IOP PUBLISHING PHYSICA SCRIPTA

Phys. Scr. T136 (2009) 014033 (4pp) doi:10.1088/0031-8949/2009/T136/014033

Asymptotic solution of the van der Pol

oscillator with small fractional damping
Feng Xie1,2 and Xueyuan Lin1
1
Department of Applied Mathematics, Donghua University, Shanghai 201620,
People’s Republic of China
2
Division of Computational Science, E-Institute of Shanghai Universities, SJTU, Shanghai 200030,
People’s Republic of China
E-mail: fxie@dhu.edu.cn

Received 27 January 2009

Accepted for publication 28 January 2009
Published 12 October 2009
Online at stacks.iop.org/PhysScr/T136/014033

Abstract
In the present paper, we investigate the van der Pol oscillator with small fractional damping.
By using the method of two-scale expansion, the asymptotic solution of the initial value
problem is obtained, which matches very well with the numerical simulation. The comparison
with the corresponding case of the integer order is also given.

PACS number: 02.3.Mv

(Some figures in this article are in colour only in the electronic version.)

1. Introduction where the fractional derivative Dα u is in the Caputo sense

defined as
The van der Pol oscillator with nonlinear damping is governed
u (s)ds
Z t 0
1
by the second-order differential equation Dα u(t) = , 0 < α < 1.
0(1 − α) 0 (t − s)α
d2 u du
2
− ε(1 − u 2 ) + u = 0, For details of the basic definitions and properties of the
dt dt fractional calculus theory, see for instance [13–16].
where u is the dynamical variable and ε is a positive The two-scale expansion method [17–19] will be used to
parameter. This model was proposed by Balthasar van der construct the asymptotic solution of problem (1) and (2). The
Pol in 1920 to describe an electronic circuit that appeared asymptotic solution obtained by the two-scale method can be
in very early radios [1]. From then on, it has been compared with that of the corresponding differential equation
widely used to model a variety of physical and biological of integer order obtained by the same method, which can let us
systems [2–8]. Recently, due to the frequent appearance know the difference between fractional and integer derivatives
of fractional derivatives in various applications in fluid in van der Pol oscillators.
mechanics, viscoelasticity, biology, physics and engineering, This paper is organized as follows. The asymptotic results
various kinds of van der Pol equations of fractional order have are presented in section 2. In section 3, we use a numerical
attracted more and more attention; see for instance [9–12] and method proposed by Diethelm et al [20] to verify the validity
references therein. of the asymptotic solution. In section 4, we compare the
In the present paper, we consider the following van der obtained result with that of the corresponding case of integer
Pol oscillator with a small fractional damping term: order. The conclusions are given in section 5.
d2 u
− ε(1 − u 2 )Dα u + u = 0, 0 < ε  1, (1)
dt2 2. Asymptotic results
with
du Let us introduce the two variables t = t and τ = εt and try
u(0) = a, (0) = b, (2)
dt to construct a formal approximation u(t) by means of the

0031-8949/09/014033+04$30.00 1 © 2009 The Royal Swedish Academy of Sciences Printed in the UK

Phys. Scr. T136 (2009) 014033 F Xie and X Lin

two-variable expansion Note that

∂u 0 t
∂u 0
Z
N 1 ds
X J 1−α = (s, τ )
u(t; ε) = u n (t, τ )ε + ε
n N +1
R N +1 (t; ε). (3) ∂t 0(1 − α) 0 ∂t (t − s)α
n=0
t
∂u 0
Z
1 ds
By the chain rule, we have = (t − s, τ ) α
0(1 − α) 0 ∂t s
du n ∂u n ∂u n
+ε , t
Z
= (4) A0 cos t sin s − sin t cos s
dt ∂t ∂τ = ds
0(1 − α) 0 sα
and
2
∂ un 2
∂ un ∂ un
2 2 t
Z
d un B0 cos t cos s + sin t sin s
= + 2ε + ε2 2 . (5) + ds
dt 2 ∂t 2 ∂t∂τ ∂τ 0(1 − α) 0 sα
= (B0 (τ )I1 − A0 (τ )I2 ) sin t
In order to determine the coefficients u n , substituting
equations (3)–(5) into equations (1) and (2), and equating the + (A0 (τ )I1 + B0 (τ )I2 ) cos t, (9)
coefficients of the same powers of ε, we obtain a set of initial
value problems u n (n = 0, 1, . . .) satisfying where
t t
∂ u0
 2 Z Z
1 sin s 1 cos s
 2 + u 0 = 0,
 I1 = ds, I2 = ds.
0(1 − α) sα 0(1 − α) sα

∂t (6) 0 0

u 0 (0, 0) = a, ∂u 0 (0, 0) = b,

 Using contour integration, it can be shown that
∂t
√
π 0(1 − (α/2))
∂ u1 ∂ 2u0 1−α ∂u 0
I1∞ ≡ lim I1 = α ,
 2
 ∂t 2 + u 1 = −2 ∂t∂τ + (1 − u 0 )J
2
, 2 0((1 + α)/2)0(1 − α)

 t→∞
∂t

(7) √
 ∂u 1 ∂u 0 π 0((1 − α)/2)
u 1 (0, 0) = 0, (0, 0) = − (0, 0), I2∞ .

 ≡ lim I2 = α
∂t ∂τ t→∞ 2 0(α/2)0(1 − α)
and
Inserting equations (8) and (9) into equation (7), we have
∂ un ∂ 2 u n−1 ∂u n−2
 2
∂ 2u1 A20 + B02

 + u n = −2 − dA0
∂t 2 ∂t∂τ ∂τ 2 (B )

+ u = 2 + I − A I −

 1 0 1 0 2
∂t 2

 dτ 4
1−α ∂u n−1 ∂u n−2

  
+(1 − u 0 )J
 2
+

A2 + B02
 
dB0

∂t ∂τ

× (3B0 I1 − A0 I2 ) sin t + −2 +(A0 I1 + B0 I2 ) − 0 .




 n−1 X l dτ 4
∂u n−1−l

 X
u k u l−k J 1−α

 −   2
B0 − A20 A0 B0
∂t

× (3A0 I1 + B0 I2 ) cos t + (B0 I1 − A0 I2 ) −

l=0 k=0
 4 2

n−2 X l
∂u n−2−l
   2 2
 X B0 − A0 A0 B0
, × (A0 I1 + B0 I2 ) sin 3t + (A0 I1 + B0 I2 ) −

− u k u l−k J 1−α


∂τ



 l=0 k=0
4 2

 

∂u n ∂u n−1

× (B0 I1 − A0 I2 ) cos 3t.

u n (0, 0) = 0, (0, 0) = − (0, 0),




 ∂t ∂τ
n = 2, 3, . . . ,

In order to avoid secular terms in u 1 (t, τ ) it is necessary that
where J 1−α is the Riemann–Liouville fractional integral the coefficients of sin t and cos t vanish. Considering that only
operator of order 1 − α, which is defined as the behaviors for t → ∞ matter, we substitute I1∞ and I2∞ for
I1 and I2 , respectively. It follows that A0 (τ ) and B0 (τ ) satisfy
Z t
1 u(s)ds the following equations:
J 1−α u(t) = .
0(1 − α) 0 (t − s)α

A2 + B02
 
dA0 1
A0 I2∞ − B0 I1∞ + 0 3B0 I1∞ − A0 I2∞ ,


From the initial value problem (6), we can easily get =
dτ 2 4
(10)
u 0 (t, τ ) = A0 (τ ) cos t + B0 (τ ) sin t (8)
A20 + B02
 
with dB0 1
A0 I1∞ B0 I2∞ ∞ ∞
.



= + − 3A0 I1 + B0 I2
dτ 2 4
A0 (0) = a, B0 (0) = b. (11)
Under the polar coordinates
Here A0 and B0 will be determined by the requirement that
the next term u 1 (t, τ ) does not exhibit secular behavior. A0 (τ ) = ρ cos θ, B0 (τ ) = ρ sin θ,

2
Phys. Scr. T136 (2009) 014033 F Xie and X Lin

equations (10) and (11) become

1.0
I2∞ ρ2
 
dρ
= 1− ρ, (12)
dτ 2 4 0.5

0.0
I∞ 3ρ 2
 
dθ
= 1 1− . (13)
dτ 2 4
−0.5
The corresponding initial conditions for equations (12)
and (13) are −1.0
√ b
ρ(0) = a 2 + b2 , θ(0) = arctan , (14) 0 20 40 60 80
a t
where we assume that a 6= 0. The case a = 0 can be
considered similarly. The solution to the initial value Figure 1. Asymptotic solution (line) and numerical solution (point)
problem (12)–(14) can be obtained by elementary integration of problem (1) and (2) for α = 21 , a = 1, b = 0 and ε = 0.01.
√
2 a 2 + b2
ρ(τ ) = p ,
a 2 + b2 + (4 − a 2 − b2 )e−I2 τ
∞

3I1∞  a 2 + b2 + (4 − a 2 − b2 )e−I2 τ 
 ∞ 

θ(τ ) = − I1∞ τ − ∞ ln 
2I2 4 
b
+ arctan .
a
Thus, we obtain a first approximation for the solution u(t; ε):

u(t; ε) ≈ ρ(τ ) cos θ(τ ) cos t + ρ(τ ) sin θ(τ ) sin t,

where ρ(τ ) and θ (τ ) are defined above.

In a similar fashion, we can obtain the higher order
approximation of u(t; ε), which is omitted here.
Figure 2. The amplitude difference for a = 1, b = 0 and ε = 0.001.
3. Numerical simulation
Consider the free van der Pol equation
In this section, we adopt an Adams–Bashforth–Moulton
predictor–corrector method [20] to verify numerically the d2 u du
validity of the asymptotic solution obtained. − ε(1 − u 2 ) + u = 0, 0 < ε  1.
dt 2 dt
In what follows, we take α = 12 in our fractional
derivative, a = 1 and b = 0. A straightforward calculation Using the two-scale expansion method, we obtain the solution
gives subject to the condition (2) having the following first
approximation:
u 0 (t; ε) = ρ cos θ cos t + ρ sin θ sin t,
u(t) = ρ(τ ) cos θ (τ ) cos t + ρ(τ ) sin θ (τ ) sin t + O(ε),
where
where
2
ρ=p √ , √
1 + 3e−(εt/ 2) 2 a 2 + b2
√  ρ(τ ) = p ,
a 2 + b2 + (4 − a 2 − b2 )e−τ

εt 3  1 + 3e(−εt/ 2) 
θ = − √ − ln  . b
2 2  4 
θ (τ ) = arctan .
a
Figure 1 gives the comparison between the asymptotic
Figures 2 and 3 show the amplitude difference and the phase
solution and the numerical solution of problem (1) and (2) for
α = 21 , a = 1, b = 0 and ε = 0.01. Agreement between them difference between these two cases for a = 1, b = 0 and ε =
is good. 0.001, respectively.

4. Comparison with the case of integer order 5. Conclusions

In this section, let us compare the result obtained in section 2 In this paper, the two-scale expansion method has been
with that of the corresponding case of integer order. successfully applied to find the asymptotic solution of the

3
Phys. Scr. T136 (2009) 014033 F Xie and X Lin

[2] van der Pol B and van der Mark J 1927 Frequency
demultiplication Nature 120 363–4
[3] Boyarsky L L 1967 Neural net analogs of rhythmic activity in
the nervous system Biosystems 1 39–46
[4] Katoh R and Mori M 1984 Control method of biped
locomotion giving asymptotic stability of trajectory
Automatica 20 405–14
[5] Oyedeji K 2005 An analysis of a nonlinear elastic forced van
der Pol oscillator equation J. Sound Vib. 281 417–22
[6] Spasic A M and Lazarevic M P 2007 A new approach to the
phenomena at the interfaces of finely dispersed systems
J. Colloid Interface Sci. 316 984–95
[7] Belhaq Mohamed and Sah Si Mohamed 2008 Horizontal fast
excitation in delayed van der Pol oscillator Commun.
Nonlinear Sci. Numer. Simul. 13 1706–13
[8] Enjieu Kadji H G, Chabi Orou J B and Woafo P 2008
Synchronization dynamics in a ring of four mutually
coupled biological systems Commun. Nonlinear Sci. Numer.
Simul. 13 1361–72
[9] Barbosa R S, Tenreiro Machado J A, Vinagre B M and
Figure 3. The phase difference for a = 1, b = 0 and ε = 0.001. Calderon A J 2007 Analysis of the Van der Pol oscillator
containing derivatives of fractional order
J. Vib. Control 13 1291–301
van der Pol oscillator with small fractional damping. The [10] Ge Z M and Zhang A R 2007 Chaos in a modified van der Pol
fractional derivative is considered in the Caputo sense. system and in its fractional order systems Chaos Solitons
Numerical results show that the asymptotic solution obtained Fractals 32 1791–822
in this paper is a good approximation of the original problem. [11] Chen J H and Chen W C 2008 Chaotic dynamics of the
In addition, the asymptotic solution obtained by the two-scale fractionally damped van der Pol equation Chaos Solitons
Fractals 35 188–98
method can be compared with that of the corresponding [12] Gafiychuk V, Datsko B and Meleshko V 2008 Analysis
differential equation of the integer order obtained by the same of fractional order Bonhoeffer–van der Pol oscillator
method, which tell us the difference between fractional and Physica A 387 418–24
integer derivatives in the van der Pol oscillators. The question [13] Oldham K B and Spanier J 1974 The Fractional Calculus
whether the present method can also be used to solve van der (New York: Academic)
Pol oscillators with small fractional damping and forced terms [14] Miller K S and Ross B 1993 An Introduction to the Fractional
Calculus and Fractional Differential Equations (New York:
should be addressed in a future study. Wiley)
[15] Samko S G, Kilbas A A and Marichev O I 1993 Fractional
Acknowledgments Integrals and Derivatives—Theory and Applications
(Yverdon: Gordon and Breach)
[16] Podlubny I 1999 Fractional Differential Equations (San Diego,
We thank the referees for valuable comments and suggestions. CA: Academic)
We were supported by the National Natural Science [17] Nayfeh A H 1981 Introduction to Perturbation Technigues
Foundation of China (grant no. 10701023) and partially (New York: Wiley)
supported by E-Institutes of Shanghai Municipal Education [18] de Jager E M and Jiang F R 1996 The Thoery of Singular
Commission (grant no. N.E03004). Perturbation (Amsterdam: North-Holland)
[19] Kevorkian J K and Cole J D 1996 Multiple Scale and Singular
Perturbation Methods (New York: Springer)
References [20] Diethelm K, Ford N J and Freed A D 2002 A
predictor–corrector approach for the numerical solution
[1] van der Pol B 1920 A theory of the amplitude of free and offractional differential equations Nonlinear Dyn.
forced triode vibrations Radio Rev. 1 701–10 29 3–22