You are on page 1of 12

Nonlinear Dyn (2008) 51:5970 DOI 10.

1007/s11071-006-9191-0

ORIGINAL ARTICLE

Spectral method for solving the equal width equation based on Chebyshev polynomials
A. H. A. Ali

Received: 10 November 2006 / Accepted: 8 December 2006 / Published online: 24 January 2007 C Springer Science + Business Media B.V. 2007

Abstract A spectral solution of the equal width (EW) equation based on the collocation method using Chebyshev polynomials as a basis for the approximate solution has been studied. Test problems, including the migration of a single solitary wave with different amplitudes are used to validate this algorithm which is found to be accurate and efcient. The three invariants of the motion are evaluated to determine the conservation properties of the algorithm. The interaction of two solitary waves is seen to cause the creation of a source for solitary waves. Usually these are of small magnitude, but when the amplitudes of the two interacting waves are opposite, the source produces trains of solitary waves whose amplitudes are of the same order as those of the initial waves. The three invariants of the motion of the interaction of the three positive solitary waves are computed to determine the conservation properties of the system. The temporal evaluation of a Maxwellian initial pulse is then studied. Comparisons are made with the most recent results both for the error norms and the invariant values. Keywords EW equation . Collocation method . Chebyshev polynomials . Spectral method

1 Introduction The regularized long-wave (RLW) equation is an alternative description of nonlinear dispersive waves to the more usual Kortewegde Vries (KdV) equation [1]. It has solitary wave solutions of a rather general type [1, 2]. A less well-known alternative, proposed by Morrison et al. [3], is the equal width equation (EWE) which also has solitary wave solutions, but of a less general type. Solitary waves are wave packets or pulses which propagate in nonlinear dispersive media. The dynamical balance between the nonlinear and the dispersive effects of these waves retain a stable waveform. A soliton is a very special type of solitary wave which also keeps its waveform after collision with other solitons. In practical physics and quantum mechanics, it is standard practice to use the term soliton to designate both solutions to wave equations integrable via the inverse scattering transform, such as KdV, and also to designate localized solutions of nonintegrable equations, such as RLW and EWE. Few analytic solutions for solving EWE are known under certain conditions. Approximate solutions for solving EWE using Galerkins method with both cubic B-spline nite elements [4, 5], a Petrov-Galerkin method using quadratic B-spline nite elements [6], Zaki [7, 8] has solved EW equation by a leastsquare technique using linear space-time nite elements and PetrovGalerkin nite element scheme with shape functions taken as quadratic B-spline functions,
Springer

A. H. A. Ali Mathematics Department, Faculty of Science, Menoua University, Shebein El-Koom, Egypt e-mail: ahaali 49@yahoo.com

60

Nonlinear Dyn (2008) 51:5970

respectively. Recently, Raslan [9] has solved the EW equation using collocation method with quartic Bspline nite elements and solved the resulting system of rst-order ordinary differential equations using the fourth-order RungeKutta method. Soliman [10] has solved the corresponding generalized regularized longwave equation by Hes variational iteration method. Also, Soliman and Hussein [11] have solved the corresponding RLW equation using the collocation method with septic spline. In this paper, we set up the spectral method based on collocation method with basis Chebyshev polynomials. The resulting system will be a system of ordinary differential equations which can be solved using the RungeKutta algorithm. 2 Governing equation and numerical method The EW equation for the long waves propagating in the positive X-direction can take the form: u t + uu X u X X t = 0, a X b, t > 0. (1)

The approximate solution u N (x, t) to the exact solution u(x, t) can be written in the form:
N

u N (x, t) =
n=0

n (t)Tn (x),

(7)

where n are the time-dependent quantities to be determined, Tn (x) are the Chebyshev functions and is dened as: Tn (x) = cos(n cos1 (x)). (8)

From Equation (7) and the property of the Chebyshev functions, we can get the following [12]:
N

ut =
n=0

n (t)Tn (x),
N 1 n=0

(9)
N 1 n=0 (1) n (t)Tn (x),

ux = u x xt =

n (t)Tn (x) =

(10)

with the boundary conditions u(a, t) = u(b, t) = 0, and the initial condition u(X, 0) = f (X ), (3) t >0 (2)

N 2 n=0

n (t)Tn (x) =

N 2 n=0

n (t)Tn (x).

(2)

(11)

Using Equations (9)(11) and substituting in Equation (4) we obtain


N

where and are positive parameters, subscripts X and t denote differentiation, and f (X ) is a localized disturbance inside the considered interval. Using a linear transformation to transfer the interval [a, b] into the interval [1, 1], Equations (1)(3) become: u t + 2uu x 4u x xt = 0, 1 x 1, t > 0,
2

2
n=0

n (t)n Tn (x)
N

+ 2
n=0

n (t) Tn (x)

N 1 k=0

(1) k (t) Tk (x)

N 2 n=0

(2) (t) Tn (x) = 0, n

(12)

(4) where x is the new variable and = b a, the boundary conditions u(1, t) = u(1, t) = 0, and the initial condition u(x, 0) = f (x),
Springer

where n denotes the derivative of n with respect to t and (1) , (2) are dened by
(1) n =

t >0

(5)

2 cn 2 cn

n+2 j1N

(n + 2 j 1)n+2 j1 ,
j=1 n+2 jN

(6)

(2) n =

j(n + j)(n + 2 j)n+2 j .


j=1

(13)

Nonlinear Dyn (2008) 51:5970

61

The nonlinear term Tn (x)Tk (x) can be expressed as a linear combination which is dened as: 1 Tn (x)Tk (x) = Tn+k (x) + T|nk| (x) 2 (14)

Equation (17) can be written in the simple form: A4 (t) = B1 (t),


(18)

where A4 = 2 I 4 A1 , B1 = A2 ( )

We substitute Equation (14) into Equation (12) and hence use the inner product with the weight function j (x). If we choose the weight function j (x) = (x xi ), where (x xi ) is the Dirac Delta function, then this equation can be deduced using the property of the Dirac Delta function to obtain [12]
N

are two matrices of order (N + 1) (N + 1). Multiplying both sides of Equation (18) by the matrix S, we get [13] S A4 (t) = S B1 (t),

2
n=0

n (t)Tn (xi )
N 1 n=0

(19)

N ,N 1 l,k=0 l+k=n

(1) l (t)k (t)Tl+k (xi )

where S = (sin ) = (Tn (xi )), i = 1, . . . , N 1, n = 0, 1, . . . , N (20)

N 1 l,k=0 |lk|=n

(1) l (t)k T|lk| (xi )

N 2 n=0

n (t) Tn (xi ) = 0,

(2)

i = 1, . . . , N 1 (15)

the matrix S of order (N 1) (N + 1). The system given in the Equation (19) consists of N 1 equations in N + 1 unknowns and to obtain a unique solution for this system we need two further equations. For this, we add the two boundary conditions:
N

where the collocation points of the Chebyshev functions are calculated from Equation (8) at x = xi . Equation (15) can be written in the recurrence relation as: 2 n +
N ,N 1 l,k=0 l+k=n (1) l k + N ,N 1 l,k=0 |lk|=n

u(1, t) =
n=0

an Tn (1) = a0 a1 + a2 a3 + a4 + + a N = 0
N

(2) (1) l k 4 n = 0.

u(1, t) =
n=0

an Tn (1) = a0 + a1 + a2 + a3 + a4 + + a N = 0

(16) Hence, we can write Equation (16) in the matrix form as a system of ordinary differential equations ( I 4 A1 ) = ( A2 ( )) ,
2

to Equation (20) to obtain a new system: A (t) = F (t, (t)),


(21)

(17)

where A1 and A2 are the coefcient matrices for the second derivative and the nonlinear term, respectively.

where (t) and F are N + 1 vectors with N + 1 com ponents and A is a matrix of order (N + 1) (N + 1). This system of rst order of ordinary differential equations can be solved numerically using the RungeKutta algorithm to get the numerical solution (t).

Springer

62

Nonlinear Dyn (2008) 51:5970

3 The initial state From the initial condition u(x, 0) on the function u(x, t), we must determine the initial vector 0 so that the time evolution of , using Equation (21), can be started. We rewrite Equation (7) for the initial condition as
N

The analytic values of the invariants can be found from: I1 = 6c/k, I2 = 12c2 /k + 48kc2 /5, (26)

I3 = 144c3 /5k.

4.1 Single solitary wave motion u N (xi , 0) =


n=0 0 n Tn (xi ),

i = 0, 1, . . . , N .

(22)

We now validate our algorithm by studying the motion of solitary waves. The solitary wave solution of the EWE Equation (1) is [3]: u(x, t) = 3c sec h 2 [k(x vt x0 )], (27)

Equation (22) gives a system of N + 1 equations which can be written in the matrix form as: S 0 = b,

(23)

where S is a matrix of order (N + 1) (N + 1) and is dened by Equation (20) and the vector is b = ( f (x0 ), f (x1 ), . . . , f (x N ))T .

4 Test problems A computer program using FORTRAN language with algorithm to calculate the nonlinear term was written for the purpose of obtaining soliton solutions and modeling the undular bore to the EW equation. Accuracy of the method is measured by the L 2 - and L - error norms
N

L 2 = ||u exact u N ||2 = h 2 2


i=0

u iexact u iN

, (24)

L = ||u exact u N || = max u iexact u iN ,


i

and the conservation properties of the proposed algorithm are examined by calculating the invariants which was shown by Olver [15] and corresponds to mass, momentum, and energy, respectively,
b b a b

I1 =
a

u d x, u 3 d x.
a

I2 =

(u 2 + u 2 ) d x, x (25)

I3 =

where v = c is the wave velocity, and k 2 = 1/4. This equation represents a single soliton of magnitude 3c and width k, initially centered at x0 . Here k depends only on and not c as does the corresponding constant for RLW equation; thus, for a given equation (xed ) all solitary waves have the same width, hence the name EW equation. Waves exist with all possible velocities c, c , unlike the RLW equation for which there is the forbidden region 0 c 1. First, we study the motion of a single solitary wave Equation (27) of amplitude 3 through a region 0 x 80 with x0 = 15 and = = 1. To make a comparison with earlier simulation results, Equation (27) is taken as the initial condition with t = 0, x = 0.4, t = 0.1. The simulation is run up to time t = 20. The analytic values of the invariant quantities are: I1 = 12, I2 = 28.8, I3 = 57.6. In the simulation of a solitary wave of amplitude 3, the present approach algorithm leads, at t = 20, to L 2 - error norm as given in Table 1 with value L 2 = 0.0055 103 which is very small, while the quantities I1 , I2 , and I3 are shown in Table 2. The constants of motion vary little from the analytic value: I1 does not change from the analytic, I2 varies by less than 0.00006% and I3 varies by less than 0.0002%. In a corresponding simulation using a collocation method with quartic spline elements [9], the L 2 -error norm at t = 20 is less than 0.6463 103 and the quantity I1 does not change from the analytic, I2 varies by less than 0.043%, and I3 varies by less than 0.005%. Second, we model the motion of a single solitary wave with the three different amplitudes 0.3, 0.09, and 0.03 and compare with results given in [6, 7, 9] at times t = 40, 80; see Tables 37, using the region 0 x

Springer

Nonlinear Dyn (2008) 51:5970 Table 1 The L 2 103 error norm for a single solitary wave of amplitude 3, 0 x 80 t Present method N = 200, t = 0.1 Method [9] N = 400, t = 0.2 0.2 0.0001 0.0111 2 0.0005 0.0940 4 0.0010 0.1567 6 0.0015 0.2009 8 0.0020 0.2441 10 0.0025 0.2938 12 0.0030 0.3502 14 0.0036 0.4135 16 0.0043 0.4835 18 0.0049 0.5608 20

63

0.0055 0.6462

Table 2 Invariant values for a single solitary wave: amplitude = 3, 0 x 80 t 0.2 2 4 6 8 10 12 14 16 18 20 Method Present N = 200 t = 0.1 I1 11.999998 12.000001 12.000000 12.000000 12.000000 12.000000 12.000000 12.000000 12.000000 12.000000 12.000000 I2 28.800000 28.799999 28.799999 28.799998 28.799996 28.799994 28.799992 28.799989 28.799987 28.799985 28.799983 I3 57.599999 57.599992 57.599984 57.599975 57.599967 57.599958 57.599951 57.599943 57.599934 57.599926 57.599918 Method [9] N = 400 t = 0.2 I1 12.000000 12.000000 12.000000 12.000000 12.000000 12.000000 12.000000 12.000000 12.000000 12.000000 12.000000 I2 28.788570 28.788500 28.788420 28.788330 28.788240 28.788150 28.788070 28.787990 28.787900 28.787810 28.787720 I3 57.599950 57.599730 57.599480 57.599210 57.598960 57.598700 57.598440 57.598170 57.597900 57.597650 57.597380

Table 3 Invariants and error for single solitary wave amplitude = 0.3, 0 x 30 Method Present x= t = 0.25 T 5 10 20 30 40 80 40 40 80 80 I1 1.199935 1.199961 1.199988 1.200002 1.200019 1.201120 1.199992 1.1967 1.1964 1.1910 I2 0.288000 0.288000 0.288000 0.288000 0.288000 0.288001 0.292159 0.2860 0.2858 0.2855 I3 0.057600 0.057600 0.057600 0.057600 0.057600 0.057600 0.057599 0.0570 0.0569 0.0558 L 2 103 0.0067 0.0041 0.0016 0.0015 0.0038 0.2052 0.0795 3.475 7.444 3.849

Collocation [9] Least-square [7] Least-square [7] PetrovGalerkin [6] Table 4 Error norms for a single solitary wave: amplitude = 0.09, 0 x 30

x x x x

= 0.03, = 0.03, = 0.03, = 0.03,

t t t t

= 0.2 = 0.03 = 0.03 = 0.05

t 0 10 20 40 80

Method Present x = 0.25 t = 0.25

L 2 103 0.0000 0.0024 0.0018 0.0010 0.0004

L 103 0.0000 0.0120 0.0089 0.0049 0.0015

Method [7] x = 0.1 t = 0.05

L 2 103 0.0000 0.0200 0.0400 0.0900 0.2200

L 103 0.0200 0.0200 0.0300 0.0700 0.1600

Table 5 Invariant values for a single solitary wave: amplitude = 0.09, 0 x 30

t 0 10 20 40 80

Method Present x = 0.25 t = 0.25

I1 0.3600 0.3600 0.3600 0.3600 0.3600

I2 0.0259 0.0259 0.0259 0.0259 0.0259

I3 0.00156 0.00156 0.00156 0.00156 0.00156

Method [7] x = 0.1 t = 0.05

I1 0.3600 0.3599 0.3599 0.3597 0.3593

I2 0.0259 0.0259 0.0259 0.0259 0.0259

I3 0.00156 0.00156 0.00155 0.00155 0.00155

Springer

64 Table 6 Error norms for a single solitary wave: amplitude = 0.03, 0 x 30

Nonlinear Dyn (2008) 51:5970

t 0 40 80

Method Present x = 0.25 t = 0.25

L 2 103 0.0000 0.0007 0.0005

L 103 0.0000 0.0036 0.0024

Method [7] x = 0.1 t = 0.1

L 2 103 0.0017 0.0084 0.0177

L 103 0.0055 0.0063 0.0127

Table 7 Invariant values for a single solitary wave: amplitude = 0.03, 0 x 30 t 0 40 80 Method Present x = 0.25 t = 0.25 I1 0.1200 0.1200 0.1200 I2 0.00288 0.00288 0.00288 I3 0.000058 0.000058 0.000058 Method [7] x = 0.1 t = 0.1 I1 0.1200 0.1200 0.1200 I2 0.00288 0.00288 0.00288 I3 0.000058 0.000058 0.000058

30, x0 = 10, = = 1, x = t = 0.25. To examine the behavior of the present numerical algorithm for the solitary wave with the three different amplitudes 0.3, 0.09, and 0.03, we use the L 2 - error norm to measure the accuracy and the quantities I1 , I2 , and I3 to measure conservation. We discuss the three cases: (i) The case with amplitude 0.3, the analytic values of the invariants are: I1 = 1.2, I2 = 0.288, and I3 = 0.0576. By the time t = 40, see Table 3 where the L 2 -error norm is less than 0.0038 103 and the constants of motion vary little from the analytic value: I1 varies by less than 0.002%, I2 and I3 are conserved during the experiment. In a corresponding simulation using a collocation method with quartic spline elements [9], the L 2 -error norm is less than 0.0796 103 and the quantity I1 changes by less than 0.0007%, I2 by less than 1.5%, and I3 varies by less than 0.002%, and using a least-squares method with linear spline elements [7] the L 2 -error norm is less than 3.476 103 and the quantity I1 changes by less than 0.28%, I2 by less than 0.7%, and I3 varies by less than 1.1%. At time t = 80, the L 2 -error norm is less than 0.2053 103 , which is smaller than the previous results [6, 7] 3.849 103 and 7.444 103 , respectively. The invariant value I1 changes by less than 0.094%, I2 by less than 0.0004%, and I3 does not change and in the corresponding simulation using PetrovGalerkin method with cubic spline nite elements [6] the invariant values I1 , I2 , and I3 change by less than 0.75, 0.87, and 3.13%, respectively, and using a least-squares method with linear spline elements [7] the quantity I1 changes by less
Springer

than 0.3%, I2 by less than 0.74%, and I3 varies by less than 1.22%. (ii) The case with amplitude 0.09 and time t = 80, the analytic values of the invariants are: I1 = 0.36, I2 = 0.02592, and I3 = 0.001555. The computed values of the L 2 -error norm and the quantities I1 , I2 , and I3 are given in Tables 4 and 5, respectively, with the corresponding previous results. This simulation of a solitary wave of amplitude 0.09 leads, with the present algorithm, to the L 2 -and L -error norms of 0.0004 103 0.0015 103 , respectively, while I1 , I2 , and I3 are constants during the experiment. In corresponding simulation using least-squares method with linear spline nite elements [7], the L 2 -and L error norms are 0.22 103 0.15 103 respectively, while the invariant values change by less than 0.65% during the experiment. (iii) The case with small amplitude 0.03 and time t = 80, the analytic values of the invariants are: I1 = 0.12, I2 = 0.00288, and I3 = 0.000058. The computed values of the L 2 -error norm and the quantities I1 , I2 , and I3 are given in Tables 6 and 7, respectively, with the corresponding previous results. This simulation of a solitary wave of amplitude 0.03 leads, with the present algorithm, to the L 2 -and L -error norms of 0.0005 103 0.0024 103 , respectively, while I1 , I2 , and I3 are constants during the experiment. In corresponding simulation using least-squares method with linear spline nite elements [7], the L 2 - and L -error norms are 0.0177 103 0.0127 103 , respectively, while the invariant values are constants during the experiment.

Nonlinear Dyn (2008) 51:5970 Table 8 Invariant values for the interaction of positive and negative solitary waves, 0 x 80 t 1 4 6 8 10 Method Present N = 320 t = 0.1 I1 2.399998 2.399999 2.400001 2.399941 2.399900 I2 97.919679 97.919605 97.919581 97.925487 97.915904 I3 58.521339 58.521151 58.521044 58.522231 58.522098 Method [9] N = 800 t = 0.1 I1 2.399999 2.399996 2.399999 2.399998 2.399998 I2 97.909970 97.908200 97.694180 97.012280 97.149240 I3 58.521330 58.521110 58.521440 58.526150 58.526940

65

4.2 The interaction of solitary waves Consider the initial condition of two solitary waves: u(x, 0) = u 1 + u 2 where u i = 3ci sec h 2 (0.5(x xi ci )), i = 1, 2 (29) (28)

and solving the EW equation over the region a x b taking = = 1 and appropriate boundary conditions. 4.2.1 The interaction of positive and negative solitary waves Gardner [5] and Raslan [9] have studied the interaction of a positive and negative solitary waves for the EW equation, and observed the collision to produce additional pairs of daughter solitary waves emanating from the point of initial contact, an observation conrmed by [1315]. We have repeated those experiments using the appropriate initial condition Equations (27)(28) and solved the EW equation over the region 0 x 80 taking, c1 = 1.2 x1 = 23, c2 =
Fig. 1 The interaction of two opposite solitary waves at t = 10
3 2 1 0 -1

1.4, x2 = 38, x = 0.25, and t = 0.1. The analytical valued of the invariant quantities are I1 = 2 2 12(c1 + c2 ) = 2.4, I2 = 28.8(c1 + c2 ) = 97.92, and 3 3 I3 = 57.6(c1 + c2 ) = 58.5216. In Fig. 1 we show the behavior of interaction of positive and negative solitary waves at time t = 10. The values of I1 , I2 , and I3 throughout the simulation are shown in Table 8 comparing with Raslan [9] and all are satisfactorily conserved; I1 changes by less than 0.0042%, I2 by less than 0.0042%, and I3 varies by less than 0.0009%. In corresponding simulation using a collocation method with quartic spline nite elements [9], the invariant quantity I1 changes by less than 0.00009%, I2 by less than 0.79%, and I3 by less than 0.0092% during the experiment. 4.2.2 The interaction of two positive solitary waves Raslan [9] has studied the interaction of two positive solitary waves for the EW equation with initial condition and observed the two waves have apparently passed through one another and emerged unchanged by the encounter. We have chosen to study a similar situation using the initial condition given by Equations (27)(28). The EW equation was solved over the region

u
-2 -3 -4 -5 0 10 20 30 40 50 60 70 80

Springer

66
5

Nonlinear Dyn (2008) 51:5970

a: t=0.5

b: t=15

u
2 1 1 0 0 10 20 30 40 50 60 70 80 0 0 10 20 30 40 50 60 70 80

X
5 5

c: t=20

d: t=25

u
2 1 1 0 0 20 40 60 80 0 0 20 40 60 80

Fig. 2 The motion of two positive solitary waves with amplitudes ratio two to one (2:1) before the interaction and after the interaction

0 x 80 with c1 = 1.5, x1 = 10, c2 = 0.75, x2 = 25, x = 0.4, and t = 0.1. The analytical valued of the invariant quantities are I1 = 27.0, I2 = 81.0, and I3 = 218.7. The conguration at time t = 25, which is sometime after the interaction is complete, is shown in Fig. 2ad. The waves have apparently passed through one another and emerged unchanged by encounter. The simulations are run to time t = 25, and the invariants I1 , I2 , and I3 are recorded in Table 9 comparing with Raslan [9] and all are satisfactorily conserved; I1 does not changes, I2 by less than 0.0004, and I3 varies by less than 0.0002%. In corresponding simulation using a collocation method with quartic spline nite elements [9], the invariant quantity I1 changes by less than 0.47%, I2 by less than 0.28%, and I3 by less than 0.0008% during the experiment. Now, we study the interaction of two solitary waves as above but with amplitudes ratio three to one (3:1), in this case it changes only the value of c1 = 2.25. In Fig. 3ad, the interaction of these two solitary waves are plotted at different times. The analytical valued of the invariants can be found as I1 = 36.0, I2 = 162.0, and

I3 = 680.4. The conguration at time t = 15, which is some time after the interaction is complete, is shown in Fig. 3ad. The waves have apparently passed through one another and emerged unchanged by encounter. The simulations are run to time t = 15, and the invariants I1 , I2 , and I3 are given in Table 10 comparing with Raslan [9] and all are satisfactorily conserved; I1 does not changes, I2 by less than 0.007%, and I3 varies by less than 0.0093%. In corresponding simulation using a collocation method with quartic spline nite elements [9], the invariant quantity I1 changes by less than 0.003%, I2 by less than 0.012%, and I3 by less than 0.0089% during the experiment. Now, we study the interaction of two solitary waves as earlier but with amplitude ratio four to one (4:1), in this case the change is only in the value of c1 = 2 and c2 = 0.5. In Fig. 4ad, the interaction of these two solitary waves are plotted at different times. The analytical valued of the invariant quantities are I1 = 30, I2 = 122.4, and I3 = 468. The conguration at time t = 15, which is sometime after the interaction is complete, is shown in Fig. 4ad. The waves have apparently

Springer

Nonlinear Dyn (2008) 51:5970 Table 9 Invariant values for the interaction of two positive solitary waves with amplitudes ratio 2:1, 0 x 80 t 0.5 5 10 15 20 25 Method Present N = 200 t = 0.1 I1 26.999961 27.000001 27.000000 27.000000 27.000000 27.000000 I2 81.000402 81.000244 81.000075 80.999990 80.999877 80.999703 I3 218.702754 218.702045 218.701267 218.700697 218.700413 218.699658 Method [9] N = 800 t = 0.1 I1 26.999730 27.000080 27.000050 27.001290 27.026770 27.124800 I2 81.010280 81.000370 80.992990 80.996670 82.408130 81.220630 I3 218.701200 218.700600 218.699700 218.699000 218.698900 218.698300

67

Table 10 Invariant values for the interaction of two positive solitary waves with amplitudes ratio 3:1, 0 x 80 t 0.5 5 10 15 Method Present N = 200 t = 0.1 I1 36.000011 36.000001 36.000000 36.000000 I2 162.000888 161.996975 161.994645 161.990234 I3 680.408994 680.382228 680.366003 680.336773 Method [9] N = 800 t = 0.1 I1 36.000060 36.000030 36.000060 36.000990 I2 162.015400 161.979500 161.980400 161.982000 I3 680.399500 680.372400 680.354600 680.339500

8 7 6 5 4 3 2 1 0 0

a : t= 1 5

8 7 6 5 4 3 2 1 0

b : t= 5

10

20

30

40

50

60

70

80

10

20

30

40

50

60

70

80

X
8 7 6 5 4 3 2 1 0 0 10 20 30 40 50 60 70 80 8 7

c : t= 1 0
u

6 5 4 3 2 1 0 0 10 20

d : t= 1 5

30

40

50

60

70

80

Fig. 3 The motion of two positive solitary waves with amplitudes ratio three to one (3:1) before the interaction and after the interaction.

passed through one another and emerged unchanged by encounter. The simulations are run to time t = 15, and the invariants I1 , I2 , and I3 are given in Table 11 comparing with Raslan [9] and all are satisfactorily conserved; I1 does not changes from the analytic value, I2 by less than 0.003%, and I3 varies by less than 0.0045%. In corresponding simulation using a collocation method with quartic spline nite elements [9],

the invariant quantities I1 changes by less than 0.029%, I2 by less than 0.012%, and I3 by less than 0.32% during the experiment. At the end of the study of the interaction of two solitary waves, we study the interaction of three positive solitary waves with amplitudes ratio 9:3:1, in this case we put c1 = 4.5, c2 = 1.5, c3 = 0.5, x = 0.4, t = 0.1, x1 = 10, x2 = 25, x3 = 35, and the

Springer

68
7 6 5 4

Nonlinear Dyn (2008) 51:5970


a : t= 5
7 6 5 4

b : t= 5

3 2 1 0 0 7 6 5 4 10 20 30 40 50 60 70 80

3 2 1 0 0 7 6 5 4 10 20 30 40 50 60 70 80

c : t= 5

d : t= 1 5

3 2 1 0 0 10 20 30 40 50 60 70 80

3 2 1 0 0 10 20 30 40 50 60 70 80

Fig. 4 The motion of two positive solitary waves with amplitudes ratio four to one (4:1) before the interaction and after the interaction

interval 0 x 100. The analytical value of the invariant quantities are I1 = 12(c1 + c2 + c3 ) = 78.0, 2 2 2 3 I2 = 28.8(c1 + c2 + c3 ) = 655.2, and I3 = 57.6(c1 + 3 3 c2 + c3 ) = 5450.4. The simulations are run to time t = 15, which is some time after the interaction is complete, as shown in Fig. 5ad. The waves have apparently passed through one another and emerged unchanged by encounter. The values taken by the invariant quantities I1 , I2 , and I3 over the period of simulation are given in Table 12 comparing with Raslan [9] and all are satisfactorily conserved; I1 varies by less than 0.00003%, I2 by less than 0.43%, and I3 by less than 0.71%. In corresponding simulation using a collocation method with quartic spline nite elements [9], the

invariant quantities I1 changes by less than 0.006%, I2 by less than 0.37%, and I3 by less than 0.72% during the experiment. 4.3 Maxwellian initial condition Consider the Maxwellian initial condition u(x, 0) = exp((x 7)2 ). (30)

We have solved the EW equation with initial condition (30) and various values of the parameter . We discuss the numerical solution in the cases: = 0.2, 0.04, 0.001.

Table 11 Invariant values for the interaction of two positive solitary waves with amplitudes ratio 4:1, 0 x 80 t 0.5 5 10 15 Method Present N = 200 t = 0.1 I1 30.000000 30.000001 30.000000 30.000000 I2 122.400616 122.399078 122.398102 122.396379 I3 468.004881 467.995550 467.989175 467.979127 Method [9] N = 800 t = 0.1 I1 29.999930 30.000040 30.000070 30.008420 I2 122.41500 122.386300 122.389500 122.789500 I3 468.000200 467.991000 467.983700 467.974600

Table 12 Invariant values for the interaction of three positive solitary waves with amplitudes ratio 9:3:1, 0 x 100 t 0.5 5 10 15 Method Present N = 250 t = 0.1 I1 78.000015 77.999999 77.999998 77.999984 I2 655.168150 654.543602 653.483686 652.411538 I3 5449.679333 5440.868793 5426.678714 5412.231849 Method [9] N = 1000 t = 0.1 I1 78.001320 78.000240 77.997310 77.995390 I2 655.118800 654.349400 653.418900 652.810400 I3 5449.065000 5438.952000 5426.081000 5411.639000

Springer

Nonlinear Dyn (2008) 51:5970 Table 13 Invariant values for Maxwellian initial condition

69

0.2

t 0.5 1.0 2.0 3.0 4.0 0.5 1.0 2.0 3.0 4.0 0.5 1.0 2.0 3.0 4.0

I1 (Present) 1.772454 1.773063 1.773604 1.774116 1.774653 1.772454 1.773063 1.773603 1.774114 1.774648 1.772453 1.772454 1.772453 1.772434 1.760800

I2 (Present) 1.503977 1.508032 1.520078 1.539771 1.566542 1.303447 1.308312 1.322768 1.346399 1.378525 1.254567 1.259631 1.274673 1.299265 1.332696

I3 (Present) 1.023327 1.023328 1.023329 1.023329 1.023330 1.023327 1.023328 1.023329 1.023329 1.023330 1.023327 1.023328 1.023329 1.023329 1.023330

I1 [9] 1.772454 1.772457 1.772531 1.774443 1.796324 1.772454 1.772453 1.772454 1.772447 1.765447 1.772453 1.772454 1.772453 1.772434 1.760800

I2 [9] 1.503339 1.503303 1.503190 1.505232 1.583276 1.303289 1.303178 1.302571 1.301551 1.301041 1.254567 1.254566 1.254304 1.253925 1.255103

I3 [9] 1.023327 1.023327 1.023328 1.023329 1.023345 1.023328 1.023331 1.023355 1.023400 1.023424 1.023326 1.023327 1.023330 1.023327 1.023311

0.04

0.001

14 12 10 8

14

a : t= 0 .5

12 10 8

b : t= 5

6 4 2 0 0 10 20 30 40 50 60 70 80 90 100

u
6 4 2 0 0 10 20 30 40 50 60 70 80 90 100

X
14 12 10 8

X
14 12 10 8

c : t= 1 0

d : t= 1 5

6 4 2 0 0 10 20 30 40 50 60 70 80 90 100

u
6 4 2 0 0 10 20 30 40 50 60 70 80 90 100

Fig. 5 The motion of three positive solitary waves with amplitudes ratio 9:3:1 before the interaction and after the interaction

For = 0.2, the simulations are run to time t = 4, while the invariant quantities I1 , I2 , and I3 are given in Table 13 comparing with Raslan [9]. The three invariant quantities change from the initial by a factor of 0.0022, 0.063, and 0.000003, respectively. In corresponding simulation using a collocation method with quartic spline nite elements [9], the three invariant quantities change from the initial by a factor of 0.024, 0.08, and 0.000018, respectively, during the experiment.

For = 0.04, the simulations are run to time t = 4, while the invariant quantities I1 , I2 , and I3 are given in Table 13 comparing with Raslan [9]. The three invariant quantities change from the initial by a factor of 0.0022, 0.075, and 0.000003, respectively. In corresponding simulation using a collocation method with quartic spline nite elements [9], the three invariant quantities change from the initial by a factor of 0.007, 0.0023, and 0.0001, respectively, during the experiment.

Springer

70

Nonlinear Dyn (2008) 51:5970

For = 0.001, the simulations are run to time t = 4, while the invariant quantities I1 , I2 , and I3 are given in Table 13 comparing with Raslan [9]. The three invariant quantities change from the initial by a factor of 0.012, 0.078, and 0.000003 respectively. In corresponding simulation using a collocation method with quartic spline nite elements [9], the three invariant quantities change from the initial by a factor of 0.012, 0.00065, and 0.00002, respectively, during the experiment.

References
1. Peregrine, D.H.: Calculations of the development of an undular bore. J. Fluid Mech. 25, 321330 (1966) 2. Abdulloev, Kh.O., Bogolubsky, H., Makhankov, V.G.: One more example of inelastic soliton interaction. Phys. Lett. A 56, 427428 (1976) 3. Morrison, P.J., Meiss, J.D., Carey, J.R.: Scattering of RLW solitary waves. Physica D 11, 324336 (1981) 4. Gardner, L.R.T., Gardner, G.A.: Solitary waves of the regularised long wave equation. J. Comput. Phys. 91, 441459 (1990) 5. Gardner, L.R.T., Gardner, G.A.: Solitary waves of the equal width wave equation. J. Comput. Phys. 101, 218223 (1992) 6. Gardner, L.R.T., Gardner, G.A., Ayoub, F.A., Amein, N.K.: Simulations of the EWE undular bore. Commun. Num. Methods Eng. 13, 583592 (1997) 7. Zaki, S.I.: A least-squares nite element scheme for the EW equation. Comput. Methods Appl. Mech. Eng. 189, 587594 (2000) 8. Zaki, S.I.: Solitary waves induced by the boundary forced Ew equation. Comput. Methods Appl. Mech. Eng. 190, 4881 4887 (2001) 9. Raslan, K.R.: Collocation method using quartic B-spline for the equal width (EW) equation. Appl. Math. Comput. (USA) 168, 785805 (2005) 10. Soliman, A.A.: Numerical simulation of the generalised regularised long wave equation by Hes variational iteration method. Math. Comput. Simul. 70(2), 119124 (2005) 11. Soliman, A.A., Hussein, M.H.: Collocation solution for RLW equation with septic spline. Appl. Math. Comput. 161, 623636 (2005) 12. Fornberg, B.: A Practical Guide to Pseudospectral Methods. Cambridge University Press, Cambridge, UK (1995) 13. Olver, P.J.: Euler operators and conservation laws of the BBM equation. Math. Proc. Cambridge Phil. Soc. 85, 143 159 (1979) 14. Elibeck, J.C., McGuire, G.R.: Numerical study of the RLW equation II: Interaction of solitary waves. J. Comput. Phys. 23, 6373 (1977) 15. Iskandar, L., El-Deen Mohamedein, M.Sh.: Solitary waves interaction for the BBM equation. Comput. Methods Appl. Mech. Eng. 96, 361372 (2001).

5 Discussion It has been shown that the numerical solution for solving the EW equation using the spectral method based on Chebyshev polynomials within the collocation method is more accurate compared to the recent results during all run to the simulations. The error norms computed by the present algorithm with different amplitudes compared to the previous results were found to be smaller. The three invariants of motion are satisfactorily constant in all the computer simulations described here, so that the algorithm can fairly be described as conservative. So, we deduce that this algorithm is accurate and more efcient than the previous algorithms and we believe that this approach will also be useful for solving similar nonlinear partial differential equations. It is worthwhile noticing that all our computations have been conducted on a 32-bit machine, which means that the accuracy could have been much better if we used a mainframe with 256 or 512-bit processors.

Springer

You might also like