Applied Mathematics and Computation 149 (2004) 15–29

www.elsevier.com/locate/amc

Exact solutions for heat-like and

wave-like equations with variable coefficients
a,* b
Abdul-Majid Wazwaz , Alice Gorguis
a
Department of Mathematics, Saint Xavier University, Chicago, IL 60655, USA
b
Department of Mathematics, Northeastern Illinois University, Chicago, IL 60625, USA

Abstract
In this paper, Adomian decomposition method is presented for solving heat-like and
wave-like models with variable coefficients. The method is demonstrated for a variety of
problems in one and higher dimensional spaces where exact solutions are obtained. The
results obtained in all cases show the reliability and the efficiency of this method.
Ó 2003 Elsevier Inc. All rights reserved.

Keywords: Adomian decomposition method; Heat-like equations; Wave-like equations

1. Introduction

In various fields of science and engineering, many physical problems can be

described by initial boundary value problems (IBVP) with variable coefficients.
These linear and nonlinear models, as well as their analytic solutions, are of
fundamental importance for applied sciences. Moreover, these models were
treated numerically and analytically. Typical examples of linear models are
Euler–Darboux equation [1], LambropoubsÕ equation [2], and Tricomi equa-
tion [3] given by
ðx
yÞuxy þ ðaux
buy Þ ¼ 0; ð1Þ
uxy þ axux þ byuy þ cxyu þ ut ¼ 0; ð2Þ
uyy ¼ yuxx ; ð3Þ

*
Corresponding author.
E-mail address: [email protected] (A.-M. Wazwaz).

0096-3003/$ - see front matter Ó 2003 Elsevier Inc. All rights reserved.
doi:10.1016/S0096-3003(02)00946-3
16 A.-M. Wazwaz, A. Gorguis / Appl. Math. Comput. 149 (2004) 15–29

respectively. On the other hand, examples of nonlinear models are introduced

in KdV equation [4] of variable coefficients and ClairautÕs equation [5] given by

ut þ atn uux þ btm uxxx ¼ 0; ð4Þ

u ¼ xux þ yuy þ f ðux ; uy Þ; ð5Þ

respectively. For more details about these models and other physical models
with variable coefficients, the reader is referred to [6] and the references therein.
Several techniques, such as the spectral methods, perturbation techniques,
finite differences method and finite element method have been used to handle
these problems numerically and analytically.
The main goal of this work is the desire to obtain exact solutions to heat-like
and wave-like equations with variable coefficients. Our approach stems mainly
from Adomian decomposition method [7–16] that has been widely used with
promising results. The method is well known in the literature and it has been
used to handle a variety of physical and engineering models. It is interesting to
point out that the method of separation of variables is not always applicable to
these problems. The Adomian decomposition method provides the solution in
a rapidly convergent series with components that are elegantly computed. The
decomposition method provides a direct scheme for solving problems without
any need for linearization or any restrictive assumptions.
The main advantage of the method is that it can be applied directly for all
types of differential and integral equations, linear or nonlinear, homogeneous
or inhomogeneous, with constant coefficients or with variable coefficients.
Another important advantage is that the method is capable of greatly reducing
the size of computational work while still maintaining high accuracy of the
numerical solution. Convergence of Adomian decomposition scheme was es-
tablished in [11] where fixed point theorems were used. The effectiveness and
the usefulness of Adomian method are demonstrated by finding exact solutions
to the models that will be investigated.

2. Analysis

In this section we demonstrate the main algorithm of Adomian decompo-

sition method on heat-like and wave-like equations independently.

2.1. Heat-like equations

Without loss of generality, we consider a heat-like equation with variable

coefficients described by a three-dimensional IBVP of the form
 A.-M. Wazwaz, A. Gorguis / Appl. Math. Comput. 149 (2004) 15–29 17

ut ¼ f ðx; y; zÞuxx þ gðx; y; zÞuyy þ hðx; y; zÞuzz ;

0 < x < a; 0 < y < b; 0 < z < c; t > 0; ð6Þ

subject to the Neumann boundary conditions

ux ð0; y; z; tÞ ¼ f1 ðy; z; tÞ; ux ða; y; z; tÞ ¼ f2 ðy; z; tÞ;

uy ðx; 0; z; tÞ ¼ g1 ðx; z; tÞ; uy ðx; b; z; tÞ ¼ g2 ðx; z; tÞ; ð7Þ
uz ðx; y; 0; tÞ ¼ h1 ðx; y; tÞ; uz ðx; y; c; tÞ ¼ h2 ðx; y; tÞ;

and the initial condition

uðx; y; z; 0Þ ¼ /ðx; y; zÞ: ð8Þ

In an operator form, Eq. (6) can be written as

Lt u ¼ f ðx; y; zÞuxx þ gðx; y; zÞuyy þ hðx; y; zÞuzz ; ð9Þ

where the differential operator Lt is given by

o
Lt ¼ : ð10Þ
ot

Assuming Lt is invertible, therefore the inverse operator L
1

t is given by the
integral
Z t
L
1
t ðÞ ¼ ðÞ dt: ð11Þ
0

Applying the inverse operator L
1 t on Eq. (9), and using the initial condition
(8), we find


uðx; y; z; tÞ ¼ /ðx; y; zÞ þ L
1
t f ðx; y; zÞuxx þ gðx; y; zÞuyy þ hðx; y; zÞuzz :
ð12Þ

The details of the Adomian decomposition method are well known now in the
literature and can be found in [7–16]. The decomposition method consists of
representing the solution uðx; y; z; tÞ by the decomposition series
X
1
uðx; y; z; tÞ ¼ un ðx; y; z; tÞ; ð13Þ
n¼0

where the components un ðx; y; z; tÞ of the solution uðx; y; z; tÞ will be determined

in a recursive manner.
18 A.-M. Wazwaz, A. Gorguis / Appl. Math. Comput. 149 (2004) 15–29

Substituting (13) into both sides of (12) gives

0 ! !
X1 X
1 X
1
un ðx; y; zÞ ¼ /ðx; y; zÞ þ L
1 @f un þg un
t
n¼0 n¼0 xx n¼0 yy
! 1
X
1
þh un A : ð14Þ
n¼0 zz

The decomposition method identifies the zeroth component u0 by all terms that
arise from the initial condition and from integrating the source term if exists.
To determine the components un , n P 0 of the solution u, Adomian decom-
position method formally introduces the use the recursive relation
u0 ðx; y; z; tÞ ¼ /ðx; y; zÞ;
  ð15Þ
ukþ1 ðx; y; z; tÞ ¼ L
1
t f ðu k Þ xx þ gðu k Þ yy þ hðu k Þ zz ; k P 0:

Consequently,
P1 we can recurrently determine every component of the series
n¼0 un . Accordingly, the series solution of u is readily obtained. However, it
was found that the series solution converges rapidly to the exact solution if
such a solution exists. It is interesting to note that the series solution was
obtained by implementing the initial condition only. The prescribed boundary
conditions can be used to justify the obtained solution.

2.2. Wave-like equations

Without loss of generality, we consider a wave-like equation with variable

coefficients obeying a three-dimensional IBVP of the form
utt ¼ f ðx; y; zÞuxx þ gðx; y; zÞuyy þ hðx; y; zÞuzz ;
0 < x < a; 0 < y < b; 0 < z < c; t > 0; ð16Þ

subject to the Neumann boundary conditions

ux ð0; y; z; tÞ ¼ f1 ðy; z; tÞ; ux ða; y; z; tÞ ¼ f2 ðy; z; tÞ;
uy ðx; 0; z; tÞ ¼ g1 ðx; z; tÞ; uy ðx; b; z; tÞ ¼ g2 ðx; z; tÞ; ð17Þ
uz ðx; y; 0; tÞ ¼ h1 ðx; y; tÞ; uz ðx; y; c; tÞ ¼ h2 ðx; y; tÞ;

and the initial conditions

uðx; y; z; 0Þ ¼ wðx; y; zÞ; ut ðx; y; z; 0Þ ¼ hðx; y; zÞ: ð18Þ
Eq. (16) can be written as
e
L t u ¼ f ðx; y; zÞuxx þ gðx; y; zÞuyy þ hðx; y; zÞuzz ; ð19Þ
 A.-M. Wazwaz, A. Gorguis / Appl. Math. Comput. 149 (2004) 15–29 19

where e
L t is given by

e o2
Lt ¼ 2 : ð20Þ
ot
Assuming e L t is invertible, hence the inverse operator e
L
1
t is given given by the
integral
Z tZ t
e
L
1 ðÞ ¼ ðÞ dt dt: ð21Þ
t
0 0

Operating with the inverse operator e L
1

t on both sides of Eq. (19), and using
the initial conditions (18), we obtain


uðx; y; z; tÞ ¼ wðx; y; zÞ þ thðx; y; zÞ þ e
L
1
t f ðx; y; zÞuxx þ gðx; y; zÞuyy

þ hðx; y; zÞuzz : ð22Þ

Adomian method defines the solution uðx; y; z; tÞ by the decomposition series

X
1
uðx; y; z; tÞ ¼ un ðx; y; z; tÞ; ð23Þ
n¼0

where the components un ðx; y; z; tÞ of the solution uðx; y; z; tÞ will be determined

recurrently.
Substituting (23) into both sides of (22) yields
0 !
X
1 X
1
e @f
un ðx; y; z; tÞ ¼ wðx; y; zÞ þ thðx; y; zÞ þ L
1
un
t
n¼0 n¼0 xx
! ! 1
X
1 X
1
þg un þh un A : ð24Þ
n¼0 yy n¼0 zz

To determine the components un , n P 0 recurrently, Adomian decomposition

method admits the use of the recursive relation

u0 ðx; y; z; tÞ ¼ wðx; y; zÞ þ thðx; y; zÞ;

  ð25Þ
ukþ1 ðx; y; z; tÞ ¼ e
L
1
t f ðuk Þxx þ gðuk Þyy þ hðuk Þzz ; k P 0:

With this relation, the components un , n P 0 are readily obtained. This in turn
leads to the solution in a series form. However the solution in a closed form
follows immediately if an exact solution exists. Notice that the series solution
was obtained by implementing the initial conditions only, where the given
boundary conditions can be used for justification only.
20 A.-M. Wazwaz, A. Gorguis / Appl. Math. Comput. 149 (2004) 15–29

3. Illustration of the analysis

In this section, three distinct models from each type will be tested by using
the proposed method.

3.1. Heat-like models

Example 1. We first consider the one-dimensional IBVP

1
ut ¼ x2 uxx ; 0 < x < 1; t > 0; ð26Þ
2

subject to the boundary conditions

uð0; tÞ ¼ 0; uð1; tÞ ¼ et ; ð27Þ

and the initial condition

uðx; 0Þ ¼ x2 : ð28Þ

In an operator form, Eq. (26) becomes

1
Lt u ¼ x2 uxx ; 0 < x < 1: ð29Þ
2

Operating with L
1

t on (29) and using the initial condition yields
 
1 2
2
uðx; tÞ ¼ x þ L
1
t x uxx ; ð30Þ
2

where L
1
t is defined above in (11).
Substituting the decomposition series (13) for u into (30) gives
! !
X
1
1 2 X1
2
un ðx; tÞ ¼ x þ L
1
t x un ðx; tÞ : ð31Þ
n¼0
2 n¼0 xx

We next use the recursive relation

u0 ðx; tÞ ¼ x2 ;
 
1 2 ð32Þ
ukþ1 ðx; tÞ ¼ L
1
t x ðuk Þxx ; k P 0:
2
 A.-M. Wazwaz, A. Gorguis / Appl. Math. Comput. 149 (2004) 15–29 21

This in turn gives the first few components

u0 ðx; tÞ ¼ x2 ;
 
1
1 2
u1 ðx; tÞ ¼ L
1
t L x ðu0 Þxx ¼ x2 t;
2
 
1
1 2 t2
u2 ðx; tÞ ¼ L
1
t L x ðu Þ
1 xx ¼ x2 ; ð33Þ
2 2!
 
1 t3
u3 ðx; tÞ ¼ L
1
t L
1 x2 ðu2 Þxx ¼ x2 ;
2 3!
..
.
The solution in a series form is given by
 
2 t2 t3
uðx; tÞ ¼ x 1 þ t þ þ þ    ; ð34Þ
2! 3!
and in a closed form by
uðx; tÞ ¼ x2 et : ð35Þ

Example 2. We next consider the two-dimensional IBVP

1
ut ¼ ðy 2 uxx þ x2 uyy Þ; 0 < x; y < 1; t > 0; ð36Þ
2
subject to the Neumann boundary conditions
ux ð0; y; tÞ ¼ 0; ux ð1; y; tÞ ¼ 2 sinh t;
ð37Þ
uy ðx; 0; tÞ ¼ 0; uy ðx; 1; tÞ ¼ 2 cosh t;
and the initial condition
uðx; y; 0Þ ¼ y 2 : ð38Þ
Proceeding as in Example 1 we write
1
Lt u ¼ ðy 2 uxx þ x2 uyy Þ: ð39Þ
2
Applying the inverse operator L
1
t on both sides of (39) gives
1
2 
uðx; y; tÞ ¼ y 2 þ L
1 y uxx þ x2 uyy : ð40Þ
2 t
Using the decomposition series for uðx; y; tÞ leads to
0 ! ! 1
X1
1 X1 X 1
un ¼ y 2 þ L
1 @y 2 un þ x2 un A : ð41Þ
t
n¼0
2 n¼0 n¼0
xx yy
22 A.-M. Wazwaz, A. Gorguis / Appl. Math. Comput. 149 (2004) 15–29

Following Adomian analysis gives the recursive relation

u0 ðx; y; tÞ ¼ y 2 ;
1 ð42Þ
ukþ1 ðx; y; tÞ ¼ L
1 ðy 2 ðuk Þxx þ x2 ðuk Þyy Þ; k P 0;
2 t
so that the first few components by
u0 ðx; y; tÞ ¼ y 2 ;
1
u1 ðx; y; tÞ ¼ L
1 ðy 2 ðu0 Þxx þ x2 ðu0 Þyy Þ ¼ x2 t;
2 t
2
1 2 t
u2 ðx; y; tÞ ¼ L
1 2 2
t ðy ðu1 Þxx þ x ðu1 Þyy Þ ¼ y ; ð43Þ
2 2!
3
1 2 t
u3 ðx; y; tÞ ¼ L
1 2 2
t ðy ðu2 Þxx þ x ðu2 Þyy Þ ¼ x ;
2 3!
4
1 2 t
u4 ðx; y; tÞ ¼ L
1 2 2
t ðy ðu3 Þxx þ x ðu3 Þyy Þ ¼ y :
2 4!
Hence, the solution in a series form is
   
t3 t2 t4
uðx; y; tÞ ¼ x2 t þ þ    þ y 2 1 þ þ þ    ; ð44Þ
3! 2! 4!
and therefore the exact solution
uðx; y; tÞ ¼ x2 sinh t þ y 2 cosh t; ð45Þ
is readily obtained.

Example 3. Consider the three-dimensional inhomogeneous IBVP

1 2
ut ¼ x 4 y 4 z 4 þ ðx uxx þ y 2 uyy þ z2 uzz Þ; 0 < x; y; z < 1; t > 0; ð46Þ
36
subject to the boundary conditions
uð0; y; z; tÞ ¼ 0; uð1; y; z; tÞ ¼ y 4 z4 ðet
1Þ;
uðx; 0; z; tÞ ¼ 0; uðx; 1; z; tÞ ¼ x4 z4 ðet
1Þ; ð47Þ
4 4 t
uðx; y; 0; tÞ ¼ 0; uðx; y; 1; tÞ ¼ x y ðe
1Þ;
and the initial condition
uðx; y; z; 0Þ ¼ 0: ð48Þ
In an operator form, Eq. (46) becomes
1 2
Lt u ¼ x4 y 4 z4 þ ðx uxx þ y 2 uyy þ z2 uzz Þ: ð49Þ
36
 A.-M. Wazwaz, A. Gorguis / Appl. Math. Comput. 149 (2004) 15–29 23

Operating with L
1

t on (49) yields
1
1
2 
uðx; y; z; tÞ ¼ x4 y 4 z4 t þ L x uxx þ y 2 uyy þ z2 uzz : ð50Þ
36 t
Using the decomposition series (8) for the solution uðx; y; z; tÞ we find
0 ! ! ! 1
X1
1 X1 X1 X 1
un ¼ x 4 y 4 z 4 t þ @ x 2 un þ y2 un þ z2 un A :
n¼0
36 n¼0 n¼0 n¼0
xx yy zz

ð51Þ

Accordingly, we set the recursive relation

u0 ðx; y; z; tÞ ¼ x4 y 4 z4 t;
0 ! !
1
1 @ 2 X
1 X
1
2
ukþ1 ðx; y; z; tÞ ¼ Lt x uk þy uk
36 n¼0 n¼0 ð52Þ
xx yy
! !
X
1
þ z2 uk ; k P 0:
n¼0 zz

The first few components of uðx; y; z; tÞ

u0 ðx; y; z; tÞ ¼ x4 y 4 z4 t;
1   t2
u1 ðx; y; z; tÞ ¼ L
1 t x2 ðu0 Þxx þ y 2 ðu0 Þyy þ z2 ðu0 Þzz ¼ x4 y 4 z4 ;
36 2!
1
1  2  t3
u2 ðx; y; z; tÞ ¼ Lt x ðu1 Þxx þ y 2 ðu1 Þyy þ z2 ðu1 Þzz ¼ x4 y 4 z4 ; ð53Þ
36 3!
1
1  2 2 2

4 4 4 t
4
u3 ðx; y; z; tÞ ¼ Lt x ðu2 Þxx þ y ðu2 Þyy þ z ðu2 Þzz ¼ x y z ;
36 4!
..
.

follow immediately. Upon summing these iterates, we find that the solution in a
series form is
 
t2 t3 t4
uðx; y; z; tÞ ¼ x4 y 4 z4 t þ þ þ þ    : ð54Þ
2! 3! 4!

The exact solution

uðx; y; z; tÞ ¼ x4 y 4 z4 ðet
1Þ; ð55Þ

follows immediately.
24 A.-M. Wazwaz, A. Gorguis / Appl. Math. Comput. 149 (2004) 15–29

3.2. Wave-like models

In what follows we illustrate our analysis by examining the following three

wave-like equations.

Example 1. We first consider the one-dimensional IBVP

1
utt ¼ x2 uxx ; 0 < x < 1; t > 0; ð56Þ
2

subject to the boundary conditions

uð0; tÞ ¼ 0; uð1; tÞ ¼ 1 þ sinh t; ð57Þ

and the initial conditions

uðx; 0Þ ¼ x; ut ðx; 0Þ ¼ x2 : ð58Þ

Eq. (56) can be written as

1
Lt u ¼ x2 uxx ; 0 < x < 1; ð59Þ
2

where Lt , a second differential operator, is assumed invertible. As a result, the

inverse operator L
1
t exists and given by the two fold integral operator
Z t Z t
L
1
t ðÞ ¼ ðÞ dt dt: ð60Þ
0 0

Applying L
1
t on (59) and using the initial conditions we obtain
1
2 
uðx; tÞ ¼ x þ x2 t þ L
1
t x uxx ; ð61Þ
2
where L
1
t is defined above in (60).
Substituting the decomposition series (13) for u into (61) gives
! !
X1
1
1 2 X 1
2
un ðx; tÞ ¼ x þ x t þ Lt x un ðx; tÞ : ð62Þ
n¼0
2 n¼0 xx

We next use the recursive relation

u0 ðx; tÞ ¼ x þ x2 t;
1
2  ð63Þ
ukþ1 ðx; tÞ ¼ L
1t x ðuk Þxx ; k P 0:
2
 A.-M. Wazwaz, A. Gorguis / Appl. Math. Comput. 149 (2004) 15–29 25

With this relation we obtain the first few components

u0 ðx; tÞ ¼ x þ x2 t;
1
2  t3
u1 ðx; tÞ ¼ L
1t x ðu0 Þxx ¼ x2 ;
2 3!
1
1
2  t5
u2 ðx; tÞ ¼ Lt x ðu1 Þxx ¼ x2 ; ð64Þ
2 5!
1
1
2  t7
u3 ðx; tÞ ¼ Lt x ðu2 Þxx ¼ x2 ;
2 7!
..
.
The solution in a series form is given by
 
t3 t5
uðx; tÞ ¼ x þ x2 t þ þ þ    ; ð65Þ
3! 5!
so that the exact solution is
uðx; tÞ ¼ x þ x2 sinh t: ð66Þ

Example 2. We next consider the two-dimensional IBVP

1 2
utt ¼ ðx uxx þ y 2 uyy Þ; 0 < x; y < 1; t > 0; ð67Þ
12
subject to the Neumann boundary conditions
ux ð0; y; tÞ ¼ 0; ux ð1; y; tÞ ¼ 4 cosh t;
ð68Þ
uy ðx; 0; tÞ ¼ 0; uy ðx; 1; tÞ ¼ 4 sinh t;
and the initial conditions
uðx; y; 0Þ ¼ x4 ; ut ðx; 0Þ ¼ y 4 : ð69Þ
Proceeding as before we write
1 2
Lt u ¼ ðx uxx þ y 2 uyy Þ: ð70Þ
12
Applying L
1
t on both sides of (70) gives
1
1
2 
uðx; y; tÞ ¼ x4 þ y 4 t þ Lt x uxx þ y 2 uyy : ð71Þ
12
Using the decomposition series for uðx; y; tÞ gives
0 ! ! 1
X1
1 X1 X
1

1 @ 2
4 4
un ¼ x þ y t þ Lt x un þ y2 un A : ð72Þ
n¼0
12 n¼0 n¼0
xx yy
26 A.-M. Wazwaz, A. Gorguis / Appl. Math. Comput. 149 (2004) 15–29

This gives the recursive relation

u0 ðx; y; tÞ ¼ x4 þ y 4 t;
1 ð73Þ
ukþ1 ðx; y; tÞ ¼ L
1 ðx2 ðuk Þxx þ y 2 ðuk Þyy Þ; k P 0:
12 t
The first few components of uðx; y; tÞ are given by
u0 ðx; y; tÞ ¼ x4 þ y 4 t;
1
1 2 t2 t3
u1 ðx; y; tÞ ¼ Lt ðx ðu0 Þxx þ y 2 ðu0 Þyy Þ ¼ x4 þ y 4 ;
12 2! 3!
1
1 2 t4 t5 ð74Þ
u2 ðx; y; tÞ ¼ Lt ðx ðu1 Þxx þ y 2 ðu1 Þyy Þ ¼ x4 þ y 4 ;
12 4! 5!
1
1 2 t6 t7
u3 ðx; y; tÞ ¼ Lt ðx ðu2 Þxx þ y 2 ðu2 Þyy Þ ¼ x4 þ y 4 :
12 6! 7!
The solution in a series form is
   
t2 t4 t3 t5
uðx; y; tÞ ¼ x4 1 þ þ þ    þ y 4 t þ þ þ    ; ð75Þ
2! 4! 3! 5!
so that the exact solution is given by
uðx; y; tÞ ¼ x4 cosh t þ y 4 sinh t ð76Þ
is readily obtained.

Example 3. Consider the three-dimensional inhomogeneous IBVP

1
2 
utt ¼ ðx2 þ y 2 þ z2 Þ þ x uxx þ y 2 uyy þ z2 uzz ; 0 < x; y; z < 1; t > 0;
2
ð77Þ
subject to the boundary conditions
uð0; y; z; tÞ ¼ y 2 ðet
1Þ þ z2 ðe
t
1Þ;
uð1; y; z; tÞ ¼ ð1 þ y 2 Þðet
1Þ þ z2 ðe
t
1Þ;
uðx; 0; z; tÞ ¼ x2 ðet
1Þ þ z2 ðe
t
1Þ;
uðx; 1; z; tÞ ¼ ð1 þ x2 Þðet
1Þ þ z2 ðe
t
1Þ;
uðx; y; 0; tÞ ¼ ðx2 þ y 2 Þðet
1Þ; uðx; y; 1; tÞ ¼ ðx2 þ y 2 Þðet
1Þ þ ðe
t
1Þ;
ð78Þ
and the initial conditions
uðx; y; z; 0Þ ¼ 0; ut ðx; y; z; 0Þ ¼ x2 þ y 2
z2 : ð79Þ
 A.-M. Wazwaz, A. Gorguis / Appl. Math. Comput. 149 (2004) 15–29 27

In an operator form, Eq. (77) becomes

1
2 
Lt u ¼ ðx2 þ y 2 þ z2 Þ þ x uxx þ y 2 uyy þ z2 uzz : ð80Þ
2
Operating with L
1 t on (80) yields
   
2 2 t2 2 t2
uðx; y; z; tÞ ¼ ðx þ y Þ t þ þz
tþ
2! 2!
1
2 
þ L
1 x uxx þ y 2 uyy þ z2 uzz : ð81Þ
2 t
Using the decomposition series (13) for uðx; y; z; tÞ we find
0 !
X1  2
  2
 X
1
t t 1 @x 2
un ¼ ðx2 þ y 2 Þ t þ þ z2
t þ þ L
1 un
n¼0
2! 2! 2 t n¼0 xx
! ! 1
X 1 X1
þ y2 un þ z2 un A : ð82Þ
n¼0 yy n¼0 zz

This leads to
   
2 2 t2 2 t2
u0 ðx; y; z; tÞ ¼ ðx þ y Þ t þ þz
tþ ;
0 2! 2!
! !
1
1 @ 2 X 1
2
X1
ukþ1 ðx; y; z; tÞ ¼ Lt x uk þy uk ð83Þ
2 n¼0 n¼0
! ! xx yy
X1
þ z2 uk ; k P 0:
n¼0 zz

The first few components of uðx; y; z; tÞ

   
t2 t2
u0 ðx; y; z; tÞ ¼ ðx2 þ y 2 Þ t þ þ z2
t þ ;
2! 2!
1  
u1 ðx; y; z; tÞ ¼ L
1 x 2
ðu 0 Þ þ y 2
ðu 0 Þ þ ðu 0 Þ zz ;
2 t  3
xx

yy
 
2 2 t t4 2 t3 t4
¼ ðx þ y Þ þ þz
þ ;
3! 4! 3! 4! ð84Þ
1  
u2 ðx; y; z; tÞ ¼ L
1 t x2 ðu1 Þxx þ y 2 ðu1 Þyy þ ðu1 Þzz ;
2  5   
2 2 t t6 2 t5 t6
¼ ðx þ y Þ þ þz
þ ;
5! 6! 5! 6!
..
.
28 A.-M. Wazwaz, A. Gorguis / Appl. Math. Comput. 149 (2004) 15–29

follow immediately. Upon summing these iterates, we find that the solution in a
series form is
 
t2 t3 t4
uðx; y; z; tÞ ¼ ðx2 þ y 2 Þ t þ þ þ þ   
2! 3! 4!
 2

t t3 t4
þ z2
t þ
þ þ    ; ð85Þ
2! 3! 4!
and therefore the exact solution
uðx; y; z; tÞ ¼ ðx2 þ y 2 Þet þ z2 e
t
ðx2 þ y 2 þ z2 Þ; ð86Þ
follows immediately.

4. Discussion

The main goal of this work was to propose a reliable method for solving
heat-like and wave-like equations with variable coefficients. The proposed
equations may not be solved by the method of separation of variables. The
Adomian decomposition method has worked effectively to handle these mod-
els, and this gives it a wider applicability.
The proposed scheme was applied directly without any need for transfor-
mation formulae or restrictive assumptions. The decomposition method is
capable of greatly reducing the volume of computational work compared to
standard methods while still maintaining high accuracy of the numerical so-
lution. This size reduction amounts to the improvement of performance of this
approach.
The approach was tested by employing the method to obtain exact solutions
for six numerical examples, three from each type. The results obtained in all
cases demonstrate the reliability and the efficiency of this method.

References

[1] W. Miller Jr., Symmetries of differential equations: The hypergeometric and Euler–Darboux
equations, SIAM J. Math. Anal. 4 (2) (1973) 314–328.
[2] R. Wilcox, Closed form solution of the differential equation uxy þ axux þ byuy þ cxyu þ ut ¼ 0
by normal-ordering exponential operators, J. Math. Phys. 11 (4) (1970) 1235–1237.
[3] A.R. Manwell, The Tricomi Equations with Applications to the Theory of Plane Transonic
Flow, Pitman, 1979.
[4] N. Nimala, M.J. Vedan, B.V. Baby, A variable coefficient Kortweg–de Vries equation:
Similarity analysis and exact solution, J. Math. Phys. 27 (11) (1986) 2644–2646.
[5] S. Iyanaga, Y. Kawada, Encyclopedia Dictionary of Mathematics, MIT, 1962.
[6] D. Zwillinger, Handbook of Differential Equations, Wiley, New York, 1992.
[7] G. Adomian, Solving Frontier Problems of Physics: The Decomposition Method, Kluwer,
Boston, 1994.
 A.-M. Wazwaz, A. Gorguis / Appl. Math. Comput. 149 (2004) 15–29 29

[8] G. Adomian, A review of the decomposition method in applied mathematics, J. Math. Anal.
Appl. 135 (1988) 501–544.
[9] G. Adomian, On the solution of partial differential equations with specified boundary
conditions, Math. Comput. Modell. 140 (2) (1994) 569–581.
[10] G. Adomian, R. Rach, Noise terms in decomposition solution series, Comput. Math. Appl. 24
(11) (1992) 61–64.
[11] Y. Cherrauault, G. Saccomandi, B. Some, New results for convergence of AdomianÕs method
applied to integral equations, Math. Comput. Modell. 16 (2) (1992) 85–93.
[12] A.M. Wazwaz, Partial Differential Equations: Methods and Applications, Balkema Publishers,
The Netherlands, 2002.
[13] A.M. Wazwaz, A First Course in Integral Equations, World Scientific, New Jersey, 1997.
[14] A.M. Wazwaz, Analytical approximations and Pad
e approximants for VolterraÕs population
model, Appl. Math. Comput. 100 (1999) 13–25.
[15] A.M. Wazwaz, The modified decomposition method and Pad
e approximants for solving the
Thomas–Fermi equation, Appl. Math. Comput. 105 (1999) 11–19.
[16] A.M. Wazwaz, On the solution of the fourth order parabolic equation by the decomposition
method, Int. J. Comput. Math. 57 (1995) 213–217.