Chapter 1

Introduction

Mathematics compares the most diverse

phenomena and discovers the secret analo-
gies that unite them

Joseph Fourier

Linear/Non-linear (NL) differential and integral equations (D&IEs) and delay differential equa-
tions have a dominant role in mathematics, science and engineering. The solutions of NL DEs
are determined by some particular techniques such as perturbation techniques, linearization tech-
niques, and using some numerical methods. In order to solve NL DEs with the help of perturbation
methods a small parameter ε is introduced. The parameter ε is taken as a coefficients of NL term.
The main concept to apply the perturbation method is the consideration of the solution in a series
form. The series is in the form of the powers of ε and the solution components, which have to
be determined with cleaver substitution and by equating the terms having identical powers of ε.
In spite of it linearization techniques been used for the solution of NL DEs. As the Bernoulli
differential equation which is NL in nature and solved by converting it in a linear form.

Although, some numerical methods like Adomian decomposition method (ADM), An iterative
method (AIM), and their modifications are useful to solve NL DEs. The numerical methods are
based on the assumption that the solution of the given problem is in a converging series form
of solution components. These solution components are determined with the help of recurrence
relation. In ADM the NL term is approximated by some special polynomials called Adomian
polynomials. In AIM the NL term is approximated with the help of a special arrangement.

Riccati differential equations (RDEs) are first order NL DEs, which appear in conventional
time dependent quantum mechanics (QM), suppersymmetric QM. The role of RDEs is not limited
in the field QM. Suppose a set of DEs describe optimal control for a system. The solution of such
system is obtained by optimizing a quadratic cost function. In order to optimize solution we have
to solve a RDEs. In spite of it the solution of RDEs is valuable in missile guided control system
and thermodynamics [4, 5, 6]. The form of RDEs solved in this work, is written as

y0 (t) − P(t)y(t) = f (t) + Q(t)y2 (t), y(a) = ya , (1.1)

1. I NTRODUCTION 3

where P(t), f (t), Q(t) are some known continuous functions in interval [a, b]. If a particular solu-
tions of Eq.(1.1) is given or known, then the associated analytical solution can be determined. As
well, every RDE of the form Eq.(1.1) can be converted into a self adjoint differential equation [7],
and the respective general solutions can be obtained in some of the particular cases. Unfortunately,
there is no any general method to obtain the analytical (or closed form) solution of Eq.(1.1). In
any case, some numerical methods like ADM, Homotopy perturbation method, iterated He’s Ho-
motopy perturbation method, He’s variation iteration method, Laplace Adomian decomposition
method with Padè approximation (LADMP) [8, 9, 10, 11, 12, 13], etc. have been presented to find
an approximate solution of Eq.(1.1).

Although, these methods give pretty good approximate solutions. However, either the compu-
tation of undetermined coefficients or implementation of multiple operations is required to solve
Eq.(1.1). Moreover, these methods are also very restrictive to solve Eq.(1.1), suppose with vari-
able coefficients, they have their own advantage and limitations. Therefore in chapter 4, Green’s
Decomposition Method with Padè approximation (GDMP) was proposed to solve eq.(1.1) and ob-
served that the computed results are more accurate and reliable than previously proposed methods.
In spite of it a Modified Green’s Decomposition Method with Padè approximation (MGDMP) is
also proposed to increase the efficiency of GDMP. The applications of RDEs are not limited in
the field of mathematics, science and engineering. The RDEs have a huge role in the field of
biotechnology.

Genome integrity of an organism has the most important role in his successful healthy life
and several immunological, developmental as well other health issues are also associated with it.
Therefore, disintegration of genome has various lethal effect on the organism. One known major
factor to disintegrate the genome is DNA double-strand breaks (DSBs). DSBs are a form of dan-
gerous lesions which have the both broken strands of DNA duplex structure. It is also considered
as a leading operator element for life threatening cancer [14, 15]. DSBs can be originated by two
major ways namely exogenous and endogenous. Exogenous sources are ionizing radiation (IR)
and some chemicals while endogenous sources like DNA replication, V(D)J recombination and
meiotic exchange may arise DSBs [16]. Various DSB repair mechanisms are evolved by organ-
isms to prevent their lethal effects. Unrepaired DSBs may cause of induced apoptosis, cell cycle
arrest or mitotic cell death [17] while incorrectly repaired DSBs may drive cancer by inversions,
deletions or translocation [18, 19]. Two extensively studied pathways to repair DSBs are homolo-
gous recombination (HR) and non-homologous end-joining (NHEJ) [20]. NHEJ characterized by
blocking of 5 end resection and close vicinity of heterodimer protein Ku70-Ku80 with DSB ends.

In NHEJ, DSB ends are directly ligated and sometimes errors make small deletions, insertions
and substitutions at breaks while translocations are made by joining of DSBs from different parts
of genome [21]. On the other hand, HR is considered as error free and commenced during the
re-sectioning of DSBs by enzymes like helicases and nucleases. It creates 3 single stranded DNA
overhangs and a complex structure forms with RAD51 recombinase. Consequently, in repair DNA
synthesis, a template homologous duplex DNA molecule is evaded by this complex structure [22].
The approach of mathematical modelling to handle such biological problems is a supplementary
venture and various associated models of biological descriptions are presented time to time, see
1. I NTRODUCTION 4

[23, 24, 25].

In addition, a few comprehensive mathematical models are also designed for DSBs [26, 27, 28].
However, DSBs are remain a puzzle and the presented solutions of aforesaid models need a more
efficient solutions. Therefore, here still a requirement for further advancement and more specific
mathematical approach to deal with DSBs in an impressive manner. For this purpose, we are using
the average number of DSBs per cell at time t derived as elsewhere [3, 29, 30],

1
U 0 (t) = δ R − U(t) − γU 2 (t), U(a) = Ua , (1.2)
τ

where U(t) average number of DSBs per cell at time t, δ is average number of DSBs induced
per unit dose, R is radiation dose rate, τ is a repair time constant, and γ is a binary reaction rate
constant in the sence of mass-action chemical kinetics. In this work, average number of double
strand breaks per cell time are given in the form of RDEs, the model formed is known as DNA-
Repair model (DRM), which have to be solved numerically. The problem is important because
in a small time interval the solution of the problem given by most renowned methods like ADM,
Laplace Adomian Decomposition Method with Padè Approximation (LADMP) gives huge error.
Therefore in this work, we use a new class of Adomian polynomials in ADM to solve DRM.
The convergence analysis of the method has also been discussed with the help of rearrangement
theorem. An example of DRM is solved numerically, Which results with good accuracy in the
comparison of aforesaid methods.

AIM [31] has been a clinical approach to solve NL D&IEs numerically with the help of
solution components. The method is easy to use as the NL term approximated by a rearrangement
of the series of solution component’s functions. A number of problems and some modifications
in the renowned methods [32, 33, 34] have been tackled with the help of AIM. Although, AIM
is effective when the computational interval of a given problem is small. However, if we have to
solve a problem in a comparatively large interval, the iterations become intricate to give subsequent
components. Therefore the method exhibits huge error. In chapter 5, two modifications in AIM
have been proposed, where one is the Two-Step Modified Iterative Method (TSMIM) and the
other is Multistage Modified Iterative Method (MMIM). In order to solve NL DE with TSMIM,
firstly we construct a recursive scheme with the help of AIM. If the initial component can be
broken into two parts and one of the parts satisfies the given problem, then that part is treated
as the first component, and the rest of the part is added to the second component. Thus, the new
recursive scheme obtained, which gives the numerical solution for the given NL D&IE. This entire
procedure is called TSMIM. An analysis of TSMIM has been construed in the chapter. Now, if
one of the parts does not satisfy NL D&IE, we use MMIM to solve that equation numerically. In
MMIM the computational interval has been fractionize into a number of subintervals, and in each
of the subinterval AIM is applied separately. These modifications have been verified through some
examples. The final results have been shown through some figures and tables.

System of delay differential equation monopolize a place of fundamental importance in all

areas of science and engineering like biology, economics, control and electrodynamics. Particu-
larly these equations come into an interest, when the models [35], which were based on ODE fail
1.1. P RELIMINARIES : 5

completely. In real, a pantograph is a device at the upper part of the train that grabs electric current
from the overhead electric wires and transfers to the train motors. Because of this process the train
moves with a constant speed. The main task was to analyze the motion of the catenary in response
to a pantograph moving with constant speed. At some distance from the supports the simple solu-
tion for the wire and pantograph displacement was obtained by neglecting the inertia and elasticity
of the pantograph. Near a support this solution starts to give a huge perturbation as it predicts dis-
continuities in the vertical pantograph velocity. That’s why pantograph equation was emanated
by Ockendon and Taylor [36], which shows the electricity collection system of the catenary for
trains to confirm that the contact is sustained throughout the overhead wire. Pantograph equations
are studied by many researchers [37, 38, 39, 40, 41, 42, 43]. A form of pantograph equation [44]
solved in the chapter 6, has been solved with the help of modified ADM (MADM)[45].
m
u0 (t) = ∑ λn (t)u(qnt) + f (t) (1.3)
n=1

where u(0) = u0 , t ∈ [0, T ], 0 < qn ≤ 1, λn0 s, and f (t) are continuous known functions on [0, T ] for
n = 1, 2, · · · , m. Convergence of the used method has been developed for eq.(1.3), in spite of it, an
absolute error bound developed for the solution of eq.(1.3). Two numerical examples are solved
with MADM and results are compared through absolute error (AE) with ADM and exact solution.

1.1 Preliminaries:

1.1.1 An Observation of ADM:

Every DE can be written in the deterministic form

F(y) = g(t), (1.4)

where, F is a NL differential operator having linear and NL terms, i.e. F can be decompose as
F = L + R + N. Then the deterministic form eq.(1) will be visible as

Ly + Ry + Ny = g(t). (1.5)

In eq.(1.5), Ly + Ry is linear term with Ly as the higher order derivative, Ry is the remaining linear
term, and Ny is the NL term. In order to solve eq.(1.5) inverse operator must exist with respect to
L, and y must lie in the Banach space. Suppose L−1 is the n-fold inverse operator with respect to
the nth order differential operator L. Operate L−1 both side

L−1 Ly + L−1 Ry + L−1 Ny = g(t),

y = g(t) + φ − L−1 Ry − L−1 Ny, (1.6)

where, φ is the term, which come into a role because of the given initial of boundary conditions.
In order to solve eq.(1.4) numerically, the solution is considered in converging series form such
that
1.1. P RELIMINARIES : 6

∞
y = ∑ yi . (1.7)
i=0

Now, eq.(1.6) will be visible as

! !
∞ ∞ ∞
∑ yi = g(t) + φ − L−1 R ∑ yi − L−1 N ∑ yi . (1.8)
i=0 i=0 i=0

Here the NL term will be decompose by some special functions named as Adomian polynomials
such that
!
∞ ∞
N ∑ yi = ∑ Ai . (1.9)
i=0 i=0

Therefore eq.(1.8) will be

! !
∞ ∞ ∞
−1 −1
∑ yi = g(t) + φ − L R ∑ yi −L ∑ Ai . (1.10)
i=0 i=0 i=0

Now, the recurrence relation is defined as

y0 = g(t) + φ ,
y1 = −L−1 R(y0 ) − L−1 (A0 ), (1.11)
yi+1 = −L−1 R(yi ) − L−1 (Ai ).

Here, recurrence relation eq.(1.11) indicates ADM, by which a numerical solution of eq.(1.4) is
obtained. For further calculation of of eq.(1.11), one must have to construct Adomian polynomials.
There are some certain methods, for the construction of Adomian polynomials.

1. Using Arbitrary Perimeter λ Approach:

According to the considered assumption, the solution of eq.(1.4) is in the converging series form
eq.(1.7). Therefore an arbitrary scalar parameter λ has been put forward such that
∞
ŷ(λ ) = ∑ yn λ n . (1.12)
i=0

Suppose the series is having a radius of convergence r. Let both ŷ and N are analytic over open
disc with centre O and radius r, then N ŷ will also be analytic over open disc(O, r).
Therefore
∞
N ŷ(λ ) = ∑ Ai λ i , (1.13)
i=0

where A0i s are called Adomian polynomials. If R = 1, ŷ(1) will converge to y. Now using Abel’s
theorem, it can be observed that
1.1. P RELIMINARIES : 7

lim ŷ(λ ) = y. (1.14)

λ →1−

Therefore

lim N ŷ(λ ) = Ny. (1.15)

λ →1−

Now using eqs.(1.12)-(1.15), we have

!!
1 dk ∞
Ak =
k! dλ k
N ∑ yi λ i , (1.16)
i=0 λ =0

Therefore for each value of k = 0, 1, 2, 3, · · · , one can easily obtain the Adomian polynomials.

2. Using Kroneker Delta Function Approach:

Let the functional form of N(y) is f (y). In order to obtain Adomian polynomials with the help of
Kroneker delta function, Taylor’s series expansion of f (y), with respect to y0 is
∞
1
f (y) = ∑ k! (y − y0 )(k) f (k) (y0 ),
k=0
!(k)
∞ ∞
1
=∑ ∑ yi f (k) (y0 ), (1.17)
k=0 k! i=1

with the help of Kroneker delta function, eq.(1.17) can be rearranged as

" ! #
∞ i i+1−k
1
f (y) = f (y0 ) + ∑ ∑ k! ∑ (δi, j1 + j2 +···+ jk ).(y j1 .y j2 . · · · .y jk ) f k (y0 ) , (1.18)
i=1 k=1 j1 , j2 ,··· , jk =1

where, δl,m is the Kroneker delta function. According to the basic definition of NL term N(y)
can be rearranged in a converging series form of Adomian polynomials as in eq.(1.9). Therefore
eq.(1.18) can be written as
" ! #
∞ ∞ i i+1−k
1 k
∑ Ai = f (y0 ) + ∑ ∑ k! ∑ (δi, j1 + j2 +···+ jk ).(y j1 .y j2 . · · · .y jk ) f (y0 ) , (1.19)
i=0 i=1 k=1 j1 , j2 ,··· , jk =1

Since, Adomian polynomials are defined by the definition

A0 = f (y0 )
!
i i+1−k
1
Ak =∑ ∑ (δi, j1 + j2 +···+ jk ).(y j1 .y j2 . · · · .y jk ) f k (y0 ) (1.20)
k=1 k! j1 , j2 ,··· , jk =1
1.1. P RELIMINARIES : 8

3. Using Taylor’s Series Method Approach:

To obtain Adomian polynomials via Taylor’s series expansion of f (y) about y0 . Therefore using
eq.(1.17), we have

f (y) = f (y0 ) + (y1 + y2 + · · · ) f (1) (y0 ) + (y1 + y2 + · · · )2 f (2) (y0 )

+(y1 + y2 + · · · )3 f (3) (y0 ) + (y1 + y2 + · · · )4 f (4) (y0 ) + · · · (1.21)

Therefore eq.(1.21) can be rearranged as

1 2 (2)
f (y) = f (y0 ) + [y1 f (1) (y0 )] + [y2 f (1) (y0 ) + y f (y0 )] + · · · (1.22)
2! 1

According to the basic definition of NL term Ny can be rearranged in a converging series form of
Adomian polynomials as in eq.(1.9). Therefore eq.(1.22) can be written as
∞
1
∑ Ai = f (y0 ) + [y1 f (1) (y0 )] + [y2 f (1) (y0 ) + 2! y21 f (2) (y0 )] + · · · (1.23)
i=0

Since the Adomian polynomials will be defined by the formula

k
Ak = ∑ c(ν, k) f (ν) (y0 ), k ≥ 0. (1.24)
ν=1

Here, we shall define define the first four Adomian polynomials given by eqs.(1.16), (1.20), and
(1.24) as




 A0 = f (y0 ),

A1 = y1 f (1) (y0 ),





 1
A2 = y2 f (1) (y0 ) + y21 f (2) (y0 ),

2! (1.25)
A = y f (1) (y ) + y y f (2) (y ) + 1 y3 f (3) (y ),


3 3 0 1 2 0 0
3! 1



 
 1 2 1 2 (3) 1 4 (4)
A4 = y4 f (1) (y0 ) + y1 y3 + y2 f (2) (y0 ) + y1 y2 f y0 + y f (y0 ).



2! 2! 4!

Although the Adomian polynomials eq.(1.25), obtained using eqs.(1.16),(1.20), and (1.24) are
not unique, because Adomian polynomials are a rearrangement of Taylor’s series expansion, of
the deterministic form f(y) of Ny. In article [46], we dissertate two new classes of Adomian
polynomials.
The class A∗n , is a class of Adomian polynomials, which can be useful to solve a problem with
n
ADM. Let Sn = ∑ yk is the sum of first n + 1 components. Also Tn is the Taylor’s series expansion
k=0
of the NL term Ny = f (y) about point y0 such that

T0 = f (y0 )
n+1
1
Tn = ∑ k! (Sn − S0 )k f (k) (y0 ), n ≥ 1. (1.26)
k=0
1.1. P RELIMINARIES : 9

Now, the class A∗n of Adomian polynomials is given as


A∗ = T = f (y )
0 0 0
(1.27)
A∗ = T − T , n ≥ 1
n n n−1

Therefore, using eq.(1.27), first four polynomials of the new class of Adomian polynomial are



A∗0 = f (y0 )


A∗ = y1 f (1) (y0 ) + 2!1 f (2) (y0 ),


 1


A∗2 = y2 f (1) y0 + 2!1 (y22 + 2y1 y2 ) f (2) (y0 ) + 3!1 (y31 + y32 + 3y21 y2 + 3y1 y22 ) f (3) y0 , (1.28)


A∗3 = y3 f (1) + 2!1 (u23 + 2y2 y3 + 2y1 y3 ) f (2) (y0 ) + 3!1 (y23 + 3y21 y3 + 3y22 y3 + 3y2 y23






 +6y y y ) f (3) (y ) + 1 (y + y + y )4 f (4) (y )

1 2 3 0 4! 1 2 3 0

The class A∗∗

n is also a class of Adomian polynomials, which is useful for the solution of a NL
problem using ADM. For A∗∗
n class of Adomian polynomials, let

T0 = A∗∗
0 = f (y0 )
" #
n−1
1
Tn = f (y0 ) + (Sn − S0 ) f (1) (y0 ) + ∑ (Sn−k − S0 )k+2 f (k+2) y0 , n ≥ 1. (1.29)
(k + 2)! k=0

Now, the class A∗∗

n of Adomian polynomials is given as

A∗∗ = T = f (y )
0 0 0
(1.30)
A∗∗ = T − T , n ≥ 1
n n n−1

Therefore, using eq.(1.30), first four polynomials of the new class of Adomian polynomial are

∗∗
A0 = f (y0 ),




A∗∗ = y1 f (1) (y0 ) + 2!1 y21 f (2) (y0 ),


 1

A∗∗
2 = y2 f
(1) (y ) + 1 (2y y + y2 ) f (2) (y ) + 1 y3 f (3) (y ),
0 2! 1 2 2 0 3! 1 0 (1.31)


A∗∗ = y f (1) (y ) + 1 (2y y + 2y y + y2 ) f (2) (y ) + 1 (3y2 y


 3 3 0 2! 1 3 2 3 3 0 3! 1 2


+3y1 y22 + y32 ) f (3) (y0 ) + 4!1 y41 f (4) (y0 ).



1.1.2 A Reliable Modification in ADM:

It is clear by the scheme eq.(1.11), obtained from the usual ADM that the zeroth component y0
depends on g(t) + φ . Therefore a slight change in y0 , may cause for a rapid convergence than
the usual ADM. In this modification ADM g(t) + φ is broken into two parts f1 and f2 , such that
g(t) + φ = f1 + f2 . If either f1 or f2 is assigned to the y0 in such a way that the convergence of the
numerical solution is much faster than the usual ADM. Since Modified Adomian Decomposition
method (MADM) is effective when there is a change in zeroth component of ADM, therefore
using eq.(1.11)
1.1. P RELIMINARIES : 10

y0 = g(t) + φ = f1 + f2 ,
y1 = −L−1 R(y0 ) − L−1 (A0 ), (1.32)
−1 −1
yi+1 = −L R(yi ) − L (Ai ).

According to MADM eq.(1.32) can be written as

y0 = f1 (or f2 ),
y1 = f2 (or f1 ) − L−1 R(y0 ) − L−1 (A0 ), (1.33)
yi+1 = −L−1 R(yi ) − L−1 (Ai ).

One must have to emphasis that the success of MADM relies upon the right choice of f1 or f2 for
the rapid convergence.

1.1.3 Green’s Decomposition Method:

As we know that every DE can be inscribed in a deterministic form of eq.(1.4) with boundary
conditions

A11 y(a) + A12 y(1) (a) + · · · + A1n y(n−1) = c1

A21 y(b) + A22 y(1) (b) + · · · + A2n y(n−1) b = c2 (1.34)

such that at least one of A1 j , A2 j 6= 0, j = 1, 2, · · · , n for the GDM. Let us consider the homoge-
neous form of eq.(1.4) and eq.(1.34) in such a way



Ly + Ry = 0

A11 y(a) + A12 y(1) (a) + · · · + A1n y(n−1) = c1 (1.35)


A y(b) + A y(1) (b) + · · · + A y(n−1) b = c ,

21 22 2n 2

the unique solution of eq.(1.35) is Y (x).

Also the Green’s function for the homogeneous boundary conditions using eq.(1.5) considered as



Ly + Ry = g(x) − Ny,

A11 y(a) + A12 y(1) (a) + · · · + A1n y(n−1) = 0 (1.36)


A y(b) + A y(1) (b) + · · · + A y(n−1) b = 0

21 22 2n

is

G (x,t)
1 a≤x<t ≤b
G(x,t) = (1.37)
G (x,t)
2 a ≤ t < x ≤ b.
1.1. P RELIMINARIES : 11

Therefore the solution of eq.(1.4), with associated boundary conditions eq.(1.34) will be
Z b
y(x) = Y (x) + G(x,t)(g(t) − Ny)dt,
a
Z x Z b
= Y (x) + G2 (x,t)(g(t) − Ny)dt + G1 (x,t)(g(t) − Ny)dt. (1.38)
a x

In order to apply GDM, using the concept of eq.(1.7) and (1.9) in eq.(1.38),

Z x
! Z b
!
∞ ∞ ∞
∑ yi (x) = Y (x) + G2 (x,t) g(t) − ∑ Ai dt + G1 (x,t) g(t) − ∑ Ai dt. (1.39)
i=0 a i=0 x i=0

Therefore the recurrence relation is defined as




 y0 (x) = Y (x)

 Z x Z b
y1 (x) = G2 (x,t) (g(t) − A0 ) dt + G1 (x,t) (g(t) − A0 ) dt. (1.40)
 aZ xZ

 x b
yn+1 (x) = G2 (x,t) (g(t) − An ) dt + G1 (x,t) (g(t) − An ) dt.

a x

Here A0i s are Adomian polynomials given by eq.(1.25). Therefore eq.(1.40) is required scheme
for the required GDM. In order to upgrade GDM into modified green decomposition method
(MGDM) to solve eq.(1.40) one may use another class of Adomian polynomials given in eq.(1.28)
and eq.(1.31).

1.1.4 An Iterative Method:

An Iterative Method [31, 47], is one of the sublime technique to solve eq.(1.4) numerically. Con-
sider a general functional equation:

y = ψ + ξ (u), (1.41)

where ξ is a NL operator form, which is defined from a Banach space B → B and ψ is a known
function/source term. Assume the solution of eq.(1.41) in convergent series form as
∞
y= ∑ yn . (1.42)
n=0

Now, the NL term can be decomposed as

!
∞
ξ (y) = ξ ∑ yi
i=0
( ! !) (1.43)
∞ i i−1
= ξ (y0 ) + ∑ ξ ∑ yj −ξ ∑ yj ,
i=1 j=0 j=0
1.1. P RELIMINARIES : 12

and using the equations eq.(1.42) and eq.(1.43), the equation eq.(1.41) can be written as
∞
y = ∑ yn
n=0
∞
(
i
!
i−1
!) (1.44)
=ψ + ξ (y0 ) + ∑ ξ ∑ yj −ξ ∑ yj ,
i=1 j=0 j=0

again, a recurrence relation is defined [47] as follow:


y0 = ψ





y = ξ (y )
1 0
(1.45)
m+1 = ξ (y0 + · · · + ym ) − ξ (y0 + · · · + ym−1 ),
y





 m = 1, 2, ...

this implies
y1 + y2 + · · · + ym+1 = ξ (y0 + y1 + · · · + ym ), (1.46)

and the equations eq.(1.46) and eq.(1.41) can be written as,

∞
y = ψ + ∑ yn . (1.47)
n=1

If ξ is contraction, i.e.,

kξ (x) − ξ (y)k ≤ Kkx − yk, 0<K<1,

then using eq.(1.45), we have

kym+1 k = kξ (y0 + y1 + · · · + ym ) − ξ (y0 + y1 + · · · + ym−1 )k (1.48)

≤ Kkym k (1.49)
m+1
≤K ky0 k (1.50)
∞
where m = 0, 1, 2, ..., and the series ∑ yn , uniformly converges to a solution eq.(1.41), using
n=0
Banach fixed point theorem [48].

1.1.5 Padè Approximation:

NL D&IEs do not have an analytic or closed form solution always. Therefore numerical solution
become necessary to solve NL DorIE. The numerical solution occurs in a series form, however the
truncated series gives a solution with a limited region of convergence (RoC). Therefore to increase
the RoC a mathematical tool Padè Approximation (PA) has been used. PA was developed by Henri
Padè in 1890. PA provides a rational approximation oa a truncated power series. PA is denoted
as M
  M
N , where M ≥ 0 and N ≥ 1 are positive integers. In N PA, M and N represents the degree
of numerator and denominator respectively. Therefore a continuous function f (x), which can be
approximated in a power series ∑∞ i
i=0 ai x such that
1.1. P RELIMINARIES : 13

∞
f (x) = ∑ ai xi
i=0
= a0 + a1 x + a2 x2 + · · · + aK xK + O(xK+1 ), (1.51)

0 M
here ai s are known arbitrary constants. Therefore N PA is defined by the rational function

M

M
 ∑ ci xi
i=0
(x) = N
,
N i
1 + ∑ di x
i=1
c0 + c1 x + · · · + cM xM
= + O(xM+N+1 ), (1.52)
1 + d0 + d1 x + · · · + dN x N
0 0 M
where M + N ≤ K and ci s, di s are unknown arbitrary constants. If M = N in eq(1.52), then M is
called diagonal PA. In most of the cases diagonal PA is more effective than usual PA. If we have
0 0
to approximate f (x), we have to determine ci s, di s. Therefore to calculate PA using [49], we have

c0 + c1 x + · · · + cM xM
a0 + a1 x + · · · + aK x K =
1 + d0 + d1 x + · · · + dN x N

(a0 + a1 x + · · · + aK xK )(1 + d0 + d1 x + · · · + dN xN ) = c0 + c1 x + · · · + cM xM . (1.53)

For the solution of eq(1.53) and on equating like powers of x, we have





 c0 = a0





 c1 = a1 + d1 a0


c2 = a2 + d1 a1 (1.54)


.
..



 M

= + ∑ di aM−i .

 c
 M
 aM
i=1

0 0
This system of equation can be solved easily and ci s, di s are calculated.