Applied Mathematical Modelling 37 (2013) 81188130

Contents lists available at SciVerse ScienceDirect

Applied Mathematical Modelling

journal homepage: www.elsevier.com/locate/apm

A new modication of the variational iteration method

for van der Pol equations
Yuan-Ju Huang, Hsuan-Ku Liu
Department of Mathematics and Information Education, National Taipei University of Education, Taiwan

a r t i c l e

i n f o

Article history:
Received 22 May 2012
Received in revised form 5 December 2012
Accepted 8 March 2013
Available online 24 March 2013
Keywords:
Van der Pol equation
Variational iteration method
Hes polynomials

a b s t r a c t
In this paper, we provide a new modication of the variational iteration method (MVIM)
for solving van der Pol equations. The modication couples the classical variational iteration method with Hes polynomials, where the Hes polynomials are applied to the approximate solution and the initial condition to eliminate secular terms. For the large e, the
numerical results demonstrate that the modication method get an accurate approximate
period than the other presented methods.
2013 Elsevier Inc. All rights reserved.

1. Introduction
Consider the van der Pol equation



U  e 1  U 2 U_ 0; e > 0;
U

1:1

where U_ dU
. This equation is a mathematical model of self-sustained oscillations of a triode electric circuit with current
dt
voltage characteristic in [27]. The van der Pol equation can be written in the system of two rst order differential equations
as phase plane (see [21]), that is,

8
< U_ V;


: V_ U e 1  U 2 V:

1:2

Such a system is called an autonomous system. By LienardLevinsonSmith Theorem, we can conclude that the van der
pol Eq. (1.1) has an essentially unique nontrivial periodic solution. In other words, this equivalent autonomous system (1.2)
has a unique closed path in the phase plane [21].
A considerable number of methods [3,4,20] are proposed to nd the approximate solution of (1.1). LindstedtPoincare
method [20] introduces a transformation of the independent variable and expands the solution in power series of e. Andersen
and Geer [3] expand the frequency in power series of e up to 163 terms by using FORTRAN and have determined that the
frequency has two complex-conjugate singularities in the complex e2-plane. Rouald [20] introduces a parameter
qe 1e e ; e P 0 to improve LindstedtPoincare method. This method is available not only for small but for all e > 0. Bunomo
2
[4] introduces another parameter qe 1e e2 derived from the Shohat transformation and obtains a series expansion for the
Corresponding author.
E-mail addresses: [email protected] (Y.-J. Huang), [email protected] (H.-K. Liu).
0307-904X/$ - see front matter 2013 Elsevier Inc. All rights reserved.
http://dx.doi.org/10.1016/j.apm.2013.03.033

8119

Y.-J. Huang, H.-K. Liu / Applied Mathematical Modelling 37 (2013) 81188130

frequency. Coefcients in the expansion are computed up to O(e402). Liu [12] applied the modied homotopy perturbation
method to consider the large parameter e.
In contrast to previous methods, the variational iteration method, which was proposed originally by He [7,9], attracted
much attention in the past few years as a promising method for solving nonlinear differential equations. The variational iteration method has been proved by many authors to be a powerful mathematics tool for dealing with various kind of linear and
nonlinear problems [1,2,5,7,9,10,1319,28,29]. Additionally, the homotopy perturbation transform method and Hes polynomials are used to solve the fractional partial differential equation [2326]. From their results, it shows that this method is
effective and simple.
So far many modications are provided of the variational iteration method. As the solution is approximated by a polynomial, Abassy et al. [1,2] analyze the procedure of the variational iteration solution and introduce the modication to overcome the calculation of unneeded terms. They use it to given an approximate power series solutions for non-linear problems.
Shao and Han [22] couple variational iteration method with the homotopy perturbation method to solve the sineGordon
equation. In their numerical results, both modications provide an accurate approach to nd an approximation for the nonlinear differential equations.
However, we nd that the variational iteration method as well as the modications is not applicable to (1.1). Observing
the variational iteration method, the expressions of the approximated solution are based in the selected linear operator
2

Lut ddt2u u and Lut ddt2u. When Lut ddt2u u, the approximation of (1.1) are expressed by the set of basis functions
 
ft m cosnt; t m sinntjm P 0; n P 1g which breaks down for t P O e2 because of the existence of the secular term as tnsint
2

or tncost. When the linear operator Lut ddt2u, the solution of (1.1) is approximated as the form of polynomials. Obviously,

this approximation tends to positive or negative innity as the time tends to innity for any order of tn. This is because that
the van der Pol equation has a nontrivial periodic solution.
We nd it has better to express a periodic solution by the set of periodic basis functions. Hence, to nding an accurate
approximation of the van der Pol equations (or other nonlinear differential equations with a periodic solution), we should
select a set of periodic basis functions and eliminate the secular terms in each iteration. However, the modication of variational iteration method with Hes polynomials that select a set of periodic basis functions and eliminate the secular terms
has not been found yet. The methods [1,2,22] do not eliminate the secular terms that makes their approximations inaccurate
for the van der Pol equations (see Figs. 1 and 2 in Section 2). On the other hand, these three methods did not provide the
frequency and the amplitude of the periodic solution directly, where the frequency and the amplitude are the major characteristics of a periodic function.
In this paper, the variational iteration method is modied by introducing a transformation such that the solution is expressed by the periodic basis functions. Precisely, we couple the classical variational iteration method with Hes polynomials
and construct a new homotopy to solve (1.1). Moreover, our method expands the solution as well as the frequency and the
amplitude of the solution and correct the secular terms by using the expansion of the frequency and the amplitude. Our
modication proposed in Section 3 extends the variational iteration method with Hes polynomials to select a set of periodic
basis functions and eliminate the secular terms in each iteration. Therefore, our modication provides an accurate approximation for the van der Pol equation. This implies that our method provides a new idea of the variational iteration method
with Hes polynomials for nding an approximation of the nonlinear differential equations with a periodic solution.
Observing the numerical results, the proposed method gets the periodic solution up to 10 terms after 6 iterations and the
approximation is close to the numerical solution when e = 1 and e = 2. We compare our approximation with other ones and
the numerical solution. We nd that our approximation with 6 iterations is more accurate than other high order

3.0

Dashed line: the 5th-order approximation

of the MVIM(I)
Solid line: the numerical solution

2.5
2.0
1.5
1.0

0.5

1.0

1.5

2.0

Fig. 1. Comparison of the numerical solution with the 5th-order approximation of MVIM (I) when e = 1 and A = 2.

8120

Y.-J. Huang, H.-K. Liu / Applied Mathematical Modelling 37 (2013) 81188130

Dashed line: the 6th-order approximation

of the MVIM(II)
Solid line: the numerical solution

10

t
Fig. 2. Comparison of the numerical solution with the 6th-order approximation of MVIM (II) when e = 1 and A = 2.

Table 1
Valuse for period T(e).
e

(1) Numerical result

(2) Our result

(3) Lius result

(4) Andersens result

(5) Bunomos result

1
2
3
4
5
10
20
100

6.6628686
7.62987448
8.85909550
10.20352369
11.61223067
19.07836957
34.6823253
162.83707

6.66219485
7.60337105
8.82517012
10.1828653
11.6152236
19.2259700
35.0490082
163.661083

6.6490
7.5535
8.7271
10.033
11.412
18.743
33.981
157.79

6.663286859
7.629874480
8.85909604
10.203911
11.61892
19.446
36.631
179.38

6.63286859
7.62987447
8.85909549
10.20352376
11.61224061
19.09846721
35.0603638
169.122629

Table 2
Error for period T(e).
e

(1) Numerical result

(2) Our result (%)

(3) Lius result (%)

(4) Andersens result (%)

(5) Bunomos result (%)

1
2
3
4
5
10
20
100

6.6628686
7.62987448
8.85909550
10.20352369
11.61223067
19.07836957
34.6823253
162.83707

0.01
0.35
0.38
0.20
0.03
0.77
1.06
0.51

0.21
1.00
1.49
1.67
1.72
1.76
2.02
3.10

0.01
0.00
0.00
0.00
0.06
1.93
5.62
10.16

0.45
0.00
0.00
0.00
0.00
0.11
1.09
3.86

approximations from perturbation methods where Andersens result uses 82 terms in the expansion; Bunomos result uses
203 terms in the expansion. For example, the error of our period T(e) for e 6 5 is less than 1% and for e P 10 is less than 1.5%,
while the error of other methods are all large than 1%. The details of the comparison results are displayed in Tables 1 and 2 in
Section 4.
This paper is organized as follows. In Section 2, we introduce the basic idea of variational iteration method and its existence modications. We apply these methods for the van der Pol equation and indicate they are not applicable. Then, in Section 3, we modify the variational iteration method by introducing a transformation to express the solution by the periodic
basis functions. We couple the classical variational iteration method with Hes polynomials to solve the van der Pol equation.
In Section 4, we compare our results with the numerical solution and other ones to show the efciency of our new
modication.
2. Variational iteration method and its existence modications
Consider the van der Pol equation as the form
d2 u
dt2



u  e 1  u2 du
0;
dt

u0 A;

u 0 0;

tP0

2:1

Y.-J. Huang, H.-K. Liu / Applied Mathematical Modelling 37 (2013) 81188130

8121

Liu [12] applied the LienardLevinsonSmith Theorem to show that the van der Pol Eq. (2.1) has an essentially unique
nontrivial periodic solution; that is, the equivalent autonomous system of (2.1) has a unique closed path in the phase plane.
In this section, we rst apply the variational iteration method as well as the modications to the van der Pol Eq. (2.1). And
then, we demonstrate that these methods could not provide an accurate approximation for the periodic solution.
2.1. Basic concept for the variational iteration method (VIM)
Consider the following differential equation:

Lu Nu gt;

2:2

where L is a linear operator, N is a nonlinear operator and g(t) is a inhomogeneous term. Then we can construct a correct
functional as follows:

un1 t un t

kfLun n Nun n  gngdn

2:3

where k is a general Lagrange multiplier which can be identied optimally via variational theory, un is the nth approximate
~ n denotes a restricted variation, i.e. du
~ n 0. This method has been applied in dealing with various kind of linsolution, and u
ear and nonlinear problems [1,2,5,7,9,10,12,14,28,29].
Applying (2.3) to (2.1), the correction functional is written down as follows:

un1 t un t

(
)
2

 du
~ n s
d un s
2
~
k
un s  e 1  un s
ds;
ds2
ds

2:4

~ n is considered as a restricted variation, i.e. du

~ n 0.
where u
We identify ks as an optimal Lagrange-multiplier

~ n su0n sgds
kfu00n s un s  e1  u
Z t
dun t ksdu0n sjst  k0 sdun sjst
k00 s ksdun jds

dun1 t dun t d

2:5

0
The multiplier, therefore, can be identied as k sins  t, and the following variational iteration formula can be
obtained:

un1 t un t

Z
0

( 2
)

 dun s
d un s
2
sins  t
un s  e 1  un s
ds
ds2
ds

2:6

Assuming that the initial approximate solution of (2.1) has the form

u0 t A cos t

2:7

For example, we take A = 2 and obtain

u1 2 cos t

3e
e
sin t  sin 3t
4
4

2:8

Substituting (2.8) into (2.6), we get

u2 2 cos t



3 sin t 1
1
 sin 3t e
13 cos t 18 cos 3t  5 cos 5t 12t sin te2
4
4
96

120t cos t  165 sin t 162 sin 3t  50 sin 5t 7 sin 7te3

1536
672 cos t  840 cos 3t 200 cos 5t  35 cos 7t 3 cos 9t  1680t sin te4

:
20480

2.2. Modied variational iteration method (I)

Abassy et al. [1,2] proposed the modied method (MVIM) and used it to given an approximate power series solutions for
non-linear problems.
MVIM is used for solving the general nonlinear initial value problem

8122

Y.-J. Huang, H.-K. Liu / Applied Mathematical Modelling 37 (2013) 81188130

Lux; t Rux; t Nux; t 0

ux; 0 f0 x
@ux;t
jt0
@t

f1 x

2:9

..
.
@ s1 ux;t
jt0
@t s1

fs1 x

@s
,
@t s

where L
s = 1, 2, 3, . . . is the highest partial derivative with respect to t, R is a linear operator which has partial derivatives with respect to x and Nu(x, t) is the nonlinear term.
Eq. (2.9) (with s = 1) is solved using the MVIM formula:

U n1 U n

kfRU n  U n1 Gn  Gn1 gds

2:10

where U1 = 0, U0 = f0 and Gn(x, t) is a polynomial of degree n in t and is obtained from NUn(x, t) = Gn(x, t) + O(tn + 1) and k 1
which is the Lagrange multiplier and is identied via variational theory; for more details and illustrative examples, see [1].
The modied variational iteration formula is obtained as form:

U n1 x; t U n x; t

ksfRU n  U n1 Gn  Gn1 gds

2:11

0
s1

s
where ks t
s1!

is called a general Lagrange multiplier, and Gn(x, t) obtained from NUn(x, t) = Gn(x, t) + O(ts(n + 1)) is a

polynomial of degree (s(n + 1)  1) in t (see [2]).

Applying (2.9) and (2.10) to (2.1), we dene
2

Lut

d u
dt

Rut ut  e

du
;
dt

Nut eut2

du
;
dt

and obtain the following MVIM iteration formula

U n1 U n

s  tfRU n  U n1 Gn  Gn1 gds;

2:12

where U1 = 0, U0 = A, and Gn(x, t) is a polynomial of degree (2n + 1) in t and is obtained from

NU n x; t Gn x; t Ot 2n1

2:13

The rst two approximations are obtained by observing (2.12):

U1 U0

A
s  tfRU 0  U 1 G0  G1 gds A  t2
2

2:14

and

U2 U1

s  tfRU 1  U 0 G1  G0 gds A 

At2 At4 1 3
1
1 3 5
A t e:

 At e A3 t3 e 
6
20
2
24 6

2.3. Modied variational iteration method (II)

This modied variational iteration method is made by the classical variational iteration method coupled with Hes polynomials [22].
Consider the following homotopy:

Z
Hu; p 1  pu0 x  u p


~  gsds 0
ksLu Nu

2:15

Let u
obtain

P1

n
n0 p un

and assume that Nu

P1

n0 p

p0 : u0 u0 x
Rx
p1 : u1 x 0 ksLu0 s H0  gsds;
Rx
p2 : u2 x u1 0 ksLu1 s H1 ds;
..
.
Rx
pn1 : un1 un 0 ksLun s Hn ds
..
.

Hn . By comparing the coefcients of the same power of p in Eq. (2.15), we

2:16

8123

Y.-J. Huang, H.-K. Liu / Applied Mathematical Modelling 37 (2013) 81188130

P
i
and get the solution u limp!1 1
i0 ui p .
Given u0 = 2 cos t. Consider the following homotopy:

Z
Hu; p 1  pu0 t  u p


sins  tLu Nuds 0

2:17

P
P1 n
n
Let u 1
n0 p un and assume that Nu
n0 p H n . By comparing the coefcients of the same power of p in Eq. (2.17), we
obtain the following recurrence formula

p0 : u0 u0 t
Rt
p1 : u1 0 sins  tLu0 s H0 ds
Rt
p2 : u2 u1 0 sins  tLu1 s H1 ds
..
.
Rt
pn1 : un1 un 0 sins  tLun s Hn ds

2:18



For the nonlinear item Nut e u2n  1 u0n , the rst Hes polynomial is given as



H0 e u20  1 u00 2e sin 3t

2:19

Substituting (2.19) into (2.18), we obtain the rst two approximations

u1

sins  t2 sin 3sds

u2 u1

3e
e
sin t  sin 3t
4
4

2:20

sins  tLu1 s H1 ds


1 
13e2 cos t 18e2 cos 3t  5e2 cos 5t  72e sin t 12t e2 sin t 24e sin 3t
96

2.4. The set of basis functions for van der Pol equations
It is well known that a real function can be expressed by some different basis functions and basic functions have close
relationships with linear operators (see [11]). From above three methods, we nd that the solutions of Eq. (2.1) have different
2

types of expressions for the second-order linear differential equations Lut ddt2u u and Lut ddt2u. For example, when
2

Lut ddt2u u, the initial approximation u0(t) governed by (2.1) is A cos t and the second-order approximation is

u  2 cos t



3 sin t 1
1
 sin 3t e
13 cos t 18 cos 3t  5 cos 5t 12t sin te2
4
4
96

120t cos t  165 sin t 162 sin 3t  50 sin 5t 7 sin 7te3

1536
672 cos t  840 cos 3t 200 cos 5t  35 cos 7t 3 cos 9t  1680t sin te4

20480

This approximation of (2.1) are expressed by the set of basis functions

ft m cosnt; t m sinntjm P 0; n P 1g
 
which breaks down for t P O e2 because of the existence of the secular term as tn sin t or tn cos t.
2
When Lut ddt2u, the initial approximation u0(t) governed by (2.1) is A and the second-order approximation is

uA

At2 At 4 1 3
1
1 3 5
A t e

 At e A3 t 3 e 
6
20
2
24 6

This solution of (2.1) is approximated as the form polynomials. Obviously, this approximation tends to positive or negative innity as the time tends to innity for any order of tn.
These approximations are valid only of small t, as shown in Figs. 1 and 2 for the following equation (e = 1):
2

d u
dt


 du
u  1  u2
0 ; u0 A;
dt

u0 0 0:

8124

Y.-J. Huang, H.-K. Liu / Applied Mathematical Modelling 37 (2013) 81188130

From the physical points of view, the solution of van der Pol equations is periodic and the rigorous proof for their periodicity can be found in Liu [12]. This implies that the above approximate solutions do not match the principal character of
van der Pol equations. Therefore, it might be better to express a periodic solution by the set of periodic basis functions

fcosnxt; sinnxtjn P 0g;

where x = 2p/T is the frequency, and T is the period of the solution.
In Section 3, we apply the transformation h = xt to (2.1) and obtain
2

x2


 du
u  xe 1  u2
0
dh
dh

d u
2

Here, u(h) can be expressed by the periodic basis functions

fcosnh; sinnhjn P 0g

3. A new modication of the variational iteration method

Our modication is coupled the classical variational iteration method with Hes polynomials. Especially, we expand
P
P1 n
n
n
n
A 1
n0 p An and x
n0 p xn to correct the secular term as e sin h and e cos h. In Section 3.1, we introduce the basic
concepts of the homotopy method as well as Hes polynomials. Our new modication will be illustrated in Section 3.2.
3.1. Basic concept of the homotopy perturbation method
Reviewing the basic concept of HPM [8], we consider the following function:

Au  f r 0;

r2K

3:1

with the boundary conditions of:



@u
0;
B u;
@n

r2C

3:2

where A is a general differential operator, B is a boundary operator, f(r) is a known analytical function and C is the boundary
of domain K. Generally, the operator A can be divided into a linear part L(u) and a nonlinear part N(u). Eq. (3.1) can be written
as:

Lu Nu  f r 0

3:3

According to HPM, we construct the a homotopyv r; p : K  0; 1 ! R which satises:

Hv ; p 1  pLv  Lu0  pAv  f r 0;

p 2 0; 1;

r2K

3:4

where u0 is an initial approximation of (2.8), which satises the boundary conditions. Thus, we have

Hv ; 0 Lv  Lu0 0;

3:5

Hv ; 1 Av  f r 0;

3:6

We use the embedding parameter p as a small parameter, and assume that v in Eq. (3.4) can be written as a power series
of p:

1
X

v i pi v 0 pv 1 p2 v 2   

3:7

i0

If p ? 1, we get an approximation solution of Eq. (3.1):

u limv v 0 v 1 v 2   

3:8

p!1

Here, the nonlinear item N(u) above is considered as the decomposed form

Nu

1
X
pi Hi H0 pH1 p2 H2   

3:9

i0

where Hns are the so-called Hes polynomials [6]

n
X
1 @n
Hn u0 ; . . . ; un
N
pi ui
n
n! @p
i0

;
p0

n 0; 1; 2; . . .

3:10

8125

Y.-J. Huang, H.-K. Liu / Applied Mathematical Modelling 37 (2013) 81188130

3.2. The modication of the variational iteration method

P
P1 n
n
To illustrate our modication, we rst introduce u 1
n0 p un and give Nu
n0 p H n .
According to the variational iteration method, the nth approximate solution dended as

U n1 h U n h

kfLU n s NU n s  gsgds;

n 0; 1; . . .

3:11

The nth approximate solution is also defended as

Un

n
X

ui

3:12

i0

Substituting (3.12) into (3.11), from rst approximate solution, we have

Rh

kfLu0 s Nu0 s  gsgds;

Rh
.
u0 u1 u2 u0 u1 0 kfLu0 u1 s Nu0 u1 s  gsgds..
!
!
n
n1
n1
n1
X
X
X
X
Rh
ui
ui 0 kfL
ui s N
ui s  gsgds

u0 u1 u0

i0

i0

i0

i0

then we get

u1

Rh
0

Rh

kfLu0 s H0 s  gsgds

kfLu0 s Lu1 Nu0 u1 s Nu0 s  Nu0 s  gsgds

Rh
u1 0 kfLu1 s Nu0 u1 s  Nu0 sgds
Rh
.
u1 0 kfLu1 s H1 sgds..
!
!
!
n2
n1
n2
n2
X
X
X
X
Rh
ui s N
ui s  N
ui s  gsgds
un 0 kfL ui s Lun1 s N
u2

i0

un1

Rh

un1

Rh

i0

i0

i0

i0

fkLun1 Hn1 sgds

Note that N(u0 + u1) = H0 + H1, N(u0) = H0 and N

In this manner, we nd

uh; p

i0

!
! )
n1
n2
X
X
kLun1 s N
ui s  N
ui s ds

Pn

i0 ui

Pn

i0 H i .

1
X
pn un u0 u1 p u2 p2 u3 p3    un pn   
n0

Z
u0 p

Z
kfLu0 s H0 s  gsgds u1 p2 p

u2 p3 p

1  pu0  u p

Z
0

ksLu Nu  gsds;

That is,


kfLu1 sp H1 spgds

kfLu2 sp2 H2 sp2 gds    un1 pn p

and obtain

u u0 pu  u0 p

ksLu Nu  gsds 0

Thus, we construct the following homotopy:

Hu; p 1  pu0 h  u p

Z
0

ksLu Nu  gsds 0


kfLun1 spn1 Hn1 spn1 gds . . .

8126

Y.-J. Huang, H.-K. Liu / Applied Mathematical Modelling 37 (2013) 81188130

3.3. Applying the modication to the van der Pol equation

Besides, in order to obtain a uniform expansion for the solution of (1.1), we give the modication of variational iteration
method for (1.1).
First, we let

h xet

3:13

where x(e) = 2p/T(e) is the frequency. Note x(0) = 1.

Eq. (1.1) becomes
2

x2


 du
0
u  xe 1  u2
dh
dh

d u

3:14

Consider the following homotopy:

Hv ; p 1  pu0 h  v p


ksLv Nv ds 0

3:15

0
2

where p 2 [0, 1]. Because we want to nd the periodic approximate solution, we assume that Lv ddhv2 v ,

 2


Nv x2  1 ddhv2 xe v 2  1 ddhv , so that we can identify k sins  h as an optimal Lagrange-multiplier and u0 = A cos
h is an initial approximation of (3.14).
The solution as well as the frequency and the amplitude, u, x and A, are expanded, respectively in power series of p as
follows:

v h; e; p

1
X

v n h; epn

3:16

n0

xe; p

1
X

xn epn

3:17

n0

Ae; p

1
X
An epn

3:18

n0

If p ? 1, we get an approximation solution of Eq. (3.14):

uh; e limv h; e; p v 0 v 1 v 2   
p!1

xe limxe; p x0 x1 x2   
p!1

Ae limAe; p A0 A1 A2   
p!1

Here, the nonlinear item N(v) above is considered as the following decomposed form

Nv

1
X
pi Hi H0 pH1 p2 H2   

3:19

i0

where Hns are the so-called Hes polynomials

n
X
1 @n
Hn v 0 ; . . . ; v n
N
pi v i
n! @pn
i0

n 0; 1; 2; :::

;
p0

Substituting (3.16)(3.19) into (3.15) and comparing the coefcients of the same power of p, we obtain

po : v 0 A0 cos h
p1 : v 1 A1 cos h
..
.

Rh
0

sins  hLv 0 s H0 ds

pn1 : v n1 v n An  An1 cos h

Rh
0

sins  hLun s Hn ds n 1; 2; . . .

 2
For the nonlinear item Nv x2  1 ddhv2 xev 2  1 ddhv , the rst Hes polynomials is given as

H0 e



ex0 A0 4  A20



ex0 A30
v 20  1 v 00
sin 3h
sin h A0 1  x20 cos h 
4
4

3:20

Y.-J. Huang, H.-K. Liu / Applied Mathematical Modelling 37 (2013) 81188130

8127

In order to ensure that no secular term appears in the next iteration, we equate the coefcients of sin h and cos h equal to
zero. Thus we require

x0 1 and A0 2
which imply

H0 2e sin 3h

3:21

Substituting (3.21) into (3.20), we obtain

v 1 A1 cos h

sins  h2e sin 3hds A1 cos h

0

3e
e
sin h  sin 3h:
4
4

3:22

Hence, the next Hes polynomials is given as

!
"
#
2
 d2 v

 dv
@ 
d v1
dv 1
0
0
2
x0 x1 p2  1 

x
p

v
p


1


p
0
1
0
1
2
2
@p
dt
dt
dt
dt
p0


2
1 2
3 2
5e
4x1  e cos h e cos 3h 
cos 5h  2A1 e sin h  3A1 2x1 e sin 3h
4
2
4

H1

In order to ensure that no secular term appears in the next iteration, we equate the coefcients of sin h and cos h equal to
zero again. Thus we require

x1

e2
and A1 0
16

which imply

H1

3 2
5
1
e cos 3h  e2 cos 5h e3 sin 3h
2
4
8

3:23

Substituting (3.23) into (3.20), we obtain

v 2 v 1 A2  A1 cos h

sins  hLv 1 H1 ds

A2 cos h 

13 2
3
5
3
1
e cos h e2 cos 3h  e2 cos 5h  e3 sin h e3 sin 3h
96
16
96
64
64

Hence, the next Hes polynomials is given as

H2

1 
1536x2 cos h 9e4 cos h  72e4 cos 3h 60e4 cos 5h  768A2 e sin h 8e3 sin h  1152A2 e sin 3h
384

768x2 e sin 3h 216e3 sin 3h  560e3 sin 5h 224e3 sin 7h

As the same reason, that is in order to ensure that no secular terms appear in the next iteration, equating the coefcient of
sin h and cos h equal to zero yields

x2

3e4
e2
and A2
512
96

In the process of calculations, we focus on nding out the coefcients of sin h and cos h to eliminate the secular terms. By
substituting v0, v1, . . ., vn into (3.20), we nd that there are no sin h and cos h in Ln but in the Hes polynomials Hn. From Hn, we
obtain xn and An iteratively.In this manner, the rest of components of the iteration formula (3.20) will be obtained by the
mathematical soft wares, such as Mathematica in this paper.
For example, in the 6-th order approximation

ue; h 2 cos h



3 sin h 1
1
 sin 3h e
12 cos h 18 cos 3h  5 cos 5he2
4
4
96

79 sin h  27 sin 3h 20 sin 5h  4 sin 7he3

2304
3285 cos h  16920 cos 3h 21700 cos 5h  10745 cos 7h 1647 cos 9he4

552960
194565 sin h  582975 sin 3h 1322125 sin 5h 1106210 sin 7h 402678 sin 9h  49797 sin 11he5

66355200
66355200
1215 cos h 10170 cos 3h  19825 cos 5h 10745 cos 7h  1647 cos 9he6

2211840

8128

Y.-J. Huang, H.-K. Liu / Applied Mathematical Modelling 37 (2013) 81188130

515 sin h 163 sin 3h  140 sin 5h 28 sin 7he7 12 cos h  18 cos 3h 5 cos 5he8

393216
49152


105 sin h 35 sin 3h 9
10
e Oe


2097152 2097152

with the amplitude A and the frequency x given by

A2

e2
96

1033e4
329e6
e8


Oe9
552960 1105920 49152

and

x1

e2
16

17e4
35e6
35e8
63e10


Oe11
3072 884736 786432 8388608

Besides, when e is large, we introduce a new parameter q, dened by the transformation qe 1e e. Then we expand our
results in the power series
of q.


e2 17e4 35e6 35e8  63e10
in power series of q and obtain
We expand ex e 1  16
3072
884736
786432
8388608

ex q q2

15q3 13q4 1937q5 1237q6 128771q7 104875q8 2558501q9 1295279q10




Oq11
16
16
3072
3072
884736
884736
7077888
2359296

which shall be compared in Section 4.

4. Numerical comparison
At rst the numerical solutions are compared with the approximate solutions of our modied variational iteration method in the cases of e = 1, A = 2 and e = 2, A = 2, where the numerical solutions and the approximated results are obtained by
using Mathematica 7. Observing Figs. 3 and 4, we nd that our 6-th iteration approximation is close to the numerical solution. In the same case of e = 1, A = 2 we demonstrated that the variational iteration method as well as other modications are
only valid of small t in Figs. 1 and 2. As a result, our modication provided a better way of approximating the solution for the
van der Pol equation when parameters are small.
We then consider the cases of large parameters. In Table 1, we compare the period of (1.1) with Andersens result [3],
Bunomos result [4] and Lius result [12] for the cases of e = 1, 2, 3, 4, 5, 10, 20, 100. We get the numerical values of other
methods from Andersen [3]. These values are also used by Bunomo [4] and Liu [12]. Our result is using 10 terms in the
expansion, Andersens result is using 82 terms, Lius result is using 8 terms and Bunomos result using 203 terms. Here,
we get the period up to 10 terms after 6 iterations while Liu [12] gets 8 terms after 8 iterations. By comparing with the
numerical results, we nd that our 6-th iteration approximation is more close to the numerical result than others. To make
nj
it clearly, we indicate the error of period in Table 2. The error in this paper is dened by the formula jT rTT
 100%, where Tr is
n
the approximate result and Tn is the numerical result. Then, we nd that our error of period T(e) for e 6 5 is less than 1%.
When e P 10, we nd that our error of period T(e) is less than 1.5% but other methods are all greater than ours. Especially,
our error of period T(e) is 0.51% while Lius error is 3.10%, Andersens error is 10.16% and Bunomos error is 3.86% when
e = 100. This implies that our proposed method is more accurate than others. In summary, our modication uses less itera-

10

20

30

40

50

2
Fig. 3. Comparison of the numerical solution with the 6 iterations approximation of modied variational iteration method when e = 1 and A = 2. (1) Dashed
line: the 6th-order approximation and (2) solid line: the numerical solution.

8129

Y.-J. Huang, H.-K. Liu / Applied Mathematical Modelling 37 (2013) 81188130

10

20

30

40

50

2
Fig. 4. Comparison of the numerical solution with the 6 iterations approximation of modied variational iteration method when e = 2 and A = 2. (1) Dashed
line: the 6th-order approximation and (2) solid line: the numerical solution.

tions to get an accurate approximation than the other methods. Therefore, our modication is effective. However, the maximum error always occurs at the points of 2 + 4n, n = 1,2,. . . in Fig. 4. This might be interesting to nd out the reason in future
studies.

5. Conclusion
In this paper, we nd the variational iteration method is not applicable for the van der Pol equation because of the existence of the secular terms. We have then introduce a simple transformation to overcome this disadvantage and modify the
variational iteration method by coupling the classical variational iteration method with Hes polynomials. We nd that our 6
iterations are more accurate than other high-order approximations from perturbation methods. That is, our modied variational iteration method is more effective. We also get an accurate approximate period of (1.1) for the large e by our proposed
method and show that our approximation is close to the numerical solution when e = 1 and e = 2. In the future research, it
will be interesting to extend our proposed method to solve the familiar van der Pol Dufng equation.
References
[1] T. Abassy, M. Eltawil, H. Elzzoheiry, Toward a modied variational iteration method, J. Comput. Appl. Math. 207 (2007) 137147.
[2] T. Abassy, Modied variational iteration method (nonlinear homogeneous initial value problem), Comput. Math. Appl. 59 (2010) 912918.
[3] C.M. Andersen, J.F. Geer, Power series expansions for the frequency and period of the limit cycle of the van der Pol equation, Soc. Ind. Appl. Math. 42
(1982) 678693.
[4] A. Buonomo, The periodic solution of van der Pols equation, Soc. Ind. Appl. Math. 59 (1998) 156171.
[5] F. Geng, A modied variational iteration method for solving Riccati differentail equations, Comput. Math. Appl. 60 (2010) 18681872.
[6] A. Ghorbani, Beyond adomian polynomials: He polynomials, Chaos Solition. Fract. 39 (2009) 14861492.
[7] J.H. He, Variational iteraion methoda kind of non-linear analytical technique: some examples, Int. J. Non-Linear Mech. 34 (1999) 699708.
[8] J.H. He, Homotopy perturbation method: a new nonlinear analytical technique, Appl. Math. Comput. 135 (2003) 7379.
[9] J.H. He, Variational iteraion methodsome recent results and new interpretaions, J. Comput. Appl. Math. 207 (2007) 317.
[10] L. Jin, Application of modied variational iteration method to the Bratu-type problems, Int. J. Contemp. Math. Sci. 5 (2010) 153158.
[11] S. Liao, Y. Tan, A general approach to obtain series solutions of nonlinear differential equations, Stud. Appl. Math. 119 (2007) 297354.
[12] K.Y. Liu, Homotopy Perturbation Method for van der Pol Equation, National Chengchi University, Taiwan, 2005.
[13] J. Lu, Variational iteration method for solving two-point boundary value problems, J. Comput. Appl. Math. 207 (2007) 9295.
[14] S. Momani, S. Abuasad, Z. Odibat, Variational iteration method for solving nonlinear boundary value problems, Appl. Math. Comput. 183 (2006) 1351
1358.
[15] M.A. Noor, S.T. Mohyud-Din, Modied variational iteration method for heat and wave-like equations, Acta Appl. Math. 104 (2008) 257269.
[16] M.A. Noor, S.T. Mohyud-Din, M. Tahir, Modied variational iteration method for ThomasFermi equation, World Appl. Sci. J. 4 (2008) 479486.
[17] M.A. Noor, S.T. Mohyud-Din, Modied variational iteration method for Goursat and Laplace problems, World Appl. Sci. J. 4 (2008) 487498.
[18] M.A. Noor, S.T. Mohyud-Din, Modied variational iteration method for a boundary layer problem in unbounded domain, Int. J. Nonlinear Sci. 7 (2009)
426430.
[19] M.A. Noor, K.I. Noor, S.T. Mohyud-Din, Modied variational iteration technique for solving singular fourth-order parabolic partial differential
equations, Nonlinear Anal. 71 (2009) 630640.
[20] Ronald E. Mickens, An Introduction to Nonlinear Oscillations, Combridge University Press, 1981.
[21] S.L. Ross, Differential Equations, Third Edition., Wiley, New york, 1984.
[22] X. Shao, D. Han, On some modied variational iteration methods for solving the one-dimensional sineGordon equation, Int. J. Comput. Math. 88 (5)
(2011) 969981.
[23] Sunil Kumar, H. Kocak, Ahmet Yildirim, A fractional model of gas dynamics equations and its analytical approximate solution using Laplace transform,
Z. Naturforsch. 67a (2012) 389396.
[24] Sunil Kumar, Ahmet Yildirim, Y. Khan, W. Leilei, A fractional model of the diffusion equation and its analytical solution using Laplace transform, Sci.
Irantica 19 (4) (2012) 11171123.

8130

Y.-J. Huang, H.-K. Liu / Applied Mathematical Modelling 37 (2013) 81188130

[25] Sunil Kumar, Yasir Khan, Ahmet Yildirim, A mathematical modelling arising in the chemical systems and its approximate numerical solution, Asia Pac.
J. Chem. Eng. (2011), http://dx.doi.org/10.1002/apj.647.
[26] Sunil Kumar, Om P. Singh, Numerical inversion of the Abel integral equation using homotopy perturbation method, Z. Naturforsch. 65a (2010) 677
682.
[27] B. Van der Pol, On relaxationosciillations, Phil. Mag. 2 (1926) 978992.
[28] S. Varedi, M. Hosseini, M. Rahimi, D. Ganji, Hes variational iteration method for solving a semi-linear inverse parabolic equation, Phys. Lett. A (2007)
16.
[29] A. Wazwaz, A comparison between the variational iteration method and Adomain decomposition method, J. Comput. Appl. Math. 207 (2006) 129136.