RAPID COMMUNICATIONS

PHYSICAL REVIEW A 66, 061602 2002

Nonlinear atom optics with bright matter-wave soliton trains

V. Y. F. Leung,* A. G. Truscott, and K. G. H. Baldwin
Atom Manipulation Project, Research School of Physical Sciences and Engineering Australian National University, Canberra, ACT 0200, Australia Received 30 August 2002; published 31 December 2002 Bright-soliton train creation and evolution in a Bose-Einstein condensate is simulated numerically using the experimental conditions of K. E. Strecker et al. Nature 417, 150 2002 as a starting point. We identify the controlling factors that dictate soliton number, phase imprinting, and stability. Reliable production of multiple phase-engineered condensed-matter pulses can have practical applications to future nonlinear atom optics experiments, which we illustrate with an example of colinear scattering between repulsive solitons. DOI: 10.1103/PhysRevA.66.061602 PACS number s : 03.75.Fi, 05.45.Yv, 03.65.Ge

I. INTRODUCTION

Recent experiments 1,2 show that a Bose-Einstein condensate BEC of atoms with an attractive interaction can produce coherent matter-wave pulses that propagate without dispersion. In such pulses, or bright solitons, the kinetic energy which causes an untrapped condensate to expand is exactly counter balanced by the inward pull of the mutually attractive atoms, an interaction that under other conditions may result in condensate instability and collapse. In this paper we numerically simulate the results of one such experiment Strecker et al. 1 to i , demonstrate that a mean-eld approach is capable of describing an attractive condensate in an area of nonlinear behavior such as soliton creation, and ii , investigate the soliton interactions by extending the experimental parameters. The rst solitons in a BEC were observed as propagating dark notches in repulsive condensates 3,4 . In these experiments phase-imprinting by nonuniform illumination of the condensate engineered a phase step in the BEC. By establishing a light block in the condensate, dark-soliton trains have also been generated 5 . Of equal interest is the creation of bright solitons, which has been the subject of several theoretical studies 6 9 . Bright solitons require an attractive interaction i.e., a negative scattering length, which only allows condensation under a very restricted range of conditions. However, success in the manipulation of the condensate scattering length through Feshbach resonances 10 has since made experimental investigations feasible, and this development has generated a great deal of recent experimental and theoretical activity 1,2,11,12 . In the experiment we study, the released BEC formed a train of multiple solitons which oscillated collectively in a weak harmonic potential for many cycles, each soliton kept apart from its neighbor by the alternating phase structure of the ensemble 1 . This result suggests the possibility of designing a series of self-focussing, phase-engineered coherent matter pulses as the basis of future experiments in nonlinear atom optics. Bose-Einstein condensates can be well described by the Gross-Pitaevskii nonlinear Schrodinger equation GPE ,

which treats the condensate as a single atomic wave function subjected to the mean-eld attraction or repulsion of the atoms. Previous investigations of the GPE have predicted that stable solitons may be found experimentally when the number of atoms per soliton is small and the condensate is quasione-dimensional 1D 6,7 . In this regime the condensate (4 na) 1/2 is larger than the transverse healing length radius of the conning potential, where n is the condensate density and a is the s-wave scattering length 13 . In addition, stationary solutions of the GPE have suggested that trains of solitons may be observed in condensates which have undergone a nonadiabatic excitation 6 . One signicant contribution of the recent experimental results is their support of the assumption that the GPE is equally applicable to condensates of both positive and negative scattering length. Experimentally, as described in Ref. 1 , a quasi-1D condensate was formed in a boxlike potential with signicantly tighter connement in the transverse than the axial dimension. Manipulation of a Feshbach resonance rapidly tuned the scattering length of the atoms from positive to negative, resulting in a condensate which became attractive at, or shortly after, its release. Most of the unconned atoms were ejected in a burst, leaving behind a small proportion of remnant atoms as solitons. These were allowed to evolve as they propagated along the axial potential of a far red-detuned laser eld for a short time before being imaged. The dynamic nature of the experimental initial conditions suggests that the rst step to understand the mechanism of soliton train formation and phase imprinting is best facilitated by a time-dependent model. We present the results of a dynamic simulation of soliton train creation, based on the numerical solution of the time-dependent GPE. By adhering closely to the experimental conditions 1 we are able to reproduce the factors which determine the stability, number, and phase relationships of the solitons. An understanding of the manipulation and control of these fundamental characteristics will assist the design of future experiments. As an example, we present a numerical simulation of a colinear scattering experiment.
II. SIMULATING A QUASI-1D CONDENSATE

*Electronic address: v.leung@iap.uni-bonn.de

1050-2947/2002/66 6 /061602 4 /$20.00

We begin with the nonlinear Schrodinger equation NLSE ,

66 061602-1 2002 The American Physical Society

RAPID COMMUNICATIONS

LEUNG, TRUSCOTT, AND BALDWIN

PHYSICAL REVIEW A 66, 061602 2002

r,t t

2m

2 V r,t

NU 0

r,t

r,t . 1

This is a mean-eld description of an evolving nonlinear macroscopic quantum state that describes the behavior of N condensate atoms when the temperature is close to zero by taking U 0 4 2 a/m, where m is the atomic mass 14 . We will assume that V is a cylindrically symmetric potential, with transverse and axial dimensions given by (r,z), respectively. Since the condensates in experiments 1,2 are in the quasi-1D regime, several simplications can be made that reduce the dimensionality of Eq. 1 . As described in Ref. 7 , a condensate under much greater transverse than axial connement with trap frequencies such that r / z 1 can be approximated as a separable wave function of the form r,z,t r z,t , 2

where within the energy regime of the low-order modes of (z,t), the most probable state for the tightly trapped component (r) is the ground state. 1 As V 2 m 2 z 2 the system has natural units harmonic oscillator units h.o.u. of characteristic length and time based on the size of the noninteracting ground-state wavefunction and the period of harmonic oscillation. Replacing ( /2m ) 1/2z and t in the 1D version of (z,t) with Eq. 1 results in a dimensionless form of the NLSE 14 , i ,
2 2

FIG. 1. Evolution of a train of three repulsive solitons as they propagate in a weak harmonic potential. The plot is of ( , ) 2 the condensate density distribution, vs normalized length and time. At 0 units of h.o.u. the nonlinear constant here was ramped from C 200 to C 22 over a period of 5, and the condensate was 1.75 into a harmonic potential V/600 where V is released at the initial trapping potential. The inset shows a plot of soliton number vs release time h.o.u. , with conditions as for the main gure.

1 4

where, in order to normalize the wave function, the nonlinear coefcient C 4 Na( /2m ) 1/2. Finally, following the treatment of Saito and Ueda 15 , Eq. 3 can be extended to include atomic losses due to inelastic collisions, i ,
2 2

1 4 , 2

C
2

i L2

L3

, 6

, 4

where L 2 and L 3 are the two-body and three-body recombination loss-rate coefcients, respectively.
III. NUMERICAL RESULTS AND ANALYSIS

Reliable methods of soliton creation and phase imprinting are required if bright matter-wave soliton trains are to be a useful source of coherent wave packets. We rst present an analysis of these processes, based on the experimental work performed by Strecker et al. We then show that within the existing experimental parameters it is possible to create scattering between matter-wave solitons.

Equation 3 was numerically integrated using a fourthorder Runge-Kutta method. To obtain the numerical equivalent of the initial repulsive condensate in Ref. 1 , we started with the analytical noninteracting ground-state solution and adiabatically ramped to a large and positive C. The ramp is then rapidly reversed to C 0, a parallel to the experimental manipulation of a Feshbach resonance. Concurrently in a time interval 1 h.o.u. before the end of the reverse ramp the original harmonic trap is replaced by a much 20 h.o.u. with respect to the weaker one displaced by rst. This is equivalent to releasing the condensate from the side of a gentle potential well. Thus the initially stationary condensate will acquire some nonzero velocity v . The resulting condensate density, Fig. 1, shows a train of three wave packets which exhibit solitonlike behavior. These are found to oscillate stably for many periods with neither observable dispersion nor particle exchange. Noninteracting particles simultaneously released from different points of a harmonic oscillator will overtake each other every half period. In contrast, as in Ref. 1 , our simulated particle trajectories do not intersect, indicating that the solitons are kept apart by a repulsive force. Mutual repulsion between two solitons arises from their opposing phase 16 . In a stationary train of equal-sized repulsive pulses this leads to a series of phase steps between neighboring solitons 6 . The condensate in Ref. 1 may acquire such a phase structure if the phase gradient is determined by velocity such that d /dz m v / . Our spatial phase proles show that nite phase steps can be expected whenever the solitons are repulsive and nonoverlapping. Lack of effective phase imprinting, such as when the reverse ramp is insufciently rapid, leads to solitons with arbitrary phase relationships, as is evident from their random relative motion.

061602-2

RAPID COMMUNICATIONS

NONLINEAR ATOM OPTICS WITH BRIGHT MATTER- . . .

PHYSICAL REVIEW A 66, 061602 2002

In agreement with Ref. 1 , we have found that N s in, enabling us to vary the number of creases linearly with solitons created Fig. 1, inset . One possible explanation is N s is required to prothat a total phase difference of duce N s solitons of alternating phase, and therefore N s v . This was observed experimentally, where v was controlled between the through manipulation of the time interval release of the condensate and the end of the decreasing ramp, 1 . In the simulation we start from N s 1 such that v 0. However, in the experiment N s 4 for 0. for We note that the predicted single solitons were not observed, perhaps because irregularities in the experimental potential prevent excitation of solely the lowest-order modes of the soliton train. Soliton stability was largely determined by the nal value for C a. As is well known 14 , when a 0 the condensate will expand. If a 0, continued increase of the magnitude of a or N will cause the condensate to contract to the extent that it can no longer be treated as quasi-1D. For our particular 23 17 . In between parameters this occurs at C critical lies a regime of stable soliton solutions, within which we nd that maximizing N s requires the magnitude of C to be large and close to C critical . It has been shown that the stability criterion restricts the size of any one soliton to 6000 atoms taking a 3a 0 ) 7 . Therefore, for a given phase gradient, a greater atom number can only be supported by the creation of more solitons. Experimentally, the number of particles in the released condensate greatly exceeds the number that can be stably contained in solitons 1 . In fact most of the released atoms are lost in a burst, and the solitons are formed from the small number of atoms in the remnant fraction. To investigate the effect of a burst event on the validity of our conclusions, we

FIG. 2. Implosion and ejection of atoms from the release of a strongly attractive condensate into a weak harmonic potential. The plot is of ( , ) 2 the condensate density distribution, vs , units of h.o.u. . At 0 the nonlinear constant here was ramped from C 200 to C 25 over a period of 5, and the condensate 1.5 into a harmonic potential V/500, where V was released at is the initial trapping potential.

employed Eq. 4 to take into account the higher peak densities experienced during the negative ramp as the condensate collapses, and performed simulations with negative values of C just outside the stability regime of the loss-less equation. Depending on the extent of the nonlinearity we observed both implosions Fig. 2 and solitonlike pulses with this model. Having studied the process of soliton formation, we then addressed the issue of soliton interactions. Collisions between solitons will occur if particles are given enough kinetic energy to overcome the short-range repulsive force, which scales exponentially with pulse separation 16 . Therefore by placing a series of colinearly propagating repulsive

FIG. 3. Colinear scattering of two repulsive solitons of approximately equal size in a harmonic potential, in which the particle momentum is more dominant than the repulsive interaction. a is a plot of ( , ) 2 vs , units of h.o.u. . b shows ( , ) 2 vs for different values of in a collision sequence. Note the change in scale for the fourth subplot. Soliton 1 overtakes and passes through soliton 2 every half a period, at the extreme ends of the potential. At 0 the nonlinear constant here was ramped from C 200 to C 10 over a period of 5, and the condensate was released at 1.5 displaced 20 from the center as for Fig. 1, into a harmonic potential of the form V/500, where V is the initial trapping potential. 061602-3

RAPID COMMUNICATIONS

LEUNG, TRUSCOTT, AND BALDWIN

PHYSICAL REVIEW A 66, 061602 2002 IV. CONCLUSION

solitons in a steep potential well, it should be possible to extend experiment 1 into a BEC soliton scattering experiment. Figure 3 shows a pair of pulses in an external potential chosen so that the repulsive interaction is weak compared to the particle momentum. From symmetry, the motion of completely noninteracting pulses will bring them into coincidence at the center of a harmonic potential. If the particles are repulsive we may expect the point of collision to occur later in the oscillation, as is indeed observed here. During the collision, the pulses are barely distinguishable, and the discrete phase steps between pulses are temporarily lost. After the scattering event each pulse regains its identity with no obvious particle exchange, as is evident from the lack of growth of any one soliton at the expense of another after many cycles. Following each collision the solitons eventually become sufciently well separated for their opposing phase relationships to again become apparent. Each collision causes a spatial displacement of the solitons similar to the scattering of elastic particles. Thus the results are akin to those found for the scattering of optical 1D solitons and 3D light bullets in a 1D potential 16,18 . Further analysis of soliton interaction should be possible through an analytical model such as presented in Ref. 19 .

As for their optical counterparts, bright matter-wave solitons may be useful as non-linear atom optical wave packets. In this paper we have attempted to demonstrate that further manipulation of bright solitons with existing technology can potentially uncover a rich variety of interesting nonlinear matter-wave phenomena. We have reproduced the soliton trains observed in Ref. 1 and also shown that experiment 1 is capable, with minimum modication, of demonstrating the scattering of colinearly propagating solitons, allowing measurement of the repulsive force between them. Spatial coherence may also be demonstrated if interference fringes are observed during the collision 20 . This has implications for the use of soliton trains as an atom laser. In addition, pulses of different particle number will have different sized nonlinearities, and hence different interaction potentials. Finally, creation of mutually attractive solitons for 0 will allow the rst studies of matter-wave soliwhich ton capture and fusion.
ACKNOWLEDGMENTS

We gratefully acknowledge Yu. S. Kivshar and E. A. Ostrovskaya for helpful comments on the manuscript.

1 2 3 4 5 6 7 8 9 10 11 12

K.E. Strecker et al., Nature London 417, 150 2002 . L. Khaykovich et al., Science 296, 1290 2002 . S. Burger et al., Phys. Rev. Lett. 83, 5198 1999 . J. Denschlag et al., Science 287, 97 2000 . Z. Dutton et al., Science 293, 663 2001 . L.D. Carr et al., Phys. Rev. A 62, 063610 2000 . V.M. Perez-Garca et al., Phys. Rev. A 57, 3837 1998 . H. Michinel et al., Phys. Rev. A 60, 1513 1999 . L.D. Carr et al., J. Phys. B 33, 3983 2000 . E.A. Donley et al., Nature London 412, 295 2001 . L.D. Carr and Y. Castin, e-print cond-mat/0205624. U.A. Khawaja et al., e-print cond-mat/0206184.

A. Gorlitz et al., Phys. Rev. Lett. 87, 130402 2001 . P.A. Ruprecht et al., Phys. Rev. A 51, 4704 1995 . H. Saito and M. Ueda, Phys. Rev. Lett. 86, 1406 2001 . J.P. Gordon, Opt. Lett. 8, 596 1983 . For a lithium condensate with a 3a 0 in a cigar-shaped trap with v 70 Hz, this corresponds to N 70 000, where a 0 is the Bohr radius. 18 D.E. Edmundson and R.H. Enns, Phys. Rev. A 51, 2491 1995 . 19 P.V. Elyutin et al., Phys. Rev. E 64, 016607 2001 . 20 M.R. Andrews et al., Science 275, 637 1997 . 13 14 15 16 17

061602-4