Applied Mathematics and Computation 218 (2011) 13711378

Contents lists available at ScienceDirect

Applied Mathematics and Computation

journal homepage: www.elsevier.com/locate/amc

Adomian decomposition and Pad approximate for solving differential-difference equation

Zhen Wang a,, Li Zou b,d, Zhi Zong c,d
a

School of Mathematical Science, Dalian University of Technology, Dalian 116085, China School of Aeronautics and Astronautics, Dalian University of Technology, Dalian 116085, China c School of Naval Architecture, Dalian University of Technology, Dalian 116085, China d The State Key Laboratory of Structure Analysis for Industrial Equipment, Dalian 116085, China
b

a r t i c l e

i n f o

a b s t r a c t
In this paper, we apply the Adomian decomposition method and Pad-approximate to solving the differential-difference equations (DDEs) for the rst time. A simple but typical example is used to illustrate the validity and the great potential of the Adomian decomposition method (ADM) in solving DDEs. Comparisons are made between the results of the proposed method and exact solutions. The results show that ADM is an attractive method in solving the differential-difference equations. 2011 Elsevier Inc. All rights reserved.

Keywords: Differential-difference equation Adomian decomposition method Pad-ADM approximation Volterra equation Discrete mKdV equation

1. Introduction The DDEs play an important role in modelling complicated physical phenomena such as particle vibrations in lattices, currents ow in electrical networks, and pulses in biological chains. The solutions of these DDEs can provide numerical simulations of nonlinear partial differential equations, queuing problems, and discretizations in solid state and quantum physics. Since the works of Fermi et al. in the 1950s [1], there were quite a number of research works developed during the last decades on DDEs. For instances, Levi and his co-workers analyzed the condition for existence of higher symmetries for a class of DDEs [2,3], Yamilov and his co-workers [4,5] made outstanding contribution to the classication of DDEs, integrability tests and connections between integrable partial differential equation (PDEs) and DDEs [6,7], and a lot of works were developed to analyze the properties of solutions of DDEs [811]. A wealth of information about integrable DDEs can be found in papers by Suris [12,13] and also in his book [14]. A method named HAM is also extended to deal DDE [15,16]. In the beginning of the 1980, a so-called ADM [1722] has been used to solve effectively, easily, and accurately a large class of linear and nonlinear equations, solutions partial, deterministic or stochastic differential equations with approximation which converge rapidly. Wu et al. [23] extended ADM to solve DDEs, and obtained good accuracy with the analytical solution. In this paper, by introducing Pad approximation we successfully use ADM method to solving NDDEs. We give a better method to enhance the convergence through applying Pad-approximate. In order to give a clear view to our study, we have chosen two examples of special properties to illustrate ADM method and Pad-approximation. This paper organized as follows. In Section 2, we extend the ADM to solving nonlinear DDEs with initial condition, an brief outline of the methodology is presented. In Section 3, consists of main results of the paper, in which ADM of the DDEs has been developed. Two simple but typical examples are used to illustrate the validity and the great potential of the Adomian decomposition method in solving DDEs. Further discussions and conclusions are presented in Section 4.

Corresponding author.
E-mail address: [email protected] (Z. Wang). 0096-3003/$ - see front matter 2011 Elsevier Inc. All rights reserved. doi:10.1016/j.amc.2011.06.019

1372

Z. Wang et al. / Applied Mathematics and Computation 218 (2011) 13711378

2. Analysis of the methodology 2.1. ADM for DDEs For the purpose of illustration of the methodology to the proposed method, using ADM, we begin by considering the nonlinear differential-difference equation. Let us consider the differential-difference equation in following general form

Lun Run ; un1 ; un1 ; . . . N un ; un1 ; un1 ; . . . g ;

with prescribed conditions, where un is the unknown function, L is the highest order derivative which is assumed to be easily invertible, R is a linear differential operator of less order than L; N un ; un1 ; un1 ; . . . represents the nonlinear term and g n is the source term. Assuming the inverse operator L1 exists and it can be taken as the dene integral with respect to t from t 0 to t, i.e.,

L1

dt :

t0

Applying L1 to both the sides of (1) and using the initial conditions, we get

un f L1 Run ; un1 ; un1 ; . . . N un ; un1 ; un1 ; . . .;

where the function f x represents the term arising from integrating the source term g and from using the given initial or boundary conditions, all are assumed to be prescribed. In the spite of Adomian decomposition, we assumes a series that the unknown function un t can be expressed by an innite series in the form of

un t

1 X m0

un;m t:

The nonlinear operator N un ; un1 ; un1 ; . . . can be decomposed by an innite series of polynomials given by

N un ; un1 ; un1 ; . . .

1 X m0

An;m un;0 ; . . . ; un;m ; un1;0 ; . . . ; un1;m ; un1;0 ; . . . ; un1;m ; . . .;

where An;m are the appropriate Adomians polynomials which depend upon un;0 ; . . . ; un;m ; un1;0 ; . . . ; un1;m ; un1;0 ; . . . ; un1;m ; . . . In order to determine the Adomian polynomials, we introduce a parameter k and (5) becomes

1 X m0

un;m tk ;

1 X m0

un1;m t k ;
then

1 X m0

! un1;m t k ; . . .
m

1 X m0

An;m km :

Let un;k

P1

m m0 un;m t k ,

An;m

! ! m 1 d 1 d m N k un; ; un1 ; un1 ; . . . m N un;k ; un1;k ; un1;k ; . . . m! dk m! dk k 0 k0 " !# m 1 1 1 X X X 1 d N un;m tkm ; un1;m t km ; un1;m t km ; . . . : m! dkm m0 m0 m0
m k 0

This formula is easy to compute by using Maple software and by setting a computer code to get as many polynomials as we need in the calculation of the numerical as well as explicit solutions. If the zeros component un 0 is given, then the remaining components where n > 1 can be determined by using recurrence relation one by one

un;0 f ; un;m1 L1 Run;m ; un1;m ; un1;m ; . . . An;m m > 0:

8 9

For numerical computation, computer cannot deal with innite operations for power series in terms of un;i for i from zero to innity, so we must truncated this series, i.e., remains only rst nite terms for computation, we use

un;k t

k X m0

un;m

10

to denotes the sum of rst n term approximation to un t .

Z. Wang et al. / Applied Mathematics and Computation 218 (2011) 13711378

1373

2.2. Pad approximation Pad-approximate extrapolation technique consists in approximating a truncated series by a rational function. In general, the latter function has the advantage of extending the range of validity of the initial polynomial. Pad- approximates have thus far been successfully applied to statistical mechanics [24] and more recently to other physical problems [25,26]. According to this technique, the truncated series is approximated by:
N X n0

a n zn %

b0 b1 z bNK zNK : 1 c1 z c2 z2 cK zK

11

It is essential is that replacing rational polynomial by polynomial to approximate the solution of nonlinear DDEs. It is easy to see that the limit of polynomial at innity is innity. But a important property of soliton is that the limit of soliton at innity is either zero or a constant. This is a conict. So we should nd another expression to replace the obtained series solution with the property of soliton. Actually, Pad-approximation just satises this property by converting the polynomial to rational polynomial. The two sets of coefcients bi and cj are determined by imposing the condition that the Taylor series expansion around z 0 of the rational function matches up to order N the original truncated series. Order K may vary from 0 to N. For N an even integer, and K N the rational function is called the diagonal Pad-approximate. 2 We can easily see that Pad-approximation convert the truncated power series into rational polynomial form. The truncated power series trends to innity, but this does not obey the asymptotic behavior of solutions to DDEs, in order to make sure that the constructed solution of DDE and the soliton solution have the same asymptotic behavior, we should introduce Pad-approximation in form of (11). Obviously that Eq. (11) has a nite limit as jxj trends to innity when K N K , it have zero limit as jxj trends to innity when K < N K , this feature coincide with the soliton solution. 3. Application 3.1. Volterra equation To verify the validity and the potential of ADM in solving differential-difference equations, we apply it to Volterra equation

u0nt un un1 un1 ;

with the initial condition

12

un 0 n;
whose exact solution can be written as

13

un t

n : 1 2t

14

This exact solution is easy to verify. According to the solution of ADM as mentioned in above section, Eq. (12) can be written in an operator form as

Lun Mun ; un1 Nun ; un1 ;

@ where L @ . t Operating with L1 on the both sides of Eq. (15) and using the initial condition (13) gives

15

un t un 0 L1 M un ; un1 Nun ; un1 ;

16

where M un ; un1 un un1 and N un ; un1 un un1 are the nonlinear terms. We could assume the expressions of un t in the decomposition form as follows

un
i.e.,

1 X m0

un;m ;

17

unk unk

1 X m0 1 X m0

unk;m ; k 2 N; unk;m ; k 2 N:

18 19

1374

Z. Wang et al. / Applied Mathematics and Computation 218 (2011) 13711378

M un ; un1 un un1 and N un ; un1 un un1 can be expressed in terms of general Adomian polynomial (5) as follows:

Mun ; un1 Nun ; un1

1 X m0 1 X m0

Am ;

20 21

Bm :

According to Eq. (7), we could get the rst four components of Adomian polynomial as follows

A0 un;0 un1;0 ; B0 un;0 un1;0 ; A1 un;1 un1;0 un;0 un1;1 ; B1 un;1 un1;0 un;0 un1;1 ; A2 un;2 un1;0 un;1 un1;1 un;0 un1;2 ; B2 un;2 un1;0 un;1 un1;1 un;0 un1;2 ; A3 un;3 un1;0 un;2 un1;1 un;1 un1;2 un;0 un1;3 ; A3 un;3 un1;0 un;2 un1;1 un;1 un1;2 un;0 un1;3 :

22 23 24 25 26 27 28 29

Identifying the zeroth components of un;0 , the remaining components un;m ; m > 1 can be determined by using recursive relation given by

un;0 n; un;m1 L1 M un ; un1 Nun ; un1 L1 Am Bm :

30 31

The remaining components un;m ; m > 1 can be completely determined such that each terms are determined by using the previous terms, and the series solutions thus entirely evaluated. We have the following solution

un;1 2nt ; un;4 16nt 4 ;

un;2 4nt 2 ;

un;3 8nt3 ; un;6 64nt 6 ;

32 33

un;5 32nt 5 ;

So, the solution to Volterra equation can be written as

un t n 2nt 4nt 2 8nt 3 16nt 4 32nt 5 64nt 6 ;

in other word, the truncated series solution can be written as

34

/n;6 t

6 X m0

un;m t n 2nt 4nt 2 8nt 3 16nt4 32nt 5 64nt6 :

Employing [1, 1] Pad-approximation to /n;6 t , we get

un t

n : 1 2t

35

Obviously, this is an exact solution of Volttera equation with the initial condition un 0 n. This simple example indicates that the ADM is effective for solving nonlinear differential-difference equation. In the following section, we will give a complex example which subject to an initial condition which admit a soliton solution. 3.2. Discrete mKdV equation We consider another differential-difference equation, discrete mKdV equation

u0nt a u2 n un1 un1 ;

with the initial condition

36

un 0

  p K 1 coshK sechKn 2; a tanh 2       p K K 1 coshK sech Kn 2a tanh 1 coshK t 2 : a tanh 2 2

37

whose exact solution can be written as [27]

un t

38

Z. Wang et al. / Applied Mathematics and Computation 218 (2011) 13711378

1375

According to ADM as mentioned in above section, Eq. (36) can be written in an operator form as

Lun aun1 aun1 M un ; un1 Nun ; un1 ;

@ where L @ . t Operating with L1 on the both sides of Eq. (39) and using the initial conditions (37) gives

39

un t un 0 L1 aun1 aun1 M un ; un1 Nun ; un1 ;

where M un ; un1 and N un ; un1 decomposition form as follows u2 n un1 u2 n un1

40

are the nonlinear terms. We assume the expressions of un t in the

un
i.e.,

1 X m0

un;m ;

41

unk unk

1 X m0 1 X m0

unk;m ; k 2 N; unk;m ; k 2 N:

42 43

2 M un ; un1 u2 n un1 and N un ; un1 un un1 can be expressed in terms of general Adomian polynomial (5) as follows:

Mun ; un1 Nun ; un1

1 X m0 1 X m0

Am ;

44 45

Bm :

According to Eq. (7), we could get the rst four components of Adomian polynomial as follows

A0 u2 n;0 un1;0 ; B0 u2 n;0 un1;0 ; A1 2un;0 un1;0 un;1 u2 n;0 un1;1 ; B1 2un;0 un1;0 un;1 u2 n;0 un1;1 ;
2 A2 u2 n;1 un1;0 2un;0 un1;1 un;1 2un;0 un1;0 un;2 un;0 un1;2 ;

46 47 48 49 50 51 52 53 u2 n;0 un1;3 : u2 n;0 un1;2 ;

B2

u2 n;1 un1;0

2un;0 un1;1 un;1 2un;0 un1;0 un;2 u2 n;1 un1;1

2 A3 2un;1 un1;0 un;2 u2 n;1 un1;1 2un;0 un1;2 un;1 2un;0 un1;1 un;2 2un;0 un1;0 un;3 un;0 un1;3 ;

A3 2un;1 un1;0 un;2

2un;0 un1;2 un;1 2un;0 un1;1 un;2 2; un;0 un1;0 un;3

Identifying the zeroth components of un;0 , the remaining components un;m ; m > 1 can be determined by using recursive relation given by

un;0 un 0 un1;0 un1;0

  p K 1 coshK sechKn 2; a tanh 2   p K 1 coshK sechK n 1 2; un 0 a tanh 2   p K 1 coshK sechK n 1 2; un 0 a tanh 2

54 55 56 57

un;m1 L1 aun1;m aun1;m M un ; un1 Nun ; un1 L1 aun1;m aun1;m Am Bm :

The remaining components un;m ; m > 1 can be completely determined such that each terms are determined by using the previous terms, and the series solutions thus entirely evaluated. The explicit form of un;1 ; un;2 and un;3 of the discrete mKdV equation could be calculate with the aid of computer language such Maple and Mathematica. As the terrible large of the expressions of un;1 ; un;2 and un;3 , we omit them for simple. If we select a 4 and K 0:3, we shows the gures of solution of mKdV equation at n 3; n 6 and n 10. Figs. 13 show the 5th order of approximation of un t and [2,4] AdomianPad approximation, comparing with the exact solution, in case of a 4; K 0:3, respectively when n 3; n 6 and n 10. In these three cases, the 5th order of approximation of un t is agrees well with the exact solution in internal [-1, 1]. When coupled with the Pad technique, it is seen that un t can be well approximated by the analytic solutions given by ADM.

1376

Z. Wang et al. / Applied Mathematics and Computation 218 (2011) 13711378

0.7 0.6 0.5 0.4 u 0.3 0.2 0.1 4 2 0 2 t exact solution [2,4] PADM approximate 5th order approximation
Fig. 1. Comparison of the exact solution with the 5th order ADM solution and [2,4] Pad-ADM solution for n 3; a 4 and K 0:3.

0.7 0.6 0.5 0.4 u 0.3 0.2 0.1 4 2 0 2 t 5th order approximation [2,4] PADM approximate exact solution
Fig. 2. Comparison of the exact solution with the 5th order ADM solution and [2,4] Pad-ADM solution for n 6; a 4 and K 0:3.

When n 3, we get 5th-order approximation solution as

/3;6 t 0:3650187680 0:711839736t 0:3050185398t2 0:8133988949t 3 1:653954216t 4 1:059129072t5 1:127364659t6 :

Employed Pad approximation, /3;6 changes into

58

P/3;6 t

0:3650187680 0:09737333183t 0:03423312949t 2 : 1:0 1:683383042t 2:353433157t2 0:9544860142t 3 0:6747396968t 4

59

When n 6, we get 5th-order approximation solution as

/6;6 t 0:5970595389 0:287089108t 1:633696998t 2 1:353491259t 3 3:457381822t 4 4:564761114t5 6:443604825t 6 : 60

Z. Wang et al. / Applied Mathematics and Computation 218 (2011) 13711378

1377

0.7 0.6 0.5 0.4 u 0.3 0.2 0.1 4 2 0 2 t exact solution 5th order approximation [2,4] PADM approximate
Fig. 3. Comparison of the exact solution with the 5th order ADM solution and [2,4] Pad-ADM solution for n 10; a 4 and K 0:3.

Employed Pad approximation, /6;6 changes into

P/6;6 t

0:5970595387 0:01975491075t 0:1162163845t2 : 1 0:4477513209t 2:756886095t 2 0:2838422095t 3 0:8742766996t 4

61

When n 10, we get 5th-order approximation solution as

/10;6 t 0:3946913551 0:732297318t 0:1874559866t2 1:100900947t3 1:834161686t4 0:6519439036t 5 2:221864030t 6 :

Employed Pad approximation, /10;6 changes into

62

P/10;6 t

0:3946913553 0:09108508980t 0:04428247296t 2 : 1:0 1:624591519t 2:427075081t 2 0:9422557918t3 0:7111606928t 4

63

4. Discussion and conclusions This present analysis exhibits the applicability of the decomposition method to solve nonlinear differential-difference equation. In this work we demonstrate that this method is well suited to solve nonlinear differential-difference equation. The decomposition method is straightforward, without restrictive assumptions, and the components of the series solution can be easily computed using any mathematical symbolic package. Moreover, this method does not change the problem into a convenient one for the use of linear theory. It, therefore, provides more realistic series solutions that generally converge very rapidly in real physical problems. When solutions are computed numerically, the rapid convergence is obvious. In this paper, we successfully employ the Adomian decomposition method to solve nonlinear DDEs. As an illustrative example, we apply Adomian decomposition method to discrete Volttera equation and discrete mKdV equation. We get the exact solutions of Volterra equation after Pad-approximation of 6th-order approximation solution by generalized Adomian decomposition method. In other hand, 5th-order approximation solution of discrete mKdV equation was given. Using Pad-approximation, we get a soliton solution of discrete mKdV equation which agrees well with the exact solution. Since the numerous DDEs in many elds of scientic and the hardness of solving them, this DDEsADM will become a much more interesting method to solving nonlinear DDEs in science and engineering. Acknowledgements This work is Supported by Natural Sciences Foundation of China under the Grant 11026165, 50909017, 50921001 and 11072053, Doctoral Fund of Ministry of Education of China under the Grant 20100041120037 and the Fundamental Research Funds for the Central Universities.

1378

Z. Wang et al. / Applied Mathematics and Computation 218 (2011) 13711378

References
[1] E. Fermi, J. Pasta, S. Ulam, Collected Papers of Enrico Fermi, Chicago Press, Chicago, IL, 1965. pp. 978. [2] D. Levi, O. Ragnisco, Extension of the spectral transform method for solving nonlinear differential- difference equations, Lett. Nuovo. Cimento 22 (1978) 691. [3] D. Levi, RI. Yamilov, Conditions for the existence of higher symmetries of evolutionary equations on a lattice, J. Math. Phys. 38 (1997) 6648. [4] R.I. Yamilov, Construction scheme for discrete Miura transformations, J. Phys. A: Gen. 27 (1994) 6839. [5] V.E. Adler, S.I. Svinolupov, R.L. Yamilov, Multi-component Volterra and Toda type integrable equations, Phys. Lett. A 254 (1999) 24. [6] I.Y. Cherdantsev, R.I. Yamilov, Master symmetries for differential-difference equations of the Volterra type, Physica D 87 (1995) 140. [7] A.B. Shabat, R.I. Yamilov, Symmetries of nonlinear lattices, Leningrad Math. J. 2 (1991) 377. [8] D.J. Zhang, Singular solutions in casoratian form for two differential-difference equations, Chaos Soliton Fract. 23 (2005) 1333. [9] K. Narita, Soliton solution for a highly nonlinear difference-differential equation, Chaos Soliton Fract. 3 (1993) 279. [10] Z. Wang, H.Q. Zhang, New exact solutions to some difference differential equations, Chin. Phys. 15 (2006) 2210. [11] Z. Wang, H.Q. Zhang, A symbolic computational method for constructing exact solutions to difference differential equations, Appl. Math. Comput. 178 (2006) 431. [12] Yu.B. Suris, New integrable systems related to the relativistic Toda lattice, J. Phys. A: Math. Gen. 30 (1997) 1745. [13] Yu.B. Suris, Integrable discretizations for lattice systems: local equations of motion and their hamiltonian properties, Rev. Math. Phys. 11 (1999) 727822. [14] Yu.B. Suris, The Problem of Integrable Discretization: Hamiltonian Approach, Progress in Mathematics, 219, Birkhuser Verlag, Basel, 2003. [15] L. Zou, Z. Zong, G.H. Dong, Generalizing homotopy analysis method to solve Lotka-Volterra equation, Comput. Math. Appl. 56 (2008) 2289. [16] Z. Wang, L. Zou, H.Q. Zhang, Applying homotopy analysis method for solving differential-difference equation, Phys. Lett. A 369 (2007) 77. [17] G. Adomian, The Decomposition Method, Kluwer Acad. Publ., Boston, 1994. [18] G. Adomian, S.E. Serrano, Stochastic contaminant transport equation in porous media, Appl. Math. Lett. 11 (1998) 5355. [19] G. Adomian, R. Rach, N.T. Shawagfeh, On the analytic solution of the. Lane-Emden equation, Phys. Lett. 2 (1995) 161181. [20] D. Kaya, S.M. El-Sayed, On the solution of the coupled Schroinger-KdV equation by the decomposition method, Phys. Lett. A 313 (2003) 8288. [21] T. Badredine, K. Abbaoui, Y. Cherruault, Convergence of Adomians method applied to integral equations, Kybernetes 28 (1999) 557564. [22] G. Adomian, Solutions of nonlinear PDE, Appl. Math. Lett. 11 (1998) 121123. [23] L. Wu, L.D. Xie, J.F. Zhang, Adomian decomposition method for nonlinear differential-difference equations, Commun. Nonlinear Sci. Numer. Simul. 14 (1) (2007) 1218. [24] G.A. Baker, Essential of Pad Approximants, Academic, New York, 1975. [25] M. Chrysos, R. Lefebvre, O. Atabek, On the self-generation of asymptotic boundary conditions in energy quantization, J. Phys B: At. Mol. Opt. Phys. 27 (1994) 3005. [26] V. Ryaboy, R. Lefebvre, N. Moiseyev, Cumulative reaction probabilities using Pade analytical continuation procedures, J. Chem. Phys. 99 (1993) 3509. [27] Z. Wang, Discrete tanh method for nonlinear difference-differential equations, Comput. Phys. Commun. 180 (2009) 1104.