BF 02355320
BF 02355320
BF 02355320
1, 1997
The problem of saturation of monochromatic wave instability in the one-dimensional focusing N S model is con-
sidered by a method unifying the ideas of the inverse scattering method and the complex Whitham deformation
theory. Bibliography: 10 titles.
INTRODUCTION
Instability of the monchromatic wave uo(t) = e 2it in the nonlinear Schrhdinger (NS) equation
is a classical phenomenon first discovered and investigated in the 1960's (here the names of Lighthill,
Benjamin, Feir, and Zakharov should be mentioned among many others).
An analytical description of instability effects was a challenge for theorists for many years (see, e.g.,
the monographs [1, 2, 7] and the references therein). The connection of this problem with the complex
Whitham modulation theory [1] was discussed in [3], but it was not clear at that time how to surpass the
instabilities in the Whitham-NS system itself.
An alternative approach to instability problems of this type, based on the Kolmogorov-Arnold-Moser
(KAM) perturbation technique, was proposed in [4], which, in particular, allowed one to explain some
interesting numerical experiments (see, e.g., [9]). The rigorous results of [4] on the KAM-stability of finite-
'
gap tori "unstable in the linear approxlmatl "on" stimulated interest in the complex Whitham deformations
'
Published in Zapiski Nauchnykh Seminarov POMI, Vol. 215, 1994, pp. 65-76. Original article submitted July 25,
1993.
1596 1072-3374/97/8501-1596518.00
9 Plenum Publishing Corporation
The present paper is the first step in this direction. Here we concentrate on studying the external region
Dex. In this region we confirm the conjecture of [5] and, moreover, we are able to calculate explicitly the
phase shift ~2 = arg u(z, t). We also prove some important qualitative results connected with the behavior
of the solution in the internal region Dint.
For simplicity we assume that u(x, 0) is C~176 and that (u(x, O) - uo(t))lt=o belongs to the Schwartz
class (e.g., has finite support).
Our aim is to find the asymptotics of the solution of the Cauchy problem (1), (2) as t ~ co. It is well
known [2, 6] that (1) can be written as a compatibility condition of the Zakharov-Shabat linear system
g(A)=-iAo.3+
(0i~ 0 ' O'a= -1
0) " (4)
Define the Jost functions r and r as the solutions of (3), (4) fixed by the asymptotics
where
eo(P,x,t) := e x p ( - i k ( A ) ( x - 2At) - it)
(1)
A-k(A) .
uo(t)
Here k(A) = x / ~ + 1 determines the spectral curve F of genus g = 0. We denote the points on r by
P = (A, k). Let F1 and F2 be the upper and lower sheets of I. We assume that i Im A > 0 for A C F~2,
where the superscripts "+" and " - " denote the upper and lower half-planes of C.
Define H+ := r + UF~- and 1I_ := F~" UF +. We choose the branch of the function k(A) so that Im k(A) > 0
for A E H+. The continuous spectrum E of the scattering problem (2)-(5) is E = R U [-i, i]. It has a
complex branch [ - i , i], which is a source of trouble in any ISM investigation.
Standard reasoning similar to that in [6] shows that r is an analytical function in H+ \ co, while
r is analytical in H_ \ co.
On the spectrum E, we have the scattering relation
r = r +r A(P) C E. (6)
Here a(P) and b(P) are scalar functions independent of x and t. Let o. : (A, k) -~ ( A , - k ) denote the sheet
permutation involution on F and let a : (A, k) ~ (A, k) be the complex conjugation anti-involution, which
does not permute the sheets of F.
The function a(P) is analytical in II_ and a(P) --* 1 as P --* 0ot,2,
1597
where a2 =
(0 i) 0 and re(P) = -i(A + k). Hence, for A(P) e E we have
Our aim now is to invert the scattering map u(x,t) --+ {r,r(P)} constructed in See. 1. Consider the
preimage E C F of the spectrum E under the canonical A-projection:
L: := R~ u R2 u ~, u o ~ , , A(s = E. (8)
Here R 1 C F1 is the "upper real axis": A(R1) = R, while IR2 := -aR1. The contour 7 C F1 surrounds
the cut [0, i] so that the sheet FI is left at the left. Standard reasoning (see, e.g., [8]) gives the following
singular integral equation (SIE) for the Jost function r P E Z::
r = eo(P) - -~-~i
1/ M(+)(P, Q)r(Q)r (9)
L
The subscript (+) in (9) means that we consider the limit P' --, P : P' E H+, P E L.
Below we perform an asymptotic analysis of (9).
3. ASYMPTOTIC ANALYSIS IN T H E E X T E R N A L R E G I O N
In Dex we try the following ansatz for the Jost function r x, t), P E ~:
Let ),(P) E IR. Then the integral in (9) contains fast oscillations, and using well-known estimates for
Cauchy integrals we get the "reduced" SIE for the function A(P) (cf. [8]). More precisely, we get
1598
A(P)eo(P)=~(P)- ~
1/
27ri
M(+)(P,Q).r(Q).r(crQ)A(Q)eo(Q).H(f(Q,~)), (12)
R~u~2
Ast(() = 8 (14)
Note that these two stationary points are real only if ( E Dex,
/~st ~ ]~ ~ 1~[ ~__-4V/'~. (15)
For the critical values (~ = 4-4v~ our two stationary points coincide,
A~ : Ast(• = +~. (16)
L e m m a . For ~ E Dex we have
, eo(P,t), t -, R. (17)
Proof. Define
;y= { A E C \ R : ImA>0, Imf(A,~)=0};
"~ is an analog of the Stokes line for the function f(A, (). Note that if [([ > 4V~t, then exactly one complex
"whisker" of ;/comes out from the real stationary point Ast((). By the Cauchy theorem the contour 7 in
(13) can be smoothly deformed to a homotopic contour "~st C zf. The function f(P, () is real for P E z/st
Thus, the integral in (13) vanishes due to the Riemann-Lebesgue theorem, which completes the proof. []
Due to (17) our reduced SIE (12) can be solved by the usual methods (of. [8]):
.4(P) : = e x p
(1/)
A(P, ( ) = .4(+)(P),
2zri
N1UN2
~(P) e R,
2~4(P,Q)logg(Q,~) .
(18)
(19)
As in Sec. 2, the subscript (+) in (18) means that in (19) we take the limit from H+. We use the notation
d,~(Q) k(Q) + k(P)
3)/(P, Q) := ,k(Q) - A(P) 2k(Q) ' g(P' ~) := 1 - r ( P ) r ( a P ) . / l ( f ( P , ~)). (20)
Formulas (18)-(20) and (11) determine all the coefficients of ansatz (10).
This leads us to the main result of this section:
P r o p o s i t i o n 1. In the "solitonless" case (7) the solution u(x,t) of the Cauchy problem (1), (2) has the
following behavior l'or t ~ oc, [~I > 4v~ + a, Va > O:
u(x,t) = exp(2it + icg(4)) + OIt -1/2) (21)
with
1
~( ~) = - -~ f log g(Q, dk((Q)
Q) . (22)
The order of decreasing of the correction term in (21) is 1/2 due to the nondegeneracy of stationary points.
1599
4. WItlTItAM DEFORMATIONS IN THE INTERNAL REGION
Now consider the region Dint: [~l --- 4V'2, t --+ oo. Following [5] we look for the solution u(x, t) of the NS
model, having the form
u(z,t)=Ul(Z, tlr(())+O(t-A), A>0, (23)
where ul is a one-phase NS solution. There are well-known explicit formulas for ul in terms of the elliptic
functions (see, e.g., [5]). Here we only note that ul in (23) is constructed by the deformed spectral curve
F(~) of genus g = 1:
g+l
with the branch points hi(() obeying the reduced Whitham equations
The characteristic speeds sj are given by the universal formula sj := s(h)]x=~, where s(h) is a rational
function of h,
s(h) = dg~2(P)/d~l(P), P 9 F((). (25)
The Abel integrals ~1 and Ft2 ( - quasimomentum and quasienergy) are standard in the NS theory. At the
infinite points their asymptotics is described by
f/,(P) , +(h + w, + O(1/h)), P ~ oo +, (26)
f~2(P) , =F(2h + w2 + O(1/A)),
and they are normalized by the conditions
dft, = ~ dft2 = 0, i=l,...,g, (27)
i i
where ai is the cycle surrounding the cut [hi, Xi], i = 1,... ,g.
Formulas (24)-(27) are generalized to the multiphase case g > 1, but here we concentrate on the case
g=l.
T h e o r e m 1. For any nonsingular curve P (*) we have the inequalities
I*jl < 0 <[0h, < vj (28)
Proof. The proof is simple. The first inequality follows from the definition (25) and the fact that the zeros
of dr/1 (P) cannot coincide with the branch points hi, as seen from consideration of the harmonic function
v(P) = I m f h ( P ) .
To prove the second inequality we use identity
Osj _ 1 0s(h) l , Vj, (20)
0h j - 2 " 0h ~=x;
which easily follows from (26), (27). []
We call the nontrivial dependence hi({) satisfying (24) the Whitham "motion" of ,kj. Theorem 1 implies
that this motion can proceed only with finite speed. Moreover, we shall see below that in contrast to
"stable" situation [3] this motion even theoretically cannot last for infinite "time" {! (We emphasize that
we are speaking only about the case g = 1.)
All properties of the Whitham motion in our "unstable" situation (here instability means that, typically,
sj have nonzero imaginary p ~rts, which leads to the exponential growth of perturbations in the generic
Whitham system, see [3, 5, 7]) are closely connected with the picture of the level set S,
S={heC : Ims(h)=0, ImA>0}.
It is easy to see that cx~ ~ S and the function Re(s(h)) is monotone along S. It is also clear that the moving
point hi({) should belong to S. It is interesting that (a) only one branch point can belong to S and (b)
the moving point hj(~) has a smaller imaginary part than the "stable" point A~ 5s hi. This follows from
1600
T h e o r e m 2. Let I m s l = 0. Then
Proof. The proof is a straightforward consequence of the explicit formulas for the speeds s d (see e.g., [SJ).
[]
Note that a priori it is not obvious that the complex motion in (24) can exist. To prove the existence
theorem, in [5] we used the implicit function theorem.
T h e o r e m 3. Fix the stable point A2 = i. Then the equation
Proof. Expand (31) into the Taylor series in the vicinity of ~ . The small parameter is ~ = IIm),ll. It is
necessary to take into account the O(84) coefficients. Then we apply the implicit function theorem to the
function IF = (Ims(A1))/(ImA1) 3. []
Using this local result and Theorem 1 we easily prove a global result.
T h e o r e m 4. Fix A2 - i. Then F_,qs. (23) uniquely define the motion of A~ (() for ~ ff Diat.
Due to (24) the evolution of the moving point A1 proceeds along the set
Note that the function 8(,~ 1) is monotone along S1. The c~rve $1 connects the po::~ts A~r and A+r and
passes through A1 = A2 = i. It is obviously symmetric with respect to the transformation A1 ~ - ~ 1 .
The picture of the deformation F(~) is shown in Figs. 1 and 2. This qualitative picture is confirmed by
T h e o r e m 5. For 0 < [~t < 4 v ~ the moving point ~1 -~ fl + i6 satisfies the inequMities
0 06{>0, (32)
aT <0,
Proof. We use the real part of (24), which has the form
I -- ~
4fl + 2. - - 7 - - -- ~. (33)
Note that by virtue of Theorem 1, 5~ = fl~ = 0 is impossible. It follows from (33) that fl~ cannot vanish
"earlier" than 5~. Then using (33) we prove that 6(() is a nondecreasing smooth function for 0 > ( > -4x/-2.
Finally, we take into account the local behavior [5] of the functions fl(~), 5(~) near A~,
1 - 4 x / 2 -F c, 0 < e << 1.
1601
To finish the proof, note that (33) is invariant with respect to )~1 ~ - - ~ 1 , ~ ~ --~. []
R e m a r k 2. Here we have not considered the problem of justifying the Whitham ansatz (23) in Dint. This
more difficult problem will be considered separately. Note that in the "stable situations" the corresponding
analytical technique is well developed [3, 8].
R e m a r k 3. In analogy with the stable situations, the Whitham deformations considered above can be
applied to some step-like problems for the focusing NS model (1). For example, it is easy to construct a
Whitham curve F(() satisfying the boundary conditions
V(~) , r• ~• (34)
For example, if a _ = i and or+ = 0, then the deformation is actually shown in Fig. 1. This case corresponds
to the solution of the step problem
1, 37 --+ --(X3,
o) , (36)
0, x --+ +oc.
See the picture of the solutions' amplitude lu(x, t)l for t ~ ~ in Fig. 4.
In conclusion, we note that in the case R e a + < R e a _ there is no genus-one Whitham solution F(~)
satisfying (34). We conjecture that the Whitham modulations of genus two wiU be of crucial importance
for this problem.
...., l, |
/
/
/
/
/
/
*Added by the author: Now this statement is proved. The proof will be published in the author's paper "Complex W h i t h a m
deformations in problems with integrable instability," to appear in Zap. Nauch. Semin. P O M I (1995).
1602
|
\
\
\
\
lul
(,
=I
lu, I-'t
! !
-q4~r
FIG 3 . S A T U R A T I O N OF INSTABILITY FOR t ---y ~ : BEHAVIOR OF THE
AMPLITUDE [U(X,t)l OF THE SOLUTION OF PROBLEM (1), (2).
-4~1; I x =
Translated by R. F. Bikbaev.
Literature Cited
1603
3. R. F. Bikbaev, "Diffraction in nonlinear defocusing medium," Zap. Nauch. Semin. LOMI, 179, 23-31
(1989).
4. R. F. Bikbaev and S. B. Kuksin, "A periodic boundary-value problem for the sine-Gordon equation,
small Hamiltonian perturbations of it, and KAM-deformations of finite-gap tori," Algebra Analiz, 4,
No. 3, 42-78 (1992).
5. R. F. Bikbaev and V. R. Kudashev, "Whitham deformations partially saturating the modulational
instability in the nonlinear SchrSdinger equation," Pisma Zh. Eksp. Teor. Fiz., 59, 741-745 (1994).
6. L. D. Faddeev and L. A. Takhtadzhyan (Takhtajan), Hamiltonian Methods in the Theory of Solitons,
Springer, Berlin (1987).
7. S. K. Zhdanov and B. A. Trubnikov, QuasistabIe Unstable Media [i~ Russian], Nauka, Moscow (1991).
8. R. F. Bikbaev, "On the shock waves in one-dimensional models with cubic nonlinearity," Teor. Mat.
Fiz., 97, No. 2, 191-212 (1993).
9. V. E. Zakharov, M. F. Ivanov, and L. N. Shur, "On an anomalously slow stochastization in some
two-dimensional field theory models," Pisma Zh. Eksp. Teor. Fiz., 30, 39-44 (1979).
10. A. Z. Patashinskii and V. L. Pokrovskii, Fluctuation Theory of Phase Transitions [in Russian], Nauka,
Moscow (1982).
1604