Author's personal copy

Physica D 238 (2009) 14391448

Contents lists available at ScienceDirect
Physica D
journal homepage: www.elsevier.com/locate/physd
Matter-wave vortices and solitons in anisotropic optical lattices
Thawatchai Mayteevarunyoo
a,b
, Boris A. Malomed
a,
, Bakhtiyor B. Baizakov
c,d
, Mario Salerno
d
a
Department of Physical Electronics, School of Electric Engineering, Faculty of Engineering, Tel Aviv University, Tel Aviv 69978, Israel
b
Department of Telecommunication Engineering, Mahanakorn University of Technology, Bangkok 10530, Thailand
c
Physical-Technical Institute of the Uzbek Academy of Sciences, 2-b G. Mavlyanov Street, 700084, Tashkent, Uzbekistan
d
Dipartimento di Fisica E.R. Caianiello and Consorzio Nazionale, Interuniversitario per le Scienze Fisiche della Materia (CNISM), Universit di Salerno, I-84081 Baronissi (SA), Italy
a r t i c l e i n f o
Article history:
Available online 19 August 2008
PACS:
03.75.Lm
05.45.Yv
42.70.Qs
Keywords:
BoseEinstein condensate
Photonic crystal
Collapse
Square vortex
Rhombus vortex
Quadrupole
Gap soliton
Dipole soliton
a b s t r a c t
Using numerical methods, we construct families of vortical, quadrupole, and fundamental solitons in
a two-dimensional (2D) nonlinear-Schrdinger/GrossPitaevskii equation which models BoseEinstein
condensates (BECs) or photonic crystals. The equation includes the attractive or repulsive cubic
nonlinearity and an anisotropic periodic potential. Two types of anisotropy are considered, accounted for
by the difference in the strengths of the 1Dsublattices, or by a difference in their periods. The limit case of
the quasi-1D optical lattice (OL), when one sublattice is missing, is included too. By means of systematic
simulations, we identify stability limits for two species of vortex solitons and quadrupoles, of the rhombus
and square types. In the attraction model, rhombic vortices and quadrupoles remain stable up to the limit
case of the quasi-1Dlattice. Inthe same model, finite stability limits are foundfor vortices andquadrupoles
of the square type, in terms of the anisotropy parameter. In the repulsion model, rhombic vortices and
quadrupoles are stable in large parts of the first finite bandgap (FBG). Another species of partly stable
anisotropic states is found in the second FBG, subfundamental dipoles, each squeezed into a single cell of
the OL. Square-shaped quadrupoles are completely unstable in the repulsion model, while vortices of the
same type are stable only in weakly anisotropic OL potentials.
2008 Elsevier B.V. All rights reserved.
1. Introduction
Asubject of great current interest in theoretical and experimen-
tal studies of the dynamical patterns formation in BoseEinstein
condensates (BECs) and photonic crystals is the existence and sta-
bility of multidimensional solitons and localized vortices in two-
and three-dimensional (2D and 3D) settings. While solitons have
been created in well-known experiments in BEC with attractive in-
teractions between atoms, loaded in effectively 1D (cigar-shaped)
traps [1,2], a basic problem impeding straightforward creation of
multidimensional solitons is the instability against collapse driven
by the cubic self-attractive nonlinearity [3]. As concerns ring-
shaped solitons with intrinsic vorticity, they are additionally sub-
ject to the symmetry-breaking azimuthal instability, which is even
stronger than the instability to the collapse [4].
It was proposed[58] that a general methodfor the stabilization
of multidimensional solitons and localized vortices in BEC may
be based on the use of optical lattices (OLs), i.e., periodic
potentials which are induced by the interference between
counterpropagating laser beams illuminating the condensate.

Corresponding author.
E-mail address: malomed@eng.tau.ac.il (B.A. Malomed).
Stable 2D and 3D solitons can be supported by full OLs (with
dimension D equal to that of the space in which the lattice was
created), and also by low-dimensional OLs, with dimension D
1, i.e., quasi-1D and quasi-2D lattices in the 2D and 3D space,
respectively [7,10]. In the latter case, the solitons naturally feature
a strongly anisotropic shape.
Although the OL breaks the rotational invariance, it can support
and stabilize not only fundamental solitons, but also vortical
ones [5,6,8,9], including localized vortices of higher orders, with
spin (the topological charge, alias vorticity) S 1 [11]. The
simplest crater-shaped vortex soliton, in the form of a single
density peak with an inner hole induced by the vorticity, is always
unstable [12,13] (the instability splits it into several pulses, one of
which survives, demonstrating a random walk across the OL [13]).
Stable vortices with S = 1 are built as sets of four [5,6] or eight [12,
23] peaks, with the phase difference, respectively, = /2 or
/4 between adjacent ones, which corresponds to the total phase
circulation of 2. There are two different species of the simplest
stable vortex solitons, which are composed of four peaks: densely
packed squares, with the center coinciding with a local maximum
of the OL potential [6,13], and porous rhombuses, featuring
a nearly empty lattice cell at the center [5,12,13] (two similar
species of vortex solitons are also known in discrete models, where
rhombuses are sometimes called crosses [14]). It was recently
0167-2789/$ see front matter 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.physd.2008.07.024
Author's personal copy
1440 T. Mayteevarunyoo et al. / Physica D 238 (2009) 14391448
demonstrated [9] that the squares and rhombuses form separate
families, each featuring a pair of branches connected at a turning
point. The modes belonging to the different branches differ by
the number of peaks (for instance, four and eight peaks that may
constitute a rhombic vortex, as mentioned above).
Stable vortices of higher orders, up to S = 6, were constructed
as ring-shaped sets consisting of up to N = 12 peaks, with
respective phase differences = 2S/N [11]. The localized
vortices formally corresponding to S = 2, which are composed
of 4 peaks, have = , i.e., these are actually real solutions
in the form of quadrupoles. Also predicted were stable topological
patterns in the form of supervortices, which are ring chains built
of 12 (or more) compact crater-shaped vortices carrying local
spins s = 1, with global vorticity S = 1 imprinted onto
the entire ring [11]. The supervortices may be stable, even if
individual crater peaks, of whichthey are composed, are unstable
in isolation, as mentioned above.
The quasi-2D lattice in the 3D space can also support stable
3D solitons with embedded vorticity [15]. On the other hand, the
existence and stability of vortices in the 2D model with the quasi-
1D lattice remains an open problem.
Solitons of a different type, namely, gap solitons (GSs), can
be supported by full OL potentials of any dimension in BEC
with repulsive interactions between atoms. The GSs result from
the balance between the repulsive nonlinearity and the negative
effective mass in parts of the linear bandgap spectrum generated
by the OL [16,17]. They are stable localized objects [18], even if
they cannot realize the ground state of the condensate trapped in
the OL. The creation of GSs containing 250 atoms of
87
Rb was
reported in the effectively 1D setting [19] (see also review [20]).
Multidimensional GSs [21], including gap-type vortices [12,22,
23], and semi-gap solitons (which are organized as GSs in one
direction, and regular solitons in the other [24]) were predicted
too. As concerns gap-soliton vortices, they may feature both the
square [22] and rhombic [12,22,23] shapes. In both 1D and 2D
settings, stable GSs may be supported not only by periodic OLs, but
also by quasi-periodic lattices [25].
Thus far, the studies of 2D and 3D solitons supported by
OLs were confined to two limit cases, viz., the full (isotropic)
lattices, and low-dimensional ones, with one sublattice missing.
In experiments, it is quite easy to create a more general setting,
with an anisotropic OL, composed of 1D sublattices with different
strengths and/or different periods. To the best of our knowledge,
fundamental solitons and vortices in anisotropic lattices were
previously studied only in the discrete model [14], that may be
considered as a model for the BEC trapped in a very deep OL
[17,26].
Similar settings are available for the experiment in nonlinear
optics, where photonic lattices can be induced in photorefractive
crystals illuminated by pump laser beams in the ordinary
polarization (in which the medium is nearly linear), while solitons
are created by probe beams launched in the extraordinary
polarization [27]. In addition to fundamental 2D solitons [28],
localized vortices [29], necklace-shaped [30] and circular [31]
solitons have been created by means of this technique. In
particular, 2D anisotropic solitons supported by an isotropic
square-shaped photo-induced lattice were reported in Ref. [34].
Asymmetrically shaped vortex solitons were predicted in that
medium too [35].
Another realization of 2D solitons [32] and vortices [33] in
nonlinear optics is possible in photonic-crystal fibers. Unlike the
saturable nonlinearity characteristic to photorefractive crystals,
they feature the same cubic (Kerr) nonlinearity as BEC.
The anisotropic-lattice settings for solitons, vortices and
quadrupoles, which are the subject of the present work, suggest
to consider several issues of evident interest. One of them is
finding stability borders for 2D vortex and quadrupole solitons
in the model with the attractive cubic nonlinearity. Another
straightforward question is to identify existence and stability
limits for 2D GSs in the model with the self-repulsion, where,
obviously, solitons cannot exist in the quasi-1D limit, when one
of 1D sublattices is switched off. The first noteworthy finding
reported below is that both the vortices (with S = 1) and
quadrupoles of the rhombus type exist and remain stable up to
the limit of the quasi-1D lattice (in the attraction model). For the
square-shaped vortices and quadrupoles, we find a critical degree
of the anisotropy of the 2Dlattice, up to which they remain stable
unlike their rhombic counterparts, they are unstable in the limit of
the quasi-1Dlattice. Generally, the stability region for quadrupoles
in the attraction model is essentially broader than for vortices. In
the repulsion model, we find the existence and stability limits for
the fundamental (S = 0), vortical (S = 1) and quadrupole GSs. In
this case too, the rhombuses are much more stable than squares,
but quadrupoles are found to be less stable than vortices, on the
contrary to the attraction model.
These results predict universal properties of fundamental,
vortical, and quadrupole 2D solitons in nonlinear periodic
media, that can be realized in BEC, photonic crystals and photonic-
crystal fibers, and other physical media. Experimental verification
of the predictions is quite feasible both in BEC and photonic lattices
in photorefractive crystals.
The rest of the paper is organized as follows. In Section 2,
we formulate the model and demonstrate spectra generated by
the OLs in its linear version. Systematic findings for the 2D
vortices and quadrupoles of the rhombus and square type are
reported in Sections 3 and 4, for the models with attraction and
repulsion, respectively. While the results are obtained by means
of systematic simulations, Section 3 includes a brief discussion
which aims to explain some findings by means of an analytical
approximation. The paper is concluded by Section 5 in which
the main results are summarized, and examples of stable three-
dimensional GSs supported by the respective anisotropic OL are
additionally displayed.
2. The model: nonlinear equations and linear spectra
2.1. The GrossPitaevskii equation
The starting point is the 3D GrossPitaevskii equation for the
mean-field wave function, (X, Y, Z, T), where the coordinates
and time denoted by capital letters are measured in physical units:
i

T
=

h
2
2m
_

X
2
+

2

Y
2
+

2

Z
2
_
+
4

h
2
a
s
m
||
2
+W (X, Y, Z) . (1)
Here m and a
s
are the atomic mass and scattering length of atomic
collisions, and W is the external potential. For the anisotropic 3D
lattice with strength W
0
and period d, the potential is
W = W
0
_
cos
_
2X
d
_
+cos
_
2Y
d
_
+cos
_
2Z
d
__
, (2)
where anisotropy factor takes values 0 < 1, with the
isotropic and quasi-2D limits corresponding, respectively, to =
1 and = 0. If the 1D components of the OL are induced by
the superposition of two counterpropagating laser beams with
wavelength and misalignment angle 2, the corresponding OL
period is d = / (2 cos ). Besides using different intensities of
light in different pairs of beams, which is accounted for by < 1 in
Eq. (2), another source of the anisotropy may be the use of different
Author's personal copy
T. Mayteevarunyoo et al. / Physica D 238 (2009) 14391448 1441
Fig. 1. (a) The edge of the SIG (semi-infinite gap), as found from Eq. (8) at different values of anisotropy coefficient and fixed = 1. (b) The same for different values of
and fixed = 1. The SIG in the spectrum of linearized equations (5) and (6) is located to the left of the edge point.
angles for the beampairs inducing X, Y, and Z-components of the
lattice potential, i.e., different periods of these components.
In the 3D case, one can rescale the coordinates, time, potential,
and wave function as follows: (X, Y, Z) (d/) (x, y, z),
T
_
2md
2
/
2

h
_
t, W
0

_

h
2
/2md
2
_
V
0
, (X, Y, Z, T)
(2d)
1

/2 |a
s
|u (x, y, z, t), which casts Eq. (1) in the normalized
form,
iu
t
=
_
u
xx
+u
yy
+u
zz
_
+|u|
2
u V
0
[ cos (2x)
+ cos (2y) +cos (2z)], (3)
where sgn (a
s
) is 1 and +1 for the model of with the self-
attraction and self-repulsion, respectively.
2.2. The two-dimensional model: the formulation and linear spectra
The setting which can be reduced to the 2D model is based
on Eq. (1) with a combination of the strong confinement in one
direction and anisotropic OL in the perpendicular plane,
W =
1
2
m
2
z
Z
2
W
0
_
cos
_
2X
d
_
+ cos
_
2Y
d
__
, (4)
where the anisotropy parameter again takes values 0 <
1. The reduction to the 2D equation is performed,
as usual, by means of substitution [36] (X, Y, Z, T) =
exp
_
(i/2)
z
T Z
2
/
_
2a
2
z
__
U (X, Y, T) in Eq. (1), where the
transverse-confinement lengthis a
z
=

h/m
z
. Further, rescaling
(X, Y) (d/) (x, y), T
_
md
2
/
2

h
_
t, W
0

_

h
2
/md
2
_
V
0
,
d
1
_
_
/2

2 |a
s
|
_
u(x, y, t) casts the resulting 2Dequation in
the normalized form,
iu
t
= (1/2)
_
u
xx
+u
yy
_
+ |u|
2
u
V
0
[cos (2x) + cos (2y)] u, (5)
with = 1 having the same meaning as in Eq. (3).
As said above, the anisotropy of the OL may also be induced
by different periods of its quasi-1D components of the OL, which
corresponds to the following modification of the 2D equation:
iu
t
= (1/2)
_
u
xx
+u
yy
_
+ |u|
2
u
V
0
[cos (2x) +cos (2y)] u, (6)
where the ratio of the two periods is defined to be 1. The
second term in this potential becomes negligible for 1, as
the strength of the interaction of a short-period potential with a
soliton of a finite width is exponentially small [5,7]. The number of
atoms in the condensate is given by
__
| (X, Y, Z)|
2
dXdYdZ
_
2

2
_
1
(a
z
/ |a
s
|) N, where
N =
___
|u(x, y)|
2
dxdy (7)
is the 2D norm of the rescaled wave function.
Stationary solutions to Eq. (5) or (6) with chemical potential
are looked for in the ordinary form, u (x, y, t) = exp(it)u (x, y),
where function u(x, y) satisfies equation
u +(1/2)
_
u
xx
+u
yy
_
+V
0
[cos (2x) + cos (2y)] u |u|
2
u = 0 (8)
(we will consider the two types of the anisotropy separately,
i.e., either < 1, = 1 or = 1, > 1). Relevant solutions
to Eq. (8) are real for fundamental solitons and quadrupoles, and
complex for vortices.
In the attraction model, we will be looking for solitons with
belonging (as usual) to the semi-infinite gap (SIG) in the
linear spectrum generated by Eq. (8), while in the model with
the repulsive nonlinearity solutions will be sought for in finite
bandgaps (FBGs). The bandgap spectrum of the isotropic version
of Eq. (8) (with = 1) is well known (see, e.g., Ref. [20]), and in the
opposite limit of the quasi-1DOL it reduces simply to the SIGof the
respective Mathieu equation, u + (1/2) u

+ V
0
cos (2x) u = 0,
and does not include any FBG.
The spectrum of the anisotropic OL should be calculated anew
for 0 < < 1, with = 1, as well as for > 1 and = 1.
We have performed this analysis by means of standard numerical
methods. The so found border of the SIG is shown, as a function of
OL strength V
0
and anisotropy parameters and , in Fig. 1, and
the intervals of occupied by the first FBG are shown in Fig. 2.
3. Stability limits for vortex and quadrupole solitons in the
model with attraction
3.1. General approach
As said above, in the model combining the self-attractive
nonlinearity and anisotropic lattice potential, the identification of
stability limits for basic topologically structuredstates, i.e., vortices
and quadrupoles, is an issue of major interest, while it was known
before that fundamental solitons are stable in the 2D and quasi-1D
Author's personal copy
1442 T. Mayteevarunyoo et al. / Physica D 238 (2009) 14391448
Fig. 2. (a) Borders of the first FBG (finite bandgap) in the spectrum of linearized equation (5). (b) The same for Eq. (6).
lattices alike. In fact, the stability for the fundamental solitons
can be easily predicted by means of the VakhitovKolokolov
(VK) criterion [7]. Formally, the same criterion predicts the
stability of vortices and quadrupoles too, but it is well known
[4] that topological solitons are prone to instabilities against
perturbation modes associated with complex eigenvalues, which
cannot be detected by the VK criterion. Therefore, a careful
analysis of the stability of the vortex and quadrupole states by
means of numerical methods is necessary. Below, we perform it
through systematic direct simulations of the evolution of slightly
perturbed vortices and quadrupoles. Another approach may be
based on the computation of eigenvalues from the linearization of
Eqs. (5) and (6). However, that approach, although being a more
mathematically rigorous one, is quite involved in the present
situation, and, on the other side, direct simulations are better
adjusted to modeling the experiment.
In addition to the numerical analysis, we will briefly present
an approximate analytical method, based on the consideration
of vortices and quadrupoles as sets of four weakly interacting
peaks (fundamental solitons). Some crude results can be obtained
by means of this approximation, but in most cases it does not
produce definite predictions, as the assumption of very weak
interactions between the peaks turns out to be inadequate. On the
other hand, for stronger interactions the approximation becomes
cumbersome.
To collect the results reported below, we first constructed
the respective solution families numerically, as localized states
generated by stationary Eq. (8). Then, as said above, their stability
was tested by means of direct simulations of Eq. (5) or (6), using
the split-step code in domain 6 < x, y < +6, covered by a
grid of size 128 128 points, with absorbers installed at its edges.
The stability was run, typically, up to t = 2000, and the respective
states were classified as stable ones if they kept no less than 99% of
the initial norm. In physical units, this time corresponds to 0.1 s,
for the BEC of
7
Li atoms and OL period d 2 m. It may be
estimated as 50 diffraction times determined by the total size of
typical stable patterns displayed below. For unstable ones, much
shorter evolution times are quite sufficient to make the instability
evident.
The numerical solution of Eq. (8) was carried out by means of a
modification of the squared-operator method [37], which provides
for quick convergence of the iteration scheme in the present
setting. The following initial guesses with arbitrary amplitude A
0
were used to construct the simplest vortex solitons of the square
and rhombus types, respectively:
u
(0)
sq
(x, y) = A
0
_
exp
_

_
x
2
+y
2
__
+i exp
_

_
(x )
2
+y
2
__
exp
_

_
(x )
2
+
_
y
n

_
2
__
i exp
_

_
x
2
+
_
y
n

_
2
___
, (9)
u
(0)
rh
(x, y) = A
0
_
exp
_

_
x
2
+
_
y +
n

_
2
__
+i exp
_

_
_
x
n

_
2
+y
2
__
exp
_

_
x
2
+
_
y
n

_
2
__
i exp
_

_
(x +)
2
+y
2
___
, (10)
where n is an integer. Actually, we set n = 1 for the anisotropic
OL with < 1 and = 1. In the case of the lattice with = 1
and > 1, we took n = 1 for 1 < 2, while for integer
values 2 we fixed n = , i.e., the configurations were square-
like in the latter case too. To generate quadrupoles, the complex
string of pre-exponential coefficients in expressions (9) and (10),
(1, i, 1, i), was replaced by a real one, (1, 1, 1, 1).
3.2. Rhombus-shaped patterns
One of essential results obtained in this work is that rhombic
quadrupoles and vortices remain stable in the limit of the quasi-1D
lattice, which is described by Eq. (5) with = 0. This conclusion is
similar to that for the fundamental solitons, whose stability in the
presence of the quasi-1D OL was established before [7,8]. Typical
examples of the stable evolution in the model with = 0 are
displayed in Fig. 3. The initial configurations included in this figure
provide for a description of the density and phase distributions in
the stationary rhombus-shaped patterns.
The patterns observed in Fig. 3 seem nearly isotropic, despite
the fact that the setting is strongly anisotropic, which is explained
by the fact that the examples are taken deep enough in the
SIG, where the model tends to make them effectively isotropic.
Examples taken close to the gaps edge seem much more
anisotropic, see, e.g., Ref. [7].
In the model with the anisotropic OL based on Eq. (6), the
rhombic vortices and quadrupoles are stable up to 10
(for 7, they become more sensitive to perturbations and
demonstrate small intrinsic oscillations, but remain overall stable).
These patterns are quite similar to those displayed in Fig. 3 (in
the case of large integer , they were built using configuration (9)
with n = , which corresponds to the same size of the rhombus
as in Fig. 3).
Author's personal copy
T. Mayteevarunyoo et al. / Physica D 238 (2009) 14391448 1443
Fig. 3. (Color online) Generic examples of the stable evolution of rhombus-shaped quadrupole (a-d) and vortex (e), in the attraction model ( = 1) with the quasi-1D
lattice ( = 0), for V
0
= 5. The chemical potential of both unperturbed states is = 7. (a, b): The evolution of the density in the quadrupole soliton along lines y = 0 and
x = 0. (c): Contour plots of the initial and final distributions of the density and phase. (d): An overview of the real solution for the stationary quadrupole. (e): The same as
in (c), but for a stable vortex soliton (the density evolution for the vortex, which is not displayed here, is very similar to that shown in (a) and (b) for the quadrupole).
3.3. Square-shaped patterns
The test of the stability of square quadrupoles and vortices in
the attraction model with the anisotropic OL, based on Eq. (5),
demonstrates that they may be both stable and unstable, see
typical examples in Figs. 4 and 5, respectively. As can be concluded
from Fig. 5 and is corroborated by systematic simulations, the
instability of the quadrupoles (if they are unstable) develops
slower than that of the vortices. Each unstable vortex eventually
collapses into a fundamental soliton, losing a part of the initial
norm. The outcome of the instability development in square-
shaped quadrupoles is less clear, as the process is very slow.
Systematic simulations performed at many values of the
parameters are summarizedinthe stability diagramfor the square-
shapedquadrupoles andvortices inthe plane of the lattice strength
and anisotropy, (V
0
, ), see Fig. 6. A noteworthy feature of the
diagram is that the stability region for the quadrupoles is much
wider than for the vortices. On the other hand, we stress that,
although in interval 3 V
0
6 the minimum value,
min
,
necessary for the stability of the quadrupoles is very small, it
remains finite (as shown above, the square-shaped quadrupoles
suffer a slow decay at extremely small values of ).
A trend evident in Fig. 6, viz., the steep increase of
min
with
the decrease of V
0
, may be realized as a manifestation of the
delocalization transition in weak OLs, which leads to the decay
of 2D solitons [38]. Another manifestation of the same trend,
revealed by detailed considerations of numerical data, is that
persistent (although non-growing) intrinsic oscillations of stable
quadrupoles and vortices, initiated by small perturbations, are
more conspicuous in weak lattices, as well at close to
min
. On
the other hand, the growth of
min
observed in Fig. 6 at V
0
>
6 is explained by moving closer to the edge of the SIG, where
the patterns themselves become strongly anisotropic, their norm
drops to small values, and it is difficult to maintain their stability.
The square-shaped modes were also investigated for the other
type of the anisotropic OL, with = 1 and > 1, see Eq. (6).
In particular, it was found that the vortices with = 7 the
same value for which the stability diagram is displayed in Fig. 6
Author's personal copy
1444 T. Mayteevarunyoo et al. / Physica D 238 (2009) 14391448
Fig. 4. (Color online) The same as in Fig. 3, but for a typical stable square-shaped quadrupole (a-d), in the strongly anisotropic (but not quasi-1D) lattice with = 0.05
and V
0
= 2, and for a stable square vortex (e), with = 0.1 and V
0
= 5. The chemical potentials of the unperturbed quadrupole and vortex are = 5 and = 7,
respectively.
remain stable in the lattice with V
0
= 5 only up to 1.15,
and are unstable at larger . In the same case, the quadrupoles are
stable upto 1.2. However, unlike the vortices, the quadrupoles
regain their stability in some intervals at higher values of first,
for 1.8 2.12. At sill larger , stability regions are found
around = 7, 12, 18, 20, alternating with unstable intervals.
3.4. Analytical estimates
In the model with the attractive nonlinearity, an analytical
insight into the existence of complexes built of four far separated
peaks, alias fundamental solitons (see Figs. 3 and 4), can be
provided by the effective potential of the interaction between
2D fundamental solitons separated by large distance R [39],
U(R) = CR
1/2
exp
_

2R
_
cos (), where C > 0 is a
constant, and is, as above, the phase shift between the solitons.
The applicability condition for the use of this potential is that

2R must be large enough. Straightforward considerations

demonstrate that the single static configuration which can be
predicted by the combination of the pair-wise interactions
mediated by this potential (with corresponding to the rhombic
or square-shaped patterns, see Eqs. (9) and (10)), and a strong
quasi-1D lattice with = 0, is the rhombus-shaped quadrupole.
While the failure of this coarse approximation to predict the
patterns of other types does not mean that they cannot exist,
the possibility to predict the rhombic quadrupoles in such a
simple way suggests that this particular type of four-peak patterns
ought to be an especially robust one. This conclusion agrees with
the above conclusions that the rhombuses are more stable than
squares, and the quadrupoles are essentially more stable than
vortices.
Author's personal copy
T. Mayteevarunyoo et al. / Physica D 238 (2009) 14391448 1445
Fig. 5. (Color online) Typical examples of the instability development of square-shaped quadrupoles (a, b) and vortices (c, d) in a very strongly anisotropic lattice, with
= 0.035 and = 0.01 in (a, b) and (c, d), respectively. The other parameters are the same as in the corresponding parts of Fig. 4.
Fig. 6. (Color online) Stability borders for square-shaped quadrupoles and vortices
in the attraction model, at a fixed chemical potential, = 7 (at other values of
, the stability diagram is quite similar to the present one). The solitons are stable
above the respective borders. The dotted line is the edge of the SIG (semi-infinite
gap), beyond which solitons cannot exist in the attraction model.
4. Stability limits for fundamental and vortical solitons in the
model with repulsion
4.1. Fundamental gap solitons and subfundamental dipoles
A comprehensive investigation of the fundamental GSs (gap
solitons) in the model with the anisotropic OL (0 < < 1 or
> 1) and repulsive nonlinearity ( = +1 in Eqs. (5) and (6))
yields a simple conclusion: a stable fundamental soliton (exactly
one) can be found for each value of belonging to the first FBG
(finite bandgap). An example, taken close to the left edge of the
FBG is displayed in Fig. 7. Naturally, this soliton features a strongly
localized (tightly bound) shape along the strong direction, x, and
a loosely bound shape, typical to GSs, along the weak direction, y.
It is relevant to mention that the 1D model combining the
repulsive nonlinearity and OL gives rise to specific solutions
in the form of subfundamental (SF) solitons in the respective
second FBG, namely, odd (antisymmetric) tightly localized states,
which are squeezed, essentially, into a single lattice cell [40]
Fig. 7. (Color online) An example of a stable strongly anisotropic fundamental gap
soliton located near the left edge of the first FBG (finite bandgap), at = 2.4, in
the model with repulsion ( = +1 in Eq. (5)) and strongly anisotropic lattice, with
= 0.05 and V
0
= 5.
(subfundamental means that the norm of the soliton is smaller
than the norm of the fundamental soliton with the same value of
the chemical potential). However, all SF solitons are unstable in
the 1D model, shedding off a part of the norm and transforming
themselves into ordinary fundamental solitons which fall into the
first FBG [40].
In the second FBG of the 2D model, we have found strongly
localized states which may be considered as SF dipoles. They are
represented by stationary real solutions which feature an even
(symmetric) shape in one direction, and antisymmetric shape in
the perpendicular direction, as shown in Fig. 8. An essential result
is that the SF dipoles may be stable in the isotropic OL ( = = 1).
This finding is relevant to report in the present context, as the SF
dipoles are essentially anisotropic states, even if the underlying
lattice is isotropic. For the isotropic OL with V
0
= 5 (this value
pertains to Fig. 8), the second FBG is 1.8 < < +0.75, the SF
dipoles being stable in a part of it, 1.8 < < 0.75.
Various complex patterns of the dipole type bifurcating from
edges of FBGs were investigated in detail in recent work [41]. In
particular, it is plausible that the tightly bound SF dipoles reported
here may represent a continuation of a solution family which starts
Author's personal copy
1446 T. Mayteevarunyoo et al. / Physica D 238 (2009) 14391448
Fig. 8. (Color online) A typical example of the stable subfundamental dipole soliton with = 1, belonging to the second FBG of the isotropic lattice with V
0
= 5, in the
repulsion model. (a) The overview of the dipole; (b) the continuous and dotted curves represent, respectively, cross sections along axes x and y.
Fig. 9. (Color online) The same as in Fig. 3(e), but for a typical stable rhombic vortex
in the repulsion model with the anisotropic 2D lattice, for = 0.13 and V
0
= 5.
The chemical potential of the vortical soliton is = 2.5, which falls in the first
FBG.
as a loosely bound dipole pattern at a gaps edge, as shown in Fig.
13(b) of [41]. However, the stability of the patterns was not studied
in that work.
4.2. Rhombic vortices and quadrupoles
We studied the existence and stability of multi-peak complexes
forming topological solitons in the repulsion model within the first
FBG. As well as in the model with attraction, rhombus-shaped
modes are found to be essentially more stable than their square
counterparts. In particular, it was found that, in the isotropic
OL, the rhombic vortices and quadrupoles exist and are stable
everywhere inside the first FBG, in agreement with the conclusions
of Refs. [22,23].
Further, systematic analysis of the model with the anisotropic
lattice has made it possible to identify stability limits for these
patterns at fixed values of the chemical potential. In particular,
for = 2.5 and V
0
= 5, the vortices are stable in interval
0.125 1, see a typical example in Fig. 9, while the
rhombic quadrupoles are stable in a somewhat narrower range,
0.22 1.
Infact, the above-mentionedstability regionfor the quadrupoles
includes an area where they seem completely stable (at 0.7),
and an interval of quasi-stability, where, at smaller values of ,
the rhombic quadrupoles develop intrinsic oscillations and thus
turn into robust breathers, keeping their overall shape. An example
of the breather, taken close to the instability border, is displayed in
Fig. 10. In the other version of the anisotropic lattice, with = 1
and > 1 (see Eq. (8)), the rhombic vortices and quadrupoles re-
main stable up to quite large values of (we did not try to find
exact stability borders in that case).
More sophisticated species of stable rhombus-shaped vortices
have been found too. For instance, a stable soliton with vorticity
S = 1 can be constructed, in a moderately anisotropic OL, as a
set of eight (rather than four) major local peaks, see an example
in Fig. 11.
4.3. Square-shaped vortices and quadrupoles
Patterns of the square type are much less stable in the repulsion
model than their rhombic counterparts. In particular, the square-
shaped quadrupoles are always unstable, even in the isotropic OL
(note that stable examples of GS quadrupoles which were reported
in Ref. [23] had the rhombic shape). The square vortices are stable
only in weakly anisotropic OLs. For instance, for V
0
= 5 and
= 4, the stability region is 0.65 < 1. In this region, the
vortices are sensitive to perturbations, responding to them with
conspicuous intrinsic vibrations, but nevertheless keep the overall
stability. The vortices are easily destabilized too in the other type
of the anisotropic OL, with = 1 and slightly exceeding 1, see
Eq. (8).
5. Conclusion
The objective of this work was to establish stability lim-
its for the two species of topologically organized rhombus-
and square-shaped solitary, vortices (with topological charge
S = 1) and quadrupoles, in the 2D model based on the
GrossPitaevskii/nonlinear-Schrdinger equation with the attrac-
tive or repulsive nonlinearity and anisotropic OL (optical-lattice)
potential. In addition to BEC, the model may describe spatial pat-
terns in photonic crystals. Two different types of the anisotropy
were considered, accounted for by different strengths of the two
sublattices, or by a difference in their periods; the limit case of
the quasi-1D potential was included too. In the repulsion model,
we also aimed to investigate the stability of fundamental GSs
(gap solitons).
Author's personal copy
T. Mayteevarunyoo et al. / Physica D 238 (2009) 14391448 1447
Fig. 10. (Color online) The same as in Fig. 3(ac), but for a quasi-stable rhombic quadrupole in the repulsion model including an anisotropic 2D lattice with = 0.25 and
V
0
= 5 (the instability border is at = 0.22, in this case). The chemical potential of the unperturbed quadrupole is = 2.5. It spontaneously rearranges itself into a
persistent breather.
Fig. 11. An example of a stable 8-peak vortex, found in the repulsion model with a
strong moderately anisotropic lattice, V
0
= 10, = 0.5.
It has been found that, in the attraction model, the rhombus-
shaped quadrupoles and vortices remain stable up to the limit
of the quasi-1D lattice (similar to the fundamental solitons,
for which this property was known before). An explanation
was proposed for this finding, based on the consideration of
the rhombus-shaped quadrupole as a weakly bound complex
of four fundamental solitons. For square-shaped quadrupoles
and vortices, finite stability limits were found, in terms of the
anisotropy parameter (). In the repulsion model, we have found
Table 1
The summary of stability and instability properties of the localized states in the
models with the attractive and repulsive nonlinearity
Type of mode Attraction Repulsion
Fundamental soliton Fully stable Fully stable
Rhombic vortex Fully stable Partly stable
Rhombic quadrupole Fully stable Partly stable
Square vortex Partly stable Chiefly unstable
Square quadrupole Partly stable Fully unstable
Subfundamental dipole n/a Partly stable
Partly stable means that the square-shaped patterns in the attraction model
become unstable in the limit case of the quasi-1D lattice, or that the rhombus-
shaped patterns and SF dipoles are unstable in parts of the respective FBGs (finite
bandgaps) of the repulsion model. Chiefly unstable means that the square-shaped
vortices in the latter model are quickly destabilized with the increase of the
anisotropy of the 2D lattice.
that stable fundamental solitons completely fill out the first finite
FBG (finite bandgap). Rhombic vortices and quadrupoles are also
stable in large parts of the first FBG (sometimes, they feature
persistent intrinsic oscillations, but maintain the overall shape). In
addition, tightly bound states in the form of subfundamental (SF)
dipoles have been found in the second FBG, in a part of which they
are stable (on the contrary to completely unstable SF solitons in
the 1D model with repulsion). As for square-shaped states in the
repulsion model, the quadrupoles are completely unstable, while
the vortices may be stable only inweakly anisotropic lattices. These
findings are summarized in Table 1.
A general inference is that rhombic patterns are always much
more robust than their square-shaped counterparts. As concerns
the role of the intrinsic structure, in the attraction model the
quadrupoles are more stable than vortices, while in the model with
repulsion the opposite is true.
Author's personal copy
1448 T. Mayteevarunyoo et al. / Physica D 238 (2009) 14391448
Fig. 12. Typical examples of stable 3D gap solitons in Eq. (3) with (a) isotropic ( = 1) and (b) anisotropic ( = 0.5) OL potential. The strength of the potential is V
0
= 6.
The plots display isosurfaces of the local density, |u (x, y, z)|
2
. Values of x and (y, z) are indicated in units of 1/ (24) and 1/ (6), respectively.
A challenging problem is to extend this analysis in a systematic
form to 3D solitons, especially ones with embedded vorticity. In
viewof the computational complexity of the 3D model, we did not
aim to carry out its systematic analysis in the framework of the
present work. Nevertheless, examples of stable GSs supported by
anisotropic OLs in three dimensions have been found, see Fig. 12.
These solutions were obtained by direct simulations of Eq. (3) in
real time, in a sufficiently large spatial domain with absorbers
installed at its borders. The initial configuration was an isotropic
Gaussian for instance, u
0
(x, y, z) = 3 exp
_

_
x
2
+y
2
+z
2
__
,
which was used to generate Fig. 12. In the course of the relaxation,
the GS typically loses 20% of the initial 3D norm, defined as
___
|u|
2
dxdydz.
Acknowledgements
We appreciate discussions with J. Yang. The work of T.M.
was supported, in a part, by a postdoctoral fellowship from the
PikovskyValazzi Foundation, by the Israel Science Foundation
through the Center-of-Excellence grant No. 8006/03, and also by
the Thailand Research Fund under grant No. MRG5080171. M.
S. acknowledges a partial financial support from MIUR through
the inter-university project PRIN-2005: Transport properties of
classical and quantum systems. B.B.B. and B.A.M. appreciate
hospitality of the Department of Physics at the University of
Salerno.
References
[1] K.E. Strecker, G.B. Partridge, A.G. Truscott, R.G. Hulet, Nature 417 (2002) 150;
L. Khaykovich, F. Schreck, G. Ferrari, T. Bourdel, J. Cubizolles, L.D. Carr, Y. Castin,
C. Salomon, Science 296 (2002) 1290;
S.L. Cornish, S.T. Thompson, C.E. Wieman, Phys. Rev. Lett. 96 (2006) 170401.
[2] K.E. Strecker, G.B. Partridge, A.G. Truscott, R.G. Hulet, New J. Phys. 5 (2003)
73.1.
[3] L. Berg, Phys. Rep. 303 (1998) 259.
[4] B.A. Malomed, D. Mihalache, F. Wise, L. Torner, J. Opt. B Quant. Semics. Opt. 7
(2005) R53;
B.B. Baizakov, B.A. Malomed, Mario Salerno, Phys. Rev. E 74 (2006) 066615.
[5] B.B. Baizakov, B.A. Malomed, M. Salerno, Europhys. Lett. 63 (2003) 642.
[6] J. Yang, Z.H. Musslimani, Opt. Lett. 28 (2003) 2094.
[7] B.B. Baizakov, B.A. Malomed, M. Salerno, Phys. Rev. A 70 (2004) 053613.
[8] B.B. Baizakov, B.A. Malomed, M. Salerno, in: F.Kh. Abdullaev, V.V. Kono-
top (Eds.), Nonlinear Waves: Classical Quantum Aspects, Kluwer
Academic Publishers, Dordrecht, 2004, pp. 6180. Also available at:
http://rsphysse.anu.edu.au/~asd124/Baizakov_2004_61_NonlinearWaves.pdf.
[9] J. Wang, J. Yang, Phys. Rev. A 77 (2007) 033834.
[10] D. Mihalache, D. Mazilu, F. Lederer, Y.V. Kartashov, L.-C. Crasovan, L. Torner,
Phys. Rev. E 70 (2004) 055603(R).
[11] H. Sakaguchi, B.A. Malomed, Europhys. Lett. 72 (2005) 698.
[12] A. Gubeskys, B.A. Malomed, Phys. Rev. A 76 (2007) 043623.
[13] R. Driben, B.A. Malomed, A. Gubeskys, J. Zyss, Phys. Rev. E 76 (2007) 066604.
[14] P.G. Kevrekidis, D.J. Frantzeskakis, R. Carretero-Gonzlez, B.A. Malomed,
A.R. Bishop, Phys. Rev. E 72 (2005) 046613.
[15] H. Leblond, B.A. Malomed, D. Mihalache, Phys. Rev. E 76 (2007) 026604.
[16] O. Zobay, S. Ptting, P. Meystre, E.M. Wright, Phys. Rev. A 59 (1999) 643;
G.L. Alfimov, V.V. Konotop, M. Salerno, Europhys. Lett. 58 (2002) 7.
[17] A. Trombettoni, A. Smerzi, Phys. Rev. Lett. 86 (2001) 2353;
F.Kh. Abdullaev, B.B. Baizakov, S.A. Darmanyan, V.V. Konotop, M. Salerno, Phys.
Rev. A 64 (2001) 043606.
[18] K.M. Hilligse, M.K. Oberthaler, K.-P. Marzlin, Phys. Rev. A 66 (2002) 063605;
D.E. Pelinovsky, A.A. Sukhorukov, Y.S. Kivshar, Phys. Rev. E 70 (2004) 036618.
[19] B. Eiermann, Th. Anker, M. Albiez, M. Taglieber, P. Treutlein, K.-P. Marzlin,
M.K. Oberthaler, Phys. Rev. Lett. 92 (2004) 230401.
[20] O. Morsch, M. Oberthaler, Rev. Mod. Phys. 78 (2006) 179215.
[21] B.B. Baizakov, V.V. Konotop, M. Salerno, J. Phys. B 35 (2002) 5105;
P.J.Y. Louis, E.A. Ostrovskaya, C.M. Savage, Y.S. Kivshar, Phys. Rev. A 67 (2003)
013602;
E.A. Ostrovskaya, Y.S. Kivshar, Opt. Exp. 12 (2004) 19; Phys. Rev. Lett. 93
(2004) 160405;
H. Sakaguchi, B.A. Malomed, J. Phys. B 37 (2004) 2225;
A. Gubeskys, B.A. Malomed, I.M. Merhasin, Phys. Rev. A 73 (2006) 023607.
[22] E.A. Ostrovskaya, Y.S. Kivshar, Phys. Rev. Lett. 93 (2004) 160405; Opt. Exp. 12
(2004) 19;
E.A. Ostrovskaya, T.J. Alexander, Y.S. Kivshar, Phys. Rev. 74 (2006) 023605.
[23] H. Sakaguchi, B.A. Malomed, J. Phys. B 37 (2004) 2225.
[24] B.B. Baizakov, B.A. Malomed, M. Salerno, Eur. Phys. J. D 38 (2006) 367.
[25] H. Sakaguchi, B.A. Malomed, Phys. Rev. E 74 (2006) 026601;
G. Burlak, A. Klimov, Phys. Lett. A 369 (2007) 510.
[26] G.L. Alfimov, P.G. Kevrekidis, V.V. Konotop, M. Salerno, Phys. Rev. E 66 (2002)
046608;
R. Carretero-Gonzlez, K. Promislow, Phys. Rev. A 66 (2002) 033610;
N.K. Efremidis, D.N. Christodoulides, Phys. Rev. A 67 (2003) 063608;
M.A. Porter, R. Carretero-Gonzlez, P.G. Kevrekidis, B.A. Malomed, Chaos 15
(2005) 015115.
[27] J.W. Fleischer, G. Bartal, O. Cohen, T. Schwartz, O. Manela, B. Freedman,
M. Segev, H. Buljan, N.K. Efremidis, Opt. Exp. 13 (2005) 1780;
A.S. Desyatnikov, N. Sagemerten, R. Fischer, B. Terhalle, D. Tr ger, D.N. Neshev,
A. Dreischuh, C. Denz, W. Krlikowski, Y.S. Kivshar, Opt. Exp. 14 (2006) 2851.
[28] J.W. Fleischer, M. Segev, N.K. Efremidis, D.N. Christodoulides, Nature 422
(2003) 147.
[29] D.N. Neshev, T.J. Alexander, E.A. Ostrovskaya, Y.S. Kivshar, H. Martin,
I. Makasyuk, Z.G. Chen, Phys. Rev. Lett. 92 (2004) 123903;
J.W. Fleischer, G. Bartal, O. Cohen, O. Manela, M. Segev, J. Hudock,
D.N. Christodoulides, Phys. Rev. Lett. 92 (2004) 123904.
[30] J. Yang, I. Makasyuk, P.G. Kevrekidis, H. Martin, B.A. Malomed,
D.J. Frantzeskakis, Z. Chen, Phys. Rev. Lett. 94 (2005) 113902.
[31] X. Wang, Z. Chen, P.G. Kevrekidis, Phys. Rev. Lett. 96 (2006) 083904.
[32] P. Xie, Z.-Q. Zhang, X. Zhang, Phys. Rev. E 67 (2003) 026607;
A. Ferrando, M. Zacars, P.F. de Cordoba, D. Binosi, J.A. Monsoriu, Opt. Exp. 11
(2003) 452.
[33] A. Ferrando, M. Zacars, P.F. de Cordoba, D. Binosi, J.A. Monsoriu, Opt. Exp. 12
(2004) 817.
[34] R. Fischer, D. Trger, D.N. Neshev, A.A. Sukhorukov, W. Krlikowski, C. Denz,
Y.S. Kivshar, Phys. Rev. Lett. 96 (2006) 023905.
[35] T.J. Alexander, A.A. Sukhorukov, Y.S. Kivshar, Phys. Rev. Lett. 93 (2004) 063901.
[36] L. Salasnich, A. Parola, L. Reatto, Phys. Rev. A 65 (2002) 043614.
[37] J.K. Yang, T.I. Lakoba, Stud. Appl. Math. 118 (2007) 153.
[38] B.B. Baizakov, M. Salerno, Phys. Rev. A 69 (2004) 013602.
[39] B.A. Malomed, Phys. Rev. E 58 (1998) 7928.
[40] T. Mayteevarunyoo, B.A. Malomed, Phys. Rev. A 74 (2006) 033616;
S.K. Adhikari, B.A. Malomed, Europhys. Lett. 79 (2007) 50003.
[41] Z. Shi, J. Yang, Phys. Rev. E 75 (2007) 056602.