Ashenafi Agizaw

ANALYTICAL AND NUMERICAL TREATMENT FOR SINGULAR
TWO-POINT BOUNDARY VALUE PROBLEMS VIA ADOMIAN AND

THE IMPROVED ADOMIAN DECOMPOSITION METHODS
MSc. PROJECT
ASHENAFI AGIZAW
OCTOBER 2015
HARAMAYA UNIVERSITY, HARAMAYA
Analytical and Numerical Treatment for Singular Two-Point Boundary
Value Problems via Adomian and the Improved Adomian Decomposition
Methods
A Project submitted to the Department of Mathematics,

Postgraduate Program Directorate
HARAMAYA UNIVERSITY
In Partial Fulfillment of the Requirements for the Degree of

MASTER OF SCIENCE IN MATHEMATICS
(NUMERICAL ANALYSIS)
Ashenafi Agizaw
October 2015
Haramaya University, Haramaya
HARAMAYA UNIVERSITY
Postgraduate Program Directorate
I hereby certify that I have read and evaluated this Project titled Analytical and numerical
treatment for singular two-point boundary value problems via Adomian and the improved
Adomian decomposition methods prepared under my guidance by Ashenafi Agizaw. I
recommend that it be submitted as fulfilling the project requirement.
Seleshi Demie (PhD) __________________ ___________________

Advisor Signature Date
As member of the board of Examiners of the MSc Project Open Defense Examination, we
certify that we have read and evaluated the Project prepared by Ashenafi Agizaw and
examined the candidate. We recommend that the Project be accepted as fulfilling the Project
requirement for the degree of Master of Science in Mathematics (Numerical Analysis).
________________ ________________
Chairperson Signature Date
________________ ________________ ________________

Internal Examiner Signature Date
________________ ________________ ________________

External Examiner Signature Date
ii
DEDICATION
I dedicate this manuscript to my cherished father Ato Agizaw Jije who passed away in 2010.
iii
STATEMENT OF THE AUTHOR
By my signature below, I declare that this Project is my own work. I have followed all ethical
and technical principles of scholarship in the preparation, and compilation of this Project. Any
scholarly matter that is included in the Project has been given recognition through citation.
This Project is submitted in partial fulfillment of the requirements for a Numerical Analysis
degree at the Haramaya University. The Project is deposited in the Haramaya University
Library and is made available to borrowers under the rules of the Library. I solemnly declare
that this Project has not been submitted to any other institution anywhere for the award of any
academic degree, diploma or certificate.
Brief quotations from this Project may be made without special permission provided that
accurate and complete acknowledgement of the source is made. Requests for permission for
extended quotations from or reproduction of this Project in whole or in part may be granted by
the Head of the School or Department when in his or her judgment the proper use of the
material is in the interest of scholarship. In all other instances, however, permission must be
obtained from the author of the Project.
Name: Ashenafi Agizaw Jije
Date 04/11/2015
Department: Mathematics
Signature
iv
ACRONYMS, ABBREVIATIONS AND NOTATIONS
Acronyms and Abbreviations

ARD Absolute Relative Deviation
ADM Adomian Decomposition Method
ADM-PT Adomian Decomposition Method-Pade Technique
CPU Central Processing Unit
DTM Differential Transformation Method
FDM Finite Difference Method
G-FEM Galerkin-Finite Element Method
GTEHSC Gambella Teachers Education and Health Science College
HAM Homotopy Analysis Method
IADM Improved Adomian Decomposition Method
MADM Modified Adomian Decomposition Method
MVIM Modified Variational Iteration Method
NA Nonlinearity A
NB Nonlinearity B
NC Nonlinearity C
PAT Pade Approximant Technique
STPBVPs Singular Two-Point Boundary Value Problems
Notations
ADM Solution
Exact local temperature
IADM Solution
Local temperature by MADM
MADM solution
Temperature determined by MADM
v
BIOGRAPHICAL SKETCH
The author was born in (G.C) on August in SNNP Regional State, Gurage Zone,
Agena wereda, Darcha kebele.
He attended his primary education at Darcha and Yewegerawe primary schools. Then after, he
joined Addis Ketema Comprehensive Secondary School found at the capital of the country to
attend his secondary education. He then joined Jimma University in and received
Bachelor of Education degree in Mathematics in July, .
He has been employed in Gambella Regional State, Gambella town as a High school teacher
in ; and working for more than three and half years in the school he then joined Gambella
Teachers Education and Health Science College (GTEHSC) in May, as a Mathematics
Instructor.
Working for more than three and half years in GTEHSC, he joined Postgraduate Program at
Haramaya University, College of Natural and Computational Sciences, Department of
Mathematics in to pursue a program of study for M.Sc. degree in Numerical Analysis.
vi
ACKNOWLEDGEMENT
I am deeply beholden to my advisor Dr. Seleshi Demie next to Almighty God for his profound
commitment and encouragements he has shown throughout my study period. I pay respect and
express indebtedness to him because of his inspiring guidance, consistent supervision and
especially, his suggestions in selection of tools from the start of the proposal and at every
phase of this project work.
My gratitude goes to Gambella Regional State Ethics and Anti-Corruption Commission for
their invaluable commitment to give the right justice for problems happened by the preceding
scholarship committee in GTEHSC during competition for postgraduate studies. It was
unlikely to be here without the involvement of the commission.
I would like to extend my gratitude to GTEHSC for the provision of the necessary support to
complete the postgraduate study at Haramaya University.
Likewise, I wish to thank all the authors whose works played an important role in the success
of this project work.
vii
TABLE OF CONTENTS
STATEMENT OF THE AUTHOR IV

ACRONYMS, ABBREVIATIONS AND NOTATIONS V
BIOGRAPHICAL SKETCH VI
ACKNOWLEDGEMENT VII
LIST OF TABLES X
LIST OF FIGURES XI
ABSTRACT XII
1. INTRODUCTION 1
1.1. Background of the study 1
1.2. Statement of the problem 5
1.3. Objectives 6
2. LITERATURE REVIEW 7
2.1. Extended surfaces 7
2.2. Numerical and Analytical treatments of the BVPs 7
2.3. Treatments of related problems using ADM and its modifications 9
3. MATERIALS AND METHODS 12

4. PRELIMINARIES 13
4.1. Singular points 13
4.2. Boundary value problems 16
4.3. Operators 18
5. SINGULAR TWO-POINT BVPS VIA ADM, IADM AND MADM 21

5.1. General Description of the Adomian Decomposition Method 21
5.1.1. Advantages and disadvantages of the ADM 30
5.1.2. MADM 31
5.1.3. IADM 35
5.2. Numerical and Analytical Illustrations 38
viii
TABLE OF CONTENTS (continued)
5.2.1. Application 38
5.2.1.1. Constant thermal conductivity 41
5.2.1.2. Power law dependence of thermal conductivity 45
5.2.2. Numerical and Analytical Treatments 47
6. SUMMARY, CONCLUSION AND RECOMMENDATION 80

6.1. Summary 80
6.2. Conclusion 81
6.3. Recommendation 82
7. REFERENCES 83
8. APPENDICES 86
8.1. Appendix A 86
8.2. Appendix B 86
8.3. Appendix C 89
ix
LIST OF TABLES
Table Page
5.1. Numerical results for example 5.3 51
x
LIST OF FIGURES
Figure Page
5.1. A wall transferring heat by convection 39

5.2. Triangular and rectangular fin profiles 40
5.3. The geometry of a straight fin of triangular profile 43
5.4. Temperature distribution along the length of the silicon fin 48
5.5. Comparison of the IADM and exact solutions for example 5.3 51
5.6. Comparison of the IADM and MADM solutions for example 5.4 56
5.7. Comparison of the IADM and MADM solutions for example 5.5 61
5.8. Comparison of the MADM and exact solutions for example 5.6 65
5.9. IADM solutions for Thomas-Fermi equation in example 5.7 69
5.10. MADM solution for example 5.8 72
5.11. Comparison of the IMADM and exact solutions for example 5.9 78
xi
Analytical and Numerical treatment for Singular Two-Point Boundary
Value Problems via Adomian and the Improved Adomian Decomposition
Methods
ABSTRACT
Singular two-point boundary value problems arise in different areas of applied sciences such
as engineering, physics and thermal management. It has been difficult to get solutions for
plenty of them because of their variety of categories and the behaviors they usually depict.
Numerous methods like DTM, HAM and MVIM have been applied to determine solutions of
these problems that require additional computational work since all boundary conditions are
not included in the canonical form. This project investigated solutions for the problems in a
direct way treating numerically and analytically. Although the classical ADM is very powerful
in treating nonlinear BVPs, it fails to treat some singular boundary value problems. This
difficulty was alleviated by the modifications MADM and IADM. The difficulty of those
singular problems, due to the existence of a singular point was overcome in these
contributions. It was demonstrated that the approaches can be well suited to attain an
accurate solution to the second-order STPBVPs, linear and nonlinear, since the methods
require no perturbation, discretization and linearization. Moreover, symbolic programming
was employed to handle a model problem representing temperature distribution in heat
dissipating fins with triangular profiles. MADM was used with a piece of computer code in
MATLAB to seek solution for the mentioned problem involving constant and a power law
dependence of thermal conductivity on temperature governed by a linear and nonlinear BVPs,
respectively, for which considerable results were obtained. More examples including a
problem formulated dealing with a triangular silicon fin were solved and analyzed using
tables and figures for better elaborations where appreciable agreement between the
approximate and exact solutions was observed. All the computations were performed using
MATHEMATICA and MATLAB.
xii
1. INTRODUCTION
1.1. Background of the study
In mathematics, the field of ordinary differential equations describe phenomena that changes
continuously. They arise in models throughout Mathematics, Science and Engineering. A
boundary value problem (BVP) in ordinary differential equations is a differential equation
together with a set of additional constraints, called the boundary conditions. If the given
conditions are given at more than one point and the differential equation is of order two or
greater, it is called a boundary value problem.
A system of ordinary differential equations has many solutions. Commonly a solution of

interest is determined by specifying the values of all its components at a single point .
This point and a direction of integration define an initial value problem (IVP). In many
applications the solution of interest is determined in a more complicated way. A boundary
value problem specifies values or equations for solution components at more than one point in
the range of the independent variable . Generally IVPs have a unique solution, but this is not
true of BVPs. Like a system of linear algebraic equations, a BVP may not have a solution at
all, or may have a unique solution, or may have more than one solution.
A differential equation that could be reduced to
for some continuous functions and on an open interval is the second order differential
equation (linear). The equation subjected to either of the common boundary conditions that
can be categorized as:
2
where, constitute a second order boundary value problem. The boundary value
problems with boundary conditions and respectively are called BVPs with Dirchlet
and Neumann boundary conditions; whereas BVPs subjected to boundary conditions and
are known as BVPs with mixed boundary conditions. A solution to a BVP is a solution
to the differential equation which also satisfies the boundary conditions.
A two-point boundary value problem of total order n on a finite interval may be written
as an explicit first order system of ordinary differential equations with boundary values
evaluated at two points as
( ) ( )
The boundary conditions defined by must be independent; that is they cannot be expressed
in terms of each other (if is linear the boundary conditions must be linearly independent). In
practice, most BVPs do not arise in the form of but instead in combination of equations
defining various orders of derivatives of the variables which sum to n.
Here, and for some value with and where each of the vector
functions and are independent. However, there are well-known, commonly arising,
boundary conditions that are not separated; for example, consider periodic boundary
conditions which, for a problem written in the form of are
The word two-point refers to the fact that the boundary condition function is evaluated at the
solution at the two interval endpoints and unlike for IVPs where the initial conditions
are all evaluated at a single point. Occasionally, problems arise where the function is also
evaluated at the solution at other points in . Such cases can result in a multipoint BVP.
Most practically arising two-point BVPs have separated boundary conditions where the
function may be split into two parts one for each endpoint.
( ) ( )
Boundary value problems for linear differential equations are solved rather easily since the
final algebraic step involves the solution of linear equations. However, if the underlying
3
equations are nonlinear, the solution could have several branches, or the arbitrary constants
from the general solution could occur in different arguments of transcendental functions. As a
result, it is not always possible to complete the final algebraic step for nonlinear problems.
Several mathematical models that explain natural phenomena are mostly formulated in terms
of nonlinear differential equations, both ordinary and partial. Many problems in applied
sciences such as nuclear physics, engineering, thermal management, gas dynamics, chemical
reaction, studies of atomic structures and atomic calculations lead to singular boundary value
problems and often only positive solutions are important. However, most of the methods
developed in mathematics are used in solving linear differential equations. For these reasons, a
semi-analytical method named Adomian decomposition method (ADM) proposed by George
Adomian have been attracting the attentions of many mathematicians,
physicists, engineers and various graduate researchers. The method has the advantage of
converging to the exact solution and can easily handle a wide class of both linear and
nonlinear differential and integral equations. The commonly existing problems in various
fields of studies mentioned earlier lead to singular BVPs of the form:
( )
subject to either of the boundary conditions where atleast one of the functions
and has a singular point and and are finite constants. For example,
when and ( ) , is known as the
Thomas-Fermi equation given by the singular equation which arises in the
study of electrical potential in an atom. Also when and ( ) ,
is known as the generalized Emden-Fowler equation with a negative exponent and it
frequently arises in applied mathematics. Another example is given by the singular equation
, sometimes written in the form which results from
an analysis of heat conduction through a solid with heat generation. The function
represents the heat generation in the solid, is the temperature and the constant is equal to
or depending on whether the solid is a plate, a cylinder or a sphere. Equation
subject to the boundary conditions or represents temperature distribution inside a
triangular extended surface (or fin) with thermal conductivity. Thus, extended surfaces are
4
governed by BVPs; and they are widely used in many engineering appliances which include,
but are not limited to, air conditioning, refrigeration, automobile and chemical processing
equipments. For this reason this project tried to consider both invariant and power-law
dependence of thermal conductivities governed respectively by linear and nonlinear BVPs.
The primary objective of using extended surfaces is to enhance the heat transfer rate between a
solid and an adjoining fluid. Such an extended surface is termed a fin. In a convectional heat
exchanger heat is transferred from one fluid to another through a metallic wall. The rate of
heat transfer is directly proportional to the extent of the wall surface, the heat transfer
coefficient and to the temperature difference between one fluid and the adjacent surface. If
thin strips (fins) of metals are attached to the basic surface, extending into one fluid, the total
surface for heat transfer is thereby increased.
This project tried to consider two triangular profile fin problems by incorporating with
MATLAB codes. The modified Adomian decomposition method (MADM) is used to get the
component decomposition terms. The Adomian decomposition method (ADM) is a well-
known systematic method for practical solution of linear or nonlinear and deterministic or
stochastic operator equations, including ODEs, PDEs, integro-differential equations, etc. The
method is a powerful technique, which provides efficient algorithms for analytic approximate
solutions and numeric simulations for real-world applications in the applied sciences and
engineering. It permits to solve both nonlinear IVPs and BVPs without restrictive assumptions
such as required by linearization, perturbation, discretization, guessing the initial term or a set
of basis functions, and so forth. The accuracy of the analytic approximate solutions obtained
can be verified by direct substitution. A key notion is the Adomian polynomials, which are
tailored to the particular nonlinearity to solve nonlinear operator equations. The decomposition
method has been used extensively to solve effectively a class of linear and nonlinear ordinary
and partial differential equations. However, a little attention was devoted for its application in
solving the singular two-point boundary value problems (STPBVPs). This project treated
some classes of singular second-order two-point boundary value problems both analytically
and numerically using the ADM and the modifications Improved Adomian Decomposition
Method (IADM) and MADM focusing on the Dirchlet and mixed boundary conditions; and
applied the symbolic softwares MATLAB and MATHEMATICA to facilitate computing.
5
1.2. Statement of the problem
Singular two-point boundary value problems arise in different areas of applied sciences such
as engineering, physics and thermal management as already pointed out in section 1.1. It is
difficult to get solutions for plenty of them because of their variety of categories and the
behaviors they usually depict. Numerous methods have been applied to determine numerical
solutions of these scientific problems which of course result in variety of approximate
solutions. Modified variational iteration method (MVIM), Differential transformation method
(DTM), Finite difference method (FDM), Adomian decomposition method-Pad technique
(ADM-PT), etc. are among the methods used to treat such problems.
This project tried to investigate analytical and numerical solutions of STPBVPs given
generally by:
( )
where either of or are singular with boundary conditions , and using
the ADM, MADM and IADM accordingly. To perform this, it was tried to answer the project
questions:
1. How do the decomposition method and its modifications treat nonlinearity behaviors of
STPBVPs?
2. How do the modifications remove the singularity nature of the STPBVPs?
3. How do the decomposition methods treat linear BVPs?
6
1.3. Objectives
The main objective of this project was to study analytical and numerical solutions of second
order singular two-point boundary value problems using standard ADM and its modifications
IADM and MADM.
The study explored the specific objectives:
to describe the ADM and derive the Lesnics approach to solve STPBVPs.
to determine approximate solutions of STPBVPs by MADM and IADM alternatively.
to determine analytical solutions of STPBVPs using ADM.
to perform error analysis to compare the results obtained accordingly.
to make comparison between the numerical and analytical solutions of the problems by
using the respective plots on the same coordinate plane.
7
2. LITERATURE REVIEW
2.1. Extended surfaces
Scientists and engineers are interested in singular BVPs because the problems arise in a wide
range of applications, such as in chemical engineering, mechanical engineering, nuclear
industry, thermal management and nonlinear dynamical systems. For this reason, these kinds
of problems have been studied by many researchers. To be specific, as extended surfaces are
found in many engineering appliances, numerous studies have been performed on different fin
configurations with constant and variable thermal properties. Fins are echo-friendly and
economic means of convective heat transfer enhancement. They are encountered quite often in
practice: from industrial compact heat exchangers to CPU heat sink modules of personal
computers. Finned structures, better known as heat sinks, have well served thermal
management of electronic systems for many years. Kraus et al. (2001) provided a review on
this subject devoted to various aspects of extended surface convective heat transfer.
Mokheimer (2003) investigated the performance of annular fins with different profiles subject
to variable heat transfer coefficient. The performance of the fin is expressed in terms of fin
efficiency as a function of the ambient and fin geometry parameters.
2.2. Numerical and Analytical treatments of the BVPs
In order to treat STPBVPs, plenty of numerical methods were investigated during the past
years. Betaineh et al. (2008) used modified homotopy analysis method (MHAM) to treat
STPBVPs analytically. Though both Homotopy analysis method (HAM) and MHAM are
effective in treating some of such problems, the methods require plenty of computational
works. For instance, if one needs to solve a STPBVP analytically, he/she is expected to
determine different order deformation equations and plot a graph of a curve called an h-curve
which needs time and effort to get the required value of .
Zavalani (2015) used Galerkin-finite element method (G-FEM) to treat two point BVPs
numerically. The method uses three basic procedures. Making weak formulation is the first
step to determine, the second step is to develop finite element formulation. Lastly, weighted
average is used for full discretization; that are all required by neither the ADM nor its
modifications. Moreover, the method fails to treat nonlinear and singular BVPs.
8
Cen (2006) investigated a numerical method based on Greens function to obtain numerical
solution of STPBVPs. Beyond complex behavior of the method it fails to treat BVPs with
inhomogeneous boundary conditions.
Khuri and Sayfy (2010) proposed a new approach, modified decomposition method in
combination with the cubic B-spline collocation technique introduced for numerical solution
of a class of singular boundary value problems arising in physiology. The main thrust of this
approach is to decompose the domain of the problem into two subintervals. The singularity,
which lies in the first subinterval, is removed via the application of a modified decomposition
procedure based on a special integral operator that is applied to surmount the singularity.
Then, in the second sub-domain, which is outside the vicinity of the singularity, the resulting
problem is treated employing the B-spline collocation technique. The approach is applicable to
treat nonlinear singular BVPs and in addition, the approximate solutions converge faster to the
exact solution but very vast computational work is needed.
Noor and Mohyud-Din (2008) used modified variational iteration method (MVIM) to
determine solutions of singular and non-singular initial and boundary value problems. The
proposed technique is applied on equations of higher order boundary layer problem, unsteady
flow of gas, third-order dispersive and fourth-order parabolic partial differential equations.
The authors made the modification by introducing ADM to VIM. They used the Pade
approximants technique (PAT) in order to make their work more concise and for better
understanding of the solution behavior but needs caution to select the initial value because the
approximants are heavily dependent on the initial value. Though they concluded that the
proposed frame work is very powerful and efficient in finding the analytical solutions for
singular and nonsingular boundary value problems, the method requires a huge computational
work as it is incorporated from the ADM, VIM, and PAT.
Ravi and Aruna used an analytical method called DTM to obtain the exact solutions
for some linear two-point boundary value problems with singular nature. Although these
analytical methods are effective in the linear case, their applicability for nonlinear problems
was not examined in the paper.
9
2.3. Treatments of related problems using ADM and its modifications
Adomian formally introduced ADM concept and formulas for generating Adomian
polynomials for all forms of nonlinearities since early 1980s. Several authors have been
focusing on this area to develop a practical method for the calculation of Adomian
polynomials.
Adomian (1988) reviewed the method in Applied Mathematics; and concluded that the
decomposition method can be an effective method for solution of a wide class of problems
providing generally a rapidly convergent series solution. Based on the author, it has some
distinct advantages over usual approximation methods in that it is computationally convenient,
provides analytic, verifiable solutions not requiring perturbation, linearization, or
discretization required by methods like FDM which results in massive computations. Since it
solves nonlinear problems rather than linearizing them, the resulting solutions are more
physically realistic.
Inc and Evans obtained an approximate solution for only one nonlinear example by
using the ADM-Pad technique. The technique they proposed requires the computation of
undetermined coefficients.
Hassan and Zhu ) introduced an efficient modification of the ADM for solving singular
initial value problem in the second-order ordinary differential equations. The scheme is tested
for some examples and the obtained results demonstrate efficiency of the proposed method.
The study showed that the decomposition method is simple and easy to use and produces
reliable results with few iterations. The results show that the rate of convergence of the
modified decomposition method, MADM is higher than the standard ADM for initial value
problems.
Several other researchers also have developed modifications to the ADM. The modifications
arise from evaluating difficulties specific for the type of problem under consideration. The
modification usually involves only a slight change and is aimed at improving the convergence
or accuracy of the series solution.
Duan et al. reviewed the ADM and its modifications including different modified and
parameterized recursion schemes, the multistage ADM for initial value problems as well as the
10
multistage ADM for boundary value problems, new developments of the method and its
applications to linear or nonlinear and ordinary or partial differential equations, including
fractional differential equations. The authors have presented a contemporary review of the
ADM and discussed its utility and advantages for solving linear or nonlinear and deterministic
or stochastic operator equations without any restrictive assumptions, including ODEs, PDEs,
integral equations and integro-differential equations for IVPs or BVPs. They also concluded
that the ADM is the method of choice for solving nonlinear differential equations with a wide
class of analytic nonlinearities including product, polynomial, exponential, trigonometric,
hyperbolic, composite, negative-power, radical and decimal-power nonlinearities.
Furthermore, the ADM has been shown to be a reliable method for the solution of nonlinear
fractional ODEs and PDEs for both IVPs and BVPs. In summary, based on Duan et al.
the ADM is a powerful and efficient technique for the solution of nonlinear ordinary,
partial and fractional differential equations. It provides the analyst with an easily computable,
readily verifiable and rapidly convergent sequence of analytic approximate functions for the
solution.
Singh et al. proposed two modified recursive schemes for solving a class of doubly
singular two-point boundary value problems. The authors illustrated how the modified
recursive schemes can be used to solve a class of doubly singular two-point boundary value
problems. Based on the authors the accuracy of the numerical results indicates that the method
is well suited for the solution of this type problem. They included that the main advantage of
the approach is that it provides a direct scheme to obtain approximate solutions, and they have
also shown graphically that these approximate solutions are almost identical to the analytic
solution. They also stated another advantage of the modified recursion scheme as: it does not
require the computation of undetermined coefficients, whereas most of previous recursive
schemes like by Inc and Evans ( ) do require the computation of undetermined
coefficients. The method provides a reliable technique which requires less work compared to
the traditional techniques such as FDM, Cubic spline method, and standard ADM. The
numerical results of the examples are presented and they believe that only a few terms are
required to obtain accurate solutions. By comparing the results with other existing methods, it
has been proved that proposed modified ADM is a more powerful method for solving the
singular problems.
11
Lin and Chen proposed a modification of the Adomian decomposition method,

MADM. The authors demonstrated that nonlinear double singular BVPs of second order can
be handled without difficulty. Furthermore, the authors stated that MADM can also be used to
solve singular boundary value problems where the traditional method ADM fails. The
numerical computation gives a more precise approximation of the solution. The reported
results show a greater improvement over the traditional method.
Kaliyappan and Hariharan recall that they presented a simple way of computing
Adomian polynomials by applying the decomposition of positive integers as a subscript of the
variable for nonlinear terms. They developed MATLAB codes that can easily generate
Adomian polynomials with single variable. They recommended the possibility of extension of
the codes for multivariable case. In this paper MATHEMATICA is used instead to generate
Adomian polynomials.
12
3. MATERIALS AND METHODS
Sources in the web and libraries were used to collect all the pieces of information about the
singular two-point boundary value problems together with the methods and recorded
subsequently. Specifically,
relevant journals and books were addressed to gather information about STPBVPs and
the methods to treat the problems.
Convectional heat transfer is considered to study temperature distribution along a
triangular fin profile by formulating a singular two-point BVP.
second order two-point BVPs of singular linear and nonlinear were identified.
the collected information was analyzed and arranged keeping coherence.
symbolic softwares, MATLAB and MATHEMATICA were applied suitably to ease
the computations by the methods and graphs were plotted using the programs.
frequent contacts with the advisor have been made throughout the study to make the
necessary amendments and modifications before the submission of the final report.
13
4. PRELIMINARIES
4.1. Singular points
It is of interest to solve the differential equation
or, in standard form,
in the neighborhood of a singular point, as the behavior of the solutions there may be among
their most important features. When the singularities are not too wild, a modification of the
technique of power series can be used to calculate the solutions there.
To simplify the discussion, attentions are restricted to equations of the form (4.1) where,
and are polynomials, that they may be assumed to have no common factors.
The solutions to the linear second order differential equations of the form (4.1) in the case with
, and real analytic in a neighborhood of a point and can be expanded as
power series of the form:
and with the knowledge of and the coefficients with can be determined
successively by certain recurrence relations.
There are many important linear second order differential equations (with non-constant
coefficients) that arise in mathematical physics which do not satisfy the above conditions, and
power series solutions would like to be able to obtained for some of these.
The method discovered by the German mathematician F. G. Frobenius in the 1870s and is
therefore often called the Frobenius method is used to determine the singular points.
Definition 4.1. A function is said to be analytic at a point if can be expanded in a

power series about which has a positive radius of convergence.
14
Definition 4.2. A point is called an ordinary point of a differential equation of the form
(4.2) if both and are analytic at .
Definition 4.3. The point is called a singular point of if
This causes difficulties in writing in the standard form
Definition 4.4. The singular point is said to be regular singular point of (4.1) if the
equation can be written as
This means that, for near , the functions and can be rewritten as and
where and are analytic at . This is the same as saying that
the multiplier of in has at most the factor in denominator; and the multiplier
of in (4.4) has at most the factor in the denominator.
A singular point which is not regular is called irregular singular point.
Remark: If and are analytic at and , then is analytic at .
Example 4.1. Given the Legendre differential equation

, let . Find the singular points of the equation and determine which regular
singular points are.
Solution
The points are the zeros of and are clearly,
To test whether example is regular, divide the given equation by the factor to
obtain
15
Here, it can be observed that the multiplier of has the form
( )
and the multiplier of has the form
( )
The denominators of the factors in the parentheses above are not zero at , so they are
analytic at . Hence, is a regular singular point. A similar test for shows
that it also is a regular singular point.
Example 4.2. Determine all the singular points of
Solution
Here, it can easily be observe that
The singular points are those points where and fail to be analytic. Observe that
and are ratios of functions that are analytic everywhere. Therefore and
are analytic except, perhaps, when their denominators are zero.
is analytic except at But in the case of recall that has the power series
expansion:
Thus
Which is analytic everywhere. Consequently the only singular point of the given DE is
16
4.2. Boundary value problems
Definition 4.5. A boundary value problem for second order ordinary differential equations is a
differential equation with conditions specified at the extremes (boundaries) of the independent
variable in the equation. Recall that an initial value problem has all of the conditions specified
at the same value of the independent variable; and that value is at the left end of the domain,
thus the term "initial" value.
A boundary value problem can generally be denoted as
The value of the dependent variable or its derivatives is specified at two different points
A differential equation together with suitable boundary conditions forms a two-point boundary
value problem.
Definition 4.6. Let be given function and and be given numbers. The
problem
is called two-point boundary value problem. Two-point boundary value problems are BVPs
with boundary conditions given at two specified points. Equation subjected to the
boundary conditions given at two particular points and as is a simple example of a
two-point boundary value problem.
The differential equation at boundary conditions is known as linear BVP.
Linear BVPs can mainly be classified as homogeneous and inhomogeneous.
The linear BVPs are called homogeneous if and are all zero. They are said to
known as inhomogeneous if atleast one of and are non-zero.
17
A boundary value problem for a differential equation consists of finding a solution of the
differential equation that also satisfies the boundary conditions. A boundary value problem
may or may not have a solution; and if it has a solution it may be unique or infinitely many.
Example 4.3. Solve the BVP
Solution
The general solution can be obtained to be
And from the boundary conditions given
Here observe that both of the boundary conditions are resulted only in the value of Thus
implies
i.e. infinitely many solutions
implies and any constant This shows that the boundary conditions are
incompatible on Hence this ceases no solution for the BVP.
In the usual linear case one fails to have continuity in one of the coefficients at the end points
of an interval where the boundary conditions are applied. This project focuses on the singular
points and nonlinear problems where either of the coefficient functions is not defined at zero,
even though zero may be prescribed at a boundary condition at an end point.
Recall that a point is called a singular point of
if
And this differential equation subjected to boundary conditions given at two specified points
and as and is a two-point boundary value problem.
18
Definition 4.7. A two-point BVP with a singular point subjected to two specified
boundary conditions at and in an interval I as and , where and
are some constants, the two-point BVP is called a STPBVP.
Example 4.4. The BVP
with and is singular at Hence the two-point BVP is a good example

of STPBVP.
4.3. Operators
An operator is a function that takes a function as an argument instead of numbers as we are

used to dealing with in functions. We already know a couple of operators even if we did not
know that they were operators. Here are some examples of operators.
Or, if a function is plugged in, say in each of the above, then the following can be
obtained.
These are all fairly simple examples of operators but the derivative and integral are operators.
A more complicated operator would be the heat operator. The heat operator can be found from
a slight rewrite of the heat equation without sources. The heat operator is then
and
( )
is another differential operator for a particular second order differential equation (see the
example below).
19
The operator in second order differential equations is a twice differentiable function. The
domain of is the twice differentiable functions on an open interval . The terminology of
the function is used to describe or or, simply and the range of the functions on
(and hence is itself a function on ). Generally, is chosen
for the order differential equations and thus its inverse follows as the -fold definite
integration operator from to The operator defined has the following basic property:
If and are twice differentiable functions on and and are constants, then
Definition 4.8. An operator satisfying property is called linear operator.
Example 4.10. The differential equation
can be re written in compact form as
( )
So from this an operator can be generated to be:
( )
so that can be written in an operator form as:
( )
Beyond this, it can be verified that the operator is a linear operator. Bearing the basic
properties of derivation of ordinary differential equations in mind and plugging in the
expression into the operator above it can be shown as:
20
( )
[ ]
( )
( ) ( )
Thus the operator at is linear operator.
An operator that is not linear is known as nonlinear operator. In this project nonlinear
operators in which nonlinear functions are plugged are symbolized by some various
representations like and . These operators are used to determine the Adomian
polynomials by the help of Adomian formula which is briefly discussed in the following
chapter.
21
5. SINGULAR TWO-POINT BVPs VIA ADM, IADM AND MADM
5.1. General Description of the Adomian Decomposition Method
In the 1980s, George Adomian introduced a new method to solve nonlinear functional
equations. This method has since been termed the ADM and has been the subject of many
investigations. The method involves splitting the equation under investigation into linear and
nonlinear portions. The linear operator representing the linear portion of the equation is
inverted and the inverse operator is then applied to the equation. The nonlinear portion is
decomposed into a series of Adomian polynomials. This method generates a solution in the
form of a series whose terms are determined by a recursive relationship using the polynomials.
In reviewing the basic methodology involved, consider a general differential equation in an
operator form:
(5.1)
where is an operator representing the linear portion which is easily invertible, is the
nonlinear operator representing the nonlinear term and is a linear operator for the remainder
of the linear portion. Applying the inverse operator , the equation then becomes
(5.2)
Since is linear, would represent integration and with any given boundary conditions, will
give an equation for incorporating these conditions. This gives
(5.3)
where represents the function generated by integrating and using the boundary
conditions.
The Adomian decomposition method introduces the solution and the nonlinear function
by the infinite series as:
and
22
where, are the Adomian polynomials that can be determined by Adomian formula:
[ ( )]
So for , reduces to
For it reduces to
For it becomes
The desired Adomian polynomials can be determined using similar procedure.
.
.
.
It can be observed that depends only on , depends only on and , depends
only on and and so on.
Optionally, a simple way of computing Adomian polynomials of any type of nonlinearity is

presented by applying the decomposition of positive integers as a subscript of the variable
for nonlinear terms through the use of MATHEMATICA software in the appendix C.
MATHEMATICA exploits general symbolic programming for generating Adomian
polynomials.
Now, substituting ( ) and ( ) into ( ), one can get:

23
The recursive relationship is found to be
(5.8)
And hence from
for . So, having determined the components the solution in a series

form follows immediately by .
Theorem 5.1. Let be a polynomial, i.e.,

is a polynomial. i.e.,
Proof see page of Lin and Chen
Based on Lin and Chen for a given polynomial
is also a polynomial and then a combination of polynomials is also a polynomial.
So, will be a polynomial. The same procedure holds to obtain as a polynomial.

24
Example 5.1. As a simple example consider the nonlinear differential equation
Recall that the ADM involves separating the equation under investigation into linear and
nonlinear portions. The linear operator representing the linear portion of the equation is
inverted and the inverse operator is then applied to the equation. Any given conditions are
taken into consideration. The nonlinear portion is decomposed into a series of Adomian
polynomials. This method generates a solution in the form of a series whose terms are
determined by a recursive relationship using these Adomian polynomials.
Following the method described above, the desired linear operator is defined as
The inverse operator is then
Rewriting the differential operator in operator form, we have
where is nonlinear operator such that
Next apply the inverse operator for to the left hand side of the equation the following way
Using the initial condition this becomes
Returning this to it becomes

25
Next the Adomian polynomials, would be generated as follows:
Let be expanded as infinite series
and define
Then
( )
So from (5.12) the ADM invites the recursive relation:
Now using the Adomian formula given by (5.6) the polynomials become:
.
.
.
26
Using these results in the recursive relation (5.13) the solution is generated as follows:
.
.
.
Hence the approximate solution for obtained using the first five components denoted
look like the following.
Example 5.2. Consider Airys equation given by
The equation can be written in operator form as:
where is a highest degree operator representing the linear portion which is easily invertible
defined by
27
is another linear operator for the remainder of the linear portion defined by .
and the corresponding two fold integral for the operator is defined by
Then operating with in similar procedure as example it is obtained that
This results in
.
.
.
is Airys equation.
The decomposition method and its modifications can be applied suitably to treat singular
BVPs. How the ADM treats a class of singular BVPs is discussed below. For illustrative
purpose, the following class of singular BVPs is considered which can be handled more easily,
quickly and elegantly by implementing ADM for the exact solutions without making massive
computational work. To begin,
28
Consider (1.6) in the form
, (5.14)
subject to the boundary conditions
(5.14) in an operator form can be rewritten as:
where the linear differential operator is defined by
( )
The inverse operator is therefore defined by
( )
Operating with (5.16), it then follows
( )
In this equation the expression may represent linear or nonlinear differential equation.
Note that a second order ordinary differential equation written in standard form over the
dependent variable and independent variable is said to be linear if it can be written in the
form:
otherwise it can be called as nonlinear.
It is not unusual to transform an equation into another equivalent form so that it would be
convenient to use better techniques of solving. The following two cases deal with the
expression in (5.19).
29
Case 1. If is linear.
Consider in the form:
. Thus, (5.19) becomes
( ) ( )
Now decompose the solution as in (5.4).
It is important to note that the standard ADM suggests that the zeroth component usually
defined by the function:
( )
that represents the terms arising from integrating the source term and from using the
given conditions; and all are assumed to be prescribed. Accordingly, the solution can be
computed by using the recurrence relation:
( )
Case 2. If is nonlinear.
where is nonlinear.
Substituting this in to ( ),
( )
The ADM is based on decomposing and the nonlinear term as

30
and
Using these results in , it then follows
( )
According to the standard ADM, the solution can be computed by using the recurrence
relation:
( )
Thus, determining the components using ( ) and/or ( ) the solution in a

series form follows immediately. The series may be summed to provide the solution in a
closed form. However, for concrete problems, the term partial sum may be used to give the
approximate solution as:
The classical ADM is very powerful in treating nonlinear BVPs, though this is one of the
qualities of the method over some other methods it has its own shortcomings like its failure to
treat some nonlinear singular boundary value problems. The following subsection addresses
some merits and demerits of the method.
5.1.1. Advantages and disadvantages of the ADM
Holmquist (2007) addressed views of different researchers who have used the ADM and have
enumerated on the many advantages that it offers and disadvantages. Most often cited is the
efficiency of the method. Many authors found that the ADM requires less computational work
than traditional approaches. Other advantages include the ability to solve nonlinear problems
31
without linearization, the wide applicability to several types of problems and scientific fields,
and the development of a reliable, analytic solution. This method does not linearize the
problem nor use assumptions of weak nonlinearity and therefore can handle nonlinearities
which are quite general and generates solutions that may be more realistic than those achieved
by simplifying the model to achieve conditions required for other techniques. Moreover, the
ADM is analytic, requiring neither linearization nor perturbation, and continuous with no
resort to discretization.
The ADM does have some disadvantages, however. To begin with, the method gives a
series solution which must be truncated for practical applications. ADM requires the use of
Adomian polynomials for nonlinear terms, and this needs more work. In addition, the rate and
region of convergence are potential shortcomings. Although the series can be rapidly
convergent in a very small region, it has very slow convergence rate in wider region and the
truncated series solution is an inaccurate solution in that region, which will restrict the
application area of the method.
5.1.2. MADM
Since the introduction of the method in early 1980s, ADM has led to several modifications
made by various researchers in an attempt to improve the accuracy and expand the application
of the original method. As pointed out above, the rate of convergence of the series solutions is
one of the potential shortcomings of the decomposition method. To improve on this, the
authors tried to introduce different modifications of the method. To begin with, based on
Wazwaz (1999) the standard ADM is modified in such a way that the function in can
be divided into two parts as follows to increase rate of convergence of the series solution and
minimize the size of computations. This modification is applicable irrespective of the types of
the BVPs under consideration.
(5.28)
Accordingly, a slight variation was proposed only on the components and . The
suggestion was that only the parts be assigned to the component , whereas the remaining
part be combined with other terms given in to define to get the recursive relation:
32
Although this variation in the formation of and is slight, however it plays a major role in
accelerating the convergence of the solution and in minimizing the size of calculations. In
many cases the modified scheme avoids unnecessary computations, especially in calculation
of the Adomian polynomials. In other words, sometimes there is no need to evaluate the so-
called Adomian polynomials required for nonlinear operators or if needed to evaluate these
polynomials the computation will be reduced very considerably by using the modified
recursive scheme. There are two important remarks related to the modified method. First, by
proper selection of the functions and , the exact solution may be obtained by using
very few iterations, and sometimes by evaluating only two components. The success of this
modification depends only on the choice of and , and this can be made through trials,
that are the only criteria which can be applied so far. Second, if consists of one term only,
the scheme should be employed in this case.
Another modification of the standard ADM which alleviates the deficiency of treating some
singular boundary value problems like, BVPs subjected to mixed boundary conditions of the
form is MADM presented by Hasan and Zhu In fact, it is a slight refinement to
the original ADM; it only modifies the involved differential operator. Generally, MADM by
the authors mentioned proposes the differential and inverse operators:
and
times
for treatment of order boundary value problem of the form:
33
In sub section the operators above are reduced so that they could be used to treat second
order BVPs and then symbolic programming is incorporated into them to obtain solutions to a
triangular fin problems involving both constant and power-law dependent thermal
conductivity (highly nonlinear).
Kim and Chun came up with another modification of the standard ADM to solve
singular order boundary value problems. This scheme is designed in such a way that
BVPs with singular nature can easily be treated. In this project the operators are used to treat
singular second-order BVPs with mixed boundary conditions of the form Generally, to
see what MADM by the authors mentioned look like, consider the singular boundary value
problem of order ordinary differential equation ( ) given the following way:
Where is nonlinear differential operator of order less than , is a given function,

are given constants, where
Now can be re written in the form
Or equivalently,
[ ( )]
can be written in the operator form
Where, the differential operator employs the first two derivatives
( )
34
in order to overcome the singularity behavior at .
In view of and , the inverse operators and are the integral operators
defined by
time
By applying on one can have
such that
By applying on one can have
such that
The standard ADM introduces the solution and the nonlinear function by infinite
series given by and where the Adomian polynomials are determined by the
formula at Substituting and in to gives
Identifying the Adomian method admits the use of

the recursive relation
35
which gives
.
.
.
This leads to the complete determination of the components of The series solution
defined by follows immediately. For numerical purposes, the term approximant
defined by can be used to approximate the exact solution.
5.1.3. IADM
Ebaid (2010) made improvements of operators developed earlier by Lesnic (2001) for the
purpose of treating the heat equation with Dirchlet boundary condition. In this work the IADM
is used to deal with linear and nonlinear STPBVPs with Dirchlet boundary conditions. The
improvement is based on the ADM and Lesnics work later developed by Ebaid (2010). Lesnic
(2001) proposed the inverse operators:
to solve the Dirchlet BVP for the heat equation
, , (5.50)
under the boundary conditions , and the initial condition

.
Using the definition in ( ) it is observed that
i.e., the boundary conditions can be used directly. However, from ) again one can see
that the lower bound of all integrations is restricted to the initial point .
36
In fact, this restriction can be avoided by using a new definition of which gives the same
result as in ( ) and given by:
where, is free lower point. This free lower point plays an important role if the equation being
solved has a singular point. So the desired operator originally designed by the author
mentioned earlier is derived as follows.
is defined as:
where is to be determined such that can be expressed only in terms of the

boundary conditions given in . Using this definition, thus:
Setting and ,
In order to express in terms of the two boundary conditions only, the coefficient
has to be eliminated by setting
assuming .
Using ( ) in ( ), the operator below proposed by Ebaid (2010) is obtained to solve the
singular two-point Dirchlet BVPs and
37
Thus is reduced to:
From ( ) it is noted that is already expressed in terms of the given boundary

conditions without any restrictions on . So, the choice of the value that can take depend
properly on the singular point of the equation under consideration. For example, if the
equation has a singular point say at , will be chosen to be any real value except the
value of . Moreover, if the equation has two singular points at and , then
will be considered any real value except these values of and . In general, if the equation
has singular points , ,..., , then takes any real value except the values of these
singular points.
For solving linear and non-linear singular two-point boundary value problems under the
Dirchlet boundary condition, IADM is established using ( ) together with the standard
ADM.
( )
Applying the operator given by ( ) on both sides of ( ), it is observed that:
[ ( )]
The Adomian decomposition method introduces the solution and the nonlinear function
by infinite series as in ( ) and ( ), respectively.
Substituting the results into ( ) and according to the ADM, the solution can be
smartly computed by using the recurrence relations constructed based on the following cases.
38
Case 1:- If , i.e., linear function, then the solution can be computed by using
the recurrence relation:
, (5.60)
Case 2:- If is nonlinear function, then the recurrence relation required to compute the
solution is:
, (5.61)
where ( ) and ( ), ( ) and ( ) improve the standard ADM, (IADM) and can be
used to solve linear and nonlinear singular two-point boundary value problems subject to
Dirchlet boundary conditions. Hence, using the recurrence relation ( ) or ( ) depending
on linearity behavior of the boundary value problem, the truncated approximate
solution can be computed by the help of ( ).
5.2. Numerical and Analytical Illustrations
In this section, scientific real life application problem about triangular fin that involves both
invariant and power-law dependent thermal conductivities is considered. The modified
operators by Hassan and Zhu shown in sub section and MATLAB codes are
assimilated to obtain the numerical solutions. For the sake of illustrations, plenty of examples
are shown.
5.2.1. Application
Consider a wall at temperature transferring heat by convection to an ambient at temperature

as shown in the figure below.
39
Figure 5.1. A wall transferring heat by convection.
The rate of heat transfer from this wall may be evaluated in terms of a heat transfer coefficient
in the form
where, is the heat flux, coefficient of heat transfer, the cross section, temperature at
the base /wall temperature/ and the ambient temperature.
One of the prime objectives of the study of heat transfer is to find ways of controlling this .
For example, the design of a heat exchanger is often based on achieving the smallest possible
heat transfer area or the largest possible amount of heat transfer for any given size heat
exchanger.
As can be observed from one of the ways of increasing is by increasing the heat
transfer area. The surface area of a wall may, in principle, be increased in two ways as shown
in figures below.
40
Figure 5.2. Triangular and rectangular fin profiles
In Figure the extended surfaces are integral parts of the base material, obtained by a
casting or extruding process. In Figures the extended surfaces, that may or may
not be made from the base material, are attached to the base by pressing, soldering, or
welding. The same geometry is obtained, though less frequently, by machining the base
material. In practice, manufacturing technology and cost dictate the selection of the most
desirable form.
Applications of extended surfaces are numerous, particularly in heat transfer to gaseous media.
Since in this case the corresponding heat transfer coefficient is low, a small, compact heat
exchanger may be achieved only by the use of extended surfaces.
Since the temperature of an extended surface does not remain constant along its length,
because of transversal heat transfer by convection to the surroundings, the heat transfer from
extended surfaces cannot be evaluated from Thus it is unquestionable to evaluate the
temperature distribution in extended surfaces, which enables to determine the heat transfer in
terms of the temperature distribution. So this project studied the temperature distribution in a
triangular profile as shown in figure below considering both constant and variable thermal
conductivities across a variable cross section as follows.
41
Figure 5.3. The geometry of a straight fin of triangular profile.
The general formulation of problems of extended surfaces with variable cross sections is given
by
( )
This equation is elaborated in two cases considering constant and variable thermal
conductivities in the following sub sections.
5.2.1.1. Constant thermal conductivity
Assuming constant thermal conductivity and measuring temperature above the ambient,
may further be re-designed to give:
( )
where, ; and noting from figure above that ( ) and
inserting these values into it becomes
( )
( )
42
Hence the differential equation governing temperature distribution inside a triangular fin with
invariant thermal conductivity can be expressed as:
( )
with boundary conditions and where is the temperature measured
above the ambient temperature, subscript stands for fin base and is the fin length. In
addition, is the fin parameter defined as:
where and are convective heat transfer coefficients of fins either sides, is the thermal
conductivity and is fins vertical dimension at its base. For this formulation, the origin of
coordinates is placed at the tapered end of the fin.
From ( ) and (5 ), appropriate differential and inverse operators to treat

respectively are,
and
Equation can be rewritten as:
This in its operator form can be expressed as:
Applying ( ) on both sides of ( ), the following is found
( )
43
Using ( ) and the boundary condition, the recursive relation can easily be formulated as:
MATLAB codes are used to determine the required decomposition components of this
recurrence relation and are available in appendix A. For the sake of demonstration, the first six
components of the solution computed by the foresaid MATLAB code are given as follows.
So by ( ) the approximate solution, say using the first ten components can be determined
as:

44
(( ( ))
( ( ))
( (
)) )
But only the first three components of the MADM solution are used to compare with the exact
solution. The relevant analytical exact solution of ) as given by Arpaci (1966) is:
where denotes Bessels function of second kind.
So from ),
A relative error which covers part of a silicon fin length is defined to see how the approximate
MADM solutions to the exact solution are.
An absolute relative deviation (ARD) from exact solution that considers ten different points of
the interval is defined and computed the following way.
45
(| | | | | |
| | | | | |
| | | | | |
| | )
= 0.0153360308460284
ARD from exact solution is defined to see how far the MADM solutions from the exact
solutions are; and it shows that the approximate solution digresses only about 1.5% from the
exact solution for a small parameter . Such a definition for ARD gives a global and average
sense of deviation from exact solution and is more valid than focusing on error at a single
fixed point. Even this gap may be filled more by taking appropriate values of the constants.
Note that larger temperature difference does not imply less heat loss. Arpaci (1966) shares this
by stating In terms of heat transfer from extended surfaces, it is more appropriate to compare
the exact and the approximate heat losses than the exact and approximate temperatures, to
determine the limitation of the approximate solution.
5.2.1.2. Power law dependence of thermal conductivity
Assuming general power-law dependence under steady condition for thermal conductivity of
the fin material in the form
46
and averaged heat transfer coefficient, the succeeding governing equation can be derived the
following way.
( )
where, denote the variable cross-sectional area, local temperature, the periphery and
the heat transfer coefficient, respectively. Substituting the appropriate terms, becomes
( ( ) ( ) )
( )
( )
Dividing by and rearranging gives:
( ) ( )
Using ( ), equation ( ) can easily be converted to its operator form equivalent as:
( ) ( )
Applying ( ) on both sides of ( ), one can achieve:
( ) ( ) ( )
Here, it is easy to observe that there are three different nonlinearities; and they can be
represented as series decomposition of three different Adomian polynomials as follows.
( )

47
Using ( ) in ( ) and following the decomposition technique,
( ) ( ) ( )
Thus, the recurrence relation needed is:
( ) ( )
To reach the very final solution, each decomposed component of the series
has to be computed recursively. For this reason, it is necessary to obtain Adomian polynomials
components of at each of the iterations. To handle this task neatly, three
functions returning symbolic representations for , a function to take inverse
transform and a core code to calculate the ultimate solution with the help of these functions are
used. All these MATLAB codes are given in the appendix B.
Using the computational code mentioned previously the MADM solution series can be
expanded upto any desired component. To demonstrate, the first three components of the
decomposition solution can be shown as:
5.2.2. Numerical and Analytical Treatments

Model problem
Silicon, being an efficient thermal conductor, has been of extensive interest in fabrication of
cooling fins and packed heat sink modules especially for thermal management in
48
microelectronics. For temperatures ranging within , a power law correlation for

thermal conductivity of silicon is given by:
( )
where,
Regarding ),
Now it can be taken the advantage of the described computational work in subsection 5.2.1.2
based on MADM to investigate the temperature distribution in a triangular silicon fin with
dimensions L=0.05m and b=0.005m, subject to a constant base temperature of
and ambient temperature of The results to this model
problem from the first eight decomposition solutions by MADM are used to plot the graph as
shown in figure 5.4.
Figure 5.4. Temperature distribution along the length of the silicon fin.
The figure is all about variant thermal conductivity (the power law dependence) of the silicon
fin considered above taking the first eight decomposition solutions of the problem. It can be
observed from the figure that the temperature distribution decreases as one goes from the
source to the tapered end of the triangular fin. So, unnecessary heat from a system can be
49
dissipated to the surrounding atmosphere through extended surfaces. Compared to the FDM by
Mokheimer (2003), the temperature along the fin determined by MADM is observed to
decline faster though only the first eight components of the decomposition were used. Hence,
overheating of operating machines can be protected by plugging in many extended surfaces to
the machines. All these confirm that fins can serve as effective coolants. Outside this, it is not
difficult to understand how powerful the scheme, modification of ADM to treat highly
nonlinear BVPs.
Example 5.3. Consider
with the conditions and
I. ADM solution
As the given equation is not inhomogeneous, ADM fails to provide the self-cancelling noise
terms that ADM favors where sum of noise terms vanishes in the limit. Self-cancelling noise
terms appear for some inhomogeneous differential equations only. Thus, if terms in are
cancelled by terms in even though includes further terms, then the remaining non
cancelled terms in are cancelled by terms in the remaining terms of are cancelled by
some terms of and so on. Finally, the exact solution of the equation is readily found for the
inhomogeneous case by determining the first two or three terms of the solution and by
keeping only the non-cancelled terms of .
Moreover, it can be observed that the standard ADM fails to treat such problems since its
operators are designed so as to give no attention to the boundary conditions of the BVPs. Such
difficulties are resolved perfectly by the IADM in which the operators are designed in such a
way that all the boundary conditions are included. But the exact solution of the problem as
given by Benabidallah and Cherruault (2004) is
II. IADM solution

Rewriting (5.90) in the standard form gives:
50
With
using ) given by
Substituting the given terms, one can get:
For and the given terms above at the second equation in reduces to:
[ ]
Similarly, can be determined as:
[ ]
[ ]
By ( ),
which is the approximate solution using the first three iterations only. But the first seven
iterative results are used to get better numerical results in the table 5.1 that shows the
numerical illustration using both the IADM and exact solutions together.
51
In order to verify how close the IADM solution to the exact solution the graph of the
two solutions are plotted in figure 5.5. For the IADM solution it is enough to use only the first
three components of the decomposition series solution. All the computations and the plot are
done by MATHEMATICA.
Figure 5.5. Comparison of the IADM and Exact solutions for example 5.3.
Table 5.1. Numerical results for example 5.3.
Approximate Exact Error

x
Solution Solution
1 1 1 0
1.1 1.024373035 1.024373062 2.6258E08
1.2 1.042179302 1.042179356 5.3793E08
1.3 1.053967377 1.053967445 6.7857E08
1.4 1.060201157 1.060201221 6.3608E08
1.5 1.061278079 1.061278124 4.4892E08
1.6 1.057542456 1.057542476 2.0090E08
1.7 1.049295464 1.049295466 1.8739E09
1.8 1.03680277 1.036802784 1.4216E08
1.9 1.030300538 1.020300552 1.3555E08
2 1 1 0
52
In figure 5.5, the IADM solution and the exact solution
are plotted together but to get better approximate solution, MATHEMATICA is used to
evaluate up to the 7th iteration as shown in table 5.1. It can be concluded from the figure that
an accurate IADM solution is obtained considering only few terms of the decomposition
solution. Also, in order to compare the approach with another modification of the ADM at
Benabidallah and Cherruault (2004) the numerical results for the absolute errors
are presented in table 5.1. In the journal mentioned, the authors used the 7-terms
of the ADM in indirect way in which a huge amount of computational work is needed to
obtain the numerical solution with absolute errors . Hence, in comparison, the
improved Adomian decomposition method is not only more accurate but also more simple and
direct.
Example 5.4. Consider the inhomogeneous singular Bessel equation
subject to the boundary conditions and .
I.IADM solution
The equation considered has a singular point at . So, by IADM the free lower point can
be chosen to be any real value except zero. As discussed in subsection 5.1.3, the appropriate
recurrence relation to determine solution of the inhomogeneous singular BVP would be:
where and
Putting the given information above, the simplified looks like:

53
But the operator which is designed so as to treat the singularity at is defined as:
where and are the two-points of the BVP, is the free constant which plays a great role in
treating the problem with singularity feature. Hence using in results in:
which is the first term of the decomposition.
For in the second equation of the recurrence relation can be determined to

be:
This way MATHEMATICA is used to determine the rest of the required components of the
solution to get better result. To demonstrate, the next four approximate solutions are found to
be the following:
54
The first ten terms of the decomposition were used to get the IADM solution of the problem
as:
II. MADM solution
Based on MADM by Wazwaz (1999) the singular BVP can be treated as follows.
Consider It can be rewritten in the form of (5.14) given by
Multiplying both sides by the above equation becomes
Writing the left hand side in its compact form becomes:

55
Now use the standard ADM recursive relation ) in specific. In addition, using MADM by
Wazwaz (1999) minimizes the size of computations and results in the exact solution the
following way.
( )
( )
By (5.4), the exact solution becomes:
MATLAB is used to plot the graphs of the IADM and MADM solutions together in figure
below considering numerical results of the first ten components of the solutions as follows.
[0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1];
[0 0.004511209 0.009022418 0.020523838 0.032025259 0.047528057 0.063030855
0.079527513 0.096024171 0.110520476 0.125016781 0.134513954 0.144011127
0.145509085 0.147007044 0.13850554 0.128004037 0.104502898 0.081001759
0.04500879 0];
[0 0.002375000 0.009000000 0.019125000 0.032000000 0.046875000 0.063000000
0.079625000 0.096000000 0.111375000 0.125000000 0.136125000 0.144000000
0.147875000 0.147000000 0.140625000 0.128000000 0.108375000 0.081000000
0.045125000 0];
56
Figure 5.6. Comparison of IADM and MADM solutions of example 5.4.

Numerical illustrations of the problem by the methods used in this project and other related
methods are shown in table 5.2.
Approximate Exact Error

x
Solution Solution / /
0 0 0 0 0
0.1 0.009022418 0.009000000 2.3E05 2.2418E05
0.2 0.032025259 0.032000000 1.1E05 2.5259E05

0.3 0.063030855 0.063000000 5.5E05 3.0855E05
0.4 0.096024171 0.096000000 2.3E04 2.4171E05
0.5 0.125016781 0.125000000 1.1E04 1.6781E05
0.6 0.144011127 0.144000000 1.2E04 1.1127E05
0.7 0.147007044 0.147000000 1.6E04 7.0442E06
0.8 0.128004037 0.128000000 1.5E04 4.0377E06
0.9 0.081001759 0.081000000 4.1E05 1.7590E06
1 0 0 0 0
th
is the approximate solution obtained at the 26 iteration by Cui and Geng (2007).
In order to verify numerically whether the approach IADM leads to accurate solutions, the
symbolic programs, MATLAB and MATHEMATICA are used to evaluate the decomposition
series solutions using the n-terms approximation given by ( ). The approximate solution
57
is compared with the exact solution obtained by MADM both in figure

5.6 and table 5.2. The numerical results show that a very good approximation is achieved
using small values of n-terms of the decomposition series solution. It is also important to note
that the approach by Lesnic (2001) fails to overcome the singularity at for this singular
problem. Moreover, a comparison of the numerical results for the absolute errors
with that of Reproducing Kernel method by Cui and Geng (2007) are shown in table
5.2. In addition, almost the same result as the result at the 26th iteration is obtained by the
method mentioned after 51 iterations. This shows that the approaches used in this project are
not only easier and confidential but also by far more accurate.
Example 5.5. Consider the linear singular equation
[ ]
with the boundary conditions and .
I. IADM solution
Using the recurrence relation given by ( ) and the same procedure as in examples above.
With
[ ]
and hence substituting the given constants and , in these equations one can easily
get
( )
Using these expressions the first component of the decomposition of the solution u(x) can be
found to be:
58
Similar procedure as in previous examples the next four approximate solutions shown below
are obtained with the help of MATHEMATICA using the recurrence relation ( )
59
II. ADM solution
Here the ADM is supported by the Maclaurin series expansion of functions to get in to the
desired solution. The equation can be rewritten as [ ]
Using the scheme in (5.22),
( [ ] )
From this recurrence relation, one can easily obtain

60
To show the possibility of obtaining the exact solution of the given equation, the approximate
solution can be established by ( ) the following way.
61
By investigating the approximate solutions and , it can be observed that they agree
with the Maclaurin expansion of the function up to , and respectively. So, by
evaluating more terms of the decomposition series, the exact solution is found to be
The following (figure 5.7) is plot by MATHEMATICA where the IADM solutions obtained
using only the first four components of the decomposition solutions and the exact solution
(for ).
Figure 5.7. Comparison of the IADM and ADM solutions for example 5.5.
62
Table 5.3 shows the numerical illustration using the IADM and exact solutions for the
problem. Unlike the figure, the first seven terms of the decomposition solution are used.
Approximate Solution Error

Exact Solution
x / /
0 1 1 0
0.2 1.221155823 1.221402758 2.47E-04
0.4 1.491930426 1.491824698 1.06E-04
0.6 1.822019687 1.8221188 9.91E-05
0.8 2.2256288 2.225540928 8.79E-05
1 2.718281828 2.718281828 0
To verify how close the approximate solution is to the exact one: , the IADM
solution considering the first four approximate results and the exact solution are plotted in
figure 5.7 for and . It is shown from this figure that the approximate solution
obtained through the IADM is very close to the exact one using very few terms. Furthermore,
numerical results are shown in the table using the first seven IADM solutions of the example
considered. From the absolute error in the table it is observed that the IADM solution is closer
to the exact solution. Kumar used a three-point finite difference method in which a
huge amount of computational work is needed to obtain the numerical solution.
Example 5.6. Consider the nonlinear boundary value problem

I. MADM solution
To treat second order BVPs with mixed boundary conditions of this form, the operators at
(5.38) and (5.40) can be reduced, respectively, to
( )
and
63
Hence, from and one can define the operators below to treat the problem
considered.
( )
and
Rearranging and multiplying the given equation by , in an operator
form it becomes
Applying on both sides of one can get
Using the decomposition series for the linear function and the Adomian polynomial
series for the nonlinear term , the following expression can be obtained.
( )
This gives the recursive relationship
( )
Now to determine the Adomian polynomials for the nonlinear term, call it the
Adomian formula given by is used in as follows.
For ,
64
For ,
For the respective Adomian polynomials can be obtained in similar fashion.
Or, optionally, MATHEMATICA is set to give desired number of decomposition terms of any
nonlinearity types as shown in the appendix C. For instance,
For ,
MATHEMATICA displayed the corresponding Adomian polynomials in terms of the

subscript of the term for as
this means
where in this case is Hence, only by substituting the appropriate derivatives of the
nonlinear function the desired result follows for as
The same procedure gives the rest of the polynomials:
.
.
.
Substituting these results of the Adomian polynomials in to gives the components

65
.
.
.
The approximate solution of the nonlinear homogeneous boundary value problem considering
only the first three iterative results is thus approximately given using by:
II. Exact solution

The exact solution of the nonlinear BVP is given by Kim and Chun (2010) as:
The graphs of the example from the MADM and exact solutions are plotted using
MATHEMATICA in figure 5.8.
Figure 5.8. Comparison of the MADM and exact solutions for example 5.6.
66
Table 5.4 shows numerical illustrations of the example that deal with the exact solution given
by Kim and Chun (2010) and the approximate solution determined by MADM to see how
close they are using the absolute error.
MADM solution Exact Solution /Exact solution /

x
0 1 1 0
0.1 0.998335705 0.998337488 1.7830E-06
0.2 0.993392139 0.993399268 7.1290E-06
0.3 0.985313275 0.985329278 1.6003E-05
0.4 0.974326442 0.974354704 2.8262E-05
0.5 0.960725402 0.960768923 4.3521E-05
0.6 0.944850222 0.944911183 6.0961E-05
0.7 0.92706639 0.927145541 7.9151E-05
0.8 0.907745218 0.907841299 9.6081E-05
0.9 0.887246455 0.887356509 1.1005E-04
1 0.865920000 0.866025404 1.0540E-04
From the example above it can easily be observed that the modification of the decomposition
method MADM can easily treat problems that the standard Adomian method fails to be
effective, even resulting in a good agreement with the exact solution. Only the first three
iterative results of the solution are used to plot the graph. Thus, as can be seen from the figure,
the graphs overlap and that implies the results obtained by MADM are as accurate as the
results from the exact solution. In addition, numerical results are shown in table 5.4
considering only the first four iterative results of the approximate solution in order to get better
numerical result. The absolute error determined show that the method is too accurate to treat
nonlinear singular two-point boundary value problems with mixed boundary conditions.
Example 5.7. Consider the Thomas-Fermi equation
subject to the boundary conditions and .

67
IADM solution
The IADM is suitable to treat the nonlinear singular boundary value problem. Equation (5.61)
is the appropriate recurrence relation to treat the problem; and hence given as
Substituting the given terms in the first equation of this recurrence relation the following is
obtained.
Here as the first term of the decomposition is combination of two terms, it is convenient to
apply MADM by Wazwaz (1999) to facilitate the speed of convergence of the solution. Thus
the term can be broken into two parts so that the first component be the initial, i.e. and the
second to be added to the result obtained in the next iteration. So this results in and as
shown below:

[ ]
Now to simplify this for the operator is defined as

Based on the results in Appendix C where the first six terms of the Adomian
polynomials all expressed in terms of are shown as follows.

68
Now use these Adomian polynomials to determine the desired series solutions. Using

in gives:

[ ]

( ) ( )
From

[ ]

[ ]

[ ( )]

( ) ( )
Similarly the rest of the decomposition components can be obtained by the help of
MATHEMATICA. The terms and are shown as follows.

69
Since the exact solution of this problem is not known, it is preferred to consider different
term approximate solutions.
Hence, the graph is plotted in Figure by MATHEMATICA using the first few approximate
solutions of the Thomas-Fermi equation.
Figure 5.9. IADM solution of Thomas-Fermi equation in example 5.7.

70
Table 5.5 shows numerical results for the Thomas-Fermi equation at the fifth and sixth
components of the decomposition solution.
Approximate Approximate
x solution at 5th solution at 6th Absolute
iteration iteration Error
0 1 1 0
0.1 0.813047773 0.812848530 1.99E-04
0.2 0.655569258 0.655697162 1.28E-04
0.3 0.515807841 0.516712006 9.04E-04
0.4 0.391164379 0.393048295 1.88E-03
0.5 0.281443238 0.284213923 2.77E-03
0.6 0.187526564 0.190818110 3.29E-03
0.7 0.110819158 0.114074208 3.26E-03
0.8 0.052938729 0.055535968 2.60E-03
0.9 0.015508609 0.016925449 1.42E-04
1 0 0 0
The fifth and sixth approximate solutions of the example considered are plotted on figure 5.9.
Furthermore, numerical results and the absolute difference of the approximate solutions at the
5th and 6th iterative results that satisfy the original equation are depicted in table 5.5. As shown
in the figure, the two successive IADM solutions plotted and found one to overlap on the
other. Moreover, the absolute error between these results show that accurate solutions for
Thomas-Fermi equation can be obtained in few iterations. Furthermore, from the graphs one
can observe that the IADM solution converges to a certain function which may point the
direction of the exact solution. Also it should be noted that the homotopy method and the
differential transformation method cannot be applied to solve this BVP due to the existence of

the nonlinear term . So, it may be concluded that the current approach has many
advantages over the others, mainly because it can be applied directly to singular two-point
BVPs with complex nonlinearities.
Example 5.8. Consider the singular boundary value problem
with boundary conditions

71
MADM solution
By ) and the operators and can be defined as
( )
and
so that
and
The problem in an operator form becomes
Applying on both sides of one can get:
( ( ))
Hence, applying on both sides of the following is observed:

72
But the left hand side of is already shown in , thus substituting,
Rearranging and using the boundary conditions,
which is the exact solution of the BVP.
Here for the equation reduces to
with boundary condition
and the corresponding exact solution would be
Whose graph is shown below.

u
12
10
x
0.2 0.4 0.6 0.8 1.0
Figure 5.10. MADM solution for example 5.8.
As shown above, the MADM solution of the singular boundary problem is given in general
cases for a constant k. For illustrative purpose; the graph of the solution for the case is
73
plotted. Here, it is observed that beyond the ability to treat BVPs with singular feature MADM
can also be able to provide exact solutions for plenty of nonlinear BVPs with singularity
nature.
Example 5.9. Consider the nonlinear singular BVP
( )
subject to the boundary conditions and
I. IADM solution
As can be expected, it is necessary to represent the nonlinear part by the Adomian

polynomials. Here the nonlinear term is
The required recurrence relation is (5.61) which is given by:
But for fast convergence to the exact solution, MADM by Wazwaz (1999) helps to rearrange
the result obtained above so that assumes to be zero and all the existing terms obtained to
be added to the following way:
For
The Adomian polynomial can be determined either using the Adomian formula or the
results provided by MATHEMATICA at appendix C to be:
74
Adding the previous result of i.e. to one can get as:
Recall that
For ,
from the second equation of the recurrence relation would be:
( )
( )
The terms of the decomposition components are getting too vast to solve by hand but
MATHEMATICA facilitates computing. So it can be used to list as many decomposition
terms as desired. Though more than ten terms of the decomposition are used in the IADM
solution, it is believed not economical to write the next iterative results, it is found important
to list only up to the fifth iteration below.
75
Hence, the result obtained by ( is used to show the numerical illustrations in table 5.6.
II. Exact solution

Consider the given equation
76
( )
Re writing it in the form of ( ) one can get:

( ) ( )
By the recursive ( ) for ADM given by
( )

Here and but the Adomian polynomials are already determined in IADM
solution above.
From the recursive relation above it can be obtained that
For reduces to

( [ ] )

( [ ] )
Similarly using one can easily observe that

77
.
.
.
Now these results by the ADM can be used to get the series expansion of . The
solution is then given by
( )
which is the exact solution.

MATLAB is used to plot the IADM and exact solutions in the figure 5.11.
78
Figure 5.11. Comparison of the IADM and exact solutions for example 5.9.
The numerical illustrations of the problem considered are shown in table below.
Approximate Exact solution Error

x
solution
0 0.693147181 0.693147181 0
0.1 0.683195177 0.683196850 1.67E-06
0.2 0.653924627 0.653926467 1.84E-06
0.3 0.606967226 0.606969484 2.26E-06
0.4 0.544725404 0.544727175 1.77E-06
0.5 0.470002399 0.470003629 1.23E-06
0.6 0.385661666 0.385662481 8.15E-07
0.7 0.294370545 0.294371061 5.16E-07
0.8 0.198450643 0.198450939 2.96E-07
0.9 0.099820206 0.099820335 1.29E-07
1 0 0 0
79
In the overlapping plots shown above, one can easily observe that the IADM and the exact
solutions are nearly identical. Furthermore, numerical results are shown in table 5.6 in which
an absolute error is obtained. This shows that the IADM converges faster to the exact
solution. Unlike that of the DTM, in this example it can easily be observed that it is not
difficult to obtain exact solutions for nonlinear inhomogeneous BVPs using the standrad
ADM.
80
6. SUMMARY, CONCLUSION AND RECOMMENDATIONS
6.1. Summary
Linear and nonlinear STPBVPs that arise in different areas of science such as gas dynamics,
chemical reactions, thermal dynamics, etc are treated both numerically and analytically via
ADM and its latest modifications MADM and IADM.
It is difficult to get solutions for plenty of them because of their variety of categories and the
behaviors they usually depict. Numerous methods have been applied to determine numerical
solutions of these problems which of course result in variety of approximate solutions. MVIM,
DTM, FDM, ADM-PT, etc are among the methods used to treat such problems.
Mainly, this project studied second order STPBVPs both numerically and analytically using
MADM, IADM and ADM. To achieve this, it was planned to derive basic operators like
Lesnics inverse operator to treat STPBVPs, to compare the results by plotting the graphs of
the approximate and exact solutions on the same coordinate plane by using the symbolic
softwares MATLAB and MATHEMATICA and perform error analysis to see how close the
approximate solutions to the exact solutions are. All these and other related tasks were
performed accordingly by referring different library sources based on the plan.
Convective triangular fins with invariant and power-law temperature-dependent thermal

conductivities were studied by MADM. Through MATLAB and MATHEMATICA, reliable
parametric solutions for both problems was offered by the help of the modification. A realistic
model problem regarding a silicon fin of specific dimensions was carried out as a numerical
illustration. The MADM solutions were compared with an exact analytical solution by Arpaci
(1966) for the constant (governed by linear BVP) and a numerical solution via FDM by
Mokheimer (2003) for power-law temperature-dependent cases (governed by nonlinear
BVP). Moreover, it was observed that ADM and its modifications; MADM and IADM are
comfortable to treat the problems. The approximate solutions obtained by MADM and IADM
converge faster to the respective exact solutions. Some numerical approximations by IADM
are compared with the results in some other methods like reproducing kernel space method by
Cui and Geng (2007) in which magnificent power of fast convergent behavior of IADM is
observed.
81
6.2. Conclusion
Classes of linear and nonlinear singular boundary value problems can be treated through the
standard Adomian decomposition method and its modifications analytically and numerically.
All the ADM and the modifications require no perturbation, discretization and linearization to
treat two-point BVPs both numerically and analytically.
Although the classical ADM is very powerful, it fails in treating some singular boundary value
problems due to the existence of singular point at . So this difficulty is alleviated by the
modifications MADM and IADM; and shown by treating STPBVPs with Dirchlet and mixed
boundary conditions holding singular feature both numerically and analytically. It is
demonstrated that the modifications can be well suited to attain an accurate solution to the
second-order singular boundary value problems, linear and nonlinear as well. The difficulty of
those singular problems, due to the existence of the singular point at x = 0, is overcome in
these contributions. The illustrative examples also show that the standard ADM and its
modifications are very effective in providing promising results. The analyses of temperature
distribution with constant and variable thermal conductivity along the triangular fin profile
governed respectively by linear and nonlinear two-point boundary value problems and the
results obtained from the model problem confirm that fins can serve as effective coolants. But
the standard ADM has drawbacks in that it fails to provide the self-cancelling noise terms for
homogeneous cases where sum of the noise terms vanishes in the limit for inhomogeneous
case which makes it inconsistent in the area. In addition, the operators of the standard
Adomian decomposition method support intervals of only the form in which the
independent variable is defined where is positive real number. To the contrary, it was noted
that in addition to their consistency, the improved decomposition methods IADM and MADM
are effective in treating both linear and nonlinear two-point boundary value problems with
singular feature.
Though the methods are very convenient for software treatments, it is noted that the modified
decomposition methods (especially IADM) may encounter difficulties only in obtaining each
component for some complex nonlinear problems even if symbolic packages are used because
each component is obtained by cumbersome definite integrals.
82
6.3. Recommendations
Based on the analysis of Analytical and Numerical treatment for singular two-point boundary
value problems via Adomian and the improved Adomian decomposition methods the
following basic recommendations are suggested:
Two-point BVPs with single singular point were treated by the modifications. It would
be worthwhile to expand application of the operators so that the method could be used
to treat two-point BVPs with multiple singularities.
To study heat transfer from extended surfaces it is undisputable to study the
temperature distribution along the fins length. So individuals who are interested to
work on heat transfer in these areas can benefit from the results and the MATLAB
codes presented.
The ADM is not only used in a straight forward manner but also requires less
computational works in comparison to some other methods like HAM and MVIM. To
the contrary, though the series can be rapidly convergent in a very small region, it has
very slow convergence rate in wider regions and the truncated series solution is an
inaccurate solution in that region which will seriously restrict the application area of
the method. An investigation into this claim would greatly benefit the scientific
community.
83
7. REFERENCES
Abdelrazec Ahmed, H. M. 2008. Adomian decomposition method: Convergence analysis and

numerical approximations. McMaster University, Hamilton, Ontario.
Adomian, G. 1988. A Review of the Decomposition Method in Applied Mathematics.
Academic Press, Inc, 135:501-544.
Adomian, G. 1994. Solving Frontier Problems of Physics: The Decomposition Method.
Kluwer Academic Publishers.
Almazmumy, M., Hendi, F.A, Bakodah, H.O and Alzumi, H. 2012. Recent modifications of
Adomian decomposition method for initial value problem in ordinary differential
equations. American Journal of Computational Mathematics, 2(1):228-234.
Arpaci, S. Vedat 1966. Conduction heat transfer. Addison Wesley Publishing Company,
Massachusetts.
Behirya, S. H., Hashisha, H., El-Kallab,I. L., Elsaid, A. 2007. A new algorithm for the
decomposition solution of nonlinear differential equations. Elsevier, 54:459-466.
Benabidallah, M. and Cherruault, Y. (2004). Application of the Adomian method for solving a
class of boundary problems. Kybernetes, 33(1):118-132
Betaineh, A. Sami, Noorani, M. S. M., Hashim I. 2008. Approximate solutions of singular two
point boundary value problems by modified homotopy analysis method. Elsevier B.V.,
372:4062-4066.
Cen Zhongi. 2006. Numerical study for a class of singular two-point boundary value problems
using Greens functions. Applied Mathematics and Computation. 183(1):1016.
Cui, M. and Geng, F. 2007. Solving singular two-point boundary value problem in reproducing
kernel space. Journal of Computations and Applied Mathematics, 205:615.
Duan, J.-S, Rach, R., Baleanu, D. and Wazwaz, A.-M. 2012. A review of the Adomian
decomposition method and its applications to fractional differential equations.
Commun. Frac. Calc., 3(2):73-99.
Ebaid, A. 2010. Modification of Lesnics Approach and New Analytic Solutions for Some
Nonlinear Second-Order Boundary Value Problems with Dirichlet Boundary
Conditions. Zeitschrift f ur Naturforschung, Tubingen, 65a:692-296.
Hasan, Y.Q. and Zhu, L.M. 2008. Modified Adomian decomposition method for singular
initial value problems in the second order ordinary differential equations. 3(1):183193.
84
Hasan, Y.Q. and Zhu, L.M. 2009. A note on the use of modified Adomian decomposition
method for solving singular boundary value problems of higher-order ordinary
differential equations. Communications in Nonlinear Science and Numerical
Simulation, 14(8):3261-3265.
Holmquist S. M. 2007. An examination of the effectiveness of the Adomian decomposition
method in fluid dynamic applications. University of Central Florida, Orlando, Florida.
Inc, M. and Evans, D.J. 2003. The decomposition method for solving a class of singular two-
point boundary value problems. International Journal of Computational Mathematics.
80(7):869882.
Kaliyappan, M. and Hariharan, S. 2015. Symbolic computation of Adomian polynomials
based on Rachs Rule. British Journal of Mathematics and Computer Science,
5(5):562-570.
Khuri, S.A. and Sayfy, A. 2010. A novel approach for the solution of a class of singular
boundary value problems arising in physiology. Mathematical and Computer
Modeling, 52(3):626-636.
Kim, W. and Chun, C. 2010. A modified Adomian decomposition method for solving higher-
order singular boundary value problems. Z. Naturforsch, 65a:1093-1100.
Kogan, D. 2002. The use of Adomian decomposition method for solving a specific nonlinear
partial differential equation. Elazig Turkey, 9:343-349.
Kraus, A.D., Aziz, A. and Welty, J. 2001. Extended Surface Heat Transfer, John Wiley and
Sons, Inc., New York.
Kumar, M. 2002. A three-point finite difference method for a class of singular two-point
boundary value problems. Journal of Computations and Applied Science, 145:8997.
Lesnic, D. 2001. A computational algebraic investigation of the decomposition method for
time-dependent problems. Elsevier Science Inc., 119:197-206.
Lin, Y. and Chen, C.K. 2014. Modified Adomian decomposition method for double singular
boundary value problems. Rom. Journal of Physics, Bucharest, 59(5-6):443-454.
Martin, Jr. 1983. Ordinary differential equations. Mc Graw-Hill book company.
Mokheimer Esmail, M.A. 2003. Heat transfer from extended surfaces subject to variable heat
transfer coefficient. Springer-Verlag, 39:131-138.
85
Noor, M.A. and Mohyud-Din, S.T. 2008. Solution of singular and non-singular initial and
boundary value problems by modified variational iteration method. Hindawi
Publishing Corporation, 1-23.
Ravi, A.S.V. and Aruna, K. 2008. Solution of singular two-point boundary value problems
using differential transformation method. (A372):4062-4066.
Singh, R., Kumar, J., and Nelakanti, G. 2012. New approach for solving a class of doubly
singular two-point boundary value problems using Adomian decomposition method.
Hindawi Publishing Corporation, 1-23.
Wazwaz, A.-M. 1999. A Reliable Modification of Adomian Decomposition Method. Applied
Mathematics Computation, 102(1):77-86.
Zavalani, G. 2015. Galerkin-finite element method for two point boundary value problems of
ordinary differential equations. Science Publishing Group, 4(2):64-68.
86
8. APPENDICES
8.1. Appendix A
MATLAB code to get the decomposition terms of the recurrence relation ( ).
clear all %clears previously declared symbols.

clc %clears page.
syms L m x xx s Thetha_b %syms declares the variables used to be symbolic.
nth=input('How many decomposition terms do you want to include in your solution? ');
%invites to insert the number of decomposition terms needed to be used in the decomposition.
f=1; s=0;
for n=1:nth
s=s+f;
disp(sprintf('%s%d', 'Thetha_', n-1,'='))%displays each component of the decomposition with
%their respective values.
disp(f*Thetha_b)
f=int((1/xx)*int(m^2*f,x,0,xx),xx,L,x);% MADM operator in the form int(expr,var,a,b); it
%computes definite integral.
end
solution=s*Thetha_b %finally this displays the sum of the results from the for loop.
8.2. Appendix B
MATLAB codes to get the decomposition terms of the recurrence relation ( ).
1. For
% Function Ak, returning the kth component of the Adomian polynomials corresponding to
% the nonlinearity A.
% Beginning
function Ak =f(k)
syms x s h
87
sym('u0(x)');sym('u1(x)');sym('u2(x)');sym('u3(x)');sym('u4(x)');sym('u5(x)');sym('u6(x)');sym(
'u7(x)');sym('u8(x)');sym('u9(x)');sym('u10(x)');sym('u11(x)');sym('u12(x)');sym('u13');sym('u1
4');sym('u15');sym('u16'); sym('u17');sym('u18'); sym('u19'); sym('u20');
s='u0(x)'+h*'u1(x)'+h^2*'u2(x)'+h^3*'u3(x)'+h^4*'u4(x)'+h^5*'u5(x)'+h^6*'u6(x)'+h^7*'u7(x)'
+h^8*'u8(x)'+h^9*'u9(x)'+h^10*'u10(x)'+h^11*'u11(x)'+h^12*'u12(X)'+h^13*'u13'+h^14*'u1
4'+h^15*'u15'+h^16*'u16'+h^17*'u17'+h^18*'u18'+h^19*'u19'+h^20*'u20';
Ak=(1/factorial(k))*subs(diff(1/s*(diff(s,x))^2,h,k),h,0);
% The End
2. For
% Function Bk, returning the kth component of the Adomian polynomials corresponding to
% the nonlinearity B.
% Beginning
function Bk =f(k)
syms x s h Z
+h^8*'u8(x)'+h^9*'u9(x)'+h^10*'u10(x)'+h^11*'u11(x)'+h^12*'u12(x)'+h^13*'u13'+h^14*'u14
'+h^15*'u15'+h^16*'u16'+h^17*'u17'+h^18*'u18'+h^19*'u19'+h^20*'u20';
Bk=(1/factorial(k))*subs(diff(s^(1-Z),h,k),h,0);
% The End
3. For
% Function Ck, returning the kth component of the Adomian polynomials corresponding to
the %nonlinearity C.
% Beginning
function Ck =f(k)
syms x s h Z
88
+h^8*'u8(x)'+h^9*'u9(x)'+h^10*'u10(x)'+h^11*'u11(x)'+h^12*'u12(x)'+h^13*'u13'+h^14*'u14
'+h^15*'u15'+h^16*'u16'+h^17*'u17'+h^18*'u18'+h^19*'u19'+h^20*'u20';
Ck=(1/factorial(k))*subs(diff(s^(-Z),h,k),h,0);
% The End
4. Code for the inverse operator
% Function Linverse, returning inverse transformed function of its input.

% Beginning
function Linverse =f(y)
syms L x xx;
Linverse = int(1/xx*int(y*x,x,0,xx),xx,L,x);
% The End
5. The main code that determines the final solution.
% The Core Code

% Beginning
clc
clear all
syms U x u0 T0 T1 T2 BETA h L k0 b Tinf Tb Z sumsol
T0=Tb;
T1=-BETA*Linverse(subs(Ak(0),{sym('u0(x)')},{T0}))+2*h*L/k0/b*Linverse(subs(Bk(0)/x,
sym('u0(x)'),{T0}))-2*h*L*Tinf/k0/b*Linverse(subs(Ck(0)/x,sym('u0(x)'),{T0}));
T2=-BETA*Linverse(subs(Ak(1),{sym('u0(x)'),sym('u1(x)')},{T0,T1}))+2*h*L/k0/b*
Linverse(subs(Bk(1)/x,{sym('u0(x)'),sym('u1(x)')},{T0,T1}))-2*h*L*Tinf/k0/b*Linverse
(subs(Ck(1)/x,{sym('u0(x)'),sym('u1(x)')},{T0,T1}));
T3=-BETA*Linverse(subs(Ak(2),{sym('u0(x)'),sym('u1(x)'),sym('u2(x)')},{T0,T1,T2}))+
2*h*L/k0/b*Linverse(subs(Bk(2)/x,{sym('u0(x)'),sym('u1(x)'),sym('u2(x)')},{T0,T1,T2}))-
2*h*L*Tinf/k0/b*Linverse(subs(Ck(2)/x,{sym('u0(x)'),sym('u1(x)'),sym('u2(x)')},{T0,T1,T2}));
T4=-BETA*Linverse(subs(Ak(3),{sym('u0(x)'),sym('u1(x)'),sym('u2(x)'),sym('u3(x)')},
{T0,T1,T2,T3}))+2*h*L/k0/b*Linverse(subs(Bk(3)/x,{sym('u0(x)'),sym('u1(x)'),sym('u2(x)'),
89
sym('u3(x)')},{T0,T1,T2,T3}))-2*h*L*Tinf/k0/b*Linverse(subs(Ck(3)/x,{sym('u0(x)'),sym
('u1(x)'),sym('u2(x)'),sym('u3(x)')},{T0,T1,T2,T3}));
T5=-BETA*Linverse(subs(Ak(4),{sym('u0(x)'),sym('u1(x)'),sym('u2(x)'),sym('u3(x)'),sym
('u4(x)')},{T0,T1,T2,T3,T4}))+2*h*L/k0/b*Linverse(subs(Bk(4)/x,{sym('u0(x)'),sym('u1(x)'),
sym('u2(x)'),sym('u3(x)'),sym('u4(x)')},{T0,T1,T2,T3,T4}))-2*h*L*Tinf/k0/b*Linverse(subs
(Ck(4)/x,{sym('u0(x)'),sym('u1(x)'),sym('u2(x)'),sym('u3(x)'),sym('u4(x)')},{T0,T1,T2,T3,T4}));
('u4(x)'),sym('u5(x)')},{T0,T1,T2,T3,T4,T5}))+2*h*L/k0/b*Linverse(subs(Bk(5)/x,{sym('u0(
x)'),sym('u1(x)'),sym('u2(x)'),sym('u3(x)'),sym('u4(x)'),sym('u5(x)')},{T0,T1,T2,T3,T4,T5}))-
2*h*L*Tinf/k0/b*Linverse(subs(Ck(5)/x,{sym('u0(x)'),sym('u1(x)'),sym('u2(x)'),sym('u3(x)'),s
ym('u4(x)'),sym('u5(x)')},{T0,T1,T2,T3,T4,T5}));
('u4(x)'),sym('u5(x)'),sym('u6(x)')},{T0,T1,T2,T3,T4,T5,T6}))+2*h*L/k0/b*Linverse(subs(Bk
(6)/x,{sym('u0(x)'),sym('u1(x)'),sym('u2(x)'),sym('u3(x)'),sym('u4(x)'),sym('u5(x)'),sym('u6(x)'
)},{T0,T1,T2,T3,T4,T5,T6}))-2*h*L*Tinf/k0/b*Linverse(subs(Ck(6)/x,{sym('u0(x)'),sym
('u1(x)'),sym('u2(x)'),sym('u3(x)'),sym('u4(x)'),sym('u5(x)'),sym('u6(x)')},{T0,T1,T2,T3,T4,T5,T6}));
L=0.05;Tinf=273+25;Tb=273+150;h=4;BETA=-1.3;Z=-1.3;k0=148*300^-BETA;b=0.005;
disp('The solution by MADM is: ');
sumsol=T0+T1+T2+T3+T4+T5+T6+T7
% The End
8.3. Appendix C
The following is a MATHEMATICA program to get any number of the Adomian

polynomials of any type of nonlinearity developed based on the subscripts of the term .
Based on the code, the first six terms of the Adomian polynomials in Wolfram language are
displayed in the following way:
90
F[0]
These results displayed above by MATHEMATICA would generally mean respectively that

Ashenafi Agizaw

Uploaded by

Copyright:

Available Formats

Ashenafi Agizaw

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Ashenafi Agizaw

Uploaded by

Copyright:

Available Formats

ANALYTICAL AND NUMERICAL TREATMENT FOR SINGULAR

TWO-POINT BOUNDARY VALUE PROBLEMS VIA ADOMIAN AND

A Project submitted to the Department of Mathematics,

In Partial Fulfillment of the Requirements for the Degree of

Seleshi Demie (PhD) __________________ ___________________

________________ ________________ ________________

________________ ________________ ________________

Name: Ashenafi Agizaw Jije

Acronyms and Abbreviations

STATEMENT OF THE AUTHOR IV

3. MATERIALS AND METHODS 12

5. SINGULAR TWO-POINT BVPS VIA ADM, IADM AND MADM 21

6. SUMMARY, CONCLUSION AND RECOMMENDATION 80

5.1. A wall transferring heat by convection 39

1.1. Background of the study

A system of ordinary differential equations has many solutions. Commonly a solution of

A differential equation that could be reduced to

, sometimes written in the form which results from

1.2. Statement of the problem

The study explored the specific objectives:

2.1. Extended surfaces

2.2. Numerical and Analytical treatments of the BVPs

2.3. Treatments of related problems using ADM and its modifications

Lin and Chen proposed a modification of the Adomian decomposition method,

3. MATERIALS AND METHODS

4.1. Singular points

It is of interest to solve the differential equation

or, in standard form,

Definition 4.1. A function is said to be analytic at a point if can be expanded in a

Definition 4.3. The point is called a singular point of if

This causes difficulties in writing in the standard form

where and are analytic at . This is the same as saying that

A singular point which is not regular is called irregular singular point.

Remark: If and are analytic at and , then is analytic at .

Example 4.1. Given the Legendre differential equation

The points are the zeros of and are clearly,

Here, it can be observed that the multiplier of has the form

and the multiplier of has the form

Example 4.2. Determine all the singular points of

Here, it can easily be observe that

4.2. Boundary value problems

A boundary value problem can generally be denoted as

The differential equation at boundary conditions is known as linear BVP.

Linear BVPs can mainly be classified as homogeneous and inhomogeneous.

Example 4.3. Solve the BVP

The general solution can be obtained to be

And from the boundary conditions given

i.e. infinitely many solutions

Recall that a point is called a singular point of

Example 4.4. The BVP

with and is singular at Hence the two-point BVP is a good example

An operator is a function that takes a function as an argument instead of numbers as we are

Definition 4.8. An operator satisfying property is called linear operator.

Example 4.10. The differential equation

can be re written in compact form as

So from this an operator can be generated to be:

so that can be written in an operator form as:

Thus the operator at is linear operator.

5. SINGULAR TWO-POINT BVPs VIA ADM, IADM AND MADM

Seleshi Demie (PhD) _