the generalized kortewed de vries (kdv)...
TRANSCRIPT
14
THE GENERALIZED
KORTEWED DE VRIES (KdV) EQUATION
2.1 INTRODUCTION
The Korteweg-de Vries (KdV) equation is a nonlinear partial differential
equation of third order as
0 xxxxt uuuu , (2.1.1)
where and are positive parameters and the subscripts denote partial
differentiation. The KdV equation first appeared in an article written by Dutch
mathematicians Korteweg and de Vries [62]. They formulated the KdV equation to
describe long wave propagation on shallow water. The KdV equation is one of the
simplest nonlinear model equations for solitary waves. In the KdV equation (2.1.1),
),( txuu measures the elevation at time t and position x , i.e. the height of the water
above the equilibrium level. The second and the third terms in the equation are the
nonlinearity and the dispersion. The term xuu describes the sharpening of the wave
and xxxu the dispersion i.e. waves with different wave lengths propagate different
velocities. Remarkably the balances between these two effects allow solutions in
terms of propagating waves with unchanged form. Indeed, the KdV equation is a
special case of the generalized Korteweg-de Vries (gKdV) equation of the form
0 xxxx
p
t uuuu , (2.1.2)
where p is a positive integer. When 2p , the gKdV equation (2.1.2) takes the form
02 xxxxt uuuu (2.1.3)
and it is known as the Modified Kortweg-de Vries equation (mKdV). In the study of
the KdV equation, Zabusky and Kruskal [99] indicated that the wave solutions
persisted after interactions and the wave solutions were called ‘solitons’. The soliton
concept is very important in the study of nonlinear wave phenomena. Physically,
15
when two solitons of different amplitudes (and hence, of different speeds) are placed
far apart on the real line, the taller (faster) wave is on the left of the shorter (slower)
wave, the taller one eventually catches up to the shorter one and then overtakes it.
When this happens they undergo a nonlinear interaction according to the KdV
equation and emerge from the interaction completely preserved in form and speed
with only a phase shift. Thus these two remarkable features: (i) steady progressive
pulse like solitons and (ii) the preservation of their shapes and speeds confirmed the
particle like property of the waves.
Many exact solutions of the KdV equation with appropriate initial condition
by using scattering theory were given by Gardner et al. [39]. They discover the first
integrable nonlinear partial differential equation. Gardner et al. [40] and Hirota
[48,49] constructed analytic solutions of the KdV equation that provide the
description of the interaction among N solitons for any positive integer N . Certain
other methods for finding the exact solutions of the nonlinear PDEs include the
Painleve analysis [62], the Lie group theoretic methods [72], the direct algebraic
method [50] and tangent hyperbolic method [68].
Kruskal and Zabusky, and then Miura [65] considered the equation of the
existence of conservation laws of the KdV equation namely equations of the form
0
x
J
t
N, (2.1.4)
where ,.....),,( xxx uuuNN is the density and ,.....),,( xxx uuuJJ is the associated
flux. If 0J as x (or if it is periodic in space) then the total number Ndx
is conserved, because
0,.....),,(
dxuuuNdt
dxxx .
16
It was conjecture that the KdV equation had an infinite number of
conservation laws, and this was latter proved by Kruskal and Miura and
simultaneously by Gardner[40].
For, 3p it was found that the gKdV equation (2.1.2) had only three
conservation laws [7] and it is also found non-integrability [21]. Miura conjecture that
Eq. (2.1.3) has an infinite number of conservation laws and this might imply that the
KdV equation (2.1.1) and the mKdV Eq. (2.1.3) are related. He proceed to obtain the
transformation between them, namely if v satisfies (2.1.3)
02 xxxxt vvvvQv ,
and xvvu 2 , (2.1.5)
then ),( txu satisfies
0 xxxxt uuuuPu .
Hence, ),( txu satisfies the KdV equation (2.1.1). The transformation formula
(2.1.5) is known as Miura Transformation.
2.2 REVIEW OF THE NUMERICAL METHODS FOR SOLVING
gKdV EQUATION
Zabusky and Kruskal [99] were the pioneers in studying the KdV equation
numerically, by using the leapfrog method as an explicit finite difference scheme.
Tappert [92] has proposed a split-step Fourier method to solve the KdV equation
numerically. One of the most important numerical methods used to solve the KdV
equation was presented and exposed by Gereig and Morris [45]; this was the
hopscotch method. This hopscotch is an implicit finite difference method.
Fornberg and Whitham [37] gave an extensive study for the gKdV equation,
by using a pseudo-spectral method or what is called the finite Fourier transform.
17
Sanz-Serna and Christe [84] has applied a spectral method, namely the
Galerkin method. Taha and Ablowitz [93] in a series of papers have studied and
applied the famous and well known finite element method to solve the KdV equation.
They also gave an extensive comparison between the different numerical methods,
especially the finite difference and finite element methods. Although this is a wide
comparison, no one is able to say which is the best numerical method, for the
numerical solution of the KdV equation.
Chan and Kerkhoven [14] have discussed alternative time discretization of the
KdV equation. They showed that, with the leapfrog method for the advection term
and Crank-Nicolson method for the linear term, the stability limit is independent of
x (the space step) for any finite time interval.
Helal [53] has applied the tau method for obtaining a semi-analytical solution
for the KdV equation. He used the Chebyshev orthogonal polynomials as a basis for
this method. A numerical comparison showed that this method is more efficient than
the hopscotch method.
Abdur Rashid [78] has applied the Fourier Pseudospectral method for solving
the KdV equation accurately. He constructed the discrete representation of the
solution through interpolate trigonometric polynomial of the solution at collocation
points. Recently Rathish Kumar and Mani Mehra [79] have solved the KdV equation
by using Wavelet Galerkin method. They verified the asymptotic stability of the
proposed scheme.
Gardner et al. [41] has obtained the solutions of the Modified Korteweg-de
Vries (mKdV) equation by Galerkin method with quadratic B-spline finite elements.
The motion, interaction and generation of solitary waves are studied using the
18
method. T. Geyikli and D. Kaya [44] had used finite element technique for the
numerical solution of the mKdV equation.
Dogan Kaya and El-Sayed [59] had used Adomain Decomposition Method
(ADM) for the generalized KdV equation. They also proved the convergence of ADM
applied to the generalized KdV equation. The remarkable accuracy had shown by
comparing the numerical solution with the known analytic solution.
2.3 EXACT SOLUTION TO THE gKdV EQUATION
The simplest mathematical wave is a function of the form )(),( ctxftxu
which for example, is a solution to the simplest partial differential equation
0 xt cuu where c denotes the speed of the wave. For the well known wave
equation 02 xxtt ucu the famous d’Alembert solution leads to two wave fronts
represented by terms )( ctxf and )( ctxf .
Hence we start here with a trial solution
)()(),( fctxftxu . (2.3.1)
Substituting the trial solution (2.3.1) into the gKdV equation (2.1.2) we are led to the
ordinary differential equation
03
3
d
fd
d
dff
d
dfc p , (2.3.2)
which can written as
01 3
31
d
fdf
d
d
pd
dfc p .
Integrating, with respect to , it follows that
Ad
fdf
pcf p
2
21
1
,
where A is the constant of integration. In order to obtain a first order equation of f a
multiplication of d
df2 is done, i.e. :
19
d
dfA
d
fd
d
df
d
dff
pd
dfcf p 22
1
22
2
21
d
dfA
d
df
d
df
d
d
ppf
d
dc p 2
)2)(1(
22
22
Integrating both sides with respect to leads to
BAfd
dff
ppcf p
2
)2)(1(
22
22
, (2.3.3)
where B is another constant of integration.
Now it required that in case x we should have 0,0,02
2
d
fd
d
dff .
From these requirements it follows that 0 BA . With 0 BA , equation (2.3.3)
can be written as
pkfcf
d
df
22
, where )2)(1(
2
ppk
.
By separation of variables we may write
d
kfcf
df
p 21
.
Using the transformation, 2sechckf p , so that dhcdfkpf p 21 sec2 , we get
ddcp
2.
0
2x
cp
, where 0x is another constant of integration.
0
2
2sec
2
)2)(1(xctx
cph
ppcf p
p xctxcp
hppc
ctxftxu
0
2
2sec
2
)2)(1()(),(
(2.3.4)
20
This is the exact solution of the gKdV equation (2.1.2). Taking 1p and 2p in
equation (2.3.4) it will give the exact solutions of the KdV equation (2.1.1) and the
mKdV equation (2.1.3) respectively.
2.4 THE PETROV-GALERKIN METHOD
For convenience, the gKdV equation (2.1.2) is rewritten in the form
01
1
xxxx
p
t uup
u
. (2.4.1)
Periodic boundary conditions on the region bxa are assumed in the form
0),(),( tbutau . The space interval bxa is discretized by uniform )1( N
grid points jhax j , where Nj ,,2,1,0 and the grid spacing is given by
Nabh /)( . Let )(tU j denote the approximation to the exact solution ),( txu j .
Using Petrov-Galerkin method, we assume the approximate solution of Eq. (2.4.1) as
N
j
jjh xtUtxu0
)()(),( . (2.4.2)
The product approximation technique [16] is used for the nonlinear term in the
following manner
N
j
j
p
j
p
h xtUtxu0
11 )()(),( , (2.4.3)
where Njxj ,,2,1,0),( are the usual piecewise linear “hat” function given by
otherwise
xxxhjhx
xxxhjhx
x jj
jj
j
,0
],[,/)(1
],[,/)(1
)( 1
1
The unknown functions NjtU j ,...,2,1,0,)( are determined from the variational
formulation
0,)(,)(1
,)( 1
jxxxhjx
p
hjth uup
u
, (2.4.4)
where j , Nj ,,2,1,0 are test functions, which are taken to be cubic B-splines
given by
21
otherwise
xxxxx
xxxxxxxhxxhh
xxxxxxxhxxhh
xxxxx
hx
jjj
jjjjj
jjjjj
jjj
j
,0
,)(
,)(3)(3)(3
,)(3)(3)(3
,)(
1)(
21
3
2
1
3
1
2
11
23
1
3
1
2
11
23
12
3
2
3
and , denote the usual inner product b
a
dxxgxfgf )()(, .
Integrating by parts and using the fact that 0)()()()( baba xx , Eq.
(2.4.4) leads to the formulation
0)(,)(,)(1
,)( 1
xxjxhjx
p
hjth uup
u
. (2.4.5)
Performing the integrations on (2.4.5) will give the following system of ordinary
differential equations (ODEs)
0)1(1260
13
CBAhhp
, (2.4.6)
where 1112 266626 jjjjj UUUUUA ,
1
1
1
1
1
1
1
2 1010
p
j
p
j
p
j
p
j UUUUB
and 2112 22 jjjj UUUUC , .,....,3,2,1,0 Nj
Now to solve the ODEs, we assume n
jU to be a fully discrete approximation to the
exact solution ),,( nj txu where tntn and t is the time step size. Using the
Crank-Nicholson approach 21 nn UUU and the forward difference scheme for
the time derivative, tUUU nn
1 , Eqn. (2.4.6) is reduced to the system of
nonlinear equations
.)1(1230
1
)1(1230
1
3
1
3
11
nnn
nnn
hhp
t
hhp
t
CBA
CBA
(2.4.7)
The nonlinear system (2.4.7) can be solved by Newton’s method and the required
solution to the gKdV equation can be found.
22
2.5 STABILITY ANALYSIS
To apply the Von Neumann stability for the system (2.4.7), we must first
linearize this system. Assuming u in the nonlinear term x
puu of the gKdV equation
(2.1.2) as locally constant, u~ , the linearized scheme is
n
j
n
j
n
j
n
j
n
j
n
j
n
j
n
j
n
j
n
j
aUbUcUdUeU
eUdUcUbUaU
2112
1
2
1
1
11
1
1
2
(2.5.1)
where 312
~
30
1
h
t
h
uta
p
,
3
2
12
~10
30
26
h
t
h
utb
p
,
30
66c ,
3
2
12
~10
30
26
h
t
h
utd
p
,
312
~
30
1
h
t
h
ute
p
.
Substituting the Fourier mode
),exp( ijU nn
j hk and 1i , (2.5.2)
where k is the mode number and h the space step size into the linearized system
(2.5.1), we obtained the amplification factor, g as
iYX
iYXg
n
n
1
, (2.5.3)
where cos2cos dbeacX
and sin2sin dbeaY .
Thus, for all values of ,, th and we have
122
22
YX
YXggg .
Hence, the proposed method is unconditionally stable in the linear sense.
2.6 NUMERICAL TESTS
It has been shown that the gKdV equation has an analytic solution of the form
p xctxcp
hppc
txu
0
2
2sec
2
)2)(1(),(
,
23
where c and 0x are constants. It has mention that the gKdV equation has three
conservation laws. These are given below:
b
a
udxI1 , b
a
dxuI 2
2 and
b
a
x dxuuI 23
3
3
.
In this section, we present some numerical experiments to find the solution of single
solitary wave, in addition to the two soliton interactions at different time levels. Also
we present the birth of solitons from an initial pulse.
2.6.1 SINGLE SOLITARY WAVE
First we take 1p , so that the gKdV equation is the KdV equation and the
analytic solution takes the form:
0
2
2
1sec
3),( xctx
ch
ctxu
, (2.6.1)
where c and 0x are constant. The 2L and L error norms are used to compare the
numerical solutions with the exact solution and the quantities 1I , 2I and 3I are shown
to measure the conservation for the scheme. We choose )3/1(c , 2.0h , 1.0t ,
1 , 1 and 200 x over the domain ]80,0[ . Thus the solitary wave has
amplitude unity and the simulations are done up to 80t . Values of the three
invariants as well as 2L and L error norms from our method has been computed and
reported in Table 2.1. The chances of the invariants 1I , 2I and 3I from their initial
values are less than 0.00012, 0.00001 and 0.00005 respectively. Error deviations are
changed in the range 000062644.00000756628.0 error . The motion of a single
solitary wave using the proposed scheme corresponding to the above set of parameters
has been computed and plotted in Fig. 2.1.
24
Table 2.1: Invariants and error norms for single solitary wave 1p , 3/1c ,
2.0h , 1.0t , 1 , 200 x , 800 x .
t
1I 2I 3I 3
2 10L 310L
0
20
40
60
80
6.92814
6.92824
6.92825
6.92820
6.92823
4.6188
4.6188
4.6188
4.6188
4.6188
2.77128
2.77129
2.77129
2.77129
2.77126
0
0.203188
0.255819
0.327420
0.404813
0
0.342689
0.472876
0.581553
0.756628
Fig. 2.1: Single solitary wave with 1p , 3/1c , 200 x , 800 x at level
time 80,40,0t
Fig. 2.2: Single solitary wave with 2p , 2/1c , 200 x , 800 x at level
time 80,40,0t .
25
Fig. 2.3: Single solitary wave with 3p , 5/1c , 200 x , 800 x at level
time 80,40,0t .
Table 2.1: Invariants and error norms for single solitary wave 1p , 3/1c ,
2.0h , 1.0t , 1 , 200 x , 800 x .
t 1I 2I 3I 3
2 10L 310L
0
10
20
30
40
50
60
70
80
6.92814
6.92823
6.92824
6.92825
6.92825
6.92821
6.92820
6.92820
6.92823
4.6188
4.6188
4.6188
4.6188
4.6188
4.6188
4.6188
4.6188
4.6188
2.77128
2.77129
2.77129
2.77127
2.77129
2.77128
2.77129
2.77127
2.77126
0
0.173189
0.203188
0.229209
0.255819
0.304573
0.32742
0.38525
0.404813
0
0.238921
0.342689
0.311464
0.472876
0.541176
0.581553
0.766973
0.756628
Secondly, we take 2p , so that the gKdV equation becomes the modified
KdV equation and the equation has the analytic solution as
0sec
6),( xctx
ch
ctxu
. (2.6.2)
In this case, we consider the following parameters for the simulation of the problem:
)2/1(c , 2.0h , 1.0t , 3 , 1 and 200 x over the domain ]80,0[ . Thus
the solitary wave has amplitude unity and the simulations are done up to 80t .
26
Values of the three invariants as well as 2L and L error norms from our method has
been computed and reported in Table 2.2. The chances of the invariants 1I , 2I and 3I
from their initial values are less than 0.00005, 0.00002 and 0.00003 respectively.
Error deviations are changed in the range of 00122617.000122794.0 error .
The motion of a single solitary wave using the proposed scheme corresponding to the
above set of parameters has been computed and plotted in Fig. 2.2.
Finally, we take 3p and consider the following set of parameters for the
simulation of the problem. We take, )5/1(c , 2.0h , 1.0t , 2 , 1 and
200 x for the simulation of the problem over the domain ]80,0[ . Thus the solitary
wave has amplitude unity and the simulations are done up to 80t . Values of the
three invariants as well as 2L and L error norms from our method has been
computed and reported in Table 2.3. The chances of the invariants 1I , 2I and 3I from
their initial values are less than 0.00014, 0.00003 and 0.00009 respectively. Error
deviations are changed in the range of 000116961.000016448.0 error . The
motion of a single solitary wave using the proposed scheme corresponding to the
above set of parameters has been computed and plotted in Fig. 2.3.
Table 2.2: Invariants and error norms for single solitary wave 2p , 2/1c ,
2.0h , 1.0t , 3 , 1 , 200 x , 800 x .
t 1I 2I 3I 3
2 10L 310L
0
10
20
30
40
50
60
70
80
4.44288
4.44288
4.44289
4.44292
4.44292
4.44291
4.44289
4.44286
4.44286
2.8284
2.8281
2.8281
2.8281
2.8281
2.8281
2.8281
2.8281
2.8281
1.74998
1.74997
1.74997
1.74997
1.74997
1.74997
1.74997
1.74996
1.74997
0
0.710642
1.35740
2.01775
2.68235
3.35304
4.00643
4.67266
5.33633
0
0.165455
0.311979
0.465156
0.618102
0.774527
0.920563
1.07373
1.22794
27
Table 2.3: Invariants and error norms for single solitary wave 3p , 5/1c ,
2.0h , 1.0t , 3 , 1 , 200 x , 800 x .
t 1I 2I 3I 3
2 10L 310L
0
10
20
30
40
50
60
70
80
6.27083
6.27083
6.27096
6.27101
6.27100
6.27088
6.27098
6.27084
6.27030
3.85663
3.85663
3.85663
3.85663
3.85663
3.85663
3.85664
3.85663
3.85663
2.48559
2.48557
2.48551
2.48556
2.48563
2.48558
2.48555
2.48552
2.48551
0
0.958627
0.973995
0.957792
0.957425
0.980087
0.957492
0.965510
0.971789
0
0.169812
0.149183
0.144506
0.151097
0.141056
0.140572
0.140047
0.116961
2.6.2 INTERACTION OF TWO gKdV SOLITARY WAVES
Here we study the interaction of two well separated solitary waves having
different amplitudes and travelling in the same direction. The initial condition is given
by
2
1
2
2sec
2
)2)(1()0,(
i
pi
ii xxcp
hppc
xu
, (2.6.3)
where ic and 2,1, ixi are arbitrary constants. In this case also we take
32,1 andp . For all values of p , the following parameters:
1,1.0,2.0 th have been chosen with range ]160,0[ .
28
Fig. 2.4 (a) Fig. 2.4 (b)
Fig. 2.4(c) Fig. 2.4(d)
Fig. 2.4(e)
Fig. 2.4 (a)-(d): Interaction of two gKdV solitary waves at different time levels and
Fig. 2.4 (e) three dimensional plot for .1p
29
First we take 1p , so that the gKdV equation (2.1.2) becomes the KdV
equation and the initial condition (2.6.3) takes the form
i
i
i
i xxc
hc
xu 2
1sec
3)0,( 2
2
1
(2.6.4)
Here, we chose the following parameters for our simulation: 1 , ,3/11 c
6/12 c , 201 x and 602 x . Then the amplitudes of the two solitary waves are in
the ratio 1:2 . The simulation is done up to the time t 320, The three invariant of
motion are tabulated in Table 2.4 and Fig. 2.4(a)-2.4(d) show the interaction of the
two solitons at different time and Fig. 2.4(e) shows its three dimensional plot.
Table 2.4: Invariants for two solitary wave interaction for 1p .
t 1I 2I 3I
0
40
80
120
160
200
240
280
320
11.8271
11.8272
11.8272
11.8272
11.8272
11.8271
11.8272
11.8272
11.8272
6.25179
6.25180
6.25180
6.25180
6.25181
6.25187
6.25182
6.25180
6.25180
3.26119
3.26116
3.26116
3.26117
3.26123
3.26123
3.26122
3.26119
3.26117
Secondly, we consider the case for 2p . In this case the initial condition
(2.22) takes the form:
i
i
i
i xxc
hc
xu
sec6
)0,(2
1
. (2.6.5)
Here, we chose the following parameters for our simulation: 1,3 , ,2/11 c
8/12 c , 201 x and 602 x . The amplitudes of the two solitary waves are in the
ratio 2:1. The simulation is done up to the time t 200, The three invariant of motion
30
are tabulated in Table 2.5 and Fig. 2.5(a)-2.5(d) show the interaction of the two
solitons at different time and Fig. 2.5(e) shows its three dimensional plot.
.
Fig. 2.5(a) Fig. 2.5(b)
Fig. 2.5(c) Fig. 2.5(d)
Fig. 2.5(e)
Fig. 2.5 (a)-(d): Interaction of two gKdV solitary waves at different time levels and
Fig. 2.5 (e): three dimensional plots for .2p
31
Fig. 2.6(a) Fig. 2.6(b)
Fig. 2.6(c) Fig. 2.6 (d)
Fig. 2.6 (e)
Fig. 2.6 (a)-(d): Interaction of two gKdV solitary waves at different time levels and
Fig. 2.6 (e) three dimensional plots for .3p
32
Finally, we take 3p . In this case the initial condition (2.6.3) takes the form:
2
1
32
2
3sec
10)0,(
i
iii xx
ch
cxu
. (2.6.6)
Here, we chose the following parameters for our simulation: 1,2 , ,5/21 c
10/12 c , 201 x and 602 x . The amplitudes of the two solitary waves are in the
ratio 2:1. The simulation is done up to time 320t , The three invariant of motion are
tabulated in Table 2.6 and Fig. 2.6(a)-2.6(d) show the interaction of the two solitons at
different time and Fig. 2.6(e) shows its three dimensional plot.
Table 2.5: Invariants for two solitary wave interaction for 2p .
t 1I 2I 3I
0
40
80
120
160
200
8.88576
8.88581
8.88574
8.88571
8.88567
8.88568
4.24263
4.24263
4.24276
4.24267
4.24263
4.24263
2.24646
2.24611
2.24832
2.24875
2.24544
2.24646
Table 2.6: Invariants for two solitary wave interaction for 3p .
t 1I 2I 3I
0
40
80
120
160
200
240
280
12.6253
12.6252
12.6253
12.6253
12.6253
12.6273
12.6200
12.6219
7.76476
7.76476
7.76476
7.76512
7.76488
7.76475
7.76480
7.76466
4.99077
4.99013
4.99077
4.99022
4.99203
4.99665
4.99257
4.99264
33
2.6.3 SPLITTING OF SOLITONS FROM A SINGLE INITIAL PULSE
Here we study the splitting of solitons from a single initial pulse. In this case
also we consider three cases viz. 1p , 2p and 3p .
First we take 1p . The initial condition in this case takes the form:
108
1sec
3
2)( 2 x
hxf (2.6.7)
The parameter chosen are: 1 , 410 , 01.0 th and the movement
of the initial pulse is consider over the interval ]3,0[ . This simulation is performed up
to the time step 3t . The three invariants of motion 1I , 2I and 3I are tabulated in
Table 2.7. Fig. 2.7(a) and 2.7(b) show the splitting of the soliton at time 0t and at
time 3t respectively. Fig. 2.7(c) shows the three dimensional plot of this case.
Fig. 2.7(a) Fig. 2.7(b)
Fig. 2.7(c)
Fig. 2.7 (a) and (b): Splitting of gKdV solitary waves from a single initial pulse and
Fig. 2.7 (c) its three dimensional plot for .1p
34
Secondly we take 2p and consider the same initial condition as given in
equation (2.6.7) for the case 1p . The parameters chosen are: 1 , 410 ,
01.0 th and the movement of the initial pulse is consider over the interval ]3,0[ .
This simulation is performed up to the time step 8t . The three invariants of motion
1I , 2I and 3I are tabulated in Table 2.8. Fig. 2.8(a)-(c) showed the splitting of the
soliton from the initial pulse at different time 4,0t and 8 respectively. Also, Fig.
2.8(d) shows the three dimensional plot of this case.
Fig. 2.8(a) Fig. 2.8(b)
Fig. 2.8(c) Fig. 2.8(d)
Fig. 2.8 (a)-(c): Splitting of gKdV solitary waves from a single initial pulse and Fig.
2.8 (d) its three dimensional plot for .2p
35
Table 2.7: Invariance for the splitting of gKdV solitary waves for .1p
t 1I 2I 3I
0
1
2
3
4
0.138564
0.138565
0.138564
0.138564
0.138563
0.0615837
0.0614930
0.0649040
0.0614902
0.0614901
0.0314758
0.0314142
0.0313945
0.0314100
0.0313999
Table 2.8: Invariance for the splitting of gKdV solitary waves for .2p
t 1I 2I 3I
0
2
4
6
8
0.217647
0.217649
0.217638
0.217720
0.217641
0.092376
0.092318
0.092312
0.092312
0.092311
0.047513
0.048489
0.049179
0.049464
0.049597
Table 2.9: Invariance for the splitting of gKdV solitary waves for .3p
t 1I 2I 3I
0
2
4
6
8
0.217647
0.217655
0.217651
0.217705
0.217609
0.092376
0.092355
0.092325
0.092331
0.092341
0.0475125
0.0493923
0.0505958
0.0507069
0.0507019
Lastly, we consider the case .3p In this case also we consider the same
initial condition as above. The parameters chosen are: 1 , 410 , 01.0 th
and the movement of the initial pulse is consider over the interval ]3,0[ . This
simulation is performed up to the time step 8t . The three invariants of motion 1I ,
2I and 3I are tabulated in Table 2.9. Fig. 2.9(a)-(c) showed the splitting of the soliton
from the initial pulse at different time 4,0t and 8 respectively. Also, Fig. 2.9(d)
36
shows the three dimensional plot of this case. It is observed from the Tables 2.7-2.9
that the third invariance,
b
a
x dxuuI 23
3
3
is little fluctuated in all the cases.
Fig. 2.9(a) Fig. 2.9(b)
Fig. 2.9(c) Fig. 2.9(d)
Fig. 2.9 (a)-(c): Splitting of gKdV solitary waves from a single initial pulse and
Fig. 2.9 (d): its three dimensional plot for .3p
2.7 CONCLUSION
The Petrov-Galerkin method using the linear hat function and cubic B-spline-
function as trial and test functions has been successfully implemented to study the
solitary waves of the gKdV equation. It has been shown that the scheme is accurate
37
and efficient. It has also been shown that the scheme so developed is unconditionally
stable. We have tested our scheme using single solitary wave for which the exact
solution is known and then extended this scheme to the study of the interaction of two
solitary waves and also breaking of the solitary waves from an initial pulse. It is also
observed that the solution of the gKdV equation does not give true soliton solution
when 3p . We have also shown the appearance of small tail after the collision of
two solitons in the case of 3p . Thus we require further research work in this area
through both analytical and numerical solution of this equation type.