Applied Mathematics and Computation 135 (2003) 73–79

www.elsevier.com/locate/amc

Homotopy perturbation method:

a new nonlinear analytical technique q
Ji-Huan He
Shanghai Institute of Applied Mathematics and Mechanics, Shanghai University,
149 Yanchang Road, Shanghai 200072, China

Abstract
In this paper, a new perturbation method is proposed. In contrast to the traditional
perturbation methods, this technique does not require a small parameter in an equation.
In this method, according to the homotopy technique, a homotopy with an imbedding
parameter p 2 ½0; 1 is constructed, and the imbedding parameter is considered as a
‘‘small parameter’’, so the method is called the homotopy perturbation method, which
can take the full advantages of the traditional perturbation methods and homotopy
techniques. To illustrate its effectiveness and its convenience, a Duffing equation with
high order of nonlinearity is used; the result reveals that its first order of approximation
obtained by the proposed method is valid uniformly even for very large parameter, and
is more accurate than the perturbation solutions.
Ó 2002 Elsevier Science Inc. All rights reserved.

Keywords: Perturbation methods; Homotopy; Duffing equation; Nonlinearity

1. Introduction

Various perturbation methods have been widely applied to solve nonlinear

problems. Unfortunately, the traditional perturbation techniques are based on
the assumption that a small parameter must exist, which is too over-strict to
find wide application, for most nonlinear equations have no small parameter at
all. Many new techniques have been proposed recently to eliminate the ‘‘small

q
This work is supported by the China National Natural Science Foundation.
E-mail address: [email protected] (J.-H. He).

0096-3003/02/$ - see front matter Ó 2002 Elsevier Science Inc. All rights reserved.
PII: S 0 0 9 6 - 3 0 0 3 ( 0 1 ) 0 0 3 1 2 - 5
74 J.-H. He / Appl. Math. Comput. 135 (2003) 73–79

parameter’’ assumption, such as the artificial parameter method proposed by

Liu [1], the homotopy analysis method proposed by Liao [2,3], and the vari-
ational iteration method proposed by He [4–6]. A review of recently developed
nonlinear analysis methods can be found in detail in [7].
In this paper we will, taking full advantage of Liu’s method and Liao’s
method, propose a new perturbation method called the homotopy perturbation
method [8,9].
To illustrate the basic ideas of the new method, we consider the following
nonlinear differential equation,
AðuÞ þ f ðrÞ ¼ 0; r 2 X; ð1Þ
with boundary conditions
Bðu; ou=onÞ ¼ 0; r 2 C; ð2Þ
where A is a general differential operator, B is a boundary operator, f ðrÞ is a
known analytic function, and C is the boundary of the domain X.
The operator A can, generally speaking, be divided into two parts L and N ,
where L is linear, while N is nonlinear. Eq. (1), therefore, can be rewritten as
follows:
LðuÞ þ N ðuÞ  f ðrÞ ¼ 0: ð3Þ
In case the nonlinear equation (1) has no ‘‘small parameter’’, we can embed
an artificial parameter [1] in the equation:
LðuÞ þ pN ðuÞ  f ðrÞ ¼ 0; ð4Þ
where p is called an artificial parameter.
We call the above equation (4) a perturbation equation with an artificial
parameter. The artificial parameter method [1] assumes that the approximation
of Eq. (4) can be expressed as a series of the power of p, i.e.
u ¼ u0 þ pu1 þ p2 u2 þ p3 u3 þ ð5Þ
When p ! 1 Eq. (4) corresponds to (3), and (5) becomes the approximate
solution of Eq. (3). Some interesting results have been achieved by this method
[1]. However, in this method, the embedding parameter is quite casual, and
often enough the approximations obtained by this method will not be uniform.
For example
y 0 þ y 2 ¼ 1; t > 0; ð6Þ
with initial condition yð0Þ ¼ 0.
We embed an artificial parameter p in Eq. (6) as follows:
y 0 þ py 2 ¼ 1: ð7Þ
 J.-H. He / Appl. Math. Comput. 135 (2003) 73–79 75

Its approximate solution in the power of p is not uniformly valid, so some

special techniques must be applied. In [1], Liu embeds the artificial parameter
as follows:

y 0  ð1  yÞð1 þ pyÞ ¼ 0; ð8Þ

yielding a uniform approximation.

So we can see clearly that, in Liu’s method, there is an uncertainty about an
appropriate artificial parameter. To construct a universal perturbation equa-
tion with appropriate artificial parameter, we will use homotopy in topology
[2,3].
By the homotopy technique proposed by Liao in [2,3], we construct a ho-
motopy of Eq. (1) vðr; pÞ : X ½0; 1 ! R which satisfies

Hðv; pÞ ¼ ð1  pÞ½LðvÞ  Lðu0 Þ þ p½AðvÞ  f ðrÞ ¼ 0;

ð9aÞ
p 2 ½0; 1; r 2 X

or

Hðv; pÞ ¼ LðvÞ  Lðu0 Þ þ pLðu0 Þ þ p½N ðvÞ  f ðrÞ ¼ 0; ð9bÞ

where p 2 ½0; 1 is an embedding parameter, and u0 is an initial approximation

which satisfies the boundary conditions.
Obviously, from (9a), (9b), we have

Hðv; 0Þ ¼ LðvÞ  Lðu0 Þ ¼ 0; ð10Þ

Hðv; 1Þ ¼ AðvÞ  f ðrÞ ¼ 0: ð11Þ

The changing process of p from zero to unity is just that of vðr; pÞ from u0 to
uðrÞ. In topology, this is called deformation, and LðvÞ  Lðu0 Þ, AðvÞ  f ðrÞ are
homotopic.
Here the embedding parameter is introduced much more naturally, unaf-
fected by artificial factors; further it can be considered as a small parameter for
0 6 p 6 1. So it is very natural to assume that the solution of (9a), (9b) can be
expressed as

v ¼ v0 þ pv1 þ p2 v2 þ ð12Þ

The approximate solution of (1), therefore, can be readily obtained:

u ¼ lim v ¼ v0 þ v1 þ v2 þ ð13Þ
p!1

The convergence of the series of (13) has been proved in the [8,9].
76 J.-H. He / Appl. Math. Comput. 135 (2003) 73–79

2. Some examples

We first consider a very simple example,

y 0 þ y 2 ¼ 0; x P 0; x 2 X; yð0Þ ¼ 1; ð14Þ
with an exact solution 1=ð1 þ xÞ.
We can construct the following homotopy X ½0; 1 ! R which satisfies
ð1  pÞðY 0  y00 Þ þ pðY 0 þ Y 2 Þ ¼ 0; p 2 ½0; 1; x 2 X; ð15Þ
with initial approximation y0 ¼ 1.
Suppose the solution of Eq. (15) has the form
Y ¼ Y0 þ pY1 þ p2 Y2 þ ð16Þ
Substituting (16) into (15), and equating the terms with the identical powers
of p,
p0 : Y00 ¼ y00 ; ð17aÞ
p1 : Y10 þ y00 þ Y02 ¼ 0; Y1 ð0Þ ¼ 0; ð17bÞ
p2 : Y20 þ 2Y0 Y1 ¼ 0; Y2 ð0Þ ¼ 0: ð17cÞ
For simplicity we always set Y0 ¼ y0 ¼ 1. Accordingly we have Y1 ¼ x, and
Y2 ¼ x2 ; thus we have a second order of approximation of Eq. (14)
y2 ¼ Y0 þ Y1 þ Y2 ¼ 1  x þ x2 : ð18Þ
Now we will consider a typical nonlinear equation, Duffing equation with
nonlinearity of high order
d2 u
þ u þ eu5 ¼ 0; t 2 X; uð0Þ ¼ A; u0 ð0Þ ¼ 0: ð19Þ
dt2
We construct a homotopy X ½0; 1 ! R which satisfies
5
LðvÞ  Lðu0 Þ þ pLðu0 Þ þ pev ¼ 0; ð20Þ
2 2
where Lu ¼ d u=dt þ u.
We assume the initial approximation of (19) has the form
u0 ðtÞ ¼ A cos at; ð21Þ
where aðeÞ is a nonzero unknown constant with að0Þ ¼ 1.
Supposing the approximate solution of Eq. (20) has the form of (12), by the
same manipulation, we have
Lðv0 Þ  Lðu0 Þ ¼ 0; ð22Þ
Lðv1 Þ þ Lðu0 Þ þ ev50 ¼ 0; v01 ð0Þ ¼ v1 ð0Þ ¼ 0: ð23Þ
 J.-H. He / Appl. Math. Comput. 135 (2003) 73–79 77

From (22) and (23), we can obtain

v0 ¼ u0 ¼ A cos at; ð24Þ


d2 v1 2 5 4 eA5
þ v 1 þ A  a þ 1 þ eA cos at þ ðcos 5at þ 5 cos 3atÞ
dt2 8 16
¼ 0: ð25Þ
The solution of (25) can be readily obtained by the so-called variational
iteration method [4–6] proposed by the present author:
Z t 

5
v1 ðtÞ ¼ sinðs  tÞ A  a2 þ 1 þ eA4 cos as
0 8
5

eA
þ ðcos 5as þ 5 cos 3asÞ ds
16


5 A eA5
¼  a2 þ 1 þ eA4 2 ðcos at  cos tÞ þ
8 a 1 16ð25a2  1Þ
5eA5
ðcos 5at  cos tÞ þ ðcos 3at  cos tÞ: ð26Þ
16ð9a2  1Þ
The constant a can be identified by various methods such as methods of
weighted residuals (least squares method, method of collocation, Galerkin
method). Here we set the coefficient of cos t to zero in order to eliminate the
secular term which may occur in the next iteration


2 5 4 A eA5 5eA5
  a þ 1 þ eA   ¼ 0: ð27Þ
8 a2  1 16ð25a2  1Þ 16ð9a2  1Þ
Thus we have


5 A eA5
v1 ðtÞ ¼  a2 þ 1 þ eA4 2 cos at þ cos 5at
8 a 1 16ð25a2  1Þ
5eA5
þ cos 3at; ð28Þ
16ð9a2  1Þ
with a defined as in (27).
If, for example, the first order of approximation is sufficient, then we have
u1 ðtÞ ¼ v0 ðtÞ þ v1 ðtÞ
5eA5 eA5 5eA5
¼ 2
cos at þ 2
cos 5at þ
8ða  1Þ 16ð25a  1Þ 16ð9a2  1Þ
cos 3at; ð29Þ
with a defined as in (27).
The approximation of a2 in Eq. (27) can be written down as
78 J.-H. He / Appl. Math. Comput. 135 (2003) 73–79

5
a2 ¼ 1 þ eA4 þ Oðe2 Þ: ð30Þ
8
Its period can be approximately expressed as follows:
2p
T ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ; ð31Þ
1 þ 5eA4 =8
while the approximation and its period obtained by perturbation method [10]
are
2p
u ¼ A cosð1 þ 5eA4 =16Þt; T ¼ ; ð32Þ
1 þ 5eA4 =16
and the exact period can be readily obtained as follows:
Z p=2
4 dx
Tex ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ; ð33Þ
1 þ ð1=3ÞeA 0 4 1 þ k cos2 x þ k cos4 x
with k ¼ ð1=3ÞeA4 =ð1 þ ð1=3ÞeA4 Þ.
It should be specially pointed out that the perturbation formula (32) is valid
only for small parameter e, whereas (31) is valid not only for the small pa-
rameter, but also for very large parameter; even in case eA4 ! 1, we have
pffiffiffiffiffiffiffiffiffiffi Z pffiffiffiffiffiffiffiffiffiffi
Tex 2 15=8 p=2 dx 2 15=8
lim ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ¼ 1:21433
eA4 !1 T p 0 1 þ cos2 x þ cos4 x p
¼ 1:05856:
Therefore, for any value of e, it can be easily proved that 0 6 jðTex  T Þj=
Tex 6 5:8%, so the approximate solution obtained by the proposed method is
uniformly valid for any value of e!

3. Conclusion

A novel method called homotopy perturbation method is proposed wherein

the results at the first order of approximations are much more accurate than
the traditional solutions at second order of approximation.

References

[1] G.L. Liu, New research directions in singular perturbation theory: artificial parameter
approach and inverse-perturbation technique, in: Conference of 7th Modern Mathematics and
Mechanics, Shanghai, 1997.
[2] S.J. Liao, An approximate solution technique not depending on small parameters: a special
example, Int. J. Non-Linear Mech. 30 (3) (1995) 371–380.
 J.-H. He / Appl. Math. Comput. 135 (2003) 73–79 79

[3] S.J. Liao, Boundary element method for general nonlinear differential operators, Eng. Anal.
Boundary Elem. 20 (2) (1997) 91–99.
[4] J.H. He, Approximate analytical solution for seepage flow with fractional derivatives in porous
media, Comput. Meth. Appl. Mech. Eng. 167 (1998) 57–68.
[5] J.H. He, Approximate solution for nonlinear differential equations with convolution product
nonlinearities, Comput. Meth. Appl. Mech. Eng. 167 (1998) 69–73.
[6] J.H. He, Variational iteration method: a kind of nonlinear analytical technique: some
examples, Int. J. Non-Linear Mech. 34 (4) (1999) 699–708.
[7] J.H. He, A review on some new recently developed nonlinear analytical techniques, Int. J.
Nonlinear Sci. Numer. Simul. 1 (1) (2000) 51–70.
[8] J.H. He, Homotopy perturbation technique, Comput. Meth. Appl. Mech. Eng. 178 (1999)
257–262.
[9] J.H. He, A coupling method of homotopy technique and perturbation technique for nonlinear
problems, Int. J. Non-Linear Mech. 35 (1) (2000) 37–43.
[10] A.H. Nayfeh, Problems in Perturbation, Wiley, New York, 1985.