310: which, with eq.~(\ref{scaling_action}), turns into $F_0=F\omega^{-4/3}$ at

3311: resonance.

3312: Hence, except for a global multiplicative factor $I^{-2},$

3313: the Hamiltonian of a hydrogen atom in an external field

3314: depends only on scaled quantities.

3315: 

3316: Finally, note that the {\em quantum} dynamics is not invariant

3317: with respect to the above scaling transformations. Indeed, the Planck constant

3318: $\hbar$ fixes an absolute scale for the various action variables. 

3319: Thus, the spectrum

3320: of the Floquet Hamiltonian will not be scale invariant,

3321: while the underlying classical phase space

3322: structure is. This latter feature 

3323: will be used to identify in the quantum spectrum 

3324: the remarkable features we are interested in.

3325:  

3326: 

3327: \subsection{Rydberg states in linearly polarized microwave fields}

3328: \label{LIN}

3329: 

3330: We are now ready to consider specific examples of non-dispersive

3331: wave-packets. We consider first the simplest, one-dimensional,

3332: driven hydrogen atom, as defined by eqs.~(\ref{h0_1d}),(\ref{v_1d}).

3333: 

3334: For a real 3D atom, this corresponds to driving the electron

3335:  initially prepared in a one-dimensional 

3336: eccentricity one orbit along the polarization axis of the field. 

3337: In fact,

3338: it turns out that the one-dimensionality of the dynamics is not

3339: stable under the external driving: the Kepler ellipse (with

3340: orientation fixed in configuration

3341: space by the Runge-Lenz vector) slowly precesses off the

3342: field polarization axis (see sections~\ref{LIN3D} and \ref{staticmw}).

3343: Thus, the 1D presentation which follows has mostly

3344: pedagogical value -- being

3345: closest to the 

3346: general case discussed later.

3347: However, a one-dimensional model allows to grasp

3348: essential features of the driven atomic dynamics and provides the simplest

3349: example for the creation of non-dispersing wave-packets by a near-resonant

3350: microwave field. 

3351: A subsequent section will describe the dynamics of the real 3D atom under

3352: linearly polarized driving, and amend on the flaws and

3353: drawbacks of the one-dimensional model.

3354: 

3355: \subsubsection{One-dimensional model}

3356: \label{LIN1D}

3357: 

3358: From eqs.~(\ref{h0_1d}),(\ref{v_1d}), the Hamiltonian of the driven 1D atom

3359: reads

3360: \begin{equation}

3361: H=\frac{p^2}{2}-\frac{1}{z}+Fz\cos (\omega t),\ z>0.

3362: \label{ham_lin1d}

3363: \end{equation}

3364: This has precisely the general form, eq.~(\ref{ham}), and we can therefore 

3365: easily

3366: derive explicit expressions for the secular Hamiltonian subject to the

3367: semiclassical quantization conditions, eqs.~(\ref{ebkpt}),

3368: (\ref{ebkires}),(\ref{ebkinres}), as well

3369: as for the quantum mechanical eigenenergies, eq.~(\ref{prediction_mathieu}), in the

3370: pendulum approximation.

3371: With the Fourier expansion, eq.~(\ref{v_1dfou}), and

3372: identifying $\lambda$ and $V(p,z)$ in eq.~(\ref{ham}) with $F$ and $z$ in

3373: eq.~(\ref{ham_lin1d}), respectively, the Fourier coefficients in

3374: eq.~(\ref{fourier_components}) take the explicit form

3375: \begin{equation}

3376: V_0=\frac{3}{2}I^2;\ \ \ \ V_m=-I^2\frac{J_m'(m)}{m},\ m\neq 0.

3377: \label{coeff_lin1d}

3378: \end{equation}

3379: The resonant action -- which defines the position of the resonance island in 

3380: fig.~\ref{pendulum} -- 

3381: is given by eq.~(\ref{scaling_action}).

3382: In a quantum 

3383: description, 

3384: the resonant action coincides with the 

3385: resonant principal quantum number:

3386: \begin{equation}

3387: n_0=\hat{I}_1 = \omega^{-1/3},

3388: \label{acteqqn}

3389: \end{equation}

3390: since the Maslov index vanishes in 1D, see section~\ref{RSEF}, and 

3391: $\hbar\equiv 1$ in atomic units.

3392: The resonant coupling is then given by:

3393: \begin{equation}

3394:  V_1=-\hat{I}^2J_1'(1),

3395: \label{v1_LIN1D}

3396: \end{equation}

3397: and the secular Hamiltonian, eq.~(\ref{hsec_ap}), reads:

3398: \begin{equation}

3399: {\cal H}_{\mathrm sec} = \hat{P}_t - \frac{1}{2\hat{I}^2} 

3400: -\omega \hat{I} -  J_1'(1)\hat{I}^2 F \cos\hat{\theta}.

3401: \label{hsec_h1d}

3402: \end{equation}

3403: This Hamiltonian has the standard form of a secular

3404: Hamiltonian with a resonance island centered around

3405: \begin{equation}

3406: \hat{I}=\hat{I}_1 = \omega^{-1/3}=n_0,\ \ \ \ \hat{\theta} = \pi,

3407: \label{posres}

3408: \end{equation}

3409: sustaining librational motion  within its boundary.

3410: 

3411: Those energy values of ${\cal H}_{\mathrm sec}$ which

3412: define contour lines (see fig.~\ref{pendulum}) 

3413: such that the contour integrals, eqs.~(\ref{ebkires}),(\ref{ebkinres}),

3414: lead to 

3415: non-negative integer values of $N$, are the semiclassical quasienergies

3416: of the 1D hydrogen atom under external driving. The non-dispersive wave-packet

3417: eigenstate of this model atom in the electromagnetic field is represented by

3418: the ground state $N=0$ of ${\cal H}_{\mathrm sec}$,

3419: localized (in phase space) near the center

3420: of the resonance island. A detailed comparison

3421: of the semiclassical energies to the exact quantum solution of our problem

3422: will be provided in the next subsection,

3423: where we treat the

3424: three-dimensional atom in the field. There it will turn out that the

3425: spectrum of the 1D model is actually neatly 

3426: embedded  in the 

3427: spectrum of

3428: the real 3D atom.

3429: 

3430: In the immediate vicinity of the resonance island, 

3431: the secular Hamiltonian can be further simplified, leading to the

3432: pendulum approximation, see section~\ref{CD} and eq.~(\ref{eqpend}).

3433: The second derivative of the unperturbed Hamiltonian with respect

3434: to the action is:

3435: \begin{equation}

3436: H^{''}_0=-\frac{3}{n_0^4},

3437: \label{h0second}

3438: \end{equation}

3439: and the pendulum Hamiltonian reads:

3440: \begin{equation}

3441: {\cal H}_{\mathrm pend} = \hat{P}_t - \frac{3}{2n_0^2}

3442: - J_1'(1)n_0^2 F \cos\hat{\theta} -\frac{3}{2n_0^4}(\hat{I}-n_0)^2.

3443: \label{hpend_lin1d}

3444: \end{equation}

3445: Remember that $n_0$ is the resonant action, not necessarily an integer.

3446: As we are interested in states deeply inside the resonance island,

3447: we can employ the harmonic approximation around the

3448: stable fixed point $(\hat{I}=n_0,\hat{\theta}=\pi)$, and finally

3449: obtain the semiclassical energies of the non-dispersive

3450: wave-packets:

3451: \begin{equation}

3452: {\cal E}_{N,k} = k\omega -\frac{3}{2n_0^2} + J_1'(1)n_0^2 F -

3453: \left(N+\frac{1}{2}\right) \omega_{\mathrm harm},

3454: \label{semiharm1dval}

3455: \end{equation}

3456: where, in agreement with eq.~(\ref{omega_harmonic})

3457: \begin{equation}

3458: \omega_{\mathrm harm}= 

3459: \sqrt{3J_1'(1)} \frac{\sqrt{F}}{n_0} = \omega \sqrt{3J_1'(1)F_0}

3460: \label{omega_h1d}

3461: \end{equation}

3462: is the classical librational frequency in the resonance island.

3463: The quantum number $k$ reflects the global $\omega$ periodicity

3464: of the Floquet spectrum, as a consequence of eq.~(\ref{ebkpt2}).

3465: 

3466: As already noted in section~\ref{scaling_laws}, the semiclassical quantization

3467: breaks the scaling of the classical dynamics. Nontheless, 

3468: the semiclassical

3469: energy levels can be written in terms of the

3470: scaled parameters introduced above, by virtue of

3471: eqs.~(\ref{scaling_F},\ref{posres}):

3472: \begin{equation}

3473: {\cal E}_{N,k=0} = \frac{1}{n_0^2}

3474: \left[ -\frac{3}{2} + J_1'(1) F_0 - \frac{N+1/2}{n_0} \sqrt{3J_1'(1)F_0}

3475: \right].

3476: \label{effective}

3477: \end{equation}

3478: Note that the term $(N+1/2)/n_0$ highlights 

3479: the role of $1/n_0$ as an effective Planck constant.

3480: 

3481: As discussed in section~\ref{section_mathieu},

3482: the fully ``quantum" quasienergies of the

3483: resonantly driven atom can be obtained using the

3484: very same pendulum approximation of the system, 

3485: together with the solutions

3486: of the Mathieu equation. In our case, the  characteristic exponent

3487: in the Mathieu equation is given by eq.~(\ref{cexp})

3488: and the Mathieu parameter is, according to 

3489: eqs.~(\ref{map_mathieu3}),(\ref{v1_LIN1D}),(\ref{h0second}):

3490: \begin{equation}

3491: q=\frac{4}{3}Fn_0^6J'_1(1)=\frac{4}{3}F_0n_0^2J'_1(1).

3492: \label{mathind_LIN1D}

3493: \end{equation}

3494: The quantum quasienergy levels are then given by eq.~(\ref{prediction_mathieu})

3495: which reads:

3496: \begin{equation}

3497: {\cal E}_{\kappa,k=0} = - \frac{3}{2n_0^2} 

3498: - \frac{3}{8n_0^4} \ a_{\kappa}(\nu,q).

3499: \label{prediction_mathieu_lin1d}

3500: \end{equation}

3501: 

3502: As discussed in section~\ref{QD}, the non-dispersive wave-packet

3503: with maximum localization is associated with $\kappa=0$ and is well

3504: localized inside the non-linear resonance island between the internal

3505: coulombic motion and the external driving, provided the parameter $q$

3506: is of the order of unity (below this value, the resonance island

3507: is too small to support a localized state). The minimum scaled 

3508: microwave amplitude is thus of the order of:

3509: \begin{equation}

3510: F_{0,\mathrm{trapping}}= F_{\mathrm{trapping}}n_0^4 \simeq \frac{1}{n_0^2}

3511: \label{ftrapping}

3512: \end{equation}

3513: which is thus {\em much smaller} -- by a factor $n_0^2$, i.e.

3514: three orders of magnitude in typical experiments --  

3515: than the electric field created

3516: by the nucleus. This illustrates that a well chosen weak perturbation

3517: may strongly influence the dynamics of a non-linear system. From the

3518: experimental point of view, this is good news, a limited

3519: microwave power is sufficient to create non-dispersive wave-packets.

3520: 

3521: We have

3522: so far given a complete description of the dynamics of the

3523: resonantly driven, one-dimensional

3524: Rydberg electron, from a semiclassical as well as from a quantum mechanical

3525: point of view, in the resonant approximation. These approximate treatments

3526: are now complemented by a numerical solution of the exact quantum mechanical

3527: eigenvalue problem described by the Floquet equation (\ref{calhq}),

3528: with $H_0$ and $V$ from eqs.~(\ref{h0_1d}),(\ref{v_1d}),

3529: as well as by the numerical integration of the classical equations of motion

3530: derived from eq.~(\ref{ham_lin1d}). Using this machinery, we 

3531: illustrate some of the essential properties of nondispersive wave-packets

3532: associated

3533: with the principal resonance in this system, whereas we 

3534: postpone the discussion of other primary resonances

3535: to section \ref{ORH}.

3536: 

3537: Fig.~\ref{lin1d_00} compares the phase space structure of the exact classical

3538: dynamics generated by the Hamilton function (\ref{ham_lin1d}), and the isovalue

3539: curves of the pendulum dynamics, eq.~(\ref{hpend_lin1d}),

3540: for the case $n_0=60,$

3541: at scaled field strength

3542: $F_0=Fn_0^4=0.01$. The Poincar\'e surface of section is taken

3543: at 

3544: phases $\omega t=0$ (mod $2\pi$) and plotted in

3545: $(I,\theta)$ variables which, for such times, coincide

3546: with the $(\hat{I},\hat{\theta})$ variables, 

3547: see eqs.~(\ref{rotframe_a}),(\ref{rotframe_b}).

3548: Clearly, the pendulum approximation predicts the

3549: structure of the invariant curves very well, 

3550: with the resonance island surrounding

3551: the stable periodic orbit at $(\hat{I}\approx 60,\hat{\theta}=\pi),$

3552: the unstable fixed point at $(\hat{I}\approx 60,\hat{\theta}=0),$

3553: the separatrix, and the rotational motion outside the resonance island.

3554: Apparently, only tiny regions of stochastic motion invade the

3555: classical phase space, which hardly affects the quality of the pendulum

3556: approximation. It should be emphasized that -- because of the scaling laws,

3557: see section~\ref{scaling_laws} -- the figure depends on

3558: the scaled field strength $F_0$ only. Choosing a different microwave

3559: frequency with the same scaled field leads, via eq.~(\ref{acteqqn}), 

3560: to a change

3561: of $n_0$ and, hence, of the scale of $I.$

3562: \begin{figure}

3563: \centerline{\psfig{figure=bdzf10.ps,width=15cm,angle=-90}}

3564: \caption{Poincar\'e surface of section for the dynamics of a 1D hydrogen

3565: atom driven by an external oscillatory electric field, see eq.~(\protect\ref{ham_lin1d}).

3566: The driving frequency is chosen as $\omega=1/60^3,$ such that the nonlinear

3567: resonance island is centered at principal action (or effective

3568: principal quantum number) $n_0=60.$ The scaled external field amplitude

3569: is set to $F_0=Fn_0^4=0.01$. Although the driving field is much weaker

3570: than the Coulomb field between the electron and the nucleus, 

3571: it suffices to create a relatively

3572: large resonance island which supports several non-dispersive

3573: wave-packets.}

3574: \label{lin1d_00}

3575: \end{figure}

3576: 

3577: Fig.~\ref{lin1d_0} compares the prediction of the Mathieu approach,

3578: eq.~(\ref{prediction_mathieu_lin1d}), for the quasienergy levels of the

3579: Floquet Hamiltonian to

3580: the exact numerical result obtained by diagonalization of the full Hamiltonian,

3581: see section~\ref{RSEF}. Because of the $\hbar\omega$ periodicity

3582: of the Floquet spectrum, the sets of energy levels of the pendulum,

3583: see fig.~\ref{fig_mathieu},

3584: are folded in one single Floquet zone.

3585: For  states located inside or in the vicinity

3586: of the resonance island -- the only ones plotted in 

3587: fig.~\ref{fig_mathieu}a --

3588: the agreement is very good for low and

3589: moderate

3590: field strengths.

3591: Stronger

3592: electromagnetic fields lead to deviations between the Mathieu and the exact

3593: result. This indicates 

3594: higer order corrections to the pendulum approximation.

3595: \begin{figure}

3596: \centerline{\psfig{figure=bdzf11.ps,width=15cm,angle=-90}}

3597: \caption{Comparison of the exact quasienergy spectrum of the 1D hydrogen atom

3598: driven by a linearly polarized microwave field (right),

3599: eq.~(\protect\ref{ham_lin1d}),

3600: with the prediction of the pendulum approximation (left),

3601: eq.~(\protect\ref{prediction_mathieu_lin1d}). Because of the $\hbar\omega$

3602: periodicity of the Floquet spectrum, the energy levels

3603: described by the pendulum approximation are folded inside one

3604: Floquet zone. The agreement between the exact quantum result and the

3605: pendulum approximation is very good. 

3606: The filled circle shows the 

3607: most localized non-dispersive wave-packet $N=0$ shown in

3608: figs.~\protect\ref{lin1d_2}-\protect\ref{lin1d_4}, while

3609: the filled square represents the hyperbolic non-dispersive wave-packet

3610: partly localized in the vicinity of the unstable

3611: equilibrium point of the pendulum, shown in

3612: figs.~ \protect\ref{lin1d_5} and \protect\ref{lin1d_7}. 

3613: The open circle and square compare the exact location of the

3614: respective quasienergy values with the Mathieu prediction, which is

3615: considerably better for the ground state as compared to the separatrix 

3616: state.}

3617: \label{lin1d_0}

3618: \end{figure}

3619: 

3620: Fig.~\ref{lin1d_1} shows a typical Poincar\'e surface of section of the

3621: classical dynamics of the driven Rydberg electron, at different values

3622: of the phase of the driving field. The field amplitude is chosen

3623: sufficiently high to

3624: induce largely chaotic dynamics,

3625: with the

3626: principal resonance as the only remnant of regular motion occupying

3627: an appreciable volume of phase space.

3628: The figure clearly illustrates

3629: the temporal evolution of the elliptic island with the phase of the driving

3630: field, i.e. the locking of the electronic motion on the external driving. The

3631: classical stability island follows the dynamics of the unperturbed electron

3632: along the 

3633: resonant trajectory.

3634: The distance from the nucleus is parametrized by the variable $\theta,$ 

3635: see eq.~(\ref{aa_lin1d}). At $\omega t = \pi,$ the 

3636: classical electron hits 

3637: the nucleus (at $\theta=0$),

3638: its velocity

3639: diverges and changes sign discontinuously.

3640: This explains the

3641: distortion of the resonance island as it approaches $\theta=0,\pi$.

3642: \begin{figure}

3643: \centerline{\psfig{figure=bdzf12.eps,width=15cm,angle=-90}}

3644: \caption{Surface of section of the classical phase space of a 1D hydrogen atom,

3645: driven by a linearly polarized microwave field of amplitude $F_0=0.053$, 

3646: for different values of

3647: the phase: $\omega t=0$ (top left), $\omega t=\pi/2$ (top right), $\omega t=\pi$ (bottom left),

3648: $\omega t=3\pi/2$ (bottom right). At this driving strength,

3649: the principal

3650: resonance remains as the

3651: only region of regular motion of appreciable size,

3652: in a globally chaotic phase space. The action-angle variables 

3653: $I,\theta$ are defined by

3654: eq.~(\protect\ref{aa_lin1d}), according to which a 

3655: collision with

3656: the nucleus occurs 

3657: at $\theta=0$.}

3658: \label{lin1d_1}

3659: \end{figure}

3660: 

3661: Quantum mechanically, we expect a non-dispersive wave-packet eigenstate 

3662: to be localized within the resonance island. 

3663: The semiclassical prediction 

3664: of its quasienergy, eq.~(\ref{prediction_mathieu_lin1d}),

3665: facilitates to identify the nondispersive wave-packet within  the 

3666: exact Floquet spectrum, after numerical diagonalization of the Floquet 

3667: Hamiltonian (\ref{calhq}).

3668: The wave-packet's configuration space

3669: representation is shown 

3670: in fig.~\ref{lin1d_2}, for the same phases of the field as

3671: in the plots of the 

3672: classical dynamics in fig.~\ref{lin1d_1}. Clearly, the wave-packet

3673: is very well localized at the outer turning point of the Kepler electron

3674: at phase $\omega t=0$

3675: of the driving field, and is reflected off the nucleus half a period of the

3676: driving field later. 

3677: On reflection, the electronic density exhibits some

3678: interference structure, as well as some transient spreading. This is a

3679: signature of the quantum mechanical uncertainty in the angle $\theta$:

3680: part of the wave-function, which still

3681: approaches

3682: the Coulomb singularity, interferes with 

3683: the other part already reflected off the nucleus.  

3684: The transient spreading is equally 

3685: manifest in the temporal

3686: evolution of the uncertainty product $\Delta z\Delta p$ itself, which is

3687: plotted in fig.~\ref{lin1d_3}.

3688: Apart from this singularity at $\omega t=\pi,$ the wave-packet

3689: is approximately Gaussian at any time, with a time-dependent 

3690: width

3691: (compare $\omega t=0$ and

3692: $\omega t=\pi/2$).

3693: \begin{figure}

3694: \centerline{\psfig{figure=bdzf13.ps,width=14cm,angle=-90}}

3695: \caption{Configuration space representation of the electronic 

3696: density of the non-dispersive wave-packet eigenstate

3697: of a 1D hydrogen atom,

3698: driven by a linearly polarized microwave field, for the same field amplitude

3699: and phases as in fig.~\protect\ref{lin1d_1}. 

3700: The eigenstate is 

3701: centered on the principal resonance of classical phase space,

3702: at action (principal quantum number) $n_0=60$.

3703: In configuration space,

3704: the wave-packet is localized at the outer

3705: turning point (with zero average velocity) at time $t=0$, then propagates

3706: towards the nucleus which it hits at $t/T=0.5$ (where $T=2\pi/\omega$ is the

3707: microwave period). 

3708: Afterwards, it propagates outward to the apocenter

3709: which is reached at time $t/T=1$. After one period, the

3710: wave-packet recovers {\em exactly} its initial shape, and

3711: will therefore propagate along the classical trajectory 

3712: forever, without spreading.

3713: The wave-packet has approximately Gaussian shape(with time-dependent width) 

3714: except at $t/T\simeq 0.5$.

3715: At this instant, the head of the wave-packet, which 

3716: already has been reflected off the nucleus, interferes with its tail,

3717: producing interference fringes. 

3718: }

3719: \label{lin1d_2}

3720: \end{figure}

3721: \begin{figure}

3722: \centerline{

3723: \psfig{figure=bdzf14a.eps,width=7cm,angle=-90}

3724: \psfig{figure=bdzf14b.eps,width=7cm,angle=-90}}

3725: \caption{(a) Dashed line: 

3726: Classical temporal evolution of the position of the Rydberg electron 

3727: resonantly driven by a linearly polarized microwave field, in the 

3728: one dimensional model, eq.~(\protect\ref{ham_lin1d}), of the hydrogen atom.

3729: ($n_0=60$; scaled field amplitude $F_0=Fn_0^4=0.053)$. 

3730: Thick line: Expectation value $\langle z\rangle$ for the non-dispersive

3731: wave-packet shown in fig.~\protect\ref{lin1d_2}, as a function of time.

3732: It follows the classical trajectory remarkably well (except for collisions

3733: with the nucleus).

3734: Dotted line: The position uncertainty $\Delta z = \sqrt{\langle z^2\rangle -

3735: \langle z \rangle ^2}$ 

3736: of the wave-packet. 

3737: $\Delta z$ being much smaller than

3738: $\langle z \rangle$ (except near collisions with the nucleus)

3739: highlights 

3740: the efficient localization of the wave-packet.

3741: (b) Uncertainty product $\Delta z\Delta p$ of the 

3742: wave-packet. The periodically repeating maxima 

3743: of this quantity indicate the collision of the electron 

3744: with the atomic nucleus.

3745: Note that the minimum uncertainty at

3746: the outer turning point of the wave-packet is very close to the Heisenberg

3747: limit $\hbar/2$. Although the wave-packet is {\em never}

3748: a minimal one, it is nevertheless well localized and

3749: an excellent 

3750: approximation of a classical particle.}

3751: \label{lin1d_3}

3752: \end{figure}

3753: 

3754: To complete the analogy between classical and quantum motion, we finally

3755: calculate the Husimi distribution -- the

3756: phase space representation of the wave-packet eigenstate

3757: defined in section~\ref{coherent_states}, eq.~(\ref{husimi_def})  --

3758: in order to obtain a direct comparison between classical and

3759: quantum dynamics in phase space. Fig.~\ref{lin1d_4} shows the resulting 

3760: phase

3761: space 

3762: picture, 

3763: again

3764: for different phases of the driving field. The association of the quantum

3765: mechanical time evolution with the classical resonance island (see fig.~\ref{lin1d_1}) 

3766: is

3767: unambiguous. The transient spreading at the collision with 

3768: the nucleus ($\omega t=\pi$) is due to the divergence

3769: of the classical velocity 

3770: upon reflection.

3771: \begin{figure}

3772: \centerline{\psfig{figure=bdzf15an.ps,width=7cm}

3773: \psfig{figure=bdzf15bn.ps,width=7cm}}

3774: \smallskip

3775: \centerline{\psfig{figure=bdzf15cn.ps,width=7cm}

3776: \psfig{figure=bdzf15dn.ps,width=7cm}}

3777: \caption{Husimi representation of the wave-packet eigenstate of

3778: fig.~\protect\ref{lin1d_2} in classical phase space, for the same 

3779: phases $\omega t$ and 

3780: scales $(0\leq \theta \leq 2\pi, 30\leq I \leq 90)$ as employed for the

3781: classical surface of section in fig.~\protect\ref{lin1d_1}. Clearly, the

3782: quantum mechanical eigenstate of the atom in the field follows the classical

3783: evolution without dispersion, except for its transient spreading when 

3784: reflected off 

3785: the nucleus (at $\omega t=\pi$, bottom left), due to the divergence

3786: of the classical velocity at that position.

3787: }

3788: \label{lin1d_4}

3789: \end{figure}

3790: 

3791: As discussed in section~\ref{CD} and visible in fig.~\ref{lin1d_00},

3792: there is an hyperbolic fixed point (i.e. an unstable equilibrium point)

3793: at $(\hat{I}=n_0,\hat{\theta}=0)$: it corresponds to the

3794: unstable equilibrium position of the pendulum 

3795:  when it points ``upwards". For the driven system,

3796: it corresponds to an unstable periodic orbit resonant with the

3797: driving frequency: it is somewhat similar to the stable

3798: orbit supporting the non-dispersive wave-packets, except that is is

3799: shifted in time by half a period. 

3800: 

3801: As discussed in section~\ref{scars}, the classical motion slows down 

3802: at the hyperbolic fixed point

3803: (the time to reach the unstable equilibrium point with zero velocity

3804: diverges~\cite{marion}), and the eigenfunction must 

3805: exhibit a maximum of the electronic density at this

3806: position. 

3807: In addition, due to the periodicity

3808: of the drive, the corresponding (``hyperbolic'') wave-packet eigenstate

3809: necessarily follows the dynamics of a classical particle 

3810: which evolves along the unstable periodic orbit.

3811: However, because the orbit is unstable, the quantum 

3812: eigenstate cannot remain 

3813: fully localized -- some probability

3814: has to flow away along 

3815: the unstable manifold of the classical flow in the vicinity of the 

3816: hyperbolic fixed point.

3817: Consequently,

3818: such an eigenstate is partially localized along 

3819: the separatrix between librational

3820: and rotational motion.

3821: For an illustration, first consider fig.~\ref{lin1d_4a}, which 

3822: shows classical surfaces

3823: of section of the driven (1D) hydrogen atom, at $F_0=0.034$, 

3824: again for different phases $\omega t$.

3825: Comparison with fig.~\ref{lin1d_1} shows a larger 

3826: elliptic island at this slightly lower field amplitude, 

3827: as well as remnants of the $s=2$ resonance island 

3828: at slightly larger

3829: actions $I/n_0\simeq 1.2\ldots 1.3$. 

3830: The time evolution of the electronic density of the

3831: eigenstate localized near the

3832: hyperbolic fixed point is displayed in fig.~\ref{lin1d_5},

3833: for different phases of the driving field.

3834: Clearly, as compared to fig.~\ref{lin1d_2}, the wave-packet moves in phase

3835: opposition to the driving field, and displays slightly irregular localization

3836: properties.  Accordingly, the Husimi representation in

3837: fig.~\ref{lin1d_7}

3838: exhibits reasonnably good localization on top of the hyperbolic point at phase

3839: $\omega t=\pi$, but the electronic probability spreads significantly

3840: along the

3841: separatrix layer at phase, as visible at $\omega t=0$.

3842: \begin{figure}

3843: \centerline{\psfig{figure=bdzf16.eps,width=14cm,angle=-90}}

3844: \caption{Surface of section of the classical phase space of a 1D hydrogen atom

3845: driven by a linearly polarized microwave field, for different values of

3846: the phase: $\omega t=0$ (top left), $\omega t=\pi/2$ (top right), $\omega t=\pi$ (bottom left),

3847: $\omega t=3\pi/2$ (bottom right). The action angle variables $I,\theta$ are defined by

3848: eq.~(\protect\ref{aa_lin1d}). At this value of the field amplitude,

3849: $F_0=0.034$, the principal resonance island 

3850: (and a small remnant of the $s=2$ resonance island) remain as the

3851: only regions of regular motion, in a globally chaotic phase space.}

3852: \label{lin1d_4a}

3853: \end{figure}

3854: \begin{figure}

3855: \centerline{\psfig{figure=bdzf17n.ps,width=14cm,angle=-90}}

3856: \caption{Wave-packet eigenstate anchored to the hyperbolic fixed point of the

3857: principal resonance of the 1D hydrogen atom driven by a linearly polarized 

3858: microwave field,

3859: in configuration space,

3860: for the same phases

3861: of the driving field as in fig.~\protect\ref{lin1d_4a}. 

3862: The wave-function is partly localized, especially close to

3863: the outer turning point at $t=0.5\times 2\pi/\omega$, but the localization

3864: is far from being perfect.

3865: Comparison to figs.~\protect\ref{lin1d_1}, \protect\ref{lin1d_2} and

3866: \protect\ref{lin1d_4a} shows

3867: that the state evolves in phase opposition with the stable, non-dispersive

3868: wave-packet, with significantly worse localization properties.}

3869: \label{lin1d_5}

3870: \end{figure}

3871: \begin{figure}

3872: \centerline{\psfig{figure=bdzf18a.ps,width=7cm}

3873: \psfig{figure=bdzf18b.ps,width=7cm}}

3874: \smallskip

3875: \centerline{\psfig{figure=bdzf18c.ps,width=7cm}

3876: \psfig{figure=bdzf18d.ps,width=7cm}}

3877: \caption{Husimi

3878: representation of the ``hyperbolic" wave-packet eigenstate of

3879: fig.~\protect\ref{lin1d_5}, for the same phases 

3880: (scales as in fig.~\protect\ref{lin1d_4a}).

3881: Clearly, the

3882: quantum mechanical eigenstate of the atom in the field follows the classical

3883: evolution. It is partly

3884: localized on top of the hyperbolic fixed point, but also

3885: spreads along the separatrix confining the principal

3886: resonance.

3887: The localization is more visible at $t=0$ (top left),

3888: the spreading more visible  at

3889: $t=0.5 \times 2\pi/\omega$ (bottom left).}

3890: \label{lin1d_7}

3891: \end{figure}

3892: 

3893: In the above discussion of the localization properties of the wave-packet 

3894: eigenstate we represented the wave-function in the $I$-$\theta$ phase space

3895: of {\em classically bounded} motion (i.e., classical motion with negative 

3896: energy). However, as we shall see in more detail in section~\ref{ION}, the 

3897: microwave driving actually induces a nonvanishing overlap of {\em all} 

3898: Floquet eigenstates \cite{graffi85,yajima82}, and, hence, of the wave-packet

3899: eigenstates, with the atomic continuum. It suffices to say here that 

3900: the associated finite decay rates induce finite life times of approx. 

3901: $10^6$ unperturbed Kepler orbits for the quantum objects considered in this 

3902: chapter, and are therefore irrelevant on the present level of our discussion.

3903: In figs.~\ref{lin1d_2}, \ref{lin1d_4}, \ref{lin1d_5}, and \ref{lin1d_7}, a

3904: finite decay rate would manifest as a slow reduction of the electronic 

3905: density, without affecting its shape or localization properties, after 

3906: $10^6$ classical Kepler periods.

3907: 

3908: \subsubsection{Realistic three-dimensional atom}

3909: \label{LIN3D}

3910: 

3911: Extending our previous analysis to the three-dimensional hydrogen atom driven

3912: by a linearly polarized microwave field, we 

3913: essentially expand the accessible 

3914: phase space. Since the Hamiltonian 

3915: \begin{equation}

3916: H_{\rm LP}=\frac{{\vec p}^2}{2}-\frac{1}{r}+Fz\cos(\omega t)

3917: \label{ham_lin_3d}

3918: \end{equation}

3919: is invariant under rotations around the field polarization axis,  

3920: the projection of the angular momentum is a conserved quantity and gives

3921: rise to a good quantum number $M$. 

3922: Hence, only two dimensions of configuration space are left, which, together

3923: with the explicit, periodic time dependence, span a five-dimensional 

3924: phase space. 

3925: 

3926: In the 1D situation described previously, the key ingredient for the existence

3927: of non-dispersive wave-packets was the phase locking of the internal

3928: degree of freedom on the external drive. In the 3D situation, there remains

3929: one single drive, but there are 

3930: several internal degrees of freedom. In the generic

3931: case, not all internal degrees of freedom can be simultaneously locked

3932: on the external drive, and one can expect only partial

3933: phase locking, i.e. only partially localized wave-packets. The non-trivial

3934: task is to understand how the phase locking of one degree of freedom

3935: modifies the dynamics along the other degrees of freedom. In atomic systems,

3936: the Coulomb degeneracy makes it possible to gain a full understanding

3937: of this phenomenon. 

3938: 

3939: The starting point is similar to the 1D analysis in

3940: sec.~\ref{LIN1D}, that is the expression of the Floquet

3941: Hamiltonian  -- whose eigenstates are of interest --

3942: as a function of action-angle coordinates

3943: $(I,\theta)$, $(L,\psi)$,

3944: $(M,\phi)$ introduced in section \ref{RSEF}.

3945: Using eqs.~(\ref{beta},\ref{rot_euler})

3946: and the Fourier expansion, eq.~(\ref{v_3d_x}-\ref{v_3d_z}),

3947: of the position operator, one obtains:

3948: \begin{equation}

3949: {\mathcal H} = P_t -\frac{1}{2I^2}+ F \sqrt{1-\frac{M^2}{L^2}} 

3950: \sum_{m=-\infty}^{+\infty}{\left[-X_m\cos\psi  \cos(m\theta-\omega t)

3951: +  Y_m \sin\psi\sin(m\theta-\omega t)\right]},

3952: \label{hamaa}

3953: \end{equation}

3954: with

3955: \begin{eqnarray}

3956: X_m(I)&=& I^2\frac{J_m'(me)}{m},\ \ \ \ m\neq 0, \label{xm} \\

3957: Y_m(I)&=& I^2 \frac{\sqrt{1-e^2}J_m(me)}{me},\ \ \ \ m\neq 0,  \label{ym}\\ 

3958: X_0(I)&=& -\frac{3}{2}eI^2, \label{x0}\\

3959: Y_0(I)&=&0.

3960: \label{y0}

3961: \end{eqnarray}

3962: where $e=\sqrt{1-L^2/I^2}$ is, as before, the eccentricity of the Kepler orbit

3963: (see 

3964: eq.~(\ref{eccentricity})).

3965: The absence of $\varphi$ in the 

3966: Hamiltonian reflects the azimuthal symmetry around the field axis and 

3967: ensures the conservation of $M$.

3968: 

3969: Precisely as in the treatment of the 

3970: one-dimensional 

3971: problem, we now transform 

3972: to slowly varying variables, given by 

3973: eqs.~(\ref{rotframe_a})-(\ref{rotframe_c}): 

3974: \begin{eqnarray}

3975: \hat{\mathcal H} & = & \hat{P}_t -\frac{1}{2 \hat I^2}- \omega\hat I +

3976: F \sqrt{1-\frac{M^2}{L^2}} \nonumber \\

3977: \times & & \sum_{m=-\infty}^{+\infty} {\left[-X_m \cos\psi  

3978: \cos(m\hat{\theta} + (m-1)\omega t)

3979: + Y_m \sin\psi \sin(m\hat{\theta} + (m-1)\omega t)\right]}.

3980: \label{hamab}

3981: \end{eqnarray}

3982: Averaging over the fast variable $t$ (over the driving field period 

3983: $T$) gives

3984: the secular Hamiltonian of the three-dimensional problem

3985: \begin{equation}

3986: {\mathcal H}_{\rm sec}  = \hat{P}_t -\frac{1}{2\hat I^2}- 

3987: \omega\hat{I}

3988:  +F \sqrt{1-\frac{M^2}{L^2}}(-X_1(\hat{I})\cos\psi \cos\hat\theta+

3989: Y_1(\hat{I})\sin\psi \sin\hat\theta).

3990: \label{hamsc2}

3991: \end{equation}

3992: Its physical interpretation is rather simple: the $X_1$ term represents

3993: the oscillating dipole (resonant with the frequency of the drive) 

3994: along the major

3995: axis of the classical Kepler ellipse, while the $Y_1$ term represents

3996: the oscillating dipole along the minor axis. As these 

3997: two components of the oscillating dipole are in quadrature, they interact

3998: with two orthogonal components of the external drive, hence

3999: the $\cos\hat\theta$ and $\sin\hat\theta$ terms.

4000: Finally, both components can be combined to produce the compact form

4001: \begin{equation}

4002: {\mathcal H}_{\rm sec}=\hat{P}_t-\frac{1}{2\hat I^2}-\omega \hat{I}

4003: +F \chi_1 \cos(\hat \theta + \delta_1),

4004: \label{hamscfin}

4005: \end{equation}

4006: with

4007: \begin{eqnarray}

4008: \chi_1(\hat I,L,\psi) & := & \sqrt{1-\frac{M^2}{L^2}}\ \sqrt{X_1^2\cos^2\psi+Y_1^2\sin^2\psi} ,

4009: \label{substc} \\

4010: \tan\delta_1 (L,\psi)& := & \frac{Y_1}{X_1}\tan\psi = 

4011: \frac{J_1(e)\sqrt{1-e^2}}{J'_1(e)e} \tan\psi.

4012: \label{substb}

4013: \end{eqnarray}

4014: In this form, the secular Hamiltonian has the same structure as

4015: the general 1D expression, eq.~(\ref{hsec_ap}), and its specialized

4016: version for the 1D hydrogen atom, eq.~(\ref{hsec_h1d}).

4017: The difference is that the additional action angle-variables

4018: $(L,\psi)$, $(M,\phi)$ {\em only} enter in the amplitude and phase of 

4019: the coupling defining the resonance island. 

4020: This allows to separate various time scales 

4021: in the system:

4022: \begin{itemize}

4023: \item The shortest time scale is associated with the Kepler

4024: motion, which is also the period of the external drive.

4025: In the resonant approximation discussed in detail

4026: in section~\ref{CD}, this time scale is eliminated by passing to the

4027: rotating frame. 

4028: \item The time scale of the secular (or pendulum) motion

4029: in the $(\hat{I},\hat{\theta})$ plane is significantly longer.

4030: It is the inverse of the classical pendulum frequency,

4031: eq.~(\ref{omega_harmonic}), of the order of

4032: $1/\sqrt{F_0}$ Kepler periods. In the regime of weak external

4033: driving we are interested in, $F_0 \ll 1$, it is thus much longer

4034: than the preceding time scale.

4035: \item The time scale of the ``transverse" (or angular) motion along the

4036: $(L,\psi)$, $(M,\phi)$ variables. Because these 

4037: are constant for the unperturbed Coulomb system, the time

4038: derivatives like $dL/dt$ and 

4039: $d\psi/dt$ generated by eqs.~(\ref{hamscfin})-(\ref{substb}) 

4040: are proportional 

4041: to $F$, and the resulting time scale is proportional

4042: to $1/F$. More precisely, it is of the order

4043: of $1/F_0$ Kepler periods, i.e., once again, significantly longer

4044: than the preceding time scale.

4045: \end{itemize}

4046: 

4047: From this separation of time scales, it 

4048: follows that we can use the following, additional secular approximation:

4049: for the motion in the $(\hat{I},\hat{\theta})$ plane, 

4050: $\chi_1$ and $\delta_1$ are adiabatic invariants,

4051: which can be considered as constant quantities. We then exactly

4052: recover the Hamiltonian discussed for the 1D model of the atom,

4053: with 

4054: a resonance island confining trajectories

4055: with librational motion in the $(\hat{I},\hat{\theta})$ plane, and

4056: rotational motion outside the resonance island.

4057: The center of the 

4058: island is located at:

4059: \begin{equation}

4060: \hat{I}=\hat{I}_1 = \omega^{-1/3}=n_0,\ \ \ \ \ \ 

4061: \hat{\theta} = -\delta_1.

4062: \label{Kures}

4063: \end{equation}

4064: As already pointed out in section \ref{CD}, 

4065: the 

4066: size of the resonance island in phase space is determined

4067: by the strength of the resonant coupling $\chi_1(\hat I,L,\psi).$

4068: In the pendulum approximation, its extension in $\hat{I}$ 

4069: scales as $\sqrt{\chi_1(\hat{I}_1,L,\psi)}$, i.e. with

4070: $\chi_1$ evaluated at the center, eq.~(\ref{Kures}), of the island.

4071: 

4072: The last step is to consider the slow motion in the $(L,\psi)$ plane.

4073: As usual, when a secular

4074: approximation is employed, the slow motion

4075: is due to an effective Hamiltonian which is 

4076: obtained by averaging of the secular Hamiltonian over the fast motion.

4077: Because the coupling $\chi_1(\hat{I},L,\psi)$ exhibits a simple scaling 

4078: with $\hat{I}$ (apart from a global $\hat{I}^2$ dependence, it depends

4079: on the scaled angular variables $L_0$ and $\psi$ only),

4080: the averaging over the fast motion results in an effective

4081: Hamiltonian for the ($L_0$,$\psi$) motion which depends

4082: on $\chi_1(\hat{I},L,\psi)$ only.  

4083: We deduce

4084: that the slow motion 

4085: follows curves of constant

4086:  $\chi_1(\hat{I}_1,L,\psi),$ at a velocity which

4087: depends on the average over the fast variables.

4088: $\chi_1(\hat{I}_1,L,\psi)$ is thus a constant of motion, both for the fast

4089: $(\hat{I},\hat{\theta})$  and the slow $(L,\psi)$ motion.  

4090: This also implies that the order of the 

4091: quantizations in the fast and slow variables

4092: can be interchanged: using the dependence of $\chi_1(\hat{I}_1,L,\psi)$ 

4093: on $(L,\psi),$ we obtain quantized values of $\chi_1$

4094: which in turn can be used as constant values to quantize the

4095: $(\hat{I},\hat{\theta})$ fast motion. 

4096: Note that the separation of time scales is here essential

4097: \footnote{If one considers non-hydrogenic atoms -- with

4098: a core potential in addition to the Coulomb potential -- the classical

4099: unperturbed ellipse precesses, adding an additional time scale,

4100: and the separation of time scales is much less obvious. See also

4101: section~\protect{\ref{SPEC}}.}.

4102: Finally,

4103: the dynamics in $(M,\phi)$ is trivial, since $M$ is a constant

4104: of motion. In the following, we will consider 

4105: the case $M=0$ for simplity. 

4106: Note that, when the eccentricity of the classical

4107: ellipse tends to $1$ -- i.e. $L\to 0$ -- and when

4108: $\psi \to 0,$ the Hamiltonian (\ref{hamsc2}) coincides

4109: exactly with the Hamiltonian of the 1D atom, eq.~(\ref{hsec_h1d}).

4110: This is to be expected, as it corresponds to 

4111: a degenerate classical Kepler ellipse along the $z$ axis.

4112: 

4113: The adiabatic separation of the 

4114: radial and of the angular motion allows the separate WKB quantization of

4115: the various degrees of freedom. In addition to the quantization conditions

4116: in $(\hat{P}_t,t)$ and $(\hat{I},\hat{\theta})$, 

4117: eqs.~(\ref{ebkpt}-\ref{ebkinres}),

4118: already formulated in our general 

4119: description of the semiclassical approach in 

4120: section~\ref{sapp},

4121: we

4122: additionally need to quantize the angular motion, according to:

4123: \begin{equation}

4124: \frac{1}{2\pi}\oint_\gamma L d\psi =  \left(p+\frac{1}{2}\right)\hbar,

4125: \label{ebkp}

4126: \end{equation}

4127: along a loop $\gamma$ of constant

4128: $\chi_1$ in the $(L,\psi)$ plane.

4129: 

4130: Importantly, the loops of constant $\chi_1$ are independent

4131: of the microwave amplitude $F$ and scale simply with $\hat{I}.$

4132: Thus, the whole quantization in the 

4133: $(L,\psi)$ plane has to be done only once.

4134: With

4135: this prescription we can unravel the semiclassical 

4136: structure of the quasienergy spectrum induced by the additional 

4137: degree of freedom spanned by $(L,\psi)$, as an amendment to the spectral 

4138: structure of the one-dimensional model discussed in section~\ref{LIN1D}.

4139: Fig.~\ref{lin3d_1} shows the equipotential curves of $\chi_1$ in the

4140: $(L,\psi)$ plane.

4141: For a comparison with quantal data, the

4142: equipotential lines plotted correspond to the quantized

4143: values of $\chi_1$ for $n_0=21$.

4144: Using the well-known properties of the Bessel functions~\cite{abramowitz72},

4145: it is easy to show that $\chi_1(L,\psi)$ has the following

4146: fixed points:

4147: \begin{itemize}

4148: \item ($L=\hat{I}_1$, arbitrary $\psi$). This corresponds to a Kepler

4149: ellipse with maximum angular momentum, i.e. a circular orbit 

4150: in a plane containing the microwave polarization axis along $\hat z$. 

4151: As such a circle corresponds to a degenerate family of elliptical

4152: orbits with arbitrary orientation of the major axis, $\psi$ is a

4153: dummy angle. This fixed point corresponds to a global

4154: maximum of $\chi_1(L=\hat{I}_1)= \hat{I}_1^2/2$,

4155: and is surrounded by ``rotational"

4156: trajectories in the $(L,\psi)$ plane.

4157: An alternative representation of the ($L,\psi$) motion on the unit sphere,

4158: spanned by $L$ and the $z$ and $\rho$-components of the Runge-Lenz vector,

4159: contracts the line representing this orbit in 

4160: fig.~\ref{lin3d_1} 

4161: to an elliptic fixed point \cite{abu97}. 

4162: \item ($L=0, \psi=\pi/2,3\pi/2).$ This corresponds to a 

4163: degenerate straight line trajectory perpendicular to the

4164: microwave field. Because of the azimuthal

4165: symmetry around the electric field axis, the two points

4166: actually correspond to the same physics.

4167: The oscillating dipole clearly vanishes there,

4168: resulting in a global minimum of $\chi_1(L=0,\psi=\pi/2,3\pi/2)=0.$

4169: This stable fixed point is surrounded by ``librational" trajectories

4170: in the $(L,\psi)$ plane.

4171: \item ($L=0, \psi=0,\pi).$ This corresponds to a 

4172: degenerate, straight line trajectory along the

4173: microwave field, i.e. the situation already considered in the

4174: 1D model of the atom. $\psi=0$ and $\pi$ correspond

4175: to the two orbits pointing up and down, which are of course

4176: equivalent. This is a saddle point of

4177: $\chi_1(L=0,\psi=0,\pi)=J'_1(1)\hat{I}_1^2.$

4178: Hence, it is an unstable equilibrium point. As an implication,

4179: in the real 3D world, the motion along the microwave axis, with the

4180: phase of the radial motion locked on the external drive,

4181: is angularly unstable (see also section \ref{staticmw}). This leads 

4182: to a slow precession of the initially degenerate Kepler 

4183: ellipse off the axis, 

4184: and will manifest itself in the localization properties

4185: of the 3D analog of the nondispersive wave-packet displayed in 

4186: fig.~\ref{lin1d_2}. This motion takes place along

4187: the separatrix between librational and rotational

4188: motion.

4189: \end{itemize}

4190: \begin{figure}

4191: \centerline{\psfig{figure=bdzf19.ps,width=10cm,angle=-90}}

4192: \caption{Isovalue curves of the angular part $\chi_1$, eq.~(\ref{substc}),

4193: of the secular

4194: Hamiltonian ${\mathcal H}_{\rm sec}$ represented in the plane

4195: of the $L_0=L/\hat{I}_1$ and $\psi$ coordinates.

4196: The slow evolution of the

4197: Kepler ellipse of a Rydberg electron driven 

4198: by a resonant, linearly polarized microwave field, takes place

4199: along such isovalue curves.

4200: $L_0=L/\hat{I}$ represents the total angular

4201: momentum (a circular trajectory in a plane containing 

4202: the field polarization axis has $L_0=1$), and

4203: $\psi$ the [canonically conjugate] angle between

4204: the field polarization axis and the major axis of the Kepler ellipse. 

4205: The separatrix emanating from the unstable fixed point

4206: ($L_0=0,\psi=0$) separates rotational and librational motion, both 

4207: ``centered" around their respective stable

4208: fixed points ($L_0=1$, $\psi$ arbitrary) and ($L_0=0$, $\psi=\pi/2$). 

4209: The former corresponds to a circular orbit

4210: centered around the nucleus. The latter

4211: represents a straight linear orbit perpendicular to the field axis. The

4212: unstable fixed point corresponds to linear motion along the

4213: polarization axis. However, this initially degenerate Kepler ellipse will

4214: slowly precess in the azimuthal plane. The equipotential curves shown here

4215: satisfy the quantization condition (\protect\ref{ebkp}), 

4216: for $n_0=21$

4217: and $p=0\ldots 20$. At lowest order, the motion in the $(L_0,\psi)$ 

4218: plane is independent of 

4219: the microwave field strength and of the resonant principal quantum

4220: number $n_0$.

4221: }

4222: \label{lin3d_1} 

4223: \end{figure}

4224: Once the quantized values of $\chi_1$ (represented by 

4225: the trajectories in fig.~\ref{lin3d_1})

4226: have been determined,

4227: we can quantize the (${\hat I},{\hat \theta}$) 

4228: motion with 

4229: these values fixed. 

4230: Fig.~\ref{lin3d_2} shows the

4231: equipotential lines of ${\mathcal H}_{\rm sec}$, 

4232: for the three values of $\chi_1$

4233: corresponding to the $p=0$, $10$ and $20$ states, see eq.~(\ref{ebkp}), 

4234: of the 

4235: $n_0=21$ manifold. In each case, the contour for the lowest state

4236: $N=0$ has been drawn, together with the separatrix between the

4237: librational and rotational (${\hat I},\hat{\theta}$) modes. 

4238: The separatrix

4239: determines the size of the principal resonance island

4240: for the different substates of the transverse

4241: motion. Note that the principal resonance is largest for the 

4242: $p=20$ state, localized closest to the stable circular

4243: orbit (hence associated with the

4244: maximum value of $\chi_1$), whereas the smallest resonance

4245: island is obtained for the $p=0$ state, localized

4246: in the vicinity  

4247: of (though not precisely at) 

4248: the straight line orbit perpendicular to the field axis

4249: (minimum value of $\chi_1$).

4250: For the latter orbit itself, the first order coupling 

4251: vanishes identically ($\chi_1=0$), which 

4252: shows that the semiclassical results obtained 

4253: from our first order approximation (in $F$) for the Hamiltonian may be quite

4254: inaccurate in the vicinity of this orbit. Higher order corrections 

4255: may become important.

4256: \begin{figure}

4257: \centerline{\psfig{figure=bdzf20.eps,width=14cm,angle=-90}}

4258: \caption{

4259: Isovalue curves of the secular Hamiltonian ${\mathcal H}_{\rm sec}$,

4260: eq.~(\ref{hamscfin}),

4261: generating the ($\hat I,\hat \theta$)

4262: motion of a Rydberg electron in a resonant microwave field. 

4263: $\hat I$ and $\hat \theta$ correspond

4264: to the atomic principal quantum number, and to the polar angle of the electron on

4265: the Kepler ellipse, respectively. The scaled microwave amplitude is fixed at

4266: $F_0=0.03$. Since the isovalues of 

4267: ${\mathcal H}_{\rm sec}$ depend on the 

4268: transverse motion in $(L,\psi)$ via the constant value

4269: of $\chi_1$, eq.~(\ref{substc}), 

4270: contours (solid lines) are shown for three characteristic values of 

4271: $\chi_1$, corresponding to fixed quantum numbers $p=0,10,20$, 

4272: eq.~(\ref{ebkp}), 

4273: of the angular

4274: motion for the $n_0=21$ resonant manifold.

4275: Only the ``ground state'' orbit satisfying eq.~(\protect\ref{ebkires}) 

4276: with $N=0$ is 

4277: shown, 

4278: together with the separatrix (dashed lines) 

4279: between librational 

4280: and rotational motion in

4281: the ($\hat I,\hat \theta$) plane. The separatrix encloses the  

4282: principal

4283: resonance island in phase space,

4284: see also 

4285: eqs.~(\ref{eqpend},\ref{area}). Panel (a) corresponds to the orbit 

4286: with

4287: $L_0=L/n_0\simeq 1$ (rotational orbit, $p=20$), panel (b) to the orbit

4288: close to the separatrix of the angular motion ($p=10$), panel (c)

4289: to the librational orbit close to the stable fixed point $L_0=0$,

4290: $\psi=\pi/2$. Note that the resonance island  is smallest for librational, 

4291: largest

4292: for rotational, and of intermediate size for separatrix modes of the

4293: angular motion.}

4294: \label{lin3d_2}

4295: \end{figure}

4296: 

4297: As discussed above, the classical motion in the $(L,\psi)$ plane is 

4298: slower than

4299: in the (${\hat I},\hat{\theta}$) plane. In the semiclassical approximation,

4300: the spacing between consecutive states corresponds to the

4301: frequency of the classical motion (see also eq.~(\ref{cprinc})). 

4302: Hence, it is to be expected that states

4303: with the same quantum number $N$, but with successive quantum numbers $p$,

4304: will lie at neighboring energies, building well-separated manifolds associated

4305: with a single value of $N$. 

4306: The energy spacing 

4307: between

4308: states in the same manifold should scale as $F_0$, while the spacing between

4309: manifolds should scale as $\sqrt{F_0}$ (remember that $F_0\ll 1$ in the case

4310: considered here). Accurate quantum calculations fully

4311: confirm this prediction, with manifolds originating from

4312: the degenerate hydrogenic energy levels at $F_0=0$, as we shall demonstrate

4313: now. 

4314: We first concentrate on the $N=0$ manifold, originating

4315: from $n_0=21$. Fig.~\ref{lin3d_3} shows 

4316: the comparison between the

4317: semiclassical and the quantum energies, for different values of the 

4318: scaled driving

4319: field amplitude $F_0=Fn_0^4$. The agreement is excellent, except for 

4320: the lowest lying states 

4321: in the manifold for $F_0=0.02$. The lowest energy level ($p=0$)

4322: corresponds to motion close to the stable fixed point $L=0$, $\psi =\pi/2$ 

4323: in fig.~\ref{lin3d_1}; the highest energy level ($p=20$)

4324: corresponds to rotational motion $L/n_0\simeq 1$. The levels with

4325: the smallest energy difference ($p=10,11$) correspond to the librational and

4326: the rotational trajectories closest to the separatrix, respectively. 

4327: The narrowing

4328: of the level spacing in their vicinity is just a consequence of the slowing 

4329: down of the

4330: classical motion~\cite{delande97}. 

4331: In the same figure, we also plot (as a dashed line) the corresponding 

4332: exact

4333: quasienergy level for the 1D model of the atom (see section~\ref{LIN1D}).

4334: As expected, it closely follows the separatrix state $p=10$. 

4335: Such good agreement is a direct proof of the validity

4336: of the adiabatic separation between the

4337: slow motion in $(L,\psi)$, and the fast motion

4338: in (${\hat I},\hat{\theta}$).

4339: \begin{figure}

4340: \centerline{\psfig{figure=bdzf21.eps,width=15cm,angle=-90}}

4341: \caption{Comparison of the semiclassical quasienergies (circles; with

4342: $N=0$, see eq.~(\protect{\ref{ebkires}})),

4343: originating from the unperturbed $n_0=21$ manifold, 

4344: to the exact quantum

4345: result (crosses), at different values of the (scaled) driving field amplitude

4346: $F_0=Fn_0^4=0.02$ (a), $0.03$ (b), $0.04$ (c). The agreement is excellent. The

4347: quantum number $p=0\ldots 20$ labels the quantized classical

4348: trajectories plotted in fig.~\protect{\ref{lin3d_1}}, starting from the

4349: librational state $\mid p=0\rangle$ at lowest energy, rising through the

4350: separatrix states $\mid p=10\rangle$ and $\mid p=11\rangle$, up 

4351: to the rotational

4352: state $\mid p=20\rangle$. The dashed line indicates the exact quasienergy of

4353: the corresponding wave-packet eigenstate of the 1D model discussed in section

4354: \ref{LIN1D}. The 1D dynamics is neatly embedded in the spectrum of

4355: the real, driven 3D atom.}

4356: \label{lin3d_3}

4357: \end{figure}

4358: 

4359: Fig.~\ref{lin3d_4} shows a global comparison of the 

4360: semiclassical prediction with

4361: the exact level dynamics (energy levels vs. $F_0$), 

4362: in a range from $F_0=0$ to $F_0=0.06$, which exceeds the

4363: typical ionization threshold ($F_0\simeq 0.05$) observed in current 

4364: experiments

4365: \cite{bayfield89,koch95b,bellermann96}. We observe that the semiclassical 

4366: prediction 

4367: tracks the exact

4368: quasienergies quite accurately, even

4369: for large $F_0$-values, 

4370: where the resonant 

4371: $n_0=21$ manifold 

4372: overlaps with other Rydberg manifolds, or with side bands of 

4373: lower or higher lying

4374: Rydberg states.

4375: The agreement becomes unsatisfactory only 

4376: in the region of very small $F_0$, where the size

4377: of 

4378: the resonance island in $(\hat I, \hat \theta)$ is very small.

4379: This is not unexpected, as semiclassics should fail when the

4380: area of the resonance island is comparable to $\hbar$, compare 

4381: eq.~(\ref{number_of_trapped_states}).

4382: In this weak driving regime, the pendulum approximation can be used

4383: to produce more accurate estimates of the energy levels.

4384: The fast $(\hat{I},\hat{\theta})$ motion is essentially identical to the one

4385: of the 1D driven hydrogen atom: thus, the Mathieu approach used

4386: in section~\ref{LIN1D} can be trivially extended to the 3D case.

4387: The only amendment is to replace the factor $J_1'(1)n_0^2$ in

4388: the expression of the Mathieu parameter $q$ by the various

4389: quantized values of $\chi_1$ for $0\leq p \leq n_0-1,$ and to use

4390: the same equation (\ref{prediction_mathieu_lin1d}) for the energy levels.

4391: 

4392:  

4393: \begin{figure}

4394: \centerline{\psfig{figure=bdzf22.ps,width=10cm,angle=-90}}

4395: \caption{Level dynamics of the numerically 

4396: exact quasienergies (dotted lines)

4397: in the vicinity of the resonant manifold emerging from $n_0=21$ ($N=0$),

4398: compared to the semiclassical prediction (full lines), for 

4399: $F_0=Fn_0^4=0\ldots 0.06$.

4400: Note that the maximum field amplitude exceeds the typical ionization

4401: thresholds measured in current experiments for $\omega n_0^3 \simeq 1$

4402: \protect\cite{bayfield89,koch95b,bellermann96}. 

4403: Nontheless, the semiclassical prediction accurately 

4404: tracks the exact

4405: solution across a large number of avoided crossings

4406: with other Rydberg manifolds.

4407: }

4408: \label{lin3d_4}

4409: \end{figure}

4410: 

4411: The semiclassical construction of the energy levels from classical

4412: orbits is -- necessarily -- reflected in the localization properties of the

4413: associated eigenstates, as demonstrated by the electronic densities of the

4414: states $\mid p=0\rangle$, $\mid p=10\rangle$, and $\mid p=20\rangle$ in

4415: fig.~\ref{lin3d_5}, for the same field amplitudes as in 

4416: fig.~\ref{lin3d_3}. Note that, in this plot, the electronic densities 

4417: are averaged over one field cycle,

4418: hence display only the angular localization properties of the eigenstates.

4419: Their localization along the classical orbits defined by the

4420: stable or unstable fixed points of the

4421: $(L,\psi)$ dynamics is obvious \cite{abu96,abu98a,abu97}.

4422: Note in particular the nodal

4423: structure of the state $\mid p=10\rangle$, associated with the unstable 

4424: fixed point:

4425: there are sharp nodal lines perpendicular to the $z$-axis, reflecting the

4426: dominant motion along the $z$-axis, but also nodal lines of low visibility in

4427: the angular direction. They are a manifestation of the slow classical

4428: precession of the Kepler ellipse, i.e. the slow secular evolution in

4429: the ($L,\psi$) plane. 

4430: The quantum state, however, dominantly

4431: exhibits the motion along the $z$-axis, as a signature of the effective

4432: separation of time scales of radial and angular motion.

4433: Finally, it should be realized from a comparison of the top to the middle and

4434: bottom row of fig.~\ref{lin3d_5}

4435: that the quasiclassical localization properties of the eigenstates are

4436: essentially unaffected as $F$ rises, despite various avoided crossings which

4437: occur at intermediate field values, see fig.~\ref{lin3d_4}. Especially,

4438: the angular structure does not depend at all on $F$, as predicted by the

4439: secular approximation.

4440: \begin{figure}

4441: \centerline{\psfig{figure=bdzf23.ps,width=15cm}}

4442: \caption{Electronic densities of the extremal librational ($p=0$, left), 

4443: separatrix ($p=10$, center), and extremal rotational ($p=20$, right) 

4444: quasienergy states of the $n_0=21$  manifold of a 3D hydrogen

4445: atom exposed to a resonant microwave field

4446: in cylindrical coordinates $(\rho,z)$, at different values of the

4447: driving field amplitude $F_0=Fn_0^4=0.02$ (top), $0.03$ (middle), $0.04$ (bottom), 

4448: averaged

4449: over one period of the driving field. Note the clear localization along the

4450: classical orbits corresponding to the respective contours in

4451: fig.~\protect\ref{lin3d_1}, for {\em all} field amplitudes. The nodal lines

4452: of the electronic densities clearly exhibit the direction of the underlying

4453: classical motion. The field-induced finite decay rate of the eigenstates 

4454: (see section~\protect\ref{ION}) is negligible on time scales shorter than 

4455: approx. $10^6$ Kepler periods. Each box extends over 

4456: $\pm 1000$ Bohr radii, in both $\rho$ (horizontal) and

4457: $z$ (vertical) directions, with the nucleus at the center of the plot. The microwave polarization axis is oriented  

4458: vertically along $z.$}

4459: \label{lin3d_5}

4460: \end{figure}

4461: 

4462: The eigenstates displayed here are localized

4463: along classical trajectories which are resonantly driven by the external

4464: field. Hence, we should expect them to exhibit wave-packet like motion along

4465: these trajectories, as the phase of the driving field is changed. 

4466: This is indeed the case as illustrated  

4467: in fig.~\ref{lin3d_8} for the state $p=20$ 

4468: with maximal angular momentum 

4469: $L/n_0\simeq 1$ \cite{kalinski95b,abu96,abu98a}. 

4470: Due to the azimuthal symmetry of the problem, the actual 3D electronic

4471: density is obtained by rotating the figure around the vertical axis.

4472: Thus, the wave-packet is  actually a doughnut moving periodically

4473: from the north to the south

4474: pole (and back)

4475: of a sphere, slightly deformed along the field direction. The interference

4476: resulting from the contraction of this doughnut to a compact wave-packet at

4477: the poles is clearly visible at phases $\omega t=0$ and $\omega t=\pi$ in the

4478: plot. Note that the creation of unidirectional wave-packet

4479: eigenstates moving along a circle in the plane containing the field

4480: polarization axis is not possible for the real 3D atom \cite{abu98a},

4481:  as opposed to the

4482: reduced 2D problem studied in \cite{kalinski95b}, due to the 

4483: abovementioned azimuthal

4484: symmetry (see also section \ref{EP}).

4485: \begin{figure}

4486: \centerline{\psfig{figure=bdzf24.ps,width=14cm}}

4487: \caption{Temporal evolution of the electronic density

4488: of the extremal rotational quasienergy state

4489: $\mid p=20\rangle$ of the $n_0=21$ resonant manifold, 

4490: for different phases $\omega t=0$ (left),

4491: $\omega t=\pi/2$ (center), $\omega t=\pi$ (right) of the driving field, at

4492: amplitude $F_0=0.03$, in cylindrical coordinates. 

4493: Each box extends over $\pm 700$ Bohr radii, in both directions, $\rho$

4494: (horizontal) and

4495: $z$ (vertical). The microwave polarization axis is oriented 

4496: along $z$. Because of the azymuthal symmetry of the problem, 

4497: the actual 3D electronic

4498: density is obtained by rotating the figure around the vertical axis. 

4499: The state represents

4500: a non-dispersive wave-packet shaped like a doughnut, moving periodically from

4501: the north to the south pole (and back) of a sphere. For higher $n_0$, the

4502: angular localization on the circular orbit should improve.}

4503: \label{lin3d_8}

4504: \end{figure}

4505: For other states in the $n_0=21$ resonant manifold, the longitudinal

4506: localization along the periodic orbit is less visible. The reason is that

4507: $\chi_1$ is smaller than for the $p=20$ state, leading to a smaller

4508: resonance island in $(\hat{I},\hat{\theta})$ (see fig.~\ref{lin3d_2}) 

4509: and, consequently, to less

4510: efficient localization. Proceeding to higher $n_0$-values should improve

4511: the situation.

4512: 

4513: Let us briefly discuss ``excited'' states in the resonance island,

4514: i.e. manifolds corresponding to $N>0$ in eq.~(\ref{ebkires}). 

4515: Fig.~\ref{lin3d_9}

4516: shows the exact level dynamics, with the semiclassical prediction for

4517: $N=1$ superimposed \cite{abu98a}.

4518:  The states in this manifold originate from $n_0=22$. We

4519: observe quite good agreement between the quantum and semiclassical results for

4520: {\it high} lying states in the manifold (for which the principal action

4521: island is large, see fig.~\ref{lin3d_2}). 

4522: For lower lying states the agreement is improved for higher

4523: values of $F_0$. If $F_0$ is too low, the states are not fully localized 

4524: inside 

4525: the resonance island and, consequently,

4526: are badly reproduced by the resonant semiclassical 

4527: approximation.  This is further

4528: exemplified in fig.~\ref{lin3d_10}, for $N=2$. 

4529: Here, the

4530: agreement is worse than for smaller values of $N$, and is observed only 

4531: for

4532: large $F_0$ and large $p$. This confirms the picture that the validity

4533: of the semiclassical approach outlined here 

4534: is directly related to the size,

4535: eqs.~(\ref{area},\ref{number_of_trapped_states}), of the

4536: resonance island in $(\hat I, \hat\theta)$ space (see also the discussion

4537: in section~\ref{PTP}).

4538: \begin{figure}

4539: \centerline{\psfig{figure=bdzf25.ps,width=10cm,angle=-90}}

4540: \caption{Comparison of the numerically 

4541: exact level dynamics (dotted lines)

4542: with the

4543: semiclassical prediction (solid lines),

4544: for the $n_0=22$ ($N=1$) manifold of a 3D hydrogen atom exposed to

4545: a microwave field with frequency $\omega=1/(21)^3,$ resonant

4546: with the $n_0=21$ manifold.

4547: For sufficiently high $F_0,$ the quantum 

4548: states originating at $F_0=0$ from the unperturbed $n_0=22$ level 

4549: are captured by the principal resonance island and then represent 

4550: the first excited state of the motion in the ($\hat I, \hat \theta$)

4551: plane (i.e., $N=1$ in eq.~(\protect{\ref{ebkires}})). Since the island's 

4552: size  

4553: depends on the angular ($L,\psi$) motion (value of $p$ in 

4554: eq.~(\protect{\ref{ebkp}}), see also fig.~\protect\ref{lin3d_2}),

4555: states with large $p$ enter the resonance zone first. For these,

4556: the agreement between quantum and semiclassical quasienergies 

4557: starts to be satisfactory at lower $F_0$ values than for low-$p$ states.}

4558: \label{lin3d_9}

4559: \end{figure}

4560: \begin{figure}

4561: \centerline{\psfig{figure=bdzf26.ps,width=10cm,angle=-90}}

4562: \caption{Same as fig.~\protect{\ref{lin3d_9}}, but for $N=2$. The quantum 

4563: states originate

4564: from the manifold $n_0=20$. The 

4565: resonance island in the ($\hat I,\hat \theta$) 

4566: coordinates

4567: is too small to support any $N=2$ states for $F_0<0.03$, as seen from the

4568: negative slope of the quasienergy levels. The quantum states 

4569: ``cross'' the separatrix as the amplitude is further increased,

4570: and successively enter the resonance zone, starting from the 

4571: largest value of $p$ in eq.~(\protect{\ref{ebkp}}) 

4572: (the resonance island size increases with $p$).

4573: Even for $F_0>0.04,$ only a minority of substates of 

4574: the $n_0=20$ manifold is well represented by the resonant semiclassical 

4575: dynamics (indicated by the solid lines).}

4576: \label{lin3d_10}

4577: \end{figure}

4578: 

4579: Finally note that, as already mentioned at the end of section~\ref{LIN1D}, all

4580: wave-packet eigenstates have a finite decay rate which induces a slow, global

4581: reduction of the electronic density localized on the resonantly driven

4582: classical periodic orbit. However, the time scale of this decay is of the

4583: order of thousands to millions of Kepler cycles, and therefore leaves our

4584: above conclusions unaffected. However, some very intriguing consequences of

4585: the nonvanishing continuum coupling will be discussed in section~\ref{ION}.

4586: 

4587: \subsection{Rydberg states in circularly polarized microwave fields}

4588: \label{CP}

4589: 

4590: As shown in the preceding section, the use of a linearly polarized

4591: microwave field is not sufficient to produce a non-dispersive

4592: wave-packet fully localized in all three dimensions,

4593: due to the azymuthal symmetry around the

4594: microwave polarization axis. To get more flexibility, one may consider

4595: the case of arbitrary polarization. It turns out that the results

4596: are especially simple in circular polarization. They are the subject of this

4597: section.

4598: 

4599: In most experiments on 

4600: microwave driven Rydberg atoms,

4601: linearly polarized (LP) microwaves have 

4602: been used \cite{bayfield89,koch95b,bayfield74,bayfield96,galvez88}. 

4603: For circular polarization (CP),

4604: first experiments were performed for alkali

4605: atoms in the late eighties \cite{fu90,cheng96}, 

4606: with hydrogen atoms 

4607: following only recently \cite{bellermann96}. The latter experiments 

4608: also studied 

4609: the general case of elliptic polarization (EP). While, 

4610: at least theoretically, different

4611: frequency regimes 

4612: were considered for CP microwaves

4613: (for a review see \cite{delande97b}) --

4614: we shall restrict 

4615: our discussion here to 

4616: resonant driving.

4617: Given a different 

4618: microwave polarization, and thus a different form 

4619: of the interaction Hamiltonian, eq.~(\ref{vcirc}),

4620: Kepler trajectories which are distinct from those considered in the LP case

4621: will be most efficiently locked on the external driving. Hence, in the sequal,

4622: we shall launch nondispersive wave-packets along periodic orbits which are

4623: distinct from those encountered above.

4624: 

4625: Historically, the creation of non-dispersive wave-packets in 

4626: CP and LP microwave fields,

4627: respectively, 

4628: has been considered quite independently.

4629: In particular, in the CP case, the notion of nondispersive 

4630: wave-packets has been introduced \cite{ibb94} 

4631: along quite 

4632: different lines 

4633: than the one adopted in this

4634: review. The original work, as well as subsequent studies of the CP situation

4635: \cite{farrelly95a,ibb95,kalinski95a,delande95,kuba95a,kalinski96b,kuba97a,kuba97b}

4636: used 

4637: the fact that, in this specific case, 

4638: the time-dependence of the

4639: Hamiltonian may be removed by a unitary transformation to the rotating frame

4640: (see below). Thus, the stable periodic orbit 

4641: at the center of the island 

4642: turns into a stable equilibrium point in the rotating frame. This allows the

4643: expansion of the Hamiltonian into a Taylor series in the vicinity of 

4644: the fixed point,

4645: and in particular a standard harmonic treatment using normal modes. 

4646: We shall review 

4647: this line of reasoning in detail below. It is, however, instructive 

4648: to first 

4649: discuss

4650: the very same system using the general resonance approach

4651: exposed in section~\ref{GM}. 

4652: 

4653: 

4654: \subsubsection{Hamiltonian}

4655: 

4656: With $\hat z$ the propagation direction of the microwave,

4657: the electric field rotates

4658: in the $x-y$ plane, and the Hamiltonian (\ref{h}) takes

4659: the

4660: following explicit form:

4661: \begin{equation}

4662: H_{\rm CP}=\frac{\vec{p}^2}{2}-\frac{1}{r}+

4663: F\left\{x\cos(\omega t)+y\sin(\omega t)\right\}.

4664: \label{ham_cp_3d}

4665: \end{equation}

4666: 

4667: In contrast to the LP 

4668: situation, there is no simplified one-dimensional

4669: model in the CP case. However, a simplified two-dimensional 

4670: model 

4671: exists, where

4672: the motion is restricted to the $x-y$ plane. As long as one is interested only

4673: in the dynamics of non-dispersive wave-packets, this motion is 

4674: stable (see below),

4675: which means that a small deviation from the $z=0$ plane does not affect the

4676: qualitative behavior. Hence, much physical insight can be obtained

4677: from the simplified 2D model. It 

4678: will be discussed in section \ref{2d_model}.

4679: 

4680: \subsubsection{Resonance analysis}

4681: \label{reso_an}

4682: We follow the general treatment exposed in sections~\ref{CD} and \ref{LIN1D}

4683: for the 1D case,

4684: and 

4685: express the external perturbation as a function of the action-angle

4686: variables 

4687: $(I,\theta)$, $(L,\psi)$ and $(M,\phi)$ introduced in section~\ref{RSEF}. 

4688: By inserting equations~(\ref{v_3d_x},\ref{v_3d_y}) in eq.(\ref{ham_cp_3d}), 

4689: after appropriate account for the projection of the body-fixed frame 

4690: $(x',y',z')$ onto the laboratory frame $(x,y,z)$, eq.~(\ref{rot_euler}),

4691: we obtain for the Floquet Hamiltonian:

4692: \begin{equation}

4693: {\mathcal H}=

4694: P_t

4695: -\frac{1}{2I^{2}}+F   \sum_{m=-\infty}^{\infty}\left[

4696:     V_m \cos(m\theta+\phi-\omega t)

4697:     -U_m

4698:     \sin(m\theta+\phi-\omega t)\right],

4699: \label{h_circ_aa}

4700: \end{equation}

4701: where the Fourier coefficients are given by 

4702: (see also eqs.~(\ref{xm}-\ref{y0})):

4703: \begin{eqnarray}

4704:  V_{m}(I,L,M,\psi) &=& \cos\psi \left(Y_m + \frac{M}{L} X_m\right), \label{Vb}\\

4705:  U_{m}(I,L,M,\psi) &=& \sin\psi \left(X_m + \frac{M}{L} Y_m\right).

4706: \label{Ub}

4707: \end{eqnarray}

4708: Once again, transformation 

4709: to the ``rotating frame", eqs.~(\ref{rotframe_a}-\ref{rotframe_c}), and

4710: averaging over one field period $T=2\pi/\omega$ (thereby neglecting all 

4711: rapidly varying terms) leaves us with the explicit form of the

4712: secular Hamiltonian:

4713: \begin{equation}

4714: {\mathcal H}_{\rm sec}=\hat{P}_t-\frac{1}{2\hat I^2}-\omega \hat I

4715: +F\left[V_1(\hat{I},L,M,\psi)\cos (\hat{\theta}+\phi) - 

4716: U_1(\hat{I},L,M,\psi)\sin (\hat{\theta}+\phi) \right],

4717: \label{hamsccir}

4718: \end{equation}

4719: similar to eq.~(\ref{hamsc2}).

4720: This can be rewritten as:

4721: \begin{equation}

4722: {\mathcal H}_{\rm sec}=\hat{P}_t-\frac{1}{2\hat I^2}-\omega \hat I

4723: +F \chi_1 \cos (\hat{\theta}+\phi+\delta_1)  ,

4724: \label{hamsccir2}

4725: \end{equation}

4726: with the effective perturbation

4727: \begin{equation}

4728: \chi_1(\hat{I},L,M,\psi)=\sqrt{V_1^2+U_1^2},

4729: \label{widthcir}

4730: \end{equation}

4731: and

4732: \begin{equation}

4733: \tan \delta_1(\hat{I},L,M,\psi) = \frac{U_1}{V_1}.

4734: \label{phasecirc} 

4735: \end{equation}

4736: 

4737: This secular Hamiltonian, which has, once again, 

4738: the same structure as the 1D secular Hamiltonian~(\ref{hsec_ap}),

4739: governs the ``slow" dynamics of the system

4740: in the vicinity of the resonance. Similarly to the LP case, the

4741: various degrees of freedom evolve on different time scales:

4742: \begin{itemize}

4743: \item In the $(\hat{I},\hat{\theta})$ plane, the situation is exactly like

4744: for a one-dimensional system. There is a resonance island around

4745: the resonant action,

4746: with a pendulum-like structure. Non-dispersive wave-packets

4747: are associated with eigenstates localized at the center of this island,

4748: at the point (see eqs.~(\ref{scaling_action},\ref{hamsccir2})):

4749: \begin{equation}

4750: \hat{I}_1 = \omega^{-1/3} = n_0,\ \ \ \ \hat{\theta} = -(\phi + \delta_1).

4751: \label{center_island_CP}

4752: \end{equation}

4753: The period of the secular 

4754: classical motion close to the resonance center scales as 

4755: $1/\sqrt{F}$. 

4756: It 

4757: defines an

4758: intermediate time scale, slower than

4759: the Kepler frequency, but faster than the transverse motion in the

4760: other coordinates $(L,\psi)$ and $(M,\phi)$. 

4761: \item In the subspace spanned by $(L,M,\psi,\phi)$, the motion is much 

4762: slower, with

4763: a time scale proportional to $1/F$. The effective Hamiltonian

4764: describing this motion is obtained by averaging the fast

4765: motion in the perturbation which describes

4766: this motion $(\hat{I},\hat{\theta})$ plane, which in turns

4767: implies that $\chi_1$ itself is constant for both the motion in 

4768: $(\hat{I},\hat{\theta})$ and $(L,M,\psi,\phi)$ space.

4769: \end{itemize}

4770: Note that

4771: $\chi_1$ does not depend on the angle $\phi$. This,

4772: in turn, implies that $M$ is a constant of the slow motion. 

4773: This is because the circular polarization

4774: does not define any preferred direction in the  $x-y$ polarization plane.

4775: 

4776: Once again, much alike our discussion in section~\ref{LIN3D},

4777: the well-known properties of Bessel functions

4778: \cite{abramowitz72}, together with eqs.~(\ref{Vb},\ref{widthcir}), imply

4779: that, for given $M$, the maximum of $\chi_1$ occurs at 

4780: $L=M$,

4781: corresponding to the situation when the electronic motion is

4782: restricted to the polarization plane. In this plane, 

4783: the maximum $\chi_1=\hat{I}^2$

4784: is reached for the circular orbit defined by $M=L=\hat{I}$ (i.e. 

4785: $L_0=1$). 

4786: This defines

4787: a resonant periodic orbit locked on the 

4788: external microwave driving, which maximizes the

4789: effective Hamiltonian in each coordinate and is, therefore,

4790: fully 

4791: stable in all phase space directions. 

4792: The orbit is a circular Kepler orbit in the polarization

4793: plane, where the electron rotates around the nucleus with exactly

4794: the angular velocity of the microwave. It is not really surprising

4795: that this orbit maximizes the interaction energy with the external

4796: field: indeed, along this orbit, the atomic dipole rotates 

4797: exactly in phase with the polarization vector of 

4798: the circularly polarized microwave field. 

4799: As in the case of linear polarization discussed in 

4800: section~\ref{LIN3D}, the angular motion in the

4801: $(L,\psi,M,\phi)$ variables (which is trivial

4802: in $(M,\phi)$, since $M$ is constant) could be studied in detail.

4803: For the sake of brevity, we will not repeat such an analysis here. We rather

4804: concentrate on the  wave-packets 

4805: which are best localized in the resonance island

4806: near the circular orbit.

4807: The simplest approximation 

4808: to describe these states is to replace

4809: the largest quantized value of $\chi_1$ by its maximum value $\hat{I}^2$

4810: estimated at the center of the resonance island, eq.~(\ref{center_island_CP}).

4811: Then, the situation is similar to the 1D model of the atom, 

4812: eq.~(\ref{hsec_h1d}),

4813: except that the strength of the coupling is $\hat{I}^2$ instead of 

4814: $-J_1'(1)\hat{I}^2.$ In complete analogy to the steps leading from 

4815: eq.~(\ref{hsec_h1d}) to eqs.~(\ref{semiharm1dval},\ref{omega_h1d}) we employ 

4816: the pendulum approximation with a subsequent harmonic expansion

4817: around the pendulum's stable equilibrium point, 

4818: deeply inside the resonance island.

4819: The harmonic frequency of the motion in the $(\hat I,\hat\theta)$ plane is:

4820: \begin{equation}

4821: \omega_{\mathrm harm}= 

4822: \frac{\sqrt{3F}}{n_0} = \omega \sqrt{3F_0},

4823: \label{omega_harmonic_cp}

4824: \end{equation}

4825: and the quasi-energy levels are:

4826: \begin{equation}

4827: {\cal E}_{N,k} = k\omega -\frac{3}{2n_0^2} + n_0^2F - \left(N+\frac{1}{2}\right) \omega_{\mathrm harm}.

4828: \label{estharm}

4829: \end{equation}

4830: Note that, by construction, 

4831: $-3/2n_0^2 + F n_0^2$ is nothing but the energy at the center

4832: of the resonance island, i.e. the energy of the resonant circular orbit.

4833: For very small $F$, the resonance island shrinks and may

4834: support only a small number of states, or even no state at all. 

4835: In this regime, the harmonic

4836: approximation, eq.~(\ref{spectrum_harmonic}), breaks down.

4837: Alternatively, one can apply a quantum treatment

4838: of the pendulum motion in the $(\hat{I},\hat{\theta})$ plane, as explained

4839: in section~\ref{section_mathieu} and discussed in section~\ref{LIN1D} 

4840: for the

4841: 1D model of the atom exposed to a linearly polarized

4842: microwave. The analysis -- essentially identical to the one in

4843: section~\ref{LIN1D} --  yields the following expression for the

4844: energy levels:

4845: \begin{equation}

4846: {\cal E}_{k,N}= k\omega- \frac{3}{2n_0^2}-\frac{3a_N(\nu,q)}{8n_0^4},

4847: \end{equation}

4848: where $a_N(\nu,q)$ are the Mathieu eigenvalues (compare with 

4849: eq.~(\ref{map_mathieu2})

4850: for the general case), with

4851: \begin{equation}

4852: q=\frac{4}{3}Fn_0^6,

4853: \end{equation}

4854: and 

4855: \begin{equation}

4856: \nu=-2n_0 \ \ ({\mathrm mod}\ 2)

4857: \end{equation}

4858: the characteristic exponent.

4859: 

4860: These expressions are valid for the states localized close to the resonant

4861: circular orbit.

4862: For the other states, the calculation is essentially identical, the

4863: only amendment being the use of the values 

4864: of $\chi_1$

4865: following from the quantization of the secular motion, instead of the maximum

4866: value $n_0^2.$ 

4867: 

4868: Finally, as the center of the resonance island corresponds to a circular

4869: trajectory in the $(x,y)$ plane, the Floquet states associated with 

4870: the non-dispersive 

4871: wave-packets

4872: will be essentially composed of combinations of circular states 

4873: $|n,\ L=M=n-1\rangle$,

4874: with coefficients described by the solutions

4875: of the Mathieu equation, as explained in section~\ref{section_mathieu}.

4876: This Mathieu formalism has been rediscovered in this particular CP situation

4877: via complicated approximations 

4878: on the exact Schr\"odinger equation in \cite{kalinski96b}. 

4879: We believe that the standard resonance analysis using the

4880: pendulum approximation leads, at the same time, to simpler calculations, 

4881: and to

4882: a much more transparent

4883: physical picture.

4884:  

4885: \subsubsection{The two-dimensional model}

4886: \label{2d_model}

4887: 

4888: We shall now discuss the simplified 2D model of the CP problem, which 

4889: amounts to restricting the motion to the

4890: $(x,y)$ plane, but retains almost all the features of the full 3D problem. 

4891: Instead of the six-dimensional phase space 

4892: spanned by the 

4893: action-angle variables $(I,\theta)$, $(L,\psi)$, $(M,\phi)$, one is left

4894: with a four-dimensional submanifold 

4895: with coordinates $(I,\theta)$, $(M,\phi),$ see section~\ref{action-angle_hydrogen}. 

4896: The secular Hamiltonian  

4897: then reads 

4898: (compare eqs.~(\ref{hamsccir2},\ref{widthcir})):

4899: \begin{equation}

4900: {\mathcal H}_{\rm sec}=\hat{P}_t-\frac{1}{2\hat I^2}-\omega \hat I

4901: +FV_1(\hat{I},M)\cos (\hat{\theta}+\phi)  ,

4902: \label{hamsccir_2d}

4903: \end{equation}

4904: with (see eq.~(\ref{Vb}))

4905: \begin{equation}

4906: V_1(\hat{I},M)= \hat{I}^2 \left[ J_{1}^{'}(e) + {\mathrm sign}(M)

4907:  \frac{\sqrt{1-e^2}}{e}J_{1}(e) \right] ,

4908: \end{equation}

4909: where (as in eq.~(\ref{eccentricity}))

4910: \begin{equation}

4911: e=\sqrt{1-\frac{M^2}{\hat{I}^2}}.

4912: \end{equation}

4913: 

4914: However, the Maslov index for the $(\hat{I},\hat\theta)$ motion is different.

4915: Indeed, the energy spectrum of 

4916: the 2D atom is given by eq.~(\ref{spectrum_2d}).

4917: Thus, quantized values of the action

4918: are half-integer multiples of $\hbar$. The relation

4919: between the resonant action $\hat{I}_1=\omega^{-1/3}$ and the

4920: corresponding principal quantum number now reads (with $\hbar=1$):

4921: \begin{equation}

4922: n_0=\hat{I}_1-\frac{1}{2}.

4923: \end{equation}

4924: As explained in section~\ref{section_mathieu}, the optimal

4925: case for the preparation of non-dispersive wave-packets --

4926: where the states are the most deeply bound

4927: inside the resonance island --  is for integer

4928: values of $n_0,$ i.e. frequencies

4929: (compare with eqs.~(\ref{resonance_condition},\ref{omk}))

4930: \begin{equation}

4931: \omega = \frac{1}{\left(n_0+\frac{1}{2}\right)^3}.

4932: \end{equation}

4933: For the energy levels of the non-dispersive wave-packets,

4934: this also implies that the characteristic exponents in the Mathieu

4935: equation -- see section~\ref{section_mathieu} -- are shifted

4936: by one unit:

4937: \begin{equation}

4938: \nu=-2n_0 \ \ ({\mathrm mod}\ 2)\ \ = -2\hat{I}_1+1\ \ ({\mathrm mod}\ 2).

4939: \end{equation}

4940: 

4941: 

4942: \subsubsection{Transformation to the rotating frame}

4943: \label{rotating_frame}

4944: 

4945: The resonance analysis developed above is restricted to first

4946: order in the amplitude $F$ of the external drive. Extensions to higher

4947: orders are possible, but tedious. For CP,

4948: an alternative approach is possible,

4949: which allows higher orders to be included quite easily.

4950: It is 

4951: applicable to CP only and thus

4952: lacks the generality of the resonance approach we used so far. 

4953: Still, it is rather

4954: simple and deserves an analysis.

4955: 

4956: In CP, one may remove the time dependence of the

4957: Hamiltonian (\ref{ham_cp_3d}) by a transformation 

4958: to the noninertial frame rotating

4959: with the external frequency $\omega$. The unitary fransformation

4960: $U=\exp(i\omega L_z t)$ leads to \cite{bunkin64,grozdanov92}

4961: \begin{equation}

4962: H_{\rm rot}= U \ H_{\rm CP}\ U^{\dagger} + 

4963: i U \frac{\partial U^{\dagger}}{\partial t}

4964: =

4965: \frac{\vec{p}^2}{2} -\frac{1}{r} +Fx -\omega L_z.

4966: \label{hrot}

4967: \end{equation}

4968: Classically, such an operation corresponds to a time dependent rotation of the

4969: coordinate frame spanned by $\overline{x} = x\cos \omega t +y\sin \omega t$, 

4970: $\overline{y} = y\cos \omega t-x \sin \omega t$ 

4971: (and dropping the bar 

4972: hereafter)\footnote{Passing to the rotating frame implies  

4973: a change of 

4974: $\phi$ to $\overline{\phi}=\phi-\omega t$ in eq.~(\ref{h_circ_aa}). 

4975: That is definitely

4976: different from the change $\theta \to \hat{\theta}=\theta-\omega t$, 

4977: eq.~(\ref{rotframe_a}), used in the resonance analysis. 

4978: Both transformations are unfortunately

4979: known under the same name of ``passing to the rotating frame". This is

4980: quite confusing, but one has to live with it. Along the resonantly driven

4981: circular orbit we are considering here, 

4982: it happens that the azimuthal angle $\phi$ and the polar angle 

4983: $\theta$ actually coincide.

4984: It follows that the two approaches are equivalent {\em in the vicinity}

4985: of this orbit.}.

4986: The Hamiltonian~(\ref{hrot}), as a time-independent operator, has some energy

4987: levels and corresponding eigenstates. Its spectrum is not 

4988: $\omega$-periodic, although the unitary transformation assures that there is

4989: a one-to-one correspondence between its spectrum (eigenstates) and the

4990: Floquet spectrum 

4991: of eq.~(\ref{ham_cp_3d})\footnote{In

4992: fact, if $|\phi_i\rangle$ is an eigenstate of $H_{\rm rot}$ with energy $E_i$,

4993:  then $U^{\dagger}|\phi_i \rangle$ is

4994: a Floquet eigenstate with quasi-energy $E_i$, while states shifted in

4995: energy by $k\omega$ are of the form $\exp (ik\omega t)U^{\dagger}|\phi_i

4996:  \rangle.$ For a more detailed discussion of this point, 

4997: see \cite{delande98}.}.

4998: It was observed \cite{klar89} that the Hamiltonian~(\ref{hrot})

4999: allows for

5000: the existence of a stable fixed (equilibrium) 

5001: point in a certain range of the microwave 

5002: amplitude $F$. Later on, it was realized \cite{ibb94}

5003: that wave-packets initially

5004: localized in the vicinity of this fixed point will not disperse

5005: (being bound by the fact that the fixed point is stable) for 

5006: at least several Kepler periods. 

5007: In the laboratory frame, these wave-packets (also

5008: called ``Trojan states" \cite{farrelly95a,ibb95,ibb94,kalinski95a})

5009: appear as wave-packets moving around the

5010: nucleus along the circular trajectory, which is nothing

5011: but the periodic orbit at the center of the resonance island discussed in 

5012: section~\ref{reso_an}.

5013: In the original formulation \cite{ibb94} and the discussion which followed

5014: \cite{farrelly95a,ibb95,kalinski95a,kalinski96b},

5015: great attention was paid to the accuracy of the harmonic

5016: approximation (see below). This was of 

5017: utmost importance for 

5018: the non-spreading

5019: character of {\it Gaussian-shaped} Trojan wave-packets considered in

5020: \cite{farrelly95a,ibb95,ibb94,kalinski95a,kalinski96b}. 

5021: As soon pointed out in \cite{delande95}, however, 

5022: the accuracy of

5023: this approximation

5024: is immaterial for the very existence of

5025: the wave-packets, which are to be identified, as shown above, with

5026: well-defined Floquet states.

5027: 

5028: Let us 

5029: recapitulate the fixed point analysis of \cite{ibb94,klar89}

5030: in the rotating frame.

5031: Inspection of the classical version of the  Hamiltonian $H_{\rm rot}$,

5032: eq.~(\ref{hrot}), 

5033: shows that, due to symmetry, 

5034: one may seek the fixed point

5035: at $z=y=0$. The condition for an equilibrium (fixed) point,

5036:  i.e., $d\vec{r}/dt=0,\ d\vec{p}/dt=0$,

5037: yields immediately that $p_{z,\rm eq}=p_{x, \rm eq}=0$, 

5038: $p_{y,\rm eq}=\omega x_{\rm eq}$,

5039: with the subscript ``$\rm eq$'' for 

5040: ``equilibrium''.

5041: The remaining equation

5042: for $d p_x/dt$ gives the condition

5043: \begin{equation}

5044:   -F+\omega^2x_{\rm eq}-\frac{|x_{\rm eq}|}{x_{\rm eq}^3}=0,

5045: \label{wp0}

5046: \end{equation}

5047: that defines the position of the fixed point as a function of $F$.  

5048: Following \cite{ibb94} let us introduce the 

5049: dimensionless parameter

5050: \begin{equation}

5051: q=\frac{1}{\omega^{2}|x_{\rm eq}|^{3}}.

5052: \label{q}

5053: \end{equation}

5054: One may easily express the fixed point position, the microwave field

5055: amplitude, as well as the corresponding energy in terms of $q$ and $\omega$.

5056: Explicitly:

5057: \begin{equation}

5058: x_{{\rm eq}1}=\frac{1}{q^{1/3}\omega^{2/3}}, 

5059: \qquad F=\frac{1-q}{q^{1/3}}\omega^{4/3},

5060:  \qquad  E_{\rm eq}=\frac{1-4q}{2}\left(\frac{\omega}{q}\right)^{2/3},

5061: \label{wps}

5062: \end{equation}

5063: and

5064: \begin{equation}

5065: x_{{\rm eq}2}=-\frac{1}{q^{1/3}\omega^{2/3}}, 

5066: \qquad F=\frac{q-1}{q^{1/3}}\omega^{4/3},

5067:  \qquad  E_{\rm eq}=\frac{1-4q}{2}\left(\frac{\omega}{q}\right)^{2/3}.

5068: \label{wpu}

5069: \end{equation}

5070: For $F=0$,  $q=1$ in eqs.~(\ref{wps},\ref{wpu}).

5071: For $F<0,$ (i.e., $q>1$ in eq.~(\ref{wps})), $x_{{\rm eq}1}$ 

5072: is an unstable fixed point, while for moderately

5073: positive $F$ (i.e., $8/9<q<1$ \cite{ibb94,klar89}) it is stable.

5074: Stability of the second equilibrium point  $x_{{\rm eq}2}$

5075: is achieved by changing the sign of $F.$ For moderate

5076: fields ($q$ close to unity), the stable and the unstable fixed points 

5077: are located on

5078: opposite sides of the nucleus, and at almost the same distance from it. 

5079: As the whole analysis is classical, it has to obey the scaling laws

5080: discussed in section~\ref{scaling_laws}. Hence, all quantities describing

5081: the equilibrium points in the preceding equations scale as powers

5082: of the microwave frequency. A consequence

5083: is that there is a very simple correspondence

5084: between the scaled microwave amplitude and the dimensionless parameter $q$:

5085: \begin{equation}

5086: F_0=F\omega^{4/3} = F n_0^4 = \frac{1-q}{q^{1/3}}.

5087: \end{equation}

5088: The parameter $q$ can thus be thought of as 

5089: a convenient parametrization (leading to simpler algebraic formula)

5090: of the scaled microwave amplitude.

5091: 

5092: A fixed point in the rotating frame corresponds

5093: to a periodic orbit with exactly the period $T=2\pi/\omega$ of the

5094: microwave driving field in

5095: the original frame. The stable fixed point

5096: (periodic orbit) thus corresponds to the center of the resonance island, and to

5097: the stable equilibrium point of the pendulum in the secular approximation.

5098: Similarly, the unstable fixed point corresponds to the

5099: unstable equilibrium point of the pendulum.

5100: Note that the stable fixed point approaches

5101: $x_{\rm eq}=\omega^{-2/3}=\hat{I}_1^2$ when $F\rightarrow 0$, 

5102: i.e. the radius of the circular classical Kepler 

5103: trajectory with frequency $\omega$.

5104: Thus,

5105: the stable fixed point

5106: smoothly reaches the location  of the circular state 

5107: of the hydrogen atom, with a classical Kepler frequency equal

5108: to the driving microwave frequency.  Its energy $E_{\rm eq}=-3\omega^{2/3}/2$

5109: is the energy of the circular orbit in the rotating frame.

5110: 

5111: Since the non-dispersive wave-packets are localized in

5112: the immediate vicinity of the stable fixed point in the rotating frame, an expansion of the

5113: Hamiltonian around that position is useful. Precisely at the fixed point, all

5114: first order terms (in position and momentum) vanish. At second order,

5115: \begin{eqnarray}

5116: H_{\mathrm rot} \simeq H_{\mathrm harmonic}&&=E_{\rm eq} +

5117: \frac{{\vec{\tilde p}}^2}{2}

5118:  - \omega(\tilde{x}\tilde{p}_y-\tilde{y}\tilde{p}_x)

5119: + \frac{\omega^2 q \tilde{y}^2}{2}

5120: \nonumber\\ && - \omega^2 q \tilde{x}^2

5121: + \frac{\omega^2 q \tilde{z}^2}{2}\! ,

5122: \label{harmonic} 

5123: \end{eqnarray}

5124: where 

5125:  $(\tilde{x},\tilde{y},

5126: \tilde{z})=(x-x_{\rm eq},y-y_{\rm eq},z-z_{\rm eq})$ 

5127: (and accordingly for the

5128: momenta) denotes the displacement with respect

5129: to the fixed point. 

5130: Thus, in the harmonic approximation

5131: the motion in the $z$ direction

5132: decouples from that in the $x-y$ plane and is an oscillation with

5133: frequency $\omega\sqrt{q}.$ 

5134: The Hamiltonian for the latter,

5135: up to the additive constant $E_{\rm eq}$,  can be expressed in the

5136: standard form for a 2D, rotating anisotropic oscillator 

5137: \begin{equation}

5138: H= \frac{\tilde{p}_x^2+\tilde{p}_y^2}{2} +\frac{\omega^2

5139: (a\tilde{x}^2+b\tilde{y}^2)}{2}-\omega(\tilde{x}\tilde{p}_y-\tilde{y}\tilde{p}_x),

5140: \label{rotaibb}

5141: \end{equation}

5142: where the two parameters $a$ and $b$ are equal to $-2q$ and $q$,

5143: respectively. 

5144: This standard form has been studied in textbooks~\cite{marion}.

5145: It may be used to describe the stability of 

5146: the Lagrange equilibrium points in

5147: celestial mechanics (see \cite{ibb94,marion} and references therein).

5148: Because this Hamiltonian mixes position and momentum coordinates,

5149: it is not straightforward to determine the stability at the origin.

5150: The result is that there are two domains of stability:

5151: \begin{equation}

5152: a,b \geq 1

5153: \label{stability_region_1}

5154: \end{equation}

5155: and

5156: \begin{equation}

5157: -3 \leq a,b \leq 1,\ \ \ {\mathrm with}\ \ (a-b)^2+8(a+b) \geq 0

5158: \label{stability_region_2}

5159: \end{equation}

5160: 

5161: For the specific CP case, where $a=-2q$ and $b=q,$ only the second

5162: stability region is relevant, and the last inequality

5163: implies $8/9\leq q\leq 1$ for the fixed point to be stable.

5164: 

5165: Alternatively, one can ``diagonalize" the Hamiltonian~(\ref{rotaibb})

5166: and construct 

5167: its normal modes. The normal modes entangle position and momentum

5168:  operators due

5169: to the presence of crossed position-momentum terms in the

5170: Hamiltonian.

5171: Only along the $z$-mode (which is decoupled from the rest), the creation 

5172: and

5173: annihilation operators, $b_z^\dagger,b_z$ are

5174:  the standard combinations of $\tilde z$

5175: and $p_{\tilde z}$ operators. 

5176: In the $(x,y)$ plane, the creation and annihilation operators in the

5177: $\pm$ normal modes have complicated explicit formulae given in~\cite{delande98}. 

5178: After some algebra, one ends up with 

5179: the frequencies of the normal modes,

5180: \begin{eqnarray}

5181:  \omega_\pm &=& \omega \sqrt{\frac{2-q\pm Q}{2}},\label{omegas}\\

5182: \omega_z &=& \omega \sqrt{q},

5183: \label{ompmz}

5184: \end{eqnarray}

5185: where

5186: \begin{equation}

5187: Q=\sqrt{9q^2-8q}

5188: \label{defQ}

5189: \end{equation}

5190: and $q\geq 8/9$ for $Q$ to be real.

5191: In terms of  

5192: creation/annihilation operators, the harmonic Hamiltonian,

5193: eq.~(\ref{harmonic}),

5194: takes the form \cite{ibb94,delande95,kuba95a,delande98}

5195: \begin{equation}

5196: H_{\rm harmonic}

5197:  = E_{\rm eq} + \left(b^\dagger_+ b_+ +\frac{1}{2}\right)\omega_+ 

5198: - \left(b^\dagger_- b_-+\frac{1}{2}\right)\omega_- 

5199:  + \left(b^\dagger_z b_z+\frac{1}{2}\right)\omega_z .

5200: \label{hh}

5201: \end{equation}

5202: A minus sign appears in front of the $\omega_-$ term. This is because

5203: the fixed point is {\em not} a minimum of the Hamiltonian, although

5204: it is fully stable\footnote{In the first stability region, 

5205: eq.~(\ref{stability_region_1}),

5206: only + signs appear.}. 

5207: This is actually due to the momentum-position

5208: coupling, hence the Coriolis force. It is the same phenomenon

5209: which is responsible for the stability of the Trojan asteroids 

5210: \cite{ibb94,yeazell00,marion}

5211: and of an ion in a magnetic trap \cite{lee95} 

5212: (in the latter case, the position-momentum

5213: coupling is due to the magnetic field).

5214: 

5215: Finally, with $n_\pm, n_z$ counting

5216: the excitations in the corresponding modes, we obtain the

5217: harmonic prediction for the energies of the eigenstates in the vicinity of the fixed

5218: point: 

5219:  \begin{equation}

5220: E(n_+,n_-,n_z) = E_{\rm eq} + 

5221: \left(n_++\frac{1}{2}\right)\omega_+

5222:  - \left(n_-+\frac{1}{2}\right)\omega_-

5223: +  \left(n_z+\frac{1}{2}\right)\omega_z .

5224: \label{enharm}

5225: \end{equation}

5226: In particular, for $n_\pm=0,\ n_z=0$, we get a prediction for the ground 

5227: state of the oscillator, a Gaussian localized on top of the fixed point, i.e.,

5228: a Trojan wave-packet. In the following, we denote eigenstates in the 

5229: harmonic approximation

5230: as $|n_+,n_-,n_z\rangle$, thus the ground state non-dispersive wave-packet

5231: as $|0,0,0\rangle$. 

5232: In a 2D model, the $\omega_z$ term is dropped, the

5233: corresponding eigenstates are denoted $|n_+,n_-\rangle$ and have energies:

5234: \begin{equation}

5235: E^{2D}(n_+,n_-) = E_{\rm eq} + 

5236: \left(n_++\frac{1}{2}\right)\omega_+

5237:  - \left(n_-+\frac{1}{2}\right)\omega_-.

5238: \label{enharm2d}

5239: \end{equation}

5240: In fig.~\ref{cp2df}, we show the probability densities of the $|0,0\rangle$

5241: wave-packets obtained by exact numerical diagonalization of the  

5242: 2D Hamiltonian (\ref{hrot}), for various

5243: values of the microwave field amplitude.

5244: Clearly, for sufficiently strong microwave amplitudes,

5245: the wave-packets are well localized around

5246: the classical stable fixed point, with banana-like shapes.

5247: At very 

5248: weak fields, the stability of the

5249: fixed point gets 

5250: weaker and weaker; for a vanishing microwave field, all points on the

5251: circle with radius $\omega^{-2/3}$ are equivalent, and one has a ring of

5252: equilibrium points. Thus, when $F$ tends to zero,

5253: the non-dispersive wave-packet progressively extends along the angular direction

5254: (with the radial extension almost unchanged), ending with a doughnut shape

5255: at vanishing field. This means that, if $n_0$ is chosen

5256: as an integer,

5257: the non-dispersive wave-packet smoothly evolves into a circular

5258: state $|n_0,M=n_0\rangle$ as $F\rightarrow 0$. 

5259: The same is true for the 3D

5260: atom, where the non-dispersive wave-packet smoothly evolves into the

5261: circular state $|n_0,L=M=n_0-1\rangle.$

5262: \begin{figure}

5263: \centerline{\psfig{figure=bdzf27.ps,width=10cm}}

5264: \caption{Non-dispersive wave-packets of the two-dimensional hydrogen atom

5265: driven by a circularly polarized microwave field, for different values 

5266: of the microwave amplitude $F$.

5267: The microwave frequency is fixed at $\omega=1/(60.5)^3$,

5268: corresponding to a resonance island centered at

5269: $n_0=60.$ With increasing

5270: microwave amplitude, more states are coupled and the wave-packet becomes

5271: better localized. The scaled microwave amplitude $F_0=F\omega^{-4/3}$ 

5272: is $0.0003$, $0.0011$,

5273: $0.0111$, $0.0333$, $0.0444$, $0.0555$, from top left to bottom right.

5274: The nucleus is at the center of the figure, which extends over

5275: $\pm 5000$ Bohr radii in each direction. The microwave field is horizontal,

5276: pointing to the right.}

5277: \label{cp2df}

5278: \end{figure}

5279: 

5280: Provided the resonance island around the fixed point 

5281: is large enough,

5282: the harmonic approximation can also be used for studying properties

5283: of ``excited" states inside the resonance island.

5284: As an example, 

5285: fig.~\ref{accharm} shows the $|n_+=1,n_-=3\rangle$ 

5286: state calculated from the harmonic approximation, compared to

5287: the state obtained by exact numerical diagonalization of the 2D Floquet

5288: Hamiltonian. Obviously, the structure of the exact state is very similar to

5289: the one obtained from 

5290: its harmonic approximation. Because the creation and annihilation

5291: operators in the $\omega_{\pm}$ modes entangle position

5292: and momentum coordinates in a complicated way~\cite{delande98}, and 

5293: although the system

5294: is then completely integrable, the wave-function in the

5295: harmonic approximation is {\em not} separable in any coordinate system 

5296: (in contrast with the usual harmonic oscillator). Actually, the wave-function

5297: can be written as a product of Gaussians and Hermite polynomials of 

5298: the position

5299: coordinates, but the Hermite polynomials have to be evaluated for 

5300: complex values.

5301: This results in the unusual pattern of the probability density displayed in

5302: fig.~\ref{accharm}. An improvement over

5303: the harmonic approximation is possible,

5304: by bending the axis in the spirit of~\cite{kalinski95a}, in order to 

5305: account for the spherical symmetry of the dominant Coulomb potential. 

5306: With this

5307: improvement, the probability density, shown in the middle row of 

5308: fig.~\ref{accharm},

5309: is almost indistinguishable from the exact result. Let us repeat that

5310: this bending -- and consequently the deviation from Gaussian character of the

5311: wave-function -- does not affect at all the non-dispersive character of the 

5312: wave-packet.

5313: \begin{figure}

5314: \centerline{\psfig{figure=bdzf28.ps,width=10cm} }

5315: \caption{Comparison of exact Floquet eigenstates (bottom row) and of

5316: their harmonic

5317: approximations, for the ground state wave-packet $|n_+=0,n_-=0\rangle$ 

5318: (left column),

5319: and for the $|n_+=1,n_-=3\rangle$ excited state (right column), 

5320: for the 2D hydrogen atom

5321: driven by a circularly polarized microwave (field pointing to the

5322: right of the figure), 

5323: with amplitude $F_0=0.0333$ and resonant frequency

5324: corresponding to $n_0=60$. The nucleus is located at the center of each plot

5325: which extend over $\pm 5000$ Bohr radii. 

5326: The top row represents eigenfunctions in the harmonic

5327: approximation in $\tilde{x},\tilde{y}$ coordinates, 

5328: eq.~(\protect\ref{rotaibb}).

5329: The eigenfunctions in the middle row are obtained from the harmonic 

5330: approximation to eq.~(\protect\ref{hrot}) in polar coordinates. 

5331: They exhibit a clear bending of the electronic density along the circular 

5332: trajectory.

5333: The excited

5334: wave-packet $|1,3\rangle$ appears in 

5335: fig.~\protect{\ref{solit2}} as a straight 

5336: line (modulo small avoided crossings) 

5337: with a negative slope, meeting the state $|0,0\rangle$ in a broad avoided

5338: crossing, around $F_0\simeq 0.036$.

5339: }

5340: \label{accharm}

5341: \end{figure}

5342: 

5343: Let us now turn to the realistic 3D model of the atom.

5344: Fig.~\ref{circwp3d} shows an isovalue contour of several 

5345: non-dispersive wave-packets

5346: for the hydrogen atom driven by a microwave

5347: field with frequency

5348: $\omega=1/60^3$, i.e., roughly resonant with 

5349: the $n_0=60\to 59,61$ transitions 

5350: (see eqs.~(\ref{cprinc},\ref{resonance_condition},\ref{omk})). 

5351: The 

5352: best localized wave-packet is the ground state $|0,0,0\rangle$, while

5353: the three other states are excited by one quantum in either of 

5354: the normal modes $\omega_{\pm,z}$, 

5355: and are therefore significantly more extended in space.

5356: \begin{figure}

5357: \centerline{\psfig{figure=bdzf29.ps,width=15cm}}

5358: \caption{Isovalue plots (at 30\% of the maximum value) 

5359: of non-dispersive wave-packets 

5360: in the three-dimensional

5361: hydrogen atom driven by a circularly polarized microwave field. 

5362: Frequency of

5363: the driving $\omega=1/60^3$, amplitude 

5364: $F_0=0.04442$. 

5365: In the laboratory frame, the wave-packets 

5366: propagate -- without changing their shapes -- along a circular

5367: trajectory 

5368: centered around the nucleus indicated by a cross.  

5369: The cube edges measure 10000 Bohr radii. The microwave polarization plane

5370: is horizontal with the field pointing to the right. The

5371: four wave-packets shown represent the ground state wave-packet $|0,0,0 

5372: \rangle$ (top left), and the excited states $|1,0,0\rangle$ (bottom left),

5373: $|0,1,0\rangle$ (top right), and $|0,0,1\rangle$ (bottom right).

5374: Eventually, the microwave field will ionize such states, but their

5375: lifetimes are extremely long, of the order of thousands to millions

5376: of Kepler periods. 

5377: }

5378: \label{circwp3d}

5379: \end{figure}

5380: Again, as already mentioned in sections~\ref{LIN1D} and \ref{LIN3D}, these

5381: wave-packet eigenstates have finite, but extremely long life-times (several

5382: thousands to millions of Kepler orbits), due to the

5383: field induced ionization. For a detailed discussion of their decay

5384: properties see section~\ref{ION}.

5385: 

5386: As already demonstrated in 

5387: the LP case (see figs.~\ref{lin3d_3},\ref{lin3d_4}), the semiclassical

5388: prediction for the energies of the non-dispersive wave-packets

5389: is usually excellent. In order to stress the (small) differences,

5390: we plot in fig.~\ref{solit2} a part of the Floquet spectrum

5391: of the two-dimensional model 

5392: atom,

5393: i.e. quasi-energy levels versus the (scaled) microwave amplitude,

5394: after substraction of the prediction of the

5395: harmonic approximation around the stable fixed point, eq.~(\ref{enharm}), 

5396: for the 

5397: ground state wave-packet $|0,0\rangle$.

5398: The result is shown in units of the mean level spacing,

5399: estimated\footnote{This estimate follows from the local energy splitting, 

5400: $\sim n_0^{-3}$, divided by the number $n_0/2$ of photons needed to ionize the initial 

5401: atomic state by a resonant driving field.} to be roughly $2/n_0^4$.

5402: If the harmonic approximation was exact,

5403: the ground state wave-packet would be represented by 

5404: a horizontal line at zero.

5405: The actual result is not very far from that, which proves that the

5406: semiclassical method predicts the correct energy with an accuracy

5407: mostly better than the mean level spacing. The other states of the system

5408: appear as 

5409: energy levels which rapidly evolve with $F_0$, and which  

5410: exhibit extremely small

5411: avoided crossings -- hence extremely small couplings -- with the wave-packet.

5412: In the vicinity of such avoided

5413: crossings, the energy levels are perturbed, 

5414: the diabatic wave-functions mix (the wave-packet eigenstates 

5415: get distorted),

5416: and, typically, the lifetime of the

5417: state decreases (induced by the coupling to the closest

5418: Floquet state \cite{abuth,abu95a}, typically much less resistant against ionization, as we shall

5419: discuss in detail in sec.~\ref{ION}).

5420: 

5421: Thus, strictly

5422: speaking, when we speak of non-dispersive wave-packets as specific

5423: Floquet states, we really have in mind a generic situation, {\em far} from

5424: any avoided crossing. In particular, the examples of wave-packet states

5425: shown in the figures above correspond to such situations.

5426: \begin{figure}

5427: \centerline{\psfig{figure=bdzf30.ps,width=12cm}}

5428: \caption{Spectrum of the two-dimensional hydrogen atom in

5429: a circularly polarized microwave field 

5430: of frequency $\omega=1/(60.5)^{3}$, 

5431: as a function of the

5432: scaled microwave amplitude $F_0=F\omega^{-4/3}$. 

5433: In order to test the accuracy of the harmonic

5434: prediction, 

5435: we substract the semiclassical energy for the ground

5436: state $|0,0\rangle$ wave-packet, eq.~(\protect{\ref{enharm}}), from the result

5437: of the exact numerical diagonalization, and rescale the energy axis in units 

5438: of the mean level spacing. The

5439: almost horizontal line slightly above zero corresponds to the

5440: non-dispersive ground state wave-packet, which typically undergoes small

5441: avoided crossings with other Floquet states. A relatively large 

5442: avoided crossing

5443: occurs when two wave-packet-like states meet, as here 

5444: happens around $F_0\simeq 0.036$.

5445: The other (``colliding'') 

5446: state is the excited wave-packet $|1,3\rangle$. The dashed

5447: line indicates the harmonic prediction, eq.~(\ref{enharm}), for this state, 

5448: which is obviously

5449: less accurate. Note, however, that the slope is correctly predicted.

5450: }

5451: \label{solit2}

5452: \end{figure}

5453: The observed accuracy of the semiclassical approximation has 

5454: important practical consequences:

5455: in order to obtain the ``exact'' wave-packets numerically, 

5456: we do not need many

5457: eigenvalues for a given set of parameters. Using the

5458: Lanczos algorithm for the partial diagonalization of a matrix, it is enough to extract

5459: few (say five) eigenvalues only, 

5460: centered on 

5461: the semiclassical prediction. The accuracy of the latter

5462: (a fraction of the mean level spacing) is

5463: sufficient for a clear identification of the appropriate

5464: quantum eigenvalue.

5465:  Actually, in a real diagonalization of the Floquet Hamiltonian, we

5466: identify the wave-packet states by both their vicinity 

5467: to the semiclassical prediction

5468: for the energy, 

5469: and the large (modulus of the) slope of the

5470: level w.r.t. changes of $F_0$, induced by its large dipole

5471: moment in the rotating frame (by virtue of the Hellman-Feynman 

5472: theorem \cite{cct92}).

5473: The latter 

5474: criterion is actually also very useful for the identification of excited

5475: wave-packets in a numerically exact spectrum.

5476: The state $|1,3\rangle$ (in the harmonic approximation)

5477: presented in 

5478: fig.~\ref{accharm} is precisely the excited

5479: wave-packet which appears in fig.~\protect{\ref{solit2}} as a ``line''

5480: with a negative slope, meeting the $|0,0\rangle$ state in a broad avoided

5481: crossing around $F_0\simeq 0.036$. 

5482: From that figure, it is

5483: apparent that  the harmonic

5484: prediction for the energy is not excellent for the  $|1,3\rangle$ state.

5485: On the other hand, the slope of the Floquet

5486: state almost matches the slope given by the harmonic

5487: approximation, which confirms that the exact wave-function is still 

5488: well approximated by its harmonic counterpart.

5489: Note that also from the experimental point of view it is important to get accurate

5490: and simple semiclassical estimates of the energies of the non-dispersive

5491: wave-packets, since it may help in their preparation and unambiguous

5492: identification. For a more detailed discussion, see section~\ref{SPEC}.

5493: 

5494: It is interesting to compare the accuracy of 

5495: the harmonic approximation

5496: to the pendulum 

5497: description outlined previously in sec.~\ref{reso_an}. The latter

5498: results from lowest order perturbation theory in $F$.

5499: Taking the small $F$ limit we get $\omega_+,\omega_z

5500: \rightarrow

5501: \omega$, and $\omega_-\rightarrow \omega\sqrt{3F_0}$, for the 

5502: harmonic modes, see 

5503: eqs.~(\ref{omegas}-\ref{defQ}). 

5504: The latter result coincides -- as it should -- with the pendulum prediction,

5505: eqs.~(\ref{omega_harmonic},\ref{omega_harmonic_cp}). Similarly, the energy of the stable

5506: equilibrium point, eq.~(\ref{wps}), becomes at first order in $F$:

5507: \begin{equation}

5508: E_{\rm eq}= - \frac{3}{2n_0^2} + n_0^2\ F + O(F^2)

5509: \end{equation}

5510: which coincides with

5511: the energy of the center of the resonance island, see section~\ref{reso_an}.

5512: Thus, the prediction of the resonance analysis agrees with the harmonic

5513: approximation in the rotating frame. 

5514: For a more accurate estimate of the validity of both approaches, we have

5515: calculated -- for the 2D model of the atom, but similar

5516: conclusions are reached in 3D -- 

5517: the energy difference between the exact quantum result  

5518: and the prediction

5519: using a semiclassical

5520: quantization of the secular motion in the $(M,\phi)$ plane together with

5521: the Mathieu method in the $(\hat{I},\hat\theta )$ plane on the one side,

5522: and the prediction of the harmonic approximation around the fixed point,

5523: eq.~(\ref{enharm}), on the other side. 

5524: In fig.~\ref{semicirc}, we compare the results coming from both approaches.

5525: As expected, the semiclassical approach

5526: based on the Mathieu equation is clearly superior for very small microwave

5527: amplitudes, as it is ``exact" at first order in $F$. On the other hand,

5528: for the harmonic approximation to work well, the 

5529: island around the fixed point has to be sufficiently large.

5530: Since the size of the island increases as $\sqrt{F},$ the harmonic 

5531: approximation

5532: may become valid only for sufficiently large microwave amplitudes, when 

5533: there is at least

5534: one state trapped in the island. As seen, however,

5535: in fig.~\ref{semicirc}, the harmonic

5536: approximation yields a satisfactory prediction for the wave-packet energy

5537: (within few \% of the mean spacing) almost everywhere. 

5538: For increasing $n_0,$ the harmonic approximation is better and better

5539: and the Mathieu approach is superior only over a smaller and smaller range

5540: of $F_0=Fn_0^4$, 

5541: close to 0. Still, both

5542: approaches give very good predictions for the typical values of $F_0$ used in

5543: the following, say $F_0\simeq 0.03$. The spikes visible in the figure are

5544: due to the many small avoided crossings visible in fig.~\ref{solit2}.

5545: 

5546: \begin{figure}

5547: \centerline{\psfig{figure=bdzf31.eps,width=14cm,bbllx=0pt,bblly=50pt,bburx=600pt,bbury=400pt}}

5548: \caption{

5549: Difference between  the exact quantum energy of a non-dispersive

5550: wave-packet of a 2D hydrogen atom in a 

5551: circularly polarized microwave

5552: field, and two different semiclassical predictions, as a function of 

5553: the scaled microwave field amplitude, $F_0$.

5554: The thick lines are obtained from 

5555: a harmonic approximation 

5556: of the motion around

5557: the stable equilibrium point (in the rotating frame), eq.~(\ref{enharm2d}) 

5558: and the 

5559: thin lines use a quantum treatment (Mathieu approach, 

5560: sec.~\protect\ref{section_mathieu})

5561: of the motion in the 

5562: resonance island,

5563: combined with a semiclassical treatment of the secular motion. 

5564: From top to bottom

5565: $\omega=1/(30.5)^3,\ 1/(60.5)^3, \ 1/(90.5)^3$, corresponding to wave-packets

5566: associated with Rydberg states of principal quantum number

5567: $n_0=30,60,90$. The energy difference is expressed in units of the mean

5568: level spacing, estimated by $n_0^4/2$.

5569: The prediction of the Mathieu approach is

5570: consistently better for low $F_0$, and the harmonic approximation

5571:  becomes clearly superior for larger $F_0.$ For larger and larger $n_0,$

5572:  the harmonic approximation is better and better. Note that both

5573:  approximations make it possible to estimate the energy of the non-dispersive

5574:  wave-packet with an accuracy better than the mean level spacing, allowing

5575:  for its simple and unambiguous 

5576: extraction from exact numerical data.}

5577: \label{semicirc}

5578: \end{figure}

5579: 

5580: While we have shown some exemplary  wave-packets for few values of

5581:  $n_0$ and $F$

5582: only, they generally look very similar provided that

5583: \begin{itemize}

5584: \item $n_0$ is sufficiently large, say $n_0>30$. For smaller $n_0$,

5585: the wave-packet looks a bit distorted and one observes some

5586: deviations from the harmonic approximation (for a more detailed 

5587: discussion of this

5588: point see secs.~\ref{ION} and \ref{SPO}, where 

5589: ionization and spontaneous emission

5590: of the wave-packets are discussed);

5591: \item $F_0$ is sufficiently large, say $F_0> 0.001$, 

5592: such that

5593: the resonance island can support at least one state. For smaller

5594: $F_0,$ the wave-packet becomes more extended in the angular coordinate, 

5595: since less

5596:  atomic circular states are significantly coupled, see fig.~\ref{cp2df};

5597: \item $F_0$ is not too large, say smaller than $F_0\simeq0.065$. 

5598: Our numerical data suggest that the upper limit is not given

5599: by the limiting value of $q=8/9$,

5600: for which the fixed point is still stable. The limiting

5601: value appears to be rather linked to the  

5602: $1:2$ resonance between the $\omega_+$ and 

5603: $\omega_-$ modes, which occurs approx. at $F_0\simeq0.065$.

5604: \item In particular, the value $q=0.9562$ (i.e. $F_0\simeq 0.04442$),

5605: corresponding to 

5606: optimal 

5607: classical stability of the fixed point, advertised in~\cite{ibb94} as the

5608: optimal one, is by no means favored. A much broader range of microwave

5609: amplitudes is available (and equivalent as far as the ``quality'' of the

5610: wave-packet is concerned). What is much more relevant, is the presence of

5611: some accidental avoided crossings with other Floquet states.

5612: \end{itemize}

5613: Still, these are no very restrictive conditions, and we are left with 

5614: a broad range

5615: of parameters favoring the existence of nondispersive wave-packets, 

5616: a range which

5617: is experimentally fully accessible (see section~\ref{EXP} for a more

5618: elaborate discussion of experimental aspects). 

5619: \begin{figure}

5620: \centerline{\psfig{figure=bdzf32.ps,width=10cm}}

5621: \caption{Floquet eigenstate of the two-dimensional hydrogen atom in a

5622: circularly polarized microwave field. This state is 

5623: partially localized on the unstable

5624: equilibrium point (in the rotating frame), 

5625: see eqs.~(\ref{wp0})-(\ref{wpu}), for $F_0=0.057$ and $n_0=60$.

5626: This localization is of purely classical origin. The nucleus is at the center

5627: of the figure which extends over $\pm 5000$ Bohr radii. The microwave

5628: field points to the right.

5629: }

5630: \label{cp2unstb}

5631: \end{figure}

5632: 

5633: Finally, in analogy with the LP case, we may consider 

5634: Floquet states localized on the unstable fixed point associated 

5635: with the principal resonance island. 

5636: From

5637: the discussion following eq.~(\ref{wpu}), this point is located 

5638: opposite to the stable fixed point, on the other side of the nucleus. 

5639: An example of

5640: such a state is shown in fig.~\ref{cp2unstb}, for an amplitude of the microwave

5641: field that ensures that most of the nearby Floquet states ionize rather 

5642: rapidly.

5643: The eigenstate displayed in the figure  

5644: lives much longer (several thousands of Kepler periods). The

5645: localization in the vicinity of an unstable fixed point, in analogy

5646: to the LP case discussed previously, is of purely classical origin.

5647: As pointed out in \cite{delande97},

5648: such a localization must not be confused with scarring \cite{heller84} --

5649: a partial localization on an unstable periodic orbit embedded in a chaotic

5650: sea -- which disappears in the semiclassical limit \cite{heller84,bogomolny88a}.

5651: 

5652: \subsection{Rydberg states in elliptically polarized microwave fields}

5653: \label{EP}

5654: 

5655: The origin of non-dispersive wave-packets being their localization inside

5656: the resonance island (locking the frequency of the electronic motion

5657: onto the external drive) suggests that

5658:  such wave-packets are quite robust and 

5659: should  exist not only for CP and LP, but

5660: also for arbitrary elliptical polarization (EP). 

5661: 

5662: The possible existence

5663: of nondispersive wave-packets for EP  was mentioned 

5664: in \cite{lee97,Bth}, using the classical ``pulsating SOS'' approach. The

5665: method, however, 

5666: did not allow for quantitative predictions, and was restricted to

5667: elliptic polarisations very close to the CP case.

5668: However, the robustness of such wave-packets for arbitrary 

5669: EP is obvious 

5670: once the localization mechanism inside the resonance island 

5671: is well understood \cite{sacha98b,sachath}.

5672: 

5673: Let us  consider an elliptically polarized driving field of constant

5674: amplitude. With the ellipticity parameter $\alpha\in[0;1]$,

5675: \begin{equation}

5676: V=F(x\cos\omega t + \alpha y\sin\omega t)

5677: \label{v_ep}

5678: \end{equation}

5679: establishes a continuous transition between linear ($\alpha =0$) and circular 

5680: ($\alpha =1$) polarization treated in the two preceding 

5681: chapters\footnote{Note, however, that $\alpha =0$ defines a linearly polarized field 

5682: along the $x$-axis, i.e., {\em in} the plane of elliptical polarization for 

5683: $\alpha >0$. In sec.~\ref{LIN}, the polarization vector was chosen along

5684: the $z$-axis. The physics is of course the same, but the algebraic expressions

5685: are slightly different, requiring a rotation by an angle $\beta=\pi/2$

5686: around the $y$-axis.}.

5687: This general case is slightly more complicated than both  

5688: limiting cases LP and CP. 

5689: For LP microwaves (see section \ref{LIN3D}), the conservation of

5690: the angular momentum projection onto the polarization axis, $M$, makes

5691: the dynamics effectively two-dimensional. For the CP case,

5692: the transformation (\ref{hrot}) 

5693: to the frame rotating

5694: with the microwave frequency removes the explicit 

5695: time-dependence (see section \ref{CP}).

5696: None of these simplifications is

5697: possible in the general EP case, and the problem

5698: is truly three dimensional {\em and} time-dependent.

5699: 

5700: To illustrate the transition from LP to CP via EP, the 

5701: two-dimensional model of the atom is sufficient, and we shall restrain 

5702: our subsequent treatment to this computationally less involved case.

5703: The

5704: classical resonance analysis for EP microwave ionization has been

5705: described in detail in~\cite{sacha97,sacha98c}. It follows closely the lines described in detail

5706: in sections~\ref{LIN} and \ref{CP}. By expanding the

5707: perturbation, eq.~(\ref{v_ep}), in the action-angle coordinates $(I,M,\theta,\phi)$

5708: of the two-dimensional atom (see sec.~\ref{action-angle_hydrogen}), 

5709: one obtains the following secular Hamiltonian:

5710: 

5711: \begin{equation}

5712: {\cal H}_{\rm sec}={\hat P_t}-\frac{1}{2{\hat I}^2}-\omega {\hat I}+

5713: F\left[{\cal V}_1(\hat{I},M,\phi;\alpha)\cos\hat\theta-

5714: {\cal U}_1(\hat{I},M,\phi;\alpha)\sin\hat\theta\right],

5715: \label{EPr_0}

5716: \end{equation}

5717: with

5718: \begin{eqnarray}

5719:  {\cal V}_{1}(\hat{I},M,\phi;\alpha) &=& \cos\phi \left(X_1 + \alpha Y_1\right),\nonumber \\

5720:  {\cal U}_{1}(\hat{I},M,\phi;\alpha) &=& \sin\phi \left(Y_1 + \alpha X_1\right).

5721: \label{calUV}

5722: \end{eqnarray}

5723: 

5724: This can be finally rewritten as:

5725: \begin{equation}

5726: {\cal H}_{\rm sec}={\hat P_t}-\frac{1}{2{\hat I}^2}-\omega {\hat I}+

5727: F\chi_1(\hat{I},M,\phi;\alpha)\cos(\hat\theta+\delta_1).

5728: \label{EPr}

5729: \end{equation}

5730: Both, $\chi_1$ and $\delta_1$,

5731: depend on the shape and orientation of the   

5732: electronic elliptical trajectory,

5733: as well as on 

5734: $\alpha$,

5735: and are given by:

5736: \begin{equation}

5737: \chi_1(\hat{I},M,\phi;\alpha)= \sqrt{{\cal V}_1^2 + {\cal U}_1^2}, 

5738: \end{equation}

5739: and

5740: \begin{equation}

5741: \tan \delta_1(M,\phi;\alpha) = \frac{{\cal U}_1}{{\cal V}_1},

5742: \end{equation}

5743: which is once more the familiar form of a system with a resonance island 

5744: in the $(\hat{I},\hat{\theta})$ plane. The expressions obtained are in fact

5745: very similar to the ones we obtained for 

5746: the three-dimensional atom exposed to a

5747: circularly polarized microwave field, eqs.~(\ref{hamsccir2}-

5748: \ref{phasecirc}), in section \ref{reso_an}. This is actually not surprising:

5749: the relevant parameter for the transverse dynamics is the magnitude

5750: of the atomic dipole oscillating with the driving field, i.e. the

5751: scalar product of the oscillating atomic dipole with the polarization vector.

5752: The latter can be seen either as the projection of the 

5753: oscillating atomic dipole 

5754: onto the polarization plane or as the projection of the polarization vector

5755: on the plane of the atomic trajectory. If one considers a three-dimensional 

5756: hydrogen

5757: atom in a circularly polarized field, the projection of the polarization 

5758: vector

5759: onto the plane of the atomic trajectory is elliptically polarized

5760: with ellipticity $\alpha=L/M.$ This is another method to rediscover 

5761: the Hamiltonian~(\ref{EPr}) from Hamiltonian (\ref{hamsccir2}).

5762: \begin{figure}

5763: \centerline{\psfig{figure=bdzf33.ps,width=12cm}}

5764: \caption{The scaled effective perturbation $\chi_1/\hat{I}^2$

5765: driving the transverse/angular

5766: motion of a two-dimensional hydrogen atom exposed to a resonant,

5767: elliptically polarized microwave field, 

5768: plotted as a function of

5769: the scaled angular momentum, $M_0=M/\hat{I}$, and of the angle 

5770: $\phi$ between the

5771: Runge-Lenz vector and the major axis of the polarization ellipse.

5772: Left and right panels correspond to

5773: $\alpha=0.1$ and $\alpha=0.6$, respectively. Non-dispersive

5774: wave-packets are localized around the maxima of this effective potential,

5775: at $M_0=\pm 1,$ and are circularly co- and contra-rotating (with

5776: respect to the microwave field) around the nucleus 

5777: (see fig.~\protect\ref{ep2}).

5778: }

5779: \label{ep1}

5780: \end{figure}

5781: 

5782: 

5783: To obtain a semiclassical estimation of the energies of the nondispersive 

5784: wave-packets 

5785:  we proceed 

5786: precisely in the same way as for LP and CP. Since the radial motion in 

5787: $(\hat{I},\hat{\theta})$ is much faster than in the transverse/angular degree 

5788: of freedom defined by $(M,\phi)$, we first quantize the effective perturbation

5789: $\chi_1(\hat I,M,\phi;\alpha)$ driving the angular motion. 

5790: Fig.~\ref{ep1} shows 

5791: $\chi_1/\hat{I}^2$, as a function of $M_0=M/\hat{I}$ and 

5792: $\phi$, for two different values of the driving field ellipticity $\alpha$. 

5793: Note that $\chi_1$ becomes more symmetric as $\alpha\rightarrow 0$, since 

5794: this limit defines the LP case, where the dynamics cannot depend on the 

5795: rotational sense of the electronic motion around the nucleus. 

5796: The four extrema 

5797: of $\chi_1$ define the possible wave-packet eigenstates. Whereas the minima

5798: at $\phi=\pi/2,3\pi/2$ correspond to elliptic orbits of intermediate 

5799: eccentricity $0<e<1$ perpendicular to the driving field major axis, the 

5800: ($\phi$-independent)

5801: maxima at $M_0=\pm 1$ define circular orbits which co- or

5802: contra-rotate 

5803: with the driving field. 

5804: In the limit $\alpha\rightarrow 0$, the $M_0=\pm1$ maxima are associated with

5805: the same value of $\chi_1$. 

5806: Hence, the

5807: actual Floquet eigenstates appear as tunneling doublets in the 

5808: Floquet spectrum, 

5809: each member of the doublet being a superposition of 

5810: the co- and 

5811: contra-rotating wave-packets.

5812:  

5813: The quantization of the fast motion in the $(\hat{I},\hat{\theta})$ plane 

5814: is similar to the one already performed in the LP and CP cases.

5815: As we already observed (see fig.~\ref{lin3d_2}), the size 

5816: of the resonance island in the $(\hat{I},\hat{\theta})$ plane is 

5817: proportional to 

5818: $\sqrt{\chi_1}$. Correspondingly, also the localization properties of the 

5819: wave-packet along the classical trajectory improve with increasing $\chi_1$.

5820: We therefore conclude from fig.~\ref{ep1} that the eigenstates corresponding 

5821: to the minima of $\chi_1$ cannot be expected to exhibit strong 

5822: longitudinal localization,

5823: whereas the eigenstates localized along the circular orbits at the maxima of 

5824: $\chi_1$ can.

5825: 

5826: \begin{figure}

5827: \centerline{\psfig{figure=bdzf34.ps,width=10cm}}

5828: \caption{Energy levels of the two dimensional hydrogen atom driven by 

5829: a resonant,

5830: elliptically polarized microwave field of scaled amplitude $F_0=0.03$,

5831: for $n_0=21$, as a function of the 

5832: field ellipticity $\alpha$

5833: (the resonant frequency of the microwave is $\omega=1/(21.5)^3)$. 

5834: Full lines: semiclassical prediction; dotted lines: exact numerical result

5835: for the states originating from the $n_0=21$ hydrogenic manifold. The 

5836: non-dispersive wave-packets are the states originating from the upper doublet

5837: at $\alpha=0.$ The ascending (resp. descending) energy level is associated

5838: with the wave-packet co- (resp. contra-) rotating with the microwave field.

5839: }

5840: \label{ep3}

5841: \end{figure}

5842: 

5843: Fig.~\ref{ep3} compares the semiclassical prediction obtained by quantization

5844: of $\chi_1$ and ${\mathcal H}_{\rm sec}$ 

5845: (following the lines already described in section~\ref{LIN3D},

5846: for the ground state $N=0$ in the resonance island) 

5847: to the exact quasienergies (determined by numerical diagonalization of the

5848: Floquet Hamiltonian), for

5849: $\alpha$ varying from LP to CP.

5850: The agreement is excellent, with slightly 

5851: larger discrepancies between the semiclassical and the exact results for the

5852: states with smallest energy. For those states the resonance island is 

5853: very small

5854: (small $\chi_1$), what explains the discrepancy.

5855: \begin{figure}

5856: \centerline{\psfig{figure=bdzf35.ps,width=12cm}}

5857: \caption{Non-dispersive wave-packets of the two dimensional hydrogen 

5858: atom exposed to an elliptically polarized, resonant microwave field. 

5859: Scaled microwave amplitude $F_0=0.03$

5860: (for resonant 

5861: principal quantum number  $n_0=21$), 

5862: and ellipticity

5863: $\alpha=0.4$. Top row: non-dispersive wave-packet moving on a 

5864: circular orbit corotating with 

5865: the microwave field, for phases $\omega t=0,\pi/4,\pi/2$ 

5866: (from left

5867: to right). This wave-packet evolves into the eigenstate represented in 

5868: fig.~\protect\ref{cp2df}, under continuous increase of the ellipticity 

5869: to $\alpha=1$.

5870: Bottom row: 

5871: non-dispersive wave-packet launched along the same circular orbit, but 

5872: contra-rotating with  the driving field 

5873: (for the 

5874: same phases).

5875: Note that, while the co-rotating wave-packet almost preserves its shape 

5876: during the 

5877: temporal evolution,

5878: the contra-rotating one exhibits significant distortions during one 

5879: field cycle, as a direct

5880: consequence of its complicated level dynamics shown in 

5881: fig.~\protect\ref{ep3}. 

5882: Still being an exactly time-periodic Floquet

5883: eigenstate, it regains its shape after

5884: every period of the microwave.

5885: The size of each box extends over 

5886: $\pm 800$ Bohr radii, in both $x$ and $y$ directions,

5887: with the nucleus in the middle. The major axis of the polarization

5888: ellipse is along the horizontal $x$ axis and the microwave field points

5889: to the right at $t=0.$  

5890: }

5891: \label{ep2}

5892: \end{figure}

5893: 

5894: The highest lying state in fig.~\ref{ep3}, ascending with $\alpha$,

5895: is a non-dispersive wave-packet state located on the circular orbit and

5896: corotating with the EP field. It is shown in fig.~\ref{ep2} for $\alpha=0.4$.

5897: As mentioned above, the corresponding counterrotating wave-packet 

5898: is energetically

5899: degenerate with the co-rotating one for $\alpha=0$. Its energy decreases with

5900: $\alpha$ (compare fig.~\ref{ep3}). It is shown in the bottom row in

5901: fig.~\ref{ep2} for $\alpha=0.4$. While the corotating wave-packet

5902: preserves its shape for all $\alpha$ values (except at isolated avoided

5903: crossings) the counter-rotating wave-packet undergoes a series of strong

5904: avoided crossings for $\alpha >0.42$, progressively loosing its localized

5905: character. This is related to a strong decrease of the maximum 

5906: of $\chi_1$ at $M_0=-1$ with $\alpha$, clearly visible in fig.~\ref{ep1}.

5907: 

5908: 

5909: While we have discussed the 2-dimensional case only, 

5910: the CP situation (compare section \ref{CP}) indicates that for

5911: sufficiently large $\alpha$, the important resonant motion occurs in the

5912: polarization plane, being stable versus small deviations in the

5913: $z$ direction.

5914: Thus the calculations presented above are also

5915: relevant for the real three dimensional world, provided $\alpha$ is not far from 

5916: unity \cite{sachath,sacha98b}. For arbitrary $\alpha$, a full 3D

5917: analysis is required. While this is clearly more involved, the general scenario

5918: of a wave-packet anchored to a resonance island will certainly prevail.

5919: 

5920: \section{Manipulating the wave-packets}

5921: \label{MA}

5922: 

5923: We have shown in the previous sections that non-dispersive wave-packets 

5924: are genuine 

5925: solutions of the Floquet eigenvalue problem, eq.~(\ref{calhq}), under resonant 

5926: driving, for arbitrary polarization of the driving field. The semiclassical 

5927: approximation used to guide our exact numerical approach directly demonstrates 

5928: the localization of the electronic density in well defined regions of phase space,

5929: which protect the atom against ionization induced by the 

5930: external field (see, however, sec.~\ref{ION}). We have also seen that classical

5931: phase space does not only undergo structural changes under changes of the 

5932: driving field amplitude (figs.~\ref{lin1d_00}, \ref{lin1d_1},\ref{lin1d_4a}), 

5933: but also under changes of the driving field ellipticity (fig.~\ref{ep1}).

5934: Therefore, the creation of non-dispersive wave-packets can be conceived as an

5935: easy and efficient means of quantum control, which allows the 

5936: manipulation and the controlled transfer of quantum population accross phase space.

5937: In particular, one may imagine the creation of a wave-packet moving along 

5938: the polarization axis of a linearly

5939: polarized microwave field. A subsequent,

5940: smooth change through elliptical to finally

5941: circular polarization allows to transfer the electron to a circular

5942: orbit.

5943: 

5944: Adding additional static fields to the Hamiltonian (\ref{calhq}) provides us 

5945: with yet another handle to control the orientation and shape of highly excited 

5946: Rydberg trajectories, and, hence, to manipulate the localization properties 

5947: of nondispersive wave-packet eigenstates in configuration and phase space.

5948: The key point is that trapping inside the nonlinear resonance

5949: island is a robust mechanism which protects the

5950: non-dispersive wave-packet very efficiently from imperfections. This allows

5951: to adjust the wave-packet's properties at will, just by adiabatically

5952: changing the properties of the island itself. Moreover, when the strength of 

5953: the external

5954: perturbation increases, chaos generically invades a large part of 

5955: classical phase space, but the resonance islands most often survives. 

5956: The reason is that the phase locking phenomenon

5957: introduces various time scales in the system, which have different

5958: orders of magnitude.

5959: That makes the system quasi-integrable (for example

5960: through some adiabatic approximation {\em \`a la Born-Oppenheimer}) and

5961: -- locally -- more resistant to chaos. 

5962: 

5963: Hereafter, we discuss two possible alternatives of manipulating the 

5964: wave-packets.

5965: One

5966: is realized by adding a static

5967: electric field to the 

5968: LP microwave drive \cite{sacha98a}.

5969: Alternatively, the addition of 

5970: a static magnetic field to CP driving enhances the region of classical 

5971: stability, and extends the range of applicability of the 

5972: harmonic approximation \cite{lee97,farrelly95a,farrelly95,brunello96,cerjan97,lee95,brunello97}. 

5973: 

5974: \subsection{Rydberg states in linearly polarized microwave and static electric

5975: fields}

5976: \label{staticmw}

5977: \label{LINF}

5978: 

5979: Let us first consider a Rydberg electron driven by a resonant, linearly

5980: polarized microwave, in the presence of a static electric field.

5981: We already realized 

5982: (see the discussion in sec.~\ref{LIN3D}) that 

5983: the classical 3D motion of the driven Rydberg electron is angularly unstable

5984: in a LP microwave field. It turns out, however, 

5985: that a stabilization

5986: of the angular motion is possible by the addition of a

5987: static

5988: electric field $F_s$ parallel to the microwave polarization axis

5989: \cite{sacha98a,leopold86,leopold87}.

5990: The corresponding Hamiltonian reads:

5991: \begin{equation}

5992: H=\frac{p_x^2+p_y^2+p_z^2}{2}-\frac{1}{r}+Fz\cos\omega t +F_s z,

5993: \label{hcart}

5994: \end{equation}

5995: which we examine 

5996: in the vicinity of the $s=1$ resonance. As in sec.~\ref{LIN3D}, the 

5997: angular momentum projection $M$ on the $z$ axis remains a constant of 

5998: motion, and we shall assume $M=0$ in the following.

5999: Compared to the situation of a pure microwave field, there is an

6000: additional time scale, directly related to the static field.

6001: Indeed, in the presence of a perturbative static field alone, it is known 

6002: that the Coulomb degeneracy of the hydrogenic energy levels (in $L$)

6003: is lifted. The resulting eigenstates are combinations of the $n_0$ substates

6004: of the $n_0$ manifold (for $0 leq L \leq n_0-1$)\footnote{These states 

6005: are called ``parabolic" states, since the

6006: eigenfunctions are separable in parabolic coordinates \protect\cite{landau2}.}.

6007: The associated energy 

6008: levels are equally spaced by a quantity proportional to $F_s$ 

6009: ($3n_0F_s$ in atomic units). Classically, the trajectories are no longer

6010: closed but rather Kepler ellipses which slowly librate around the static 

6011: field axis,

6012: periodically changing their shapes, at a (small) frequency  

6013: $3n_0F_s\ll \omega_{\rm Kepler}=1/n_0^3.$

6014: Thus, the new time scale associated with the static electric field

6015: is of the order of $1/F_{s,0}$ Kepler periods, where

6016: \begin{equation}

6017: F_{s,0} = F_s n_0^4

6018: \end{equation}

6019: is the scaled static field. This is to be compared to the time scales $1/F_0$

6020: and $1/\sqrt{F_0}$, which characterize the secular time evolution in the

6021: $(L,\psi)$ and $(\hat{I},\hat{\theta})$ coordinates, respectively 

6022: (see discussion in section~\ref{LIN3D}), in the

6023: presence of the microwave field alone. To achieve confinement of the

6024: electronic trajectory in the close vicinity of the field polarization axis, we

6025: need $1/F_0\simeq 1/F_{s,0}$, with both, $F_s$ and $F$ small enough 

6026: to be treated at first order.

6027: 

6028: If we now consider the $s=1$ resonance, we deduce the secular Hamiltonian

6029: by keeping only the term which does not vanish after

6030: averaging over one Kepler period. For the microwave field, this 

6031: term was already identified in eq.~(\ref{hamscfin}).

6032: For the static field, only the static Fourier component of the atomic dipole,

6033: eqs.~(\ref{x0},\ref{y0}),

6034: has a non-vanishing average over one period.

6035: Altogether, this finally leads to

6036: \begin{equation}

6037: {\mathcal H}_{\rm sec}=\hat P_t -\frac{1}{2\hat{I}^2} -\omega \hat{I}

6038: + F_s X_0(\hat{I},L) \cos \psi  

6039:  + F \chi_1(\hat{I},L,\psi) \cos(\hat \theta+\delta_1).

6040: \end{equation}

6041: Since the last two terms of this Hamiltonian depend differently on

6042: $\hat{\theta}$, it is no more possible, as it was in the pure LP case (see

6043: section \ref{LIN3D}), to perform the quantization of the slow LP motion

6044: first. Only the secular approximation \cite{cct92} which consists in

6045: quantizing first the fast variables $(\hat{I},\hat{\theta})$, and subsequently

6046: the slow variables $(L,\psi)$, remains an option for the general

6047: treatment. However, since we are essentially interested in the wave-packet

6048: eigenstate with optimal localization properties, we shall focus on the ground

6049: state within a sufficiently large resonance island induced by a microwave

6050: field of an appropriate strength. This motivates the harmonic expansion of the

6051: secular Hamiltonian around the stable fixed point at 

6052: \begin{equation}

6053: \hat{I}=\hat{I}_1=\omega^{-1/3},\ \ \ \ \hat\theta = -\delta_1.

6054: \end{equation}

6055: with the characteristic

6056: frequency, see eq.~(\ref{omega_harmonic}):

6057: \begin{equation}

6058: \Omega(\hat{I}_1L,\psi) = \frac{\sqrt{3F\chi_1(\hat{I}_1,L,\psi)}}{\hat{I}_1^2}.

6059: \end{equation} 

6060: Explicit evaluation of the ground state energy of the locally harmonic

6061: potential yields 

6062: the effective Hamiltonian for the slow motion in 

6063: the $(L,\psi)$ plane:

6064: \begin{equation}

6065: H_{\rm eff}= -\frac{\Omega(L,\psi)}{2} +F\chi_1(L,\psi)

6066: +F_sX_0(L) \cos \psi ,

6067: \label{slow}

6068: \end{equation}

6069: where all quantities are evaluated at $\hat{I}=\hat{I_1}.$

6070: For the determination of the angular localization properties of the wave-packet

6071: it is now sufficient to inspect the extrema of $H_{\rm eff}$.

6072: 

6073: \begin{figure}

6074: \centerline{\hbox{

6075: \psfig{figure=bdzf36.ps,width=12cm}}}

6076: \caption{Contours of the effective Hamiltonian,

6077: eq.~(\protect{\ref{slow}}), in the

6078: $(L_0,\psi)$ plane (with $L_0=L/n_0$ the scaled angular momentum,

6079: and $\psi$ the angle between the major axis of the elliptical trajectory

6080: and the field axis). 

6081: The potential surface generates the slow 

6082: evolution

6083: of the angular coordinates of the 

6084: Kepler trajectory of a Rydberg electron exposed to collinear,

6085: static and resonant microwave electric fields.  

6086: Initial atomic principal quantum number $n_0=60$; 

6087: scaled microwave amplitude $F_0=Fn_0^4=0.03$, and scaled static field amplitude

6088: $F_{s,0}=0.12F_0<F_{s,c}$ (a),

6089: $F_{s,0}=0.25F_0>F_{s,c}$ (b), with $F_{s,c}$ the critical static field 

6090: amplitude defined in eq.~(\protect\ref{ec}). 

6091: The lighter the background, the higher

6092: the effective energy. 

6093: The contours are plotted at the semiclassical energies which 

6094: quantize $H_{\rm eff}$

6095: according to eq.(\ref{ebkp}), and thus represent the

6096: 60 eigenenergies shown in fig.~\protect\ref{f2}.

6097: Observe the motion of the stable island along the

6098: $\psi=\pi$ line (corresponding to the energetically highest 

6099: state in the manifold), under changes of $F_s$.

6100:  }

6101: \label{f1}

6102: \end{figure}

6103: For $F_{s,0}=0$, 

6104: we recover the pure LP case with a maximum along the line $L_0=1$ 

6105: (circular state), 

6106: and a minimum at $L_0=0$, $\psi=\pi/2$ (see fig.~\ref{lin3d_1}), 

6107: corresponding to 

6108: a straight line orbit perpendicular to the field. 

6109: For increasing

6110: $F_{s,0}$, the maximum moves towards lower values of $L_0$, and contracts

6111: in $\psi$, whereas the minimum approaches $\psi=0$ for constant $L_0=0$, 

6112: see fig.~\ref{f1}. It is easy to show that there exists

6113: a critical value $F_{s,c}$ of the static field, depending on $\hat{I}_1$,

6114: \be

6115: F_{s,c}=\frac{2}{3}\left\vert F_0J'_1(1)

6116: -\frac{\sqrt{3F_0J'_1(1)}}{4\hat{I}_1}\right\vert

6117:    \simeq 0.217F_0 -0.164\frac{\sqrt{F_0}}{\hat{I}_1},

6118: \label{ec}

6119: \ee

6120: above which both fixed

6121: points reach $L_0=0.$ Then, in particular, the maximum 

6122: at $L_0=0$, $\psi=\pi$, corresponds to a straight line orbit {\em parallel} 

6123: to

6124: $\vec{F}_s$. Note that in the classical limit, $\hat{I}_1\rightarrow\infty$,

6125: eq.~(\ref{ec}) recovers the purely

6126: classical value \cite{leopold87} for angular stability of the 

6127: straight line orbit 

6128: along the polarization axis, as it should.

6129: \begin{figure}

6130: \centerline{\psfig{figure=bdzf37.ps,width=10cm}}

6131: \caption{Semiclassical energy levels of the resonantly driven manifold 

6132: $n_0=60$ of the hydrogen atom, as a function of the ratio of the scaled static 

6133: electric field strength $F_s$ to the scaled 

6134: microwave amplitude $F_0$, for fixed $F_0=Fn_0^4=0.03$. The insert shows 

6135: the scaled angular momentum $L_0=L/n_0$ of the stable fixed point 

6136: $(L_0,\psi=0)$, see fig.~\protect\ref{f1},

6137: as a function of the same variable. The corresponding trajectory evolves

6138: from a circular orbit coplanar with the polarization axis to a straight

6139: line orbit stretched along this axis, via orbits of intermediate eccentricity.

6140: For $F_{s,0}>F_{s,c}$, eq.~(\ref{ec}), the stable fixed point is stationary

6141: at $L_0=0$. The corresponding wave-packet state, localized in

6142: the vicinity of the fixed point, is the energetically 

6143: highest state in the spectrum.

6144: For $F_{s,0}>F_{s,c}$, it is a completely localized wave-packet

6145: in the three dimensional space, propagating back and forth along the

6146: polarization axis, without spreading (see fig.~\ref{linf_wp}).

6147: }

6148: \label{f2}

6149: \end{figure}

6150: Therefore, by variation of 

6151: $F_{s,0}\in[0,F_{s,c}]$, we are able to continuously tune 

6152: the position of the maximum in 

6153: the $(L_0,\psi)$

6154: plane.

6155: Consequently, application of an additional static electric field gives us 

6156: control over the trajectory traced by 

6157: the wave-packet.

6158: This is further illustrated in fig.~\ref{f2}, through the 

6159: {\em semiclassical} level dynamics of the

6160: resonantly driven manifold originating from the $n_0=60$ energy shell, as a 

6161: function of $F_s$. In the limit $F_s=0$, the spectrum is equivalent to the one

6162: plotted in fig.~\ref{lin3d_3}(b). 

6163: As the static field is ramped up, the highest lying state

6164: (maximum value of the semiclassical quantum number $p$, for $F_s=0$, see 

6165: fig.~\ref{lin3d_3}, and the right column of fig.~\ref{lin3d_5}) gets stretched 

6166: along the static field direction and finally, for $F_{s,0}>F_{s,c}$, collapses onto

6167: the quasi one dimensional wave-packet eigenstate bouncing off the nucleus along

6168: a straight line Kepler trajectory. Likewise, the energetically lowest 

6169: state of the manifold (at $F_s=0$, minimum value of the semiclassical 

6170: quantum number $p$, left column of fig.~\ref{lin3d_5}) is equally rotated 

6171: towards the direction defined by $\vec{F_s}$, but stretched in the opposite 

6172: direction.

6173:  

6174: \begin{figure}

6175: \centerline{\psfig{figure=bdzf38.ps,width=12cm}}

6176: \caption{Temporal dynamics of the non-dispersive wave-packet of a

6177: three-dimensional hydrogen atom exposed to a linearly polarized,

6178: resonant microwave field, in the presence of a parallel static electric field. 

6179: $F_0=0.03$, $F_{s,0}=0.009$, $F_{s,0}/F_0=0.3$, $n_0=60$. 

6180: Driving field phases at the different stages of the wave-packet's evolution:

6181: $\omega t= 0$ (top left), $\pi/2$ (top right),

6182: $3\pi/4$ (bottom left), $\pi$ (bottom right). 

6183: It is well localized in the three dimensions of space and repeats its shape

6184: periodically (compare fig.~\protect\ref{lin1d_2} for the analogous dynamics in 

6185: the restricted 1D model, where no additional static electric field is needed). 

6186: The nucleus is at the center of the plot which extends

6187: over $\pm 8000$ Bohr radii. The microwave polarization axis and the

6188: static field are oriented along the vertical axis.

6189: }

6190:  \label{linf_wp}

6191: \end{figure}

6192: 

6193: The existence of a non-dispersive wave-packet localized in all three

6194: dimensions of space is confirmed by a pure quantum calculation, using

6195: a numerically exact  

6196: diagonalization of the Floquet Hamiltonian. Figure~\ref{linf_wp}

6197: shows the electronic density of a single Floquet eigenstate (the

6198: highest one in fig.~\ref{f2}, for $F_{s,0}/F_0=0.3$), 

6199: at various phases of the driving field.

6200: The wave-packet is clearly localized along the field axis, and

6201: propagates along a straight line classical 

6202: trajectory, repeating its shape periodically. Its dynamics precisely reproduces 

6203: the dynamics of the 1D analogue illustrated in fig.~\ref{lin1d_2}. 

6204: Once again, as for previous

6205: examples, the finite ionization rate (see section \ref{ION}) of the 

6206: 3D wave-packet is of

6207: the order of some million Kepler periods.

6208: 

6209: 

6210: 

6211: \subsection{Wave-packets in the presence of a static magnetic field}

6212: \label{MAG}

6213: 

6214: Similarly to a static electric field which may stabilize an angularly

6215: unstable wave-packet, 

6216: the properties of non-dispersive wave-packets

6217: under circularly polarized driving, in the presence of an additional 

6218: static magnetic field normal to 

6219: the polarization plane,

6220: has been studied in a series of papers

6221: \cite{lee95,farrelly95a,farrelly95,brunello96,cerjan97,lee97,brunello97}.

6222: The Hamiltonian of the system in the coordinate frame corotating with the CP

6223: field 

6224: reads (compare with eq.~(\ref{hrot}), for the pure CP case)

6225: \be

6226: H=\frac{p_x^2+p_y^2+p_z^2}{2}-\frac{1}{r}-

6227: (\omega-\omega_{\mathrm c}/2)L_z +Fx

6228: +\frac{\omega_{\mathrm c}^2}{8}(x^2+y^2),

6229: \label{fareq1}

6230: \ee

6231: where $\omega_{\mathrm c}$ is the cyclotron frequency. In atomic 

6232: units, the cyclotron frequency $\omega_{\mathrm c}=-qB/m$

6233: equals  the magnetic field value. It can be both positive or negative,

6234: depending on the direction of the magnetic field\footnote{A different

6235: convention is used (quantization axis defined by the orientation of

6236: the magnetic field) in many papers on this subject. It leads to

6237: unnecessarily complicated equations.}.

6238: This additional parameter modifies the dynamical properties which characterize 

6239: the

6240: equilibrium points, the analysis of which may be carried out alike

6241: the pure CP case treated in sec.~\ref{CP}. 

6242: A detailed stability analysis can be found in 

6243: \cite{lee97,farrelly95,rakovic98} and we summarize here

6244: the main results only.

6245:   

6246: Since changing the sign of $F$ in eq.~(\ref{fareq1}) is equivalent to changing

6247: the sign of $x$ 

6248: from positive to negative, we only consider the 

6249: equilibrium position at $x_{\rm eq}>0$ (compare eqs.~(\ref{wp0}-\ref{wpu})).

6250: For nonvanishing magnetic field, its position is given by

6251: \be

6252: \omega(\omega-\omega_{\mathrm c})x_{\rm eq}-\frac{1}{x_{\rm eq}^2}-F=0.

6253: \label{fareq2}

6254: \ee

6255: Redefining the dimensionless parameter $q$ (see eq.~(\ref{q})) via 

6256: \be

6257: q=\frac{1}{\omega(\omega-\omega_{\mathrm c})x_{\rm eq}^3},

6258: \label{mq}

6259: \ee

6260: we obtain for the

6261: microwave amplitude 

6262: \be 

6263: F=[\omega(\omega-\omega_{\mathrm c})]^{2/3}(1-q)/q^{1/3},

6264: \label{fm}

6265: \ee

6266: and  

6267: \be 

6268: E_{\mathrm eq}=[\omega(\omega-\omega_{\mathrm c})]^{1/3}(1-4q)/2q^{2/3}

6269: \label{me}

6270: \ee

6271: for the equilibrium energy.

6272: 

6273: Harmonic expansion of

6274: eq.~(\ref{fareq1}) around 

6275: the equilibrium point $(x_{\rm eq},y_{\rm eq}=0,z_{\rm eq}=0)$ allows for 

6276: a linear stability analysis

6277: in its vicinity.

6278: Alike the pure CP case,

6279: the $z$ motion decouples from the motion in

6280: the $(x,y)$ plane. For the latter, we recover 

6281: the generic harmonic Hamiltonian 

6282: discussed

6283: in section \ref{CP}, eq.~(\ref{rotaibb}), provided we substitute

6284: \be

6285: \tilde{\omega}=\omega-\omega_{\mathrm c}/2.

6286: \label{om}

6287: \ee

6288: 

6289: When expanded at second order around the equilibrium point, the

6290: Hamiltonian (\ref{fareq1}) takes the standard form

6291: of a rotating anisotropic oscillator, eq.~(\ref{rotaibb}), with

6292: $\tilde{\omega}$ replacing $\omega$, and with the stability parameters:

6293: \bea

6294: a&=&\frac{1}{\tilde{\omega}^2}\left(\frac{\omega_{\mathrm c}^2}{4}-\frac{2}

6295: {x_{\rm eq}^3}

6296: \right), \nonumber \\

6297: b&=&\frac{1}{\tilde{\omega}^2}\left(\frac{\omega_{\mathrm c}^2}{4}+\frac{1}

6298: {x_{\rm eq}^3}

6299: \right).

6300: \label{ab_mag}

6301: \eea

6302: 

6303: The regions of 

6304: stability of the equilibrium point $(x_{\rm eq},y_{\rm eq},z_{\rm eq})$

6305: are thus obtained from the domains of stability of the

6306: 2D rotating anisotropic oscillator, given by 

6307: eqs.~(\ref{stability_region_1},\ref{stability_region_2}). They

6308: are visualized in terms of the 

6309: physical parameters $F$ and $\omega_{\mathrm c}$,

6310: (using the standard scaled electric field $F_0=Fn_0^4=F\omega^{-4/3}$)

6311: in fig.~\ref{farfig1}, with the black region corresponding

6312: to eq.~(\ref{stability_region_1}), and the grey region to eq.~(\ref{stability_region_2}).

6313: Observe that the presence of the magnetic field tends to enlarge the 

6314: region

6315: of stability in parameter space; 

6316: for $\omega_{\mathrm c}=0$ (pure CP case, no magnetic field) the

6317: stability region is quite tiny, in comparison to large values of 

6318: $|\omega_{\mathrm c}|$

6319: \footnote{As long as we are interested in long-lived wave-packets,

6320: the region of small $F_0$ is of interest only. At higher $F_0$ and for

6321: $\omega_{\mathrm c}<\omega$, the

6322: strong driving field will ionize the atom rather fast -- see section \ref{ION}.

6323:  This makes the gray

6324: region $F_0>0.1$ of little practical

6325: interest.}.

6326: On the other hand, the stability diagram does not provide us any detailed 

6327: information on the actual size of the resonance island surrounding the 

6328: equilibrium point.

6329: However, 

6330: it is precisely the size of the resonance island which is crucial for 

6331: anchoring

6332: nondispersive quantum wave-packets close to the classical equilibrium point

6333: (see sec.~\ref{sapp}). 

6334: 

6335: An alternative approach to characterize the stability properties of

6336: the classical motion near $(x_{\rm eq},y_{\rm eq},z_{\rm eq})$ has been 

6337: advertized in 

6338: \cite{lee97,farrelly95a,farrelly95,brunello96,cerjan97,lee95,brunello97}: the 

6339: concept of zero velocity surfaces (ZVS).

6340: \begin{figure}

6341: \centerline{\psfig{figure=bdzf39.eps,width=10cm,angle=-90}}

6342: \caption{(Shaded) Regions of stability of the equilibrium point 

6343: $(x_{\rm eq},y_{\rm eq},z_{\rm eq})$ for circularly polarized 

6344: driving (amplitude $F$, 

6345: frequency $\omega$) of a hydrogen atom, in the presence of a magnetic

6346: field (corresponding cyclotron frequency $\omega_{\mathrm c}$). The black and grey regions correspond 

6347: to the two regions of stability, described by 

6348: eqs.~(\protect{\ref{stability_region_1}}) and (\protect{\ref{stability_region_2}}),

6349: respectively.

6350: }

6351: \label{farfig1}

6352: \end{figure}

6353: In order to construct a ZVS, 

6354: the Hamilton function is expressed in terms of velocities  

6355: rather than canonical momenta. For the harmonic Hamiltonian

6356: (\ref{rotaibb}), the calculation yields

6357: \begin{equation}

6358: H=\frac{v_x^2+v_y^2}{2}+\frac{\omega^2}{2}\left [(a-1)x^2+ (b-1)y^2

6359: \right ].

6360: \label{ZVS1}

6361: \end{equation}

6362: Thus, the ``kinetic energy'' becomes a positive function of velocities, and one

6363: can define the ZVS as

6364: \begin{equation}

6365: S=H-\frac{v_x^2+v_y^2}{2},

6366: \label{ZVS2}

6367: \end{equation}

6368: the generalization of an effective potential for interactions which mix 

6369: position and momentum coordinates.

6370: Note that, when the 

6371: velocities coincide with the canonical momenta, $S$ is nothing 

6372: but the potential energy surface. We prefer to denote it $S$ instead of $V$,

6373: to stress the difference. As discussed in detail in \cite{lee97}, a ZVS

6374: may be used to locate the equilibrium points. However, 

6375: their stability properties are not obvious 

6376: ({\em contrary} to the potential surface, where 

6377: minima define stable

6378: fixed points, while maxima and saddle points are unstable). For a ZVS, saddles

6379: are also unstable, but maxima may either be stable {\em or} 

6380: unstable.

6381: For example, the first stability region, eq.~(\ref{stability_region_1}), 

6382: of the rotating 2D anistropic

6383: Hamiltonian is associated with

6384: a stable minimum of the ZVS. The second region of stability,

6385: eq.~(\ref{stability_region_2}), corresponds to $a,b\leq 1$ and thus

6386: to a maximum of the ZVS. However, the ZVS does not show

6387: any qualitative change whether $(a-b)^2+8(a+b)$ is positive or negative,

6388: i.e. whether the equilibrium point is stable or unstable.

6389: Thus, a ZVS is clearly inappropriate, or at least potentially

6390: dangerous, for the discussion of the classical motion close to

6391: equilibrium.

6392: As a matter of fact, this difficulty with the ZVS is crucial

6393: in our case, even for the pure CP case, without additional magnetic field.

6394: Indeed, the ZVS

6395: becomes

6396: \begin{equation}

6397: S=\left(-\frac{2q+1}{2}x^2+\frac{q-1}{2}y^2\right) \omega^2,

6398: \label{ZVS3}

6399: \end{equation}

6400: where we use the single parameter $q$ to parametrize $S$. At $q=1$

6401: (i.e. $F=0$, see eq.~(\ref{fm})), the equilibrium point 

6402: turns from a saddle

6403: (for $q>1$) into a maximum. Consequently, the ZVS correctly reflects 

6404: the change of the equilibrium

6405: point from unstable ($q>1$) to stable ($q<1$).

6406: However, for any  $q\in(0,1)$, the equilibrium remains a maximum of the ZVS,

6407: which completely misses the change of stability at $q=8/9$.

6408: Thus, the very same maximum

6409: may change its stability (which fundamentally affects the classical 

6410: motion in its

6411: vicinity) without being noticed by inspection of the 

6412: ZVS. The latter

6413: evolves very smoothly around $q=8/9$. Thus, the ZVS contours provide

6414: {\em no} information on the nature of the classical 

6415: motion in the vicinity of the

6416: equilibrium point, in {\em disaccord} with   

6417: \cite{lee97,cerjan97,farrelly95a,farrelly95,brunello96,lee95,brunello97}.

6418: Similarly, the isovalue contours of the ZVS (which are ellipses in the

6419: harmonic approximation) have no relation

6420: with the isovalue  contours of the ground-state wave-packet

6421: localized

6422: around the equilibrium point (these contours

6423: are also ellipses in the harmonic approximation where the

6424: wave-packet is a Gaussian),

6425: contrary to what is stated in \cite{lee97,farrelly95a}.

6426: For example, the aspect ratio (major axis/minor axis)

6427: of the ZVS contour lines is $\sqrt{(2q+1)/(1-q)}$ which varies smoothly 

6428: around $q=8/9$, while the aspect ratio of the isocontours 

6429: of the ground state wave-packet diverges when $q\to 8/9.$ 

6430: \footnote{While this argument has been presented here

6431: for the simplest case of the harmonic

6432: oscillator hamiltonian (\ref{rotaibb}), it carries over to 

6433: the full, nonlinear model,

6434: eq.~(\ref{fareq1}).}

6435: 

6436: Nonwithstanding, a ZVS may be used for other purposes \cite{lee95}, e.g., 

6437: to show the existence of an ionization threshold 

6438: for the Hamiltonian (\ref{fareq1}), when $\omega_{\mathrm c}>\omega$

6439: (area coded in black in fig.~\ref{farfig1}). 

6440: Clearly, due to the parabolic confinement in the 

6441: $x-y$-plane, ionization is only possible along the 

6442: $z$ direction.

6443: The threshold  is given by

6444: \cite{lee95}

6445: \be

6446: E_{\mathrm ion}=F^2/2\omega(\omega-\omega_{\mathrm c}),

6447: \label{thresh_m}

6448: \ee

6449: which 

6450: lies above the equilibrium energy $E_{\rm eq}$. Thus, for parameters 

6451: in that region,

6452: the electron -- initially placed close to the stable fixed point --

6453: cannot ionize. One may expect, therefore, that wave-packets built

6454: around the equilibrium point for $\omega_{\mathrm c}>\omega$ lead to

6455: {\em discrete} Floquet states. In other cases, e.g., for pure

6456: CP driving, non-dispersive wave-packets are rather represented by 

6457: long-living resonances (see section \ref{ION}).

6458: 

6459: Finally, it has been often argued  

6460: \cite{lee97,farrelly95a,farrelly95,brunello96,cerjan97,lee95,brunello97}

6461: that the presence of the magnetic field is absolutely necessary for the

6462: construction of non-dispersive wave-packets. The authors consider 

6463: non-dispersive

6464: wave-packets 

6465: as equivalent 

6466: to Gaussian shaped wave-functions (using

6467: equivalently the notion of coherent states). Then it is vital that the

6468: motion in the vicinity of the fixed point is locally harmonic within a

6469: region of size $\hbar$. This leads the authors to conclude that

6470: non-dispersive wave-packets may not exist for the pure CP case except in the

6471: extreme semiclassical regime. 

6472: As opposed to that, the diamagnetic term in

6473: eq.~(\ref{fareq1}) gives a stronger weight to 

6474: the harmonic term, which is the basis of the above claim\footnote{Note

6475:  that the non-harmonic terms, being entirely due

6476: to the Coulomb field, are not removed or decreased by the addition

6477: of a magnetic field. They are just hidden by a larger harmonic term.}.

6478: From our point of view, which, as already stated above, 

6479: attributes 

6480: the non-dispersive character of the wave-packet 

6481: to a classical nonlinear

6482: resonance, the accuracy of the harmonic approximation 

6483: (which, anyway, always

6484: remains an approximation) is 

6485: irrelevant for the existence of

6486: non-dispersive wave-packets. 

6487: The best proof is that \cite{farrelly95a}

6488: concludes, on the

6489: basis of the validity of the harmonic approximation,

6490: that non-dispersive wave-packets should not exist for

6491: $n_0\simeq 60$ in CP field, in complete

6492: contradiction to numerically exact experiments showing their

6493: existence down to $n_0=15$ \cite{delande95}.

6494: On the other hand, it is an interesting question

6495: how good 

6496: the harmonic approximation actually is in the pure CP case. The interested

6497: reader may find a more quantitative discussion of this point in section

6498: \ref{SPO}.

6499: 

6500: 

6501: 

6502: \section{Other resonances}

6503: \label{HOR}

6504: 

6505: \subsection{General considerations}

6506: \label{ORG}

6507: 

6508: We have so far restricted our attention to nondispersive wave-packets anchored

6509: to the principal resonance of periodically driven Hamiltonian systems. In

6510: section~\ref{GM}, we already saw that {\em any} harmonic of the unperturbed 

6511: classical motion can dominate the harmonic expansion (\ref{hamfou}) of the 

6512: classical Hamilton function, provided it is resonantly driven by the external

6513: perturbation, i.e.

6514: \begin{equation}

6515: s\theta -\omega t\simeq {\mathrm const},\ s>0\ {\mathrm integer}.

6516: \label{rescond}

6517: \end{equation}

6518: This is the case when the $s^{\mathrm th}$ 

6519: harmonic of the classical internal frequency $\Omega$ is resonant with

6520: the external driving $\omega$. As $\Omega$ depends on the classical unperturbed

6521: action, the corresponding classical

6522: resonant action is defined by

6523: \begin{equation}

6524: \Omega(I_s) = \frac{\partial H_0}{\partial I}(I_s) = \frac{\omega}{s}.

6525: \label{Is}

6526: \end{equation}

6527: At this action, 

6528: the period of the classical motion is $s$ times the period 

6529: of the external drive.

6530: Precisely like in the $s=1$ case (the principal resonance), for any integer

6531: $s>1,$ Floquet eigenstates of the driven system exist which are 

6532: localized on the 

6533: associated classical stability islands in phase space. The energy of 

6534: these eigenstates 

6535: can again be estimated through the semiclassical quantization of the secular

6536: dynamics. 

6537: To do so, we start from eqs.~(\ref{hamfou},\ref{hamext}) and transform to 

6538: slowly varying variables (the ``rotating frame") defined by

6539: \begin{eqnarray}

6540: \hat{\theta}&=&\theta-\frac{\omega t}{s}, \label{trafo_s_a} \\

6541: \hat{I}&=&I, \nonumber \label{trafo_s_b} \\

6542: \hat{P_t}&=&P_t+\frac{\omega I}{s} \nonumber.

6543: \label{trafo_s_c}

6544: \end{eqnarray}

6545: The Floquet Hamiltonian in this rotating frame now reads:

6546: \begin{eqnarray}

6547: \hat{\mathcal H} =  \hat{P}_t+H_0(\hat{I})-\frac{\omega\hat{I}}{s} 

6548: + \lambda

6549: \sum_{m=-\infty}^{+\infty}V_m(\hat{I})\left\{\cos\left(m\hat{\theta}+

6550: \left(\frac{m-s}{s}\right)\omega t\right)\right\}, 

6551: \label{ham_trafo_lin1d}

6552: \end{eqnarray}

6553: which is periodic with period $\tau=sT$ with $T=2\pi/\omega$.

6554: Passing to the rotating frame 

6555: apparently destroys the $T$-periodicity of the original Hamiltonian.

6556: This, however, is of little importance, the crucial point

6557: being to keep the periodicity  $sT$ of the internal motion.

6558: If we now impose the resonance condition (\ref{rescond}), the major 

6559: contribution to the sum in eq.~(\ref{ham_trafo_lin1d}) will come 

6560: from the slowly evolving resonant term $m=s$, while the other 

6561: terms vanish upon averaging the fast variable $t$ over one period $\tau$, 

6562: leading to the secular Hamilton function

6563: \begin{equation}

6564: {\cal H}_{\rm sec}=\hat{P}_t+H_0(\hat{I})-\frac{\omega\hat{I}}{s}

6565: +\lambda V_s(\hat{I})\cos(s\hat{\theta}).

6566: \label{sec_s}

6567: \end{equation}

6568: This averaging procedure eliminates the explicit time dependence of 

6569: $\hat{\mathcal H}$, and is tantamount to restricting the validity of 

6570: ${\mathcal H}_{\rm sec}$ to the description of those classical 

6571: trajectories 

6572: which comply with eq.~(\ref{rescond})

6573: and, hence, exhibit a periodicity with period $\tau$. This will have an 

6574: unambiguous signature in the quasienergy spectrum, as we shall see further

6575: down.

6576: The structure of the secular Hamiltonian is simple and reminds us of the

6577: result for 

6578: the principal $s=1$ resonance, eq.~(\ref{hsec_ap}). However,

6579: due to the explicit appearance of the factor $s>1$ in the argument of the

6580: $\cos$ term,  a juxtaposition of $s$ resonance

6581: islands close to the resonant action, eq.~(\ref{Is}), is created.

6582: It should be emphasized that these $s$ resonance islands

6583: are actually $s$ clones of the same island. Indeed, a trajectory trapped

6584: inside a resonance island will successively visit all the islands:

6585: after one period of the drive, $\theta$ is approximately increased

6586: by $2\pi/s,$ corresponding to a translation to the next island. 

6587: At the center of the islands, there is a single, stable resonant

6588: trajectory whose period is exactly $\tau=sT.$

6589: 

6590: At lowest order in $\lambda,$ all the quantities of interest can

6591: be expanded in the vicinity of $I_s$,

6592: exactly as for the principal resonance in section~\ref{CD}.

6593: $V_s$ is consistently evaluated at the resonant action. 

6594: The pendulum approximation of the secular Hamiltonian then reads:

6595: \begin{equation}

6596: {\cal H}_{\mathrm pend} = \hat{P}_t + H_0(\hat{I_s}) 

6597: -\frac{\omega}{s}\hat{I}_s 

6598: + \frac{1}{2}

6599: H^{''}_0(\hat{I_s})\ (\hat{I}-\hat{I_s})^2 + \lambda

6600: V_s(\hat{I_s}) \cos s\hat{\theta},

6601: \label{eqpend_s}

6602: \end{equation}

6603: with the centers of the islands located at:

6604: \begin{equation}

6605: \hat{I}=\hat{I}_s=I_s,

6606: \end{equation}

6607: \begin{equation}

6608: \left\{

6609: \begin{array}{l}

6610: \displaystyle \hat\theta = k \frac{2\pi}{s},\ \ \ \ \ {\mathrm if}\ \lambda V_s(\hat{I}_s)H^{''}_0(\hat{I}_s)<0,\\

6611: \displaystyle \hat\theta = k \frac{2\pi}{s} + \frac{\pi}{s},\ \ \ \ \ {\mathrm if}\ \lambda V_s(\hat{I}_s)H^{''}_0(\hat{I}_s)>0,

6612: \end{array}

6613: \right.

6614: \end{equation}

6615: where $k$ is an integer running from 0 to $s-1.$

6616: For $H^{''}_0(\hat{I}_s)$ -- see eq.~(\ref{second_derivative}) -- positive,

6617: these are minima of the secular Hamiltonian, otherwise they are

6618: maxima. 

6619: The extension of each resonance island is, as a direct generalization of the

6620: results of section \ref{CD}:

6621: \begin{eqnarray}

6622: \Delta\hat\theta& = &\frac{2\pi}{s},\\

6623: \Delta\hat{I} & = & 4\sqrt{\left|\frac{\lambda V_s(\hat{I_s})}{ H^{''}_0(\hat{I_s})}\right|,}

6624: \end{eqnarray}

6625: and its area

6626: \begin{equation}

6627: A_s(\lambda) = \frac{16}{s} \sqrt{\left|

6628: \frac{\lambda V_s(\hat{I_s})}{ H^{''}_0(\hat{I_s})}\right|}.

6629: \label{area_s}

6630: \end{equation}

6631: Again, the dependence of $A_s(\lambda)$ on $\sqrt{|\lambda|}$ implies that even

6632: small perturbations may

6633: induce

6634: significant changes in the phase space structure, provided the perturbation

6635: is resonant with a harmonic of the unperturbed classical motion\footnote{

6636: The situation is very different in the opposite case, when $\omega$

6637: is the $s^{\mathrm th}$ SUB-harmonic \cite{marion} of the internal frequency. A resonance

6638: island may then exist but it is typically much smaller as it comes into

6639: play only at order $s$ in perturbation theory, with a

6640: size scaling as $|\lambda|^{s/2}.$}.

6641: 

6642: The construction of a non-dispersive wave-packet is simple once the

6643: $s$-resonance structure is understood: indeed, any set of initial

6644: conditions trapped in one of the $s$-resonance islands will classically remain

6645: trapped forever. Thus, a quantum wave-packet

6646: localized initially inside a resonance island is a good candidate for

6647: building a  non-dispersive wave-packet. There remains, however, a difficulty:

6648: the wave-packet can be initially placed in any of the $s$ resonance islands.

6649: After one period of the driving, it will have jumped to

6650: the next island, meaning that it will be far from its initial position.

6651: On the other hand, the Floquet theorem guarantees the existence

6652: of states which are strictly periodic with the period of the drive 

6653: (not the period of the resonant internal motion). 

6654: The solution to this

6655: difficulty is to build eigenstates which simultaneously occupy

6656: {\em all} 

6657: $s$-resonance islands, that is, which are composed

6658: of $s$ wave-packets each localized on a different resonance

6659: island. After one period of the drive, each individual wave-packet replaces

6660: the next one, resulting in globally $T$-periodic motion of this ``composite''

6661: Floquet state.

6662:  If the system has a macroscopic size (i.e. in the

6663: semiclassical limit), 

6664: individual wave-packets will appear extremely well localized

6665: and lying far from the other ones

6666: while maintaining a well-defined phase coherence with them.

6667: For $s=2$, the situation

6668: mimics a symmetric double well potential, where even and odd solutions

6669: are linear combinations of nonstationary states, each localized in either one

6670: well \cite{landau2}.

6671: 

6672: In order to get insight in the structure of the Floquet quasi-energy spectrum,

6673: it is useful to perform the semiclassical 

6674: EBK quantization of the secular Hamiltonian (\ref{sec_s}).

6675: Quantization of the motion in 

6676: $(\hat{I},\hat{\theta})$,

6677: see section~\ref{sapp}, provides 

6678: states trapped within the

6679: resonance islands (librational motion),

6680: and states localized outside 

6681: them (rotational motion). As usual, the number of

6682: trapped states is given 

6683: by the size, eq.~(\ref{area_s}) of the

6684: resonance island:

6685: \begin{equation}

6686: {\mathrm Number\ of\ trapped\ states} \simeq \frac{8}{\pi \hbar s}

6687: \sqrt{\left|\frac{\lambda V_s(\hat{I_s})}{ H^{''}_0(\hat{I_s})}\right|}.

6688: \label{number_of_trapped_states_s}

6689: \end{equation}

6690: The quantization can be performed along the contours of 

6691: any of the $s$ clones of the resonance island,

6692: giving of course the same result. However, this does {\em not}

6693: result in a $s$-degeneracy of the spectrum: indeed, the $s$ clones

6694: belong to the same torus in phase space (see above) and do

6695: not generate $s$ independent states.

6696: 

6697: If the number of trapped states is sufficiently large,

6698: the harmonic approximation to the pendulum (or secular) Hamiltonian

6699: can be used, with the frequency of the harmonic motion

6700: around the stable resonant orbit given by:

6701: \begin{equation}

6702: \omega_{\mathrm harm} = s\sqrt{|\lambda V_s(\hat{I_s})H^{''}_0(\hat{I_s})|}.

6703: \label{omega_harmonic_s}

6704: \end{equation}

6705: 

6706: In order to get the complete semiclassical Floquet spectrum,

6707: we additionally have to perform the semiclassical quantization in the 

6708: $(t,\hat{P}_t)$ plane,

6709: giving:

6710: \begin{equation}

6711: \frac{1}{2\pi}\int_0^{\tau}\hat{P}_tdt=\frac{\hat{P}_t\tau}{2\pi}

6712: =\frac{s\hat{P}_t}{\omega} = \left(

6713: j+\frac{\mu}{4}\right)\hbar,

6714: \label{wkb_k1d}

6715: \end{equation}

6716: with $j$ integer.

6717: This finally yields the semiclassical Floquet levels (in the 

6718: harmonic approximation):

6719: \begin{equation}

6720: {\cal E}_{N,j} = H_0(\hat{I_s}) - \frac{\omega}{s} \hat{I}_s + \left(j+\frac{\mu}{4}\right)\hbar \frac{\omega}{s}

6721: - \ {\mathrm sign}(H^{''}_0(\hat{I_s}))

6722: \left[ |\lambda V_s(\hat{I_s})| -

6723: \left(N+\frac{1}{2}\right) \hbar \omega_{\mathrm harm} \right],

6724: \label{spectrum_harmonic_s}

6725: \end{equation}

6726: with $N$ a non-negative integer.

6727: The wave-packet with optimum localization in the

6728: $(\hat{I},\hat{\theta})$ plane, i.e. optimum localization

6729: along the classical unperturbed orbit, is the $N=0$ state.

6730: According to eq.~(\ref{spectrum_harmonic_s}), the

6731: semiclassical quasi-energy spectrum has

6732: a periodicity $\hbar\omega/s,$ whereas the ``quantum" Floquet theory

6733: only enforces 

6734: $\hbar\omega$ periodicity.

6735: Thus, inside a Floquet zone of width $\hbar\omega,$

6736: each state appears $s$ times (for $0\leq j < s$), at energies

6737: separated by  $\hbar\omega/s.$ Note that this property is a direct

6738: consequence of the possibility of eliminating the time dependence of

6739: $\mathcal H$ in eq.~(\ref{ham_trafo_lin1d}) by averaging over $\tau$,

6740: leading to the time independent expression (\ref{sec_s}) for

6741: ${\cal H}_{\rm sec}$. 

6742: Therefore, it will 

6743: be only approximately valid for the exact quantum Floquet

6744: spectrum. In contrast, the $\hbar\omega$ periodicity holds exactly,

6745: as long as the system Hamiltonian is time-periodic.

6746: 

6747: We will now recover the $\hbar\omega/s$ 

6748: periodicity in a quantum description of our problem,

6749: which will provide us with the formulation of an eigenvalue problem 

6750: for the wave-packet eigenstates anchored to the $s$-resonance, in terms of a 

6751: Mathieu 

6752: equation. In doing so, we shall extend the 

6753: general concepts outlined in section \ref{section_mathieu} above.

6754: 

6755: Our starting point is eq.~(\ref{coupled_equations}), which we again consider

6756: in the regime where the eigenenergies $E_n$ of the 

6757: unperturbed Hamiltonian $H_0$ are locally approximately spaced 

6758: by $\hbar\Omega.$

6759: The resonance condition (\ref{rescond}) implies

6760: \begin{equation}

6761: \left. \frac{dE_n}{dn}\right|_{n=n_0}=\hbar\frac{\omega}{s},

6762: \label{deriven}

6763: \end{equation}

6764: where, again, $n_0$ is not necessarily an integer, and is related

6765: to the resonant action and its associated Maslov index through:

6766: \begin{equation}

6767: \hat{I}_s = \left(n_0+\frac{\mu}{4}\right) \hbar.

6768: \label{def_n0_s}

6769: \end{equation}

6770: When the resonance condition is met, the only efficient

6771: coupling in eq.~(\ref{coupled_equations}) connects 

6772: states with the same value of $n-sk.$ In other words, 

6773: in the secular approximation, a given state $(n,k)$ only 

6774: couples to $(n+s,k+1)$ and $(n-s,k-1).$

6775: We therefore consider a given ladder of coupled states labeled 

6776: by $j=n-sk.$ Because of the overall $\omega$ perodicity

6777: of the spectrum, changing $j$ by $s$ units (i.e., shifting all $k$-values

6778: by 1) is irrelevant, so that it is enough to consider the $s$ independent

6779: ladders $0\leq j \leq s-1.$

6780: Furthermore,

6781: in analogy to section~\ref{section_mathieu}, eq.~(\ref{matel_mat}), 

6782: we can replace the coupling matrix

6783: elements in eq.~(\ref{ham_trafo_lin1d}) by the resonantly driven Fourier

6784: coefficients of the classical motion, 

6785: \begin{equation}

6786: \langle \phi_n|V|\phi_{n+s}\rangle\simeq\langle \phi_{n-s}|V|\phi_n\rangle\simeq V_s(\hat{I}_s).

6787: \label{coupling_s}

6788: \end{equation}

6789: With these approximations, and the shorthand notation $r=n-n_0$,

6790: eq.~(\ref{coupled_equations}) takes the form of $s$ independent

6791: sets of coupled equations, 

6792: identified by the integer $j$:\footnote{For $s=1$, this equation reduces of course

6793: to eq.~(\protect\ref{tridiag}). We here use $n_0$ instead of $\hat{I}_s$; 

6794: the two quantities differ only by the Maslov index, eq.~(\protect\ref{def_n0_s}).}

6795: \begin{equation}

6796: \left[

6797: {\cal E}-E_{n_0}+\frac{n_0-j}{s}\hbar\omega-\frac{\hbar^2 r^2}{2}H_0^{''}

6798: (\hat{I}_s)\right]

6799: d_{r}=\frac{\lambda}{2}V_s[d_{r+s}+d_{r-s}],

6800: \label{coupled_s}

6801: \end{equation}

6802: where 

6803: \begin{equation}

6804: d_{r}=c_{n_0+r,\frac{n_0+r-j}{s}},

6805: \label{d_s}

6806: \end{equation}

6807: as a generalization of the notation in eq.~(\ref{diag_mat}). Again, $r$ is

6808: not necessarily an integer, but the various $r$ values involved

6809: in eq.~(\ref{coupled_s}) are equal modulo $s$.

6810: Precisely as in the case of the principal resonance, eq.~(\ref{coupled_s}) 

6811: can be

6812: mapped on its dual space expression, via eq.~(\ref{dual}):

6813: \begin{equation}

6814: \left[

6815: -\frac{\hbar^2}{2}H_0^{''}(\hat{I}_s)\frac{d^2}{d\phi^2}+E_{n_0}

6816: -\frac{n_0-j}{s}\hbar\omega+\lambda

6817: V_s\cos(s\phi)\right] f(\phi) = {\cal E} f(\phi),

6818: \label{mathieu_s}

6819: \end{equation}

6820: and identified with the Mathieu equation (\ref{mat_eq}) through

6821: \begin{eqnarray}

6822: s\phi & = & 2v, \\

6823: a & = & \frac{8}{\hbar^2s^2H_0^{''}(\hat{I}_s)}\left[{\cal E}-E_{n_0}+(n_0-j)\hbar

6824: \frac{\omega}{s}\right],\ j=0,\ldots, s-1, \nonumber \\

6825: q & = & \frac{4\lambda V_s(\hat{I}_s)}{s^2\hbar^2H_0^{''}(\hat{I}_s)}. \nonumber

6826: \label{ident_s}

6827: \end{eqnarray}

6828: The quasienergies associated with the $s$ resonance in the pendulum

6829: approximation then follow immediately as

6830: \begin{equation}

6831: {\cal E}_{\kappa,j}=E_{n_0}-(n_0-j)\hbar\frac{\omega}{s}+\frac{\hbar^2s^2}{8}

6832: H_0^{''}(\hat{I}_s)a_{\kappa}(\nu,q),

6833: \label{qu_mat_s}

6834: \end{equation}

6835: where the index $j$ runs from $0$ to $s-1$, and $\kappa$	labels the 

6836: eigenvalues of the Mathieu equation \cite{abramowitz72}.

6837: Again, the boundary condition for the solution

6838: of the Mathieu equation is incorporated via the characteristic exponent, which

6839: reads

6840: \begin{equation}

6841: \nu=-2\frac{n_0-j}{s}\ ({\mathrm mod}\ 2),\ \ \ \ \ j=0,\ldots s-1.

6842: \label{char_s}

6843: \end{equation}

6844: The structure of this quasi-energy spectrum apparently displays the

6845: expected $\hbar\omega/s$ periodicity. However, the characteristic 

6846: exponent $\nu$

6847: -- and consequently the $a_{\kappa}(\nu,q)$ eigenvalues --

6848: depend on $j$, what makes the periodicity approximate only. It is

6849: only far inside

6850: the resonance island that the $a_{\kappa}(\nu,q)$ eigenvalues are almost

6851: independent

6852: of $\nu$ and the periodicity is recovered. Deviations from this

6853: periodicity are further discussed in section~\ref{OGB}.

6854: Finally, the asymptotic expansion of the $a_{\kappa}(\nu,q)$ for large $q$,

6855: eq.~(\ref{math_asy}),

6856: gives again (compare section \ref{section_mathieu}) 

6857: the semiclassical estimate of the energy levels in the

6858: harmonic approximation, eq.~(\ref{spectrum_harmonic_s}), where the

6859: indices $\kappa$ and $N$ coincide.

6860:  

6861: 

6862: 

6863: \subsection{A simple example in 1D: the gravitational bouncer}

6864: \label{OGB}

6865: 

6866: As a first example of non-dispersive wave-packets localized on $s>1$ primary

6867: resonances, 

6868: we consider

6869: the particularly simple 1D model of

6870: a particle

6871: moving vertically in the gravitational field, and bouncing off 

6872: a periodically driven horizontal plane.

6873: This system is known as the Pustylnikov model \cite{lichtenberg83} (or,

6874: alternatively, the ``gravitational bouncer", or the ``bubblon model"

6875: \cite{holthaus95,benvenuto91,oliveira94,flatte96})  

6876: and represents a standard example of chaotic motion. Moreover,

6877: despite its simplicity, it may find possible 

6878:  applications in the dynamical manipulation of cold atoms

6879: \cite{steane95}.

6880: A gauge transformation shows its equivalence to a periodically

6881: driven particle moving in a triangular potential well, with the Hamiltonian

6882: \begin{equation}

6883: H=\frac{p^2}{2}+V(z)+\lambda z \sin(\omega t),

6884: \label{boueq1}

6885: \end{equation}

6886: where 

6887: 

6888: \begin{equation}

6889: V(z)=\left\{ 

6890: \begin{array}{ll}

6891: z & {\rm for}\ z\geq 0 , \\

6892: \infty & {\rm for}\ z<0 .

6893: \end{array}

6894: \right.

6895: \end{equation}

6896: The strength $\lambda$ of the periodic driving is proportional 

6897: to the maximum

6898: excursion 

6899: of

6900: the oscillating surface.

6901: Classically, this system is 

6902: well approximated by the standard map \cite{lichtenberg83}

6903: (with the momentum and the phase of the driving

6904:  at the moment of the bounce as variables), with kicking amplitude

6905: $K=4\lambda$. 

6906: 

6907: Apart from a (unimportant) phase shift $\pi/2$ in $\omega t$, 

6908: eq.~(\ref{boueq1}) 

6909: is of the general type of eq.~(\ref{h_gen}) 

6910: and the scenario for the creation of 

6911: non-dispersive wave-packets described in sections~\ref{CD} and

6912: \ref{QD} is applicable. 

6913: As a matter

6914: of fact, a  careful analysis of the problem 

6915: using the Mathieu approach described

6916: in sec.~\ref{section_mathieu}, as well as the semiclassical

6917: quantization of the Floquet Hamiltonian 

6918: were already outlined in \cite{holthaus95}, 

6919: where the associated Floquet eigenstates were baptized ``flotons".

6920: We recommend \cite{holthaus95,flatte96} for 

6921: a detailed discussion of the problem,

6922: reproducing here only the main results,

6923: with some minor modifications.

6924: 

6925: Solving the classical equations of motion for the unperturbed Hamiltonian

6926: is straightforward 

6927: (piecewise uniformly accelerated

6928: motion alternating with 

6929: bounces off the mirror) 

6930: and it is easy to express the unperturbed Hamiltonian and the

6931: classical internal frequency in terms of action-angle variables:

6932: \begin{eqnarray}

6933: \displaystyle H_0&=&\frac{(3\pi I)^{2/3}}{2},\\

6934: \displaystyle \Omega &=& \frac{\pi^{2/3}}{(3I)^{1/3}},

6935: \end{eqnarray}

6936: while the full, time-dependent Hamiltonian reads:

6937: \begin{equation}

6938:  H=\frac{(3\pi I)^{2/3}}{2}+\frac{\lambda\pi I^{2/3}}{(3\pi)^{1/3}}

6939: \sin(\omega t)

6940: -\frac{2\lambda (3I)^{2/3}}{\pi^{4/3}}\sin(\omega t)\sum_{n=1}^\infty

6941:  \frac{\cos(n\theta)}{n^2}.

6942: \label{boueq2}

6943: \end{equation}

6944: 

6945: Thus, the resonant action (\ref{Is}) is given by

6946: \begin{equation}

6947: \hat{I}_s = \frac{\pi^2s^3}{3\omega^3},

6948: \label{act-s-res}

6949: \end{equation}

6950: with the associated strength of the effective coupling

6951: \begin{equation}

6952: V_s = \frac{(3I)^{2/3}}{s^2\pi^{4/3}}.

6953: \end{equation}

6954: Using the framework of

6955: sections~\ref{CD}, \ref{QD} (for $s=1$),

6956: and \ref{ORG} (for $s>1$), the reader

6957: may easily compute the various  properties  of non-dispersive wave-packets

6958: in this system\footnote{There is, however, a tricky point:

6959: the Maslov index in this system

6960: is 3, with a contribution  1 coming from the outer turning point,

6961: and 2 from $z=0$, since the oscillating plane 

6962: acts as a hard wall. Hence, the relation

6963: between the principal quantum number and the action is 

6964: $I_1=n+3/4$, see eq.~(\protect\ref{WKB}).}.  

6965: An example for $s=1$

6966: is presented in fig.~\ref{boufig2},

6967: for the resonant principal quantum number $n_0=1000,$ (i.e., 

6968: $\hat{I}_1=1000.75$) where both,

6969: the (time-periodic) probability densities in 

6970: configuration and phase space 

6971: are shown. Note that such 

6972: high $n_0$ values (or even higher) correspond to typical experimental 

6973: falling heights (around 0.1 mm for $n_0=1000$) in experiments on cold atoms \cite{steane95}.

6974: Therefore, the creation of an atomic wave-packet in such an experiment would 

6975: allow to store 

6976: the atom in a quasi classical ``bouncing mode" over arbitrarily long times, 

6977: and might find some application in the field 

6978: of atom optics \cite{oberthaler99}.

6979: 

6980: \begin{figure}

6981: \centerline{\psfig{figure=bdzf40.ps,width=14cm}}

6982: \caption{A non-dispersive $s=1$ wave-packet of 

6983: the gravitational bouncer, eq.~(\protect\ref{boueq1}). The quantum number 

6984: of the resonant state is chosen as $n_0=1000$, to match typical 

6985: experimental dimensions \protect\cite{steane95}.

6986: The left

6987: column shows the time evolution of the wave-packet 

6988: for $\omega t=0,\pi/2,\pi,3\pi/2$ (from top to bottom).

6989: The right column shows the

6990: corresponding phase space (Husimi, see eq.~(\ref{husimi_def})) 

6991: representation ($z$ axis horizontal as in the left column,

6992: momentum $p$ on the vertical axis).

6993: The parameters are 

6994: $\omega\simeq 0.1487$, $\lambda=0.025$. 

6995: The periodic, nondispersive dynamics of 

6996: the wave-packet bouncing off the mirror in the gravitational field is 

6997: apparent.}

6998: \label{boufig2}

6999: \end{figure}

7000: 

7001: 

7002: \begin{figure}

7003: \centerline{\psfig{figure=bdzf41.ps,width=12cm}}

7004: \caption{Floquet eigenstates anchored to the $\omega=2\Omega$ 

7005: resonance in the gravitational bouncer (left column), in configuration space,

7006: at $\omega t=0$. Each eigenstate exhibits two wave-packets shifted by a phase

7007: $\pi$ along the classical trajectory, to abide the Floquet

7008: periodicity imposed by eq.~(\ref{flstate}). Symmetric and antisymmetric 

7009: linear combinations 

7010: of these states isolate either one 

7011: of the wave-packets, which now evolves precisely like a classical particle

7012: (right column), periodically bouncing off the wall at $z=0$.

7013: The parameters are 

7014: $\omega\simeq 0.2974$,

7015: (corresponding to 

7016: $n_0=1000$, for the $2:1$ resonance), 

7017: $\lambda=0.025$.}

7018: \label{othfig1}

7019: \end{figure}

7020: 

7021: For $s=2$, we 

7022: expect, following the general discussion in section \ref{ORG},

7023:  two quasi-energy levels, separated 

7024: by $\omega/2$, according to the semiclassical result, 

7025: eq.~(\ref{spectrum_harmonic_s}), which are

7026: {\em both} associated with the $s=2$ resonance. 

7027: As a matter of fact, such states are born out from an exact numerical 

7028: diagonalization of the Floquet Hamiltonian derived from eq.~(\ref{boueq1}). 

7029: An exemplary situation is shown in fig.~\ref{othfig1}, for $n_0=1000$ (i.e., 

7030: $\hat{I}_2=1000.75$, in eq.~(\ref{act-s-res})). The tunneling coupling between the 

7031: individual wave-packets shown in the right column of fig.~\ref{othfig1} is 

7032: given by the tunneling splitting $\Delta$ between the energies of both 

7033: associated Floquet states (left column of fig.~\ref{othfig1}) modulo 

7034: $\omega/2$. From the Mathieu approach -- 

7035: eqs.~(\ref{qu_mat_s},\ref{char_s}) --

7036: this tunneling coupling is directly related to the variations of the

7037: Mathieu eigenvalues when the characteristic exponent is changed.

7038: In the limit where the resonance island is big enough, $q\gg 1$ in

7039: eq.~(\ref{qu_mat_s}), asymptotic expressions \cite{abramowitz72}

7040: allow for the following estimate~\cite{holthaus95}:

7041: \be

7042: \Delta=\frac{8\sqrt{2}\lambda^{3/4}}{\pi\sqrt{\omega}}

7043: \exp\left(-\frac{16\pi\sqrt{\lambda}}{\omega^3}\right)=

7044: \frac{8[3(n_0+3/4)]^{1/6}\lambda^{3/4}}{\pi^{4/3}}

7045: \exp\left(-6(n_0+3/4)\sqrt{\lambda}/\pi\right),

7046: \label{otheqn}

7047: \ee

7048: which we can test with our numerically exact quantum treatment. 

7049: The result is shown in fig.~\ref{othfig2}, for two different values of $n_0$.

7050: 

7051: \begin{figure}

7052: \centerline{\psfig{figure=bdzf42.eps,width=14cm,bbllx=0pt,bblly=50pt,bburx=600pt,bbury=400pt}}

7053: \caption{Tunneling splitting between the energies (modulo

7054: $\omega/2$) of two Floquet states of the gravitational bouncer, anchored to the

7055: $s=2$ resonance island, as a function of the driving amplitude 

7056: $\lambda$.

7057: Driving frequency 

7058: $\omega\simeq 1.0825$ (top) and $\omega\simeq 0.8034$ (bottom),

7059: corresponding to resonant states $n_0=20$ and $n_0=50$, respectively. 

7060: The dashed line reproduces   

7061: the prediction of Mathieu theory \protect\cite{holthaus95}, 

7062: eq.~(\protect{\ref{otheqn}}).

7063: Observe that the latter fits the exact 

7064: numerical data only for small values of $\lambda$. At larger 

7065: $\lambda,$ small avoiding crossings between one member of the

7066: doublet and eigenstates originating from other manifolds dominate

7067: over the pure tunneling contribution and the Mathieu prediction 

7068: is not accurate.}

7069: \label{othfig2}

7070: \end{figure}

7071: 

7072: Observe the 

7073: excellent agreement for small $\lambda$, 

7074: with an almost

7075: exponential decrease of the splitting with $\sqrt{\lambda}$, as expected from

7076: eq.~(\ref{otheqn}). However,

7077: for larger values of $\lambda$ (nontheless still in the regime of predominantly

7078: regular classical motion) the splitting saturates and then starts to fluctuate

7079: in

7080: an apparently random way. While the phenomenon has not been completely 

7081: clarified so far, 

7082: we are inclined to attribute it

7083:  to tiny avoided crossings with Floquet states localized is 

7084:  some other resonance islands

7085: (for a discusion

7086:  of related phenomena see

7087: \cite{bonci98,brodier01}).

7088: Comparison of the two 

7089: panels of fig.~\ref{othfig2} additionally indicates that the region of 

7090: $\lambda$ values where 

7091: avoided crossings become important 

7092: increases in the semiclassical limit, and that eq.~(\ref{otheqn})

7093: remains valid for small $\lambda$ only. This is easily understood: in the semiclassical

7094: limit $n_0\rightarrow \infty,$ 

7095: the tunneling splitting decreases exponentially, while the density of states

7096: increases.

7097: 

7098: Finally, fig.~\ref{othfig3} shows a Floquet eigenstate anchored to the $s=11$

7099: resonance island chain. 

7100: \begin{figure}

7101: \centerline{\psfig{figure=bdzf43.ps,width=12cm}}

7102: \caption{Single Floquet state of the gravitational bouncer at phases 

7103: $\omega t=0$ (top) and $\omega t=\pi/2$ (bottom), anchored to

7104: the $11:1$ resonance island chain in classical phase space.

7105: Driving frequency $\omega=11\Omega\simeq 0.6027$, 

7106: $n_0=20000$, and

7107: $\lambda=0.025$. In the upper plot,

7108: among 11 individual

7109: wave-packets which constitute the eigenstate,

7110: five pairs (with partners moving in opposite

7111: directions) interfere at distances $z>0$ from the mirror, whereas the 

7112: 11th wave-packet bounces off the wall and interferes with itself. At

7113: later times (bottom), the 11 wave-packets are well separated in space.}

7114: \label{othfig3}

7115: \end{figure}

7116: It may

7117: be thought of as a linear combination of 11 non-dispersive

7118: wave-packets which, at a given time, may interfere with each other, or,

7119: at another time (bottom panel), are spatially well separated. 

7120: For a Helium atom bouncing off an atom mirror in the earth's gravitational

7121: field (alike the setting in \cite{steane95}),

7122: the $z$ values for such a 

7123: state reach 5 millimeters. This non-dispersive wave-packet is

7124: thus a macroscopic object composed of 11 individual components

7125: keeping a well-defined phase coherence. 

7126: 

7127: 

7128: \subsection{The $s=2$ resonance in atomic hydrogen under linearly 

7129: polarized driving}

7130: \label{ORH}

7131: 

7132: 

7133: Let us now return to the  hydrogen atom driven by LP microwaves.

7134: The highly nonlinear character of the Coulomb interaction favours 

7135: non-dispersive wave-packets anchored to the $s:1$ resonance island, since the

7136: Fourier components $V_s$ of the coupling between the atom and the

7137: microwave decay 

7138: slowly\footnote{The very same behavior characterizes the gravitational

7139: bouncer discussed in the previous section. For the bouncer the slow

7140: inverse square dependence of $V_s$ on $s$ is due to a hard 

7141: collision with the oscillating surface.

7142:  For the Coulomb problem, the singularity at the origin

7143: is even stronger.}

7144:  with $s$ -- compare eqs.~(\ref{v_1dfou},\ref{v_3d_x}).

7145: Consider first the simpler

7146: 1D model of the atom. We  discuss the 

7147: $s=2$ case only, since similar conclusions can be obtained

7148: for higher $s$ values.

7149: The left panel in fig.~\ref{orh_0} shows the classical phase space 

7150: structure (Poincar\'e surface of section) 

7151: for $F_0=0.03$,

7152:  with the $s=2$

7153: resonance completely embedded in the chaotic sea, and well separated from the 

7154: much larger principal resonance island.

7155: \begin{figure}

7156: \centerline{

7157: \psfig{figure=bdzf44.ps,width=14cm}

7158: }

7159: \caption{Left: Poincar\'e surface of section of the one-dimensional hydrogen atom 

7160: under linearly polarized driving, eq.~(\protect\ref{ham_lin1d}), for resonant

7161: driving at twice the Kepler frequency.

7162: The scaled field strength is $F_0=0.03$, 

7163: and the phase is fixed at $\omega t=0.$

7164: The $s=2$ resonance islands are 

7165: apparent, embedded in the chaotic sea, and separated from the $s=1$ resonance

7166: by invariant tori. Right: Husimi representation \protect\cite{abu95a}

7167: of a Floquet eigenstate (for $n_0=60$)

7168: anchored to the $s=2$ resonance displayed on the left.}

7169: \label{orh_0} 

7170: \end{figure}

7171: From our experience with the principal 

7172: resonance, and from the general considerations on $s:1$ resonances above, 

7173: we expect to find 

7174: Floquet eigenstates which are localized on this classical phase space

7175: structure and mimic the temporal evolution of the corresponding 

7176: classical trajectories.

7177: Indeed, the right plot in fig.~\ref{orh_0} 

7178: displays a Floquet eigenstate obtained by 

7179: ``exact" numerical diagonalization, which precisely exhibits 

7180: the desired properties.

7181: 

7182: Again, this observation has its direct counterpart in the realistic 

7183: 3D atom, where the $2:1$ resonance allows for the

7184: construction of non-dispersive wave-packets along elliptic trajectories,

7185: as we shall demonstrate now. 

7186: We proceed as for the $s=1$ case (sec.~\ref{LIN3D}): 

7187: the secular Hamiltonian is obtained by averaging the full Hamiltonian,

7188: eq.~(\ref{hamaa}), after transformation to the ``rotating frame",

7189: eq.~(\ref{trafo_s_b}),

7190: over one period $\tau=sT$ of the resonantly driven classical

7191: trajectory:

7192: \begin{equation}

7193: {\mathcal H}_{\rm sec}=\hat{P}_t-\frac{1}{2\hat I^2}

7194: -\frac{\omega \hat I}{s}

7195: +F\sqrt{1-\frac{M^2}{L^2}}\left[-X_s(\hat{I}) \cos\psi\cos\hat{\theta} + Y_s(\hat{I}) \sin\psi\sin\hat{\theta}\right],

7196: \label{hamscfin_s}

7197: \end{equation}

7198: where $X_s$ and $Y_s$ are given by eqs.~(\ref{xm},\ref{ym}). 

7199: This can be condensed into 

7200: \begin{equation}

7201: {\mathcal H}_{\rm sec}=\hat{P}_t-\frac{1}{2\hat I^2}

7202: -\frac{\omega \hat I}{s}

7203: +F\chi_s\cos(s\hat{\theta}+\delta_s) ,

7204: \label{hamscfin2}

7205: \end{equation}

7206: with

7207: \begin{eqnarray}

7208: \chi_s(\hat{I},L,\psi) & := & \sqrt{1-\frac{M^2}{L^2}}\sqrt{X_s^2\cos^2\psi+Y_s^2\sin^2\psi},

7209: \label{substc2} \\

7210: \tan\delta_s (L,\psi)& := & \frac{Y_s}{X_s}\tan\psi = \frac{J_s(se)\sqrt{1-e^2}}{eJ_s'(se)} \tan\psi.

7211: \label{substb2}

7212: \end{eqnarray}

7213: 

7214: For simplicity, we will now discuss the case $M=0,s=2.$

7215: Fig.~\ref{orh_2} shows the equipotential lines of $\chi_2$

7216: in the $(L,\psi)$ plane, calculated from 

7217: eqs.~(\ref{hamscfin2}-\ref{substb2})\footnote{Since $\chi_s$ scales globally 

7218: as $\hat{I}^2$,

7219: the equipotential lines in fig.~\protect\ref{orh_2} do not depend

7220: on $\hat{I}.$}.

7221: For a comparison with quantum data, the equipotential lines represent

7222: the values of $\chi_2$ for $n_0=42$, quantized from the WKB prescription

7223: in the $(L,\psi)$ plane, exactly as done for the principal

7224: $s=1$ resonance in section~\ref{LIN3D}.

7225:  

7226: \begin{figure}

7227: \centerline{\psfig{figure=bdzf45.ps,width=12cm,angle=-90}}

7228: \caption{Contour plot of the effective perturbation $\chi_2$, 

7229: eq.~(\protect\ref{substc2}), generating the slow evolution of the 

7230: electronic trajectory in the $(L_0=L/\hat{I},\psi)$ plane. The secular

7231: motion in this case is topologically different from that corresponding to

7232: the $s=1$ resonance (compare with fig.~\protect\ref{lin3d_1}). 

7233: In particular, 

7234: there appear new fixed points at $L_0\simeq 0.77,\psi=\pi/2,3\pi/2$ (unstable,

7235: corresponding to unstable elliptic

7236: orbits with major axis perpendicular to the polarization axis) and at

7237: $L_0\simeq 0.65,\ \psi=0,\pi$ (stable, corresponding

7238: to stable elliptic orbits with major axis parallel and antiparallel

7239: to the polarization axis). The resonance island in the $(\hat I, \hat \theta)$ 

7240: plane is quite large for the latter stable orbits, and the associated

7241: eigenstates are

7242: non-dispersive wave-packets localized both longitudinally along

7243: the orbit (locked on the microwave phase), and in the transverse direction, 

7244: see figs.~\protect\ref{orh_5}, \protect\ref{orh_6}, \protect\ref{orh_7}.}

7245: \label{orh_2} 

7246: \end{figure}

7247: 

7248: One immediately notices that the secular

7249: motion is in this case topologically different from that corresponding to

7250: the $s=1$ resonance (compare to fig.~\ref{lin3d_1}), 

7251: with the following features:

7252: \begin{itemize}

7253: \item three different types of motion coexist, with separatrices

7254: originating from the straight line orbits parallel 

7255: ($L_0=0,\psi=0,\pi)$ to the

7256: polarization axis.

7257: \item The straight line orbits perpendicular to the polarization axis

7258: ($L_0=0,\psi=\pi/2,3\pi/2)$ lie at minima -- actually zeros -- of

7259: $\chi_2.$ At lowest order, they exhibit vanishing coupling to 

7260: the external field,

7261: as for the $s=1$ resonance. Hence, the resonance island in the

7262: $(\hat{I},\hat{\theta})$ plane will be small, and the

7263: wave-packets  localized along the corresponding orbits 

7264: are not expected to exist for moderate excitations.

7265: \item The circular orbit (in the plane containing the polarization axis,

7266: $L_0=1,$ arbitrary $\psi$) also exhibits vanishing coupling (since the 

7267: circular motion is purely harmonic, no coupling is possible

7268: at $\omega=2\Omega$).

7269: \item There are ``new" fixed points at $L_0\simeq 0.77,\psi=\pi/2,3\pi/2$

7270: (unstable), and at $L_0\simeq 0.65,\psi=0,\pi$ (stable), corresponding

7271: to elliptical orbits with major axis perpendicular and parallel to the

7272: polarization axis, respectively. The latter ones correspond

7273: to {\em maxima} of $\chi_2$, and are associated with a large

7274: resonance island in the $(\hat{I},\hat{\theta})$ plane. The motion in

7275: their vicinity is strongly confined,

7276: {\it both} in the angular ($L_0,\psi$) and 

7277: in the ($\hat I,

7278: \hat \theta$) coordinates: the corresponding eigenstates can be characterized

7279: as non-dispersive wave-packets, localized both longitudinally along

7280: the orbit (locked on the microwave phase), and in the transverse direction.

7281: \end{itemize}

7282: 

7283: In order to separate quantum states localized in different regions

7284: of the $(L_0,\psi)$ space, we

7285: show in fig.~\ref{orh_3} a comparison between the semiclassical prediction and

7286: the numerically exact 

7287: Floquet energies (obtained as

7288: in section \ref{LIN3D} for the $s=1$ resonance) originating from 

7289: this manifold, with  

7290: $N=0$ in eqs.~(\ref{ebkires},\ref{ebkinres}), at $F_0=0.04$.

7291: Observe that the 16 upmost states appear in eight quasi-degenerate pairs

7292: differing by parity. 

7293: Exact degeneracy does not happen because of 

7294: tunneling effects: the lower the doublet in energy, the larger its

7295: tunneling splitting. The tunneling process 

7296: involved here is a ``transverse" tunneling in the $(L,\psi)$ plane, 

7297: where the electron jumps from the elliptic $(L_0\simeq 0.65,\psi=0)$ 

7298: Kepler 

7299: trajectory to its image under $z$-parity, the $(L_0\simeq 0.65,\psi=\pi)$ 

7300: trajectory (compare fig.~\ref{orh_2}). This tunneling process is entirely due 

7301: to the specific

7302: form of $\chi_2$,

7303: with two {\em distinct} maxima.

7304: 

7305: \begin{figure}

7306: \centerline{\psfig{figure=bdzf46.eps,width=12cm,angle=-90}}

7307: \caption{Comparison of numerically 

7308: exact quasienergies originating from the 

7309: $n_0=42$ manifold (depicted by pluses) to the semiclassical prediction 

7310: (open symbols) based

7311: on the quantization of the $s=2$ resonance island

7312: (microwave frequency $=2 \times$ Kepler frequency), for 

7313: scaled microwave field $F_0=0.04$. Circles

7314: correspond to doubly degenerate states localized in the vicinity of maxima of 

7315: $\chi_2$, around the elliptic fixed points at $(L_0\simeq 0.65,\psi=0,\pi)$

7316: in fig.~\protect\ref{orh_2}. Triangles correspond to almost 

7317: circular states in the vicinity of the stable minimum at 

7318: ($L_0=1,\psi\ \rm arbitrary)$, while diamonds correspond to states localized 

7319: around the stable minima at $L_0=0,\psi=\pi/2,3\pi/2$. 

7320: The agreement between 

7321: the semiclassical and quantum energies is very good, provided the size of the 

7322: resonance island in the $(\hat{I},\hat{\theta})$ plane 

7323: is sufficiently large

7324: (high lying states in the manifold). For low 

7325: lying states in the manifold, the discrepancies between quantum and 

7326: semiclassical 

7327: results are significant, due to the insufficient size of the island.

7328: }

7329: \label{orh_3} 

7330: \end{figure}

7331: 

7332: The energetically highest doublet in fig.~\ref{orh_3} 

7333: corresponds to states

7334: localized as close as possible 

7335: to the fixed points $L_0\simeq 0.65,\ \psi=0,\pi$.

7336: For these states (large resonance island in the ($\hat I, \hat \theta$)

7337: plane),

7338: semiclassical quantization nicely

7339: agrees with the quantum results. On the other hand, the agreement

7340: between quantum and semiclassical results progressively degrades 

7341: for lower energies,

7342: as the size of the  island

7343: in the ($\hat I, \hat \theta$) plane becomes smaller. Still, 

7344: the disagreement between

7345: semiclassical and quantum results is at most of the order of

7346: the spacing between adjacent levels\footnote{

7347: A quantum approach based on the pendulum approximation and 

7348: the Mathieu equation would give a much better prediction

7349: for such states.}.

7350: Below the energy of the unstable fixed points at 

7351: $L_0\simeq 0.77,\psi=\pi/2,3\pi/2$,

7352: there are no more pairs of classical trajectories in the $(L,\psi)$ plane

7353: corresponding to distinct classical dynamics related by $z$-parity.

7354: Hence, the doublet structure 

7355: has to disappear, as confirmed by the

7356: exact quantum results shown in fig.~\ref{orh_3}. On the other hand, there

7357: are two disconnected regions in the $(L,\psi)$ plane which can give rise to 

7358: quantized values of

7359: $\chi_2$ (and, consequently, to quasienergies) within the same quasienergy 

7360: range: the neighbourhood of

7361: the stable fixed points $(L_0=0,\psi=\pi/2,3\pi/2)$, and the region close to

7362: $L_0=1$. In the semiclassical quantization 

7363: scheme, these regions are completely decoupled  

7364: and induce two independent, 

7365: non-degenerate series of quasienergy levels. 

7366: Consequently, the complete spectrum exhibits a rather

7367: complicated structure, caused by the interleaving of these two series.

7368: 

7369: As discussed in section~\ref{ORG}, for a $s:1$ resonance, in a Floquet 

7370: zone of width

7371: $\omega=s\Omega$, there is

7372: not a single manifold of states, but rather a set

7373: of $s$ different manifolds approximately identical and

7374: separated by $\Omega.$

7375: Deviations from the exact $\omega/s$ periodicity are due

7376: to tunneling~\cite{holthaus95}. This tunneling process is however

7377: {\em completely} different from the 

7378: ``transverse" one in the $(L,\psi)$ plane described above. 

7379: It is a case of 

7380: ``longitudinal" tunneling, where the electron jumps from one location

7381: on a Kepler orbit to another, shifted along 

7382: the {\em same} orbit. This longitudinal tunneling is 

7383: similar in origin to the tunneling described in section \ref{OGB}. 

7384: It has to be stressed that it represents a  general

7385: phenomenon in the vicinity of a $s:1$ resonance 

7386: (with $s\ge2)$, due to the

7387: phase space structure in the $(\hat I, \hat \theta$) plane, see section

7388: ~\ref{ORG},

7389:  in contrast

7390: to the ``transverse" quasi-degeneracy (discussed in fig.~\ref{orh_3})

7391: due to the specific form of $\chi_2.$

7392: Inspecting the numerically exact quantum quasienergy spectrum,

7393: we indeed find the manifold 

7394: shown in fig.~\ref{orh_4} (compared to the semiclassical prediction). 

7395: Observe that the agreement between quantum and

7396: semiclassical quasienergies is similar to that observed in fig.~\ref{orh_3},

7397: except for the low lying states. Here, incidentally, the states

7398: anchored to the resonance island are strongly perturbed by another

7399: Rydberg manifold; proper identification of the individual 

7400: quantum states is very difficult in this region, and therefore no quantum 

7401: data are shown at low energies.

7402: 

7403: \begin{figure}

7404: \centerline{\psfig{figure=bdzf47.eps,width=12cm,angle=-90}}

7405: \caption{Same as fig.~\protect\ref{orh_3}, but for the mirror manifold 

7406: shifted in energy by $\omega/2$. While, for most states, the agreement with 

7407: semiclassics is of the same quality as in fig.~\protect\ref{orh_3}, no quantum

7408: data are plotted at the bottom of the manifold. Indeed, at those energies, 

7409: another Rydberg manifold strongly perturbs the spectrum due to close 

7410: accidental degeneracy. Consequently, the unambiguous identification of 

7411: individual states is very difficult.}

7412: \label{orh_4} 

7413: \end{figure}

7414: 

7415: Finally, let us consider the localization properties of the wave-functions 

7416: associated with 

7417: the upmost states of the manifolds in figs.~\ref{orh_3} and \ref{orh_4}. 

7418: These wave-functions should localize in the vicinity of 

7419: stable trajectories of period 2, i.e. they should be 

7420: strongly localized, both in angular and orbital coordinates, 

7421: along an elliptic Kepler orbit of intermediate eccentricity. 

7422: However, because of the longitudinal quasi-degeneracy, we expect the

7423: associated Floquet eigenstates to be composed of

7424: two wave-packets on the ellipse, exchanging their positions

7425: with period $T.$ 

7426: Furthermore, due to the transverse quasi-degeneracy,

7427: we should 

7428: have combinations of the elliptic orbits labeled by 

7429: $\psi=0$ and $\psi=\pi.$ Altogether, this makes four individual

7430: wave-packets represented by each Floquet state. 

7431: Due to the azimuthal symmetry of the problem around the field polarization

7432: axis, each wave-packet actually is doughnut-shaped (compare fig.~\ref{lin3d_8}

7433: for the simpler $s=1$ case).

7434: 

7435: Exact quantum calculations fully confirm this prediction. Fig.~\ref{orh_5}

7436: shows the electronic density of the upmost Floquet state in the

7437: $n_0=42$ manifold (fig.~\ref{orh_3}), averaged over one field period. 

7438: As expected,

7439: it is localized along two symmetric Kepler ellipses ($\psi=0$ and $\psi=\pi$,

7440: respectively), but

7441: longitudinally delocalized because of the time average.

7442: In fact, there are four such Floquet states displaying very similar electronic

7443: densities. These are the energetically highest doublet 

7444: in the $n_0=42$ manifold,

7445: and the upmost doublet in the ``mirror" manifold displayed in fig.~\ref{orh_4}.

7446: 

7447: \begin{figure}

7448: \centerline{\psfig{figure=bdzf48.ps,width=12cm,angle=0}}

7449: \caption{Electronic density of the upmost eigenstate of the $n_0=42$ manifold

7450: of fig.~\protect\ref{orh_3}, averaged over one microwave period. This state 

7451: presents localization along a pair of Kepler ellipses oriented along the field

7452: polarization axis. The box measures $\pm 3500$ Bohr radii in both $\rho$ and 

7453: $z$ directions, with the nucleus at the center. 

7454: The microwave polarization axis along $z$ is 

7455: parallel to the vertical 

7456: axis of the figure. The orientation and eccentricity of the ellipse

7457: are well predicted by the classical resonance analysis.}

7458: \label{orh_5} 

7459: \end{figure}

7460: 

7461: Fig.~\ref{orh_6} shows the electronic densities of these four

7462: Floquet eigenstates at phase $\omega t=0$ of the driving field:

7463: the four doughnuts  are now clearly visible,

7464: as well as the orbital and radial 

7465: localizations along the two elliptic trajectories. 

7466: Very much in the same way as for a double well potential

7467: (or for the bouncer discussed in section \ref{OGB}, compare 

7468: fig.~\ref{othfig1}),

7469: a linear combination 

7470: of these four states allows for the selection of one single doughnut, 

7471: localized along one single classical Kepler ellipse. 

7472: This wave-packet then evolves along this 

7473: trajectory without dispersion, as demonstrated in fig.~\ref{orh_7}. 

7474: \begin{figure}

7475: \centerline{\psfig{figure=bdzf49.ps,width=12cm,angle=0}}

7476: \caption{Electronic densities of the eigenstates of the upmost doublet states 

7477: (top) of the $n_0=42$ manifold of fig.~\protect\ref{orh_3}, and of their mirror

7478: states (bottom), shifted in energy by $\omega/2$ (fig.~\ref{orh_4}), 

7479: at driving field phase $\omega t=0$. 

7480: The longitudinal localization

7481: on the Kepler ellipses (similar for all states) is apparent. On each ellipse,  

7482: four different individual wave-packets (or rather, due to azimuthal symmetry,

7483: two doughnut wave-packets) 

7484: can be distinguished, 

7485: propagating along the Kepler ellipse.

7486: Notice the phase shift of $\pi$ in the temporal evolution on the two ellipses,

7487: implied by $z$-inversion. The microwave polarization axis along $z$ is 

7488: given by the

7489: vertical 

7490: axis of the figure, with the nucleus at the center of the figure.}

7491: \label{orh_6} 

7492: \end{figure}

7493: \begin{figure}

7494: \centerline{\psfig{figure=bdzf50.ps,width=12cm,angle=0}}

7495: \caption{Temporal evolution of a convenient linear combination of the four 

7496: eigenstates 

7497: of fig.~\protect\ref{orh_6}, for phases $\omega t=0$ (top left), 

7498: $\pi/2$ (top center), $\pi$ (top right), $3\pi/2$ (bottom left), $2\pi$ (bottom

7499: right) of the driving field. Clearly, a single doughnut propagating along a 

7500: single trajectory has been selected by the linear combination. This wave-packet

7501: essentially repeats its periodic motion with period $2T=4\pi/\omega$. 

7502: It slowly 

7503: disperses, because the four states it is composed of 

7504: are not exactly degenerate 

7505: (tunneling effect), and because it ionizes (see sec.~\ref{ION}). 

7506: The microwave polarization axis 

7507: along $z$ is 

7508: parallel to the

7509: vertical 

7510: axis of the figure, with the nucleus at the center of the plot.}

7511: \label{orh_7} 

7512: \end{figure}

7513: Note, however, that 

7514: this single wave-packet is {\em not} a single Floquet state, and

7515: thus does not exactly

7516: repeat itself periodically. It slowly disappears

7517: at long times, for 

7518: at least two 

7519: reasons: firstly, because of

7520: longitudinal and transverse tunneling, the phases of the four Floquet

7521: eigenstates accumulate small differences as time evolves, what induces

7522: complicated oscillations between the four possible

7523: locations of the wave-packet, and secondly, 

7524: the ionization rates of the individual

7525: Floquet states lead to ionization and loss of phase coherence,

7526: especially if the ionization rates (see sec.~\ref{ION})

7527: of the four states are not equal.

7528: 

7529: 

7530: 

7531: \section{Alternative perspectives}

7532: \label{ALTP}

7533: 

7534: 

7535: There are several known systems where 

7536: an oscillating field is used to

7537: stabilize a specific mode of motion, such as 

7538: particle accelerators \cite{lichtenberg83}, Paul traps 

7539: \cite{paul92} for ions, etc. 

7540: In these cases, the 

7541: stabilization is a completely

7542: classical phenomenon based on 

7543: the notion of nonlinear resonances. What 

7544: distinguishes our

7545: concept of non-dispersive wave-packets discussed in the preceding chapters 

7546: from those situations is

7547: the necessity to use quantum (or semi-classical) mechanics to describe a given

7548: problem, due to relatively low quantum numbers. Still, the principle of 

7549: localization 

7550: remains the same, and consists in locking the motion of the system

7551: on the external drive.

7552: However, it is not essential that the drive be provided externally, it 

7553: may well be supplied 

7554: by a (large) part of the system to the (smaller) remainder. Note that, 

7555: rather formally, also 

7556: an atom exposed to a microwave field can be understood as one large 

7557: quantum system -- a 

7558: dressed atom, see sec.~\ref{SQ} -- where the field-component 

7559: provides the drive for 

7560: the atomic part \cite{cct92}.

7561: In the present section, we shall therefore briefly recollect 

7562: a couple of related phase-locking phenomena in slightly more complicated 

7563: quantum systems, which open additional perspectives for creating

7564: non-dispersive 

7565: wave-packets in the microscopic world.

7566: 

7567: 

7568: 

7569: \subsection{Non-dispersive wave-packets in rotating molecules}

7570: \label{MOL}

7571: 

7572: A situation closely related to atomic hydrogen exposed to CP microwaves 

7573: (sec.~\ref{CP}) is met when considering the dynamics of a single, 

7574: highly

7575: excited Rydberg electron in a

7576: rotating molecule~\cite{benvenuto94}. 

7577: In \cite{ibb96}, the following  

7578: model Hamiltonian has been proposed:

7579: \be

7580: H=\frac{\vec{p}^2}{2}-\frac{1}{|\vec{r} + \vec{a}(t)|},

7581: \label{molibb0}

7582: \ee

7583: where $\vec{a}(t)$ denotes the position of the center of the Coulomb field

7584: w.r.t. the molecular center of mass, and is assumed to rotate in the 

7585: $x-y$ plane 

7586: with constant frequency $\omega$:

7587: \begin{eqnarray}

7588: {\vec a}(t) & = & \left (

7589: \begin{array}{lc}

7590: \cos\omega t & \ \ -\sin\omega t \\

7591: \sin\omega t & \ \ \cos\omega t

7592: \end{array} \right ){\vec a}.

7593: \label{rotvec}

7594: \end{eqnarray}

7595: In the rotating frame, one obtains the Hamiltonian \cite{benvenuto94} 

7596: \be

7597: H=\frac{p_x^2+p_y^2+p_z^2}{2}-\frac{1}{r} +a\omega^2 x -\omega L_z +

7598: \frac{a^2\omega^2}{2},

7599: \label{molibb}

7600: \ee

7601: which, apart from the constant term $a^2\omega^2/2$, is equivalent to the one 

7602: describing an atom driven by a CP field

7603: (compare eq.~\ref{hrot})).  Note that the

7604: role of the microwave amplitude (which can be arbitrarily tuned in the 

7605: CP problem)

7606: is taken by $a\omega^2$, i.e., a combination of molecular parameters, 

7607: what, of course, 

7608: restricts the

7609: experimental realization of 

7610: non-dispersive wave-packets in the molecule to 

7611: properly selected molecular species \cite{ibb96}. 

7612: 

7613: With the help of the stability analysis outlined in sec.~\ref{rotating_frame}, 

7614: eqs.~(\ref{wp0}-\ref{wpu}), the 

7615: equilibrium position $x_{\mathrm eq}$ 

7616: of the molecular Rydberg electron is easily estimated according to 

7617: (assuming a small value of $a$, limited by the size 

7618: of the molecular core)

7619: \be

7620: x_{\mathrm eq}\simeq ({\cal I}/{\cal J})^{2/3},

7621: \label{x0mol}

7622: \ee

7623:  with ${\cal I}$ 

7624: the molecular momentum of inertia, 

7625: ${\cal J}=\omega{\cal I}$

7626: the rotational quantum number.

7627: To optimize the angular localization of the wave-packet, 

7628: it is necessary that $x_{\mathrm eq}$ 

7629: be sufficiently large 

7630: (from sec.~\ref{RSEF}, $x_{\mathrm eq}\sim n_0^2$, where

7631: $n_0$ is the electronic principal quantum number). Thus, for given ${\cal I}$,

7632: ${\cal J}$

7633: should be 

7634: small. In 

7635: \cite{ibb96}, 

7636: a hydrogen-tritium molecule is considered, which yields 

7637: $n_0\simeq 18$ for ${\cal J}=1$. 

7638: 

7639: Note, however, that such reasoning 

7640: is {\em not} justified. 

7641: The effective Hamiltonian (\ref{molibb0})

7642: implies a {\em classical} description of 

7643: the molecular rotation (much as the {\em classical} treatment 

7644: of the periodic drive in 

7645: eq.~(\ref{h_gen}), with a well-defined phase) defined by the position vector

7646: ${\vec a}(t).$

7647: For such an approach to

7648: be valid, ${\cal J}$ must be sufficiently 

7649: large. The molecular rotation plays

7650: the role of the microwave field in the analogous CP problem,

7651: the number of rotational quanta is just equivalent to the average number of

7652: photons defining the amplitude of the (classical) coherent state 

7653: of the driving 

7654: field. Clearly, if ${\cal J}$ is too small, the  

7655: effect  of an exchange

7656: of angular momentum between the Rydberg electron and the core

7657: on 

7658: the {\em quantum state} 

7659: of the core (and, hence, on $a\omega^2$ assumed to be constant

7660:  in eq.~(\ref{molibb})) cannot be

7661: neglected and, therefore, 

7662: precludes any semiclassical treatment, see also \cite{delande98,shirley65}. 

7663: In other words, if ${\cal J}$ is too small, the number of rotational 

7664: states of the core which are coupled 

7665: via the interaction is too small to mimic a quasi-classical evolution as 

7666: suggested by eq.~(\ref{rotvec}).

7667: 

7668: Nontheless, this caveat does not completely

7669: rule out the existence of 

7670:  molecular non-dispersive wave-packets,

7671: provided a fast

7672: rotation of a core  

7673: with large momentum of inertia (to render $x_{\mathrm eq}$

7674: sufficiently large, eq.~(\ref{x0mol}),

7675: such that the electronic wave-packet gets localized 

7676: far away from the molecular core) can be 

7677: realized, as also suggested

7678: in \cite{ibb96}. 

7679: 

7680: \subsection{Driven Helium in a frozen planet configuration}

7681: \label{HE}

7682: 

7683: 

7684: In the previous examples of non-dispersive wave-packets,

7685: the key point has been the generic appearance of a nonlinear

7686: resonance 

7687: for periodically driven quantum

7688: systems whose  unperturbed dynamics is integrable.

7689: A natural question to ask is whether the concept of non-dispersive wave-packets 

7690: can be generalized to systems which exhibit mixed regular-chaotic dynamics 

7691: even in the absence of the external perturbation. In the atomic realm, 

7692: such a situation is realized for the helium atom, where 

7693: electron-electron interactions provide 

7694: an additional source of 

7695: nonlinearity.

7696: The corresponding Hamiltonian writes in atomic units

7697: \begin{equation}

7698: H_{\rm He} = \frac{{\vec p}_1^2}{2}+\frac{{\vec p}_2^2}{2}-\frac{2}{r_1}

7699: -\frac{2}{r_2}+\frac{1}{|{\vec r}_1-{\vec r}_2|}.

7700: \label{he_free}

7701: \end{equation}

7702: As a matter of fact, the classical and quantum 

7703: dynamics of the three-body Coulomb problem generated by Hamiltonian~(\ref{he_free}) 

7704: has been a largely unexplored

7705: ``terra incognita'' until very recently \cite{tanner00}, since the dimensionality 

7706: of the phase space dynamics increases from effectively two to effectively eight 

7707: dimensions when a second electron is added to the familiar Kepler problem.

7708: Furthermore, the exact quantum mechanical treatment of the helium atom 

7709: remains a formidable task since the early days of quantum mechanics, and 

7710: considerable advances could be achieved only very recently, with the 

7711: advent of modern semiclassical and group theoretical methods 

7712: \cite{wintgen93,gremaud97,gremaudth,puttner01}.

7713: Already the classical dynamics of this system exhibits 

7714: a largely chaotic phase space structure, which typically 

7715: leads to the rapid autoionization 

7716: of the associated {\em doubly excited} 

7717: quantum states of the atom. One of the major 

7718: surprises in the analysis of the three body Coulomb problem during 

7719: the last decades has therefore been the discovery of a new, highly correlated

7720: and classically globally stable electronic configuration, the 

7721: ``frozen planet'' \cite{eichmann90,richter90}. 

7722: The appeal of this configuration resides in its 

7723: counterintuitive, asymmetric character where both electrons are located on the

7724: same side of the nucleus. Furthermore, this configuration turns out to 

7725: be the most robust of all known doubly excited two-electron configurations, in the 

7726: sense that it occupies a large volume in phase space. Its stability is 

7727: due to the strong coupling of the two electrons by the $1/|\vec{r}_1-\vec{r}_2|$ 

7728: term in the Hamiltonian (\ref{he_free}), which enforces their highly correlated motion.

7729: 

7730: The frozen planet is an ideal candidate to test the prevailance of the 

7731: concept of nondispersive wave-packets in systems with intrinsically

7732: mixed dynamics.

7733: In a recent study~\cite{schlagheck98a,schlagheck99,schlagheck99b,schlagheckth} 

7734: the response of this highly correlated two-electron 

7735: configuration to a periodic force has been investigated from a classical

7736: and from a quantum mechanical point of view. 

7737: The Hamiltonian for the driven problem writes, in the length gauge,

7738: \begin{equation}

7739: H=H_{\rm He}+F\cos(\omega t)(z_1+z_2).

7740: \label{he_driven}

7741: \end{equation}

7742: Guided by the experience 

7743: on non-dispersive wave-packets in one electron Rydberg states,

7744: the driving frequency $\omega$ 

7745: was chosen near resonant with the natural frequency

7746: $\Omega_{\rm FP}\approx 0.3n_i^{-3}$ of the frozen planet, where $n_i$ denotes the 

7747: principal

7748: quantum number of the inner electron. It was found that, for a suitably chosen 

7749: driving field amplitude $F$, a nonlinear resonance between the correlated 

7750: electronic motion and the external drive can be induced in the classical 

7751: dynamics, at least for 

7752: the collinear frozen planet where the three particles (two electrons and the

7753: nucleus) are aligned 

7754: along the polarization

7755: axis of the driving field. 

7756: 

7757: However, contrary to the situation for the 

7758: driven hydrogen atom discussed in sections \ref{LIN1D} and \ref{LIN3D}, 

7759: there is a fundamental difference between the one 

7760: dimensional model of the driven three body Coulomb problem

7761: and the full 3D problem. For the one-electron system,

7762: we have seen that the classical Kepler ellipse performs

7763: a slow precession in the angular variables, though

7764: remains bounded and does not ionize. In contrast, if one permits deviations

7765: from collinearity in the driven frozen planet dynamics, it is found that

7766: the transverse

7767: direction is generally unstable and leads to rapid ionization. This transverse

7768: ionization is simply due to the fact that the external field destroys the 

7769: intricate electron electron correlation which creates the unperturbed 

7770: frozen planet. 

7771: Notwithstanding, 

7772: it has been shown that the application of an additional, weak 

7773: static electric field allows to compensate 

7774: for the transverse instability, and to establish a classically globally 

7775: stable dynamical 

7776: situation for the frozen planet. The transverse confinement 

7777: through the static field again justifies the collinear model, 

7778: and first quantum calculations performed 

7779: for this restricted model show the existence of a 

7780: wave-packet associated with the principal resonance between the frozen planet orbit and 

7781: the driving field,

7782: which faithfully traces the classical trajectory at

7783: the period of the drive. 

7784: As for driven one-electron systems, 

7785: these nondispersive two-electron 

7786: wave-packets exhibit life times of typically $10^6$ driving field periods

7787: \footnote{Again, in contrast to the driven one electron 

7788: problem, nothing guarantees that the life times obtained for the 1D model 

7789: carry over

7790: to the real 3D object. On the contrary, first results on the bare 3D Coulomb 

7791: problem 

7792: \protect\cite{schlagheckth} indicate a strong dependence 

7793: of the life times on 

7794: the dimension of the accessible configuration space.}.

7795: 

7796: Hence, there is strong evidence that a resonant external forcing allows

7797: for the creation of quantum eigenstates with a quasi-classical

7798: temporal evolution, even in the presence of strong two-particle correlations.

7799: 

7800: 

7801: \subsection{Non-dispersive wave-packets in isolated core excitation of

7802: multielectron atoms}

7803: \label{LAMBRO}

7804: 

7805: Another example of non-dispersive wave-packets in a two-component 

7806: atomic system has 

7807: recently been proposed for two-electron 

7808: atoms \cite{hanson95,zobay96,mecking98}. 

7809: The scheme

7810: uses an isolated-core excitation in which one of the electrons is transfered

7811: to a Rydberg trajectory 

7812: by a {\em short} laser pulse, forming 

7813: an initially well-localized wave-packet.

7814: A second 

7815: source {\em continuously} drives a transition between two discrete states 

7816: of the remaining atomic 

7817: core. The latter

7818: induces Rabi oscillations (or a coherent superposition) between 

7819: two 

7820: Rydberg series to which the

7821: first electron is

7822: excited. If the 

7823: Rabi 

7824: frequency (controlled by the continuous drive of the core) is matched with the

7825: Kepler frequency of the orbit of  

7826: the 

7827: outer electron,

7828: the autoionization rate of the latter

7829: may be strongly suppressed, provided the respective

7830: phases are also matched properly: 

7831: if 

7832: the electron approaches its 

7833: inner turning radius (where 

7834: the configuration-interaction between Rydberg electron and core -- 

7835: leading to autoionization --

7836: is 

7837: strongest) while 

7838: the core is in its ground state, 

7839: autoionization becomes impossible since the  

7840: configuration-interaction 

7841: does not compensate for the ionization potential of the Rydberg electron. 

7842: On the other

7843: hand, when the electron is far from the nucleus 

7844: (and electron-electron

7845: interaction is weak), the core may be in its excited state, 

7846: without ejecting the Rydberg 

7847: electron.

7848: 

7849: Consequently, autoionization is supressed for the center of the 

7850: Rydberg wave-packet.

7851: During 

7852: time evolution, however, the wave-packet spreads, 

7853: its head and

7854: its tail desynchronize with the Rabi evolution of the core, 

7855: and eventually approach   

7856: the region close

7857: to the nucleus (where configuration-interaction is most pronounced)  

7858: when the core is not

7859: in its ground state. Then these parts of the wave-packet autoionize, 

7860: and the remaining Rydberg population 

7861: is reshaped into a localized wave-packet, since the spreading tails have been 

7862: chopped off.

7863: Hence, these wave-packets exhibit a rather rapid ``melting'' (on a time scale

7864: of at most some hundred Kepler periods) --

7865: to be compared 

7866: to hundreds of thousands or even more Kepler cycles performed 

7867: by non-dispersive 

7868: wave-packets in microwave driven hydrogen atoms studied above (which also

7869: ionize, however {\em very slowly}, see sec.~\ref{ION}).

7870: 

7871: 

7872: The present scenario is in some sense reminiscent of the one 

7873:  in sec.~\ref{MOL}, with a (quantum) two-level core replacing

7874: the rotating molecular core.

7875: As mentioned above, a two-level system alone can only exchange one quantum

7876: with the the outer electron and thus cannot provide an

7877:  exact phase locking mechanism for the highly excited Rydberg electron.

7878: However, the two-level core

7879: is here {\em driven} by an external electromagnetic field and consequently

7880: gains an additional degree of freedom which can be used for the

7881: phase locking mechanism. The drawback is that this phase locking

7882: implies losses (through autoionization). Nevertheless,

7883: the quasi-classical evolution over $\sim 100$ Kepler cycles 

7884: is still quite impressive, and presumably stems from the relatively sharp 

7885: confinement of efficient configuration-interaction within a spatial region close

7886: to the inner turning point of the Rydberg wave-packet.

7887: 

7888: 

7889: 

7890: \section{Characteristic properties of non-dispersive wave-packets}

7891: \label{CHPROP}

7892: 

7893: After presenting several examples of non-dispersive wave-packets in the previous

7894: chapters, we now study their specific properties in more detail. 

7895: Especially, several important physical

7896: processes which may affect the existence of wave-packets 

7897: have so far been hidden under the carpet~\cite{carpetmen}. 

7898: The two most important ones, at least for

7899: driven atoms, are ionization and spontaneous emission, and they will 

7900: be discussed

7901: in detail below. First, let us briefly discuss the general properties of

7902: wave-packet eigenstates 

7903: under the variation of various parameters of 

7904: the driven system 

7905: (e.g.,

7906: microwave amplitude 

7907: and frequency, the strength of 

7908: an external static 

7909: field, etc.).

7910: 

7911: 

7912: \subsection{Ionization rates and chaos assisted tunneling}

7913: \label{ION}

7914: 

7915: 

7916: Atoms driven by microwaves will eventually ionize. Therefore, the

7917: non-dispersive wave-packet states discussed up till now cannot be,

7918: rigorously speaking, discrete states, 

7919: they are rather resonances~\cite{goldberger50} with

7920: some finite life-times. Importantly, as we shall discuss in detail below,

7921: these life-times may be extremely long, of the order of millions of microwave

7922: periods. 

7923: In that sense, they are comparable to 

7924: those of highly 

7925: excited atomic Rydberg states,

7926: which also decay, by spontaneous emission, 

7927: on time scales of few millions of classical periods.

7928: Even more importantly, the life-times of the 

7929: non-dispersive wave-packets are typically orders of

7930: magnitude larger than the life-times of other 

7931: states in the Floquet spectrum: the wave-packets

7932: are particularly resistant

7933: to ionization. This is due to the 

7934: classical confinement

7935: of the

7936: electron inside the regular island. 

7937: To ionize, the electron has no other option but

7938: to tunnel

7939: out of the classically confining island,

7940: before gaining energy by diffusive excitation~\cite{casati88}. 

7941: The resonance island is strictly confining only

7942: for a one-dimensional system. For multi-dimensional systems, the 

7943: tori in the resonance islands are not fully isolating and a very

7944: slow classical diffusion process might eventually lead to ionization. 

7945: This, however, takes place on extremely long time scales and is

7946: completely negligible in atomic systems. 

7947: In practice, ionization of the wave-packet is essentially mediated by a pure

7948: quantum process, exponentially unlikely 

7949: in the

7950: semiclassical limit. As we shall see below, this tunneling

7951: process has quite interesting properties which may be quantitatively

7952: described 

7953: for microwave driven atoms. 

7954: More details 

7955: can be found in \cite{kuba95a,kuba98,kuba98a}. 

7956: 

7957: Due to the initial tunneling step, the life-times of 

7958: non-dispersive wave-packets will typically be 

7959: much longer than those of 

7960: Floquet states 

7961: localized in the chaotic

7962: sea surrounding the island \cite{abu95a}. Moreover, 

7963: since the ionization mechanism involves 

7964: chaotic

7965: diffusion, many quantum mechanical paths link the initial wave-packet to

7966: the final continuum. Thus, the life-time of the wave-packet will reflect 

7967: the interferences between those different possible paths,

7968: and will sensitively depend on parameters such as the microwave frequency or

7969: amplitude, that affect the interfering paths 

7970: through the chaotic sea. 

7971: These fluctuations,

7972: reported first in \cite{kuba95a}, are perfectly deterministic

7973: and

7974: resemble the conductance fluctuations observed in mesoscopic systems 

7975: \cite{stone85} . 

7976: \begin{figure}

7977: \centerline{\psfig{figure=bdzf51.eps,width=8cm,bbllx=100pt,bblly=150pt,bburx=520pt,bbury=600pt,angle=-90}}

7978: \caption{Typical fluctuations of the width (ionization rate)

7979: and of the energy (with respect to its averaged, smooth behavior)

7980: of the non-dispersive wave-packet of a two-dimensional

7981: hydrogen atom in a circularly polarized microwave field.

7982:  The data presented

7983: are obtained for small variations

7984: of the effective principal quantum number $n_0=\omega^{-1/3}$ around 40,

7985: and a scaled

7986: microwave electric field $F_0=0.0426$.

7987: To  show that the fluctuations 

7988: cover

7989: several orders of magnitude, we use a logarithmic vertical scale, and

7990: plot the absolute value of the shift.}

7991: \label{ionfig1}

7992: \end{figure}

7993: In fig.~\ref{ionfig1}, we show the fluctuations of the

7994: ionization rate (width) of the non-dispersive wave-packet

7995: of the two dimensional hydrogen atom in a circularly

7996: polarized microwave field. The energy levels

7997: and widths

7998: are obtained as explained in sec.~\ref{RSEF}, by numerical

7999: diagonalization of the complex rotated Hamiltonian. 

8000: All the data presented in this section have been obtained 

8001: in the regime where the typical ionization rate is smaller than the

8002: mean energy spacing between consecutive levels, so that the

8003: ionization can be thought as a small perturbation acting

8004: on bound states.

8005: The width (although 

8006: very small) displays

8007: strong fluctuations over several orders of magnitude.

8008: Similarly, the real part of the energy (i.e., the center of the

8009: atomic resonance) displays wild fluctuations. The latter

8010: can be observed only if the smooth variation

8011: of the energy level with the control parameter (following 

8012: approximately the semiclassical prediction given

8013: by eq.~(\ref{estharm}))

8014: is substracted.

8015: Therefore, we fitted the numerically obtained energies by a smooth function

8016: and substracted this fit to obtain the displayed fluctuations. Note that

8017: these fluctuations are so small that an accurate fit is 

8018: needed\footnote{In particular, 

8019: the semiclassical expression is not sufficiently

8020: accurate for such  a fit.}. This can be easily seen in fig.~\ref{solit2}

8021: where, on the scale of the mean level spacing, these fluctuations

8022: are invisible by eye (the level appears as a straight horizontal line).

8023: 

8024: The explanation  

8025: for the fluctuations is the following: 

8026: in a quantum language, they are due to the coupling

8027: between the localized wave-packet and states

8028: localized in the chaotic sea surrounding the resonance island. 

8029: While the energy 

8030: of the wave-packet is a smooth function of the

8031: parameters $F$ and $\omega,$  the energies

8032: of the chaotic states  display a complicated behavior characterized

8033: by level repulsion and large avoided crossings. It happens often

8034: that -- for some parameter values -- there is a quasi-degeneracy

8035: between the wave-packet eigenstate and a chaotic state, see the numerous

8036: tiny avoided crossings in fig.~\ref{solit2}. 

8037: There, the two

8038: states are efficiently mixed, the wave-packet captures

8039: some part of the coupling of the chaotic state to the continuum

8040: and its ionization

8041: width increases (see also \cite{abu95a}). 

8042: This is the very origin

8043: of the observed fluctuations. Simultaneously, the 

8044: chaotic state repels the wave-packet state leading to a deviation of the 

8045: energy from its smooth behavior,

8046: and thus to the observed fluctuations. 

8047: This mechanism is similar 

8048: to 

8049: ``chaos assisted tunneling'', described in the literature

8050: \cite{LB90,LB92,grossmann91,GDJH91b,GDJH93,plata92,bohigas93,bohigas93b,tomsovic94,shudo95,leyvraz96}

8051: for both, driven one-dimensional and two-dimensional autonomous systems.

8052: There, the tunneling rate between two symmetric

8053: islands -- which manifests itself through the splitting between

8054: the symmetric and antisymmetric states of a doublet -- 

8055: may be strongly enhanced by the

8056: chaotic transport between the islands. We have then a ``regular''

8057: tunneling escape 

8058: from one island, a chaotic diffusive transport from the vicinity of

8059: one island to the other (many paths, leading to interferences and resulting

8060: in large fluctuations of the splitting),

8061:  and another ``regular'' tunneling penetration into the

8062: second island. In our case, the situation is even simpler -- we have a

8063: ``regular'' tunneling escape supplemented by a chaotic diffusion and

8064: eventual ionization. Thus, instead of the level splitting, we observe 

8065: a shift of the energy level and a finite width.

8066: 

8067: Since these fluctuations stem from the coupling between the regular

8068: wave-packet state and a set of chaotic states, it is quite natural

8069: to model such a situation via a Random Matrix model \cite{kuba98}, 

8070: the

8071: approach being directly motivated by a similar treatment of the

8072: tunneling splitting in \cite{tomsovic94}. For details,  

8073: we refer the 

8074: reader to the original work~\cite{kuba98}. It suffices to say here that

8075: the model is characterized by three real parameters: $\sigma$ -- which

8076: characterizes the mean strength of the coupling between the regular state 

8077: and the chaotic levels, $\gamma$ -- which measures the decay of the chaotic

8078: states (due to ionization;  direct ionization transitions from

8079: the wave-packet state to the continuum are negligible),

8080: and $\Delta$ -- which is the mean level spacing of chaotic levels.

8081: The two physically relevant, dimensionless 

8082: parameters are  $\gamma/\Delta$ and $\sigma/\Delta$.

8083: In the perturbative regime $(\gamma/\Delta,\sigma/\Delta\ll 1)$ it is possible

8084: to

8085: obtain analytical \cite{kuba98} 

8086: predictions for the statistical distribution of the energy

8087: shifts $P(s)$ (of the 

8088: wave-packet's energy from its unperturbed value)

8089: and 

8090: for the distribution of its widths $P(\Gamma)$. $P(s)$ turns out to

8091: be a Cauchy distribution (Lorentzian), similarly to the tunneling

8092: splitting distribution found in \cite{tomsovic94}. The distribution of the 

8093: widths 

8094: is a bit more complicated (it is the square root of $\Gamma$ which

8095: is approximately 

8096: Lorentzian distributed). 

8097: The perturbative approach fails for the 

8098: asymptotic behavior of the 

8099: tails of the distributions, where an exponential cut-off is expected and

8100: observed in numerical studies \cite{kuba98,tomsovic94}. By fitting 

8101: the predictions of the Random Matrix model to the numerical data of 

8102: fig.~\ref{ionfig1}, we may finally extract the values of 

8103: $\gamma/\Delta$, the strength of the decay, and of $\sigma/\Delta,$  

8104: the coupling between the regular and the chaotic states.

8105: \begin{figure}

8106: \centerline{\psfig{figure=bdzf52.eps,width=8cm,bbllx=100pt,bblly=120pt,bburx=320pt,bbury=500pt,angle=-90}}

8107: \caption{The distribution of energy shifts

8108: (a) and 

8109: ionization widths 

8110: (b) for the non-dispersive wave-packet of a two-dimensional hydrogen atom

8111: in a circularly polarized field, obtained by numerical diagonalization

8112: of the Hamiltonian (large bins), compared to 

8113: the random matrix model (small bins). 

8114: Both distributions

8115: are shown on a double logarithmic scale to better visualize the behavior

8116: over a large range of shift and width values. 

8117: Since the energy shift 

8118: may be 

8119: positive or negative, we show the distribution of its modulus. 

8120: The random matrix model fits very well the numerical results,

8121: with both distributions showing 

8122: regions of algebraic behavior followed by an exponential cut-off.}

8123: \label{ionfig2}

8124: \end{figure}

8125: An example of such a fit is shown in fig.~\ref{ionfig2}. 

8126: The numerical data

8127: are collected around some mean values of $n_0$ and $F_0$, typically

8128: 1000 data points were used for a single fit \cite{kuba98}. This

8129: allowed us to study the dependence of the parameters 

8130: $\gamma/\Delta,\sigma/\Delta$ on $n_0$ and $F_0.$

8131: \begin{figure}

8132: \centerline{\psfig{figure=bdzf53.eps,width=8cm,bbllx=100pt,bblly=150pt,bburx=520pt,bbury=600pt,angle=-90}}

8133: \caption{ 

8134: Effective tunneling rate $\sigma/\Delta$  

8135: of the wave-packet (a), 

8136: as a function of the

8137: effective quantum number $n_0=\omega^{-1/3}$

8138: (the inverse of the effective Planck constant), 

8139: for fixed classical dynamics, $F_0=0.0426$. Note the

8140: exponential decrease for sufficiently high $n_0$ (the vertical scale is

8141: logarithmic). The corresponding effective chaotic 

8142: ionization rate $\gamma/\Delta$

8143: (b)

8144: smoothly increases with $n_0$, approximately as $n_0^2$.}

8145: \label{ionfig3}

8146: \end{figure}

8147: The dependence 

8148: on $n_0$

8149: is shown in fig.~\ref{ionfig3}.

8150: Clearly, the tunneling rate $\sigma/\Delta$ decreases 

8151: {\em exponentially} with $n_0.$

8152: Since $n_0$ is the inverse of the effective Planck constant in our problem

8153: (see the discussion in section~\ref{LIN1D} and eq.~(\ref{effective})), 

8154: this shows that

8155: \begin{equation}

8156: \sigma/\Delta \propto \exp \left (-\frac{S}{\hbar_{\mathrm eff}}\right),

8157: \label{dec_exp_tun}

8158: \end{equation}  

8159: where $S$, corresponding to some effective imaginary action 

8160: \cite{shudo95}, is found to be given for our specific choice of

8161: parameters by 

8162:  $S\simeq 0.06 \pm 0.01$ (as fitted from the plot). Such an exponential

8163:  dependence is a hallmark of a tunneling process, thus confirming that

8164:  the wave-packets are strongly localized in the island and communicate

8165:  with the outside world via tunneling.

8166: The $n_0$ dependence of the dimensionless chaotic ionization

8167: rate $\gamma/\Delta$ is very different: it shows a slow, algebraic 

8168: increase with $n_0$. A simple analysis based on a Kepler map

8169: \cite{casati88} description 

8170: would yield a linear increase with $n_0$, whereas our 

8171: data seem to suggest a quadratic function of $n_0$. This discrepancy

8172: is not very surprising, bearing in mind the simplicity of the

8173: Kepler map approach.

8174: 

8175: Similarly, we may study, for fixed $n_0$, the dependence of $\gamma/\Delta$

8176: and $\sigma/\Delta$ on $F_0$, i.e. on the microwave field strength. Such

8177: studies, performed for both linear and circular polarizations, have 

8178: indicated that, not very surprisingly, the chaotic 

8179: ionization rate

8180: $\gamma/\Delta$ increases rather smoothly with the microwave amplitude

8181: $F_0$. On the other hand, the tunneling rate $\sigma/\Delta$ shows

8182: pronounced non-monotonic variations with $F_0,$ see fig.~\ref{ionfig4}.

8183: \begin{figure}

8184: \centerline{\psfig{figure=bdzf54.eps,width=8cm,bbllx=100pt,bblly=150pt,bburx=520pt,bbury=600pt,angle=-90}}

8185: \caption{The tunneling rate $\sigma/\Delta$ (panel (a)) and the chaotic

8186: ionization rate $\gamma/\Delta$ (panel (b)), as a function of the scaled

8187: microwave amplitude $F_0$, for wave-packet eigenstates of a two-dimensional 

8188: hydrogen atom in a circularly polarized

8189: microwave field. 

8190: Observe the oscillatory behavior of the tunneling rate. The bumps

8191: are due to secondary nonlinear resonances in the classical dynamics of the

8192: system.}

8193: \label{ionfig4}

8194: \end{figure}

8195: This unexpected behavior can nontheless be explained \cite{kuba98}. 

8196: The bumps in $\sigma/\Delta$ occur at microwave field

8197: strengths where secondary nonlinear resonances emerge 

8198: within the 

8199: resonance island in

8200: classical phase space. 

8201: For circular polarization, this 

8202: corresponds to some resonance

8203: between two eigenfrequencies $\omega_+$ and $\omega_-$ (see

8204: sec.~\ref{CP} and figs.~\ref{solit2},\ref{solfig}) of the dynamics in the classical resonance island. 

8205: Such resonances 

8206: strongly perturb

8207: the classical dynamics and necessarily affect the

8208: quantum transport 

8209: from 

8210: the island.

8211: 

8212: Let us

8213: stress finally that, even for rather strong microwave fields (say $F_0=0.05$),

8214: where most of the other Floquet states have life-times of few tens or hundreds

8215: of microwave

8216: periods, and irrespective of the polarization of the driving field or of the 

8217: dimension of the accessible configuration space (1D, 2D or 3D), the life-time 

8218: (modulo fluctuations) 

8219: of a non-dispersive wave-packet is typically of the order of

8220: $10^5$ Kepler periods, for $n_0\simeq 60$. This may be used for their possible

8221: experimental detection, see section~\ref{LIFE}. 

8222: 

8223: 

8224: \subsection{Radiative properties}

8225: \label{SPO}

8226: 

8227: 

8228: So far, we have considered the interaction of 

8229: the atom with 

8230: the coherent driving field

8231: only. However, this is not the full story. 

8232: Since the driving field couples

8233: excited atomic states, it remains to be seen to which extent 

8234: spontaneous emission 

8235: (or, more precisely, the coupling to other, initially unoccupied modes of

8236: the electromagnetic field) affects the wave-packet properties. 

8237: This is very important, since the non-dispersive wave-packets

8238: are supposed to be long living objects, and spontaneous

8239: emission obviously limits their life-time.

8240: Furthermore, we have here an example of decoherence effects due to 

8241: interaction with

8242: the environment.

8243: More generally, the interaction of non-dispersive wave-packets

8244: with an additional weak external electromagnetic field may

8245: provide 

8246: a useful tool to probe 

8247: their properties.

8248: In particular, 

8249: their localization within the resonance island

8250: implies that 

8251: an external probe will couple them efficiently 

8252: only to neighboring states within the island. In turn, that   should

8253: make their experimental characterization easy and unambiguous.

8254: Of course, 

8255: external drive (microwave field) and probe must 

8256: not be treated

8257: on the same footing. 

8258: One should rather consider the 

8259: atom 

8260: dressed by the external 

8261: drive

8262: as a strongly coupled system,

8263: or use the Floquet picture described above,

8264: and 

8265: treat the additional mode(s) of the probe (environment) 

8266: as a perturbation. 

8267: We first start with the simplest situation, where a single mode

8268: of the environment is taken into account.

8269: 

8270: \subsubsection{Interaction of a non-dispersive wave-packet

8271: with a monochromatic probe field}

8272: \label{spofloq}

8273: Let us first consider the addition of a monochromatic probe field

8274: of frequency $\omega_p.$ The situation is very similar to the 

8275: probing of a time-independent system by a weak monochromatic field,

8276: with the only difference  

8277: that the Floquet Hamiltonian replaces

8278: the usual time-independent Hamiltonian. Thus, the

8279: weak probe field may induce a transition between two Floquet states

8280: if it is {\it resonant} with this transition, i.e. if

8281: $\hbar \omega_p$ is equal to the quasi-energy difference between

8282: the two Floquet states. According to Fermi's Golden Rule,

8283: the transition probability is proportional to the

8284: square of the matrix element coupling the initial Floquet state

8285: $|{\cal E}_i\rangle$ to the final one $|{\cal E}_f \rangle .$

8286: 

8287: Using 

8288: the Fourier representation of Floquet states,

8289: \begin{equation}

8290: |{\cal E}(t)\rangle = \sum_k{\exp (-ik\omega t) |{\cal E}^k\rangle},

8291: \end{equation}

8292: and averaging over one driving field cycle $2\pi/\omega$ we get

8293: \begin{equation}

8294: \langle {\cal E}_f | {\cal T} | {\cal E}_i\rangle =

8295: \sum_{k=-\infty}^{\infty}

8296: {\langle {\cal E}_f^k | {\cal T} | {\cal E}_i^k \rangle}.

8297: \label{matelm_floquet}

8298: \end{equation}

8299: $\cal T$ denotes the transition operator, usually some

8300: component of the dipole operator depending on the polarization

8301: of the probe beam.

8302: If the quasi-energy levels are not bound states but

8303: resonances with finite life-time (for example because of

8304: multiphoton transition amplitudes to the continuum) 

8305: this approach is easily extended \cite{cct92}, yielding

8306: the following expression for the photoabsorption cross-section of the

8307: probe field at frequency $\omega_p$:

8308: \begin{equation}

8309: \sigma(\omega_p) = \frac{4 \pi \omega_p \alpha}{c} 

8310: {\mathrm Im} \sum_{f}{|\langle {\cal E}_f | {\cal T} | {\cal E}_i \rangle |^2

8311: \left [\frac{1}{{\cal E}_f-{\cal E}_i-\omega_p}+\frac{1}{{\cal E}_f-{\cal E}_i+\omega_p} \right ]}

8312: \label{sigma_probe}

8313: \end{equation}

8314: where $\alpha$ is the fine structure constant, 

8315: and where ${\cal E}_f$ and ${\cal E}_i$ are the complex energies 

8316: of the initial and final

8317: Floquet states. The sum extends over all the Floquet states of the system.

8318: Although the Floquet energy spectrum is itself $\omega$-periodic

8319: (see sec.~\ref{QD}), this is {\em not} the case for the photoabsorption

8320: cross-section. Indeed, the Floquet states at energies ${\cal E}_f$ and

8321: ${\cal E}_f+\omega$ have the same Fourier components, but shifted

8322: by one unit in $k$, resulting in different matrix elements. 

8323: The photoabsorption spectrum is thus

8324: composed of series of lines separated by $\omega$ with unequal intensities.

8325: When the driving is weak, each Floquet state has a dominant Fourier component.

8326: The series then appears as a dominant peak accompanied by side bands

8327: shifted in energy by an integer multiple of the driving frequency $\omega.$

8328: In the language of the scattering theory \cite{cct92,goldberger50,faisal87}, 

8329: these side bands

8330: can be seen as the scattering of the probe photon assisted

8331: by one or several photons of the drive. In any case, the Floquet

8332: formalism is well suited, since it contains 

8333: this weak driving regime as a limiting case, as well as the strong 

8334: driving regime

8335: needed to 

8336: generate a non-dispersive wave-packet.

8337:    

8338: \subsubsection{Spontaneous emission from a non-dispersive wave-packet}

8339: \label{spospo}

8340: We now 

8341: address the situation where no probe field is added to

8342: the microwave field. 

8343: Still, photons of the driving field can be scattered in the 

8344: (initially empty) 

8345: remaining modes of

8346: the electromagnetic field. This is thus some kind of spontaneous emission

8347: or rather resonance fluorescence of the atom under coherent driving. 

8348: It can be seen as spontaneous

8349: emission of the dressed atom, where 

8350: an initial

8351: Floquet state decays spontaneously to another Floquet 

8352: state with a lower quasi-energy,

8353: the energy difference being carried by the spontaneous photon.

8354: As an immediate consequence, the spectrum of the emitted photons

8355: is composed of the resonance frequencies of the Floquet system,

8356: the same that are involved in eq.~(\ref{sigma_probe}).

8357: The decay rate along a transition depends

8358: on the dipole matrix element connecting the initial and the final states,

8359: but also on the density of modes for the emitted photons.

8360: If we consider, for simplicity, the case of free atoms,

8361: one obtains:

8362: \begin{equation}

8363: \Gamma_{if} = \frac{4\alpha^3 ({\cal E}_i-{\cal E}_f)^3}{3} 

8364: |\langle {\cal E}_f|{\cal T}|{\cal E}_i\rangle|^2

8365: \label{gamma}

8366: \end{equation}

8367: where ${\cal E}_i-{\cal E}_f$ is the positive energy difference 

8368: between the initial 

8369: and final Floquet states. As the matrix element of the dipole 

8370: operator $\cal T$

8371: is involved, clearly the localization properties of the Floquet states 

8372: will be of 

8373: primordial importance for the spontaneous emission process.

8374: 

8375: The total decay rate (inverse of the life-time) of a 

8376: state $|{\cal E}_i\rangle$

8377: is obtained by summing the partial rates

8378: $\Gamma_{if}$ connecting the initial state to all states with lower

8379: energy. It is not straightforward to determine

8380: which Floquet states contribute most to the decay rate --

8381: the two factors in eq.~(\ref{gamma}) compete:

8382: while $|\langle {\cal E}_f|{\cal T}|{\cal E}_i\rangle|^2$ tends to favor states localized close

8383: to the initial state (maximum overlap), the factor $({\cal E}_i-{\cal E}_f)^3$

8384: (due to the density of modes in free space)

8385: favors transitions to much less excited states. Which factor wins depends

8386: on the polarization of the driving field.

8387: 

8388: \subsubsection{Circular Polarization}

8389: \label{spocp} 

8390: Consider first a circularly polarized microwave field. 

8391: A first analysis of spontaneous emission

8392: has been given in \cite{ibb97b}, where the rotating frame 

8393: (see sec.~\ref{CP}) approach was used. The driven problem becomes

8394: then time-independent, and the analysis of spontaneous emission 

8395: appears to be simple.

8396: This is, however, misleading, and it is quite easy to omit some

8397: transitions with considerable 

8398: rate. 

8399: The full and correct analysis, both in the rotating and in the standard

8400: frame \cite{delande98},  discusses this problem extensively. The

8401: reader should consult the original papers for details.

8402:  

8403: A crucial point is to realize that the Floquet spectrum of the Hamiltonian

8404: in CP splits into separate blocks, all of  them being identical,

8405: except for a shift by an integer multiple of the driving frequency $\omega.$ 

8406: Each block corresponds to a fixed

8407: quantum number $\kappa=k+M$ where $k$ labels the photon block 

8408: (Fourier component) in the Floquet approach, while $M$ is the azimutal 

8409: quantum number. This merely signifies that the absorption

8410: of a driving photon of circular polarization $\sigma^+$ increases $M$ by

8411: one unit. In other words, $\kappa$ is nothing but the total

8412: angular momentum (along the direction of propagation of the microwave field)

8413: of the 

8414: entire system comprising the atom and the driving field.

8415: The separate $\kappa$ blocks 

8416: are coupled by spontaneous emission.

8417: Since, again, the spontaneously emitted photon carries 

8418: one quantum of angular

8419: momentum, spontaneous emission couples states within the 

8420: same $\kappa$-block (for

8421: $\pi$ polarization of the emitted photon w.r.t. the $z$ axis, 

8422: which leaves $M$ invariant) or 

8423: in neighboring $\kappa$-blocks $\kappa'=\kappa \pm 1$ (see fig.~\ref{spon1}).

8424: \begin{figure}

8425: \centerline{\psfig{figure=bdzf55.eps,width=9cm,bbllx=100pt,bblly=100pt,bburx=420pt,bbury=400pt}}

8426: \caption{Spontaneous emission transitions for the non-dispersive wave-packet

8427: of an atom 

8428: in

8429: a circularly polarized microwave field. The quasi-energy levels of the

8430: Floquet Hamiltonian can be split in series labelled by $\kappa$

8431: (total angular momentum of the atom and of the microwave field). 

8432: The various series

8433: are identical, except for an energy shift equal to an integer multiple

8434: of the microwave frequency $\omega.$

8435: The arrows indicate possible spontaneous transitions 

8436: leaving the initial state $|0,0,0\rangle$.

8437: Only arrows drawn with solid lines are allowed in the harmonic approximation.

8438: The position of the $|0,0,0\rangle$ wave-packet in each Floquet ladder is

8439: indicated by 

8440: the fat lines.}

8441: \label{spon1}

8442: \end{figure}

8443: $\sigma^+$ polarization of the emitted photon gives rise

8444: to higher frequency photons

8445: since -- for the same initial and final Floquet states --

8446: the energy difference in the $\sigma^+$ channel is larger by

8447: $\hbar \omega$ than in the $\pi$ channel (and by

8448: $2\hbar \omega$ than in the $\sigma^-$ channel), as immediately

8449: observed in fig.~\ref{spon1}. As the emission

8450: rate, eq.~(\ref{gamma}), changes with the cubic power of the energy difference,

8451: spontaneous photons with $\sigma^+$ polarization

8452: are expected to be dominant.

8453: 

8454: In the absence of any further approximation, the spontaneous emission

8455: spectrum is  

8456: fairly complicated - it consists of three series with

8457: different polarizations, $\sigma^{\pm}$ and $\pi$.

8458: We may use, however, the harmonic approximation,

8459: discussed in detail in sec.~\ref{CP}. The Floquet

8460: states localized in the vicinity of the stable fixed point

8461: may be labelled

8462: by three quantum numbers $|n_+,n_-, n_z\rangle$, 

8463: corresponding to the various excitations in

8464: the normal modes. The non-dispersive wave-packet we are most interested in

8465: corresponds to the ground state $|0,0,0\rangle$. The dipole operator (responsible for the spontaneous transition)

8466: may be expressed as a linear combination of the creation and annihilation 

8467: operators in these normal modes. Consequently, we obtain strong selection

8468: rules for dipole transitions between $|n_+,n_-,n_z\rangle$ states belonging

8469: to different ladders (at most $\Delta n_i=0,\pm 1$ with not all possibilities

8470: allowed -- for details see \cite{delande98}). The situation is even simpler

8471: for 

8472: $|0,0,0\rangle$,

8473: which may 

8474: decay 

8475: only via three transitions, all 

8476: $\sigma^+$ polarized 

8477: (i.e., from the $\kappa$ block 

8478: to the $\kappa-1$ block):

8479: \begin{itemize}

8480: \item a transition to the 

8481: $|0,0,0\rangle$ state 

8482: in the $\kappa-1$

8483: block. By definition, this occurs precisely at the microwave frequency 

8484: of the drive. One

8485: can view this process as elastic scattering of the microwave photon;

8486: \item a transition to the state $|1,0,0\rangle$,  

8487: at frequency $\omega - \omega_+$;

8488: \item a transition to the state $|0,1,0\rangle$, 

8489: at frequency $\omega + \omega_-$.

8490: \end{itemize}

8491: In the harmonic approximation, 

8492: explicit analytic expressions can be obtained for

8493: the corresponding transition rates \cite{delande98}. It suffices to say 

8494: here that in the semiclassical

8495: limit the elastic component becomes dominant, since its intensity scales as 

8496: $\omega^{5/3}$,

8497: while the intensities of the other two components 

8498: are proportional to $\omega^2,$ i.e., are typically

8499: weaker by a factor $n_0=\omega^{-1/3}$.

8500: This implies that the non-dispersive wave-packet decays exclusively

8501: (in the harmonic approximation) to its immediate neighbor

8502: states, 

8503: emitting a photon with frequency in the microwave range, comparable

8504: to the driving frequency. Direct decay to the atomic $|n=1,L=M=0\rangle$

8505: ground state

8506: or to weakly excited states of the system is forbidden by the selection rules

8507: of the dipole operator. This is easily understood: the CP nondispersive ground

8508: state wave-packet $|0,0,0\rangle$

8509: is built essentially from states with large angular momentum (of the

8510: order of $n_0$), and as it can lose only one unit of angular

8511: momentum per spontaneous emission event, it can decay only to similar

8512: states. When the harmonic approximation breaks down, additional lines

8513: may appear, but, for the same reason, in the microwave range only. 

8514: Another important observation is that the inelastic component at

8515: $\omega + \omega_-$ is by far stronger than the one at $\omega - \omega_+$.

8516: This is entirely due to the cubic power of the transition frequency entering

8517: the expression for the rate (\ref{gamma}). Note the sign difference, due to

8518: the sign difference between $\pm$ modes 

8519: in the harmonic hamiltonian, eq.~(\ref{hh}). 

8520: 

8521: In the semiclassical limit $\omega=n_0^{-3} \to 0,$ the decay is dominated

8522: by the elastic component, and the total decay rate is \cite{delande98}:

8523: \begin{equation}

8524: \Gamma = \frac{2 \alpha^3 \omega^{5/3} q^{-2/3}}{3},

8525: \label{jacks}

8526: \end{equation}

8527: what, multiplied by the energy $\omega$ of the 

8528: spontaneous photon, gives the energy loss 

8529: due to spontaneous emission:

8530: \begin{equation}

8531: \frac{dE}{dt} = \frac{2 \alpha^3 \omega^{8/3}

8532: q^{-2/3}}{3}=\frac{2\alpha^3\omega^4|x_{\rm eq}|^2}{3},

8533: \label{jacks2}

8534: \end{equation}

8535: where we used eq.~(\ref{q}).

8536: This is nothing but the result 

8537: obtained from classical electrodynamics 

8538: \cite{jackson} for a point charge moving

8539: on a circular orbit of radius $|x_{\rm eq}|$ with frequency $\omega$.

8540: Since the charge loses energy, it cannot survive on a circular orbit and

8541: would eventually fall onto the nucleus following a spiral trajectory. 

8542: This model

8543: stimulated Bohr's original 

8544: formulation

8545: of quantum mechanics. 

8546: Let us notice that the non-dispersive

8547: wave-packet is the first physical realization of the Bohr model. There is no

8548: net loss of energy since, in our case, the electron is driven by the microwave

8549: field and an emission at frequency $\omega$ occurs in fact

8550: as an elastic scattering of a microwave photon. Thus

8551: the non-dispersive wave-packet is a cure of the long-lasting

8552: Bohr paradox.

8553: 

8554: Figure \ref{spon2} shows the square of the dipole matrix elements

8555: connecting the non-dispersive wave-packet, for

8556: $n_0=60$ (i.e., microwave frequency $\omega=1/60^3$) and

8557: scaled microwave field 

8558: $F_0=0.04446$, to other Floquet states

8559: with lower energy. These are the results 

8560: of an exact numerical diagonalization of the full Floquet Hamiltonian.

8561: They are presented as a 

8562: stick spectrum because the widths of the important lines are very narrow

8563: on the scale of the figure, which is given as a function of 

8564: the energy difference

8565: between the initial and the final state, that is the frequency of the 

8566: scattered 

8567: photon. Thus, this figure shows the lines that could be observed

8568: when recording the photoabsorption of a weak microwave probe field.

8569: As expected, there is a dominant line at the frequency $\omega$ of the

8570: microwave, and two other lines at frequencies $\omega-\omega_-$

8571: and $\omega+\omega_+$ with comparable intensities, while

8572: all other lines are at least 10 times weaker. This means that

8573: the harmonic approximation works here very well; its predictions,

8574: indicated by the crosses in the figure, are in good quantitative

8575: agreement with the exact result (apart from tiny shifts  

8576: recognizable in the figure,

8577: which correspond to the mismatch between the exact and the 

8578: semiclassical energies 

8579: observed in figs.~\ref{solit2} and \ref{solfig}). This is not completely

8580: surprising as the energy levels themselves are well reproduced

8581: by this harmonic approximation, see sec.~\ref{rotating_frame}.

8582: However, the photoabsorption  spectrum probes the {\em wave-functions} 

8583: themselves 

8584: (through the overlaps) which are well known to be much more sensitive

8585: than the energy levels. The good agreement for both the energy spectrum

8586: and the matrix elements is a clear-cut proof of the

8587: reliability of the harmonic approximation for physically accessible principal

8588: quantum numbers, say $n_0<100$; in fact, it is good down to

8589: $n_0 \simeq 30$, and the non-dispersive wave-packet exists even

8590: for lower $n_0$ values (e.g. $n_0=15$ in \cite{delande95}) although

8591: the harmonic approximation is not too good at such low quantum numbers. 

8592: There were repeated claims in the literature 

8593: \cite{lee97,farrelly95a,farrelly95,brunello96,cerjan97,lee95}

8594: that the stability island 

8595: as well as the effective potential 

8596: are necessarily

8597: unharmonic in the vicinity of the equilibrium point, and that the

8598: unharmonic terms will destroy the stability of the non-dispersive wave-packets.

8599: The present results prove that these claims are doubly wrong:

8600: firstly, as explained in sec.~\ref{rotating_frame}, harmonicity is

8601: {\em not} a requirement for non-dispersive wave-packets to exist 

8602: (the only condition is the existence of a sufficiently large

8603: resonance island); secondly, the harmonic approximation

8604: is clearly a very good approximation even for moderate values

8605: of $n_0.$

8606: 

8607: \begin{figure}

8608: \centerline{\psfig{figure=bdzf56.eps,width=12cm,angle=-90}}

8609: \caption{Square of the dipole matrix element (scaled w.r.t. $n_0$, i.e.

8610: divided by $n_0^2$)

8611: connecting the $|0,0,0\rangle$ non-dispersive 

8612: wave-packet of a three-dimensional hydrogen atom

8613: in a circularly polarized microwave field with

8614: other Floquet states, as a function of the

8615: energy difference between

8616: the two states. The stick spectrum  is the exact result obtained

8617: from a numerical diagonalization 

8618: with $\omega=1/60^3$, corresponding to a principal

8619: quantum number $n_0=60$, and scaled amplitude

8620: $F_0=0.04446$; in natural units, the microwave frequency $\omega/2\pi$ 

8621: is 30.48 GHz,

8622: and the

8623: microwave amplitude 17.6 V/cm. The crosses represent  the

8624: analytic prediction within the harmonic approximation \protect\cite{delande98}.

8625: There are three dominant lines ($\sigma^+$ polarized) discussed in the text,

8626: other transitions (as well as transitions with

8627: $\sigma^-$ or $\pi$ polarizations) are negligible, what proves

8628: the validity of the harmonic approximation. 

8629: If a weak probe field (in the microwave domain) is applied to the system 

8630: in addition

8631: to the driving field, its absorption spectrum should therefore

8632: show the three dominant lines, allowing an unambiguous

8633: characterization of the non-dispersive wave-packet.

8634: }

8635: \label{spon2}

8636: \end{figure}

8637: 

8638: \begin{figure}

8639: \centerline{\psfig{figure=bdzf57.eps,width=12cm,angle=-90}}

8640: \caption{

8641: Same as figure \protect{\ref{spon2}}, but for the

8642: decay rates along the different transitions and in natural units.

8643: The density of modes

8644: of the electromagnetic field completely kills the transition at 

8645: frequency $\omega-\omega_+$,

8646: invisible in the figure. One can see a  small line at frequency approximately

8647: $\omega+2\omega_-$ (see arrow), an indication of a weak breakdown of the harmonic

8648: approximation.

8649: }

8650: \label{spon3}

8651: \end{figure} 

8652: 

8653: Multiplication by the free space density of states transforms fig.~\ref{spon2}

8654: in fig.~\ref{spon3}, which shows that the corresponding 

8655: spontaneous decay rates 

8656: are very low, of the order of 100 Hz at most. They are

8657: few orders of magnitude smaller than the ionization rates and thus may be

8658: difficult to observe. With increasing $n_0,$ the spontaneous rate decreases

8659: algebraically while the ionization rate decreases {\it exponentially}, 

8660: see sec.~\ref{spolp}. 

8661: Thus for large $n_0,$ the spontaneous emission may be the

8662: dominant process. For $F_0\simeq 0.05$ the cross-over may be expected around

8663: $n_0=200$. However, for smaller $F_0,$ the ionization rate decreases

8664: considerably, and for $F_0\simeq 0.03$ both rates become comparable around

8665: $n_0=60$. Still, a rate of few tens of photons (or electrons

8666: in the case of ionization) per second 

8667: may be quite hard to observe experimentally.

8668: 

8669: To summarize, resonance fluorescence of 

8670: non-dispersive wave-packets in 

8671: circularly polarized microwave occurs only in the microwave range

8672: (close to the driving frequency). In 

8673: particular, the elastic component (dominant in the semiclassical limit)

8674: does not destroy the wave-packet, the wave-packet merely converts the 

8675: microwave photon into a photon emitted with the same polarization, but in

8676: a different direction.

8677: Let us stress that we assumed the free space density of modes in this

8678: discussion. Since 

8679: the microwave field may be also 

8680: supplied to the atom by

8681: putting the latter in a microwave cavity, it should be 

8682: interesting to 

8683: investigate how the density of modes in such a cavity affects the

8684: spontaneous emission rate either by increasing

8685: or decreasing it (see \cite{haroche92} for a review)

8686: or, for special cavities (waveguides), even

8687: invalidates the concept of a decay rate \cite{lewenstein88,lewenstein88b}. 

8688:   

8689: \subsubsection{Linearly polarized microwave}

8690: \label{spolp}

8691:   

8692: Let us now discuss the spontaneous emission of non-dispersive wave-packets

8693: driven by a linearly polarized microwave field. 

8694: The situation becomes complicated since we should consider

8695: different wave-packets corresponding to (see fig.~\ref{lin3d_5})

8696: extreme librational states ($p=0$, located perpendicularly to the polarization

8697: axis), separatrix states elongated along the polarization axis, and 

8698: extreme rotational (maximal $p$,  doughnut shaped) states

8699: of the resonantly driven manifold. Clearly, all these wave-packet states have

8700: different localization properties and spontaneous emission will couple

8701: them to different final states. No systematic analysis of the effect has

8702: been presented until now, only results based on the simplified 

8703: one-dimensional model are available \cite{hornberger98}. 

8704: Those are of relevance 

8705: for the spontaneous emission of the separatrix based wave-packet

8706: and are reviewed below. Quantitatively we may, however, expect that

8707: the spontaneous emission properties of the extreme rotational wave-packet

8708: will resemble those of the non-dispersive wave-packet in circular

8709: polarization. Indeed, the linearly polarized wave 

8710: can be decomposed into two circularly

8711: polarized waves, and the extreme

8712: rotational state (fig.~\ref{lin3d_8}) is a coherent superposition of

8713: two circular wave-packets -- each locked on one 

8714: circularly polarized component -- moving in the opposite sense. 

8715: The decay of each

8716: component can then be

8717: obtained from the preceding discussion.

8718:  

8719: A completely different picture emerges for other wave-packets. Their electronic

8720: densities averaged over one period are concentrated along either the $\rho$

8721: (extreme librational) or $z$ axes and do not vanish close to the 

8722: nucleus (fig.~\ref{lin3d_5}).

8723: They have non-negligible dipole elements with Floquet states 

8724: built on low lying

8725: atomic states (i.e. states practically unaffected by the driving field).

8726: Because of the cubic power dependence of the decay rate, eq.~(\ref{gamma}),

8727: on the

8728: energy of the emitted photon, these will dominate the spontaneous emission. 

8729: Thus, in contrast to the CP case, 

8730: spontaneous decay will lead to the destruction of the wave-packet. 

8731: A quantitative analysis confirms this qualitative picture.

8732: To this end, 

8733: a general master equation formalism can be developed \cite{hornberger98}, which

8734: allows 

8735: to treat 

8736: the

8737: ionization process induced by the driving field exactly,

8738: while 

8739: the spontaneous emission is treated perturbatively, 

8740: as in the preceding section. 

8741: Applied 

8742: to the one-dimensional model of the atom (sec.~\ref{LIN1D}), 

8743: with the density of

8744: field modes of the real three-dimensional world, it is possible 

8745: to approximately model 

8746: the behaviour of the separatrix states of the three-dimensional atom.

8747:  

8748: A non-dispersive wave-packet may decay

8749: either by ionization or by spontaneous emission, the total 

8750: decay rate being the sum of the two rates

8751: \cite{hornberger98}. 

8752: Like

8753: in the CP case, the  

8754: decay rate to the atomic continuum decreases exponentially with $n_0$ 

8755: (since it is essentially a tunneling process, see sec.~\ref{ION}, eq.~(\ref{dec_exp_tun}))

8756: while the spontaneous decay rate depends 

8757: algebraically on $n_0$ \cite{bethe77}. The wave-packet is a coherent

8758: superposition of

8759: atomic states with principal quantum number close to $n_0=\omega^{-1/3}$, and

8760: the dipole matrix element between an atomic state $n$ and a weakly 

8761: excited state scales as $n^{-3/2}$\cite{bethe77}. 

8762: Since the energy of the emitted

8763: photon is of order one (in atomic units), eq.~(\ref{gamma})

8764: shows

8765: that the spontaneous emission rate should decrease like 

8766: $\alpha^3/n_0^3.$

8767: The numerical results, presented in fig.~\ref{spon4}, fully confirm

8768: this $1/n_0^3$ prediction.

8769: \begin{figure}

8770: \centerline{\psfig{figure=bdzf58.eps,width=12cm,angle=-90}}

8771: \caption{

8772: Comparison of the spontaneous decay rate, the 

8773: ionization rate, and their sum, for a

8774: non-dispersive wave-packet in a linearly polarized microwave field,

8775: as a 

8776: function of the principal quantum number $n_0$. 

8777: Microwave amplitude $F_0=0.04442,$ decay rates in atomic units. 

8778: The full decay rate exhibits a cross-over 

8779: from a dominantly coherent (ionization)

8780: to a dominantly incoherent (spontaneous emission) regime. 

8781: The fluctuations of the

8782: rate present in the coherent regime are suppressed in the incoherent regime.

8783: The data presented here are obtained by an exact numerical calculation

8784: on the one-dimensional model of the atom \protect\cite{hornberger98}, 

8785: see sec.~\ref{LIN1D}.

8786: }

8787: \label{spon4}

8788: \end{figure}

8789: However, the spontaneous decay of real, 3D wave-packets with near 1D

8790: localization properties (see fig.~\ref{lin3d_5}, middle column, and 

8791: fig.~\ref{linf_wp}) is

8792: certainly slower. 

8793: Indeed, these states are combinations of atomic states

8794: with various total angular momenta $L$; among them, only the

8795: low-$L$ values decay rapidly to weakly excited states, 

8796: the higher $L$

8797: components being coupled only to higher excited states. 

8798: In other words, they are dominantly composed by extremal parabolic Rydberg

8799: states, which have well-known decay properties \cite{bethe77}.

8800: Altogether,

8801: their decay rate is decreased by a factor of the order of $n_0,$  

8802: yielding a $n_0^{-4}$ law instead of  $n_0^{-3}$. 

8803: 

8804: On the other hand, since the ionization process is dominated by tunneling in

8805: the direction of the microwave polarization axis

8806: \cite{abuth,abu98a,abu97,abu97b}, the ionization rate in 3D remains globally

8807: comparable to the ionization rate in 1D, for the wave-packet launched along

8808: straight line orbits. This remains true even if the generic

8809: fluctuations of the ionization rate (see section~\ref{ION}) may induce locally

8810: (in some control parameter) large deviations between individual 3D and 1D

8811: decay rates\footnote{A similar behaviour is observed

8812: in circular polarization for the 2D and 3D non-dispersive wave-packets: they

8813: exhibit comparable ionization rates, but distinct fluctuations

8814: \protect{\cite{kuba95a}}.}. 

8815: Therefore, the transition from dominant ionization to dominant spontaneous

8816: decay will shift to slightly higher values of $n_0$ in 3D.

8817: As in the CP case studied above, this cross-over may be moved to smaller 

8818: values of $n_0$  by reducing the

8819: ionization rate, i.e., by decreasing $F_0$.

8820: 

8821: \subsection{Non-dispersive wave-packet as a soliton}

8822: \label{SOLI}

8823: 

8824: 

8825: 

8826: The non-dispersive character of the wave-packets discussed in this review

8827: brings to mind solitons, i.e. solutions of {\it nonlinear} wave equations

8828: that propagate without deformation: the non-linearity is there

8829: essential to overcome the spreading of the solution. The non-dispersive

8830: wave-packets discussed by us are, on the other hand, solutions of

8831: the {\it linear} Schr\"odinger wave equation, and it is not some

8832: non-linearity of the wave equation which protects them from spreading, but

8833: rather the periodic driving. Thus, at first glance,

8834: there seems to be no link  between both phenomena. This is

8835: not fully correct. One may 

8836: conceive non-dispersive wave-packets as solitonic

8837: solutions of particular nonlinear equations, 

8838: propagating not in time, but in parameter 

8839: space \cite{kuba97b}.

8840: The 

8841: evolution of energy levels in such a space, 

8842: called ``parametric level dynamics'',

8843: has been extensively studied (see 

8844: \cite{haake90,nakamura93} for 

8845: reviews), both for time-independent

8846: and  for periodically time-dependent systems. In the latter case, 

8847: the energy levels are the quasi-energies 

8848: of the Floquet Hamiltonian (see sec.~\ref{QD}). 

8849: 

8850: For the sake of simplicity, 

8851: we consider here the two-dimensional hydrogen atom 

8852: exposed to a circularly polarized microwave (sec.~\ref{2d_model}),

8853: where the explicit time dependence can be removed by 

8854: transforming to

8855: the rotating frame (see sec.~\ref{rotating_frame}), but completely

8856: similar results are obtained for the Floquet Hamiltonian of any 

8857: periodically time-dependent system.

8858: The Hamiltonian, given by eq.~(\ref{hrot}),

8859: \be

8860: H=\frac{{\vec p}^2}{2}-\frac{1}{r}+Fx-\omega L_z,

8861: \label{solieq0}

8862: \ee

8863: may be thought of as an example of a generic system of the form

8864: \be

8865: H(\lambda)=H_0+\lambda V,

8866: \label{solieq1}

8867: \ee

8868: where $\lambda$ is a parameter. In our case, 

8869: for example, the

8870: microwave amplitude may be tuned, leading to $V=x$ and $\lambda=F$. 

8871: The interesting quantities are then 

8872: the eigenvalues $E_i(\lambda)$ and the eigenfunctions

8873: $|\psi_i(\lambda)\rangle$ of eq.~(\ref{solieq1}). 

8874: Differenciating the

8875: Schr\"odinger equation with respect to $\lambda$,

8876: one shows (with some algebra)

8877: \cite{haake90} that the behavior of $E_i(\lambda)$ with

8878: $\lambda$ may be viewed as the motion of $N$ fictitious

8879: classical particles

8880: (where $N$ is

8881: the dimension of the Hilbert space) with positions $E_i$ and momenta

8882: $p_i=V_{ii}=\langle\psi_i|V|\psi_i\rangle$, governed by the

8883: Hamiltonian

8884: \begin{equation}

8885: {\cal H}_{\rm cl}=\sum_{i=1}^N\frac{p_i^2}{2}+\frac{1}{2}\sum_{i=1}^N\sum_

8886: {j=1,j\neq i}^N

8887: \frac{\mid{\cal L}_{ij}\mid^2}{(E_j-E_i)^2},

8888: \label{hclas}

8889: \end{equation}

8890: where

8891: ${\cal L}_{ij}=(E_i-E_j)\langle\psi_i|V|\psi_j\rangle$

8892: are additional independent variables obeying the general Poisson brackets

8893: for angular momenta. The resulting dynamics,

8894: although nonlinear, is integrable \cite{haake90}.

8895: 

8896: Let us now consider the parametric motion of some eigenstate 

8897: $\mid n_+,n_- \rangle$, 

8898: for example of the ground state wave-packet $|0,0\rangle$.

8899: Its coupling to other states is quite weak -- because

8900: of its localization in a well defined region of phase space -- 

8901: and the corresponding ${\cal L}_{ij}$ are consequently very small.

8902: If we 

8903: first suppose

8904: that the wave-packet state is well isolated (in energy)

8905: from other wave-packets (i.e., states with low values of $n_+,n_-$), 

8906: the fictitious particle associated with 

8907: $|0,0\rangle$ 

8908: basically ignores the other 

8909: particles and propagates freely at constant

8910: velocity. It preserves its properties across the

8911: successive interactions with neighboring states, 

8912: in particular its shape:

8913: in that sense, it is a solitonic solution of the

8914: equations of motion generated by Hamiltonian~(\ref{hclas}).

8915: 

8916: Suppose that, in the vicinity of some $F$ values, 

8917: another wave-packet 

8918: state (with low $n_+,n_-$ quantum numbers) becomes quasi-degenerate 

8919: with $|0,0\rangle$. 

8920: In the harmonic approximation, see sec.~\ref{rotating_frame}, the two states

8921: are completely uncoupled; 

8922: it implies that the corresponding

8923: ${\cal L}_{ij}$ vanishes and the two levels cross. 

8924: The coupling

8925: between

8926: the two solitons 

8927: stems from the {\em difference} between the exact hamiltonian

8928: and its harmonic approximation, 

8929: i.e. from third order or higher terms, beyond the harmonic

8930: approximation. Other

8931: $|n_+,n_-\rangle$ states having different slopes w.r.t. $F$ induce 

8932: ``solitonic collisions'' 

8933: at some other values of $F$.  

8934: To illustrate the effect,

8935: part of the spectrum of the two-dimensional hydrogen atom in 

8936: a CP microwave

8937: is shown in fig.~\ref{solfig}, as a function of the scaled microwave

8938: amplitude $F_0$.

8939: For the sake of clarity, the

8940: energy of the ground state wave-packet 

8941: $|0,0\rangle$ calculated

8942: in the harmonic approximation, eq.~(\ref{estharm}) is substracted,

8943: such that it appears as an almost

8944: horizontal line.

8945: Around $F_0=0.023$, it is crossed by another solitonic solution, 

8946: corresponding to

8947: the $|1,4\rangle$ wave-packet\footnote{Similarly to the 

8948: wave-packet $|1,3\rangle$ discussed in fig.~\protect\ref{solit2},

8949: the semiclassical harmonic

8950: prediction for the energy of $|1,4\rangle$ is not satisfactory.

8951: However, the slope of the energy level is well reproduced as a 

8952: function of $F_0$.}, 

8953: what represents the collision

8954: of two solitons. Since this

8955: avoided crossing 

8956: is narrow and well isolated from

8957: other avoided crossings, the wave-functions before and after 

8958: the crossing preserve

8959: their shape and character, as typical for an isolated two-level system.

8960: This may be further verified by wave-function plots before and after

8961: the collision (see 

8962: \cite{kuba97b} for more details). 

8963: \begin{figure}

8964: \centerline{\psfig{figure=bdzf59.ps,width=10cm}}

8965: \caption{

8966: The quasi-energy spectrum of a two-dimensional hydrogen atom 

8967: in a circularly polarized microwave field,

8968: as a function of the scaled microwave amplitude $F_0$ (for

8969: $n_0=60$). In order to emphasize the dynamics of wave-packet states,

8970: the semiclassical prediction, eq.~(\protect\ref{estharm}) for the

8971: ground state wave-packet energy is substracted from the

8972: numerically calculated energies. Consequently, the

8973: ground state wave-packet $|0,0\rangle$ is represented by 

8974: the almost horizontal line. 

8975: The dashed line represents the

8976: semiclassical prediction for the $|1,4\rangle$ wave-packet. Although it is 

8977: rather far from the exact result, the slope

8978: of the energy level is well reproduced. The size

8979: of the avoided crossing between the ``solitonic"

8980: levels $|0,0\rangle$ and  $|1,4\rangle$ is a direct measure

8981: of the failure of the harmonic approximation.

8982:  } 

8983: \label{solfig}

8984: \end{figure}

8985: The avoided crossings become larger 

8986: (compare fig.~\ref{solit2}) with increasing $F_0$. In fact, as mentioned 

8987: in sec.~\ref{rotating_frame},

8988: we have numerically verified that the solitonic character of the ground

8989: state wave-packet practically disappears at

8990: the  $1:2$ resonance, close to $F_0\simeq 0.065$ \cite{kuba97b}. For larger

8991: $F_0,$ while one may still find nicely

8992: localized wave-packets for {\em isolated} values of $F_0,$ 

8993: the increased size of the avoided crossings makes it difficult

8994: to follow the wave-packet 

8995: when sweeping $F_0.$ 

8996: For such strong fields, the ionization rate of

8997: wave-packet states becomes appreciable, comparable to the level

8998: spacing between consecutive states and the simple solitonic model breaks down.

8999: In order to understand the variations of the 

9000: (complex) energies of the resonances with $F$, a slightly

9001: more complicated model -- level dynamics in the complex plane --

9002: should be used~\cite{haake90}.

9003: 

9004: 

9005: 

9006: 

9007: \section{Experimental preparation and detection of non-dispersive wave-packets}

9008: \label{EXP}

9009: 

9010: In the preceding chapters, we have given an extensive

9011: theoretical description of the characteristic 

9012: properties of non-dispersive wave-packets in driven Rydberg

9013: systems. We have seen that these surprisingly robust ``quantum particles''

9014: are ubiquitous in the interaction of electromagnetic radiation with 

9015: matter. However, any theoretical analysis needs to be confronted with 

9016: reality, and we have

9017: to deal with the question of creating and identifying non-dispersive

9018: wave-packets in a laboratory experiment.

9019: In our opinion, none of the currently operational experiments on the

9020: interaction of Rydberg atoms with microwave fields 

9021: allows for an unambiguous identification of non-dispersive wave-packets,

9022: although some of them \cite{koch95b} 

9023: certainly have already populated such states.

9024: In the following, we shall therefore start out with 

9025: a brief description of the typical approach of state of the art experiments, 

9026: and subsequently extend on various alternatives to create and to probe  

9027: non-dispersive wave-packets in a real experiment.

9028: We do not aim at a comprehensive 

9029: review on the interaction 

9030: of Rydberg atoms with microwave fields, but rather refer to 

9031: \cite{abuth,delande94,sachath,koch95b,casati87b}

9032: for a detailed treatment of various aspects of this intricate

9033: problem. Here, we strictly focus on

9034: issues pertinent to our specific purpose.

9035: 

9036: \subsection{Experimental status}

9037: \label{EXSQ}

9038: 

9039: 

9040: The theoretical interest in the interaction of Rydberg states of atomic

9041: hydrogen

9042: with low frequency electromagnetic

9043: fields has been triggered by early experiments \cite{bayfield74}

9044: which showed a surprisingly

9045: efficient excitation and subsequent ionization of the atoms

9046: by the field. More precisely, a microwave field

9047: of frequency $\omega$

9048: comparable to the energy

9049: difference between the initial atomic state and its nearest neighbor

9050: was observed to induce appreciable ionization, 

9051: for atom-field interaction times of 

9052: approx. $100$ driving field cycles, and for 

9053: field amplitudes 

9054: beyond a certain threshold value

9055: (of the order of $5-10\%$ of the

9056: Coulomb field experienced by the Rydberg electron on its unperturbed

9057: Kepler orbit). This threshold behavior of the

9058: ionization probability as a function of the driving field amplitude

9059:  rather than

9060: of the driving frequency -- in apparent contradiction to the photoeffect --

9061: motivated a theoretical analysis of the classical dynamics of the Rydberg

9062: electron under external driving. It turned out that

9063: the ionization threshold marks the transition

9064: from  regular

9065: to  chaotic {\em classical dynamics} of the driven electron \cite{leopold78}.

9066: 

9067: The microwave ionization of atomic Rydberg states was thus identified

9068: as

9069: an experimental testing ground for quantum transport

9070: under the conditions of classically mixed regular chaotic

9071: dynamics, where the transport was simply measured by the experimentally

9072: observed ionization yield, or -- with some additional experimental effort --

9073: by the time dependent redistribution of the atomic population over the

9074: bound states \cite{bayfield88a,bluemel89a,bluemel91a,abu91}. 

9075: Depending on the precise value of the scaled frequency $\omega_0$ -- the ratio

9076: of the microwave frequency $\omega$ 

9077: to the Kepler frequency $\Omega_{\rm Kepler}$

9078: of the initially excited Rydberg atom, eq.~(\ref{omega0}) --

9079: of the driving field, theory soon predicted essentially

9080: ``classical'' ionization yields ($\omega_0<1.0$), or some quantum suppression

9081: of chaotic ionization ($\omega_0>1.0$) 

9082: \cite{casati84}, 

9083: mediated by the quantum mechanical

9084: interference effect known as 

9085: {\em dynamical localization}, 

9086: analogous to 

9087: Anderson localization in the electronic transport

9088: through disordered solids  

9089: \cite{fishman82,grempel84,brenner96,abu98b,wimbergerda,wimberger01}.

9090: The physical process involved in chaotic ionization is 

9091: classically deterministic

9092: diffusion, 

9093: therefore essentially statistical in nature, and 

9094: insensitive to the details of the transport process.

9095: Correspondingly, the mere ionization probability

9096: condenses all details of the ionization process in one single number,

9097: without revealing details on 

9098: individual local

9099: structures in phase space. It reflects the statistical characteristics

9100: of the excitation process, rather than the population of some well defined

9101: individual atomic  states in its course \cite{abuth,abu95a}.

9102: Hence, state of the art experiments are ``blind'' for the details of the

9103: atomic excitation process on the way to ionization, and therefore not

9104: suitable for the unambiguous identification of individual eigenstates of the

9105: atom in the field, notably of non-dispersive wave-packets. 

9106: The case is getting

9107: worse with additional complications which are unavoidable in a real experiment,

9108: such as the unprecise definition of the initial state the atoms are prepared

9109: in 

9110: \cite{koch95b,galvez88,sirko96,sirko93a,sauer92b,leeuwen85,koch95a,koch92b,koch92a,koch89}, 

9111: the experimental uncertainty on the envelope of the amplitude of the

9112: driving field experienced by the atoms as they enter the interaction

9113: region with the microwave 

9114: (typically a microwave cavity or wave guide) \cite{bluemel91a,sauer92b,noel00},

9115: stray electric fields due to contact potentials in the interaction region,

9116: and finally uncontrolled noise sources which may affect the

9117: coherence effects involved in the quantum mechanical transport process

9118: \cite{bayfield91}.

9119: On the other hand, independent experiments on the microwave ionization

9120: of Rydberg states of atomic hydrogen \cite{bayfield89,galvez88}, 

9121: as well as on hydrogenic initial states 

9122: of lithium \cite{noel00}, did indeed provide hard evidence

9123: for the relative stability of the atom against ionization when driven by a

9124: resonant field of scaled frequency $\omega_0\simeq 1.0$.

9125: Furthermore, in the hydrogen experiments, this stability was observed 

9126: to be insensitive

9127: to the polarization of the driving field, be it linear,

9128: circular or elliptical \cite{bellermann96}.

9129: These experimental findings suggest that some atomic dressed states 

9130: anchored to

9131: the principal resonance island in the classical phase space

9132: are populated by switching on the microwave field,

9133: since these states  tend to be more stable

9134: against ionization than  states localized in the chaotic sea~\cite{abu95a}, 

9135: see section \ref{ION}.

9136: 

9137: Consequently, for an unambiguous preparation and identification of

9138: non-dispersive wave-packets launched along well defined classical trajectories

9139: the experimental strategy has to be refined. 

9140: We suggest

9141: two 

9142: techniques for their preparation:

9143: \begin{itemize}

9144: \vskip -10pt

9145: \item The direct, selective optical excitation from

9146: a low lying state {\em in the presence of the microwave field}. 

9147: This approach actually 

9148: realizes some kind of  

9149: ``Floquet state absorption spectroscopy'' \cite{abu96}.

9150: \item The preparation of the appropriate atomic initial 

9151: state -- optionally in the presence of a static field --

9152: followed by  

9153: switching  

9154: the microwave field on the appropriate time scale 

9155: to the desired maximum field amplitude \cite{kuba95b}.

9156: \end{itemize}

9157: 

9158: Two, possibly complementary methods should allow 

9159: for an efficient detection of such wave-packets: 

9160: \begin{itemize}

9161: \item Floquet spectroscopy  -- this time 

9162: involving either microwave or optical transitions between states dressed

9163: by the microwave field;

9164: \item Measurements of the time-dependence of the ionization yield 

9165: of the non-dispersive

9166: wave-packet. This 

9167: requires the ability to vary the

9168: interaction time between the atoms and the microwave by more 

9169: than one order of magnitude

9170: \cite{abuth,abu95a,noel00,abu95b,arndt91,benson95}.

9171: \end{itemize}

9172: 

9173: All these techniques 

9174: are experimentally well-developed and actually realized in different, 

9175: currently

9176: operational experimental settings \cite{noel00,abu95b,noel00b}.

9177: The only prerequisite for an unambiguous identification of non-dispersive 

9178: wave-packets therefore remains an experimental set-up which allows 

9179: to follow these

9180: complementary strategies simultaneously.

9181: 

9182: 

9183: \subsection{Direct preparation}

9184: \label{SPEC}

9185: 

9186: 

9187: The 

9188: most straightforward way to populate a non-dispersive

9189: wave-packet state is its

9190: direct optical excitation

9191: in the presence of the driving field,

9192: from a 

9193: weakly excited state of the system at energy $E_0$.

9194: In section \ref{spofloq}, we have discussed how a weak electromagnetic probe 

9195: field can induce transitions between Floquet states.

9196: This is particularly easy if  

9197: one of the states involved is in an energetically low

9198: lying state. 

9199: Such a state

9200: is practically unaffected by the driving

9201: microwave field (which is weak as compared to the Coulomb field 

9202: experienced by a deeply 

9203: bound state, and very  far

9204: from any resonance), such that the corresponding Floquet state is almost

9205: exactly 

9206: identical with the time-independent atomic state. In other words,

9207: all the Fourier components of the Floquet state vanish,

9208: except the $k=0$ component, 

9209: which represents the 

9210: unperturbed atomic state

9211: $|\phi_0\rangle$.

9212: In such a case, the photoexcitation cross-section (\ref{sigma_probe}) becomes

9213: (neglecting the anti-resonant term):

9214: 

9215: \begin{equation}

9216: \sigma(\omega_p) = \frac{4 \pi \omega_p \alpha}{c}

9217: {\mathrm Im} \sum_{f}{|\langle {\cal E}_f^0 | {\cal T} | \phi_0 \rangle |^2

9218: \frac{1}{{\cal E}_f-E_0-\omega_p}},

9219: \label{sigma_probe_simple}

9220: \end{equation}

9221: where the sum extends over all Floquet states with energy ${\cal E}_f$ and

9222: involves only their $k=0$ Fourier component

9223: \footnote{Alternatively, the sum could be rewritten as a sum

9224: over one Floquet zone only, with all the Fourier

9225: components $|\phi_f^k\rangle$ involved, with the denominator

9226: replaced by ${\cal E}_f+k\omega-E_0-\omega_p.$}.

9227: 

9228: Eq.~(\ref{sigma_probe_simple}) shows that the excitation probability 

9229: exhibits a maximum any

9230: time the laser is scanned across a frequency which is resonant with

9231: the transition from the 1s ground state $|\phi_0\rangle$ 

9232: to a specific dressed state of the atom

9233: in the field. Fig.~\ref{abs_spec} shows

9234: an example for the photoabsorption probability from the ground state

9235: of atomic hydrogen in the presence of 

9236: a microwave field and a parallel static electric field. The microwave

9237: frequency is resonant with

9238: atomic transitions in the region of $n_0\simeq 60$.

9239: Clearly, the

9240: cross-section shows extremely narrow peaks each of which corresponds to

9241: a Floquet eigenstate of the atom in the field. As a matter of fact, the

9242: state marked by the arrow is similar to the dressed state of the 3D atom

9243: displayed in fig.~\ref{linf_wp}, 

9244: a wave-packet periodically moving along the

9245: field polarization axis. As obvious from the figure, this state can be

9246: efficiently reached by direct excitation from the ground state.

9247: Furthermore, due to its sharp signature in

9248: $\sigma(\omega_p)$, it is easily and unambiguously identified.

9249: On the other hand, this kind of preparation of the wave-packet is

9250: obviously reserved to those dressed states 

9251: which

9252: have nonvanishing overlap with the deeply bound atomic states. 

9253: For wave-packets tracing circular or elliptical orbits far from the

9254: nucleus, another strategy is needed, that

9255: is discussed below.

9256: 

9257: \begin{figure}

9258: \centerline{\psfig{figure=bdzf60.eps,width=9cm,angle=-90}}

9259: \caption{Photo-excitation of highly excited Rydberg states of the

9260: hydrogen atom in the presence of a linearly polarized microwave field

9261: of frequency $\omega/2\pi=30.48$ GHz and amplitude

9262: $F= 7.93$ V/cm and a parallel static electric field $F_s=2.38$ V/cm.

9263: The initial state is the ground state of the atom, and the

9264: polarization of the probe beam is parallel to the static

9265: and microwave fields. The spectrum 

9266: is displayed

9267: in the region where Floquet states are mainly composed of Rydberg states

9268: with principal quantum number around $n_0=60$, while the microwave driving 

9269: is resonant with the Kepler frequency of such states. Hence, some of

9270: the Floquet states are trapped in the non-linear resonance island

9271: and behave as non-dispersive wave-packets. Of special interest

9272: is the state marked with an arrow, which is similar to the non-dispersive

9273: wave-packet displayed in fig.~\protect{\ref{linf_wp}}, localized

9274: in all three dimensions of space (the field amplitude is

9275: slightly different). Its large photo-excitation probability

9276: should make a direct experimental preparation possible. At the scale of the figure,

9277: the width of the various lines is very small, and the spectrum is

9278: almost a pure stick spectrum.

9279: }

9280: \label{abs_spec}

9281: \end{figure}

9282: A slightly adapted spectroscopic approach should 

9283: be equally 

9284: useful for the unambiguous identification of wave-packet eigenstates.

9285: Instead of probing the dressed spectrum from a weakly excited

9286: state using a laser field, one may equally well probe the local structure

9287: of the dressed spectrum in the vicinity of the wave-packet eigenstate

9288: by inducing transitions from the wave-packet to neighboring states

9289: by a second, weak microwave field of linear or circular 

9290: polarization \cite{bluemel90a}.

9291: Such stimulated transitions will be mediated by the dipole matrix elements

9292: given in equation~(\ref{matelm_floquet}), and allow to measure

9293: the energy spacings in the

9294: immediate vicinity of the wave-packet state directly. Hence, microwave probe

9295: spectroscopy should be an extremely

9296: sensitive probe, since it allows for the unambiguous

9297: identification of the wave-packet via the characterization of its local

9298: spectral environment.

9299: 

9300: Given the spectroscopic

9301: resolution which is nowadays available in the optical as well as in the

9302: microwave domain, the spectroscopic approach outlined above seems to be the method

9303: of choice for an unambiguous identification, and -- where possible --

9304: for an efficient launch of nondispersive wave-packets along a periodic

9305: orbit of the classical dynamics. What it requires, however, is a

9306: precise determination of Floquet spectra from the accompanying

9307: quantum calculation. Fortunately, both for hydrogen and for alkali

9308: atoms, the necessary theoretical quantum data may be obtained from 

9309: already existing

9310: software \cite{krug00,krug01}.

9311: 

9312: Although we elaborated in this review paper only the case of the hydrogen

9313: atom, the general concepts are also fruitful for non-hydrogenic atoms.

9314: Indeed, the major difference between the Rydberg electron in a hydrogen

9315: atom and in a non-hydrogenic atom is the existence in the latter case

9316: of an ionic core which affects the classical and quantum dynamics of the

9317: Rydberg electron. On the scale of a Rydberg atom, the ionic core is 

9318: a extremely small object which will thus induce a very local perturbation.

9319: As long as the Rydberg electron does not approach the ionic core,

9320: it behaves completely similarly in hydrogen or non-hydrogenic atoms.

9321: Thus, the properties of non-dispersive wave-packets tracing circular

9322: or elliptical classical orbits are essentially independent of the ionic core,

9323: and the hydrogenic analysis holds. 

9324: For orbits which come close to the nucleus, the ionic core may scatter

9325: the Rydberg electron. Thus, instead of being indefinitely trapped on a

9326: torus inside a resonance island, it may happen that the Rydberg electron

9327: hops from a torus to another one when it gets close to the nucleus.

9328: This of course will affect the long time classical and quantum dynamics.

9329: Nevertheless, it remains true that most of the time the classical

9330: dynamics -- and consequently the phase locking phenomenon responsible

9331: for the existence of non-dispersive wave-packets -- is identical

9332: to the hydrogenic dynamics. From the quantum point of view, the ionic

9333: core is responsible for the existence of non-zero quantum defects

9334: in the low angular momentum channels. The energy levels, mixed by the microwave

9335: driving, will thus be significantly shifted from their

9336: hydrogenic positions. However, the {\em structure} of the energy levels --

9337: grouped in manifolds -- will essentially survive, see \cite{krug00}.

9338: It is likely that some non-dispersive wave-packets also exist in

9339: non-hydrogenic atomic species.

9340: 

9341: 

9342: \subsection{Preparation through tailored pulses}

9343: \label{PTP}

9344: 

9345: Another, indirect method for preparing  non-dispersive wave-packets

9346: is also available. This will be the method of choice for wave-packets

9347: moving along classical orbits of large angular momentum (small

9348: eccentricity). Such states, obviously, are not accessible to a direct

9349: optical excitation from weakly excited, low angular momentum states.

9350: The same general scheme may be also applicable to high eccentricity

9351: wave-packets although in that case we expect that the direct

9352: excitation may be more efficient and flexible.

9353: The method to be discussed here

9354: consists of two stages.

9355: We first prepare the atom in a well chosen and well defined initial

9356: highly excited state, and then turn the microwave field on relatively slowly,

9357: from zero amplitude to its plateau value $F_{\rm max}$.

9358: 

9359: A non-dispersive wave-packet is a single eigenstate $|{\cal E}\rangle$

9360: of the Floquet Hamiltonian

9361: describing the driven system at fixed driving field amplitude

9362: $F.$

9363: As shown in sections  \ref{LIN3D}, \ref{CP} and \ref{SOLI},

9364:  the evolution of the quasienergies

9365: of the driven atom with an external control parameter like the driving field

9366: amplitude is rather complicated.

9367:  It reflects the dramatic transformation

9368: of the structure of classical phase space, manifesting in an

9369: abundance of avoided crossings of various sizes in the level dynamics.

9370: Still, as exemplified in figs.~\ref{fig_mathieu}, \ref{solit2}, and \ref{solfig},

9371: the wave-packet states may be followed 

9372: rather easily 

9373: under changes of $F$ (parametrized by $t$, during the switching of the pulse)

9374: in the level dynamics, 

9375: in agreement with their ``solitonic''

9376: character (see section \ref{SOLI}).

9377: Nontheless, the very same figures illustrate clearly that the targeted 

9378: wave-packet state undergoes

9379: many avoided crossings as the  microwave amplitude is 

9380: swept. To remain in 

9381: a single eigenstate, the avoided crossings should be passed

9382: either adiabatically or diabatically, with a

9383: branching ratio at an individual crossing

9384: being described by the well known Landau-Zener scenario

9385: \cite{landau2,breuer89b}.

9386: Consequently, if we want to populate an individual wave-packet eigenstate

9387: from a field free atomic state $|\psi_0\rangle$,

9388: we need some knowledge

9389: of the energy level dynamics. Then it is possible to identify those

9390: field free states which are connected to the wave-packet via

9391: adiabatic and/or diabatic transitions in the network of energy levels,

9392: and subsequently

9393: to design $F(t)$ such as to transfer population

9394: from $|\psi_0\rangle$ to $|{\cal E}\rangle$ most efficiently.

9395: A precise experimental preparation of $|\psi_0\rangle$

9396: is the prerequisite of any such approach.

9397: 

9398: When the driving field is increased from zero, the major modification

9399: in the classical phase space is the emergence

9400:  of the resonance

9401: island (see figs.~\ref{lin3d_4}, \ref{lin3d_9}, \ref{lin3d_10}). 

9402: Quantum mechanically, the states with initial

9403: principal quantum number close to $n_0=\omega^{-1/3}$ will

9404: enter progressively inside the resonance island.

9405: For a one-dimensional system, the Mathieu equation, discussed

9406: in section \ref{section_mathieu}, fully describes the evolution of the

9407: energy levels in this regime. As shown for example

9408: in figure~\ref{fig_mathieu}, the non-dispersive wave-packet

9409: with the best localization, i.e. $N=0,$ 

9410: is -- in this simple situation -- adiabatically connected 

9411: to the field-free state with principal quantum number closest

9412: to $n_0,$ i.e. the eigenstate $\kappa=0$ of the

9413: Mathieu equation. When the Mathieu parameter $q,$ eq.~(\ref{map_mathieu3}), 

9414: is of the order

9415: of unity, the state of interest is trapped in the resonance island, which

9416: happens at field amplitudes given by eq.~(\ref{ftrapping})

9417: for the one-dimensional atom, and for 

9418: linear polarization of the microwave field.

9419: A similar scaling

9420: is expected for other polarizations, too.

9421: In the interval

9422: $F\leq F_{\rm trapping}$, the field has to be increased slowly enough such

9423: as to avoid losses from the ground state to the excited states of

9424: the Mathieu equation, at an energy separation of the order of $n_0^{-4}$.

9425: The most favorable situation is then the case of

9426: ``optimal'' resonance (see section~\ref{QD}),

9427: when $n_0$ is an integer, the situation in figure~\ref{fig_mathieu}.

9428: The wave-packet state is always separated from the other

9429: states by an energy gap comparable to its value at $F=0$,

9430: i.e. of the order of $3/(2n_0^4)$.

9431: The situation is less favorable if $n_0$ is not an integer,

9432: because the energy gap between the wave-packet

9433: of interest and the other states is smaller when $F \to 0.$

9434: The worst case is met when $n_0$ is half-integer: the free

9435: states $n_0+1/2$ and $n_0-1/2$ are quasi-degenerate, and selective

9436: excitation of a single wave-packet is thus more difficult.

9437: 

9438: The appropriate time scale for switching on the field is given 

9439: by the inverse of the

9440: energy splitting, i. e. for ``optimal" resonance

9441: \begin{equation}

9442: \tau_{\rm trapping}\sim n_0^4=n_0\times 2\pi/\omega,

9443: \label{t_smallF}

9444: \end{equation}

9445: or $n_0$ driving field periods.

9446: 

9447: Once trapped in the resonance island, the coupling to states localized

9448: outside the island will be residual -- mediated by quantum mechanical

9449: tunneling, see section \ref{ION} -- and the size of the

9450: avoided crossings between the trapped and the untrapped states is exponentially

9451: small. After adiabatic switching into the resonance island on a time 

9452: scale of $n_0$

9453: Kepler orbits, we now have to switch diabatically from $F_{\rm trapping}$ to

9454: some final

9455: $F$ value, in order to avoid adiabatic

9456: losses from

9457: the wave-packet into other states while passing through the avoided crossings.

9458: 

9459: 

9460: The preceding discussion is based on a one-dimensional model

9461: and the Mathieu equation. Taking into account the other

9462: ``transverse'' degrees of freedom is not too difficult. Indeed,

9463: as noticed in sections \ref{LIN3D}, \ref{CP}, 

9464: the various time scales of the problem

9465: are well separated. The transverse motion is slow and can be

9466: adiabatically separated from the fast motion in $(\hat{I},\hat\theta)$. 

9467: Instead of getting

9468: a single set of energy levels, one gets a family of sets,

9469: the various families being essentially uncoupled. An example

9470: for the 3D atom in a linearly polarized microwave field is shown

9471: in fig.~\ref{lin3d_4}.

9472:  It follows that the estimate for the trapping field

9473: and the

9474: switching time are essentially the same as for 1D systems.

9475: Inside the resonance island, the situation becomes slightly more

9476: complicated, because there is not a single frequency for

9477: the secular motion, but several frequencies along the

9478: transverse degrees of freedom. For example, in CP,

9479: it has been shown that there are three eigenfrequencies, 

9480: eqs.~(\ref{omegas},\ref{ompmz}), in the harmonic

9481: approximation -- see section \ref{rotating_frame}.

9482:  This results in a large number of excited energy levels

9483: which may have avoided crossings with the ``ground state", i.e.

9484: the nondispersive wave-packet we want to prepare.

9485: As shown in section \ref{SOLI}, most of these avoided

9486: crossings are extremely small and can be easily crossed diabatically.

9487: However, some of them are rather large, especially when there

9488: is an internal resonance between two eigenfrequencies. Examples are

9489: given in figs.~\ref{solit2} and \ref{solfig}, where $\omega_+=3\omega_-$ and 

9490: $\omega_+=4\omega_-$, respectively. 

9491: 

9492: These

9493: avoided crossings are large and dangerous, because the states

9494: involved lie {\em inside} the resonance island, which thus

9495: loses its protective character.

9496: They are mainly due to the unharmonic character of the

9497: Coulomb potential.

9498: Their size  may be qualitatively analyzed as we do below on the

9499: CP example, 

9500: expecting similar

9501: sizes of the avoided crossings for any polarization.

9502: 

9503: The unharmonic corrections to the

9504: harmonic approximation around the stable fixed point

9505: $x_e,p_e$ in the center of the nonlinear resonance -- as outlined in

9506: section \ref{CP} -- are due to the higher order terms

9507: $(-1)^j\tilde{x}^j/j!x_e^{j+1}$ in the

9508: Taylor series of the Coulomb potential, where

9509: $\tilde{x},\tilde{y}$ are excursions from

9510: the equilibrium position.

9511: $\tilde{x}$ and $\tilde{y}$

9512: can be expressed as linear combinations of $a_{\pm}^{\dagger}$ and

9513: $a_{\pm}$ operators \cite{delande98} giving

9514: $\tilde{x},\tilde{y}\sim \omega^{-1/2}\sim n_0^{3/2}$. Furthermore,

9515: the equilibrium distance from the nucleus scales as the size of the atom,

9516: $x_e\sim n_0^2$, and, therefore,

9517: \begin{equation}

9518: \tilde{x}^j/x_e^{j+1}\sim n_0^{\left(-2-\frac{j}{2}\right)}.

9519: \label{abu_ptp_scaleac}

9520: \end{equation}

9521: 

9522: A state $|n_+,n_-\rangle$ is obtained

9523: by the excitation of

9524: $N_+$ quanta in the $\omega_+$-mode and of $N_-$ quanta in the

9525: $\omega_-$-mode, respectively, i. e.

9526:  by the application of the

9527: operator product $(a_+^{\dagger})^{n_+}(a_-^{\dagger})^{n_-}$ on

9528: the wave-packet state $|0,0\rangle$, which -- by virtue of

9529: eq.~(\ref{abu_ptp_scaleac}) -- will be subject to an unharmonic correction

9530: scaling like

9531: \begin{equation}

9532: \Delta E_{\rm unharmonic}\sim n_0^{\left(-2-\frac{n_++n_-}{2}\right)}.

9533: \label{abu_ptp_unharm}

9534: \end{equation}

9535: Hence, the size of the

9536: avoided crossings between the wave-packet eigenstate and

9537: excited states of the local potential around the stable fixed point

9538: decreases with the number of quanta in the excited modes.

9539: 

9540: In addition, we can determine the width $\Delta F_{\rm unharmonic}$

9541: of such avoided crossings in

9542: the driving field amplitude $F$, by differentiation of the energy

9543: (\ref{enharm})

9544: of $|n_+,n_-\rangle$ with respect to $F$.

9545: Then, the difference between the energies of two eigenstates localized

9546: in the resonance island is found to scale like $Fn_0$. Defining

9547: $\Delta F_{\rm unharmonic}$ by the requirement that $Fn_0$ be of the order

9548: of $\Delta E_{\rm unharmonic}$, we find

9549: \begin{equation}

9550: \Delta F_{\rm unharmonic}\sim n_0^{\left(-3-\frac{n_++n_-}{2}\right)}.

9551: \label{abu_ptp_fun}

9552: \end{equation}

9553: Since we want to switch the field to a maximum value

9554: $F_{\rm max}\sim n_0^{-4}$, the Landau-Zener formula

9555: \begin{equation}

9556: \tau\sim \frac{F_{\rm max}}{\Delta E \Delta F}

9557: \label{abu_ptp_lz}

9558: \end{equation}

9559: yields

9560: \begin{equation}

9561: \tau_{\rm unharmonic}\sim n_0^{n_-+n_++1}\sim n_0^{n_-+n_+-2}\ \ \

9562: {\mathrm microwave\ periods} 

9563: \label{abu_ptp_tunharm}

9564: \end{equation}

9565: for the scaling behavior of the time scale which guarantees diabatic

9566: switching through avoided crossings of the wave-packet eigenstate with

9567: excited states of the elliptic island.

9568: Let us stress that this is only a very rough estimate of the

9569: switching time,  some numerical

9570: factors (not necessarily close to unity) are not taken into account.

9571: 

9572: \begin{figure}

9573: \centerline{\psfig{figure=bdzf61.ps,width=14cm}}

9574: \caption{

9575: Snapshots of the electronic density for a two-dimensional

9576: hydrogen atom exposed to a circularly polarized microwave field

9577: with increasing amplitude. The  microwave

9578: field amplitude is switched on

9579: according to eq.~(\protect\ref{abu_ptp_pulse}), with maximum scaled field

9580: $F_{0,\rm max}= 0.03$ and

9581: $T_{\rm switch}=400\times 2\pi/\omega$, where $\omega$ is resonant with

9582: $n_0=60$ (frequency $\omega=1/(60.5)^3$).

9583: The evolution of the initial circular state $n=M=60$ is numerically

9584: computed by solving the time dependent Schr\"odinger equation in a

9585: convenient Sturmian basis.

9586: Top-left -- $t=0$ (initial circular state);

9587: top-middle -- $t=20$ microwave periods;

9588: top-right -- $t=60$ periods, bottom-left -- $t=100$ periods;

9589: bottom-middle -- the final state, $t=400$ periods;

9590: bottom-right -- the non-dispersive wave-packet (exact Floquet

9591: eigenstate): it is almost indistinguishable from the previous wave-function,

9592: what proves that the excitation process efficiently

9593: and almost exclusively populates the state of interest. The box extends

9594: over 10000 Bohr radii in both directions, with the nucleus

9595: at the center. The microwave field is along the horizontal axis,

9596: pointing to the right.}

9597: \label{ptp_fig1}

9598: \end{figure}

9599: 

9600: The above predictions can be checked, e.g., by a numerical integration of the

9601: time dependent Schr\"odinger equation for a  microwave-driven 

9602: atom, taking into account the  

9603: time-dependent amplitude of the field. An

9604: exemplary calculation on the two-dimensional model atom 

9605: (see sec.~\ref{2d_model}) 

9606: can be found in \cite{kuba97a}, for 

9607: CP driving.

9608: Fig.~\ref{ptp_fig1} shows the evolution

9609: of the electronic density

9610: of the atomic wave-function (initially prepared in the circular Rydberg

9611: state $n=M=60$) during the rising part of the driving

9612: field envelope, modeled by 

9613: \begin{equation}

9614: F(t)=F_{\rm max}\sin^2\left(\frac{\pi t}{2T_{\rm switch}}\right).

9615: \label{abu_ptp_pulse}

9616: \end{equation}

9617: 

9618: The driving field frequency was chosen according to the resonance condition

9619: with the $n_0=60$ state,

9620: with a maximum scaled amplitude $F_{0,\rm max}=0.03$.

9621: Inspection of 

9622: fig.~\ref{solfig} shows that, for this value of $F_{\rm max}$, the crossing

9623: between the wave-packet eigenstate and the state $|n_+=1,n_-=4\rangle$ 

9624: has to

9625: be passed diabatically after adiabatic

9626: trapping within the principal resonance. By virtue of the above estimations

9627: of the adiabatic and the diabatic time scales, the switching

9628: time (measured in driving field cycles)

9629: has to be chosen such that $n_0<T_{\rm switch}<n_0^3$

9630: (in microwave periods).

9631: Clearly, the pulse 

9632: populates the desired wave-packet 

9633: once the driving field

9634: amplitude 

9635: reaches its maximum value. More quantitatively, 

9636: the overlap of the final state after  

9637: propagation of the time-dependent Schr\"odinger equation 

9638: with the

9639: wave-packet eigenstate of the driven atom in the field

9640: (bottom-right panel) amounts to $94\%$. Since losses of atomic

9641: population due to ionization

9642: are negligible on the time scales considered in the figure, $6\%$ of the

9643: initial atomic population is lost during the switching process.

9644: The same calculation, done for the realistic three-dimensional

9645: atom with $n_0=60$ gives the same result, proving that

9646: the $z$-direction (which is neglected in the 2D model) 

9647: is essentially irrelevant in this problem.

9648: \begin{figure}

9649: \centerline{\psfig{figure=bdzf62.eps,width=12cm,angle=-90}}

9650: \caption{Overlap between the wave-function obtained at the end of

9651: the microwave turn-on and the exact target state representing the

9652: non-dispersive wave-packet, as a function of

9653: the switching time $T_{\rm switch},$ for the two-dimensional

9654: hydrogen atom (circles). The filled squares

9655: indicate the results obtained for a fully

9656: three-dimensional atom. $F_{\rm max}$, $\omega$, and $n_0$  

9657: as in fig.~\protect\ref{ptp_fig1}.}

9658: \label{ptp_fig2}

9659: \end{figure}

9660: Figure~\ref{ptp_fig2}

9661: shows the efficiency of the proposed switching scheme

9662: as a function of the switching time $T_{\rm switch}$, expressed

9663: in units of microwave periods. Observe that too

9664: long switching times tend to be less effective, since the avoided crossings

9665: passed during the switching stage  

9666: are not traversed diabatically. The rough estimate, eq.~(\ref{abu_ptp_tunharm}),

9667: overestimates the maximum switching time by one order of magnitude.

9668: On the other hand, 

9669: too short switching times do not allow the wave-packet 

9670: to localize inside the resonance island. However, a wide range

9671: of switching times remains where good efficiency is achieved.

9672: 

9673: For a given initial state $|\psi_0\rangle$

9674: of the atom, only the

9675: resonance condition defining the driving field frequency is to some extent

9676: restrictive, as depicted in fig.~\ref{ptp_fig3}.

9677: It is crucial that the initially excited field-free state

9678: is adiabatically connected 

9679: (through the Mathieu equation) to the ground state wave-packet. 

9680: The best choice is ``optimal resonance", but the 

9681: adiabaticity is preserved 

9682: if $n_0$ is changed by less than one half, see section~\ref{QD}. 

9683: This corresponds to a relative 

9684: change of $\omega$ of the order of $3/2n_0.$

9685: Given the spectral resolution of presently available

9686: microwave generators, the definition of the

9687: frequency with an accuracy of less than 1\%

9688: is not a limitation. 

9689: The exact numerical calculation displayed in fig.~\ref{ptp_fig3}

9690: fully confirms that efficient excitation is possible as long as 

9691: $n_0=\omega^{-1/3}-1/2$ matches the effective principal 

9692: quantum number of the

9693:  initially excited field-free state within a margin of $\pm 1/2$ 

9694: (in the range [59.5,60.5]).

9695: 

9696: \begin{figure}

9697: \centerline{\psfig{figure=bdzf63.eps,width=12cm,angle=-90}}

9698: \caption{Overlap between the wave-function obtained at the end of

9699: the microwave turn-on and the exact target state representing the

9700: non-dispersive wave-packet as a function of $n_0$,

9701: obtained for the two-dimensional

9702: hydrogen atom (circles). $F_{\rm max}$ is 

9703: as in fig.~\protect\ref{ptp_fig1}, and the switching time 

9704: is $T_{\rm switch}=250$ microwave periods. 

9705: The initial state corresponds to a circular

9706: $n=60$ state of a 2D hydrogen atom.}

9707: \label{ptp_fig3}

9708: \end{figure}

9709: 

9710: In conclusion,

9711: the preparation of non-dispersive wave-packets by excitation

9712: of a Rydberg state followed by careful switching of the

9713: microwave field can be considered as an efficient method, provided

9714: a clean experimental preparation of the atomic initial state can be

9715: achieved. Furthermore, the boundaries --

9716: eqs.~(\ref{t_smallF}) and (\ref{abu_ptp_tunharm}) --

9717: imposed on the time scale for the

9718: switching process leave a sufficient flexibility for the experimentalist 

9719: to efficiently 

9720: prepare the wave-packet.

9721: A final word is in place on the homogeneity of the driving field amplitude

9722: experienced by the atoms in the ``flat top region'' of the interaction,

9723: i.e. after the switching from the field free state into

9724: the wave-packet state at $F(t)=F_{\rm max}$. In any laboratory experiment,

9725: a slow drift of the amplitude will be unavoidable over the interaction volume.

9726: Hence, slightly different non--dispersive wave-packets will coexist

9727: at various spatial positions. Since the ionization rate

9728: of nondispersive wave-packets is rather sensitive

9729: with respect to detailed values of the parameters, see

9730: section \ref{ION}, this should manifest itself

9731: by a deviation of the time dependence of the ionization yield from purely

9732: exponential decay.

9733: 

9734: 

9735: \subsection{Life time measurements}

9736: \label{LIFE}

9737: 

9738: 

9739: Given the above, rather efficient experimental 

9740: schemes for the population of non-dispersive wave-packet eigenstates --

9741: either via direct optical Floquet absorption or through an appropriate

9742: switching procedure --

9743: we still need some means to prove that we

9744: really {\em did} populate the wave-packet. As a matter of fact, to provide 

9745: unambiguous experimental evidence, one has to test various 

9746: characteristic properties of the wave-packet, so as to

9747: exclude accidental coincidences. A natural way is Floquet spectroscopy 

9748: (see sec.~\ref{SPEC}),

9749: i.e. probing the structure of the Floquet quasi-energy levels,

9750: in either

9751: the optical or the microwave regime

9752: (via absorption, stimulated emission, Raman spectroscopy etc.). Another

9753: possibility is to explore unique properties of wave-packet Floquet states.

9754: For example, as discussed in sec.~\ref{ION}, these states exhibit 

9755: extremely small ionization rates. 

9756: Hence, an experimentally accessible quantity to identify

9757: these states is the time-dependence of their survival probability, i.e. 

9758: of the probability not to ionize during an interaction time $t.$

9759: It is

9760: given by \cite{abuth,abu95c}

9761: \begin{equation}

9762: P(t)=\sum_{\epsilon}|c_{\epsilon}|^2\exp(-\Gamma_{\epsilon}t),

9763: \label{abu_lt_pion}

9764: \end{equation}

9765: where

9766: the  $c_{\epsilon}$ denote the expansion coefficients of the initial 

9767: field-free 

9768: state in 

9769: the Floquet basis,

9770: at a given value of microwave amplitude $F$.

9771: 

9772: If the selective population of the wave-packet is successful, 

9773: only one Floquet state contributes to $P(t)$, and the 

9774: decay of the population to the atomic continuum should manifest in 

9775: its exponential decrease, as opposed to a multiexponential  

9776: decrease in the case of a broad distribution of the 

9777: $c_{\epsilon}$ over the Floquet states 

9778: \cite{abuth,abu95a,abu98b,wimbergerda,wimberger01,abu95b}. 

9779: Of course, the distinction between 

9780: an exponential and an algebraic decay law requires the variability of 

9781: the experimental interaction time over more than one order of magnitude. 

9782: This is a nontrivial task in experiments on atomic Rydberg states of 

9783: hydrogen, since the typical velocities of the atomic beam are of the order

9784: of $1000\ \rm m/s$. That significantly restricts the interval on which

9785: the interaction time may be changed, taking into account the typical

9786: size (in the cm-range) of

9787: the atom-field interaction region 

9788: \cite{koch95b,bayfield85}. 

9789: However, the feasibility of such measurements has already been demonstrated

9790: in microwave experiments on rubidium Rydberg states, where the interaction

9791: time has been scanned from approx. $100$ to approx. $100000$ field cycles,

9792: i.e., over three orders of magnitude \cite{abu95b,arndt91}.

9793: Note that, whereas the dynamics

9794: of the driven Rydberg electron 

9795: along a Kepler ellipse of large eccentricity will 

9796: certainly be affected by the presence of a non-hydrogenic core, 

9797: non-dispersive wave-packets as the ones discussed in 

9798: sections \ref{CP} and \ref{EP} 

9799: can certainly be launched along circular trajectories, since the 

9800: Rydberg electron of the rubidium atom essentially experiences a Coulomb

9801: field on such a circular orbit. 

9802: 

9803: To use the character of the decay as a means 

9804: to identify the wave-packet,

9805: the microwave field amplitude should be sufficiently large

9806: to guarantee that other Floquet states localized in the chaotic sea (see sec.~\ref{ION}) 

9807: decay rapidly. Otherwise, the observation

9808: of a mono-exponential decay simply suggests that we succeeded in populating

9809: a single Floquet state - not necessarily a wave-packet \cite{abu95b}.

9810: The appropriate choice of the driving 

9811: field amplitude $F$, such that appreciable ionization  

9812: is achieved for the 

9813: longer experimentally accessible 

9814: interaction times,

9815: should therefore allow for the experimental identification of the 

9816: mono-exponential decay from the wave-packet to the atomic continuum, but also

9817: -- by varying $F$ -- of the variations of the decay rate with $F,$

9818: which is predicted to fluctuate wildly over several orders of magnitude,

9819: see section \ref{ION}. 

9820: Note, however, that this requires an excellent homogeneity of the

9821: microwave field (e.g., provided by a high quality microwave cavity), 

9822: as the fluctuations take place over

9823: rather small intervals of $F.$

9824: 

9825: 

9826: \section{Conclusions}

9827: \label{CONC}

9828: 

9829: 

9830: In this report, we have shown that novel and highly robust eigenstates 

9831: of periodically 

9832: driven quantum systems -- non-dispersive wave-packets -- are born 

9833: out of classically mixed regular-chaotic dynamics. As much as a mixed phase 

9834: space is generic for classical Hamiltonian systems, non-dispersive 

9835: wave-packets are a generic manifestation thereof on the quantum level, given

9836: a sufficiently high density of states (needed to resolve finite-size phase 

9837: space structures). While we described their semiclassical properties and

9838: their experimental preparation, manipulation, and identification during the 

9839: largest part of this report for a specific system -- atomic Rydberg states

9840: driven by a microwave field -- it is clear from our approach 

9841: that such ``quantum particles'' can be anchored to any nonlinear resonance 

9842: between a periodic drive and a periodic trajectory of a Hamiltonian 

9843: system. As an alternative example, we have briefly touched upon the 

9844: atomic realization of the gravitational bouncer, though many other 

9845: realizations in simple quantum optical or atomic and molecular systems can be

9846: thought of. Let us only mention unharmonic traps for ions, atoms, or BEC

9847: condensates, periodically kicked atoms \cite{reinhold01},

9848: as well as molecular dynamics \cite{gerber98,gerber01} 

9849: on adiabatic potential surfaces

9850: (the driven frozen planet briefly discussed in section \ref{HE} may 

9851: be conceived

9852: as opening a perspective in this direction). Nontheless, atomic 

9853: Rydberg states remain arguably the best objects to study the 

9854: fundamental properties of non-dispersive wave-packets as the realization of

9855: Schr\"odinger's dream \cite{schroe26}.

9856:  On one hand, they are 

9857: microscopic realizations of the Keplerian motion and of Bohr's orbitals using

9858: a well understood non-linear dynamical system. 

9859: On the other hand, they possess the essential complication which open 

9860: quantum systems add to bounded Hamiltonian dynamics -- the 

9861: driving-induced, coherent coupling to the atomic continuum of free electronic

9862: states. On top of that, all these features can be controlled in 

9863: real laboratory experiments, and we might actually dream of probing the

9864: characteristic properties of nondispersive wave-packets on single, trapped 

9865: atoms or ions, using novel experimental approaches yet to come. Let us finally 

9866: dare to speculate on the potential use of non-dispersive wave-packets in 

9867: coherent control: given their spectacular robustness -- which we abundantly 

9868: illustrated in this report -- it is clear that they provide a means to store 

9869: and to ``ship'' quantum probability densities in and across phase space,

9870: e.g., under adiabatic changes of the driving field polarization and/or of the 

9871: strength or orientation of additional static fields. Given the recent 

9872: advances in coherent control of molecular reactions employing laser fields

9873: \cite{gerber98} -- which so far do not explore the unique perspectives of 

9874: nonlinear dynamics -- it looks like a promising (and challenging) program to 

9875: systematically study non-dispersive wave-packets in molecular reaction 

9876: dynamics.

9877: 

9878: 

9879: \section{Acknowledgments}

9880: 

9881: It is a pleasure 

9882: to acknowledge a longstanding and fruitful collaboration with Robert

9883: G\c{e}barowski, Beno\^{\i}t Gr\'emaud, 

9884: Klaus Hornberger, Andreas Krug, Romek Marcinek, 

9885: Krzysiek Sacha, Peter Schlagheck, and Sandro

9886: Wimberger on non-dispersive wave-packets and related topics 

9887: over the past five years.

9888: 

9889: We acknowledge support of bilateral collaborations via

9890: programmes Procope (German-French) and Polonium (Polish-French).

9891: J.Z. acknowledges support by Polish Committee for Scientific Research

9892: under grant 2P03B00915. Laboratoire Kastler Brossel is laboratoire 

9893: de l'Universit{\'e} Pierre et Marie

9894: Curie et de l'Ecole Normale Sup{\'e}rieure, unit{\'e} mixte de

9895: recherche 8552 du CNRS. 

9896: CPU time on various computers has been provided by IDRIS (Orsay) and 

9897: RZG (Garching).

9898: 

9899: 

9900: 

9901: \newpage

9902: \begin{thebibliography}{100}

9903: 

9904: \bibitem{englert95}

9905: J.~A. Bergou and B.~G. Englert, J. Mod. Opt. {\bf 45},  701  (1998).

9906: 

9907: \bibitem{schroe26}

9908: E. Schr\"odinger, \natw {\bf 14},  664  (1926).

9909: 

9910: \bibitem{lichtenberg83}

9911: A.~J. Lichtenberg and M.~A. Lieberman, {\em Regular and Stochastic Motion},

9912:   Vol.~38 of {\em Applied Mathematical Sciences} (Springer, Berlin, 1983).

9913: 

9914: \bibitem{raman97}

9915: C. Raman, T.~C. Weinacht, and P.~H. Bucksbaum, \pr A {\bf 55},  R3995  (1997).

9916: 

9917: \bibitem{yeazell90}

9918: J.~A. Yeazell, M. Mallalieu, and J. C.~R.~Stroud, \prL {\bf 64},  2007  (1990).

9919: 

9920: \bibitem{alber91}

9921: G. Alber and P. Zoller, \phr {\bf 199},  231  (1991).

9922: 

9923: \bibitem{landau2}

9924: L.~D. Landau and E.~M. Lifschitz, {\em Quantum Mechanics} (Pergamon, Oxford,

9925:   1977).

9926: 

9927: \bibitem{parker86}

9928: J. Parker and J. C.~R.~Stroud, \prL {\bf 56},  716  (1986).

9929: 

9930: \bibitem{alber86}

9931: G. Alber, H. Ritsch, and P. Zoller, \pr {\bf 34},  1058  (1986).

9932: 

9933: \bibitem{averbukh89}

9934: I.~S. Averbukh and N.~F. Perelman, \pla {\bf 139},  449  (1989).

9935: 

9936: \bibitem{yeazell91}

9937: J.~A. Yeazell and J. C.~R.~Stroud, \pr {\bf 43},  5153  (1991).

9938: 

9939: \bibitem{hillery84}

9940: M. Hillery, R.~F. O'Connell, M.~O. Scully, and E.~P. Wigner, \phr {\bf 106},

9941:   12  (1984).

9942: 

9943: \bibitem{moyal49}

9944: J.~E. Moyal, Proc. Camb. Phil. Soc. Math. Phys. Sci. {\bf 45},  99  (1949).

9945: 

9946: \bibitem{landau1}

9947: L.~D. Landau and E.~M. Lifschitz, {\em Mechanics} (Pergamon, Oxford, 1994).

9948: 

9949: \bibitem{haake90}

9950: F. Haake, {\em Quantum Signatures of Chaos}, Vol.~54 of {\em Springer Series in

9951:   Synergetics} (Springer, Berlin, 1991).

9952: 

9953: \bibitem{glauber63}

9954: R.~J. Glauber, \pr {\bf 131},  2766  (1963).

9955: 

9956: \bibitem{mandel90}

9957: L. Mandel and E. Wolf, {\em Coherence and Quantum Optics} (Cambridge University

9958:   Press, Cambridge, 1995).

9959: 

9960: \bibitem{cct92}

9961: C. Cohen-Tannoudji, J. Dupont-Roc, and G. Grynberg, {\em Atom-Photon

9962:   Interactions: Basic Processes and Applications} (John Wiley and Sons, New

9963:   York, 1992).

9964: 

9965: \bibitem{cct73}

9966: C. Cohen-Tannoudji, B. Diu, and F. Lalo\"e, {\em M\'ecanique quantique}

9967:   (Hermann, Paris, 1973).

9968: 

9969: \bibitem{liboff80}

9970: R.~L. Liboff, {\em Introductory Quantum Mechanics} (Holden-Day, San Francisco,

9971:   1980).

9972: 

9973: \bibitem{husimi40}

9974: K. Husimi, Proc. Phys. Math. Soc. Japan {\bf 22},  264  (1940).

9975: 

9976: \bibitem{bethe77}

9977: H.~A. Bethe and E.~E. Salpeter, {\em Quantum Mechanics of One- and Two-Electron

9978:   Atoms} (Plenum Publishing Corp., New York, 1977).

9979: 

9980: \bibitem{goldberger50}

9981: M.~S. Goldberger and K.~M. Watson, {\em Collision Theory} (Wiley, New York,

9982:   1964).

9983: 

9984: \bibitem{faisal87}

9985: F.~H.~M. Faisal, {\em Theory of Multiphoton Processes} (Plenum Press, New York,

9986:   1987).

9987: 

9988: \bibitem{yeazell88}

9989: J.~A. Yeazell and J. C.~R.~Stroud, \prL {\bf 60},  1494  (1988).

9990: 

9991: \bibitem{marmet94}

9992: L. Marmet {\it et~al.}, \prL {\bf 72},  3779  (1994).

9993: 

9994: \bibitem{mallalieu94}

9995: M. Mallalieu and J. C.~R.~Stroud, \pr {\bf A49},  2329  (1994).

9996: 

9997: \bibitem{weinacht99}

9998: T.~C. Weinacht, J. Ahn, and P.~H. Bucksbaum, Nature {\bf 397},  233  (1999).

9999: 

10000: \bibitem{ottb}

10001: E. Ott, {\em Chaos in Dynamical Systems} (Cambridge University Press,

10002:   Cambridge, 1993).

10003: 

10004: \bibitem{lee97}

10005: E. Lee, A.~F. Brunello, and D. Farrelly, \pr A {\bf 55},  2203  (1997).

10006: 

10007: \bibitem{schroe68}

10008: E. Schr\"odinger,  in {\em Sources of Quantum Mechanics}, edited by B.~L.

10009:   van~der Waerden (Dover, New York, 1968).

10010: 

10011: \bibitem{delande94}

10012: D. Delande and A. Buchleitner, \amop {\bf 35},  85  (1994).

10013: 

10014: \bibitem{abu95d}

10015: A. Buchleitner and D. Delande, \prL {\bf 75},  1487  (1995).

10016: 

10017: \bibitem{ibb94}

10018: I. Bia{\l}ynicki-Birula, M. Kalinski, and J.~H. Eberly, \prL {\bf 73},  1777

10019:   (1994).

10020: 

10021: \bibitem{ahn00}

10022: J. Ahn, T.~C. Weinacht, and P.~H. Bucksbaum, Science {\bf 287},  463  (2000).

10023: 

10024: \bibitem{meyer00}

10025: D.~A. Meyer, Science {\bf 289},  1431a  (2000).

10026: 

10027: \bibitem{kwiat00}

10028: P.~G. Kwiat and R.~J. Hughes, Science {\bf 289},  1431a  (2000).

10029: 

10030: \bibitem{bucksbaum00}

10031: P.~H. Bucksbaum, J. Ahn, and T.~C. Weinacht, Science {\bf 289},  1431a  (2000).

10032: 

10033: \bibitem{berman77}

10034: G.~P. Berman and G.~M. Zaslavsky, Phys. Lett. {\bf 61A},  295  (1977).

10035: 

10036: \bibitem{henkel92}

10037: J. Henkel and M. Holthaus, \pr {\bf A45},  1978  (1992).

10038: 

10039: \bibitem{holthaus94}

10040: M. Holthaus, Prog. Theor. Phys. Supplement {\bf 116},  417  (1994).

10041: 

10042: \bibitem{holthaus95}

10043: M. Holthaus, \csf {\bf 5},  1143  (1995).

10044: 

10045: \bibitem{abuth}

10046: A. Buchleitner, Ph.D. thesis, Universit\'e Pierre et Marie Curie, Paris, 1993.

10047: 

10048: \bibitem{farrelly95a}

10049: D. Farrelly, E. Lee, and T. Uzer, \prL {\bf 75},  972  (1995).

10050: 

10051: \bibitem{ibb95}

10052: I. Bia{\l}ynicki-Birula, M. Kalinski, and J.~H. Eberly, \prL {\bf 75},  973

10053:   (1995).

10054: 

10055: \bibitem{farrelly95}

10056: D. Farrelly, E. Lee, and T. Uzer, \pla {\bf 204},  359  (1995).

10057: 

10058: \bibitem{kalinski95a}

10059: M. Kalinski, J.~H. Eberly, and I. Bia{\l}ynicki-Birula, \pr {\bf A52},  2460

10060:   (1995).

10061: 

10062: \bibitem{kalinski95b}

10063: M. Kalinski and J.~H. Eberly, \pr {\bf A52},  4285  (1995).

10064: 

10065: \bibitem{delande95}

10066: D. Delande, J. Zakrzewski, and A. Buchleitner, Europhys. Lett. {\bf 32},  107

10067:   (1995).

10068: 

10069: \bibitem{kuba95a}

10070: J. Zakrzewski, D. Delande, and A. Buchleitner, Phys. Rev. Lett. {\bf 75},  4015

10071:    (1995).

10072: 

10073: \bibitem{ibb96}

10074: I. Bia{\l}ynicki-Birula and Z. Bia{\l}ynicka-Birula, Phys. Rev. Lett. {\bf 77},

10075:    4298  (1996).

10076: 

10077: \bibitem{eberly96}

10078: J.~H. Eberly and M. Kalinski,  in {\em Multiphoton Processes 1996, Proceedings

10079:   of the 7th International Conference on Multiphoton Processes,

10080:   Garmisch-Partenkirchen, Germany, October 1996}, Vol.~154 of {\em Institute of

10081:   Physics Conference Series}, edited by P. Lambropoulos and H. Walther

10082:   (Institute of Physics, Bristol and Philadelphia, 1997), pp.\ 29--36.

10083: 

10084: \bibitem{kalinski96a}

10085: M. Kalinski and J.~H. Eberly, \prL {\bf 77},  2420  (1996).

10086: 

10087: \bibitem{brunello96}

10088: A.~F. Brunello, T. Uzer, and D. Farrelly, \prL {\bf 76},  2874  (1996).

10089: 

10090: \bibitem{kalinski96b}

10091: M. Kalinski and J.~H. Eberly, \pr {\bf A53},  1715  (1996).

10092: 

10093: \bibitem{kalinski97}

10094: M. Kalinski and J.~H. Eberly, \prL {\bf 79},  3542  (1997).

10095: 

10096: \bibitem{kalinski98}

10097: M. Kalinski, \pr {\bf 57},  2239  (1998).

10098: 

10099: \bibitem{ibb97a}

10100: I. Bia{\l}ynicki-Birula and Z. Bia{\l}ynicka-Birula, Phys. Rev. Lett. {\bf 78},

10101:    2539  (1997).

10102: 

10103: \bibitem{ibb97b}

10104: Z. Bia{\l}ynicka-Birula and I. Bia{\l}ynicki-Birula, \pr {\bf A56},  3629

10105:   (1997).

10106: 

10107: \bibitem{delande97}

10108: D. Delande, J. Zakrzewski, and A. Buchleitner, \prL {\bf 79},  3541  (1997).

10109: 

10110: \bibitem{kuba97a}

10111: J. Zakrzewski and D. Delande, \jpb {\bf 30},  L87  (1997).

10112: 

10113: \bibitem{cerjan97}

10114: C. Cerjan, E. Lee, D. Farrelly, and T. Uzer, \pr {\bf A55},  2222  (1997).

10115: 

10116: \bibitem{kuba97b}

10117: J. Zakrzewski, A. Buchleitner, and D. Delande, Z. Phys. {\bf B103},  115

10118:   (1997).

10119: 

10120: \bibitem{abu96}

10121: A. Buchleitner, D. Delande, and J. Zakrzewski,  in {\em Multiphoton Processes

10122:   1996, Proceedings of the 7th International Conference on Multiphoton

10123:   Processes, Garmisch-Partenkirchen, Germany, October 1996}, Vol.~154 of {\em

10124:   Institute of Physics Conference Series}, edited by P. Lambropoulos and H.

10125:   Walther (Institute of Physics, Bristol and Philadelphia, 1997), pp.\ 19--28.

10126: 

10127: \bibitem{kuba98}

10128: J. Zakrzewski, D. Delande, and A. Buchleitner, Phys. Rev. {\bf E57},  1458

10129:   (1998).

10130: 

10131: \bibitem{kuba98a}

10132: J. Zakrzewski, D. Delande, and A. Buchleitner, Acta Physica Polon. {\bf A 93},

10133:   179  (1998).

10134: 

10135: \bibitem{abu98a}

10136: A. Buchleitner, K. Sacha, D. Delande, and J. Zakrzewski, Eur. Phys. J. D {\bf

10137:   5},  145  (1999).

10138: 

10139: \bibitem{delande98}

10140: D. Delande and J. Zakrzewski, \pr A {\bf 58},  466  (1998).

10141: 

10142: \bibitem{sachath}

10143: K. Sacha, Ph.D. thesis, Jagellonian University, Krak\'ow, 1998, unpublished.

10144: 

10145: \bibitem{hornbergerda}

10146: K. Hornberger, Master's thesis, Ludwig-Maximilians-Universit\"at, M\"un\-chen,

10147:   1998.

10148: 

10149: \bibitem{hornberger98}

10150: K. Hornberger and A. Buchleitner, Europhys. Lett. {\bf 41},  383  (1998).

10151: 

10152: \bibitem{sacha98a}

10153: K. Sacha, J. Zakrzewski, and D. Delande, Eur. Phys. J. {\bf D1},  231  (1998).

10154: 

10155: \bibitem{sacha98b}

10156: K. Sacha and J. Zakrzewski, \pr A {\bf 58},  3974  (1998).

10157: 

10158: \bibitem{sacha99a}

10159: K. Sacha and J. Zakrzewski, \pr A {\bf 59},  1707  (1999).

10160: 

10161: \bibitem{yeazell00}

10162: D. Farrelly,  in {\em Physics and Chemistry of Wave Packets}, edited by J.~A.

10163:   Yeazell and T. Uzer (John Wiley \& Sons, New York, 2000).

10164: 

10165: \bibitem{Berry_WKB}

10166: M.~V. Berry and K.~E. Mount, Rep. Prog. Phys. {\bf 35},  315  (1972).

10167: 

10168: \bibitem{ozorio88}

10169: A.~M.~O. de~Almeida, {\em Hamiltonian Systems: Chaos and Quantization}

10170:   (Cambridge University Press, Cambridge, 1988).

10171: 

10172: \bibitem{einstein17}

10173: A. Einstein, Verh. d. Dtsch. Phys. Ges.  82  (1917).

10174: 

10175: \bibitem{bornwolf}

10176: M. Born and E. Wolf, {\em Principles of Optics} (Academic Press, Oxford, 1999).

10177: 

10178: \bibitem{breuer91}

10179: H.~P. Breuer and M. Holthaus, \anp {\bf 211},  249  (1991).

10180: 

10181: \bibitem{heller89}

10182: E.~J. Heller,  in {\em Chaos and Quantum Physics}, Vol.~Session LII of {\em Les

10183:   Houches} (North-Holland, Amsterdam, 1991), p.\ 547.

10184: 

10185: \bibitem{stoeckmann99}

10186: H.~J. St\"ockmann, {\em Quantum Chaos: An Introduction} (Cambridge University

10187:   Press, Cambridge, 1999).

10188: 

10189: \bibitem{jensen89b}

10190: R.~V. Jensen, M.~M. Sanders, M. Saraceno, and B. Sundaram, \prL {\bf 63},  2771

10191:    (1989).

10192: 

10193: \bibitem{leopold94}

10194: J.~G. Leopold and D. Richards, \jpb {\bf 27},  2169  (1994).

10195: 

10196: \bibitem{zaslavsky81}

10197: G.~M. Zaslavsky, \phr {\bf 80},  157  (1981).

10198: 

10199: \bibitem{ringot00}

10200: J. Ringot, P. Szriftgiser, J.~C. Garreau, and D. Delande, \prL {\bf 85},  2741

10201:   (2000).

10202: 

10203: \bibitem{abu97}

10204: A. Buchleitner and D. Delande, Phys. Rev. {\bf A55},  R1585  (1997).

10205: 

10206: \bibitem{chirikov59}

10207: B.~V. Chirikov, Doc. Ac. Sci. USSR {\bf 125},  1015  (1959).

10208: 

10209: \bibitem{chirikov79}

10210: B.~V. Chirikov, \phr {\bf 52},  263  (1979).

10211: 

10212: \bibitem{floquet1883}

10213: M.~G. Floquet, Ann. \'Ecole Norm. Sup. {\bf 12},  47  (1883).

10214: 

10215: \bibitem{mermin76}

10216: N.~W. Ashcroft and N.~D. Mermin, {\em Solid State Physics} (Saunders College

10217:   Publishing, Fort Worth, 1976), college edition.

10218: 

10219: \bibitem{shirley65}

10220: J.~H. Shirley, \pr {\bf 138},  B979  (1965).

10221: 

10222: \bibitem{marion}

10223: J.~B. Marion, {\em Classical Dynamics of Particles and Systems}, 2nd ed.

10224:   (Academic Press, New York, 1970).

10225: 

10226: \bibitem{loudon}

10227: R. Loudon, {\em The Quantum Theory of Light}, 2nd ed. (Clarendon Press, Oxford,

10228:   1986).

10229: 

10230: \bibitem{abramowitz72}

10231:  in {\em Handbook of Mathematical Functions}, edited by M. Abrammowitz and

10232:   I.~A. Stegun (Dover, New York, 1972).

10233: 

10234: \bibitem{shakeshaft88}

10235: R. Shakeshaft, Z. Phys. D {\bf 8},  47  (1988).

10236: 

10237: \bibitem{abu95c}

10238: A. Buchleitner, D. Delande, and J.~C. Gay, \josab {\bf 12},  505  (1995).

10239: 

10240: \bibitem{cormier96}

10241: E. Cormier and P. Lambropoulos, J. Phys. B {\bf 29},  1667  (1996).

10242: 

10243: \bibitem{javanainen88}

10244: J. Javanainen, J.~H. Eberly, and Q. Su, \pr {\bf A38},  3430  (1988).

10245: 

10246: \bibitem{su91}

10247: Q. Su and J.~H. Eberly, \pr {\bf A44},  5997  (1991).

10248: 

10249: \bibitem{reed91}

10250: V.~C. Reed, P.~L. Knight, and K. Burnett, \prL {\bf 67},  1415  (1991).

10251: 

10252: \bibitem{burnett92}

10253: K. Burnett, V.~C. Reed, J. Cooper, and P.~L. Knight, \pr {\bf A45},  3347

10254:   (1992).

10255: 

10256: \bibitem{grobe93}

10257: R. Grobe and J.~H. Eberly, \pr {\bf A47},  R1605  (1993).

10258: 

10259: \bibitem{grobe94}

10260: R. Grobe, K. Rz\c{a}\.zewski, and J.~H. Eberly, J. Phys. B {\bf 27},  L503

10261:   (1994).

10262: 

10263: \bibitem{kusta65}

10264: P. Kustaanheimo and E. Stiefel, J. Reine Angew. Math. {\bf 218},  204  (1965).

10265: 

10266: \bibitem{delande84}

10267: D. Delande and J.-C. Gay, \jpb {\bf 17},  335  (1984).

10268: 

10269: \bibitem{rath88}

10270: O. Rath and D. Richards, \jpb {\bf 21},  555  (1988).

10271: 

10272: \bibitem{griffiths92}

10273: J.~A. Griffiths and D. Farrelly, \pr {\bf A45},  R2678  (1992).

10274: 

10275: \bibitem{gebarowski95}

10276: R. G{\c{e}}barowski and J. Zakrzewski, \pr {\bf A51},  1508  (1995).

10277: 

10278: \bibitem{schlagheck99}

10279: P. Schlagheck and A. Buchleitner, \phd {\bf 131},  110  (1999).

10280: 

10281: \bibitem{barut79}

10282: A.~O. Barut, C.~K.~E. Schneider, and R. Wilson, J.\ Math.\ Phys.\ {\bf 20},

10283:   2244  (1979).

10284: 

10285: \bibitem{chen80}

10286: A.~C. Chen, \pr {\bf A 22},  333  (1980).

10287: 

10288: \bibitem{chen81}

10289: A.~C. Chen, \pr {\bf A 23},  1655  (1981).

10290: 

10291: \bibitem{delandeth}

10292: D. Delande, Ph.D. thesis, Universit\'e de Paris, Paris, 1988, th\`ese de

10293:   doctorat d'etat.

10294: 

10295: \bibitem{lanczos}

10296: C. Lanczos, J. Res. Nat. Bur. Standards, Sect B {\bf 45},  225  (1950).

10297: 

10298: \bibitem{delande91}

10299: D. Delande, A. Bommier, and J.-C. Gay, \prL {\bf 66},  141  (1991).

10300: 

10301: \bibitem{ericsson80}

10302: T. Ericsson and A. Ruhe, Math. Comput. {\bf 35},  1251  (1980).

10303: 

10304: \bibitem{grimes94}

10305: R.~G. Grimes, J.~G. Lewis, and H.~D. Simon, SIAM J. Matrix Anal. Appl. {\bf

10306:   15},  228  (1994).

10307: 

10308: \bibitem{balslev71}

10309: E. Balslev and J.~M. Combes, Commun.\ math.\ Phys. {\bf 22},  280  (1971).

10310: 

10311: \bibitem{graffi85}

10312: S. Graffi, V. Grecchi, and H.~J. Silverstone, Ann.\ Inst.\ Henri Poincar\'e

10313:   {\bf 42},  215  (1985).

10314: 

10315: \bibitem{yajima82}

10316: K. Yajima, Comm.\ Math.\ Phys. {\bf 87},  331  (1982).

10317: 

10318: \bibitem{nicolaides78}

10319: C.~A. Nicolaides and D.~R. Beck, Int. J. Quant. Chem. {\bf XIV},  457  (1978).

10320: 

10321: \bibitem{reinhardt83}

10322: B.~R. Johnson and W.~P. Reinhardt, Phys. Rev. A {\bf 28},  1930  (1983).

10323: 

10324: \bibitem{ho83}

10325: Y.~K. Ho, Phys. Rep. {\bf 99},  1  (1983).

10326: 

10327: \bibitem{moiseyev98}

10328: N. Moiseyev, Phys. Rep. {\bf 302},  211  (1998).

10329: 

10330: \bibitem{abu94}

10331: A. Buchleitner, B. Gr\'emaud, and D. Delande, \jpb {\bf 27},  2663  (1994).

10332: 

10333: \bibitem{englefield72}

10334: M.~J. Englefield, {\em Group Theory and the Coulomb Problem} (Wiley, New-York,

10335:   1972).

10336: 

10337: \bibitem{goldstein80}

10338: H. Goldstein, {\em Classical Dynamics} (Addison-Wesley, Reading, Ma., 1980),

10339:   p.146.

10340: 

10341: \bibitem{jensen84}

10342: R.~V. Jensen, \pr {\bf A30},  386  (1984).

10343: 

10344: \bibitem{meerson82}

10345: B.~I. Meerson, E.~A. Oks, and P.~V. Sasorov, \jpb {\bf 15},  3599  (1982).

10346: 

10347: \bibitem{casati88}

10348: G. Casati, I. Guarneri, and D.~L. Shepelyansky, IEEE J.\ Quantum Electron. {\bf

10349:   24},  1420  (1988).

10350: 

10351: \bibitem{bayfield89}

10352: J.~E. Bayfield, G. Casati, I. Guarneri, and D.~W. Sokol, \prL {\bf 63},  364

10353:   (1989).

10354: 

10355: \bibitem{koch95b}

10356: P.~M. Koch and K.~A.~H. van Leeuwen, \phr {\bf 255},  289  (1995).

10357: 

10358: \bibitem{bellermann96}

10359: M.~R.~W. Bellermann, P.~M. Koch, D. Mariani, and D. Richards, \prL {\bf 76},

10360:   892  (1996).

10361: 

10362: \bibitem{bayfield74}

10363: J.~E. Bayfield and P.~M. Koch, \prL {\bf 33},  258  (1974).

10364: 

10365: \bibitem{bayfield96}

10366: J.~E. Bayfield, S.~Y. Luie, L.~C. Perotti, and M.~P. Skrzypkowski, \pr {\bf

10367:   A53},  R12  (1996).

10368: 

10369: \bibitem{galvez88}

10370: E.~J. Galvez {\it et~al.}, \prL {\bf 61},  2011  (1988).

10371: 

10372: \bibitem{fu90}

10373: P. Fu, T.~J. Scholz, J.~M. Hettema, and T.~F. Gallagher, \prL {\bf 64},  511

10374:   (1990).

10375: 

10376: \bibitem{cheng96}

10377: C.~H. Cheng, C.~Y. Lee, and T.~F. Gallagher, \pr {\bf A54},  3303  (1996).

10378: 

10379: \bibitem{delande97b}

10380: D. Delande and J. Zakrzewski,  in {\em Classical, Semiclassical and Quantum

10381:   Dynamics in Atoms}, No.~485 in {\em Lecture Notes in Physics}, edited by H.

10382:   Friedrich and B. Eckhardt (Springer, New York, 1997), p.\ 205.

10383: 

10384: \bibitem{bunkin64}

10385: F.~V. Bunkin and A.~M. Prokhorov, Sov.\ Phys.\ JETP {\bf 19},  739  (1964).

10386: 

10387: \bibitem{grozdanov92}

10388: T.~P. Grozdanov, M.~J. Rakovi\'c, and E.~A. Solovev, \jpb {\bf 25},  4455

10389:   (1992).

10390: 

10391: \bibitem{klar89}

10392: H. Klar, \zpd {\bf 11},  45  (1989).

10393: 

10394: \bibitem{lee95}

10395: E. Lee, A.~F. Brunello, and D. Farrelly, \prL {\bf 75},  3641  (1995).

10396: 

10397: \bibitem{abu95a}

10398: A. Buchleitner and D. Delande, \csf {\bf 5},  1125  (1995).

10399: 

10400: \bibitem{heller84}

10401: E.~J. Heller,  in {\em Classical and Quantum Chaos}, edited by A. Voros, M.

10402:   Giannoni, and A. Zinn-Justin (Elsevier, Amsterdam, 1991), Chap.~Scars.

10403: 

10404: \bibitem{bogomolny88a}

10405: E.~B. Bogomolny, \phd {\bf 31},  169  (1988).

10406: 

10407: \bibitem{Bth}

10408: A. Brunello, Ph.D. thesis, State University of New York at Stony Brook, Stony

10409:   Brook, 1997, unpublished.

10410: 

10411: \bibitem{sacha97}

10412: K. Sacha and J. Zakrzewski, \pr {\bf A56},  719  (1997).

10413: 

10414: \bibitem{sacha98c}

10415: K. Sacha and J. Zakrzewski, \pr A {\bf 58},  488  (1998).

10416: 

10417: \bibitem{brunello97}

10418: A.~F. Brunello, T. Uzer, and D. Farrelly, \pr A {\bf 55},  3730  (1997).

10419: 

10420: \bibitem{leopold86}

10421: J.~G. Leopold and D. Richards, \jpb {\bf 19},  1125  (1986).

10422: 

10423: \bibitem{leopold87}

10424: J.~G. Leopold and D. Richards, \jpb {\bf 20},  2369  (1987).

10425: 

10426: \bibitem{rakovic98}

10427: M.~J. Rakovi\'c, T. Uzer, and D. Farrelly, \pr A {\bf 57},  2814  (1998).

10428: 

10429: \bibitem{benvenuto91}

10430: F. Benvenuto, G. Casati, I. Guarneri, and D.~L. Shepelyansky, \zp B {\bf 84},

10431:   159  (1991).

10432: 

10433: \bibitem{oliveira94}

10434: C.~R. de~Oliveira, G. Casati, and I. Guarneri, \epl {\bf 27},  187  (1994).

10435: 

10436: \bibitem{flatte96}

10437: M.~E. Flatt\'e and M. Holthaus, \anp {\bf 245},  113  (1996).

10438: 

10439: \bibitem{steane95}

10440: A. Steane, P. Szriftgiser, P. Desbiolles, and J. Dalibard, \prL {\bf 74},  4972

10441:    (1995).

10442: 

10443: \bibitem{oberthaler99}

10444: M.~K. Oberthaler {\it et~al.}, \prL  4447  (1999).

10445: 

10446: \bibitem{bonci98}

10447: L. Bonci, A. Farusi, P. Grigilini, and R. Roncaglia, \pr E {\bf 58},  5689

10448:   (1998).

10449: 

10450: \bibitem{brodier01}

10451: O. Brodier, P. Schlagheck, and D. Ullmo, \prL {\bf 87},  64101  (2001).

10452: 

10453: \bibitem{paul92}

10454: W. Paul, Rev. Mod. Phys. {\bf 62},  531  (1992).

10455: 

10456: \bibitem{benvenuto94}

10457: F. Benvenuto, G. Casati, and D.~L. Shepelyansky, \prL {\bf 72},  1818  (1994).

10458: 

10459: \bibitem{tanner00}

10460: G. Tanner, K. Richter, and J.~M. Rost, \rmp {\bf 72},  497  (2000).

10461: 

10462: \bibitem{wintgen93}

10463: D. Wintgen and D. Delande, \jpb {\bf 26},  L399  (1993).

10464: 

10465: \bibitem{gremaud97}

10466: B. Gr\'emaud and D. Delande, \epl {\bf 40},  363  (1997).

10467: 

10468: \bibitem{gremaudth}

10469: B. Gr\'emaud, Ph.D. thesis, Universit\'e Paris 6, 1997.

10470: 

10471: \bibitem{puttner01}

10472: R. P\"uttner {\it et~al.}, \prL {\bf 86},  3747  (2001).

10473: 

10474: \bibitem{eichmann90}

10475: U. Eichmann, V. Lange, and W. Sandner, \prL {\bf 64},  274  (1990).

10476: 

10477: \bibitem{richter90}

10478: K. Richter and D. Wintgen, \prL {\bf 65},  1965  (1990).

10479: 

10480: \bibitem{schlagheck98a}

10481: P. Schlagheck and A. Buchleitner, \jpb {\bf 31},  L489  (1998).

10482: 

10483: \bibitem{schlagheck99b}

10484: P. Schlagheck and A. Buchleitner, \epl {\bf 46},  24  (1999).

10485: 

10486: \bibitem{schlagheckth}

10487: P. Schlagheck, Ph.D. thesis, Technische Universit\"at, M\"unchen, 1999,

10488:   published by Herbert Utz Verlag.

10489: 

10490: \bibitem{hanson95}

10491: L.~G. Hanson and P. Lambropoulos, \prL {\bf 74},  5009  (1995).

10492: 

10493: \bibitem{zobay96}

10494: O. Zobay and G. Alber, \pr {\bf 54},  5361  (1996).

10495: 

10496: \bibitem{mecking98}

10497: B. Mecking and P. Lambropoulos, \pr {\bf A57},  2014  (1998).

10498: 

10499: \bibitem{carpetmen}

10500: M. Berry, I. Marzoli, and W.~P. Schleich, Physics World {\bf 14},    (2001).

10501: 

10502: \bibitem{stone85}

10503: P.~A. Lee and A.~D. Stone, \prL {\bf 55},  1622  (1985).

10504: 

10505: \bibitem{LB90}

10506: W.~A. Lin and L.~E. Ballentine, \prL {\bf 65},  2927  (1990).

10507: 

10508: \bibitem{LB92}

10509: W.~A. Lin and L.~E. Ballentine, \pr {\bf 45},  3637  (1992).

10510: 

10511: \bibitem{grossmann91}

10512: F. Grossmann, T. Dittrich, P. Jung, and P. H\"anggi, \prL {\bf 67},  516

10513:   (1991).

10514: 

10515: \bibitem{GDJH91b}

10516: F. Grossmann, T. Dittrich, P. Jung, and P. H\"anggi, Z.\ Phys.\ B\ {\bf 84},

10517:   315  (1991).

10518: 

10519: \bibitem{GDJH93}

10520: F. Grossmann {\it et~al.}, J.\ Stat.\ Phys.\ {\bf 70},  229  (1993).

10521: 

10522: \bibitem{plata92}

10523: J. Plata and J.~M.~G. Llorente, \jpa {\bf 25},  L303  (1992).

10524: 

10525: \bibitem{bohigas93}

10526: O. Bohigas, S. Tomsovic, and D. Ullmo, \phr {\bf 223},  43  (1993).

10527: 

10528: \bibitem{bohigas93b}

10529: O. Bohigas, D. Boos\'e, R.~E. de~Carvalho, and V. Marvulle, Nucl.\ Phys.\ {\bf

10530:   A560},  197  (1993).

10531: 

10532: \bibitem{tomsovic94}

10533: S. Tomsovic and D. Ullmo, \pr {\bf E50},  145  (1994).

10534: 

10535: \bibitem{shudo95}

10536: A. Shudo and K.~S. Ikeda, \prL {\bf 74},  682  (1995).

10537: 

10538: \bibitem{leyvraz96}

10539: F. Leyvraz and D. Ullmo, \jpa {\bf 29},  2529  (1996).

10540: 

10541: \bibitem{jackson}

10542: J.~D. Jackson, {\em Classical Electrodynamics} (Wiley, New York, 1975).

10543: 

10544: \bibitem{haroche92}

10545: S. Haroche,  in {\em Fundamental Systems in Quantum Optics, Le Houches, Session

10546:   LIII,1990}, edited by J. Dalibard, J.~M. Raimond, and J. Zinn-Justin

10547:   (Elsevier, New York, 1992).

10548: 

10549: \bibitem{lewenstein88}

10550: M. Lewenstein, J. Zakrzewski, T.~W. Mossberg, and J. Mostowski, \jpb {\bf 21},

10551:   L9  (1988).

10552: 

10553: \bibitem{lewenstein88b}

10554: M. Lewenstein, J. Zakrzewski, and T.~W. Mossberg, \pr {\bf A38},  808  (1988).

10555: 

10556: \bibitem{abu97b}

10557: D.~D.~R. Buchleitner, D. Delande, and J. Zakrzewski,  in {\em Non-spreading

10558:   Kn\"odel-packets in Microsoft fields, Proceedings of the Third International

10559:   Bavarian Conference on Kn\"odel-Packets}, No.~007 in {\em Food, physics and

10560:   politics}, edited by H. Schalther and W. Weich (Stoiber Comp., Aidling,

10561:   1997), p.\ 294561.

10562: 

10563: \bibitem{nakamura93}

10564: K. Nakamura, {\em Quantum chaos, a new paradigm of nonlinear dynamics}, {\em

10565:   Cambridge Nonlinear Science Series 3} (Cambridge University Press, Cambridge,

10566:   1993).

10567: 

10568: \bibitem{casati87b}

10569: G. Casati, I. Guarneri, and D.~L. Shepelyansky, \pr {\bf A36},  3501  (1987).

10570: 

10571: \bibitem{leopold78}

10572: J.~G. Leopold and I.~C. Percival, \prL {\bf 41},  944  (1978).

10573: 

10574: \bibitem{bayfield88a}

10575: J.~E. Bayfield and D.~W. Sokol, \prL {\bf 61},  2007  (1988).

10576: 

10577: \bibitem{bluemel89a}

10578: R. Bl\"umel {\it et~al.}, \prL {\bf 62},  341  (1989).

10579: 

10580: \bibitem{bluemel91a}

10581: R. Bl\"umel {\it et~al.}, Phys. Rev. {\bf A44},  4521  (1991).

10582: 

10583: \bibitem{abu91}

10584: A. Buchleitner, L. Sirko, and H. Walther, \epl {\bf 16},  35  (1991).

10585: 

10586: \bibitem{casati84}

10587: G. Casati, B.~V. Chirikov, and D.~L. Shepelyansky, \prL {\bf 53},  2525

10588:   (1984).

10589: 

10590: \bibitem{fishman82}

10591: S. Fishman, D.~R. Grempel, and R.~E. Prange, \prL {\bf 49},  509  (1982).

10592: 

10593: \bibitem{grempel84}

10594: D.~R. Grempel, R.~E. Prange, and S. Fishman, \pr {\bf A29},  1639  (1984).

10595: 

10596: \bibitem{brenner96}

10597: N. Brenner and S. Fishman, \prL {\bf 77},  3763  (1996).

10598: 

10599: \bibitem{abu98b}

10600: A. Buchleitner, I. Guarneri, and J. Zakrzewski, Europhys. Lett. {\bf 44},  162

10601:   (1998).

10602: 

10603: \bibitem{wimbergerda}

10604: S. Wimberger, Master's thesis, Ludwig-Maximilians-Universit\"at M\"unchen,

10605:   2000.

10606: 

10607: \bibitem{wimberger01}

10608: S. Wimberger and A. Buchleitner, \jpa {\bf 34},  7181  (2001).

10609: 

10610: \bibitem{sirko96}

10611: L. Sirko, A. Haffmans, M.~R.~W. Bellermann, and P.~M. Koch, \epl {\bf 33},  181

10612:    (1996).

10613: 

10614: \bibitem{sirko93a}

10615: L. Sirko {\it et~al.}, \prL {\bf 71},  2895  (1993).

10616: 

10617: \bibitem{sauer92b}

10618: B.~E. Sauer, M.~R.~W. Bellermann, and P.~M. Koch, \prL {\bf 68},  1633  (1992).

10619: 

10620: \bibitem{leeuwen85}

10621: K.~A.~H. van Leeuwen {\it et~al.}, \prL {\bf 55},  2231  (1985).

10622: 

10623: \bibitem{koch95a}

10624: P.~M. Koch, Physica D {\bf 83},  178  (1995).

10625: 

10626: \bibitem{koch92b}

10627: P.~M. Koch,  in {\em Chaos and Quantum Chaos}, Vol.~411 of {\em Lecture Notes

10628:   in Physics}, edited by W.~D. Heiss (Springer, Berlin, 1992).

10629: 

10630: \bibitem{koch92a}

10631: P.~M. Koch, \ch {\bf 2},  131  (1992).

10632: 

10633: \bibitem{koch89}

10634: P.~M. Koch {\it et~al.}, \psc {\bf T26},  51  (1989).

10635: 

10636: \bibitem{noel00}

10637: M.~W. No\"el, M.~W. Griffith, and T.~F. Gallagher, \pr {\bf A62},  063401

10638:   (2000).

10639: 

10640: \bibitem{bayfield91}

10641: J.~E. Bayfield, \ch {\bf 1},  110  (1991).

10642: 

10643: \bibitem{kuba95b}

10644: J. Zakrzewski and D. Delande, J. Phys. B {\bf 28},  L667  (1995).

10645: 

10646: \bibitem{abu95b}

10647: A. Buchleitner {\it et~al.}, Phys. Rev. Lett. {\bf 75},  3818  (1995).

10648: 

10649: \bibitem{arndt91}

10650: M. Arndt, A. Buchleitner, R.~N. Mantegna, and H. Walther, \prL {\bf 67},  2435

10651:   (1991).

10652: 

10653: \bibitem{benson95}

10654: O. Benson {\it et~al.}, \pr {\bf A51},  4862  (1995).

10655: 

10656: \bibitem{noel00b}

10657: M.~W. No\"el, M.~W. Griffith, and T.~F. Gallagher, \prL {\bf 87},  043001

10658:   (2001).

10659: 

10660: \bibitem{bluemel90a}

10661: R. Bl\"umel and U. Smilansky, \josab {\bf 7},  664  (1990).

10662: 

10663: \bibitem{krug00}

10664: A. Krug and A. Buchleitner, \epl {\bf 49},  176  (2000).

10665: 

10666: \bibitem{krug01}

10667: A. Krug and A. Buchleitner, \prL {\bf 86},  3538  (2001).

10668: 

10669: \bibitem{breuer89b}

10670: H.~P. Breuer and M. Holthaus, \pla {\bf 140},  507  (1989).

10671: 

10672: \bibitem{bayfield85}

10673: J.~E. Bayfield and L.~A. Pinnaduwage, \prL {\bf 54},  313  (1985).

10674: 

10675: \bibitem{reinhold01}

10676: C.~O. Reinhold {\it et~al.}, \jpb {\bf 34},  L551  (2001).

10677: 

10678: \bibitem{gerber98}

10679: A. Assion {\it et~al.}, Science {\bf 282},  919  (1998).

10680: 

10681: \bibitem{gerber01}

10682: T. Brixner, N.~H. Damrauer, P. Niklaus, and G. Gerber, Nature {\bf 414},  57

10683:   (2001).

10684: 

10685: \end{thebibliography}

10686: 

10687: \end{document}

10688: