Anda di halaman 1dari 44

I NVI TED PAPER

Bright and dark spatial solitons


in non-Kerr media
YU. S. KI VSHAR
Australian Photonics Cooperative Research Centre, Optical Sciences Centre,
Research School of Physical Sciences and Engineering, The Australian
National University, Canberra ACT 2000, Australia
An overview of the theory of self-guided optical beams, spatial optical solitons supported
by non-Kerr non-linearities, is presented. This includes bright and dark solitons in optical
media with intensity-dependent non-linear response as well as two-component solitary
waves supported by parametric wave mixing in quadratic or cubic media. The properties
of non-linear spatially localized waves are discussed for qualitatively different types of
soliton bearing non-integrable non-linear models, including the scalar model described
by a generalized non-linear Schrodinger equation and the models of the second- and
third-harmonic generation. Special attention is paid to the recent advances of the theory
of soliton stability and soliton internal modes.
1. Introduction
Recent years have shown increased interest from dierent experimental and theoretical
groups in the study of self-guided optical beams that propagate in slab waveguides or bulk
non-linear media without supporting waveguide structures. Such beams are commonly
referred to by physicists as spatial optical solitons even though they do not possess all
properties of the formal mathematical denition of solitons, spatially localized non-linear
waves of integrable non-linear models. The property of integrability is valid only for the
so-called Kerr solitons, the (1 1)-dimensional beams of cubic non-linearity described in
the paraxial approximation.
Simple physics explains the existence of spatial optical solitons in a generalized self-
focusing non-linear medium. First, we recall the physics of optical waveguides (see, for
example [1] and references therein). Optical beams have an innate tendency to spread
(diract) as they propagate in a homogeneous medium. However, the beam's diraction
can be compensated for by beam refraction if the refractive index is increased in the region
of the beam. An optical waveguide is an important means to provide a balance between
diffraction and refraction if the medium is uniform in the direction of propagation. The
corresponding propagation of the light is conned in the transverse direction of the
waveguide, and it is described by the so-called linear guided modes, spatially localized
eigenmodes of the electric eld in the waveguide.
Optical and Quantum Electronics 30 (1998) 571614
03068919 1998 Chapman & Hall 571
As was discovered a long time ago [2], a similar eect, i.e. suppression of diraction
through a local change of the refractive index, can be produced solely by non-linearity. As
has been well established in many experiments (for the most recent overview of experi-
mental observations of dierent types of spatial optical solitons see [3]), some materials
can display considerable optical non-linearities when their properties are modied by the
light propagation. In particular, if the non-linearity leads to a change of the refractive
index of the medium in such a way that it generates an eective positive lens to the beam,
the beam can become self-trapped and propagates unchanged without any external
waveguiding structure [2]. Such stationary self-guided beams are called these days spatial
optical solitons, they exist with proles of a certain form allowing a local compensation of
the beam diffraction by the non-linearity-induced change in the material refractive index.
Until recently, the theory of spatial optical solitons has been based on the non-linear
Schro dinger (NLS) equation with a cubic non-linearity, which is exactly integrable by
means of the inverse scattering (IST) technique [4]. Generally speaking, the integrability
means that any localized input beam will be decomposed into stable solitary waves (or
solitons) and radiation, and also that interaction of solitons is elastic. From the physical
point of view, the integrable NLS equation describes the (1 1)-dimensional (i.e. one
transverse and one longitudinal dimensions) beams in a Kerr (cubic) non-linear medium in
the framework of the so-called paraxial approximation. The cubic NLS equation is known
to be a good model for temporal optical solitons propagating enormous distances along
existing waveguides, optical bres. In application to spatial optical solitons, the cubic NLS
equation becomes an inappropriate model. First, for spatial optical solitons much higher
input powers are required to compensate for diffraction, meaning that the refractive index
experiences large deviations from a linear (Kerr) dependence. Second, as was recognized a
long time ago (see, for example [5]), radially symmetric stationary localized solutions of
the (2 1)-dimensional NLS equation are unstable and may display collapse (for a
comprehensive overview of the wave collapse phenomenon see [6]). To deal with realistic
optical models, saturation had been suggested as a way to stabilize such a catastrophic
self-focusing and produce stable solitary waves of higher dimensions (see, for example [7,
8]), the effect also discussed in some other applications [9, 10]. Accounting for this effect
immediately leads to non-integrable models of generalized non-linearities, not possessing the
properties of integrability and elastic soliton collisions. Another mechanism of non-Kerr
non-linearities and enhanced non-linear properties of optical materials is a resonant,
phase-matched interaction between the modes of different frequencies. In this latter case,
multi-component solitary waves are created, and the mutual beam coupling can modify
drastically the properties of single beams, as it occurs in the case of the so-called quadratic
solitons of cascaded non-linearities (see, for example, the recent excellent review paper on
cascaded effects in v
(2)
materials [11]).
In spite of the fact that, generally speaking, there exist no universal analytical tools for
analysing solitary waves and their interactions in non-integrable models, recent advances
of the theory suggest that many of the properties of optical solitons in non-Kerr media are
similar, and they can be approached with the help of rather general physical concepts. One
of the recent advances in this eld was the developments of the unifying conceptual
approach (see [12] and references therein) based on the notion that a self-guided beam
induces a waveguide. This allows one to employ the theory of linear guided waves sup-
ported by prescribed waveguiding structures to understand, at least qualitatively, the
properties and physics of self-guided waves existing without supporting waveguides.
572
YU. S. Kivshar
The observation that a beam creates a waveguide and guides itself in it is, as a matter of
fact, not new, this concept is known from the very rst days of non-linear optics [13], and
it was also clearly stated in some earlier papers on the light self-focusing [2, 14]. The same
concept, in a dierent context, has been rediscovered often since that time (see, for
example [15] as a typical example), and it was also realized recently that this concept alone
could be useful to make novel predictions, extending our knowledge of self-guided beams
in non-Kerr media [12].
Some of the recent advances in the theory of self-guided beams based on the physics of
the induced waveguides suggest that the behaviour of beams in non-Kerr materials are
qualitatively similar [16]. This has led to a number of predictions about spatial solitons
and their dynamical evolution which do not follow from the analysis based solely on the
(1 1)-dimensional integrable cubic NLS equation. From this perspective we understand
that there is no simple mapping between temporal and spatial optical solitons. Spatial
solitons are a much richer and more complicated phenomenon, and this has already been
demonstrated by a number of elegant experiments in this eld.
In particular, it has recently been demonstrated experimentally, that self-guided beams
can be observed in materials with strong photorefractive and photovoltaic eect [1720],
in vapours with a strong saturation of the intensity-dependent refractive index [21, 22],
and also as a result of the mutual self-focusing due to the phase-matched three-wave
mixing in quadratic (or v
(2)
) non-linear crystals [23]. In all these cases, propagation of self-
guided waves is observed in non-Kerr materials which are described by the models more
general than the cubic NLS equation.
The main purpose of this paper is to discuss the most recent advances in the theory of
self-guided beams in non-integrable models of non-Kerr non-linearities. Because a com-
prehensive review of Kerr spatial solitons is available in the same issue [24], here we
concentrate mostly on the general properties of solitary waves of non-Kerr (e.g. saturable
or v
(2)
) non-linearities described by non-integrable non-linear models. We also discuss the
different physical mechanisms leading to such non-Kerr solitary waves, including the
effect of parametric interaction between phase-matched beams of different frequencies.
2. Models of non-Kerr spatial solitons
2.1. Why temporal and spatial solitons are different
Because the phenomenon of the long-distance propagation of temporal optical solitons in
optical bres is known to a much broader community of researchers in optics and non-
linear physics, rst we emphasize the dierence between spatial and temporal solitons.
Indeed, stationary beam propagation in planar waveguides has been considered somewhat
similar to the pulse propagation in bres. This is a direct manifestation of the so-called
spatio-temporal analogy in wave propagation [25, 26], meaning that the propagation co-
ordinate : is treated as the evolution variable and the spatial beam prole along the
transverse direction, for the case of waveguides, is similar to the temporal pulse prole, for
the case of bres. This analogy has been employed for many years, and it is based on a
simple notion that both beam evolution and pulse propagation can be described by the
cubic NLS equation (see, for example, [27, 28]). However, contrary to this widely accepted
opinion, we point out a crucial difference between these two phenomena. Indeed, in the
case of the non-stationary pulse propagation in bres, the operation wavelength is usually
selected near the zero of the group-velocity dispersion. This means that the absolute value
of the bre dispersion is small enough to be compensated by a week non-linearity such as
Bright and dark spatial solitons
573
that produced by the (very weak) Kerr effect in optical bres which leads to a relative non-
linearity-induced change in the refractive index very small. Therefore, non-linearity in such
systems is always weak and it should be well modelled by the same form of the cubic NLS
equation, which is known to be integrable be means of the IST technique [4, 29, 30].
However, for very short (fs) pulses the cubic NLS equation describing the long-distance
propagation of pulses should be corrected to include some additional (but still small)
effects such as higher-order dispersion, Raman scattering, etc. (for example, [31, 32]). All
these corrections can be taken into account by a perturbation theory (for an overview of
the perturbation theory of solitons see [33]). Thus, in bres non-linear effects are very
weak and they become important only when dispersion is small (near the zero-dispersion
point) affecting the pulse propagation over large distances (of order of hundreds of metres
or even kilometres).
In contrary to pulse propagation in optical bres, the physics underlying stationary
beam propagation in planar waveguides and bulk media is dierent. In this case the non-
linear change in the refractive index should compensate for the beam spreading caused by
diraction which is not a small eect. That is why to observe spatial solitons, much larger
non-linearities are usually required, and very often such non-linearities are not of the Kerr
type (e.g. they saturate at higher intensities). This leads to the models of generalized non-
linearities with the properties of solitary waves different from those described by the
integrable cubic NLS equation. For example, unlike the solitons of the cubic NLS equa-
tion, solitary waves of generalized non-linearities may be unstable, they also show some
interesting features, such as fusion due to collision, inelastic interactions and spiralling in a
bulk, wobbling, amplitude oscillation, etc. Propagation distances usually involved in the
phenomenon of beam self-focusing and spatial soliton propagation are of the order of
millimetres or centimetres. As a result, the physics of spatial solitary waves is very rich and
it should be understood in the framework of the theory of non-integrable models.
It is worth noting that such a distinction between spatial and temporal solitary waves is
less clear for the so-called gap solitons, non-linear localized modes exciting in a frequency
gap produced by the Bragg reection of light on periodic structures such as bre gratings
[34] or multilayered dielectric mirrors such as Bragg reectors [35]). Because of a relatively
large dispersion induced by the periodic variation of the refractive index, the soliton period
in bre gratings is about 110 cm, requiring much larger input powers (see, for example,
[34] for an overview of the most recent results), and the propagation distances become
comparable with those for spatial solitons. That is why, an experimental realization of the
concept of light bullets [36] unifying the spatial and temporal self-trapping, would require
materials with large dispersion that can be achieved in the periodic structures.
2.2. Basic equations
To describe spatial optical solitons in the framework of the simplest scalar model with
non-resonant non-linearities, we consider the propagation of a monochromatic scalar
electric eld E in a bulk optical medium with an intensity-dependent refractive index,
n = n
0
n
nl
(I), where n
0
is the linear refractive index, and n
nl
(I) describes the variation in
the index due to the eld with the intensity I = [E[
2
. The function n
nl
(I) is assumed to be
dependent only on the light intensity only, and it may be introduced phenomenologically.
Solutions of the governing Maxwell's equation can be presented in the form
E(
~
R
l
Y ZY t) = E(
~
R
l
Y Z) exp(ib
0
Z ixt) c.c. (1)
574
YU. S. Kivshar
where c.c. denotes complex conjugate, x is the source frequency, and b
0
= k
0
n
0
= 2pn
0
ak
is the plane-wave propagation constant for the uniform background medium, in terms of
the source wavelength k = 2pcaxY c being the free-space speed of light. Usually, the spatial
solitons are discussed for two geometries. For the beam propagation in a bulk, we assume
a (2 1)-dimensional model, so that the Z-axis is parallel to the direction of propagation,
and the X- and Y -axes are two transverse directions. For the beam propagation in a planar
waveguide, the eective eld is found by integrating Maxwell's equations over the
transverse structure dened by the waveguide connement, and therefore the model
becomes (1 1)-dimensional.
The function E(
~
R
l
Y Z) describes the wave envelope which in the absence of non-linear
and diraction eects E would be a constant. If we substitute Equation 1 into the two-
dimensional, scalar wave equation, we obtain the generalized non-linear parabolic equa-
tion,
2ik
0
n
0
dE
dZ

d
2
E
dX
2

d
2
E
dY
2
_ _
2n
0
k
2
0
n
nl
(I)E = 0 (2)
In dimensionless variables, Equation 2 becomes the well-known generalized NLS equation,
where local non-linearity is introduced by the function n
nl
(I).
For the case of the Kerr (or cubic) non-linearity we have n
nl
(I) = n
2
IY n
2
being the
coecient of the Kerr eect of an optical material. Now, introducing the dimensionless
variables, i.e. measuring the eld amplitude in the units of k
0
(n
0
[n
2
[)
1a2
and the propa-
gation distance in the units of k
0
n
0
, we obtain the (2 1)-dimensional NLS equation in the
standard form,
i
dW
d:

1
2
d
2
W
dx
2

d
2
W
dy
2
_ _
[W[
2
W = 0 (3)
where the complex W stands for a dimensionless envelope, and the sign () is dened by
the type of non-linearity, self-defocusing (`minus', for n
2
` 0) or a self-focusing (`plus', for
n
2
b 0).
For propagation in a slab waveguide, the eld structure in one of the directions, say Y , is
dened by the linear guided mode of the waveguide. Then, the solution of the governing
Maxwell's equation has the structure
E(
~
R
l
Y ZY t) = E(XY Z)A
n
(Y ) exp( ib
(0)
n
: ixt) (4)
where the function A
n
(Y ) describes the corresponding fundamental mode of the slab
waveguide. Similarly, substituting this ansatz into Maxwell's equations and averaging over
Y , we come again to the renormalized equation of the form 3 with the Y -derivative
omitted, which in the dimensionless form becomes the standard cubic NLS equation
i
dW
d:

1
2
d
2
W
dx
2
[W[
2
W = 0 (5)
Equation 5 coincides formally with the equation for the pulse propagation in dispersive
non-linear optical bres, and it is known to be integrable by means of the inverse scat-
tering transform (IST) technique.
For the case of non-linearities more general than the cubic one, we should consider the
generalized NLS equation,
575
Bright and dark spatial solitons
i
dW
d:

d
2
W
dx
2
) ([W[
2
)W = 0 (6)
where the function ) (I) describes a non-linearity-induced change of the refractive index,
usually ) (I) I for small I.
2.3. Nonresonant optical non-linearities
The generalized NLS equation 2 (or Equation 6) has been considered in many papers for
analysing beam self-focusing and the properties of spatial bright and dark solitons (see, for
example, [3755]. All types of non-Kerr non-linearities discussed in relation to the exis-
tence of solitary waves in non-linear optics can be divided, generally speaking, into three
main classes: (i) competing non-linearities, e.g. focusing (defocusing) cubic and defocusing
(focusing) quintic non-linearity (see, for example, [39, 40, 43, 44, 50]) and also general-
ization to a power non-linearity (for example, [5355]); (ii) saturable non-linearities (see,
for example, [4749] and also [51, 52]), and (iii) transiting non-linearities (see, for example,
[3941, 46]).
Usually, the non-linear refractive index of an optical material deviates from the linear
(Kerr) dependence for larger light intensities. Non-ideality of the non-linear optical
response is known for semiconductor (for example, AlGaAs, CdS, CdS
1x
Se
x
) waveguides
and semiconductor-doped glasses (see, for example, [5658]). Larger deviation from the
Kerr non-linearity is observed for non-linear polymers. For example, recently the mea-
surements of a large non-resonant non-linearity in single crystal PTS (p-toluene sulpho-
nate) at 1600 nm [59, 60] revealed a variation of the non-linear refractive index with the
input intensity which can be modelled by competing, cubic-quintic non-linearities,
n
nl
(I) = n
2
I n
3
I
2
(7)
This model describes a competition between self-focusing (n
2
b 0), at smaller intensities,
and self-defocusing (n
3
` 0), at larger intensities. Similar models are usually employed to
describe the stabilization of wave collapse in the (2 1)-dimensional NLS equation (for
example, [61] and references therein).
In a more general case, the models with competing non-linearities can be described by a
power-law dependence on the beam intensity,
n
nl
(I) = n
p
I
p
n
2p
I
2p
(8)
where p is a positive constant and usually n
p
n
2p
` 0.
Models with saturable non-linearities are the most typical ones in non-linear optics. For
higher powers, saturation of the non-linearity has been measured in many materials and
consequently the maximum refractive index change has been reported (see, for example,
[62]). We do not linger on the physical mechanisms behind the saturation but merely note
that it exists in many non-linear media being described by phenomenological models
introduced more than 25 years ago (see, for example, [7, 8, 63, 64]). The effective gener-
alized NLS equation with saturable non-linearity is also the basic model to describe the
recently discovered (1 1)-dimensional photovoltaic spatial solitons in photovoltaic
photorefractive materials such as LiNbO
3
(see [51]). Unlike the phenomenological models
usually used to describe saturation of non-linearity, for the case of photovoltaic solitons
this model can be justied rigorously [51, 52].
576
YU. S. Kivshar
Several models are usually considered to describe saturating non-linearities. From a
general point of view, the function n
nl
(I) describing the saturating non-linearity should be
characterized by three independent parameters: the saturation intensity, I
sat
, the maximum
change in the refractive index, n

, and the Kerr coefcient n


2
which appears for small I. In
particular, the phenomenological model
n
nl
(I) = n

1
1
(1 IaI
sat
)
p
_ _
(9)
satises these criteria, provided n
2
= n

paI
sat
. In the particular case p = 1, model 9
reduces to the well-known expression derived from the two-level model, which is used very
frequently. For the case p = 2, model 9 possesses localized solutions for bright and dark
solitons in an explicit analytical form [48, 49].
Finally, bistable solitons introduced by Kaplan [39, 40] usually require a special type of
the intensity-dependent refractive index which changes from one type to another one, e.g.
it varies from one kind of the Kerr non-linearity, for small intensities, to another kind
with a different value of n
2
, for larger intensities. This type of non-linearity is known to
support bistable dark solitons [41, 42] as well. One of the simplest models of such
transiting non-linearities describes a change from one type of the Kerr dependence to the
other one, i.e.
n
nl
(I) =
n
(1)
2
I I ` I
cr
n
(2)
2
I I b I
cr
_
(10)
A smooth transition of this kind can be modelled by the function [41]
n
nl
(I) = n
2
I1 a tanh[c(I
2
I
2
cr
)[ (11)
where for I I
cr
Y n
nl
(I) n
(1)
2
I, where n
(1)
2
= n
2
[1 a tanh
2
(cI
2
cr
)[, and for I I
cr
,
n
nl
(I) n
(2)
2
I, where n
(2)
2
= (1 a). Unfortunately, examples of non-linear optical mate-
rials with such dependencies are not yet known, but the bistable solitons possess attractive
properties useful for their possible futuristic applications in all-optical logic and switching
devices.
At last, we would like to mention the model of logarithmic non-linearity,
n
2
(I) = n
2
0
ln(IaI
0
), that allows close-form exact expressions not only for stationary
Gaussian beams (or Gaussons, as they were introduced in [65]), but also for periodic and
quasi-periodic regimes of the beam evolution [66]. The main features of this model are the
following: (i) the stationary solutions do not depend on the maximum intensity (quasi-
linearization) and (ii) radiation from the periodic solitons is absent (the linearized problem
has a discrete spectrum only). Such unusual properties persist in any dimension [65, 66].
Similar localized solutions exist in a model of highly non-local media [67] allowing a
simple analysis based on a linear equation for a quantum harmonic oscillator.
2.4. Validity of the NLS models
As we have discussed above, the scalar generalized NLS equation is a rather universal
model for spatial optical solitons. It is derived from the rst principles on the basis of very
general assumptions about non-linear properties of an optical medium. However, the NLS
model may fail in a number of cases (see also [68] for examples from other elds), and
therefore one should be aware of the validity limits of this simple model, introducing
577
Bright and dark spatial solitons
instead more general and more appropriate models for describing self-guided beams. Here
we discuss two such generalizations.
First of all, as standard derivation of the NLS equation is based on the so-called multi-
scale asymptotic technique, sometimes called the reductive perturbation method (for
example, [69, 70]). It assumes non-resonant non-linearities when the most important
effects are described by an envelope of the eld of the fundamental frequency x prop-
agating with the carrier wave number k. All higher-order harmonics, even if they have
been excited a priori, are assumed to be very small and, therefore, they do not modify the
eld evolution of the main frequency which, in the case of the cubic non-linearity, is
described by the NLS equation. However, when some multiple frequencies are generated,
they may strongly affect the wave propagation at the fundamental harmonic provided
the so-called matching conditions are satised. For example, strong interaction between
the main frequency x and two other frequencies x
1
and x
2
occurs provided x = x
1
x
2
and the phase mismatch Dk = k (k
1
k
2
) vanishes. This kind of three-wave mixing is
possible in a medium where the lowest-order non-linearity is quadratic. When the
medium non-linearity is cubic, wave coupling is possible in the form of a four-wave
mixing process. When any such resonance condition is satised, the envelope of the
fundamental eld becomes strongly coupled to a secondary eld (or more than one eld)
and the single NLS equation is no longer valid. Instead, a coupling between the modes
may support multi-component solitary waves which differ drastically from the conven-
tional solitons of the scalar NLS equation. Below we consider two such examples, the
two-wave mixing solitons due to second-harmonic interaction (quadratic media) and
third-harmonic interaction (cubic media). Importantly, the intermode interaction pro-
vides an efcient mechanism for non-Kerr non-linearities.
The second class of problems when the NLS model should be generalized is closely
related to spatial optical solitons described by non-Kerr non-linearities. Indeed, it is well
known that the NLS equation with non-linearity stronger than cubic, e.g. a power-law
focusing non-linearity [u[
2q
u, has localized solutions which blow-up, so that a singularity
appears at nite : (see [6] for a review). This phenomenon occurs for negative values of the
system Hamiltonian under the condition qD _ 2, where q is the power of the non-linearity
and D stands for the (D 1)-dimensional model (see, for example, [71]). Blow-up
(or collapse) in nite : means that the NLS model of this dimension fails as an envelope
equation since it breaks the scales on which it was derived in the framework of the multi-
scale asymptotic technique. For spatial solitons this condition means that if D = 2, then
the cubic non-linearity [u[
2
u is already sufcient to induce collapse. If D = 1, then one
needs the quintic (or higher-order) non-linearity to induce collapse. Blow-up indicates also
that the primary NLS model should be corrected, e.g. by taking into account the effects of
non-paraxiality in the beam self-focusing [72, 73]. Because the catastrophic beam self-
focusing is also corrected by any kind of saturable non-linearity, we do not discuss in
detail the phenomenon of the wave collapse. Also, all solitons of a quadratic medium do
not display collapse.
3. Existence of solitary waves
3.1. Stationary soliton solutions
From the mathematical point of view, non-linear self-guided waves are described by
stationary localized solutions of the corresponding partial dierential equations. Scalar
bright solitons have the form [2, 4, 16]
578
YU. S. Kivshar
E(~rY :) = E(~rY b) exp(ib:) (12)
where ~r = (xY y) and E(~rY b) is real, and they depend on the propagation constant b.
Propagation constant b denes both the shape and stability of the stationary solutions.
The problem for the stationary proles E(~rY b) does not involves : and it reduces to a
simple eigenvalue equation described by one or several coupled ordinary dierential
equations (ODEs) in the case of soliton propagation in waveguides, for the (1 1)-
dimensional geometry, and in the case of the beams of circular symmetry, for the (2 1)-
dimensional geometry. Proles of bright solitary waves, i.e. those with vanishing
asymptotics, can be found as separatrix trajectories which start from and return to a
critical point corresponding to the vanishing boundary conditions, i.e. E(~rY b) 0 and
\E(~rY b) 0 for [~r[ . Therefore, the existence regions for solitary wave solutions can
be found by the analysis of the existence and types of critical points. For scalar solitons,
the corresponding ODE can often be solved analytically, because the corresponding
equation describes an effective system with one degree of freedom. Additionally, the
simple structure of the critical points does not allow the existence of complicated (e.g.
multi-hump) localized waves.
Two-component solitons, or more general cases of vector solitons, were rst studied by
Berkhoer and Zakharov [74], and later Manakov [75] who demonstrated the existence
of exact analytical solutions for the special kind of two-component solitary waves, the
so-called Manakov solitons. Vector solitary waves were discussed for both temporal and
spatial solitons [7679] and they were also demonstrated experimentally [80, 81]. For the
spatial case, a vector soliton brings a distinctly new concept. The vector soliton is a very
particular case of a more general `dynamic soliton', which can be described as two coupled
modes of an effective waveguide that they induce [79].
When solitons consist of several interacting modes, the stationary waves are
described by a system of coupled (and generally non-integrable) ordinary differential
equations which may possess many exotic (including multi-hump) localized solutions
[79, 82, 83], solitons with oscillating tails, etc. Usually, all these exotic solitons are
unstable. The existence of such solitons is a direct consequence of the complex critical
points of the corresponding ODE dynamical system describing the stationary localized
modes.
The structure of dark solitons is more complicated (see, for example, the recent review
paper on dark solitons [84]). From the physical point of view, dark solitons are reec-
tionless radiation modes of the waveguides they induce [85], they also have a localized
shape similar to bright solitons but with a complex envelope and non-vanishing
asymptotics. For example, in the case of the (1 1) geometry, a plane dark soliton has
the form
E(xY :) = E(nY J ) exp(ib:) exp[i/(nY J )[ (13)
where E(nY J ) and /(nY J ) are the real amplitude and phase of the wave, and n = x J:.
This time, the propagation constant b is determined by the background eld, E(nY J ) E
0
for [n[ . Parameter J denes the steering angle (or `velocity') of the dark soliton. The
main feature of a dark soliton is its non-trivial phase, so that the parameter J is related to
the total phase shift D/(J ) = /() /(). The stationary prole of a dark soliton is
described by two coupled ODEs for E(n) and /(n), and the corresponding separatrix
trajectory connects two critical points corresponding to non-vanishing boundary condi-
579
Bright and dark spatial solitons
tions [84]. The structure of dark solitons of circular symmetry in (2 1) dimension
+
, i.e.
vortex solitons, is more complicated [84, 88],
E(~rY uY :) = E(r) exp(ib:) exp(iu) (14)
where b is the background propagation constant, r = [~r[ is the radius and u is the azi-
muthal angle in the cylindrical system of coordinates. As above, the vortex prole is
dened by the corresponding ODE for the function E(r). Apart from the vortex soliton, no
other spatial soliton is mathematically stable in a Kerr (cubic) non-linearity.
Therefore, the regions of the existence of solitary waves solutions with decaying (bright)
or constant (dark) asymptotics can be determined by the analysis of the critical points of
the corresponding system of ODEs. For scalar (bright and dark) solitons, the ODEs are
integrable (an effective mechanical system with one degree of freedom) and the structure of
the critical points usually does not allow the existence of complicated (e.g. multi-hump)
solutions. For vector solitons, e.g. two different orthogonal modes, the stationary waves
are described by a system of coupled (and generally non-integrable) ODEs which may
display many exotic solutions which are usually unstable. The existence of such solitons is a
direct consequence of the complex structure of the critical points of the corresponding
dynamical system describing the stationary localized modes.
3.2. Concept of an induced waveguide
The notion that a self-guided beam can be regarded as a mode of the waveguide it induces
is a useful tool for understanding spatial optical solitons. This concept, as a matter of fact,
is somewhat self-evident, and it has been known for more than 30 years from the rst
prediction of self-guided beams [2, 13, 14]. Recently, Snyder and co-workers [12, 79, 89]
have suggested this concept for the so-called self-consistency method (also called the linear
perspective approach) and interpretation of both stationary and even some non-stationary
self-guided beams.
The linear perspective concept is based on the fact that all spatial solitary waves are
qualitatively the same. In particular, all stationary self-guided beams can be treated as the
modes of a (linear) axially uniform waveguide. This waveguide is induced by the inter-
action of light with the non-linear medium and is in general anisotropic. This elementary
concept allows us to borrow physics and exact solutions for guided waves directly from the
pages of waveguide theory providing a physical insight into the physics of guided waves.
In the simplest case, a soliton is one mode of the waveguide it induces [2, 79]; more
generally, it is any two (or more) modes of the induced waveguide [79] which explains the
coexistence of different classes of multi-component solitons such as dark and bright.
Vector solitons are the special case when the modes are degenerate.
4. Solitons in v
(2)
media
4.1. Models for quadratic solitons
When the fundamental frequency becomes phase matched with one of its harmonics, the
NLS equation fails and the wave propagation should be described by some other models.
In the case of quadratic (or the so-called v
(2)
) non-linearity, the simplest eect of resonant
*
As spatially localized solutions of the defocussing cubic NLS equation, vortex solitons were introduced, for the
rst time, in a paper by Ginzburg and Pitaevski [86] (see also [87]), as topological excitations in super uids.
However the term `vortex' had been used much earlier for dierent (linear) physical processes and for dierent
dening equations.
580
YU. S. Kivshar
wave interaction is the generation of the second harmonics 2x by the fundamental fre-
quency x, the process being a particular case of a more general process of three-wave
mixing. Such an interaction is efcient provided the matching conditions between the wave
propagation constants are satised. This is well known in the theory of the second-har-
monic generation (SHG) (see, for example, [90, 91]). For example, in an anisotropic
medium, for any wave vector direction
~
kak, two dierent corresponding values of k(x) can
be found. In other words, for any direction of propagation there are two normal waves
(which are called ordinary and extraordinary waves) which have different polarizations and
travel with dierent phase velocities. For the ordinary wave the direction of wave vector
~
k
coincides with the direction of the Poynting vector~s (i.e. with the direction of energy ow),
whereas for the extraordinary wave the directions of
~
k and ~s do not coincide.
To derive the model of three-wave mixing in a diractive medium, one should consider
parametric interaction between three stationary quasi-plane monochromatic waves with
the envelopes E

(where = 1Y 2Y 3) and assume x


1
x
2
= x
3
with the corresponding wave
vectors to be nearly matched, i.e. k
1
(x
1
) k
2
(x
2
) k
3
(x
3
) = Dk, where Dk k
3
. If all
three vectors
~
k

have the same direction, there is no phase velocity walk-o. However, if


some of the three waves are extraordinary then their energy ows diverge and this should
be taken into account in the structure of the slowly varying envelopes E

. Choosing the
:-axis as the direction of
~
k

, and the x-axis being in the plane dened by


~
k

and the direction


of the energy walk-o, we consider the electric eld as a sum of three elds of the reso-
nantly interacting frequencies. As a result, in the approximation of slowly varying enve-
lopes, we can derive a system of equations that describes the type II SHG,
2ik
1
dE
1
d:
2ik
1
q
1
dE
1
dx

d
2
E
1
dx
2

d
2
E
1
dy
2

8px
2
1
c
2
~ v
(2)
1
E
3
E
+
2
exp(iDk:) = 0
2ik
2
dE
2
d:
2ik
2
q
2
dE
2
dx

d
2
E
2
dx
2

d
2
E
2
dy
2

8px
2
2
c
2
~ v
(2)
2
E
3
E
+
1
exp(iDk:) = 0
2ik
3
dE
3
d:
2ik
3
q
3
dE
3
dx

d
2
E
3
dx
2

d
2
E
3
dy
2

8px
2
3
c
2
~ v
(2)
3
E
1
E
2
exp(iDk:) = 0
(15)
Equations 15 describe the case when the spatial walk-o of all waves occurs in the same
plain. Formally, this is true only for single-axis crystals, but the corresponding general-
ization is trivial. For a slab waveguide, the structure of the linear guided modes in the
direction of the trapping provided by the waveguide is known. Then, using an approxi-
mate separation of variables, E

Y E

(xY yY :) = F

(y)E

(xY :), and integrating out the depen-


dencies in y, we can obtain the similar system of non-linear coupled equations but with the
normalized (scaled) coecients.
In the limiting case, when x
1
= x
2
= x
3
a2, only one characteristic frequency x
0
= x
1
is
involved. This requires only one source of coherent radiation at the fundamental fre-
quency x
0
and a wave of the double frequency 2x
0
is generated due to SHG phenomenon
(type I SHG). In this case, we put E
1
= E
2
and, therefore, the spatial solitons due to type I
SHG in a v
(2)
slab waveguide are described by the system of two coupled equations:
2ik
0
dE
1
d:

d
2
E
1
dx
2

8px
2
0
c
2
~ v
(2)
1
E
3
E
+
1
exp(iDk:) = 0
4ik
0
dE
3
d:
4ik
0
q
3
dE
3
dx

d
2
E
3
dx
2

32px
2
0
c
2
~ v
(2)
3
E
2
1
exp(iDk:) = 0
(16)
581
Bright and dark spatial solitons
where x
0
Y k
0
and E
1
are the frequency, wave number and the electric eld intensity of the
rst harmonic wave, respectively; E
3
and q
3
are the electric eld intensity and the walk-o
angle for the second harmonic wave, and Dk = 2k
0
k
3
is the wave vector mismatch.
In the case of spatial solitons, we normalize Equations 16 measuring the transverse
coordinate x in units of the input beam size r
0
, and the propagation coordinate :, in units
of the diraction length :
d
= r
2
0
k
2
= r
2
0
ac
1
. Introducing the dimensionless elds,
E
1
= [t([c
1
c
2
[)
1a2
a(2v
1
v
2
r
4
0
)
1
2
exp(ib:)[ and E
2
= (x[c
1
[av
1
r
2
0
) exp[i(2b D):[, we nally
obtain
i
dt
d:
r
d
2
t
dx
2
bt wt
+
= 0
ir
dw
d:
id
dw
dx
s
d
2
w
dx
2
r(2b D)w
t
2
2
= 0
(17)
where D = :
d
DkY d = d
2
r
0
a[c
2
[Y r = sign(c
1
)Y s = sign(c
2
) and r = [c
1
ac
2
[. The dimen-
sionless parameter b is proportional to the non-linearity-induced phase velocity shift.
4.2. Solitons in the cascading limit
In the simplest case of no walk-o, system 17 can be further normalized to scale out the
propagation constant b,
i
dt
d:
r
d
2
t
dx
2
t wt
+
= 0
ir
dw
d:
s
d
2
w
dx
2
aw
t
2
2
= 0
(18)
where the dimensionless parameter a = 2r rDab includes the mismatch parameter.
Equation 18 is the generic model of v
(2)
solitons in the absence of walk-o. Its solutions
have been analysed by many authors, in the (1+1)-dimensional case [9298] and in a more
general (2+1)-dimensional case [99, 100]. Solitary waves in the case of more general, non-
degenerated three-wave mixing has also been investigated (see, for example, [101, 102]).
Experimental observation of quadratic solitons has been reported for both waveguide
geometry [103] and for two-frequency beam propagation in a bulk [23, 104] (see also a
review paper [11]).
It is straightforward to see why (and when) we expect to nd spatially localized solu-
tions of Equations 18. Indeed, let us consider the limit of large a, which corresponds to
large positive values of the mismatch Dk. In this case, the second equation of the system 18
can be approximately reduced to the form w ~ t
2
a(2a). The substitution of this expression
into the rst equation of system 18 results in the standard NLS equation for the rst
harmonic,
i
dt
d:
r
d
2
t
dx
2
t
1
2a
[t[
2
t = 0 (19)
The NLS equation possesses stable bright (at r = 1) or dark (at r = 1) soliton solu-
tions. We will call the limit of large a the cascading limit. In this limit the effective Kerr-like
behaviour due to cascaded v
(2)
effects is clearly seen and the second harmonic component
w is much weaker than the rst harmonic t.
Using such a simple reduction to the NLS equation, we may look for a stationary (i.e.
:-independent) localized solutions of Equations 18 in the form of an asymptotic series in
582
YU. S. Kivshar
the parameter a
1
and nd the real functions t(x) and w(x) in the form of asymptotic
series,
t(x) = 2a
1a2
sech x 4sa
1a2
tanh
2
x sech x
w(x) = 2 sech
2
x sa
1
(16 sech
2
x 20 sech
4
x)
(20)
for bright solitons at r = 1, and
t(s) = 2
1a2
a
1a2
tanh s 2
1a2
sa
1a2
(s sech
2
s tanh s sech
2
s)
w(s) = tanh
2
s sa
1
(2s tanh s sech
2
s 4 sech
2
s 5 sech
4
s)
(21)
where s = xa2
1a2
, for dark solitons at r = 1.
The properties of Kerr solitons of Equation 19 are well known. Existence of the as-
ymptotic solutions 20 and 21 obtained in the cascading limit suggests that for a 1 the
system 18 should have stable bright solitons for r = 1Y s = 1, and stable dark solitons
for r = 1Y s = 1, similar to Equation 19. However, this conclusion is not satisfactory,
because: (i) formal localized solutions 20 and 21 can be non-stationary for system 18 due
to their resonance with linear waves; (ii) in the case of dark solitons, the solutions 21 can
also be unstable due to parametric modulational instability, as has been rst demonstrated
in [105, 106].
4.3. Families of bright solitons
As follows from the NLS limit valid for large a, bright solitons of the two coupled
equations 18 should exist for r = 1 in the form of one-hump localized proles for the real
functions t(x) and w(x). Such solutions were rst found by Buryak and Kivshar [95, 96, 98]
using the numerical shooting technique, for any positive value of a.
Examples of two-wave localized solutions of Equations 18 are presented in Fig. 1a for
a = 0X2 and a = 10X0. For a 1 the maximum amplitude of the fundamental component
t
max
is much larger than the similar value w
max
for the second harmonic component and
this case corresponds to the asymptotic solution 20, t ~ 2a
1a2
sech xY w ~ 2 sech
2
x. The
ratio w
max
at
max
characterizing the whole family is plotted in Fig. 1a where the lled circle
corresponds to the exact solution [92]
t(x) = 2
1a2
w(x) = (3a2
1a2
) sech
2
(xa2) (22)
that exists at a = 1, and the asymptotic (dashed) curve corresponds to the NLS limit,
w
max
at
max
~ 1aa
1a2
.
The numerical analysis of the stability of this soliton family based on the direct inte-
gration of Equation 18 and the eigenvalue analysis of the corresponding linearized
problem has shown that both stable and unstable solitons exist, depending on the values of
system parameters a and r [107]. In the cascading limit a 1, the solitons shown in Fig.
1a are stable, whereas in the other limit (a 0) solitons become unstable (e.g. left-bottom
soliton of Fig. 1a is unstable for r = 2X0). In spite of detected instability for a 0, for
a ~ 1 the parametric solitons are stable even under the action of very strong perturbations.
These two-wave solitons can be generated from a rather broad class of initial conditions.
Figure 1b shows the soliton generation from a rst harmonic sech-form input pulse. Due
to diraction, the input pulse becomes broader, but it also generates the second harmonics
and, after a rather short transition period, a two-component bright soliton forms. This
583
Bright and dark spatial solitons
Figure 1 (a) Characteristic proles of two-wave bright solitons at r = 1 and s = 1 as the functions of the
dimensionless mismatch parameter a. The lled circle at a = 1 corresponds to the exact solution 22. (b) Self-
trapping of an initial sech-prole input into a two-wave bright soliton. Shown are the peak intensities of the
fundamental and second harmonics.
584
YU. S. Kivshar
kind of behaviour is possible only due to the existence of the continuous family of stable
bright solitons.
Additionally to the one-hump localized solutions described above, the numerical
analysis indicates the existence of continuous (in a) families of two-hump (and even multi-
hump) bright solitons, which can be treated as bound states of one-hump solitons [108,
109]. These solitons exist only for 0 ` a ` 1. At a 1 the distance between the neigh-
bouring solitons increases to innity. Numerical stability analysis indicates that these
multi-hump bright solitons are unstable and either split into partial stable solitons or
disintegrate completely, for suciently small values of a where stab!e single solitons do not
exist.
In spite of the fact that in the cascading limit the eective NLS equation 19 does not
depend on the sign s, localized solutions are very dierent for s = 1 and s = 1 in
Equations 18. A simple analysis of the soliton tails indicates that for s = 1 one-hump
solitons do not exist due to the resonance with linear waves. However, the numerical
analysis still allows one to nd bound states of such solitons existing as discrete sets of
two- (and multi-)soliton radiationless states where radiation is suppressed outside, but
exists between the solitons in the form of trapped oscillations. Such bound states of
solitons in the presence of radiation may occur in other non-linear systems where single
solitons in the presence of radiation may occur in other non-linear systems where single
solitons do not exist (see, for example, [110] and references therein). Figure 2 gives the
results related to such two-soliton states as solutions of Equations 18 at r = 1 and
s = 1.
It is interesting, that in this case one solution is also known in an explicit analytical form
[111]. It exists at a = 2 and has the form,
t(x) = 6(2
1a2
) tanh x sech x w(x) = 6 sech
2
x (23)
Solution 23 represents one member of the family of two-soliton bound states of an integer
order. Two-soliton bound states of the third and tenth orders are shown in the bottom
part of Fig. 2. Because of a delicate balance between the solitons and radiation for such
stationary solutions to exist, all these bound states are unstable, they either split into single
radiating solitons, or disintegrate in a more complicated fashion.
4.4. Families of dark solitons
Following the preliminary results of the cascading limit when the eective NLS equation
19 is valid, we expect to nd dark solitons in the case r = 1, that corresponds to a
defocusing eective cubic non-linearity. Indeed, the numerical results obtained by Buryak
and Kivshar [95, 96, 98] indicate that single dark radiationless solitons exist for
r = s = 1 as localized solutions of Equations 18. In the case r = s = 1, a continuous
family of parametric dark solitons exists for 0 ` a ` and in the interval 0 ` a ` 8 these
solitons have non-monotonic radiationless oscillatory tails. Examples of these two-wave
dark solitons are presented in Fig. 3 for a = 1X0 (nonmonotonic tails) and a = 10X0
(monotonic tails).
For the cascading limit (a 1) the solution can be presented in the asymptotic form 21.
When a is not large, the asymptotic solution 21 fails (e.g. it does not describe oscillatory
tails for a ` 8), but the dark soliton family still exists for a b 0, and it can be charac-
terized, e.g. by the minimum amplitude of the second harmonic w
min
. The dependence of
w
min
versus a is shown in Fig. 3. For large a it approaches the asymptotic dashed curve
585
Bright and dark spatial solitons
w
min
~ 1aa which corresponds to the NLS-like solitons of the cascading limit. Dark sol-
itons of Fig. 3 are stable if their backgrounds are modulationally stable (i.e. the stability
domain is 2 ` a ` , and it does not depend on r).
Due to the existence of decaying oscillating tails, a dark soliton can trap another dark
soliton to form a bound state, a twin-hole dark soliton [106]. This mechanism is well known
for other types of solitons [112], and the v
(2)
dark solitons are just a particular example.
Figure 2 Discrete set of two-wave bright solitons for r = 1 and s = 1 in the form of soliton pairs. These
solitons appear as two radiative bright solitons with radiation trapped between them. The lled circle at a = 2
corresponds to the exact solution 23.
586
YU. S. Kivshar
For a 8 the distance between the neighbouring dark solitons in a bound state increases
to innity. To the best of our knowledge, this is the rst example when stable twin-hole
dark solitons have been identied.
The dark solitons presented above exist for the case r = s = 1 in Equations 18.
Recently other continuous families of dark solitons have been found for r = s = 1 in
the interval 0 ` a ` 2 [109]. These dark solitons (similarly to the solitons shown in Fig. 3)
Figure 3 Characteristic proles of two-wave dark solitons for r = 1 and s = 1 versus the dimensionless
mismatch parameter a. Oscillating tails of the dark solitons appear for a ` 8.
587
Bright and dark spatial solitons
possess oscillating tails and thus can form bound states. Stability of these solutions is still
an open problem.
Similar to the case of bright solitons, for r = s = 1 in Equations 18 single dark
radiationless solitons do not exist due to the resonance with linear waves. However, a
discrete set of two- (and multi-) soliton radiationless bound states can be found, these
solutions appear due to trapping of radiation. Figure 4 gives the results related to two-
soliton radiationless bound states of dark radiative solitons. One can see that, similar to
the cases discussed above, there exists an exact analytic solution [113]
t(x) = 2
1a2
w(x) = 2
1a2
1
3
2
sech
2
(xa2)
_ _
(24)
which is a two-soliton bound state of the rst order. The two-soliton bound states of the
third and sixth orders are shown in the bottom part of Fig. 4.
Analytical results [105, 106] indicate that all radiationless bound states of radiative dark
solitons are unstable due to the development of parametric modulational instability. This
result has been conrmed by the direct numerical simulations.
5. Solitons due to third-harmonic generation
5.1. Model and motivations
In the case of solitary waves of quadratic media, the coupling between the fundamental
and second harmonics produces an eective non-linearity that allows the beam self-
trapping via mutual self-focusing. When the harmonics are decoupled, no solitary waves
exist due to solely quadratic non-linearity. Here we consider the qualitative dierent
example. We consider the case of the Kerr-type non-linear response near the point of the
third-harmonic generation. As is well known (see, for example, [90, 91]), the third-order
contribution to the polarization is made up of two components, a response at the fun-
damental frequency of the beam (described through an intensity-dependent refractive
index), and a response at the frequency 3x, which is known to lead to third-harmonic
generation under the condition of phase matching. Then, launching a monochromatic
beam near the point of phase matching with the third harmonics can result in co-prop-
agating beams at the fundamental and third-harmonic frequencies drastically modifying
the plane waves [114] and solitons of Kerr non-linearity [115, 116]. Such a mechanism of
resonant wave coupling can be responsible for the generation of enhanced non-linearities
and non-Kerr solitary waves even in the systems with a weak cubic response.
The process of the generation of non-Kerr solitons due to the third- harmonic inter-
action can be regarded as another example of the cascading eect whereby eective higher-
order non-linearities are generated [117], the mechanism responsible for the existence of
quadratic solitons. In a cubic medium, this mechanism drastically aects the propagation
of spatial solitary waves of the fundamental frequency under the condition of phase
matching with the third harmonic. The physics of this eect is rather simple. Indeed, it can
be shown [114] that a slightly mismatched process of third-harmonic generation leads to
an eective quintic non-linearity and, therefore, to the cubic-quintic NLS equation. As a
result, we expect that solitary waves cease to exist when the effective quintic non-linearity
becomes strongly defocusing (see, for example, [55] and references therein, for the most
recent comprehensive analysis of solitary waves of the cubic-quintic NLS equation).
To describe the solitons modied by the third-harmonic generation, we follow the
original papers by Sammut et al. [115, 116] and consider resonant interaction between a
588
YU. S. Kivshar
linearly-polarized beam of frequency x and its third harmonic (assuming a slab waveguide
geometry) presenting the electric eld E in the form E =
1
2
[E
1
exp[i(k
1
: xt)[
E
3
exp[i(k
3
: 3xt)[ c.c., where k

= xn

ac and n

= n(x) for = 1Y 3. Each frequency


component of the eld then satises that scalar wave equation where we assume the cubic
non-linear response. For the slowly varying envelopes E
1
and E
3
, we obtain
Figure 4 Discrete set of two-wave dark soliton pairs existing for r = 1 and s = 1. These solitons appear as
two dark solitons with radiation trapped between them. The lled circle at a = 1 corresponds to the exact
solution 24.
589
Bright and dark spatial solitons
2ik
1
dE
1
d:

d
2
E
1
dx
2
v [E
1
[
2
2[E
3
[
2
_ _
E
1
E
+2
1
E
3
exp(iDk:)
_ _
= 0 (25)
2ik
3
dE
3
d:

d
2
E
3
dx
2
9v [E
3
[
2
2[E
1
[
2
_ _
E
3

1
3
E
3
1
exp(iDk:)
_ _
= 0 (26)
where Dk = 3k
1
k
3
is the phase mismatch, the non-linearity parameter
v = (3px
2
ac
2
)[v
(3)
[ is dened to be always positive, whereas the sign of the Kerr non-
linearity depends on whether the material is self-focusing (positive) or defocusing (nega-
tive). These equations describe a special case of a more general four-wave mixing process
(see, for example, [118] as an example of solitary waves).
Equations 25 and 26 can be further normalized using the scales of a beam width :
0
and
diraction length :
d
= 2:
2
0
k
1
, and introducing the dimensionless elds U = 3(k
1
:
2
0
v)
1a2
E
1
and W = (k
1
:
2
0
v)
1a2
exp(iDk:)E
3
. For the stationary solutions we then substitute
U = ub
1a2
exp(ibZ) and W = wb
1a2
exp(i3bZ), where b is the beam propagation constant
which is dened by the beam total power and therefore can be treated as an external
parameter of families of solitary waves. As a result, the system of coupled equations for
the solitary wave proles can be rewritten in the following dimensionless form [115, 116]
i
du
d:

d
2
u
dx
2
u
1
9
[u[
2
2[w[
2
_ _
u
1
3
u
+2
w = 0 (27)
ir
dw
d:

d
2
w
dx
2
aw (9[w[
2
2[u[
2
)w
1
9
u
3
= 0 (28)
where : = bZ and x = b
1a2
X. Stationary beams are described by real solutions, u(x) and
w(x), dened by the system 27 and 28 with the :-derivatives omitted. These localized solu-
tions depend only on a single dimensionless parameter, a = r(3 Dab) with two dimen-
sionless parameters, D = 2k
1
Dk:
2
0
and r = k
3
ak
1
, where r = 3 for the case of spatial solitons.
5.2. Multistability of bright solitons
The basic structure of Equations 27 and 28 is qualitatively similar to the equations derived
above for the case of parametric solitary waves supported by two-wave mixing in v
(2)
media. Moreover, the denition of the eective mismatch parameter a is almost identical
to that case, in spite of the dierent structure and physical meaning of the non-linear
coupling terms. We believe this is a general property of dierent systems with cascaded
non-linearities.
First of all, far away from the point of phase matching (i.e. for large a), the energy
conversion from the fundamental to the third harmonic is small, i.e. [w[ [u[. For a 1,
from Equation 28 we have approximately w u
3
a9a, and Equation 27 becomes the cubic-
quintic NLS equation allowing the solutions for solitary waves in an explicit form (see, for
example, [55]). This suggests the general structure of an asymptotic expansion for the
localized solutions of Equations 27 and 28 in powers of = a
1
, as was rst presented by
Sammut et al. [115]
u(x) =
3(2
1a2
)
cosh x
1 2 2
1
cosh
2
x
_ _
2
2
70
29
cosh
2
x

12
cosh
4
x
_ _ _ _
O(
3
) (29)
w(x) =
6(2
1a2
)
cosh
3
x
1 3 1
10
cosh
3
x
_ _ _ _
O(
3
) (30)
590
YU. S. Kivshar
The asymptotic analytical solutions 29 and 30 are a limiting case of a family of localized
solutions described by the real functions u(x) and w(x) which have been found numerically
[115, 116]. All types of these solutions can be characterized by the normalized total power
P
tot
=
_

([u[
2
3r[w[
2
) dx (31)
which is a conserved quantity of the system 27 and 28.
Figure 5 shows the normalized total power P
tot
dened by Equation 31 as a function of
the mismatch parameter a, for dierent types of two-wave localized solutions of the system
27 and 28. The inset gure shows an expanded portion of the dependence P
tot
(a) for the
range 8.2 _ a _ 9.2. It can be seen from the form of P
tot
(a) that, near the point of phase
matching between the fundamental and third harmonics (i.e. at a = 9 when r = 3), there
exist three distinct types of localized solutions for bright solitary waves.
The most important class of two-wave bright solitons is described by a family of lo-
calized solutions for coupled fundamental and third harmonic elds. The distribution of
power between the two frequencies varies from being predominantly in the third har-
monic, for smaller a, to predominantly in the fundamental, at larger values of a. In this
latter case, we can apply the cascading approximation to nd P
tot
(a) =
36 192 (27072a5)
2
O(
3
), which is shown as a dashed line in Fig. 5.
For large values of the normalized mismatch parameter a, the amplitude of the beam at
the fundamental frequency grows whereas that of the third harmonic vanishes, in agree-
ment with the prediction of the analytical results obtained in the cascading approximation.
An example of this kind of solution is presented in Fig. 6a, it corresponds to the point A in
Fig. 5.
A simple analysis done by Sammut et al. [115, 116] shows that the family of two-
frequency solitary waves bifurcates from the one-frequency solution for the third har-
monic, w(x) = [(2a)
1a2
a3 [sech (a
1a2
x) and u(x) = 0, at the point of exact phase matching
(i.e. a = 9 at r = 3 and D = 0). This family of the one-frequency solitary waves is char-
acterized by the normalized power, P
tot
= 4a
1a2
, and it is described by the standard cubic
NLS equation which follows from Equation 28 at u = 0. It is clear that this type of solitary
wave is possible only due to the self-phase modulation effect taken into account for the
third harmonic.
Figure 5 Variation of the normalized total power,
P
tot
versus the dimensionless mismatch parame-
ter a for the three distinct families of solitary wave
solutions of Equations 27 and 28. The dashed
curve corresponds to the asymptotic expansion.
Lower curves merge at the bifurcation point
O (a = 9). Points A to D indicate the particular
examples presented in Fig. 6ad, respectively.
591
Bright and dark spatial solitons
Finally, the third family of localized solutions shown in Fig. 5 includes the simplest
analytical solution [119, 120]. This solution exists only at a = 1 and it has the following
form u
s
(x) = a sech x and w
s
(x) = bu
s
(x), where the parameter b is dened by a real root of
a cubic equation. In sharp contrast with the theory of v
(2)
parametric solitons, the ana-
lytical solution of model 27 and 28 and the asymptotic solution of the cascading limit,
a 1, do not belong to the same family. Moreover, varying continuously the eective
mismatch parameter a along this family of localized solutions shows that this class of
solitary waves corresponds to multi-hump solitary waves, as is shown in Fig. 6b for a = 8X2
(point B in Fig. 5). Then, it is not surprising that all solutions of this family are unstable,
the conclusion veried by employing direct numerical simulations [116].
The most interesting feature of this class of two-wave parametrically coupled solitary
waves manifests itself near the point of the phase matching. In the case of negative phase
mismatch, corresponding to the part of the curve P
tot
(a) from the left of the bifurcation
point O shown in Fig. 5, for any xed value of the parameter a we reveal the simultaneous
existence of three localized solutions. Therefore, in this case the propagation character-
istics of two-frequency coupled self-guided beams become multi valued. Two characteristic
proles of the solutions in this region are shown in Fig. 6c and d corresponding to the
points C and D in Fig. 5, respectively.
Therefore, in the problem of third-harmonic generation there exists more than one
possible propagation constant (and, consequently, more than one possible shape) of the
Figure 6 Examples of the fundamental (thin line) and third harmonic (thick line) proles for several solitary
wave solutions which belong to different families. Proles (a) to (d) correspond to points A to D in Fig. 5,
respectively.
592
YU. S. Kivshar
parametric spatial soliton for the same value of the total power P
tot
dened by Equation
31. This phenomenon is usually associated with soliton multi-stability predicted for non-
Kerr non-linearities [39, 40, 121]. For the problem of the third-harmonic generation, a
novel physical mechanism, resonant parametric wave mixing, can also lead to similar
eects indicating that eective non-Kerr non-linearities and multi-stable solitary waves
may become possible even in a Kerr medium.
Similar results have been found by Sammut et al. [116] for two-frequency dark solitons,
and it has been shown that these types of solitary waves are also drastically modied near
the point of phase matching. The most important effect discovered by Sammut et al. [116]
is the existence of parametric modulational instability that strongly affects some types of
dark solitons existing on a non-vanishing background. Additionally, numerical simula-
tions carried out by Sammut et al. [116] indicate that for the values of a selected close to
the phase-matching point a = 9, the dark solitons display an instability similar to that
existing for saturable non-linearities.
All these eects demonstrate that non-Kerr non-linearities arising from phase-matched
interaction between the fundamental beam and its third harmonic can have a dramatic
eect on the spatial solitary waves. First, such induced non-Kerr non-linearities restrict the
existence region of allowed values of the beam power, and also they lead to multi-stable
beam propagation when more than one possible beam prole and propagation constant
exist for a xed value of power near the phase matching. For dark solitary waves, para-
metric coupling generates two types of instabilities, parametric modulational instability of
the background and also inherent instability of black solitons, similar to that earlier
predicted for scalar dark solitons of saturable non-linearities.
6. Soliton internal modes
As has been shown above, dierent types of spatial optical solitons require the models of
non-Kerr non-linearities which are generally non-integrable. Soliton instability is one of the
major novel features of such solitary waves (see Section 7 below). Then, for the case of
stable non-Kerr solitons, the question is: What kind of novel physical properties can be
expected for solitary waves of non-integrable models? For many years, it was believed that
solitary waves of non-integrable models differed from solitons of integrable models only in
the character of soliton interactions: unlike proper solitons, interaction of solitary waves is
accompanied by radiation. However, the recent analysis of solitary wave solutions of the
generalized NLS models (see, for example, [122124]) has revealed that a small pertur-
bation to an integrable model may create an internal mode of a solitary wave. This effect is
beyond a regular perturbation theory, because solitons of integrable models do not possess
internal modes. But in non-integrable models such modes may introduce qualitatively new
features into the system dynamics being responsible for long-lived oscillations of the
solitary wave amplitude and resonant soliton interaction.
Internal modes have been earlier analysed only for the so-called kink solitons, topo-
logical solitary waves of the KleinGordon type models (see, for example, [125127]). The
internal modes of kinks, usually called `shape modes', are known to modify drastically the
kink dynamics because they can temporarily store energy taken away from the kink's
translational motion and later restore the energy back. This mechanism gives rise to
resonant structures in the kink-antikink collisions [125, 126].
The important common feature of non-Kerr solitons observed in numerical simulations
is that they display long-lived persistent oscillations of their amplitude [37, 107, 128, 129].
593
Bright and dark spatial solitons
Many of such features observed numerically for dierent types of envelope solitons can be
naturally explained in the framework of the concept of the soliton internal mode, gener-
ically similar to the kink's shape mode. Therefore, the existence of internal modes is a
common feature of solitary waves in many different non-integrable non-linear models.
First of all, we demonstrate that any small perturbation of the integrable NLS equation
may lead to an internal mode of a solitary wave. We follow the original work by Kivshar
et al. [124] and consider the perturbed NLS equation 6 in a slightly modied form
i
dW
d:

d
2
W
dx
2
2[W[
2
W q([W[
2
)W = 0 (32)
where, in general, q(X) is an operator describing corrections due to non-Kerr non-linear-
ities. A localized solution of Equation 32 for solitary waves can be found in the form
W(xY :) = U(x) exp(i:), where for simplicity we put b = 1 that may always be scaled out.
The real function U(x) is expressed asymptotically as U(x) = U
0
(x) U
1
(x) O(
2
),
where U
0
= sech x is the soliton of the cubic NLS equation, and U
1
is a localized cor-
rection dened from Equation 32. The linearized problem for the perturbed NLS equation
arises upon the substitution W(xY :) = U(x) [U(x) W (x)[ exp(iX:) [U
+
(x) W
+
(x)[
exp (iX:) exp(i:) and it has the form,
d
2
U
dx
2

6
cosh
2
x
1
_ _
U XW

)
1
(x)U = 0 (33a)
d
2
W
dx
2

2
cosh
2
x
1
_ _
W XU

)
2
(x)W = 0 (33b)
where

)
1
(x) = q(U
2
0
) 2U
2
0
q
/
(U
2
0
) 12U
0
U
1
and

)
2
(x) = q(U
2
0
) 4U
0
U
1
.
The linear eigenvalue problem 33 can be solved exactly at = 0 (see, for example,
[130]). Its spectrum consists of two (symmetric) branches of the continuous modes with
the eigenvalues X = X(k) = (1 k
2
), and discrete spectrum modes corresponding to
the degenerated eigenvalue X = 0. In the presence of a perturbation, the eigenvalue
problem 33 can possess an additional discrete eigenvalue that appears due to a bifur-
cation from the continuum spectrum band. To nd that bifurcating eigenvalue, we
assume, for deniteness, that the cut-o frequencies X
min
= 1 are not aected by the
perturbation. Then, the internal mode frequency can be presented in the form,
X
0
= 1
2
j
2
. The analytical approach developed in [124] allows one to obtain the
expression for the parameter [j[,
[j[ =
1
4
sign()
_

U(xY 0)

)
1
(x)U(xY 0)
_
W(xY 0)

)
2
(x)W (xY 0)
_
dx (34)
where the non-oscillatory eigenfunctions are dened in the limit k 0, i.e.
U(xY 0) = 1 2 sech
2
x and W(xY 0) = 1. The positiveness of the parameter j gives the
criterion for the existence of the soliton internal mode.
It is interesting to note that this criterion for the existence of a soliton internal mode has
an analogy with the famous Peierls problem in quantum mechanics [131] stating that a
one-dimensional potential well always possesses at least one discrete eigenvalue. For the
case of the soliton internal mode, an additional eigenvalue appears with no threshold
provided j b 0, due to a deformation of the reectionless soliton potential.
594
YU. S. Kivshar
In a general case, we should solve the eigenvalue problem 33 numerically. The case of
the cubic-quintic non-linearity, where the soliton solutions are known in an explicit form
[55], was analysed in [123]. For this kind of non-linearity, the non-linear refractive index in
Equation 6 is modelled by the function ) ([W[
2
) = 4[W[
2
3r[W[
4
, where the quintic term
describes a higher-order correction to the Kerr law, focusing (r = 1) or defocusing
(r = 1). The eigenvalue problem 33 was solved numerically by means of the shooting
technique.
A simple analysis of the bifurcation criterion 34 shows that for r = 1 an additional
(discrete) eigenvalue X = X
0
appears (with its symmetric counterpart, X = X
0
), and it
lies inside the spectrum gap, X
0
` b. The bifurcation theory allows one to nd this
eigenvalue near the spectrum band edge,
X
0
= b 1
1
36
b
2
O(b
4
)
_ _
(35)
The eigenfunctions corresponding to this eigenvalue decay exponentially for large x. Thus,
in the cubic-quintic NLS equation, any small soliton perturbation should transform into
three components: localized correction to the stationary shape of the solitary wave and
radiation (those are the same as for the Kerr solitons), and the third component, which we
call the soliton internal mode, that manifests itself as oscillations of the soliton width and
amplitude. This mode is associated with a novel (discrete) eigenvalue of the linearized
problem 33.
In contrast, for r = 1 the bifurcation analysis does not predict any additional discrete
eigenvalue of the linear problem 33. Therefore, the soliton internal mode does not appear,
so that in this case long-lived oscillations of the soliton width and amplitude are not
expected.
To illustrate these conclusions, Fig. 7a and b shows two distinct types of the evolution of
the soliton peak intensity when the soliton amplitude is perturbed by 107 of its value. For
r = 1, this perturbation excites a long-lived periodic variation of the soliton peak in-
tensity and width, see Fig. 7a. However, no soliton internal mode exists for r = 1, that is
why the initial oscillations decay rapidly for any value of b. An example is shown in Fig. 7b
for b = 0X8, where also a shift of the soliton peak intensity due to the perturbation can be
clearly seen.
We would like to emphasize an important consequence of the existence of the soliton
internal mode, which can be probably observed experimentally, as has been suggested in
[123]. Indeed, because the period of the internal soliton oscillation depends on the total
power of the beam, the output width of the solitary wave created by an input beam at a
xed propagation distance depends non-monotonously on the input power. Figure 8 shows
two examples of the dependence of the output beam width as a function of the input power
for the scaled soliton, i.e. the input beam of the form W(xY 0) = AU(x) with A ,= 1. It is clear
that this dependence becomes oscillatory for the case when the soliton internal mode exists
being excited by the input beam (thin solid, Fig. 8), whereas it is an almost straight line for
the other case, when the soliton internal mode does not exist (thick solid, Fig. 8).
7. Soliton stability: a general overview
Spatial optical solitons are of both fundamental and technological importance if they are
stable under propagation. Existence of stationary solutions of the dierent models of non-
Kerr non-linearities, in particular those presented above, does not guarantee their
595
Bright and dark spatial solitons
stability. Therefore, the soliton stability is a key issue in the theory of self-guided optical
beams.
For temporal solitons in optical bres, non-linear eects are usually weak and the model
based on the cubic NLS equation and its deformations is valid in most of the cases [31, 32].
Therefore, being described by integrable or nearly integrable models, solitons are always
stable, or their dynamics can be aected by (generally small) external perturbations which
can be treated in the framework of perturbation theory.
As has been discussed in Section 2.1, much higher powers are usually required for
spatial solitons in waveguides or bulk media, so that real optical materials demonstrate
Figure 7 Dynamics of a perturbed soliton in the cubic-quintic NLS model: (a) r = 1 and b = 5 (soliton
internal mode exists and its width is of the order of the soliton width); (b) r = 1 and b = 0X8 (mode does not
exist). Dashed lines in (a) and (b) show the corresponding unperturbed values of the soliton intensity.
596
YU. S. Kivshar
essentially a non-Kerr change of the non-linear refractive index with increasing light
intensity. Generally speaking, the non-linear refractive index always deviates from Kerr
for larger input powers, e.g. it saturates. Therefore, models with a more general intensity-
dependent refractive index are employed to analyse spatial solitons and, as we discuss
below, solitary waves in such non-Kerr materials can become unstable. Importantly, very
often the stability criteria for solitary waves can be formulated in a rather general form
using the system invariants.
7.1. Stability of one-parameter solitons
7.1.1. Bright solitons
Stability of bright solitons of the NLS equation with a generalized non-linearity has been
extensively investigated for many years, and the criterion for the soliton stability has been
derived by dierent methods (see, for example, [132137]). Stability of bright solitons in
the generalized NLS equation of any dimension is given by the simple integral criterion
[132]
dP
db
=
d
db
_
J
[E(~rY :)[
2
d~r
_ _
b 0 (36)
Figure 8 Output beam width w
out
, normalized to the input beam width w
0
, versus the input beam power P,
normalized to the soliton power P
0
, both measured at z = 10. Thin solid: the case when the soliton internal
mode exists (r = 1Y b = 2); thick solid: the case when the soliton internal mode does not exist
(r = 1Y b = 0X6).
597
Bright and dark spatial solitons
where P is the total beam power and b is the soliton propagation constant. The validity of
the VakhitovKolokolov criterion 36 is based on the specic properties of the eigenvalue
problem that appears after linearizing the NLS equation 6 near the solitary wave solutions
[134, 135], for example, the condition 36 is associated with the existence of only one
negative eigenvalue of that problem. If this latter condition is not fullled, the stability
criterion may not be directly formulated in terms of the beam power P, as we have in the
case of non-linear guided waves. Indeed, it has already been established that stability of
self-guided waves in non-linear waveguide structures can be given in some cases by the
same criterion (see, for example, [137139]) but, in general, it is more complicated and
depends on a particular structure and the type of non-linearity.
Criterion of the soliton stability 36 is usually valid for the bright solitons (and non-linear
guided waves) which constitute a one-parameter family of localized solutions, i.e. their
shape is solely dened by the beam propagation constant b. The similar criterion is valid
even in the case of two-component solitons governed by the only power invariant. For
example, it has been shown that a similar criterion applies for two-wave solitons in v
(2)
materials [107].
Linear stability analysis does not allow one to predict the subsequent evolution of
unstable solitons. Dierent scenarios of the instability-induced dynamics of bright solitons
have been found numerically (see, for example, [133] and references therein). Recently,
Pelinovsky et al. [53, 54] have developed an asymptotic analytical method to describe the
dynamics of unstable solitons (e.g. their diffraction-induced decay, collapse, or switching
to a novel stable state). Some of these results are summarized below in Section 8.
7.1.2. Dark solitons
In contrast to bright solitons, the stability criterion for dark solitons of the generalized
NLS equation has not been understood until recently and, even more, this issue created a
lot of misunderstanding in the past. In particular, we notice some eorts to employ, by
analogy with bright solitons, the soliton complementary power (see, for example, [41]) and
to use this invariant for analysing the soliton bistability [42, 46], a statement that a black
dark soliton (a dark soliton with zero intensity at its centre) is always stable [137], etc. As
was noticed in [50] after analysing the results of numerical simulations, the complimentary
power does not dene the stability of dark solitons.
From a historical point of view, the rst eorts to analyse the stability of dark solitons
were stimulated by numerical simulations done by Barashenkov and Kholmurodov [140]
who observed instability of the so-called `bubbles', localized waves of rarefaction of the
Bose gas condensate. These non-topological solitary waves belong, in the (1 1)-
dimensional case, to the family of dark solitons of the NLS equation with the cubic-quintic
non-linearity and they survive in higher dimensions [141]. Although quiescent bubbles
were found to be always unstable regardless of dimension [142], numerical simulations
revealed that moving bubbles can be stabilized at non-zero velocities [140]. Later Bogdan
et al. [143] (see also [144]) explained this phenomenon through the multi-valued depen-
dence of the bubble's energy versus the remormalized bubble's momentum.
However, it was believed for a long time that kink-type dark solitons (in particular,
black solitons) are always stable (see, for example, [137]). Instability of black solitons was
observed for the rst time by Kivshar and Krolikowski [145] in numerical simulations of
the NLS equation with the saturable non-linearity 9 at p = 2. These authors used the
variational principle to link the instability to the existence of multivalued solutions in
598
YU. S. Kivshar
terms of the system invariants, and also derived the instability threshold condition (i.e. the
marginal stability condition) by using the multi-scale asymptotic expansions.
Eventhough it was known for some time that the stability criterion for dark solitons
should be dened through the renormalized soliton momentum,
dM
r
dt
=
d
dt
i
2
_

u
du
+
dx
u
+
du
dx
_ _
1
u
2
0
[u[
2
_ _
dx
_ _
b 0 (37)
and this was shown to be consistent with the results of numerical simulations [140, 142,
146] and the variational principle for bubble-type [143] and kink-type [145] dark solitons,
the rigorous proof of this stability criterion was presented only recently by Barashenkov
[147], with the help of the Lyapunov function, and Pelinovsky et al. [54], by using the
asymptotic expansion near the instability threshold. The rst approach does not allow one
to describe the instability itself but it is more general to prove the global stability if it
exists, whereas the second method is valid in the vicinity of the instability threshold being
sufcient to determine the instability domain.
All basic models of optical non-linearities, i.e. the models with saturable, transiting and
competing non-linearity discussed in Section 2.3, display instabilities of dark solitons in
some regions of the parameters. For example, in a saturable medium dark solitons become
unstable provided the saturation intensity is below a certain threshold value [54, 146].
Instability displayed by dark solitons is qualitatively different from the instability of bright
solitons, it is a drift instability (see [146]).
7.2. Stability of two-parameter solitons
Many of the `novel' soliton states demonstrated for case of two interacting modes are
unstable. Stability of bright and dark vector solitons (see, for example, [12, 7678]),
rotating or dynamic solitons [79] is still an open problem. However, as has been demon-
strated for the rst time for the case of non-degenerated three-wave mixing in a diffractive
medium [101], the threshold of the instability for two-parameter solitons (marginal sta-
bility condition) is given by the following criterion
J b
1
Y b
2
( ) =
d F
1
Y F
2
( )
d b
1
Y b
2
( )
=
dF
1
db
1
dF
2
db
2

dF
1
db
2
dF
2
db
1
= 0 (38)
where F

( = 1Y 2) are two invariants of the system (total and complimentary powers, for
the problem of three-wave mixing) describing two-parameter solitary waves, and b

are
two independent parameters of the stationary localized solutions. The stability criterion
itself, i.e. the sign of the function J, depends on the model under consideration.
Result 38 has rst been derived by the asymptotic expansion technique [101] and then
veried by the analysis of the global structure of the system invariants [102, 148]. It is a
direct consequence of the topology of the invariant surface H(F
1
Y F
2
), and it seems to
be valid for dierent types of vectorial and coupled solitons described by two indepen-
dent parameters introduced by two non-trivial invariants of the model. For example, in
the case of coupled bright-dark solitons, F
1
is the power of the bright component P and
b
1
is the propagation constant of the bright component, whereas F
2
is the total
momentum M and b
2
is the soliton velocity J [149]. For incoherently interacting bright
solitons, the invariants F

are two powers corresponding to two scalar components


[150]. The same result holds for the stability of solitons in the models with the absence of
599
Bright and dark spatial solitons
the Galilean invariance, such as two-wave parametric solitons with the walk-off effect. In
this latter case the second parameter is the soliton velocity J and the second invariant is
the soliton momentum M; the same criterion has been recently demonstrated for the
so-called walking solitons [151].
In general, the marginal stability criterion 38 is valid provided the conditions similar to
those for the validity of the VakhitovKolokolov criterion 36 are satised, namely the
instability is associated with a kind of translational bifurcation of localized eigenmodes of
an associated linear eigenvalue problem when the value k
2
, where k is an eigenvalue,
remains real but it changes its sign passing zero. For more complicated models, the
invariant criterion may be not valid and other types, e.g. oscillatory instabilities may occur,
as has been recently demonstrated for gap solitons in a non-integrable deformation of the
Thirring model [152]. Another example is given by a system where there exist, for the same
values of the system parameters, a number of different soliton families corresponding to
bifurcations of invariant surfaces. An example of this kind has been recently found for the
problem of non-degenerated four-wave mixing where stable multi-colour solitons
correspond to the lowest invariant surface [153].
7.3. Symmetry-breaking and transverse soliton instabilities
The soliton criteria discussed above are a special case of more general wave instabilities
that may occur due to non-linear properties of physical systems [154]. The simplest type of
such an instability is modulational instability characterizing a breakup of a continuous-
wave (or plane wave) eld of a large intensity. Modulational instability was rst predicted
and analysed in the context of non-linear wave propagation in uids [155, 156] and, even
earlier, in non-linear optics [157, 158]. One of the important physical processes associated
with the development of modulational instability is the generation of a train of localized
waves (beams or pulses) [159], the effect observed experimentally in different physical
systems, e.g. in optical bres [160].
Modulational instability can be viewed as the simplest case of the so-called symmetry-
breaking instability, when a solution of a non-linear system of a certain dimension (e.g.
a plane wave or soliton) is subjected to a broader class of perturbations. An example of the
same type of instability is a breakup of low-dimensional solitary waves under the action of
perturbations involving higher dimensions. This is a typical case of the so-called transverse
instability of plane solitary waves, rst discussed for long-wave solitons of the Korteveg
de Vries (KdV) equation [161] and envelope solitons of the NLS equation [162, 163], and
then investigated by different methods for a variety of non-linear soliton-bearing models
(see, for example, [164170]), with demonstrations in numerical simulations [171, 172].
Many of these results are now well-known in the soliton theory (see, for example, [173]).
In application to spatial optical solitons this type of soliton instability has attracted
special attention and discussion these days. First of all, this is because of the recent
progress in developing and employing non-linear optical materials, that led to several
important discoveries of self-focusing, self-trapped beam propagation, and solitary waves
in dierent non-linear materials, including photorefractive [17, 18], quadratic [23], and
saturable non-Kerr [21] media. It is expected that the development of novel band-gap
materials based on quadratic or cubic non-linearities will eventually lead to the experi-
mental observation and manipulation of the so-called light bullets, self-focused states of
light localized in both space and time, the building blocks of the future all-optical
photonics devices.
600
YU. S. Kivshar
Secondly, dierent types of symmetry-breaking instabilities have been recently described
theoretically and even observed experimentally in non-linear optics. This includes the
observation of breakup of bright soliton stripes in a bulk photorefractive medium due to
transverse modulational instability and formation of two-dimensional bright soliton
beams [174, 175]; the theoretical study [176179] and experimental observation [180, 181]
of generation of pairs of optical vortex solitons due to the transverse instability of dark-
soliton stripes; the theory and experimental demonstration of spatial modulational
instability in quadratically non-linear v
(2)
optical media [104, 182]; a decay of ring-shape
optical beams with non-zero angular momentum into higher-dimensional solitary waves,
observed experimentally in rubidium vapour (a saturable defocusing medium) [21] and
investigated theoretically as a general solitonic eect [183]. As a result of the development
of the soliton transverse instability, the (1 1)-dimensional soliton stripe decays either
into solitary waves of higher dimensions (e.g. (2 1)-dimensional bright or dark solitons
of radial symmetry) or it decays into radiation. For the case of quadratic solitons, both
such scenarios have been discussed theoretically [182] and also observed experimentally
[104].
8. Nonlinear theory of soliton instabilities
Linear stability analysis does not allow one to predict the subsequent evolution of unstable
(bright and dark) solitons. However, recently the so-called non-linear theory of soliton
instabilities has been suggested and demonstrated for several soliton bearing non-linear
models [53, 54, 107]. This theory is based on the multi-scale asymptotic technique and it
allows one to describe the soliton evolution near the stability threshold (i.e. the marginal
stability curve).
According to this approach, unstable bright solitons display three general scenarios of
their evolution [53], i.e. they diffract, collapse, or switch to a stable state with long-lived
oscillations of their amplitude. These oscillating solitons are possible due to the existence of
soliton internal modes [122124], one of the major properties that differentiate integrable
and non-integrable models. Unstable dark solitons display much simpler types of evolu-
tion which is always accompanied by emission of radiation which can generate new dark
solitons (soliton splitting). Generally, this leads to a drift of a dark soliton that becomes
`greyer' [54, 146].
In this section we present a brief summary of analytical results of the asymptotic theory
that describes not only linear instabilities but also the non-linear long-term evolution of
the unstable solitons. This approach is based on a non-trivial modication of the soliton
perturbation theory [33] near the instability threshold. We consider the instabilities of
bright solitons in a non-Kerr medium described by the generalized NLS equation 6 and
two-component bright solitons supported by resonant parametric interactions in a v
(2)
optical material, and also the drift instability of dark solitons in a non-Kerr defocusing
medium. All cases briey considered below are for the (1 1)-dimensional solitary waves,
but the analysis can be readily extended to cover similar problems for solitary waves of
higher dimensions.
8.1. Bright solitons in non-Kerr media
We consider the generalized NLS equation 6 where the function ) ([W[
2
) characterizes a
non-linear correction to the material refractive index. The model (equation 6) has been
intensively investigated in the context of the existence and stability of bright optical
601
Bright and dark spatial solitons
solitons which are stationary solutions of Equation 6 of the form
W
s
(xY :) = U(xY b) exp (ib:), where the real function U(xY b) vanishes for [x[ , and b is
the non-linearity-induced shift of the propagation constant. As has been already discussed
above, for generalized non-linearities spatial solitons may become unstable, and a standard
approach is to analyse the soliton instability using the linear stability analysis. For the case
of bright solitons of the generalized NLS equation 6, the soliton stability is given by the
VakhitovKolokolov criterion [132]. However, the linear stability analysis does not allow
one to understand the subsequent evolution of unstable solitons when the linearized
equations, describing exponentially growing perturbations on the soliton prole, become
invalid, and numerics is used for every particular problem.
The standard soliton perturbation theory [33] is usually applied to analyse the soliton
dynamics under the action of external perturbations. Here we deal with a qualitatively
dierent physical problem when an unstable bright soliton evolves under the action of its
`own' perturbations. As a result of the development of the instability, the soliton frequency
b varies in time, b = b(:). The linear stability of bright solitons is determined by the slope
of the derivative dP
s
adb, where P
s
(b) is the soliton energy
P
s
(b) =
1
2
_

U
2
(xY b) dx
Near the instability threshold, when the derivative P
/
s
(b) vanishes, the growth rate is small,
and we can assume that the instability-induced evolution of the perturbed soliton is, rst,
slow in : and, second, almost adiabatic (i.e. self-similar). Therefore, we can develop an
asymptotic theory representing the solution to the original model (Equation 6) as
W = /(xY bY Z) exp[ib
0
: i
_
1
0
X(Z
/
) dZ
/
[, where b = b
0

2
X(Z), Z = :, and 1. Here
the constant value b
0
is chosen near the critical value b
c
where the derivative dP
s
(b)adb
vanishes. Then, using the asymptotic multi-scale expansion for /(xY bY Z),
/ = U(xY b)
3
/
3
(xY bY Z) O(
4
), we obtain the following equation for the soliton
propagation constant X (details can be found in [53]),
M
s
b
c
( )
d
2
X
dZ
2

1

2
dP
s
db

b=b
0
X
1
2
d
2
P
s
db
2

b=b
c
X
2
= 0 (39)
where M
s
(b) is calculated through the stationary soliton solution,
M
s
=
_

1
U(xY b)
_
x
0
U(x
/
Y b)
dU x
/
Y b ( )
db
dx
/
_ _
2
dx b 0
Thus, the remarkable result of this asymptotic analysis is the following. In the gener-
alized NLS equation 6 the dynamics of solitons near the instability threshold can be
described by Equation 39 which is equivalent to the equation for motion of an eective
inertial and conservative particle of the mass M
s
(b
c
) and coordinate X moving under the
action of a potential force which is proportional to the difference P
0
P
s
(b), where
P
0
= P
s
(b
0
).
The rst two terms in Equation 39 cover the result of the linear stability analysis [132,
134, 135, 137], i.e. the soliton of Equation 6 is linearly unstable provided dP
s
(b
0
)adb ` 0.
The non-linear term in Equation 39 allows us to consider not only linear but also long-
term (non-linear) dynamics of unstable solitons and, moreover, to identify all scenarios of
the soliton instability dynamics.
602
YU. S. Kivshar
As an example, we consider the generalized NLS equation with two power-law non-
linearities, ) (I) = I
pa2
cI
p
, which possesses an explicit soliton solution for any p b 0
[55]. For 1 ` p ` 2 the function P
s
(b) always has a minimum, and the derivative
d
2
P
s
(b
c
)adb
2
is positive (see Fig. 9a). It is convenient to analyse the dynamical system 39
on the phase plane (XY

X) or, equivalently, on the plane (bY

b). When the value P
0
exceeds
the extremum value of P
s
(x
c
), the corresponding phase plane is presented in Fig. 9b.
The standard initial-value problem to the generalized NLS model (Equation 6) corre-
sponds to the initial condition for equation 39 lying on the axis

b(0) = 0. For such initial
conditions we reveal three dierent regimes of the soliton dynamics. They are all depicted
by the curves 1, 2 and 3 in Fig. 9a and b.
If the amplitude of the initially perturbed soliton is taken to be smaller than the am-
plitude of the (unstable) stationary solution (i.e. b(0) ` b
0
, curves 1 in Fig. 9a and b), the
instability leads to a decrease of b (proportional to the soliton amplitude), and this process
results in the soliton spreading into small-amplitude waves which nally decay into linear
waves due to diraction.
On the other hand, if the initially perturbed unstable soliton has the amplitude slightly
bigger than that of the stationary soliton solution (curves 2 in Fig. 9b), the exponentially
growing instability `pushes' the soliton into the stability region where there exists a stable
stationary soliton solution with b = b
f
, corresponding to the same value of the power
invariant P
0
(see Fig. 9a). However, due to the inertial nature of the soliton evolution, the
transition from an unstable to stable state is accompanied by the long-lived periodic
oscillations of the soliton amplitude. These oscillations can be explained by the existence
of a non-trivial internal mode of a bright soliton in the generalized NLS model (Equa-
tion 6), recently observed numerically and discussed analytically for the solitons of dif-
ferent types of non-linear response [122124, 129].
We note that, according to our analytical theory, the periodic oscillations of the soliton
amplitude must disappear for larger deviations from the stable equilibrium state
(b(0) b b
1
, curves 3 in Fig. 9a). In spite of the fact that larger deviations of the soliton
amplitude do not remove the soliton from the stability region, the evolution of pertur-
Figure 9 (a) Characteristic dependence of the soliton power P
s
(b) in the model (Equation 6) with
f (I) = I
pa2
cI
p
for 1 ` p ` 2. Minimum point b
c
corresponds to the instability threshold, b
0
and b
f
, to the
unstable and stable solitons at P
s
(b
0
) = P
0
= 1X96, respectively. (b) Phase plane (bY

b) of the asymptotic
model describing dierent regimes of the soliton dynamics: soliton decay (curves 1 and 3) and amplitude
oscillations (curve 2).
603
Bright and dark spatial solitons
bations is predicted to lead again to decreasing of the soliton amplitude. As soon as the
soliton enters the unstable region, it nally spreads out due to diraction.
All predicted regimes of the soliton instability have been found in numerical simulations
[54]. Figure 10 shows several examples of the soliton evolution for the soliton propagation
constant b(:) determined as the rst-order derivative of the soliton phase calculated
numerically at the pulse maximum. The results are in an excellent agreement with the
predictions of the analytical model (Equation 39) (cf. curves in Fig. 10 with the trajectories
on the phase plane shown in Fig. 9b).
8.2. Two-wave solitons in v
(2)
media
As has been mentioned above in Section 2.4, the NLS equation becomes invalid near
resonances with the higher harmonics excited when the phase matching condition is sat-
ised. In a medium with quadratic response, non-trivial eects can be observed already for
two interacting waves due to parametric two-wave mixing in a diractive medium. In
Section 4, we have discussed dierent types of two-wave parametric solitons which have
been recently observed experimentally in non-linear planar waveguides [103]. Here we
discuss briey the stability of bright two-component solitons in such systems.
We recall that resonant interaction between the fundamental and second harmonics in a
diractive v
(2)
medium can be described by the coupled equations for the dimensionless
variables (see also Section 4.1)
i
dt
d:

d
2
t
dx
2
t t
+
w = 0
ir
dw
d:

d
2
w
dx
2
aw
1
2
t
2
= 0
(40)
where : is the propagation distance, x stands for the beam transverse coordinate. The
parameter r describes the ratio of the wave vectors (for spatial solitons, r ~ 2), and we
consider it positive. The parameter a can be presented as a = 2r D, where D is pro-
portional to the wave vector mismatch Dk = k
2
2k
1
between the harmonics.
Figure 10 Results of numerical simulations of the soliton instabilities in the generalized NLS equation with
f (I) = I
pa2
cI
p
at p = 1X35. (a) Initial conditions are selected near the unstable soliton with b = b
0
. (b) Initial
conditions are selected near the stable soliton with b = b
f
.
604
YU. S. Kivshar
Diraction described by the second-order derivatives in Equations 40 is crucial for the
existence of two-component soliton solutions. The stationary soliton solutions of Equa-
tions 40 are presented by real functions t = t
0
(xY a) and w = w
0Ya
(x) which are independent
of :. For the particular value a = 1 these solutions were rst found by Karamzin and
Sukhorukov [92]. Numerical analysis done in [95, 96] revealed the existence of the soliton
solutions of system 40 for any positive value of a.
To analyse the stability of the soliton solutions t
0
and w
0
, we linearize Equations 40 on
the soliton background as the following,
t(xY :) = t
0
(x) J
r
(x) iJ
i
(x) [ [ exp(k:) w(xY :) = w
0
(x) W
r
(x) iW
i
(x) [ [ exp(2k:)
and investigate the corresponding linear eigenvalue problem for the corrections (J
r
Y J
i
)
and (W
r
Y W
i
). Similar to the case of bright NLS solitons discussed above in Section 8.1, this
problem can be solved by the asymptotic method for non-zero but small k. Near the in-
stability threshold, such solutions are expected to exist only for special values of the
parameters a near the critical curve a = a
c
(r). The instability threshold, as well as the
general dependence k(aY r), can be found from the corresponding solvability conditions to
the linear problem. This analysis was rst presented in [107], and it was shown that the
instability threshold is given by the generalized VakhitovKolokolov criterion,
dQadb = 0, where b(aY r) = (2r a)
1
is the renormalized soliton propagation constant
and
Q =
1
2(2r a)
3a2
_

t
2
0
(xY a) 2rw
2
0
(xY a)
_
dx
is the renormalized power (MenleyRowe) invariant of Equations 40. In all these formulas
the parameter r is considered as an arbitrary parameter. Using this criterion and the
numerical results on the stationary soliton solutions, the instability threshold curve
a = a
c
(r) has been calculated in [107]. Moreover, it has been shown that the two-wave
parametric solitons are unstable for a ` a
c
.
The important physical question is the development of linear instability in the subse-
quent dynamics of the two-wave solitons. In order to describe this analytically, we should
take into account non-linear eects. We select a = a
0
close to a
c
, so that the small
parameter characterizes the deviation (a
0
a
c
) ~ O(
2
). Then, it follows from the linear
theory that the growth rate has the order O() and, therefore, the unstable linear per-
turbations grow on the `slow' scale Z = :. This allows us to introduce the slowly varying
complex phase S = S(Z) and look for the perturbed solutions of Equations 40 in the form
of asymptotic series t = [t
0
(xY a)
2
XJ
r
(xY a) O(
3
)[ exp(iS), w = [w
0
(xY a)

2
XW
r
(xY a) O(
3
)[ exp(2iS), where the functions J
r
and W
r
are the solutions of the
linearized problem, and X = (2r a) dSadZ describes a correction to the soliton propa-
gation constant. Using the asymptotic multi-scale technique we nd the non-linear
equation for the function X,
M
s
a
c
( )
d
2
X
dZ
2

1

2
dQ
db

a=a
0
X
1
2
d
2
Q
db
2

a=a
c
X
2
= 0 (41)
where the renormalized soliton mass is dened as
M
s
= (2r a)
1a2
_

w
2
0
W
i
w
0
_ _
2
x
t
2
0
J
i
t
0
_ _
2
x

1
2w
0
t
0
W
i
2w
0
J
i
( )
2
_ _
dx b 0
605
Bright and dark spatial solitons
where the index x stand for the corresponding partial derivatives, and the functions J
i
and
W
i
are the solutions of the linearized problem. We mention that the derivative
d
2
Qadb
2
[
a=a
c
is always positive for the two-wave parametric solutions.
It turns out that the asymptotic theory applied to the two-wave v
(2)
solitons gives, in a
renormalized form, essentially the same equation of motion of the equivalent particle as
Equation 39 for the (one-component) bright solitons in non-Kerr optical materials. As we
discussed above, this equation describes two main scenarios of the instability of bright
solitons, either long-lived periodic amplitude oscillations or soliton decay.
Indeed, according to Equation 41, the exponential growth of linear perturbations with
X(0) b 0 is stabilized by non-linearity leading to oscillations around a novel stable equi-
librium state. This equilibrium state corresponds to a stationary soliton which can also be
described by Equations 40 but for a renormalized parameter a lying inside the stability
region. Therefore, for a slightly increased amplitude of an unstable soliton our analytical
model (Equation 41) predicts in-phase pulsations of the fundamental and second har-
monics around a novel stable soliton, and this exactly corresponds to the evolution
observed numerically, see Fig. 11. For X(0) ` 0, according to Equation 41, such a sta-
bilization is not possible and, as a result, a slightly decreased amplitude of an unstable
soliton gradually decreases further.
Figure 11 Periodic oscillations of the two-wave soliton of the model 40 after the development of instability
(r = 2Y a = 0X05). (a) Evolution of the harmonics amplitudes v
m
= [v(0Y z)[ (solid) and w
m
= [w(0Y z)[ (da-
shed). (b), (c) Propagation of the soliton components v and w, respectively.
606
YU. S. Kivshar
8.3. Dark solitons in non-Kerr media
The asymptotic analysis presented above for one- and two-component bright solitons has
been also developed for dark solitons [54]. Here we consider the dark solitons described by
the generalized NLS equation 6 and derive an asymptotic equation governing their non-
linear dynamics near the instability threshold.
The dark soliton of Equation 6 can be presented in the form,
W
s
= U
s
(nY tY q) exp(i) (q):) U
s
= Uexp(ih)
where n = x t:, dhadn =
1
2
t(1 qaU
2
), and the real function U(nY tY q) tends to the non-
zero boundary conditions at innity, U
2
q as [n[ . We consider, for simplicity, the
case when U
2
` q for any n. Thus, the dark soliton depends on two parameters, the
velocity t and the background intensity q. For t c, where c = [2q)
/
(q)[
1a2
is the phase
velocity of linear waves propagating on the constant background [W[
2
= q, the dark sol-
iton has small amplitude, while for t 0 the soliton amplitude (or `darkness') reaches its
maximum (but always I
min
b 0).
In spite of the fact that the background intensity is determined by the boundary con-
ditions, its local variations are still possible, e.g. due to radiation emitted by the perturbed
dark soliton during its non-stationary evolution. Therefore, the dynamics of unstable dark
solitons is much more complicated than that of bright solitons. However, we show below
that even in this case, but near the instability threshold, the effective analytical model for
the non-linear dynamics of the dark solitons can be derived with the help of the asymptotic
method.
As has been already mentioned in Section 7.1.2, the renormalized eld momentum
dened as
M
s
(tY q) = 2 Im
_

dW
+
s
dn
W
s
1
q
W
s
[ [
2
_ _
dn = t
_

U
2
q
_ _
2
U
2
dn (42)
plays a key role in the formulation of the variational principle for dark solitons, and also it
denes the criterion for their linear instability, given by the slope of the derivative dM
s
adt.
Thus, both linear and non-linear stability analysis for dark solitons is based on the
renormalized momentum (Equation 42).
Following the approach outlined above for bright solitons, we present the perturbed
dark soliton of the generalized NLS equation 6 in the form W = /(nY tY qY XY Z)
exp[i) (q): R(XY Z)[, where n = x t
0
:
_
Z
0
J (Z
/
) dZ
/
, t = t
0
J (Z), X = x, Z = :,
and 1. The constant value t
0
stands for the velocity of the (unstable) unperturbed
dark soliton which is selected near the critical value t
c
where the derivative dM
s
adt van-
ishes. Then, we expand / in the asymptotic series,
/ = U
s
(nY tY q)
2
/
2
(nY tY qY XY Z) O
3
_ _
and reduce the generalized NLS equation 6 to a sequence of linear inhomogeneous
equations, similar to the case of bright solitons. However, the important dierence be-
tween the asymptotic procedure for bright and dark solitons is the existence of the linear
correction term to the dark soliton which does not vanish at innity, (1a
2
)([/[
2
1) u

as n . This non-localized part of /


2
in the region of the soliton [n ~ O(1)[ indicates
the generation of the radiation eld escaping the perturbed dark soliton. This radiation
eld can be found as a solution to the generalized NLS equation 6 in the outer region
X ~ O(1) where [W[
2
= q
2
U

r
(XY Z) O(
3
). Analysis shows that, in the leading order,
607
Bright and dark spatial solitons
the radiation propagates to the right and to the left with the limiting phase velocity c,
U

r
= U

r
(X (cZ). At the more slower time scale, the evolution of the radiation eld
obeys the KdV equations [179],
(4c
dU

r
df
aU

r
dU

r
dX

d
3
U

r
dX
3
= 0 (43)
where f =
3
: and a = 4[3)
/
(q) q)
//
(q)[. Then, the matching conditions produce the
prole of the radiation eld generated by the perturbed dark soliton, U

r
= u

as
n , X 0, and Z = Xat
0
.
After such a technical procedure, in the third-order approximation, we can derive the
following dierential equation for J (Z)
l
s
t
c
( )
dJ
dZ

1

dM
s
dt

t=t
0
J
1
2
d
2
M
s
dt
2

t=t
c
J
2
= 0 (44)
where the positive coecient l
s
(tY q) has the form,
l
s
(tY q) =
q
c
dS
s
dt
_ _
2

2c
q
dP
s
dt
_ _
2
b 0
Here
S
s
(tY q) = t
_

1
q
U
2
_ _
dn
and
P
s
(tY q) =
1
2
_

U
2
q
_ _
dn
are proportional to the total phase shift across the dark soliton and its complementary
power, respectively. Thus, due to the radiation the dark solitons near the instability
threshold can be described by an equation for motion of an eective inertial and dissipative
particle of mass l
s
and velocity J under the action of the frictional force which is pro-
portional to the difference M
0
M
s
(t), where M
0
= M
s
(t
0
).
As follows from Equation 44, the dark soliton becomes unstable for dM
s
adt[
t=t
0
` 0
and the instability dynamics essentially depends on the sign of the perturbation. Usually,
the instability is observed for small velocities so that d
2
M
s
adt
2
[
t=t
c
b 0. Therefore, if we
decrease the intensity of the unstable dark soliton, the instability results in a monotonic
transition from an unstable to a stable soliton realized at the same value of the renor-
malized momentum M
0
. This scenario of the soliton instability is described asymptotically
by the following solution of Equation 44,
J =
J
0
J
f
J
f
J
0
( ) exp(kZ) J
0
(45)
where J
0
is the initial velocity of the dark soliton, and the instability growth rate k and the
nal soliton velocity J
f
are dened through the coecients of Equation 44. For the other
sign of the perturbation, the instability cannot be described by bounded solutions. This
implies that an increase of the intensity of the dark soliton cannot be suppressed by weak
non-linear eects described by Equation 44, and it can lead to an essential transformation
of the unstable dark soliton. As a result of this transformation, the intensity of the dark
608
YU. S. Kivshar
soliton may fall to zero at the minimum point with the subsequent evolution depending on
the global non-linear properties of the particular case of the generalized NLS model.
The radiation eld excited due to the transition of the dark soliton from unstable to
stable states can be also found,
U

r
= (
kDJ
4c(c (t)
q
2
dS
s
dt
c
dP
s
dt
_ _
sech
2
(kXa2t)
Since the radiation eld develops into a soliton of the KdV equation 43 only for negative
amplitude, the secondary dark soliton is formed in the wave U

r
which propagates in the
opposite direction with respect to the original perturbed soliton. On the other hand, the
radiation in a co-propagating wave U

r
decays into quasi-linear dispersive waves.
To conrm the analytical theory of the dark soliton instability, the generalized NLS
equation 6 with the non-linearity in the form, ) (I) = I cI
2
was integrated numerically
[54]. Figure 12 presents two scenarios of the instability development, namely, the decay of
the unstable dark soliton into two stable solitons of smaller amplitudes which propagate in
the opposite directions (Fig. 12a) and the `collapse'-type dynamics of a dark soliton with a
formation of two kink-like fronts (Fig. 12b). Thus, the basic predictions of the asymptotic
analytical theory have been shown to be in good agreement with the results of the
numerical simulations.
9. Concluding remarks
We have discussed dierent aspects of the theory of spatial optical solitons. From the
mathematical point of view, these kinds of non-linear waves demonstrate a number of very
Figure 12 Evolution of an unstable dark soliton in the model 6 with f (I) = I cI
2
. (a) Transformation to a stable
dark soliton via decay of an unstable soliton into two stable ones. (b) Collapse of an unstable dark soliton into
two kink-type fronts. Top: initial (solid) and nal (dashed) states; bottom: the corresponding contour plots.
609
Bright and dark spatial solitons
interesting properties which seem generic to spatially localized solutions of non-integrable
non-linear models, e.g. the existence of soliton instabilities, the long-lived oscillations of
the soliton amplitude due to the soliton internal modes, etc. From the physical point of
view, it seems important that in many cases spatial optical solitons cannot only be
described analytically and numerically for a variety of nonlinear models, but, and this is
the most amazing fact, they can also be veried, with relatively good accuracy, by direct
measurements of the beam propagation through non-linear media. This suggests the
validity of dierent phenomenological models for describing self-guided beams in optical
media with non-resonant and resonant non-linearities.
We would like to emphasize that for many years the properties of non-linear waves
in non-linear models of dierent physical context have been described by reducing the
original and complex model equations to integrable ones. The concept of spatial sol-
itons does not allow, in principle, such a simplication. As a result, in presenting the
analytical results, we have tried to avoid the traditional restrictions associated with
consideration of only the cases of integrable models (i.e. the cubic NLS equation for
scalar solitons, or the Manakov equations for vector solitons) and their deformations,
or even analytically solvable models, as has been done in many previous review papers
and books on optical solitons. Instead, we have emphasized the physics of the
underlying phenomena and general concepts associated with more realistic (and usually
non-integrable) physical models. This approach has involved discussion of soliton
stability and instability-induced dynamics of non-Kerr solitons, since in non-integrable
models solitary waves can become unstable. We believe that the concept of spatial
optical solitons gives us one of the examples where, even dealing with a variety of
non-integrable models, we are able to develop more general physical approaches and
analytical tools for explaining, at least qualitatively, a number of eects observed in
experiments.
As for the future application of spatial solitons, many problems involving the soliton-
based concept of light guiding light are still not solved experimentally, and this denitely
will require more eort to make this concept a practical reality. Searching and employing
new materials with strong non-linear properties may suciently speed up this process.
However, even now we can say that the rapid progress of the physics of spatial optical
solitons has already resulted in the development of novel mathematical techniques and
conceptual approaches to deal with solitary waves of non-integrable systems, and these
tools are expected to be useful for many other elds where non-linear properties of wave
propagation become important.
Acknowledgements
I would like to thank Professor G. Stegeman for his kind invitation to present my view of
the recent advances in the theory of non-Kerr spatial solitons in this special issue. Dierent
parts of this paper are based on the original research I have been carrying out during the
last years in collaboration with my colleagues and PhD students, including Dr V. V.
Afanasjev, Dr O. Bang, Dr A. V. Buryak, Dr D. E. Pelinovsky, Professor R. A. Sammut,
Mrs V. V. Steblina, and Dr S. Trillo. I thank them all for a very productive collaboration.
I also wish to acknowledge many useful and encouraging discussions with Professor
B. Luther-Davies, Professor M. Segev, and Professor A. W. Snyder, who contributed
enormously into my understanding of the properties of self-guided beams in non-Kerr
optical media.
610
YU. S. Kivshar
This work was supported by the Australian Photonics Cooperative Research Centre, the
Australian Research Council, the Department of Industry, Sciences, and Tourism, the
Australian Academy of Sciences, and the Research School of Physical Sciences and
Engineering, through research projects and research grants on nonlinear waveguides,
spatial optical solitons, and light guiding light.
References
1. A. W. SNYDER and J. D. LOVE, Optical Waveguide Theory (Chapman and Hall, New York and London,
1985).
2. R. Y. CHIAO, E. GARMIRE and C. H. TOWNES, Phys. Rev. Lett. 13 (1964) 479.
3. G. I. STEGEMEN, Opt. Appl. 26 (1996) 240.
4. V. E. ZAKHAROV and A. B. SHABAT, Zh. Eksp. Teor. Fiz. 61 (1971) 118 [Sov. Phys. JETP 34 (1972) 62].
5. P. L. KELLEY, Phys. Rev. Lett. 15 (1965) 1005.
6. J. J. RASMUSSEN and K. RYPDAL, Phys. Scripta 33 (1986) 481.
7. J. H. MARBURGER and E. L. DAWES, Phys. Rev. Lett. 21 (1968) 556.
8. E. L. DAWES and J. H. MARBURGER, Phys. Rev. 179 (1969) 862.
9. P. K. KAW, K. NISHIKAWA, Y. YOSHIDA and A. HASEGAWA, Phys. Rev. Lett. 35 (1975) 88.
10. J. Z. WILCOX and T. J. WILCOX, Phys. Rev. Lett. 34 (1975) 1160.
11. G. I. STEGEMAN, D. J. HAGAN and L. TORNER, J. Opt. Quantum Electron. 28 (1996) 1691.
12. A. W. SNYDER, D. J. MITCHELL and YU. S. KIVSHAR, Mod. Phys. Lett. B 9 (1995) 1479.
13. G. A. ASKAR'YAN, Sov. Phys. JETP 15 (1962) 1088 [Zh. Eksp. Teor. Fiz. 42 (1962) 1567].
14. S. A. AKHMANOV, A. P. SUKHORUKOV and R. V. KNOKHLOV, Sov. Phys. JETP 23 (1966) 1025 [Zh.
Eksp. Teor. Fiz. 50 (1966) 1537].
15. J. T. MANASSAH, Opt. Lett. 16 (1991) 587.
16. A. W. SNYDER and YU. S. KIVSHAR, J. Opt. Soc. Am. B 14 (1997) 3025.
17. G. C. DUREE, J. L. SHULTZ, G. J. SALAMO, M. SEGEV, A. YARIV, B. CROSIGNANI, P. DIPORTO, E. J.
SHARP and R. R. NEURGAONKAR, Phys. Rev. Lett. 71 (1993) 533.
18. M. SEGEV, G. C. VALLEY, B. CROSIGNANI, P. DIPORTO and A. YARIV, Phys. Rev. Lett. 73 (1994) 3211.
19. M. SHIH, P. LEACH, M. SEGEV, M. H. GARRETT, G. SALAMO and G. C. VALLEY, Opt. Lett. 21 (1996)
324.
20. M. D. ITURBE-CASTILLO, P. A. MARQUEZ AGUILAR, J. J. SANCHEZ-MONDRAGON, S. STEPANOV
and V. VYSLOUKH, Appl. Phys. Lett. 64 (1994) 408.
21. V. TIKHONENKO, J. CHRISTOU and B. LUTHER-DAVIES, J. Opt. Soc. Am. B 12 (1995) 2046.
22. V. TIKHONENKO, J. CHRISTOU and B. LUTHER-DAVIES, Phys. Rev. Lett. 76 (1996) 2698.
23. W. E. TORRUELLAS, Z. WANG, D. J. HAGAN, E. W. VANSTRYLAND, G. I. STEGEMAN, L. TORNER
and C. R. MENYUK, Phys. Rev. Lett. 74 (1995) 5036.
24. N. N. AKHMEDIEV, Opt. Quantum Electron. 30 (1998) 535.
25. S. A. AKHMANOV, A. P. SUKHORUKOV and R. V. KHOKHLOV, Usp. Fiz. Nauk 93 (1967) 19 [Sov. Phys.
Uspekhi 10 (1968) 609].
26. O. SVELTO, Progress in Optics, Vol. XII, edited by E. Wolf (North-Holland, Amsterdam, 1974).
27. A. D. BOARDMAN and K. XIE, Radio Sci. 28 (1993) 891.
28. R. Y. CHIAO, I. H. DEUTSCH, J. C. GARRISON and E. M. WRIGHT, Frontiers in Nonlinear Optics (The
Sergei Akhmanov Memorial Volume) edited by H. Walter et al. (IOP, Bristol, 1993) p. 151.
29. V. E. ZAKHAROV and A. B. SHABAT, Zh. Eksp. Teor. Fiz. 64 (1973) 1627 [Sov. Phys. JETP 37 (1973) 823].
30. V. E. ZAKHAROV, S. V. MANAKOV, S. P. NOVIKOV and L. P. PITAEVSKI, Theory of Solitons: The Inverse
Scattering Transform (Nauka, Moscow, 1980) [English translation by Consultant Bureau, New York, 1984].
31. A. HASEGAWA, Solitons in Optical Fibers (Springer, Berlin, 1989).
32. A. HASEGAWA and Y. KODAMA, Solitons in Optical Communications (Oxford University Press, Oxford,
1995).
33. YU. S. KIVSHAR and B. A. MALOMED, Rev. Mod. Phys. 61 (1989) 763.
34. B. J. EGGLETON, C. M. DE STERKE and R. E. SLUSHER, J. Opt, Soc. Am. B 14 (1997) 2980.
35. A. KOZHEKIN and G. KURIZKI, Phys. Rev. Lett. 74 (1995) 5020.
36. Y. SILBERBERG, Opt. Lett. 15 (1990) 1282.
37. V. E. ZAKHAROV, V. E. SOBOLEV and V. S. SYNAKH, Zh. Eksp. Teor. Fiz. 60 (1971) 136 [Sov. Phys. JETP
33 (1971) 77].
611
Bright and dark spatial solitons
38. V. E. ZAKHAROV and V. S. SYNAKH, Zh. Eksp. Teor. Fiz. 68 (1975) 940 [Sov. Phys. JETP 41 (1975) 465].
39. A. E. KAPLAN, Phys. Rev. Lett. 78 (1985) 1291.
40. A. E. KAPLAN, IEEE J. Quantum Electron. 21 (1985) 1538.
41. R. H. ENNS and L. J. MULDER, Opt. Lett. 14 (1989) 509.
42. L. J. MULDER and R. H. ENNS, IEEE J. Quantum Electron. 25 (1989) 2205.
43. S. GARTZ and J. HERRMANN, J. Opt. Soc. Am. B 8 (1991) 2296.
44. S. GARTZ and J. HERRMANN, Opt. Lett. 17 (1992) 484.
45. J. HERRMANN, Opt. Commun. 91 (1992) 337.
46. F. G. BASS, V. V. KONOTOP and S. A. PUZENKO, Phys. Rev. A 46 (1992) 4185.
47. A. W. SNYDER and A. P. SHEPPARD, Opt. Lett. 18 (1993) 499.
48. W. KROLIKOWSKI and B. LUTHER-DAVIES, Opt. Lett. 17 (1992) 1414.
49. W. KROLIKOWSKI and B. LUTHER-DAVIES, Opt. Lett. 18 (1993) 188.
50. W. KROLIKOWSKI, N. N. AKHMEDIEV and B. LUTHER-DAVIES, Phys. Rev. E 48 (1993) 3980.
51. G. C. VALLEY, M. SEGEV, B. CROSIGNANI, A. YARIV, M. M. FEJER, M. C. BASHAW, Phys. Rev. A 50
(1994) R4457.
52. D. N. CHRISTODOULIDES and M. I. CARVALHO, J. Opt. Soc. Am. B 12 (1995) 1628.
53. D. E. PELINOVSKY, V. V. AFANASJEV, and YU. S. KIVSHAR, Phys. Rev. E 53 (1996) 1940.
54. D. E. PELINOVSKY, YU. S. KIVSHAR and V. V. AFANASJEV, Phys. Rev. E 54 (1996) 2015.
55. R. W. MICALLEF, V. V. AFANASJEV, YU. S. KIVSHAR and J. D. LOVE, Phys. Rev. E 54 (1996) 2936.
56. P. ROUSSIGNOL, D. RICARD, J. LUKASIK and C. FLYTZANIS, J. Opt. Soc. Am. B 4 (1987) 5.
57. L. H. ACIOLI, A. S. L. GOMES, J. M. HICKMANN and C. B. DE ARAUJO, Appl. Phys. Lett. 56 (1990) 2279.
58. F. LEDERER and W. BIEHLIG, Electron. Lett. 30 (1994) 1871.
59. B. LAWRENCE, M. CHA, W. E. TORRUELLAS, G. I. STEGEMAN, S. ETEMAN, G. BAKER and F. KAJ-
ZER, Appl. Phys. Lett. 64 (1994) 2773.
60. B. LAWRENCE, W. E. TORRUELLAS, M. CHA, M. L. SUNDHEIMER, G. I. STEGEMAN, J. METH, S.
ETEMAN and G. BAKER, Phys. Rev. Lett. 73 (1994) 597.
61. C. JOSSERAND and S. RICA, Phys. Rev. Lett. 78 (1997) 1215.
62. J. L. COUTAZ and M. KULL, J. Opt. Soc. Am. B 8 (1991) 95.
63. T. K. GUSTAFSON, P. L. KELLEY, R. Y. CHIAO and R. G. BREWER, Appl. Phys. Lett. 12 (1968) 165.
64. J. D. REICHERT and W. G. WAGNER, IEEE J. Quantum. Electron. QE-4 (1968) 221.
65. I. BIALYNICKI-BIRULA and J. MYCIELSKI, Phys. Scripta 20 (1979) 539.
66. A. W. SNYDER and D. J. MITCHELL, Opt. Lett. 22 (1997) 16.
67. A. W. SNYDER and D. J. MITCHELL, Science 276 (1997) 1538.
68. J. D. GIBBON, Soliton Theory: A Survey of Results, edited by A. P. FORDY (Manchester University Press,
Manchester, 1990) pp. 133151.
69. A. JEFFREY and T. KAWAHARA, Asymptotic Methods in Nonlinear Wave Theory (Pitman, London, 1982).
70. T. TANIUTI and K. NISHIHARA, Nonlinear Waves (Pitman, Boston, 1983).
71. V. E. ZAKHAROV, Sov. Phys. JETP 72 (1972) 908.
72. M. FEIT and J. FLECK, J. Opt. Soc. Am. B 5 (1988) 633.
73. G. FIBICH, Phys. Rev. Lett. 76 (1996) 4356.
74. A. L. BERKHOER and V. E. ZAKHAROV, Sov. Phys. JETP 31 (1970) 486 [Zh. Eksp. Teor. Fiz. 58
(1970) 903].
75. S. V. MANAKOV, Sov. Phys. JETP 38 (1974) 248 [Zh. Eksp. Teor. Fiz. 65 (1973) 505].
76. D. N. CHRISTODOULIDES and R. I. JOSEPH, Opt. Lett. 13 (1988) 53.
77. M. V. TRATNIK and J. E. SIPE, Phys. Rev. A 38 (1988) 2011.
78. S. TRILLO, S. WABNITZ, E. M. WRIGHT and G. STEGEMAN, Opt. Lett. 13 (1988) 871.
79. A. W. SNYDER, S. HEWLETT and D. J. MITCHELL, Phys. Rev. Lett. 72 (1994) 1012.
80. M. SHALABY and A. J. BARTHELEMY, IEEE J. Quantum Electron. 28 (1992) 2736.
81. J. U. KANG, G. I. STEGEMAN, J. S. AITCHISON and N. AKHMEDIEV, Phys. Rev. Lett. 76 (1996) 3699.
82. N. N. AKHMEDIEV, V. M. ELEONSKY, N. E. KULAGIN and L. P. SHIL'NIKOV, Pis'ma Zh. Tekh. Fiz. 15
(1989) 19 [Sov. Tech. Phys. Lett. 15 (1989) 587].
83. M. HAELTERMAN and A. P. SHEPPARD, Phys. Rev. E 49 (1994) 3376.
84. YU. S. KIVSHAR and B. LUTHER-DAVIES, Phys. Rep. 298 (1998) 81.
85. A. W. SNYDER, D. J. MITCHEL, and B. LUTHER-DAVIES, J. Opt. Soc. Am. B 10 (1993) 2341
86. V. L. GINZBURG and L. P. PITAEVSKI, Sov. Phys. JETP 7 (1959) 858 [Zh. Eksp. Teor. Fiz. 34 (1958)
1240]
612
YU. S. Kivshar
87. L. P. PITAEVSKI, Sov. Phys. JETP 13 (1961) 451 [Zh. Eksp. Teor. Fiz. 40 (1961) 646]
88. A. W. SNYDER, L. POLADIAN and D. J. MITCHELL, Opt. Lett. 17 (1992) 789.
89. A. W. SNYDER, D. J. MITCHELL and YU. S. KIVSHAR, Physics and Applications of Optical Solitons in
Fibers, edited by A. Hasegawa (Kluwer Academic Press, Amsterdam, 1996) p. 263.
90. D. L. MILLS, Nonlinear Optics: Basic Concepts (Springer-Verlag, Berlin, 1991).
91. R. W. BOYD, Nonlinear Optics (Academic Press, San Diego, 1992).
92. YU. N. KARAMZIN and A. P. SUKHORUKOV, Zh. Eksp. Teor. Fiz. 68 (1975) 834 [Sov. Phys. JETP 41
(1976) 414].
93. A. A. KANASHOV and A. M. RUBENCHIK, Physica D 4 (1981) 122.
94. R. SCHIEK, J. Opt. Soc. Am. B 10 (1993) 1848.
95. A. V. BURYAK and YU. S. KIVSHAR, Opt. Lett. 19 (1994) 1612.
96. A. V. BURYAK and YU. S. KIVSHAR, Opt. Lett. 20 (1995) 1080.
97. L. TORNER, C. R. MENYUK and G. I. STEGEMAN, Opt. Lett. 19 (1994) 1615.
98. A. V. BURYAK and YU. S. KIVSHAR, Phys. Lett. A 197 (1995) 407.
99. L. TORNER, C. R. MENYUK, W. E. TORRUELLAS and G. I. STEGEMAN, Opt. Lett. 20 (1995) 13.
100. A. V. BURYAK, YU. S. KIVSHAR and V. V. STEBLINA, Phys. Rev. A 52 (1995) 1670.
101. A. V. BURYAK, YU. S. KIVSHAR and S. TRILLO, Phys. Rev. Lett. 77 (1996) 5210.
102. A. V. BURYAK and YU. S. KIVSHAR, Phys. Rev. Lett. 78 (1997) 3286.
103. R. SCHIEK, Y. BAEK and G. I. STEGEMAN, Phys. Rev. E 53 (1996) 1138.
104. R. A. FURST, D. M. BABOIU, B. L. LAWRENCE, W. E. TORRUELLAS, G. I. STEGEMAN, S. TRILLO and
S. WABNITZ, Phys. Rev. Lett. 78 (1997) 2756.
105. P. FERRO and S. TRILLO, Phys. Rev. E 51 (1995) 4994.
106. A. V. BURYAK and YU. S. KIVSHAR, Phys. Rev. A 51 (1995) R41.
107. D. E. PELINOVSKY, A. V. BURYAK and YU. S. KIVSHAR, Phys. Rev. Lett. 75 (1995) 591.
108. A. D. BOARDMAN, K. XIE and A. SANGARPAUL, Phys. Rev. A 52 (1995) 4099.
109. H. HE, M. J. WERNER and P. D. DRUMMOND, Phys. Rev. E 54 (1996) 896.
110. A. V. BURYAK, Phys. Rev. E 52 (1995) 1156.
111. M. J. WERNER and P. D. DRUMMOND, J. Opt. Soc. Am. B 10 (1993) 2390.
112. K. A. GORSHKOV and L. A. OSTROVSKY Physica D 3 (1981) 428.
113. K. HAYATA and M. KOSHIBA, Phys. Rev. A 50 (1994) 675.
114. S. SALTIEL, S. TANEV and A. D. BOARDMAN, Opt. Lett. 22 (1997) 148.
115. R. A. SAMMUT, A. V. BURYAK and YU. S. KIVSHAR, Opt. Lett. 22 (1997) 1385.
116. R. A. SAMMUT, A. V. BURYAK and YU. S. KIVSHAR, J. Opt. Soc. Am. B 15 (1998) 1488.
117. I. V. TOMOV and M. C. RICHARDSON, IEEE J. Quantum Electron. 12 (1976) 521.
118. N. A. ANSARI, R. A. SAMMUT and H. T. TRAN, J. Opt. Soc. Am. B 13 (1996) 1419.
119. Y. CHEN, Phys. Rev. A 50 (1994) 5145.
120. K. HAYATA, Y. UEHIRA and M. KOSHIBA, Phys. Rev. A 51 (1995) 1549.
121. A. W. SNYDER, D. J. MITCHELL and A. V. BURYAK, J. Opt. Soc. Am. B 13 (1996) 1146.
122. D. E. PELINOVSKY, YU. S. KIVSHAR and V. V. AFANASJEV, Physica D 116 (1998) 121.
123. V. V. AFANASJEV, P. K. CHU and YU. S. KIVSHAR, Opt. Lett. 22 (1997) 1388.
124. YU. S. KIVSHAR, D. E. PELINOVSKY, T. CRETEGNY and M. PEYRARD, Phys. Rev. lett. 80 (1998) 5032.
125. D. K. CAMPBELL, J. F. SCHONFELD and C. A. WINGATE, Physica D 9 (1983) 1.
126. M. PEYRARD and D. K. CAMPBELL, Physica D 9 (1983) 33.
127. D. K. CAMPBELL, M. PEYRARD and P. SODANO, Physica D 19 (1986) 165.
128. A. W. SNYDER, S. HEWLETT and D. J. MITCHELL, Phys. Rev. E 51 (1995) 6297.
129. C. ETRICH, U. PESCHEL, F. LEDERER, B. A. MALOMED and YU. S. KIVSHAR, Phys. Rev. E 54 (1996)
4321.
130. D. J. KAUP, Phys. Rev. A 42 (1990) 5689.
131. L. D. LANDAU and E. M. LIFSHITZ, Quantum Mechanics (Pergamon Press, New York, 1977).
132. M. G. VAKHITOV and A. A. KOLOKOLOV, Radiophys. Quantum Electron 16 (1973) 783 [Izv. Vyssh. Uch.
Zav. Radioz. 16 (1973) 1020].
133. E. A. KUZNETSOV, A. M. RUBENCHIK and V. E. ZAKHAROV, Phys. Rep. 142 (1986) 103.
134. M. I. WEINSTEIN, Comm. Pure. Appl. Math. 39 (1986) 51.
135. M. I. WEINSTEIN, SIAM J. Math. Anal. 16 (1985) 472.
136. F. V. KUSMARTSEV, Phys. Rep. 183 (1989) 2.
137. D. J. MITCHELL and A. W. SNYDER, J. Opt. Soc. Am. B 10 (1993) 1572.
613
Bright and dark spatial solitons
138. C. K. R. T. JONES and J. MOLONEY, Phys. Lett. A 117 (1986) 175.
139. D. HART and E. M. WRIGHT, Opt. Lett. 17 (1992) 121.
140. I. V. BARASHENKOV and KH. T. KHOLMURODOV, JINR Preprint P17-86-698, Dubna (in Russian).
141. I. V. BARASHENKOV and V. G. MAKHANKOV, Phys. Lett. A. 128 (1988) 52.
142. I. V. BARASHENKOV, A. D. GOSHEVA, V. G. MAKHANKOV and I. V. PUZYNIN, Physica D 34 (1989)
240.
143. M. M. BOGDAN, A. S. KOVALEV and A. M. KOSEVICH, Fiz. Nizk. Temp 15 (1989) 511 [Sov. J. Low Temp.
Phys. 15 (1989) 288].
144. I. V. BARASHENKOV and E. YU. PANOVA, Physica D 69 (1993) 114.
145. YU. S. KIVSHAR and W. KROLIKOWSKI, Opt. Lett. 20 (1995) 1527.
146. YU. S. KIVSHAR and V. V. AFANASJEV, Opt. Lett. 21 (1996) 1135.
147. I. V. BARASHENKOV, Phys. Rev. Lett. 77 (1996) 1193.
148. U. PESCHEL, C. ETRICH, F. LEDERER and B. A. MALOMED, Phys. Rev. E 55 (1997) 7704.
149. A. V. BURYAK, YU. S. KIVSHAR and D. F. PARKER, Phys. Lett. A 215 (1996) 57.
150. D. E. PELINOVSKY and R. H. J. GRIMSHAW, Nonlinear Instability Analysis, Advances in Fluid Mechanics
Series, Vol 12, edited by L. Debnath and S. R. Choudhury (Comput. Mech. Publ., Southampton, 1997)
Chap. 8.
151. L. TORNER, D. MIHALACHE, D. MAZILU and N. N. AKHMEDIEV, Opt. Commun. 138 (1997) 105.
152. I. V. BARASHENKOV, D. E. PELINOVSKY and E. V. ZEMLYANAYA, Phys. Rev. Lett. 80 (1998) 5117.
153. P. B. LUNDQUIST, D. R. ANDERSEN and YU. S. KIVSHAR, Phys. Rev. 57 (1998) 3551.
154. E. INFELD and R. ROWLANDS, Nonlinear Waves, Solitons and Chaos (Cambridge University Press,
Cambridge, 1990).
155. T. B. BENJAMIN and J. E. FEIR, J. Fluid Mech. 27 (1967) 417.
156. V. E. ZAKHAROV, J. Appl. Mech. Tech. Phys. 2 (1968) 190.
157. L. A. OSTROVSKY, Zh. Exp. Teor. Fiz. 51 (1966) 1189 [Sov. Phys. JETP 24 (1967) 797].
158. V. I. BESPALOV and V. I. TALANOV, Pisma Zh. Eksp. Teor. Fiz. 3 (1966) 471 [JETP Lett. 3 (1966) 307].
159. H. C. YUEN and W. E. FERGUSON, Phys. Fluids 21 (1978) 2116.
160. K. TAI, A. HASEGAWA and A. TOMITA, Phys. Rev. Lett. 56 (1986) 135.
161. B. B. KADOMTSEV and V. I. PETVIASHVILI, Dokl. Acad. Nauk SSSR 192 (1970) 753 [Sov. Phys. Doklady
15 (1970) 539].
162. V. E. ZAKHAROV, Zh. Eksp. Teor. Fiz. 53 (1967) 1735 [Sov. Phys. JETP 26 (1968) 994].
163. V. E. ZAKHAROV and A. M. RUBENCHIK, Zh. Eksp. Teor. Fiz. 65 (1973) 997 [Sov. Phys. JETP 38 (1974)
494].
164. N. YAJIMA, Prog. Theor. Phys. 52 (1974) 1066.
165. G. SCHMIDT, Phys. Rev. Lett. 34 (1975) 724.
166. K. H. SPATSCHEK, P. K. SHUKLA and M. Y. YU, Phys. Lett. A 54 (1975) 419.
167. YU. V. KATYSHEV and V. G. MAKHANKOV, Phys. Lett. A 57 (1976) 10.
168. D. ANDERSEN, A. BONDESON and M. LISAK, J. Plasma Phys. 21 (1979) 259.
169. M. J. ABLOWITZ and H. SEGUR, J. Fluid Mech. 92 (1979) 691.
170. M. J. ABLOWITZ and H. SEGUR, Stud. Appl. Math. 62 (1980) 249.
171. L. M. DEGTYAREV, V. E. ZAKHAROV and L. I. RUDAKOV, Zh. Eksp. Teor. Fiz. 68 (1975) 115 [Sov. Phys.
JETP 41 (1975) 57].
172. N. R. PEREIRA, R. N. SUDAN and J. DENAVIT, Phys. Fluids 20 (1977) 271, 936.
173. M. J. ABLOWITZ and H. SEGUR, Solitons and the Inverse Scattering Transformation (SIAM, Philadelphia,
1981).
174. A. V. MAMAEV, M. SAFFMAN and A. A. ZOZULYA, Europhys. Lett. 35 (1996) 25.
175. A. V. MAMAEV, M. SAFFMAN, D. Z. ANDERSON and A. A. ZOZULYA, Phys. Rev. A 54 (1996) 870.
176. E. A. KUZNETSOV and S. K. TURITSYN, Zh. Eksp. Teor. Fiz. 94 (1988) 119 [Sov. Phys. JETP 67 (1988)
1583].
177. G. S. MCDONALD, K. S. SYED and W. J. FIRTH, Opt. Commun. 95 (1993) 281.
178. C. T. LAW and G. A. SWARTZLANDER JR, Opt. Lett. 18 (1993) 586.
179. D. E. PELINOVSKY, YU. A. STEPANYANTZ and YU. S. KIVSHAR, Phys. Rev. E 51 (1995) 5016.
180. V. TIKHONENKO, J. CHRISTOU, B. LUTHER-DAVIES and YU. S. KIVSHAR, Opt. Lett. 21 (1996) 1129.
181. A. V. MAMAEV, M. SAFFMAN and A. A. ZOZULYA, Phys. Rev. Lett. 76 (1996) 2262.
182. A. D. ROSSI, S. TRILLO, A. V. BURYAK and YU. S. KIVSHAR, Phys. Rev. E 56 (1997) 4959.
183. W. J. FIRTH and D. V. SKRYABIN, Phys. Rev. Lett. 79 (1997) 2450.
614
YU. S. Kivshar

Anda mungkin juga menyukai