Chapter 3

Corrected Radau Quadrature and

its Applications

3.1 Introduction

Gauss quadrature rules are more useful than Newton-Cote’s quadrature rules. One

advantage of Gauss quadrature rules is, they don’t need mesh points to be equispaced

and the other is its degree of precision is 2n − 1, which is higher than that of Newton-

Cote’s quadrature rules. Among the several Gauss type quadrature rules available in

the literature, Gauss Lobatto and Gauss Radau quadratures are remarkable. As these

rules includes end points as mesh points, they have many applications in boundary

value problems. These rules are studied and generalized by many researchers. W.

Gautschi constructed high order Gauss Lobatto [81] and Gauss Radau [82] quadrature

rules for Jacobi’s weight functions ((1 − t)α (1 + t)β , α, β > −1) and Lagurre weight

functions (tα e−t , α > −1). He used modified Jacobi matrix for the computation of

weights and nodes. Later in [85], these rules are generalized by Gautschi for the

end points with arbitrary multiplicity. Gautschi and Li [84], developed corrected
3.1 Introduction 24

quadrature rules of Gauss Lobatto and Gauss Radau type, with end points having

multiplicity two, for the Chebyshev weight functions. These rules involve first deriva-

tive of function at end points. In [70] S. Li proposed formulae for Kronrod extension

of Gauss Lobatto and Gauss Radau quadrature rules having end points of multiplicity

two. These rules are constructed for the Chebyshev weight functions. Bokhari and

others [41], derived corrected Gauss Radau and Gauss Lobatto quadrature rules with

respect to the weight function (1 − t) using Jacobi matrix. Eslahchi and Dehghan

[46], developed three new quadrature rules of order 2n + m − 1 using the method of

integration by parts. These rules involve derivative of a function of an arbitrarily

fixed order m. Nodes, weights and errors of newly derived quadrature rules are com-

puted using Gauss-Jacobi and Gauss-Lagurre quadrature rules. Aljinovi and others

[1] constructed error bounds for two point Gauss Radau quadrature as well as three

point Gauss-Radau and Gauss-Lobatto quadrature rules for functions of bounded

variations. Milovanović and Cvetković [27], developed a new corrected quadrature

rule for a finite positive Borel measure whose support is an unbounded set.

Maleknejad and Lotfi [37], proposed a method for the evaluation of Fredholm

and Volterra type integral equations of first and second kind. They interpolated the

unknown function, occurred in the integral equation, by an interpolating polynomial

at zeros of Chebyshev polynomial and then applying Gauss quadrature rules, solution

of FIE and VIE is obtained. In [72], solution of FIE is obtained by using Gauss

quadrature rule with multiple nodes. Hossain and Ahmed [42], formulated a new

scheme for the computation of initial value problems (IVP) of the form

y 0 (x) = f (x, y), x ∈ [a, b]

with initial condition y(a) = y0 . They converted the IVP into an integral equation of
3.2 Corrected Radau Quadrature rule 25

the type
Z xk+1
yk+1 = yk + f (x, y)dx,
xk

which is evaluated using Gauss-Legendre quadrature rule. Taylor’s series approxi-

mation is used to transform the function of two variables into the function with one

variable.

In the present chapter we derive third order corrected quadrature rule. Optimal

error bounds are obtained for the quadrature rule via double integral inequalities.

This third order quadrature rule is used to construct numerical method of solution of

first order ODE. It is proved that the method so obtained is optimal. The corrected

quadrature rule is also used to determine approximate solutions of FIE, FIDE and

VIDE.

The chapter is organized as follows. In second section, a new third order corrected

quadrature rule is derived by using method of undetermined coefficients. In third

section, piecewise continuous functions are defined to obtain upper and lower bounds

for the error term in the corrected quadrature rule. Optimality of error bounds is also

established in this section. Last section is devoted to the applications of the corrected

quadrature rule.

3.2 Corrected Radau Quadrature rule

Let
Z 1
f (x)dx = λ0 f (0) + λ1 f (x1 ) + λ2 f 0 (0). (3.2.1)
0

The unknowns x1 , λ0 , λ1 and λ2 are determined in such a way that equation (3.2.1)

is identically satisfied for all polynomials. We assume that f (x) is a polynomial of

degree less than or equal to three, i.e. f (x) = xi , i = 0, 1, 2, 3.

3.2 Corrected Radau Quadrature rule 26

Then equation (3.2.1) gives a system of four equations in four unknowns

λ0 + λ1 = 1,
1
λ 1 x1 + λ 2 = ,
2
1
λ 1 x1 2 = ,
3
1
λ 1 x1 3 = .
4

Solving these equations simultaneously we get

3 11 16 1
x1 = , λ0 = , λ1 = and λ2 = .
4 27 27 18

With these values, equation (3.2.1) becomes,

Z 1  
11 16 3 1
f (x)dx = f (0) + f + f 0 (0). (3.2.2)
0 27 27 4 18

Error in formula (3.2.2) is given by

Z 1
E = f (x)dx − Qc , (3.2.3)
0
f (4) (ξ)
= C , 0 < ξ < 1.
4!

where
11 16 3 1
Qc = f (0) + f ( ) + f 0 (0)
27 27 4 18

and
Z 1  4
11 16 3 1
C = x dx − (04 ) −
4
− (4 × 03 ),
0 27 27 4 18
1
= .
80

Equation (3.2.2) represents the corrected quadrature rule of order three.

3.3 Error Analysis 27

3.3 Error Analysis

To obtain the optimal error bounds we define piecewise continuous polynomials,

11 3
P1 (x) = x − , 0≤x≤ ;
27 4
3
= x − 1, < x ≤ 1.
4

 
1 2 22x 1 3
P2 (x) = x − + , 0≤x≤ ;
2! 27 9 4
(x − 1)2 3
= , < x ≤ 1.
2! 4

 
x 2 11x 1 3
P3 (x) = x − + , 0≤x≤ ;
3! 9 3 4
(x − 1)3 3
= , < x ≤ 1.
3! 4

x2 2 44x 2
 
3
P4 (x) = x − + , 0≤x≤ ;
4! 27 3 4
(x − 1)4 3
= , < x ≤ 1.
4! 4

Note that
Z 1
1
P4 (x)dx = (3.3.1)
0 1920

and
1
max P4 (x) = . (3.3.2)
x[0,1] 6144

Consider
Z 1
I= P4 (x)f (4) (x)dx.
0

Using integration by parts, above integral reduces to

Z 1
I=− P3 (x)f (3) (x)dx.
0
3.3 Error Analysis 28

Continuous integration of above integral gives

Z 1 Z 1
(4)
P4 (x)f (x)dx = f (x)dx − Qc . (3.3.3)
0 0

Thus from equations (3.2.3) and (3.3.3) we can write,

Z 1
E= P4 (x)f (4) (x)dx. (3.3.4)
0

Thus equation (3.3.4) represent the error in the corrected quadrature formula.

We define two real numbers,

γ4 = min f (4) (x) and Γ4 = max f (4) (x). (3.3.5)

x[0,1] x[0,1]

Clearly,

γ4 ≤ f (4) (x) ≤ Γ4 for x  [0, 1]. (3.3.6)

From equation (3.3.1) and equation (3.3.4) we can write,

Z 1h i Z 1 Z 1
f (4) (x) − γ4 P4 (x)dx = f (4) (x)P4 (x)dx − γ4 P4 (x)dx
0 0 0
γ4
= E− . (3.3.7)
1920

Similarly
Z 1h
(4)
i Γ4
Γ4 − f (x) P4 (x)dx = − E. (3.3.8)
0 1920
But
Z 1h i Z 1
(4)
f (x) − γ4 P4 (x)dx ≤ max |P4 (x)| |f (4) (x) − γ4 |dx (3.3.9)
0 x[0,1] 0

and
Z 1h i Z 1
(4)
Γ4 − f (x) P4 (x)dx ≤ max |P4 (x)| |Γ4 − f (4) (x)|dx. (3.3.10)
0 x[0,1] 0

Since |f (4) (x)−γ4 | = f (4) (x)−γ4 and |Γ4 −f (4) (x)| = Γ4 −f (4) (x), from equations

(3.3.7), (3.3.9) and (3.3.2) we get

Z 1h
γ4 i
E− = f (4) (x) − γ4 P4 (x)dx
1920 0
3.3 Error Analysis 29

Z 1
1
≤ |f (4) (x) − γ4 |dx
6144 0
1 h (3) i
≤ f (1) − f (3) (0) − γ4
6144
1
≤ [S3 − γ4 ] ,
6144
where S3 = f (3) (1) − f (3) (0). (3.3.11)

Thus we have
1 1
E≤ S3 + γ4 . (3.3.12)
6144 30720

Similarly from equations (3.3.8), (3.3.10) and (3.3.2) we have

1 1
S3 + Γ4 ≤ E. (3.3.13)
6144 30720

Equations (3.3.12) and (3.3.13) together with equation (3.3.4) gives

1 1 1 1
S3 + Γ4 ≤ E ≤ S3 + γ4 , (3.3.14)
6144 30720 6144 30720

where Γ4 , γ4 and S3 are defined in the equations (3.3.5) and (3.3.11).

It is observed that error bounds derived in equation (3.3.14) are sharp.

Consider the function f (x) = x4 .

Then, γ4 = Γ4 = 4! and S3 = f (3) (1) − f (3) (0) = 4!.

With these values and from equation (3.3.14) we get

1 1 1
Lower bound = S3 + Γ4 = , (3.3.15)
6144 30720 80

1 1 1
Upper bound = S3 + γ4 = (3.3.16)
6144 30720 80

and
Z 1 Z 1  4
c 11 16 3 1
E= f (x)dx − Q = x dx − (04 ) −
4
− (4 × 03 )
0 0 27 27 4 18
3.4 Applications of Corrected Radau Quadrature 30

1
∴E = . (3.3.17)
80

Thus equations (3.3.15)-(3.3.17) all together prove that upper and lower bound of

the error function E are same for the function f (x) = x4 .

Thus the bounds given in equation (3.3.14) are optimal.

3.4 Applications of Corrected Radau Quadrature

3.4.1 Numerical Method for Ordinary Differential Equations

In this section we propose a numerical scheme to solve an ODE of the form

y 0 (x) = f (x, y(x)). (3.4.1)

We convert the ODE (3.4.1) into an integral equation by integrating it w.r.t. x

between the limits xn and xn+1 . Thus we get,

Z xn+1
y(xn+1 ) − y(xn ) = f (x, y(x))dx. (3.4.2)
xn

By substituting x = xn + ht, where h = xn+1 − xn , equation (3.4.2) becomes

Z 1
yn+1 = yn + h f (xn + ht, y(xn + ht))dt,
0

where yi = y(xi ).

On applying the corrected quadrature formula (3.2.2) to above equation we get,

11 1 16
yn+1 = yn + K1 + K2 + K3 , (3.4.3)
27 18 27

where

K1 = hf (xn , yn ), (3.4.4)

K2 = h2 (fx (xn , yn ) + f (xn , yn )fy (xn , yn )) , (3.4.5)

3.4 Applications of Corrected Radau Quadrature 31

  
3h 3h
K3 = hf xn + , y xn + .
4 4

We approximate K3 as
 
3h
K3 = hf xn + , y(xn ) + a31 K1 + a32 K2 . (3.4.6)
4

From equations (3.4.4) and (3.4.5), equation (3.4.6) can be written as,
 
3h 0 2 00
K3 = hf xn + , yn + ha31 yn + h a32 yn . (3.4.7)
4

Using Taylor’s series expansion of equation (3.4.3) and equation (3.4.7) about (xn , yn )

and equating the coefficients of equal powers of h, we get

3 9
a31 = and a32 = .
4 32

Thus we get a numerical scheme

11 1 16
yn+1 = yn + K1 + K2 + K3 , (3.4.8)
27 18 27

where 
K1 = hf (xn , yn ), 






2
K2 = h (fx (xn , yn ) + f (xn , yn )fy (xn , yn )) , (3.4.9)
   
3h 3 9


K3 = hf xn + , yn + K1 + K2 . 


4 4 32
The truncation error T is

T = y(xn+1 ) − yn+1
h4
= fy (xn , yn ) y 000 (xn , yn ). (3.4.10)
4!

From equation (3.4.10), it is observed that method (3.4.8) is third order optimal

method.

For the test equation y 0 = λy , equation (3.4.8) becomes

 
11 1 2 2 16 3 9 2 2
yn+1 = yn + hλyn + h λ yn + hλ yn + hλyn + h λ yn ;
27 18 27 4 32
3.4 Applications of Corrected Radau Quadrature 32

 
1 2 2 1 3 3
= 1 + hλ + h λ + h λ yn ,
2 3!
= R(hλ)yn ,

where
z2 z3
 
R(z) = 1+z+ + .
2! 3!

From the growth function R(z), we conclude that the stability region of (3.4.8) coin-

cides with the stability region of third order explicit Runge-Kutta (R-K) method.

Example 3.4.1. Consider an ODE y 0 = xy with initial condition y(0) = 1. An exact

x2
solution of this ODE is y = e 2 . Following table presents numerical solution of given
1
ODE for h = 41 , 20 obtained by (3.4.8) as well as by third order R-K method and exact
solution.

xn Proposed method Third ordered R-K method Exact Solution

1 1 1 1
h= 4 h= 20 h= 4 h= 20
0.25 1.031738281 1.031743325 1.031901042 1.031744786 1.031743407
0.5 1.133071219 1.133147697 1.133470403 1.133151773 1.133148453
0.75 1.324452185 1.324781616 1.325287761 1.324791303 1.324784759
1.0 1.647683512 1.648711352 1.649383605 1.648733282 1.648721271

Table 3.1: Numerical and Exact solution of an ODE y 0 = xy

From table (3.1) it is clear that the proposed method gives better results than

third order R-K method.

3.4.2 Approximate Solution of Integral equation

In this section we use corrected Radau quadrature rule to obtain approximate solu-

tions of Fredholm and Volterra integral equations.

3.4 Applications of Corrected Radau Quadrature 33

3.4.2.1 Fredholm Integral equation

Consider a linear Fredholm integral equation of second kind,

Z b
φ(x) = f (x) + λ k(x, t)φ(t)dt, (3.4.11)
a

where f (x) and k(x, t) are known functions, λ is non-zero constant and φ(x) is an

unknown function. To determine φ(x), we assume that φ(x), f (x) and k(x, t) are

differentiable functions.

To convert the integral in equation (3.4.11) in standard form (3.2.2) substitute t =

a + (b − a)y . Equation (3.4.11) now becomes

Z 1
φ(x) = f (x) + λ(b − a) k(x, a + (b − a)y)φ(a + (b − a)y)dy. (3.4.12)
0

Applying the quadrature rule (3.2.2) to the equation (3.4.12), we get

    
11 16 3 3
φ(x) = f (x) + λ(b − a) k(x, a)φ(a) + k x, a + (b − a) φ a + (b − a)
27 27 4 4

b−a
k(x, a)φ0 (a) + φ(a)kt (x, a) .


+ (3.4.13)
18

Differentiating equation (3.4.13) w.r.t. x, we get

    
0 0 11 16 3 3
φ (x) = f (x) + λ(b − a) kx (x, a)φ(a) + kx x, a + (b − a) φ a + (b − a)
27 27 4 4

b−a
kx (x, a)φ0 (a) + φ(a)kxy (x, a) .


+ (3.4.14)
18

From above equation we have

  
0 18 0 11 b−a
φ (a) = f (a) + λ(b − a) kx (a, a) + kxy (a, a) φ(a)
18 − λ(b − a)2 kx (a, a) 27 18
   
16λ(b − a) 3 3
+ kx a, a + (b − a) φ a + (b − a) . (3.4.15)
27 4 4

From equation (3.4.15) and equation (3.4.13) we get,

16
φ(x) = f (x) + Af 0 (a) + λ(b − a)(k(x, c) + Akx (a, c))φ(c)
27
3.4 Applications of Corrected Radau Quadrature 34

  
11 11 b−a
+λ(b − a) k(x, a) + A k(a, a) + kxt (a, a)
27 27 18

(b − a)
+ kt (x, a) φ(a), (3.4.16)
18

λ(b−a)2 k(x,a)
where A = 18−λ(b−a)2 kx (a,a)
, c = a + 34 (b − a) and
λ is non-zero constant.

Equation (3.4.16) contains two unknown values φ(a) and φ a + 43 (b − a) . These values

are calculated by solving the system of equations obtained from equation (3.4.16) at

x = a and x = a + 34 (b − a).

Example 3.4.2. Consider a linear FIE,

Z 1
y(x) = 1 + x ty(t)dt,
0

where f (x) = 1, k(x, t) = xt, λ = 1, a = 0 and b = 1.

Analytic solution of the problem is y(x) = 1 + 43 x.

From equation (3.4.16), we get

  
1 4 3
y(x) = 1 + x y(0) + y . (3.4.17)
18 9 4

At x = 0 and x = 43 , equation (3.4.17) gives y(0) = 1 and y 3 25

4 = 16 .

Thus from equation (3.4.17), solution of given FIE is

3
y(x) = 1 + x.
4

3.4.2.2 Volterra Integral equation

Consider a linear Volterra integral equation of second kind,

Z x
φ(x) = f (x) + λ k(x, t)φ(t)dt, (3.4.18)
a

where f (x) and k(x, t) are known functions and φ(x) is an unknown function.

From equation (3.4.18), observe that φ(a) = f (a).

3.4 Applications of Corrected Radau Quadrature 35

To convert the integral in equation (3.4.18) in standard form (3.2.2) substitute t =

a + (x − a)y . Then equation (3.4.18) becomes

Z 1
φ(x) = f (x) + λ(x − a) k(x, a + (x − a)y)φ(a + (x − a)y)dy.
0

Since φ(a) = f (x), applying the quadrature rule (3.2.2) to above equation, we get
    
11 16 3 3
φ(x) = f (x) + λ(x − a) f (a)k(x, a) + k x, a + (x − a) φ a + (x − a)
27 27 4 4

x−a
k(x, a)φ0 (a) + f (a)kt (x, a) .


+ (3.4.19)
18

Differentiating equation (3.4.19) w.r.t. x, we get

11
φ0 (x) = f 0 (x) + λf (a) ((x − a)kx (x, a) + k(x, a))
27
λ 0
+ φ (a)(x − a)((x − a)kx (x, a) + 2k(x, a))
18    
16 3 3
+ λφ a + (x − a) k x, a + (x − a)
27 4 4
   
16 3 3
+ λ(x − a)φ a + (x − a) kx x, a + (x − a)
27 4 4
   
16 3 3
+ λ(x − a)φx a + (x − a) k x, a + (x − a)
27 4 4
λ
+ f (a)(x − a)((x − a)ktx (x, a) + 2kt (x, a)).
18

At x = a above equation becomes

φ0 (a) = f 0 (a) + λf (a)k(a, a). (3.4.20)

At x = a + 43 (x − a), from equation (3.4.19) and equation (3.4.20) we get,

   
3 3
φ a + (x − a) = f a + (x − a)
4 4
   
11 x−a 0 3
+λ(x − a) f (a) + (f (a) + λf (a)k(a, a)) k a + (x − a), a
36 32 4
 
λ 3
+ (x − a)2 f (a)kt a + (x − a), a
32 4
   
4 3 9 9
+ λ(x − a)k a + (x − a), a + (x − a) φ a + (x − a) (3.4.21) .
9 4 16 16
3.4 Applications of Corrected Radau Quadrature 36

Let
 
9 9
φ a + (x − a) = φ(a) + (x − a)φ0 (a),
16 16
 
9 9
= 1 + (x − a)k(a, a) f (a) + (x − a)f 0 (a),
16 16
= A.(say) (3.4.22)

Thus equation (3.4.21) becomes

 
3 3
φ a + (x − a) = f (a + (x − a))
4 4
   
11 x−a 0 3
+λ(x − a) f (a) + (f (a) + λf (a)k(a, a)) k a + (x − a), a
36 32 4
 
λ 3
+ (x − a)2 f (a)kt a + (x − a), a
32 4
 
4 3 9
+ λ(x − a)Ak a + (x − a), a + (x − a)
9 4 16
= B.(say) (3.4.23)

From equations (3.4.20), (3.4.22) and (3.4.23), equation (3.4.19) becomes

 
11 x−a 0
φ(x) = f (x) + λ(x − a) f (a) + (f (a) + λf (a)k(a, a)) k(x, a)
27 18
 
16 3 1
+ λ(x − a)Bk x, a + (x − a) + λ(x − a)2 f (a)kt (x, a). (3.4.24)
27 4 18

Equation (3.4.24) represents the solution of VIE (3.4.18).

Example 3.4.3. Consider a linear VIE,

Z 1
φ(x) = x + (t − x)φ(t)dt,
0

where f (x) = x, k(x, t) = t − x, λ=1 and a = 0.

Analytic solution of the problem is φ(x) = sinx.

From equation (3.4.24), we get

x3 9 5
φ(x) = x − + x .
3! 128
3.4 Applications of Corrected Radau Quadrature 37

3.4.3 Approximate Solution of Integro-Differential Equations

3.4.3.1 Fredholm Integro-Differential Equations

Consider Linear Fredholm integro-Differential equation,

Z b
0
y (x) + αy(x) = f (x) + k(x, t)y(t)dt (3.4.25)
a

with initial condition y(a) = y0 .

To convert the integral in standard form (3.2.2), substitute t = a + (b − a)v , then

equation (3.4.18) becomes,

Z 1
0
y (x) + αy(x) = f (x) + (b − a) k(x, a + (b − a)v)y(a + (b − a)v)dv.
0

Applying the quadrature rule (3.2.2) to above equation we get

(b − a)2
 
0 11 b−a 0
y (x) + αy(x) = f (x) + y0 + y (a) (b − a)k(x, a) + y0 kt (x, a)
27 18 18
   
16 3 3
+ (b − a)y a + (b − a) k x, a + (b − a) . (3.4.26)
27 4 4

Multiplying equation (3.4.26) by eαx and then integrating the resulting equation be-

tween the limits a to x we get

x   Z x
b−a 0
Z
αx αa αs 11
e y(x) − e y0 = e f (s)ds + y0 + y (a) (b − a) eαs k(s, a)ds
a 27 18 a
Z x
(b − a)2
+ y0 eαs kt (s, a)ds
18 a
 Z x  
16 3 3
+ (b − a)y a + (b − a) eαs k s, a + (b − a) ds.
27 4 a 4
(3.4.27)

Solution of equation (3.4.25) is given by equation (3.4.27). Equation (3.4.27) repre-

sents the required solution of equation (3.4.25) if the values of y 0 (a) and y a + 43 (b − a)

are known. These values are obtained by evaluating equation (3.4.26) at x = a and
3.4 Applications of Corrected Radau Quadrature 38

equation (3.4.27) at x = a + 34 (b − a) respectively.

Let

3
c = a + (b − a),
4
A = 18 − (b − a)2 k(a, a),
Z c Z c
αc 16 3 αs 16
B = e − (b − a) k(a, c) e k(s, a)ds − (b − a) eαs k(s, c)ds,
27A a 27A a
11 (b − a)2
D = f (a) + (b − a)k(a, a)y0 + kt (a, a)y0 − αy0 .
27 18

Equation (3.4.27) gives

 
0 18 16
y (a) = D + (b − a)k(a, c)y(c) . (3.4.28)
A 27

From equation (3.4.27) and (3.4.28) we have

eαa y0 1 c αs (b − a)2 y0 c αs
Z Z
y(c) = + e f (s)ds + e kt (s, a)ds
B B a 18B a
 Z c
(b − a) b − a 11
+ D + y0 eαs k(s, a)ds. (3.4.29)
B A 27 a

From equations (3.4.28) and (3.4.29), equation (3.4.27) becomes

Z x  Z x
αx αa αs 11 D
e y(x) = e y0 + e f (s)ds + (b − a) y0 + (b − a) eαs k(s, a)ds
a 27 A a
Z x
(b − a)2
+ y0 eαs kt (s, a)ds
18
a Z x Z x 
16(b − a)
+ (b − a)2 k(a, c) eαs k(s, a)ds + eαs k(s, c)ds ×
27AB a a
 Z c 2 Z x
(b − a)
eαa y0 + eαs f (s)ds + y0 eαs kt (s, a)ds
18
a Z c a

11 D αs
+(b − a) y0 + (b − a) e k(s, a)ds . (3.4.30)
27 A a

Thus equation (3.4.30) represents the solution of FIDE (3.4.25).

Example 3.4.4. Consider a linear FIDE,

Z 1
1
y 0 (x) + y(x) = 1 + x t y(t)dt,
3 0
3.4 Applications of Corrected Radau Quadrature 39

with initial condition y(0) = 0.

Here f (x) = 1, k(x, t) = xt, α = 1/3, a = 0 and b = 1.

Analytic solution of the problem is y(x) = x.

From equation (3.4.30), approximate solution of the problem is y(x) = x.

3.4.3.2 Volterra Integro-Differential Equation

Consider a linear VIDE

Z x
0
y (x) = f (x) + k(x, t)y(t)dt, (3.4.31)
a

with initial condition y(a) = y0 .

Observe that

y 0 (a) = f (a).

By substitution t = a + (x − a)v equation (3.4.31) becomes

Z 1
0
y (x) = f (x) + (x − a) k(x, a + (x − a)v)y(a + (x − a)v)dv.
0

Applying the quadrature rule (3.2.2) to above equation we get

 
0 11 x−a (x − a)
y (x) = f (x) + (x − a) y0 + f (a) k(x, a) + kt (x, a)y0
27 18 18
   
16 3 3
+ k x, a + (x − a) y a + (x − a) .
27 4 4

Integrating above equation between the limits a to x we get,

x x 
s−a
Z Z
11
y(x) − y0 = f (s)ds + y0 + f (a) (s − a)k(s, a)ds
a a 27 18
16 x
Z    
3 3
+ (s − a)k s, a + (s − a) y a + (s − a) ds
27 a 4 4
Z x
1
+ y0 (s − a)2 kt (s, a)ds. (3.4.32)
18 a
3.4 Applications of Corrected Radau Quadrature 40

Equation (3.4.32) represents the solution y(x) if last but one term in equation (3.4.32)

is known explicitly.

Consider
Z x    
3 3
I= (s − a)k s, a + (s − a) y a + (s − a) ds.
a 4 4

Converting the above integral into standard form (3.2.2), by using substitution s =

a + (x − a)w and then applying quadrature rule (3.2.2) to resulting equation we get

(x − a)2
   
4 2 3 9 9
I= y0 k(a, a) + (x − a) k a + (x − a), a + (x − a) y a + (x − a) .
18 9 4 16 16

Let
 
9 9
y a + (x − a) = y(a) + (x − a)y 0 (a),
16 16
9
= y0 + (x − a)f (a).
16

Thus

(x − a)2
  
3 9
I = k(a, a) + 8k a + (x − a), a + (x − a) y0
18 4 16
3
 
(x − a) 3 9
+ k a + (x − a), a + (x − a) f (a).
4 4 16

With above value of I , equation (3.4.32) becomes

 
4 3 3 9
y(x) = (x − a) k a + (x − a), a + (x − a) f (a)
27 4 16
3
   
2 2 3 9
+ 1 + 5 (x − a) k(a, a) + 8k a + (x − a), a + (x − a) y0
3 4 16
Z x Z x 
11 s−a
+ f (s)ds + y0 + f (a) (s − a)k(s, a)ds
a a 27 18
Z x
1
+ y0 (s − a)2 kt (s, a)ds. (3.4.33)
18 a

Thus equation (3.4.33) gives approximate solution of VIDE (3.4.31).

Example 3.4.5. Consider a linear VIDE

Z x
y 0 (x) = 1 + y(t)dt, (3.4.34)
0
3.5 Conclusions 41

with initial condition y(0) = 1.

Analytic solution of the problem is y(x) = ex .

Here we have, f (x) = 1, k(x, t) = 1 and a = 0.

Then from equation (3.4.33) approximate solution of above problem is

x2 19 x3
y(x) = 1 + x + + .
2! 9 3!

3.5 Conclusions

A third order corrected Radau quadrature rule is derived by method of undetermined

coefficients. Upper and lower error bounds are obtained by defining piecewise contin-

uous polynomials. With suitable example optimality of the error bounds is proved.

A third order numerical method for first order ODE is proposed. The method is

suggested in the light of corrected quadrature formula. It is proved that the proposed

method is optimal. By the test example it is shown that the numerical results ob-

tained from the method give better approximations as compared to third order R-K

method. Method of approximate solutions for FIE, FIDE and VIDE are also derived

from corrected Radau quadrature rule. Proposed methods are illustrated by some

examples.