Stochastic Soliton Solutions of the High-Order Nonlinear Schrödinger Equation in the Optical Fiber with Stochastic Dispersion and
Nonlinearity
Hui Zhong, Bo Tian, Hui-Ling Zhen, and Wen-Rong Sun
State Key Laboratory of Information Photonics and Optical Communications, and School of Science, Beijing University of Posts and Telecommunications, Beijing 100876, China Reprint requests to B. T.; E-mail:tian_bupt@163.com
Z. Naturforsch.69a,21 – 33 (2014) / DOI: 10.5560/ZNA.2013-0071
Received June 3, 2013 / revised September 7, 2013 / published online December 18, 2013
In this paper, the high-order nonlinear Schrödinger (HNLS) equation driven by the Gaussian white noise, which describes the wave propagation in the optical fiber with stochastic dispersion and non- linearity, is studied. With the white noise functional approach and symbolic computation, stochastic one- and two-soliton solutions for the stochastic HNLS equation are obtained. For the stochastic one soliton, the energy and shape keep unchanged along the soliton propagation, but the velocity and phase shift change randomly because of the effects of Gaussian white noise. Ranges of the changes increase with the increase in the intensity of Gaussian white noise, and the direction of velocity is inverted along the soliton propagation. For the stochastic two solitons, the effects of Gaussian white noise on the interactions in the bound and unbound states are discussed: In the bound state, periodic oscillation of the two solitons is broken because of the existence of the Gaussian white noise, and the oscillation of stochastic two solitons forms randomly. In the unbound state, interaction of the stochastic two solitons happens twice because of the Gaussian white noise. With the increase in the intensity of Gaussian white noise, the region of the interaction enlarges.
Key words:Stochastic Solitons; Gaussion White Noise; High-Order Nonlinear Schrödinger Equation; Symbolic Computation; White Noise Functional Approach.
1. Introduction
Nonlinear Schrödinger (NLS) equations describe the wave propagation in different nonlinear media, such as the nonlinear fibers [1], photonic crystals [2], Bose-Einstein condensates [3], and ion plasmas [4].
Some NLS solitons have been studied to analyse the interaction between the nonlinearity and dispersion [5, 6]. In reality, such models in the nonlinear media in- volve certain uncertainties [7]. Having considered the effects of stochastic coefficients or initial conditions on the soliton solutions, people are also able to work on the NLS equations driven by the noises [8–14]. Evo- lution of the NLS solitons with stochastic cubic non- linearity has been investigated via the numerical sim- ulations [9]. NLS solitons with white noise dispersion which describe the propagation of a signal in an opti- cal fiber with dispersion management have been stud- ied [10], and dispersion-managed vector solitons in the
birefringent optical fibers with stochastic birefringence have been discussed numerically [11].
Propagation of the soliton pulse in a single-mode fiber with delayed Raman response and random param- eters is described by the stochastic NLS equation with high-order nonlinearity and dispersion [1,15–17],
∂U
∂z +i
3 k≥2
∑
ikβk(z) k!
∂kU
∂tk =iγ(z)
1+ i ω0
∂
∂t
·
U(z,t) Z t
−∞
R t0
|U z,t−t0
|2dt0
,
(1)
whereU(z,t)is the slowly varying normalized enve- lope of the ultrashort pulse, z is the normalized dis- tance along the direction of the propagation,t is the retarded time,t0is the formal variable,ω0is the center frequency of the ultrashort pulse spectrum,R(t)is the Raman response function, while thekth-order group- velocity-dispersion (GVD) coefficientβk(z)and non-
© 2014 Verlag der Zeitschrift für Naturforschung, Tübingen·http://znaturforsch.com
linearity coefficient γ(z) are considered as stochastic functions:
βk(z) =βk0
1+mβk(z)
, γ(z) =γ0
1+mγ(z) , whereβk0 andγ0 are respectively the mean values of βk(z) and γ(z), mβ
k(z) and mγ(z) are the zero-mean stochastic processes of the Gaussian white noise, hmβ
ki=hmγi=0, hmβ
k(z)mβ
k(z0)i=2σβ2
kδ(z−z0), hmγ(z)mγ(z0)i=2σγ2δ(z−z0),
where h· · · i denotes the statistical average, z0 is the formal variable, and σβk and σγ are respectively the variances of stochastic processes mβ
k and mγ. Mod- ulational instability of the periodic pluse arrays de- scribed by (1) in the optical fiber with stochastic pa- rameters of high-order nonlinearity and dispersion has been studied [15]. In the nonlocal focusing and defo- cusing Kerr media with stochastic dispersion and non- linearity, effects of the noises on the modulational in- stability for (1) have been discussed [16,17]. In the case when the pulse envelope evolves slowly along the fiber, (1) can be approximately written as [1]
∂U
∂z +iβ2(z) 2
∂2U
∂t2 −β3(z) 6
∂3U
∂t3 = iγ(z)
|U|2U+ i ω0
∂
∂t |U|2U
−iτRU∂|U|2
∂t
, (2)
where τR is the Raman resonant time constant. Al- though the periodic-like solutions [18] and numerical simulation [19] of (2) without high-order nonlinear- ity and dispersion have been obtained, to our knowl- edge, stochastic soliton solutions for (2), in the opti- cal fiber with stochastic high-order nonlinearity and dispersion, have not been obtained as yet. Effects of the noises on the optical solitons have been discussed in the optical fiber communication systems [20], fiber lasers [21,22], and fiber optical parametric ampli- fiers [23], so that the stochastic optical solitons, which we will obtain based on (2), might mirror the effects of the Gaussian white noise on those studies with the high-order dispersion and nonlinearity.
In this paper, we will work on (2) via the white noise functional approach1 and symbolic computation [5,
1With the white noise functional approach [24], certain ana- lytic solutions for the stochastic Korteweg–de Vries (KdV) [25–27], KdV–Burgers [28,29], and (2+1)-dimensional Broer–Kaup equa- tions [30] have been obtained.
31,32]. In Section2, the Wick product, Hida test func- tion space, and Hida distribution space will be con- structed, then (2) will be transformed into the Wick- type stochastic NLS equation. Via the Hermite trans- formation, wick-type stochastic NLS equation will be transformed into an equation under certain conditions to obtain the solutions of (2). In Section3, stochas- tic one- and two-soliton solutions will be obtained via the symbolic computation and inverse Hermite trans- form, and the effects of Gaussian white noise on the dynamic properties of the solitons will be discussed.
In Section4, the stabilities of stochastic solitons will be studied through the numerical simulation. Finally, our conclusions will be given in Section5.
2. White Noise Analysis of Equation (2) In (2), the Gaussian-white-noisemβ
2(z),mβ
3(z), and mγ(z)have the following properties:
hmβ2(z)i=hmβ3(z)i=hmγ(z)i=0, mβ2(z) =h1W(z), mβ3(z) =h2W(z), mγ(z) =h3W(z),
whereh1,h2, andh3are all non-zero constants, which represent the intensity coefficients of the Gaussian white noises, respectively,W(z)is the standard Gaus- sian white noise,W(z) =dB(z)dz , andB(z)is the standard Brownian motion [24].
The white noise functional approach to study the stochastic partial differential equations in the Wick versions has been given in [24]. For (2),U=U(z,t,W) is the generalized stochastic process andt∈Rd. Let (S(Rd))and(S(Rd))∗be the Hide test function space and Hide function space onRd, respectively [33]. Let hn(x) be the dth-order Hermite polynomials. Setting ξn(x) =e−12x2hn(√
2x)/(π(n−1)!)12, we denoteα= (α1,· · ·,αd) being the d-dimensional multi-indices withα0js(j=1,· · ·,d)∈N(Ndenotes the set of the natural numbers), and letα(i)= (α1(i),· · ·,αd(i))be the ith multi-index number in some fixed ordering of all the d-dimensional multi-indices α = (α1,· · ·,αd)∈ Nd[34–36]. Forα, we define
Hα(ω) =
∞
∏
i=1
hαi(hω,ηii), ω∈ S
Rd ∗
ηi=ξα(i) =ξ
α1(i)⊗ · · · ⊗ξ
αd(i), where⊗is the tensor product.
Fix n∈N. Let(S)n1 consist of thosex=∑αcαHα
with cα ∈ Rn such that kxk =∑αc2α(α!)2(2N)kα,
∀k ∈ N with c2α = |cα|2 = ∑nk=1(c(k)α )2 if cα = (c(1)α ,· · ·,c(n)α )∈Rn.
The Wick product can be defined as [24,37,38]
FG=
∑
α,β
(aα,bβ)Hα+β, (3) where β is the d-dimensional multi-indices, F =
∑αaαHα,G=∑βbβHβ∈(S)n−1withaα,bβ∈Rn. By interpreting Wick versions, we can write (2) as
∂U
∂z +i
2H1(z)∂2U
∂t2 −1
6H2(z)∂3U
∂t3 = iH3(z)
"
|U|2U+ i ω0
∂
∂t |U|2U
−iτRU∂|U|2
∂t
#
, (4)
where H1(z) = β20[1+h1W(z)], H2(z) = β30[1+
h2W(z)], andH3(z) =γ0[1+h3W(z)].
ForF=∑αaαHα ∈(S)n−1withaα ∈Rn, the Her- mite transform ofFis defined as [24,37,38]
F(z) =e H(F) =
∑
α
aαωα∈Cn, (5) whereω∈Cnis a complex variable. ForF,G∈(S)n−1, by the Wick product definition, we have
H(F(W)G(W)) =F(W^)G(W)
=F(ω])·G(ω]).
(6) With the Hermite transformation of (4), the Wick products are turned into the ordinary products and (4) is written as
∂eu(z,t,ω)
∂z +i
2Hf1(z,ω)∂2eu(z,t,ω)
∂t2
−1
6Hf2(z,ω)∂3eu(z,t,ω)
∂t3 = ifH3(z,ω)
|ue(z,t,ω)|2eu(z,t,ω) + i
ω0
∂
∂t |ue(z,t,ω)|2ue(z,t,ω)
−iτRue(z,t,ω)∂|eu(z,t,ω)|2
∂t
,
(7)
where u(z,e t,ω) = H[U(z,t,W)], Hf1(z,ω) = β20[1 +h1We(z,ω)], Hf2(z,ω) = β30[1 +h2We(z,ω)], and Hf3(z,ω) =γ0[1+h3We(z,ω)] with We(z,ω) = H[W(z)] =∑∞k=1ηk(z)ω. Once the solutionsu(z,e t,ω) for (7) is obtained, the solutions U(z,t,W) for (2) can be obtained by the inverse Hermite transform of u(z,t,e ω).
3. Stochastic Soliton Solutions and Discussions 3.1. Bilinear Forms and Stochastic Soliton Solutions
To obtain the bilinear forms for (7), we introduce the dependent variable transformation,
ue(z,t,ω) =g(z,t,ω)
f(z,t,ω), (8)
where g(z,t,ω) is a complex differentiable function andf(z,t,ω)is a real one. Withh2=h3andβ30=τR=
1
ω0, the bilinear forms for (7) can be obtained as
Dz+i
2Hf1(z,ω)D2t −1
6Hf2(z,ω)D3t
g·f =0, (9) Hf1(z,ω)D2t f·f=Hf3(z,ω)g·g∗, (10) with “*” representing the complex conjugate and the Hirota operatorsDzandDt defined by [31,32]
Dmz Dnt(a·b) = ∂
∂z− ∂
∂z0 m
∂
∂t− ∂
∂t0 n
·a(z,t)b(z0,t0) z0=z,t0=t
,
(11)
z0andt0being the formal variables,a(z,t)as the func- tion of z andt, b(z0,t0) as the function of z0 andt0, m=0,1,2,· · ·andn=0,1,2,· · ·.
Via the Hirota method [31,32], g(z,t,ω) and f(z,t,ω)can be expanded as
g(z,t,ω) =εg1(z,t,ω) +ε3g3(z,t,ω)
+ε5g5(z,t,ω) +· · ·, (12) f(z,t,ω) =1+ε2f2(z,t,ω) +ε4f4(z,t,ω)
+ε6f6(z,t,ω) +· · ·, (13) withε is a small parameter.gn(z,t,ω)and fn(z,t,ω) (n =1,2,· · ·) are the functions to be determined.
Substituting Expressions (12) and (13) into Bilinear Forms (9) and (10) and equating the coefficients of the same powers ofεto zero yield the recursion relations forgn(z,t,ω)and fn(z,t,ω) (n=1,2,· · ·).
To obtain the one-soliton solutions for (7), we trun- cate Expressions (12) and (13) for g1(z,t,ω) and
f2(z,t,ω), respectively. Setting
g1(z,t,ω) =n1eθ, f2(z,t,ω) =m1eθ+θ∗, (14) whereθ=k(z,ω) +b1t=k(z,ω) + (b11+ib12)tand n1=n11+in12 with n11, n12, b11, b12, and m1 as
the real constants, k(z,ω) as the complex function.
Substituting Expressions (14) into Bilinear Forms (9) and (10), we can get the constraints on the parame- ters,
k(z,ω) = Z z
−∞ b31−ib21
Hf2(ξ,ω)dξ, m1=− n211+n212
8b211 , α=2γ0
β20 h3=h1. Withε=1, the one-soliton solutions for (7) can be expressed as
eu(z,t,ω) = n1eb1t+k(z,ω) 1−(n211+n212)γ0
4β20b211 e(b1+b∗1)t+k(z,ω)+k∗(z,ω)
= n1eb1t+
Rz
0(b31−ib21)fH2(ξ,ω)dξ
1−(n211+n212)γ0
4β20b211 e(b1+b∗1)t+R0z(b31−ib21)Hf2(ξ,ω)dξ+R0z(b∗31 +ib∗21 )Hf2(ξ,ω)dξ
.
(15)
To obtain the two-soliton solutions for (7), we as- sume that
g1(z,t) =m1eθ1+m2eθ2,
f2(z,t) =n1eθ1+θ1∗+n2eθ2+θ2∗+n3eθ1+θ2∗ +n4eθ2+θ1∗,
g3(z,t) =m3eθ1+θ2+θ1∗+m4eθ1+θ2+θ2∗, f4(z,t) =O1eθ1+θ2+θ1∗+θ2∗,
whereO1=O11+iO12,mj=mj1+imj2,nj=nj1+ inj2, andθl=kl(z,ω) +blt=kl(z,ω) + (bl1+bl2)t withO11,O12,mj1,mj2,nj1,nj2, bl1, andbl2 as the real constants, and kl(z,ω)as the complex functions (j=1,2· · ·4,l=1,2). Substituting them into Bilinear Forms (9) and (10), we can get the constraints on the parameters,
k1(z,ω) = Z z
−∞
(b11+ib12)2(b11+ib12−iα1)
·Hf2(ξ,ω)dξ,
k2(z,ω) = Z z
−∞
(b21+ib22)2(b21+ib22−iα1)
·fH2(ξ,ω)dξ, n1=− m211+m212
α2
8b211 , n2=− m221+m222
α2
8b221 ,
n3=−(m11+im12) (m21−im22)α2 2(b11+ib12+b21−ib22)2 , n4=−(m11−im12) (m21+im22)α2
2(b11−ib12+b21+ib22)2 , m3=−m2 m211+m212
α2(b1−b2)2 8b211 b1+b∗22 , m4=−m1 m221+m222
α2(b1−b2)2 8b221 b1+b∗22 , α1=3β20
β30 α2=6γ0
β30 h1=h2,
O1=
m211+m212
m221+m222 α22
h
(b11−b21)2+ (b12−b22)2i2
64b211b221 h
(b11+b21)2+ (b12−b22)2i2 .
Withε=1, the two-soliton solutions for (7) can be expressed as
eu(z,t,ω) = m1eb1t++m2eθ2+m3eθ1+θ2+θ1∗+m4eθ1+θ2+θ2∗ n1eθ1+θ1∗+n2eθ2+θ2∗+n3eθ1+θ2∗+n4eθ2+θ1∗+O1eθ1+θ2+θ1∗+θ2∗
=
m1eb1t+b21(b1−iα1)R−∞z Hf2(ξ,ω)dξ+m2eb2t+b22(b2−iα1)R−∞z Hf2(ξ,ω)dξ +m3e(b2+2b11)t+[2b11(b211+2α1b12−3b212)+b22(b2−iα1)]R−∞z Hf2(ξ,ω)dξ
+m4e(b1+2b21)t+[2b21(b221+2α1b22−3b222)+b21(b1−iα1)]R−z∞Hf2(ξ,ω)dξ
n1e2b11t+2b11(b211+2α1b12−3b212)
Rz
−∞Hf2(ξ,ω)dξ
+n2e2b21t+2b21(b221+2α1b22−3b222)R−∞z fH2(ξ,ω)dξ +n3e(b1+b∗2)t+[b21(b1−iα1)+b∗22 (b∗2+iα1)]R−z∞Hf2(ξ,ω)dξ
+n4e(b2+b∗1)t+[b22(b2−iα1)+b∗21 (b∗1−iα1)]R−∞z Hf2(ξ,ω)dξ
+O1e2(b11+b21)t+2[b311+b321+b11(2α1−3b12)b12+b21(2α1−3b22)b22]R−∞z Hf2(ξ,ω)dξ . (16) Taking the inverse Hermite transformation of solutions (15), the one-soliton solutions for (4) are obtained as
U(z,t) = n1e
Rz
0(b31−ib21)H2(ξ)dξ+b1t
1−(n211+n212)γ0
4β20b211 e[R0z(b31−ib21)H2(ξ)dξ+R0z(b∗31 +ib∗21 )H2(ξ)dξ]+(b1+b∗1)t
= n1e(b31−ib21)β30z+b1teβ30h2(b31−ib21)R−∞z W(ξ)dξ+b1t
1−(n211+n212)γ0
4β20b211 e(b31−ib21+b∗31 +ib∗21 )β30z+(b1+b∗1)teβ30h2(b31−ib21+b∗31 +ib∗21 )R−∞z W(ξ)dξ
= n1e(b31−ib21)β30z+b1teβ30h2(b31−ib21)B(z)
1−(n211+n212)γ0
4β20b211 e(b31−ib21+b∗31 +ib∗21 )β30z+(b1+b∗1)teβ30h2(b31−ib21+b∗31 +ib∗21 )B(z)
. (17)
Since eB(z)=eB(z)−12z2 withrepresenting the Wick products andB(z)representing the standard Brownian motion [24], the stochastic one-soliton solutions for (2) can be expressed as
U(z,t) = n1e(b31−ib21)β30z+(b31−ib21)β30h2B(z)−12(b31−ib21)β30h2z2+b1t
1−(n211+n212)γ0
4β20b211 e(b31−ib21+b∗31 +ib∗21 )β30z+β30h2(b31−ib21+b∗31 +ib∗21 )[B(z)−12z2]+(b1+b∗1)t
. (18)
Similarly, through the inverse Hermite transformation, Solutions (16) for (4) is transformed into U(z,t) =
m1eb1t+b21(b1−iα1)β30zeb21(b1−iα1)β30h2B(z) +m2eb2t+b22(b2−iα1)β30zeb22(b2−iα1)β30h2B(z)
+m3e(b2+2b11)t+[2b11(b211+2α1b12−3b212)+b22(b2−iα1)]β30z
·e[2b11(b211+2α1b12−3b212)+b22(b2−iα1)]β30h2B(z)
+m4e(b1+2b21)t+[2b21(b221+2α1b22−3b222)+b21(b1−iα1)]β30z
·e[2b21(b221+2α1b22−3b222)+b21(b1−iα1)]β30h2B(z)
n1e2b11t+2b11(b211+2α1b12−3b212)β30ze2b11(b211+2α1b12−3b212)β30h2B(z) +n2e2b21t+2b21(b221+2α1b22−3b222)β30ze2b21(b221+2α1b22−3b222)β30h2B(z)
+n3e(b1+b∗2)t+[b21(b1−iα1)+b∗22 (b∗2+iα1)]β30ze[b21(b1−iα1)+b∗22 (b∗2+iα1)]β30h2B(z)
+n4e(b2+b∗1)t+[b22(b2−iα1)+b∗21 (b∗1−iα1)]β30ze[b22(b2−iα1)+b∗21 (b∗1−iα1)]β30h2B(z)
+O1e2(b11+b21)t+2[b311+b321+b11(2α1−3b12)b12+b21(2α1−3b22)b22]β30z
·e2[b311+b321+b11(2α1−3b12)b12+b21(2α1−3b22)b22]β30h2B(z) . (19) The stochastic two-soliton solutions for (2) can be expressed as
U(z,t) =
m1eb1t+b21(b1−iα1)β30zeb21(b1−iα1)β30h2[B(z)−12z2] +m2eb2t+b22(b2−iα1)β30zeb22(b2−iα1)β30h2[B(z)−12z2] +m3e(b2+2b11)t+[2b11(b211+2α1b12−3b212)+b22(b2−iα1)]β30z
·e[2b11(b211+2α1b12−3b212)+b22(b2−iα1)]β1h2[B(z)−12z2] +m4e(b1+2b21)t+[2b21(b221+2α1b22−3b222)+b21(b1−iα1)]β30z
·e[2b21(b221+2α1b22−3b222)+b21(b1−iα1)]β30h2[B(z)−12z2]
n1e2b11t+2b11(b211+2α1b12−3b212)β30ze2b11(b211+2α1b12−3b212)β30h2[B(z)−12z2] +n2e2b21t+2b21(b221+2α1b22−3b222)β30ze2b21(b221+2α1b22−3b222)β30h2[B(z)−12z2]
+n3e(b1+b∗2)t+[b21(b1−iα1)+b∗22 (b∗2+iα1)]β30ze[b21(b1−iα1)+b∗22(b∗2+iα1)]β30h2[B(z)−12z2] +n4e(b2+b∗1)t+[b22(b2−iα1)+b∗21 (b∗1−iα1)]β30ze[b22(b2−iα1)+b∗21(b∗1−iα1)]β30h2[B(z)−12z2] +O1e2(b11+b21)t+2[b311+b321+b11(2α1−3b12)b12+b21(2α1−3b22)b22]β30z
·e2[b311+b321+b11(2α1−3b12)b12+b21(2α1−3b22)b22]β30h2[B(z)−12z2] . (20)
3.2. Discussions for Solutions (18) and (20)
For the effects of the Gaussian white noise on the stochastic one soliton, solution (18) is rewritten as U(z,t) =Asech
η1z+η2t+η3B(z)−1
2η3z2+C
·eiη4z+iη5t+iη6B(z)−2iη6z2, with
A=2n1C, η1= b311+2b11b12−3b11b212 β30, η2=b11,
η3= b311+2b11b12−3b11b212 β30h2, η4= −b211+3b211b12+b212−b312
β30, η5=b12,
η6= −b211+3b211b12+b212−b312 β30h2, C=
s −β20b211 4 n211+n212
γ0
,
where, of the stochastic one soliton,A andCare the amplitude and initial position, η2and η5are respec- tively related to the pulsewidth and frequency,η1and
50 250
−2 2
z
Amplitude
Fig. 1. Plot of the stochastic function B(z) via stochastic number sequence.
η3are both related to the velocity, andη4andη6are related to the phase shifting. Therefore, the velocity and phase shifting of the stochastic one soliton are ob- tained asη1+η3B(z)˙ −η3zandη4z+η6B(z)−12η6z2. It shows that the Gaussian white noise affects the ve-
Fig. 2. (a) Stochastic one soliton via Solutions (18). (b) Energy of the stochastic one soliton via Solutions (18). Parameters are:n1=1−i,b1=1−2i,α=−20,β30=1, andh2=0.
Fig. 3. (a) Stochastic one soliton via Solutions (18). (b) Energy of the stochastic one soliton via Solutions (18). Parameters are:n1=1−i,b1=1−2i,α=−20,β30=1, andh2=0.2.
Fig. 4. (a) Stochastic one soliton via Solutions (18) withh2=0.5. (b) Stochastic one soliton via Solutions (18) withh2=0.8.
Parameters are:n1=1−i,b1=1−2i,α=−20, andβ30=1.
locity and phase shifting of the stochastic one soliton, while the amplitude, energy, and shape of the stochas- tic one soliton are unaffected by the Gaussian white noise. The average value of the velocity can be ob- tained asη1−η3z. It means that the acceleration is ex-