The Nonlinear Schrodinger Equation

Eran Bouchbinder
December 2003,
I. INTRODUCTION
The nonlinear Schrodinger equation is an example of a universal nonlinear model that
describes many physical nonlinear systems. The equation can be applied to hydrodynamics,
nonlinear optics, nonlinear acoustics, quantum condensates, heat pulses in solids and various
other nonlinear instability phenomena. In this short review I will present a derivation of the
nonlinear Schrodinger equation in the framework of the general Hamiltonian formalism for
nonlinear waves and analyze some of its remarkable features. The modulation instability of
plane waves and the appearance of solitary waves will be analyzed. In addition, conservation
laws will be discussed.
II. DERIVATION OF THE EQUATION
Let us consider a spectrally narrow wave packet propagating in an isotropic nonlinear
medium. Expanding the Hamiltonian of the system in terms of the amplitude a
k
of the
packet one obtains
H = H
2
+H
3
+H
4
, (1)
where H
2
=


k
|a
k
|
2
dk is the free Hamiltonian, H
3
is the cubic interaction term responsible
for three waves processes like conuence and decay and H
4
is the forth order interaction term
responsible for four waves processes which include scattering. Our wave packet occupies a
narrow region around k
0
in k-space, which means that the amplitude does not vanish only
for k k
0
. Therefore, processes that change the number of waves are non-resonant and all
the interaction terms, except the scattering term in H
4
, can be eliminated by a canonical
transformation. In that case, Hamiltons equation
i
a
k
t
=
H
a

k
, (2)
1
takes the form
i
a
k
t

k
a
k
=
1
2

T
k123
a

1
a
2
a
3
(k + k
1
k
2
k
3
) dk
1
dk
2
dk
3
. (3)
Now, since the wave packet is narrow we have k = k
0
+q with |q| |k
0
|. Then one can
expand the dispersion relation around k
0
to get
(k) =
0
+ q v +
1
2
q
i
q
j

k
i
k
j

, (4)
where v is the group velocity. In the case of isotropic medium the dispersion relation depends
only on the magnitude of k, so we obtain
q
i
q
j

k
i
k
j

= q
i
q
j

k
j

k
i
k

= q
i
q
j

k
i
k
j

k
2
+

ij

k
i
k
j
k
2

v
k

q
2

+
q
2

v
k
, (5)
where q

and q

are the projections of q parallel and perpendicular to the direction of prop-

i

t
q v
q
2

2

q
2

v
2k

q
= T

3
(q + q
1
q
2
q
3
) dq
1
dq
2
dq
3
, (6)
where T
1
2
T
0123
. Transforming this equation to the r-space and performing all the -
functions integrations on the RHS we obtain
i

t
iv

z
+

z
2
+
v
2k

= T||
2
, (7)
where we have chosen the coordinates system such that v z and

x
2
+

2

y
2
. (8)
Let us give a physical interpretation to the terms appearing in this equation. The term
proportional to v describes a propagation with the group velocity, the term involving the
second derivative in the direction of propagation describes the dispersion, the term involving
the transverse derivatives describes diraction and the term proportional to T describes the
nonlinearity of the system. Note that the nonlinear term could be regarded from the very
beginning as part of a nonlinear frequency with a contribution that depends on the square
of the amplitude. In that case
T =

|a|
2
. (9)
2
This observation will be useful in understanding the self-focusing property for transverse
perturbations. By transforming to a co-moving frame of reference z z vt (this is the
reason why we expanded
k
to second order in q) and rescaling the transverse coordinates
x x

/v, y y

/v we obtain the famous nonlinear Schrodinger equation

t
+

2

2
T||
2
= 0. (10)
III. MODULATION INSTABILITY
The nonlinear Schrodinger equation has a plane wave stationary solution with an ampli-
tude dependent phase

0
(t) = A
0
exp [iTA
2
0
t]. (11)
Let us study the stability of this solution. Consider a longitudinal perturbation of both the
amplitude and the phase
(z, t) = A(z, t) exp [i(z, t)], A(z, t) =

A(z, t) + A
0
, (z, t) =

(z, t) TA
2
0
t. (12)
Substitution of the perturbed solution into the nonlinear Schrodinger equation (10) and
separating real and imaginary parts we obtain
A
t
+

A
z

z
+

2

zz
A = 0, A
t
+

2
A
zz

2

2
z
A TA
3
= 0, (13)
where dierentiation with respect to a variable is denoted by a subscript. Representing the
amplitude and the phase of the perturbed solution as sums of the perturbation and the
plane wave solution, then linearizing with respect to the perturbation, we obtain

A
t
+

2
A
0

zz
= 0,

t
= 2TA
0

A +

2A
0

A
zz
. (14)
We can eliminate the phase to obtain one equation for the perturbation amplitude

A
tt

TA
2
0

A
zz
+

2
4

A
zzzz
= 0. (15)
Therefore, if we assume that

A exp [i(kz t)] we obtain

2
= T

A
2
0
k
2
+
2
k
4
/4. (16)
The second term in the RHS is positive, therefore the solution will be unstable against long
wavelength perturbations (small ks) for T

< 0. Note that the maximal growth rate

2T/

. For long water waves we have

k which implies

< 0.
Therefore, if T > 0, the waves will be modulated. In passing, we consider briey transverse
perturbations. In that case, we should replace

with v/k which is positive. Thus, the

|a|
2
< 0, (17)
which means that instability wave velocity decreases with increasing amplitude. This situ-
ation leads to self focusing. The parts of the wave front where the amplitude is larger will
propagate slower and this eect will enhance itself by the focusing from adjacent parts.
IV. SOLITARY WAVES AND CONSERVATION LAWS
In the previous section we studied the modulation instability of the nonlinear Schrodinger
equation. Next we will consider the outcome of this instability, meaning that we are looking
for new stationary solutions. Physically, we expect to nd such solutions if the pressure
of waves can balance the attraction force. In other words, the fact that we introduced
nonlinearity into the equation might lead to qualitatively new phenomena since dispersion
is no longer the only relevant quantity as in the linear equation. This physical intuition also
implies that the solution cannot be obtained by small nonlinear perturbation of the linear
problem. If the nonlinearity cannot balance the dispersion, we expect to nd collapse (i.e.
nite time singularities) or dispersion that leads to decay. It turns out that the existence
of such a solution is dimensionally dependent (a calculation was presented in the lectures).
Let us focus on the one-dimensional case. Consider a traveling wave solution of the form
(z, t) = A(z ut) exp

if(z ut) iTA

, (18)
where A
0
and u are parameters of the solution. Substituting this form into the nonlinear
Schrodinger equation and separating real and imaginary parts yields
uA
z
=

A
z
f
z
+
f
zz
A
2

, A
zz
= f
z
A

f
z

2u

+
2T

(A
3
A
2
0
A). (19)
First consider u = 0 (a standing wave solution). In that case, the rst equation can be
solved by f = f
z
= 0 and we are left with
A
zz
=
dU
dA
, U(A) =
2T

(A
4
2A
2
0
A
2
), (20)
4
where we introduced the potential U(A) so that the equation will have the form of Newton
equation with A representing space coordinate and z representing time. Recall that T/

<
0 as required by the modulation instability. In full analogy to the Newtonian case we have
a conserved quantity, the energy of the system,
E =
1
2

A
z

2T

(A
4
2A
2
0
A
2
) . (21)
U
A
Solitary Wave
Periodic
Wave
Periodic
Wave
Plane
Wave
FIG. 1: The gure shows the potential U(A) and several possible total energy levels. The rst
case E < 0 (or E > 0) (dashed red lines) corresponds to periodic waves. The second case E = 0
(black line) corresponds to a solitary wave. The green solid line at the level of the minima of U(A)
corresponds to a plane wave solution that was discussed earlier and was added for completeness.
Fig. 1 shows the dependence of the solution on the value of E. For E < 0 (or E > 0) the
particle oscillates in time, which corresponds to a periodic wave in z. For E = 0 it takes
innite time to reach zero, which means that A 0 as z . This localized solution
is called a standing soliton whose form can be obtained by integrating Eq. (21) with E = 0
(z, t) =

2A
0
cosh
1

A
0

2T

exp (iTA
2
0
t). (22)
In order to derive the moving solitary wave solution we multiply the rst equation in (19)
by A and integrate to get
f
z
=
u

A
2
(A
2
A
2
0
), (23)
where we xed the constant of integration such that f
z
= 0 for A = A
0
. Substituting
this equation into the second equation in (19), we obtain an equation for A, which can be
5
integrated to yield
(z, t) =

2A
0
cosh
1

A
0

2T

(z ut)

exp

iTA
2
0
t + i(z ut)
u
2

. (24)
Note that this solution reduce to Eq. (22) for u = 0 and that the velocity and amplitude
are independent. The features of the solution are remarkable. In the linear case there are
non-local modes (plane waves) that one should superimpose in order to construct a localized
wave packet that spreads out during the time evolution due to dispersion. In the nonlinear
case, the solution (24) is a localized wave that preserves it shape during the time evolution.
The fact that the soliton solution is localized allows us to superimpose N-solitons that are
suciently far from each other to yield a new solution. In that sense, one can think of the
solitons as the analogs of plane waves in the linear case. Nevertheless, the analogy cannot
be full. Consider two highly separated moving solitons and assume that they are moving
one towards the other (or that they move in the same direction where the one behind is
faster). Therefore, whatever the initial distance between the solitons they will get closer
such that the nonlinear interaction is important and clearly the superposition solution is no
longer valid. Further analysis is needed to answer the following questions: What is the form
of the solution during the interaction? What happens after the interaction? or generally,
how an arbitrary initial envelope evolves? The method that enables one to answer these
questions in known as the Inverse Scattering Transform (IST). A detailed report on the
method is far beyond the scope of this review. Here, I will just sketch the philosophy behind
the method [1]. The method might be regarded as the nonlinear Fourier analysis since it
shows a remarkable analogy to the linear Fourier analysis and actually reduces to it in the
linear limit. Since linear problems are characterized by their dispersion relations, one can
solve them using Fourier Transforms with the following scheme
Suppose now that one can construct a transformation (direct scattering transformation)
from (z, 0) to a new eld S(, 0) (scattering data) whose time evolution is relatively simple.
Moreover, suppose one can recover the evolved eld (z, t) from the scattering data (inverse
scattering transform). To be more concrete, the eigenvalues of the linear 2 2 eigenvalue
problem
i

1
t
+
2
= E
1
i

2
t

1
= E
2
(25)
6
(z,0)
k
(k,0)
(z,t)
Fourier Transform
Inverse Fourier
Transform

k
(k,t)=

k
(k,0)e
i(k)t

(k)
are conserved when evolves according to the one-dimensional nonlinear Schrodinger equa-
tion. actually plays the role of the potential in Eqs. (25). This is the source of the name
scattering in that context; information on the scattering data enables one to recover the
scattering potential , which is the solution of the required nonlinear problem. In general
one has the scheme
(z,0) S(,0)
(z,t)
Direct Scattering
Inverse Scattering
S(,t)
Time evolution
of the
scattering data
Before going back to the problem of two interacting solitons, let us consider conservation
laws for the nonlinear Schrodinger equation. The equation has innite number of integrals
of motion. In that sense, the equation can be regarded as describing an innite dimensional
Hamiltonian system [1]. Starting with Eq. (25) it is relatively easy to show that [1]
C
n
=

n
dz, (26)
with
n
satisfying the recursion relation

n+1
=

z
+
n1

k=0

nk1
, n 1

0
=

,
1
=

z
, (27)
7
is time independent, i.e. an integral of motion. Note that in order to conform with the
notation of [1] we set

= 2 and T = 1. The three rst integrals of motion

||
2
dz, C
1
=

z
dz, C
2
=

(|
z
|
2
+||
4
)dz, (28)
are trivial and correspond to the number of waves, the momentum and the energy, respec-
tively. Using Eq. (27) we obtain for the rst non-trivial integral of motion
C
3
=

zzz
+ 4||
2

z
+||
2

)dz . (29)
Note that by taking the time derivative of C
n
and using the nonlinear Schrodinger equation
one can check directly the validity of any conservation law. Following the same procedure
one can generate innitely many polynomial integrals of motion. The deep reason for this
surprising result stem from the fact that the nonlinear Schrodinger equation (as well as other
universal nonlinear models like KdV equation) are innite dimensional Hamiltonian systems
for which the number of integrals of motion equals the number of degrees of freedom.
The method of IST enables one to obtain N-solitons solutions. Before analyzing such
solutions let us understand physically the scenario of the interaction using the conservations
laws described above. When the two solitons are highly separated, as was explained above,
the nonlinear interaction can be neglected and the solution is a superposition of two isolated
solitons. When the solitons are close enough the nonlinear interaction can no longer be
neglected and the solution should be derived using IST. What happens after the interaction?
The total number of waves is conserved (recall that only the scattering part of H
4
, a process
that conserves the number of waves, was taken into account in the initial Hamiltonian H).
Therefore we expect the interaction to reproduce two waves. Moreover, energy-momentum
conservation and the existence of innitely many other polynomial integrals of motions (in
the language of IST we would say that the spectrum of Eqs. (25) is conserved). Therefore we
expect the velocities and the shapes of the solitons to be conserved when they are far enough
from the interaction region. This is a streaking prediction. It seems that even though the
equation is nonlinear the solitons collide elastically. The only trace that the interaction can
leave on the solitons, i.e. a measurable eect that is consistent with the conservation laws,
is a phase shift relative to the motion in the absence of interaction. In order to verify these
predictions we should analyze the exact two-solitons solution. Let us set, for simplicity,

= 1 and T = 1 in the nonlinear Schrodinger equation (10). Then the two-solitons

(1 + v
2
)(1 + a
2
)

exp (z) cosh (x

i) +

exp (z) cosh (x + + i)

1 + iav

, x = az + (1 a
2
)(1 vt)
= avz + (1 + a
2
)(1 v
2
)
t
2
, = vz + a(1 + v
2
)t, (31)
and a is related to the average amplitude.
This solution describes the collision in the center of mass frame of reference such that
the relative velocity is 2v, with one soliton larger and slower than the other. The solution
is shown in Fig. 2 for a = 0.5, v = 1 and t = 10, 0, 10. It is clear that the picture that was
predicted before is indeed reproduced by the two-solitons solution. Note that the interaction
is a short range interaction. Moreover, taking the limit t in Eq. (30) shows that the
-20 -10 0 10 20
z
-20 -10 0 10 20
z
-20 -10 0 10 20
z
FIG. 2: Three snapshots of the two-soliton solution for a = 0.5 and v = 1. The gures from
top-left, in a clockwise direction, show |(z, t)| for t = 10, 0, 10 respectively.
one-soliton solutions are recovered with the only dierence from the interaction-free case is
phase shifts. Fig. 3 ,taken from [1], depicts schematically the eect of the interaction.
9
FIG. 3: A typical interaction of two solitons at succeeding times.
V. SUMMARY
In this brief review I introduced the nonlinear Schrodinger equation and discussed some
of its remarkable features. The modulation instability in which a plane wave solution is
unstable against longitudinal perturbations was presented and the criterion of instability
derived. Other solutions were discussed using the analogy to Newton equation. The one
soliton solution separates between periodic wave solutions. Basic features of the one soliton
solution were discussed. The basic philosophy of the Inverse Scattering Transformation
was presented, emphasizing its structure as a generalization of the Fourier Transform. The
integrals of motion of the equation were presented and their consequences regarding solitons
interaction were discussed. An exact two soliton solution was cited and its properties were
discussed and displayed, including the phase shifts accumulated during an interaction event.
[1] M.J. Ablowitz and H. Segur, Solitons and the inverse Scattering Transform, (Siam, Philadel-
phia, 1981).
[2] J.P. Gordon, Opt. Lett. 8, 596 (1983).
10