1. Introduction
The nonlinear Schrödinger (NLS) equation,
as one of the most famous integrable systems, is known as a ‘universal’ model [
1], which means it appears as a governing model in various physical phenomena. Here ‘i’ is the imaginary unit,
ε = ± 1, ∣
q∣
2 =
qq* and
* stands for complex conjugate. It emerges in describing wave packages in deep water [
2,
3], plasma physics [
4], optical fiber [
5,
6], etc. In addition, the NLS equation with various external potentials (known also as the Gross–Pitaevskii (GP) equation [
7–
9]) is also the governing model in nonlinear optics and Bose–Einstein condensates (BEC) [
10]. One can refer to [
11] for more references and applications of the NLS equation and its extensions.
The NLS equation with
x-coefficient can describe nonlinear waves in non-uniformity media [
12–
15]. Such equations have been shown to be integrable in the sense of having Lax pairs, with the spectral parameter
η satisfying
ηt ≠ 0, which are referred to non-isospectral nonlinear Schrödinger equations (NNLSEs). In this paper, we will investigate the following three NNLSEs:
where
α and
β are real constants, and ∂
−1 stands for the integration operator with respect to
x. We denote these equations NNLSE-I, NNLSE-II, and NNLSE-III for short, respectively. They correspond to time-dependant spectral parameter
η with time evolutions
ηt =
α,
ηt = − i
βη and
ηt = −2
η2, respectively, where , e.g. [
16]. Although the NNLSE-I and NNLSE-II can be converted to the NLS equation (
1.1a) via gauge transformations, e.g. [
16], they are physically useful in BEC: the NNLSE-I is the GP equation with a linear potential [
17], while the NNLSE-II can provide solutions to the GP equation with a parabolic potential and a gain term e.g. [
18]. The NNLSE-III can provide space-time localized soliton waves on zero background [
16,
19]. So far, integrable methods, such as the inverse scattering transform [
12–
14] and bilinear method [
16], have been applied to obtain explicit solutions of the above equations. Yet in this paper, we construct their solutions by means of a completely direct method, namely, the Cauchy matrix approach.
The Cauchy matrix approach is a method to construct and study integrable equations by means of the Sylvester-type equations. It is first systematically introduced in [
20] to investigate integrable quadrilateral equations and later developed in [
21,
22] to more general cases. It has also been applied to the Zakharov–Shabat–Ablowitz–Kaup–Newell–Segur (ZS-AKNS) system [
23], equations with self-consistent sources [
24], and the self-dual Yang-Mills equation [
25,
26], etc. The purpose of this paper is not only to construct solutions to the three NNLSEs in (
1.2), but also to extend the Cauchy matrix approach to the non-isospectral case, as the Sylvester-type equation in the Cauchy matrix scheme of the ZS-AKNS system is a typical type (see [
22,
27] for the Korteweg–de Vries (KdV) and Kadomtsev–Petviashvili (KP) type equations). One will see that the non-isospectral extension of the Cauchy matrix scheme is quite non-trivial compared with the isospectral case [
23].
This paper is organized as follows. Our plan is in the first step to derive three unreduced non-isospectral Schrödinger systems using the Cauchy matrix approach. This will be described in section 2. Then in section 3, we present solution formulae of these unreduced systems. These formulae guide us to implement reduction so that solutions of the NNLSEs are obtained, which will be done in section 4. Dynamics of these solutions are illustrated also in this section. Finally, conclusions are given in section 5. There is an appendix section where solutions of the Sylvester equation with lower triangular Toeplitz matrices are presented.
2. Cauchy matrix approach to unreduced NNLSEs
In this section, we describe the Cauchy matrix approach for unreduced NNLSEs.
2.1. Sylvester equation and master functions
We start from the Sylvester equation of the following type (see [
23,
25]):
in which the involved elements are block matrices in the form of
where , for
i = 1, 2. An equivalent form of (
2.1) is given by
We assume matrices
K1 and
K2 do not share any eigenvalues, so that the Sylvester equation (
2.1) has a unique solution
M for given
K,
r,
s [
28]. By these elements we define master functions
where
I2N is the 2
N th-order unit matrix and of which the more explicit versions are
In addition, a difference-product formula can be formulated from (
2.1) as
The proof can be found in [
23,
25].
2.2. Unreduced NNLSE-I
To derive an unreduced form of the NNLSE-I equation, let us introduce dispersion relations of
r and
s as follows,
where
and matrix
K obeys the evolution
In addition, we assume
Kt and
K commute. Then one can derive the evolution of
M by taking a derivative of the Sylvester equation (
2.1), which gives rise to
Substituting (
2.7b) and (
2.9) into it yields
which leads us to a Sylvester equation,
It has a unique zero solution in the light of assumption that
K1 and
K2 do not share any eigenvalues. Thus we have
One can also find [
23]
Next we are going to derive the derivative of
S(i,j). Let us define the auxiliary vector,
which is connected with
S(i,j) by
The derivative of
u(i) with respect to
x reads [
23]
To derive the derivative of
u(i) with respect to
t, taking
t-derivative in (
2.12) yields
and we substitute(
2.7b), (
2.9) and (
2.10) into (
2.15), and then left-multiplied by , we get
Using the relation (
2.13), it is easy to get
x-derivative of
S(i,j), which reads [
23]
For the
t-derivative of
S(i,j), from (
2.13) we have
Then, substituting (
2.7b), (
2.9) and (
2.16) into (
2.18), we have
One can repeatedly use (
2.17) and get second-order derivatives of
S(i,j) with respect to
x, which which reads
Next, let us define
one has the following by taking
i =
j = 0 in (
2.17) and (
2.19):
where [
A,
B] =
AB −
BA. Besides, by the difference-product formula (
2.6), we obtain the following relations by choosing (
i,
j) = (0, 1), (1, 0), (0, 0), respectively:
Then by direct calculation we find
Unfolding (
2.21) we obtain a closed system of
u2 and
u3 as the unreduced form of the NNLSE-I:
2.3. Unreduced NNLSE-II
To obtain an unreduced form of the NNLSE-II, we introduce the following dispersion relations:
where
a is defined as (
2.8) and
Again, we assume
Kt and
K commute. The evolutions of
M and
S(i,j) with respect to
x are the same as (
2.11) and (
2.17), so we only consider the
t-derivatives. Using (
2.23b) and (
2.24) and employing a similar procedure as we derived (
2.10), we have
Note that (
2.15) and (
2.18) are generic, and can use them for this case as well. We substitute (
2.25), (
2.23b) and (
2.24) into (
2.15), similar to the treatment in section
2.2, we have
Then, substituting (
2.23b), (
2.24) and (
2.26) into (
2.18), we get the derivative of
S(i,j) with respect to
t:
Thus we have
Through a direct calculation, we find
which reveals an equation set of
u2 and
u3 as the unreduced NNLSE-II:
2.4. Unreduced NNLSE-III
In this case, the dispersion relations are:
where
and we assume
KKt =
KtK. The evolution relations of
M, and
S(i,j) with respect to
t are presented as below:
Let
i =
j = 0 in (
2.32c) we find
By a direct calculation, we have
which leads to
where the relation
has been utilized, which has been proved in [
23].
3. Explicit solutions for the unreduced NNLSEs
We have derived the unreduced NNLSEs. Their solutions are given by
S(0,0) which is determined by
K,
M,
r and
s through the formula (
2.4). In isospectral case (see [
23])
K is a constant matrix and one can equivalently consider its canonical form, i.e. diagonal or Jordan forms or their combinations, and the resulted solutions can be classified by the canonical forms of
K. However, it is much different in non-isospectral case as
K is no longer a constant matrix and must obey evolutions such as (
2.9), (
2.24) and (
2.31). That means, in principle, we can not classify solutions by considering the canonical forms of
K.
In this section, for convenience, we only consider the case of K being diagonal. There will be a case of K composed by lower Toeplitz matrices to be presented in appendix A.
In the following, let us take the unreduced NNLSE-I (
2.22) as an example. Consider
K = diag(
K1,
K2) with
K1,
K2 being the following diagonal forms
where
such that
K satisfies (
2.9). The dispersion relation (
2.7) yields
(
3.3a) has the following solutions:
where
and are constants. Then, the set of Sylvester equations (
2.3) allow solutions and where
Finally, we reach to the explicit expressions of
u2 and
u3:
which satisfy the unreduced NNLSE-I (
2.22).
For the unreduced NNLSE-II (
2.29) and the unreduced NNLSE-III (2.34), there solutions can be expressed through the formulae (
3.7) with (
3.4) and (
3.6) but where
ρj,
ϱj,
σj,
ϖj and
kj,
lj are defined differently. For the unreduced NNLSE-II (
2.29), we have
where
For the unreduced NNLSE-III (2.34), we have
where
4. Reduction to three NNLSEs and their solutions
4.1. General case
The reduction from the unreduced NNLSEs (i.e. (
2.22), (
2.29), (
2.34)) to the three NNLSEs in (
1.2) is
together with replacing
t → i
t. To achieve the above relation, we introduce constraint on
K1 and
K2 such that
Then, from the dispersion relations (
2.7), (
2.23) and (
2.30), one can always get
4(
4In the diagonal case one should suitably take constants such that .)
Next, the original Sylvester equations (
2.3) yield
and hence we have
thanks to the uniqueness of the solutions of the Sylvester equation. Here . It then follows that
In conclusion, for the three NNLSEs in (
1.2), their solutions can be expressed in the form
with
r1,
s1 and
M1 accordingly.
In the following, for the three NNLSEs, we will look at their explicit solutions and illustrate their dynamics.
4.2. Explicit solutions of the NNLSE-I and dynamics
4.2.1. Formulation of solitons
With the above results, we rewrite the Sylvester equation and dispersion relations as
where
Then, solutions of the NNLSE-1 (
1.2a) are given by (
4.5) where
M1,
r1,
s1 satisfy the above settings. In particular, when
we have
where
Note that , and throughout this section we write
4.2.2. One-soliton solution
For
N = 1 case, we have the following:
where
ρ1,
σ1 are defined as in (
4.10). Substitute them into (
4.5) one obtains
where for complex number , . Then the square module of one-soliton solution, namely, the envelope of the soliton, becomes:
For the shape and motion of ∣
q1∣
2 given by (
4.13), we illustrate them in figure
1.
Figure 1. The shape and motion of the envelope of one-soliton solution given by (4.13) for . (a) 3D plot. (b) 2D plot of (a) at t = 5 (red dot-dashed curve), t = −3 (blue dashed curve) and t = −4 (green solid curve). |
Full size|PPT slide
The soliton ∣
q1∣
2 travels with a fixed amplitude . The top trace of ∣
q1∣
2 is a parabola, which can be derived as
of which the vertex point is
and the velocity of the soliton is
4.2.3. Two-soliton solution
For
N = 2 case, we have
and the dressed Cauchy matrix comes to be
Then the explicit formula of the two-soliton solution (
4.5) becomes:
where
For the shape and motion of ∣
q1∣
2, we illustrate them in figure
2.
Figure 2. The shape and motion of the two-soliton solution given by ∣q1∣2 with (4.16) for . (a) 3D plot. (b) 2D plot of (a) at t = 2 (red dot-dashed curve), t = 0 (blue dashed curve) and t = −2 (brown solid curve). |
Full size|PPT slide
4.2.4. Double-pole solution
The matrix
K1 can also be a triangle Toeplitz matrix (see appendix
A), which will lead to the so-called multiple-pole solutions. We will present solution formula in appendix
A for the Sylvester equations in (
2.3) where
K1 and
K2 are triangular Toeplitz matrices. For the double-pole solution, it can be obtained by setting
where
k1,
ρ1,
σ1 are defined as in (
4.10). Then the dressed Cauchy matrix can be constructed via appendix
A as
The explicit formula of the double-pole solution (
4.5) reads
where
The shape and motion of ∣
q1∣
2 are illustrated in figure
3.
Figure 3. The shape and motion of the double-pole soliton ∣q1∣2 with (4.19) for . (a) 3D plot. (b) 2D plot (a) at t = 4 (red dashed curve), t = 2 (green solid curve) and t = −2 (blue dot-dashed curve). |
Full size|PPT slide
4.3. Explicit solutions of the NNLSE-II and dynamics
4.3.1. Formulation of solitons
Solution
q2 of the NNLSE-II equation (
1.2b) can be expressed through (
4.5) where
M1,
r1,
s1 are determined by
and
In the case of
K1 being a diagonal matrix (
4.8),
r1,
s1,
M1 are given as (
4.9) where
4.3.2. One-soliton solution
The one-soliton solution corresponds to the
N = 1 case, in which we have
where
k1,
ρ1,
σ1 are defined as in (
4.21). Substituting them into (
4.5) one obtains
and then the envelope reads
equation (
4.25) describes a solitary wave traveling with an amplitude , top trace
and velocity
When , the top trace has a similar shape to while when , the top trace has a similar shape to . From (
4.26) we can also find that when , there is
x(
t) = 0, which corresponds to a stationary soliton as shown in figure
4. Note that the non-isospectral effects affect amplitude, velocity and shape of (
4.25).
Figure 4. The shape and motion of stationary one-soliton solution given by (4.25) for . (a) 3D plot. (b) 2D plot of (a) at t = −6 (blue dashed curve), t = 0 (green solid curve) and t = 5 (red dot-dashed curve). |
Full size|PPT slide
4.3.3. Two-soliton solution
When
N = 2 we have
where
kj,
ρj,
σj are defined as in (
4.21). Then, the explicit formula the two-soliton solution (
4.5) becomes:
where
The shape and motion of ∣
q2∣
2 are illustrated in figure
5.
Figure 5. The shape and motion of the envelope of two-soliton solution (4.28) for . (a) 3D plot. (b) 2D plot of (a) at t = −15 (green solid curve), t = −12 (red dashed curve) and t = −6 (blue dashed curve). |
Full size|PPT slide
4.3.4. Double-pole solution
Double-pole solution of the NNLSE-II can be given by the formula (
4.5) with the setting (
4.17) and
M1 takes the form (
4.18), where
k1,
ρ1,
σ1 are defined as in (
4.21). The explicit formula of such a solution reads
where
Shape and motion of ∣
q2∣
2 are illustrated in figure
6.
Figure 6. The shape and motion of the envelope of the double-pole soliton (4.29) for . (a) 3D plot. (b) 2D plot (a) at t = 24 (red dot-dashed curve), t = 20 (green solid curve) and t = 15 (blue dashed curve). |
Full size|PPT slide
4.4. Explicit solution of the NNLSE-III and dynamics
4.4.1. Formulation of solitons
For the NNLSE-III equation (
1.2c), its solutions are formulated by (
4.5) where
M1,
r1,
s1 are determined by
and
In the case of
K1 being a diagonal matrix (
4.8),
r1,
s1,
M1 are given as (
4.9) where
4.4.2. One-soliton solution
For one-soliton, we have
where
k1,
ρ1,
σ1 are defined as in (
4.32). From (
4.5) we have
which yields
The envelope is depicted in figure
7. It is interesting that ∣
q3∣
2 has a time-dependent amplitude , which indicates that the soliton is a localized wave with respect to both space and time. In addition, the top trace for (
4.35) reads
which, in general, is a parabola curve, and along which the soliton travels and changes its direction at the vertex (
t,
x) = (
b1,
Aa1) where ∣
q3∣
2 takes maximum value , see, e.g. Figure
7(c). A special case takes place when
A = 0, which yields a stationary soliton, as depicted in figure
7(a).
Figure 7. The shape and motion the envelope of one-soliton solution given by (4.35) for (a) . (b) 2D plot of (a) at t = 0 (blue dashed curve), t = 0.3 (red dot-dashed curve) and t = 0.5 (green solid curve). (c) , (d) 2D plot of (c) at t = −0.5 (black dashed curve), t = 0 (red solid curve) and t = 0.5 (blue dashed curve). |
Full size|PPT slide
4.4.3. Two-soliton solution
The two-soliton solution is given by (
4.5) where
K1,
r1,
s1,
M1 are given as in (
4.27) but
kj,
ρj,
σj are defined as in (
4.32). The solution is written as
where
For the shape and motion of the envelope ∣
q3∣
2, we illustrate them in figure
8, where (a) shows the scattering of two solitons and (b) describes interactions of two different stationary solitons.
Figure 8. The shape and motion of the envelope of two-soliton solution (4.36) for (a) . (b) . |
Full size|PPT slide
4.4.4. Double-pole solution
Double-pole solution of the NNLSE-III is given by the formula (
4.5) with the setting (
4.17) and
M1 takes the form (
4.18), where
k1,
ρ1,
σ1 are defined as in (
4.32). Its explicit formula is
where
The shape and motion of ∣
q3∣
2 are illustrated in figure
9. When we get two solitons moving along the same parabolic top trace, as shown in figure
9(a). When we get two overlapped solitons as shown in figure
9(c).
Figure 9. The shape and motion of the envelope of the double-pole solution (4.37) for (a) . (b) 2D plot (a) at t = 3 (blue dot-dashed curve), t = 2 (red dashed curve) and t = 1.5 (purple solid curve). (c) . (d) 2D plot (c) at t = 0.2 (green solid curve), t = 0.1 (red dot-dashed curve) and t = −0.4 (blue dashed curve). |
Full size|PPT slide
5. Conclusions
In this paper we have developed the Cauchy matrix approach to the NNLSEs, which serve as example models in the ZS-AKNS hierarchy. We believe that solutions of other order equations in the non-isospectral ZS-AKNS hierarchy, such as the non-isospectral sine-Gordon equation, the non-isospectral modified KdV (mKdV) equation and the non-isospectral Hirota equation (combination of the NLS and the mKdV) can be obtained along with this line.
In the Cauchy matrix approach, the Sylvester equation (e.g. (
2.1)) plays an central role, which defines a dressed Cauchy matrix to provide
τ functions (i.e. ∣
I2N +
M∣) for the investigated equations. In non-isospectral case, one needs to suitably select dispersion relations of the time part (e.g. (
2.7b), (
2.23b) and (
2.30b)) according to the time-evolution of the spectral parameters. One needs also to formulate special relations (e.g. (
2.35)) to figure out the integration term (e.g. in (
2.34)). Apparently, compared with the isospectral case [
23], the non-isospectral extension of the Cauchy matrix scheme is quite non-trivial. In addition, in the isospectral case (see [
23])
K is a constant matrix and it can be proved that
K and its any similar form lead to same
S(i,j) therefore one only needs to consider its canonical form and the resulted solutions can be classified by the canonical forms of
K. However, as we have seen that in non-isospectral case
K is no longer a constant matrix, usually
K and its similar form do not obey the same evolutions (e.g. (
2.9), (
2.24) and (
2.31)). Therefore in the non-isospectral case, we can not classify solutions by considering the canonical forms of
K. In appendix
A we will formulate solutions of the Sylvester equations in (
2.3) where
K1 and
K2 are triangular Toeplitz matrices, which are used to get multiple-pole solutions. Note that the Sylvester equation (
2.1) to formulate the ZS-AKNS system is different from the one for the KdV type and KP type equations (see [
22,
27]). Extension of the Cauchy matrix approach to the non-isospectral KdV and KP type equations (as well as the non-isospectral equations with sources, e.g. [
30]) will be considered elsewhere.
Appendix A. Solutions to (2.3) with triangular Toeplitz matrices
We sketch a procedure to construct solutions to the Sylvester equations in (
2.3) when
K1 and
K2 are lower triangular Toeplitz matrices. A lower triangular Toeplitz matrix is a square matrix in the following form and can be considered to be generated by a certain function:
where
f(
c) is the generating function. Note that the subindex ‘
c’ indicates that the lower triangular Toeplitz matrix is generated by taking derivatives with respect to
c. We also introduce a symmetric matrix generated by
f(
c), denoted as
As a special property of such two types of matrices, we mention that [
26]
For more properties, one can refer to [
29] and proposition 3 in [
26].
In the following we will use the notations
k1,
l1,
ρ1,
ϱ1,
σ1 and
ϖ1 that we introduced in section
3 but we do not specify them since the following description is generic and true for all the three NNLSEs. We consider
k1 and
l1 to be functions of
c1 and
d1, respectively, e.g. (
3.2), (
3.9) and (
3.11). Let
Then, it can be verified that
K = diag(
K1,
K2) satisfies the evolutions (
2.9), (
2.24) and (
2.31) when
k1,
l1 are defined as (
3.2), (
3.9) and (
3.11), respectively. Next, define
where
Then, the above defined elements satisfy the dispersion relations (
2.7), (
2.23) and (
2.30) when
k1,
l1 are defined as (
3.2), (
3.9) and (
3.11), respectively.
Next, we look for solution
M1 and
M2 of the Sylvester equations in (
2.3) in the form
where
G1 and
G2 are unknowns. In these settings, equation (
2.3a) can be rewritten as
Using the relations [
26]
(
A.6) reduces to
For convenience, we write
G1 = (
g1,
g2, …,
gN) where . Then, the first column of (
A.7) reads
which gives rise to
from which one can successively determine
g2,1,
g3,1, ...,
gN,1. For example, we have
Once with
g1 in hand, we can look at the second column of (
A.7), which is
Element of
g2 can be calculated as:
The first few elements are
For the
n-th column (
n > 1) of (
A.7), we have
which indicates that
Finally,
G1 = (
g1,
g2, …,
gN) can be derived.
G2 can be solved from equation (2.3b) in a similar way. Here we skip the details.
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}