Adama Science and Technology University

School of Applied Natural Science

Department of Applied Mathematics

ACCELERATED FINITE DIFFERENCE METHOD FOR

SINGULARLY
PERTURBED TWO-PARAMETER BOUNDARY VALUE
PROBLEMS
BY : BEREKET BOGALE
ID:Ugr/23281/13

PROPOSAL SUBMITTED TO DEPARTMENT OF APPLIED MATHEMATICS

IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR DEGREE OF
BACHELOR OF ARTS (BA) IN MATHEMATICS

February 2024
Adama, Ethiopia

Contents
Chapter 1......................................................................................................................................................1
Introduction..................................................................................................................................................1
1.1 Background of the Study.......................................................................................................................1
1.3 Objectives of the Study..........................................................................................................................3
1.3.1 General Objective...........................................................................................................................3
1.3.2 Specific Objectives..........................................................................................................................3
1.4 Significance of the Study.......................................................................................................................3
1.5 Delimitation of the Study......................................................................................................................3
Chapter 2......................................................................................................................................................4
Literature Review........................................................................................................................................4
2.1 Two Parameter Singularly Perturbed Problems.................................................................................4
2.2 Finite Difference Method (FDM)......................................................................................................5
2.3 Richardson Extrapolation Method...................................................................................................5
2.4 Numerical versus Analytical Methods.............................................................................................6
Chapter 3......................................................................................................................................................7
Methodology.................................................................................................................................................7
3.1 Study design........................................................................................................................................7
3.2 Sources of Information......................................................................................................................7
3.3 Mathematical Procedures..................................................................................................................7
Chapter four.................................................................................................................................................8
Work and Budget Plan................................................................................................................................8
4.1 Work plan...........................................................................................................................................8
4.2 Budget plan.........................................................................................................................................9
Reference....................................................................................................................................................10
Chapter 1

Introduction

1.1 Background of the Study

A cliche among experts in the subject, numerical analysis is a discipline of mathematics that combines
science and art. However, non-specialists frequently misunderstand it. As a scientific discipline,
numerical analysis studies how arithmetic can be used to solve mathematical issues. Approximations are
used to replace quantities that cannot be calculated arithmetically, such as integrals or derivatives,
allowing an approximate solution to be determined. Numerical analysis is an art form that involves
selecting the process (and appropriately implementing it) that best fits the solution to a given problem.

Numerical analysis is important because it allows us to solve issues that are hard to answer analytically
and provides an approximation of a solution. Numerical techniques are most frequently used in
computational mathematics to solve equations from the sciences, engineering, and physics. Numerical
algorithm creation and computing is one of mathematics’ problems. However, researchers in
computational mathematics are working to create computer-based numerical techniques for future use
(Burden and Faires, 2011).

Differential equations can be classified as boundary value problems or initial value issues based on how
many points have conditions applied to them. The distinction between initial value problems and
boundary value problems does not exist for first order differential equations because only one condition
is stated. We looked at boundary value problems, which are differential equations with conditions
applied at various places, for our purposes.

Boundary value problems can be solved numerically using methods employing finite differences, which
substitute an appropriate difference approximation for each derivative in the differential equation. To
preserve a given order of truncation error, the specific difference approximation and the step size h are
used.

Furthermore, the way the FDM functions is by substituting a finite grid, also known as a mesh, of points
where the dependent variable is estimated for the region over which the independent variable in the
differential equation is defined. The most crucial instrument for analyzing finite difference approaches is
the Taylor expansion. Here, we use the finite difference formulae at grid points to represent (about) the

1
derivatives of the differential equation using Taylor expansion, creating a linear algebraic system that can
be solved using the Thomas Algorithm. There are many various types of formulas to take into account,
but generally speaking, the accuracy of these formulas is directly proportional to the magnitude of h,
which is usually small (Li et al., 2018).

1.2 Statement of the Problem

In the later part of the 1970s, there was a concerted investigation into numerical techniques for solving
solitary perturbation issues. The research horizon has continued to grow since then, but the
presentation of fresh advancements in the analysis of these numerical techniques has not gotten the
attention it deserves.

The fact that singly perturbed differential equations are present in many cases is one of the primary
motivations for studying them. Formula In fact, the actual occurrence of boundary layers is theoretically
explained by this characteristic of equations. Boundary layer mathematical manifestations occur when
minor parameters multiply the coefficients of some or all of the differential equations' highest-order
derivatives; these equations are referred to as singularly perturbed. Singular perturbation problems
frequently arise in a variety of applied mathematics and engineering fields. These kinds of issues are
commonly identified by multiplying some or all of the higher-order terms in the differential equation by
an arbitrarily small quantity known as the perturbation parameter.

The differential equation's nature entirely changes in the limit scenario, when the singular perturbation
parameter equals zero, which is why the term "singular perturbation" was coined. For instance,
nonlinear hyperbolic equations replace nonlinear parabolic equations in the conservation of energy and
momentum equations (Miller et al. 2012). OMalley (OMalley, 1967) asymptotically analyzed the nature
of the two-parameter issue, finding that the ratio of µ to ε plays a substantial role in the solution by
forming a relationship that directly influences the layer structure at the boundary points x = 0 and x = 1.
(Shishkin and Tivot, 1976) solved the two-parameter singular perturbed boundary value problem using
an exponential fitted finite difference scheme on a uniform grid.

The name "singular perturbation" was coined because, in the limit case, the differential equation's
nature completely changes when the unique perturbation parameter equals zero. For example, in the
equations for the conservation of momentum and energy, nonlinear hyperbolic equations take the place
of nonlinear parabolic equations (Miller et al. 2012). By directly influencing the layer structure at the
boundary points x = 0 and x = 1, the ratio of µ to ε plays a significant role in the solution, according to

2
OMalley's asymptotic analysis of the two-parameter problem (OMalley, 1967). (Shishkin and Tivot, 1976)
used an exponential fitted finite difference technique on a uniform grid to solve the two-parameter
singular perturbed boundary value problem.

This means that there is always room for improvement in the numerical approach to singularly
perturbed boundary value issues. In this work, the singularly perturbed boundary value problem with
two parameters was solved using the accelerated finite difference approach. To improve the
approximation solution's accuracy and rate of convergence, the Richardson extrapolation technique is
used to an approximation scheme produced by the finite difference method.

1.3 Objectives of the Study

1.3.1 General Objective
The general objective of this study is to formulate an accelerated finite difference method for solving the
second-order two-parameter singularly perturbed boundary value problem.

1.3.2 Specific Objectives

The specific objectives of the study are

 To develop an order finite difference method for solving two parameters singularly perturbed
boundary value problems.
 To apply the Richardson extrapolation method to increase the rate convergence of the method.

1.4 Significance of the Study

The results obtained in this research might help to:
 Provide some background information for other researchers who work in this area.
 Introduce the development of an accelerated finite difference method for solving two parameter
singularly perturbed boundary value problems.

1.5 Delimitation of the Study

We are going to delimit to develop an accelerated finite difference method for solving two
parameters singularly perturbed boundary value problems of the form
−εy00(x) + µa(x)y
(x) + b(x)y(x) = f(x), x ∈ (0, 1),
subject to the boundary conditions

3
y(0) = α, y(1) = β,
where ε and µ are perturbation parameters such that 0 < ε, µ 1, a(x), b(x) and f(x) are sufficiently smooth
and bounded functions with a(x) ≥ a
∗ > 0 and b(x) ≥ b
∗ > 0 for
x ∈ [0, 1].

Chapter 2

Literature Review

2.1 Two Parameter Singularly Perturbed Problems

At a non-linear vibrations seminar at New York University in 1946, Friedrichs and Wasow coined the
phrase "singular perturbation." Numerous fields of applied mathematics and engineering, such as fluid
mechanics, chemical reactor theory, elasticity, gas porous media, electrode theory, and heat and mass
transfer processes in composite materials with limited heat conduction and diffusion, frequently
encounter singular perturbation problems (SPPs). In most cases, the higher order terms in the
differential equation are multiplied by an arbitrary tiny parameter known as the perturbation to identify
these kinds of difficulties. The analytical solutions to these kinds of problems display multi-scale
phenomena, meaning that they behave smoothly away from the boundary layer and fluctuate quickly in
a small area known as the boundary layer (Pratibhamoy,2013)

Mathematicians, engineers, and physicists are interested in singularly perturbed differential equations
with arbitrarily small parameters, µ, ε (say), where 0 < ε, µ ~ 1. It is commonly known that slowly and
rapidly varying components characterize the solution of singularly perturbed boundary value issues.
According to Akram and Afia (2013), there are thin transition layers where the solution can leap abruptly,
but outside of the layers, the solution fluctuates gradually and behaves consistently. Higher-order issues
that are singularly perturbed are categorized according to how the original equation's order is impacted
if one sets ε = 0, where ε is a tiny positive parameter that multiplies the differential equation's maximum
derivative. We classify the problem as being of the convection-diffusion type if the order is decreased by
one, and as being of the reaction-diffusion type if the order is lowered by two. In this research, we have
examined a differential equation that is frequently referred to as a singularly perturbed boundary value
issue. This problem involves multiplying both the first and second derivatives by two tiny parameters, as

4
seen in equations (1.1) − (1.2). We have used an accelerated finite difference method to provide an
approximate solution to this problem.

2.2 Finite Difference Method (FDM)

Boundary Value problems (BVPs) may be simple enough for analytical solutions in the case of basic
physical systems; nevertheless, most physical processes of interest may be rather sophisticated, and the
mathematical descriptions of these processes (BVPs) may be complex enough to prevent analytical
solutions. In this situation, the mathematical description could be simplified to the point where
analytical answers are feasible. The simplified forms in this method might not accurately represent the
behavior in real life, and it might not always be able to simplify things.

In the second option, we completely give up on analytical solutions as a workable way to resolve difficult
real-world issues and instead use numerical techniques to produce numerical BVP answers. The
numerical solution of BVPs can be obtained in a variety of ways. One of these techniques for solving
BVPs numerically is the finite difference approach, which is the focus of this investigation. This is perhaps
the reason that all numerical methods for BVPs are referred to as methods of approximation.

All the other approaches, regardless of which one is used to generate the numerical solutions of the
BVPs, have one thing in common: the numerical solution that is obtained in each case is an
approximation of the real or theoretical solution of the BVP The breadth of BVPs in the sciences,
engineering, and mathematical physics is immense. The research, creation, and use of problem-
dependent computational algorithms are completely non-general and never-ending. In addition to being
approximation methods, numerical methods for BVPs share another characteristic with other methods:
the BVPs defined by the differential operators need to be transformed in some way into an algebraic
system of equations. How various approximation techniques do this vary. The kind of differential
operators and the approximation technique determine the type of coefficient matrix in the resulting
algebraic systems. The approximate solution is given by the algebraic system of equations' solution.
Therefore, while this procedure appears straightforward in theory, how this is carried out distinguishes
one approximation approach from the others in terms of inventiveness (Surana et al., 2017).

2.3 Richardson Extrapolation Method

Richardson extrapolation is a method that yields a better approximation by averaging two computed
answers from different values of the parameter (step size), h. When it is known that an approximation
method has an error component with a predictable form, one that depends on a parameter h,

5
extrapolation can be used (Burden and Faires, 2011). This method can be used to solve multidimensional
problems. This section will demonstrate the Richardson extrapolation process, which is used for finite
difference approximations to derivatives.

Investigating the use of Richardson extrapolation as a way to speed up convergence and increase the
precision of the finite difference method we utilized is one of the goals of this study (Munyakazi et al.
2018).

This method was used for a singularly perturbed reaction-diffusion problem using a Bakhvalov mesh by
Vulanovic et al. in 1986. (Natividad and Stynes, 2003) looked at its application to the numerical solution
of the singularly perturbed convection-diffusion problem.

(Mohapatra and Natesan, 2008) developed a second-order post-processing method based on this
concept to solve a delay differential equation that is singly disturbed. A Richardson extrapolation was
also used by Reinhard (1989) to improve the precision of the difference solutions of partial differential
equations that can be enlarged by even powers of the discretization parameter h.

A fascinating article about the background and uses of extrapolation can be found in Joy (1971).

In this case, fourth-order Finite difference with the Richardson extrapolation technique in the singularly
perturbed two-parameter boundary value problems with uniform mesh to reduce the nodal error, and
improve the accuracy and rate of convergence.

2.4 Numerical versus Analytical Methods

An analytical solution is a precise answer to an issue that may be determined symbolically by equation
manipulation. Manipulating is the act of working, operating, or handling by hand. However, it becomes
extremely difficult to discover an accurate solution, or an analytical solution, for higher-order or non-
linear differential equations with complex coefficients. Therefore, to solve these equations, numerical
techniques are required. An equation's approximate solution is provided via numerical methods. The
ability to derive a numerical solution from an analytical problem even in the absence of an analytical
answer is a significant benefit of numerical analysis. It is crucial to understand that whereas analytical
approaches typically produce a conclusion in terms of mathematical functions that can be assessed for
cases, numerical solutions are always numerical. Nonetheless, a plot of the numerical data helps
illustrate some of the solution's behavior.

6
The fact that numerical analysis yields approximate conclusions even if results can be made as accurately
as needed is another significant feature. The singularly perturbed boundary value issue for ordinary
differential equations can be solved numerically using a range of techniques. To solve singular perturbed
boundary value problems with two parameters, we have developed a numerical method in this study
that we refer to as the accelerated finite difference method. The Richardson extrapolation technique has
been used to accelerate the rate of convergence from the fourth order to the sixth order.

Chapter 3

Methodology
This chapter consists of the study area and period, study design, source of information, and mathematical
procedures.

3.1 Study design

This research is going to be conducted by mixed design, i.e. documentary review design and experimental
design.

3.2 Sources of Information

Books, journals, published articles, and the internet are some of the sources that we are going use to
perform this research. Experimental results will be obtained by MATLAB software.

3.3 Mathematical Procedures

We collected the necessary information using documentary review and design, and then. we analyzed I.
Thus, to achieve the stated objectives, the study followed the following steps.
1. Describing the problem,
2. Discretizing the solution domain,
3. Replacing derivatives in differential equations by finite difference approximations and writing in tri-
diagonal system form,
4. Apply Richardson extrapolation method,
5. Establishing the stability and convergence of the obtained scheme,
6. Writing MATLAB code for the obtained scheme,
7. Verification of the scheme by using numerical examples and results.

7
 Chapter four

Work and Budget Plan

4.1 Work plan

The periodical evaluation of the development of the activities of the thesis ,designing a main time
schedule is worthwhile
Table 4.1 the time schedule of thesis work

No Task Sep- Nov Dec Jan Feb Mar -apr May -jun
1 Reading different literature √
and looks to the title
2 Title selection and reading √ √
relevant materials
3 Preparation of the proposal √ √
4 Submission of final proposal √ √
5 Proposal presentation √
6 Analyzing and writing the √
whole thesis
7 Submission of the first draft of √
the thesis
8 Editing the draft of the thesis √
based on the thesis
9 Final report submission √
10 Presentation of the thesis √

8
4.2 Budget plan
No Material unit Quantity Unit Total
required price(ETB) price(ETB)
Stationary services
Final binding copy 5 227 1,135.00
Subtotal 1 1,135.00
1 Personal expense and transportation
Per diem for Days 25 549 13,725.00
2 researcher
Starting from
Adama to
Addis Ababa
Payment for persons 3 200 6,000.00
data collectors
for 10 days
Vehicle rent Terms 3 1380 4,140.00
Subtotal 2 23,865.00
Total = subtotal 1+ subtotal 2 25,000.00

9
Reference
 D. Kumar (2012), Finite difference scheme for a singularly perturbed convection-diffusion
problem with two small parameters. Math. Aeterna,Vol. 2,no. 5, 441-458.
 D.C Joy (1971), Survey of extrapolation process in numerical analysis. SIAM Review 13,No.4,
435-490,QAI.S2 178.
 D. Pratibhamoy (2013), Robust Numerical schemes for singularly perturbed boundary value
problems on Adaptive meshes. Indian Inst. Technology, Math. Dept. No.08612305.
 E.P. Doolan, J.J.H. Miller, and W.H.A.Schilders (1980), Uniform Numerical Methods For
Problems With Initial And Boundary Layers. Dublin Boole Press.
 F. Reinhard (1989), On Richardson extrapolation for finite difference methods on regular grids,
Numer.Math. 55, 451-462.
 G. Akram, and N. Afia (2013), Solution of fourth order singularly perturbed boundary value
problem using septic spline, Middle-east Journal of Scientific Research,15 (2): 302- 311.
 G.I. Shishkin, and V.A. Titov (1976), A difference scheme for a differential equation with two
small parameters multiplying the derivatives, Numer. Method Contin. Medium Mech. 7, 145155.
 G.D. Smith (1985), Numerical solution of Partial differential equations Finite difference methods,
Third edition, Oxford University Press, New york.
 H.G. Roos, and Z. Uzelac (2003) The SDFEM for a convection-diffusion problem with two small
parameters, Comput. Method Appl. Math. 3(3), 443458.
 JJ. Miller, E.ORiordan and G.I.Shishkin (2012), Fitted numerical method for singular
perturbation problems, Revised edition, World Scientific, Singapore.
 K.O. Friedrichs, and W.Wasow (1946), Singular perturbation of non-linear oscillations. Duke
Math.J.,13: 367-381.
 M.C. Natividad, M. Stynes (2003), Richardson extrapolation for a convection-diffusion problem
using a shishkin mesh. Appl. Numer. Math. 45, 315329

10