1

CHAPTER 6
Numerical Differentiation and Integration

By: Habtamu Garoma 9/8/21

 Numerical Differentiation
2

Suppose ( xi , yi ) ; i  0,1,2,..., n are given. We want to find the derivative of

y  f (x) passing through the (n+1) points, at a point x .

If x is near to the starting value x0 , then we use Newton’s forward

difference interpolation formula.

u  u  1 2 u  u  1 u  2 3
y  x   y  x0  uh   y0  uy0   y0   y0  ... 1
2 3!

x  x0
where y (x) is a polynomial of degree n in x and u 
h

By: Habtamu Garoma 9/8/21

 Cont’d
3
Now differentiate y (x) with respect to x

dy dy du 1 dy du 1
   since 
dx du dx h du dx h

1 2u  1 2 3u 2  6u  2 3 4u 3  18u 2  22u  6 4 
= y0   y0   y0   y0  ... ... 2
h 2 6 24 

dy
This gives the value of at any x which may be any where in the
dx

interval.

If x  x0  u  0, then (2) reduces to

dy dy 1 1 2 1 3 1 4 
  y   y   y   y  ...
h  
0 0 0 0
dx x  x0 dx u 0 2 3 4 

By: Habtamu Garoma 9/8/21

 Cont’d
4
Differentiating (2) again with respect to x , and we get

d2y d  dy  du d  dy  1
      
dx 2 du  dx  dx du  dx  h

1  2 6u 2  18u  11 4 
   y 0   u  1 3
y 0   y0  ...
h2  12 

d3y 1  3 12u  18 4 
  y   y  ...
h3  
0 0
dx 3 12 

Setting x  x 0 , i.e. u  0

d2y 1  2 11 4 
 2
 2 
 y 0  3
y 0   y0  ... and
dx h  12 

d3y 1  3 3 4 
3
 3   y 0   y 0  ...
dx h  2 

By: Habtamu Garoma 9/8/21

 Cont’d
5

If x is near to the end value x n , then we use Newton’s backward

difference formula to compute the derivative.

v v  1 2 v v  1 v  2 3
y  x   y  x n  vh   y n  vyn   yn   y n  ... ... 
2 3!

v  xn
where v 
h

Differentiating (*) with respect to x

dy dy dv 1 dy
  
dx dv dx h dv
By: Habtamu Garoma 9/8/21
 Cont’d
6
dy 1  2v  1 2 3v 2  6v  2 3 4v 3  18v 2  22v  6 4 
 y n   yn   yn   y n  ...
dx h  2 6 24 

d2y 1  2 6v 2  18v  11 4 
  y n   v  1  3
y n   y n  ...
dx 2 h 2  24 

d3y 1  3 12v  18 4 
  y n  12  y n  ...
dx 3 h 3  

Setting x  x n

dy 1 1 1 1 
 yn  2 yn  3 yn   4 yn  ...
dx x  xn h  2 3 4 

d2y 1 2 11 
 2 
 yn  3 yn   4 yn  ...
dx 2 x  xn
h  12 

d3y 1 3 3 
 3 
 yn   4 yn  ...
dx 3 x  xn
h  2 

By: Habtamu Garoma 9/8/21

 Cont’d
7

If higher differences in the equation are very small, we can

take only the first few terms in the interpolating polynomial

and solve it for . The interpolating polynomial we find is an
approximate function to the original function i.e. is small.
But may not be small even if is small. This means the error
involved in obtaining through may be great. So, it can be
used with confidence only if the differences of some order
are constants.
By: Habtamu Garoma 9/8/21
 Maxima and Minima
8
We want to find the value of x at which the curve is maximum or

minimum given the tabular values. Using any one of the methods

developed so far, we can get the interpolating polynomial of degree n, and

then using the concepts how to find maximum and minimum in calculus

find the derivative of this polynomial and equate it to zero.

dy 1  2u  1 2 3u 2  6u  2 3 
 y0   y0   y0  ...
dx h  2 6 

dy 2u  1 2 3u 2  6u  2 3
0  y0   y0   y0  ...  0
dx 2 6
By: Habtamu Garoma 9/8/21
 Example 1
9

. The population of a certain town is given below. Find the

rate of growth of the population in 1931, 1941, 1961

and 1971.

Year x : 1931 1941 1951 1961 1971

Population in: 40.62 60.80 79.95 103.56 132.65

Thousands

By: Habtamu Garoma 9/8/21

 Solution
10

x p p 2 p 3 p 4 p

1931 40.62
20.18
-1.03
1941 60.80 5.49
19.15
4.46 -4.47
1951 79.95 1.02
23.61
5.48
1961 103.56
29.09
1971 132.65

To get the rate of growth of the population we find p ( x ) when

x  1931

1931  1931
x 0  1931  u  0
10

By: Habtamu Garoma 9/8/21

 Cont’d
11

dp dp 1 1 1 1 
( )   p0  2 p0  3 p0  4 p0  ... 
dx x  1931 dx u  0 h  2 3 4 

1  1 1 1 
 20.18    1. 03   5. 49     4. 47 
10  2 3 4 


 2.36425

When x  1941

1941  1931
u 1
10
By: Habtamu Garoma 9/8/21
 Cont’d
12

Putting u = 1, in

dp 1  2u  1 2 3u 2  6u  2 3 4u 3  18u 2  22u  6 4 
 p0   p0   p0   p0  ... and
dx h  2 6 24 

dp 1 1 1 1 
We get,   20.18    1.03   5.49    4.47  
dx u  1 10  2 6 12 

= 1.83775

When x  1961 , we use the backward interpolating polynomial

1961  1971
So v   1
10
By: Habtamu Garoma 9/8/21
 Cont’d
13

dp 1 2v  1 2 3v 2  6v  2 3 

  p n   pn   p n  ... 
dx v  1 h 2 6  v  1

1 

1 1
 29.09   5.48  1.02  
1
  4.47   = 2.65525
10  2 6 12 

1971  1971
When x  1971 so v  0
10

 dp  1 1 1 1 
    pn   2 pn   3 pn   4 pn  ... 
 dx u  0 2  2 3 4 

1  1 1 1 
  29.09   5.48  1.02     4.47   = 3.10525
10  2 3 4 

By: Habtamu Garoma 9/8/21

 Example 2
14

The table below gives the results of an observation;

 is the observed temperature in degrees centigrade of

a vessel of cooling water and t is the time in minutes

from the beginning of the observation.

t: 1 3 5 7 9

: 85.3 74.5 67.0 60.5 54.3

Find the approximate rate of cooling at t = 3 and t = 3.5.

By: Habtamu Garoma 9/8/21

 Solution
15

x   2p 3p 4p

1 85.3
10.8
3 74.4 3.3
-7.5 -2.3
5 67.0 1.0 1.6
-6.5 -0.7
7 60.5 0.3
-6.2
9 54.3

d
Since represents the rate of cooling, we have
dt

t  t0 t 1
u 
2 2

By: Habtamu Garoma 9/8/21

 Cont’d
16
At t  3, u  1 and at t  3.5, u  1.25

Using u  1 in (2), we get

 d  1  2u  1 2 3u 2  6u  2 3 4u 3  18u 2  22u  6 4 
    0   0   0    0  ...
 dt  2  2 6 24 

 d  1 1  1 1 
    10.8  (3.3)     2.3  (1.6)
 dt u 1 2  2  6  12 

1
   10.8  1.65  0.383333333  0.1333333333 = - 4.316666667
2

 d  1
ii)  
 (1.25)  1 
  10.8   2 (3.3) 
(0.8125
  2.3  1.1875 (1.6)
 dt u 1.25 2   2  6 24 

= - 3.9671875
By: Habtamu Garoma 9/8/21
 Example 3
17

Given the following data, find y (6) and the maximum

value of y.

x: 0 2 3 4 7 9

y: 4 26 58 112 466 922

By: Habtamu Garoma 9/8/21

 Solution
18
Since the x ’s are not equally spaced, we will use Newton’s divided

difference formula. Now we form the difference table

x y f  x  2 f  x  3 f  x  4 f  x 

0 4
11
2 26 7
1
32
3 58 11 0
54 1
4 112 16 0
118 1
7 466
22
228
9 922
By: Habtamu Garoma 9/8/21
 Cont’d
19
By Newton’s divided difference formula

y  f  x   f  x0    x  x0  f  x0 , x1    x  x0  x  x1  f  x0 , x1 , x2   ...

= 4  ( x  0)(11)  ( x  0)( x  2)(7)  ( x  0)( x  2)( x  3)(1)

= x  2 x  3x  4
3 2

 y ' ( x)  3x 2  4 x  3

y '  6   3 6   4 6   3  135
2

To find the maximum value of y , consider y ( x)  0 .

 3x 2  4 x  3  0

There is no value of x that satisfies this equation since the roots are

imaginary. Thus, there is no an extremum value in the range.

By: Habtamu Garoma 9/8/21
 Numerical Integration
20

Consider  f  x dx .
b

This integral represents the area between y  f (x) , the x –axis and the

lines x  a and x  b . This integration is possible as far as f (x) is explicitly

given and the function is integrable.

Now suppose set of (n+ 1) paired values are given. First as we did in

the case of numerical differentiation, we find f (x) by an interpolating

 p  x  dx
b
polynomial Pn (x) and obtain n which can approximate the value for
a

 p  x  dx .
b
n
a
By: Habtamu Garoma 9/8/21
 Cont’d
21
A general quadrature formula for equidistant spacing

(Newton-Cote’s formula)

For equally spaced intervals, we have Newton’s forward difference

formula as

u  u  1 2 u  u  1 u  2 3
y  x   y0  uy0   y0   y0  ...1
2! 3!

x  x0
where u  , h is the interval of differencing.
h

By: Habtamu Garoma 9/8/21

 Cont’d
22
Now, instead of f (x) , we will replace it by this interpolating polynomial

y (x) of Newton.

x  x0 x  x0
Since xn  x0  nh and u  un
h h
x 0  nh
f  x dx   f  x dx
xn
Thus x0 x0

x 0  nh
 Pn  x dx where Pn (x) is interpolating polynomial of degree n.
x0

n u  u  1 2 u  u  1 u  2  3 
   y0  uy0   y0   y0  ...hdu
0
 2! 3! 

 n u  u  1 2 u 3  3u 2  2u 3 
 h   y0  uy0   y0   y0  ...du
0
 2! 3! 
By: Habtamu Garoma 9/8/21
 Cont’d
23
n
 u u 3 2
u 4

   u 3
 u 2

u2 3 2 4
 h uy0  y0   y0 
2
 y0  ...
3

 2 2! 3! 
  0

 n2 1  n3 n 2  2 1  n4 3 2  3
 h ny0  y0     y0    n  n  y0  ...(2)
 2 2 3 2  6 4 

This equation is called Newton-cote’s quadrature formula and is a

general quadrature formula.

By: Habtamu Garoma 9/8/21

 Cont’d
24

 By taking different values for n we get a number of

special formulas.
Trapezoidal Rule

Put n = 1, in the quadrature formula

x 0  nh x0  h  1 
 f  x dx   f  x dx  h 1. y0  y0 
x0 x0
 2 

Since other differences do not exist if n = 1.

 1 
 h y0   y1  y0  
 2 

h
  y0  y1 
2
By: Habtamu Garoma 9/8/21
 Cont’d
25
x 0  nh
f  x  dx   f  x dx
xn
x0 x0

x 0  nh x0  2h x 0  nh
 f  x dx   f  x dx  ...   f  x dx
x0 x0  h x 0   n 1 h

h
  y0  y1   h  y1  y2   ...  h  yn 1  yn 
2 2 2

h
   y0  yn   2 y1  y2  ...  yn 1  
2

This is known as a trapezoidal rule.

Even though this method is very simple for calculation, the error in
this case is significant
By: Habtamu Garoma 9/8/21
 Truncation error in Trapezoidal Rule
26
In the neighborhood of x  x0 , we can expand y  f (x ) by Taylor series in

powers of x  x0 . That is,

y  x   y0   x  x0  y '

 x  x0 
2
y0''  ... 1
0
2!

x1 
ydx    y0   x  x0  y0 
x1  x  x0  ''
2

 y0  ...dx
'
x0 x0
 2! 

 x  x0  '  x  x0  ''
x1
 2 3

  y0 x  y0  y0  ...
 2! 3!  x0

 y0  x1  x0   x1  x0 
2
'
y 
 x1  x0 
3
y0''  ...
0
2! 3!

h 2 ' h3 ''
 hy0  y0  y0  ... (2)
2! 3!
By: Habtamu Garoma 9/8/21
 Cont’d
27
If h is the equidistant length, then also

h
 y0  y1   Area of the first trapezium = A0 say
x1
x0
ydx 
2

Putting x  x1 in (1), we get

y  x1   y1  y0   x1  x0  y 
'  x1  x0 
2
y0''  ...
0
2!

h 2 ''
 y0  hy  y0  ...
'
0
2!

h h h 2 '' 
A0   y0  y1    y0  y0  hy0 
'
y0  ... 
2 2 2! 

h2 ' h3 ''
 hy0  y0  y0  ...
2 2.2!
By: Habtamu Garoma 9/8/21
 Cont’d
28

Subtracting A0 from (2), we obtain

x1 1 1 
x0 ydx  A0  h y  3!  2.2!   ...
3 ''
0

1
  h3 y0''  ...
2

1 3 ''
 The error made in the first interval ( x0 , x1 ) is  h y0  ...
2

1 3 ''
Similarly the error in the ith interval   h yi 1
2

By: Habtamu Garoma 9/8/21

 Cont’d
29

Hence, the total cumulative error E is

E   h3  y0''  y1''  y2''  ...  yn'' 1 

1
2

 E 
nh 3
12
M 
where M  max y 0'' , y1'' , y 2'' ,... 

 b  a  h2
M if the interval is (a, b) and h
ba
12 n

 The error in the trapezoidal rule is of order h 2 .

The accuracy of the result can be improved by increasing the number

of intervals and decreasing the value of h .

By: Habtamu Garoma 9/8/21
 Simpson’s one-third Rule
30
Setting n = 2 in Newton-cotes quadrature formula, we have

 18 4 
f  x  dx  h 2 y0  4y0    2 y0  since other terms vanish
x2
x
 23 2 

 1 
 h 2 y0  2 y1  y0    E  1 y0 
2

 3 

 1 
 h 2 y0  2 y1  2 y0   y2  2 y1  y0  
 3 

1 4 1 
 h  y2  y1  y0 
3 3 3 

h
  y2  4 y1  y0 
3
By: Habtamu Garoma 9/8/21
 Cont’d
31
h h
f  x dx   y2  4 y3  y4  and f  x dx   yi  4 yi 1  yi  2 
x4 xi  2
Similarly, x2 3 xi 3

If n is an even integer, then the last integral will be

h
f  x dx   y n2  4 y n1  y n 
xn
xn  2 3

Adding all these integrals, if n is even positive integer, then

y 0 , y1 , y 2 , y 3 ,..., y n are odd in number; we have

xn x2 x
f ( x) dx   4 f  x dx  ...   f  x  dx
xn
x0
f ( x) dx  
x0 x2 xn  2

h
   y0  4 y1  y2    y2  4 y3  y4   ...   yn  2  4 yn 1  yn  
3

h
   y0  yn   2 y2  y4  ...  yn  2   4 y1  y3  ...  yn 1  
3
By: Habtamu Garoma 9/8/21
 Simpson’s three-eighths Rule
32

Putting n = 3 in Newton-cotes formula, we get

x3  9 19 1  81  

x0
f ( x)dx  h 3 y0  y0   2 y0    27  9 3 y0 
 2 22 6 4  

 9 9 3 
 h 3 y0   y1  y0    E  1 y0   E  1 y0 
2 3

 2 4 8 

 9 9 9 3 
 h 3 y0  y1  y0   y2  2 y1  y0    y3  3 y2  3 y1  y0  
 2 2 4 8 

3h
  y3  3 y2  3 y1  y0 
8

By: Habtamu Garoma 9/8/21

 Cont’d
33
If n is a multiple of 3,

x0  nh x  3h x0 6 h x  nh
 f ( x)dx   0 f ( x)dx   f  x dx  ...  0
 f  x  dx
x0 x0 x  3h x0 ( n3) h
0

3h
   y0  3 y1  3 y2  y3    y3  3 y4  3 y5  y6   ...   yn 3  3 yn  2  3 yn 1  yn  
8

3h
   y0  yn   3 y1  y2  y4  y5  ...  yn  2  yn 1   2 y3  y6  yn  
8

This is Simpson’s three-eighths rule and is applicable only when n is a

multiple of 3.

By: Habtamu Garoma 9/8/21

 Truncation error in Simpson’s
formula
34

By Taylor expansion of y  f (x) in the neighborhood of x  x0 , we

obtain

y  y0   x  x0  y ' x  x0 
2
y ''

 x  x0  '''
3
y0  ... ...1
0 0
2! 3!

x  x  x  x   x  x  2
 x  x  3

x0 x0 
2 ydx  2 y0     dx
0 ' 0 ' ' 0 '''
y 0 y 0 y 0 ... 
 1! 2! 3! 

 x  x0  '  x  x0  ''  x  x0  ''' 

x2
 2 3 4

  y0 x  y0  y0  y0  ..
 21! 3! 4!  x0

 y0  x2  x0   x2  x0 
2
y '

 x2  x0  ''  x2  x0 
3
y 
4
y0'''  ...
0 0
2! 3! 4!
By: Habtamu Garoma 9/8/21
 Cont’d
35

4h 2 ' 8h3 '' 16 2 '''

 2hy0  y0  y0  h y0  ...
2! 3! 4!

4 3 ' ' 2 h 4 ' ' ' 4h 5  4 

 2hy0  2h y  h y0 
2 '
0 y0  y0  ...
3 3 15

h
 y0  4 y1  y 2  , by Simpson’s rule
x2
Now let A1  area  x ydx 
0 3

Putting x  x1 in (1), to get

y1  y0   x1  x0  y0' 
 x1  x2  y ''  ...
0
2!

h 2 '' h3 '' '

 y0  hy  '
0 y0  y0  ...
2! 3!

By: Habtamu Garoma 9/8/21

 Cont’d
36
Putting x  x 2 in (1), we have

4h 2 '' 8h3 '''

y2  y0  2hy '
0 y0  y0  ...
2! 3!

h  h 2 ' ' h3 '''  

 A1   y0  4 y0  hy0 
'
y0  y0  ...   y0  2hy0'  2h 2 y0''  ... 
3  2! 3!  

4 2 5
 2hy0  2h 2 y0'  h3 y0''  h 4 y0'''  h5 y0 4   ...
3 3 18

x2 4 5
  ydx  A1    h5 y0 4   ...
x0
 15 18 

 h5  4 
 y0  ... Omitting the remaining terms involving h 6 and
90

higher powers of h .
By: Habtamu Garoma 9/8/21
 Cont’d
37
 h5  4 
This means that the error made in  x0 , x2  is  y0  ...
90

 h5  4 
Similarly, the error made in  x2 , x4  is y2 and so on.
90

Hence the total error E is  x0 , xn  is  

9

h5  4 

y0  y2 4   ...

nh5
E  M where M is the maximum value of y0 4  , y2 4   ..., y2 4n 2
90

Since  x2 n , y2 n  is the last paired value because we require odd number

of ordinates to apply Simpson’s one-third rule.

If the interval is (a, b) , then b  a  h(2n) , using this

E 
 b  a  h4 M
180

Hence, the error in Simpson’s one-third rule is of the order h 4 .

By: Habtamu Garoma 9/8/21

 Example 1
38

7
Evaluate 
1
x 2 dx by using

a) Trapezoidal rule

b) Simpson’s rule and verify your results by actual integration

By: Habtamu Garoma 9/8/21

 Solution
39

Here y ( x)  x 2 and the interval length is 7-1 = 6 so we divide this

6
interval into 6 equal parts with h  1
6

x: 1 2 3 4 5 6 7

y: 1 4 9 16 25 36 49

a) By Trapezoidal rule

h
  y1  y7   2 y2  y3  y4  y5  y6  
7

1
x 2 dx 
2

1
 1  49  2 4  9  16  25  36   115
2
By: Habtamu Garoma 9/8/21
 Cont’d
40
a) By Simpson’s one-third rule

1
  y1  y7   2 y3  y5 )  4( y2  y4  y6  
7

1
x 2 dx 
3

1
  1  49  2 9  25  4 4  16  36 
3

= 114

Since n = 6, we can also use Simpson’s three-eighth rule.

3
  y1  y7   3 y2  y3  y5  y6   2 y4 
7
So  1
x 2 dx 
8

3
  1  49  3 4  9  25  36  216 
8

= 114
7
7 x3 1 1 342
 x dx   73    114
2
1 3 1 3 3 3
By: Habtamu Garoma 9/8/21
 Example 2
41

dx
h  0.2 , and hence
1
2. Evaluate 0 1  x 2 , using Trapezoidal rule with
obtain an approximate value of  .

By: Habtamu Garoma 9/8/21

 Solution
42
1
Let y x  
1  x2

x: 0 0.2 0.4 0.6 0.8 1

y: 1 0.961538461 0.862068965 0.735294117 0.609756097 0.5

dx h
0 1  x 2 2   y0  y5   2 y1  y2  y3  y4  
1


0. 2
  1  0.5  2 0.961538461  0.862068965  0.735294117  0.6097566097  
2

= 0.783731528

By: Habtamu Garoma 9/8/21

 Cont’d
43

But by actual integration

1 dx 
0 1  x 2
1 1
 tan x 0

4


  0.783731528
4

   3.134926112

To compare this approximated value of  with its actual value using

calculator 3.141592654 , the error is 0.00666654159 which is 0.6% .

By: Habtamu Garoma 9/8/21

 Example 3
44

3. From the table below, find the area bounded by the curve and

the x -axis from x  7.47 to x  7.53

x: 7.47 7.48 7.49 7.50 7.51 7.52 7.53

y: 1.93 1.95 1.98 2.01 2.03 2.06 2.08

By: Habtamu Garoma 9/8/21

 Solution
45

i) By Trapezoidal rule

0.01
f  x dx   1.93  2.08  21.95  1.98  2.01  2.03  2.06 
7.53

7.47 2

= 0.12035

ii) By Simpson’s one-third rule

0.01
f  x dx   1.93  2.08  21.98  2.03  4  41.95  2.01  2.06 
7.53

7.47 3

 0.120366667

iii) By Simpson’s three-eighths rule

3(0.01)
f  x dx   1.93  2.08  31.95  1.98  2.03  2.06  2 2.01 
7.53
 7.47 8

 0.1203375
By: Habtamu Garoma 9/8/21
 Example 4
46

1
4. Evaluate 
0
e x dx by Simpson’s one-third rule correct to five decimal

places.

By: Habtamu Garoma 9/8/21

 Solution
47
The interval b  a  1

Since error E 
 b  a  h 4 M , where M = Max (e x ) in the range
180

1 4
 h e
180

Now we require E  106

h 4e
  10  6
180

1
 180  10 6  4
 h     0.090207886  0.1
 e 

Hence we take h  0.1

By: Habtamu Garoma 9/8/21
 Cont’d
48

1
  e x dx 
0
0.1
3

1  e  2 e0.2  e0.4  e0.6  e0.8   4 e0.1  e0.3  e0.5  e0.7  e0.9  
 1.718282782

By the actual integration,

1 1
0
e x dx  e x  e  1  1.718281828
0

1
Correct to five decimal places 0
e x dx  1.71828 which is the same as the

exact value.

By: Habtamu Garoma 9/8/21

 Example 5
49

4. The table below gives the velocity v of a moving particle at time t

second. Find the distance (S) covered by the particle in 12

seconds and also the acceleration at t  2 seconds.

t: 0 2 4 6 8 10 12

v: 4 6 16 34 60 94 136

By: Habtamu Garoma 9/8/21

 Solution
50

ds dv
We know that v  and a 
dt dt

To get S

2
  4  136  216  60  4 6  34  94 
12
S  0
vdt 
3

= 552 meters

 dv 
To find a, a    first form the difference table
 dt  t  2

By: Habtamu Garoma 9/8/21

 Cont’d
51

t v v 2v 3v

0 4
2
8
2 6 0
10
8
4 16
18 0
8
6 34
26 0
8 60 8
34
0
10 94
42 8
12 136

 dv  1 1 1 
   v0  2v0  3v0 
 dt t  2 h 2 3 
By: Habtamu Garoma 9/8/21
 Cont’d
52

 dv  1 1 2 1 3 
   v0   v0   v0  taking v0 = 6
 dt t  2 h  2 3 

1 1 
 10   8   3m / sec 2
2 2 

By: Habtamu Garoma 9/8/21

 53

THE END

By: Habtamu Garoma 9/8/21