1: % Version 3.22 18/12/2001
2: \documentstyle[pra,aps,epsfig,twocolumn]{revtex}
3:
4:
5: %\documentstyle[preprint,aps,epsfig]{revtex}
6: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
7: %TCIDATA{OutputFilter=LATEX.DLL}
8: %TCIDATA{LastRevised=Tue Oct 16 12:06:50 2001}
9: %TCIDATA{<META NAME="GraphicsSave" CONTENT="32">}
10: %TCIDATA{CSTFile=revtex.cst}
11:
12: \begin{document}
13: \twocolumn[\hsize\textwidth\columnwidth\hsize\csname@twocolumnfalse\endcsname
14: \title{Theoretical analysis of quantum dynamics in 1D lattices:
15: Wannier-Stark description}
16: \author{Quentin Thommen, Jean Claude Garreau and V{\'e}ronique Zehnl{\'e}}
17: \address{Laboratoire de Physique des Lasers, Atomes et Mol{\'e}cules,
18: UMR 8523, \\ and
19: Centre d'Etudes et de Recherches Laser et Applications,\\
20: Universit\'{e} des Sciences et Technologies de Lille, \\
21: F-59655 Villeneuve d'Ascq Cedex, France\\
22: }
23: \date{\today}
24: \maketitle
25:
26: \begin{abstract}
27: This papers presents a formalism describing the dynamics of
28: a quantum particle in a one-dimensional tilted time-dependent lattice.
29: The description uses the Wannier-Stark states,
30: which are localized in each site of the lattice and provides
31: a simple framework leading to
32: fully-analytical developments. Particular attention is devoted to
33: the case of a time-dependent potential, which results in a rich variety of
34: quantum coherent dynamics is found.\\
35: \vspace{0.5cm}
36: Pacs number(s): 03.65.-w, 03.75.-b, 32.80.Lg, 32.80.Pj
37: \end{abstract}
38: ]
39:
40: \section{Introduction}
41: \label{sec:intro}
42: Quantum dynamics in a periodic lattice is one of the oldest
43: problems of quantum mechanics, whose basis have been
44: settled by Bloch and Zener \cite{ref:Bloch,ref:Zener}, in the 30's. Aimed at
45: the description of the electron motion in crystalline lattices,
46: this problem has largely been considered, for about half a century, as
47: an academic one, because dissipation effects forbid the observation
48: of most quantum effects in the motion of a crystalline electron.
49: Laser cooling of atoms has brought a revival of the interest on such
50: problems, as it produces atoms
51: whose de Broglie wavelength is comparable to the wavelength of the light
52: interacting with the atoms, and whose resulting kinetic energy is comparable
53: to the typical lightshift induced by the radiation. The latter
54: feature means that the cold atoms can be trapped in light potentials (or dipole
55: potentials). The former means that the atom dynamics in such a potential
56: is, in absence of dissipation, essentially quantum. Moreover, the
57: main source of dissipation is spontaneous emission, which can be arbitrarily
58: reduced (if one disposes of a powerful enough laser), whereas keeping a
59: constant light potential, just by an increase of the laser-atom detuning
60: \cite{ref:LesHouches}.
61:
62: Light potentials are a consequence of the displacement of atomic levels
63: resulting from the interaction with light, corresponding to a process in
64: which a photon is absorbed transferring the atom to an (virtual)
65: excited state from which the atoms de-excites back to the
66: ground-state by {\it stimulated} emission. Such a process induces an
67: energy shift
68: of the atomic levels that can be deduced from second order perturbation
69: theory and which is proportional to light intensity, to the square of the
70: coupling (that is, to $|{\bf d_{eg} \cdot \epsilon}|^2$, where ${\bf d_{eg}}$
71: is the dipole matrix element between the states $g$ and $e$, and $%
72: {\bf \epsilon}$ is the polarization vector of the light), and to the inverse
73: of the laser-atom detuning $\delta_L=\omega_L-\omega_{eg}$
74: ($\omega_{eg}$ is the Bohr frequency between states $g$ and
75: $e$). Any spatial
76: gradient of this energy shift produces a (conservative) force, and thus a
77: potential. A simple example is that of a standing wave formed by two
78: counter-propagating parallel-polarized beams. Placed in such a standing wave,
79: an atom perceives a periodic one-dimensional potential whose strenght
80: varies sinusoidally in the space. Standing waves (with little variations)
81: form the model potential considered in the present work.
82:
83: Light potentials generated by standing waves have been used
84: in many experimental studies of quantum dynamics. For example, Bloch
85: oscillations have been observed both with single atoms \cite{ref:BlochOsc} and
86: with a Bose-Einstein condensate (BEC) \cite{ref:BlochOscBEC} in an accelerated
87: standing wave. Wannier-Stark ladders \cite{ref:WSLadders} and collective
88: tunneling effects \cite{ref:CollTunnel} have also been studied with such
89: a system. Atoms placed in an intense,
90: phase-modulated, or pulsed, standing wave realize a paradigmatic system for
91: theoretical and experimental studies of quantum chaos, the so-called
92: Quantum Kicked Rotor
93: \cite{ref:QKRRaizen,ref:QKRChristiensen,ref:QKRLille,ref:QKROxford}.
94:
95: In this paper, we consider the quantum dynamics of an atom (of mass $M$)
96: placed in a tilted sinusoidal potential whose phase (that is, the position of
97: its nodes) can be modulated in an arbitrary way, corresponding to
98: the Hamiltonian
99: \begin{equation}
100: H={\frac{p^{2}}{2M}}+v_{0} \cos \{ 2k_{L}
101: \left[x-x_{0}(t)\right] \} +f(t)x
102: \label{eq:H}
103: \end{equation}
104: where $x_{0}(t)$ is a phase and $f(t)$ a
105: force, both being (eventually) time-dependent, and
106: $k_{L}=2\pi/\lambda_L$ is the wavenumber
107: of the standing wave. Different temporal
108: dependences of $x_{0}(t)$ can be considered. For instance, the
109: accelerated case, $x_{0}(t)=(1/2)at^{2}$, which has been studied in \cite
110: {ref:BlochOsc,ref:WSLadders,ref:CollTunnel},
111: is equivalent to an inertial force $F=Ma$ in the frame of the
112: potential (see appendix \ref{app:UnitTransf}).
113:
114: A natural energy unit in such a context is the ``recoil energy'',
115: defined as the change in kinetic energy of the atom corresponding
116: to the absorption a photon, given by
117: \begin{equation}
118: E_{R}={\frac{\hbar ^{2}k_{L}^{2}}{2M}} \label{eq:RecoilEnergy}
119: \end{equation}
120: to which one can associate a recoil frequency $\omega _{R}=E_{R}/\hbar $ and
121: a recoil momentum $p_{R}=\hbar k_{L}$, etc. It is also useful to re-scale the
122: variables: $X \equiv x/(\lambda_L/2)$, where $\lambda_L/2$ is the step of the
123: periodic lattice, and $\tau \equiv \omega _{R}t$. With these
124: definitions the above Hamiltonian takes the following form, which is
125: retained in the rest of the paper:
126: \begin{equation}
127: H={\frac{P^{2}}{2m^*}}+V_{0}\cos \{ 2\pi[X-X_{0}(\tau )]\}+F(\tau)X
128: \label{H_ren}
129: \end{equation}
130: where $V_{0} \equiv v_{0}/E_{r}$, $F \equiv f\lambda_L/ 2E_R$
131: (note that in this
132: system the momentum operator in the real space is
133: $P=-i(\partial/\partial X)$, $\hbar=1$, the reduced mass is
134: $m^*=\pi^2/2$ and $d=1$ is the step of the lattice).
135: For simplicity, in what follows, we shall write the rescaled
136: variables $x$, $p$ and $t$.
137:
138: In the next section, we briefly review the time-independent Hamiltonian case
139: in Eq.~(\ref{H_ren}), introduce the
140: Wannier-Stark basis, and show that it leads to a very simple description
141: of the Bloch oscillation. In the following sections, we shall discuss the
142: more complicated dynamics that arises when a harmonic modulation of the lattice is
143: applied.
144:
145: \section{The Bloch oscillation in the Wannier-Stark description}
146: \label{sec:WSBasis}
147: The Bloch oscillation (BO) is a well known phenomenon
148: discovered by Zener while studying the quantum properties of an electron in
149: a (perfect) crystal submitted to a constant electric field \cite{ref:Zener}.
150: The BO arises when a small spatial tilt is added to the lattice:
151: Quantum particles do not fall
152: along the slope of the potential, but perform a periodic, space-limited,
153: oscillation. BO is thus a strictly quantum behavior. We shall call
154: ``lattice'' the untilted potential, and ``tilted lattice''
155: the sum of the lattice and the tilted potential.
156:
157: The BO is usually described in the basis of the so-called {\em Bloch
158: states} \cite{ref:Bloch},
159: i.e., the eigenstates of the lattice.
160: In this paper, we use another framework corresponding to
161: the eigenstates of the {\em tilted} potential. While the lattice
162: is invariant under spatial translations by a multiple of
163: the spatial period $d$, the tilted lattice has a more complicate symmetry:
164: it is invariant under simultaneous spatial translation by a lattice
165: period $d$ {\em and}
166: energy translation by $\omega_B = Fd$ ($\omega_B$ is the ``Bloch
167: frequency''). One then expects the eigenenergies to form
168: ``ladder'' structures separated by $\omega_B$, the so-called Wannier-Stark
169: ladders, introduced by Wannier in connection with
170: the problem of electrons in a crystal submitted to a homogeneous electric
171: field \cite{ref:Wannier}.
172: Each element of the ladder corresponds to eigenfunctions (the
173: Wannier-Stark states) centered at a given well, and thus separated
174: by an integer multiple of $d$. The
175: form, and even the existence of these eigenstates has been the object of a
176: long controversy, that has been settled only recently \cite{ref:Nenciu}. In
177: the present paper, we consider a spatially limited lattice, extending over many
178: periods, and limited by an infinite-height box. This changes only
179: very slightly the ``bulk'' properties of the system, and the
180: eigenenergies and eigenstates obtained numerically display (to a very good
181: approximation) the expected ladder structure described above.
182: In this framework, Wannier-Stark states (WSS) are
183: the eigenfunctions of the system.
184: The relation to the case of an infinite
185: tilted lattice makes no problem if the corresponding states
186: (Wannier-Stark resonances) have long enough life-times compared
187: to the experimental times. The existence of these states has been evidenced in 1988 in a
188: semi-conductor superlattice \cite{ref:WSLaddersSL}, and 1996 with cold atoms in
189: an optical lattice \cite{ref:WSLadders}. Note that, being
190: spatially localized, Wannier-Stark states provide an ideal tool for
191: the description of the wave function of a cold atom (as produced by
192: a ``Sisyphus-boosted" MOT) placed in the
193: potential, whose de Broglie wavelength (around $\lambda_L/3$) is
194: of the order of the lattice period $\lambda_L/2$.
195:
196: Consider the properties of the time-independent Hamiltonian $H_{0}$
197: \begin{equation}
198: H_{0}={\frac{p^{2}}{2m^*}}+V_{0}\cos (2\pi x)+Fx
199: \label{eq:H0}
200: \end{equation}
201: where $F$ is a constant force. The eigenfunctions of $H_{0}$ are
202: Wannier-Stark states forming an energy ladder whose separation is
203: $\omega_B=Fd$, $E_{nm}=E_m+nFd$.
204:
205: \vspace{0.5cm}
206: %Figure 1
207: \begin{figure}[tbp]
208: \begin{center}
209: \psfig{figure=fig1.eps,width=8cm,clip=}
210: \caption{(a) Periodic potential with a tilt. (b) The WSS
211: state $\varphi_n$ is localized in the well of index $n=1$
212: and has appreciable overlap with neighbor lattice sites $n=0,2$.
213: This eigenfunction is obtained for $V_0=2.5$ and $F=0.5$.
214: The ``numerical'' box includes 64 lattice sites.}
215: \label{fig:TiltedPotential}
216: \end{center}
217: \end{figure}
218: \vspace{0.5cm}
219:
220: The BO can be advantageously described by using
221: the Wannier-Stark state (WSS) localized inside a given
222: individual lattice well, corresponding to the lowest energy of
223: the states associated to this well (see Fig.~\ref{fig:TiltedPotential}).
224: Note that considering only the
225: ground state of each well is equivalent to the restriction to
226: the first Bloch band in the description based on Bloch states.
227: We also choose strong enough $F$ and $V_0$ to produce well-localized WSS
228: \cite{ref:Fukuyama}. The WSS associated to the lattice well labeled
229: $n$ is noted $\varphi_{n}(x)$ (supposed real) and the corresponding
230: eigenenergy is $E_{n}$ (we drop the index $m$). The symmetries of the
231: potential discussed above then imply:
232: \begin{equation}
233: \varphi _{n+p}(x) = \varphi _{n}(x-pd)
234: \label{eq:WSSTranslation}
235: \end{equation}
236: and
237: \begin{equation}
238: E_{n+p}=E_{n}+ pFd \text{ .}
239: \label{eq:EnergyTranslation}
240: \end{equation}
241:
242: Let us describe the atomic wave function by a superposition of WSS
243: \begin{equation}
244: \Psi (x,t)=\sum_n c_{n}(t)\varphi _{n}(x)
245: \label{eq:Psi_WS}
246: \end{equation}
247: with $c_n(t) = c_{n} e^{-iE_{n}t}$; $c_n$ being the amplitude at $t=0$.
248: The dynamical quantities of the atom can easily calculated. For
249: instance, the mean value of the atomic position operator is:
250: \begin{equation}
251: \langle x\rangle =\sum_n X_{n,n}\left| c_{n}\right| ^{2}+
252: \sum_{n\ < m} \left( X_{n,m}c_{n}^{\ast }c_{m}e^{i(E_{n}-E_{m})t}+c.c \right)
253: \label{eq:xmean}
254: \end{equation}
255: where $X_{n,m} \equiv \left\langle \varphi _{n}\left| x\right| \varphi
256: _{m}\right\rangle$. As long as the WSS $\varphi_n$ is well localized in
257: the respective well, we can keep only
258: nearest-neighbors contributions ($n=m\pm 1)$ and derive a simplified
259: expression:
260: \begin{equation}
261: \langle x\rangle =\overline{x}+ X_{n,n+1} \left( \sum_{n}c_{n}^{\ast
262: }c_{n+1}
263: e^{-i\omega_{B}t}+c.c \right)
264: \end{equation}
265: where $\overline{x} = \sum X_{n,n} \left| c_{n}\right|^{2} $ is the mean
266: position and $X_{n,n+1}=X_{0,1}=X_{0,-1}=X_{n,n-1}$ is independent of
267: $n$ \cite{note:Identities}. This
268: result evidences a counter-intuitive property of the quantum
269: motion in a tilted lattice: instead of falling along the slope,
270: the atom performs an oscillation
271: with frequency $\omega_B$ (the Bloch frequency). The amplitude of this
272: Bloch oscillation is proportional to $X_{0,1}$ and grows with the overlap
273: between neighbors WSS, i.e for a small slope $F$
274: and small lattice depth $V_{0}$. The physical origin of the BO
275: appears here clearly as an interference effect between neighbor sites, as
276: $c_n^* c_{n+1}$ is the coherence between the sites $n$ and
277: $n+1$. Note the lack of oscillations if
278: the atom is localized in only one well.
279: Eq.~(\ref{eq:xmean}) also predicts the occurrence
280: of harmonics with frequency $p\omega_B$ ($p$ integer)
281: but with smaller amplitudes since they involve the coupling
282: strength $X_{n,n+p}$ \cite{note:amplitude}.
283:
284: The description of the BO given here is
285: very different of the usual ``solid-state'' approach. There, the Bloch
286: states (eigenstates of the untilted lattice)
287: are taken as the reference basis. The oscillation is described in a
288: semi-classical frame as the periodic evolution of the atom's
289: quasi-momentum $Ft$ (in the first Brillouin zone) with period $T=2\pi/(Fd)$
290: \cite{ref:Mermin,ref:Holthaus}.
291: Although intuitive, this approach does not make clear the
292: role of quantum interference produced by the periodic lattice
293: structure as the basic
294: mechanism underlying the Bloch oscillation, which is evidenced in our approach.
295:
296: \section{The modulated potential in the Wannier-Stark description:
297: Resonant dynamics}
298: \label{sec:TimeDepWSS}
299:
300: With the existence of the natural frequency $\omega_B$ of the system
301: in mind, one is tempted
302: to investigate the quantum dynamics in presence of a harmonic
303: external forcing. The WSS approach proves to be very efficient,
304: since it allows a fully-analytical description.
305: After some general considerations, we study in this section the case
306: of resonant forcing, and show that it leads to a very rich and
307: interesting dynamics. The general (non-resonant) case will be treated in the
308: next section.
309:
310: Consider the Hamiltonian of Eq.~(\ref{H_ren}) with a constant force
311: $F$ and a lattice phase modulation
312: \begin{equation}
313: x_{0}(t )=a\sin (\omega t) \text{.}
314: \label{eq:modulation}
315: \end{equation}
316: The developments are simpler if we use a unitary transformation that
317: transforms the modulation in a time-dependent force. Physically,
318: this is equivalent to move to an accelerated
319: reference frame in which the lattice is at rest \cite{ref:BlochOsc,ref:Ben},
320: adding thus an inertial force (Appendix \ref{app:UnitTransf}). In this
321: frame, the new Hamiltonian is given by Eq.~(\ref{eq:H0}) plus a
322: time-dependent force $F'(t)$:
323: \begin{equation}
324: F^\prime(t )=m^*\frac{d^{2}x_{0}(t )}{dt^{2}}=
325: -m^*a\omega^{2}\sin (\omega t ) = -F_0 \sin(\omega t)
326: \label{eq:Finertial}
327: \end{equation}
328: where $F_0 \equiv m^*a\omega^2$ is the amplitude of the inertial force.
329: The harmonic time-dependence in Eq.~(\ref{eq:Finertial}) is the analog of an
330: AC electric field for electrons in a (perfect) crystal. The dynamics
331: in such a system
332: can be described in a quite simple fashion by writing the state of the atom
333: as a superposition of WSS, Eq.~(\ref{eq:Psi_WS}). The coefficients
334: $c_n(t)$ can be obtained by reporting Eq.~(\ref{eq:Psi_WS})
335: into the Schr{\"o}dinger equation
336: \begin{equation}
337: [ H_0-F_0 x \sin (\omega t) ] \Psi(x,t) =
338: i { \partial \Psi(x,t) \over \partial t}
339: \label{eq:ModulatedH}
340: \end{equation}
341: where $H_0$ is given by Eq.~(\ref{eq:H0}), with eigenstates
342: $\varphi_n(x)$. This produces the following set of
343: coupled differential equations for the $c_n(t)$:
344: \begin{equation}
345: \dot{c}_{n}(t)=-iE_{n}c_{n}(t)+i F_0 \sin(\omega t)
346: \sum_{m}X_{n,m}c_{m}(t)
347: \end{equation}
348: where $\dot{c}_n \equiv dc_n/dt$.
349: Neglecting temporarily the coupling
350: between different WSS, (i.e putting $X_{n,m}=X_{n,n} \delta_{m,n}$),
351: the amplitudes are obtained simply as $c_n=\exp [i\phi_n(t)]$, with the
352: time-dependent phase
353: \begin{equation}
354: \phi_n(t)=-E_n t-{F_0 X_{n,n} \over \omega} \cos(\omega t)
355: \label{eq:phi_n}
356: \end{equation}
357: where
358: $X_{n,n}=X_{0,0}+nd$, depends on the site index $n$ \cite{note:Identities}.
359: We now write $c_n(t)\equiv d_n(t) e^{i\phi_n(t)}$. The amplitudes
360: $d_n$ obey the following system of differential equations:
361: \begin{equation}
362: \dot {d}_{n}=i F_0 \sum_{m \neq n}X_{n,m}d_{m}(t)
363: \exp \{ i \left[\phi_{m}(t)-\phi_{n}(t) \right]\} \sin(\omega t)
364: \text{ .}
365: \label{eq:dn0}
366: \end{equation}
367: After Eq.~(\ref{eq:phi_n}), the phase difference $\phi_m(t)-\phi_n(t)$ is:
368: \begin{equation}
369: \phi_{m}(t)-\phi_{n}(t)=(n-m) \left[\omega_Bt +{ F_0 d \over \omega }
370: \cos(\omega t) \right]
371: \end{equation}
372: where we used Eq. (\ref{eq:EnergyTranslation}).
373: Eq.~(\ref{eq:dn0}) can be recast as
374: \begin{mathletters}
375: \begin{equation}
376: \dot {d}_{n} =
377: i F_0 \sum_{p \neq 0} X_p d_{n+p} \left[ e^{-ip\omega_Bt}
378: e^{-i(pF_0d/\omega) \cos(\omega t)} \right] \sin(\omega t)
379: \label{eq:dn1}
380: \end{equation}
381: \begin{eqnarray}
382: = { F_0 \over 2 } \sum_{p \neq 0}X_p d_{n+p}
383: \sum_{l} (-i)^l J_l \left(p{ F_0 d\over \omega}\right) \nonumber \\
384: \{ e^{ i \left[(l+1)\omega -p\omega_B \right] t }-
385: e^{ i \left[(l-1)\omega -p\omega_B \right] t }\}
386: \label{eq:dn2}
387: \end{eqnarray}
388: \label{eq:dn}
389: \end{mathletters}
390: where, $X_p\equiv X_{n,n+p}$ \cite{note:Identities}. $J_n(x)$ is the Bessel
391: function of the first kind, and we have used the well-known property
392: of Bessel functions:
393: \begin{equation}
394: e^{-iz \cos(\omega t)} = \sum_{l=-\infty}^{+\infty}
395: J_l(z) (-i)^l e^{il\omega t} \text{ .}
396: \label{eq:BesselGenerator}
397: \end{equation}
398: Note that the sum over lattice sites (i.e over $p$) extends
399: only over a few neighbors sites, since the coupling coefficients $X_p$ rapidly
400: shrink to zero \cite{note:amplitude}. If the
401: modulation is smooth enough to avoid projections on other
402: states of the system, the sum over
403: the harmonics of the modulation (i.e. $l$) is also limited to a few terms
404: close to $l=0$ (typically, $l_{max} \sim p F_0 d/ \omega =
405: pm^*a\omega d \sim O(1)$).
406: Therefore, only a finite number of terms are to be retained in the above
407: expression. On the other hand, the evolution of $d_n$ described by
408: Eqs. (\ref{eq:dn2}) is a sum of oscillations with frequencies
409: $\left[(l\pm 1)\omega-p\omega_B \right]$. In the following we keep
410: only the so-called {\em secular} terms, that is,
411: terms that oscillate slowly or do not oscillate at all. The resulting
412: ``close to resonance" dynamics is observed when $(l\pm 1)\omega
413: \approx p\omega_B $, i.e when the forcing frequency $\omega$ is
414: commensurable (or almost) with the system's natural frequency $\omega_B$.
415:
416: Let us consider the simpler resonant case, $\omega = \omega _{B}$.
417: Due to the relative strength of the factors $X_{p}$ we can keep,
418: to a good accuracy, only the contribution of the
419: next-neighbor site ($p=1$), which leads to the
420: following expression:
421: \begin{equation}
422: \dot {d}_{n}(t)=\Omega_1 \left[ d_{n+1}-d_{n-1} \right]
423: \label{eq:dnsimple}
424: \end{equation}
425: where
426: \begin{eqnarray}
427: \Omega_1 = { F_0 X_1 \over 2 } \left[
428: J_0\left( { F_0 d \over \omega_B } \right)
429: + J_2\left({ F_0 d \over \omega_B }\right) \right] \nonumber \\
430: = { \omega_B X_1 \over d } J_1 \left({F_0 d \over \omega_B} \right){ .}
431: \label{eq:Omega1}
432: \end{eqnarray}
433: This equation is similar to a ``dipole coupling" between sites
434: $n$ and $n \pm 1$ where $\Omega_1$ plays the role of a Rabi frequency.
435: Note that, contrary to intuition, the coupling towards the left
436: or towards the right neighbor is the same.
437:
438: The meaning of Eq.(\ref{eq:dnsimple}) can be better appreciated
439: by searching for the plane-wave solutions of the form:
440: \begin{equation}
441: d_{n}(t)=e^{i\left( k_{0}d n+\omega t\right) } \text{ .}
442: \label{eq:WSWave}
443: \end{equation}
444: The (dimensionless) wavenumber $k_0$ takes into account the
445: phase difference between neighbor
446: sites. Substitution into Eq.~(\ref{eq:dnsimple})
447: leads to the dispersion relation
448: \begin{equation}
449: \omega =2 \Omega_1 \sin(k_{0}d)
450: \label{eq:dispers}
451: \end{equation}
452: and to the group velocity $v_{g}\equiv d\omega /dk_{0}:$
453: \begin{equation}
454: v_{g}=2\Omega_1 d\cos(k_{0}d) \text{ .}
455: \label{eq:vg}
456: \end{equation}
457: This result shows that the dynamics is different
458: depending on the wave number $k_{0}$. For instance, if $k_{0}d=\pm \pi /2$
459: (phase-quadrature from site to site), $v_{g}=0$ and there is no global motion.
460: If $k_{0}d=\pi$, $v_{g}=-2\Omega_1 d$ and the global motion is a fall along
461: the slope of the potential with the maximum speed
462: $2\Omega_1 d$. More interesting is the case $k_{0}=0$, where
463: $v_{g}=2\Omega_1 d$: the atom then climbs up the slope of the potential with
464: a constant maximum speed: there is, in this case, {\em coherent transfer
465: of energy} from the modulation to atom, thanks to
466: the particular phase relations between neighbor sites. Note also that,
467: contrary to the motion of a classical particle, the speed
468: $\left| v_{g}\right|$ is independent of the sense of displacement:
469: the wavepacket climbs the slope up or down at the same speed.
470:
471: More detailed information on the wavepacket motion can be grabbed by
472: writing the amplitude $d_n$ in the more general form:
473: \begin{equation}
474: d_{n}(t)=f_{n}(t)e^{i\left( k_{0}dn+\omega t \right) }
475: \end{equation}
476: where $f_{n}$ are complex amplitudes describing the envelope of the
477: atomic wavepacket, assumed to vary slowly in time as
478: compared to the frequency $\omega_B$, and in space as compared
479: to the lattice period $d$. Reporting the
480: above expression into Eq.~(\ref{eq:dnsimple}), we get:
481: \begin{eqnarray}
482: \dot{f}_{n}+i\omega f_{n}=\Omega_1 \left[ \cos (k_{0}d)\left(
483: f_{n+1}-f_{n-1}\right) \right. \nonumber \\
484: \left. +i\sin (k_{0}d)\left(f_{n+1}+f_{n-1}\right) \right]
485: \text{ .}
486: \end{eqnarray}
487: Using the dispersion relation Eq.~(\ref{eq:dispers}) and
488: keeping only slowly varying contributions, we obtain:
489: \begin{eqnarray}
490: \dot {f}_{n}=\Omega_1 \left[ \cos (k_{0}d)\left(
491: f_{n+1}-f_{n-1}\right) \right. \nonumber \\
492: \left.
493: +i\sin (k_{0}d)\left( f_{n+1}+f_{n-1}-2f_{n}\right) \right]
494: \text{ .}
495: \label{eq:slowfn}
496: \end{eqnarray}
497: Since $f_{n}$ vary slowly in space, we can take the continuous limit
498: (with respect to the variable $x=nd$) and deduce the following
499: equation for the wavepacket envelope:
500: \begin{equation}
501: \dot{f}(x,t) =\Omega_1 \left( 2d\cos (k_{0}d) {\frac{\partial
502: }{\partial x}}
503: +id^2\sin(k_{0}d) {\frac{ \partial^2 }{\partial x^2 }} \right) f(x,t)
504: \text{ .}
505: \label{eq:PDE}
506: \end{equation}
507: This equation is an interesting piece of information. If
508: $k_{0}d=\pm \pi/2$, one has a diffraction equation of the form $\dot{f}
509: =\pm i\Omega_1 d^2 \partial_{x}^{2}f$, which describes the spreading of
510: the atomic wavepacket (i.e. diffraction) with no global
511: displacement. The case $k_{0}d=0,\pi$ gives the wave equation $%
512: \dot{f}=v_g \partial_x f$ ($v_g=2d\Omega_1$), describing a
513: wavepacket traveling with constant velocity $v_{g}$ and no deformation: the
514: wavepacket presents an ascending or descending coherent motion. Mixed
515: behaviors, i.e. spreading at a diffusion rate $\Omega_1 d^2 \sin (k_{0}d)$
516: and uniform displacement with group velocity $v_g=2\Omega_1 d\cos (k_0d)$,
517: are found for other values of $k_{0}$. The general
518: solution can be easily obtained by performing a spatial Fourier transform
519: of Eq.~(\ref{eq:PDE}).
520:
521: \vspace{0.5cm}
522: %Figure 2
523: \begin{figure}[tbp]
524: \begin{center}
525: \psfig{figure=fig2.eps,width=8cm,clip=}
526: \caption{The atomic
527: wavepacket is obtained from a full integration of the
528: Schr{\"o}dinger equation corresponding to the modulated
529: tilted lattice. The wavepacket at $t=0$ has a site-to-site
530: phase of $\pi/2$. The wavepacket spreads in time with
531: no displacement, in an agreement with the theoretical prediction.
532: Parameters are $V_0=2.5$, $F=0.5$, $\omega=\omega_B=0.5$
533: and $a=0.2$}
534: \label{fig:ResDynQuadrature}
535: \end{center}
536: \end{figure}
537: \vspace{0.5cm}
538: %Figure 3
539: \begin{figure}[tbp]
540: \begin{center}
541: \psfig{figure=fig3.eps,width=8cm,clip=}
542: \caption{Same as Fig.~\protect\ref{fig:ResDynQuadrature}, except
543: that the site-to-site phase is $k_0=0$. The wavepacket has a
544: shape-preserving motion upwards the slope of the potential. The
545: observed group velocity is 0.030, in good agreement with the
546: prediction of Eq.~(\protect\ref{eq:vg}), $v_g=0.032$ ($X_{0,1}=0.13$).}
547: \label{fig:ResDynInPhase}
548: \end{center}
549: \end{figure}
550: \vspace{0.5cm}
551:
552: The above result shows that, depending on the initial wavenumber $k_{0}$,
553: i.e. on the way the initial wave packet is prepared, the atom behaves in
554: qualitatively different ways. Figs. \ref{fig:ResDynQuadrature} and
555: \ref{fig:ResDynInPhase} are obtained by a direct
556: integration of the Schr\"{o}dinger equation, with the Hamiltonian given by
557: Eq.~(\ref{eq:ModulatedH}), for different evolution times.
558: In the first case the initial wavepacket is prepared
559: with site-to-site phase $k_0d=\pi/2$ and diffraction is clearly visible,
560: as predicted. In Fig.~\ref{fig:ResDynInPhase}, the wavepacket, prepared
561: with $k_0=0$ has a uniform displacement while preserving its shape.
562: Its group velocity obtained from the numerical simulations is $v_g=0.030$,
563: in very good agreement with the theoretical
564: value from Eq.~(\ref{eq:vg}) which is $v_g=0.032$.
565: \vspace{0.5cm}
566: %Figure 4
567: \begin{figure}[tbp]
568: \begin{center}
569: \psfig{figure=fig4.eps,width=8cm,clip=}
570: \caption{Evolution of a wave packet for $\omega=2\omega_B$. The initial
571: wavepacket is a superposition of $\varphi_n$ ($n$ odd) which are
572: in phase, and of $\varphi_n$ ($n$ even) with phase difference $\pi$.
573: The initial wavepacket separates in two packets with opposite velocities.}
574: \label{fig:HarmonicDyn}
575: \end{center}
576: \end{figure}
577: \vspace{0.5cm}
578: Other resonant behaviors are observed if $\omega=q\omega_B$
579: ($q$ integer). From the general expression of Eq.~(\ref{eq:dn}),
580: one finds for instance a next-to-neighbor ($n \rightarrow n\pm 2$)
581: resonant interaction if $\omega=2\omega_B$, leading to:
582: \begin{equation}
583: \dot{d}_{n}(t)=\Omega_2 \left[ d_{n+2}-d_{n-2}\right]
584: \label{eq:DynNextNeig2}
585: \end{equation}
586: with $\Omega_2 = (\omega_B/d) X_2 J_1(F_0 d/\omega_B)$.
587: Note that $X_{2} \ll X_1$ \cite{note:amplitude}.
588: We illustrate in Fig.~\ref{fig:HarmonicDyn} the temporal evolution
589: of a wavepacket, obtained by numerical integration of the Schr\"{o}dinger
590: equation. The initial state is prepared as a superposition
591: of two packets: one packet is constructed with in-phase amplitudes
592: (that is, $c_n$ and $c_{n+2}$ have the same phase for $n$ odd),
593: and the other one is constructed with amplitudes in phase-opposition
594: (that is, $c_n$ and $c_{n+2}$ have opposite sign for $n$ even).
595: The first packet
596: moves with velocity $v_g=4\Omega_2 d$, and the second with $v_g=-4\Omega_2 d$.
597: The figure displays an original behavior showing each of these
598: two initially inter-penetrated packets moving independently
599: in opposite directions, creating a highly delocalized state.
600:
601: \section{The modulated potential: general case}
602: \label{sec:GeneralCase}
603: In this section we generalize the results of the preceding section
604: to the case of a non-resonant modulation. We follow
605: essentially the same steps as in Sec.~\ref{sec:TimeDepWSS}, and
606: we shall skip algebraic details of the calculations.
607: Coming back to Eq.~(\ref{eq:dn1}) and looking for a solution of the form:
608: \begin{equation}
609: d_n(t)=e^{i[(k_0 d n +\phi(t)]}
610: \label{eq:WaveHR}
611: \end{equation}
612: one gets the instantaneous frequency:
613: \begin{eqnarray}
614: \dot{\phi}
615: &=& 2 F_0 \sum_{p > 0}X_{p}
616: \cos \{p[k_0 d -\theta(t)]\} \sin(\omega t)
617: \end{eqnarray}
618: where $\theta(t) = \omega_Bt +(F_0 d/\omega) \cos(\omega t)$, and
619: we used $X_{p}=X_{-p}$. The group velocity
620: $v_g=d\dot{\phi}/dk_0$ is thus
621: \begin{equation}
622: v_g= 2F_0 d \sum_{p > 0}pX_{p} \sin \{p[\theta(t)-k_0 d]\}
623: \sin(\omega t)
624: \label{eq:vegeneral}
625: \end{equation}
626: As in the preceding section, more detailed behavior is
627: obtained by putting
628: \begin{equation}
629: d_n(t)=f_n(t)e^{i[k_0dn+\phi(t)]}
630: \end{equation}
631: where $f_n$ are slowly-varying amplitudes.
632: The generalization of Eq.~(\ref{eq:slowfn}) is then:
633: \begin{equation}
634: \dot{f}_{n}=
635: i F_0 \sum_{p \neq 0} X_{p} [f_{n+p}-f_n] e^{ip(k_0d-\theta)}
636: \sin(\omega t)
637: \end{equation}
638: or,
639: \begin{eqnarray}
640: \dot{f}_{n} = F_0 \sum_{p > 0}X_{p} \sin(\omega t) \nonumber \\
641: \{
642: - \left[f_{n+p}-f_{n-p} \right] \sin[p(k_0d-\theta)] \nonumber \\
643: +i\left[ (f_{n+p}+f_{n-p}-2f_n) \right] \cos[p(k_0d-\theta)] \} \text{ .}
644: \end{eqnarray}
645: Taking the continuous limit of the above expression then
646: produces an equation describing both the propagation and the
647: diffraction of the wavepacket:
648: \begin{eqnarray}
649: \dot{f}(x,t) =\left(v_g(t) {\frac{\partial }{\partial x}}
650: +iD(t) {\frac{ \partial^2 }{\partial x^2 }} \right) f(x,t) \nonumber\\
651: +2i F_0 \sum_{p \ge 2} X_p \cos[p(\theta-k_0d)]\sin(\omega t) f(x,t)
652: \label{eq:PDEHR}
653: \end{eqnarray}
654: where
655: \begin{equation}
656: D(t)= F_0 d^2 \sum_{p>0} p^2 X_p \cos[p(\theta-k_0d)]
657: \sin(\omega t)
658: \end{equation}
659: and the group velocity $v_g$ is given by Eq.~(\ref{eq:vegeneral}).
660: Note that the last term in Eq.(\ref{eq:PDEHR}) is a phase term which
661: is $O(X_2)<<1$ and does not
662: contribute to the probability density $|f(x,t)|^2$. For the sake of
663: lightness, it is not considered in the following.
664:
665: The Fourier transform of Eq.~(\ref{eq:PDEHR}) with respect to $x$
666: produces an algebraic equation for the Fourier transform
667: $\tilde{f}(k,t)$ of $f(x,t)$, whose solution is:
668: \begin{equation}
669: \tilde{f}(k,t) = e^{ik x^\prime(t)} e^{ik^2 \Delta(t)} \tilde{f}(k,0)
670: \end{equation}
671: and thus:
672: \begin {equation}
673: f(x,t)= { 1 \over \sqrt{2\pi} } \int e^{ik[x+x^\prime(t)]} e^{ik^2 \Delta(t)}
674: \tilde{f}(k,0)
675: \end{equation}
676: with
677: \begin{equation}
678: x^\prime(t)= \int_0^t v_g(\tau) d\tau
679: \end{equation}
680: and
681: \begin{equation}
682: \Delta(t) = \int_0^t D(\tau) d\tau \text{ .}
683: \end{equation}
684: This expression describes a coherent motion of the wavepacket formed of
685: an oscillatory motion with the time-dependent group velocity
686: Eq.~(\ref{eq:vegeneral}), and a diffusive motion with a time-dependent
687: diffusion coefficient $D(t)$. For example, if one builds an initial
688: gaussian packet of width $a_0$, $f(x,0)= \exp(-x^2/a_0^2)$
689: one finds, after some straightforward calculations:
690: \begin{equation}
691: |f(x,t)|^2 =\frac{a_0}{a(t)}
692: \exp\left(-{\frac{2x(t)^2}{a(t)^2}}\right)
693: \text{ ,}
694: \end{equation}
695: where
696: \begin{equation}
697: a(t)=a_0 \left[1+16{ \Delta(t)^2 \over a_0^4}
698: \right]^{1/2} \text{ .}
699: \end{equation}
700: \vspace{0.5cm}
701: %Figure 5
702: \begin{figure}[tbp]
703: \begin{center}
704: \psfig{figure=fig5.eps,width=6cm,clip=}
705: \caption{Spatio-temporal behavior of
706: $|\Psi(x,t)|^2$ obtained numerically by integration
707: of the Schr\"{o}dinger equation (gray level convention: the maximum values
708: of $|\Psi(x,t)|^2$ are depicted in black), from $t=0$ to
709: $t=4\pi/\delta$ (that is, 2 periods of the beat frequency
710: $\delta=\omega-\omega_B=0.02$). Other parameters are the same as in
711: Fig.~\protect \ref{fig:ResDynQuadrature}. }
712: \label{fig:EvolutionGeneral}
713: \end{center}
714: \end{figure}
715: \vspace{0.5cm}
716: The physical meaning of our developments can be evidenced by
717: considering the case where $\omega$ differs from $\omega_B$ by
718: a small detuning $\delta=\omega -\omega_B$, $|\delta| \ll \omega_B$,
719: and keep only leading-order terms of order $O(\delta^{-1})$.
720: The wavepacket then undergoes a harmonic oscillation at the beat frequency
721: $\delta$ with a group
722: velocity given by Eq.~(\ref{eq:vegeneral})
723: \begin {equation}
724: v_g(t)=2\Omega_1 d\cos(k_0d + \delta t)
725: \end{equation}
726: corresponding to a periodic mean position displacement
727: \begin{equation}
728: \langle x(t)\rangle=x(0)-\frac{2\Omega_1 d}{\delta}
729: [\sin(k_0d + \delta t)-\sin(k_0d)]
730: \end{equation}
731: The width of the wavepacket oscillates in a breathing mode which
732: is governed by:
733: \begin{equation}
734: \Delta(t) = \frac{\Omega_1 d^2}{\delta} [\cos(k_0d +\delta
735: t)-\cos(k_0d)] \text{ .}
736: \end{equation}
737: The results of Sec. \ref{sec:TimeDepWSS} for a resonant excitation
738: are naturally recovered in the limit $\delta\rightarrow 0$.
739: \vspace{0.5cm}
740: %Figure 6
741: \begin{figure}[tbp]
742: \begin{center}
743: \psfig{figure=fig6.eps,width=8cm,clip=}
744: \caption{Mean values (a) $\langle x(t) \rangle$ and (b)
745: $\sqrt{\langle x^2(t) \rangle - \langle x(t) \rangle^2}$ obtained
746: from the wavepacket dynamics depicted in Fig.~\protect
747: \ref{fig:EvolutionGeneral}.}
748: \label{fig:MeanValues}
749: \end{center}
750: \end{figure}
751: \vspace{0.5cm}
752: We have performed the integration of the Schr\"{o}dinger equation
753: for a detuning $\delta=0.02$. Fig.~\ref{fig:EvolutionGeneral} shows
754: the spatio-temporal dynamics
755: of the wavepacket and clearly evidences the periodic oscillations
756: and the breathing at frequency $\delta$ predicted above. Figure
757: \ref{fig:MeanValues} shows
758: the mean position and width of the wavepacket
759: as a function of time.
760: The comparison with the theory is very good. For instance, the
761: amplitude of $\langle x(t)\rangle$ is found numerically as 1.55
762: compared to the theoretical value 1.73.
763:
764: \section{conclusion}
765: We have studied, in a fully-analytical way, the dynamics of a wavepacket in a
766: static and time-modulated tilted potential in the framework of the
767: Wannier-Stark states. This
768: basis is well suited to the description of
769: the state of a cold atom (as produced by a Sisyphus-boosted MOT).
770: Moreover, it provides a
771: simple description the atomic dynamics, which is proved to be very rich:
772: A variety of coherent motions are obtained depending
773: on the preparation of the initial wavepacket and its site-to-site
774: quantum coherence. We can note that the description
775: introduced here is, in its principle, independent of the details of
776: the lattice, provided it presents localized states in the lattice sites.
777: It is therefore generalizable to other kinds of lattices.
778:
779: Let us finally, mention that the present work distinguishes from
780: the more usual ``solid-state''
781: approach which is based on Bloch functions.
782: We postpone for a forthcoming work the detailed comparison
783: between the Wannier-Stark and Bloch approaches.
784:
785: \section{acknowledgments}
786: Laboratoire de Physique des Lasers, Atomes et Mol{\'e}cules (PhLAM)
787: is UMR 8523 du CNRS et de l'Universit{\'e} des Sciences
788: et Technologies de Lille. Centre d'Etudes et Recherches
789: Lasers et Applications (CERLA) is supported by Minist\`{e}re de la
790: Recherche, R\'{e}gion Nord-Pas de Calais and Fonds Europ\'{e}en de
791: D\'{e}veloppement Economique des R\'{e}gions (FEDER).
792:
793: \appendix
794: \section{Unitary transformation}
795: \label{app:UnitTransf}
796: Eq.~(\ref{eq:Finertial}) is obtained if we perform a unitary transformation
797: \begin{equation}
798: U(t)=e^{iX_{0}(t)P}e^{-i\beta (t)X}e^{i\gamma (t)}
799: \end{equation}
800: where we have included translation operators in space and momentum with $%
801: \beta =2m^*\dot{X}_{0}$(i.e momentum of a particle of mass $%
802: m^*$). In this framework, following \cite{ref:Ben}, we obtain (with $%
803: UXU^{+}=X+X_{0}$ and $UPU^{+}=P+\beta $)
804: \begin{eqnarray}
805: H^{\prime } = UHU^{+}+i\left( d_{t}U\right) U^{+} \nonumber \\
806: = \frac{(P+\beta )^{2}}{2m^*}+V_{0}\cos (2\pi X)
807: +F(X+X_{0}(t))- \dot{X}_{0}(t)P+ \nonumber \\
808: \dot{\beta }(t)\left( X+X_{0}\right) -
809: \dot{\gamma } \nonumber
810: \end{eqnarray}
811: with $\dot{\gamma}=FX_{0}+m^* \stackrel{..}{X}_0 X_0+(m^*/2)
812: \dot{X}_{0}^{2}:$
813: \begin{equation}
814: H^{\prime }=\frac{P^{2}}{2m^*}+V_{0}\cos (2\pi X)+
815: (F+m^*\stackrel{..}{X_{0}}(t))X
816: \end{equation}
817: Therefore, in the frame of the periodic potential, the Hamiltonian contains
818: an inertial force proportional to $\stackrel{..}{X_{0}}(t).$
819:
820: \begin{references}
821:
822: \bibitem{ref:Bloch} F. Bloch, Z. Phys {\bf 52}, 555 (1928).
823:
824: \bibitem{ref:Zener} C. Zener, Proc. R. Soc. London A {\bf 145}, 523
825: (1934).
826:
827: \bibitem{ref:LesHouches} See for example {\it Fundamental Systems in
828: Quantum Optics}, {\'E}cole d'{\'e}t{\'e} des Houches, Session LIII, 1990,
829: edited by J. Dalibard, J. M. Raimond, and J. Zinn-Justin (North-Holland,
830: Amsterdam, 1992).
831:
832: \bibitem{ref:BlochOsc} M. Ben Dahan, E. Peik, J. Reichel, Y. Castin, and
833: C. Salomon, Phys. Rev. Lett. {\bf 76}, 4508 (1996).
834:
835: \bibitem{ref:BlochOscBEC} O. Morsch, J. H. M{\"u}ller, M. Cristiani, D.
836: Ciampini, and E. Arimondo, Phys. Rev. Lett. {\bf 87}, 140402 (2001).
837:
838: \bibitem{ref:WSLadders} S. R. Wilkinson, C. F. Bharucha, K. W. Madison, Q.
839: Niu, and M. G. Raizen, Phys. Rev. Lett. {\bf 76}, 4512 (1996).
840:
841: \bibitem{ref:CollTunnel} B. P. Anderson and M. Kasevich, Science {\bf \ 282}%
842: , 1686 (1998).
843:
844: \bibitem{ref:QKRRaizen} F. L. Moore, J. C. Robinson, C. F. Bharucha, P. E.
845: Williams, and M. G. Raizen, Phys. Rev. Lett. {\bf 73}, 2974 (1994); B. G.
846: Klappauf, W. H. Oskay, D. A. Steck, and M. G. Raizen, Phys. Rev. Lett. {\bf %
847: 81}, 1203 (1998).
848:
849: \bibitem{ref:QKRChristiensen} H. Amman, R. Gray, I. Shvarchuck, and N.
850: Christiensen, Phys. Rev. Lett. {\bf 80}, 4111 (1998).
851:
852: \bibitem{ref:QKRLille} J. Ringot, P. Szriftgiser, J. C. Garreau, and D.
853: Delande, Phys. Rev. Lett. {\bf 85}, 2741 (2000).
854:
855: \bibitem{ref:QKROxford} M. B. D'Arcy, R. M. Godum, M. K. Oberthaler, M. K.
856: Cassettari, G. S. Summy, Phys. Rev. Lett. {\bf 87}, 074102 (2001).
857:
858: \bibitem{ref:Wannier} G. Wannier, Rev. Mod. Phys. {\bf 62}, 645 (1962).
859:
860: \bibitem{ref:Nenciu} For a complete review of the subject, see G. Nenciu,
861: Rev. Mod. Phys. {\bf 63}, 91 (1991).
862:
863: \bibitem{ref:WSLaddersSL} J. Bleuse, G.Bastard and P. Voisin, Phys. Rev.
864: Lett {\bf 60}, 220 (1988).
865:
866: \bibitem{ref:Fukuyama} H. Fukuyama, R. A. Bari, and H. C. Fogdeby,
867: Phys. Rev. B {\bf 8}, 5579 (1973).
868:
869: \bibitem{note:Identities}
870: By virtue of the translational
871: properties of the WSS, Eqs.~\protect(\ref{eq:WSSTranslation}) and
872: (\protect\ref{eq:EnergyTranslation}), $X_{n,n+p}$ ($p \neq 0$)
873: does not depend on $n$:
874: $X_{n,n+p} = \int \varphi_n^*(x) x \varphi_{n+p}(x) dx =
875: \int \varphi_0^*(x-nd) x \varphi_{p}(x-nd) dx =
876: \int \varphi_0^*(x^\prime) (x^\prime+nd) \varphi_{p}(x^\prime)
877: dx^\prime = X_{0,p}$, where we used the orthogonality of the WSS.
878: Note however that $X_{n,n}=X_{0,0} + nd$ does depend on $n$.
879:
880: \bibitem{note:amplitude} While the nearest neighbors interaction is
881: proportional to $X_{01}=0.13$ the next to neighbor interaction is
882: $X_{02}=7.8$ $10^{-3}$ i.e the amplitude of the oscillation at
883: $2\omega_B$ is roughly 20 times smaller than that at $\omega_B$ (the
884: numerical values were obtained for $F=0.5$ and $V_0=2.5$).
885:
886: \bibitem{ref:Mermin} N. W. Ashcroft N.W and N. D. Mermin,
887: {\it Solid State Physics}, Holt Rinehart and Winston, 1976.
888:
889: \bibitem{ref:Holthaus} M. Holthaus, J. Opt. B: Quantum Semiclass.
890: Opt. {\bf 2}, 589 (2000).
891:
892: \bibitem{ref:Ben} M. Ben Dahan, {\it Th{\`e}se de doctorat} (unpublished)
893: Paris, 1997.
894:
895: \end{references}
896:
897: \end{document}
898: