Numerical Method (2067- second batch)

1. Discuss methods of Half-interval and Newton’s for solving the non-linear equation f(x)
=0. Illustrate the methods by figures and compare them stating their advantages and
disadvantages.
Half-Interval method:
Suppose that 𝑓(𝑥) is continuous function in the interval [𝑎0 , 𝑏0 ] and 𝑓(𝑎0 ) 𝑓(𝑏0 ) < 0, then by
intermediate value theorem, there exists a root of 𝑓(𝑥) in the interval (𝑎0 , 𝑏0 ). We calculate the
(𝑎 +𝑏 )
first approximation of this root as 𝑐0 = 0 2 0 . If 𝑓(𝑐0 ) = 0, then 𝑐0 is the root of 𝑓(𝑥). If not
then we bisect the interval [𝑎0 , 𝑏0 ] into two equal length sub-intervals [𝑎0 , 𝑐0 ]& [𝑐0 , 𝑏0 ] and set
𝑎1 = 𝑎0 , 𝑏1 = 𝑐0 if 𝑓(𝑎0 )𝑓(𝑐0 ) < 0 and 𝑎1 = 𝑐0 , 𝑏1 = 𝑏0 if 𝑓(𝑐0 )𝑓(𝑏0 ) < 0. The second
(𝑎1 +𝑏1 )
approximation of the root is now calculated as 𝑐1 = . If 𝑓(𝑐1 ) = 0, then 𝑐1 is the root of
2
𝑓(𝑥). If not then we again bisect the interval [𝑎1 , 𝑏1 ] into two equal length sub-intervals
[𝑎1 , 𝑐1 ]& [𝑐1 , 𝑏1 ] and set 𝑎2 = 𝑎1 , 𝑏2 = 𝑐1 if 𝑓(𝑎1 )𝑓(𝑐1) < 0 & 𝑎2 = 𝑐1 , 𝑏2 = 𝑏1 if
(𝑎2 +𝑏2 )
𝑓(𝑐1 )𝑓(𝑏1 ) < 0 and then calculate the third approximation as 𝑐2 = and continuing the
2
above process.

al

Fig: Half Interval Method

ep
itn

Source: www.csitnepal.com Page 1

cs
 Numerical Method (2067- second batch)

This process of calculating the approximations 𝑐0 , 𝑐1 , 𝑐2 , … is repeated until we find a root of

𝑓(𝑥) or a satisfactory approximation of it.
Advantages:

 This method is guaranteed to work for any continuous function 𝑓(𝑥) on the interval [𝑎, 𝑏]
with 𝑓(𝑎)𝑓(𝑏) < 0.
 The number of iterations required to achieve a specified accuracy is known in advance.
Disadvantage:
 The method converges slowly, i.e., it requires more iterations to achieve the same
accuracy when compared with some other methods for solving non-linear equations.
Newton’s Method:
Let 𝑓(𝑥) be a differentiable function and let 𝑥0 be an initial points which is sufficiently close to
the root of 𝑓(𝑥). Let (𝑥1 , 0) be the point of intersection of the x-axis and the tangent drawn to
the curve 𝑓(𝑥) at (𝑥0 , 𝑓(𝑥0 )). Newton’s method takes this point as the first approximation for
the root of 𝑓(𝑥). To calculate this point we note that the slope of the tangent to 𝑓(𝑥) at 𝑥 = 𝑥0 is
equal to the slope of the line through the points (𝑥1 , 0) and (𝑥0 , 𝑓(𝑥0 )). i.e.
(𝑓(𝑥0 ) − 0) 𝑓(𝑥0 )
𝑓 ′ (𝑥0 ) = ⇒ 𝑥1 = 𝑥0 − ′
𝑥0 − 𝑥1 𝑓 (𝑥0 )

al

Fig: Newton-Raphson Method

ep
itn

Source: www.csitnepal.com Page 2

cs
 Numerical Method (2067- second batch)

If 𝑓(𝑥1 ) = 0, then 𝑥1 is the required root of 𝑓(𝑥). If not, then we take the point of intersection
(𝑥2 , 0) of the x-axis and the tangent to the 𝑓(𝑥) at 𝑥 = 𝑥1 as the next approximation of the root.
As above, we have
𝑓(𝑥1 )
𝑥2 = 𝑥1 −
𝑓 ′ (𝑥1 )
In general, the (𝑛 + 1)𝑡ℎ approximation of the root of 𝑓(𝑥) is given by the formula:
𝑓(𝑥𝑛 )
𝑥𝑛+1 = 𝑥𝑛 − ; 𝑛≥0
𝑓 ′ (𝑥𝑛 )
We continue to calculate the approximations 𝑥1 , 𝑥2 , 𝑥3 , … using the above formula until we find
the root or its satisfactory approximation.
Advantages:
 Unlike the incremental search and bisection methods, the Newton-Raphson method isn’t
fooled by singularities.
 Also, it can identify repeated roots, since it does not look for changes in the sign of 𝑓(𝑥)
explicitly.
 It can find complex roots of polynomials, assuming you start out with a complex value
for 𝑥1 .
 For many problems, Newton-Raphson converges quicker than either bisection or
incremental search.
Disadvantages:
 The Newton-Raphson method only works if you have a functional representation of
𝑓 ′ (𝑥). Some functions may be difficult or impossible to differentiate. You may be able to
𝑓(𝑥+∆𝑥)−𝑓(𝑥)
work around this by approximating the derivative 𝑓 ′ (𝑥) ≈ .
∆𝑥
 The Newton-Raphson method is not guaranteed to find a root.

2. Derive the equation for Lagrange’s interpolating polynomial and find the value of f(x) at
x=1 for the following:
x -1 -2 2 4
f(x) -1 -9 11 69
al

Solution: Here,
ep
itn

Source: www.csitnepal.com Page 3

cs
 Numerical Method (2067- second batch)

(𝑥 − 𝑥1 )(𝑥 − 𝑥2 )(𝑥 − 𝑥3 ) (𝑥 + 2)(𝑥 − 2)(𝑥 − 4)

𝑙0 (𝑥) = =
(𝑥0 − 𝑥1 )(𝑥0 − 𝑥2 )(𝑥0 − 𝑥3 ) (−1 + 2)(−1 − 2)(−1 − 4)
1
= (𝑥 + 2)(𝑥 − 2)(𝑥 − 4)
15
(𝑥 − 𝑥0 )(𝑥 − 𝑥2 )(𝑥 − 𝑥3 ) (𝑥 + 1)(𝑥 − 2)(𝑥 − 4)
𝑙1 (𝑥) = =
(𝑥1 − 𝑥0 )(𝑥1 − 𝑥2 )(𝑥1 − 𝑥3 ) (−2 + 1)(−2 − 2)(−2 − 4)
1
= − (𝑥 + 1)(𝑥 − 2)(𝑥 − 4)
24
(𝑥 − 𝑥0 )(𝑥 − 𝑥1 )(𝑥 − 𝑥3 ) (𝑥 + 1)(𝑥 + 2)(𝑥 − 4) 1
𝑙2 (𝑥) = = = − (𝑥 + 1)(𝑥 + 2)(𝑥 − 4)
(𝑥2 − 𝑥0 )(𝑥2 − 𝑥1 )(𝑥2 − 𝑥3 ) (2 + 1)(2 + 2)(2 − 4) 24
(𝑥 − 𝑥0 )(𝑥 − 𝑥1 )(𝑥 − 𝑥2 ) (𝑥 + 1)(𝑥 + 2)(𝑥 − 2) 1
𝑙3 (𝑥) = = = (𝑥 + 1)(𝑥 + 2)(𝑥 − 2)
(𝑥3 − 𝑥0 )(𝑥3 − 𝑥1 )(𝑥3 − 𝑥2 ) (4 + 1)(4 + 2)(4 − 2) 60
And Lagrange’s interpolating polynomial is given by:
𝑃3 (𝑥) = 𝑓0 𝑙0 (𝑥) + 𝑓1 𝑙1 (𝑥) + 𝑓2 𝑙2 (𝑥) + 𝑓3 𝑙3 (𝑥)
1 (−1)
𝑃3 (𝑥) = (−1) (𝑥 + 2)(𝑥 − 2)(𝑥 − 4) + (−9) (𝑥 + 1)(𝑥 − 2)(𝑥 − 4)
15 24
(−1) 1
+ 11 (𝑥 + 1)(𝑥 + 2)(𝑥 − 4) + 69 (𝑥 + 1)(𝑥 + 2)(𝑥 − 2)
24 60
1
∴ 𝑃3 (𝑥) = 𝑥 3 + 𝑥 +
4
Now, the value of 𝑓(𝑥) at 𝑥 = 1 is given as;
1 9
𝑃3 (1) = 13 + 1 + =
4 4

3. Write Newton-cotes integration formulas in basic form for x=1, 2, 3 and give their
𝟏 𝟐
composite rules. Evaluate ∫𝟎 𝒆−𝒙 𝒅𝒙 using the Gaussian integration three point formula.
𝑏
To find the value of ∫𝑎 𝑓(𝑥)𝑑𝑥 numerically using the Newton-Cotes method, we first of all
divide the interval [𝑎, 𝑏] into 𝑛 equal parts of length ℎ by points 𝑥𝑖 = 𝑎 + 𝑖ℎ, 𝑖 = 0, 1, 2, … , 𝑛
(𝑏−𝑎)
where ℎ = . Then 𝑎 = 𝑥0 < 𝑥1 < 𝑥2 < ⋯ < 𝑥𝑛 = 𝑏 forms a partition of [𝑎, 𝑏]. Let 𝑃𝑛 (𝑥)
𝑛
be the interpolating polynomial of 𝑓(𝑥) interpolating at 𝑛 + 1 points (𝑥𝑖 , 𝑓𝑖 ), 𝑖 = 0, 1, 2, … , 𝑛
where 𝑓𝑖 = 𝑓(𝑥𝑖 ). Then 𝑃𝑛 (𝑥) is given by the formula
𝑆(𝑆 − 1) 2 𝑆(𝑆 − 1) … (𝑆 − 𝑛 + 1) 𝑛
al

𝑃𝑛 (𝑥) = 𝑓0 + 𝑆∆𝑓0 + ∆ 𝑓0 + ⋯ + ∆ 𝑓0
2 𝑛!
ep

𝑥−𝑥0
Where, 𝑆 = & ∆𝑗 𝑓0 = ∆𝑗−1 𝑓1 − ∆𝑗−1 𝑓0 are the 𝑗 𝑡ℎ forward differences.
𝑛
itn

Source: www.csitnepal.com Page 4

cs
 Numerical Method (2067- second batch)
𝑏 𝑏
We now approximate the value of ∫𝑎 𝑓(𝑥)𝑑𝑥 by ∫𝑎 𝑃𝑛 (𝑥)𝑑𝑥.
𝑏 𝑏
Therefore, ∫𝑎 𝑓(𝑥)𝑑𝑥 ≈ ∫𝑎 𝑃𝑛 (𝑥)𝑑𝑥
𝑏
𝑆(𝑠 − 1) 2 𝑆(𝑠 − 1) … (𝑆 − 𝑛 + 1) 𝑛
= ∫ [𝑓0 + 𝑆∆𝑓0 + ∆ 𝑓0 + ⋯ + ∆ 𝑓0 ] 𝑑𝑥
𝑎 2! 𝑛!
𝑏
Which is the Newton-Cotes formula for numerically evaluated ∫𝑎 𝑓(𝑥)𝑑𝑥.

Numerical:
(1−0)𝑦+1+0
Let 𝑥 = = 0.5𝑦 + 0.5
2

Then the limit of integration are changed from (0, 1) to (−1, 1) so that
1
−𝑥 2
1 − 0 1 −(0.5𝑦+0.5)2
∫ 𝑒 𝑑𝑥 = ∫ 𝑒 𝑑𝑦
0 2 −1
Using the Gaussian 3-point formula, we get
1
2
∫ 𝑒 −(0.5𝑦+0.5) 𝑑𝑦
−1
2 2
= 0.55556 × 𝑒 −(0.5×(−0.77460)+0.5) + 0.88889 × 𝑒 −(0.5×0+0.5) + 0.55556
2
× 𝑒 −(0.5×0.77460+0.5)
= 0.54855 + 0.69227 + 0.25282 = 1.49364
1
2 1−0
∴ ∫ 𝑒 −𝑥 𝑑𝑥 = × 1.49364 = 0.74682
0 2

4. Solve the following system of algebraic linear equation using Gauss-Jordan algorithm.
𝟎 𝟐 𝟎 𝟏 𝐱𝟏 𝟎
𝟐 𝟐 𝟑 𝟐 𝐱𝟐 −𝟐
( ) (𝐱 ) = ( )
𝟒 −𝟑 𝟎 𝟏 𝟑 −𝟕
𝟔 𝟏 −𝟔 −𝟓 𝐱𝟒 𝟔
The augmented matrix of the system is as follow:
al
ep
itn

Source: www.csitnepal.com Page 5

cs
 Numerical Method (2067- second batch)

0 2 0 1 0
2 2 3 2 −2
[ ]
4 −3 0 1 −7
6 1 −6 −5 6
Interchanging first row with second row: 𝑅1 ↔ 𝑅2
2 2 3 2 −2
0 2 0 1 0
[ ]
4 −3 0 1 −7
6 1 −6 −5 6
1
Normalize the first row: 𝑅1 → 2 𝑅1

3
1 1 1 −1
2
0 2 0 1 0
4 −3 0 1 −7
[6 1 −6 −5 6 ]
Eliminate 𝑥1 from 2nd, 3rd and 4th row: 𝑅2 → 𝑅2 ; 𝑅3 → 𝑅3 − 4𝑅1 ; 𝑅4 → 𝑅4 − 6𝑅1
3
1 1 1 −1
2
0 2 0 1 0
0 −7 −6 −3 −3
[0 −5 −15 −11 12 ]
1
Normalize the second row: 𝑅2 → 2 𝑅2

3
1 1 1 −1
2
1
0 1 0 0
2
0 −7 −6 −3 −3
[0 −5 −15 −11 12 ]
Eliminate 𝑥2 from 1st, 3rd and 4th row: 𝑅1 → 𝑅1 − 𝑅2 ; 𝑅3 → 𝑅3 + 7𝑅2 ; 𝑅4 → 𝑅4 + 5𝑅2
3 1
1 0 −1
2 2
1
0 1 0 0
2
1
0 0 −6 −3
2
17
al

[0 0 −15 −
2
12 ]
ep

1
Normalize the third row: 𝑅3 → − 6 𝑅3
itn

Source: www.csitnepal.com Page 6

cs
 Numerical Method (2067- second batch)

31
1 0 −1
22
1
0 1 0 0
2
1 1
0 0 1 −
12 2
17
[0 0 −15 −
2
12 ]

3
Eliminating 𝑥3 from 1st, 2nd and 4th row: 𝑅1 → 𝑅1 − 2 𝑅3 ; 𝑅2 → 𝑅2 ; 𝑅4 → 𝑅4 + 15𝑅3

5 7
1 0 0 −
8 4
1
0 1 0 0
2
1 1
0 0 1 −
12 2
117 39
[0 0 0 −
12 2 ]
12
Normalize the fourth row: 𝑅4 → − 117 𝑅4

5 7
1 0 0 −
8 4
1
0 1 0 0
2
1 1
0 0 1 −
12 2
[0 0 0 1 −2 ]
5 1 1
Eliminating 𝑥4 from 1st, 2nd and 3rd row: 𝑅1 → 𝑅1 − 8 𝑅4 ; 𝑅2 → 𝑅2 − 2 𝑅4 ; 𝑅3 → 𝑅3 + 12 𝑅4

1
1 0 0 0 −
2
0 1 0 0 1
1
0 0 1 0
3
[0 0 0 1 −2 ]
1 1
Therefore, the solution is 𝑥1 = − 2 ; 𝑥2 = 1 ; 𝑥3 = 3 ; 𝑥4 = −2
al
ep
itn

Source: www.csitnepal.com Page 7

cs
 Numerical Method (2067- second batch)

5. Write an algorithm and computer program to solve system of linear equation using
Gauss-Seidel iterative method.
Algorithm:
Input:
A diagonally dominant system of linear equations 𝐴𝑥 = 𝑏
Process:
𝑏
FOR 𝑖 = 1 TO 𝑛 SET 𝑥𝑖 = 𝑎 𝑖
𝑖𝑖

BEGIN: SET 𝑘𝑒𝑦 = 0

FOR 𝑖 = 1 TO 𝑛
{ SET 𝑠𝑢𝑚 = 𝑏𝑖
FOR 𝑗 = 1 TO 𝑛 AND 𝑗 = 1

{ SET 𝑠𝑢𝑚 = 𝑠𝑢𝑚 − (𝑎𝑖𝑗 ∗ 𝑥𝑗 )

}
SET 𝑑𝑢𝑚𝑚𝑦 = 𝑠𝑢𝑚/𝑎𝑖𝑖
𝑑𝑢𝑚𝑚𝑦−𝑥𝑖
IF 𝑘𝑒𝑦 = 0 AND | | > 𝑒𝑟𝑟𝑜𝑟
𝑑𝑢𝑚𝑚𝑦

THEN
SET 𝑘𝑒𝑦 = 1
SET 𝑥𝑖 = 𝑑𝑢𝑚𝑚𝑦
}
IF 𝑘𝑒𝑦 = 1 THEN
GOTO BEGIN
Output:
Approximate solution 𝑥𝑖 ; 𝑖 = 1, 2, 3, … , 𝑛 of 𝐴𝑥 = 𝑏

Computer program:
al

#include<iostream.h>
#include<conio.h>
ep
itn

Source: www.csitnepal.com Page 8

cs
 Numerical Method (2067- second batch)

#include<iomanip.h>
#include<math.h>
#define MAXIT 50
#define EPS 0.000001
void gaseid(int n, float a[10][10], float b[10], float x[10], int *count, int *status);
void main()
{ float a[10][10], b[10], x[10];
int i, j, n, count, status;
cout<<"** SOLUTION BY GUASS SEIDEL ITERATION METHOD **"<<endl;
cout<<"input the size of the system:"<<endl;
cin>>n;
cout<<"input coefficients, a(i.j)"<<endl;
cout<<"one row on each line"<<endl;
for(i=1; i<=n; i++)
for(j=1; j<=n; j++)
cin>>a[i][j];
cout<<"input vector b:"<<endl;
for(i=1; i<=n; i++)
cin>>b[i];
gaseid(n, a, b, x, &count, &status);
if(status==2)
{ cout<<"no CONVERGENCE in "<<MAXIT<<"
iterations."<<endl<<endl<<endl;
}
else
{ cout<<"SOLUTION VECTOR X"<<endl;
al

for(i=1; i<=n; i++)

cout<<setw(15.6)<<x[i]<<endl;
ep
itn

Source: www.csitnepal.com Page 9

cs
 Numerical Method (2067- second batch)

cout<<"iterations= "<<count;
}
getch();
}
void gaseid(int n, float a[10][10], float b[10], float x[10], int *count, int *status)
{ int i, j, key;
float sum, x0[10];
for(i=1; i<=n; i++)
x0[i]=b[i]/a[i][i];
*count=1;
begin:
key=0;
for(i=1; i<=n; i++)
{ sum=b[i];
for(j=1; j<=n; j++)
{ if(i==j)
continue;
sum=sum-a[i][j]*x0[j];
}
x[i]=sum/a[i][i];
if(key==0)
{ if(fabs((x[i]-x0[i])/x[i])>EPS)
key=1;
}
}
if(key==1)
al

{ if(*count==MAXIT)
ep

{ *status=2;
itn

Source: www.csitnepal.com Page 10

cs
 Numerical Method (2067- second batch)

return;
}
else
{ *status=1;
for(i=1; i<=n; i++)
x0[i]=x[i];
}
*count=*count+1;
goto begin;
}
return;
}

6. Explain the Picard’s proves of successive approximation. Obtain a solution up to the fifth
𝒅𝟐 𝒚
approximation of the equation = 𝒚 + 𝒙 such that 𝒚 = 𝟏 when 𝒙 = 𝟎 using Picard’s
𝒅𝒙
process of successive approximation.
𝑑𝑦
Suppose that we are given a differential equation of the form 𝑑𝑥 = 𝑓(𝑥, 𝑦) ; 𝑦(𝑥0 ) = 𝑦.

Then 𝑑𝑦 = 𝑓(𝑥, 𝑦)𝑑𝑥

Integrating both sides of above equation in the interval (𝑥0 , 𝑥), we get
𝑥 𝑥
∫ 𝑑𝑦 = ∫ 𝑓(𝑡, 𝑦(𝑡))𝑑𝑡
𝑥0 𝑥0
𝑥
𝑦(𝑥) − 𝑦(𝑥0 ) = ∫ 𝑓(𝑡, 𝑦(𝑡))𝑑𝑡
𝑥0
𝑥
𝑦(𝑥) = 𝑦(𝑥0 ) + ∫ 𝑓(𝑡, 𝑦(𝑡))𝑑𝑡 … … … (𝑖)
𝑥0

Now to solve equation (i), we use the method of iteration as follows:

We replace 𝑦(𝑡) on the right of equation (i) by 𝑦, and calculate the first approximation 𝑦1 (𝑥) of
al

𝑥
𝑦(𝑥) as 𝑦1 (𝑥) = 𝑦0 + ∫𝑥 𝑓(𝑡, 𝑦0 )𝑑𝑡
ep

0
itn

Source: www.csitnepal.com Page 11

cs
 Numerical Method (2067- second batch)

The second approximation 𝑦2 (𝑥) of 𝑦(𝑥) is calculated by substituting 𝑦(𝑡) on the right of
𝑥
equation (i) as 𝑦2 (𝑥) = 𝑦0 + ∫𝑥 𝑓(𝑡, 𝑦1 (𝑡))𝑑𝑡
0

Proceeding similarly, the 𝑛𝑡ℎ approximation of 𝑦(𝑥) is given by the iteration 𝑦𝑛 (𝑥) = 𝑦0 +
𝑥
∫𝑥 𝑓(𝑡, 𝑦𝑛−1 (𝑡))𝑑𝑡
0

This iterative method of solving the differential equation is known as Picard’s Method.
Numerical:
𝑑𝑦
For 𝑑𝑥 = 𝑦 + 𝑥 when 𝑦 = 1 & 𝑥 = 0 ; 𝑖. 𝑒. 𝑦(𝑥) = 𝑦(0) = 1

Picard’s iteration method is given by

𝑥 𝑥
𝑦𝑛 (𝑥) = 𝑦0 + ∫ 𝑓(𝑡, 𝑦𝑛−1 (𝑡))𝑑𝑡 = 1 + ∫ 𝑓(𝑡, 𝑦𝑛−1 (𝑡))𝑑𝑡
𝑥0 0

When 𝑥 = 1, we get
𝑥 𝑥 𝑥
(𝑡 + 1)2 (𝑥 + 1)2 1
𝑦1 (𝑥) = 1 + ∫ 𝑓(𝑡, 𝑦0 (𝑡))𝑑𝑡 = 1 + ∫ (𝑡 + 1)𝑑𝑡 = 1 + [ ] =1+ −
0 0 2 0
2 2
𝑥2
=1+𝑥+
2
When 𝑥 = 2, we get
𝑥 𝑥 𝑥
𝑡2 𝑡3
𝑦2 (𝑥) = 1 + ∫ 𝑓(𝑡, 𝑦1 (𝑡))𝑑𝑡 = 1 + ∫ (𝑡 + 1 + 𝑡 + ) 𝑑𝑡 = 1 + [𝑡 + 𝑡 2 + ]
0 0 2 6 0
𝑥3 2
=1+𝑥+𝑥 +
6
When 𝑥 = 3, we get
𝑥 𝑥 𝑥
𝑡3 𝑡3 𝑡4
𝑦3 (𝑥) = 1 + ∫ 𝑓(𝑡, 𝑦2 (𝑡))𝑑𝑡 = 1 + ∫ (𝑡 + 1 + 𝑡 + 𝑡 + ) 𝑑𝑡 = 1 + [𝑡 + 𝑡 2 + + ]
2
0 0 6 3 24 0
𝑥3 𝑥4
= 1 + 𝑥 + 𝑥2 + +
3 24
When 𝑥 = 4, we get
𝑥 𝑥
𝑡3 𝑡4
𝑦4 (𝑥) = 1 + ∫ 𝑓(𝑡, 𝑦3 (𝑡))𝑑𝑡 = 1 + ∫ (𝑡 + 1 + 𝑡 + 𝑡 2 + + ) 𝑑𝑡
0 0 3 24
3 4 𝑥
𝑡 𝑡 𝑡5 𝑥3 𝑥4 𝑥5
al

2
= 1 + [𝑡 + 𝑡 + + + ] = 1 + 𝑥 + 𝑥2 + + +
3 12 120 0 3 12 120
ep

When 𝑥 = 5, we get
itn

Source: www.csitnepal.com Page 12

cs
 Numerical Method (2067- second batch)
𝑥 𝑥
𝑡3 𝑡4 𝑡5
𝑦5 (𝑥) = 1 + ∫ 𝑓(𝑡, 𝑦4 (𝑡))𝑑𝑡 = 1 + ∫ (𝑡 + 1 + 𝑡 + 𝑡 2 + + + ) 𝑑𝑡
0 0 3 12 120
3 4 𝑥
𝑡
2
𝑡 𝑡5 𝑡6 2
𝑥3 𝑥4 𝑥5 𝑥6
= 1 + [𝑡 + 𝑡 + + + + ] = 1+𝑥+𝑥 + + + +
3 12 60 720 0 3 12 60 720

7. Derive a difference equation to represent a Laplace’s equation. Solve the following

Laplace equation:
𝝏𝟐 𝒖 𝝏𝟐 𝒖
+ 𝝏𝒚𝟐 = 𝟎 Within 𝟎 ≤ 𝒙 ≤ 𝟑, 𝟎 ≤ 𝒚 ≤ 𝟑
𝝏𝒙𝟐

For the rectangular plate given as:

Difference equation to represent Laplace’s equation:

Let 𝑢 = 𝑢(𝑥, 𝑦) be a function of two independent variables 𝑥 & 𝑦. Then by Taylor’s formula:
ℎ2 ℎ3
𝑢(𝑥 + ℎ, 𝑦) = 𝑢(𝑥, 𝑦) + ℎ𝑢𝑥 (𝑥, 𝑦) + 𝑢 (𝑥, 𝑦) + 𝑢𝑥𝑥𝑥 (𝑥, 𝑦) + ⋯ … … … (𝑖)
2! 𝑥𝑥 3!
ℎ2 ℎ3
𝑢(𝑥 − ℎ, 𝑦) = 𝑢(𝑥, 𝑦) − ℎ𝑢𝑥 (𝑥, 𝑦) + 𝑢𝑥𝑥 (𝑥, 𝑦) − 𝑢𝑥𝑥𝑥 (𝑥, 𝑦) + ⋯ … … … (𝑖𝑖)
2! 3!
𝑘2 𝑘3
𝑢(𝑥, 𝑦 + 𝑘) = 𝑢(𝑥, 𝑦) + 𝑘𝑢𝑦 (𝑥, 𝑦) + 𝑢𝑦𝑦 (𝑥, 𝑦) + 𝑢𝑦𝑦𝑦 (𝑥, 𝑦) + ⋯ … … … (𝑖𝑖𝑖)
2! 3!
𝑘2 𝑘3
𝑢(𝑥, 𝑦 − 𝑘) = 𝑢(𝑥, 𝑦) − 𝑘𝑢𝑦 (𝑥, 𝑦) + 𝑢𝑦𝑦 (𝑥, 𝑦) − 𝑢𝑦𝑦𝑦 (𝑥, 𝑦) + ⋯ … … … (𝑖𝑣)
2! 3!
Adding equations (i) & (ii) and ignoring the terms containing ℎ4 and higher powers, we get
𝑢(𝑥 + ℎ, 𝑦) + 𝑢(𝑥 − ℎ, 𝑦) = 2𝑢(𝑥, 𝑦) + ℎ2 𝑢𝑥𝑥 (𝑥, 𝑦)
al

1
𝑜𝑟, 𝑢𝑥𝑥 (𝑥, 𝑦) = [𝑢(𝑥 + ℎ, 𝑦) − 2𝑢(𝑥, 𝑦) + 𝑢(𝑥 − ℎ, 𝑦)] … … … (𝐴)
ep

ℎ2
itn

Source: www.csitnepal.com Page 13

cs
 Numerical Method (2067- second batch)

Adding equations (iii) & (iv) and ignoring the terms containing 𝑘 4 and higher powers, we get
𝑢(𝑥, 𝑦 + 𝑘) + 𝑢(𝑥, 𝑦 − 𝑘) = 2𝑢(𝑥, 𝑦) + 𝑘 2 𝑢𝑦𝑦 (𝑥, 𝑦)
1
𝑜𝑟, 𝑢𝑦𝑦 (𝑥, 𝑦) = [𝑢(𝑥, 𝑦 + 𝑘) − 2𝑢(𝑥, 𝑦) + 𝑢(𝑥, 𝑦 − 𝑘)] … … … (𝐵)
𝑘2
Now if 𝑢𝑥𝑥 + 𝑢𝑦𝑦 = 0 is the given Laplace’s equation, then from equation (A) & (B) we have
1 1
[𝑢(𝑥 + ℎ, 𝑦) − 2𝑢(𝑥, 𝑦) + 𝑢(𝑥 − ℎ, 𝑦)] + [𝑢(𝑥, 𝑦 + 𝑘) − 2𝑢(𝑥, 𝑦) + 𝑢(𝑥, 𝑦 − 𝑘)] = 0
ℎ2 𝑘2
Choosing ℎ = 𝑘, we get
𝑢(𝑥 + ℎ, 𝑦) + 𝑢(𝑥, 𝑦 + ℎ) + 𝑢(𝑥 − ℎ, 𝑦) + 𝑢(𝑥, 𝑦 − ℎ) − 4𝑢(𝑥, 𝑦) = 0
1
∴ 𝑢(𝑥, 𝑦) = [𝑢(𝑥 + ℎ, 𝑦) + 𝑢(𝑥, 𝑦 + ℎ) + 𝑢(𝑥 − ℎ, 𝑦) + 𝑢(𝑥, 𝑦 − ℎ)]
4
is the difference equation for Laplace’s equation.

Numerical:
From the difference equation for Laplace’s equation, we have
200 + 200 + 𝑢2 + 𝑢3 − 4𝑢1 = 0 ⇒ −4𝑢1 + 𝑢2 + 𝑢3 = −400 … … … (𝑖)
200 + 100 + 𝑢4 + 𝑢1 − 4𝑢2 = 0 ⇒ 𝑢1 − 4𝑢2 + 𝑢4 = −300 … … … (𝑖𝑖)
𝑢1 + 200 + 100 + 𝑢4 − 4𝑢3 = 0 ⇒ 𝑢1 − 4𝑢3 + 𝑢4 = −300 … … … (𝑖𝑖𝑖)
𝑢2 + 𝑢3 + 100 + 100 − 4𝑢4 = 0 ⇒ 𝑢2 + 𝑢3 − 4𝑢4 = −200 … … … (𝑖𝑣)
Solving the equations (𝑖), (𝑖𝑖), (𝑖𝑖𝑖) & (𝑖𝑣), we get
𝑢1 = 175
𝑢2 = 𝑢3 = 150
𝑢4 = 125

(OR) 7. Derive a difference equation to represent Poisson’s equation. Solve the Poisson’s
equation 𝛁 𝟐 𝒇 = 𝟐𝒙𝟐 𝒚𝟐 over the square to main 𝟎 ≤ 𝒙 ≤ 𝟑, 𝟎 ≤ 𝒚 ≤ 𝟑 with 𝒇 = 𝟎 on the
boundary and 𝒉 = 𝟏.
al

Difference equation to represent Poisson’s equation:

ep

Let 𝑢 = 𝑢(𝑥, 𝑦) be a function of two independent variables 𝑥 & 𝑦. Then by Taylor’s formula:
itn

Source: www.csitnepal.com Page 14

cs
 Numerical Method (2067- second batch)

ℎ2 ℎ3
𝑢(𝑥 + ℎ, 𝑦) = 𝑢(𝑥, 𝑦) + ℎ𝑢𝑥 (𝑥, 𝑦) + 𝑢𝑥𝑥 (𝑥, 𝑦) + 𝑢𝑥𝑥𝑥 (𝑥, 𝑦) + ⋯ … … … (𝑖)
2! 3!
ℎ2 ℎ3
𝑢(𝑥 − ℎ, 𝑦) = 𝑢(𝑥, 𝑦) − ℎ𝑢𝑥 (𝑥, 𝑦) + 𝑢𝑥𝑥 (𝑥, 𝑦) − 𝑢𝑥𝑥𝑥 (𝑥, 𝑦) + ⋯ … … … (𝑖𝑖)
2! 3!
𝑘2 𝑘3
𝑢(𝑥, 𝑦 + 𝑘) = 𝑢(𝑥, 𝑦) + 𝑘𝑢𝑦 (𝑥, 𝑦) + 𝑢𝑦𝑦 (𝑥, 𝑦) + 𝑢𝑦𝑦𝑦 (𝑥, 𝑦) + ⋯ … … … (𝑖𝑖𝑖)
2! 3!
𝑘2 𝑘3
𝑢(𝑥, 𝑦 − 𝑘) = 𝑢(𝑥, 𝑦) − 𝑘𝑢𝑦 (𝑥, 𝑦) + 𝑢𝑦𝑦 (𝑥, 𝑦) − 𝑢𝑦𝑦𝑦 (𝑥, 𝑦) + ⋯ … … … (𝑖𝑣)
2! 3!
Adding equations (i) & (ii) and ignoring the terms containing ℎ4 and higher powers, we get
𝑢(𝑥 + ℎ, 𝑦) + 𝑢(𝑥 − ℎ, 𝑦) = 2𝑢(𝑥, 𝑦) + ℎ2 𝑢𝑥𝑥 (𝑥, 𝑦)
1
𝑜𝑟, 𝑢𝑥𝑥 (𝑥, 𝑦) = [𝑢(𝑥 + ℎ, 𝑦) − 2𝑢(𝑥, 𝑦) + 𝑢(𝑥 − ℎ, 𝑦)] … … … (𝐴)
ℎ2
Adding equations (iii) & (iv) and ignoring the terms containing 𝑘 4 and higher powers, we get
𝑢(𝑥, 𝑦 + 𝑘) + 𝑢(𝑥, 𝑦 − 𝑘) = 2𝑢(𝑥, 𝑦) + 𝑘 2 𝑢𝑦𝑦 (𝑥, 𝑦)
1
𝑜𝑟, 𝑢𝑦𝑦 (𝑥, 𝑦) = [𝑢(𝑥, 𝑦 + 𝑘) − 2𝑢(𝑥, 𝑦) + 𝑢(𝑥, 𝑦 − 𝑘)] … … … (𝐵)
𝑘2
Now if 𝑢𝑥𝑥 + 𝑢𝑦𝑦 = 𝑔(𝑥, 𝑦) is the given Poisson’s equation, then from equation (A) & (B)
choosing ℎ = 𝑘 we have,
𝑢(𝑥 + ℎ, 𝑦) + 𝑢(𝑥, 𝑦 + ℎ) + 𝑢(𝑥 − ℎ, 𝑦) + 𝑢(𝑥, 𝑦 − ℎ) − 4𝑢(𝑥, 𝑦) = ℎ2 𝑔(𝑥, 𝑦)
which is the difference equation for Poisson’s equation.
Numerical:
The domain is divided as follows with 𝑓 = 0 at the boundary

al

Now, from the difference equation for the Poisson’s equation, we have
ep

0 + 0 + 𝑓2 + 𝑓3 − 4𝑓1 = 12 × 2 × 12 × 22
itn

Source: www.csitnepal.com Page 15

cs
 Numerical Method (2067- second batch)

𝑜𝑟, 𝑓2 + 𝑓3 − 4𝑓1 = 8 … … … (𝑖)

0 + 0 + 𝑓1 + 𝑓4 − 4𝑓2 = 12 × 2 × 22 × 22
𝑜𝑟, 𝑓1 + 𝑓4 − 4𝑓2 = 32 … … … (𝑖𝑖)

0 + 0 + 𝑓1 + 𝑓4 − 4𝑓3 = 12 × 2 × 1 × 1
𝑜𝑟, 𝑓1 + 𝑓4 − 4𝑓3 = 2 … … … (𝑖𝑖𝑖)

0 + 0 + 𝑓2 + 𝑓3 − 4𝑓4 = 12 × 2 × 22 × 1
𝑜𝑟, 𝑓2 + 𝑓3 − 4𝑓4 = 8 … … … (𝑖𝑣)

Solving these equations, we get

11
𝑓1 = −
2
43
𝑓2 = −
4
13
𝑓3 = −
4
11
𝑓4 = −
2

al
ep
itn

Source: www.csitnepal.com Page 16

cs