
Self-accelerating solitons
Boris A. Malomed1,2
1Department of Physical Electronics, School of Electrical Engineering, Faculty of Engineering,
and Center for Light-Matter Interaction, Tel Aviv University, Tel Aviv 69978, Israel
2Instituto de Alta Investigaci´on, Universidad de Tarapac´a, Casilla 7D, Arica, Chile
Basic models which give rise to one- and two-dimensional (1D and 2D) solitons, such as the
Gross-Pitaevskii (GP) equations for BEC, feature the Galilean invariance, which makes it possible
to generate families of moving solitons from quiescent ones. A challenging problem is to find models
admitting stable self-accelerating (SA) motion of solitons. SA modes are known in linear systems
in the form of Airy waves, but they are poorly localized states. This brief review presents two-
component BEC models which make it possible to predict SA solitons. In one system, a pair of
interacting 1D solitons with opposite signs of the effective mass is created in a binary BEC trapped
in an optical-lattice potential. In that case, opposite interaction forces, acting on the solitons with
positive and negative masses, produce equal accelerations, while the total momentum is conserved.
The second model is based on a system of GP equations for two atomic components, which are
resonantly coupled by a microwave field. The latter model produces an exact transformation to an
accelerating references frame, thus predicting 1D and 2D stable SA solitons, including vortex rings.
Introduction. – A basic property of one- and two-dimensional (1D and 2D) equations which produce solitons is the
Galilean invariance, which generates solitons moving with an arbitrary velocity from a quiescent one. A challenging
issue is to construct localized states moving at a constant acceleration, rather than constant velocity [1, 2]. A well-
known fact is that the linear Schr¨odinger equation, iuz+ (1/2)uxx = 0 (it is written in the form of the paraxial
propagation equation in optics, with propagation distance zand transverse coordinate x) admits self-accelerating
(SA) solutions in the form of Airy waves [3]. Later, it was predicted [4] and experimentally demonstrated, originally
in optics [5], and then in electron beams [6], plasmonics [7], Bose-Einstein condensates (BECs) [8], acoustics [9], gas
discharge [10], and water waves [11] that truncated Airy waves (TAWs) can be created in these media, the respective
exact solution of the linear Schr¨odinger equation being
uTAW (x, z) = u0Ai αx −α4
4z2+iℵα2zexp (ℵαx)
×exp −iα6
12 z3+iα3
2xz −ℵα4
2z2+iℵ2α2
2z,(1)
where Ai is the Airy function, and constants u0,α, and ℵ>0 define, respectively, the amplitude, internal scale, and
truncation of the Airy wave. Accordingly, the input which generates solution (1) is u(x;z= 0) = u0Ai (αx) exp (ℵαx).
The truncation factor is necessary to make the total norm (power, in terms of optics) of the input finite:
NTAW ≡Z+∞
−∞ |u(x)|2dx =u2
0√8πℵα−1exp 2ℵ3/3,(2)
while the TAW’s momentum,
P=iZ+∞
−∞
u∗
xudx, (3)
is zero, in spite of the self-acceleration featured by this wave (both Nand Pare dynamical invariants of the Schr¨odinger
equation).
In fact, the truncation leads to gradual degradation of the TAW, as shown, in particular, by factor exp −ℵα4z2/2
in solution (1). Another source of degradation is the action of nonlinearity, as TAW is the eigenmode of the linear
medium. Effects of nonlinearity on the Airy waves were considered in many works [12]-[18], chiefly demonstrating
decay into solitons.
A possibility to create well-localized (unlike the Airy waves) SA two-component pulses in optics was elaborated
in terms of a system of coupled nonlinear Schr¨odinger (NLS) equations with opposite signs of the group-velocity
dispersion (GVD) in them [19], aiming to create two pulse components with opposite signs of their effective masses.
In this case, the opposite interaction forces with which the coupled components act on each other give rise to identical
signs of the acceleration. A solution for such an SA bound state was constructed approximately, taking an unperturbed
NLS soliton in the anomalous-GVD component, and applying the Thomas-Fermi approximation to the normal-GVD
arXiv:2210.01507v1 [nlin.PS] 4 Oct 2022