1:
2: \documentclass[onecolumn,showpacs,pre]{revtex4}
3: \usepackage{graphicx}
4: \usepackage{subfigure}
5:
6: \begin{document}
7:
8: \title{A model of a dual-core matter-wave soliton laser }
9: \author{Peter Y. P. Chen}
10: \affiliation{School of Mechanical and Manufacturing Engineering, University of New South
11: Wales, Sydney 2052, Australia}
12: \author{Boris A. Malomed}
13: \affiliation{Department of Interdisciplinary Studies, School of Electrical Engineering,
14: Faculty of Engineering, Tel Aviv University, Tel Aviv 69978, Israel}
15:
16: \begin{abstract}
17: We propose a system which can generate a periodic array of
18: solitary-wave pulses from a finite reservoir of coherent
19: Bose-Einstein condensate (BEC). The system is built as a set of
20: two parallel quasi-one-dimensional traps (the reservoir proper and
21: a pulse-generating cavity), which are linearly coupled by the
22: tunneling of atoms. The scattering length is tuned to be negative
23: and small in the absolute value in the cavity, and still smaller
24: but positive in the reservoir. Additionally, a parabolic potential
25: profile is created around the center of the cavity. Both edges of
26: the reservoir and one edge of the cavity are impenetrable.
27: Solitons are released through the other cavity's edge, which is
28: semi-transparent. Two different regimes of the intrinsic operation
29: of the laser are identified: circulations of a narrow
30: wave-function pulse in the cavity, and oscillations of a broad
31: standing pulse. The latter regime is stable, readily providing for
32: the generation of an array containing up to $10,000$
33: permanent-shape pulses. The circulation regime provides for no
34: more than $40$ cycles, and then it transforms into the oscillation
35: mode. The dependence of the dynamical regime on parameters of the
36: system is investigated in detail.
37: \end{abstract}
38:
39: \pacs{03.75.-b; 03.75.Lm; 05.45.Yv}
40: \maketitle
41:
42: \section{Introduction}
43:
44: The concept of matter-wave lasers, which are intended to generate
45: strong coherent beams of atoms, and many applications that such
46: beams can find, are well known. It is commonly assumed that a
47: Bose-Einstein condensate (BEC) may serve as a source of the
48: matter-wave laser\ beam. Atomic lasing was considered in detail
49: theoretically (see Refs. \cite{laser} and references therein) and
50: demonstrated in various experimental settings \cite{experiment};
51: the topic was recently reviewed in Ref. \cite{review}. Of special
52: interest is a possibility to construct a laser operating in a
53: pulsed (quasi-soliton) regime, i.e., periodically releasing narrow
54: localized pulses of coherent atom waves, as proposed in Refs.
55: \cite{soliton-laser} and \cite{Carr}. In that connection, it is
56: relevant to mention that (effectively) one-dimensional solitons in
57: BEC with very weak attraction between atoms (in $^{7}$Li) were
58: created in well-known experiments \cite{soliton}.
59:
60: The first models of the pulsed matter-wave lasers assumed release
61: of quasi-solitons from an elongated (``cigar-shaped") trap filled
62: by a self-attractive BEC. This scheme, while probably realizable
63: in the experiment, does not offer sufficiently effective means to
64: control the operation regime, and does not secure generation of a
65: very large number of pulses in a nearly periodic fashion. A more
66: sophisticated scheme was very recently proposed in Ref.
67: \cite{Spain}, which assumes zero nonlinearity coefficient (i.e.,
68: zero scattering length of inter-atomic collisions) in a larger
69: part of the elongated trap, and attraction between atoms (negative
70: scattering length)\ in its smaller part, where the pulses are to
71: be formed and released. The sign of the scattering length may be
72: controlled and altered along the length of the trap by means of
73: the Feshbach resonance\ \cite{Feshbach} (this mechanism was used
74: for the experimental creation of the BEC solitons \cite{soliton}).
75:
76: In this work, we aim to propose a different model of a pulsed matter-wave
77: laser, in which the BEC reservoir and the pulse-generating cavity are
78: separated. As shown in Fig. \ref{setup} below, they are elongated parallel
79: traps with linear coupling between them, due to tunneling of atoms. Using
80: the aforementioned possibility to control the scattering length by means of
81: the Feshbach resonance, we assume that the interaction between atoms is
82: (very weakly) repulsive in the reservoir, while in the cavity it is
83: attractive (and weak too, see below). Additionally, the negative scattering
84: length is assumed to be altered along the cavity's axis, which can be
85: achieved using an appropriate configuration of the magnetic field
86: responsible for the Feshbach resonance (we do not assume any time dependence
87: of the scattering length or other parameters).
88:
89: Besides using the magnetic field,\ it was predicted theoretically
90: \cite{optFRtheory} and demonstrated experimentally
91: \cite{optFRexperiment} that\ the Feshbach resonance can be
92: provided too by an adequately tuned optical field. Accordingly, a
93: spatially modulated stationary distribution of the light intensity
94: may also be employed to provide for necessary spatially
95: inhomogeneous Feshbach-resonance configurations in the reservoir
96: and cavity.
97:
98: It is relevant to mention that various soliton solutions in dual-core traps,
99: with the linear coupling between the cores and \emph{opposite} signs of the
100: scattering length in them, were recently studied in detail \cite{Valery},
101: following an earlier work which was dealing with a model of dual-core
102: nonlinear optical fibers with opposite signs of the group-velocity
103: dispersion (rather than nonlinearity) in the two cores \cite{Dave}.
104:
105: In the model introduced below, both edges of the reservoir, and
106: the left edge of the cavity are impenetrable to atoms, while the
107: right edge of the cavity (a ``valve")\ allows the release of
108: pulses into an outcoupling atomic waveguide, see Fig. \ref{setup}.
109: The model was originally designed with an intention to provide for
110: formation of a narrow quasi-soliton pulse in the cavity, that
111: would periodically circulate in it, bouncing from the edges. Each
112: time that the soliton hits the right edge (valve), a pulse is
113: released into the outcoupling waveguide. After being slashed this
114: way, the intrinsic solitary pulse is supposed to replenish itself
115: in the course of the subsequent cycle of the circulation by
116: collecting atoms tunneling from the reservoir. Accordingly, the
117: initial condition is taken with a large number of atoms in the
118: reservoir, and a small number in the cavity.
119:
120: As described in detail below, numerical simulations demonstrate
121: that the circulation regime outlined above may be observed, but it
122: is not a really stable one: the intrinsic pulse would perform no
123: more than $40$ circulations, quickly coming to a halt and getting
124: broad. Nevertheless, a very robust regime of the periodic release
125: of pulses can be found in a large parameter region. In this
126: regime, a broad ``lump" stays immobile in the cavity, performing
127: periodic cycles of stretching and compression. When its right wing
128: reaches the valve, in the course of each cycle of the vibrations,
129: a new outcoupling pulse is released. The respective loss in the
130: number of atoms in the cavity is compensated by the influx of
131: atoms tunneling from the reservoir. The pulse-release cycles
132: repeat very many times, till the reservoir gets essentially
133: depleted. We stress that the model is a \emph{passive }one, i.e.,
134: it supports the stable pulse generation regime without any
135: externally applied active control, such as periodically opening
136: and shutting the valve by some clock signal.
137: \begin{figure}[tbp]
138: \includegraphics[width=4.00in]{fig0.eps}
139: \caption{A scheme of the proposed matter-wave laser composed of two nearly
140: one-dimensional traps coupled by tunneling of atoms, with negative and
141: positive scattering lengths in the cavity and reservoir, respectively. The
142: opening at the right edge of the cavity ($x=1$) depicts the valve releasing
143: pulses into an outcoupling waveguide. Shading symbolically shows an
144: instantaneous density distribution, $\left\vert \protect\psi \right\vert
145: ^{2} $ and $\left\vert \protect\phi \right\vert ^{2}$, in both traps (larger
146: density corresponds to more intense shading). The upper part shows the
147: potential profile in the cavity, with the drop $\Delta U$ at $x=1$
148: corresponding to the valve.}
149: \label{setup}
150: \end{figure}
151:
152: The paper is organized as follows. In Section 2, the model is formulated.
153: Basic results, which demonstrate both unstable circulation-based and stable
154: vibration-based regimes of the periodic generation of solitary pulses, are
155: presented in Section 3. In Section 4, we analyze the dependence of the
156: operation of the matter-wave laser on its parameters, and give an estimate
157: for the actual number of atoms in outcoupled pulses generated by the laser.
158: The paper is concluded by Section 4.
159:
160: \section{The model}
161:
162: According to what was said above, the model of the matter-wave laser is
163: based on a system of linearly coupled Gross-Pitaevskii equations (GPEs) for
164: the wave functions $\psi (x,t)$ and $\phi (x,t)$ of atoms in the
165: (effectively) one-dimensional cavity and reservoir, with opposite signs in
166: front of the nonlinear terms. We scale $\hbar $, the atomic mass $m$, and
167: the nonlinearity coefficient at $x=0$ in the cavity (in physical units, the
168: latter is $4\pi \hbar ^{2}a/m$, where $a$ is the $s$-wave scattering length
169: of atomic collisions around $x=0$) to be $1$. Then, the coupled equations
170: are cast in a normalized form,
171: \begin{eqnarray}
172: i\psi _{t} &=&-\frac{1}{2}\psi _{xx}-\left( 1+gx\right) |\psi |^{2}\psi
173: \nonumber \\
174: &&+f(x-L/2)^{2}\psi -\kappa \phi , \label{psi} \\
175: i\phi _{t} &=&-\frac{1}{2}\phi _{xx}+\epsilon |\phi |^{2}\phi -\kappa \psi ,
176: \label{phi}
177: \end{eqnarray}where the coordinate $x$ takes values $0<x<L$ (i.e., $L$ is the length of
178: both the cavity and reservoir). In accordance with what was said above, we
179: assume a small positive scattering length (weak repulsion between atoms) in
180: the reservoir, $\epsilon >0$, while in the cavity the interaction is
181: attractive, with the strength linearly increasing with $x$ ($g>0$ is
182: assumed), following a corresponding profile of the static magnetic or
183: optical field that controls the scattering length in the cavity via the
184: Feshbach resonance.
185:
186: The coefficient $g$ will be small (see below), but it is needed to
187: create an effective potential slope in the cavity: as the
188: intra-cavity pulse gets ``heavier" due to the agglomeration of
189: atoms tunneling from the reservoir, the term proportional to $g$
190: creates a gradually increasing effective potential which pushes
191: atoms to the right (since $g$ is positive). In the aforementioned
192: circulation regime, this effective (nonlinear) potential pushes
193: the intra-cavity localized pulse to the right. After hitting the
194: right edge (at $x=L$), where the ``valve" is placed (see details
195: below), the pulse loses a part of its mass, generating the
196: outcoupled soliton. Therefore, the force induced by the nonlinear
197: potential drops, allowing the intra-cavity pulse to move in the
198: opposite direction after the bounce from the right edge, until it
199: will bounce (without loss) from the left edge, $x=0$. This way,
200: the nonlinear term proportional to $g$ may support periodic
201: circulations of the internal pulse (although, as said above, in
202: the domain of the model's parameter space explored in this work
203: the circulations are not really stable, eventually switching into
204: the regime of vibrations; the latter regime is possible with $g=0$
205: as well).
206:
207: Further, we assume that an external parabolic potential acts in the cavity,
208: \begin{equation}
209: U_{\mathrm{pot}}(x)=f(x-L/2)^{2}, \label{Upot}
210: \end{equation}with the center set at the middle of the cavity. It is well known that such
211: a potential corresponds to a magnetic or optical trap applied to the
212: condensate \cite{optical}. It will be shown below that the results are not
213: sensitive to the exact shape of the intra-cavity potential; in fact,
214: essentially the same stable vibration regime can be obtained in the model
215: with no potential. \ Note that Eq. (\ref{phi}) for the reservoir does not
216: include any potential, as it is not necessary to generate the dynamical
217: regime sought for. We also explored a model including a potential in the
218: reservoir, but it did not affect the operation regime in any remarkable
219: manner.
220:
221: The linear coupling between the cavity and reservoir is accounted
222: for by the positive coefficient $\kappa $, which determines the
223: tunneling time $1/\kappa $ for atoms. Keeping $\epsilon ,g,f$ and
224: $\kappa $ as free parameters of the model, the remaining scaling
225: invariance of Eqs. (\ref{psi}) and (\ref{phi}) makes it possible
226: to set $L\equiv 1$, which is fixed below.
227:
228: As said above, the model assumes that the reservoir at its both edges, and
229: the cavity at its left edge are closed by impenetrable lids, which gives
230: rise to the respective boundary conditions (b.c.),
231: \begin{equation}
232: \phi (x=0)=\phi (x=1)=\psi (x=0)=0. \label{0}
233: \end{equation}The pulses (solitons) are to be released into an external waveguide from the
234: cavity at its right edge, $x=1$ (recall we have set $L\equiv 1$).
235: For this purpose, a valve is set at $x=1$, assuming that the
236: corresponding b.c. for the $\psi $ field is linear. Actually, the
237: single possible form of such a linear b.c. is\begin{equation} \psi
238: _{x}(x=1)=iq\psi (x=1), \label{tau}
239: \end{equation}with a positive constant $q$ (its physical meaning is discussed below).
240: Indeed, using the continuity equation for the total density, $\rho
241: \equiv |\psi |^{2}+|\phi |^{2}$, and current, $j=(i/2)\left( \psi
242: _{x}^{\ast }\psi -\psi _{x}\psi ^{\ast }\right) +(i/2)\left( \phi
243: _{x}^{\ast }\phi -\phi _{x}\phi ^{\ast }\right) \equiv
244: j_{1}+j_{2}$, in the coupled GPEs (\ref{psi}) and (\ref{phi}),
245: \begin{equation}
246: \frac{\partial \rho }{\partial t}=\frac{i}{2}\left( \psi _{xx}\psi ^{\ast
247: }-\psi _{xx}^{\ast }\psi \right) \equiv -\frac{\partial j}{\partial x},
248: \label{rho}
249: \end{equation}integrating Eq. (\ref{rho}) over the length of the system, $0<x<1$, and
250: taking into regard that $j_{1}(x=0)=j_{2}(x=0)=j_{2}(x=1)=0$
251: pursuant to b.c. (\ref{0}), one can derive a \textit{depletion
252: equation} for the solution's norm $N$, which is proportional to
253: the total number of atoms in the system:\begin{eqnarray} N
254: &=&\int_{0}^{1}\left( |\psi (x)|^{2}+|\phi (x)|^{2}\right)
255: dx\equiv
256: N_{1}+N_{2}, \label{N} \\
257: -\frac{dN}{dt} &=&j_{1}(x=1,t)=q\left\vert \psi (x=1,t)\right\vert
258: ^{2}\equiv E(t), \label{E}
259: \end{eqnarray}where b.c. (\ref{tau}) was substituted for $\psi _{x}$ in the expression for
260: $j_{1}(x=1)$. The latter result is what might be expected: the
261: instantaneous rate at which the density is released through the
262: valve is proportional to the local density $\left\vert \psi
263: (x=1)\right\vert ^{2}$, which justifies the adoption of b.c.
264: (\ref{tau}). Initial values of the norms $N_{1}^{(0)}$ and
265: $N_{2}^{(0)}$ of the $\psi $ and $\phi $ fields, and the b.c.
266: constant $q $ are supplementary control parameters of the model,
267: in addition to the above-mentioned set of $\left( \epsilon
268: ,g,f,\kappa \right) $. Below, it will be demonstrated that really
269: important parameters are $\epsilon $, $\kappa $, and $q$.
270:
271: The physical meaning of b.c. (\ref{tau}) can be readily
272: understood. Indeed, it implies that the local gradient of the
273: phase of the wave function $\psi $ at the point $x=1$, which is
274: proportional to the atom's momentum, is fixed to be $q$. This can
275: be realized by assuming that, to the right of the point $x=0$ (in
276: the outcoupling waveguide), the cavity potential drops by $\Delta
277: U=q^{2}/2$ (recall we have set $m=\hbar =1$). The drop $\Delta U$
278: must be essentially larger than the potential and kinetic energy
279: of atoms to the left of the point $x=1$ (in the cavity). The
280: latter condition, which amounts to\begin{equation} U(x=1)=f/4\ll
281: q^{2}/2 \label{condition}
282: \end{equation}[see Eq. (\ref{Upot})], provides for the possibility to neglect small
283: deviations in the kinetic energy of the released atoms from $q^{2}/2$. We
284: notice that a soliton-releasing valve in the matter-wave laser model
285: recently proposed in Ref. \cite{Spain} includes a similar element (an
286: effective potential step).
287:
288: The form of the function $E(t)$, as defined in Eq. (\ref{E}), determines the
289: shape of the density pattern (actually, an array of pulses) released into
290: the outcoupling waveguide, the same way as the temporal shape of the input
291: signal at the point $z=0$ determines the shape of a temporal soliton or
292: solitonic array propagating along the coordinate $z$ in the optical fiber
293: \cite{Agrawal}. Note that, as follows from Eq. (\ref{tau}), the limit cases
294: of $q=0$ and $q=\infty $ correspond, respectively, to a reflecting mirror
295: and impenetrable lid, i.e., $\psi _{x}(x=1)=0$ and $\psi (x=1)=0$. In either
296: case, the system does not release anything. In addition to yielding the
297: final result in the form of $E(t)$, Eq. (\ref{E}), with $dN/dt$ computed
298: directly from numerical data, provides for a means to monitor the accuracy
299: of simulations.
300:
301: Finally, we note that the model does not include an explicit form of the
302: outcoupling waveguide (the extension of the cavity beyond $x=1$), which is
303: responsible for the eventual shaping the released pulses into solitons.
304: Actually, the soliton-formation problem in a uniform waveguide has been
305: studied in detail before (see, e.g., Ref. \cite{Carr} and references
306: therein).
307:
308: \section{Operation regimes of the pulsed matter-wave laser}
309:
310: In the numerical analysis of the model, Eqs. (\ref{psi}) and
311: (\ref{phi}) were solved by means of a finite-element
312: pseudospectral method, which implies division of the integration
313: domain into several finite elements. Solution in each element is
314: approximated by a polynomial, and a pseudo-spectral method, based
315: on Chebyshev collocation points \cite{Peter}, is used in each
316: element. We have checked that a numerical algorithm based on
317: domain stretching, rather than domain division into finite
318: elements \cite{Chebyshev}, produces the same results.
319:
320: In cases when the stationary version of the GPEs, i.e., a system of two
321: linearly coupled ordinary differential equations, had to be solved (see
322: below), the finite-element pseudospectral method reduces them to a set of
323: algebraic equations, which were treated by means of the Newton's method. In
324: the general case, when Eqs. (\ref{psi}) and (\ref{phi}) are partial
325: differential equations, the time derivatives in the equations were
326: approximated by the finite difference as per the Crank-Nicholson scheme.
327: This, together with the finite-element pseudospectral method, yields a set
328: of nonlinear algebraic equations, that were also solved by the Newton's
329: algorithm.
330:
331: In all the cases considered, the integration domain $0\leq x\leq
332: 1$ was divided into four equal elements, with a fifth-order
333: polynomial used in each one. The procedure leads to a set of $48$
334: algebraic equations, and it was checked that increase of the
335: polynomial's order did not change the results in any tangible
336: aspect. We used the time step of $\Delta t=0.001$, checking that
337: smaller time steps gave virtually the same results (while larger
338: $\Delta t$ would generally be insufficient to produce accurate
339: results).
340:
341: Simulations started with initial conditions which, by themselves,
342: were stationary solutions to Eqs. (\ref{psi}) and (\ref{phi}), in
343: the form of $\left\{ \psi (x,t),\phi (x,t)\right\} =\exp \left(
344: -i\mu t\right) \left\{ \Psi (x),\Phi (x)\right\} $ with real $\Psi
345: (x)$ and $\Phi (x)$. These stationary solutions were found, in a
346: numerical form, for given values of $g,\epsilon $ and $\kappa $.
347: While doing so, b.c. (\ref{tau}) at $x=1$ was temporarily replaced
348: by $\psi =0$ [as said above, this formally corresponds to
349: $q=\infty $ in Eq. (\ref{tau})], since a stationary solution is
350: not possible otherwise, due to the nonzero current $j_{1}$ across
351: the point $x=1$, see Eq. (\ref{rho}). Further, both equations
352: (\ref{psi}) and (\ref{phi}) used to generate the stationary
353: solutions were modified by including \textit{ad hoc} potentials
354: linear in $x$, as the so generated initial configurations were
355: found to quickly produce stable operation regimes, even though the
356: actual potentials in Eqs. (\ref{psi}) and (\ref{phi}) are
357: different [recall that Eq. (\ref{phi}) has no potential at all]. A
358: set of typical initial configurations obtained this way is
359: displayed in Fig. \ref{fig1}.
360: \begin{figure}[tbp]
361: $\begin{array}{cc}
362: \includegraphics[width=3.7in]{fig1a.eps} &
363: \includegraphics[width=3.7in]{fig1b.eps}\end{array}$
364: \caption{Typical examples of the initial real profiles of $\Psi (x)$ (a) and
365: $\Phi (x)$ (b) for $g_{1}=0.02$, $\protect\epsilon =0.1$, and
366: $\protect\kappa =0.5$, used in simulations of the coupled
367: equations (\protect\ref{psi}) and (\protect\ref{phi}). In all the
368: cases, the norm of the $\protect\phi $-component is fixed (in the
369: normalized units) to be $N_{2}=363.4$, while the norm $N_{1}$ of
370: the $\protect\psi $-component varies as shown in the figure.}
371: \label{fig1}
372: \end{figure}
373:
374: Starting from such initial configurations, simulations produced two types of
375: dynamical regimes. The first one features a relatively narrow pulse which
376: performs shuttle motion, circulating in the cavity. Each collision of the
377: pulse with the valve at the right edge of the cavity gives rise to a
378: quasi-soliton released by the matter-wave laser. However, as mentioned
379: above, the circulation regime was never found to be stable (although we
380: cannot claim that the investigation of the system's parameter space was
381: exhaustive, as it is difficult to perform complete exploration of the
382: seven-paramater space). The pulse would perform no more than $40$
383: circulations (typically, fewer), and this regime would transform itself into
384: a stable vibration regime.
385:
386: An example of the transition from circulations to vibrations is displayed in
387: Fig. \ref{fig2}. Actually, the transition may be much shorter at other
388: values of the parameters. In most cases, as shown below, the vibration
389: regime sets in directly from the initial configuration, without going
390: through the transient stage of circulations.
391: \begin{figure}[tbp]
392: $\includegraphics[width=4.00in]{fig2a.eps}$ \newline
393: $\begin{array}{cc}
394: \includegraphics[width=3.7in]{fig2b.eps} &
395: \includegraphics[width=3.7in]{fig2c.eps}\end{array}$
396: \caption{Transition of the circulation regime into a vibration one, shown in
397: terms of a three-dimensional plot for the density $|\protect\psi
398: (x,t)|^{2}$ in the cavity (a), contour plots of the density (b),
399: and the outcoupling rate $E(t)$ (c). The parameters are $g=0.02$,
400: $\protect\epsilon =0.1$, $f=3$, $\protect\kappa =0.5$, and $q=50$.
401: The initial (normalized) numbers of atoms in the reservoir and
402: cavity are $N_{2}^{(0)}=360$ and $N_{1}^{(0)}=12$. } \label{fig2}
403: \end{figure}
404:
405: The instability of the circulation regime can be understood. Indeed, one may
406: consider a small fluctuation which makes the norm of the pulse released into
407: the outcoupling waveguide slightly smaller than in the established regime,
408: hence the norm remaining in the cavity soliton becomes slightly larger.
409: Consequently, the strength of the effective nonlinear potential, that tries
410: to push the soliton to the right, drops (as a results of the bounce from the
411: valve) by an amount which is a bit smaller than in the unperturbed state,
412: and the soliton will therefore slide to the left slower than in the absence
413: of the perturbation. As a result, spending more time in the cavity, the
414: soliton will absorb more atoms in the course of the next cycle of the
415: circulations, and will return to the right edge with a still larger norm.
416: The mechanism by which a fluctuation slightly increasing the soliton's norm
417: leads to its still larger increase (or, conversely, a fluctuation decreasing
418: the norm causes its further decrease) implies the instability. It may happen
419: that the circulation regime can be stabilized in a model including an active
420: element (i.e., \textit{forced circulations} might be stable), but in this
421: work we focus on the more fundamental passive model.
422:
423: In the regime of vibrations, a broad pulse stays at the center of
424: the cavity and performs a very large number of internal
425: oscillations. At a stage of the vibration cycle when the pulse
426: spreads out, its right wing hits the valve and generates an
427: outcoupling localized matter-wave packet. A typical example of
428: such a stable regime is displayed in Figs. \ref{fig3} and
429: \ref{fig4}. The second panel in Fig. \ref{fig3} shows (for a long
430: interval of time) the temporal dependence of the outcoupling rate
431: $E(t)$, as defined by Eq. (\ref{E}). As said above, this
432: dependence actually determines the shape of the pulse array
433: released by the matter-wave laser.
434: \begin{figure}[tbp]
435: $\begin{array}{cc}
436: \includegraphics[width=3.7in]{fig3a.eps} & \includegraphics[width=3.7in]{fig3b.eps} \\
437: \includegraphics[width=3.7in]{fig3c.eps} & \includegraphics[width=3.7in]{fig3d.eps}\end{array}$
438: \caption{A typical example of the establishment of a stable vibration regime
439: in the cavity, shown in terms of the outcoupling rate $E(t)$,
440: which determines the shape of the pulse array generated by the
441: matter-wave laser [$E$ is defined as per Eq. (\protect\ref{E})].
442: (a) The initial stage; (b) beginning of the established regime;
443: (c) a fragment from panel (b), showing 15 cycles in detail; (d) a
444: single typical cycle (shown in terms of obviously rescaled
445: variables $t^{\ast }$ and $E^{\ast }$), together with a simplest
446: analytical fit to it. In panels (a) and (b), contour plots in the
447: lower subpanels (ii) additionally illustrate the spatiotemporal
448: evolution of the atom density, $\left\vert \protect\psi
449: (x,t)\right\vert ^{2}$. Parameters are $g=0.02$, $\protect\epsilon
450: =0.10$, $f=3$, $\protect\kappa =0.5$, and $q=200$. The initial
451: normalized amounts of matter in the reservoir and cavity are
452: $N_{2}=363.4$ and $N_{1}=6$. In the established regime, the period
453: is $T=0.135$, with the duty cycle of $50\%$.} \label{fig3}
454: \end{figure}
455: \begin{figure}[tbp]
456: \includegraphics[width=4.00in]{fig4.eps}
457: \caption{The evolution of the density distribution in the cavity,
458: $|\protect\psi (x,t)|^{2}$, for the same case as shown in Fig.
459: \protect\ref{fig3}.} \label{fig4}
460: \end{figure}
461:
462: Figure \ref{fig4} depicts vibrations of the density distribution
463: in the cavity, for the same case as displayed in Fig. \ref{fig3}.
464: As seen from these figures, after a transient period, a very
465: robust regime (with some residual long-period modulations) sets
466: in. With the maximum value of the release rate $E$ in this regime
467: $E_{\max }\simeq 0.4$, the oscillation period $T=0.135$ and the
468: \textit{duty cycle} (the share of the period within which the
469: value of the density at the right edge of the cavity, $|\psi
470: (x=1,t)|^{2}$, exceeds half of its maximum value) being $0.5$
471: imply that the number of cycles which are expected before
472: depletion of the reservoir will become appreciable (say, $N$ will
473: drop to two thirds of $N_{2}^{(0)}=363.4$) can be easily
474: estimated:\begin{equation} \frac{(2/3)N_{2}^{(0)}}{E_{\max }\cdot
475: (T/2)}\simeq 10,000\text{ \textrm{cycles}.} \label{cycles}
476: \end{equation}Direct verification of this prediction requires too long simulations
477: (up to $t\simeq 1500$). An estimate for the actual number of atoms in the generated
478: pulses (which may range between $10$ and $1000$) is given in the next
479: section.
480:
481: It is relevant to stress that, in the cases presented in Figs.
482: \ref{fig2} and \ref{fig3}, \ref{fig4}, with $f=3$ and $q=200$, the
483: condition (\ref{condition}), which substantiates the use of b.c.
484: (\ref{tau}), is satisfied with a huge margin. In fact, this
485: condition holds equally well in all the cases where stable
486: operation of the matter-wave laser model was observed. It is also
487: relevant to mention that, in physical units, the value of $q\sim
488: 100$ corresponds, for $^{7}$Li, to the velocity $\sim 1$ cm/s of
489: the released atoms, the respective kinetic energy being $\sim 0.1$
490: nK, on the temperature scale.
491:
492: It is relevant to compare the depletion rate in the stable operation mode
493: [as given by Eq. (\ref{E})] with the rate at which the cavity and reservoir
494: exchange the matter. A generic example of this comparison is displayed in
495: Fig. \ref{fig5}. As is seen, the exchange rate is much higher than the speed
496: of depletion. It is interesting to note that the long-period beatings in the
497: dependence of $N_{1}(t)$, which are observed in \ref{fig5}, practically do
498: not manifest themselves in the oscillations of the outcoupling rate $E(t)$
499: [nor in the evolution of $N_{1}+N_{2}$, as the beatings in $N_{1}(t)$ are
500: almost exactly compensated by anti-phase beatings in $N_{2}(t)$]. In other
501: words, the beatings do not affect the quality of the pulse array generated
502: by the matter-wave laser.
503:
504: The line showing $N_{1}+N_{2}$ vs. $t$ in Fig. \ref{fig5} seems to
505: have finite thickness due to fast oscillations of $E(t)$, such as
506: those in Fig. \ref{fig3}. Its comparison with the respective
507: dependence predicted by averaging of the depletion equation
508: (\ref{E}) (a bold dashed line) is also shown in Fig. \ref{fig5}.
509: An error due to the averaging may account for a small discrepancy
510: with the actual evolution of $N_{1}+N_{2}$.
511: \begin{figure}[tbp]
512: \includegraphics[width=4.00in]{fig5.eps}
513: \caption{The evolution of the norm (amount of matter) in the
514: reservoir, $N_{1}$ (the rapidly oscillating dashed curve), and the
515: total norm, $N_{1}+N_{2}$ (the continuous finite-thickness line),
516: in a stable operation regime. The bold dashed curve displays the
517: prediction following from the depletion equation
518: (\protect\ref{E}), averaged over rapid oscillations of $E(t)$.
519: Parameters are the same as in Fig. \protect\ref{fig3}.}
520: \label{fig5}
521: \end{figure}
522:
523: \section{Dependence of the operation regime on parameters}
524:
525: The matter-wave laser model proposed above depends on several
526: parameters. It is necessary to identify those which are really
527: important to the stability of the operation mode, and which are
528: not. First of all, simulations demonstrate that there is virtually
529: no dependence on the potential's strength $f$ in Eq. (\ref{psi}):
530: at least within the interval of $1<f<3$, all dynamical
531: characteristics of the matter-wave laser\ model display very
532: little variation, if other parameters are fixed (taking typical
533: values of physical parameters given below, we conclude that, in
534: this parameter region, the range of $1<f<3$ actually corresponds
535: to the range of the longitudinal trapping frequencies $0.1-1$ Hz).
536: The dependence on the initial amount of matter in the cavity,
537: $N_{1}^{(0)}$, is very weak too, at least in the interval of
538: $5<N_{1}^{(0)}<20$. Also negligible is sensitivity of the
539: established regime to the parameter $g$ in Eq. (\ref{psi}). In
540: fact, additional simulations have demonstrated that virtually the
541: same stable oscillation regime sets in if we take Eqs. (\ref{psi})
542: and (\ref{phi}) with $f=g=0$.
543:
544: A more important role is played by the constant $q$ in Eq.
545: (\ref{tau}) (recall it determines the momentum of atoms in the
546: outcoupling pulse), and the linear-tunneling constant $\kappa $
547: that determines the strength of the linear coupling between the
548: reservoir and cavity in Eqs. (\ref{psi}) and (\ref{phi}). Figures
549: \ref{fig6} and \ref{fig7} illustrate effects of the variation of
550: these parameters on the operation regime. This is shown via the
551: change of the time dependences of the outcoupling rate $E(t)$ and
552: total amount of matter remaining in the system, $N_{1}+N_{2}$.
553: \begin{figure}[tbp]
554: $\begin{array}{cc}
555: \includegraphics[width=3.7in]{fig6a.eps} & \includegraphics[width=3.7in]{fig6b.eps}
556: \end{array}$\caption{The influence of the variation of the boundary-condition parameter
557: $q$ (the momentum of atoms in pulses released by the matter-wave laser), see
558: Eq. (\protect\ref{tau}), on the time dependence of the outcoupling rate (a),
559: and on the depletion dynamics (b). Except for $q$, other parameters are the
560: same as in Fig. \protect\ref{fig3}.}
561: \label{fig6}
562: \end{figure}
563: \begin{figure}[tbp]
564: $\begin{array}{cc}
565: \includegraphics[width=3.7in]{fig7a.eps} & \includegraphics[width=3.7in]{fig7b.eps}
566: \end{array}$\caption{The same as in Fig. \protect\ref{fig6}, but showing the effect of
567: the variation of the strength of the linear coupling between the reservoir
568: and cavity. }
569: \label{fig7}
570: \end{figure}
571:
572: The essential dependence of the dynamical regimes on the
573: parameters $q$ and $\kappa $ is quite natural. In addition to
574: this, and somewhat unexpectedly, the system demonstrates strong
575: sensitivity to the value of the nonlinearity coefficient $\epsilon
576: $ in Eq. (\ref{phi}), which is proportional to the positive atomic
577: collision length in the reservoir. In fact, stable operation
578: regimes, providing for the generation of very long periodic arrays
579: of the outcoupling pulses, are possible in a relatively narrow
580: interval of values of $\epsilon $, if the other parameters are
581: fixed. The dependence of the dynamical states on $\epsilon $ is
582: illustrated, by displaying segments of the established dependence
583: $E(t)$, in Fig. \ref{fig8}. Panel (a) in this figure shows that,
584: at $\epsilon =0.08$, the generated array of pulses has a high
585: amplitude, but with shallow dips between pulses, so that they are
586: not well separated. The same panel demonstrates that the optimum
587: shape of the array of well-separated pulses is achieved between
588: $\epsilon =0.09$ and $0.12 $ (which implies the positive
589: scattering length in the reservoir being $\simeq 10$ times smaller
590: than the absolute value of the negative scattering length in the
591: cavity). The situation in the latter interval is additionally
592: illustrated by a set of curves in panel (b) of Fig. \ref{fig8}.
593: Physically, the necessary fine-tuning of $\epsilon $ in the
594: reservoir may be achieved by means of the Feshbach resonance.
595:
596: \begin{figure}[tbp]
597: $\begin{array}{cc}
598: \includegraphics[width=3.7in]{fig8a.eps} & \includegraphics[width=3.7in]{fig8b.eps}
599: \end{array}$\caption{The effect of variation of the nonlinearity coefficient
600: $\protect\epsilon $ in the reservoir on the operation regime of the matter-wave laser
601: model. Other parameters are the same as in Fig. 3.}
602: \label{fig8}
603: \end{figure}
604:
605: Undoing normalizations that cast the coupled GPEs in the rescaled form of
606: Eqs. (\ref{psi}) and (\ref{phi}), it is easy to derive a relation between
607: the actual number of atoms, $N_{\mathrm{phys}}$, and the norm $N_{1}$ of the
608: $\psi $ function in the normalized units:
609: \begin{equation}
610: N_{\mathrm{phys}}=\frac{r^{2}}{8|a|L}N_{1}, \label{Nphys}
611: \end{equation}where $r$ is the transverse radius of the cigar-shaped trap, which plays the
612: role of the cavity, $a$ and $L$ being, respectively (as defined above), the
613: Feshbach resonance-controlled atomic scattering length in the cavity and the
614: cavity's length, both taken in physical units. For the cases displayed in
615: Fig. \ref{fig8}, a characteristic value of the norm of an individual
616: released pulse, in the normalized units, is $N_{1}\sim 0.1$. Assuming the
617: aspect ratio of the cylindrical trap to be $L/(2r)\simeq 10$, and the actual
618: length $L\simeq 100$ $\mu $m, Eq. (\ref{Nphys}) yields the number of atoms
619: per outcoupling pulse between $10$ and $1000$, if the scattering length is
620: kept at a low level, between $a\simeq -0.1$ nm and $a\simeq -0.01$ nm (a
621: very small value of $|a|$ is necessary to prevent collapse in the cavity
622: trap). These values may be sufficient for physical applications. In
623: particular, the threshold number of $^{7}$Li atoms, necessary for the
624: creation of a stable three-dimensional solitons in an optical lattice, was
625: recently shown to be $60$ \cite{3D}. The number of atoms $\sim 100$ may also
626: be sufficient for direct optical detection of the matter-wave pulse. As for
627: the total number of atoms originally stockpiled in the reservoir, the above
628: estimates, including Eq. (\ref{cycles}), yield the range of $N_{0}\sim
629: 10^{5}-10^{7}$ for it.
630:
631: \section{Conclusion}
632:
633: The objective of this work was to propose a model of a matter-wave
634: laser which can provide for stable periodic generation of a
635: sequence of solitary-wave pulses from a reservoir containing a
636: large amount of matter in the form of a coherent Bose-Einstein
637: condensate. The system includes two parallel cigar-shaped traps,
638: \textit{viz}., the reservoir and the work cavity, which are
639: coupled by the tunneling of atoms between them. The scattering
640: length of atomic collisions is tuned to be positive and negative
641: in the reservoir and cavity, respectively. Both ends of the
642: reservoir, and the left edge of the cavity are closed by lids.
643: Solitary pulses are released through a ``valve" at the right edge
644: of the cavity, which is described by the linear boundary condition
645: (\ref{tau}), and may be implemented as a potential step separating
646: the cavity and the outcoupling waveguide. Two different regimes of
647: the operation of the matter-wave laser were identified:
648: circulations of a narrow solitary pulse in the cavity, and
649: vibrations of a broad standing lump. Only the latter mode is
650: stable, while the circulation regime spontaneously rearranges
651: itself into the vibration mode. Dependence of the stability and
652: characteristics of the pulse-generating regime on parameters of
653: the model was explored. The regime is sensitive to the
654: boundary-condition parameter $q$ and strength $\kappa $ of the
655: linear coupling between the core and reservoir, and especially
656: sensitive to the value of the nonlinearity coefficient (i.e.,
657: scattering length) in the reservoir.
658:
659: The bottom line is that the model can guarantee stable generation of a
660: pattern consisting of up to $10^{4}$ permanent-shape pulses, each containing
661: between $10$ and $1000$ atoms, under typical experimental conditions. It is
662: relevant to stress that the parameter space of the model, which is huge
663: (seven-dimensional, as the full set of the parameters includes $\epsilon
664: ,g,f,\kappa ,q$, and the initial norms $N_{1,2}^{(0)}$), is far from being
665: fully explored, so still more promising regimes may be hidden in the model.
666:
667: \section*{Acknowledgements}
668:
669: We appreciate valuable discussions with J. Brand, L. Carr, and V.
670: M. P\'{e}rez-Garc\'{\i}a. The work of B.A.M. was partially
671: supported by the Israel Science Foundation through the
672: Excellence-Center grant No. 8006/03.
673:
674: \bigskip \newpage
675:
676: \begin{thebibliography}{99}
677: \bibitem{laser} Holland M, Burnett K, Gardiner C, Cirac J I, and Zoller P
678: 1996 \textit{Phys. Rev}. \textit{A} \textbf{54} R1757; Moy G M and
679: Savage C M 1997 \textit{ibid. }\textbf{56} R1087; Kneer B, Wong T,
680: Vogel K, Schleich W P, and Walls D F 1998 \textit{ibid
681: }\textbf{58} 4841; Band Y B, Julienne P S and Trippenbach M 1999
682: \textit{ibid. }\textbf{59} 3823; Martin J L, McKenzie C R, Thomas
683: N R, Sharpe J C, Warrington D M, Manson P J, Sandle W J and Wilson
684: A C 1999 \textit{J. Phys. B -- At. Mol. Opt. Phys}. \textbf{32}
685: 3065; Trippenbach M, Band Y B, Edwards M, Doery M, Julienne P S,
686: Hagley E W, Deng L, Kozuma M, Helmerson K, Rolston S L and
687: Phillips W D 2000 \textit{ibid. }\textbf{33} 47;~Robins N, Savage
688: C and Ostrovskaya E A 2001 \textit{Phys. Rev. A} \textbf{64}
689: 043605; Gerbier F, Bouyer P, and Aspect A 2001 \textit{Phys. Rev.
690: Lett.} \textbf{86} 4729; Floegel F, Santos L and Lewenstein M 2003
691: \textit{Europhys. Lett}. \textbf{63} 812; Ashkin A 2004 Proc. Nat.
692: Acad. Sci. USA \textbf{101} 12108
693:
694: \bibitem{experiment} Mewes M O, Andrews M R, Kurn D M, Durfee D S, Townsend
695: C G, and Ketterle W 1997 \textit{Phys. Rev. Lett.} \textbf{78} 582
696: Miesner H J, Stamper-Kurn D M, Andrews M R, Durfee D S, Inouye S
697: and Ketterle W 1998 \textit{Science} \textbf{279} 1005; Bloch I,
698: Hansch T W and Esslinger T 1999 \textit{Phys. Rev. Lett}.
699: \textbf{82} 3008; Deng L, Hagley E W, Wen J, Trippenbach M, Band
700: Y, Julienne P S, Simsarian J E, Helmerson K, Rolston S L and
701: Phillips W D 1999 \textit{Nature} \textbf{398} 218; Hagley E W,
702: Deng L, Kozuma M, Wen J, Helmerson K, Rolston S L and Phillips W D
703: 1999 \textit{Science} \textbf{283} 1706; Kohl M, Hansch T W and
704: Esslinger T, 2001 \textit{Phys. Rev. Lett}. \textbf{87} 160404; Le
705: Coq Y, Thywissen J H, Rangawala S A, Gerbier F, Richard S,
706: Delannoy G, Bouyer P and Aspect A, 2001 \textit{ibid }\textbf{87}
707: 170403; Bloch I, Kohl M, Greiner M, Hansch T W and Esslinger T
708: 2001 \textit{ibid }\textbf{87}, 030401; Chikkatur A P, Shin Y,
709: Leanhardt A E, Kielpinski D, Tsikata E, Gustavson T L, Pritchard D
710: E, and Ketterle W 2002 \textit{Science} \textbf{296} 2193
711:
712: \bibitem{review} Bongs K and Sengstock K 2004 \textit{Rep. Progr. Phys}.
713: \textbf{67} 907
714:
715: \bibitem{soliton-laser} Leboeuf P, Pavloff N, and Sinha S 2003 \textit{Phys.
716: Rev. A} \textbf{68} 063608
717:
718: \bibitem{Carr} Carr L D and Brand J 2004 \textit{Phys. Rev. A} \textbf{70}
719: 033607
720:
721: \bibitem{soliton} Strecker K E, Partridge G B, Truscott A G, and Hulet R G
722: 2002 \textit{Nature} \textbf{417} 150; Khaykovich L, Schreck F,
723: Ferrari G, Bourdel T, Cubizolles J, Carr L D, Castin Y and Salomon
724: C 2002 \textit{Science} \textbf{296} 1290
725:
726: \bibitem{Spain} Rodas-Verde M I, Michinel H and P\'{e}rez-Garc\'{\i}a V M
727: 2005 \textit{Phys. Rev. Lett.} \textbf{95} 153903
728:
729: \bibitem{Feshbach} Kagan Y , Surkov E L, and Shlyapnikov G V 1997
730: \textit{Phys. Rev. Lett.} \textbf{79} 2604; Roberts J L, Claussen N R, Burke J P,
731: Jr, Greene C H, Cornell E A, and Wieman C E 1998 \textit{ibid. }\textbf{81}
732: 5109; Inouye S, Andrews M R,~Stenger J, Miesner H-J, Stamper-Kurn D M, and
733: Ketterle W 1998 \textit{Nature} \textbf{392} 151
734:
735: \bibitem{optFRtheory} Fedichev P O, Kagan Yu, Shlyapnikov G V, and Walraven
736: J T M 1996 \textit{Phys. Rev. Lett.} \textbf{77} 2913
737:
738: \bibitem{optFRexperiment} Theis M, Thalhammer G, Winkler K, Hellwig M, Ruff
739: G, Grimm R, and Denschlag J H 2004 \textit{Phys. Rev. Lett. }\textbf{93}
740: 123001
741:
742: \bibitem{Valery} Shchesnovich V S, Malomed B A, and Kraenkel R A 2004
743: \textit{Physica D} \textbf{188} 213; Shchesnovich V S, Cavalcanti S B, and
744: Kraenkel R A 2004 \textit{Phys. Rev. A} \textbf{69} 033609; Shchesnovich V S
745: and Cavalcanti S B 2005 \textit{ibid} \textbf{71}, 023607
746:
747: \bibitem{Dave} Kaup D J and Malomed B A 1998 \textit{J. Opt. Soc. Am. B}
748: \textbf{15} 2838
749:
750: \bibitem{optical} Stamper-Kurn D M, Andrews M R , Chikkatur A P, Inouye S,
751: Miesner H J, Stenger J and Ketterle W 1998 \textit{Phys. Rev.
752: Lett. }\textbf{80} 2027
753:
754: \bibitem{Agrawal} Agrawal G P 1995 \textit{Nonlinear Fiber Optics} (Academic
755: Press: San Diego)
756:
757: \bibitem{Peter} Chen P Y P 1982 \textit{Electron. Lett. }\textbf{16} 1048;
758: \textbf{18} 736
759:
760: \bibitem{Chebyshev} Renault R and Frohlich J 1996
761: %%A pseudospectral Chebychev method for the 2D
762: %%wave equation with domain streching and absorbing boundary conditions
763: J. Comp. Phys. \textbf{124} 324 %%-336
764:
765: \bibitem{3D} Mihalache D, Mazilu D, Lederer F, Malomed B A, Crasovan L-C,
766: Kartashov Y V and Torner L 2005 \textit{Phys. Rev. A} \textbf{72} 021601(R)
767: \end{thebibliography}
768:
769: \end{document}
770: