COMMUNICATIONS IN NUMERICAL METHODS IN ENGINEERING, VOl.

9, 579-586 (1993)

NUMERICAL SOLUTION OF REGULARIZED LONG-WAVE

EQUATION

P. c. JAIN’, RAMA SHANKAR AND T. v. SINGH

Deportment of Mathematics, Indian Institute of Technologv. Hauz Khas, New Delhi - I IOOI6, India

SUMMARY
A cdmbination of the splitting method and the cubic spline technique has been used to solve a non-linear
regularized long wave (RLW)equation. The accuracy and the stability of the proposed method are
discussed. Then, two numerical examples are solved to illustrate the robustness of the algorithm. The
results are in good agreement with the available results.

INTRODUCTION
Non-linear systems of partial differential equations subject to certain physically motivated
initial/boundary conditions usually govern the wave phenomena. Burgers’ equation
1
ut + uux
.
= - u,,
Re
and the Korteweg-de Vries (K-dV) equation
ut + ux + uux + u x x x = 0 (2)
where, as usual, subscripts denote partial derivatives and Re is the Reynolds Number, are the
best examples of model equations governing the wave phenomena.
Benjamin et al. have advocated the study of the regularized long-wave (RLW) equation as
an alternative to the K-dV equation (2) which may be expressed as
Ur + uX+ auu, - buxxt= 0, XE IR, t >0
(3)
u(x, 0)= g ( x ) , X€ m
where a and b are real constants. They’ have also shown that equation (3), with a = b = 1, can
be used as a modified model for various phenomena represented by the K-dV equation. The
existence and uniqueness properties of this problem under suitable conditions have been
discussed by Bona and Pryant.’ Albert et al.’ have obtained sufficient conditions for stability
of solitary wave solutions of model equations for long waves. These conditions are shown to
be applicable to the RLW equation.
Peregrine4 was the first to use the RLW equation to describe the development of an undular
bore. He used the finite-difference scheme to solve this equation, which is only first-order-
accurate in time and second-order-accurate in space. Eilbeck and McGuire’ proposed a
difference scheme to solve the RLW equation, which is second-order-accurate in space and

*Present Address: 3389 D/3 Vasant Kunj, New Delhi-110030, India.

0748-8025/93/070579-08$09.00 Received 25 March 1992

0 1993 by John Wiley & Sons, Ltd. Revised 15 September 1992
580 P. C. JAIN, R. SHANKAR AND T. V. SINGH

time as well. But this scheme is a three-level scheme and requires another approximation
initially. Jain and Iskander described a combined approach of quasilinearization and
invariant imbedding to find the numerical solution of the RLW equation (3). Owing to the
simplifications introduced in that paper, the numerical results had some limitations. These
'
and other finite-difference schemes are critically examined by Bona et al. Jain et al. * proposed
a general-purpose algorithm for obtaining the solution of Burgers' equation with a semilinear
boundary condition. In the present work we apply a similar technique as proposed in
Reference 8 for solving the RLW equation. The accuracy and the stability of the proposed
method are discussed. Then, two numerical examples are solved to illustrate the robustness of
the algorithm.

NUMERICAL METHOD
Consider the RLW equation
ut + u, + af, - bu,,, =0 (4)
where f = u2/2.
We use equation (4) as follows:
~1 ( u - b u ~ ~-afx
)t= (5)
1
7 (u - bu,,), = - u, (6)
Approximating the time derivative by forward difference, the derivative uxxby second-order
derivative Mi" of the cubic spline function interpolating u f s and f , by the first derivative m,!'
of the cubic spline function interpolating f / s , equation ( 5 ) can be written as
u/+l/2 - bMT+l/Z= u y - b M / - k61am:+'/2- ka(1 - & ) m y (7)
Here k is the increment in time and 01 E [0,1] is a cubic spline parameter.
Now the continuity conditions9 for the derivatives of the splines are:

where
p6f/=f?+r-f?-1
62ui"= u?+1- 2ui" + 2.47-1
and h is the space mesh size.
Eliminating Mi" and mi" from (7), (8) and (9), we obtain the difference scheme for equation
( 5 ) as

Similarly, from equation (6) we obtain the difference scheme

 NUMERICAL SOLUTION OF REGULARIZED LONG-WAVE EQUATION 58 1

Equations (10) and (1 1) constitute a multistep implicit finite-difference scheme for solving
equation (4).
For computational work, we remove the non-linearity in the difference equation (10) by
using the Taylor series expansion:

where A = aflau.
f;+l/'=f;+k af
2 at 0 " + O ( k ' ) = f i " +A:(U:+l/'-
i
u:) + O(k2) (12)

Thus equation (10) takes the form

(
6' b 6' +-ak 81pliA:)U,!"L/2= (1
1 ---
6 2 2 h
+--
6'
6
b p)u:--
6'
2h (1 - el)ccsfi"
ak (13)

The algorithm for the numerical solution of equation (4) is given by equations (11) and (13).

LOCAL TRUNCATION ERROR AND STABILITY OF THE METHOD

Local truncation errors in equations (10) and (1 1) are given by

and

respectively, by neglecting the terms of higher order. Therefore equations (10) and (11) give
a truncation error of O(k2+ kh2) for d1 and 82 E [0,1], whereas for 81 = 82 = 112 equations (10)
and (11) provide a method of O(k2+ h').
The local truncation error in calculating u?+' from U? for equation (4) is given by

It is clear from equation (16) that our numerical method is of O(k2+ h2) for 81 = 82 = 4. Also,
using the Von-Neumann method of stability on the basic equations (11) and (13), we find that
the proposed technique is unconditionally stable for all 81 and 82 c [i,11.

NUMERICAL EXAMPLES AND RESULTS

In this Section we solve two test examples to illustrate the accuracy and efficiency of the
proposed numerical method.
582 P. C. JAIN. R. SHANKAR AND T. V. SlNGH

Example 1
Consider R L W equation (3) for a = b = 1 with the boundary condition .

~(+25,t)=O, t>O (17)

and the initial condition as
u(x, 0) = 3C sech’(px) (18)
where C is an arbitrary constant and p = f(C/(l + C))”’.The solitary-wave (soliton) solution
of this problem is given by
u(x, t) = 3C sech’(px - w t + 4) (19)
where w = f(C(C+ 1))”’ and $C is an arbitrary constant. (The solution (19) is obtained by
using the initial condition (18) when 4 = 0). For computational work we have chosen C = 0.01 ,
0.07 and 0.1 as representative values of the parameter C. Mesh constants are chosen as
h = k = 0.1, = 0 and the domain of the solution as - 25 < x - t < 25. The numerical results
are found to be in good agreement for C=O-O3 (cf. Figure 1) and C=O-O7, while the
proposed method yields some deviation when compared with the soliton solution for C = 0.1
(cf. Figure 2). The reason for this deviation between the solitary wave solution and the
numerical solution may be as pointed out by Benjamin et a1.” and Bona et al.” that the
validity of the RLW equation and its solitary wave solution in real systems depends on the
magnitude I u I of the solution being small everywhere. In other words, it is required that
C 4 1; our technique yields satisfactory results under this restriction.

0.10-

0.08 1

0.06-
-
t
0.04-

0.02-

-20.0 -100 0.0 10.0 20.0 3ao

Figure 1. Soliton solution of example 1: C=0-03;---solitary, -numerical solution

 NUMERICAL SOLUTION OF REGULARIZED LONG-WAVE EQUATION 583

k = 0.1
h = 0.1
t = 20

-20.0 -10.0 0.0 10.0 20.0 300

c x - t -

Figure 2. Soliton solution of example 1: C- 0- I; ---solitary. -numerical solution

Example 2
Consider equation (3) for the undular bore ( a = and b = i), with the initial and boundary
conditions as

=I
u ( x , 0) 2 uo (1 - tan(:))

u ( - 20,t) = uo and the right-hand boundary condition is set at u = 0, where uo and d are
constants. The numerical calculations are stopped if the non-zero part of the solution
overtakes this moving boundary within some numerical tolerance error e.
Figure 3(a) shows the development of the undulation profile for the initial gentle slope with
d = 5.0 and ug = 0.1. The computational work is done up to t = 200. It is observed that the
undular bore almost reaches the steady state up to t Q 150, but it becomes unstable for t > 150
(cf. Figure 4). For a stable region it is found that the rate of growth of the first undulation
is constant in the beginning and then this initial rate of growth varies with the square of the
initial amplitude. The stable part of the solution is in conformity with the results obtained by
Peregrine.
Similar results are obtained for the undular bore with gentle slope when d = 5.0 and
uo = 0.2; in this case the steady-state region is found up to t = 70. The rate of growth of
the undular bore is observed to be fast in the beginning. This change in the behaviour of the
undular bore is due to the fact that the results of computations depend on the form of the
initial wave.4
584 P. C. JAIN, R. SHANKAR AND T. V. SINGH

0 175

=150

=125

rn = 100

I=Yu
&lo Elz
t=75 t r75

t =50 t550

0.1
0 -10 0 10 2 0t - 2 5 10 20t ~ 2 5

-20 -20 -10 0

(a) -x-t- (b) CX- t ---c

Figure 3. Example 2. development of undulation profile for (a) gentle slope, d = 5, uo = 0 - 1 and (b) steep slope,
d = 2 . 0 , uo=o.1

0.2 5r

0.2 0- -

0.05 - -
0 I I I

Figure 4. Variation of max u in development of undulation profile with respect to t: (a) uo = 0.1, d = 5; (b) uo= 0 - 1,
d=2.O
 NUMERICAL SOLUTION OF REGULARIZED LONG-WAVE EQUATION 585

Figure 3(b) shows the development of the undulation profile for the initial steep bore with
d = 2.0 and u = 0.1. Computational work is done up to = 200. Peregrin4 has observed that if
the initial wave is relatively steep then the rate of growth of the first undulation is not constant
but decreases slowly after a rapid start; this is confirmed by our computational results up to
t < 150 (cf Figure 4). After t = 150, the numerical results show instability. Similar results were
obtained for the initial steep bore with d = 2.0 and uo = 0.2. Again, slight changes in the rate
of growth of the undular bore have been observed, and this is again due to the reason that
the initial steep wave affects the development of the undulation bore p r ~ f i l e . ~
Albert et al.3 have given sufficient conditions under which the results of the RLW equation
are stable. It is found that if we use the initial condition
$'o(x) = 3c sech*(fJcx)
or

which satisfy the sufficient condition of Albert et then we obtain stable results (cf. Figures
5(a) and (b)). Figure 6 shows the development of undular bores.
It is interesting to note that under the following transformations:
u = - (b'" +b1/4- 1)
a
x = bl/ZXJ
t = b1I4tr

t=175

t =150

t.125

t =loo
t
1
t=75

t=50

t = 25

O 00 10
O 20 3 30 L1

Figure 5. Example 2. development of undulation profile with u(x.0) =(a) 96(x), C = O * O 1 and (b) +l(x), c = 0.01
586 P. C. JAIN, R. SHANKAR AND T. V. SINGH

Figure 6. Variation of max u in development of undulation profile: (a) u(x. 0) = &f,(x); (b) u(x, 0) = & 5 ( x )

equation (3) reduces to the RLW equation discussed by Albert ei aL3 and that the initial
condition (20) does not satisfy the sufficient condition given in Reference 3. This confirms the
occurrence of the instabilities observed in our numerical results for t > 150 (cf. Figure 4).
Thus, the proposed numerical scheme, which is of general nature, yields satisfactory numerical
solutions of the R L W problem.

ACKNOWLEDGEMENT

The authors are thankful to the referees for constructive criticisms which have helped in
improving the paper.

REFERENCES
1. T. B. Benjamin, J. L. Bona and J. J. Mohony, ‘Model equations for long waves in nonlinear
dispersive systems’, Phil. Trans. Roy SOC., London, 272A, 47 (1972).
2. J. L. Bona and P. J. Pryant, ‘A mathematical model for long waves generated by wave makers in
nonlinear dispersive systems’, Proc. Cambridge Phil. SOC.,73, 391 (1973).
3. J. P. Albert, J. L. Bona and D. B. Henry, ‘Sufficient conditions for stability of solitary wave
solutions of model equations for long waves’, Physica, 24D, 343 (1987).
4. D. H. Peregrine, ‘Calculations of the development of an undular bore’, J. Fluid Mech., 25(2), 321
(1%).
5. J. C. Eilbeck and G. R. McGuire, ‘Numerical study of regularized long wave equation; I: Numerical
methods’, J. Comput. Phys., 19, 43 (1975).
6. P. C. Jain and Labib Iskander, ‘Numerical solution of the RLW equation’, Comput. Methods Appl.
Mech. Eng., 20, 195 (1979).
7. J. L. Bona, W. G. Pritchard and L. R. Scott, ‘Numerical schemes for a model for nonlinear
dispersive waves’, J. Comput. Phys., 60, 167 (1985).
8. P. C. Jain, Rama Shankar and T. V. Singh, ‘Cubic spline technique for the solution of Burgers’
equation with a semilinear boundary condition’, Commun. appl. numer methods, 8, 235 (1992).
9. J. H.Ahlberg, E. N. Nilson and J. L. Walsh, The Theory of Splines and their Applications,
Academic Press Inc., London, 1967.
10. T. B. Benjamin, ‘The stability of solitary waves’, Roc. Roy. SOC., London, 328A, 153 (1972).
11. J. L. Bona, W. G. Pritchard and L. R. Scott, ‘Long wave interaction’, Phys. Fluids, 23, (3), 438
(1980).