PHYSICAL REVIEW E 83, 016604 (2011)

Localization and delocalization of two-dimensional discrete solitons pinned to linear

and nonlinear defects

Valeriy A. Brazhnyi1,* and Boris A. Malomed2,†

1
Centro de Fı́sica do Porto, Faculdade de Ciências, Universidade do Porto, R. Campo Alegre 687, P-4169-007 Porto, Portugal
2
Department of Physical Electronics, Faculty of Engineering, Tel Aviv University, Tel Aviv 69978, Israel
(Received 30 September 2010; published 31 January 2011)
We study the dynamics of two-dimensional (2D) localized modes in the nonlinear lattice described by the
discrete nonlinear Schrödinger equation, including a local linear or nonlinear defect. Discrete solitons pinned
to the defects are investigated by means of the numerical continuation from the anticontinuum limit and also
using the variational approximation, which features a good agreement for strongly localized modes. The models
with the time-modulated strengths of the linear or nonlinear defect are considered too. In that case, one can
temporarily shift the critical norm, below which localized 2D modes cannot exist, to a level above the norm of
the given soliton, which triggers the irreversible delocalization transition.

DOI: 10.1103/PhysRevE.83.016604 PACS number(s): 05.45.Yv, 03.75.Lm, 03.75.Kk

I. INTRODUCTION II. THE MODEL

The discrete nonlinear Schrödinger (DNLS) equations We consider the following model based on the 2D DNLS
constitute a universal class of models which are profoundly equation with a local defect:
interesting in their own right, as dynamical systems, and
serve as models of a number of physical systems in nonlinear i u̇n,m + J 2 un,m + Vn,m un,m + (σ + Wn,m )|un,m |2 un,m = 0,
optics [1], studies of Bose-Einstein condensates [2,3], and (1)
other contexts [4]. In particular, soliton solutions to the
where the overdot stands for the time derivative, 2 un,m ≡
DNLS equation in one, two, and three dimensions represent
un,m+1 + un,m−1 + un+1,m + un−1,m − 4un,m is the 2D dis-
fundamental dynamically localized modes in discrete media.
crete Laplacian, the coupling constant of the lattice will be
Experimentally, one- and two-dimensional (1D and 2D)
fixed by scaling J ≡ 1, unless another choice of J is specified
discrete solitons have been created in nonlinear optical systems
explicitly, and σ = +1 and −1 correspond to the attractive
of several types [1].
and repulsive nonlinearities, respectively (σ = 0 for the linear
An important ingredient of DNLS models is represented
lattice). Further, functions Vn,m and Wn,m , which account for
by local defects. They are interesting as additional elements
the linear and nonlinear components of the defect, are taken
of the lattices [4] and find important physical realizations.
as Gaussian profiles,
In particular, they may describe various strongly localized
structures in photonic crystals [5], including nanocavities [6], Vn,m = v exp[−(n2 + m2 )/v ],
as well as microresonators [7] and quantum dots [8]. (2)
The objective of the present work is to consider the Wn,m = w exp[−(n2 + m2 )/w ],
interaction of 2D discrete solitons with local linear and with respective strengths v,w and widths v ,w . In this
nonlinear defects in DNLS lattices. The defect may be notation, positive and negative strengths correspond to the
concentrated at a single site on the lattice or it may be shaped attractive and repulsive defect, respectively. In fact, we
as a Gaussian of a finite width. After introducing the model in consider the linear and nonlinear defects separately. Models
Sec. II, we consider stationary solitons pinned by the defects in of 2D nonlinear lattices with other types of local defects were
Sec. III. The analysis is performed by means of the variational considered earlier [11], including defects induced by edges of
approximation (VA) and numerical methods. In particular, the the lattice [12].
pinned-soliton families feature a specific combination of stable Looking for the stationary solutions,
and unstable parts. In Section IV, we consider possibilities of
the application of the “management” [9] to two-dimensional un,m = Un,m exp(−iωt), (3)
(2D) discrete solitons, using the linear and nonlinear defects
we arrive at the nonlinear eigenvalue problem,
whose amplitudes vary slowly in time. In this direction, we
analyze the possibility of triggering a delocalization transition ωUn,m + 2 Un,m + Vn,m Un,m
by means of this method, which may be used in the design
+ (σ + Wn,m )|Un,m |2 Un,m = 0, (4)
of switching schemes in photonics. Previously, the induced
transition to the delocalization was demonstrated in uniform for the real frequency ω and the profile of the stationary discrete
lattices with the intersite coupling strength subject to the time mode Um,n , which may be complex, in the general case. In the
modulation [10]. The paper is concluded in Sec. V. absence of the defect, Eq. (4) gives rise to the linear dispersion
relation featuring the phonon band, above and below which
one can find nonlinear
 modes, described by respective curves
*
[email protected] N (ω), with N = m,n |um,n |2 being the norm (power) of the
†
[email protected] nonlinear state. The important difference between 2D and 1D

1539-3755/2011/83(1)/016604(6) 016604-1 ©2011 American Physical Society

VALERIY A. BRAZHNYI AND BORIS A. MALOMED PHYSICAL REVIEW E 83, 016604 (2011)

settings is that, in the latter case, the fundamental single-peak e−2a

mode (the discrete soliton of the Sievers-Takeno type) persists Lcoupl = 8|A|2 (3e−a − e−3a − 2) (11)
(1 − e−2a )2
in the limit of N → 0, while all 2D solitons are bounded
by a critical norm, Ncr , below which localized modes do + 4(AB ∗ + A∗ B)e−a − 4|B|2 , (12)
w 4
not exist [4]. Accordingly, an abrupt delocalization (decay) Ldefect = v|B| + |B| .
2
(13)
of discrete 2D solitons was predicted in the case when the 2
intersite coupling constant exceeds a certain critical value [10]. The substitution of the harmonic time dependence, A(t) =
In the following we show that introducing a defect with linear Ae−iωt ,B(t) = Be−iωt , where A and B are real amplitudes,
and nonlinear components and varying their strengths, or the casts the Lagrangian into the following stationary form, with
lattice coupling constant, one can change the critical norm, C ≡ e−a :
Ncr , and thus control the transition to the delocalization. A2 C 2
Lkin = 4ω + ωB 2 , (14)
(1 − C 2 )2
III. STATIONARY DISCRETE SOLITONS PINNED TO THE
C4 σ
DEFECT: THE VARIATIONAL APPROACH AND Lself = 2σ A4 + B 4, (15)
NUMERICAL RESULTS (1 − C 4 )2 2
2
C
A. The variational approximation Lcoupl = 8A2 (3C − C 3 − 2) (16)
(1 − C 2 )2
The VA was successfully used for the study of 2D discrete
solitons in uniform (defect-free) DNLS lattices [13,14]. Here, + 8ABC − 4B 2 , (17)
w
we start with application of the VA to the 2D lattice in the Ldefect = vB 2 + B 4 . (18)
presence of the δ-defect localized at the origin: 2
Fixing the frequency ω and numerically solving the ensuing
i u̇m,n + 2 um,n + σ |um,n |2 um,n system of the Euler-Lagrange equations, which follows from
+ (v + w|um,n |2 )δm,0 δn,0 um,n = 0; (5) the effective Lagrangian, ∂L/∂A = ∂L/∂B = ∂L/∂C = 0,
one can find the set of variational parameters (A,B,a) and
compare the more general form of the defect in Eq. (2). the norm corresponding to ansatz (7):
Generally, the strength of the defect may be time dependent; 
v = v(t),w = w(t). N = B 2 + A2 (1 − δm,0 δn,0 )C 2(|m|+|n|) . (19)
Equation (5) can be derived from the Lagrangian, m,n

i  ∗ B. The linear defect

L= (u u̇m,n − um,n u̇∗m,n )
2 m,n m,n Curves N (ω) for the discrete solitons produced by the VA,
 along with their counterparts, obtained by dint of the direct
+ [u∗m,n (um+1,n + um,n+1 ) numerical solution based on the continuation from the anti-
m,n continuum limit (see Appendix A), are displayed in Fig. 1, for
+ um,n (u∗m+1,n + u∗m,n+1 ) − 4|um,n |2 ] the pure linear defect.
Figure 1(a) shows that, in the case of v = 0.1, which is
σ  1
+ |um,n |4 + v|u0,0 |2 + w|u0,0 |4 . (6) practically tantamount to the defect localized at the single site
2 m,n 2 [cf. Eq. (5)], the VA-generated norm is in good agreement
with the numerical results far from the limit of ω → 0. At
To apply the VA, we adopt the following ansatz (previously, it small ω, the soliton spreads out, and its actual shape deviates
was used for the analysis of discrete solitons in the 2D DNLS from the exponential ansatz (7), which leads to a discrepancy
equation with the cubic-quintic nonlinearity [14]): between the variational and the numerical results, as shown in
 Fig. 1(a). Also, due to restrictions on the size of the domain
B if m = n = 0;
um,n = −a(|m|+|n|)
(7) of calculation (we used 41 × 41 grid points), the numerical
Ae otherwise. curves cannot be continued to extremely small values of ω;
Substituting ansatz (7) into Lagrangian (6), one can therefore, they are not presented in Fig. 1(b).
perform the respective calculations and find the effective Naturally, the increase in the width of the defect to v = 1,
Lagrangian as a sum of four terms which depend on three which makes the shape of defect (2) essentially different from
dynamical parameters, (A,B,a) (which may be functions of that in Eq. (5), leads to a stronger discrepancy in the limit
time) and represent, respectively, the kinetic part, on-site ω → 0, as shown in Fig. 2.
self-interaction, intersite couplings, and effects induced by the We have checked the linear stability of the solutions along
defect: their existence curves (see details in Appendix B). The results
are shown by means of solid and dashed portions in the figures
L = Lkin + Lcoupl + Lint + Lself , (8) displaying the respective numerically generated N(ω) curves.
−2a Note that, here and in similar subsequent plots, the soliton
2ie i
Lkin = −2a
(A∗ Ȧ − AȦ∗ ) + (B ∗ Ḃ − B Ḃ ∗ ), (9) families may contain two distinct stability segments separated
(1 − e ) 2 2 by an instability region. It should be noted that, far from
−4a
e σ the limit of ω = 0, the Vakhitov-Kolokolov (VK) criterion,
Lself = 2σ |A|4 −4a
+ |B|4 , (10)
(1 − e ) 2 2 dN/dω < 0 [4], correctly predicts the transition from stability

016604-2
LOCALIZATION AND DELOCALIZATION OF TWO- . . . PHYSICAL REVIEW E 83, 016604 (2011)

(a) (b) 20 18
20 Δv =0.1 Δw=0.1 (a) Δw=0.1 (b)
20 N
N 16
N N

10 10 14
10

12

0 0 0 10
-2 -1.5 -1 -0.5 0 -0.04 -0.02 0
-2 -1.5 -1 -0.5 0 -0.03 -0.02 -0.01 0
ω ω
ω ω
FIG. 1. (Color online) (a) Comparison of the dependence of the
FIG. 3. (Color online) Same as Fig. 1, but for the pure nonlinear
norm on the frequency for the fundamental discrete solitons, as
defect (v = 0), at different values of its strength: w = −0.5, −0.2, 0,
predicted by the VA, with the use of Eq. (19) [dotted (blue) lines], and
and +0.5 (from top to bottom). The width of the defect corresponding
its counterpart, produced by the numerical solution of the stationary
to the numerical curves is w = 0.1.
version of Eq. (5) [solid (black) and dashed (red) lines], for three
strengths of the linear defect (w = 0): v = −1, 0, and +1 (from top
C. The nonlinear defect
to bottom; recall that v > 0 and v < 0 correspond to the attractive
and repulsive defects, respectively). Numerical results were obtained Similar results were obtained for the pure nonlinear defect,
for a narrow defect, with width v = 0.1 in Eq. (2). The nonlinearity that is, with v = 0 in Eqs. (2) and (5), as shown in Figs. 3 and 4.
is attractive (σ = +1). (b) Zoom-in on the plot in (a), corresponding In particular, analyzing the results for linear and nonlinear
to the VA prediction for the norm given by Eq. (19), in the vicinity defects of different widths, we have found that there is a
of ω = 0. In this and all other figures, solid and dashed portions of particular intermediate value of the width in the interval
the numerically generated curves depict subfamilies of stable and of v,w ⊂ (0.1,1), at which the discrepancy between the
unstable solitons, respectively. numerical and the variational solutions attains a minimum (in
particular, it is w ≈ 0.7 for the nonlinear defect; see Fig. 5).
For the sake of comparison, the VA-predicted and numerical
to instability at points where dN/dω changes its sign from curves N (ω) are also shown in Fig. 6 for discrete solitons in a
negative to positive. However, at small ω, the formally applied defect-free 2D lattice (of course, these results are not different
VK criterion only partially complies with the linear-stability from those reported in earlier works which were dealing with
results, which may be an effect of boundary conditions on the 2D DNLS equation without defects [4,13]). Comparing
properties of very broad modes corresponding to small ω. them to Figs. 1 and 3, we conclude that the discrepancy
It is noteworthy that, as one can conclude from comparison between the VA and numerical findings is actually smaller
of Figs. 1(a) and 2, the increase in width of the attractive linear in the presence of the linear or nonlinear defect.
defect leads to a significant reduction of the instability region. Finally, it is also interesting to consider the case of the linear
This trend is also corroborated by the VK criterion. lattice (σ = 0), with all the nonlinearity concentrated only in
the form of a narrow defect (2), with v = 0 and w > 0. In Fig. 7,

20
Δv = 1 Δ w =1
20
N
N

10
10

0 0
-2 -1.5 -1 -0.5 0 -2 -1.5 -1 -0.5 0
ω ω
FIG. 2. (Color online) Same as Fig. 1(a), but with the norm found FIG. 4. (Color online) Same as Fig. 3(a), but for defect width
numerically for the broad defect with width v = 1. w = 1.

016604-3
VALERIY A. BRAZHNYI AND BORIS A. MALOMED PHYSICAL REVIEW E 83, 016604 (2011)

20 14
Δw =0.7
18 12
N
16 N
10 σ=1, w=0
σ=0, w=1, Δ w =0.1
14
8
12
6
10
-2 -1.5 -1 -0.5 0 -2 -1.5 -1 -0.5 0
ω ω
FIG. 5. (Color online) Comparison between variational and FIG. 7. (Color online) Comparison of the numerically found
numerical results for the pure nonlinear defect with amplitude dependence N (ω) in a linear lattice (σ = 0), with the nonlinearity
w = −0.5 and width w = 0.7. In this case, the overall discrepancy represented solely by the narrow nonlinear defect [v = 0, w = 0.1;
between the VA-predicted and the numerically found curves [which the lower (red) line], with its counterpart in the usual defect-free
are defined as in Fig. 3(a)] attains its minimum. DNLS lattice [the upper black line].

the linear defect, v(t), which eventually returns to the initial

the corresponding dependence N (ω) is shown for w = 0.1, value, vi ≡ v(t = 0) (cf. Ref. [15]):
which implies that the nonlinearity is actually concentrated at
a single site, n = m = 0. v(t) = vf + (vi − vf )|1 − 2t/tf |. (20)
Here vf ≡ v(t = tf /2) corresponds to the turning point, t =
IV. LOCALIZATION-DELOCALIZATION TRANSITION tf /2, while the return time is t = tf . Note that the total norm
CONTROLLED BY TEMPORAL MODULATIONS OF remains a conserved quantity in the DNLS equation with the
LINEAR AND NONLINEAR DEFECTS time-dependent strength of the local defect.
A. Variation of the linear defect First, we consider the soliton for vi = 0, with norm Ns
(point A in Fig. 8), which precisely coincides with the
In this section, we consider effects of hysteresis for discrete minimum of the existence curve, but for v = −1 (point B).
solitons, induced by the adiabatic variation of the strength of This means that, by taking vi = 0 and vf = −1 in Eq. (20) at
t = tf /2, we may transfer the soliton from point A to point B.
As long as the soliton stays on the existence curve between
these two points, it remains localized, featuring some changes

15 12
N v=
-1 .
N v= 1
10 -1
v=
0
8v
=1
5 Ns
C A B

0
-2 -1.5 -1 -0.5 0 4
ω
-5 -4 -3 -2 -1 0
FIG. 6. (Color online) Same as Figs. 1(a) and 3(a), but for ω
fundamental discrete solitons in a 2D lattice without defects (v =
w = 0), at three values of the coupling constant in Eq. (1): J = 1.2, FIG. 8. (Color online) Transformation of a discrete soliton under
1, and 0.5 (from top to bottom). the action of the adiabatic variation of the strength of the linear defect.

016604-4
LOCALIZATION AND DELOCALIZATION OF TWO- . . . PHYSICAL REVIEW E 83, 016604 (2011)

8 8
C (a) C (a)
A A A A
6 6
B 2
|u 0,0(t)| B
|u0,0(t)|24 4

2 2

0 0
0 50 100 150 200 0 50 100 150 200
t t
1 (b) C (b)
C 0.2

v (t) A
A
w (t ) 0 A A
0
B
-0.2
B
-1
0 50 100 150 200 -0.4
0 50 100 150 200
t t

FIG. 9. (Color online) (a) Evolution of the density at the origin, FIG. 11. (Color online) Same as Fig. 9, but for the temporal
|u0,0 (t)|2 , corresponding to the time-modulation scenarios displayed modulation of a nonlinear defect.
in Fig. 8. The (dark) solid and (blue) dashed-dotted lines represent
scenarios A-B-A (with vf = −1) and A-C-A (with vf = 1), which B. Variation of the nonlinear defect
are shown in (b) and in Fig. 8. The (red) dashed line corresponds to
Here we aim to consider effects produced by the adiabatic
the case where the strength v(t) temporarily falls to a value below the
critical value, vf = −1.1 < vcr , which leads to delocalization. At the
variation of the strength of the nonlinear defect, following the
right, initial (top) and final density profiles, with vf = −1 (middle) same scenario as in Eq. (20):
and vf = −1.1 (bottom), are displayed.
w(t) = wf + (wi − wf )|1 − 2t/tf |, (21)
with wi = w(t = 0) and wf = w(t = tf /2) corresponding to
in the profile; see details of dynamics in Fig. 9 (this is also valid the initial strength and its value at the turning point, t = tf /2,
for the dynamics between point A and point C). However, if, with the return to the initial value at t = tf . As in the previous
in the framework of the same scenario, the amplitude of the case, the norm of the solution remains conserved in the course
defect falls below the critical value, vf < vcr , which in the of the evolution.
present case is vcr = −1, then the minimum of the existence Again, we start with the solution at wi = 0 (point A in
curve turns out to be higher than the norm of the evolving Fig. 10), with a norm which coincides with the minimum
mode, which is expected to trigger the delocalization. of the existence curve for w = −0.2 (point B). This means
These assumptions were verified by means of direct that, by taking wi = 0 and wf = −0.2 in Eq. (21) at time
integration of Eq. (1), with the strength of the linear defect t = tf /2, we will transfer the soliton from point A to point B
v varying in time according to Eq. (20). The result is presented (see Fig. 10). If the soliton stays on the existence curve between
in Fig. 9, where the evolution of the density at the origin, these two points, the solution always remains localized (see the
|u0,0 (t)|2 , is shown for three values of vf , along with the dynamical picture in Fig. 11). However, if, for the same initial
corresponding initial and final profiles. condition, the strength of the nonlinear defect is allowed to
drop below the critical value, wf < wcr , which, in the present
case, is wcr = −0.2, then the minimum of the existence curve
15 becomes higher than the norm of the solution, which, again, is
expected to trigger the transition to the delocalization.
w=
N -0.
3
These assumptions were verified through the direct inte-
w= gration of Eq. (1) with the strength of the nonlinear defect, w,
-0.2 varying in time according to Eq. (21). The results are presented
10 in Fig. 11, where the evolution of the density at the origin,
w=
0 |u0,0 (t)|2 , is shown for three values of wf .

Ns V. CONCLUSION
C A B
w= 0 In this work, we have considered the static and dynamical
.2
5 properties of 2D discrete solitons in a nonlinear lattice
described by the DNLS equation, which includes a local
linear or nonlinear defect. The solitons trapped around the
-5 -4 -3 -2 -1 0 defects were investigated using both the VA (variational
ω approximation) and numerical methods, the VA showing
good agreement with the numerical findings for sufficiently
FIG. 10. (Color online) Soliton transformation under the action narrow solitons. Then the analysis was extended to the
of the adiabatic variation of the strength of the nonlinear defect. model with linear and nonlinear defects whose strength was

016604-5
VALERIY A. BRAZHNYI AND BORIS A. MALOMED PHYSICAL REVIEW E 83, 016604 (2011)

subject to the slow variation in time. In the latter case, one needed [to proceed from the narrow defect, (5), to the broader
of the possibilities is controlled onset of the transition to one, (2)]. After that, by changing the frequency ω of the
delocalization. solution, we obtain the dependence of norm N on ω and, also,
The analysis reported in this work can be extended by analyze the linear stability of the corresponding solutions; see
considering combinations of linear and nonlinear defects, Appendix B.
searching for symmetric, antisymmetric, and asymmetric
modes trapped by symmetric pairs of defects, and analyzing
other “management” schemes [16–18], with the defect strength APPENDIX B
subject to a time-periodic modulation. To analyze the stability of solitons, we consider perturbed
solutions,
ACKNOWLEDGMENTS um,n (t) = [Um,n + am,n (t)]e−iωt . (B1)
V.A.B. acknowledges support from FCT Grant No. PTDC/ Representing the small perturbations as am,n (t) = (αm,n +
FIS/64647/2006. B.A.M. appreciates the hospitality of Centro iβm,n )eλt , substituting this into Eq. (1), and separating real
de Fı́sica do Porto (Porto, Portugal). and imaginary parts, we derive the system
    
α 0 L1 α
APPENDIX A λ = , (B2)
β −L3 0 β
Here we discuss some details on the continuation of
solutions from the anti-continuum limit, which corresponds where λ is the stability eigenvalue, and operators L1 and L3
to the uncoupled lattice described by Eqs. (1) and (5) with are
J = 0 [4]. Looking for stationary modes in the form of L1 = 2 + Vm,n − ω + (σ + Wm,n ) (Um,n )2 , (B3)
Eq. (3), in the anticontinuum limit one obtains an obvi-
L3 = 2 + Vm,n − ω + 3(σ + Wm,n ) (Um,n )2 . (B4)
ous exact solution for the one-site fundamental discrete
soliton: Multiplying the first equation in system (B2) by −λ and
 then using the second equation, one arrives at the eigenvalue
±A if m = n = 0;
Um,n = (A1) problem,
0 otherwise.
√ α = L1 L3 α, (B5)
In Eqs. (A1), A ≡ −(ω + v)/(σ + w) (see Ref. [19] and
references therein). Then, using the standard Newton-Raphson where = −λ . The underlying stationary solution, Um,n , is
2

method, we gradually increase the coupling constant from unstable if the spectrum of includes complex or negative
J = 0 to J = 1, also increasing the width of the defect if real values.

[1] F. Lederer, G. I. Stegeman, D. N. Christodoulides, G. Assanto, [9] B. A. Malomed, Soliton Management in Periodic Systems
M. Segev, and Y. Silberberg, Phys. Rep. 463, 1 (2008). (Springer, New York, 2006).
[2] A. Trombettoni and A. Smerzi, Phys. Rev. Lett. 86, 2353 (2001); [10] G. Kalosakas, K. Ø. Rasmussen, and A. R. Bishop, Phys. Rev.
F. Kh. Abdullaev, B. B. Baizakov, S. A. Darmanyan, V. V. Lett. 89, 030402 (2002).
Konotop, and M. Salerno, Phys. Rev. A 64, 043606 (2001); [11] P. L. Christiansen, Yu. B. Gaididei, K. Ø. Rasmussen, V. K.
G. L. Alfimov, P. G. Kevrekidis, V. V. Konotop, and M. Salerno, Mezentsev, and J. Juul Rasmussen, Phys. Rev. B 54, 900 (1996).
Phys. Rev. E 66, 046608 (2002). [12] M. I. Molina and Y. S. Kivshar, Phys. Rev. A 80, 063812 (2009).
[3] O. Morsch and M. Oberthaler, Rev. Mod. Phys. 78, 179 (2006). [13] M. I. Weinstein, Nonlinearity 12, 673 (1999); J.-K. Xue, A.-X.
[4] P. G. Kevrekidis, ed., The Discrete Nonlinear Schrödinger Zhang, and J. Liu, Phys. Rev. A 77, 013602 (2008).
Equation: Mathematical Analysis, Numerical Computations, [14] C. Chong, R. Carretero-González, B. A. Malomed, and P. G.
and Physical Perspectives (Springer, Berlin, 2009). Kevrekidis, Physica D 238, 126 (2009).
[5] T. Hattori, N. Tsurumachi, and H. Nakatsuka, J. Opt. Soc. Am. [15] Yu. V. Bludov, V. A. Brazhnyi, and V. V. Konotop, Phys. Rev. A
B 14, 348 (1997). 76, 023603 (2007).
[6] Y. Akahane, T. Asano, B. S. Song, and S. Noda, Nature 425, 944 [16] Z. Chen, H. Martin, E. D. Eugenieva, J. Xu, and J. Yang, Opt.
(2003). Express 13, 1816 (2005); D. N. Christodoulides and R. I. Joseph,
[7] R. Colombelli, K. Srinivasan, M. Troccoli, O. Painter, C. F. Opt. Lett. 13, 794 (1988).
Gmachl, D. M. Tennant, A. M. Sergent, D. L. Sivco, A. Y. Cho, [17] T. Mayteevarunyoo, B. A. Malomed, and M. Krairiksh, Phys.
and F. Capasso, Science 302, 1374 (2003). Rev. A 76, 053612 (2007).
[8] H. Nakamura, Y. Sugimoto, K. Kanamoto, N. Ikeda, Y. Tanaka, [18] G. Burlak and B. A. Malomed, Phys. Rev. 77, 053606 (2008).
Y. Nakamura, S. Ohkouchi, Y. Watanabe, K. Inoue, H. Ishikawa, [19] G. L. Alfimov, V. A. Brazhnyi, and V. V. Konotop, Physica D
and K. Asakawa, Opt. Express 12, 6606 (2004). 194, 127 (2004).

016604-6