Ann. Phys. (Berlin) 528, No. 6, 493–503 (2016) / DOI 10.1002/andp.

201500322

Original Paper
Analytic soliton solutions of cubic-quintic Ginzburg-Landau
equation with variable nonlinearity and spectral filtering in
fiber lasers
Long-Gang Huang1 , Li-Hui Pang2 , Pring Wong1 , Yan-Qing Li1 , Shao-Yi Bai1 , Ming Lei1,∗ , and
Wen-Jun Liu1,2,∗

Received 27 October 2015, revised 21 December 2015, accepted 8 January 2016

Published online 15 February 2016

tons are self-confined and have been successfully used to

In fiber lasers, the study of the cubic-quintic complex carry the logical information for optical communications
Ginzburg-Landau equations (CGLE) has attracted much at- due to their stable propagation, as a result, attracting a
tention. In this paper, four families (kink solitons, gray soli- lot of attention of researchers [9–11, 14]. To model the
tons, Y-type solitons and combined solitons) of exact soli- optical soliton propagation in optical fibers, the complex
Ginzburg-Landau equation (CGLE) is more generalized
ton solutions for the variable-coefficient cubic-quintic CGLE
than the nonlinear Schrödinger one. When it comes to
are obtained via the modified Hirota method. Appropriate
long-distance communication systems, the attenuation
parameters are chosen to investigate the properties of soli- of pulses inherent in optical fibers can not be ignored [3,
tons. The influences of nonlinearity and spectral filtering 17–19]. It is encouraging that compensation for the at-
effect are discussed in these obtained exact soliton solu- tenuation can be realized by the fiber amplifier, in which
tions, respectively. Methods to amplify the amplitude and the soliton propagation can be modeled by the cubic-
compress the width of solitons are put forward. Numerical quintic CGLE [14, 20, 21].
simulation with split-step Fourier method and fourth-order As the canonical equation describing the weakly non-
linear and dissipative systems, the cubic-quintic CGLE
Runge-Kutta algorithm are carried out to validate some
has been involved in a variety of phenomena includ-
of the analytic results. Transformation from the variable-
ing second-order phase transitions, superconductivity,
coefficient cubic-quintic CGLE to the constant coefficients super-fluidity, Bose-Einstein condensation, liquid crys-
one is proposed. The results obtained may have certain ap- tals, nonlinear optics, plasmas and even string theory
plications in soliton control in fiber lasers, and may have [22–26]. Many kinds of soliton solutions have been re-
guiding value in experiments in the future. ported so far [27–31]. However, most researches for the
cubic-quintic CGLE have been focused on numerical
solutions including localized fixed-shape solutions and
pulsating solutions [27–29]. The former type of solu-
1 Introduction tion refers to stable stationary pulses (solitons), com-
posite pulses, and moving pulses [27], while the lat-
Since the concept of solitons was created by Zabusky ter one can be plain pulsating, creeping, snaking, and
and Kruskal in 1965 [1], considerable progress has been exploding (erupting) solutions [28, 29]. Researches on
made in the study of various types of solitons as well as
their applications in such fields as nonlinear optics, fluid
∗ Corresponding authors E-mail: [email protected];
dynamics, plasma physics, condensed matter physics,
solids, particle physics and astrophysics [2–11]. The bal- [email protected]
1 State Key Laboratory of Information Photonics and Optical Com-
ance of nonlinear and dispersive effects enables solitons
to possess the stable, localized, and particle-like char- munications, School of Science, Beijing University of Posts and
acters, which can describe various physical phenomena Telecommunications, Beijing, 100876, China
2 Beijing National Laboratory for Condensed Matter Physics, In-
[12, 13]. As one of the most important branches of soli-
tons, optical solitons have been theoretically and exper- stitute of Physics, Chinese Academy of Sciences, Beijing, 100190,
imentally studied [14–16]. In optical fibers, optical soli- China


C 2016 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 493
 L.-G. Huang et al.: Analytic soliton solutions of cubic-quintic Ginzburg-Landau equation
Original Paper

those different kinds of solitons can help to understand In this paper, equation (1) will be analytically stud-
abundant physical laws involved in realistic applications. ied via the modified Hirota method for the first time,
For instance, the conventional bright and dark solitons with analytic soliton solutions being derived. It should
can be used in data transmission, readdressing as well be noted that some new types of soliton solutions in the
as switching [32–34]. Besides, exploding solutions in the CGLE with variable-coefficient nonlinear coefficient and
cubic-quintic CGLE can be experimentally observed in spectral filtering will be proposed. And in subsequent
mode-locked fiber lasers, which can guide the design of discussion, physical effects of the nonlinearity as well as
mode-locked fiber lasers [35, 36]. Moreover, the effect spectral filtering affecting the properties and characters
of nonlinear gradient terms on pulsating, exploding and of different solitons will be investigated.
creeping solitons has been investigated [37]. Chaotic pul- This paper is organized as follows. In the second sec-
sating and period-doubling solutions have also been ob- tion, the analytic solution for the variable-coefficient
tained [38]. cubic-quintic CGLE will be derived with the modified
Some physical effects in optics can be modeled by the Hirota method [47]. In the third section, four families
cubic-quintic CGLE such as pulse propagation in opti- of analytic soliton solutions will be obtained, and in-
cal fibers with gain and spectral filtering, transverse soli- fluences of nonlinearity as well as spectral filtering on
ton effects in wide aperture lasers [39–41], and passively solitons will be discussed along with the verification of
mode-locked laser systems [42, 43]. Some results about numerical simulations. In addition, transformation from
the solutions of cubic-quintic CGLE have been reported the variable-coefficient cubic-quintic CGLE to the con-
in recent years. Namely, the influence of intra-pulse Ra- stant coefficients one will be provided. Finally, conclu-
man scattering (IRS) on the stable stationary pulses in sions will be drawn in the fourth section.
the presence of constant linear and nonlinear gain as
well as spectral filtering has been numerically examined
[44]. Two categories (pulses and snakes) of dissipative 2 Analytic soliton solution for the
solitons have been found, with the dependence of both variable-coefficient cubic-quintic CGLE
their shape and stability on the physical parameters of
the cubic-quintic CGLE being analyzed [45]. Soliton dy- Transformations facilitating the application of the modi-
namics from dissipative soliton to dissipative soliton res- fied Hirota method are
onance have been classified by a combined analytical ap-
proach, based on the established laser models [46]. g (x, t)
u(x, t) = , (2)
f (x, t) 2 +iα
1
Although a lot of works have been done on the cubic-
quintic CGLE, and some soliton solutions have been
found by numerical methods, the analytic soliton solu- where g (x, t) is a complex differentiable function, and
tions are still rare. Moreover, for the variable-coefficient α, f (x, t) are assumed to be real. By virtue of symbolic
cubic-quintic CGLE, the analytic soliton solutions ob- computation, bilinear forms for equation (2) can be
tained by the modified Hirota method have not been re- derived as
ported before. Thus, via the modified Hirota method, we
study the variable-coefficient cubic-quintic CGLE as the i Dx,α g · f + [ pr (x) + i pi (x)]Dt,α
2
g · f − i[γr (x) + iγi (x)]g
master equation [13]: · f =0 (3)
∂u ∂ 2u
i + [ pr (x) + i pi (x)] 2 + [qr (x) + iqi (x)]|u|2 u 1
∂x ∂t − [ pr (x) + i pi (x)](1 + 2iα)(3 + 2iα)Dt2 f · f + [qr (x)
8
+ [cr (x) + ici (x)]|u|4 u = i[γr (x) + iγi (x)]u, (1)
+ iqi (x)]|g |2 f + [cr (x) + ici (x)]|g |4 = 0 (4)
where u(x, t) is the envelope amplitude of the elec-
tric field, t is the retarded time and x is the prop-
agation distance. pr (x) and pi (x) represent the wave Dt2 f · f and Dx,α , Dt,α
2
are respectively the generalized
dispersion and spectral filtering, qr (x) and qi (x) are the Hirota’s bilinear operator and modified Hirota’s bilinear
nonlinear coefficient and the nonlinear gain-absorption operators, which can be defined by [47–49]:
coefficient, respectively. cr (x) and ci (x) stand for the    
higher-order correction terms to the nonlinear refrac- ∂ 1 ∂ m
m
Dx,α n
Dt,α (g · f ) = − + iα
tive index and the nonlinear amplification absorption. ∂x 2 ∂ x
   n
And γr (x) and γi (x) are related to the linear gain and fre- ∂ 1 ∂
× − + iα g (x, t) f (x , t )|x =x,t =t (5)
quency shift. ∂t 2 ∂t

494 www.ann-phys.org 
C 2016 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Ann. Phys. (Berlin) 528, No. 6 (2016)

Original Paper
 m n
∂ ∂ ∂ ∂ partial differential equation (10) becomes the partial
Dxm Dtn(g · f ) = −  − 
∂x ∂x ∂t ∂t differential equation of the only unknown function f2 .
Thus, we can obtain the form of function f2 according
×g (x, t) f (x , t )|x =x,t =t (6)
to equation (10) as follows
Bilinear forms (3), (4) can be solved by the following
power series expansions for g (x, t) and f (x, t) f2 (x, t) = σ (x) exp(θ + θ ∗ ), (15)

g (x, t) = εg 1 (x, t) + ε 3 g 3 (x, t) + ε 5 g 5 (x, t) + · · · , (7) where θ ∗ is the conjugate of θ .

Substituting g 1 (x, t) and f2 (x, t) into the resulting set
f (x, t) = 1 + ε2 f2 (x, t) + ε 4 f4 (x, t) + ε 6 f6 (x, t) + · · · , (8) of linear partial differential equations (9–12), which refer
to the above recursion relations yielded by equating co-
where ε is a formal expansion parameter. Substituting ex- efficients of the same powers of ε to zero and after some
pressions (7), (8) into bilinear forms (3), (4), and equating calculations, we can get the constraints on the parame-
coefficients of the same powers of ε to zero yield the re- ters:
cursion relations for the functions g n(x, t), s and fn(x, t), s 
as follows, [−2b1 b2 pr (x) + (b22 − b12 ) pi (x) + γr (x)]dx
a1 (x) = ,
x
∂2g1 ∂g1 
[−iγr (x) + γi (x)]g 1 (x, t) + [ pr (x) + i pi (x)] +i [−2b1 b2 pi (x) + (b12 − b22 ) pr (x) + γi (x)]dx
∂t2 ∂x a2 (x) = ,
x
= 0, (9)
qr (x)
σ (x) = − 2 , pr (x)
1 b1 (−3 pr (x) + 4α 2 pr (x) + 8α pi (x))
[qr (x) + iqi (x)]|g 1 |2 + (−3 − 8iα + 4α 2 )[ pr (x)
4 2(b2 − b1 α) pi (x)
= ,
∂ 2 f2 b1
+i pi (x)] 2 = 0, (10)
∂t qi (x) = 8b12 ασ (x) pr (x) + 3b12 σ (x) pi (x) − 4b12 α 2 σ (x) pi (x),
  
1 ∂ f2 ∂ g 1 qr (x) ∂ pi (x)
(−i + 2α) (−i pr (x) + pi (x)) 4 − (1 + 2iα) γr (x) = −(2b1 α − b2 )2 pi (x) + , cr (x)
4 ∂t ∂t 2 pi (x) ∂ x qr (x)
 
∂ 2 f2 ∂ f2 = −σ (x)qr (x), ci (x) = −σ (x)qi (x),
g 1 2 + 2g 1 + 4 f2 (g 1 (−iγr (x) + γi (x))
∂t ∂x
g n(x, t) = 0 (n = 3, 5, 7, . . .), fn(x, t) = 0

∂2g1 ∂g1
+ ( pr (x) + i pi (x)) 2 + i = 0, (11) (n = 4, 6, 8, . . .).
∂t ∂x
4(cr (x) + ici (x))|g 1 |4 − (−3 + 4α 2 − 8iα)( pr (x) + i pi (x)) At the end, analytic solutions for equation (1) can be
  obtained under the condition of the above constraints.
∂ f2 2
+ f2 (4|g 1 |2 (qr (x) + iqi (x)) + (−3 + 4α 2 − 8iα) Without loss of generality, we set ε = 1, and the analytic
∂t
 soliton solution can be expressed as:
∂ 2 f2
( pr (x) + i pi (x)) 2 = 0. (12)
∂t g (x, t) g 1 (x, t)
u(x, t) = = . (16)
2 +iα [1 + f2 (x, t)] 2 +iα
1 1
f (x, t)
For the analytic soliton solution of equation (1), and
according to the linear partial differential equation (9),
g 1 (x, t) can be assumed to be in the following generalized
exponential form: 3 Results and discussions
g 1 (x, t) = exp(θ ), (13) By choosing different types of variable nonlinearity and
spectral filtering profiles and appropriate values of the
θ = [a1 (x) + ia2 (x)]x + (b1 + ib2 )t + (k1 + ik2 ), (14) arbitrary parameters b j , k j (j = 1, 2), α and γi (x), several
kinds of solitons of the variable-coefficient cubic-quintic
with a j (x), ( j = 1, 2) are differentiable functions of x to CGLE are obtained. Furthermore, their pulse widths and
be determined, and b j , k j ( j = 1, 2) are real constants. amplitudes can be adjusted through changing the related
When we substitute g 1 (x, t) = exp(θ ) back into (10), the parameters, as is shown in Figs. 1–12.


C 2016 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.ann-phys.org 495
 L.-G. Huang et al.: Analytic soliton solutions of cubic-quintic Ginzburg-Landau equation
Original Paper

Figure 1 Evolution of kink solitons affected by nonlinearity and spectral filtering. Parameters adopted here are: b1 = 0.22, b2 =
0.097, k1 = 0.131, k2 = 0.28, α = 0.69, γi (x) = 1 with (a) qr (x) = −2, pi (x) = 3; (b) qr (x) = −2.5, pi (x) = 3; (c) qr (x) =
−2, pi (x) = 3.5; (d) Comparison between kink solitons of (a) (blue solid line), (b) (red dotted line) and (c) (green dash-dotted line) at
t = 0.

Figure 2 Numerical simulation of kink soli-

tons corresponding to Fig. 1(a) and (b) with the
consistent parameters as: b1 = 0.22, b2 =
0.097, α = 0.69, γi (x) = 1 with (a) qr (x) =
−0.002, pi (x) = 0.003; (b) qr (x) = − 0.022,
pi (x) = 0.003.

Figure 3 Phase shift and propagation change of kink solitons affected by the arbitrary parameters b1 , k1 . Parameters are the same with
Fig. 1(a) but (a) k1 = 0.431; (b) b1 = −0.22; (c) Phase shift between kink solitons in Fig. 1(a) ((blue solid line)) and Fig. 3(a) (red dotted line).

496 www.ann-phys.org 
C 2016 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Ann. Phys. (Berlin) 528, No. 6 (2016)

Original Paper
Figure 4 Gray solitons and influence of
nonlinear coefficient. Parameters are:
b1 = 0.031, b2 = −0.44, k1 = 2, k2 =
2.6, α = −0.22, γi (x) = 1, pi (x) =
−5.38 Tanh (0.23x) with (a)
qr (x) = 2.1 Tanh (2x); (b) qr (x) =
2.4 Tanh (2x); (c) Comparison between
gray solitons of (a) (blue solid line) and (b)
(red dotted line) at t = 0; (d) Correspond-
ing nonlinear coefficient profiles of (a)
(blue solid line) and (b) (red dotted line).

3.1 Kink soliton solutions With the higher absolute value of the nonlinearity
(qr (x) = −2.5), the amplitude of kink soliton in Fig. 1(b) is
In this subsection, we choose all the parameters as con- about 0.36, which is smaller than that in Fig. 1(a) (qr (x) =
stants. Kink solitons, waves with invariant profile in the −2, amplitude is about 0.45). However, with the bigger
shape of a twist, are obtained as shown in Fig. 1. value of spectral filtering ( pi (x) = 3.5), the amplitude of

Figure 5 Influence of nonlinear coefficient. Parameters are the same with Fig. 4(a) but (a) qr (x) = 2.1 Tanh (10x); (b) Comparison
between gray solitons of Fig. 4(a) (blue solid line) and Fig. 5(a) (red dotted line) at t = 0; (c) Corresponding nonlinear coefficient profiles
qr (x) of Fig. 4(a) (blue solid line) and Fig. 5(a) (red dotted line).

Figure 6 Influence of spectral filtering. (a) Parameters are the same with Fig. 4(a) but pi (x) = −5.68 Tanh (0.23x); (b) Comparison
between gray solitons of Fig. 4(a) (blue solid line) and Fig. 6(a) (red dotted line) at t = 0; (c) Corresponding spectral filtering profiles of
Fig. 4(a) (blue solid line) and Fig. 6(a) (red dotted line).


C 2016 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.ann-phys.org 497
 L.-G. Huang et al.: Analytic soliton solutions of cubic-quintic Ginzburg-Landau equation
Original Paper

Figure 7 Influence of spectral filtering. (a) Parameters are the same with Fig. 4(a) but pi (x) = −5.38 Tanh (0.63x); (b) Comparison
between gray solitons of Fig. 4(a) (blue solid line) and Fig. 7(a) (red dotted line) at t = 0; (c) Corresponding spectral filtering profiles of
Fig. 4(a) (blue solid line) and Fig. 7(a) (red dotted line).

qr (x) = −0.022 in Fig. 2(b)) we choose, the smaller the

amplitude of kink soliton will be (amplitude is about 0.42
in Fig. 2(a), while about 0.38 in Fig. 2(b)). At the same
time, when we change the value of spectral filtering, in-
fluence on the amplitude of kink solitons is also consis-
tent with the analytic ones.
Note that the steepness of kink soliton can be weak-
ened or enhanced at the same time. The greater the spec-
Figure 8 Numerical simulation of gray solitons corresponding tral filtering (or the smaller the nonlinearity) is, the larger
to Fig. 4 with the appropriate parameters as: b1 = 0.031, b2 = the amplitude is and the steeper the kink soliton will be,
−0.44, α = −0.22, γi (x) = 1, pi (x) = −5.38 Tanh (0.23(x − obviously can be seen in Fig. 1(d). That’s because the
δ)) × 10−4 with (a) qr (x) = 2.1 Tanh (2(x − δ)) × 10−4 ; nonlinearity loss can bring about attenuation, leading to
(b) qr (x) = 2.4 Tanh (2(x − δ)) × 10−4 , δ = 40. the decrease. And the spectral filtering effect can help
to obtain higher energy and narrower pulses. That is to
say, we can increase the amplitude and steepness of kink
kink soliton can be amplified in Fig. 1(c) (amplitude is solitons by increasing the value of spectral filtering or de-
about 0.52). Besides, the influence of nonlinearity can be creasing the absolute value of nonlinearity.
confirmed by numerical calculations through the fourth- On one hand, when we change the value of the ar-
order Runge-Kutta algorithm, as shown in Fig. 2 [50]. bitrary parameters k1 , the phase shift of kink solitons
In order to better contrast with the analytic results, we occurs without other characters (such as amplitude,
choose the consistent parameters given in Fig. 1(a) and steepness, propagation, etc) of kink solitons influenced,
(b). Numerical results also show that the higher the abso- compared between Fig. 1(a) (k1 = 0.131) and Fig. 3(a)
lute value of the nonlinearity (qr (x) = −0.002 in Fig. 2(a); (k1 = 0.431). And the kink soliton has a rightward shift as

Figure 9 Solitonic solutions. (a) Parameters adopted here are: b1 = 0.5, b2 = −0.34, k1 = −0.13, k2 = −0.88, α = −0.31, γi (x) =
1 qr (x) = −0.3 Tanh (0.1x), and pi (x) = −4 Tanh (0.25x); (b) Parameters are the same with (a) but qr (x) = −0.15 Tanh (0.1x);
(c) Parameters are the same with (a) but pi (x) = −8 Tanh (0.25x).

498 www.ann-phys.org 
C 2016 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Ann. Phys. (Berlin) 528, No. 6 (2016)

Original Paper
Figure 10 (a) Bright soliton via solution (16). Parameters are: b1 = 1, b2 = 1, k1 = 1, k2 = 1, α = −0.6, γi (x) = 1, qr (x) = −e−x −
e x , and pi (x) = 0.15e−0.1x − e0.2x ; (b) Cross section plot of (a); (c) Corresponding numerical simulation of (a) with the same parameters
except the adjustments of qr (x) = −e−(x−ς ) − e(x−ς ) , pi (x) = 0.15e−0.1(x−ς ) − e0.2(x−ς ) , ς=8.

spintronic shift registers, in which magnetic kink soliton

exists. We hope our research on kink solitons can be ob-
served in the future experiments such as optical quan-
tum storage or other significant experiments.

3.2 Gray soliton solutions

According to solution (16), gray solitons are obtained

Figure 11 (a) Solitonic solutions. Parameters are the same with and propagate stably when we choose both the nonlin-
Fig. 10(a) but pi (x) = e−0.1x − e0.1x ; (b) Cross section plot of (a). ear coefficient and spectral filtering profile as Tanh-type
functions with other appropriate parameters as shown
shown in Fig. 3(c) (red dashed line). On the other hand, in Fig. 4(a). By increasing the coefficient before Tanh
the direction of soliton propagation can be changed by function in nonlinear coefficient qr (x), soliton amplitude
altering the sign of parameter b1 (b1 = 0.22 in Fig. 1(a); will be decreased. The value of the coefficient before
b1 = −0.22 in Fig. 3(b)). Because, parameter k1 is a free Tanh function in Fig. 4(a) is about 2.1, but is about 2.4
phase factor for the exponential function in the expres- in Fig. 4(b). As a result, the amplitude of gray solitons
sion of u(x, t). Its value can affect the phase of the soliton. in Fig. 4(b) is smaller than that in Fig. 4(a). And from
And b1 is the coefficient of t in the exponential function, Fig. 4(c), we can find that the larger the maximum value
which can affect the propagation of soliton pulse directly. of nonlinearity is, the smaller the gray soliton amplitude
That is to say, we can control the propagation and phase is.
shift of kink solitons by choosing proper coefficients of k1 This means that the amplitude is inversely propor-
and b1 . tional to the maximum value of nonlinearity. It may be
As we can see that the kink solitons have a twist struc- explained as when the nonlinear effect enhanced, the
ture and can jump into a relatively high amplitude from nonlinear loss increases at the same time, as a result,
the zero background sharply, it can be applied in vertical the amplitude of soliton decreases. Whereas, when we

Figure 12 (a) Coexistence of bright and dark solitons. Parameters are the same with Fig. 10(a) but pi (x) = 0.2e−x − e0.1x ; (b) Cross
section plot of (a); (c) Corresponding spectral filtering profile.


C 2016 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.ann-phys.org 499
 L.-G. Huang et al.: Analytic soliton solutions of cubic-quintic Ginzburg-Landau equation
Original Paper

change the counterpart value in spectral filtering pi (x), the retarded time exceeds the value and a bright soli-
the result is just the opposite (as shown in Fig. 6). That is ton is formed. By decreasing the absolute value of coef-
to say, we can improve the energy of gray soliton through ficient of Tanh fuction in nonlinearity (Fig. 9(a): qr (x) =
decreasing the coefficient before Tanh-type functions −0.3Tanh(0.1x), and (b): qr (x) = −0.15Tanh(0.1x)), the
of nonlinear coefficient qr (x) or through increasing the amplitude of solitons is amplified with the opening
ones in pi (x). mouth not be changed as is shown in Fig. 9(b). How-
When we increase the coefficient of x in qr (x), the am- ever, when we change the counterpart in spectral filter-
plitude of gray soliton also increases, as shown in Fig. 5. ing, the amplitude is amplified as well as the opening
Comparison between Fig. 4(a) (qr (x) = 2.1Tanh(2x)) and mouth shrinks obviously as is shown in Fig. 9(c). This
Fig. 5(a) (qr (x) = 2.1Tanh(10x)) shows that the bigger shows that we can alter the energy distribution by ad-
the coefficient of x in qr (x), the larger the amplitude of justing the nonlinearity and spectral filtering, which is
gray soliton is (shown in Fig. 5(b)). We should point out helpful to understand the propagation of solitons. As we
that unlike the former condition of soliton amplification know, the magnitude of soliton amplitude can represent
that the non-zero continuous wave background varies, the energy of pulse, the soliton we observed may be used
the level of the background keeps fixed but with valley to realize the focus of energy in fiber laser or the signal
value becoming lower than that in Fig. 4(a) (shown in reversal in the electromagnetic storage.
Fig. 5(a) and (b)). Besides, the corresponding maximum
value of nonlinearity keeps unchanged but the slope of
it increases around x = 0, as shown in Fig. 5(c). This im- 3.4 Coexistence of bright and dark solitons
plies that soliton amplitude is proportional to the value
(or the absolute value) of slope of nonlinear coefficient When we choose the nonlinear coefficient and spec-
profile near x = 0. Nevertheless, when we increase the tral filtering as double-exponential functions (the form
value of coefficient of x in spectral filtering pi (x), the soli- of n1 e−m1 x − n2 em2 x ). Bright and dark soliton can be
ton width can be obviously compressed without the am- observed. First, by choosing appropriate parameters, the
plitude changed as shown in Fig. 7. The reason is that conventional bright soliton is obtained in Fig. 10(a), with
spectral filtering effect can filter the unwished pulses, as the corresponding numerical results affiliated behind in
a result, achieving high energy and narrower pulses in Fig. 10(c).
dissipative systems. To sum up, a certain technique by When we change the spectral filtering profile into
changing the nonlinear coefficient and spectral filtering symmetric one (i. e. n1 = n2 = 1, m1 = m2 = 0.1), the
can realize soliton control and improve the soliton qual- solitonic solutions arise in Fig. 11. It seems that bright
ities to some extent. soliton is split into two parts, symmetric to each other.
To validate the accuracy of the solution of gray soli- And it may also be regarded as two bright solitonic parts
tons, in this part, we also add numerical calculations that we hope to enhance the capacity of signal transmis-
discussed above via the split-step Fourier method [51]. sion in optical fibers.
In order to keep the finite pulse evolution and get the One can see from Fig. 10(a), bright soliton propagate
appropriate shape of gray solitons, we reduce the same stably, without any distortion. That is due to the balance
multiple of pi (x) and qr (x), and conduct the coordinate between dispersion and nonlinearity. If we look care-
translation (x − δ) as shown in Fig. 8. Similar to Fig. 4, fully at Fig. 10(b), we find the shape of soliton is not ex-
we obtain gray solitons through numerical simulation. In actly symmetric. In other words, the left side of soliton is
addition, when we increase the coefficient before Tanh sharper than the right side. That is because the structure
function in nonlinear coefficient, soliton amplitude can of two exponential functions in spectral filtering profile
be also decreased, which coincides with the analytic one. are not symmetric (i. e. pi (x) = n1 e−m1 x − n2 em2 x , where
n1 = n2 and m1 = m2 ). It represents that the asymmetri-
cal distribution of spectral filtering effect can shape the
3.3 “Y-type” solitons soliton into asymmetric form.
In order to examine the new phenomenon, numer-
Similar with sub-section 3.2, Tanh-type nonlinearity and ical simulation is also proposed. When we choose the
spectral filtering and another series of arbitrary parame- similar parameters with appropriate scaling and coordi-
ters are applied, as a result, a new type of soltion is ob- nate translation of spectral filtering pi (x) and nonlinear
served as is shown in Fig. 9. It looks like the letter “Y”. coefficient qr (x) from Fig. 10(a), similar results can be
The energy of soliton is scattered when the retarded time obtained shown in Fig. 10(c). We should point out that
t is less than a constant value, while it is focused when the bright soliton from the numerical simulation is also

500 www.ann-phys.org 
C 2016 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Ann. Phys. (Berlin) 528, No. 6 (2016)

Original Paper
asymmetric with the left side of soliton sharper than the our paper. Though some constraints have to be consid-
right side, conforming to the analytic results. ered, solutions of some constant coefficients equations
Increasing the coefficient of x in the spectral filter- can be obtained to some extent by imposing to the vari-
ing profile (i. e. pi (x) = 0.2e−x − e0.1x ), bright and dark able coefficients to be the constant ones.
solitons emerge simultaneously. Unlike the previous re- On the other hand and for a general condition, by
ported combined solitons, the valley of dark soliton is virtue of the similarity transformation mentioned in refs.
very sharp. Interesting phenomenon is that the lowest [26, 52], we can transform equation (18) into (17). First,
point of dark soliton (at x = −1.4) coincides with the zero we set the format for transformation:
point of corresponding spectral filtering profile, shown
in Fig. 12(c). Perhaps because the sign of spectral filter- u(x, t) = ψ(X, T )ρ(x, t) exp[iϕ(x, t)], X = X(x), T
ing profile changed at x = −1.4, and when we select the
= T (x, t), (19)
modulus of it, there is an abrupt change of value at that
point, leading the sharp valley of dark soliton as is shown where ρ(x, t), ϕ(x, t), X(x), and T (x, t) are real-valued
in Fig. 12(b). In addition, we also find that the size of functions and ψ(X, T ) is a complex-valued function of
the slope of spectral filtering profile can also influence variables of X and T. Attention should be paid that all the
the properties of combined soliton. From Fig. 12, we can functions are continuous derivable functions. Substitut-
see that the bright soliton propagates parallel to the dark ing the transformation (19) into equation (18), we can get
soliton with no impact on each other. And they can also the following result:
be transmitted without distortion, which can find po-
 
tential applications in optical communication systems. dX ∂ψ ∂ T 2 ∂ 2ψ
Overall, we should point out that when we change the iρ + pρ + qρ 3 |ψ|2 ψ + cρ 5 |ψ|4 ψ
dx ∂ X ∂t ∂2T
coefficients in nonlinearity profile, similar results can be 
obtained. ∂T ∂2T ∂ T ∂ρ ∂ T ∂ϕ
+ iρ + pρ 2 + 2 p + ipρ
∂x ∂ t ∂t ∂t ∂t ∂t
 
∂ T ∂ϕ ∂ψ ∂ρ ∂ϕ ∂ 2ρ
3.5 Transformation from Eq. (1) to the constant + ipρ = i γρ − −iρ + ip 2
∂t ∂t ∂ T ∂x ∂x ∂ t
coefficients one   
∂ρ ∂ϕ ∂ 2ϕ ∂ϕ 2
− 2p − pρ 2 −ipρ , (20)
In this part, we will focus on the transformation from Eq. ∂t ∂t ∂ t ∂t
(1) to the constant coefficients one:
In order to be consistent with the constant coeffi-
∂ψ ∂ 2ψ cients CGLE of equation (17), we equate the correspond-
i + P 2 + Q|ψ|2 ψ + C|ψ|4 ψ = iψ, (17)
∂x ∂t ing coefficients of the resulting equation (20) to those of
equation (17). Similar to ref. [26], the following results
where P = 0, Q, C,  are known complex constants and
can be derived:
ψ = ψ(X, T ) is the complex function of X and T.
For simplicity sake, equation (1) can be abbreviated dX
as the following form: iρ(x, t) = 1,
dx
 
∂u ∂ 2u ∂T 2
i + p 2 + q|u|2 u + c|u|4 u = iγ u, (18) pρ(x, t) = P,
∂x ∂t ∂t
qρ 3 (x, t) = Q,
where p = pr (x) + i pi (x), q = qr (x) + iqi (x), c = cr (x) +
ici (x), γ = γr (x) + iγi (x) are complex-valued functions of cρ 5 (x, t) = C,
x ∂ρ ∂ϕ ∂ 2ρ ∂ρ ∂ϕ
On one hand, when we choose the independent vari- γρ − −iρ + ip 2 −2 p
∂x ∂x ∂ t ∂t ∂t
able coefficients (those are pr (x), pi (x), qr (x), γr (x), γi (x))  2
in solution (16) as constants, the rest of the variable pa- ∂ ϕ
2
∂ϕ
− pρ 2 − ipρ = ,
rameters (those are qi (x), cr (x), ci (x)) will degenerate into ∂ t ∂t
the constants. In other words, Eq. (1) turns into the con- ∂T ∂2T ∂ T ∂ρ ∂ T ∂ϕ
iρ + pρ 2 +2 p +ipρ
stant coefficients one. And solution (16) becomes the so- ∂x ∂ t ∂t ∂t ∂t ∂t
lution of constant coefficients equation, which can be il- ∂ T ∂ϕ
lustrated by kink soliton solutions in sub-section 3.1 in + ipρ = 0, (21)
∂t ∂t


C 2016 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.ann-phys.org 501
 L.-G. Huang et al.: Analytic soliton solutions of cubic-quintic Ginzburg-Landau equation
Original Paper

Detailed investigations for solutions of partial dif- parameters. Methods to amplify the amplitude and
ferential equations (21) have been put forward in ref. shrink the opening mouth have been discussed. At last,
[26]. It means that if the constant coefficients of P = bright and dark solitons exist at the same time when
0, Q, C and  are given (or equivalently, equation (17) is we choose the nonlinearity and spectral filtering as the
given), we can obtain the corresponding variable coef- double-exponential functions (as the form: n1 e−m1 x −
ficients p = pr (x) + i pi (x), q = qr (x) + iqi (x), c = cr (x) + n2 em2 x ). We find that the symmetric structures of expo-
ici (x), γ = γr (x) + iγi (x) as well as the real-valued func- nential functions in the nonlinearity and spectral filter-
tions ρ(x, t) and ϕ(x, t) according to equations (21) re- ing can form the symmetric solitons, while the asym-
ferred to ref. [26]. Moreover, solutions of the related metric ones form the opposite ones. What’s more, the
variable-coefficient cubic-quintic CGLE (18) can be de- exponential nonlinearity and spectral filtering profiles
rived via the modified Hirota method, which is precisely with asymmetric structures can form the combined soli-
what we have done in our work. Finally solutions of Eq. ton (coexistence of bright and dark solitons). Results ob-
(17) can be obtained by the inverse transformation of tained in this paper might be of value in various research
fields modeled by cubic-quintic CGLE, such as nonlinear
u(x, t) optical systems, quantum storage, etc.
ψ(X, T ) = . (22)
ρ(x, t) exp[iϕ(x, t)]

Thereby, solutions of constant coefficients equation Acknowledgements. We express our sincere thanks to the Editors
can be achieved by transforming the variable equation and Referees for their valuable comments. This work has been
into the constant one. supported by the National Natural Science Foundation of China
(NSFC) (Grant Nos. 61205064 and 61574020), by the Fund of State
Key Laboratory of Information Photonics and Optical Communica-
4 Conclusions tions (Beijing University of Posts and Telecommunications, grant
No. IPOC2015ZC07).
As a universal model in various physical phenomena, the
cubic-quintic CGLE with variable nonlinearity and spec- Key words. Cubic-quintic complex Ginzburg-Landau equation,
tral filtering has been studied analytically. With the help modified Hirota method, spectral filtering, soliton control, simi-
of modified Hirota method, the bilinear forms (3)-(4) for larity transformation.
Eq. (1) have been derived, and analytic soliton solution
(16) has been obtained after some symbolic computa-
tion. On the basis of analytic soliton solution, four fami- References
lies of solitons have been obtained by choosing different
kinds of nonlinearity and spectral filtering profiles, in- [1] N. J. Zabusky and M. D. Kruskal, Phys. Rev. Lett. 15 (6),
240–243 (1965).
cluding gray solitons, kink solitons, “Y-type” solitons as
[2] L. F. Mollenauer and J. P. Gordon, Solitons in Optical
well as combined solitons (see Figs. 1–12). Fibers, Academic Press, Burlington, 2006.
When we choose all the parameters as constants, we [3] G. P. Agrawal, Nonlinear Fiber Optics, 4th Edition,
find the kink solitons. Our research shows that the am- Academic Press, San Diego, 2007.
plitude and steepness of kink solitons can be reduced by [4] N. Fusaki, H. Ban, A. Nishiyama, K. Saeki, and M.
nonlinearity, while they can be increased by spectral fil- Hasegawa, Proc. Japan. Acad. Ser. B 85 (8), 348–362
tering. In addition, the free parameters b1 , k1 in solution (2009).
[5] G. A. Maugin, Mech. Res. Commun. 38 (5), 341–349
can affect the soliton propagation and phase shift.
(2011).
Gray solitons have been observed with Tanh type [6] Z. G. Chen, M. Segev, and D. N. Christodoulides, Rep.
nonlinearity and spectral filtering, and their amplitude Prog. Phys. 75 (8), 086401 (2012).
and pulse width can be adjusted through choosing ap- [7] L. Wang, Y. T. Gao, and F. H. Qi, Ann. Phys. 327 (8),
propriate parameters. Results show that the nonlinearity 1974–1988 (2012).
qr (x) can decrease the amplitude while the spectral filter- [8] F. Haas, B. Eliasson, and P. K. Shukla, Phys. Rev. E 85
ing effect pi (x) can enhance the amplitude and compress (5), 056411 (2012).
[9] B. Tang, D. J. Li, and Y. Tang, Chaos 24 (2), 023113
the width of solitons. Thus spectral filtering can help to
(2014).
improve the properties of solitons. [10] C. Q. Dai and H. P. Zhu, Ann. Phys. 341, 142–152
With the same type of nonlinearity and spectral fil- (2014).
tering, a new type of solitons that looks like the letter [11] C. Q. Dai, X. G. Wang, and G. Q. Zhou, Phys. Rev. A 89
“Y” has been obtained by choosing another series of (1), 013834 (2014).

502 www.ann-phys.org 
C 2016 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Ann. Phys. (Berlin) 528, No. 6 (2016)

Original Paper
[12] S. K. Turitsyn, B. G. Bale, and M. P. Fedoruk, Phys. Rep. [32] D. Y. Tang, L. Li, Y. F. Song, L. M. Zhao, H. Zhang, and
521 (4), 135–203 (2012). D. Y. Shen, Phys. Rev. A 88 (1), 013849 (2013).
[13] S. I. Fewo, A. Kenfack-Jiotsa, and T. C. Kofane, J. Phys. [33] Y. F. Song, L. Li, H. Zhang, D. Y. Shen, D. Y. Tang,
A: Math. Gen. 39 (6), 1449–1461 (2006). and K. P. Loh, Opt. Express 21 (8), 10010–10018
[14] Z. G. Chen, M. Segev, and D. N. Christodoulides, Rep. (2013).
Prog. Phys. 75 (8), 086401 (2012). [34] W. J. Liu, L. H. Pang, H. N. Han, W. L. Tian, H. Chen,
[15] V. N. Serkin, A. Hasegawa, and T. L. Belyaeva, Phys. M. Lei, P. G. Yan, and Z. Y. Wei, Opt. Express 23 (20),
Rev. Lett. 98 (7), 074102 (2007). 26023–26031 (2015).
[16] V. N. Serkin, A. Hasegawa, and T. L. Belyaeva, Phys. [35] S. T. Cundiff, J. M. Soto-Crespo, and N. Akhmediev,
Rev. A 81 (2), 023610 (2010). Phys. Rev. Lett. 88 (7), 073903 (2002).
[17] V. V. Konotop, V. S. Shchesnovich, and D. A. Zezyulin, [36] J. M. Soto-Crespoa and N. Akhmediev, Math. Comput.
Phys. Lett. A 76 (42-43), 2750–2753 (2012). Simulat. 69 (5-6), 526–536 (2005).
[18] R. Y. Hao, L. Li, Z. H. Li, and G. S. Zhou, Opt. Commun. [37] H. P. Tian, Z. H. Li, J. P. Tian, G. S. Zhou, and J. Zi, Appl.
236 (1-3), 79–86 (2004). Phys. B 78 (2), 199–204 (2004).
[19] V. N. Serkin and A. Hasegawa, IEEE J. Sel. Top. Quan- [38] N. N. Akhmediev, J. M. Soto-Crespo, and G. Town,
tum Electron. 8 (3), 418–431 (2002). Phys. Rev. E 63 (5), 056602 (2001).
[20] L. F. Mollenaur, J. P. Gordon, and P. V. Mamyshev, Opti- [39] A. Mekozzi, J. D. Moores, H. A. Haus, and Y. Lai, Opt.
cal Fiber Telecommunications, 3th Edition, Academic Lett. 16 (23), 1841–1843 (1991).
Press, San Diego, 1997. [40] Y. Kodama and A. Hasegawa, Opt. Lett. 17 (1), 31–33
[21] E. Iannone, F. Matera, A. Mecozzi, and M. Settember, (1992).
Nonlinear Optical Communication Networks, Wiley- [41] V. V. Afanasjev, Opt. Lett. 18 (10), 790–792 (1993).
Interscience, 1998. [42] H. A. Haus, J. G. Fujimoto, and E. P. Ippen, J. Opt. Soc.
[22] M. C. Cross and P. C. Hohenberg, Rev. Mod. Phys. 65 Am. B 8 (10), 2068–2076 (1991).
(3), 851–1112 (1993). [43] F. X. Kartner, J. Ausder Au, and U. Keller, IEEE
[23] I. S. Aranson and L. Kramer, Rev. Mod. Phys. 74 (1), J. Sel. Top. Quantum Electron. 4 (2), 159–168
99–143 (2002). (1998).
[24] E. L. Rempel, A. C. L. Chian, A. J. Preto, and S. [44] I. M. Uzunov, Z. D. Georgiev, and T. N. Arabadzhiev,
Stephany, Nonlinear Proc. Geoph. 11 (5-6), 691–700 Phys. Rev. E 90 (4), 042906 (2014).
(2004). [45] S. C. Mancas and R. S. Choudhury, Nonlinear Dyn. 79
[25] N. R. Pereira and L. Stenflo, Phys. Fluids 20 (10), 1733– (1), 549–571 (2015).
1734 (1977). [46] W. Lin, S. M. Wang, S. H. Xu, Z. C. Luo, and Z. M. Yang,
[26] E. Kengne, A. Lakhssassi, R. Vaillancourt, and W. M. Opt. Express 23 (11), 14860–14875 (2015).
Liu, J. Math. Phys. 53 (12), 123703 (2012). [47] K. Nozaki and N. Bekki, J. Phys. Soc. 53 (5), 1581–1582
[27] N. N. Akhmediev and A. Ankiewicz, Solitons: Nonlin- (1984).
ear Pulses and Beams, Chapman and Hall, London, [48] G. A. Zakeri and E. Yomba, J. Phys. Soc. 82 (8), 084002
1997. (2013).
[28] J. M. Soto-Crespo, N. Akhmediev, and A. Ankiewicz, [49] R. Hirota, Phys. Rev. Lett. 27 (18), 1192–1194
Phys. Rev. Lett. 85 (14), 2937–2940 (2000). (1971).
[29] W. Chang, A. Ankiewicz, N. N. Akhmediev, and J. M. [50] Z. X. Zhang, L. Chen, and X. Y. Bao, Opt. Express 18 (8),
SotoCrespo, Phys. Rev. E 76 (1), 016607 (2007). 8261–8276 (2010).
[30] W. Van Saarloos and P. C. Hohenberg, Phys. Rev. Lett. [51] G. Agrawal, Nonlinear Fiber Optics, 4th ed., Elsevier,
64 (7), 749–752 (1990). Singapore, 2009.
[31] P. Marcq, H. Chate, and R. Conte, Physica D 73 (4), [52] S. W. Xu, J. S. He, and L. H. Wang, Europhys. Lett. 97
305–317 (1994). (3), 30007 (2012).


C 2016 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.ann-phys.org 503