1: \documentstyle[pra,aps,psfig]{revtex}
2: %\documentstyle[preprint,aps,psfig]{revtex}
3: %%%%%%%%%%%%%%%%%%
4: \newcommand{\beq}{\begin{equation}}
5: \newcommand{\eeq}{\end{equation}}
6: \newcommand{\beqa}{\begin{eqnarray}}
7: \newcommand{\eeqa}{\end{eqnarray}}
8: \newcommand{\ba}{\begin{array}}
9: \newcommand{\ea}{\end{array}}
10:
11: \begin{document}
12:
13: \draft
14:
15: %%%%%%%%%%%%%%%%%%
16: \twocolumn[\hsize\textwidth\columnwidth\hsize
17: \csname@twocolumnfalse\endcsname
18: %%%%%%%%%%%%%%%%%%
19:
20: \widetext
21:
22: \title{Dynamics of a BEC bright soliton in an expulsive potential}
23: \author{Luca Salasnich}
24: \address{$^{1}$Istituto Nazionale per la Fisica della Materia,
25: Unit\`a di Milano, \\
26: Dipartimento di Fisica, Universit\`a di Milano, \\
27: Via Celoria 16, 20133 Milano, Italy}
28:
29: \maketitle
30:
31: \begin{abstract}
32: We theoretically investigate the dynamics of a
33: matter-wave soliton created in a harmonic potential, which is
34: attractive in the transverse direction but expulsive in the
35: longitudinal direction. This Bose-Einstein-condensate (BEC)
36: bright soliton made of $^7$Li atoms has been observed
37: in a recent experiment (Science {\bf 296}, 1290 (2002)).
38: We show that the non-polynomial Schr\"odinger equation, an
39: effective one-dimensional equation we derived from
40: the three-dimensional Gross-Pitaevskii equation, is able
41: to reproduce the main experimental features of this
42: BEC soliton in an expulsive potential.
43: \end{abstract}
44:
45: \pacs{PACS Numbers: 03.75.Fi, 05.45.Yv}
46:
47: %%%%%%%%%%%%%%%%%%
48: ]
49: %%%%%%%%%%%%%%%%%%
50:
51: \narrowtext
52:
53: \newpage
54:
55: \section{Introduction}
56:
57: Bright and dark solitons are localized waves that travel over
58: large distances without spreading. Bright solitons are local maxima
59: in the field while dark solitons are local minima.
60: Solitons are ubiquitous: they appear in systems as diverse as
61: oceans, shallow waters in narrow channels,
62: electric circuits and optical fibers \cite{1}.
63: Recently solitons have been experimentally produced
64: in Bose-Einstein condensates (BECs) of alkali-metal atoms
65: \cite{2,3,4,5}.
66: Dark solitons have been
67: obtained with repulsive $^{87}$Rb atoms \cite{2}, while bright solitons
68: have been observed with attractive $^7$Li atoms \cite{3,4}
69: and also in a optical lattice with $^{87}$Rb
70: atoms (gap bright solitons) \cite{5}.
71: \par
72: BEC solitons have been theoretically investigated in various papers but
73: only in few of them there is a detailed comparison
74: with experiments \cite{6,7,8}. For instance,
75: by using the Gross-Pitaevskii equation \cite{9}, in Ref. \cite {8}
76: we have successfully simulated the formation and dynamics of the train
77: of bright solitons observed in the experiment of
78: the Rice University \cite{3}.
79: In this paper we consider instead the experiment of the Ecole Normale
80: Sup\'erieure \cite{4}, where Khaykovich {\it et al.} have produced
81: and detected a single BEC bright soliton in a expulsive
82: potential. The stable configurations
83: of this BEC bright soliton have been analyzed by Carr
84: and Castin \cite{7} by means of an analytical variational method.
85: Here we complete their theoretical investigation by
86: studying numerically the full time-evolution of the bright soliton
87: by using the non-polynomial Schr\"odinger equation (NPSE) \cite{10},
88: an effective one-dimensional (1D) equation we have recently derived
89: from the 3D Gross-Pitaevskii equation \cite{9}.
90: We find that NPSE reproduces remarkably well the experimental data,
91: in particular the center of mass dynamics of the BEC cloud
92: and also the time dependence of its longitudinal width.
93:
94: \section{Expulsive potential and NPSE}
95:
96: In the experiment of Khaykovich {\it et al.} \cite{4} the Bose-Condensed
97: $^7$Li atoms are confined in a trap that can be modelled
98: by a harmonic potential
99: \beq
100: U({\bf r}) = {1\over 2}m \left[
101: \omega_{\bot}^2 (x^2+y^2) + \omega_z^2 z^2 \right] \; ,
102: \eeq
103: where $m$ is the atomic mass,
104: $\omega_{\bot} = 2\pi \times 710$ Hz is the transverse
105: frequency and $\omega_z = 2\pi i \times 78$ Hz is the {\it imaginary}
106: longitudinal frequency. This imaginary frequency is due to an
107: offset magnetic field that produces the slightly expulsive
108: harmonic potential $-(m/2)|\omega_z|^2 z^2$
109: in the longitudinal direction.
110: The main effect of this expulsive term is that the center of mass
111: of the BEC accelerates along the longitudinal direction \cite{4}.
112: \par
113: The dynamics of a BEC at zero temperature can be described by the
114: time-dependent 3D Gross-Pitaevskii equation (3D GPE), given by
115: \beq
116: \left[ i\hbar {\partial \over \partial t} +{\hbar^2\over 2 m}
117: \nabla^2 - U - g N |\psi|^2 \right] \psi = 0 \; ,
118: \eeq
119: where $\psi({\bf r},t)$ is the macroscopic wave function
120: (order parameter) of the Bose-Einstein condensate \cite{9},
121: $g=4\pi\hbar^2 a_s/m$ is the interatomic strength, $a_s$ is the s-wave
122: scattering length, and $N$ is the number of condensed atoms.
123: \par
124: In the case of strong cylindric radial confinement,
125: i.e. when the BEC travsverse energy $E_{\bot}$
126: of the BEC is equal to the transverse
127: harmonic energy $\hbar \omega_{\bot}$, the 3D GPE reduces to a 1D GPE.
128: This result can be easily obtained with a variational approach
129: by using the following trial wave function $\psi(x,y,z,t) =
130: f(z,t) \exp{[-(x^2+y^2)/(2\eta^2)]}/\sqrt{\pi \eta^2}$,
131: where $f(z,t)$ is the longitudinal wave function normalized to one and
132: $\eta$ is the transverse width. Actually one finds
133: $$
134: \left[ i\hbar {\partial \over \partial t} + {\hbar^2\over 2m}
135: {\partial^2\over \partial z^2} - {\hbar \omega_{\bot}\over 2}
136: \left( {a_{\bot}^2\over \eta^2 } + {\eta^2 \over a_{\bot}^2}
137: \right) - {m\over 2}{\omega_z^2} z^2
138: \right.
139: $$
140: \beq
141: \left.
142: - {gN \over (2 \pi) \eta^2} |f|^2
143: \right] f = 0 \; ,
144: \eeq
145: where $E_{\bot} = (\hbar \omega_{\bot}/2)
146: (a_{\bot}^2/\eta^2 + \eta^2/a_{\bot}^2)$
147: is the transverse energy
148: of the Bose condensate. In the 1D GPE one has $\eta = a_{\bot}$
149: for which $E_{\bot}=\hbar\omega_{\bot}$. Instead,
150: choosing a space-time dependent $\eta$, one finds \cite{10} that
151: $\eta$ satisfies the equation
152: \beq
153: \eta(z,t) = a_{\bot} \;
154: \large( 1+2a_sN|f(z,t)|^2 \large)^{1/4} \; .
155: \eeq
156: By inserting this formula in the previous differential
157: equation one gets a nonpolynomial Schr\"odinger equation (NPSE)
158: \cite{10}, that reduces to the 1D GPE in the weakly-interacting
159: limit $a_s N|f|^2 <<1$.
160: The NPSE has been found to be very accurate in the descripition
161: of BECs under transverse harmonic confinement
162: and a generic longitudinal external potential. The NPSE is accurate
163: with both repulsive and attractive scattering length \cite{8,10,11}.
164: In our last paper \cite{12} we have extended the NPSE to include also
165: beyond mean-field effects, like the formation of a gas of impenetrable
166: bosons in the Tonks-Girardeau regime.
167:
168: \section{BEC bright solitons}
169:
170: For simplicity we consider first the case with $\omega_z=0$
171: in Eq. (3). In this case, it is well known that the 1D GPE
172: (namely the Eq. (3) with $\eta = a_{\bot}$)
173: with negative scattering length
174: ($a_s<0$) admits bright soliton solutions \cite{1}, that set up when
175: the negative inter-atomic energy of the BEC compensates the positive
176: kinetic energy such that the BEC is self-trapped. Scaling $z$ in units
177: of $a_{\bot}$ and $t$ in units of $\omega_{\bot}^{-1}$,
178: with the position \beq f(z,t)=\Phi(z-vt) e^{iv(z-vt)}
179: e^{i(v^2/2 - \mu)t} \; ,
180: \eeq
181: from 1D GPE one finds the text-book bright soliton
182: \beq
183: \Phi(z-vt)=\sqrt{\gamma\over 2} \; sech\left[{\gamma}(z-vt) \right] \; ,
184: \eeq
185: where $\gamma=|a_s|N/a_{\bot}$ and the chemical potential $\mu$
186: is given by $\mu= 1 - \gamma^2/2$, while the velocity $v$ of the bright
187: soliton remains arbitrary. Given a value of the scattering length $a_s$,
188: the number $N$ of condensed atoms fixes the chemical potential
189: $\mu$ of the bright soliton.
190: \par
191: The solitary wave solution of Eq. (6),
192: that we call {\it 1D bright soliton}, exists for any positive
193: value of $\gamma$. On the other hand, it has been theoretically
194: predicted \cite{10,13} and experimentally shown \cite{3,4} that above a
195: critical interaction strength the BEC bright soliton, that we call
196: {\it 3D bright soliton}, will collapse.
197: To study the properties of a 3D bright soliton it is not necessary to solve
198: the full 3D GPE; in fact, the 3D bright solitons can be very accurately
199: described by NPSE \cite{8,10}. In particular, from the NPSE
200: (namely the Eq. (3) with $\eta$ given by Eq. (4)) one finds the
201: 3D bright soliton written in implicit form
202: $$
203: z - v t = {1\over \sqrt{2}} {1\over \sqrt{1-\mu}} \;
204: arcth\left[ \sqrt{ \sqrt{1-2\gamma\Phi^2}-\mu \over 1-\mu }
205: \right]
206: $$
207: \beq-{1\over \sqrt{2}} {1\over \sqrt{1+\mu}} \;
208: arctg\left[ \sqrt{ \sqrt{1-2\gamma\Phi^2}-\mu \over 1+\mu } \right] \; .
209: \eeq
210: It is important to observe that this equation is well defined only
211: for $\gamma \Phi^2 <1/2$; at $\Phi^2 = 1/(2\gamma)$ the transverse
212: width $\eta$ becomes zero. Moreover, by imposing the normalization
213: condition, one finds \beq (1-\mu)^{3/2} - {3\over2} (1-\mu)^{1/2}
214: + {3\over 2 \sqrt{2}} \gamma = 0 \; ,
215: \eeq ù
216: which admits solutions only for $0 < \gamma < 2/3$.
217: \par
218: There are thus two kind of collapse: the {\it transverse collapse},
219: when the condensate probability density $\rho=|\Phi|^2$ exceeds
220: $\rho_c = 1/(2\gamma)$, and the {\it 3D collapse}, when $\gamma$
221: reaches the value $2/3$. Note that for a single 3D bright soliton
222: at $\gamma_c=2/3$ the transverse size of the condensate soliton is
223: not yet zero, in fact $\gamma_c \rho <1/2$. Nevertheless, if we consider
224: two colliding 3D bright solitons with $\gamma<2/3$, at the impact the
225: total density of the cloud can become equal to $\rho_c$ and the system
226: will collapse.
227:
228: \begin{figure}
229: \centerline{\psfig{file=ens-f1.eps,height=2.5in}}
230: \caption{Maximum probability density $\rho_M$ of bright solitons as
231: a function of the nonlinear strength $\gamma =|a_s|N/a_{\bot}$.
232: Length $z$ in units $a_{z}=(\hbar /m\omega_z)^{1/2}$, density
233: $\rho$ in units $1/a_{\bot}$, where $a_{\bot}=(\hbar /m\omega_{\bot})^{1/2}$
234: is the harmonic length of transverse confinement.}
235: \end{figure}
236:
237: In Figure 1 we show the maximum probability density $\rho_M$ of bright
238: solitons as a function of the strength $\gamma$. The solid line
239: is the density of a single 3D bright soliton that ends at $\gamma =2/3$.
240: The dot-dashed line is the probability density of the 1D bright soliton,
241: that coincides with the solid line for small values of $\gamma$
242: (in fact, for $\gamma\ll 1$ Eq. (7) reduces to Eq. (6)).
243: The dashed line is the maximum probability density of two colliding
244: equal 3D bright solitons,
245: that ends when the curve meets the dotted line, which is the density
246: $\rho_c=1/(2\gamma)$ of the transverse collapse. Figure 1 shows that
247: two colliding and equal 3D bright solitons produce a transverse collapse
248: of the condensate for $\gamma =0.472$, which corresponds to a single
249: soliton probability density equal to $0.265$. Only for lower values
250: of $\gamma$ the collapse is avoided during collision and interference.
251:
252: \section{Bright soliton in the expulsive potential}
253:
254: We now consider the case $\omega_z\neq 0$ in Eq. (3) and
255: try to simulate the experiment of the Ecole Normale
256: Sup\'erieure \cite{4} with $^7$Li alkali-metal vapors.
257: In that experiment a BEC with positive
258: scattering length ($a_s>0$) has been condensed into a cigar-shaped
259: harmonic trap. The scattering length has been then tuned to a small
260: and negative value ($a_s<0$) via a Feshbach resonance.
261: The resulting condensate, projected onto the expulsive
262: harmonic potential of Eq. (1), has been observed to propagate
263: over a distance of $1$ mm.
264:
265: \begin{figure}
266: \centerline{\psfig{file=ens-f2.eps,height=4.in}}
267: \caption{Density of the $^7$Li BEC in the expulsive
268: potential obtained by solving the NPSE. The BEC cloud
269: propagates over $1$ mm.
270: Case with $a_s=0$ (ideal gas). There are
271: $N=4\times 10^3$ atoms. Six frames from bottom to top:
272: $t=2$ ms, $t=3$ ms, $t=4$ ms, $t=5$ ms, $t=6$ ms,
273: $t=7$ ms. External harmonic potential given by
274: Eq. (1). Red color corresponds to highest density.}
275: \end{figure}
276:
277: \par
278: Following the paper of Khaykovich {\it et al.} \cite{4},
279: we choose as initial condition for a fixed $a_s$ the ground-state
280: of the stationary Eq. (3) with a fully confining
281: harmonic potential: $\omega_{\bot} = 2\pi \times 710$ Hz and
282: $\omega_z = 2\pi \times 50$ Hz. Then the time-dependent wave
283: function is obtained by the numerical integration of the Eq. (3)
284: with the expulsive potential: $\omega_{\bot} = 2\pi \times 710$ Hz but
285: $\omega_z = 2\pi i \times 78$ Hz.
286: Note that, as in the experiment, the initial position of
287: the BEC cloud is shifted of $50$ $\mu$m on the left with
288: respect to the maximum of the expulsive potential.
289: A critical point is the choice of the number $N$ of atoms.
290: To avoid the collapse
291: we choose $N=4\times 10^3$, a value compatible
292: with the experimental data and suggested by the
293: variational theory \cite{13}. The NPSE is solved by using
294: the finite-difference numerical algorithm described
295: in \cite{14}.
296:
297: \begin{figure}
298: \centerline{\psfig{file=ens-f3.eps,height=4.in}}
299: \caption{Density of the $^7$Li BEC in the expulsive
300: potential obtained by solving the NPSE.
301: The BEC cloud propagates over $1$ mm.
302: Case with $a_s=-0.21$ nm (``bright soliton'').
303: There are $N=4\times 10^3$ atoms. Six frames from bottom to top:
304: $t=2$ ms, $t=3$ ms, $t=4$ ms, $t=5$ ms, $t=6$ ms,
305: $t=7$ ms. External harmonic potential given by
306: Eq. (1). Red color corresponds to highest density.}
307: \end{figure}
308:
309: In Fig. 2 we plot six frames of the density of the BEC
310: with $a_s=0$. This is the case of an ideal gas and the figure
311: shows that the center of mass $z_{cm}$ of the bosonic cloud follows
312: the law $z_{cm}(t)=z_0 \exp{(|\omega_z|t)}$. In addition,
313: the longitudinal width of the cloud grows due to the
314: dispersive kinetic term. In Fig. 3 we plot instead
315: the time-dependent density of the BEC with $a_s=-0.21$ nm.
316: Also here the center of mass $z_{cm}$ follows the exponential
317: grow in time but the longitudinal width does not show
318: an appreciable enlargement: the matter wave travels
319: without spreading. It is important to stress that
320: Fig. 2 and Fig. 3 are in remarkable good agreement with the
321: absorption images shown in Fig. 3 of the experimental
322: paper \cite{4} with the same values of the parameters.
323: \par
324: In their experiment, Khaykovich {\it et al.} \cite{4} have
325: measured the root mean square size $\sigma$ of the longitudinal
326: width versus the propagation time for three values of $a_s$:
327: $a_s=0$, $a_s=-0.11$ nm, and $a_s=-0.21$ nm.
328: Figure 4 shows their experimental data and our numerical results
329: obtained with the NPSE.
330: For $a_s=0$ the attractive interaction between atoms is zero
331: and the expansion of the cloud is governed by the kinetic
332: energy and the negative curvature of the longitudinal axial
333: potential. For $a_s=-0.11$ nm the size $\sigma$
334: of the axial width is consistently below that of a non-interacting gas
335: but the attractive interaction is not strong enough to
336: avoid an appreciable enlargement of the cloud.
337:
338: \begin{figure}
339: \centerline{\psfig{file=ens-f4.eps,height=3.4in}}
340: \caption{Root mean square size $\sigma$
341: of the longitudinal width of the BEC as a function
342: of the propagation time $t$. The filled circles are
343: the experimental data taken from Ref. [4]. The dashed
344: line is the ideal gas ($a_s=0$) curve.
345: The solid line is obtained from the numerical solution
346: of the NPSE. }
347: \end{figure}
348:
349: For $a_s=-0.21$ nm the enlargement is further reduced but our
350: numerical results suggest that $\sigma$ is not truly constant.
351: This is not surprising because a truly shape-invariant solitary
352: wave is expected only in the absence of a longitudinal potential
353: and with an appropriate initial condition.
354: Nevertheless Fig. 4 shows that our results are fully compatible
355: with the experimental data, which have the resolution limit
356: $\Delta \sigma = 9$ $\mu$m of the imaging system.
357: \par
358: As previously discussed, in the calculations we have choosen
359: $N=4\times 10^3$ to avoid the collapse of the BEC with
360: $a_s =-0.21$ nm. In the experiments the collapse actually implies
361: that a fraction of atoms is expelled from the Bose condensate via
362: three-body recombination. This effect can be phenomenologically
363: modelled by including a dissipative term in the GPE or in the NPSE,
364: but a more rigorous treatment requires the solution of coupled
365: time-dependent equations for the Bose condensate and
366: the thermal cloud.
367: \par
368: Finally, we note that a sudden shift of the inter-atomic
369: strength $|a_s|N/a_{\bot}$ to a very large value (for instance
370: via a Feshbach resonance) produces a set of bright solitons
371: (soliton train), which can travel in the expulsive potential.
372: A detailed analysis of the formation of these multi-soliton
373: configurations can be found in our recent papers \cite{8,15}.
374:
375: \section{Conclusions}
376:
377: We have shown that the 1D non-polynomial Schr\"odinger
378: equation (NPSE) is able to accurately describe the
379: time evolution of an attractive Bose-condensed cloud of
380: $^7$Li atoms in an expulsive harmonic potential.
381: Contrary to the 1D Gross-Pitaevskii equations,
382: the NPSE takes also into account the space-time variations of the
383: transverse width of the Bose-Einstein condensate and its numerical
384: integration is much faster than the numerical solution of
385: the 3D Gross-Pitaevskii equation.
386: The theoretical results of the NPSE have been compared with
387: the experimental data of a recent experiment performed at the
388: Ecole Normale Sup\'erieure of Paris. Both experiment and theory
389: show that, when the inter-atomic strength is sufficiently
390: strong, a localized matter wave is obtained, which travels
391: for a large distance without an appreciable spreading.
392: Our numerical calculations suggest that this solitary wave
393: is not fully shape-invariant but the experimental resolution
394: does not allow one to clarify this point.
395: It is important to stress that the limitations in
396: the current resolution of the experimental imaging system
397: can be overcome by creating a bright soliton
398: with a larger longitudinal width. This can be achieved by using
399: a smaller attractive inter-atomic strength in addition
400: with a weaker expulsive potential.
401: \par
402: In conclusion, we observe that to make clear
403: the particle-like nature of a soliton
404: it is essential to investigate its interaction with another
405: soliton. The interference of BEC bright solitons is an
406: important subject which has to be analyzed in detail,
407: having also implications for atom interferometry \cite{16}.
408:
409: \section*{Acknowledgments}
410:
411: We thank Alberto Parola, Davide Pini and Luciano Reatto for many
412: useful discussions.
413:
414: \begin{thebibliography}{99}
415:
416: \bibitem{1} P.G. Drazin and R.S. Johnson, {\it Solitons:
417: An Introduction} (Cambridge University Press, Cambridge, 1988).
418:
419: \bibitem{2} S. Burger {et al.}, Phys. Rev. Lett. {\bf 83}, 5198 (1999);
420: J. Denshlag {\it et al.}, Science {\bf 287}, 97 (2000).
421:
422: \bibitem{3} K.E. Strecker {\it et al.}, Nature {\bf 417}, 150 (2002).
423:
424: \bibitem{4} L. Khaykovich {\it et al.}, Science {\bf 296}, 1290 (2002).
425:
426: \bibitem{5} B. Eiermann {et al.}, Phys. Rev. Lett.
427: {\bf 92}, 230401 (2004).
428:
429: \bibitem{6} U. Al Khawaja, H.T.C. Stoof, R.G. Hulet, K.E. Strecker,
430: and G.B. Partridge, Phys. Rev. Lett. {\bf 89}, 200404 (2002).
431:
432: \bibitem{7} L.D. Carr and Y. Castin,
433: Phys. Rev. A {\bf 66}, 063602 (2002).
434:
435: \bibitem{8} L. Salasnich, A. Parola, and L. Reatto,
436: Phys. Rev. A {\bf 66}, 043603 (2002);
437: L. Salasnich, A. Parola, and L. Reatto,
438: Phys. Rev. Lett. {\bf 91}, 080405 (2003);
439: L. Salasnich, Laser Phys. {\bf 14}, 291 (2004).
440:
441:
442: \bibitem{9} E.P. Gross, Nuovo Cimento {\bf 20}, 454 (1961);
443: L.P. Pitaevskii, Zh. Eksp. Teor. Fiz. {\bf 40}, 646 (1961)
444: [English Transl. Sov. Phys. JETP {\bf 13}, 451 (1961)].
445:
446: \bibitem{10} L. Salasnich, Laser Phys. {\bf 12}, 198 (2002);
447: L. Salasnich, A. Parola, and L. Reatto,
448: Phys. Rev. A {\bf 65}, 043614 (2002).
449:
450: \bibitem{11} L. Salasnich, A. Parola, and L. Reatto,
451: J. Phys. B {\bf 35}, 3205 (2002); L. Salasnich,
452: Laser Phys. {\bf 13}, 543 (2003); L. Salasnich, A. Parola,
453: and L. Reatto, Phys. Rev. A {\bf 69}, 045601 (2004).
454:
455: \bibitem{12} L. Salasnich, A. Parola, and L. Reatto,
456: Phys. Rev. A {\bf 70}, 013606 (2004).
457:
458: \bibitem{13} V.M. Perez-Garcia, H. Michinel, and H. Herrero,
459: Phys. Rev. A {\bf 57}, 3837 (1998).
460:
461: \bibitem{14} L. Salasnich, A. Parola, and L. Reatto,
462: Phys. Rev. A {\bf 64}, 023601 (2001).
463:
464: \bibitem{15} L. Salasnich, cond-mat/0407653.
465:
466: \bibitem{16} Y. Torii {\it et al.}, Phys. Rev. A {\bf 61},
467: 041602 (2000); Y.-Ju Wang {\it et al.}, cond-mat/0407689.
468:
469: \end{thebibliography}
470:
471: \end{document}
472:
473:
474:
475:
476: