TT 75

"'

IC/95/322 INTERNAL REPORT (Limited Distribution)

1

Introduction

International Atomic Energy Agency and United Nations Educational Scientific and Cultural Organization INTERNATIONAL CENTRE FOR THEORETICAL PHYSICS

A NEW DROMION SOLUTION OF THE DAVEY-STEWARTSON EQUATION1

Jizong Lu2 International Centre for Theoretical Physics, Trieste, Italy.

By using the Backlund transformation, Boiti et al [1] found that for the Davey-Stewartson (DS) equation there existed a kind solution which localized at the intersection of two plane waves. This kind solution was later named as the dromion solution. After then dromions were also found by using other methods, for instance, the inverse scattering transformation (1ST) method [2] and the bilinear (BL) method [3]. Why call them dromions? This is because they can be thought to form tracks (dromos in Greek). The system for which dromion solutions have been mostly studied is the DS equation. Using the BL method, dromion solutions for other types of (2+l)-dimensional integrabte models like the generic (2-M)-dimensional nonlinear equations of Schrodinger (NLS), Korteweg-de Vries (KdV) [4] and the Nizhnib-Novikov-Veselov (NNV) equations[5] were also found. These dromions are constructed from plane waves or ghost solitons. Besides ghost solitons, however, there also exist curved line soliton solutions which are finite on a curved line and exponentially (or algebraically) decaying apart from the curve and periodic solitons which are periodic on one direction and decaying in others for some (2+l)-dimensional integrable systems. Can one construct dromions from these solitons? Most recently, it has been shown that dromion solutions can be also made out of curved line solitons for the potential breaking soliton (BS) equation and a (2+l)-dimensionaI KdV type equation [6]. In this letter, the similar kind dromions for the DS equation will be discussed.

ABSTRACT A new dromion solution is obtained for a (2+1)-dimensional integrable model: the Davey- Stewartson equation. Some interesting questions which emerge in the procedure of getting the solution are also discussed.

2

Special solutions of the DS equation

out an The DS equation describes the evolutions of surface waves of slowly varying amplitude and has been widely applied to many fields, such as water waves, plasma physics and nonlinear optics etc. In solution which localized at the intersection of two plan was named as the dromion solution. It can be written in several variant but equivalent forms. Here we use the same form as in [3] [4]:
. -I- uYV - 4u|u| 2 - 2uu = 0,

vxx+vYy+A(\u\2}xx=0.
MIRAMARE - TRIESTE October 1995 Introducing new dependent variables F (real) and G (complex) by

(2)

u~j,

v = -2d\\ogF

(3)

and rotating the coordinate axes by 45°, eqs. (1) and (2) can be written as
(4)

DxDyF-F = 2|G| , 'Submitted to II Nuovo Cimento. 2 Permanent address: Shanghai Teachers University, Shanghai 200234, People's Republic of China. where D is the usual bilinear operator defined by = (dt - dr)m{dx - dx,r(dv -

2

(5)

•,*=V,«=f). (6)

In order to solve eqs. (4) and (5), let us expand G and F in the forms of power series

where C is a constant introduced in the variable separation procedure and Ci(t), C2(i) are integral functions. On the other hand, eq. (17) implies that the only possible form of f(x, t/,t) is , y, t) =

y, 0 + c'(t) f fl-»dz f

(20)

where e is a small parameter. Substituting eqs. (7) and (8) into eqs. (4) and (5) and comparing the coefficients of various powers of t, we have

:

g0

0,

where £i(z,t) = ^i, ^(j/,*) = £2 and c'(t) = d axe functions of indicated variables. Thus eq. (17) is separated to two equations: (21)
9t =

41'

=

0,
2c' f gyldy + c(t),

(22)

This equation system can be solved in principle due to its linearity. It is very difficulty to do so, however, because there are infinite equations. In order to look for its special solutions, we cut off the equation series by setting if>W and ^*' = 0 for all k > N(N = 2,3, - • •). Thus the equation system becomes

where c(t) is another function of (. Substituting eqs. (20) ~ (22) into eq. (15), we find that it is satisfied only for d = d{t) = 0 and there is no further restriction on /, g, £.1 and £j. Substituting eq. (20) into eqs. (18) and (19) and comparing them with eqs. (21) and (22), we get /« = c{t) - 2 / ^ , , , (23) gt = c(t) - 2 ^ 6 , . (24) Particularly, a special case for c(t) = 0 is interesting. In that case eqs. (23) and (24) become

<

= 0, 0,

(10) (U) (12)

F
2gs

{S 2)

(26)

The solution of eq. (10) is

<l> =f{x,t)-g(y,t).
Substituting eq. (13) into eq. (12), we have

M

(13)

If we substitute eqs. (25) and (26) into eq. (16) and separate variables, then eq. (16) will be reduced to two trilinear equations:

+ fL = 0.
(14)

(27)

\

l

9m ~ °-

( 28 )

where p'1 = fx{x,t)gv(y,t) and £ is an arbitrary function of x,y and t. Substituting eqs. (13) and (14) into eqs. (9) and (11) and vanishing both real and imaginary parts respectively yield only three equations to restrict functions f,g and £:
fx9vt

The integrability of eq. (27) (or eq. (28)) is guaranteed by the one of the DS equation. That means that special solutions of the DS equation can be obtained by those from two new (l+l)-dimensional integrable models (eqs. (27) and (28)). However, to give out detail solutions of eqs. (27) and (28) is still very difficult. Here we only write down some special solutions. The simplest solutions possess exponential forms / =
g =

(29) (30)

= 0,

(16)

Separating variables in eq. (15) and integrating with respect to x and y respectively, we get: (18) (19)

with arbitrary constants OQ, bo. 1, b.

3 Dromion solutions
If we were able to obtained the dromion solution by substituting eqs. (29), (30) with eqs. (7), (8), (13) and (14) into eq. (3), no doubt it would be a new kind of dromions. Unfortunately, the solution made out in this way decays in most directions but one, so it is not a dromion solution. We should consider higher order terms of e. After doing so, F and G are modified as follows: F = f'(x, t) - g'{y, t) + Cf(x, t)g'{y, t), G = p(x, y,t}expi[£i(x, t) + £.2(y,t)\} (31) (32)

4 Conclusion
In summary, we would like to point out: 1) Usually one uses the bilinear form (4) and (5) to construct the dromion solution by assuming £, = fox + Uy + W + 6^ In our discussion, d(£ 2 ) is only an arbitrary function j of {x,t}({y,t}). This will make our solutions are more general than usual ones. Even in equation (40), u is different from that given in refs. [1], [2], [3] and [4] due to the arbitrary functions of time t. In deed, if we restrict c3(t) and c4(t) being linear in t and dx(t) = 0 {B = 1) in u, then « identify to them. No doubt, this is a new kind dromion solution of the DS equation, although it does not possess the obvious meaning of curved line dromions as in ref. [6j. 2) Both eqs. (27) and (28) are of trilinear form. This is a new kind of (1+1)dimensional integrable model. Although some other types of trilinear equations have been discussed [7], the knowledge about them are still much less than those of bilinear ones. The further studies about trilinear equations must be very important and valuable. 3) In eqs. (27) and (28) the variables are completely separated. / (or g) is a function of {i,t} (or {y,i}). This implies that the variable separation approach may be applied in non-linear partial differential equations. Recently, some mathematical physicists have paid their attention to this subject. For instance, by using symmetry constrain, Chang and Li [8] got the solution of the (2+l)-dimensional K-P equation from two solutions: one of the (l + l)-dimensional NLS equation with variables {x,y} and the other of the modified (l + l)-dimensional KdV equation with variables {x, t}. However, in that case the variables are not separated completely. We have discussed this subject in ref. [9].

where p2 — fx(x,t)g'y(y,t) and C" is an arbitrary constant. After substituting eqs. (31) and (32) into eq. (4) and doing straight calculation like in Section 2, we obtain the following equations, = 0, (33)
= 0,
(34)
2

• / 2 - 2 / ; / ; M - c ' 2 ( j ) / ; = o, + g'y\ - 2g'ygy
= 0,

(35)
(36)

where c[(t) and c2(t) introduced in the variable separation procedure are two arbitrary functions of t. Comparing eqs. (33) and (34) with eqs. (25 and (26), we find the relation between / , j and f\g'

(38)

Acknowledgments
The author would like to thank Prof. S. Randjbar-Daemi for hospitality at the International Centre for Theoretical Physics, Trieste, where the work was done. The author would also like to thank Prof. S.-y. Lou for his very helpful discussion.

where B = exp J' C'd(t')dt' and c'(t) is another arbitrary function of time t. From eqs. (35) and (36), we get the equations satisfied by / and g. They are just eqs. (27) and (28). Therefore eqs. (27) and (28) do probably possess some universality, which will make them quite meaningful. Now we can construct solutions of the DS equation by using solutions (29) and (30). The solutions made in this way are completely a new kind ones. In order to see this point, let us further assume Sl{x,t) = kx + at> &(!/,'} = < + »*• & (39)

Substituting eqs. (31) and (32) with eqs. (29), (30) and (39) into eq. (3), we have {akbl)i expi(-<^r/2fc - Uy/2l + c3{t) + ct(t)} aexp[-(6 + 6)/2] + /?exp[(£ - fc)/2] + 7 « P [ ( 6 - fi)/2] + ««p[(£, + &)/2] (40) with a = aoB-l-boB + (2/c} + aobo, 0= B~1a(l + bo), -y = bB(a0B~l - I), 6 = cab. The solution (40) decays exponentially in all directions after selecting a, j3,7 and S to possess same sign and abkl > 0, so it obviously is a dromion solution.

T

References
[lj M. Boiti, J, J. -P. Leon, L. Martina and F. Pampinnelli, Phys. Lett. A 132 (1988) 432 [2] A. S. Fokas and P. M. Santini. Phys. Rev. Lett. 63 (1989) 1329 [3] J- Hietarinta and R. Hirota, Phys. Lett. A 145 (1990) 237 [4] J. Hietarinta, Phys. Lett. A 149 (1990) 113 [5] R. Radha and M. Lakshmanan, J. Math. Phys. 35 (1994) 4746 [6] S-y. Lou, Preprint (1995) [7] J. Matsukolaira, J. Satsuma and W. Strampp, Phys. Lett. A 147 (1990) 467; B. Grammat.ices, A. Ramam and .1. Hietarinta, Phys. Lett. A 190 (1994) 6 [8] Y. Chang and Y-s Li, Phys. Lett. A 157 (1991) 22 [9] S-y, Lou and J. Lu, Preprint (1995)