CHINESE JOURNAL OF PHYSICS

VOL. 41 , NO. 5

OCTOBER 2003

Coherent Soliton Structures with Chaotic and Fractal Behaviors in a Generalized (2+1)-Dimensional Korteweg de-Vries System Chun-Long Zheng∗
Department of Physics, Zhejiang Lishui Normal College, Lishui 323000, China Shanghai Institute of Mathematics and Mechanics, Shanghai University, Shanghai 200072, China and Department of Physics, Zhejiang University, Hangzhou 310027, China (Received October 21, 2002) In higher dimensions there are abundant coherent soliton excitations. In this work, we reveal a novel phenomenon whereby the localized excitations show chaotic and fractal behaviors in some (2+1)-dimensional physical models. To clarify this interesting phenomenon, we take the generalized (2+1)-dimensional Korteweg de-Vries systems: vt +avxxx +bvyyy +cvx +dvy = 3a(uv)x + 3b(vw)y , vx = uy , vy = wx as a concrete example. By means of a new variable separation approach, a quite general variable separation solution of this system is derived. Along with the usual localized coherent soliton excitations such as dromions, lumps, rings, peakons, and oscillating soliton excitations, some new excitations like chaos and fractals are derived by introducing some types of lower dimensional chaotic and fractal patterns.
PACS numbers: 03.40.Kf, 03.65.Ge, 05.45.Yv

I. INTRODUCTION

Three important aspects of nonlinear science are solitons, chaos, and fractals [1]. They are widely applied in many natural sciences such as chemistry, biology, mathematics, communications, and especially in almost all branches of physics such as fluid dynamics, plasma physics, field theory, optics, and condensed matter physics [2]. Conventionally, the three aspects are treated independently, since one often considers that solitons are the basic excitations in an integrable model while chaos and fractals are elementary behaviors of a non-integrable system. In other words, one does not analyze the possibility of the existence of chaos and fractals in a soliton system. However, the above consideration may not be complete, especially in some higher dimensions. In our recent study of soliton systems, we have found that some characteristic lower dimensional arbitrary functions exist in the exact excitation of certain two dimensional integrable models. This means that any lower dimensional chaotic and/or fractal solution can be used to construct an exact solution of

http://PSROC.phys.ntu.edu.tw/cjp

442

c 2003 THE PHYSICAL SOCIETY OF THE REPUBLIC OF CHINA

VOL. 41

CHUN-LONG ZHENG

443

a higher dimensional integrable model, which also implies that any exotic behavior may propagate along this characteristic. To verify the above viewpoint, we take the generalized (2+1)-dimensional Korteweg de-Vries system (GKdV) as a concrete example [3]: vt + avxxx + bvyyy + cvx + dvy − 3a(uv)x − 3b(vw)y = 0, vx = u y , vy = wx , (1) (2)

where a, b, c and d are arbitrary constants. When c = d = 0, the (2+1)-dimensional GKdV system degenerates to the usual two dimensional Korteweg de-Vries(2DKdV) equation [4]: vt + avxxx + bvyyy − 3a(uv)x − 3b(vw)y = 0, vx = u y , vy = wx , (3) (4)

which is an isotropic Lax extension of the well known (1+1)-dimensional KdV equation. Many types of soliton solutions of the 2DKdV equation have now been studied by many authors. For instance, Boiti, Leon, Manna and Pempinelli [3] solved the 2DKdV equation via the inverse scattering transformation. Tagami [5], and Hu and Li [6] obtained its solitonlike solution by means of the B¨cklund transformation. Hu [7] has also given a nonlinear a superposition formula for the 2DKdV equation. Ohta [8] obtained the Pfaffian solutions for this mode. Radha and Lakshmanan [9] constructed its dromion solution from its bilinear form after analyzing its integrability aspects. Lou [10] obtained some new special types of multi-soliton solutions of this equation, by using the standard truncated Painlev´ analysis, e and localized excitations by means of a variable separation approach. However, the (2+1)-dimensional GKdV system has not been studied so well, although Boiti, Leon, Manna and Pempinelli [3] solve this system via the inverse scattering transformation, and Radha and Lakshmanan have constructed its multidromion solutions [9] while Zhang [11] obtained an exact solution for this system based on an extended homogeneous balance approach. In this paper, we investigate the GKdV system further. More attention is paid to its localized coherent structures, especially the localized coherent soliton structure with some novel properties like fractal and chaotic properties. In Sec.2, we outline the main procedures of the variable separation approach and apply this approach to the GKdV system. In Sec.3, we discuss the abundant localized solution structures based on results obtained by the variable separation approach. A brief discussion and a summary are given in the last section.

II. VARIABLE SEPARATION APPROACH AND ITS APPLICATION TO THE (2+1)-DIMENSIONAL GKDV SYSTEM

II-1. General theory of the variable separation approach It is well known that solving nonlinear physical models is much more difficult than solving linear ones. In linear physics, the Fourier transformation (FT) and the variable

444

COHERENT SOLITON STRUCTURES WITH . . .

VOL. 41

separation approach (VAS) are the two most important solution methods. The celebrated inverse scattering transformation (IST) can be viewed as an extension of the Fourier transformation method. However, it is difficult to extend the VAS to nonlinear physics. Recently, two kinds of “variable separating” procedures have been established. The first method is called symmetry constraints or nonlinearization of the Lax pairs [12], because this type of procedure is used only for integrable models which possess Lax pairs. Lou and Chen extended the method to some nonintegrable models. This method can be equivalently called the “formal variable separation approach” (FVSA) [13]. The independent variables of a reduced field in the FVSA have not been totally separated; the reduced field satisfies some lower-dimensional equations. The second kind of variable separation method was first established for the (2+1)-dimensional Davey-Stewartson (DS) equations and the asymmetric DS equation [14], and then recently revisited and developed for various (2+1)-dimensional models, like the generalized Ablowitz-Kaup-Newell-Segur (AKNS) system, the nonlinear Schr¨dinger equation (NLS), the usual and asymmetric Nizhnik-Novikov-Vesselov (NNV) o system, and the long dispersive wave system and Maccari system [15-21]. The main idea is that by solving the bilinear equations or higher multi-linear equations of the original models and introducing a prior ansatz, some special types of exact solutions of (2+1)-dimensional nonlinear models can be obtained from some (1+1)-dimensional variable separation fields. Here we briefly describe the basic procedures of the variable separation approach. For a general nonlinear physical system, P (v) ≡ P (x0 = t, x1 , x2 , · · · , xn , v, vxi , vxi xj , · · · ), (5) where v = v(v1 , v2 , · · · , vq )T (T indicates the transposition of a matrix), P (v) = (P1 (v), P2 (v), · · · , Pq (v))T , and Pi (v) are polynomials of vi and their derivatives. First, we make a B¨cklund transformation a
αi

vi =
j=0

vij f j−αi ,

i = 1, 2, · · · , q,

(6)

where the viαi are arbitrary known seed solutions of Eq. ( 5). In the usual cases, α i should be as small as possible, since substituting Eq. (6) into Eq. (5) can yield relatively simple multi-linear equations in this situation and is determined by the leading term analysis (suppose f ∼ 0 ). Inserting Eq. (6) with α i into Eq. (5) and requiring the leading and sub-leading terms to vanish, we can then derive {v ij , j = 0, 1, 2, · · · , αi − 1}. If the original model is integrable, this procedure will result in its bilinear or higher multi-linear equations. Second, after obtaining multi-linear equations for the original system, we select an appropriate variable separation hypothesis. For integrable models, it can be chosen as modifying Horita’s multi-soliton forms. For instance for some celebrated physical models [22-24] we often take for f an ansatz such as f = a0 + a1 p(x, t) + a2 q(y, t) + a3 p(x, t)q(y, t), (7)

where the variable separation functions p(x, t) ≡ p and q(y, t) ≡ q are only functions of (x, t) and (y, t), respectively, and a 0 , a1 , a2 , and a3 are arbitrary constants. When p and q are set as exponential functions, Eq. (7) is just Hirota’s two-soliton form.

VOL. 41

CHUN-LONG ZHENG

445

Finally, we determine the variable separation equations, which the variable separation functions p and q should satisfy after substituting the ansatz (7) into the multi-linear equations. It is worth mentioning that this procedure is tedious, since different physical models must be disposed in different ways. In the next subsection, we apply this approach to a (2+1)-dimensional GKdV system. The variable separation approach shows its efficiency in searching for localized excitations of nonlinear physical models. II-2. Variable separation approach for the (2+1)-dimensional GKdV system According to the above procedures, we first take the following B¨cklund transformaa tion of v, u, and w in Eqs. (1) and (2)
α1 α2 α3

v=
j=0

vj f j−α1 , u =
j=0

uj f j−α2 , w =
j=0

wj f j−α3 ,

(8)

where vα1 , uα2 , and wα3 are arbitrary seed solutions of the GKdV system. By using the leading term analysis, we obtain α1 = α2 = α3 = 2. (9)

Substituting Eqs. (8) and (9) directly into the Eqs. (1) and (2), and considering the fact that the functions v2 , u2 and w2 are seed solutions of the model, yields
4

P1i f i−5 = 0,
i=0 2 2

(10)

P2i f
i=0

i−3

= 0,
i=0

P3i f i−3 = 0,

(11)

where P1i , P2i , P3i are functions of {vj , uj , wj f, j = 0, 1} and their derivatives. Due to the complexity of the expressions for P 1i , P2i , and P3i , we omit their concrete forms. By requiring the leading and sub-leading terms of Eqs. (10) and (11) to vanish, the functions {vj , uj , wj , j = 0, 1} are determined. Inserting all the results into Eq. (8) and rewriting its form, the B¨cklund transformation becomes a v = −2(ln f )xy + v2 , u = −2(ln f )xx + u2 , w = −2(ln f )yy + w2 . For convenience of discussion, we choose the seed solutions v 2 , u2 and w2 as v2 = 0, u2 = p0 (x, t). w2 = q0 (y, t), (13) (12)

where u2 (x, t) and w2 (y, t) are arbitrary functions of the indicated arguments. Substituting Eqs. (12) and (13) into Eqs. (1) and (2), Eq.(2) is now an identical equation, while Eq. (1) assumes the following symmetric form
2 2 (c−3ap0 )(2f fx fxy −f 2 fxxy −2fx fy +f fxxfy )+(d−3bq0 )(2f fy fxy −f 2 fyyx −2fy fx +f fyy fx )

446

COHERENT SOLITON STRUCTURES WITH . . .

VOL. 41

2 +a[2fx (2f fxxxy + 3fxx fxy − 2fxxx fy ) − f 2 fxxxxy + f fxxxxfy − 2fxxx fxy − 6fx fxxy ]

2 +b[2fy (2f fyyyx + 3fyy fxy − 2fyyy fx ) − f 2 fyyyyx + f fyyyy fx − 2fyyy fxy − 6fy fyyx ]

+f (fxy ft + fxt fy + fyt fx ) − f 2 fxyt − 2fx fy ft + 3(ap0x + bq0y )(f 2 fxy − f fxy ) = 0, (14) which can be rewritten as a multi-linear equation with respect to f . It is a difficult task to determine the variable separated equations after substituting the ansatz (7) into Eq. (14). Fortunately, in this special case, we can treat Eq. (14) in an easy way. First, we set the coefficient of the f 2 to zero, namely a[3(p0 fxy )x − fxxxxfy ] + b[3(q0 fxy )y − fyyyy fx )] − cfxxy − dfyyx − fxyt = 0. Substituting the ansatz (7) into Eq. (15) yields (qt − 3bq0 qy + bqyyy + dqy )y px + (pt − 3ap0 px + apxxx + cpx )x qy = 0. According to Eq. (16), we simply take the following variable separated equations: pt − 3apx p0 + apxxx + cpx = 0, qt − 3bqy q0 + bqyyy + dqy = 0. (17) (18) (16) (15)

Second, inserting the ansatz.(7) with Eqs. (17) and (18) into Eq. (14), we find that Eq. (14) is satisfied identically. Also it is not easy to obtain general solutions to Eqs. (17) and (18) for any fixed p0 and q0 . However, we can treat the problem in an alternate way. Because p0 and q0 are arbitrary seed solutions, we can view p and q as arbitrary functions of {x, t} and {y, t}, respectively, then the seed solutions p 0 and q0 can be fixed by Eqs. (17) and (18). Finally, we obtain a quite general excitation of the GKdV system v= 2(a1 a2 − a3 a0 )px qy , (a0 + a1 p + a2 q + a3 pq)2 2(a1 + a3 q)pxx 2(a1 + a3 q)2 p2 pt + apxxx + cpx x − , + 2 (a0 + a1 p + a2 q + a3 pq) a0 + a1 p + a2 q + a3 pq 3apx
2 2(a2 + a3 p)2 qy 2(a2 + a3 p)qyy qt + bqyyy + dqy − + , 2 (a0 + a1 p + a2 q + a3 pq) a0 + a1 p + a2 q + a3 pq 3bqy

(19)

u=

(20)

w=

(21)

with two arbitrary functions p(x, t), q(y, t), and a 0 , a1 , a2 , a3 being arbitrary constants.

VOL. 41

CHUN-LONG ZHENG

447

III. SOME COHERENT SOLITON STRUCTURES WITH CHAOTIC AND FRACTAL BEHAVIORS IN THE (2+1)-DIMENSIONAL GKDV SYSTEM

It is interesting that expression (19) is valid for many (2+1)-dimensional models such as the DS equation, the NNV system, the ANNV equation, the ADS model, the GAKNS, the dispersive long wave equation, etc. [15, 16, 19, 21]. Because of the arbitrariness of the functions p and q included in Eq. (19), the quantity v possesses quite rich structures. For instance, as mentioned in Refs. [15, 16, 19, 21], when the functions p and q are selected appropriately, we can obtain many kinds of localized solutions, like multi-solitoff solutions, multi-dromion and dromion lattice solutions, multiple ring soliton solutions, multiple peakon solutions, and so on. Since these types of localized solutions have been discussed widely for other models, we omit all the stable localized coherent soliton structures here. An important matter is whether we can find some new types of solitons which possess chaotic and/or fractal behavior, say, the chaotic and/or fractal localized excitations for the soliton system. The answer is apparently positive, since p(x, t) and q(y, t) are arbitrary functions. There are various chaotic and fractal dromion and lump excitations, because any types of (1+1)- and/or (0+1)-dimensional chaos and fractal models can be used to construct localized excitations of higher dimensional models. Some interesting possible chaotic and fractal patterns are cited here. For simplification in the following discussion, we set a0 = 0, a1 = a2 = 1, and a3 = 2 in the expression (19). III-1. Chaos (1) Chaotic dromions In (2+1)-dimensions, one of the most important nonlinear solutions is the dromion excitation, which is localized in all directions. Now we set p and q to be p = exp(x), q = 1 + (100 + f (t)) exp(y), (22)

where f (t) is an arbitrary function of time t. From the excitation (19) with equation (22) one knows that the amplitude of the dromion is determined by the function f (t). If we select the function f (t) as a solution of a chaotic system, then we can obtain a type of chaotic dromion solution. In Fig. 1a, we plot the shape of the dromion for the physical quantity v shown by expression (19) at a fixed time (for f (t) = 0) with condition (22). The amplitude A of the dromion related to (a) is changed chaotically with f (t), as depicted in Fig. 1b, where f (t) is a solution of the following Lorenz system [25] ft = −10(f − g), gt = f (60 − h) − g, 8 ht = f g − h. 3 (23)

This fact hints that the functions p or q may take a more general form. For example, if we set p and q to be [f3 (t) > 0, f7 (t) > 0] p= f1 (t) , f2 (t) + exp(f3 (t)(x + f4 (t))) q= f5 (t) , f6 (t) + exp(f7 (t)(x + f8 (t))) (24)

448

COHERENT SOLITON STRUCTURES WITH . . .

VOL. 41

Fig.1a
0.0024 0.0022 0.002 0.0018

Fig.1b

0.001 v 0.0005 0 –6 –2 –4 –6

A 0.0016 0.0014 0.0012 0.001 0.0008 15 20 25 t 30 35

–4

–2

y0

2

4

6

6

4

2

0x

FIG. 1: (a) A plot of a single dromion structure for the physical quantity v given by the solution (19) with the conditions (22) and f (t) = 0. (b) Evolution of the amplitude A of a chaotic dromion related to (a), with f (t) being a solution of the Lorenz system (23) at different times.

with fi (t), i = 1, 2, · · · , 8 being chaotic solutions, then solution (19) becomes a chaotic dromion which may be chaotic in different ways. The amplitude of the dromion (19) with Eq. (24) will be chaotic if f1 (t), f2 (t), f5 (t) and/or f6 (t) are chaotic, while f4 (t) and/or f8 (t) are chaotic if the position of the dromion location becomes chaotic. The shape (width) of the dromion may be chaotic if the functions f 3 (t) and/or f7 (t) are chaotic. Since a detailed physical discussion has been given in Ref. [26], we omit the related plots here. (2) Chaotic line solitons It is interesting that the localized excitations are not only chaotic with time t, but also with space, say, in the direction x and/or y. If either one of p and q is selected as a localized function while other one is a chaotic solution of some (1+1)-dimensional (or (0+1)dimensional) nonintegrable models, then the excitation (19) becomes a chaotic line soliton which may be chaotic in the x or y direction. For example, we take p or q as the solution of (ς = x + ωt, η = y + νt) pςςς = pςς pς + (c + 1)p2 ς − (p2 + b(c + 1))pς − (b + c + 1)pςς + (b(a − 1) − p2 )cp, (25) p

qηηη =

2 qηη qη + (γ + 1)qη − (q 2 + β(γ + 1))qη − (β + γ + 1)qηη + (β(α − 1) − q 2 )γq, (26) q

where a, b, c, α, β, γ, ω, and ν are all arbitrary constants. Actually, Eq. (25) or Eq. (26) is also equivalent to the celebrated Lorenz system pς = −c(p − g), gς = p(a − h) − g, hς = pg − bh, (27)

upon eliminating the functions g and h in Eq.(27). Fig. 2(a) is a plot of the chaotic line

VOL. 41

CHUN-LONG ZHENG

449

Fig.2a

Fig.2b 30

0.02

20 p
0.01 v 0

10 0 2 4 6 8 10 x 12 14 16 18 20

–0.01 –10 20 18 16 14 12 10 x 8 6 4 2 10 0 y

–10 –20

FIG. 2: (a). A plot of the chaotic line soliton structure for the physical quantity v given by expression (19) with the conditions (27), (28), and (29). (b) A typical plot of the chaotic solution p in the Lorenz system (27).

soliton solution expressed by Eq. (19) with the selections: p is a chaotic solution of the Lorenz system (27) and q = tanh(y). while the parameters are taken as 8 , c = γ = 1, t = 0. (29) 3 A typical plot of the chaotic solution p in the above Lorenz system is presented in Fig. 2(b). a = α = 60, b=β= (3) Non-localized chaotic patterns Certainly, if both p and q are selected as chaotic solutions of some lower dimensional nonintegrable models, then the excitation (19) becomes chaotic in both the x and y directions, which yields non-localized chaotic-chaotic patterns. Figure 3(a) shows a plot of a special chaotic-chaotic pattern expressed by Eq. (19), p and q being given by Eqs. (25) and (26), with the selections Eq. (29) and a0 = 200, a1 = a2 = 1, a3 = 0. (30) (28)

Figure 3(b) shows the corresponding plot for two typical chaotic solutions of the Lorenz model (27) with Eq. (29). If only one of p and q is selected as a chaotic solution for some lower dimensional nonintegrable models, while the other is some periodic function, then the excitation (19) becomes chaotic in one direction and periodic in the other direction, which yields a nonlocalized chaotic-periodic pattern. Figure 4(a) shows a plot of the special chaotic–periodic pattern expressed by Eq. (19), with Eq. (30) and p being the chaotic solution of Eq. (25), also a = 60, b = 8/3, c = 10, and q is the periodic solution of Eq. (26) with α = 350, β = 8/3, γ = 10. (31)

450

COHERENT SOLITON STRUCTURES WITH . . .

VOL. 41

Fig.3a

Fig.3b

2 1 v 0

–1 –2 0 20 40 x 60 60 40 y

90 80 70 h60 50 40 30 20 100 –40 80

–20

80

100

20

0 g

20

40

0f –10 –20

10

20

FIG. 3: (a) A plot of the special chaotic-chaotic pattern expressed by Eq. (19), with p and q being chaotic solutions to Eqs. (25) and (26) under the selections Eqs. (29) and (30). (b) A plot for a typical chaotic solution of the Lorenz system (27) with Eq. (29).
Fig.4a
Fig.4b

10 5 v0 –5 –10 0

20

40 x 60

80

100

0

20

y 60 40

80

420 400 380 360 h 340 320 300 100 280 –100

–50

0 g

50

100

0f –20 –40

20

40

FIG. 4: (a). A plot of the special chaotic–periodic pattern expressed by Eq. (19), with Eq. (30) and p being the chaotic solution of Eq. (25), with a = 60, b = 8/3, c = 10, while q is the periodic solution of Eq. (26) with condition Eq. (31). (b) A plot of a typical periodic Lorenz system solution (27) with Eq. (31).

Figure 4(b) is a plot of two typical periodic solutions of the Lorenz model (27) with Eq. (31). III-2. Fractals (1) Regular fractal lumps and dromions Now we discuss the localized coherent excitation with fractal properties. The lump solution (algebraically localized in all directions) is one type of significant localized excitation. In Refs. [26–28], the authors found that many lower-dimensional piecewise smooth functions with a fractal structure can be used to construct exact solutions of a higher-dimensional soliton system, which also possesses a fractal structure. This situation also occurs in a

VOL. 41

CHUN-LONG ZHENG

451

Fig.5a
0.04 0.02

Fig.5b

0.4 0.2 v0 –0.2 –0.4 3 2 1 –1 –2 –3

y0 –0.02

x0

–1

–2

–3

3

2

1

0y

–0.04 –0.04 –0.02 0 x 0.02 0.04

FIG. 5: (a) A fractal lump structure for the potential v with the conditions (32)-(33) at t = 0. (b) A density plot of the fractal lump related to (a) in the region {x = [−045, 045], y = [−045, 045]}.

(2+1)-dimensional GKdV system. If the functions p and q are selected appropriately, we can find some types of lump solutions with fractal behaviors. Figure 3(a) shows a fractal lump structure for the physical quantity v in Eq. (19), where p and q in expression (19) are simply selected as p=1+ |x − c1 t| (cos(ln(x − c1 t)2 ))2 , 1 + (x − c1 t)4 |y − c2 t| (cos(ln(y − c2 t)2 ))2 , 1 + (y − c2 t)4 (32)

q =1+

(33)

at t = 0. In figure 5(a), we can see that the solution is localized in all directions. Near the center there are infinitely many peaks which are distributed in a fractal manner. In order to investigate the fractal structure of the lump, we should look at the structure more carefully. Figure 5(b) presents a density plot of the structure of the fractal lump in the region {x = [−0.045, 0.045], y = [−0.045, 0.045]}. More detailed studies will show us the interesting self-similar structure of the lump. For example, if we reduce the region in figure 5(b) to {x = [−0.002, 0.002], y = [−0.002, 0.002]}, {x = [−0.000085, 0.000085], y = [−0.000085, 0.000085]}, {x = [−3.6 × 10 −6 , 3.6 × 10−6 ], y = [−3.6 × 10−6 , 3.6 × 10−6 ]} and so on, we will find structures totally similar to those plotted in figure 5(b). The dromion solution (exponentially localized in all directions) is another type of significant localized excitation. If the arbitrary functions p and q are appropriately selected, we find that some special types of fractal dromions for the physical quantity v (19) are revealed. For example, if we take p = 1 + exp[ q = 1 + exp[ (x − c1 t)2 (1 + sin(ln((x − c1 t)2 )))], (y − c2 t)2 (1 + sin(ln((y − c2 t)2 )))], (34) (35)

452

COHERENT SOLITON STRUCTURES WITH . . .

VOL. 41

Fig.6a
0.1 0.05

Fig.6b

0.1 v0 –0.1 –4 –2 y0 2 4 2 4

y0

–4 –2 0x
–0.05 –0.1 –0.1 –0.05 0 x 0.05 0.1

FIG. 6: (a) A plot of a fractal dromion structure for the potential v given by the solution (19), with the conditions (34)-(35), at t = 0. (b) A density plot of the fractal structure of the dromion in the region {x = [−0.12, 0.12], y = [−0.12, 0.12]}.

then we can obtain a simple fractal dromion. Figure 6(a) shows a plot of this special type of fractal dromion structure for the potential v given by Eq. (19), with the conditions in Eqs. (34)-(35) at time t = 0. Figure 6(b) is a density plot of the fractal structure of the dromion in the region {x = [−0.12, 0.12], y = [−0.12, 0.12]}. To observe the self-similar structure of the fractal dromion more clearly, one may enlarge a small region near the center of figure 6(b). For instance, if we reduce the region of figure 4(b) to {x = [−0.002, 0.002], y = [−0.002, 0.002]}, {x = [−2.5 × 10−14 , 2.5 × 10−14 ], y = [−2.5 × 10−14 , 2.5 × 10−14 ]} and so on, we find structure totally similar to that presented in figure 6(b). (2) Stochastic fractal dromions and lumps In addition to self-similar regular fractal dromions and lumps, the lower-dimensional stochastic fractal functions may also be used to construct high-dimensional stochastic fractal dromion and lump excitations. For instance, one of the best known stochastic fractal functions is the Weierstrass function:
N

Γ ≡ Γ(ξ) =
k=0

(3/2)−k/2 sin((3/2)k ξ), N → ∞,

(36)

where the independent variable ξ may be a suitable function of {x + at} and/or {y + bt}, say ξ = x + at and ξ = y + bt, as in the functions p and q, respectively, for the following selection (37). If the Weierstrass function is included in the dromion or lump excitations, then we can derive stochastic fractal dromions and lumps. Figure (7) shows a plot of a typical stochastic fractal lump solution, determined by Eq. (19), with the conditions Eq. (36) and p = Γ(x + at) + (x + at)2 + 102 , q = Γ(y + bt) + (y + bt)2 + 102 , (37)

VOL. 41

CHUN-LONG ZHENG

453

Fig.7

2 V0 –2 60

40

20

y0 –20

–40

–60

–60

–40

–20

0x

20

40

60

FIG. 7: A plot of a typical stochastic fractal lump solution determined by Eq.(19) with the selections (36) and (37).

at t = 0. In figure (7), the vertical axis denotes the quantity V , which is only a re-scaling of the potential v: v = V × 10−7 .

IV. SUMMARY AND DISCUSSION

In summary, with help of the variable separation approach, the (2+1)-dimensional GKdV system is solved. Abundant localized coherent soliton structures of the solution (19), such as multi-dromion, multi-ring, multi-lump solutions, peakons, breathers, instantons, etc., can be easily constructed by selecting appropriate arbitrary functions. In addtion to the usual localized coherent soliton structures, we find some new localized excitations–the chaotic and fractal soliton solutions for the (2+1)-dimensional GKdV system. As is known, chaos and fractals not only belong to the realms of mathematics and computer graphics, but also exist nearly everywhere in nature, such as in fluid turbulence, crystal growth patterns, human veins, fern shapes, galaxy clustering, cloud structures and in numerous other examples. Conventionally, chaos and fractals are opposite to solitons in nonlinear science, since solitons are the representatives of an integrable system while chaos and fractals represent non-integrable systems. However, in this paper, we find some chaotic and fractal structures for localized solutions of the (2+1)-dimensional integrable GKdV model. Why do localized excitations possess some kinds of chaotic and/or fractal behaviors? If one considers the boundary and/or initial conditions of the chaotic and/or fractal solutions obtained here, one can find that the initial and/or boundary conditions possess chaotic and/or fractal properties. In other words, the chaotic and/or fractal behaviors of the localized excitations of the integrable models come from the nonintegrble boundary and/or initial conditions. Because of the numerous applications of the soliton, chaos, and fractal theories in many physics fields, such as fluid dynamics, plasma physics, field theory, optics, etc., it is important to learn more about the localized chaotic and/or fractal excitations of an

454

COHERENT SOLITON STRUCTURES WITH . . .

VOL. 41

“integrable” physical model with “nonintegrable” boundary and/or initial conditions, such as chaotic and/or fractal boundaries and/or initial conditions.

Acknowledgments

The author is in debt to Professors J. F. Zhang, L. Q. Chen, Z. M. Sheng, and Doctor X. Y. Tang for their helpful discussions, and also expresses sincere thanks to Professor S. Y. Lou for his useful references. This project was supported by the Chinese National Basic Research Project for Nonlinear Science and the Natural Science Foundation of China under Grant No. 10172056.

References
Electronic address: zjclzheng@yahoo.com.cn [1] G. I. Stegemant and M. Segev, Science 286, 1518 (1999). Y. Skivshar and B. A. Melomend, Rev. Mod. Phys. 61, 765 (1989). J. P. Gollub and M. C. Cross, Nature 404, 710 (2000). H. S. Chen, J. Wang, and Y. Gu, Chin. Phys. Lett. 17, 85 (2000). R. A. Jalabert and H. M. Pastawski, Phys. Rev. Lett. 86, 2490 (2001). [2] I. Loutsenko and D. Roubtsov, Phys. Rev. Lett. 78, 3011 (1997). M. Tajiri and H. Maesono, Phys. Rev. E55, 3351 (1997). M. Gedalin, T. C. Scott, and Y. B. Band, Phys. Rev. Lett. 78, 448 (1997). [3] M. Boiti, J. J. P. Leon, M. Manna, and F. Pempinelli, Inv. Problems 2, 271 (1986). [4] L. P. Nizhnik, Sov. Phys. Dokl. 25, 706 (1980). S. P. Novikov and A. P. Veselov, Physica D 18, 267 (1986). A. P. Veselov and S. P. Novikov, Sov. Math. Dokl. 30, 88 (1984). [5] Y. Tagami, Phys.Lett. A41, 116 (1989). [6] X. B. Hu and Y. S. Li, J. Phys. A : Math. Gen. 24, 1979 (1991). [7] X. B. Hu, J. Phys. A: Math. Gen. 24, 1331 (1991). [8] Y. Ohta, J. Phys. Soc. of Japn. 61, 3928 (1992). [9] R. Radha and M. Lakshmanan, J. Math. Phys. 35, 4746 (1994). [10] S. Y. Lou, Chin. Phys. Lett. 17, 781 (2000). [11] J. F. Zhang, Chin. Phys. 11, 651 (2002). [12] B. G. Konopelchenko, J. Sidorenko, and W. Strampp, Phys. Lett. A: 175, 17(1991). Y. Cheng and Y. S. Li, Phys. Lett. A175, 22 (1991). C. W. Cao, Sci. China A33, 528 (1990). [13] S. Y. Lou and L. L. Chen , J. Math. Phys. 40, 6491 (1999). L.L. Chen, Acta Physica Sinica 48, 2149 (1999). [14] S. Y. Lou and J. Z. Lu, J. Phys. A: Math. Gen. 29, 4209 (1996). [15] S. Y. Lou, Phys. Lett. A277, 94 (2000). [16] S. Y. Lou and H. Y. Ruan, J. Phys. A: Math. Gen. 34, 305 (2001). [17] S. Y. Lou, Physica Scripta, 65, 7 (2002). [18] C. L. Zheng and J. F. Zhang, Chin. Phys. Lett. 19, 1399 (2002). J. F. Zhang , C. L. Zheng, J. P. Meng, and J. P. Fang, Chin. Phys. Lett. 20, 448 (2003). [19] C. L. Zheng and J. F. Zhang, Chin. Phys. 12, 11 (2003).

VOL. 41

CHUN-LONG ZHENG

455

C. L. Zheng, J. F. Zhang, F. M. Wu , Z. M. Sheng, and L. Q. Chen, Chin. Phys. 12, 472 (2003). [20] C. L. Zheng, W. H. Huang, and J. F. Zhang, Commun. Theor. Phys. 38 653 (2002). Zheng C. L. and J. F. Zhang, Commun. Theor. Phys. 39, 9 (2003). [21] C. L. Zheng, J. M. Zhu, J. F. Zhang, and L. Q. Chen, Commun. Theor. Phys. 39 261 (2003). [22] M. J. Ablowit, D. J. Kaup, A. C. Newell, and H. Segur, Phys. Rev. Lett. 30, 1262 (1973). [23] B. G. Konopelchenko, J. Sidorenko, and W. Strampp, Phys. Lett. A175, 17 (1991). [24] S. Y. Lou and X.B. Hu, J. Math. Phys. 38, 6401 (1997). [25] F. W. Lorenz, J. Atomos. Sci. 20, 130 (1963). M. Clerc, P. Coullet and E. Tirapegui, Phys. Rev. Lett. 83, 3820 (1999). [26] X. Y. Tang and S. Y. Lou, Chaos, Solitons and Fractals 14, 1451 (2002). X. Y. Tang, S. Y. Lou, and Y. Zhang, Phys. Rew. E66, 46601 (2002). [27] S. Y. Lou, C. L. Chen, and X. Y. Tang, J. Math. Phys 43, 4078 (2002). [28] S. Y. Lou, X. Y. Tang, and C. L. Chen, Chin. Phys. Lett. 19, 769 (2002).