t o o | o o | o
(3)
Stable manifolds ( ) and unstable manifolds ( ) of homoclinic orbits are indicated in Fig. 2.
Periodic orbits are nested within and outside the homoclinic orbits. For , 0 = c the stable and unstable branches
of homoclinic orbits join smoothly. When the dissipative perturbation is included the stable manifold
s
W and
unstable manifold
u
W do not join. However, above certain critical amplitude of the external periodic force
transverse intersections of
s
W and
u
W occurs. The presence of such intersections implies that the Poincar map
Prediction of Horseshoe Chaos in Duffing-Van
19
Figure 1: Wave form of various periodic forces (a) sine wave (b) square wave (c) rectified sine wave (d)
symmetric saw-tooth wave (e) asymmetric saw-tooth wave and (f) modulus of sine wave. For all the forces
period is 2/, = 1 and amplitude f is 1.
has the so-called horseshoe chaos [8, 30]. Eventhough the orbits created by the horseshoe mechanism
are unstable they can exert a dramatic influence on the behavior of orbits which pass close to the point
of intersection. The appearance of transverse intersections of homoclinic orbits can be predicted analytically by
the Melnikov technique. In order to apply the Melnikov technique, in general, we rewrite the given equation of
motion into the following standard form
) , , (
1
) , (
1
t y x g y x f x c + =
(4a)
) , , (
2
) , (
2
t y x g y x f y c + =
(4b)
where
1
g and
2
g are periodic in t with period T, the Melnikov function is given by [30]
}
+
}
. =
T
d ds S
h
X f
X
D trace t
h
X g
h
X f
o
t M
0
] )) ( ( [ exp[ ) ), ( ( )) ( ( ) ( t t t
(5)
where ) , (
h h h
y x X = represents homoclinic orbits of the unperturbed systems,
1 2 2 1
g f g f g f = . and
X
D denotes the partial derivatives with respect to ). , ( y x X =
Figure 2: Phase trajectories of the unperturbed system. The stable ( W
s
) parts of
homoclinic orbits connecting saddle to itself are indicated. The analytical expression for the homoclinic orbits
is given byEq. (3).
For the DVP Eq. (2) the Melnikov function is
. )]
0
( )
2
1 ( [ )
0
( t + t +
}
+
= d t F
h
x
h
py
h
y t M
(6)
In the following we calculate the Melnikov function for the system (2) with different periodic forces.
Sine wave
For the system (2) driven by the force , sin ) (
sin
t f t F e = the Melnikov integral is worked out as
0
cos
sin
)
0
(
sin
t B f A t M e =
(7a)
where
|
.
|
\
|
(
(
o
te
te
|
=
|
o
|
o
=
2
sec
2
,
5
2
4
1
3
3
4
h B
p
A
(7b)
Prediction of Horseshoe Chaos in Duffing-Van
20
Square wave
For the square wave ) sgn(sin ) 2 ( ) ( t
sq
f t
sq
F t F e = e t + = where ) sgn( y is sign of y . The Fourier series
is
=
e
t
=
1
) 1 2 (
) 1 2 sin( 4
) (
n
n
t n f
t
sq
F
(8a)
Using its Fourier series we obtain
=
=
1
0
) 1 2 cos( )
0
(
n
t n
n
B
sq
f A t
Msq
e
(8b)
where
(
o
e t
|
e
=
2
) 1 2 (
sec
2 4 n
h
n
B
(8c)
Symmetric saw-tooth wave
The numerical implementation of the force is given by
e t s s e t
e t s s e t
e t s s
, 2 2 3 , 4
4
, 2 3 2 ,
4
, 2 0 ,
4
t
sst
f
T
t
sst
f
t
T
t
sst
f
t
T
t
sst
f
(9a)
where T is the period of the force and t is taken as mod . 2 e t Its Fourier series
.
1
2
) 1 2 (
) 1 2 sin(
) 1 (
) 1 (
2
8
) (
sst
=
e
+
t
=
n
n
t n
n
f
t F
(9b)
The Melnikov function is
=
e =
1
0
) 1 2 cos( )
0
(
n
t n
n
B
sst
f A t
sst
M
(9c)
where
(
o
e t
t
e
=
2
) 1 2 (
sec
) 1 2 (
1
) 1 ( 2 8 n
h
n
n
B
n
B
(9d)
Asymmetric saw-tooth wave
For the asymmetric saw-tooth wave,
|
.
|
\
|
s s
s s
= + =
, 2 , 2
2
, 0 ,
2
2
) (
ast
e t e t
e t
e
t
t
ast
f
T
t
ast
f
t
T
t
ast
f
t
ast
F t F
(10a)
where t is taken as mod 2 / . The Fourier series of ) (t F
ast
is given by
= ) (
sst
t F
Prediction of Horseshoe Chaos in Duffing-Van
21
=
e
+
t
=
1
sin
1
) 1 ( 2
) (
n n
t n
n
f
t
ast
F
(10b)
We obtain
=
e =
1
0
cos )
0
(
n
t n
n
B
ast
f A t
ast
M
(10c)
where
(
o
te
+
|
e
=
2
sec
1
) 1 (
2 2 n
h
n
n
B
(10d)
Rectified sine wave
For the rectified sine wave
= e t + = ) 2 ( ) ( t F t F
rec rec
e t s s e t
e t s s e
2 , 0
0 , sin
t
t t f
rec
(11a)
The Fourier series of rectified sine wave is
=
e
e
t
e
+
te
=
1
cos
2 2
1 2
) (
n
t n
n
rec
f
rec
f
t
rec
F
11b)
for e t < < t 0 and 0 for e t < < e t 2 t
The Melnikov function is
=
e =
1
0
sin 2 )
0
(
n
t n
n
C
n
B
rec
f A t
rec
M
(11c)
where
)
2 2
(
1 2
2
2
n
n
B
e
| t
eo
=
11d)
and
}
e t
e t
t et ot ot =
) 1 2 (
) 2 2 (
sin tanh sec
n
n
d n h C
n
(11e)
n
C can be evaluated numerically.
Modulus of sine wave
For the modulus of sine wave
) 2 sin( ) 2 ( ) ( t
msw
f t
msw
F t
msw
F e = e t + =
(12a)
Its Fourier series is
=
e
t
t
=
1
) 1
2
4 (
cos 4 2
) (
n
n
t n n f f
t
msw
F
(12b)
Melnikov function is
|
|
.
|
\
| e
=
2
0
cos )
0
(
t
B
msw
f A t M
msw
(12c)
where
|
.
|
\
|
o
te
|
te
=
4
sec
2
h B
(12d)
Prediction of Horseshoe Chaos in Duffing-Van
22
III. Effect of Shape Of Periodic Forces On Horseshoe Chaos
In this section we compute the Melnikov threshold values for the onset of horseshoe chaos and
compare it with numerical predictions. For the sine wave, the condition for transverse intersection of stable and
and unstable manifolds is
|
.
|
\
|
o
te
| te
|
|
.
|
\
|
|
o
o
= >
2
cosh
3
5
2
4
1
3
2 2
sin
p
m
f f
(13)
Eq. (13) is the necessary condition for the occurrence of horseshoe chaos. The sufficient condition
requires the existence of simple zeros of ). (
0
t M Equality sign in Eq. (13) corresponds to tangential intersection.
For other forces except sine and modulus of sine waves we can not write the sufficient condition for the
existence of simple zeros of ) (
0
t M as like Eq. (13) since ) (
0
t M is a convergent series. However, the
occurrence of homoclinic bifurcation can be studied by measuring the time
M
elapsed between two successive
zeros of ) (
0
t M . We consider sufficiently large number of terms say, 50 terms in the summation of the equation
for ) (
0
t M . For a fixed value f ,
0
t is varied from 0 to 200T where
e
t
=
2
T is period of the external force.
If the sign of ) (
0
t M remains same in this time interval then there is no zero of ) (
0
t M and
M
is assumed as
Figure 3: f
M
versus n, the number of terms in the summation in Eq. (8b ) for = 1,
=5.0, p = 0.4 and = 1 when the system (1) is driven by the square wave force. Variation in f
M
converges
to a constant value with increase in n.
infinity.
M
is calculated for a range of amplitude of the external force . The value of f , at which first time
) (
0
t M changes sign and thereby giving finite
M
is the Melnikov threshold value for homoclinic
bifurcation.Figure 3 shows the plot of
m
f versus n, the number of terms in the summation in Eq. (8b) for the
square wave force.
m
f converses to a constant value with increase in n. For n > 10, the variation in
m
f is
negligible. Similar result is found for the other forces also. In our numerical calculation of
m
f we fix n =50.
Figure 4 shows the variation of 1/
M
versus f for the system (2) diven by different periodic forces. Solid curve
represents the inverse of first intersection time, 1/
M
of stable and unstable branches of homoclinic orbits
+
W .
Dashed curve corresponds to the orbit