Вы находитесь на странице: 1из 9

Solutions to the Van der Pol Equation

Stephen W Mason

Summer 2017

We determine to find a solution to the Van der Pol Equation:

𝑥̈ − 𝜀 (1 − 𝑥 2 )𝑥̇ + 𝑥 = 0 (1)

For this problem, we use Matrices to find a general solution of the form:
𝑥
𝑋 = 𝑃 −1 𝐴𝑃 [𝑦] + 𝑃 −1 𝑔(𝑡) (2)

So, put 𝑥̇ = 𝑦, then 𝑦̇ = −𝑥 + 𝜀𝑦 − 𝜀𝑥 2 𝑦, yielding:

𝑥̇ 0 1 𝑥 0
𝑋=[ ]=[ ] [𝑦 ] + [ ] (3)
𝑦̇ −1 𝜀 −𝜀𝑥 2 𝑦

where we see that, for periodic motion, by substituting 𝑥 = 𝐴 cos(𝛽𝑡) and 𝑦 =


𝐴 sin(𝛽𝑡), we have:
0 0
𝑔 (𝑡 ) = [ 2 ]=[ 3 2( ) ] (4)
−𝜀𝑥 𝑦 −𝜀𝐴 cos 𝛽𝑡 sin(𝛽𝑡)

Let:

0 1
𝐴=[ ] (5)
−1 𝜀
Then:

|𝜆𝐼 − 𝐴| = [𝜆 −1 ] = [0] (6)


1 𝜆−𝜀 0
which yields the characteristic equation:

𝜆 (𝜆 − 𝜀 ) + 1 = 0 (7)

or:

𝜆2 − 𝜀𝜆 + 1 = 0 (8)

for which the eigenvalues are:


𝜀±√𝜀 2 −4 𝜀±𝑖√4−𝜀 2
𝜆= = (9)
2 2

There are at least three pairs of eigenvalues, and at least three cases to consider:

(i) 𝜀2 < 4 which yields complex eigenvalues, 𝜆1 and 𝜆2 ⟹ Stable Spiral

Page | 1
(ii) 𝜀2 = 4 which yields real and equal eigenvalues, 𝜆1 and 𝜆2 ⟹ Stable Node
(iii) 𝜀2 > 4 which yields real and different eigenvalues, 𝜆1 and 𝜆2 ⟹ Unstable
Node or Spiral

In case (iii), both eigenvalues may be negative, positive, or they may be of opposite sign,
which implies further non-linear dynamical factors.

Consider case (i), in which the eigenvalues are complex, thus signifying periodicity in
the solution. Then:
𝜀+𝛽𝑖 𝜀−𝛽𝑖
𝜆1 = and 𝜆2 = ( 10 )
2 2

where:

𝛽2 = 4 − 𝜀 2 ( 11 )

𝜀+𝛽𝑖
So, substituting in 𝜆1 = to the matrix into equation (4), we have:
2

𝜀+𝛽𝑖
−1
2
𝐵1 = [ −(𝜀−𝛽𝑖)
]=0 ( 12 )
1
2

which is of the form of the simultaneous equations:


(𝜀+𝛽𝑖)
𝑘1 − 𝑘2 = 0 ( 13 )
2

and:
(𝜀−𝛽𝑖)
𝑘1 − 𝑘2 = 0 ( 14 )
2

Since equations (13) and (14) are the complex conjugates of each other, we only need
to analyse equation (14), say, so we have:
(𝜀−𝛽𝑖)
𝑘1 − 𝑘2 = 0 ( 15 )
2

where we see that:


1
𝑘1 = (𝜀 − 𝛽𝑖 ) ( 16 )
2

and:

𝑘2 = 1 ( 17 )

which thus yields an eigenvector of:


1
(𝜀 − 𝛽𝑖 )
𝑣1 = [2 ] ( 18 )
1
and a solution vector of:

Page | 2
1 𝜀+𝛽𝑖
(𝜀 − 𝛽𝑖 )
𝑋1 = 𝐶1 [2 ] 𝑒( 2
)𝑡
( 19 )
1
𝜀−𝛽𝑖
Substituting in the eigenvalue of 𝜆2 = , we now get:
2

𝜀−𝛽𝑖
−1
2
𝐵1 = [ −(𝜀+𝛽𝑖)
]=0 ( 20 )
1 2

so that:
1
(𝜀 + 𝛽𝑖 )
𝑣2 = [2 ] ( 21 )
1
and:
1 𝜀−𝛽𝑖
(𝜀 + 𝛽𝑖 )
𝑋2 = 𝐶2 [2 ] 𝑒( 2
)𝑡
( 22 )
1

[Alternative solution vectors are:


1 𝜀+𝛽𝑖
𝑋1 = [ 𝜀 + 𝛽𝑖 ] 𝑒 ( 2 )𝑡
2
and:
1 𝜀−𝛽𝑖
( )𝑡
𝑋2 = [𝜀 − 𝛽𝑖 ] 𝑒 2
2
as verified by Mathematica]

Thus, a general complementary linearised solution is:


1 𝜀−𝛽𝑖 1 𝜀+𝛽𝑖
(𝜀 + 𝛽𝑖 ) (𝜀 − 𝛽𝑖 )
𝑋𝑐 = 𝐶1 [2 ] 𝑒( 2
)𝑡
+ 𝐶2 [2 ] 𝑒( 2
)𝑡
( 23 )
1 1
or:
1 1
𝜀
(𝜀 + 𝛽𝑖 ) 𝛽𝑖
( ) 𝛽𝑖
𝑋𝑐 = 𝐶1 𝑒 2𝑡 [2 ] 𝑒 − 2 𝑡 + 𝐶2 [2 𝜀 − 𝛽𝑖 ] 𝑒 2 𝑡 ( 24 )
1 1
which, if one converts this to the polar form using the relationship:
1
𝜀
𝛽𝑡 𝛽𝑡 (𝜀 − 𝛽𝑖 )
𝑋𝑐 = 𝑒 2𝑡 (cos ( 2 ) + 𝑖 sin ( 2 )) [2 ] ( 25 )
1
𝛽
where we only need to use the part of the eigenvalue 𝑟 = 2 𝑖, then the linearised
solution may be written as:

Page | 3
𝜀 𝛽𝑡 𝛽 𝛽𝑡 𝜀 𝛽𝑡 𝛽 𝛽𝑡
cos ( 2 ) + 2 sin ( 2 )
𝜀 𝜀 sin ( 2 ) − 2 cos ( 2 )
𝑡 2 𝑡 2
𝑋𝑐 = 𝐶1 𝑒 [ 2
𝛽𝑡
] + 𝐶2 𝑒 2 [
𝛽𝑡
] ( 26 )
cos ( 2 ) sin ( 2 )

Thus, a fundamental matrix is:


𝜀 𝜀
𝜀 𝛽𝑡 𝛽 𝛽𝑡 𝜀 𝛽𝑡 𝛽 𝛽𝑡
𝑒 2𝑡 ( cos ( ) + sin ( )) 𝑒 2𝑡 ( sin ( ) − cos ( ))
2 2 2 2 2 2 2 2
𝑃=[ 𝜀
𝛽𝑡 𝜀
𝛽𝑡
] ( 27 )
𝑡 𝑡
𝑒 2 cos ( 2 ) 𝑒 2 sin ( 2 )

The inverse of 𝑃 is 𝑃−1 , where:


𝜀 𝜀
𝛽𝑡 𝛽 𝛽𝑡 𝜀 𝛽𝑡
2 −𝜀𝑡
𝑒 2𝑡 sin ( 2 )
𝑒 2𝑡 (2 cos ( 2 ) − 2 sin ( 2 ))
𝑃−1 = 𝛽𝑒 [ 𝜀 𝛽𝑡 𝜀
𝜀 𝛽𝑡 𝛽 𝛽𝑡
] ( 28 )
−𝑒 2𝑡 cos ( 2 ) 𝑒 2𝑡 (2 cos ( 2 ) + 2 sin ( 2 ))

so that:
𝜀 𝜀
2 𝛽𝑡 𝛽𝑡 𝜀 𝛽𝑡
𝛽
𝑒 −2𝑡 sin ( 2 )
𝑒 −2𝑡 (cos ( 2 ) − 𝛽 sin ( 2 ))
𝑃 −1 =[ 2 𝜀 𝛽𝑡 𝜀
𝜀 𝛽𝑡 𝛽𝑡
] ( 29 )
− 𝛽 𝑒 −2𝑡 cos ( 2 ) 𝑒 −2𝑡 (𝛽 cos ( 2 ) + sin ( 2 ))

To find the particular-solution, we need to first determine 𝑃 −1 𝑔, so set this equal to


𝑢̇ , hence:
𝜀 𝜀
2 𝛽𝑡 𝛽𝑡 𝜀 𝛽𝑡
𝑒 −2𝑡 sin ( 2 ) 𝑒 −2𝑡 (cos ( 2 ) − 𝛽 sin ( 2 )) 0
−1 𝛽
𝑢̇ = 𝑃 (𝑡)𝑔(𝑡) = [ ][ ] ( 30 )
2 − 𝑡
𝜀
𝛽𝑡 − 𝑡
𝜀
𝜀 𝛽𝑡 𝛽𝑡 −𝜀𝐴3 cos 2 (𝑡) sin(𝑡)
−𝛽 𝑒 2 cos ( 2 ) 𝑒 2 (𝛽 cos ( 2 ) + sin ( 2 ))

or:

𝜀 𝜀
𝛽𝑡 𝛽𝑡 𝜀2 𝛽𝑡 𝛽𝑡
𝐴3 (−𝜀𝑒 −2𝑡 cos 3 ( 2 ) sin ( 2 ) + 𝑒 −2𝑡 cos 2 ( 2 ) sin2 ( 2 ))
𝛽
𝑢̇ = [ 𝜀2 𝜀
𝛽𝑡 𝛽𝑡 𝜀
𝛽𝑡 𝛽𝑡
] ( 31 )
3 − 𝑡 − 𝑡
𝐴 (− 𝑒 2 cos 3 ( ) sin ( ) − 𝜀𝑒 2 cos 2 ( ) sin2 ( ))
𝛽 2 2 2 2

On integrating this with respect to time, 𝑡, we obtain:


𝜖 𝛽 3𝛽 5𝛽 5𝛽 7𝛽
− 𝑡 2 2 (3𝛽2 +𝜖2 )cos( 5𝛽2 cos( 𝜖2 cos( 7𝛽2 cos(
𝜖𝐴3 ⅇ 2 (𝛽 −𝜖 )cos( 2 𝑡) 2
𝑡)
2
𝑡)
2
𝑡)
2
𝑡)
[ + + − + +
4𝛽 𝛽2 +𝜖2 9𝛽2 +𝜖2 25𝛽 +𝜖2
2 25𝛽 +𝜖2
2 49𝛽 +𝜖2
2

7𝛽 𝛽 3𝛽 5𝛽 7𝛽
𝜖2 cos( 𝑡) 2𝛽𝜖sin( 𝑡) 2𝛽𝜖sin( 𝑡) 6𝛽𝜖sin( 𝑡) 6𝛽𝜖sin( 𝑡)
2 2 2 2 2
+ − + − ]
49𝛽2 +𝜖2 𝛽2 +𝜖2 9𝛽2 +𝜖2 25𝛽2 +𝜖2 49𝛽2 +𝜖2

𝜖 𝛽 𝛽 𝛽 𝛽
− 𝑡 2𝛽(𝜖cos( 𝑡)−𝛽sin( 𝑡)) 2𝜖(𝛽cos( 𝑡)+𝜖sin( 𝑡)) 3𝛽 3𝛽
𝜖𝐴3 ⅇ 2 2 2 2 2 𝛽(2𝜖 cos(
2
𝑡)−6𝛽 sin( 𝑡))
2
[ + − +
𝑢= 8𝛽 𝛽2 +𝜖2 𝛽2 +𝜖2 2
9𝛽 +𝜖 2
( 32 )

5𝛽 5𝛽
3𝛽 3𝛽 2𝛽(𝜖cos( 𝑡)−5𝛽sin( 𝑡)) 5𝛽 5𝛽
2𝜖(3𝛽cos( 𝑡)+𝜖sin( 𝑡)) 2 2 2𝜖(5𝛽cos( 𝑡)+𝜖sin( 𝑡))
2 2 2 2
2 2 + + −
9𝛽 +𝜖 25𝛽2 +𝜖2 2
25𝛽 +𝜖 2

7𝛽 7𝛽 7𝛽 7𝛽
2𝛽(𝜖cos( 𝑡)−7𝛽sin( 𝑡)) 2𝜖(7𝛽cos( 𝑡)+𝜖sin( 𝑡))
2 2 2 2
+ ]
49𝛽2 +𝜖2 49𝛽2 +𝜖2
[ ]

Page | 4
Since the particular-solution is determined by calculating the relationship 𝑋𝑝 = 𝑃𝑢,
then we have:
𝜀 𝜀
𝜀 𝛽𝑡 𝛽 𝛽𝑡 𝛽 𝛽𝑡 𝜀 𝛽𝑡
𝑒 2𝑡 (2 cos ( 2 ) − 2 sin ( 2 )) 𝑒 2𝑡 ( 2 cos ( 2 ) + 2 sin ( 2 ))
𝑋𝑝 = [ 𝜀
𝛽𝑡 𝜀
𝛽𝑡

𝑒 2𝑡 cos ( 2 ) 𝑒 2𝑡 sin ( 2 )
𝜖 𝛽 3𝛽 5𝛽 5𝛽 7𝛽
− 𝑡 2 2 (3𝛽2 +𝜖2 )cos( 5𝛽2 cos( 𝜖2 cos( 7𝛽2 cos(
𝜖𝐴3 ⅇ 2 (𝛽 −𝜖 )cos( 2 𝑡) 2
𝑡)
2
𝑡)
2
𝑡)
2
𝑡)
[ + + − + +
4𝛽 𝛽 2 +𝜖2 9𝛽2 +𝜖2 25𝛽2 +𝜖2 25𝛽2 +𝜖2 49𝛽 2 +𝜖2
7𝛽 𝛽 3𝛽 5𝛽 7𝛽
𝜖2 cos( 𝑡) 2𝛽𝜖sin( 𝑡) 2𝛽𝜖sin( 𝑡) 6𝛽𝜖sin( 𝑡) 6𝛽𝜖sin( 𝑡)
2 2 2 2 2
2 2 + − + − ]
49𝛽 +𝜖 𝛽2 +𝜖2 2
9𝛽 +𝜖 2 2
25𝛽 +𝜖 2 2
49𝛽 +𝜖 2

− 𝑡
𝜖 𝛽
2𝛽(𝜖cos( 𝑡)−𝛽sin( 𝑡))
𝛽 𝛽
2𝜖(𝛽cos( 𝑡)+𝜖sin( 𝑡))
𝛽
3𝛽 3𝛽 ( 33 )
𝜖𝐴3 ⅇ 2 2 2 2 2 𝛽(2𝜖 cos(
2
𝑡)−6𝛽 sin( 𝑡))
2
[ + − +
8𝛽 𝛽2 +𝜖2 𝛽2 +𝜖2 9𝛽2 +𝜖2

5𝛽 5𝛽
3𝛽 3𝛽 2𝛽(𝜖cos( 𝑡)−5𝛽sin( 𝑡)) 5𝛽 5𝛽
2𝜖(3𝛽cos( 𝑡)+𝜖sin( 𝑡)) 2 2 2𝜖(5𝛽cos( 𝑡)+𝜖sin( 𝑡))
2 2 2 2
+ + −
9𝛽2 +𝜖2 25𝛽2 +𝜖2 25𝛽2 +𝜖2
7𝛽 7𝛽 7𝛽 7𝛽
2𝛽(𝜖cos( 𝑡)−7𝛽sin( 𝑡)) 2𝜖(7𝛽cos( 𝑡)+𝜖sin( 𝑡))
2 2 2 2

49𝛽2 +𝜖2
+ 49𝛽2 +𝜖2
]
[ ]

Thus, the full particular-solutions are:


𝐴3 𝜖
{− [4𝛽𝜖(1225𝛽4 + 74𝛽2 𝜖 2 + 𝜖 4 )cos(𝛽𝑡) +
4(9𝛽2 +𝜖2 )(25𝛽2 +𝜖2)(49𝛽 2 +𝜖2 )
12𝛽𝜖(9𝛽4 + 10𝛽2 𝜖 2 + 𝜖 4 )cos(3𝛽𝑡) − 3675𝛽6 sin(𝛽𝑡) + 1003𝛽4 𝜖 2 sin(𝛽𝑡) + 71𝛽2 𝜖 4 sin(𝛽𝑡) +
𝜖 6 sin(𝛽𝑡) − 315𝛽6 sin(3𝛽𝑡) − 341𝛽4 𝜖 2 sin(3𝛽𝑡) − 25𝛽2 𝜖 4 sin(3𝛽𝑡) + 𝜖 6 sin(3𝛽𝑡)}
𝑋𝑝 = 𝐴3 𝛽𝜖
( 34 )
{(𝛽 2 [(3675𝛽6 − 1003𝛽4 𝜖 2 − 71𝛽2 𝜖 4 − 𝜖 6 )cos(𝛽𝑡) +
+𝜖2)(9𝛽2 +𝜖2 )(25𝛽2 +𝜖2 )(49𝛽 2 +𝜖2 )
8𝛽𝜖(653𝛽4 + 82𝛽2 𝜖 2 + 5𝜖 4 + 9(9𝛽4 + 10𝛽2 𝜖 2 + 𝜖 4 )cos(2𝛽𝑡))sin(𝛽𝑡)) +
[ 3(315𝛽6 + 341𝛽4 𝜖 2 + 25𝛽2 𝜖 4 − 𝜖 6 )cos(3𝛽𝑡)]} ]

Adding on the respective complementary function to each of these particular-


solutions yields the two general solutions.

Thus, we have:
𝜀 𝜀
𝜀 𝛽𝑡 𝛽 𝛽𝑡 𝜀 𝛽𝑡 𝛽 𝛽𝑡
𝐶1 𝑒 2𝑡 ( cos ( ) + sin ( )) + 𝐶2 𝑒 2𝑡 ( sin ( ) − cos ( ))
2 2 2 2 2 2 2 2
𝐴3 𝜖
+
− [4𝛽𝜖(1225𝛽4 + 74𝛽2 𝜖 2 + 𝜖 4 )cos(𝛽𝑡) +
4(9𝛽 2 +𝜖 2)(25𝛽 2 +𝜖 2 )(49𝛽 2 +𝜖 2 )
12𝛽𝜖(9𝛽4 + 10𝛽 𝜖 + 𝜖 2 2 4 )cos(3𝛽𝑡)
− 3675𝛽 sin(𝛽𝑡) + 1003𝛽4 𝜖 2 sin(𝛽𝑡) + 71𝛽2 𝜖 4 sin(𝛽𝑡) +
6

{ 𝜖 sin(𝛽𝑡) − 315𝛽 sin(3𝛽𝑡) − 341𝛽4 𝜖 2 sin(3𝛽𝑡) − 25𝛽2 𝜖 4 sin(3𝛽𝑡) + 𝜖 6 sin(3𝛽𝑡)]


6 6
}
𝑋= 𝜀
𝛽𝑡 𝜀
𝛽𝑡
( 35 )
𝐶1 𝑒 2𝑡 cos ( ) + 𝐶2 𝑒 2𝑡 sin ( ) +
2 2
𝐴3 𝛽𝜖
[(3675𝛽6 − 1003𝛽4 𝜖 2 − 71𝛽2 𝜖 4 − 𝜖 6 )cos(𝛽𝑡) +
(𝛽 2 +𝜖 2 )(9𝛽 2 +𝜖2 )(25𝛽 2 +𝜖 2 )(49𝛽 2 +𝜖 2 )
8𝛽𝜖(653𝛽 + 82𝛽 𝜖 + 5𝜖 + 9(9𝛽 + 10𝛽2 𝜖 2 + 𝜖 4 )cos(2𝛽𝑡))sin(𝛽𝑡)) +
4 2 2 4 4

[ { 3(315𝛽6 + 341𝛽4 𝜖 2 + 25𝛽2 𝜖 4 − 𝜖 6 )cos(3𝛽𝑡)] } ]

The following graphs in figures (1) and (2) illustrate the asymptotically stable steady
state of each solution for 𝜀 < 0, and the asymptotically unstable steady state for 𝜀 > 0,
in figure 3, whereby the solution blows up. What each of these three cases suggest,
according to this model, is that if a small, negative, perturbation is made with the effect

Page | 5
of nudging the orbit inwards towards the origin, then the solution will tend to spiral
inwards towards a stable limit of zero. If, however, the perturbation effects a very
small, positive, nudge away from the origin, and similarly away from the orbital path,
then the solution will continue to spiral outwards, ultimately approaching infinity. To
this effect, this simulation is perhaps at odds with other more established models,
which otherwise illustrate that a small positive outward perturbation for the Van der
Pol oscillator apparently results in the orbit gradually spiralling back towards the
original limit cycle (of amplitude equal to 2), with only substantial perturbations
having the effect of causing ‘blow-up’ resulting from an irrevocable outward spiral
towards infinity in such a case.
y
3

x
10 20 30 40 50

Figure 1: the two solutions for 𝐴 = 2, 𝛽 = 3.99 and 𝜀 = −0.1 (spiralling inwards
towards the limit cycle of amplitude 2.
y

x
5 10 15 20 25

Figure 2: the two solutions for 𝐴 = 2, 𝛽 = 1.75 and 𝜀 = −1.5 (spiralling inwards
towards the limit cycle of amplitude 2)

Page | 6
y

200

100

x
5 10 15 20 25

100

200

Figure 3: the two solutions for 𝐴 = 2, 𝛽 = 3.75 and 𝜀 = 0.5 (spiralling outwards
towards infinity)
y
1.5

1.0

0.5

x
2 1 1 2 3
0.5

1.0

Figure 4: the solution spirals inwards towards the limit cycle, and does not appear to
veer below it. This appears to be a convergent limit cycle for the relatively small
parameter of 𝜀 = −0.1, for 𝐶1 = 𝐶2 = 1, as 𝑡 becomes large
y

x
2 1 1

Figure 5: the solution spirals outwards form the limit cycle, slowly at first, but
eventually ‘blows-up’ for the relatively small parameter of 𝜀 = −1.5, for 𝐶1 = 𝐶2 = 1,
as 𝑡 becomes large

Page | 7
y

150

100

50

x
300 200 100 100 200 300
50

100

150

Figure 6: the solution spirals outwards from the limit cycle at a greater rate than in
figure (4), and eventually ‘blows-up’ completely, this time for 𝜀 = 0.5, for 𝐶1 = 𝐶2 =
1, as 𝑡 becomes large yet again
y

1.0

0.5

x
1.0 0.5 0.5 1.0

0.5

1.0

Figure 4: the solution spirals inwards towards the limit cycle, and does not appear to
veer below it. This appears to be a convergent limit cycle for the relatively small
parameter of 𝜀 = −0.02, with 𝑥 (0) = 0 and 𝑥̇ (0) = −0.2, as 𝑡 becomes large
y

x
3 2 1 1 2 3

Figure 5: the solution spirals outwards form the limit cycle, slowly at first, but
eventually ‘blows-up’ for the relatively small parameter of 𝜀 = 0.02, with 𝑥(0) = 0
and 𝑥̇ (0) = −0.2, as 𝑡 becomes large

Page | 8
y

6 10 7

4 10 7

2 10 7

x
4 10 7 2 10 7 2 10 7 4 10 7 6 10 7

2 10 7

4 10 7

6 10 7

Figure 6: the solution spirals outwards from the limit cycle at a greater rate than in
figure (4), and eventually ‘blows-up’ completely, this time for 𝜀 = 0.5, with 𝑥 (0) = 0
and 𝑥̇ (0) = −0.2, as 𝑡 becomes large yet again

Page | 9

Вам также может понравиться