See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/228644456

Some Asymptotic Methods for Strongly Nonlinear Equations

Article in International Journal of Modern Physics B · April 2006

DOI: 10.1142/S0217979206033796

CITATIONS READS

2,204 4,742

1 author:

Ji-Huan He
Soochow University (PRC)
689 PUBLICATIONS 51,648 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

I am now editing a journal: Nonlinear Science Letters A. View project

adamuy View project

All content following this page was uploaded by Ji-Huan He on 30 December 2014.

The user has requested enhancement of the downloaded file.

April 13, 2006 10:4 WSPC/140-IJMPB 03379

International Journal of Modern Physics B

Vol. 20, No. 10 (2006) 1141–1199
c World Scientific Publishing Company

SOME ASYMPTOTIC METHODS FOR STRONGLY

NONLINEAR EQUATIONS

JI-HUAN HE
College of ience, Donghua University,
Shanghai 200051, People’s Republic of China
jhhe@dhu.edu.cn

Received 28 February 2006

This paper features a survey of some recent developments in asymptotic techniques,

which are valid not only for weakly nonlinear equations, but also for strongly ones.
Further, the obtained approximate analytical solutions are valid for the whole solution
domain. The limitations of traditional perturbation methods are illustrated, various
modified perturbation techniques are proposed, and some mathematical tools such as
variational theory, homotopy technology, and iteration technique are introduced to over-
come the shortcomings.
In this paper the following categories of asymptotic methods are emphasized:
(1) variational approaches, (2) parameter-expanding methods, (3) parameterized pertur-
bation method, (4) homotopy perturbation method (5) iteration perturbation method,
and ancient Chinese methods.
The emphasis of this article is put mainly on the developments in this field in China
so the references, therefore, are not exhaustive.

Keywords: Analytical solution; nonlinear equation; perturbation method; variational

theory; variational iteration method; homotopy perturbation; ancient Chinese mathe-
matics.

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1142
2. Variational Approaches . . . . . . . . . . . . . . . . . . . . . . . 1144
2.1. Ritz method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1144
2.2. Soliton solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1147
2.3. Bifurcation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1149
2.4. Limit cycle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1150
2.5. Variational iteration method . . . . . . . . . . . . . . . . . . . . . . . 1153
3. Parameter-Expanding Methods . . . . . . . . . . . . . . . . . . . . 1156
3.1. Modifications of Lindstedt–Poincare method . . . . . . . . . . . . . . . 1156
3.2. Bookkeeping parameter . . . . . . . . . . . . . . . . . . . . . . . . . . 1161
4. Parametrized Perturbation Method . . . . . . . . . . . . . . . . . . 1165
5. Homotopy Perturbation Method . . . . . . . . . . . . . . . . . . . 1171
5.1. Periodic solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1173
5.2. Bifurcation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1178

1141
April 13, 2006 10:4 WSPC/140-IJMPB 03379

1142 J.-H. He

6. Iteration Perturbation Method . . . . . . . . . . . . . . . . . . . . 1179

7. Ancient Chinese Methods . . . . . . . . . . . . . . . . . . . . . . 1185
7.1. Chinese method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1185
7.2. He Chengtian’s method . . . . . . . . . . . . . . . . . . . . . . . . . . 1188
8. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1193
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1193

1. Introduction
With the rapid development of nonlinear science, there appears an ever-increasing
interest of scientists and engineers in the analytical asymptotic techniques for non-
linear problems. Though it is very easy for us now to find the solutions of linear
systems by means of computer, it is, however, still very difficult to solve nonlinear
problems either numerically or theoretically. This is possibly due to the fact that
the various discredited methods or numerical simulations apply iteration techniques
to find their numerical solutions of nonlinear problems, and nearly all iterative
methods are sensitive to initial solutions,14,27,40,81,101,102,107 so it is very difficult
to obtain converged results in cases of strong nonlinearity. In addition, the most
important information, such as the natural circular frequency of a nonlinear oscil-
lation depends on the initial conditions (i.e. amplitude of oscillation) will be lost
during the procedure of numerical simulation.
Perturbation methods8,11,15,18,22,94,98 – 100 provide the most versatile tools avail-
able in nonlinear analysis of engineering problems, and they are constantly being
developed and applied to ever more complex problems. But, like other nonlinear
asymptotic techniques, perturbation methods have their own limitations:
(1) Almost all perturbation methods are based on such an assumption that a small
parameter must exist in an equation. This so-called small parameter assumption
greatly restricts applications of perturbation techniques, as is well known, an
overwhelming majority of nonlinear problems, especially those having strong
nonlinearity, have no small parameters at all.
(2) It is even more difficult to determine the so-called small parameter, which seems
to be a special art requiring special techniques. An appropriate choice of small
parameter may lead to ideal results, however, an unsuitable choice of small
parameter results in badly effects, sometimes serious ones.
(3) Even if a suitable small parameter exists, the approximate solutions solved by
the perturbation methods are valid, in most cases, only for the small values
of the parameter. For example, the approximations solved by the method of
multiple scales are uniformly valid as long as a specific system parameter is
small. However, we cannot rely fully on the approximations because there is no
criterion on how small the parameters should be. Thus checking numerically
and/or experimentally the validity of the approximations is essential.83 For
example, we consider the following two equations:113

u0 + u − ε = 0 , u(0) = 1 (1.1)
April 13, 2006 10:4 WSPC/140-IJMPB 03379

Some Asymptotic Methods for Strongly Nonlinear Equations 1143

and
u0 − u − ε = 0 , u(0) = 1 . (1.2)
The solution of Eq. (1.1) is u(t) = ε + (1 − ε)e−t, while the unperturbed solution
of Eq. (1.1) is u0 (t) = e−t . In case ε  1, the perturbation solution leads to high
accuracy, for |u(t) − u0 (t)| ≤ ε.
By similar analysis for Eq. (1.2), we have
|u(t) − u0 (t)| = ε|1 − et | . (1.3)
It is obvious that u0 can never be considered as an approximate solution of
Eq. (1.2) in case t  1, so the traditional perturbation method is not valid for this
case even when ε  1.
In what follows, we should introduce some new developed methods for problem
solving in areas where traditional techniques have been unsuccessful.
There exist some alternative analytical asymptotic approaches, such as the non-
perturbative method,33 weighted linearization method,10,105 Adomian decompo-
sition method,1,9,35,118,119 modified Lindstedt–Poincare method,28,65 – 67,91 varia-
tional iteration method,49,70,74 – 77 energy balance methods,63 tanh-method,5,13,36
F -expansion method106,115,116 and so on. Just recently, some new perturbation
methods such as artificial parameter method,69,88 δ-method,11,16 perturbation in-
cremental method,26 homotopy perturbation method,44 – 48,50 – 52,62,72 parameter-
ized perturbation method,73 and bookkeeping artificial parameter perturbation
method64 which does not depend on on the small parameter assumption are pro-
posed. A recent study68 also reveals that the numerical technique can also be pow-
erfully applied to the perturbation method. A review of perturbation techniques in
phase change heat transfer can be found in Ref. 12, and a review of recently de-
veloped analytical techniques can be found in Ref. 71. Recent advance in nonlinear
dynamics, vibrations and control in China can be found in Ref. 93.
A wide body of literature dealing with the problem of approximate solutions
to nonlinear equations with various different methodologies also exists. Although
a comprehensive review of asymptotic methods for nonlinear problems would be
in order at this time, this paper does not undertake such a task. Our emphasis is
mainly put on the recent developments of this field in China, hence, our references
to literature will not be exhaustive. Rather, our purpose in this paper is to present a
comprehensive review of the past and current work on some new developed asymp-
totic techniques for solving nonlinear problems, with the goal of setting some ideas
and improvements which point toward new and interesting applications with these
methods. The content of this paper is partially speculative and heuristic, but we
feel that there is enough theoretical and computational evidence to warrant further
investigation and most certainly, refinements and improvements, should the reader
choose to pursue the ideas.
In this review article, we limit ourselves to nonlinear order differential equations,
most nonlinear partial differential equations can be converted into order differential
April 13, 2006 10:4 WSPC/140-IJMPB 03379

1144 J.-H. He

ones by some linear translation, for example, the KdV equation

ut + uux + εuxxx = 0 . (1.4)
If the wave solution is solved, we can assume that u(x, t) = u(ξ) = u(x − ct), where
c is the wave speed. Substituting the above traveling wave solution into Eq. (1.4),
we have
−cuξ + uuξ + εuξξξ = 0 . (1.5)
Integrating Eq. (1.5), we have an ordinary differential equation:
1
−cu + u2 + εuξξ = D , (1.6)
2
where D is an integral constant.
In Sec. 2, variational approaches to soliton solution, bifurcation, limit cycle, and
period solutions of nonlinear equations are illustrated, including the Ritz method,
energy method, variational iteration method; In Sec. 3, we first introduce some
modifications of Lindstedt–Poincare method, then parameter-expanding methods
are illustrated where a parameter (or a constant) is expanded into a series in ε which
can be a small parameter in the equation, or an artificial parameter, or a book-
keeping parameter; Sec. 4 introduces parametrized perturbation method where the
expanding parameter is introduced by a linear transformation. Homotopy perturba-
tion method is systemically illustrated in Sec. 5. In Sec. 6, the iteration technology
is introduced to the perturbation method; Sec. 7 briefly introduces some ancient
Chinese methods and their modern applications are illustrated. Finally, conclusions
are made in Sec. 8.

2. Variational Approaches
2.1. Ritz method
Variational methods38,55 – 57,89,90,121 such as Raleigh–Ritz and Bubnov–Galerkin
techniques have been, and continue to be, popular tools for nonlinear analysis.
When contrasted with other approximate analytical methods, variational methods
combine the following two advantages:
(1) They provide physical insight into the nature of the solution of the problem.
For example, we consider the statistically unstable Duffing oscillator, defined
by

u00 − u + εu3 = 0 . (2.1)

Its variational formulations can be obtained with ease
Z  
1 02 1 2 1 4
J(u) = − u − u + εu dt . (2.2)
2 2 4
If we want to search for a period solution for Eq. (2.1), then from the energy integral
(2.2) it requires that the potential − 12 u2 + 14 εu4 must be positive for all t > 0, so
April 13, 2006 10:4 WSPC/140-IJMPB 03379

Some Asymptotic Methods for Strongly Nonlinear Equations 1145

an oscillation about the origin will occur only if εA2 > 2, where A is the amplitude
of the oscillation.

(2) The obtained solutions are the best among all the possible trial-functions. As
an illustration, consider the following chemical reaction92

nA → C + D

which obeys the equation

dx
= k(a − x)n , (2.3)
dt
where a is the number of molecules A at t = 0, and x is the number of molecules C
(or D) after time t, and k is a reaction constant. At the start of reaction (t = 0),
there are no molecules C (or D) yet formed so that the initial condition is x(0) = 0.
In order to obtain a variational model, we differentiate both sides of Eq. (2.3)
with respect to time, resulting in
d2 x dx
= −kn(a − x)n−1 . (2.4)
dt2 dt
Substituting Eq. (2.3) into Eq. (2.4), we obtain the following 2-order differential
equation
d2 x
= −k 2 n(a − x)2n−1 . (2.5)
dt2
The variational model for the differential equation (2.5) can be obtained with ease
by the semi-inverse method,55 which reads
Z ∞ (  2 )
1 dx 1 2 2n
J(x) = + k (a − x) dt . (2.6)
0 2 dt 2

It is easy to prove that the stationary condition of the above functional, Eq. (2.4),
is equivalent to Eq. (2.3).
We apply the Ritz method to obtain an analytical solution of the discussed
problem. When all molecules A are used up, no further increase in the number of
molecules C can occur, so dx/dt = 0. Putting this into Eq. (2.3), the final number
of molecules C is x∞ = a. By the above analysis, we choose a trial-function in the
form

x = a(1 − e−ηt ) (2.7)

where η is an unknown constant to be further determined.

It is obvious that the funtion (2.7) satisfies the initial condition. Substituting
the function (2.7) into Eq. (2.6), we obtain
Z ∞ 
1 2 2 −2ηt 1 2 2n −2nηt 1 1 2 2n
J(η) = a η e + k a e dt = a2 η + k a . (2.8)
0 2 2 4 4nη
April 13, 2006 10:4 WSPC/140-IJMPB 03379

1146 J.-H. He

Minimizing the functional, Eq. (2.6), with respect to x, is approximately equivalent

to minimizing the above function, Eq. (2.8), with respect to η. We set
∂J 1 1 2 2n
= a2 − k a =0 (2.9)
∂η 4 4nη 2
which leads to the result
1
η = √ kan−1 . (2.10)
n
So we obtain an explicit analytical solution for the discussed problem
x = a[1 − exp(−n−1/2 kan−1 t)] . (2.11)
The unknown constant, η, in the function (2.7) can be identified by various
other methods, for example various residual methods,38 but the result (2.10) is the
most optimal one among all other possible choices.
The solution (2.11) states that x increases approximately exponentially from 0
to a final value x = a, with a time constant given by the result (2.10). Chemists
and technologists always want to know its halfway through the change, i.e. x = a/2
when t = t1/2 . An approximate halfway time obtained from the problem (2.11)
reads
1 1
t1/2 = − ln 0.5 = − −1/2 n−1 ln 0.5 . (2.12)
η n ka
To compare with the exact solution, we consider the case when n = 2. Under such
a case, we can easily obtain the following exact solution:
 
1
xexact = a 1 − . (2.13)
1 − akt
Its halfway time reads
1
t̃1/2 = − . (2.14)
ak
From the half-way time (2.12) we have
1 0.98
t1/2 = − ln 0.5 = − . (2.15)
2−1/2 ka ak
The 2% accuracy is remarkable good. Further improvements in accuracy can be
achieved by suitable choice of trial-functions, and are outside the purpose of the
paper.
The above technology can be successfully applied to many first order nonlinear
differential equations arising in physics. Now we consider the Riccati equation117
in the form
y 0 + p(x)y + q(x)y 2 + r(x) = 0 . (2.16)
Differentiating Eq. (2.16) with respect to x, and eliminating y 0 in the resulted
equation, Eq. (2.16) becomes
y 00 + (p0 + p2 + 2qr)y + (q 0 + pq + 2pq)y 2 + 2q 2 y 3 + r0 + pr = 0 . (2.17)
April 13, 2006 10:4 WSPC/140-IJMPB 03379

Some Asymptotic Methods for Strongly Nonlinear Equations 1147

Its variational formulation can be expressed as

Z 
1 1 1
J(y) = − y 02 + (p0 + p2 + 2qr)y 2 + (q 0 + pq + 2pq)y 3
2 2 3

1
+ q 2 y 4 + (r0 + pr)y dx . (2.18)
2
Variational approach can also be applied to simplify nonlinear equation, consider
the Lambert equation:
k2 y 02
y 00 + y = (1 − n) . (2.19)
n y
By the semi-inverse method,55 its variational formulation reads
Z  
1 2 2n−2 02 1 2 2n
J(y) = − n y y + k y dx . (2.20)
2 2
Hinted by the above energy form, Eq. (2.20), we introduce a transformation:

z = yn . (2.21)

Functional (2.20) becomes

Z  
1 1
J(z) = − z 02 + k 2 z 2 dx . (2.22)
2 2
Its Euler–Lagrange equation is

z 00 + k 2 z = 0 , (2.23)

which is a linear equation. The solution of Eq. (2.19), therefore, has the form:

y = (C cos kx + D sin kx)1/n . (2.24)

2.2. Soliton solution

There exist various approaches to the search for soliton solution for nonlinear wave
equations.19 – 21,39 Here, we suggest a novel and effective method by means of vari-
ational technology. As an illustrative example, we consider the KdV equation
∂u ∂u ∂ 3 u
− 6u + = 0. (2.25)
∂t ∂x ∂x3
We seek its traveling wave solutions in the following frame

u(x, t) = u(ξ) , ξ = x − ct , (2.26)

Substituting the solution (2.26) into Eq. (2.25) yields

−cu0 − 6uu0 + u000 = 0 , (2.27)

where prime denotes the differential with respect to ξ.

April 13, 2006 10:4 WSPC/140-IJMPB 03379

1148 J.-H. He

Integrating Eq. (2.27) yields the result

−cu − 3u2 + u00 = 0 . (2.28)

By the semi-inverse method,55 the following variational formulation is established

Z ∞  2 !
1 2 3 1 du
J= cu + u + dξ . (2.29)
0 2 2 dξ

By Ritz method, we search for a solitary wave solution in the form

u = p sech2 (qξ) , (2.30)

where p and q are constants to be further determined.

Substituting Eq. (2.30) into Eq. (2.29) results in
Z ∞ 
1 2 4 3 6 1 2 2 4 2
J= cp sech (qξ) + p sech (qξ) + 4p q sech (qξ) tanh (qξ) dξ
0 2 2

cp2 ∞ p3 ∞
Z Z Z ∞
= sech4 (z)dz + sech6 (z)dz + 2p2 q {sech4 (z) tanh4 (z)}dz
2q 0 q 0 0

cp2 8p3 4p2 q

= + + . (2.31)
3q 15q 15
Making J stationary with respect to p and q results in

∂J 2cp 24p2 8pq

= + + = 0, (2.32)
∂p 3q 15q 15

∂J cp2 8p3 4p2

=− 2 − + = 0, (2.33)
∂q 3q 15q 2 15
or simplifying

5c + 12p + 4q 2 = 0 , (2.34)

−5c − 8p + 4q 2 = 0 . (2.35)

From Eqs. (2.34) and (2.35), we can easily obtain the following relations:
r
1 c
p = − c, q = . (2.36)
2 4
So the solitary wave solution can be approximated as
r
c c
u = − sech2 (x − ct − ξ0 ) , (2.37)
2 4
which is the exact solitary wave solution of KdV equation (2.25).
April 13, 2006 10:4 WSPC/140-IJMPB 03379

Some Asymptotic Methods for Strongly Nonlinear Equations 1149

2.3. Bifurcation
Bifurcation arises in various nonlinear problems.45,47,48,79,122 – 125 Variational
method can also be applied to the search for dual solutions of a nonlinear equation,
and to approximate identification of the bifurcation point. Consider the Bratu’s
equation in the form:

u00 + λeu = 0 , u(0) = u(1) = 0 , (2.38)

where λ is a positive constant called as the eigenvalue. Equation (2.38) comes orig-
inally from a simplification of the solid fuel ignition model in thermal combustion
theory,42 and it has an exact solution
  
0.5(x − 0.5)θ
u(x) = −2 log cosh ,
cosh(θ/4)
where θ solves
√
θ= 2λ cosh(θ/4) .

Due to its importance, the Bratu problem has been studied extensively. First, it
arises in a wide variety of physical applications, ranging from chemical reaction the-
ory, radiative heat transfer and nanotechnology114 to the expansion of universe.42
Second, because of its simplicity, the equation is widely used as a benchmarking tool
for numerical methods. Khuri42 applied Adomian decomposition method to study
the problem, the procedure is tedious though the expression is a lovely closed-form,
furthermore only one solution is found when 0 < λ < λc , and error arises when
λ → λc .
The present problem can be easily solved with high accuracy by the variational
method, furthermore, two branches of the solutions can be simultaneously deter-
mined, and the bifurcation point can be identified with ease. It is easy to establish
a variational formulation for Eq. (2.38), which reads
Z 1 
1 02 u
J(u) = u − λe dx . (2.39)
0 2
In view of the Ritz method, we choose the following trial-function:
u = Ax(1 − x) , (2.40)

where A is an unknown constant to be further determined. Substituting the function

(2.40) into the formula (2.39) results in:
Z 1 
1 2 2 Ax(1−x)
J(A) = A (1 − 2x) − λe dx . (2.41)
0 2
Minimizing the formula (2.41) to determine the constant, this yields
Z 1
∂J
= {A(1 − 2x)2 − λx(1 − x)eAx(1−x) }dx = 0 . (2.42)
∂A 0
April 13, 2006 10:4 WSPC/140-IJMPB 03379

1150 J.-H. He

Equation (2.42) can be readily solved by some mathematical software, for example,
MATHEMATICA, Matlab. There are two solutions to Eq. (2.42) for values of 0 <
λ < λc , and no solutions for λ > λc . There is only one solution for a critical value
of λ = λc , which solves
Z 1
{(1 − 2x)2 − λc x2 (1 − x)2 eAc x(1−x) }dx = 0 . (2.43)
0

Dividing Eq. (2.43) by Eq. (2.42) produces

Z 1
{x(1 − x)eAc x(1−x) }dx
Ac = Z 01 . (2.44)
{x2 (1 − x)2 eAc x(1−x) }dx
0

Solving Eq. (2.44), we have Ac = 4.72772. Substituting the result into Eq. (2.43)
results in λc = 3.56908. The exact critical value is λ̃c = 3.51383071. The 1.57%
accuracy is remarkable good in view of the crude trial-function, Eq. (2.40).
In order to compare with Khuri’s results, we consider the case λ = 1. In such
case, Eq. (2.43) becomes
Z 1
{A(1 − 2x)2 − x(1 − x)eAx(1−x) }dx = 0 . (2.45)
0

Solving this equation by Mathematica:

Find Root [Integrate [A∗ (1 − 2∗ x)∧ 2 − x∗ (1 − x)Exp[A∗ x∗ (1 − x)], {x, 0, 1}]
= 0, {A, 1}] ,
we have A1 = 0.559441, and A2 = 16.0395.
In the case λ = 1, the Bratu equation has two solutions u1 and u2 with
u01 (0) = 0.549 and u02 (0) = 10.909. Our solutions predict that u01 (0) = 0.559441,
and u02 (0) = 16.0395, with accuracies of 1.9% and 46.94%, respectively. Accu-
racy can be improved if we choose a more suitable trial-function, for example,
u = Ax(1 − x)(1 + Bx + Cx2 ).

2.4. Limit cycle

Energy approach (Variational approach) to limit cycle is suggested by He, 58,59 which
can be applied to not only weakly nonlinear equations, but also strongly ones, and
the obtained results are valid for whole solution domain. D’Acunto30 called the
energy approach as He’s variational method.
To illustrate the basic idea of the method, we consider the following nonlinear
oscillation:
ẍ + x + εf (x, ẋ) = 0 , (2.46)
In the present study the parameter ε need not be small.
April 13, 2006 10:4 WSPC/140-IJMPB 03379

Some Asymptotic Methods for Strongly Nonlinear Equations 1151

Generally speaking, limit cycles can be determined approximately in the form

m
X
x = b + a(t) cos ωt + (Cn cos nωt + Dn sin nωt) , (2.47)
n=1

where b, Cn and Dn are constant.

Substituting Eq. (2.47) into Eq. (2.46) results in the following residual
R(t) = ẍ + x + εf (x, ẋ) . (2.48)
In general, the residual might not be vanishingly small at all points, the error
depends upon the infinite “work”, dw, done by the “force” R in the infinite distance
dx:
dw = Rdx . (2.49)
We hope the totally work done in a period is zero, that requires
Z A1
Rdx = 0 , (2.50)
A0

where A0 and A1 are minimum and maximum amplitudes, respectively.

Equation (2.50) can be equivalently written in the form
Z T
Rẋdt = 0 , (2.51)
0

where T is the period, this technique is similar to the method of weighted residuals.
We assume that the amplitude weakly varies with time, so we write the ampli-
tude in the form
a(t) = Aeαt ≈ A , |α|  1 . (2.52)
Accordingly we approximately have the following expressions
p
ẋ ≈ αx − ω A2 − x2 , (2.53)
p
ẍ ≈ (α2 − ω 2 )x − 2αω A2 − x 2 . (2.54)
Now we consider the van der Pol equation:
ẍ + x − ε(1 − x2 )ẋ = 0 , (2.55)
We will not re-illustrate the solution procedure, but we identify α in Ref. 58 instead
with
ε2 (A2 − 4)ωπ
α= , (2.56)
4(2πω − εA2 )
and the frequency equation (33) in Ref. 58 should be replaced by
ε2 (A2 − 4)2 ω 2 π 2 ε2 (A2 − 4)ωπ
1 − ω2 + − = 0. (2.57)
16(2πω − εA2 )2 4(2πω − εA2 )
April 13, 2006 10:4 WSPC/140-IJMPB 03379

1152 J.-H. He

Now consider the limit ε  1, by assumption ω = 1 + aε2 + O(ε3 ), Eq. (2.57)

becomes
ε2 (A2 − 4)2 (A2 − 4) 2
 
−2a + + ε + O(ε3 ) = 0 . (2.58)
64 8
After identification the constant, a, we have
 2 2
A2 − 4

2 (A − 4)
ω =1+ε + + O(ε3 ) . (2.59)
128 16
In the case ε → ∞, the frequency can be approximated as
 
b 1
ω = +O , (2.60)
ε ε2
From Eq. (2.57), we have
b2 (A2 − 4)2 π 2 εb(A2 − 4)π
   
b 1
1− + − + O = 0. (2.61)
ε2 16(2πb/ε − εA2 )2 4(2πb/ε − εA2 ε2
If we keep the order of 1/ε, then Eq. (2.61) can be simplified as
b(A2 − 4)π
1+ = 0, (2.62)
4A2
from which the constant, b, can be readily determined. As a result, we have:
4A2
 
1
ω=− 2 +O . (2.63)
(A − 4)πε ε2
If we choose A = 2 which is identified by perturbation method,104 then the last
two terms in Eq. (2.57) vanish completely, leading to the invalidity of the method
as pointed out by Rajendran et al.104 The perturbation method provides us with
a choice of the approximate value of A, and many alternative approaches to the
identification of A exist.
Assuming that period solution has the form: x = A cos ωt, we obtain the fol-
lowing residual:
R = (1 − ω 2 )A cos ωt + Aεω(1 − A2 cos2 ωt) sin ωt . (2.64)
By substituting Eq. (2.59) or Eq. (2.63) into Eq. (2.64) respectively, we have
R = Aε(1 − A2 cos2 ωt) sin ωt + O(ε2 ) , ε → 0, (2.65)
and
R/ε = A(1 − A2 cos2 ωt) sin ωt + O(1/ε) , 1/ε → 0 . (2.66)
The value of A can be determined in the parlance of vanishing small residual by
the method of weighted residuals.
Applying the Galerkin method
Z T /4 Z T /4
R sin ωtdt = A(1 − A2 cos2 ωt) sin2 ωtdt = 0 , T = 2π/ω , (2.67)
0 0
April 13, 2006 10:4 WSPC/140-IJMPB 03379

Some Asymptotic Methods for Strongly Nonlinear Equations 1153

leading to the result: A = 2, which agrees with the standard result of the pertur-
bation method.
The simplest method among the method of weighted residuals is the method of
collocation. Collocating at ωt = π/3, ωt = π/4, and ωt = π/6 gives, respectively,
the approximate values of A: A = 2, A = 1.414, and A = 1.155. The average value
of A reads A = 1.522.
We can also apply the least squares method to find the approximate value of A
by the formulation:
Z T /4
∂R
R dt = 0 , (2.68)
0 ∂A
i.e.
Z T /4
A(1 − A2 cos2 ωt)(1 − 3A2 cos2 ωt) sin2 ωtdt = 0 , (2.69)
0
p
which leads to the result A = 8/3 = 1.633.
Now we choose A = 1.633, then Eq. (2.59) becomes
5 2
ε + O(ε3 ) = 1 − 0.0694ε2 + O(ε3 ) .
ω =1− (2.70)
72
We write down the perturbation solution for reference, which reads
1 2
ω =1− ε + O(ε3 ) = 1 − 0.0625ε2 + O(ε3 ) . (2.71)
16
Equation (2.63) becomes
     
8 1 8 1 2.566 1
ω= +O = + O = + O . (2.72)
πε ε2 πε ε2 ε ε2
The exact frequency under the limit ε → ∞ is
 
3.8929 1
ω= +O . (2.73)
ε ε2
So the obtained solution is valid for both ε → 0 and ε → ∞.

2.5. Variational iteration method

The variational iteration method70,74 – 77 has been shown to solve effectively, easily,
and accurately a large class of nonlinear problems with approximations converg-
ing rapidly to accurate solutions. Most authors found that the shortcomings aris-
ing in Adomian method can be completely eliminated by the variational iteration
method.6,7,17,96,97,103,111,112
To illustrate the basic idea of the variational iteration method, we consider the
following general nonlinear system:
Lu + N u = g(x) , (2.74)
where L is a linear operator, and N is a nonlinear operator.
April 13, 2006 10:4 WSPC/140-IJMPB 03379

1154 J.-H. He

According to the variational iteration method, we can construct the following

iteration formulation:
Z x
un+1 (x) = un (x) + λ{Lun (s) + N ũn (s) − g(s)}ds (2.75)
0

where λ is called a general Lagrange multiplier, which can be identified optimally

via the variational theory, ũn is considered as a restricted variation, i.e. δũn = 0.
We consider Duffing equation as an example to illustrate its basic computational
procedure.
d2 u
+ u + εu3 = 0 , (2.76)
dt2
with initial conditions u(0) = A, u0 (0) = 0.
Supposing that the angular frequency of the system (2.76) is ω, we have the
following linearized equation:
u00 + ω 2 u = 0 . (2.77)
So we can rewrite Eq. (2.76) in the form
u00 + ω 2 u + g(u) = 0 , (2.78)
where g(u) = (1 − ω 2 )u + εu3 .
Applying the variational iteration method, we can construct the following
functional:
Z t
un+1 (t) = un (t) + λ{u00n (τ ) + ω 2 un (τ ) + g̃(un )}dτ , (2.79)
0

where g̃ is considered as a restricted variation, i.e. δg̃ = 0.

Calculating variation with respect to un , and noting that δg̃(un ) = 0, we have
the following stationary conditions:
 00 2
 λ (τ ) + ω λ(τ ) = 0 ,

λ(τ )|τ =t = 0 , (2.80)

 0
1 − λ (τ )|τ =t = 0 .
The multiplier, therefore, can be identified as
1
λ= sin ω(τ − t) . (2.81)
ω
Substituting the identified multiplier into Eq. (2.79) results in the following iteration
formula:
1 t
Z
un+1 (t) = un (t) + sin ω(τ − t){u00n (τ ) + un (τ ) + εu3n (τ )}dτ . (2.82)
ω 0
Assuming its initial approximate solution has the form:
u0 (t) = A cos ωt , (2.83)
April 13, 2006 10:4 WSPC/140-IJMPB 03379

Some Asymptotic Methods for Strongly Nonlinear Equations 1155

and substituting Eq. (2.83) into Eq. (2.76) leads to the following residual
 
3 1
R0 (t) = 1 − ω 2 + εA2 A cos ωt + εA3 cos 3ωt . (2.84)
4 4
By the formulation (2.82), we have
t
1
Z
u1 (t) = A cos ωt + R0 (τ ) sin ω(τ − t)dτ . (2.85)
ω 0

In order to ensure that no secular terms appear in u1 , resonance must be avoided.

To do so, the coefficient of cos ωt in Eq. (2.84) requires to be zero, i.e.
r
3
ω = 1 + εA2 . (2.86)
4
So, from Eq. (2.85), we have the following first-order approximate solution:
εA3 t
Z
u1 (t) = A cos ωt + cos 3ωt sin ω(τ − t)dτ
4ω 0
εA3
= A cos ωt + (cos 3ωt − cos ωt) . (2.87)
32ω 2
It is obvious that for small ε, i.e. 0 < ε  1, it follows that
3
ω = 1 + εA2 . (2.88)
8
Consequently, in this limit, the present method gives exactly the same results
as the standard Lindstedt–Poincare method.98,100 To illustrate the remarkable ac-
curacy of the obtained result, we compare the approximate period
2π
T =p (2.89)
1 + 3εA2 /4
with the exact one
π/2
4 dx
Z
Tex = √ p , (2.90)
1 + εA2 0 1 − k sin2 x
where k = 0.5εA2 /(1 + εA2 ).
What is rather surprising about the remarkable range of validity of Eq. (2.89)
is that the actual asymptotic period as ε → ∞ is also of high accuracy.
2 3/4 π/2
p
Tex dx
Z
lim = = 0.9294 .
ε→∞ T π
p
0 1 − 0.5 sin2 x
Therefore, for any value of ε > 0, it can be easily proved that the maximal
relative error of the period Eq. (2.89) is less than 7.6%, i.e. |T − Tex |/Tex < 7.6%.
Due to the very high accuracy of the first-order approximate solution, we always
stop before the second iteration step. That does not mean that we cannot obtain
higher order approximations.
April 13, 2006 10:4 WSPC/140-IJMPB 03379

1156 J.-H. He

3. Parameter-Expanding Methods
We consider the following nonlinear oscillator
mu00 + ω02 u + εf (u) = 0 . (3.1)
Various perturbation methods have been applied frequently to analyze Eq. (3.1).
The perturbation methods are, however, limited to the case of small ε and mω02 > 0,
that is, the associated linear oscillator must be statically stable in order that linear
and nonlinear response be qualitatively similar.
Classical methods which are applied to weakly nonlinear systems include the
Lindstedt–Poincare, Krylov–Bogoliubov–Mitropolski averaging, and multiple time
scale methods. These are described by many textbooks by Bellman,15 Nayfeh,98
and Nayfeh and Mook,100 among others. These methods are characterized by ex-
pansions of the dependent variables in power series in a small parameter, resulting
in a collection of linear differential equations which can be solved successively. The
solution to the associated linear problem thus provides a starting point for the gen-
eration of the “small perturbation” due to the nonlinearity, with various devices
employed to annul secularities.
Classical techniques which have been used for both weakly and strongly nonlin-
ear systems include the equivalent linearization method,105 the harmonic balance
method.79 A feature of these methods is that the form of solution is specified in
advance. The methods work well provided that the filter hypothesis is satisfied,
that is, higher harmonics output by the nonlinearities are substantially attenuated
by the linear part of the system.

3.1. Modifications of Lindstedt–Poincare method

A number of variants of the classical perturbation methods have been proposed to
analyze conservative oscillators similar to Eq. (3.1). The Shohat expansion 108 for the
van der Pol equation is of remarkable accuracy not only for small parameter ε, but
also for large ε. But recently Mickens95 shows that it does not have a general validity
as a “perturbation” scheme that can be extended to large values of ε for Duffing
equation, as was originally thought to be. Burton23,24 defined a new parameter
α = α(ε) in such a way that asymptotic solutions in power series in α converge
more quickly than do the standard perturbation expansions in power series in ε.
εA2
α= . (3.2)
4 + 3εA2
It is obvious that α < 1/3 for all εA2 , so the expansion in power series in α is
quickly convergent regardless of the magnitude of εA2 .
The Lindstedt–Poincare method98,100 gives uniformly valid asymptotic expan-
sions for the periodic solutions of weakly nonlinear oscillations, in general, the
technique does not work in case of strongly nonlinear terms.
For the sake of comparison, the standard Lindstedt–Poincare method is briefly
recapitulated first. The basic idea of the standard Lindstedt–Poincare method is to
April 13, 2006 10:4 WSPC/140-IJMPB 03379

Some Asymptotic Methods for Strongly Nonlinear Equations 1157

introduce a new variable

τ = ω(ε)t , (3.3)
where ω is the frequency of the system. Equation (3.1) then becomes
ω 2 mu00 + ω02 u + εf (u) = 0 , (3.4)
where prime denotes differentiation with respect to τ . The frequency is also ex-
panded in powers of ε, that is
ω = ω0 + εω1 + ε2 ω2 + · · · . (3.5)
Various modifications of Lindstedt–Poincare method appeared in open literature.
It is of interest to note that a much more accurate relation for frequency ω can be
found by expanding ω 2 , rather than ω, in a power series in ε.
ω 2 = 1 + εω1 + ε2 ω2 + · · · . (3.6)
Cheung et al. introduced a new parameter
εω1
α= 2 . (3.7)
ω0 + εω1
and expand the solution and ω 2 in power series in α.
All those modifications are limited to conservation system, instead of linear
transformation (3.3), a nonlinear time transformation can be successfully annul
such shortcoming.
We can re-write the transformation of standard Lindstedt–Poincare method in
the form67
τ = t + f (ε, t) , f (0, t) = 0 , (3.8)
where f (ε, t) is an unknown function, not a functional. We can also see from the
following examples that the identification of the unknown function f is much easier
than that of unknown functional F in Dai’s method.31,32
From the relation (3.8), we have
2 2
∂2u ∂2u

∂f ∂ u
2
= 1 + 2
=G 2 . (3.9)
∂t ∂t ∂τ ∂τ
where
 2
∂f
G(ε, t) = 1+ . (3.10)
∂t
Applying the Taylor series, we have
G(ε, t) = G0 + εG1 + · · · , (3.11)
where Gi (i = 1, 2, 3, . . .) can be identified in view of no secular terms. After iden-
tification of Gi (i = 1, 2, 3, . . .), from the Eq. (3.10), the unknown function f can
be solved. The transformation Eq. (3.8) can be further simplified as
τ = f (t, ε) , f (t, 0) = t . (3.12)
April 13, 2006 10:4 WSPC/140-IJMPB 03379

1158 J.-H. He

So we have
2
∂2u ∂2u

∂f
= . (3.13)
∂t2 ∂t ∂τ 2
We expand (∂f /∂t)2 in a series of ε:
 2
∂f
= 1 + εf1 + ε2 f2 + · · · . (3.14)
∂t
Now we consider the van der Pol equation.
u00 + u − ε(1 − u2 )u0 = 0 . (3.15)
By the transformation τ = f (t, ε), Eq. (3.15) becomes
 2 2
∂f ∂ u ∂f ∂u
2
+ u = ε(1 − u2 ) . (3.16)
∂t ∂τ ∂t ∂τ
Introducing a new variable ω, which is defined as
 2
∂f 1
= 2. (3.17)
∂t ω
So Eq. (3.16) reduces to
∂2u ∂u
2
+ ω 2 u = εω(1 − u2 ) . (3.18)
∂τ ∂τ
By simple operation, we have
 2
∂f 1 1
= 2 = . (3.19)
∂t ω 1 + 81 ε2
Solving Eq. (3.19), subject to the condition f (t, 0) = t, gives
t
τ =f = q . (3.20)
1 + 81 ε2
Its period can be expressed as
r
1
T = 2π 1 + ε2 . (3.21)
8
It is obvious that when ε  1, the results are same as those obtained by standard
perturbation methods.98
However, for large values of ε (ε  1), we have the following approximation 98
for exact period
Z 1/√3  
dv
Tex ≈ 2ε √ − 3vdv = 1.614ε , (ε  1) . (3.22)
2/ 3 v
From Eq. (3.20), it follows
π
T ≈ √ ε = 2.22ε , (ε  1) . (3.23)
2
April 13, 2006 10:4 WSPC/140-IJMPB 03379

Some Asymptotic Methods for Strongly Nonlinear Equations 1159

It is obvious that for large ε, the approximate period has the same feature as
the exact one, and we have
Tex 1.614
lim = = 0.727 .
ε→∞ T 2.22
So the maximal relative error for all 0 < ε < ∞ is less than 37.5%. We can
readily obtain higher-order approximations by MATHEMATICA.
Now we return to Eq. (3.1). Most methods are limited to odd nonlinearities and
require that ω02 > 0 and m > 0. We must devise a method to solve Eq. (3.1) for
the case when (1) m > 0, ω02 ≤ 0; (2) ω02 ≥ 0, m ≤ 0; (3) even nonlinearities ex-
ist. Parameter-expanding methods including modified Lindsted–Poincare method67
and bookkeeping parameter method64 can successfully deal with such special cases
where classical methods fail. The methods need not have a time transformation
like Lindstedt–Poincare method, the basic character of the method is to expand
the solution and some parameters in the equation. If we search for periodic solution
of Eq. (3.1), then we can assume that
u = u0 + εu1 + ε2 u2 + · · · , (3.24)

ω02 = ω 2 + pω1 + p2 ω2 + · · · , (3.25)

m = 1 + pm1 + p2 m2 + · · · . (3.26)
Hereby ε can be a small parameter in the equation, or an artificial parameter or a
bookkeeping parameter.
As an illustration, consider the motion of a ball-bearing oscillating in a glass
tube that is bent into a curve such that the restoring force depends upon the cube
of the displacement u. The governing equation, ignoring frictional losses, is8
d2 u
+ εu3 = 0 , u(0) = A , u0 (0) = 0 . (3.27)
dt2
The standard Lindstedt–Poincare method does not work for this example. Now
we re-write Eq. (3.27) in the form
u00 + 0 · u + εu3 = 0 . (3.28)
Supposing the solution and the constant zero in Eq. (3.28) can be expressed as
u = u0 + εu1 + ε2 u2 + · · · , (3.29)

0 = ω 2 + εc1 + ε2 c2 + · · · . (3.30)
Substituting Eqs. (3.29) and (3.30) into Eq. (3.28), and processing as the standard
perturbation method, we have
u000 + ω 2 u0 = 0 , u0 (0) = A , u00 (0) = 0 , (3.31)

u001 + ω 2 u1 + c1 u0 + u30 = 0 , u1 (0) = 0 , u01 (0) = 0 . (3.32)

April 13, 2006 10:4 WSPC/140-IJMPB 03379

1160 J.-H. He

Solving Eq. (3.31), we have u0 = A cos ωt. Substituting u0 into Eq. (3.32) results
into
 
3 1
u001 + ω2 u1 + A c1 + A2 cos ωt + A3 cos 3ωt = 0 . (3.33)
4 4
Eliminating the secular term needs
3
c1 = − A2 . (3.34)
4
If only the first-order approximate solution is searched for, then from Eq. (3.30),
we have
3
0 = ω 2 − εA2 , (3.35)
4
which leads to the result
√
3 1/2
ω= ε A. (3.36)
2
Its period, therefore, can be written as
4π
T = √ ε−1/2 A−1 = 7.25ε−1/2 A−1 . (3.37)
3
Its exact period can be readily obtained, which reads
Tex = 7.4164ε−1/2A−1 . (3.38)
It is obvious that the maximal relative error is less than 2.2%, and the obtained
approximate period is valid for all ε > 0.
Now consider another example:99
u
u00 + = 0. (3.39)
1 + εu2
with initial conditions u(0) = A and u0 (0) = 0.
We rewrite Eq. (3.39) in the form
u00 + 1 · u + εu00 u2 = 0 . (3.40)
Proceeding with the same manipulation as the previous example, we obtain the
following angular frequency:
1
ω=q . (3.41)
1 + 43 εA2
It is obvious that for small ε, it follows that
3
ω = 1 − εA2 . (3.42)
8
Consequently, in this limit, the present method gives exactly the same results as
the standard perturbations.99 To illustrate the remarkable accuracy of the obtained
results, we compare the approximate period
r
3
T = 2π 1 + εA2 (3.43)
4
April 13, 2006 10:4 WSPC/140-IJMPB 03379

Some Asymptotic Methods for Strongly Nonlinear Equations 1161

with the exact one

Z A Z A
√ du √ du
T =4 ε p =4 ε p .
ln(1 + εA 2 ) − ln(1 + εu2 ) ln[(1 + εA 2 )/(1 + εu2 )]
0 0
(3.44)
In the case εA2 → ∞, Eq. (3.44) reduces to
Z A √ Z 1
√ du du
lim T = 4 ε p = 2 2εA p
εA2 →∞ 0 2(ln A − ln u) 0 ln(1/s)
√ Z ∞ √
= 4 2εA exp(−x2 )dx = 2 2πεA . (3.45)
0
It is obvious that the approximate period (3.43) has the same feature as the exact
one for ε  1. And in the case ε → ∞, we have
√
Tex 2 2πεA
lim = √ = 0.9213 .
A→∞ T 3επA
Therefore, for any values of ε, it can be easily proven that the maximal relative
error is less than 8.54% on the whole solution domain (0 < ε < ∞).

3.2. Bookkeeping parameter

In most cases the parameter, ε, might not be a small parameter, or there is no
parameter in the equation, under such conditions, we can use a bookkeeping pa-
rameter. Consider the mathematical pendulum:
u00 + sin u = 0 . (3.46)
We re-write Eq. (3.46) in the following form
u00 + 0 · u + sin u = 0 . (3.47)
Supposing the solution and the constant, zero, can be expressed in the forms:
u = u0 + pu1 + p2 u2 + · · · , (3.48)

0 = ω 2 + pc1 + p2 c2 + · · · . (3.49)
Here, p is a bookkeeping parameter, ci can be identified in view of no secular
terms in ui (i = 1, 2, 3, . . .). For example, the constant c1 in Eq. (3.49) can also be
identified in view of no secular terms in u1 , that is
Z T
A cos ωt{c1 A cos ωt + sin(A cos ωt)}dt = 0 , (3.50)
0
where T = 2π/ω.
From Eq. (3.50), c1 can be determined explicitly by Bessel function. The angular
frequency can be determined as
r
2J1 (A)
ω= , (3.51)
A
April 13, 2006 10:4 WSPC/140-IJMPB 03379

1162 J.-H. He

where J1 is the first-order Bessel function of the first kind.

The parameter-expanding methods can also be applied to non-periodic solu-
tions. Consider another simple example.
u0 + u 2 = 1 , u(0) = 0 , (3.52)
which has an exact solution
1 − e−2t
uex = . (3.53)
1 + e−2t
No small parameter exits in the equation, so in order to carry out a straightforward
expansion, Liu88 introduced an artificial parameter ε embedding in Eq. (3.52) so
that he obtained the following equation
u0 + (u + 1)(εu − 1) = 0 , ε = 1. (3.54)
Supposing that the solution can be expanded in powers of ε, and processing in
a traditional fashion of perturbation technique, Liu obtained the following result88
u = (1 − e−t ) + εe−t (e−t + t − 1) . (3.55)
Let ε = 1 in Eq. (3.55). We obtain the approximate solution of the original
Eq. (3.52):
u(t) = (1 − e−t ) + e−t (e−t + t − 1) . (3.56)
The obtained approximation (3.56) is of relative high accuracy.
In order to illustrate the basic idea of the proposed perturbation method, we
rewrite Eq. (3.52) in the form
u0 + 0 · u + 1 · u 2 = 1 , u(0) = 0 . (3.57)
Supposing that the solution and the coefficients (0 and 1) can be expressed in the
forms:64
u = u0 + pu1 + p2 u2 + · · · , (3.58)

0 = a + pa1 + p2 a2 + · · · , (3.59)

1 = pb1 + p2 b2 + · · · . (3.60)
Substituting Eqs. (3.58)–(3.60) into Eq. (3.57), collecting the same power of p, and
equating each coefficient of p to zero, we obtain
u00 + au0 = 1 , u0 (0) = 0 , (3.61)

u01 + au1 + a1 u0 + b1 u20 = 0 , (3.62)

u02 + au2 + a1 u1 + a2 u0 + b2 u20 + 2b1 u0 u1 = 0 . (3.63)

We can easily solve the above equations sequentially for ui (i = 1, 2, 3, . . .), the
P
initial conditions for ui (i = 1, 2, 3, . . .) should satisfy the condition i=1 ui (0) = 0.
April 13, 2006 10:4 WSPC/140-IJMPB 03379

Some Asymptotic Methods for Strongly Nonlinear Equations 1163

Solving Eq. (3.61), we have

1
u0 =(1 − e−at ) . (3.64)
a
Substituting u0 into Eq. (3.62) results in
a1 b1
u01 + au1 + (1 − e−at ) + 2 (1 − e−at )2 = 0 . (3.65)
a a
or
   
a1 b1 a1 2b1 b1 −2at
u01 + au1 + + 2 − + 2 e−at + e = 0. (3.66)
a a a a a2
The requirement of no term te−at in u1 needs
a1 2b1
+ 2 = 0. (3.67)
a a
If we keep terms to O(p2 ) only, from Eqs. (3.59) and (3.60), we have
a + a1 = 0 and b1 = 1 . (3.68)
From the Eqs. (3.67) and (3.68), we obtain
√
a = 2. (3.69)
Solving Eq. (3.66) subject to the initial condition u1 (0) = 0 gives
 
1 1 1
u1 = − − + 3 (1 − e−at ) + 3 (e−2at − e−at ) . (3.70)
a a a
Setting p = 1, we have the first-order approximation:
 
2 1 1
u(t) = u0 (t) + u1 (t) = − (1 − e−at ) + 3 (e−2at − e−at ) , (3.71)
a a3 a
√
where a = 2. It is easy to prove that the obtained result (3.71) is uniformly valid
on the whole solution domain.
Consider the Thomas–Fermi equation41,56,82
u00 (x) = x−1/2 u3/2 , u(0) = 1 , u(∞) = 0 . (3.72)
The energy (in atomic units) for a neutral atom of atomic number Z is
 2/3
6 4
E= Z 7/3 u0 (0) .
7 3π
So it is important to determine a highly accurate value for the initial slope of the
potential u0 (0).
The equation is studied by δ-method.11 The basic idea of the δ-method is to
replace the right hand side of the Thomas–Fermi equation by one which contains
the parameter δ, i.e.
u00 (x) = u1+δ x−δ . (3.73)
April 13, 2006 10:4 WSPC/140-IJMPB 03379

1164 J.-H. He

The solution is assumed to be expanded in a power series in δ

u = u0 + δu1 + δ 2 u2 + · · · . (3.74)
This, in turn, produces a set of linear equations for un :
u000 − u0 = 0 ,
u001 − u1 = u0 ln(u0 /x)
1
u002 − u2 = u1 + u1 ln(u0 /x) + u0 ln2 (u0 /x)
2
with associated boundary conditions u0 (0) = 1, u0 (∞) = 0 and un (0) = un (∞) = 0
for n > 1.
To solve un for n > 1, we need some unfamiliar functions, so it might meet some
difficulties in promoting this method.
We rewrite the equation in the forma
u00 (x) − 0 · u = 1 · x−1/2 u3/2 . (3.75)
Supposing that the solution and the coefficients (0 and 1) can be expressed in the
forms
u = u0 + pu1 + p2 u2 + · · · , (3.76)

0 = β 2 + pa1 + p2 a2 + · · · , (3.77)

1 = pb1 + p2 b2 + · · · . (3.78)
By simple operation, we have the following equations
u000 (x) − β 2 u0 = 0 , (3.79)
3/2
u001 (x) − β 2 u1 = b1 x−1/2 u0 + a 1 u0 , (3.80)
with associated boundary conditions u0 (0) = 1, u0 (∞) = 0 and un (0) = un (∞) = 0
for n > 1. The first-order approximate solution is64
u0 = e−1.19078255x . (3.81)
64
The second-order approximate solution is approximately solved as
 
1
u(x) = 1 − x e−1.19078255x . (3.82)
3
In 1981, Nayfeh98 studied the free oscillation of a nonlinear oscillation with
quadratic and cubic nonlinearities, and the governing equation reads
u00 + u + au2 + bu3 = 0 , (3.83)
where a and b are constants.

a We can also assume that u00 (x) + 0 · u = 1 · x−1/2 u3/2 , then the constant β 2 will be proven to

be negative.
April 13, 2006 10:4 WSPC/140-IJMPB 03379

Some Asymptotic Methods for Strongly Nonlinear Equations 1165

In order to carry out a straightforward expansion for small but finite ampli-
tudes for Eq. (3.83), we need to introduce a small parameter because none appear
explicitly in this equation. To this end, we seek an expansion in the form98
u = qu1 + q 2 u2 + q 3 u3 + · · · , (3.84)
where q is a small dimensionless parameter that is a measure of the amplitude of
oscillation. It can be used as a bookkeeping device and set equal to unity if the am-
plitude is taken to be small. As illustrated in Ref. 98, the straightforward expansion
breaks down due to secular terms. And Nayfeh used the method of renormaliza-
tion to render the straightforward expansion uniform. Unfortunately, the obtained
results are valid only for small amplitudes.
Hagedorn43 also used an artificial parameter to deal with a mathematical pen-
dulum. For Eq. (3.83), we can assume that64
u = u0 + pu1 + p2 u2 + · · · , (3.85)

ω02 = ω 2 + pc1 + p2 c2 + · · · , (3.86)

a = pa1 + p2 a2 + · · · , (3.87)

b = p 2 b1 + p 3 b2 + · · · . (3.88)
We only write down the first-order angular frequency, which reads
v q
u
u 1 + 3 bA2 + (1 + 3 bA2 )2 − 5 a2 A2
t 4 4 3
ω= . (3.89)
2
We write down Nayfeh’s result98 for comparison:
 
3 5
ω =1+ b − a2 A2 , (3.90)
8 12
which is obtained under the assumption of small amplitude. Consequently, in this
limit, the present method gives exactly the same results as Nayfeh’s.98

4. Parametrized Perturbation Method

In Refs. 73 and 71, an expanding parameter is introduced by a linear transforma-
tion:
u = εv + b , (4.1)
where ε is the introduced perturbation parameter, b is a constant.
We reconsider Eq. (3.52). Substituting Eq. (4.1) into Eq. (3.52) results in
(
v 0 + εv 2 + 2bv + (b2 − 1)/ε = 0 ,
(4.2a)
v(0) = −b/ε .
April 13, 2006 10:4 WSPC/140-IJMPB 03379

1166 J.-H. He

Setting b = 1 for simplicity, so we have the following equation with an artificial

parameter ε:
(
v 0 + εv 2 + 2v = 0 ,
(4.2b)
v(0) = −1/ε .
Assume that the solution of Eq. (4.2b) can be written in the form
v = v0 + εv1 + ε2 v2 + · · · . (4.3)
Unlike the traditional perturbation methods, we keep
X
v0 (0) = v(0) , and vi (0) = 0 . (4.4)
i=1

We write the first-order approximate solution of Eq. (4.2b)

1 1
v = v0 + εv1 = − e−2t + (e−4t − e−2t ) . (4.5)
ε 2ε
In view of the transformation (4.1), we obtain the approximate solution of the
original Eq. (3.52):
1
u = εv + 1 = 1 − e−2t + (e−4t − e−2t ) . (4.6)
2
It is interesting to note that in Eq. (4.6) the artificial parameter ε is eliminated
completely. It is obvious that even its first-order approximate solution (4.6) is of
remarkable accuracy.
We now consider the following simple example to take the notation
1
u0 + u − u 2 = 0 , u(0) = (4.7)
2
where there exists no small parameter. In order to apply the perturbation tech-
niques, we introduce a parameter ε by the transformation
u = εv (4.8)
so the original Eq. (4.7) becomes
v 0 + v − εv 2 = 0 , v(0) = A (4.9)
where εA = 1/2.
Supposing that the solution of Eq. (4.9) can be expressed in the form
v = v0 + εv1 + ε2 v2 + · · · (4.10)
and processing in a traditional fashion of perturbation technique, we obtain the
following equations
v00 + v0 = 0 , v0 (0) = A (4.11)

v10 + v1 − v02 = 0 , v0 (0) = 0 . (4.12)

April 13, 2006 10:4 WSPC/140-IJMPB 03379

Some Asymptotic Methods for Strongly Nonlinear Equations 1167

Solving Eqs. (4.11) and (4.12), we obtain

v0 = Ae−t (4.13)

v1 = A2 (e−t − e−2t ) . (4.14)

Using the fact A = 1/(2ε), we obtain the first-order approximate solution of
Eq. (4.9)
v = v0 + εv1 + O(ε2 ) = Ae−t + εA2 (e−t − e−2t ) + O(ε2 )
1 −t 1
= e + (e−t − e−2t ) + O(ε2 ) . (4.15)
2ε 4ε
In view of the transformation (4.8), we have the approximate solution of the
original Eq. (4.7)
3 −t 1 −2t
u = εv =
e − e (4.16)
4 4
which is independent upon the artificial parameter ε, as it ought to be. The exact
solution of Eq. (4.7) reads
e−t
uex = = e−t (1 − e−t + e−2t + · · ·) . (4.17)
1 + e−t
It is obvious that the obtained approximate solution (3.16) is of high accuracy.

Example 1. Now we consider the Duffing equation with 5th order nonlinearity,
which reads

u00 + u + αu5 = 0 u(0) = A , u0 (0) = 0 (4.18)

where α needs not be small in the present study, i.e. 0 ≤ α < ∞.
We let
u = εv (4.19)
in Eq. (4.18) and obtain
v 00 + 1 · v + αε4 v 5 = 0 , v(0) = A/ε , v 0 (0) = 0 . (4.20)
Applying the parameter-expanding method (modified Lindstedt–Poincare
method67 ) as illustrated in Sec. 3.1, we suppose that the solution of Eq. (4.20)
and the constant 1, can be expressed in the formsb
v = v 0 + ε 4 v1 + ε 8 v2 + · · · (4.21)

1 = ω 2 + ε 4 ω1 + ε 8 ω2 + · · · . (4.22)

b Note: If we assume that

v = v0 + εv1 + ε2 v2 + ε3 v3 + εv4 · · · and 1 = ω 2 + εω1 + ε2 ω2 + ε3 ω3 + ε4 ω4 · · ·
then it is easy to find that v1 = v2 = v3 = 0 and ω1 = ω2 = ω3 = 0, so that the secular terms will
not occur.
April 13, 2006 10:4 WSPC/140-IJMPB 03379

1168 J.-H. He

Substituting Eqs. (4.21) and (4.22) into Eq. (4.20) and equating coefficients of
like powers of ε yields the following equations
v000 + ω 2 v0 = 0 , v0 (0) = A/ε , v00 (0) = 0 (4.23)

v100 + ω 2 v1 + ω1 v0 + αv05 = 0 , v1 (0) = 0 , v10 (0) = 0 . (4.24)

Solving Eq. (4.23) results in
A
v0 = cos ωt . (4.25)
ε
Equation (4.24), therefore, can be re-written down as
5αA4 5αA5 αA5
 
A
v100 + ω 2 v1 + 4
+ ω 1 cos ωt + 5
cos 3ωt + cos 5ωt = 0 . (4.26)
8ε ε 16ε 16ε5
Avoiding the presence of a secular term needs
5αA4
ω1 = − (4.27)
8ε4
Solving Eq. (4.26), we obtain
αA5 αA5
u1 = − 5 2
(cos ωt − cos 3ωt) − (cos ωt − cos 5ωt) . (4.28)
128ε ω 384ε5ω 2
If, for example, its first-order approximation is sufficient, then we have
αA5
u = εv = ε(v0 + ε4 v1 ) = A cos ωt − (cos ωt − cos 3ωt)
128ω 2
αA5
− (cos ωt − cos 5ωt) (4.29)
384ω 2
where the angular frequency can be written in the form
r
p 5
ω = 1 − ε ω1 = 1 + αA4 ,
4 (4.30)
8
which is valid for all α > 0.

Example 2. Consider the motion of a particle on a rotating parabola, which is

governed by the equation (Exercise 4.8 in Ref. 99)

2
d2 u d2 u

(1 + 4q 2 u2 ) + α2 u + 4q 2 u = 0, t∈Ω (4.31)
dt2 dt2
with initial conditions u(0) = A, and u0 (0) = 0.
Herein α and q are known constants, and need not to be small. We let u = εv
in Eq. (4.31) and obtain
 2
d2 v dv
(1 + 4ε2 q 2 v 2 ) 2 + α2 v + 4q 2 ε2 v = 0 , v(0) = A/ε , v 0 (0) = 0 . (4.32)
dt dt
April 13, 2006 10:4 WSPC/140-IJMPB 03379

Some Asymptotic Methods for Strongly Nonlinear Equations 1169

By parameter-expanding method(the modified Lindsted–Poincare method67 ),

we assuming that α2 and solution of the Eq. (4.32) can be written in the form
α2 = Ω 2 + ε 2 ω1 + ε 4 ω2 + · · · (4.33)

v = v 0 + ε 2 v1 + ε 4 v2 + · · · . (4.34)
Substituting Eqs. (4.33) and (4.34) into Eq. (4.32), and equating the coefficients
of like powers of ε results in the following equations
v000 + Ω2 v0 = 0 , v0 (0) = A/ε , v00 (0) = 0 (4.35)

v100 + Ω2 v1 + ω1 v0 + 4q 2 (v02 v000 + v0 v002 ) = 0 , v1 (0) = 0 , v10 (0) = 0 . (4.36)

Solving Eq. (4.35) yields
A
cos Ωt .
v0 (τ ) = (4.37)
ε
Equation (4.36), therefore, can be re-written down as
Aω1 4q 2 A3 Ω2
v100 + Ω2 v1 + cos Ωt + (− cos3 Ωt + cos Ωt sin2 Ωt) = 0 . (4.38)
ε ε3
Using trigonometric identities, Eq. (4.38) becomes
2q 2 A3 Ω2 2q 2 A3 Ω2
 
00 2 Aω1
v1 + Ω v1 + − cos Ωt − cos 3Ωt = 0 . (4.39)
ε ε3 ε3
Eliminating the secular term demands that
2q 2 A2 Ω2
ω1 = . (4.40)
ε2
Solving Eq. (4.39), we obtain
q 2 A3 Ω2
v1 = (cos Ωt − cos 3Ωt) . (4.41)
4ε3
If, for example, its first-order approximation is sufficient, then we have the first-
order approximate solution of Eq. (4.32)
q 2 A3 Ω2
u = ε(v0 + ε2 v1 ) = A cos ωt + (cos ωt − cos 3ωt) (4.42)
4
where the angular frequency ω can be solved from Eqs. (4.33) and (4.40)
α2 = Ω2 + ε2 ω1 = Ω2 + 2q 2 A2 Ω2 (4.43)
which leads to
α
Ω= p . (4.44)
1 + 2(qA)2
Its period can be approximately expressed as follows
2π p
Tapp = 1 + 2q 2 A2 . (4.45)
α
April 13, 2006 10:4 WSPC/140-IJMPB 03379

1170 J.-H. He

In case qA is sufficiently small, i.e. 0 < qA  1, it follows that

2π
(1 + q 2 A2 ) .
Tapp =
α
Consequently, in this limit, the present method gives exactly the same results
as the standard Lindstedt–Poincare method.98 However, in our present study, qA
need not be small, even in case qA → ∞, the present results also have high accuracy
2 π/2
R p √
Tex π 0 1 + 4q 2 A2 cos2 t dt 2 2
lim = lim = = 0.900 (4.46)
|qA|→∞ Tapp
p
|qA|→∞ 1 + 2q 2 A2 π
where Tex is the exact period, which reads
Z π/2 p
2
Tex = 1 + 4q 2 A2 cos2 t dt . (4.47)
απ 0
Therefore, for any values of qA, it can be easily proven that
|Tex − Tapp |
0≤ ≤ 10% .
Tapp
Example 3. Consider the equation (Exercise 4.4 in Ref. 99)

(1 + u2 )u00 + u = 0 , u(0) = A , ¬u0 (0) = 0 . (4.48)

We let u = εv in Eq. (4.48) and obtain
v 00 + 1 · v + ε2 v 2 v 00 = 0 , v(0) = A/ε , v 0 (0) = 0 . (4.49)
2
Supposing that the solution of Eq. (4.49) and ω can be expressed in the form
v = v 0 + ε 2 v1 + ε 4 v2 + · · · (4.50)

1 = ω 2 + ε 2 ω1 + ε 4 ω2 + · · · . (4.51)
Substituting Eqs. (4.50) and (4.51) into Eq. (4.49) and equating coefficients of
like powers of ε yields the following equations
v000 + ω 2 v0 = 0 , v0 (0) = A/ε , v00 (0) = 0 (4.52)

v100 + ω 2 v1 + ω1 v0 + v02 v000 = 0 , v1 (0) = 0 , v10 (0) = 0 . (4.53)

Solving Eq. (4.52) results in
A
cos ωt .
v0 = (4.54)
ε
Equation (4.53), therefore, can be re-written down as
ω1 A ω 2 A3
v100 + ω 2 v1 + cos ωt − cos3 ωt = 0 (4.55)
ε ε3
or
ω1 A 3ω 2 A3 ω 2 A3
 
v100 + ω 2 v1 + − cos ωt − cos 3ωt = 0 . (4.56)
ε 4ε3 4ε3
April 13, 2006 10:4 WSPC/140-IJMPB 03379

Some Asymptotic Methods for Strongly Nonlinear Equations 1171

We let
3ω 2 A2
ω1 = (4.57)
4ε2
in Eq. (4.56) so that the secular term can be eliminated. Solving Eq. (4.56) yields
A3
v1 = (cos ωt − cos 3ωt) (4.58)
32ε3
Thus we obtain the first-order approximate solution of the original Eq. (4.48),
which reads
A3
u = ε(v0 + ε2 v1 ) = A cos ωt + (cos ωt − cos 3ωt) . (4.59)
32
Substituting Eq. (4.57) into Eq. (4.51) results in
3ω 2 A2
1 = ω 2 + ε 2 ω1 = ω 2 + (4.60)
4
which leads to
1
ω=q . (4.61)
1 + 43 A2

To compare, we write Nayfeh’s result,98 which can be written in the form

  
3
u = ε1/2 a cos 1 − εa2 t + θ (4.62)
8
where ε is a small dimensionless parameter that is a measure of the amplitude
of oscillation, the constants a and θ can be determined by the initial conditions.
Imposing the initial conditions, Nayfeh’s result can be re-written as
 
3 2
u = A cos 1 − A t . (4.63)
8
It is also interesting to point out that Nayfeh’s result is only valid only for small
amplitude A, while the present results (4.59) and (4.61) are valid not only for small
amplitude A, but also for very large values of amplitude.

5. Homotopy Perturbation Method

The homotopy perturbation method44 – 48,50 – 52,62,72 provides an alternative ap-
proach to introducing an expanding parameter.
To illustrate its basic ideas of the method, we consider the following nonlinear
differential equation
A(u) − f (r) = 0 , r∈Ω (5.1)
with boundary conditions
B(u, ∂u/∂n) = 0 , r∈Γ (5.2)
April 13, 2006 10:4 WSPC/140-IJMPB 03379

1172 J.-H. He

where A is a general differential operator, B is a boundary operator, f (r) is a known

analytic function, Γ is the boundary of the domain Ω.
The operator A can, generally speaking, be divided into two parts L and N ,
where L is linear, while N is nonlinear. Equation (5.1), therefore, can be rewritten
as follows
L(u) + N (u) − f (r) = 0 . (5.3)
By the homotopy technique, we construct a homotopy v(r, p) : Ω × [0, 1] → R
which satisfies
H(v, p) = (1 − p)[L(v) − L(u0 )] + p[A(v) − f (r)] = 0 , (5.4a)
or
H(v, p) = L(v) − L(u0 ) + pL(u0 ) + p[N (v) − f (r)] = 0 , (5.4b)
where p ∈ [0, 1] is an embedding parameter, u0 is an initial approximation of
Eq. (5.1), which satisfies generally the boundary conditions.
Obviously, from Eq. (5.4a) or Eq. (5.4b) we have
H(v, 0) = L(v) − L(u0 ) = 0 , (5.5)
and
H(v, 1) = A(v) − f (r) = 0 . (5.6)
It is obvious that when p = 0, Eq. (5.4a) or Eq. (5.4b) becomes a linear equation;
when p = 1 it becomes the original nonlinear one. So the changing process of p
from zero to unity is just that of L(v) − L(u0 ) = 0 to A(v) − f (r) = 0. The
embedding parameter p monotonically increases from zero to unit as the trivial
problem L(v)−L(u0 ) = 0 is continuously deformed to the problem A(v)−f (r) = 0.
This is a basic idea of homotopy method which is to continuously deform a simple
problem easy to solve into the difficult problem under study.
The basic assumption is that the solution of Eq. (5.4a) or Eq. (5.4b) can be
written as a power series in p:
v = v0 + pv1 + p2 v2 + · · · . (5.7)
Setting p = 1 results in the approximate solution of Eq. (5.1):
u = lim v = v0 + v1 + v2 + · · · . (5.8)
p→1

Recently, some rather extraordinary virtues of the homotopy perturbation

method have been exploited. The method has eliminated limitations of the tra-
ditional perturbation methods. On the other hand it can take full advantage of
the traditional perturbation techniques so there has been a considerable deal of
research in applying homotopy technique for solving various strongly nonlinear
equations,2 – 4,25,29,37,109,110,120 furthermore the differential operator L does not
need to be linear, see Ref. 29.
April 13, 2006 10:4 WSPC/140-IJMPB 03379

Some Asymptotic Methods for Strongly Nonlinear Equations 1173

Liao84 – 87 proposes a kind of approximate solution technique based on the ho-

motopy technique. In Liao’s approach, the solution is expanded into a more gener-
alized Taylor series with a free parameter ~, a careful choice of the parameter ~, as
illustrated in Liao’s publications, leads to fast convergence.
Different from Liao’s method, the homotopy perturbation method uses the
imbedding parameter p as a small parameter, only few iterations (2 or 3) are needed
to search for the needed asymptotic solution, unlike Liao’s method where infinite
series is needed, the suitable choice of the free parameter ~ can guarantee the
convergence of the obtained series. Comparison between homotopy perturbation
method and Liao’s method is systematically discussed in Ref. 51.

5.1. Periodic solution

As an illustrative example, we consider a problem of some importance in plasma
physics concerning an electron beam injected into a plasma tube where the magnetic
field is cylindrical and increases toward the axis in inverse proportion to the radius.
The beam is injected parallel to the axis, but the magnetic field bends the path
toward the axis. The governing equation for the path u(x) of the electrons is8
u00 + u−1 = 0 , (5.9)
with initial conditions u(0) = A, u0 (0) = 0.
Physical considerations show that this equation has a periodic solution. In order
to look for the periodic solution, we suppose that the frequency of Eq. (5.9) is ω,
and assume that the nonlinear term in Eq. (2.9) can be approximated by a linear
term:60
u−1 ∼ ω 2 u . (5.10)
So we obtain a linearized equation for Eq. (5.9)
u00 + ω 2 u = 0 . (5.11)
We re-write Eq. (5.9) in the form
u00 + ω 2 u = ω 2 u − u−1 . (5.12)
We construct the following homotopy
u00 + ω 2 u = p(ω 2 u − u−1 ) . (5.13)
The embedding parameter p monotonically increases from zero to unit as the trivial
problem, so Eq. (5.11) is continuously deformed to the original problem, Eq. (5.9).
So if we can construct an iteration formula for Eq. (5.13), the series of approxi-
mations comes along the solution path, by incrementing the imbedding parameter
from zero to one; this continuously maps the initial solution into the solution of the
original Eq. (5.9).
Supposing that the solution of Eq. (5.13) can be expressed in a series in p:
u = u0 + pu1 + p2 u2 + · · · . (5.14)
April 13, 2006 10:4 WSPC/140-IJMPB 03379

1174 J.-H. He

We, therefore, obtain the following linear equations for u0 and u1 :

u000 + ω 2 u0 = 0 , u0 (0) = A , u00 (0) = 0 , (5.15)

u001 + ω 2 u1 = ω 2 u0 − u−1
0 , u1 (0) = 0 , u01 (0) = 0 . (5.16)
The solution of Eq. (5.15) is u0 (t) = A cos ωt. Substituting u0 into Eq. (5.16)
results in
1
u001 + ω 2 u1 = A ω 2 cos ωt − . (5.17)
A cos ωt
A rule of avoiding secular term in u1 leads to the condition
Z π/(2ω)  
2 1
cos ωt A ω cos ωt − dt = 0 , (5.18)
0 A cos ωt
which leads to the result
√
ω= 2/A . (5.19)
Applying the variational iteration method as illustrated in Sec. 2.5, we can solve
u1 from Eq. (5.17) under the initial conditions u1 (0) = 0 and u01 (0) = 0,
1 t
 
1
Z
2
u1 (t) = sin ω(s − t) Aω cos ωs − ds . (5.20)
ω 0 A cos ωs
So we obtain the first-order approximate solution by setting p = 1, which reads
1 t
 
1
Z
2
u = u0 + u1 = A cos ωt + sin ω(s − t) Aω cos ωs − ds . (5.21)
ω 0 A cos ωs
The period of the system can be expressed in the form
√
T = 2πA = 4.442A . (5.22)
The exact period can be easily computed,
√ Z A du √ Z A
du
Tex = 2 2 s =2 2 √ . (5.23)
0
Z A 0 ln A − ln u
1/sds
u

By transformation u = As, Eq. (5.23) leads to

√ Z 1 du √
Tex = 2 2A p = 2 2πA = 5.01325A . (5.24)
0 ln(1/s)
There exist alternative approaches to the construction of the needed homotopy. We
rewrite Eq. (5.9) in the form
u00 u2 + u = 0 . (5.25)
We construct the following homotopy:
0 · u00 + 1 · u + pu00 u2 = 0 . (5.26)
April 13, 2006 10:4 WSPC/140-IJMPB 03379

Some Asymptotic Methods for Strongly Nonlinear Equations 1175

Please note that we recover the main term u00 by multiplying zero. Assuming the
solution, the coefficients of u and u00 , 1 and 0, respectively, can be expanded, in the
following form
u = u0 + pu1 + p2 u2 + · · · (5.27)

1 = ω 2 + pa1 + p2 a2 + · · · (5.28)

0 = 1 + pb1 + p2 b2 + · · · . (5.29)
Substituting Eqs. (5.27)–(5.29) into Eq. (5.26) results in
(1 + pb1 + · · ·)(u000 + pu001 + · · · ·) + (ω 2 + pa1 + · · ·)(u0 + pu1 + · · · ·)
+p(u000 + pu001 + · · · ·)(u0 + pu1 + · · · ·)2 = 0 . (5.30)
Collecting terms of the same power of p, gives
u000 + ω 2 u0 = 0 , (5.31)

u001 + ω 2 u1 + b1 u000 + a1 u0 + u000 u20 = 0 . (5.32)

The solution of Eq. (5.31) is u0 (x) = A cos ωx. Substituting u0 into Eq. (5.32), by
simple manipulation, yields
 
3 1
u001 + ω 2 u1 − b1 ω 2 − a1 + A2 ω 2 A cos ωx − A3 ω 2 cos 3ωx = 0 . (5.33)
4 4
No secular term requires that
3
b1 ω 2 − a 1 + A2 ω 2 = 0 . (5.34)
4
If the first-order approximate solution is enough, then from Eqs. (5.28) and (5.29),
we have
1 = ω 2 + a1 (5.35)

0 = 1 + b1 . (5.36)
Solving Eqs. (5.34)–(5.36) simultaneously, we obtain
2
ω = √ A−1 = 1.1547A−1 . (5.37)
3
The accuracy 7.8% is remarkable good.
We can also rewrite Eq. (5.9) in the form
1 + u00 u = 0 . (5.38)
00
Multiplying both sides of Eq. (5.38) by u , we have
u00 + u002 u = 0 . (5.39)
We construct a homotopy in the form
u00 + 0 · u + pu002 u = 0 . (5.40)
April 13, 2006 10:4 WSPC/140-IJMPB 03379

1176 J.-H. He

Similarly, we expand the solution and the coefficient of the middle term, 0, into the
forms
u = u0 + pu1 + p2 u2 + · · · (5.41)

0 = ω 2 + pa1 + p2 a2 + · · · . (5.42)
As a result, we obtain
u000 + ω 2 u0 = 0 , (5.43)

u001 + ω 2 u1 + a1 u0 + u002
0 u0 = 0 . (5.44)
The solution of Eq. (5.43) is u0 (x) = A cos ωx. Substituting u0 into Eq. (5.44), by
simple manipulation, yields
 
3 4 2 1
u1 + ω u1 + a1 + ω A A cos ωx + ω 4 A3 cos 3ωx = 0 .
00 2
(5.45)
4 4
No secular terms in u1 requires that
3
a1 = − ω 4 A2 . (5.46)
4
If we stop at the first-order approximate solution, then Eq. (5.42) becomes
3
0 = ω 2 − ω 4 A2 , (5.47)
4
from which we can obtain the same result as Eq. (5.37).
Now we consider Duffing equation
u00 + u + εu3 = 0 , ε>0 (5.48)
with initial conditions u(0) = A, and u0 (0) = 0.
We construct the following homotopy:
u00 + ω 2 u + p(εu3 + (1 − ω 2 )u) = 0 , p ∈ [0, 1] . (5.49)
When p = 0, Eq. (5.49) becomes the linearized equation, u00 + ω 2 u = 0, when
p = 1, it turns out to be the original one. We assume that the periodic solution to
Eq. (5.49) may be written as a power series in p:
u = u0 + pu1 + p2 u2 + · · · . (5.50)
Substituting Eq. (5.50) into Eq. (5.49), collecting terms of the same power of p,
gives
u000 + ω 2 u0 = 0 , u0 (0) = A , u00 (0) = 0 (5.51)

u001 + ω 2 u1 + εu30 + (1 − ω 2 )u0 = 0 , u1 (0) = 0 , u01 (0) = 0 . (5.52)

Equation (5.51) can be solved easily, giving u0 = A cos ωt. If u0 is substituted into
Eq. (5.52), and the resulting equation is simplified, we obtain
 
3 1
u001 + ω 2 u1 + 1 + εA2 − ω 2 A cos ωt + εA3 cos 3ω = 0 . (5.53)
4 4
April 13, 2006 10:4 WSPC/140-IJMPB 03379

Some Asymptotic Methods for Strongly Nonlinear Equations 1177

No secular terms in u1 requires that

r
3
ω= 1 + εA2 . (5.54)
4
The accuracy of the expression (5.54), is 7.6% when ε → ∞.
Secular terms arise for higher-order approximate solutions so we always stop
before the second iteration. If higher-order approximate solution is required, the
parameter-expanding method (the modified Lindstedt–Poincare method67 ) can be
applied. For example, we can construct the following homotopy

u00 + 1 · u + pεu3 = 0 , (5.55)

and we expand the coefficient of the linear term into a series of p:

1 = ω 2 + pω1 + p2 ω2 + · · · (5.56)

where ωi can be identified in view of absence of secular terms in ui :

3 3
ω1 = − εA2 , ω2 = − ε2 A4 . (5.57)
4 128ω 2
Substituting in Eq. (5.56) and setting p = 1, we have
3 3
1 = ω 2 − εA2 − ε2 A4 , (5.58)
4 128ω 2
or
v q
u
u 1 + 3 εA2 + 1 + 3 εA2 + 5 ε2 A4
t 4 2 8
ω= . (5.59)
2
The accuracy of Eq. (5.59) reaches 5.9% when ε → ∞.
We can also construct a homotopy in the form

(1 − p)(v 00 + ω 2 v − u000 − ω 2 u0 ) + p(v 00 + v + εv 3 ) = 0 . (5.60)

By simple operation, we have

v000 + ω 2 v0 − u000 − ω 2 u0 = 0 , (5.61)

v100 + ω 2 v1 − (v000 + ω 2 v0 − u000 − ω 2 u0 )(v000 + v0 + εv03 ) = 0 . (5.62)

We begin with

v0 = u0 = A cos ωt . (5.63)

Substituting Eq. (5.63) into Eq. (5.62), we have

 
3 2 1
v1 + ω v1 + a −ω + 1 + εA cos ωt + εA3 cos 3ωt = 0 ,
00 2 2
(5.64)
4 4
which is identical to Eq. (5.53).
April 13, 2006 10:4 WSPC/140-IJMPB 03379

1178 J.-H. He

5.2. Bifurcation
In this section, we consider the following Duffing oscillator in space86
u00 + ε(u − A2 u3 ) = 0 , u(0) = u(π) = 0 , u(π/2) = 1 , ε ≥ 0. (5.65)
For any ε ≥ 0, the above equation has the trivial solution u(t) = 0. The so-called
bifurcation occurs when a nontrivial solution exists for some of values of ε.
The traditional perturbation method predicts that86
r
ε−1
A = ±2 , ε ≥ 1, (5.66)
3
which is only valid for small values of ε.
According to the homotopy perturbation, we construct the following simple
homotopy:
u00 + εu − εpA2 u3 = 0 . (5.67)
Assume that the solution can be expressed in the form
u = u0 + pu1 + p2 u2 + · · · . (5.68)
In the parlance of the parameter-expanding method (the modified Lindstedt–
Poincare method67 ), we expand the coefficient of u in the middle term of Eq. (5.67)
in the form
ε = ω 2 + pω1 + p2 ω2 + · · · . (5.69)
Substituting Eqs. (5.68) and (5.69) into Eq. (5.67), and equating the terms with
the identical powers of p, we can obtain a series of linear equations and write only
the first two linear equations:
u000 + ω 2 u0 = 0 , u0 (0) = u0 (π) = 0 , u0 (π/2) = 1 , (5.70)

u001 + ω 2 u1 + ω1 u0 − εA2 u30 = 0 , u1 (0) = u1 (π) = 0 , u1 (π/2) = 0 . (5.71)

The general solution of Eq. (5.70) is
u0 (t) = a cos ωt + b sin ωt , (5.72)
where a and b are constant. Adjoining to the boundary conditions: u0 (0) = u0 (π) =
0, and the additional condition u0 (π/2) = 1, we have a = 1, b = 0, and ω = 1. So
the solution of u0 reads
u0 = sin t . (5.73)
Substituting u0 into Eq. (5.71), we obtain a differential equation for u1 ,
u001 + u1 + ω1 sin t − εA2 sin3 t = 0 . (5.74)
Or
 
3 1
u001 + u1 + ω1 − εA2 sin t − εA2 sin 3t = 0 . (5.75)
4 4
April 13, 2006 10:4 WSPC/140-IJMPB 03379

Some Asymptotic Methods for Strongly Nonlinear Equations 1179

No secular terms in u1 requires that

3 2
ω1 = εA . (5.76)
4
The solution of Eq. (5.75), considering the boundary conditions u1 (0) = u1 (π) = 0,
and the additional condition u1 (π/2) = 0, is
1
u1 = − εA2 (sin 3t + sin t) . (5.77)
32
We, therefore, obtain the first-order approximate solution which reads
1
u = u0 + u1 = sin t − εA2 (sin 3t + sin t) , (5.78)
32
and
3
ε = 1 + εA2 . (5.79)
4
Since ε ≥ 0, the above equation has no solution when ε < 1. However, when ε > 1,
Eq. (5.79) has the solution
r
2 1
A = ±√ 1− . (5.80)
3 ε
Therefore, the so-called simple bifurcation occurs at ε = 1. Our result is the same
with that obtained by Liao.86

6. Iteration Perturbation Method

In this section, we will illustrate a new perturbation technique coupling with the
iteration method.68 Consider the following nonlinear oscillation:
u00 + u + εf (u, u0) = 0 , u(0) = A , u0 (0) = 0 . (6.1)
We rewrite Eq. (6.1) in the following form:
u00 + u + εu · g(u, u0 ) = 0 , (6.2)
where g(u, u0 ) = f /u.
We construct an iteration formula for the above equation:
u00n+1 + un+1 + εun+1 g(un , u0n ) = 0 , (6.3)
where we denote by un the nth approximate solution. For nonlinear oscillation,
Eq. (6.3) is of Mathieu type. We will use the perturbation method to find approxi-
mately un+1 . The technique is called iteration perturbation method.68,71
Consider Eq. (3.27), and assume that initial approximate solution can be ex-
pressed as u0 = A cos ωt, where ω is the angular frequency of the oscillation, we
rewrite Eq. (3.27) approximately as follows
d2 u
+ εA2 u cos2 ωt = 0 , (6.4a)
dt2
April 13, 2006 10:4 WSPC/140-IJMPB 03379

1180 J.-H. He

or
d2 u 1 2 1
+ εA u + εA2 u cos 2ωt = 0 , (6.4b)
dt2 2 2
which is Mathieu type. Suppose that

u = u0 + εu1 + ε2 u2 + · · · , (6.5)

1 2
εA = ω 2 + εc1 + ε2 c2 + · · · (6.6)
2
and substituting Eqs. (6.5) and (6.6) into Eq. (6.4b), and equating the coefficients
of the same power of ε, we have the following two differential equations for u1 and
u2 :
1
u001 + ω 2 u1 + c1 u0 + A2 cos 2ωtu0 = 0 , u1 (0) = 0 , u01 (0) = 0 , (6.7)
2

1
u002 + ω 2 u2 + c2 u0 + c1 u1 + A2 cos 2ωtu1 = 0 , u2 (0) = 0 , u02 (0) = 0 , (6.8)
2
where u0 = A cos ωt. Substituting u0 into Eq. (6.7), the differential equation for
u1 becomes
 
1 2 1
u1 + ω u1 + A c1 + A cos ωt + A3 cos 3ωt = 0 .
00 2
(6.9)
4 4
The requirement of no secular term requires that
1
c1 = − A2 . (6.10)
4
Solving Eq. (6.9), subject to the initial conditions u1 (0) = 0 and u01 (0) = 0, yields
the result:
1
u1 = A3 (cos 3ωt − cos ωt) . (6.11)
32ω 2
We, therefore, obtain its first-order approximate solution, which reads
εA3 A
u = A cos ωt + 2
(cos 3ωt − cos ωt) = A cos ωt + (cos 3ωt − cos ωt)(6.12)
,
32ω 24
where the angular frequency is determined from the relations (6.6) and (6.10), which
reads
√
3 1/2
ω= ε A. (6.13)
2
To obtain the second-order approximate solution, we substitute u0 and u1 into
Eq. (6.8), in the parlance of no secular, we find
c1 A2 A4
c2 = 2
=− . (6.14)
32ω 128ω 2
April 13, 2006 10:4 WSPC/140-IJMPB 03379

Some Asymptotic Methods for Strongly Nonlinear Equations 1181

Substituting the determined c1 and c2 into Eq. (6.6), we have

1 2 1 ε2 A4
εA = ω 2 − εA2 − , (6.15)
2 4 128ω 2
which leads to the result:
ω = 0.8174ε1/2A . (6.16)
−1/2 −1
However, the second-order period T = 7.6867ε A is not more accurate than
the first-order one, because Eq. (6.4b) is an approximate one, the obtained result
(6.16) is the approximate angular frequency of Eq. (6.4b), not that of the original
one. So we need not search for higher-order approximations. To obtain approximate
solutions with higher accuracy, we replace zeroth-order approximate solution by the
following
A
u0 = A cos ωt + (cos 3ωt − cos ωt) . (6.17)
24
So the original Eq. (3.27) can be approximated by the linear equation
2
d2 u

A
+ ε A cos ωt + (cos 3ωt − cos ωt) u = 0. (6.18)
dt2 24
By the same manipulation, we can identify the angular frequency ω = 0.8475ε1/2A,
and obtain the approximate period T = 7.413ε−1/2A−1 . The relative error is about
0.05%.
Now we re-consider Eq. (5.9) in a more general form:
d2 u c
+ = 0. (6.19)
dt2 u
We approximate the above equation by
d2 u c
+ 2 u = 0, (6.20)
dt2 A cos2 ωt
or
2c
u00 + u + u00 cos 2ωt = 0 . (6.21)
A2
We introduce an artificial parameter in Eq. (6.21)
2c
u00 + u + εu00 cos 2ωt = 0 . (6.22)
A2
And suppose that
2c
= ω 2 + εc1 + ε2 c2 + · · · , (6.23)
A2
u = u0 + εu1 + ε2 u2 + · · · , (6.24)
we obtain the differential equation for u1 :
u001 + ω 2 u1 + c1 u0 + u000 cos 2ωt = 0 , (6.25)
April 13, 2006 10:4 WSPC/140-IJMPB 03379

1182 J.-H. He

or

u001 + ω 2 u1 = −Ac1 cos ωt + Aω 2 cos ωt cos 2ωt

Aω 2 Aω 2
 
= −Ac1 + cos ωt + cos 3ωt . (6.26)
2 2
The requirement of no secular term needs
ω2
c1 = . (6.27)
2
So we identify the angular frequency as follows
r
4c
ω= . (6.28)
3A2
The approximate period can be written as
√
3πA 5.44A
T = 1/2 = 1/2 . (6.29)
c c
Acton and Squire,8 using the method of weighted residuals, obtained the following
result:
16A 5.33A
T = 1/2
= 1/2 . (6.30)
3c c
Now we consider two-dimensional viscous laminar flow over an finite flat-plain gov-
erned by a nonlinear ordinary differential equation:61,86,87
1
f 000 + f f 00 = 0 , x ∈ [0, +∞] , (6.31)
2
with boundary conditions f (0) = f 0 (0) = 0, and f 0 (+∞) = 1.
The prime in Eq. (6.31) denotes the derivatives with respect to x which is defined
as
r
U
x=Y ,
νX
and f (x) is relative to the stream function Ψ by
Ψ
f (x) = √ .
νU X
Here, U is the velocity at infinity, ν is the kinematic viscosity coefficient, X and
Y are the two independent coordinates.
In order to obtain a perturbation solution of Blasius equation, we introduce an
artificial parameter ε in Eq. (6.31)
1
f 000 + εf f 00 = 0 . (6.32)
2
April 13, 2006 10:4 WSPC/140-IJMPB 03379

Some Asymptotic Methods for Strongly Nonlinear Equations 1183

Processing in a traditional way of perturbation technique, and supposing that

f 00 (0) = σ, we obtain a solution of Eq. (6.32) in the form of a power series87
+∞  k
X 1 Ak σ k+1 3k+2
f (x) = − x
2 (3k + 2)!
k=0

where


 1 (k = 0 and k = 1)
 !
Ak = k−1
P 3k − 1

 Ar Ak−r−1 (k ≥ 2)
3r

r=0

which is valid only for small x.

Now we begin with the initial approximate solution
1
f0 (x) = x − (1 − e−bx ) , (6.33)
b
where b is an unknown constant.
Equation (6.31) can be approximated by the following equation:
 
000 1 1
f + x − (1 − e ) f 00 = 0 ,
−bx
(6.34)
2 b
We rewrite Eq. (6.34) in the form
 
1 1
f 000 + bf 00 + x − (1 − e−bx ) − 2b f 00 = 0 , (6.35)
2 b
and embed an artificial parameter ε in Eq. (6.35):
 
1 1
f + bf + ε x − (1 − e ) − 2b f 00 = 0 .
000 00 −bx
(6.36)
2 b
Suppose that the solution of Eq. (6.36) can be expressed as

f = f0 + εf1 + · · · , (6.37)

we have the following linear equations:

f0000 + bf000 = 0 , f0 (0) = f00 (0) = 0 , and f00 (+∞) = 1 , (6.38)

 
1 1 1
f1000 + bf100 + x + e−bx − 2b − f000 = 0 ,
2 b b (6.39)
f1 (0) = f10 (0) = 0, and f10 (+∞) = 0.

The solution of Eq. (6.38) is f0 (x) = x − (1 − e−bx )/b, substituting it into

Eq. (6.39) results in
1
f1000 + bf100 = − (bx + e−bx − 2b2 − 1)e−bx . (6.40)
2
April 13, 2006 10:4 WSPC/140-IJMPB 03379

1184 J.-H. He

The constant b can be identified by the following expression:

Z +∞
e−bx (bx + e−bx − 2b2 − 1)e−bx dx = 0 . (6.41)
0

So we have
1 1 1 + 2b2
Γ(2) + − = 0, (6.42)
4b 3b 2b
which leads to the result:
√
b = 1/ 12 = 0.28867 . (6.43)
The exact solution of Eq. (6.40) is not required, the expression (6.41) requires no
terms of xe−βx (n = 1, 2, 3, . . .) in f1 . So we can assume that the approximate
solution of Eq. (6.40) can be expressed as
1
f1 (t) = Ae−bx + e−2bx + B , (6.44)
8
where the constants A and B can be identified from the initial conditions f1 (0) =
f10 (0) = 0, i.e. A = −1/4 and B = 1/8.
By setting p = 1, we obtain the first-order approximate solution, i.e.
f 00 (x) = f000 (x) + f100 (x) , (6.45)
where f1 and f2 are defined by Eqs. (6.33) and (6.44) respectively.
A highly accurate numerical solution of Blasius equation has been provided by
Howarth,78 who gives the initial slop fex
00
(0) = 0.332057. Comparing the approxi-
mate initial slop:
f 00 (0) = f000 (0) + f100 (0) = 0.3095
we find that the relative error is 6.8%.
We can improve the accuracy of approximate solution to a high-order by the
iteration technique. Now we use Eq. (6.45) as the initial approximate solution of
Eq. (6.31):
1 1 1 1
f˜0 (x) = x − (1 − e−bx ) − e−bx + e−2bx + , (6.46)
b 4 8 8
where b is undermined constant.
By parallel operation, the constant b can be identified by the following relation:
Z ∞  
−bx −bx 2 1 −bx 1 −2bx 1 −bx
e bx + e − 2b − 1 − e + e + e dx = 0 (6.47)
0 4 8 8
which leads to the result
b = 0.3062 . (6.48)
It is obvious that f˜000 (0) = b + 0.25b2 = 0.3296 reach a very high accuracy, and the
0.73% accuracy is remarkable good.
April 13, 2006 10:4 WSPC/140-IJMPB 03379

Some Asymptotic Methods for Strongly Nonlinear Equations 1185

7. Ancient Chinese Methods

China is one of the four countries with an ancient civilization. The ancient Chinese
mathematicians had made greatest contributions to the development of human cul-
ture, however, essentially nothing of a primary nature has come down to West con-
cerning ancient Chinese mathematics, little has been discussed on ancient Chinese
mathematics in some of the most famous monographs on history of mathematics.
So the present author feels strongly necessary to give a basic introduction to the
great classics of ancient Chinese mathematics for our Western colleagues, who are
unfamiliar with the Chinese language. This section concerns briefly some famous
ancient Chinese algorithms, which might find wide applications in modern science.
Jiu Zhang Suan Shu, Nine Chapters on the Art of Mathematics, comprises nine
chapters and hence its title is the oldest and most influential work in the history
of Chinese mathematics. It is a collection of 246 problems on agriculture, business
procedure, engineering, surveying, solution of equations, and properties of right tri-
angles. Rules of solution are given systematically, but there exists no proofs in Greek
sense. As it was pointed by Dauben34 that the Nine Chapters can be regarded as a
Chinese counterpart to Euclid’s Elements, which dominated Western mathematics
in the same the Nine Chapters came to be regarded as the seminal work of ancient
Chinese mathematics for nearly two millennia. Great classics, when revisited in the
light of new developments, may reveal hidden pearls, as is the case with ancient
Chinese method and He Chengtian’s inequality, which will be discussed below.

7.1. Chinese method 53

The Chinese algorithm is called the Ying Buzu Shu in Chinese, which was proposed
in about 2nd century BC, known as the rule of double false position in West after
1202 AD. Chapter 7 of the Nine Chapters is the Ying Buzu Shu (lit. Method of
Surplus and Deficiency; too much, too less), an ancient Chinese algorithm, which is
the oldest method for approximating real roots of a nonlinear equation. To illustrate
the basic idea of the method, we consider the 15th example in this chapter, which
reads:
The weight of a jade with a cubic chun is 7 liang, while the weight of a stone
with a cubic chun is 6 liang. There is a cubic bowlder with a length of 3 chun, and
a mass of 11 jin (1 jin = 16 liang). What is mass of the jade and the stone in the
bowlder?
Answer: the volume of the jade in the bowlder is 14 cubic chun, the mass is
6 jin and 2 liang; the volume of the stone is 13 cubic chun, the mass is 4 jin and
14 liang.
The solution procedure: Assume that the bowlder is made of jade only, then
the mass is overestimated by 13 liang. If we assume that bowlder is made of stone
only, the mass is underestimated by 14 liang. According to the Ying Buzu Shu, the
mass of jade can be determined.
April 13, 2006 10:4 WSPC/140-IJMPB 03379

1186 J.-H. He

In the language of modern mathematics, let x and y be the volumes of the jade
and stone, respectively, then we have
x + y = 33 = 27
and
7x + 6y = 11 × 16 = 176 .
We write R(x, y) = 7x + 6y − 198, and let x1 = 27, y1 = 0 (assume that the bowlder
is made of jade only), we have R1 (27, 0) = 13; if let x2 = 0, y2 = 27 (assume that
bowlder is made of stone only), then we obtain R2 (0, 27) = −14. According the
ancient Chinese method, we can calculate the value of x in the form
x2 R 1 − x 1 R 2 27 × 14
x= = = 14 .
R1 − R 2 13 + 14
The ancient Chinese method can also be powerfully applied to nonlinear prob-
lems. We consider the last example of Chapter 7. It reads:
There is a wall with a thickness of 5 chi. Two mice excavate a hole in the wall
from both sides. The big mouse burrows 1 chi in first day, and doubles its headway
everyday afterwards; while the small one burrows 1 chi in first day also, but halves
its headway everyday afterwards. When the hole can be completed? What is the
length for each mouse?
Answer: 2 and 2/17 days. The big mouse penetrates totally 3 and 12/17 chi,
while the small one 1 and 5/17 chi.
The solution procedure: Assume that it needs two days, then 5 chun (1 chun
= 0.1 chi) is left. If we assume that it needs three days, then 3 chi and 0.75 chun
is redundant. According to the Ying Buzu Shu, the days needed for excavation can
be determined.
30.75 × 2 + 5 × 3 2
≈2 .
5 + 30.75 17
Consider an algebraic equation,
f (x) = 0 . (7.1)
Let x1 and x2 be the approximate solutions of the equation, which lead to the
remainders f (x1 ) and f (x2 ) respectively, the ancient Chinese algorithm leads to
the result
x2 f (x1 ) − x1 f (x2 )
x= . (7.2)
f (x1 ) − f (x2 )
Now we rewrite Eq. (7.2) in the form
x2 f (x1 ) − x1 f (x2 ) f (x1 )(x1 − x2 )
x3 = = x1 − . (7.3)
f (x1 ) − f (x2 ) f (x1 ) − f (x2 )
If we introduce the derivative f 0 (x1 ) defined as
f (x1 ) − f (x2 )
f 0 (x1 ) = , (7.4)
x1 − x 2
April 13, 2006 10:4 WSPC/140-IJMPB 03379

Some Asymptotic Methods for Strongly Nonlinear Equations 1187

then we have
f (x1 )
x = x1 − , (7.5)
f 0 (x1 )
which is the well-known Newton iteration formulation proposed by Newton (1642–
1727). Chinese mathematicians had used the theory for more than one millennium.
The Chinese algorithm seems to have some advantages over the well-known Newton
iteration formula if the two points (x1 and x2 ) locate two sides of the exact root,
i.e. f (x1 ) · f (x2 ) < 0. Consider an example
f (x) = sin x . (7.6)
We begin with x1 = 0.1 and x2 = 4.7. It is obvious that f1 = 0.1 > 0 and
f2 = −1.0 < 0, so there is a root locating x ∈ (x1 , x2 ). By Eq. (7.2), we have
x2 f (x1 ) − x1 f (x2 )
x3 = = 1.43 .
f (x1 ) − f (x2 )
Calculating f (x3 ) results in f3 = 0.99 > 0, so the root locates at x ∈ (x2 , x3 ). Using
Eq. (7.2), we have
x2 f (x3 ) − x3 f (x2 )
x4 = = 3.06 .
f (x3 ) − f (x2 )
We see the procedure converges very fast to the exact root, the accuracy reaches
2.5% by only two iterations even for very poor initial prediction.
Consider the Duffing equation
u00 + u + εu3 = 0 , u(0) = A , u0 (0) = 0 . (7.7)
We use the trial functions u1 (t) = A cos t and u2 = A cos ωt, which are, respec-
tively, the solutions of the following linear equations
u00 + ω12 u = 0 , ω12 = 1 , (7.8)
and
u00 + ω22 u = 0 , ω22 = ω 2 . (7.9)
The residuals are
R1 (t) = εA3 cos3 t , (7.10)

R2 (t) = A(1 − ω 2 ) cos ωt + εA3 cos3 ωt . (7.11)

By the Chinese method, we have
ω12 R2 (0) − ω22 R1 (0) 1 − ω 2 + εA2 − ω 2 εA2
ω2 = = = 1 + εA2 , (7.12)
R2 (0) − R1 (0) 1 − ω2
i.e.,
p
ω= 1 + εA2 . (7.13)
April 13, 2006 10:4 WSPC/140-IJMPB 03379

1188 J.-H. He

The approximate period reads

2π
T =√ . (7.14)
1 + εA2
Table 1 illustrates the comparison between the approximate period with the exact
one when εA2 < 1.
The obtained solution obtained by the Chinese method is valid for the whole
solution domain, even in the case εA2 → ∞, the accuracy reaches 7.6%.

7.2. He Chengtian’s method 54

In a history book, it writes80
He Chengtian uses 26/49 as the strong, and 9/17 as the weak. Among the strong
and the weak, Chengtian tries to find a more accurate denominator of the fractional
day of the Moon. Chengtian obtains 752 as the denominator by using 15 and 1 as
weighting factors, respectively, for the strong and the weak. No other calendar can
reach such a high accuracy after Chengtian, who uses heuristically the strong and
weak weighting factors.
The above statement is generally called Tiao Ri Fa (Lit. Method for Modifica-
tion of Denominator). Here, we call it as He Chengtian’s interpolation or He Cheng-
tian’s method. There exists many other famous interpolations in ancient Chinese
mathematics.
In modern mathematical view, we illustrate the statement as follows:
According to the observation data, He Chengtian finds that
26 9
29 days > 1 Moon > 29 days .
49 17
Using the weighting factors (15 and 1), He Chengtian obtains
26 × 15 + 9 × 1 399
the fractional day = = ,
49 × 15 + 17 × 1 752
so
399
1 Moon = 29
days .
752
The error is about 0.1 second, it is the most accurate in He Chengtian’s time.
He Chengtian establishes Yuan-Jia Calendar, which is the most famous lunar cal-
endar in ancient China. The calendar was lunar with intercalary months to keep in
approximate synchronization with the seasons.

Table 1. Comparison of approximate period with the exact one.

εA2 0 0.042 0.087 0.136 0.190 0.25

Tex 6.283 6.187 6.088 5.986 5.879 5.767
T (Eq. 7.14) 6.283 6.155 6.026 5.895 5.760 5.620
April 13, 2006 10:4 WSPC/140-IJMPB 03379

Some Asymptotic Methods for Strongly Nonlinear Equations 1189

He Chengtian uses 26/49 as the strong weighing factor because it is much more
accurate than that of the weak weighing factor, 9/17. The observation shows that
the accuracy of the former is about 0.005%, while the accuracy for the latter is
about 0.22%. So the weighting factor for 26/49 is much larger than that for 9/17.
He Chengtian actually uses the following inequality:
If
a d
<x< , (7.15)
b c
where a, b, c and d are real numbers, then
a ma + nd d
< < , (7.16)
b mb + nc c
and x is approximated by
ma + nd
x(m, n) = , (7.17)
mb + nc
where m and n are weighting factors. We call x(1, 1) = (a + d)/(b + c) the He
Chengtian’s average of a/b and d/c.
Generally speaking m and n can be freely chosen, and the accuracy of the
obtained result is better than a/b or d/c. For example, if we choose m = 100 and
n = 1, then the improved fractional day can be calculated as
26 × 100 + 9 × 1 2609
the fractional day = = .
49 × 100 + 17 × 1 4917
The accuracy is about 0.0043%, and is improved compared with that of 26/49.
Equation (7.16) can be rewritten equivalently in the form
ma + nd ka + d
x= = , (7.18)
mb + nc kb + c
where k = m/n.
It is obvious that
d
lim x = , (7.19)
k→0 c
and
a
lim x = . (7.20)
k→∞ b
The changing process of k from zero to infinite is just that of x from d/c to a/b.
There must exist a certain value of k, while the corresponding value of x locates at
its exact solution.
There exists no general rule of choosing the value of k. It should be larger than 1
when a/b is more accurate than a/b. The more accurates a/b is, the larger value of
k. If the accuracy of a/b is lower than that of d/c, the value of k should be smaller
than 1. The less accurate a/b is, the smaller value of k.
April 13, 2006 10:4 WSPC/140-IJMPB 03379

1190 J.-H. He

In He Chengtian’s time, ancient Chinese mathematicians knew that 157/50 <

π < 22/7 (π = 3.14 was suggested by Liu Hui (263), and π = 22/7 was obtained
by Zu Chongzhi and/or his son Zu Geng.)
Using the weighting factors 1 and 9c , we have
157 + 9 × 22 355
π= = = 3.1415929 .
50 + 9 × 7 113
So we obtain
355 22
<π< ,
113 7
which was obtained by Zu Chongzhi (430–501).

Example 1. Consider the complete elliptic integral of the first kind

Z π/2 p
1 − k sin2 x dx .
0

It is obvious that

π/2
π π√
Z p
< 1 − k sin2 x dx < 1−k. (7.21)
2 0 2
By He Chengtian’s method, we have
Z π/2 p √
π
1 − k sin2 x dx = (p + q 1 − k)
0 2(p + q)
π √
= (1 + ξ 1 − k) , (7.22)
2(1 + ξ)
where p and q are weighting factors, ξ = q/p.
π−2
In the case k = 0, we have ξ = 0, and when k = 1, it follows ξ = 2 , so we
can approximately determine ξ in the form:
π−2
ξ= k. (7.23)
2
We, therefore, have
Z π/2 p
π−2 √
 
π
1 − k sin2 x dx = 1+ k 1−k . (7.24)
0 2 + (π − 2)k 2
Example 2. Consider the equation
(1 + u2 )u00 + u = 0 , u(0) = A , u0 (0) = 0 . (7.25)
We rewrite Eq. (7.25) in the form
1
u00 = − u. (7.26)
1 + u2
c 22/7 is more accurate than 157/50, so the value of weight factor for the former is bigger than

that of the latter.

April 13, 2006 10:4 WSPC/140-IJMPB 03379

Some Asymptotic Methods for Strongly Nonlinear Equations 1191

If we choose the trial-function in the form u = A cos ωt, where ω is the frequency,
then the maximal and minimal values of 1/(1 + u2 ) are, respectively, 1 and 1/(1 +
A2 ). So we immediately obtain
1 1
< ω2 < . (7.27)
1 1 + A2
According to He Chengtian’s interpolation, we set
m+n 1
ω2 = = , (7.28)
m + n(1 + A2 ) 1 + kA2
where m and n are weighting factors, k = n/(m + n).
So the frequency can be approximated as
1
ω= √ . (7.29)
1 + kA2
To compare, we write Nayfeh’s result,98 which can be written in the form
 
3
u = A cos 1 − A2 t , (7.30)
8
which is valid only for small amplitude A. To match Nayfeh’s result, we set k = 3/4
in Eq. (7.29), yielding the result
1
ω=q . (7.31)
1 + 43 A2

The accuracy reaches 8.5% even when A → ∞, which is remarkably good.

Example 3. Consider Duffing equation, which reads

u00 + u + εu3 = 0 , u(0) = A , u0 (0) = 0 (7.32)

where ε needs not be small in the present study, i.e. 0 ≤ ε < ∞.

We rewrite Eq. (7.32) in the form

u00 = −(1 + εu2 )u , (7.33)

which has a similar form of u00 = −ω 2 u. Assume that period solution of Eq. (7.32)
can be written in the form

u = A cos ωt , (7.34)

where ω is the frequency.

Observe that the square of frequency, ω 2 , is never less than that in the solution

ϕ1 (t) = A cos t (7.35)

of the following oscillation

u00 = −(1 + εu2min)u = −u . (7.36)

April 13, 2006 10:4 WSPC/140-IJMPB 03379

1192 J.-H. He

In addition, ω 2 never exceed the square of frequency of the solution

p
ϕ2 (t) = A cos 1 + εA2 t (7.37)
of the following oscillation
u00 = −(1 + εu2max )u = −(1 + εA2 )u . (7.38)
Hence, it follows that
r r
1 1 + εA2
<ω< . (7.39)
1 1
According to He Chengtian’s interpolation, we have
r
m + n(1 + εA2 ) p
ω= = 1 + kεA2 (7.40)
m+n
where k = n/(m + n).
In the case ε  1, the perturbation technique gives
3
ωpert = 1 + εA2 . (7.41)
8
Matching Eq. (7.40) with Eq. (7.41) when ε  1, we obtain k = 3/4. So the
frequency can is obtained as follows
r
3
ω = 1 + εA2 . (7.42)
4
Example 4. Consider the motion of a particle on a rotating parabola, which is
governed by the equation
(1 + 4q 2 u2 )u00 + ε2 u + 4q 2 uu02 = 0 , (7.43)
with initial conditions u(0) = A, and u0 (0) = 0.
We rewrite Eq. (7.43) in the form
u00 = −f (u)u , (7.44)
where
ε2 + 4q 2 u02
f (u) = . (7.45)
1 + 4q 2 u2
We begin with the trial function u(t) = A cos ωt, then we have
ε2 + 4q 2 ω 2 A2 sin2 ωt
f (u) = . (7.46)
1 + 4q 2 A2 cos2 ωt
The maximum and minimum of f (u) are
ε2
fmax (u) = , (7.47)
1 + 4q 2 A2

fmin (u) = ε2 , (7.48)

April 13, 2006 10:4 WSPC/140-IJMPB 03379

Some Asymptotic Methods for Strongly Nonlinear Equations 1193

respectively. So we have the following inequality

r s
ε2 ε2
<ω< . (7.49)
1 1 + 4q 2 A2
By He Chengtian’s method, we have
s s
(m + n)ε2 ε2
ω= 2 2
= . (7.50)
m + n(1 + 4q A ) 1 + 4kq 2 A2

The perturbation solution of Eq. (7.43) when ε  1 reads

ε
ωpert = . (7.51)
1 + 2q 2 A2
Comparison of Eq. (7.50) and (7.51) leads to the result: k = 1/2, so we obtain
ε
ω= p . (7.52)
1 + 2(qA)2
The accuracy reaches 10% even when qA → ∞.

8. Conclusions
We have reviewed a few new asymptotic techniques with numerous examples. All
reviewed methods can be applied to various kinds of nonlinear problems, and the
examples studied in the present review article can be used as paradigms for real-
life physics problems. For the nonlinear oscillators, all the reviewed methods yield
high accurate approximate periods, but the accuracy of the amplitudes cannot be
ameliorated by iteration, the discussion of the problem is available in Journal of
Sound and Vibration, Vol. 282, pages 1317–1320 (2005).

Acknowledgment
This work is supported by the Program for New Century Excellent Talents in
University of P.R. China and the William M.W. Wong Engineering Research Fund.

References
1. T. A. Abassy, M. A. El-Tawil and H. K. Saleh The solution of KdV and mKdV
equations using Adomian Pade approximation, Int. J. Nonl. Sci. Num. Simulation
5(4), 327 (2004).
2. S. Abbasbandy, Application of He’s homotopy perturbation method to functional
integral equations, Chaos Solitons Fractals (in press).
3. S. Abbasbandy, A numerical solution of Blasius equation by Adomian’s decomposi-
tion method and comparison with homotopy perturbation method, Chaos Solitons
Fractals (in press).
4. S. Abbasbandy, Application of He’s homotopy perturbation method for Laplace
transform, Chaos Solitons Fractals (in press).
April 13, 2006 10:4 WSPC/140-IJMPB 03379

1194 J.-H. He

5. H. A. Abdusalam, On an improved complex tanh-function method, Int. J. Nonl. Sci.

Num. Simulation 6(2), 99 (2005).
6. E. M. Abulwafa, M. A. Abdou and A. A. Mahmoud, The solution of nonlinear
coagulation problem with mass loss, Chaos Solitons Fractals 29(2), 313 (2006).
7. E. M. Abulwafa, M. A. Abdou and A. A. Mahmoud, Nonlinear fluid flows in pipe-
like domain problem using variational-iteration method, Chaos Solitons Fractals (in
press).
8. J. R. Acton and P. T. Squire Solving Equations with Physical Understanding (Adam
Hilger Ltd, Boston, 1985).
9. G. Adomian, A review of the decomposition method in applied mathematics, J.
Math. Anal. & Appl. 135, 501 (1988).
10. V. P. Agrwal and H. Denman, Weighted linearization technique for period approxi-
mation in large amplitude nonlinear oscillations, J. Sound Vib. 57, 463 (1985).
11. J. Awrejcewicz, I. V. Andrianov and L. I. Manevitch, Asymptotic Approaches in
Nonlinear Dynamics: New Trends and Applications (Springer-Verlag, Heidelberg,
1998).
12. A. Aziz and V. J. Lunardini, Perturbation techniques in phases change heat transfer,
Appl. Mech. Rev. 46(2), 29 (1993).
13. C. L. Bai and H. Zhao, Generalized extended tanh-function method and its applica-
tion, Chaos Solitons Fractals 27(4), 1026 (2006).
14. F. B. Bao and J. Z. Lin, Linear stability analysis for various forms of one-dimensional
burnett equations, Int. J. Nonl. Sci. Num. Simulation 6(3), 295 (2005).
15. R. Bellman, Perturbation Techniques in Mathematics, Physics and Engineering
(Holt, Rinehart and Winston, Inc., New York, 1964).
16. C. M. Bender, K. S. Pinsky and L. M. Simmons, A new perturbative approach to
nonlinear problems, J. Math. Phys. 30(5), 1447 (1989).
17. N. Bildik and A. Konuralp, The use of variational iteration method, differential
transform method and Adomian decomposition method for solving different types
of nonlinear partial differential equations, Int. J. Nonl. Sci. Num. Simulation 7(1),
65 (2006).
18. N. N. Bogoliubov and Yu. A. Mitropolsky, Asymptotic Method in the Theory of
Nonlinear Oscillations (Gordon and Breach, London, 1985).
19. J. R. Bogning, A. S. Tchakoutio-Nguetcho and T. C. Kofane, Gap solitons coexisting
with bright soliton in nonlinear fiber arrays, Int. J. Nonl. Sci. Num. Simulation 6(4),
371 (2005).
20. S. Bowong, R. Yamapi and P. Woafo, Adaptive synchronization of coupled self-
sustained electrical systems, Int. J. Nonl. Sci. Num. Simulation 6(4), 387 (2005).
21. S. Bowong A new adaptive chaos synchronization principle for a class of chaotic
systems, Int. J. Nonl. Sci. Num. Simulation 6(4), 399 (2005).
22. M. Braun, Differential equations and their application, an introduction to applied
mathematics, Applied Mathematical Sciences (3rd edn.), Vol. 15, (Springer-Verlag,
New York, 1983), pp. 120–122.
23. T. D. Burton, A perturbation method for certain nonlinear oscillators, Int. J. Nonl.
Mech. 19(5), 397 (1984).
24. T. D. Burton and Z. Rahman, On the multi-scale analysis of strongly nonlinear
forced oscillators, Int. J. Nonl. Mech. 21(2), 125 (1986).
25. X. C. Cai, W. Y. Wu and M. S. Li, Approximate period solution for a Kind of
nonlinear oscillator by He’s perturbation method, Int. J. Nonl. Sci. Num. Simulation
7(1), 109 (2006).
April 13, 2006 10:4 WSPC/140-IJMPB 03379

Some Asymptotic Methods for Strongly Nonlinear Equations 1195

26. H. S. Y. Chan, K. W. Chung and Z. Xu, A perturbation incremental method for

strongly nonlinear oscillators, Int. J. Nonl. Mech. 31, 59 (1996).
27. X. P. Chen, W. X. He and B. T. Jin, Symmetric boundary knot method for mem-
brane vibrations under mixed-type boundary conditions, Int. J. Nonl. Sci. Num.
Simulation 6(4), 421 (2005).
28. Y. K. Cheung, S. H. Chen and S. L. Lau, A modified Lindstedt–Poincare method
for certain strongly nonlinear oscillators, Int. J. Nonl. Mech. 26(3/4), 367 (1991).
29. L. Cveticanin, Homotopy-perturbation method for pure nonlinear differential equa-
tion, Chaos Solitons Fractals (in press).
30. M. D’Acunto, Determination of limit cycles for a modified van der Pol oscillator,
Mech. Res. Comm. 33(1), 93 (2006).
31. S. Q. Dai, G. F. Sigalov and A. V. Diogenov, Approximate analytical solutions for
some strongly nonlinear problems, Science in China (Series A) 33, 153 (1990).
32. S. Q. Dai On the generalized PLK method and its applications, Acta Mechanica
Sinica 6(2), (1990).
33. B. Delamotte, Nonperturbative method for solving differential equations and finding
limit cycles, Phys. Rev. Lett. 70, 3361 (1993).
34. J. W. Dauben, Ancient Chinese mathematics: the (sic) (Jiu Zhang Suan Shu) vs
Euclid’s Elements. Aspects of proof and the linguistic limits of knowledge, Int. J.
Engrg. Sci. 36, 1339 (1998).
35. T. S. El-Danaf, M. A. Ramadan and F. E. I. A. Alaal, The use of adomian de-
composition method for solving the regularized long-wave equation, Chaos Solitons
Fractals 26(3), 747 (2005).
36. M. F. El-Sabbagh and A. T. Ali, New exact solutions for (3+1)-dimensional
Kadomtsev-Petviashvili equation and generalized (2+1)-dimensional Boussinesq
equation, Int. J. Nonl. Sci. Num. Simulation 6(2), 151 (2005).
37. M. El-Shahed, Application of He’s Homotopy perturbation method to Volterra’s
integro-differential equation, Int. J. Nonl. Sci. Num. Simulation 6(2), 163 (2005).
38. B. A. Finlayson, The Method of Weighted Residual and Variational Principles
(Academic Press, New York, 1972).
39. Z. M. Ge, C. L. Hsiao and Y. S. Chen, Nonlinear dynamics and chaos control for a
time delay Duffing system, Int. J. Nonl. Sci. Num. Simulation 6(2), 187 (2005).
40. B. H. Gu and H. P. Zhang, Ballistic impact damage of laminated composite pene-
trated by cylindrical rigid projectile: a finite element simulation, Int. J. Nonl. Sci.
Num. Simulation 6(3), 215 (2005).
41. A. O. Govorov, Thomas-Fermi model and ferromagnetic phases of magnetic semi-
conductor quantum dots, Phy. Rev. B 72(7), Art. No. 075358 (2005).
42. S. A. Khuri, A new approach to Bratu’s problem, Appl. Math. Comput. 147(1), 131
(2004).
43. P. Hagedorn, Nonlinear Oscillations (translated by Wolfram Stadler) (Clarendon
Press, Oxford, 1981).
44. J. H. He, Homotopy perturbation method for solving boundary value problems, Phy.
Lett. A 350(1–2), 87 (2006).
45. J. H. He, Periodic solutions and bifurcations of delay-differential equations, Phy.
Lett. A 347(4–6), 228 (2005).
46. J. H. He Application of homotopy perturbation method to nonlinear wave equations,
Chaos Solitons Fractals 26(3), 695 (2005).
47. J. H. He, Limit cycle and bifurcation of nonlinear problems, Chaos Solitons Fractals
26(3), 827 (2005).
April 13, 2006 10:4 WSPC/140-IJMPB 03379

1196 J.-H. He

48. J. H. He, Homotopy perturbation method for bifurcation of nonlinear problems, Int.
J. Nonl. Sci. Num. Simulation, 6(2), 207 (2005).
49. J. H. He, Y. Q. Wan and Q. Guo An iteration formulation for normalized diode
characteristics, Int. J. Circ. Theor. Appl. 32(6), 629 (2004).
50. J. H. He, Asymptotology by homotopy perturbation method, Appl. Math. Comput.
156(3), 591 (2004).
51. J. H. He, Comparison of homotopy perturbation method and homotopy analysis
method, Appl. Math. Comput. 156(2), 527 (2004).
52. J. H. He, The homotopy perturbation method for nonlinear oscillators with discon-
tinuities, Appl. Math. Comput. 151(1), 287 (2004).
53. J. H. He, Solution of nonlinear equations by an ancient Chinese algorithm, Appl.
Math. Comput. 151(1), 293 (2004)
54. J. H. He, He Chengtian’s inequality and its applications, Appl. Math. Comput.
151(3), 887 (2004).
55. J. H. He, Variational principles for some nonlinear partial differential equations with
variable coefficients, Chaos Solitons Fractals 19(4), 847 (2004).
56. J. H. He, Variational approach to the Thomas-Fermi equation, Appl. Math. Comput.
143(2–3), 533 (2003).
57. J. H. He, Variational approach to the Lane-Emden equation, Appl. Math. Comput.
143(2–3), 539 (2003).
58. J. H. He, Determination of limit cycles for strongly nonlinear oscillators, Phys. Rev.
Lett. 90(17), Art. No. 174301 (2003).
59. J. H. He, Determination of limit cycles for strongly nonlinear oscillators, Phys. Rev.
Lett. 91(19), Art. No. 199902 (2003).
60. J. H. He, Linearized perturbation technique and its applications to strongly nonlinear
oscillators, Comput. Math. Appl. 45(1–3), 1 (2003).
61. J. H. He, A simple perturbation approach to Blasius equation, Appl. Math. Comput.
140(2–3), 217 (2003).
62. J. H. He, Homotopy perturbation method: a new nonlinear analytical technique,
Appl. Math. Comput. 135(1), 73 (2003).
63. J. H. He Preliminary report on the energy balance for nonlinear oscillations, Mech.
Res. Comm. 29(2–3), 107 (2002).
64. J. H. He, Bookkeeping parameter in perturbation methods, Int. J. Nonl. Sci. Num.
Simulation 2(3), 257 (2001).
65. J. H. He, Modified Lindsted-Poincare methods for some strongly nonlinear oscilla-
tions Part III: double series expansion, Int. J. Nonl. Sci. Num. Simulation 2(4), 317
(2001).
66. J. H. He, Modified Lindstedt–Poincare methods for some strongly non-linear oscil-
lations Part I: expansion of a constant, Int. J. Nonl. Mech. 37(2), 309 (2002).
67. J. H. He, Modified Lindstedt–Poincare methods for some strongly non-linear oscil-
lations Part II: a new transformation, Int. J. Nonl. Mech. 37(2), 315 (2002).
68. J. H. He, Iteration perturbation method for strongly nonlinear oscillations, J. Vib.
Cont. 7(5), 631 (2001).
69. J. H. He A modified perturbation technique depending upon an artificial parameter,
Meccanica 35(4), 299 (2000).
70. J. H. He, Variational iteration method for autonomous ordinary differential systems,
Appl. Math. Comput. 114(2–3), 115 (2000).
71. J. H. He, A review on some new recently developed nonlinear analytical techniques,
Int. J. Nonl. Sci. Num. Simulation 1(1), 51 (2000).
April 13, 2006 10:4 WSPC/140-IJMPB 03379

Some Asymptotic Methods for Strongly Nonlinear Equations 1197

72. J. H. He, Homotopy perturbation technique, Comput. Methods Appl. Mech. Engrg.
178, 257 (1999).
73. J. H. He, Some new approaches to Duffing equation with strongly and high or-
der nonlinearity (II) parametrized perturbation technique, Comm. Nonl. Sci. Num.
Simulation 4(1), 81 (1999).
74. J. H. He, Variational iteration method — a kind of non-linear analytical technique:
Some examples, Int. J. Nonl. Mech. 34(4), 699 (1999).
75. J. H. He, Approximate analytical solution for seepage flow with fractional derivatives
in porous media, Comput. Methods Appl. Mech. Engrg. 167(1–2), 57 (1998).
76. J. H. He, Approximate solution of nonlinear differential equations with convolution
product nonlinearities, Comput. Methods Appl. Mech. Engrg. 167(1–2), 69 (1998).
77. J. H. He and X. H. Wu, Construction of solitary solution and compacton-like solution
by variational iteration method, Chaos Solitons Fractals 29(1), 108 (2006).
78. L. Howarth, On the solution of the laminar boundary layer equation, Proc. R. Soc.
Lond. A 164, 547 (1938).
79. G. R. Itovich and J. L. Moiola, On period doubling bifurcations of cycles and the
harmonic balance method, Chaos Solitons Fractals 27(3), 647 (2006).
80. Z. Ji, Mathematics in the Northern-Southern, Sui and Tang Dynasties (Hebei Science
and Technology Publishing House, Shijiazhuang, 1999) (in Chinese).
81. Y. H. Kim, K. Kim and H. S. Kim, Effects of moment arms on spinal stability
subjected to follower loads: A finite element analysis study, Int. J. Nonl. Sci. Num.
Simulation 6(3), 311 (2005).
82. B. J. Laurenzi, An analytical solution to Thomas-Fermi equation, J. Math. Phys.
30(10), 2535 (1990).
83. W. K. Lee et al., Validity of the multiple-scale solution for a subharmonic resonance
response of a bar with a nonlinear boundary condition, J. Sound Vib. 208(4), 567
(1997).
84. S. J. Liao, A second-order approximate analytical solution of a simple pendulum by
the process analysis method, ASME J. Appl. Mech. 59, 970 (1992).
85. S. J. Liao and A. T. Chwang, Application of homotopy analysis method in nonlinear
oscillations, ASME J. Appl. Mech. 65, 914 (1998).
86. S. J. Liao, Beyond Perturbation (CRC Press, Boca Raton, 2003).
87. S. J. Liao, A uniformly valid analytic solution of 2-D viscous flow over a semi-infinite
flat plate, J. Fluid Mech. 385, 101 (1999).
88. G. L. Liu, New research directions in singular perturbation theory: artificial pa-
rameter approach and inverse-perturbation technique, National Conf. of 7th Modern
Mathematics and Mechanics, 1997, Shanghai, 47–53 (in Chinese).
89. H. M. Liu, Generalized variational principles for ion acoustic plasma waves by He’s
semi-inverse method, Chaos Solitons Fractals 23(2), 573 (2005).
90. H. M. Liu Variational approach to nonlinear electrochemical system, Int. J. Nonl.
Sci. Num. Simulation 5(1), 95 (2004).
91. H. M. Liu Approximate period of nonlinear oscillators with discontinuities by mod-
ified Lindstedt–Poincare method, Chaos Solitons Fractals 23(2), 577 (2005).
92. H. M. Liu and J. H. He, Variational approach to chemical reaction, Compt. Chem.
Engrg. 28(9), 1549 (2004).
93. Q. S. Lu, Special issue on advances of nonlinear dynamics, vibrations and control in
China, Int. J. Nonl. Sci. Num. Simulation 6(1), preface (2005).
94. R. E. Mickens, An Introduction to Nonlinear Oscillations (Cambridge University
Press, Cambridge, 1981).
April 13, 2006 10:4 WSPC/140-IJMPB 03379

1198 J.-H. He

95. R. E. Mickens, Comments on the Shohat expansion, J. Sound Vib. 193(3), 747
(1996).
96. S. Momani and S. Abuasad, Application of He’s variational iteration method to
Helmholtz equation, Chaos Solitons Fractals 27(5), 1119 (2006).
97. S. Momani and Z. Odibat, Numerical comparison of methods for solving linear dif-
ferential equations of fractional order, Chaos Solitons Fractals (in press).
98. A. H. Nayfeh, Introduction to Perturbation Methods (John Wiley, New York, 1981).
99. A. H. Nayfeh, Problems in Perturbation (John Wiley, New York, 1985).
100. A. H. Nayfeh and D. T. Mook, Nonlinear Oscillations (John Wiley, New York, 1979).
101. J. G. Ning and K. T. Wu, A novel space-time MmB difference scheme with second-
order accuracy for 3 dimensional flows, Int. J. Nonl. Sci. Num. Simulation 6(4), 417
(2005).
102. J. G. Ning, C. Wang and T. B. Ma, Numerical analysis of the shaped charged jet
with large cone angle, Int. J. Nonl. Sci. Num. Simulation 7(1), 71 (2006).
103. Z. M. Odibat and S. Momani Application of variational iteration method to nonlinear
differential equations of fractional order, Int. J. Nonl. Sci. Num. Simulation 7(1),
27 (2006).
104. S. Rajendran, S. N. Pandey and M. Lakshmanan, Comment on “Determina-
tion of limit cycles for strongly nonlinear oscillators”, Phys. Rev. Lett. 93(6),
Art. No. 069401 (2004).
105. J. I. Ramos, Piecewise-linearized methods for oscillators with limit cycles, Chaos
Solitons Fractals 27(5), 1229 (2006).
106. Y. J. Ren and H. Q. Zhang, A generalized F-expansion method to find abundant fam-
ilies of Jacobi Elliptic Function solutions of the (2+1)-dimensional Nizhnik–Novikov–
Veselov equation, Chaos Solitons Fractals 27(4), 959 (2006).
107. O. E. Rossler and R. Movassagh, Bitemporal dynamic Sinai divergence: an energetic
analog to Boltzmann’s entropy? Int. J. Nonl. Sci. Num. Simulation 6(4), 349 (2005).
108. J. Shohat, On Van der Pol’s and related nonlinear differential equations, J. Appl.
Phys. 15 568 (1944).
109. A. M. Siddiqui, R. Mahmood and Q. K. Ghori, Thin film flow of a third grade fluid
on a moving belt by He’s homotopy perturbation method, Int. J. Nonl. Sci. Num.
Simulation 7(1), 7 (2006).
110. A. M. Siddiqui, M. Ahmed and Q. K. Ghori, Couette and Poiseuille flows for Non-
Newtonian fluids, Int. J. Nonl. Sci. Num. Simulation 7(1), 15 (2006).
111. A. A. Soliman, A numerical simulation and explicit solutions of KdV-Burgers’ and
Lax’s seventh-order KdV equations, Chaos Solitons Fractals 29(2), 294 (2006).
112. N. H. Sweilam and M. M. Khader, Variational iteration method for one dimensional
nonlinear thermoelasticity, Chaos Solitons Fractals (in press).
113. F. Verhulst, Nonlinear Differential Equations and Dynamical Systems (Springer-
Verlag, New York, 1990).
114. Y. Q. Wan, Q. Guo and N. Pan Thermo-electro-hydrodynamic model for electro-
spinning process, Int. J. Nonl. Sci. Num. Simulation 5(1), 5 (2004).
115. D. S. Wang and H. Q. Zhang, Further improved F-expansion method and new exact
solutions of Konopelchenko–Dubrovsky equation, Chaos Solitons Fractals 25(3), 601
(2005).
116. M. L. Wang and X. Z. Li, Applications of F-expansion to periodic wave solutions for
a new Hamiltonian amplitude equation, Chaos Solitons Fractals 24(5), 1257 (2005).
117. Q. Wang, Y. Chen and H. Q. Zhang, A new Riccati equation rational expansion
method and its application to (2+1)-dimensional Burgers equation, Chaos Solitons
Fractals 25(5), 1019 (2005).
 April 13, 2006 10:4 WSPC/140-IJMPB 03379

Some Asymptotic Methods for Strongly Nonlinear Equations 1199

118. A. M. Wazwaz, A reliable modification of Adomian’s decomposition method, Appl.

Math. Comput. 92, 1 (1998).
119. A. M. Wazwaz Two reliable methods for solving variants of the KdV equation with
compact and noncompact structures, Chaos Solitons Fractals 28(2), 454 (2006).
120. T. M. Wu, The secant–homotopy continuation method, Chaos Solitons Fractals (in
press).
121. J. Zhang, J. Y. Yu and N. Pan, Variational principles for nonlinear fiber optics,
Chaos Solitons Fractals 24(1), 309 (2005).
122. C. R. Zhang, Y. G. Zu and B. D. Zheng, Stability and bifurcation of a discrete red
blood cell survival model, Chaos Solitons Fractals 28(2), 386 (2006).
123. S. W. Zhang, D. J. Tan and L. S. Chen, Dynamic complexities of a food chain model
with impulsive perturbations and Beddington–DeAngelis functional response, Chaos
Solitons Fractals 27(3), 768 (2006),
124. W. Zhang, M. H. Yao and X. P. Zhan, Multi-pulse chaotic motions of a rotor-active
magnetic bearing system with time-varying stiffness, Chaos Solitons Fractals 27(1),
175 (2006).
125. Z. D. Zhang and Q. S. Bi, Bifurcations of a generalized Camassa–Holm equation,
Int. J. Nonl. Sci. Num. Simulation 6(1), 81 (2005).

View publication stats