Solution of Non-Homogeneous Fredholm Integral Equation of Second

kind by Gauss-Chebyshev Quadrature Method

Aastha, Moumita, Muskan, Vishal

Instructor: Dr.Subhash Chandra Martha

Department of Mathematics
Indian Institute of Technology Ropar

April 23, 2024

Aastha, Moumita, Muskan, Vishal 1 / 18

Contents

Introduction
Gauss Quadrature Method
Gauss-Chebyshev Quadrature Method
Problem and Solution

Aastha, Moumita, Muskan, Vishal 2 / 18

Introduction

Gauss Quadrature Method

Gauss Quadrature method aims to find the “least” number of fixed
points to approximate the integral of a function f ∶ [−1, 1] → R
such that:
1 n
I =∫ f (x)dx = ∑ wi f (xi )
−1 i=1

where 1 ≤ i ≤ n ∶ xi ∈ [−1, 1] and wi ∈ R.

Since, a general polynomial of degree 2n − 1 has 2n coefficients, it
is possible to find a Gauss integration scheme with n number of
integration points and n number of associated weights to exactly
integrate that polynomial function on the interval [−1, 1].

Aastha, Moumita, Muskan, Vishal 3 / 18

Introduction (contd..)
Gauss-Chebyshev Quadrature Method
Gauss-Chebyshev Quadrature is an extension of Gaussian
quadrature method for approximating the value of integrals of the
following kind:
1 f (x)
∫ √ dx
−1 1 − x2
We approximate
1 f (x) n
∫ √ dx ≈ ∑ wi f (xi ) ..(1)
−1 1 − x2 i=1

Aastha, Moumita, Muskan, Vishal 4 / 18

Introduction (contd..)
Gauss-Chebyshev Quadrature Method
Gauss-Chebyshev Quadrature is an extension of Gaussian
quadrature method for approximating the value of integrals of the
following kind:
1 f (x)
∫ √ dx
−1 1 − x2
We approximate
1 f (x) n
∫ √ dx ≈ ∑ wi f (xi ) ..(1)
−1 1 − x2 i=1

where the nodes xi′ s are the roots of the Chebyshev polynomial:

Tn (x) = cos(ncos −1 (x))

i.e
cos(ncos −1 (x)) = 0
Aastha, Moumita, Muskan, Vishal 4 / 18
Introduction (contd..)
Gauss-Chebyshev Quadrature Rule (contd..)
(2i − 1)π
Ô⇒ cos −1 (x) = , for i = 1, 2, ....n.
2n

(2i − 1)π
Ô⇒ x = cos ( ), for i = 1, 2, ....n.
2n

Aastha, Moumita, Muskan, Vishal 5 / 18

Introduction (contd..)
Gauss-Chebyshev Quadrature Rule (contd..)
(2i − 1)π
Ô⇒ cos −1 (x) = , for i = 1, 2, ....n.
2n

(2i − 1)π
Ô⇒ x = cos ( ), for i = 1, 2, ....n.
2n
Substituting the above values of xi in equation (1), we get
+1 f (x) n
(2i − 1)π
∫ √ dx ≈ ∑ wi f (cos ( ))
−1 1−x 2
i=1 2n
After solving this, we get
π
wi = , i = 1, 2, ...n.
n
which are called ’weights’.

Aastha, Moumita, Muskan, Vishal 5 / 18

Problem
Question
Find out the solution of the Non-homogeneous Fredhlom Integral
Equation of second kind
1
ϕ(x) − ∫ (xt + x 2 t 2 ) ϕ(t)dt = 1, −1 ≤ x ≤ 1
−1

numerically using Gauss Chebyshev quadrature rule.

1 The exact solution of given Integral Equation is
ϕ(x) = 1 + 10x 2 /9. For what values of n (degree of the
polynomials) the approximate and exact values coincide upto
five decimal places. Give a table for this.
2 Then draw the graphs for approximate solutions and exact
solutions verses x for comparison.
3 Then draw the graph for the absolute error (=absolute
difference between exact and approximate solutions) verses x.

Aastha, Moumita, Muskan, Vishal 6 / 18

Solution

Given IE:
1
ϕ(x) − ∫ (xt + x 2 t 2 ) ϕ(t)dt = 1, −1 ≤ x ≤ 1.
−1

Or
1 1 √
ϕ(x) − ∫ √ 1 − t 2 (xt + x 2 t 2 ) ϕ(t)dt = 1, −1 ≤ x ≤ 1.
−1 1 − t2

Aastha, Moumita, Muskan, Vishal 7 / 18

Solution

Given IE:
1
ϕ(x) − ∫ (xt + x 2 t 2 ) ϕ(t)dt = 1, −1 ≤ x ≤ 1.
−1

Or
1 1 √
ϕ(x) − ∫ √ 1 − t 2 (xt + x 2 t 2 ) ϕ(t)dt = 1, −1 ≤ x ≤ 1.
−1 1 − t2
Now, by Gauss Chebyshev formula, we can approximate it as
n
ϕ(x) − ∑ wi k(x, ti )ϕ(ti ) = 1 ...(1)
i=1

Where √
k(x, ti ) = 1 − ti 2 (xti + x 2 ti 2 )

Aastha, Moumita, Muskan, Vishal 7 / 18

Solution (contd..)

where
(2i − 1)π
ti = cos ( ), i = 1, 2...n.
2n
and the weight
π
wi = , i = 1, 2...n..
n

Aastha, Moumita, Muskan, Vishal 8 / 18

Solution (contd..)

where
(2i − 1)π
ti = cos ( ), i = 1, 2...n.
2n
and the weight
π
wi =
, i = 1, 2...n..
n
Now, replacing x by tj in equation (1), we get
n
π
ϕ(tj ) − ∑ k(tj , ti )ϕ(ti ) = 1 j = 1, 2...n.
i=1
n

This gives a system of linear equations in n unknowns as

AX = b.

Aastha, Moumita, Muskan, Vishal 8 / 18

Solution (contd..)

Where A is a matrix of order n given as

⎡ 1 − π k(t1 , t1 ) − π k(t1 , t2 ) ⋯ − π k(t1 , tn ) ⎤

⎢ ⎥
⎢ − π nk(t , t ) 1 −nπ k(t , t ) ⋯ − πn k(t , t ) ⎥
⎢ ⎥
A=⎢ n 2 1 n 2 2 n 2 n
⎥
⎢ ⋮ ⋮ ⋮ ⋮ ⎥
⎢ ⎥
⎢ − π k(tn , t1 ) − πn k(tn , t2 ) ⋯ 1 − πn k(tn , tn ) ⎥
⎣ n ⎦

Aastha, Moumita, Muskan, Vishal 9 / 18

Solution (contd..)

Where A is a matrix of order n given as

⎡ 1 − π k(t1 , t1 ) − π k(t1 , t2 ) ⋯ − π k(t1 , tn ) ⎤

⎢ ⎥
⎢ − π nk(t , t ) 1 −nπ k(t , t ) ⋯ − πn k(t , t ) ⎥
⎢ ⎥
A=⎢ n 2 1 n 2 2 n 2 n
⎥
⎢ ⋮ ⋮ ⋮ ⋮ ⎥
⎢ ⎥
⎢ − π k(tn , t1 ) − πn k(tn , t2 ) ⋯ 1 − πn k(tn , tn ) ⎥
⎣ n ⎦
And X and b are n × 1 column vector given as
⎡ ϕ(t1 ) ⎤ ⎡ ⎤
⎢ ⎥ ⎢ 1 ⎥
⎢ ϕ(t ) ⎥ ⎢ ⎥
⎢ 2 ⎥ ⎢ 1 ⎥
X =⎢ ⎥, b=⎢ ⎥
⎢ ⋮ ⎥ ⎢ ⋮ ⎥
⎢ ⎥ ⎢ ⎥
⎢ ϕ(tn ) ⎥ ⎢ ⎥
⎣ ⎦ ⎣ 1 ⎦

Aastha, Moumita, Muskan, Vishal 9 / 18

Matlab Code

Aastha, Moumita, Muskan, Vishal 10 / 18

Matlab Code (contd..)

Aastha, Moumita, Muskan, Vishal 11 / 18

Matlab Code (contd..)

Aastha, Moumita, Muskan, Vishal 12 / 18

Solution (contd..)
(1) The exact solution of given Integral Equation is
ϕ(x) = 1 + 10x 2 /9. For what values of n (degree of the
polynomials) the approximate and exact values coincide upto five
decimal places. Give a table for this.

Aastha, Moumita, Muskan, Vishal 13 / 18

Solution (contd..)
(1) The exact solution of given Integral Equation is
ϕ(x) = 1 + 10x 2 /9. For what values of n (degree of the
polynomials) the approximate and exact values coincide upto five
decimal places. Give a table for this.

Aastha, Moumita, Muskan, Vishal 13 / 18

Solution (contd..)
(2) Draw the graphs for approximate solutions and exact solutions
verses x for comparison.
(3) Draw the graph for the absolute error (=absolute difference
between exact and approximate solutions) verses x.

Aastha, Moumita, Muskan, Vishal 14 / 18

Solution (contd..)
(2) Draw the graphs for approximate solutions and exact solutions
verses x for comparison.
(3) Draw the graph for the absolute error (=absolute difference
between exact and approximate solutions) verses x.

For n = 2, we get

(c) Numerical vs exact (d) Absolute error

Aastha, Moumita, Muskan, Vishal 14 / 18

Solution (contd..)

For n = 5, we get

(e) Numerical vs exact (f) Absolute error

Aastha, Moumita, Muskan, Vishal 15 / 18

Solution (contd..)

For n = 10, we get

(g) Numerical vs exact (h) Absolute error

Aastha, Moumita, Muskan, Vishal 16 / 18

Solution (contd..)

For n = 20, we get

(i) Numerical vs exact (j) Absolute error

Aastha, Moumita, Muskan, Vishal 17 / 18

 Thank You

Aastha, Moumita, Muskan, Vishal 18 / 18