
2
underlying velocity field, associated with the phase of
the condensate, and to follow its subsequent evolution in
time by means of swarms of trajectories. It was shown
by Benseny et al. [25] that these trajectories constitute
a convenient tool to determine how each element of the
condensate (not to be confused with each individual atom
itself) moves apart, thus providing some clues on its dy-
namical evolution beyond more conventional-type infor-
mation, such as the density distribution. In other words,
it is possible to determine which portions in the BEC are
going to separate faster or slower at each time by simply
observing local variations in the underlying velocity field.
In this work we investigate by means of a series of nu-
merical simulations the formation of dark solitons both
from two freely released condensates and also under the
action of an underlying harmonic trap in order to study
the appearance and persistence of recurrences in time.
The appearance of soliton-type solutions is associated
with the value of the scattering length, as. Specifically,
if as<0, as it happens in 85Rb [26] or 7Li [27], atomic
interactions are attractive and bright soliton solutions
(spike-type deformations in the density) can appear. On
the contrary, for as>0, as it is the case of 87Rb [28]
or 23Na [29], the interactions are repulsive, which corre-
sponds to a situation where dark solitons (dips in the den-
sity) can be observed. The first experimental observation
of dark solitons dates back to 25 years ago, where these
solitons were produced in an elongated (cigar shaped)
BEC by the so-called phase imprinting method [30]. A
year before, though, Scott et al. [31] identified the for-
mation of persistent dark fringes in the the case of the
collision of two separated condensates under the influ-
ence of a harmonic trap. This work showed evidence of
the relationship between the dark solitons and the phase
change, an idea that is implicit in the phase imprinting
method devised to generate vortices in the condentasate
[32], but that also led to the experimental generation of
dark solitons [30]. An overview of the advances in dark
soliton dynamics both experimentally and theoretically
can be found in [33, 34]. Such phase imprinting methods
are actually strongly connected to the Bohmian trajec-
tories here. As it is shown below, these trajectories will
allow us to monitor in real time some features that are
not evident at a first glance from the usual density distri-
butions, and also to elucidate some important differences
between the nonlinear and linear dynamical regimes that
characterize the BEC dynamics in harmonic traps. This
is by virtue of the relationship between the trajectories
and the local phase of the condensate. Indeed, pioneering
work by Tsuzuki [35] in the early 1970s on the derivation
of soliton-like solutions of the Gross-Pitaevskii equation
are strongly connected to this formulation (Tsuzuki’s is
only a first approximation to the exact approach here
considered).
The work has been organized as follows. In Sec. II the
model and computational details are introduced as well
as some elementary notions of the Bohmian formulation.
In Sec. III the results obtained from the numerical simu-
lations carried out for the two scenarios mentioned above,
namely, free propagation and motion inside a harmonic
trap, are presented and discussed. To conclude, some
final remarks are summarized in Sec. IV.
II. THEORY
A. The reduced one-dimensional GPE
Within the Hartree-Fock approximation, the dynamics
of a BEC consisting of Nidentical atoms with mass mis
described by the Gross-Pitaevskii equation [5],
iℏ∂Ψ(r, t)
∂t =−ℏ2
2m∇2+Vext(r, t) + gN(r, t)Ψ(r, t),
(1)
where Vext describes any external interaction acting on
the BEC (e.g., the confining optical trap), while the
nonlinear term gN(r, t) accounts for the self-interaction
contribution, i.e., the effective interaction with the re-
maining N−1 atoms in the cloud. The strength of
such interaction is determined by the coupling constant
g= 4πℏ2as/m, with asbeing the scattering length [7].
For convenience, the wave function (order parameter) can
be recast in polar form, as
Ψ(r, t) = pN(r, t)eiθ(r,t),(2)
which enables a description of the condensate in terms
of its density distribution, N(r, t) = |Ψ(r, t)|2, with
R|Ψ(r, t)|2dr=N, and its local phase variations, ac-
counted for by θ(r, t). Here, for numerical convenience,
we have chosen the wave function to be normalized
to unity, which implies that the factor Narising from
N(r, t) will be included as a multiplicative constant in g,
that is, from now on g= 4πℏ2asN/m.
Consider that a prolate configuration for the con-
densate, which arises assuming typical frequency ranges
fz∼20−60 Hz and f⊥∼400−900 Hz. Accordingly, the
transverse harmonic trap values will be much larger than
the longitudinal one, since ω2
⊥≫ω2
z, and hence, for the
times considered, the transverse degrees of freedom can
be assumed to be frozen. This thus allows us to provide
an effective description of the condensate dynamics along
the z-direction by means of a reduced one-dimensional
(1D) GPE [7], which reads as
iℏ∂ψ(z, t)
∂t =−ℏ2
2m
∂2
∂z2+Vext(z) + g1Dn(z, t)ψ(z, t),
(3)
where g1D=g/2πa2
⊥= 2ℏω⊥as, with a⊥=pℏ/mω⊥,
is an effective 1D coupling constant that arises from
the dimensional reduction of Eq. (1) [36, 37] (see fur-
ther details about the potential parameters in Sec. II C).
From now on, for computational convenience we assume
that the wave function ψ(z, t) is normalized to unity,
i.e., R|ψ(z, t)|2dz =Rn(z, t)dz = 1, with n(z, t) being
a linear density distribution (measured in µm−1). Ac-
cordingly, the expression for the coupling constant g1D