COMMUNICATIONS ON doi:10.3934/cpaa.2011.10.

1377
PURE AND APPLIED ANALYSIS
Volume 10, Number 5, September 2011 pp. 1377–1391

DUFFING–VAN DER POL–TYPE OSCILLATOR SYSTEM AND

ITS FIRST INTEGRALS

Zhaosheng Feng, Guangyue Gao and Jing Cui

Department of Mathematics
University of Texas–Pan American, Edinburg, Texas 78539, USA

Abstract. In this paper, under certain parametric conditions we are con-

cerned with the first integrals of the Duffing–van der Pol–type oscillator sys-
tem, which include the van der Pol oscillator and the damped Duffing oscillator
etc as particular cases. We apply the Lie symmetry method to find two non-
trivial infinitesimal generators and use them to construct canonical variables.
Through the inverse transformations we obtain the first integrals of the original
oscillator system under the given parametric conditions, and some particular
cases such as the damped Duffing equation and the van der Pol oscillator sys-
tem are discussed accordingly.

1. Introduction. Many models of dynamical problems appearing as differential

equations arise in physical, chemical and biological phenomena. Problems in dy-
namics have fascinated applied mathematicians, physicists and engineers for a long
time. Over the past few decades applications in solid and structural mechanics
as well as fluid mechanics have appeared, and there is now widespread interest in
the engineering and applied science communities in nonlinear oscillators, strange
attractors, chaos, and dynamical systems theory. In spite of the great elegance
and simplicity of such equations, the solutions of specific problems proved remark-
ably difficult [15]. Finding innovative methods to analyze and solve these equations
has been an interesting subject in the field of differential equations and dynami-
cal systems [19, 23]. For many nonlinear problems, it is not always possible and
sometimes not even advantageous to express exact solutions of nonlinear differen-
tial equations explicitly in terms of elementary functions, but it is possible to find
elementary functions that are constant on solution curves, that is, elementary first
integrals. These first integrals allow us to occasionally deduce properties that an
explicit solution would not necessarily reveal. In the pioneering work [22], Prelle
and Singer introduced a procedure to find the first integrals of first-order ordinary
differential equations (ODEs) of the form y = P (x, y)/Q(x, y), with both P (x, y)
and Q(x, y) polynomials whose coefficients lie in the field of complex numbers C.
Duarte et al. [5] extended this procedure to second-order ODEs which is based on
a conjecture that if the given second-order ODE has an elementary solution, then
there exists at least one elementary first integral I(x, y, y 0 ) whose derivatives are
all rational functions of x, y and y 0 . Another powerful technique for seeking the

2000 Mathematics Subject Classification. Primary: 76M60; Secondary: 37K10, 34L30.

Key words and phrases. First integral, Duffing oscillator, van der Pol oscillator, diffeomor-
phism, Lie symmetry method, Lie point symmetry, prolonged infinitesimal operator, parametric
solution.
This work is supported by UTPA Faculty Research Council Grant 145MATH04.

1377
1378 Z. FENG, G. GAO AND J. CUI

first integrals of various differential equations is described in [7–12, 17,29] by means

of the Lie reduction method etc. Recently, considerable attention has been received
to various nonlinear oscillator systems for finding the first integrals and related
dynamical properties [4, 15, 19, 26]. Special types of first integrals and dynamical
behaviors are of fundamental importance to our understanding of physical, chemical
and biological phenomena modelled differential equations.
In this paper, we consider a more general nonlinear oscillator system of the form

ü + (δ + βum )u̇ − µu + αun = 0. (1)

where an over-dot represents differentiation with respect to the independent variable

x, and all coefficients δ, β, µ and α are real. It is referred as to the Duffing–van der
Pol–type oscillator, since the choices α = 0 and m = 2 lead equation (1) to the van
der Pol oscillator
ü + (δ + βu2 )u̇ − µu = 0, (2)
which was originally discovered by the Dutch electrical engineer van der Pol in
electrical circuits [27, 28]. The choices β = 0 and n = 3 lead equation (1) to the
damped Duffing equation [6, 15]

ü + δ u̇ − µu + αu3 = 0. (3)

When we take β = 0 and n = 2, equation (1) becomes the damped Helmholtz

oscillator [2, 25]
ü + δ u̇ − µu + αu2 = 0. (4)
Moreover, the choices m = 2 and n = 3 lead equation (1) to the standard form of
the Duffing-van der Pol oscillator equation, whose autonomous version (force-free)
takes the form
ÿ + (δ + βy 2 )ẏ − µy + αy 3 = 0. (5)
Equation (5) arises in a model describing the propagation of voltage pulses along a
neuronal axon and has recently received much attention from many authors. A large
amount of literature exists on this equation; for details and applications, see [16,20]
and references therein. It is well known that there are a great number of theoretical
works to deal with equations (2)–(5) [13, 15, 16, 19], and applications of these four
equations and the related systems can be seen in quite a few scientific areas [2,3,14].
In the present paper, we study the oscillator system (1) to obtain its first integrals
under certain parametric conditions by applying the Lie symmetry method. The
paper is organized as follows. In the next section, in order to make this paper well
self-contained and present our results in a straightforward way, we summarize the
Lie symmetry method for constructing the first integrals of second-order ODEs. In
Section 3, we use the Lie symmetry method to derive the associated determining
system, and apply it further to find two nontrivial infinitesimal generators. After
reducing them to canonical variables, we change system (1) to a simple autonomous
equation. Through the inverse transformations we obtain the first integral of the
original oscillator system (1) under the given parametric conditions. In Section 4,
the first integrals of some particular cases such as the damped Duffing equation and
the van der Pol oscillator system are discussed accordingly. In Section 5, we present
a brief conclusion.
 DUFFING–VAN DER POL–TYPE OSCILLATOR SYSTEM 1379

2. Preliminaries. In order to understand symmetries of ordinary differential equa-

tions, we give a brief introduction of prolonged infinitesimal generators and deter-
mining equations for the Lie point symmetries [17, 21]. Consider an ODE of the
form
  dk y
y (n) = ω x, y, y 0 , · · ·, y (n−1) , y (k) ≡ k , (6)
dx
where ω is (locally) a smooth function of all of its arguments. A symmetry of
equation (6) is a diffeomorphism that maps the set of solutions of the ODE to itself.
For a diffeomorphism:
Γ : (x, y) 7→ (x̂, ŷ),
it maps smooth planar curves to smooth planar curves. The diffeomorphism Γ on
the plane induces an action on the derivatives y (k) which is the mapping as
   
Γ : x, y, y 0 , · · ·, y (n) 7→ x̂, ŷ, ŷ 0 , · · ·, ŷ (n) ,

where
dk ŷ
y (k) = , k = 1, 2, · · · , n.
dx̂k
This mapping is usually called the nth prolongation of the diffeomorphism Γ. The
functions ŷ (k) can be found recursively through the chain rule

dŷ (k−1) Dx ŷ k−1

ŷ (k) = = , ŷ (0) ≡ ŷ, (7)
dx̂ Dx x̂
where Dx is the total derivative with respect to x

Dx = ∂x + y 0 ∂y + y 00 ∂y0 + · · ·.

The symmetry condition for the ODE (6) is given by

 
ŷ (n) = ω x̂, ŷ, ŷ 0 , · · ·, ŷ (n−1) , (8)

where the functions ŷ (k) (k = 1, 2, · · · , n) is given by formula (7).

For almost all ODEs, the symmetry condition (8) is nonlinear. Lie symmetries
are obtained by linearizing (8) when ε = 0. No such linearization is possible for
discrete symmetries, which makes them hard to find. However, it is usually easy to
find out whether or not a given diffeomorphism is a symmetry of a particular ODE.
The trivial symmetry corresponding to ε = 0 leaves every point unchanged. Thus,
for ε sufficiently close to zero, the prolonged Lie symmetries take the form

x̂ = x + εξ + O(ε2 ),
ŷ = y + εη + O(ε2 ), (9)
(k) (k) (k) 2
ŷ =y + εη + O(ε ), k ≥ 1.

After inserting (9) into the symmetry condition (8), the O(ε) terms give the lin-
earized symmetry condition

η (k) = ξωx + ηωy + η (1) ωy0 + · · · + η (k−1) ωy(k−1) , (10)

1380 Z. FENG, G. GAO AND J. CUI

where the functions ŷ (k) and η (k) (k = 1, 2, · · · , n) can be derived recursively from
formula (7). That is,
Dx ŷ y 0 + εDx η + O(ε2 )
ŷ (1) = = = y 0 + ε(Dx η − y 0 Dx ξ) + O(ε2 ).
Dx x̂ 1 + εDx ξ + O(ε2 )
y (k) + εDx η (k−1) + O(ε2 )
ŷ (k) =
1 + εDx ξ + O(ε2 )
where from (9) we have
η (1) = Dx η − y 0 Dx ξ, (11)
0
(k)


η x, y, y , · · ·, y (k) = Dx η (k−1) − y (k) Dx ξ. (12)
In order to find the symmetry group G admitted by a differential equation with
infinitesimal operator
X = ξ∂x + η∂y ,
we introduce the prolonged infinitesimal generator
X (n) = ξ∂x + η∂y + η (1) ∂y0 + · · · + η (n) ∂y(n) . (13)
This can be used to express the linearized symmetry condition (10) in a compact
form:
 
X (n) y (n) − ω(x, y, y 0 , · · ·, y (n−1) ) = 0 when equation (6) holds.
Consider a diffeomorphism of the form
(x̂, ŷ) = (x̂(x, y), ŷ(x, y)) .
This type of diffeomorphism is called a point transformation. Any point transfor-
mation that is also a symmetry is called a point symmetry. To find the Lie point
symmetries of an ODE (6), one needs to find η (k) (k = 1, 2, · · ·, n). The functions
ξ and η depend upon x and y only. It follows from (11) and (12) that
η (1) = ηx + (ηy − ξx )y 0 − ξy y 02 , (14)
(2) 0 02 03
η = ηxx + (2ηxy − ξxx )y + (ηyy − 2ξxy )y − ξyy y
+{ηy − 2ξx − 3ξy y 0 }y 00 , (15)
(3) 0 02 03
η = ηxxx + (3ηxxy − ξxxx )y + 3(ηxyy − ξxxy )y + (ηyyy − 3ξxyy )y
−ξyyy y 04 + 3{ηxy − ξxx + (ηyy − 3ξxy )y 0 − 2ξyy y 02 }y 00
−3ξy y 002 + {ηy − 3ξx − 4ξy y 0 }y 000 .
The number of terms in η (k) increases exponentially with k. Hence, to study the
high-order ODEs, computer algebra or symbolic package is recommended.
Since our system (1) is a second-order ODE, we restrict our attention to
y 00 = F (x, y, y 0 ).
The linearized symmetry condition can be deduced by plugging (14) and (15) into
(10) and replacing y 00 by F (x, y, y 0 ). This yields
ηxx + (2ηxy − ξxx )y 0 + (ηyy − 2ξxy )y 02 − ξyy y 03 + {ηy − 2ξx − 3ξy y 0 }F
= ξFx + ηFy + {ηx + (ηy − ξx )y 0 − ξy y 02 }Fy0 . (16)
Although equation (16) looks complicated, in some cases it can be solved without
much trouble. Both ξ and η are independent of y 0 , and therefore (16) can be
 DUFFING–VAN DER POL–TYPE OSCILLATOR SYSTEM 1381

decomposed into a system of PDEs, which are called the determining equations for
the Lie point symmetries.
1382 Z. FENG, G. GAO AND J. CUI

3. First integrals of nonlinear oscillator systems.

3.1. Infinitesimal operators. Rewrite the oscillator equation (1) as the following
form:
ÿ = −(δ + βy m )ẏ + µy − αy n = F (x, y, y 0 ). (17)
Throughout this work we focus on the case where n = m + 1. For the general case
of n we will present our results in a subsequent work somewhere. To investigate the
integrability of this equation, we will apply the Lie theory of differential equations
[17,21]. We are aware that there are quite a few innovative methods proposed on the
integrability of various differential equations, for example, the Painlevé test [18], the
inverse scattering method [1], the Prelle-Singer method [13] and the Divisor theorem
method [8,12]. Here we apply the Lie theory to study equation (17) because the Lie
symmetry method not only provides us useful information about when the equation
is integrable, but also enables us to simplify the problem to canonical variables which
eases the integration of the equation in a more effective way without complicated
calculations.
In order to find the symmetry group G admitted by a differential equation with
the infinitesimal operator
∂ ∂
X = η(x, y) + ξ(x, y) ,
∂y ∂x
one needs to find an infinitesimal operator X (2) such that
X (2) ÿ + (δ + βy m )ẏ − µy + αy m+1 = 0.


(18)
The operator X (2) takes the form
∂ ∂ ∂ ∂
X (2) = ξ(x, y) + η(x, y) + H1 (x, y, ẏ) + H2 (x, y, ẏ, ÿ) ,
∂x ∂y ∂ ẏ ∂ ÿ
where from (13)–(15) we have A(x, y, ẏ) and B(x, y, ẏ, ÿ) as
H1 (x, y, ẏ) = ηx + ẏ(ηy − ξx ) − ẏ 2 ξy ,
H2 (x, y, ẏ, ÿ) = ηxx + ẏ(2ηxy − ξxx ) + ẏ 2 (ηyy − 2ξxy )
−ẏ 3 ξyy + ÿ(ηy − 2ξx − 3ẏξy ).
Both ξ(x, y) and η(x, y) that satisfy equation (18) generate infinitesimal operators
X (2) as in equation (18) which comprise the symmetries of the differential equation.
We know that if an ordinary differential equation admits an infinitesimal generator,
then the order of the equation can be reduced by one. Thus, the Duffing-van der Pol
oscillator system will be integrated only in the case where we can find two linearly
independent infinitesimal operators generated by ξ(x, y) and η(x, y).
Following the procedure to determine the symmetries of a differential equation
introduced in the preceding section, we have
ηxx + (2ηxy − ξxx )yx0 + (ηyy − 2ξxy )(yx0 )2 − ξyy (yx0 )3
= (2ξx − ηy + 3ξy yx0 )F + ξFx + ηFy + [ηx + (ηy − ξx )yx0 − ξy (yx0 )2 ]Fyx0 .(19)
This is a second-order partial differential equation for two unknown functions ξ(x, y)
and η(x, y). Although equation (19) looks complicated, it is not difficult to solve
(19) for ξ(x, y) and η(x, y). Since the unknown functions do not depend on the
derivative y 0 , after setting the coefficients of the powers (y 0 )i (i = 0, 1, 2, 3) in (19)
 DUFFING–VAN DER POL–TYPE OSCILLATOR SYSTEM 1383

to zero, one can get the determining equations which can be split and represented
as the system

[y 0 ]0 : ηxx = (µη − δηx ) + (2µξx − µηy )y −

((m + 1)αη + βηx ) y m + (ηy − 2ξx )αy m+1 , (20)
0 1 m−1 m m+1
[y ] : 2ηxy − ξxx = −δξx + 3µξy y − mβηy − βξx y − 3αξy y , (21)
[y 0 ]2 : ηyy − 2ξxy = −2δξy − 2βξy y m , (22)
0 3
[y ] : ξyy = 0. (23)

The last equation (23) gives

ξ = a(x)y + b(x). (24)

After substituting (24) to equation (22) yields

2βa(x)
η = a0 (x)y 2 − δa(x)y 2 − y m+2 + c(x)y + d(x), m 6= −1, m 6= −2,
(m + 1)(m + 2)
(25)
where a(x), b(x), c(x) and d(x) are functions of x to be determined. Plugging (24)
and (25) into equation (20), we obtain a polynomial of y with degree 2m + 2 which
is zero if and only if each variable coefficient is set to zero

[y 2m+2 ] : β 2 a0 − αβa = 0,
2βµa 2βδa0 2βa00
[y m+2 ] : + +
m + 2 (m + 1)(m + 2) (m + 1)(m + 2)
−α(m + 1)a0 + α(m − 1)aδ − βa00 + βδa0 = 0,
[y m+1 ] : αcm + c0 β + 2αb0 = 0,
[y m ] : (m + 1)αd + d0 β = 0, (26)
2 000 0 2 0
[y ] : a − a µ − δ a − δaµ = 0,
[y 1 ] : c00 + δc0 − 2b0 µ = 0,
[y 0 ] : d00 − dµ + δd0 = 0.

Similarly, substituting (24) and (25) into equation (21), we obtain a polynomial
of y with degree 2m + 1 which is zero if and only if the following equations are
satisfied
2maβ 2
[y 2m+1 ] : = 0,
(m + 1)(m + 2)
4β 0
[y m+1 ] : −(m + 1)a0 β + maδβ − 3αa = − a,
m+1
[y m ] : −cmβ − βb0 = 0, (27)
m−1
[y ] : −mdβ = 0,
[y ] : 3a00 − 3δa0 − 3aµ = 0,
1

[y 0 ] : 2c0 − b00 = −δb0 .

In our study we require α 6= 0 and β 6= 0. Analyzing the above two resultant

equation systems, we deduce that a(x) = 0 and d(x) = 0, and the determining
1384 Z. FENG, G. GAO AND J. CUI

system for b(x) and c(x) is reduced to

cm + b0 = 0, (28)
c0 β + αb0 = 0, (29)
2c0 − b00 + δb0 = 0, (30)
c00 + δc0 − 2b0 µ = 0. (31)
Through analyzing equations (28) and (29), we solve for b(x) and c(x) as
−c0 αm
b(x) = βe β x + b0 , (32)
α
αm
c(x) = c0 e β x , (33)
where b0 and c0 are arbitrary constants.
There are two possibilities here. The first case is to set c0 = 0. In this case b0
can be an arbitrary constant, and this means that
ξ = 1, η = 0.
Hence, only one infinitesimal operator is generated as χ1 = ∂x.
The other case, in order to get two symmetries, assumes that c0 6= 0. Substituting
equations (32) and (33) into equation (30), we obtain one parametric condition:
δβ
m= − 2. (34)
α
Substituting equations (32) and (33) into equation (31), we obtain another para-
metric condition:
α2 m αδ
2
=− − 2µ. (35)
β β
Combining conditions (34) and (35) yields
α µβ
δ= − . (36)
β α
Because b0 and c0 are arbitrary constants, for simplicity, we may assume b0 = 0
and c0 = 1. Then we find
1 αm αm
b(x) = − βe β x , c(x) = e β x ,
α
which is equivalent to
1 αm αm
ξ = − βe β x , η = e β x y.
α
Using ξ and η, we obtain two infinitesimal generators as follows
1 αm αm
χ1 = ∂x, χ2 = − βe β x ∂x + e β x y∂y. (37)
α
Each infinitesimal generator is of the form
χ = c1 χ 1 + c2 χ 2 ,
where χ1 is a homothety operator and χ2 is a translation operator. Note that
two infinitesimal generators in (37) are nontrivial. Consequently, only under the
parametric constraint (36), the oscillator system (17) is completely integrable. Oth-
erwise, the oscillator is only partially integrable and there is no way to write out
the solution in terms of elementary functions.
 DUFFING–VAN DER POL–TYPE OSCILLATOR SYSTEM 1385

3.2. Reduction to canonical variables. We know that if an ordinary differential

equation admits an infinitesimal generator, then there exists a pair of variables:
t = f (x, y), u = g(x, y),
which are called canonical variables, with f and g (g 6= 0) being arbitrary particular
solutions of the first-order linear partial differential equations [24]
∂f ∂f
ξ(x, y) + η(x, y) = χ, (38)
∂x ∂y
∂g ∂g
ξ(x, y) + η(x, y) = 0, (39)
∂x ∂y
where χ is a nonzero constant and can be chosen arbitrarily. Suppose that the
general solution of the characteristic equation
dx dy
= ,
ξ(x, y) η(x, y)
takes the form U (x, y) = C, where C is arbitrary, then the general solutions of (38)
and (39) can be expressed by
Z
dx
f (x, y) = χ + Φ1 (U ), (40)
ξ ∗ (x, U )
g(x, y) = Φ2 (U ), U = U (x, y), (41)
where Φ1 (U ) and Φ2 (U ) are the arbitrary functions, ξ ∗ (x, U (x, y)) ≡ ξ(x, y), and
U in the integral is regarded as a parameter later. Choosing χ = m in (38), and
using (40), (41) and (37), we obtain a particular solution
αm α
f (x, y) = e− β x , g(x, y) = ye β x . (42)
Since t = f (x, y) and u = g(x, y), formulas (42) can be rewritten to the parametric
form
β 1
x=− ln t, y = ut m . (43)
αm
Using the nonlinear transformations (43) yields
∂y αm 0 m+1 α 1
= − u t m − ut m , (44)
∂x β t β
∂2y 2 2
α m 00 2m+1 α2 m(m + 2) m+1 0 α2 1
= utt t m + t m ut + 2 t m u. (45)
∂x2 β2 β2 β
Substituting equations (44) and (45) into equation (17), we obtain
 2 2   2 
2m+1 α m 00 0 m 1 α αδu
t m u − αmu u + tm u− − µu
β2 β2 β
 2 
m+1 α m(m + 2) 0 δαm 0 m+1 m+1
+t m u − u − αu + αu = 0. (46)
β2 β
Under the parametric condition (36), equation (46) is simplified to an autonomous
equation
αm 00
u = u0t um ,
β 2 tt
which is easily integrated as
β2
u0t = um+1 + I, (47)
αm(m + 1)
1386 Z. FENG, G. GAO AND J. CUI

where I is an arbitrary constant. Utilizing the inverse transformations of (43) we

have
∂u β 0 α(m+1) 1 α(m+1)
= − y e β x − ye β x . (48)
∂t αm m
Substituting equations (48) into (47), under the parametric condition (36), conse-
quently we obtain the first integral to the oscillator system (1) when n = m + 1 as
follows  
0 α β α(m+1)
y + y+ y m+1
e β x = I. (49)
β m+1
4. Particular cases. Following the procedure presented in the preceding section,
we consider several special cases of the Duffing-van der Pol-type oscillator system
(1) in this section.
4.1. Force-free duffing-van der Pol oscillator. We know that the choices m = 2
and n = 3 lead equation (1) to the standard form of the Duffing-van der Pol oscil-
lator equation (5). In [4, 26], the first integral of the Duffing-van der Pol oscillator
when α = 1 is considered, namely
ÿ + (δ + βy 2 )ẏ − µy + y 3 = 0. (50)
Following the parametric condition (36), that is,
1
δ = − µβ, (51)
β
and using formula (49), we can obtain immediately that the Duffing-van der Pol
equation (50) has one first integral of the form
 
1 β 3 3x
ẏ + y + y e β = I1 . (52)
β 3
It is remarkable that in view of our parametric condition (51) and formula (52),
it shows that our parametric constraint (51) is weaker than the corresponding ones
described in the literature [4, 26], and the first integral presented in [4, 26] is just a
particular case of (52). The results on the first integral established in [13] by using
the Prelle-Singer method agree well with (52).
4.2. Duffing-type oscillator. Assume that α 6= 0 and β ≡ 0, then equation (1)
changes into the form
ÿ + δ ẏ − µy + αy m+1 = 0. (53)
We suppose that m 6= 0 and m 6= −1. By an analogous argument in Section 3, we
obtain a(x) = 0 and d(x) = 0, as well as a determining system about b(x) and c(x)
as follows
cm + 2b0 = 0, (54)
2c0 − b00 + δb0 = 0, (55)
00 0 0
c + δc − 2b µ = 0. (56)
From equations (54) and (55), we find b(x) and c(x) be of the form
−c0 (m + 4) m+4δm
x
b= e + b0 , (57)
2δ
δm
c = c0 e m+4 x , (58)
where b0 and c0 are arbitrary constants.
 DUFFING–VAN DER POL–TYPE OSCILLATOR SYSTEM 1387

There are two options here too. The first one is c0 = 0, which gives
ξ = 1, η = 0.
So, only one corresponding infinitesimal operator is generated as χ1 = ∂x. The
other option, in order to get two symmetries, is to assume c0 6= 0. Substituting
equations (57) and (58) into equation (56), we derive the parametric condition as
2m + 4 2
µ=− δ . (59)
(m + 4)2
For simplicity, we choose b0 = 0 and c0 = 1, and then have
−(m + 4) m+4
δm
x δm
b= e , c = e m+4 x .
2δ
Combining this with a(x) = 0 and d(x) = 0, we deduce
−(m + 4) m+4 δm
x δm
ξ= e , η = e m+4 x y.
2δ
Thus, an infinitesimal generator is found, namely
−(m + 4) m+4
δm
x δm
χ= e ∂x + e m+4 x y∂y.
2δ
That is, under the parametric condition (59), the oscillator system (53) is completely
integrable.
Following the procedure (38)-(41) and choosing χ = m 2 in (38), we obtain par-
ticular solutions for f and g as
δm 2δ
f (x, y) = e− m+4 x , g(x, y) = ye m+4 x . (60)
Since t = f (x, y) and u = g(x, y), formulas (60) are equivalent to the parametric
form
m+4 2
x=− ln t, y = ut m , (61)
δm
which gives
∂y δm 0 m+2 2δ 2
= − ut t m − ut m , (62)
∂x m+4 m+4
∂2y δ 2 m2 00
2(m+1) δ 2 m 2+m 0 4δ 2 2
= utt t m + t m ut + t m u. (63)
∂x2 (m + 4) 2 (m + 4) (m + 4)2
Substituting equations (62) and (63) into equation (53), we obtain
4δ 2 2δ 2
 
2
u− u − µu t m
(m + 4)2 m+4
mδ 2 0
 
m(m + 2) 2 0 2m 2 0 m+2
+ δ ut + δ ut − u t m
(m + 4)2 (m + 4)2 m+4 t
 2 !
mδ 00 m+1 2m+2
+ utt + αu t m = 0. (64)
m+4

Under the parametric condition (59), equation (64) changes into an autonomous
equation as
m2 δ 2
u00 = −αum+1 ,
(m + 4)2 tt
1388 Z. FENG, G. GAO AND J. CUI

which is integrated as
2(m + 4)2 α
(u0t )2 = − um+2 + I2 . (65)
m2 δ 2 m + 2
Using the inverse transformation of (61) yields
 
∂u m + 4 0 2y (m+2)
= − y − e m+4 δx . (66)
∂t δm m
Substituting equations (66) into (65), under the parametric condition (59), we ob-
tain the first integral of equation (53) as
(m + 4)2 0 2 2α(m + 4)2 m+2 2δ(m+2)
 
4 2 4(m + 4) 0
(y ) + y + yy + y e m+4 x = I2 .
(δm)2 m2 m2 δ m2 δ 2 (m + 2)
(67)

4.3. Damped duffing equation. The choices β = 0 and n = 3 lead equation (1)
to the damped Duffing equation [6, 9, 15]
ÿ + δ ẏ − µy + αy 3 = 0. (68)
Substituting m = 2 into equation (67), we can derive the first integral of equation
(68) immediately
 
9 0 2 2 6 0 9α 4 4 δx
(y ) + y + yy + y e 3 = I2 , (69)
δ2 δ 2δ 2
under the parametric condition µ = − 92 δ 2 .
It is noted that the corresponding results on equation (68) in [26] are identical
to our formula (69) when α = 1.

4.4. Helmholtz Oscillator. In this subsection, we assume β = 0 and n = 2 in

equation (1) and it becomes the Holmholtz oscillator
ÿ + δ ẏ − µy + αy 2 = 0, (70)
which is a nonlinear oscillator with a quadratic nonlinearity [2, 4]. Under the con-
dition β = 0 and n = 2 , it follows from equations (24) and (25) that
ξ = a(x)y + b(x),
and
η = a0 (x)y 2 − δa(x)y 2 + c(x)y + d(x),
where a(x), b(x), c(x) and d(x) are arbitrary functions. Following (26) and (27),
we have
[y 3 ] : αa0 = 0,
[y 2 ] : a000 − a0 µ − a0 δ 2 − µaδ + αc + 2αb0 = 0,
[y 1 ] : 2αd + c00 + δc0 − 2b0 µ = 0,
[y 0 ] : d00 − dµ + δd0 = 0,
and
[y 2 ] : −3αa = 0,
[y 1 ] : 3a00 − 3δa0 − 3aµ = 0,
[y 0 ] : 2c0 − b00 = −δb0 ,
 DUFFING–VAN DER POL–TYPE OSCILLATOR SYSTEM 1389

respectively. Analyzing the above two resultant systems gives that a(x) = 0, so we
obtain a determining system about b(x) and c(x) as
c + 2b0 = 0, (71)
0 00 0
2c − b + δb = 0, (72)
00 0 0
2αd + c + δc − 2b µ = 0, (73)
d00 − dµ + δd0 = 0. (74)
Equations (71) and (72) indicate that
δ
c(x) = c0 e 5 x ,
5 δ
b(x) = − c0 e 5 x + b0 ,
2δ
where b0 and c0 are constants. Substituting b(x) and c(x) into equation (73) we
derive a parametric condition
 
1 6 2
d=− δ + µ c.
2α 25
Similarly, substituting d(x) and c(x) into equation (74) we derive another parametric
condition   
1 6 2 6 2
− c δ +µ δ − µ = 0.
2α 25 25
Assumption c0 = 0 produces an infinitesimal operator χ1 = ∂x. If c0 6= 0, we
separate our discussions into two cases:
6 2
Case 1. When µ = 25 δ , because b0 and c0 are arbitrary constants, for our
convenience, we assume b0 = 0 and c0 = 1 that gives
−5 δ x δ −6 2 δ x
b= e 5 , c = e 5 x, d = δ e5 .
2δ 25α
That is,
6δ 2
 
−5 δ x δ
ξ= e5 , η = e5x y − .
2δ 25α
The two associated infinitesimal generators are generated as
6δ 2
 
−5 δ x δ
x
χ1 = ∂x, χ2 = e ∂x + e
5 5 y− ∂y.
2δ 25α
1
Choosing χ = 2 in (38) and using (40) and (41), we find a particular solution of f
and g as
6δ 2
 
δ 2δ
f (x, y) = e− 5 x ,g(x, y) = e 5 x y − ,
25α
which enables us to construct nonlinear transformations
5 6δ 2
x = − ln t, y = ut2 + , (75)
δ 25α
∂y δ 2δ
= − u0t t3 − ut2 , (76)
∂x 5 5
∂2y δ 2 00 4 δ 2 0 3 4δ 2 2
= u t + ut t + ut . (77)
∂x2 25 t 5 25
Substituting equations (75)-(77) into equation (70), we obtain an autonomous equa-
tion
δ 2 00
u = −αu2 ,
25 tt
1390 Z. FENG, G. GAO AND J. CUI

which is integrated as
50α 3
(u0t )2 = − u + I3 .
3δ 2
Using the inverse transformations of (75)-(77), we obtain one first integral of equa-
tion (70) as
24δ 2
 
50α 20 24δ 0 25 0 2 6δ x
y − 8y 2 + 2 y 3 + yy 0 − y + 2 (y ) e 5 = I3 , (78)
25α 3δ δ 5α δ
under the parametric condition
6 2
µ= δ .
25
6 2
Case 2. When µ = − 25 δ , using the same argument as that of Case 1, we can
obtain another first integral of equation (70) as
 
25 0 2 2 20 0 50α 3 6 δx
(y ) + 4y + yy + 2 y e 5 = I3 , (79)
δ2 δ 3δ
under the parametric condition
6
µ = − δ2 .
25
It is notable that the corresponding results presented in [2, 4, 10, 12] agree well
with formulas (78) and (79).
4.5. van der Pol oscillator. The choices α = 0 and m = 2 leads the equation (1)
to the van der Pol oscillator
ÿ + (δ + βy 2 )ẏ − µy = 0.
It follows from (26) and (27) that the determining system about b(x) and c(x) is
c0 β = 0,
cm + b0 = 0,
2c0 − b00 + δb0 = 0,
c00 + δc0 − 2b0 µ = 0.
Observing this system, we only have
b = b0 , c = 0.
This indicates that only one infinitesimal operator is found, namely X = ∂x, and
as a consequence the differential equation is partially integrable.

5. Conclusion. Quite a few approaches have been proposed to analyze and find
first integrals (conservation laws) and exact solutions for various nonlinear differ-
ential equations. It has been an interesting subject in mathematical and physical
communities. Notable among such approaches are the Prelle–Singer procedure and
the Lie symmetry method. In 1983, Prelle and Singer presented a deductive method
for solving first–order ODEs that presents a solution in terms of elementary func-
tions if such a solution exists. This technique has attracted many researchers from
diverse groups and has been extended to autonomous systems of ODEs of higher
dimensions for finding the first integrals and exact solutions under certain assump-
tions. From illustrative examples in the literature, the obtained first integrals of
autonomous systems are usually of rational or quasi-rational forms and searching for
solution sets (S, R) usually involves complicated calculations. On the other hand,
it is shown that the integrability properties of some nonlinear systems such as the
 DUFFING–VAN DER POL–TYPE OSCILLATOR SYSTEM 1391

Lorenz model and KdV-Burgers equation can be treated through the Lie symmetry
method in an effective manner.
In this paper, we were concerned with the first integrals and integrability proper-
ties of the Duffing–van der Pol–type oscillator equation (1), for which it seems that
symmetry analysis has not been undertaken before as far as our knowledge goes.
The new first integrals of the Duffing–van der Pol–type oscillator equation (1) were
established by means of the Lie symmetry method. To reach our goal, we used the
Lie symmetry method to derive the associated determining system. Through solv-
ing this system we found two nontrivial infinitesimal generators. After constructing
canonical variables, we changed system (1) to a simple autonomous equation. By
virtue of the inverse transformations we obtained the first integral of the original
oscillator system (1) under the given parametric conditions. The first integrals and
integrability of some particular cases such as the damped Duffing equation and the
van der Pol oscillator system were also discussed.

REFERENCES

[1] M. J. Ablowitz and H. Segur, “Solitons and the Inverse Scattering Transform,” SIAM,
Philadelphia, 1981.
[2] J. A. Almendral and M. A. F. Sanjuán, Integrability and symmetries for the Helmholtz oscil-
lator with friction, J. Phys. A (Math. Gen.), 36 (2003), 695–710.
[3] A. Canada, P. Drabek and A. Fonda, “Handbook of Differential Equations: Ordinary Differ-
ential Equations,” Volumes 2-3, Elsevier, 2005.
[4] V. K. Chandrasekar, M. Senthilvelan and M. Lakshmanan, On the complete integrability and
linearization of certain second-order nonlinear ordinary differential equations, Proc. R. Soc.
Lond. Ser. A, 461 (2005), 2451–2476.
[5] L. G. S. Duarte, S. E. S. Duarte, A. C. P. da Mota and J. E. F. Skea, Solving the second-order
ordinary differential equations by extending the Prelle–Singer method, J. Phys. A (Math.
Gen.), 34 (2001), 3015–3024.
[6] G. Duffing, “Erzwungene Schwingungen bei Veränderlicher Eigenfrequenz,” F. Vieweg u.
Sohn, Braunschweig, 1918.
[7] Z. Feng, On traveling wave solutions of the Burgers-Korteweg-de Vries equation, Nonlinearity,
20 (2007), 343–356.
[8] Z. Feng, The first-integral method to the Burgers-Korteweg-de Vries equation, J. Phys. A
(Math. Gen.), 35 (2002), 343–350.
[9] Z. Feng, G. Chen, and S. B. Hsu, A qualitative study of the damped Duffing equation and
applications, Discrete Contin. Dyn. Syst. Ser. B, 6 (2006), 1097–1112.
[10] Z. Feng and Q. G. Meng, Exact solution for a two-dimensional KdV-Burgers-type equation
with nonlinear terms of any order, Discrete Contin. Dyn. Syst. Ser. B, 7 (2007), 285–291.
[11] Z. Feng and Q. G. Meng, First integrals for the damped Helmholtz oscillator, Int. J. Comput.
Math. 87 (2010), 2798–2810.
[12] Z. Feng, S. Zheng and D. Y. Gao, Traveling wave solutions to a reaction-diffusion equation,
Z. angew. Math. Phys., 60 (2009), 756–773.
[13] G. Gao and Z. Feng, First integrals for the Duffng-van der Pol-type oscillator, E. J. Diff.
Equs., 2010 (2010), 1–12.
[14] M. Gitterman, “The Noisy Oscillator: the First Hundred Years, from Einstein until Now,”
World Scientific Publishing, Singapore, 2005.
[15] J. Guckenheimer and P. Holmes, “Nonlinear Oscillations, Dynamical Systems, and Bifurca-
tions of Vector Fields,” Springer-Verlag, New York, 1983.
[16] P. Holmes and D. Rand, Phase portraits and bifurcations of the non-linear oscillator: ẍ +
(α + γx2 )ẋ + βx + δx3 = 0, Int. J. Non-Linear Mech., 15 (1980), 449–458.
[17] P. E. Hydon, “Symmetry Methods for Differential Equations,” Cambridge University Press,
New York, 2000.
[18] E. I. Ince, “Ordinary Differential Equations,” Dover, New York, 1956.
[19] D. W. Jordan and P. Smith, “Nonlinear Ordinary Differential Equations: An Introduction
for Scientists and Engineers,” Oxford University Press, New York, 2007.
1392 Z. FENG, G. GAO AND J. CUI

[20] M. Lakshmanan and S. Rajasekar, “Nonlinear Dynamics: Integrability, Chaos and Patterns,”
Springer Verlag, New York, 2003.
[21] P. J. Olver, “Applications of Lie Groups to Differential Equations,” Springer Verlag, New
York, 1993.
[22] M. Prelle and M. Singer, Elementary first integrals of differential equations, Trans. Am. Math.
Soc., 279 (1983), 215–229.
[23] A. D. Polyanin and V. F. Zaitsev, “Handbook of Exact Solutions for Ordinary Differential
Equations,” 2nd edition, London: CRC Press, 2003.
[24] A. D. Polyanin, V. F. Zaitsev and A. Moussiaux, “Handbook of First Order Partial Differential
Equations,” Taylor & Francis, London, 2002.
[25] S. N. Rasband, Marginal stability boundaries for some driven, damped, non-linear oscillators,
Int. J. Non-Linear Mech., 22 (1987), 477–495.
[26] M. Senthil Velan and M. Lakshmanan, Lie symmetries and infinite-dimensional Lie algebras
of certain nonlinear dissipative systems, J. Phys. A (Math. Gen.), 28 (1995), 1929–1942.
[27] B. van der Pol, A theory of the amplitude of free and forced triode vibrations, Radio Review,
1 (1920), 701–710, 754–762.
[28] B. van der Pol and J. van der Mark, Frequency demultiplication, Nature, 120 (1927), 363–364.
[29] V. F. Zaitsev and A. D. Polyanin, “Handbook of Ordinary Differential Equations,” Fizmatlit,
Moscow, 2001 (in Russian).

Received March 2009, revised February 2010.

E-mail address: [email protected]
E-mail address: [email protected]
E-mail address: [email protected]