Physics Letters A 373 (2009) 2501–2505

Contents lists available at ScienceDirect

Physics Letters A

www.elsevier.com/locate/pla

Exp-function method for N-soliton solutions of nonlinear evolution equations

in mathematical physics ✩
Sheng Zhang a,b,∗ , Hong-Qing Zhang a
a
School of Mathematical Sciences, Dalian University of Technology, Dalian 116024, PR China
b
Department of Mathematics, Bohai University, Jinzhou 121013, PR China

a r t i c l e i n f o a b s t r a c t

Article history: In this Letter, the Exp-function method is generalized to construct N-soliton solutions of a KdV equation
Received 3 April 2009 with variable coefficients. As a result, 1-soliton, 2-soliton and 3-soliton solutions are obtained, from which
Received in revised form 12 May 2009 the uniform formula of N-soliton solutions is derived. It is shown that the Exp-function method may
Accepted 13 May 2009
provide us with a straightforward and effective mathematical tool for generating N-soliton solutions of
Available online 15 May 2009
Communicated by R. Wu
nonlinear evolution equations in mathematical physics.
© 2009 Elsevier B.V. All rights reserved.
PACS:
02.30.Jr
02.30.Ik
05.45.Yv

Keywords:
Nonlinear evolution equations
Exp-function method
N-soliton solutions

1. Introduction

Recently, He and Wu [1] proposed a new method called the Exp-function method to seek solitary wave solutions, periodic solutions
and compact-like solutions of nonlinear evolution equations (NLEEs). The basic idea of the Exp-function method was presented in He’s
monograph [2]. The method was used by many researchers to study various NLEEs in the straightforward way or in the sub-equation
way, such as the Double sine-Gorden equation [3], Burgers’ equation [4], Maccari’s system [5], Klein–Gordon equation [6], combined
KdV–mKdV equation [7], variant Boussinesq equations [8], Broer–Kaup–Kupershmidt equations [9], variable-coefficient equations [10–12],
high-dimensional equations [13–15], and discrete equations [16–18]. It was shown in these studies that with the free parameters in the
solutions obtained by the Exp-function method one can recover some known solutions found by the most existing methods.
More recently, Marinakis [19] did very interesting work to generalize the Exp-function method for constructing N-soliton solutions of
NLEEs. Marinakis chose the famous Korteweg–de Vries (KdV) equation to illustrate the generalized work and successfully obtained the
known 2-soliton and 3-soliton solutions in a simple and straightforward way.
In the present Letter, we would like to generalize the Exp-function method for constructing N-soliton solutions of a KdV equation with
variable coefficients [20]:

ut + α (t )uu x + β(t )u xxx = 0, (1)

which is well known as a model equation describing the propagation of weakly nonlinear and weakly dispersive waves in inhomogeneous
media, here α (t ) and β(t ) are functions of t.
The rest of this Letter is organized as follows. In Section 2, we give the description of the Exp-function method for constructing
N-soliton solutions of NLEEs. In Section 3, we apply the method to Eq. (1). In Section 4, some conclusions and discussions are given.

✩
This work was supported by the Natural Key Basic Research Project of China under Grant No. 2004CB318000.
* Corresponding author at: School of Mathematical Sciences, Dalian University of Technology, Dalian 116024, PR China. Tel.: +86 411 84708350.
E-mail address: zhshaeng@yahoo.com.cn (S. Zhang).

0375-9601/$ – see front matter © 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.physleta.2009.05.019
2502 S. Zhang, H.-Q. Zhang / Physics Letters A 373 (2009) 2501–2505

2. Basic idea of the Exp-function method for N -soliton solutions of NLEEs

For a given NLEE, say, in two variables x and t:

P (u , ut , u x , utx , utt , u xx , . . .) = 0, (2)

the Exp-function method [1] is based on the assumption that its solutions can be expressed as:
 p1 i 1 ξ1
a e
i =0 i 1
u (x, t ) = q11 , ξ1 = k1 x + c 1 t + w 1 , (3)
j 1 =0
b j 1 e j 1 ξ1

where ai 1 , b j 1 , k1 , c 1 and w 1 are unknown constants, the values of p 1 and q1 can be determined by balancing the linear term of highest
order in Eq. (2) with the highest order nonlinear term. As we know that Eq. (3) can be used to construct 1-soliton solution.
In order to seek N-soliton solutions for integer N > 1, we generalize Eq. (3) to the following form:
 p1  p2 p N
i 1 =0
· · · i NN=0 ai 1 i 2 ···i N e g =1 i g ξ g
i 2 =0
u (x, t ) =  q2 q N , ξg = k g x + c g t + w g , (4)
q1
j 1 =0 j 2 =0
· · · j NN =0 b j 1 j 2 ··· j N e g =1 j g ξ g
given the value of N = 2, it becomes:
 p1  p2 2
a e g =1 i g ξ g
i 1 =0 i 2 =0 i 1 i 2
u (x, t ) =  q2 2 , (5)
q1 g =1 j g ξg
j 1 =0 j 2 =0
b j1 j2 e

which can be used to construct 2-soliton solution.

When N = 3, Eq. (4) changes into:
 p1  p2  p3 3
a e g =1 i g ξ g
i 1 =0 i 2 =0 i 3 =0 i 1 i 2 i 3
u (x, t ) =  q2 q3 3 , (6)
q1 g =1 j g ξg
j 1 =0 j 2 =0 j 3 =0
b j1 j2 j3 e

which can be used to obtain 3-soliton solution.

Substituting Eq. (5) into Eq. (2), and using Mathematica, then equating to zero each coefficient of the same order power of the ex-
ponential functions yields a set of equations. Solving the set of equations, we can determine the 2-soliton solution, and the following
3-soliton solution by means of Eq. (6), provide they exist. If possible, we may conclude with the uniform formula of N-soliton solutions
for any N  1.

3. Applications to the KdV equation with variable coefficients

To begin with, we suppose that Eq. (1) admits the 1-soliton solution of the form:

a1 eξ1
u (x, t ) = , (7)
(1 + b1 eξ1 )2
where ξ1 = k1 x + s1 (t ) + w 1 . Obviously, Eq. (7) is embedded in the same form as Eq. (3). Substituting Eq. (7) into Eq. (1), and using
Mathematica, then equating to zero each coefficient of the same order power of the exponential functions yields a set of equations as
follows:
     
a1 k31 β(t ) + s1 (t ) = 0, a1 a1 k1 α (t ) − 11b1 k31 β(t ) + b1 s1 (t ) = 0, −a1 b1 a1 k1 α (t ) − 11b1 k31 β(t ) + b1 s1 (t ) = 0,
 
−a1 b31 k31 β(t ) + s1 (t ) = 0.
Note that the velocity of the soliton will exist provided the time-dependent coefficients are Riemann integrable. Under this condition,
which is always supposed to be true in the similar circumstances thereafter, we have

a1 = 12b1 k21 δ, s1 (t ) = −k31 δ α (t ) dt , β(t ) = δ α (t ), δ = constant. (8)

We, therefore, obtain the 1-soliton solution of Eq. (1):

12b1 k21 δ eξ1

u (x, t ) = , (9)
(1 + b1 eξ1 )2

where ξ1 = k1 x − k31 δ α (t ) dt + w 1 , b1 , k1 and w 1 are free constants, under the constraint β(t ) = δ α (t ). A spatial structure of the 1-soliton
solution (9) is plotted in Fig. 1. Solution (9) is a special case of that found in [10].
Next, we suppose that Eq. (1) has the 2-soliton solution in the form:

a10 eξ1 + a01 eξ2 + a11 eξ1 +ξ2 + a21 e2ξ1 +ξ2 + a12 eξ1 +2ξ2
u (x, t ) = , (10)
(1 + b1 eξ1 + b2 eξ2 + b3 eξ1 +ξ2 )2
where ξi = ki x + si (t ) + w i , i = 1, 2. Clearly, Eq. (10) possesses the same form as Eq. (5). Substituting Eq. (10) into Eq. (1), and using the
similar manipulations as illustrated above, we gain:
 S. Zhang, H.-Q. Zhang / Physics Letters A 373 (2009) 2501–2505 2503

Fig. 1. Parameters: b1 = 1, k1 = 1, δ = 0.5, w 1 = 0, α (t ) = sn(t , 0.5) + tanh(t ), Fig. 2. Parameters: b1 = 1, b2 = 2, k1 = 1, k2 = 1.6, δ = 0.5, w 1 = 0, w 2 = 0,
here sn(t , 0.5) is the Jacobi’s elliptic function with the modulus m = 0.5. α (t ) = sn(t , 0.5) + tanh(t ), here sn(t , 0.5) is the Jacobi’s elliptic function with the
modulus m = 0.5.

a10 = 12b1 k21 δ, a01 = 12b2 k22 δ, a11 = 24b1 b2 δ(k1 − k2 )2 , (11)
− k2 )
12b21 b2 k22 δ(k1 2
12b1 b22 k21 δ(k1 − k2 ) 2
b1 b2 (k1 − k2 ) 2
a21 = , a12 = , b3 = , (12)
(k1 + k2 )2 (k1 + k2 )2 (k1 + k2 )2

si (t ) = −k3i δ α (t ) dt , β(t ) = δ α (t ), δ = constant, i = 1, 2. (13)

Thus, with the help of Eqs. (11)–(13) we determine the 2-soliton solution (10), where b1 , b2 , k1 , k2 , w 1 and w 2 are free constants. Fig. 2
shows a spatial structure of the 2-soliton solution (10). Setting α (t ) = −6 and δ = − 16 , we obtain β(t ) = 1, then solution (10) becomes
that of the famous KdV equation in [19].
We now suppose that the 3-soliton solution of Eq. (1) can be expressed as follows:
f 1 (ξ1 , ξ2 , ξ3 )
u (x, t ) = , (14)
f 22 (ξ1 , ξ2 , ξ3 )

where ξi = ki x + si (t ) + w i , i = 1, 2, 3, and

f 1 (ξ1 , ξ2 , ξ3 ) = a100 eξ1 + a010 eξ2 + a001 eξ3 + a110 eξ1 +ξ2 + a101 eξ1 +ξ3 + a011 eξ2 +ξ3 + a210 e2ξ1 +ξ2 + a120 eξ1 +2ξ2 + a201 e2ξ1 +ξ3 + a102 eξ1 +2ξ3
+ a021 e2ξ2 +ξ3 + a012 eξ2 +2ξ3 + a111 eξ1 +ξ2 +ξ3 + a211 e2ξ1 +ξ2 +ξ3 + a121 eξ1 +2ξ2 +ξ3
+ a112 eξ1 +ξ2 +2ξ3 + a221 e2ξ1 +2ξ2 +ξ3 + a212 e2ξ1 +ξ2 +2ξ3 + a122 eξ1 +2ξ2 +2ξ3 ,
f 2 (ξ1 , ξ2 , ξ3 ) = 1 + b1 eξ1 + b2 eξ2 + b3 eξ3 + b4 eξ1 +ξ2 + b5 eξ1 +ξ3 + b6 eξ2 +ξ3 + b7 eξ1 +ξ2 +ξ3 .

It is easy to see that Eq. (14) has the same form as Eq. (6). By the similar manipulations mentioned above, we get

a100 = 12b1 k21 δ, a010 = 12b2 k22 δ, a001 = 12b3 k23 δ, a110 = 24b1 b2 δ(k1 − k2 )2 , (15)
12b21 b2 k22 δ(k1 − k2 )2
a101 = 24b1 b3 δ(k1 − k3 )2 , a011 = 24b2 b3 δ(k2 − k3 )2 , a210 = , (16)
(k1 + k2 )2
12b1 b22 k21 δ(k1 − k2 )2 12b21 b3 k23 δ(k1 − k3 )2 12b1 b23 k21 δ(k1 − k3 )2
a120 = , a201 = , a102 = , (17)
(k1 + k2 )2 (k1 + k3 )2 (k1 + k3 )2
12b22 b3 k23 δ(k2 − k3 )2 12b2 b23 k22 δ(k2 − k3 )2 12b1 b23 k21 δ(k1 − k3 )2
a021 = , a012 = , a102 = , (18)
(k2 + k3 )2 (k2 + k3 )2 (k1 + k3 )2
48b1 b2 b3 δ[k22 k23 (k22 − k23 )2 + k61 (k22 + k23 ) − 2k41 (k42 + k43 ) + k21 (k62 + k63 )]
a111 = , (19)
(k1 + k2 )2 (k1 + k3 )2 (k2 + k3 )2
24b21 b2 b3 δ(k1 − k2 )2 (k1 − k3 )2 (k2 − k3 )2 b1 b2 (k1 − k2 )2
a211 = , b4 = , (20)
(k1 + k2 )2 (k 1 + k3 )2 (k1 + k2 )2
24b1 b2 b23 δ(k1 − k2 )2 (k1 − k3 )2 (k2 − k3 )2 b1 b3 (k1 − k3 )2
a112 = , b5 = , (21)
(k1 + k3 )2 (k2 + k3 )2 (k1 + k3 )2
12b21 b22 b3 k23 δ(k1 − k2 )4 (k1 − k3 )2 (k2 − k3 )2 b2 b3 (k2 − k3 )2
a221 = , b6 = , (22)
(k1 + k2 )4 (k 1 + k3 )2 (k 2 + k3 )2 (k2 + k3 )2
2504 S. Zhang, H.-Q. Zhang / Physics Letters A 373 (2009) 2501–2505

Fig. 3. Parameters: b1 = 1, b2 = 2, b3 = 3, k1 = 1, k2 = 1.6, k3 = 2.2, δ = 0.5, w 1 = 0, w 2 = 0, w 3 = 0, α (t ) = sn(t , 0.5) + tanh(t ), here sn(t , 0.5) is the Jacobi’s elliptic function
with the modulus m = 0.5.

12b21 b2 b23 k22 δ(k1 − k2 )2 (k1 − k3 )4 (k2 − k3 )2 b1 b2 b3 (k1 − k2 )2 (k1 − k3 )2 (k2 − k3 )2

a212 = , b7 = , (23)
(k1 + k2 )2 (k 1 + k3 )4 (k 2 + k3 )2 (k1 + k2 )2 (k1 + k3 )2 (k2 + k3 )2

12b1 b22 b23 k21 δ(k1 − k2 )2 (k1 − k3 )2 (k2 − k3 )4
a122 = , si (t ) = −k3i δ α (t ) dt , β(t ) = δ α (t ), (24)
(k1 + k2 )2 (k1 + k3 )2 (k2 + k3 )4
24b1 b22 b3 δ(k1 − k2 )2 (k1 − k3 )2 (k2 − k3 )2
a121 = , δ = constant, i = 1, 2, 3. (25)
(k1 + k2 )2 (k1 + k3 )2
Employing Eqs. (15)–(25), we finally determine the 3-soliton solution (14), where b1 , b2 , b3 , k1 , k2 , k3 , w 1 , w 2 and w 3 are free constants.
In Fig. 3, a spatial structure of the 3-soliton solution (14) is shown. If we set α (t ) = −6 and δ = − 16 , solution (14) becomes that of the
famous KdV equation in [19].
If we continue to construct the N-solition solutions for any N  4, the following similar manipulations becomes rather complicated
since equating to zero the coefficients of the exponential functions implies a highly nonlinear system as pointed out in [19]. Fortunately,
we can find a formula of the N-solition solutions by analyzing the obtained solutions (9), (10) and (14). With this purpose, we re-write
solutions (9), (10) and (14) in an alternative form:
  
u (x, t ) = 12δ ln 1 + b1 eξ1 xx
, (26)

where ξ1 = k1 x − k31 δ α (t ) dt + w 1 .
  
u (x, t ) = 12δ ln 1 + b1 eξ1 + b2 eξ2 + b1 b2 eξ1 +ξ2 + B 12 xx
, (27)
 (k1 −k2 )2
where ξi = ki x − k3i δ α (t ) dt + w i (i = 1, 2), e B 12 = (k1 +k2 )2
.

 
u (x, t ) = 12δ ln 1 + b1 eξ1 + b2 eξ2 + b3 eξ3 + b1 b2 eξ1 +ξ2 + B 12 + b1 b3 eξ1 +ξ3 + B 13

+ b2 b3 eξ2 +ξ3 + B 23 + b1 b2 b3 eξ1 +ξ2 +ξ3 + B 12 + B 13 + B 23 xx
, (28)

 (ki −k j )2
where ξi = ki x − k3i δ α (t ) dt + w i (i = 1, 2, 3), e B i j = (ki +k j )2
(1  i < j  3).
Along this way, the uniform formula of N-soliton solutions can be constructed as follows:


 
N N 
i =1 μi ξi +
μi 1 i < j  N μ i μ j B i j
u (x, t ) = 12δ ln bi e , (29)
μ=0,1 i =1 xx
 
where ξi = ki x − k3i δ α (t ) dt + w i (i = 1, 2, . . . , N ), the sum μ=0,1 refers to all combinations of each μi = 0, 1 for i = 1, 2, . . . , N, and
(ki −k j )2
e Bij
= (ki +k j )2
(1  i < j  N ). If we set α (t ) = −6 and δ = − 16 , solution (29) gives the N-soliton solutions of the famous KdV equation
in [19], which was not obtained there.

Remark 1. Solutions (9), (10) and (14) obtained above have been checked with Mathematica by putting them back into the original Eq. (1).
According to the Hirota’s bilinear method [21], solution (29) can also be verified.
 S. Zhang, H.-Q. Zhang / Physics Letters A 373 (2009) 2501–2505 2505

4. Conclusions and discussions

In this Letter, 1-soliton, 2-soliton and 3-soliton solutions of the KdV equation with variable coefficients have successfully been ob-
tained, from which the uniform formula of N-soliton solutions is derived. This is due to the generalization of the Exp-function method.
Figs. 1–3 imply that these obtained solutions have rich local structures. It may be important to explain some physical phenomena. Though
the N-soliton solutions (29) can be constructed by the Hirota’s bilinear method [21], the proposed method with the help of Mathematica
for generating 1-soliton, 2-soliton and 3-soliton solutions (9), (10) and (14) is more simple and straightforward, without employing the
bilinear operator defined in the Hirota’s bilinear method. The Letter shows that the Exp-function method may provide us with a straight-
forward and effective mathematical tool for generating N-soliton solutions or testing its existence and can be extended to other NLEEs in
mathematical physics.

Acknowledgements

We express our sincere thanks to the editors and the referees for their valuable suggestions and comments. This work was supported
by the Natural Key Basic Research Project of China under Grant No. 2004CB318000 and the “Mathematics + X” Key Project of Dalian
University of Technology.

References

[1] J.H. He, X.H. Wu, Chaos Solitons Fractals 30 (2006) 700.
[2] J.H. He, Non-Perturbative Methods for Strongly Nonlinear Problems, Dissertation. de-Verlag im Internet GmbH, Berlin, 2006.
[3] J.H. He, M.A. Abdou, Chaos Solitons Fractals 34 (2007) 1421.
[4] A. Ebaid, Phys. Lett. A 365 (2007) 213.
[5] S. Zhang, Phys. Lett. A 371 (2007) 65.
[6] A. Bekir, A. Boz, Phys. Lett. A 372 (2008) 1619.
[7] X.H. Wu, J.H. He, Comput. Math. Appl. 54 (2007) 966.
[8] X.H. Wu, J.H. He, Chaos Solitons Fractals 38 (2008) 903.
[9] S. Zhang, Phys. Lett. A 372 (2008) 1873.
[10] S. Zhang, Phys. Lett. A 365 (2007) 448.
[11] S.A. El-Wakil, M.A. Madkour, M.A. Abdou, Phys. Lett. A 369 (2007) 62.
[12] S. Zhang, Nonlinear Dyn. 52 (2008) 11.
[13] S. Zhang, Chaos Solitons Fractals 38 (2008) 270.
[14] E. Yusufoglu, Phys. Lett. A 372 (2008) 442.
[15] A. Boz, A. Bekir, Comput. Math. Appl. 56 (2008) 1451.
[16] S.D. Zhu, Int. J. Nonlinear Sci. Numer. Simul. 8 (2007) 461.
[17] S.D. Zhu, Int. J. Nonlinear Sci. Numer. Simul. 8 (2007) 465.
[18] S. Zhang, Appl. Math. Comput. 199 (2008) 242.
[19] V. Marinakis, Z. Naturforsch. 60a (2008) 53.
[20] S.S. Liu, Z.T. Fu, S.D. Liu, Q. Zhao, Acta Phys. Sin. 51 (2002) 1923.
[21] D.Y. Chen, Soliton, Introduction, Science Press, Beijing, 2006.