Perturbation theory for

royalsocietypublishing.org/journal/rspa dark–bright solitons of the

Manakov system
A. Chernyavskiy1,2,3 , D. J. Frantzeskakis4 ,
Research
T. P. Horikis5 , G. N. Koutsokostas4 and B. Prinari1
Cite this article: Chernyavskiy A, 1
Department of Mathematics, University at Buffalo, Buffalo, NY 14260, USA
2
Frantzeskakis DJ, Horikis TP, Koutsokostas GN, Department of Mathematics & Statistics, Texas A&M Corpus Christi, Corpus Christi,
Downloaded from https://royalsocietypublishing.org/ on 27 January 2025

TX 78412, USA
Prinari B. 2025 Perturbation theory for 3
Department of Mathematics & Statistics, Brock University, St Catharines, L2S 3A1
dark–bright solitons of the Manakov system. ON, Canada
4
Proc. R. Soc. A 481: 20240082. Department of Physics, National and Kapodistrian University of Athens, Athens
https://doi.org/10.1098/rspa.2024.0082 15784, Greece
5
Department of Mathematics, University of Ioannina, Ioannina, 45110, Greece

TPH, 0000-0003-0731-6107; BP, 0000-0001-8809-2656

Received: 31 January 2024
Accepted: 24 October 2024 We present a direct perturbation method to study
the dynamics of dark–bright (DB) solitons of the
Manakov system under the influence of perturbations.
Our methodology combines a multi-scale expansion
Subject Category: method, perturbed conservation laws and a boundary-
Mathematics layer approach, which breaks the problem into an
inner region, pertinent to the soliton core and an outer
region, which evolves independently of the dark–
Subject Areas: bright soliton. We find that a shelf emerges around
applied mathematics, mathematical physics, the dark soliton component, which propagates with
optics a speed depending on the background intensity.
Conserved quantities of the Manakov system are
employed to determine the properties of the perturbed
Keywords: solutions. We focus on dissipative perturbations, such
perturbation theory, nonlinear Schrödinger as diffusion, as well as linear and nonlinear loss,
and show that the effect of the bright (filling)
systems, solitons
soliton component is to partially stabilize ‘bare’
dark solitons of the scalar case against perturbation-
induced dissipation. Our analytical predictions
Author for correspondence: are corroborated by results of direct numerical
Barbara Prinari simulations.
e-mail: [email protected]

1. Introduction
The mathematical modelling of physical phenomena
often leads to a certain class of nonlinear partial
differential equations (PDEs) known as integrable

© 2025 The Authors. Published by the Royal Society. All rights reserved.
 systems. One of the distinguished features of integrable systems is that they admit exact soliton
2
solutions, i.e. stable, exponentially localized travelling waves, which interact with one another
non-destructively, preserving their shape and velocity in the interaction [1]. Moreover, integrable

royalsocietypublishing.org/journal/rspa Proc. R. Soc. A 481: 20240082

...........................................................................................................
systems possess an infinite number of conserved quantities, as well as a Lax pair that allows one
to linearize them using the inverse scattering transform (IST), a nonlinear analogue of the Fourier
transform [2]. One of the prototypical integrable equations is the nonlinear Schrödinger (NLS)
equation:

iqt + qxx − 2𝜈|q|2 q = 0, (1.1)

with 𝜈 = ∓1 corresponding to the ‘focusing’ and ‘defocusing’ regimes, where bright or dark
solitons can be supported, respectively. The NLS equation is a universal model for weakly
Downloaded from https://royalsocietypublishing.org/ on 27 January 2025

dispersive nonlinear wave trains, and it has been derived in such diverse fields as deep water
waves, plasmas, nonlinear fibre optics, Bose–Einstein condensates (BECs) and spin waves [3–10].
Vector generalizations of the scalar NLS equations (VNLS equations for short) arise as relevant
physical models, under conditions similar to those described by the NLS equation, whenever there
are suitable multiple wavetrains moving with nearly the same group velocity. This may happen,
e.g., in nonlinear optics, when two waves of different polarization or two waves of different
frequencies feature a nonlinear interaction [11]. In such situations, of particular relevance is the
so-called Manakov system [12], which is a coupled NLS system:

iqt + qxx − 2𝜈‖q‖2 q = 0, (1.2)

where q(x, t) is a two-component, complex vector function. Notably, the Manakov system is
completely integrable and, like its scalar counterpart, it admits a Lax pair, soliton solutions,
infinite number of conserved quantities, etc.
Menyuk showed in [13] that in optical fibres with constant birefringence, assuming certain
nonlinear (four-wave mixing) terms are neglected, the two polarization components of the
complex electromagnetic field envelope orthogonal to direction of propagation along a fibre
satisfy asymptotically the following non-dimensional equations:

i(ut + 𝛿ux ) + duxx + (|u|2 + 𝛼|v|2 ) = 0, i(vt + 𝛿vx ) + dvxx + (𝛼|u|2 + |v|2 ) = 0, (1.3)

where 𝛿 represents the group velocity ‘mismatch’ between the components, 2d is the group
velocity dispersion (the sign of which accounts for focusing vs defocusing regimes) and 𝛼 is
a constant depending on the polarization properties of the fibre. The physical phenomenon of
birefringence implies that the phase and group velocities of the electromagnetic wave are different
for each polarization component. When 𝛼 ≠ 1 the above system is not integrable. However,
averaging over the fast birefringence fluctuations that are normally observed in a communications
environment [14] yields 𝛿 = 0 and 𝛼 = 1—that is, the system reduces to the VNLS equation (1.2).
Notice that, besides its relevance to nonlinear fibre optics, more recently, the Manakov system
has attracted considerable attention in studies related to homogeneous multicomponent BECs,
composed by, e.g., different spin states of the same atom species [15–18]; in this context, a
physically relevant situation refers to the case where the repulsive inter- and intra-species
interactions are of (approximately) equal strength, a fact rendering equation (1.2) the appropriate
model. Vector solitons of the VNLS model have also attracted much attention, especially in the
defocusing setting (𝜈 = +1). Of particular interest in this setting are the dark–bright (DB) soliton
solutions; in these states, the bright soliton—which is not supported by the scalar defocusing
NLS equation—only emerges because of an effective potential well created by the dark soliton
through the inter-component interaction; as such, DB solitons are commonly referred to as
‘symbiotic’ solitons. Predicted in the seminal work by Busch and Anglin [19], DB solitons were
first experimentally realized by the phase imprinting method [20], followed by the experimental
observation of trains of DB solitons generated by the counterflow of two superfluids [21].
While the above discussion refers to integrable systems, in most physically relevant settings the
pertinent model PDEs are usually non-integrable. Nevertheless, the theoretical predictions for the
 soliton solutions in the corresponding integrable cases have proved to be an extremely valuable
3
tool for the investigation of non-integrable solitary waves in regimes that are reasonably close to
the integrable ones. As such, in many works, perturbation-based techniques of nearly integrable

royalsocietypublishing.org/journal/rspa Proc. R. Soc. A 481: 20240082

...........................................................................................................
systems have been employed for the investigation of the evolution of solitons in the presence of
small perturbations, such as linear or nonlinear loss or gain, diffusion and higher-order dispersion
or nonlinearity effects. The goal of this work is to present a rigorous direct perturbation theory for
the study of DB solitons of the defocusing Manakov system over a constant background under
small perturbations, that is

iqt + qxx − 2(‖q‖2 − q2o )q = 𝜀R[q], (1.4)

where q(x, t) = (q1 , q2 )T , R[q] = (R1 [q], R2 [q])T , 0 < 𝜀 ≪ 1 is a perturbation parameter, and
Downloaded from https://royalsocietypublishing.org/ on 27 January 2025

q(x, t) ∼ q± (t) = (qo ei𝜃± , 0)T x → ±∞, (1.5)

are boundary conditions that correspond to solutions which are ‘dark’ in the first component
and ‘bright’ in the second one. The linear term proportional to the background amplitude qo
in equation (1.4) has been introduced to make the boundary conditions independent of t. This
can be achieved by a gauge transformation replacing q by exp(2iq2o t)q, or, more generally, by
t
exp[2i ∫0 qo (s)ds]q when the background amplitude qo depends on t.
We consider arbitrary perturbations satisfying R[0] = 0, which are asymptotically phase-
( )
invariant, such that limx→±∞ Rj [qei𝜙 ] = ei𝜙± Rj [q± ] for j = 1, 2, for any phase 𝜙(x, t), with 𝜙± ∶=
limx→±∞ 𝜙(x, t) (of course, 𝜙± ≡ 𝜙 in the case of a constant phase). Note that these are fairly general
assumptions, as a wide class of physically relevant perturbations satisfies these conditions; these
include, for example, diffusion: Rj [q] = i𝛾j 𝜕x2 qj , linear loss: Rj [q] = i𝛾j qj , nonlinear loss (pertinent to
two-photon absorption in optics): Rj [q] = i𝛾j qj |qj |2 , etc. Note that the gauge transformation used
above to remove the fast evolution of the background phase is also phase-invariant, justifying the
use of equation (1.4) instead of perturbing equation (1.2).
As mentioned above, the unperturbed defocusing Manakov system, namely equation (1.4),
with 𝜀 = 0, admits exact DB soliton solutions of the form
√ √
q1 (x, t) = [sgn(V) q2o − A2d − iAd tanh( A2d − A2b (x − Vt − xo ))] ei𝜎o , (1.6a)
√
V V2
q2 (x, t) = Ab exp [i ( x( − (A2d − A2b )) t + 𝜑o )] sech( A2d − A2b (x − Vt − xo )), (1.6b)
2 4

where Ab , Ad determine the amplitudes of the bright/dark component, respectively, qo is the

background, V is the soliton velocity, xo is the soliton centre and 𝜎o , 𝜑o are arbitrary phases [22–28].
Note that the velocity V is related to the amplitudes by

V 2 A2d = 4(q2o − A2d )(A2d − A2b ), (1.7)

and for a stationary DB soliton V = 0, Ad = qo and Ab < qo is arbitrary. The scalar dark soliton can
be obtained by setting Ab = 0, i.e. q2 ≡ 0.
While perturbation theory for solitons that decay rapidly at infinity has been widely
studied since the late 1970s—with a variety of methods ranging from multi-scale perturbation
analysis, IST-based perturbation techniques, perturbation of conserved quantities and direct
numerical simulations [29–33]—the non-vanishing background of dark solitons introduces severe
complications when applying the perturbative methods developed in the rapidly decaying
case. For the scalar defocusing NLS equation, dark solitons are completely determined by the
four parameters qo , Ad , xo and 𝜎o and in this case equation (1.7) reduces to V 2 = q2o − A2d . In
some early works, the perturbation of ‘black’ (i.e. stationary dark) solitons in lossy fibres was
studied numerically [34] and analytically [35,36]. The method developed in [36] was subsequently
extended to ‘grey’ (i.e. non-stationary or moving dark) solitons and to generic perturbations,
but only two of the four main soliton parameters, qo and Ad , were determined. In [37], it was
 shown that under a perturbation, the background evolves independently of the soliton, and after
4
separating the background amplitude from the soliton ‘core’, it is possible to determine the dark
soliton’s amplitude and width upon using the Hamiltonian approach of the adiabatic perturbation

royalsocietypublishing.org/journal/rspa Proc. R. Soc. A 481: 20240082

...........................................................................................................
theory (which is based on perturbed conservation laws). A similar approach was also used for DB
solitons in BECs [18].
It should be noted, however, that for dark solitons the adiabatic evolution of the soliton
parameters alone is not sufficient to fully characterize the perturbed solution. The reason is that,
in many cases, the perturbation generates a moving ‘shelf’, that is, a linear wave emerging on
either side of the soliton; the existence of this shelf, which was confirmed both numerically and
analytically, was in fact used to explain observed discrepancies in the perturbed conservation
laws [38], though without determining analytically the core soliton parameters.
Downloaded from https://royalsocietypublishing.org/ on 27 January 2025

An alternative approach to soliton perturbation theory in the rapidly decaying case was
pioneered in [29], as a way to determine the effects of small perturbations on the evolution of
the soliton spectral (and hence physical) parameters. The method, which might be referred to as
‘integrable perturbation theory’, relies on the IST and on completeness of squared eigenfunctions,
namely quadratic combinations of Jost eigenfunctions and their adjoints which satisfy the
linearized version of the integrable PDE. There have been many attempts at generalizing the IST-
based perturbation theory to the case of dark solitons since the early nineties. For instance, in
[39], orthogonality conditions were derived from a set of squared Jost functions for the scalar
defocusing NLS equation over a constant background, and from these conditions one can in
principle obtain all of the soliton parameters. This early work, however, did not account at all
for the evolution of the background induced by the perturbation. Subsequent works presented
proofs of the completeness of the squared eigenfunctions using different approaches [40–45], but
the results were not consistent with each other. As an example, the proof in [40] was claimed to be
incorrect in [44,45], based on the observation that the complete set should have two, not just one,
continuous spectrum basis vectors, which resulted in different predictions for the soliton velocity
and the first-order correction. In [46], the results of [40,42] and [44,45] were then declared to be
‘equivalent’ under some kind of ‘transformation between two integral variables’. On the other
hand, in [43] squared eigenfunctions were used (though without explicitly referring to them,
or to their completeness) to develop an IST-based perturbation theory for the defocusing NLS
on a background. The main drawbacks of all these works are that none of them accounted for
perturbative contributions from the shelf that develops around the dark soliton, or presented
comparisons of the theoretical predictions with numerical simulations.
To date, the most comprehensive analysis of dark-soliton perturbation for the scalar defocusing
NLS is found in [47], where a multi-scale expansion method and perturbed conservation laws
were used to find both the growth in magnitude and phase of the shelf and the adiabatic evolution
of all soliton parameters. This work also highlighted the emergence of a moving boundary layer
connecting the inner soliton core to the outer background.
There are many papers available in the literature on bright soliton perturbation theory,
but, as we mention above, less so for dark soliton, and only a handful that address vector/
multicomponent problems on a non-trivial background. In light of the difficulties that scalar
problems for dark solitons have presented, it is not surprising that few attempts have been
made so far to develop perturbative approaches for multicomponent integrable systems on a
background. In particular, to the best of our knowledge, the only works on soliton perturbation
theory for the defocusing Manakov system on a non-trivial background are [18,48,49]. The
approach in these papers relies on the adiabatic approximation, and evolution equations for the
soliton parameters of a DB soliton are derived by expanding the solution into a set of complete
eigenfunctions of the linearized operator. We note, however, that the completeness result for
the eigenfunctions appears to be an unsettled issue even in the scalar case, and in the Manakov
system, the defect of analyticity of the scattering eigenfunctions further complicates the problem.
Furthermore, it is not clear how or even if the soliton shelf can be incorporated into the description,
how the results in these papers compare with the scalar reduction on this account, and no
comparisons with direct numerical simulations are offered to corroborate the results. Arguably,
 the perturbation theory for DB Manakov solitons is to a large extent still an open problem, and the
5
present work aims at filling this gap. We stress that the great practical importance of this problem,
besides the obvious relevance of being able to include, at least perturbatively, physical effects such

royalsocietypublishing.org/journal/rspa Proc. R. Soc. A 481: 20240082

...........................................................................................................
as dissipation or loss, in optical fibres described by equation (1.3) is that one would also be able to
account for more general polarization properties of the fibre (e.g. for values of 𝛼 ≠ 1 but reasonably
close to it), and/or include four-wave mixing effects as perturbations of the integrable case. Also,
in the recent applications to BECs, the harmonic trap necessary to achieve confinement of the
atoms in multicomponent repulsive condensates is typically devised in such way that the ratio
of longitudinal vs transverse trapping frequencies is very small (𝛺 ∼ 10−2 ). In these cigar-shaped
geometries for the condensates, the BEC dynamics for the longitudinal part of the wave function
is indeed described as a perturbation of the integrable defocusing Manakov system. Moreover,
Downloaded from https://royalsocietypublishing.org/ on 27 January 2025

in [50,51], the coupling coefficients for ‘symmetric’ spin-independent and ‘antisymmetric’ spin-
dependent interaction terms are also such that their ratio is a small parameter up to which the
model equation can be considered a small perturbation of a three-component Manakov system.
These are just some examples of the practical applicative relevance of this work. Specifically,
the goal of this article is a highly non-trivial generalization of the methods developed in [47] to
describe perturbations of DB solitons.
The plan of the paper is the following. In §2, we discuss the evolution of the background, and
in §3, we use multiple scales in time to formulate the perturbation problem for a stationary DB
soliton. It is shown that, similarly to the scalar case, a shelf emerges, propagating with a speed
determined by the background intensity. The problem is then broken into an inner region, where
the core of the soliton resides, and an outer region which evolves independently of the soliton.
Here it is noted that even for stationary DB solitons the amplitude of the bright soliton, Ab , is
an additional physical parameter one needs to determine. Therefore, the perturbed conservation
laws (for Hamiltonian, energy and momentum) used in the scalar case are not sufficient to
determine all the soliton parameters. We are able to suitably augment the set of conserved
quantities of the unperturbed Manakov system and use them to determine the properties of
the moving shelf, as well as the adiabatic evolution of the soliton parameters. Our results for
a stationary DB soliton coincide with the ones in [47] in the reduction to the scalar case and
show good agreement with direct numerical simulations. In §4, we generalize the multi-scale
perturbation theory to a moving DB soliton. Although in this case we are not able to obtain a
complete solution for the amplitudes and phases of the DB soliton at O(𝜀), we use the augmented
perturbed conservation laws to obtain a nonlinear system of adiabatic evolution equations for
the soliton and shelf parameters (eight coupled ODEs in the slow time T1 = 𝜀t for the DB soliton
parameters qo , Ad , Ab and 𝜎o , and for the asymptotic amplitudes and phases of the shelf as
x → ±∞). In §5, we discuss the boundary-layer region. In §6, we present the solution of the
system of equations for the adiabatic evolution of the soliton parameters for specific perturbations
(diffusion, linear and nonlinear loss in both components). This is a highly non-trivial result, since,
a priori, one would have no guarantee that the system of equations for the above mentioned DB
soliton parameters could be solved in closed form. Furthermore, we show that our results are
consistent with [47] in the scalar reduction, and we also compare them in the general vector case
with some direct numerical simulations, with excellent agreement. Finally, §7 is devoted to some
concluding remarks and a discussion for future work.

2. The boundary at infinity

Let us consider the perturbed Manakov system (1.4) with boundary conditions (1.5), consistent
with a DB soliton. We seek the solution in the form of the following asymptotic expansions in 𝜀:
q1 = q10 + 𝜀q11 + O(𝜀2 ), q2 = q20 + 𝜀q21 + O(𝜀2 ), (2.1)
𝜕 𝜕 𝜕
and introduce two time scales, To = t and T1 = 𝜀t, so that → +𝜀 + O(𝜀2 ). The O(1)
𝜕t 𝜕To 𝜕T1
equations are of course the unperturbed Manakov system (1.4) with 𝜀 = 0 for q10 and q20 . Collecting
 O(𝜀) terms yields the first-order correction system of equations:
6
2
𝜕q11 𝜕q10 𝜕 q11
i +i + − 2(|q10 |2 + |q20 |2 − q2o )q11 − 4q10 Re(q10 q∗11 + q20 q∗21 ) = R1 [q10 ], (2.2a)

royalsocietypublishing.org/journal/rspa Proc. R. Soc. A 481: 20240082

...........................................................................................................
𝜕To 𝜕T1 𝜕x2
𝜕q21 𝜕q20 𝜕 2 q21
i +i + − 2(|q10 |2 + |q20 |2 − q2o )q21 − 4q20 Re(q10 q∗11 + q20 q∗21 ) = R2 [q20 ], (2.2b)
𝜕To 𝜕T1 𝜕x2
where ∗ denotes complex conjugate, and where we have assumed that Rj [q] = Rj [qj ], that is, that
the perturbation of each component only depends on the corresponding component. Note that
this assumption only has the purpose of simplifying the description and the notation, but it is not
essential in any of the following. We also assume 𝜕x2 q11 , 𝜕x2 q21 → 0 as |x| → ∞ (continuous wave
background), and the perturbative corrections q11 , q21 to be functions of T1 only. Under these
Downloaded from https://royalsocietypublishing.org/ on 27 January 2025

assumptions, taking the limit as x → ±∞ in equation (2.2) yields

𝜕q±10 𝜕q±20
i = R1 [q±10 ], i = R2 [q±20 ], (2.3)
𝜕T1 𝜕T1
where the superscripts ± denote limits as x → ±∞, and we note that since q1 , q2 have to satisfy the
boundary conditions to all orders, q±11 , q±21 are both zero, together with all higher-order terms.
±
Let us express the asymptotic behaviour of the dark component as q±10 = qo ei𝜙 , and take into
±
account that in the DB soliton case q20 = 0. Then, separating real and imaginary parts in the first
of equation (2.3) yields
𝜕qo ± ± 𝜕𝜙± ± ±
= Im(R1 [qo ei𝜙 ]e−i𝜙 ), qo = Re(R1 [qo ei𝜙 ]e−i𝜙 ), (2.4)
𝜕T1 𝜕T1
i.e. using the asymptotic phase invariance of the perturbation
𝜕qo 𝜕𝛥𝜙∞
= Im(R1 [qo ]), = 0. (2.5)
𝜕T1 𝜕T1
Here, 𝛥𝜙∞ = 𝜙+ − 𝜙− is the asymptotic phase difference of the background. Since the second
component is assumed to be rapidly decaying, not surprisingly, equation (2.5) terms for the
boundary at infinity are the same as in the scalar case and are independent of the perturbation in
the bright component.

3. The first-order correction for a stationary DB soliton

For a stationary DB soliton of the unperturbed Manakov system, in equations (1.6), we set Ad = qo
and V = 0:
√ √
2 2
qd = −iqo tanh( q2o − A2b (x − xo ))ei𝜎o , qb = Ab ei(qo −Ab )t sech( q2o − A2b (x − xo ))ei𝜑o . (3.1)

Note that, unlike black solitons in the scalar NLS, here the amplitude of the bright soliton, Ab < qo ,
is an additional free parameter.
Let us use Madelung coordinates for both components, that is, write q1 = uei𝜙 for the dark
component, and q2 = vei𝜓 for the bright one, with u, v, 𝜙, 𝜓 ∈ ℝ. Thus, the system (1.4) becomes

−u𝜙t + uxx − 𝜙x2 u − 2(u2 + v2 − q2o )u = 𝜀Re(R1 [uei𝜙 ]e−i𝜙 ), (3.2a)

ut + u𝜙xx + 2ux 𝜙x = 𝜀Im(R1 [uei𝜙 ]e−i𝜙 ), (3.2b)
−v𝜓t + vxx − 𝜓x2 v 2
− 2(u + v − 2
q2o )v = 𝜀Re(R2 [vei𝜓 ]e−i𝜓 ), (3.2c)
i𝜓 −i𝜓
vt + v𝜓xx + 2vx 𝜓x = 𝜀Im(R2 [ve ]e ). (3.2d)

As before, we introduce multiple scales in time To = t, T1 = 𝜀t and expand u, v, 𝜙, 𝜓 as

u = uo + 𝜀u1 + O(𝜀2 ), v = vo + 𝜀v1 + O(𝜀2 ), 𝜙 = 𝜙o + 𝜀𝜙1 + O(𝜀2 ), 𝜓 = 𝜓o + 𝜀𝜓1 + O(𝜀2 ). (3.3)

 At O(1), the system (3.2) is satisfied by the stationary DB soliton (3.1), with
7
√
uo = qo tanh( q2o − A2b (x − xo )), 𝜙o = 𝜎o , (3.4a)

royalsocietypublishing.org/journal/rspa Proc. R. Soc. A 481: 20240082

...........................................................................................................
√
vo = Ab sech( q2o − A2b (x − xo )), 𝜓o = 𝜑o + (q2o − A2b )To . (3.4b)

At O(𝜀), the system (3.2) yields

2
−uo 𝜙oT1 − u1 𝜙oTo − uo 𝜙1To + u1xx − 2𝜙ox 𝜙1x uo − 𝜙ox u1
−4u2o u1 − 4uo vo v1 − 2(u2o + v2o − q2o )u1 = Re(R1 [uo ei𝜙o ]e−i𝜙o ), (3.5a)
u1To + uoT1 + uo 𝜙1xx + u1 𝜙oxx + 2u1x 𝜙ox + 2uox 𝜙1x = Im(R1 [uo ei𝜙o ]e−i𝜙o ), (3.5b)
Downloaded from https://royalsocietypublishing.org/ on 27 January 2025

2
−vo 𝜓oT1 − v1 𝜓oTo − 𝜓1To vo + v1xx − 2𝜓ox 𝜓1x vo − 𝜓ox v1
−4v2o v1 − 4uo vo u1 − 2(u2o + v2o − q2o )v1 = Re(R2 [vo ei𝜓o ]e−i𝜓o ), (3.5c)
v1To + voT1 + vo 𝜓1xx + v1 𝜓oxx + 2v1x 𝜓0x + 2vox 𝜓1x = Im(R2 [vo ei𝜓o ]e−i𝜓o ). (3.5d)

Here, and in the following subscripts, To and T1 denote derivatives with respect to the
corresponding time-scales. Note that equations (3.4) imply

𝜙oTo = 0, 𝜓oTo = A2b − q2o , 𝜙ox = 𝜙oxx = 𝜓ox = 𝜓oxx = 0. (3.6)

Furthermore, seeking a stationary solution, we assume

u1To = 𝜙1To = v1To = 𝜓1To = 0.

In addition, we assume that qo and Ab are independent of To , but both in general have O(𝜀) terms
depending on T1 . Then, we can write the system (3.5) in matrix form as

⎡uo d𝜎o + Re(R1 [uo ei𝜙o ]e−i𝜙o )⎤

𝜕x2 − 6u2o − 2v2o + 2q2o −4uo vo u1 ⎢ dT1 ⎥,
[ ] [ ] = (3.7a)
−4uo vo 𝜕x2 − 6v2o − 2u2o + q2o + A2b v1 ⎢ v d𝜑o + Re(R [v ei𝜓o ]e−i𝜓o ) ⎥
o 2 o
dT1
⎣ ⎦
uo 𝜕x2 + 2uox 𝜕x 0 𝜙 −u + Im(R1 [uo ei𝜙o ]e−i𝜙o )
[ ] [ 1 ] = [ oT1 ]. (3.7b)
0 vo 𝜕x2 + 2vox 𝜕x 𝜓1 −voT1 + Im(R2 [vo ei𝜓o ]e−i𝜓o )

The expressions for uoT1 , voT1 which contribute to the forcing in the equations (3.7b) for the phases
can be easily computed from equations (3.4):
dq dA
⎛ qo o − Ab b ⎞
dxo dq
uoT1 = −uox ⎜ −
dT1 dT1
(x − x o ) ⎟ + uo o , (3.8a)
⎜ dT1 2 2
qo − Ab ⎟ qo dT1
⎝ ⎠
dq dA
⎛ qo o − Ab b ⎞
⎜ dxo dT1 dT1 vo dAb
voT1 = −vox − 2 2
(x − xo )⎟ + . (3.8b)
⎜ dT1 qo − Ab ⎟ Ab dT1
⎝ ⎠
In turn, the T1 -dependence of qo is given by equations (2.4), while dAb ∕dT1 is related to dqo ∕dT1
via
dAb dqo
Ab (2q2o − A2b ) − qo (4q2o − 3A2b )
dT1 dT1
√ dqo ∞
= q2o − A2b (Eo + (q2o − A2b )Im ∫ R2 [vo ei𝜓o ]vo e−i𝜓o dx), (3.9)
dT1 −∞

which is obtained from the perturbed conservation laws, and specifically equation (6.15c) in §6. In
table 1, expressions for dqo ∕dT1 and dAb ∕dT1 are listed for various types of perturbations, which
are then used to compute particular solutions to equation (3.7b) for the phases in each case.
 Table 1. dqo ∕dT1 and dAb ∕dT1 for perturbations corresponding to diffusion, linear loss and nonlinear loss.
8
diffusion linear loss nonlinear loss

royalsocietypublishing.org/journal/rspa Proc. R. Soc. A 481: 20240082

...........................................................................................................
R [q ] i𝛾 𝜕 2 q i𝛾 q i𝛾 |q |2 q
. . . . . . . . . .1. . . . .1. . . . . . . . . . . . . . . . . . . . . .1. . . . x. . . . 1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1. . .1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1. . . . . .1. . . . . . .1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
R [q ] i𝛾 𝜕 2 q i𝛾 q i𝛾 |q |2 q
. . . . . . . . . .2. . . . .2. . . . . . . . . . . . . . . . . . . . . .2. . . . x. . . . 2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2. . .2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2. . . . . .2. . . . . . .2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
dqo
0 𝛾 1 qo 𝛾1 qo3
. . . . . . . .dT
. . . .1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2 2 2
dAb 2 Ab (Ab − qo ) Ab (𝛾1 qo2 + 2𝛾2 (qo2 − A2b )) Ab (3𝛾1 qo4 + 4𝛾2 (qo2 − A2b ))
− 𝛾2
dT1 3 (2qo2 − A2 ) 2qo2 − A2b 3(2qo2 − A2b )
b
.........................................................................................................................................................................................................
Downloaded from https://royalsocietypublishing.org/ on 27 January 2025

Ideally, one should determine the general solution of the linear, non-homogeneous system (3.7)
for the first-order corrections to amplitudes and phases of the stationary DB soliton. The situation
in the Manakov system, however, is significantly more complicated than in the scalar case.
Indeed, while equation (3.7b) for the phases is fully decoupled, and one can obtain the general
homogeneous solution, as well as a particular solution, the system (3.7a) for the amplitudes
remains fully coupled, and we are only able to obtain explicit expression for the particular
solutions, as well as certain asymptotic information on the general homogeneous solution.

(a) First-order corrections for the phases

The system (3.7b) for the phases 𝜙1 , 𝜓1 is decoupled and has homogeneous solutions:
√
⎛ coth ( q2o − A2b (x − xo )) ⎞
𝜙1h = c1 + c2 ⎜x − √ ⎟, (3.10a)
⎜ ⎟
q2o − A2b
⎝ ⎠
√
⎛ 2 2
sinh (2 qo − Ab (x − xo )) ⎞
𝜓1h = d1 + d2 ⎜x + √ ⎟, (3.10b)
⎜ ⎟
2 q2o − A2b
⎝ ⎠
where c1 , c2 , d1 , d2 are arbitrary constants. It is clear that to avoid exponential growth in the phase
one needs to choose d2 = 0. Also, the coth term in the expression of 𝜙1h becomes singular at
x = xo , and at first glance one might assume that this requires c2 = 0 as well. However, the phase
(√ )
term in the dark component is multiplied by tanh q2o − A2b (x − xo ) [see equation (3.4)], which
vanishes for x = xo ; this indicates that more general solutions could exist with c2 ≠ 0. On the other
hand, particular solutions are determined by the forcing, and the right-hand sides (RHSs) of
equation (3.7b) for the phases for various choices of perturbations are listed in table 2.
From the explicit expressions of the forcing terms in table 2, one can compute particular
solutions for equation (3.7b). It is important to point out that for all three perturbations considered
here as examples (diffusion, linear loss and nonlinear loss), phase invariance can be used to
simplify the RHSs of equations (3.7b). Indeed, linear and nonlinear loss are phase-invariant for all
x, not just asymptotically; in the case of the diffusion perturbation, in general first- and second-
order derivatives of the phase of the bright component would appear, but these are zero in the
case of a stationary DB soliton according to equations (3.6).
Specifically, for a perturbation of the form of diffusion in both components, we find
√
1 dxo A2b q2o − A2b √
𝜙1p = x − 𝛾2 (x − xo ) coth ( q2o − A2b (x − xo ))
2 dT1 3(2q2o − A2b )
√
2
− 𝛾1 ln cosh ( q2o − A2b (x − xo )), (3.11a)
3
 Table 2. RHSs of equation (3.7b) for perturbations of the form of diffusion, linear loss and nonlinear loss (cf. table 1).
9
perturbation RHSs of (3.7b) for 𝜙1 , 𝜓1

royalsocietypublishing.org/journal/rspa Proc. R. Soc. A 481: 20240082

...........................................................................................................
dxo 2A2b (qo2 − A2b )
diffusion uox ( − 𝛾2 (x − xo )) + 𝛾1 uoxx
dT1 3(2qo2 − A2 ) b
2A2b (qo2 − A2b ) 2(qo2 − A2b )2
vox (xoT1 − 𝛾2 (x − xo )) + 𝛾2 vo + 𝛾2 voxx
3(2qo2 − A2b ) 3(2qo2 − A2b )
.........................................................................................................................................................................................................
2 2
dxo 2(𝛾1 qo − 𝛾2 Ab )
linear loss uox ( − (x − xo ))
dT1 2qo2 − A2b
Downloaded from https://royalsocietypublishing.org/ on 27 January 2025

2 2
dxo 2(𝛾1 qo − 𝛾2 Ab ) 𝛾2 A2b − 𝛾1 qo2
vox ( − (x − x o )) + vo
dT1 2qo2 − A2b 2qo2 − A2b
.........................................................................................................................................................................................................
4 4
dxo 2(3𝛾1 qo − 2𝛾2 Ab )
nonlinear loss uox ( − (x − xo )) − 𝛾1 qo2 uo + 𝛾1 uo3
dT1 3(2qo2 − A2b )
4 4
dxo 2(3𝛾1 qo − 2𝛾2 Ab ) 3𝛾1 qo4 + 4𝛾2 A2b (qo2 − A2b )
vox ( − (x − x o )) − vo + 𝛾2 vo3
dT1 3(2qo2 − A2b ) 3(2qo2 − A2b )
.........................................................................................................................................................................................................

x(x − 2xo )A2b (q2o − A2b ) 2 √

1 dxo
𝜓1p = x − 𝛾2 − 𝛾 2 ln cosh ( q2o − A2b (x − xo )), (3.11b)
2 dT1 6(2q2o − A2 ) 3
b

for linear loss, we obtain

𝛾1 q2o − 𝛾2 A2b √
1 dxo
𝜙1p = x − (x − xo ) √ coth ( q2o − A2b (x − xo )), (3.12a)
2 dT1
(2q2o − A2b ) q2o − A2b

1 dxo x(x − 2xo )(𝛾1 q2o − 𝛾2 A2b )

𝜓1p = x − , (3.12b)
2 dT1 2(2q2o − A2 ) b

and, finally, in the case of nonlinear loss we find

3𝛾1 q6o + 2𝛾2 A4b (q2o − A2b ) √

1 dxo
𝜙1p = x + (x − xo ) √ coth ( q2o − A2b (x − xo )), (3.13a)
2 dT1
3(2q4o − 3A2b q2o + A4b ) q2o − A2b

x(x − 2xo )(3𝛾1 q4o − 2𝛾2 A4b ) 1 A2b √

1 dxo
𝜓1p = x − + 𝛾 ln cosh ( q2o − A2b (x − xo )). (3.13b)
2 dT1 6(2q2o − A2 ) 3 2 q2o − A2
b b

One can then write the full O(𝜀) contributions to the phases by adding the homogeneous solutions
in equation (3.10) and the perturbation-dependent particular solutions in equations (3.11), (3.12)
and (3.13). Note that in all cases the phase of the bright component, 𝜓, exhibits a chirp-like
quadratic growth as x → ±∞. However, we should point out that the limit |x| → ∞ is generically
outside the range of validity of the asymptotic expansion in small time scales, since for large x the
O(𝜀) terms are no longer necessarily smaller than the O(1) terms. Also, the asymptotic behaviour
of the x-derivative of the O(𝜀) contributions to the phase of the dark soliton, 𝜙1x , as x → ±∞ is listed
in table 3 below. Furthermore, a 𝜋 jump as x → ±∞ in the O(1) term phase of the dark component
√
has to be included on account of the limit of tanh ( q2o − A2b (x − xo )) in equations (3.4).
 Table 3. Asymptotic behaviour of the x-derivative of the O(𝜀) solutions for the phases of dark component as x → ±∞ for
10
the various perturbations considered.

royalsocietypublishing.org/journal/rspa Proc. R. Soc. A 481: 20240082

...........................................................................................................
perturbation (𝜙1 )±x
√
1 dxo A2b
qo − Ab 2 2 2 √
diffusion + c2 ∓ 𝛾 2 ∓ 𝛾1 qo2 − A2b
2 dT1 3(2qo2 − A2b ) 3
.........................................................................................................................................................................................................

1 dxo 𝛾1 qo2 − 𝛾2 A2b

linear loss + c2 ∓ √
2 dT1
(2qo2 − A2b ) qo2 − A2b
Downloaded from https://royalsocietypublishing.org/ on 27 January 2025

.........................................................................................................................................................................................................

1 dxo 𝛾1 qo2 (5qo2 − A2b ) − 2𝛾2 A4b

nonlinear loss + c2 ∓ √
2 dT1
(2qo2 − A2b ) qo2 − A2b
.........................................................................................................................................................................................................

(b) First-order corrections for the amplitudes

We now proceed to obtain information on the O(𝜀) perturbative terms for the amplitudes of dark
and bright components. As mentioned above, the system (3.7a) for the amplitudes is fully coupled,
and the complete set of homogeneous solutions is not available. Note, however, that as |x| → ∞
the homogeneous system reduces to

𝜕x2 u1h ∼ 4q2o u1h , 𝜕x2 v1h ∼ (q2o − A2b )v1h , (3.14)

and since both RHSs are positive (recall that 0 ≤ Ab < qo ), the homogeneous solutions for the
amplitudes decay exponentially (exponentially growing terms would be unphysical), and their
contributions can therefore be neglected compared to the particular solutions.
The particular solutions at O(𝜀) for the system (3.7a) for u1 , v1 can be obtained as follows.
Assume that in the RHS of system (3.7a), one has Re(R1 [uo ei𝜙o ]e−i𝜙o ) = Re(R2 [vo ei𝜓o ]e−i𝜓o ) = 0. This
condition is satisfied by all the perturbations considered here, namely, diffusion as well as linear
and nonlinear loss, in the case of a stationary DB soliton. Now recall the O(1) equations for uo , vo ,
and write the corresponding homogeneous system in the matrix form

𝜕x2 − 2u2o + 2q2o −2uo vo u 0

[ ] [ o ] = [ ]. (3.15)
−2uo vo 𝜕x2 − 2v2o + q2o + A2b vo 0

Taking derivatives of this system with respect to qo and Ab , we obtain two systems

𝜕x2 − 6u2o − 2v2o + 2q2o −4uo vo u −4qo uo

[ ] [ oqo ] = [ ], (3.16)
−4uo vo 𝜕x2 − 6v2o − 2u2o + q2o + A2b voqo −2qo vo

𝜕x2 − 6u2o − 2v2o + 2q2o −4uo vo u 0

[ ] [ oAb ] = [ ]. (3.17)
−4uo vo 𝜕x2 − 6v2o − 2u2o + q2o + A2b voAb −2Ab vo

Notice that both systems have exactly the same matrix as in system (3.7a), and since qo , Ab are
space-independent, we can seek solutions u1 , v1 to system (3.7a) in the following form:

u1 = A1 uoqo + A2 uoAb , v1 = A1 voqo + A2 voAb , (3.18)

where A1 , A2 , A3 and A4 are varying coefficients. Taking linear combinations of the two
aforementioned systems, we can match the RHS of system (3.7a) and obtain the u1 , v1 solution
 in terms of derivatives of unperturbed solutions with respect to their parameters, that is,
11
𝜎oT1
𝜕uo 𝜎oT − 2𝜑oT1 𝜕uo 𝜎oT1
𝜕vo 𝜎oT − 2𝜑oT1 𝜕vo
u1 = − ( ) +( 1 ) , v1 = − ( ) +( 1 ) . (3.19)

royalsocietypublishing.org/journal/rspa Proc. R. Soc. A 481: 20240082

...........................................................................................................
4qo 𝜕qo 4Ab 𝜕Ab 4qo 𝜕qo 4Ab 𝜕Ab
We note that to completely determine the O(𝜀) contributions to the amplitudes u1 , v1 one needs
to obtain the dependence on the slow time T1 of the phase parameters of the dark and bright
components of the soliton, namely d𝜎o ∕dT1 and d𝜑o ∕dT1 . The former will be obtained from the
O(𝜀) perturbed conservation laws in §6, but d𝜑o ∕dT1 is not determined at this order. On the other
hand, the terms involving d𝜑o ∕dT1 do not contribute to u1 , v1 in the limit as x → ±∞, and one can
obtain from the above equations the asymptotic behaviour:
𝜎oT1
u1 → ∓ , v1 → 0 as x → ±∞. (3.20)
Downloaded from https://royalsocietypublishing.org/ on 27 January 2025

4qo
Notice that the particular solutions above are obtained under the assumption that the real parts
of the perturbations in both components are zero. If one has Re(R1 [uo ]) = 𝛼1 uo , Re(R2 [vo ]) = 𝛼2 vo ,
where 𝛼1 , 𝛼2 are x-independent, then a particular solution can be built in a similar way.

4. The first-order correction for a moving DB soliton

For a moving DB soliton, we have 0 < Ad < qo , 0 < Ab < Ad and V ≠ 0, with all parameters being,
𝜏
in general, functions of the slow time T1 . In a co-moving reference frame: X = x − ∫0 V(𝜀s)ds − xo ,
𝜏 = t, the DB soliton solution is written as
√ √
qd = (sgn(V) q2o − A2d − iAd tanh(X A2d − A2b )) ei𝜎o , (4.1)

V ( 𝜏 ) V2 √
i(𝜑o + X+∫0 V(𝜀s)ds+xo −( −(A2d −A2b ))𝜏)
qb = Ab e 2 4 sech(X A2d − A2b ), (4.2)
and the perturbed Manakov system (1.4) takes the form
iq𝜏 − i(V + xo𝜏 )qX + qXX − 2(‖q‖2 − q2o )q = 𝜀R[q]. (4.3)
To avoid dealing with singularities in the derivatives with respect to the soliton parameters, in
the following, we will assume that V(0) > 0, and that V remains sign definite as a function of T1 .
After applying Madelung ansatz for each component, q1 = uei𝜙 , q2 = vei𝜓 , and separating real
and imaginary parts, equation (4.3) yields
−u𝜙𝜏 + Vu𝜙X + uXX − 𝜙X2 u − 2(u2 + v2 − q2o )u = 𝜀Re(R1 [uei𝜙 ]e−i𝜙 ), (4.4a)
i𝜙 −i𝜙
u𝜏 − VuX + u𝜙XX + 2uX 𝜙X = 𝜀Im(R1 [ue ]e ), (4.4b)
−v𝜓𝜏 + Vv𝜓X + vXX − 𝜓X2 v 2
− 2(u + v − 2
q2o )v = 𝜀Re(R2 [vei𝜓 ]e−i𝜓 ), (4.4c)
v𝜏 − VvX + v𝜓XX + 2vX 𝜓X = 𝜀Im(R2 [vei𝜓 ]e−i𝜓 ). (4.4d)
Expanding u, v, 𝜙, 𝜓 in powers of 𝜀 as in equation (3.3), and letting as before 𝜕𝜏 = 𝜕To + 𝜀𝜕T1 + O(𝜀2 ),
we find for the O(1) terms in the co-moving coordinates:
√ ⎛ ⎞
√ √
2 Ad
uo = q2o − A2d sech (X A2d − A2b ), 𝜙o = 𝜎o + arctan ⎜− √ tanh (X A2d − A2b )⎟, (4.5a)
⎜ ⎟
q2o − A2d
⎝ ⎠
√ To
V V2
vo = Ab sech(X A2d − A2b ), 𝜓 o = 𝜑o + (X + ∫ V ds + xo ) − ( − (A2d − A2b )) To . (4.5b)
2 0
4

Then, stationary solutions at O(𝜀) are governed by the following, fully coupled system of ODEs:
2
− uo 𝜙oT1 + Vu1 𝜙oX + Vuo 𝜙1X + u1XX − 2𝜙oX 𝜙1X uo − 𝜙oX u1 − 4u2o u1 − 4uo vo v1 − 2(u2o + v2o − q2o )u1

= Re(R1 [uo ei𝜙o ]e−i𝜙o ), (4.6a)

 uoT1 − Vu1X + uo 𝜙1XX + u1 𝜙oXX + 2u1X 𝜙oX + 2uoX 𝜙1X = Im(R1 [uo ei𝜙o ]e−i𝜙o ), (4.6b)
12
2
− vo 𝜓oT1 + Vv1 𝜓oX + Vvo 𝜓1X + v1XX − 2𝜓oX 𝜓1X vo − 𝜓oX v1 − 4v2o v1 − 4uo vo u1 − 2(u2o + v2o − q2o )v1

royalsocietypublishing.org/journal/rspa Proc. R. Soc. A 481: 20240082

...........................................................................................................
= Re(R2 [vo ei𝜓o ]e−i𝜓o ), (4.6c)
i𝜓o −i𝜓o
voT1 − Vv1X + vo 𝜓1XX + v1 𝜓oXX + 2v1X 𝜓0X + 2voX 𝜓1X = Im(R2 [vo e ]e ). (4.6d)

In the scalar case, for which vo = 𝜓o = 0, the above system reduces to

𝜕X2 + V𝜙oX − 𝜙oX

2
− 6u2o + 2q2o uo (V − 2𝜙oX )𝜕X u1 J [u , 𝜙 ]
[ ] [ ] = [ 1 o o ], (4.7)
(−V + 2𝜙oX )𝜕X + 𝜙oXX uo 𝜕X2 + 2uoX 𝜕X 𝜙1 J2 [uo , 𝜙o ]

J1 [uo , 𝜙o ] = Re(R1 [uo ei𝜙o ]e−i𝜙o ) + uo 𝜙oT1 , J2 [uo , 𝜙o ] = Im(R1 [uo ei𝜙o ]e−i𝜙o ) − uoT1 .
Downloaded from https://royalsocietypublishing.org/ on 27 January 2025

Now, from equation (4.5), we easily find

V2
𝜙oTo = −𝜙oX V, 𝜓oTo = − + A2d − A2b , (4.8)
4
√
2 2
V A d
sech (X q2o − A2b ) V
𝜙oX = √ , 𝜓oX = , (4.9)
2 2 2 2 2 2 2
Ad sech (X qo − Ab ) − qo

as well as

Ad AdT1 − Ab AbT1 AdT1 q2o − u2o qo qoT1

uoT1 = uoX (X − xoT1 ) − + , (4.10a)
A2d − A2b Ad uo uo

Ad AdT1 − Ab AbT1 AbT1

voT1 = voX (X − xoT1 ) + vo , (4.10b)
A2d − A2b Ab
√
⎛
Ad AdT1 − Ab AbT1 sinh (2X A2d − A2b ) q A − A q ⎞
qo o dT1 d oT1
𝜙oT1 = 𝜎oT1 + 𝜙oX ⎜X − xoT1 + √ ⎟, (4.10c)
⎜ A2d − A2b 2Ad 2 2
Ad − Ab q 2
o − A2
d
⎟
⎝ ⎠
To
V 1 V
𝜓oT1 = 𝜑oT1 + x + V (X + ∫ V ds + xo ) − ( V − 2(Ad AdT1 − Ab AbT1 )) To . (4.10d)
2 oT1 2 T1 0
2 T1

To this end, in the limit X → ±∞, equation (4.6a) yield

⎛ ⎞
1 qo AdT1 − Ad qoT1 ⎟
−qo ⎜𝜎oT1 ∓ √
±
+ Vqo 𝜙1X − 4q2o u±1 = Re(R1 [uo ei𝜙o ]e−𝜙o )± , (4.11a)
⎜ qo 2 2 ⎟
qo − Ad
⎝ ⎠
−Vu±1X + qo 𝜙1XX
±
+ qoT1 = Im(R1 [uo ei𝜙o ]e−𝜙o )± , (4.11b)
V2 ±
v = Re(R2 [vo ei𝜓o ]e−i𝜓o )± , (4.11c)
4 1
±
0 = Im(R2 [vo ei𝜓o ]e−i𝜓o )± , (4.11d)

where we have used u±1XX = v±1XX = 0 for a continuous wave background. Using equation (2.5),
the second equation simplifies to Vu±1X = qo 𝜙1XX±
. Assuming that u1X → 0 as X → ±∞, then both
u1 and 𝜙1X tend to constants as X → ±∞, which corresponds to the shelf developing around the
soliton. In this case, the second equation in the system is automatically satisfied and does not yield
any additional information. Also, for all perturbations considered in this work the third equation
implies v±1 = 0.
 5. Boundary layer 13
Since u1 , 𝜙1 do not vanish as x → ±∞, the solution to order 𝜀 does not match the boundary

royalsocietypublishing.org/journal/rspa Proc. R. Soc. A 481: 20240082

...........................................................................................................
conditions at infinity. Thus, our problem is now broken into two regions: the region that matches
imposed non-decaying boundary conditions at infinity, which is unaffected by the soliton, and the
region in which the O(𝜀) correction term is valid and the solution is quasi-stationary. To resolve
this mismatch, we introduce a boundary layer in which there is a transition from the non-zero
value in the perturbation term to the boundary conditions at infinity.
In the boundary-layer region, we can write

u = qo + 𝜀w + O(𝜀2 ), v = 𝜀p + O(𝜀2 ), 𝜙 = 𝜙± + 𝜀𝜃 + O(𝜀2 ), 𝜓 = 𝜓 ± + 𝜀𝜇 + O(𝜀2 ), (5.1)

where w, p, 𝜃, 𝜇 are real-valued functions of x and t. As before, utilizing two time scales, with
Downloaded from https://royalsocietypublishing.org/ on 27 January 2025

𝜕t → 𝜕To + 𝜀𝜕T1 + O(𝜀2 ), we can write

dqo d𝜙± dqo d𝜓 ±
= 0, = 0, = Im(R1 [qo ]), −qo = −Re(R2 [qo ]). (5.2)
dTo dTo dT1 dT1
At O(1), the system (3.2) is satisfied exactly, and at O(𝜀), we have
2
𝜕𝜙± 𝜕𝜙± 𝜕𝜃 𝜕2 w 𝜕𝜙± 𝜕𝜃 𝜕𝜙±
−qo −w − qo + 2 − 2 qo − ( ) w − 4q2o w = ReR1 [qo + 𝜀w], (5.3a)
𝜕T1 𝜕To 𝜕To 𝜕x 𝜕x 𝜕x 𝜕x
𝜕w dqo 𝜕2 𝜃 𝜕 2 𝜙± 𝜕w 𝜕𝜙± 𝜕qo 𝜕𝜃
+ + qo 2 + w 2 + 2 +2 = ImR1 [qo + 𝜀w], (5.3b)
𝜕To dT1 𝜕x 𝜕x 𝜕x 𝜕x 𝜕x 𝜕x
2
d𝜓± 𝜕 2 p 𝜕𝜓 ±
−p + 2 −( ) p = ReR2 [𝜀p], (5.3c)
dTo 𝜕x 𝜕x
𝜕p 𝜕2 𝜓± 𝜕p 𝜕𝜓 ±
+p + 2 = ImR2 [𝜀p]. (5.3d)
𝜕To 𝜕x2 𝜕x 𝜕x
Using equations (5.2), we obtain
𝜕𝜙±
ReR1 [qo + 𝜀w] + qo =ReR1 [qo + 𝜀w] − ReR[qo ] = O(𝜀), (5.4a)
𝜕T1
dqo
ImR1 [qo + 𝜀w] − = ImR1 [qo + 𝜀w] − ImR1 [qo ] = O(𝜀), (5.4b)
dT1
therefore these terms go to higher order. In addition, using the fact that 𝜙± , 𝜓 ± only depend on
the slow time T1 , we obtain the system of equations:

𝜕𝜃 𝜕2 w 𝜕w 𝜕2 𝜃 𝜕2 p 𝜕p
− qo + 2 − 4q2o w = 0, + qo 2 = 0, −2pq2o + = 0, = 0, (5.5)
𝜕To 𝜕x 𝜕To 𝜕x 𝜕x2 𝜕To
√ √
2qo x 2qo x
from which it follows that p = C1 e + C2 e − , with C1 , C2 independent of To . In turn, w, 𝜃
satisfy
𝜕𝜃 𝜕2 w 𝜕w 𝜕2 𝜃
= −4q2o w + 2 , = −qo 2 . (5.6)
𝜕To 𝜕x 𝜕To 𝜕x
The above system can be decoupled by taking two derivatives in x, which yields two identical
equations for w and 𝜃:
𝜕2 w 2
2𝜕 w 𝜕4 w 𝜕2 𝜃 2
2𝜕 𝜃 𝜕4 𝜃
= 4qo − , = 4qo − . (5.7)
𝜕To2 𝜕x2 𝜕x4 𝜕To2 𝜕x2 𝜕x4

Note that the O(𝜀) perturbation 𝜇 of 𝜓 ± (cf. equations (5.1)) is not determined in this order.
Assuming a long-wave approximation, the above equations become
𝜕2 w 2
2𝜕 w 𝜕2 𝜃 2
2𝜕 𝜃
= 4qo , = 4qo , (5.8)
𝜕To2 𝜕x2 𝜕To2 𝜕x2
 that is, the O(𝜀) perturbative terms of amplitude and phase of the dark soliton satisfy the wave
14
equation. This allows us to conclude that, similarly to the scalar case, the edges of the shelf move
with the velocities W = ±2qo .

royalsocietypublishing.org/journal/rspa Proc. R. Soc. A 481: 20240082

...........................................................................................................
6. Perturbed conservation laws
(a) Stationary DB soliton
To find the slow time evolution of a stationary DB soliton parameters, Ab , 𝜎o and xo , we employ
the first three conservation laws for the unperturbed Manakov system, involving
Downloaded from https://royalsocietypublishing.org/ on 27 January 2025

∞
[ ]
E=∫ ‖q‖2 − q2o dx (Energy), (6.1a)
−∞
∞
[ ]
I = 2 Im ∫ q ⋅ qx dx − q2o 𝛥𝜙o (Renormalized momentum), (6.1b)
−∞
∞
[ ]
H =−∫ ‖qx ‖2 + (‖q‖2 − q2o )2 dx (Hamiltonian), (6.1c)
−∞

where ⋅ denotes the usual complex vector dot product: (a, b)T ⋅ (c, d)T = ac∗ + bd∗ , and 𝛥𝜙o is the
phase difference across the soliton. The evolution equations for these integrals of motion take the
form
∞
dE dqo
= 2𝜀 (−qo + Im [∫ R[q] ⋅ qdx]), (6.2a)
dt dT1 −∞
∞
dI dq2o d𝛥𝜙o
= −4𝜀Re ∫ R[q] ⋅ qx dx − 𝜀𝛥𝜙o − 𝜀q2o , (6.2b)
dt −∞ dT 1 dT1
∞
dH d 2
= 2𝜀 (E q + Re ∫ R[q] ⋅ qt dx). (6.2c)
dt dT1 o −∞

Matching O(𝜀) terms in dE∕dt and equation (6.2a), we obtain

∞ ∞
d dqo
∫ 2[uo u1 + 2vo v1 ]dx = −2qo + 2Im [∫ R[q] ⋅ qdx]. (6.3)
dt −∞ dT1 −∞

Since the edges of the shelf are moving with velocity ±2qo (see end of the previous section), we
change the interval of integration in the LHS to (−2qo t, 2qo t), and the Fundamental Theorem of
Calculus yields
∞
[ ] dqo
4qo uo (2qo t)u1 (2qo t) + uo (−2qo t)u1 (−2qo t) = −2qo + 2Im [∫ R[q] ⋅ qdx]. (6.4)
dT1 −∞

In the limit x → ±∞, using uo → ±qo , u1 → u±1 , the above equation reduces to
∞
dqo
4q2o (u+
1
− u−
1
) = −2qo + 2Im [∫ R[q] ⋅ qdx]. (6.5)
dT1 −∞

Recalling the definitions of u±1 in equation (3.20), we finally get

∞
d𝜎o dqo 1
= − Im [∫ R[q] ⋅ qdx]. (6.6)
dT1 dT1 qo −∞

Matching O(𝜀) terms in dI∕dt and equation (6.2b) and proceeding in a similar way we obtain
+ −
𝜙1x + 𝜙1x = 0, (6.7)
 (a) (b)
20 20 15

0.8 0.9

royalsocietypublishing.org/journal/rspa Proc. R. Soc. A 481: 20240082

...........................................................................................................
15 15
0.6 0.8
10 10
t

t
0.7
0.4
5 5 0.6
0.2
0.5
0 0
–40 –20 0 20 40 –40 –20 0 20 40
x x
Downloaded from https://royalsocietypublishing.org/ on 27 January 2025

Figure 1. (a) Contour

√ plot of the modulus of the dark component of a stationary DB soliton under nonlinear loss. Here, qo (0) =
1, Ab (0) = 1∕ 2, Ad = qo , x0 = 0 and 𝜖𝛾1 = 𝜖𝛾2 = −0.01. (b) Contour plot of the modulus √ of the dark component
of a moving DB soliton under linear loss. Here, parameter values are qo (0) = 1, Ab (0) = 1∕ 2, Ad = 0.8, x0 = 0 and
𝜖𝛾1 = 𝜖𝛾2 = −0.01. In both plots, the dashed red lines correspond to the analytical prediction for the evolution of the shelf’s
edge.

which in turn, in light of the asymptotic behaviour of 𝜙1x in table 3, implies

dxo
+ 2c2 = 0. (6.8)
dT1
The last equation relates the arbitrary phase constant c2 for a stationary DB soliton to the slow-time
evolution of the soliton centre. Although in this work we do not determine xo (T1 ), our numerical
simulations indicate that for all three types of perturbations considered in table 1 the soliton centre
does not move, which indicates that c2 = 0. We note that the phase constant can be independently
determined by analysing the radiation field associated with the soliton, as in [52–54] in the context
of instability-induced dynamics of dark solitons.
Finally, matching O(𝜀) terms in dH∕dt and (6.2c), yields
√ √ dqo ∞
2 d
− ( q2o − A2b (4q2o − A2b )) = −8qo q2o − A2b + 2(q2o − A2b ) Im ∫ R2 [vo ei𝜓o ]vo e−i𝜓o dx, (6.9)
3 dT1 dT1 −∞

from where, knowing dqo ∕dT1 (see equation (2.5)), and the form of the perturbation R[q], we can
find dAb ∕dT1 (cf. table 1 for some specific perturbations).
Next, we proceed to compare the above findings with direct numerical simulations. We
present here the case of nonlinear loss as an example, but we note that for the other considered
perturbations the results, and the agreement between the analytical predictions and simulations,
are similar. For this example, we evolve equation (1.4) using equation (1.6a) as initial conditions
for the case of a stationary DB soliton, with Ad = qo , using a fourth-order Runge–Kutta method in
time. In the left panel of figure 1, we depict the complete evolution of the black soliton, so as to
demonstrate the evolution of the shelf, under the nonlinear loss. The evolution of the DB soliton
parameters Ad and qo is depicted in figure 2. It is observed that the analytical predictions are in
very good agreement with the results of the direct numerical simulations.
Below we will consider the perturbed conservation laws for the general case of a moving
DB soliton, and obtain from them explicit equations for the adiabatic evolution of the soliton
parameters for the various perturbations. The reductions to stationary DB solitons and to the
scalar case will also be given.

(b) Moving DB soliton

For a moving DB soliton, we have eight independent parameters: u±1 , 𝜙1X ±
and slow evolution of
variables Ab , Ad , qo and 𝜎o . Other parameters of the first-order correction of the solution, namely
 1
Numerical Numerical 16
Analytical Analytical
0.7

royalsocietypublishing.org/journal/rspa Proc. R. Soc. A 481: 20240082

...........................................................................................................
Ab 0.95

qo
0.65
0.9

0.6 0.85
0 5 10 15 20 0 5 10 15 20
t t
Downloaded from https://royalsocietypublishing.org/ on 27 January 2025

Figure 2. The evolution of the stationary DB soliton parameters Ad and qo under nonlinear loss.The solid blue lines correspond
to the numerical
√ simulations results, whereas the dashed red lines correspond to the analytical predictions. Here, qo (0) = 1,
Ab (0) = 1∕ 2, Ad = qo , x0 = 0 and 𝜖𝛾1 = 𝜖𝛾2 = −0.01.

±
𝜓1X , are not determined at this order of perturbation. Determining the slow time evolution of xo
requires the explicit expression of the first-order corrections for amplitudes and phases and will
be the subject of future investigation.
As in the case of stationary DB soliton discussed in §5, the edge of the shelf still propagates
with velocity W(T1 ) = 2qo (T1 ), and the velocity V of the soliton now may also depend on the slow
t
time T1 . In the moving frame of reference X = x − ∫0 V(𝜖s)ds − xo , the boundaries of the shelf are
given by

t t
SL (t) = − ∫ [2qo (𝜖s) + V(𝜖s)]ds, SR (t) = ∫ [2qo (𝜖s) − V(𝜖s)]ds, (6.10)
0 0

where SL and SR give the position in X of the left and right boundaries of the shelf, respectively.
In addition to E, I and H used above, here we will also employ the conserved quantity:

∞
1
F = Im ∫ [‖q‖2 (q ⋅ qX ) + qXX ⋅ qX ] dX, (6.11)
−∞
3

which was found through symbolical software [55,56]; the time evolution of F is given by

∞
dF 1
= −2𝜀Re ∫ R[q] ⋅ ((‖q‖2 + qX ⋅ q)qX − qXXX ) dX. (6.12)
dt −∞
3

Note that each term in the integrand of the conserved quantity (6.11) involves spatial derivatives
of the field q, which decay at infinity, thus ensuring the convergence of the integral. Furthermore,
unlike what happens for the momentum equation (6.1b) (see e.g. [53]), since there is no
straightforward physical quantity associated with this higher moment F, there is no underlying
physical motivation that requires a renormalization of the integral.
Furthermore, we will use the so-called centre of energy:

∞
ℛ=∫ X(‖q‖2 − q2o )dX (Centre of energy), (6.13)
−∞

the time evolution of which is related to the momentum I:

∞
dℛ
= I + 2𝜀Im ∫ X(R[q] ⋅ q − qo R[qo ])dX. (6.14)
dt −∞
 Repeating the same steps as in the previous subsection while restricting the interval of integration
17
to X ∈ [SR , SL ] for the d∕dTo terms, we obtain four evolution equations:

royalsocietypublishing.org/journal/rspa Proc. R. Soc. A 481: 20240082

...........................................................................................................
∞
dEo +
dqo
= −2qo ((2qo + V)u−
1
+ (2qo − V)u1
) − 2qo + 2Im ∫ R[qo ] ⋅ qo dX, (6.15a)
dT1 dT1 −∞
∞
dIo +
d(q2o 𝛥𝜙) 𝜕qo
= 2q2o ((2qo + V)𝜙1X
−
+ (2qo − V)𝜙1X )− − 4Re ∫ R[qo ] ⋅ dX, (6.15b)
dT1 dT1 −∞ 𝜕X
∞
dHo dqo 𝜕qo
= 4Eo qo + 2Re ∫ R[qo ] ⋅ dX, (6.15c)
dT1 dT1 −∞ 𝜕To
dFo ( + )
= −q4o 𝜙1X −
(2qo − V) + 𝜙1X (2qo + V)
Downloaded from https://royalsocietypublishing.org/ on 27 January 2025

dT1
∞ 3
𝜕qo 𝜕qo 1 𝜕 qo
− 2Re ∫ R[qo ] ⋅ ((‖qo ‖2 + ⋅ qo ) − ) dX, (6.15d)
−∞ 𝜕X 𝜕X 3 𝜕X3

where Eo , Io , Ho and Fo are the values of the corresponding conservation laws integrals computed
for 𝜀 = 0:
√
Eo = −2 A2d − A2b , (6.16a)
√
4
Io = (A2 − A2b ) q2o − A2d , (6.16b)
Ad d
√
2 A2d − A2b ( )
Ho = − 4A2d (A2d − A2b ) + 3q2o A2b , (6.16c)
3A2d
( )√ 2 ( ( ) )
2 A2d − A2b qo − A2d A2b q2o − 2A2d − 3q2o A2d + 2A4d
Fo = . (6.16d)
3A3d

In addition, we obtain one non-evolution equation coming from the leading order of dℛ∕dt:

(2qo − V)(2u+
1
+
+ 𝜙1X ) − (2qo + V)(2u−
1
−
− 𝜙1X ) = 0. (6.17)

Putting everything together, equation (6.15), along with equations (6.17), (4.11a) and (2.5) give us
a linear system of equations for eight unknowns: u±1 , 𝜙1X±
, and slow time derivatives of 𝜎o , Ab , Ad
and qo . In what follows, we provide solutions of these equations for each of the three perturbations
considered in the paper.

(c) DB soliton parameters

Diffusion perturbations: Rj = i𝛾j qjxx , j = 1, 2. In this case, the adiabatic evolution of dAb ∕dT1 ,
dAd ∕dT1 , d𝜎o ∕dT1 is described by the following equations:
( )( ( ) ( ) )
dAb 2𝛾2 Ab A2b − A2d A2b 7q2o A4d − 4A6d + A4b 2q2o A2d − 3q4o − 6q2o A6d + 4A8d
=− ( ) , (6.18a)
dT1 3A2d −4A2b A4d + q2o A4b + 4A6d
2
( 2 2
)( 2 2
)( 2 ( 2 2
) 4
)
dAd 2𝛾2 Ab Ab − Ad Ad − qo Ab 4Ad − 3qo − 4Ad
= ( ) , (6.18b)
dT1 3 q2o A4b Ad − 4A2b A5d + 4A7d
d𝜎o √ [ ( )]−1 [ 4 ( )
= 2 A2d − A2b 3qo −4A2b A4d + q2o A4b + 4A6d × Ab 8𝛾2 A4d + 2 (𝛾1 − 6𝛾2 ) q2o A2d + 3𝛾2 q4o
dT1
( ) ]
−2A2b 4 (𝛾1 + 𝛾2 ) A6d − 5𝛾2 q2o A4d + 8𝛾1 A8d . (6.18c)
 Notice that in the scalar limit (Ab = 0), the full system of equations reduces to
18
d𝜎o 4𝛾1 A3d dAd dqo dV 1 d𝜎o 1
= , = = = 0, u±1 = − , ±
𝜙1X = ±2u±1 , (6.19)

royalsocietypublishing.org/journal/rspa Proc. R. Soc. A 481: 20240082

...........................................................................................................
dT1 3qo dT1 dT1 dT1 2 dT1 2qo ∓ V

consistently with the results [47], whereas in the stationary limit (V = 0, Ad = qo ), we find
√
d𝜎o 2qo q2o − A2b (4𝛾1 q2o − (2𝛾1 − 𝛾2 )A2b ) dAb 2𝛾2 Ab (q2o − A2b )2
= , =− , (6.20a)
dT1 3(2q2o − A2b ) dT1 3(2q2o − A2 ) b

dV dqo dAd 1 d𝜎o 1

= = = 0, u±1 = − , ±
𝜙1X = ±2u±1 . (6.20b)
dT1 dT1 dT1 2 dT1 2qo
Downloaded from https://royalsocietypublishing.org/ on 27 January 2025

Linear loss perturbations: Rj = i𝛾j qj , j = 1, 2. The adiabatic evolution of dAb ∕dT1 , dAd ∕dT1 ,
d𝜎o ∕dT1 now reads
6 3
( 4 2 2
) 2 5
dAb 2 (𝛾1 + 2𝛾2 ) Ab Ad − Ab 6𝛾2 Ad + 𝛾1 qo Ad + 2𝛾2 qo Ab
= , (6.21a)
dT1 −4A2b A4d + q2o A4b + 4A6d
2 3
( 2 2
) 4
( 2 3
) 7
dAd −2𝛾1 Ab Ad Ad + qo + Ab (𝛾1 + 2𝛾2 ) qo Ad − 2𝛾2 Ad + 4𝛾1 Ad
= , (6.21b)
dT1 −4A2b A4d + q2o A4b + 4A6d
2
( 2 2
)( 2 ( 2 2
) 4
)
d𝜎o 2Ad 𝛾1 Ad − 𝛾2 Ab Ab qo − 2Ad + 2Ad
= √ . (6.21c)
dT1 ( )
qo A2d − A2b −4A2b A4d + q2o A4b + 4A6d

In the scalar limit, this gives

d𝜎o 𝛾1 Ad dAd dV
= , = 𝛾1 A d , = 𝛾1 V, (6.22a)
dT1 qo dT1 dT1
dqo 1 d𝜎o 1
= 𝛾1 qo , u±1 = − , ±
𝜙1X = ±2u±1 , (6.22b)
dT1 2 dT1 2qo ∓ V

again consistently with [47], and in the stationary limit, we get

d𝜎o 2qo (𝛾1 q2o − 𝛾2 A2b ) 2 2 2

dAb Ab (𝛾1 qo + 2𝛾2 (qo − Ab ))
=√ , = , (6.23a)
dT1
q2o − A2b (2q2o − A2b )
dT1 2q2o − A2 b

dV dAd dqo 1 d𝜎o 1

= 0, = = 𝛾1 qo , u±1 = − , ±
𝜙1X = ±2u±1 . (6.23b)
dT1 dT1 dT1 2 dT1 2qo

Nonlinear loss perturbations: Rj = i𝛾j |qj |2 qj . In this case, the adiabatic evolution of dAb ∕dT1 ,
dAd ∕dT1 , d𝜎o ∕dT1 is given by
( 4 4 6
( 2 2
) 4 2 2 2 6 8
)
dAb Ab −12𝛾2 Ab Ad + 2Ad 4𝛾2 Ab + 5𝛾1 qo − 3𝛾1 qo Ab Ad + 4𝛾2 qo Ab − 4𝛾1 Ad
= ( ) , (6.24a)
dT1 3 −4A2b A4d + q2o A4b + 4A6d
( 6
( 2 2
) 2 2
( 2 2 4 4
) 4 4 6
( 2 2
))
dAd Ad 4𝛾2 Ab qo − Ad + 2𝛾1 Ab Ad −5qo Ad + 2Ad − 3qo + 3𝛾1 qo Ab + 4𝛾1 Ad 5qo − 2Ad
= ( ) ,
dT1 3 −4A2 A4 + q2o A4 + 4A6 b d b d
(6.24b)

d𝜎o √ ( ) −1 [ ( ( ) )
= 2A2d [3qo A2d − A2b −4A2b A4d + q2o A4b + 4A6d ] −2𝛾2 A4b A2b q2o − 2A2d + 2A4d +
dT1
( ( ) ( ) ( ))]
+ 𝛾1 −2q2o A2d 7A2b A2d + A4b − 7A4d + 3q4o A2b A2b + A2d + 4A6d A2b − A2d . (6.24c)
 0.7 0.8 1
Numerical Numerical Numerical
Analytical Analytical Analytical 19
0.95
0.65 0.75

royalsocietypublishing.org/journal/rspa Proc. R. Soc. A 481: 20240082

...........................................................................................................
Ab

Ad

qo
0.9
0.6 0.7
0.85

0.55 0.65 0.8

0 5 10 15 20 0 5 10 15 20 0 5 10 15 20
t t t
Figure 3. The evolution of the moving DB soliton parameters under linear loss.The solid blue lines correspond to the numerical
√
simulations results, whereas the dashed red lines correspond to the analytical predictions. Here, qo (0) = 1, Ab (0) = 1∕ 2,
Ad = 0.8, x0 = 0 and 𝜖𝛾1 = 𝜖𝛾2 = −0.01.
Downloaded from https://royalsocietypublishing.org/ on 27 January 2025

In the scalar limit, this yields

d𝜎o 𝛾1 Ad 1 2 dAd 1 dV 𝛾 V
= ( V + 5q2o ) , = 𝛾1 Ad ( V 2 + q2o ) , = 1 (V 2 + 2q2o ), (6.25a)
dT1 3qo 2 dT1 6 dT1 6
d𝜎o d𝛥𝜙o
( ± )
dqo 1 dT1 dT1 2
= 𝛾1 q3o , u±1 =− , ±
𝜙1X = ±2u±1 ∓ 𝛾1 Ad , (6.25b)
dT1 2 2qo ∓ V 3

where

d𝛥𝜙o 1 qo AdT1 − Ad qoT1

= √ , (6.26)
dT1 qo
q2o − A2d

with 𝛥𝜙o being the phase difference across the soliton (which is different from the asymptotic
phase difference of the background, 𝛥𝜙∞ ). Finally, in the stationary limit, we have
2 2 2 4 4 2 2 2
d𝜎o 2qo (𝛾1 qo (5qo − Ab ) − 2𝛾2 Ab ) dAb Ab (3𝛾1 qo + 4𝛾2 Ab (qo − Ab ))
= √ , = 2 2
, (6.27a)
dT1 dT1 3(2qo − A )
3(2q2o − A2b ) q2o − A2b b

dV dAd dqo 1 d𝜎o 1

= 0, = = 𝛾1 q3o , u±1 = − , ±
𝜙1X = ±2u±1 . (6.27b)
dT1 dT1 dT1 2 dT1 2qo

As in the stationary case, it is relevant to compare the above findings with results of direct
numerical simulations. Here, as an example, we consider the case of linear loss (for the other
considered perturbations, the results are similar). For the present example, we again evolve
equation (1.4), using equations (1.6) as initial conditions, employing a fourth-order Runge–Kutta
method in time. In figure 1, we depict the complete evolution of the grey component of the moving
DB soliton, so as to demonstrate the emergence and evolution of the shelf in the presence of linear
loss. Furthermore, the evolution of the DB solitons amplitudes Ad and Ab and of the background
parameter qo are depicted in figure 3. Once again, the analytical results are found to be in very
good agreement with the numerical simulations.
Finally, it is relevant to compare the evolution of the grey soliton component of a moving DB
soliton with that of a grey soliton in the scalar case. An important conclusion stemming from such
a comparison is that the coupling (or the ‘symbiosis’) between the grey and the bright soliton has
a profound effect. Indeed, it seems that the bright soliton tends to reduce the decaying effects
of the dissipative perturbations as compared to the scalar equation. While the dark soliton will
eventually decay to the background, and the background will vanish in both cases, in the Manakov
system this is delayed significantly as shown in figure 4. This result is in a qualitative agreement
with the analysis [57], where it was found that the effect of the bright (filling) soliton component
is to partially stabilize ‘bare’ dark solitons against temperature-induced dissipation in BECs, thus
providing longer lifetimes.
 0.8
Numerical 20
Analytical

royalsocietypublishing.org/journal/rspa Proc. R. Soc. A 481: 20240082

...........................................................................................................
0.75

Ad
Coupled
system
0.7

Single
equation
0.65
0 5 10 15 20
Downloaded from https://royalsocietypublishing.org/ on 27 January 2025

t
Figure 4. The evolution of the solitons’ parameters under linear loss. The solid lines correspond to√the numerical simulations
results, whereas the dashed lines to the analytical predictions. Here, qo (0) = 1, Ab (0) = 1∕ 2, Ad = 0.8, x0 = 0 and
𝜖𝛾1 = 𝜖𝛾2 = −0.01.

7. Conclusion
In conclusion, we have presented a direct perturbation theory to study the evolution of the dark–
bright solitons of the Manakov system under the action of perturbations. Our approach relies on
the combination of a multi-scale expansion method and a boundary-layer theory, such that the
problem is broken into an inner region—pertinent to the soliton core—and a shelf, namely a linear
wave emerging owing to the perturbation, which matches the boundary conditions at infinity. Our
analysis was first performed for stationary DB solitons, and then was generalized for moving ones.
Various typical perturbations were considered, and in particular physically relevant dissipative
ones, namely diffusion, as well as linear and nonlinear loss (or gain, depending on the sign of the
relevant coefficients).
Our methodology is similar to the one that was used for the study of perturbed dark solitons in
the scalar defocusing NLS equation [47]. Our results, however, extend beyond the ones presented
in that work. For example, in the simpler case of stationary DB solitons, we were able to completely
determine the asymptotic phases of the DB solitons (whereas in [47] the particular solution of
the linear, non-homogeneous equations for the O(𝜀) terms in the phases had not been obtained).
Furthermore, the additional effect of the perturbation in the bright component on the asymptotic
phases of both dark and bright soliton components was studied.
For the typical dissipative perturbations we considered, the asymptotic approximations
showed very good agreement with results of direct numerical simulations. The relevant numerical
computations confirmed the emergence of the analytically predicted shelf and, in all cases, were
found to fully support our analytical predictions. Our analysis and computations also revealed an
important difference between scalar grey solitons and the grey solitons of the Manakov system:
it was found that the presence of the bright (filling) component hinders the perturbation-induced
dissipation associated with the grey soliton, thus offering a partial stabilization—i.e. a longer
lifetime—to the corresponding Manakov DB soliton structure, in comparison to its bare scalar
dark soliton counterpart.
Our analysis and results suggest further interesting studies. Indeed, first we note that it would
be relevant to further extend our analysis to determine the evolution of the centre xo of the DB
soliton. To do this, one needs to employ the O(𝜀) correction terms for amplitudes and phases
explicitly. In principle, one could follow a similar strategy as in [47], but in the Manakov system,
the ODEs for the O(𝜀) ODEs for the amplitudes and phases do not decouple, and the solution of
this problem will require finding homogeneous and particular solutions for a fourth-order linear,
non-homogeneous ODE, with coefficients given by the O(1) dark and bright soliton solutions and
their derivatives, while the non-homogeneous term depends on the perturbation. In any case, the
determination of the evolution of the soliton centre would also be relevant for studies involving
 conservative perturbations, as, e.g. is the case with external potentials, which are particularly
21
relevant to the physics of BECs. Solution of the relevant problem for the soliton centre could
then bridge our analysis with other perturbative studies of the Manakov system, relying on the

royalsocietypublishing.org/journal/rspa Proc. R. Soc. A 481: 20240082

...........................................................................................................
adiabatic approximation [18].

Data accessibility. This paper has no additional data.

Declaration of AI use. We have not used AI-assisted technologies in creating this article.
Authors’ contributions. A.C.: Conceptualization, Formal analysis, Investigation, Methodology, Software,
Validation, Visualization, Writing—original draft, Writing—review & editing; D.J.F.: Conceptualization,
Formal analysis, Investigation, Methodology, Software, Validation, Visualization, Writing—original draft,
Writing—review & editing; T.P.H.: Conceptualization, Formal analysis, Investigation, Methodology,
Software, Validation, Visualization, Writing—original draft, Writing—review & editing; G.N.K.: Conceptuali-
zation, Formal analysis, Investigation, Methodology, Software, Validation, Visualization, Writing—review
Downloaded from https://royalsocietypublishing.org/ on 27 January 2025

& editing; B.P.: Conceptualization, Formal analysis, Investigation, Project administration, Supervision,
Validation, Visualization, Writing—original draft, Writing—review & editing.
All authors gave final approval for publication and agreed to be held accountable for the work performed
therein.
Conflict of interest. We declare we have no competing interests.
Funding. B.P. gratefully acknowledges partial support for this work from the US National Science Foundation
under grant DMS-2406626. The research work of G.N.K. was supported by the Hellenic Foundation for
Research and Innovation (HFRI) under the HFRI PhD Fellowship grant (Fellowship Number: 5860).
Acknowledgements. A.C. and B.P. would like to acknowledge very fruitful conversations on the subject of this
work with Dmitry Pelinovsky, Willy Hereman, Stephen Anco and Panos Kevrekidis.

References
1. Zabusky NJ, Kruskal MD. 1965 Interaction of solitons in a collisionless plasma and the
recurrence of initial states. Phys. Rev. Lett. 15, 240–243. (doi:10.1103/physrevlett.15.240)
2. Ablowitz MJ, Segur H. 1981 Solitons and the inverse scattering transform. Philadelphia, PA: SIAM.
3. Zakharov VE. 1968 Stability of periodic waves of finite amplitude on the surface of a deep
fluid. Sov. Phys. J. Appl. Mech. Tech. Phys. 4, 190–194. (doi:10.1007/bf00913182)
4. Benney DJ, Roskes GJ. 1969 Wave instabilites. Stud. App. Math. 48, 377–385. (doi:10.1002/
sapm1969484377)
5. Zakharov VE. 1972 Collapse of Langmuir waves. Sov. Phys. JETP 35, 908–914.
6. Hasegawa A, Tappert F. 1973 Transmission of stationary nonlinear optical pulses in dispersive
dieletric fibres I. Anomalous dispersion. Appl. Phys. Lett. 23, 142–144. (doi:10.1063/1.1654836)
7. Hasegawa A, Tappert F. 1973 Transmission of stationary nonlinear optical pulses in dispersive
dieletric fibres II. Normal dispersion. Appl. Phys. Lett. 23, 171–173. (doi:10.1063/1.1654847)
8. Kalinikos BA, Kovshikov NG, Patton CE. 1977 Decay-free microwave envelope soliton
pulse trains in yittrium iron garnet thin films. Phys. Rev. Lett. 78, 2827–2830. (doi:10.1103/
physrevlett.78.2827)
9. Zvezdin AK, Popkov AF. 1983 Contribution to the nonlinear theory of magnetostatic sprin
waves. Sov. Phys. JETP 57, 350–355.
10. Pethick CJ, Smith H. 2002 Bose–Einstein condensation in dilute gases. Cambridge, UK: Cambridge
University Press. (doi:10.1017/cbo9780511802850)
11. Kivshar YS, Agrawal GP. 2003 Optical solitons: from fibres to photonic crystals. San Diego, CA:
Academic Press.
12. Manakov SV. 1974 On the theory of two-dimensional stationary self-focusing electromagnetic
waves. Sov. Phys. JETP 38, 248–253.
13. Menyuk CR. 1987 Nonlinear pulse propagation in birefringent optical fibres. IEEE J. Quant.
Elect. 23, 174–176. (doi:10.1109/jqe.1987.1073308)
14. Menyuk CR. 1999 Application of multiple-length-scale methods to the study of optical fibre
transmission. J. Eng. Math. 36, 113–136.
15. Pitaevskii L, Stringari S. 2003 Bose–Einstein condensation. Oxford, UK: Oxford University Press.
(doi:10.1023/b:joss.0000028243.07395.b3)
16. Kevrekidis PG, Frantzeskakis DJ, Carretero-González R. 2009 Emergent nonlinear phenomena in
Bose–Einstein condensates: theory and experiment. Berlin, Germany: Springer. (doi:10.1007/978-
3-540-73591-5_1)
 17. Kevrekidis PG, Frantzeskakis DJ, Carretero-González R. 2015 The defocusing nonlinear
22
Schrödinger equation. Philadelphia, PA: SIAM. (doi:10.1137/1.9781611973945)
18. Kevrekidis PG, Frantzeskakis DJ. 2016 Solitons in coupled nonlinear Schrödinger models: a

royalsocietypublishing.org/journal/rspa Proc. R. Soc. A 481: 20240082

...........................................................................................................
survey of recent developments. Rev. Phys. 1, 140–153. (doi:10.1016/j.revip.2016.07.002)
19. Busch Th, Anglin JR. 2001 Dark-bright solitons in inhomogeneous Bose–Einstein condensates.
Phys. Rev. Lett. 87, 010401. (doi:10.1103/physrevlett.87.010401)
20. Becker C, Stellmer S, Soltan-Panahi P, Dörscher S, Baumert M, Richter EM, Kronjäger J,
Bongs K, Sengstock K. 2008 Oscillations and interactions of dark and dark–bright solitons
in Bose–Einstein condensates. Nat. Phys. 4, 496. (doi:10.1038/nphys962)
21. Hamner C, Chang JJ, Engels P, Hoefer MA. 2011 Generation of dark-bright soliton trains
in superfluid-superfluid counterflow. Phys. Rev. Lett. 106, 065302. (doi:10.1103/physrevlett.
106.065302)
Downloaded from https://royalsocietypublishing.org/ on 27 January 2025

22. Kivshar YS, Turitsyn SK. 1993 Vector dark solitons. Opt. Lett. 18, 337–339. (doi:10.1364/
ol.18.000337)
23. Radhakrishnan R, Lakshmanan M. 1995 Bright and dark soliton solutions to coupled nonlinear
Schrödinger equations. J. Phys. A. 28, 2683–2692. (doi:10.1088/0305-4470/28/9/025)
24. Sheppard AP, Kivshar YS. 1997 Polarized dark solitons in isotropic Kerr media. Phys. Rev. E
55, 4773–4782. (doi:10.1103/physreve.55.4773)
25. Nakkeeran K. 2001 Exact dark soliton solutions for a family of N coupled nonlinear
Schrödinger equations in optical fibre media, Phys. Rev. E 64, 046611. (doi:10.1103/physreve.
64.046611)
26. Prinari B, Ablowitz MJ, Biondini G. 2006 Inverse scattering transform for the vector nonlinear
Schrödinger equation with nonvanishing boundary conditions. J. Math. Phys. 47, 063508.
(doi:10.1063/1.2209169)
27. Dean G, Klotz TK, Prinari B, Vitale F. 2013 Dark-dark and dark-bright soliton interactions
in the two-component defocusing nonlinear Schrödinger equation. Applicable Analysis 92,
379–397. (doi:10.1080/00036811.2011.618126)
28. Romero-Ros A, Katsimiga GC, Kevrekidis PG, Prinari B, Biondini G, Schmelcher P. 2022 On-
demand generation of dark-bright soliton trains in Bose-Einstein condensates. Phys. Rev. A
105, 023325. (doi:10.26226/m.6275705866d5dcf63a3113f5)
29. Kaup DJ. 1976 Perturbation expansion for Zakharov-Shabat inverse scattering transform.
SIAM J. Appl. Math. 31, 121–133. (doi:10.1137/0131013)
30. Karpman VI, Maslov EM. 1977 Perturbation-theory for solitons. Zh. Eksp. Teor. Fiz. 73, 537–559.
31. Kodama Y, Ablowitz MJ. 1980 Perturbations of solitons and solitary waves. Stud. App. Math.
64, 225–245. (doi:10.1002/sapm1981643225)
32. Herman RL. 1990 A direct approach to studying soliton perturbations. J. Phys. A Math. Gen.
23, 2327–2362. (doi:10.1088/0305-4470/23/12/017)
33. Yang J. 1999 Multisoliton perturbation theory for the Manakov equations and its applications
to nonlinear optics. Phys. Rev. E 59, 2393–2405. (doi:10.1103/physreve.59.2393)
34. Zhao W, Bourkoff E. 1989 Propagation properties of dark solitons. Opt. Lett. 14, 703–705.
(doi:10.1364/ol.14.000703)
35. Giannini JA, Joseph RI. 1990 The propagation of bright and dark solitons in lossy optical fibres,
IEEE J. Quantum Electron. 26, 2109–2114. (doi:10.1109/3.64346)
36. Lisak D, Anderson M, Malomed BA. Dissipative damping of dark solitons in optical fibres.
Opt. Lett. 16, 1936–1937. (doi:10.1364/ol.16.001936)
37. Kivshar YS, Yang XP. 1994 Perturbation-induced dynamics of dark solitons. Phys. Rev. E 49,
1657–1670. (doi:10.1103/physreve.49.1657)
38. Burtsev S, Camassa R. 1997 Nonadiabatic dynamics of dark solitons. J. Opt. Soc. Am. B 14,
1782–1787. (doi:10.1364/josab.14.001782)
39. Konotop VV, Vekslerchik VE. 1994 Direct perturbation-theory for dark solitons. Phys. Rev. E
49, 2397–2407. (doi:10.1103/physreve.49.2397)
40. Chen X-J, Chen Z-D, Huang N-N. 1998 A direct perturbation theory for dark solitons
based on a complete set of the squared Jost solutions. J. Phys. A Math. Gen. 31, 6929–6947.
(doi:10.1088/0305-4470/31/33/005)
41. Chen X-J, Chen Z-D. 1998 Dark optical solitons on influence of the self-steepening term. Chin.
Phys. Lett. 15, 504–506. (doi:10.1088/0256-307x/15/7/013)
42. Huang N-N, Chi S, and Chen X-J. 1999 Foundation of direct perturbation method for dark
solitons. J. Phys. A: Math. Gen. 32, 3939–3945. (doi:10.1088/0305-4470/32/21/307)
 43. Lashkin VM. 2004 Perturbation theory for dark solitons: inverse scattering transform approach
23
and radiative effects. Phys. Rev. E 70, 066620. (doi:10.1103/physreve.70.066620)
44. Ao S-M, Yan J-R. 2005 A perturbation method for dark solitons based on a complete set of the

royalsocietypublishing.org/journal/rspa Proc. R. Soc. A 481: 20240082

...........................................................................................................
squared Jost solutions. J. Phys. A Math. Gen. 38, 2399–2413. (doi:10.1088/0305-4470/38/11/006)
45. Ao S-M. 2006 CORRIGENDUM: a perturbation method for dark solitons based on a complete
set of the squared Jost solutions. J. Phys. A Math. Gen. 39, 1979–1980. (doi:10.1088/0305-
4470/39/8/c01)
46. Ao S-M, Yan J-R. 2007 Three types of expression in dark-soliton perturbation theory based on
squared Jost solutions. Commun. Theor. Phys. 47, 15–18. (doi:10.1088/0253-6102/47/1/003)
47. Ablowitz MJ, Nixon SD, Horikis TP, Frantzeskakis DJ. 2011 Perturbations of dark solitons.
Proc. R. Soc. A 467, 2597–2621. (doi:10.1098/rspa.2010.0663)
48. Mylonas IK, Rothos VM, Kevrekidis PG, Frantzeskakis DJ. 2015 Direct perturbation theory
Downloaded from https://royalsocietypublishing.org/ on 27 January 2025

for dark-bright solitons: application to Bose–Einstein condensates. J. Phys. A: Math. Theor. 49,
015202. (doi:10.1088/1751-8113/49/1/015202)
49. Rothos VM. 2024 Adiabatic perturbation theory for the vector nonlinear Schrödinger
equation with nonvanishing boundary conditions. Theor. Math. Phys. 220, 164–190.
(doi:10.1134/s0040577924070110)
50. Bersano T, Gokhroo V, Khamehchi M, D’Ambroise J, Frantzeskakis D, Engels P, Kevrekidis
P. 2018 Three-component soliton states in spinor F = 1 Bose–Einstein condensates, Phys. Rev.
Lett. 120, 063202. (doi:10.1103/physrevlett.120.063202)
51. Lannig S, Schmied C, Prüfer M, Kunkel P, Strohmaier R, Strobel H, Gasenzer T, Kevrekidis
PG, Oberthaler MK. 2020 Collisions of three-component vector solitons in Bose–Einstein
condensates, Phys. Rev. Lett. 125, 170401. (doi:10.1103/physrevlett.125.170401)
52. Pelinovsky DE, Stepanyants, Yu A, Kivshar Yu S. 1995 Self-focusing of plane dark solitons in
nonlinear defocusing media, Phys. Rev. E 51, 5016–5026. (doi:10.1103/physreve.51.5016)
53. Pelinovsky DE, Kivshar Yu S, Afanasjev VV. 1996 Instability-induced dynamics of dark
solitons. Phys. Rev. E 54, 2015–2032. (doi:10.1103/physreve.54.2015)
54. Pelinovsky DE, Kevrekidis PG, Frantzeskakis DJ. 2005 Oscillations of dark solitons in trapped
Bose–Einstein condensates. Phys. Rev. E. 72, 016615. (doi:10.1103/physreve.72.016615)
55. Poole D, Hereman W. 2011 Symbolic computation of conservation laws for nonlinear
partial differential equations in multiple space dimensions. J. Sym. Comp. 46, 1355–1377.
(doi:10.1016/j.jsc.2011.08.014)
56. Hereman W, Anco S. 2023 Private communications with BP and AC.
57. Achilleos V, Yan D, Kevrekidis PG, Frantzeskakis DJ. 2012 Dark-bright solitons in Bose–
Einstein condensates at finite temperatures, New J. Phys. 14, 055006. (doi:10.1088/1367-
2630/14/5/055006)