Fractional Integrable and Related Discrete Nonlinear Schr odinger Equations

2025-04-22 0 0 1.45MB 11 页 10玖币
侵权投诉
Fractional Integrable and Related Discrete Nonlinear Schr¨odinger Equations
Mark J. Ablowitz,1Joel B. Been,2, 3, and Lincoln D. Carr2, 3, 4
1Department of Applied Mathematics, University of Colorado, Boulder, Colorado 80309, U.S.A.
2Department of Applied Mathematics and Statistics,
Colorado School of Mines, Golden, Colorado 80401, U.S.A.
3Department of Physics, Colorado School of Mines, Golden, Colorado 80401, U.S.A.
4Quantum Engineering Program, Colorado School of Mines, Golden, Colorado 80401, U.S.A.
Integrable fractional equations such as the fractional Korteweg-deVries and nonlinear Schr¨odinger
equations are key to the intersection of nonlinear dynamics and fractional calculus. In this
manuscript, the first discrete/differential difference equation of this type is found, the fractional inte-
grable discrete nonlinear Schr¨odinger equation. This equation is linearized; special soliton solutions
are found whose peak velocities exhibit more complicated behavior than other previously obtained
fractional integrable equations. This equation is compared with the closely related fractional aver-
aged discrete nonlinear Schr¨odinger equation which has simpler structure than the integrable case.
For positive fractional parameter and small amplitude waves, the soliton solutions of the integrable
and averaged equations have similar behavior.
Keywords: Integrable equations; Fractional calculus; Discrete nonlinear Schr¨odinger equation; Fourier split
step
INTRODUCTION
Integrable systems play a central role in nonlinear dy-
namics because they provide exactly solvable models for
important physical systems. Notable examples of inte-
grable equations are the Korteweg-deVries (KdV), ap-
plicable to shallow water waves, plasma physics, and
lattice dynamics among others [1–3], and the nonlinear
Schr¨odinger (NLS) equation, which finds applications in
nonlinear optics, Bose-Einstein condensates, spin waves
in ferromagnetic films, plasma physics, water waves, etc.
[2–5]. These integrable nonlinear evolution equations
have an infinite number of conservation laws and soli-
ton solutions [6]. Solitons, the fundamental solutions
of such equations, are stable, localized nonlinear waves
which propagate without dispersing and interact elas-
tically with other solitons. Nonlinear integrable evolu-
tion equations have these surprising properties because
of their deep mathematical structure described by the
inverse scattering transform (IST).
IST is a method of solving nonlinear equations which
generalizes Fourier transforms. It solves these equations
in three steps: mapping the initial condition into scat-
tering space, evolving the intial data in scattering space
in time, and mapping the evolved scattering data back
to physical space; i.e., inverse scattering. This pro-
cess gives the solution to nonlinear equations solvable
by IST in terms of linear integral equations; such non-
linear equations are called integrable. Recently, we used
the mathematical structure of IST associated with for
the Korteweg-deVries (KdV) and nonlinear Schr¨odinger
(NLS) equations to develop a method of finding and solv-
ing the fractional KdV (fKdV) and fractional NLS (fNLS)
Corresponding author: joelbeen@mines.edu
equations [7]. We also showed that this method could
be applied to find fractional extensions of the modified
KdV, sine-Gordon, and sinh-Gordon equations [8]. These
equations represent the first known fractional integrable
nonlinear evolution equations with smooth (physical) so-
lutions and deeply connect the fields of nonlinear dynam-
ics and fractional calculus.
Fractional calculus is a mathematical structure origi-
nally designed to define non-integer derivatives and inte-
grals. It has sense become an effective way of modeling
many physical processes that exist in multi-scale media
[9, 10] or exhibit non-Gaussian statistics or power law
behavior [11–13]. A particularly important example is
anomalous diffusion, where the mean squared displace-
ment is proportional to tα,α > 0 [11, 14–16]. Trans-
port that follows this rule has been observed in biology
[17–20], amorphous materials [21–23], porous media [24–
27], climate science [28], and attenuation in materials [29]
amongst others. As we have shown, the merger of frac-
tional and nonlinear characteristics in integrable equa-
tions such as fKdV and fNLS predict anomalous disper-
sion, where the velocity and amplitude of solitonic solu-
tions are related by a power law [7].
In this article, we demonstrate how the method in-
troduced in Ref. [7] can be applied to discrete (or
differential-difference) systems to define integrable dis-
crete fractional nonlinear evolution equations by present-
ing a fractional generalization of the integrable discrete
nonlinear Schr¨odinger (IDNLS) equation. We do this by
demonstrating the three key mathematical ingredients of
our method — IST, power law dispersion relations, and
completeness relations — for the Ablowitz-Ladik (AL)
discrete scattering problem. In the linear limit, the frac-
tional IDNLS (fIDNLS) equation is a discretization of the
fractional Schr¨odinger equation which was derived with
Feynman path integrals over L´evy flights [30, 31].
arXiv:2210.01229v1 [nlin.SI] 3 Oct 2022
2
The KdV equation was the first equation shown to be
solvable by IST in Ref. [32]; it was soon followed by
the NLS equation in Ref. [33]. These two equations were
then found to be contained in a general class of equations
solvable by IST when associated to the Ablowitz-Kaup-
Newell-Segur (AKNS) system [3, 34]. Shortly thereafter
IST was used to solve families of discrete (or differential
difference) problems like the self-dual network [35]. In
particular, it was discovered that the AKNS system could
be discretized while maintaining integrability, leading to
the AL scattering problem which was used to solve a fam-
ily of discrete nonlinear evolution equations [36]. This
family contained important discrete evolution equations
— continuous in time but discretized in space — such as
integrable discretizations of the nonlinear Schr¨odinger,
KdV, modified KdV, and sine Gordon equations. Fur-
ther, this family of equations was shown to have soliton
solutions and an infinite number of conservation laws [36].
We derive the fIDNLS equation from the AL scatter-
ing problem using three key components: linear disper-
sion relations, completeness relations, and IST. IST is
used to linearized the equation and obtain special soliton
solutions.
We also show how the characteristics of the fractional
IDNLS (fIDNLS) equation reach beyond integrability by
comparing the one-soliton solution of the fIDNLS equa-
tion to the solitary wave solution of the fractional aver-
aged discrete nonlinear Schr¨odinger (fADNLS) equation.
This equation is a different fractional generalization of
the IDNLS equation in which the linear second order dif-
ference is replaced by the discrete fractional Laplacian
[37–40]. The fADNLS equation can be understood as a
discretization of a fractional NLS equation involving the
Riesz derivative which has been extensively studied in,
e.g., [41–45]; it is also is also closely related to the (likely)
non-integrable fractional DNLS equation, recently stud-
ied in [37, 46]. Though the fADNLS equation is likely not
integrable to our knowledge (apart from the limiting case
when fADNLS reduces to IDNLS), the similarity between
the two equations suggests that some of the physical pre-
dictions of fractional integrable equations are shared by
equations which are simpler to realize computationally.
THE DISCRETE FRACTIONAL LINEAR
SCHR ¨
ODINGER EQUATION
Consider the family of discrete linear evolution equa-
tions
tqn+γ(n)qn= 0 (1)
for the function qn(t) which depends on the discrete vari-
able nZand the continuous variable tR. Here, γ
is a sufficiently regular function of the discrete laplacian,
n, defined by
(n)qn(t) = 1
h2qn+1(t)+2qn(t)qn1(t)(2)
where his the distance between lattice sites. Using the
Z-transform, which is equivalent to the discrete Fourier
transform, the solution to Eq. (1) can be explicitly writ-
ten as
qn(t) = 1
2πZπ/h
π/h
dkˆq(k, 0)eiknhγ(4 sin2(kh/2)/h2)t(3)
where ˆq(k, 0) = hP
n=−∞ qn(0)eiknh is the Z-transform
of qn(t) at t= 0 and 4 sin2(kh/2)/h2is the Fourier sym-
bol of n. Note that the Z-transform is often written
in terms of zwith the substitution z=eikh where in-
tegration in kbecomes integration with respect to zon
the unit circle. If we choose γto be power law, then
Eq. (1) becomes a fractional discrete equation in terms
of the discrete fractional laplacian. For example, if we
put γ(n) = i(n)1+,||<1, then we obtain the
linear fractional discrete Schr¨odinger equation
i∂tqn+ (n)1+qn= 0.(4)
Here, (n)1+is the discrete fractional laplacian of or-
der 1 + which is defined in terms of its Fourier sym-
bol [4 sin2(hk/2)/h2]1+and the Z-transform/discrete
Fourier transform as
(n)1+qn=1
2πZπ/h
π/h
dkˆq(k)eiknh[4 sin2(kh/2)/h2]1+.
(5)
Notice that the kintegral above can be evaluated to ex-
press the discrete fractional laplacian as a summation
over mof qmmultiplied by a weight vector. The solution
to Eq. (4) can still be written in the form Eq. (3) with
γ(4 sin2(kh/2)/h2) = i[4 sin2(kh/2)/h2]1+
and, because 4 sin2(kh/2)/h2is real and positive, the
solution to equation (1) with this choice of γis well
posed. In defining and solving the linear fractional dis-
crete Schr¨odinger equation, we used a power law disper-
sion relation, ingredient 1 of our method, and we defined
the fractional operator using completeness of the discrete
Fourier transform/Z-transform, ingredient 2. Then we
solve the equation by the inverse discrete Fourier trans-
form, the analog of ingredient 3.
THE FRACTIONAL INTEGRABLE DISCRETE
SCHR ¨
ODINGER EQUATION
To develop the fIDNLS equation, the integrable non-
linear analog of Eq. (4), and solve it, we apply the three
3
key ingredients of our method, starting with writing the
equation in terms of a linear dispersion relation. Note
that h= 1 is taken in this section without loss of gener-
ality; to recover the scaling factor for h6= 1, replace qn
by hqnand rnby hrn.
As in the linear case, Eq. (1), we have a family of
nonlinear evolution equations for the solutions qn(t) and
rn(t) [47], see also [48],
σ3
dun
dt +γ+)un= 0,un= (qn,rn)T(6)
where Trepresents transpose, σ3= diag(1,1), and Λ+
is
Λ+xn=hnE+
n0
0E
n x(1)
k
x(2)
k!(7)
+qnP+
n1rk1qnP+
n2qk+1
rnP+
n1rk1rnP+
n2qk+1 x(1)
k
x(2)
k!(8)
+hn qn+1 P+
n+1
rk
hkqn+1 P+
n+1
qk
hk
rn1P+
n
rk
hkrn1P+
n
qk
hk! x(1)
k
x(2)
k!(9)
where hn= 1 rnqn,P+
n=P
k=n, and E±
nx(q)
k=x(q)
n±1
with q= 1,2. The inverse of this operator is
Λ1
+xn=hnE
n0
0E+
n x(1)
k
x(2)
k!(10)
+qnP+
nrk+1 qnP+
n+1 qk1
rnP+
n1rk+1 rnP+
nqk1 x(1)
k
x(2)
k!(11)
+hn qn1P+
n
rk
hkqn1P+
n
qk
hk
rn+1 P+
n+1
rk
hkrn+1 P+
n+1
qk
hk! x(1)
k
x(2)
k!.(12)
Here, γis a sufficiently regular function of the operator
Λ+and is connected with the linearized dispersion re-
lation. Specifying this dispersion relation, or γdirectly,
picks out particular equations from this family. For ex-
ample, if we take
γ+) = i(2 Λ+Λ1
+)
and let rn=q
n, then we obtain the IDNLS equation
i∂tqn+ ∆nqn± |qn|2(qn+1 +qn1)=0.(13)
We can relate γto the dispersion relation of the lineariza-
tion of (6) by considering the linear limit qn0. In this
limit, we have
Λ+E+
n0
0E
nDn,(14)
so the linearization of the nonlinear evolution equation is
σ3
dun
dt +γ(Dn)un= 0.(15)
Because Dnis a diagonal matrix, we have
γ(Dn) = γ(E+
n) 0
0γ(E
n).(16)
Taking the first component of (15) with
qn=z2ne(z)t
gives
γ(z2) = (z).(17)
Therefore, by specifying the linear limit of the nonlinear
evolution equation, we obtain the nonlinear equation it-
self. To define the fIDNLS equation, we choose the linear
limit to be the discrete linear fractional Schr¨odinger equa-
tion in (4), which gives the dispersion relation ω(z) =
(2 z2z2)1+and, hence, γ(z2) = i(2 z2
z2)1+. So, the fIDNLS equation is
i∂tun+ (2 Λ+Λ1
+)1+un(t) = 0.(18)
In fact, by choosing γ(z2) = i(2 z2z2)m+, for
integer m, we generate a hierarchy of fractional equations
i∂tun+ (2 Λ+Λ1
+)m+un(t)=0.(19)
It can be shown that the limit of (18) as 0 is the
IDNLS equation (13). Notice that to define the fIDNLS
equation, we used a power law dispersion relation, ingre-
dient 1 of the method. However, this dispersion relation
leads to the operator (2 Λ+Λ1
+)1+the meaning of
which is currently unclear. To define this operator, we
will need to use the 2nd ingredient: appropriate com-
pleteness relations. The third ingredient will be making
use of IST to find solutions of the fIDNLS equation.
COMPLETENESS OF SQUARED
EIGENFUNCTIONS AND FRACTIONAL
OPERATORS
In this section we define the fIDNLS equation in (18)
and, in fact, any equation of the form (1) that is well-
posed in physical space. We do this using the observation
that γ+) is a multiplication operator when acting on
the eigenfunctions of Λ+and the fact that the eigen-
functions of Λ+are complete. This result is known as
completeness of squared eigenfunctions, and is the sec-
ond ingredient in our method. The resulting represen-
tation of γ+) will be similar to that of the discrete
fractional laplacian in (5). The eigenfunctions of Λ+are
Ψn(z) and Ψn(z) each with eigenvalue z2(note that time
tis suppressed throughout this section). Therefore, the
operation of γ+) on these eigenfunctions is given by
γ+)Ψn=γ(z2)Ψn, γ+)Ψn=γ(z2)Ψn.(20)
摘要:

FractionalIntegrableandRelatedDiscreteNonlinearSchrodingerEquationsMarkJ.Ablowitz,1JoelB.Been,2,3,andLincolnD.Carr2,3,41DepartmentofAppliedMathematics,UniversityofColorado,Boulder,Colorado80309,U.S.A.2DepartmentofAppliedMathematicsandStatistics,ColoradoSchoolofMines,Golden,Colorado80401,U.S.A.3Dep...

展开>> 收起<<
Fractional Integrable and Related Discrete Nonlinear Schr odinger Equations.pdf

共11页,预览3页

还剩页未读, 继续阅读

声明:本站为文档C2C交易模式,即用户上传的文档直接被用户下载,本站只是中间服务平台,本站所有文档下载所得的收益归上传人(含作者)所有。玖贝云文库仅提供信息存储空间,仅对用户上传内容的表现方式做保护处理,对上载内容本身不做任何修改或编辑。若文档所含内容侵犯了您的版权或隐私,请立即通知玖贝云文库,我们立即给予删除!
分类:图书资源 价格:10玖币 属性:11 页 大小:1.45MB 格式:PDF 时间:2025-04-22

开通VIP享超值会员特权

  • 多端同步记录
  • 高速下载文档
  • 免费文档工具
  • 分享文档赚钱
  • 每日登录抽奖
  • 优质衍生服务
/ 11
客服
关注