6.

NUMERICAL INTEGRATION

n
We will use the methods called numerical quadrature which use a sum  a f (x )
i 0
i i to

approximate

b
a
f ( x) dx. (6.1)

The method that involves integrating the Lagrange interpolating polynomial can be used to

derive what are known as Newton-Cotes formulas for evaluating (6.1). For example, using

the first and second Lagrange interpolating polynomials with equally-spaced nodes gives the

Trapezoidal rule and Simpson’s rule, respectively. To derive these, we need the following

theorem:

Theorem 6.1.1 (Weighted Mean Value Theorem for Integrals)

If f  C[a, b], g is integrable on [ a, b] and g ( x) does not change sign on [ a, b] , then there

exists a number  , a    b, such that

b b
a
f ( x) g ( x) dx  f ( )  g ( x) dx.
a

Let x0  a, x1  b and h  b  a. Using the linear Lagrange interpolating polynomial

( x  x1 ) ( x  x0 ) f ( ( x))
f ( x)  f ( x0 )  f ( x1 )  ( x  x0 )( x  x1 )
( x0  x1 ) ( x1  x0 ) 2!

we get

b x1  ( x  x ) ( x  x0 )  x1  f ( ( x )) 
a
f ( x) dx   
x0 ( x  x )
 0 1
1
f ( x0 ) 
( x1  x0 )
f ( x1 )  dx   

x0  2!
( x  x0 )( x  x1 )  dx

x1
( x  x1 ) 2 ( x  x0 ) 2 1 x1
 f ( x0 )  f ( x1 )   f ( ( x ))( x  x0 )( x  x1 ) dx
2( x0  x1 ) 2( x1  x0 ) x
2 x0
0

h 1
 f ( x1 )  f ( x0 )  x0 f ( ( x))( x  x0 )( x  x1 ) dx.
x1

2 2

1
We can now apply Theorem 6.1.1, since ( x  x0 )( x  x1 ) does not change sign on [ x0 , x1 ] to get

x1 x1  x3 ( x  x ) x 2 x1

x0
f ( ( x))( x  x0 )( x  x1 ) dx  f ( ( x))  ( x  x0 )( x  x1 ) dx  f ( ( x ))  
x0  3
1

2
0
 x0 x1 x 

 x0 

 x 3 x x 2 x 3 x 2x 
 f ( ( x))   1  0 1  0  0 1 
 6 2 6 2 
   x1  x0 3 

 f ( ( x))  
 6 
3
h
 f ( ( x)).
6
Thus, the Trapezoidal rule is

h 1 h3 h h3
 1 0   1 0 
b
 a
f ( x) dx 
2
f ( x )  f ( x )  .
2 6
f ( ( x )) 
2
f ( x )  f ( x ) 
12
f ( ( x))

y y  f ( x)

a  x0 b  x1 x

ba
Letting x0  a, x1  a  h and x2  b with h  and using the second Lagrange
2
polynomial, we can derive the Simpson’s rule with an O (h 4 ) error term involving f (3) .

y y  f ( x)

y  P2 ( x)

a  x0 x1 b  x2 x

We can use the third Taylor polynomial about x1 to obtain a higher-order term involving

f (4) . Then, for each x in [ x0 , x2 ], there exists a number 1 ( x) in  x0 , x2  such that

2
 f ( x1 ) f ( x1 ) f (4) (1 ( x))
f ( x)  f ( x1 )  f ( x1 )( x  x1 )  ( x  x1 ) 
2
( x  x1 ) 
3
( x  x1 ) 4
2! 3! 4!
for some number 1 ( x) in  x0 , x2  , so that
b x2  f ( x1 ) f ( x1 ) 
a
f ( x) dx    f ( x1 )  f ( x1 )( x  x1 ) 
x0
 2!
( x  x1 ) 2 
3!
( x  x1 )3  dx

x2  f
(4)
(1 ( x)) 
  ( x  x1 ) 4  dx
x0
 4! 
x
 f ( x1 ) f ( x1 ) f ( x1 ) 
2

  f ( x1 )( x  x1 )  ( x  x1 ) 2  ( x  x1 )3  ( x  x1 ) 4 
 2 6 24  x0
x2  f
(4)
(1 ( x)) 
  ( x  x1 ) 4  dx
x0
 4! 

 f ( x1 ) f ( x1 ) f ( x1 )
  f ( x1 )( x2  x1 )  ( x2  x1 ) 2  ( x2  x1 )3  ( x2  x1 ) 4
 2 6 24
f ( x1 ) f ( x1 ) f ( x1 ) 
 f ( x1 )( x1  x0 )  ( x1  x0 ) 2  ( x1  x0 )3  ( x1  x0 ) 4 
2 6 24 
x2  f
(4)
(1 ( x)) 
  ( x  x1 ) 4  dx
x0
 4! 
f ( x1 ) 3 x2  f
(4)
(1 ( x)) 
 2hf ( x1 )  h   ( x  x1 ) 4  dx, since h  x2  x1  x1  x0 .
3 x0
 4! 
Thus, by Theorem 6.1.1, we have that
x2
x2  f (4) (1 ( x)) 4 f (4) (1 ( x)) x2 f (4) (1 ( x))
      ( x  x1 )5
4
 ( x x1 )  dx ( x x1 ) dx
x0
 4!  24 x0 120 x0

f (4) (1 ( x)) f (4) (1 ( x))

 ( x2  x1 )5  ( x0  x1 )5   ( x2  x1 )5  ( x1  x0 )5 
120 120
f (1 ( x)) 5
(4)
 h
60
b h3 h5 (4)
  a
f ( x) dx  2hf ( x1 ) 
3

f ( x1 ) 
60
f (1 ( x)).

Using the second derivative midpoint formula, we have that

1 h 2 (4) 1 h 2 (4)
f ( x1 )  [ f ( x1  h )  2 f ( x1 )  f ( x1  h )]  f (1 ( x ))  [ f ( x0 )  2 f ( x1 )  f ( x2 )]  f (1 ( x))
h2 12 h2 12
so that
b h3  1 h 2 (4)  h5 (4)
a
f ( x) dx  2hf ( x1 ) 
3  h2
[ f ( x0 )  2 f ( x1 )  f ( x2 )] 
12
f (1 ( x ))   60 f (1 ( x)).


3
Therefore, the Simpson’s rule is

h h5
 f ( x0 )  4 f ( x1 )  f ( x2 )  f (4) (1 ( x)).
b
 a
f ( x) dx 
3 90

The error term in Simpson’s rule involves the fourth derivative of f , so it gives exact

results when applied to any polynomial of degree three or less.

Example 6.1.1

2
Use the Trapezoidal rule and the Simpson’s rule to evaluate  0
1  x2 .

Solutions:

Using Trapezoidal rule, we have that x0  0, x1  2 and h  2 so that

2
 f ( x0 )  f ( x2 )  f (0)  f (2)  1  5  3.236067977
2
0
1  x2 
2

20
Using Simpson’s rule, we have that x0  0, x1  1, x2  2 and h   1 so that
2

h 1
 f ( x0 )  4 f ( x1 )  f ( x2 )   f (0)  4 f (1)  f (2)
2
 0
1  x2 
3 3

1
1  4 2  5   2.964307409.
3

The Newton-Cotes formulas, however, are generally unsuitable for use over large integration
intervals because of the oscillatory nature of high-degree polynomials.

Theorem 6.1.2

ba
Let f  C 2 [a, b], h  and x j  a  jh for each j  0,1, 2,..., n. There exists a   (a, b)
n
for which the Composite Trapezoidal rule for n subintervals can be written with its error
term as

h n 1  ba 2

b
a
f ( x) dx  
2
f ( a )  2
j 1
f ( x j )  f (b)  
 12
h f (  ).

4
Example 6.1.2

(a) Find an approximation to the area of the region bounded by the normal curve

2
x
 
 
1
f ( x) 
2
e
 2

and the x  axis on the interval [ ,  ] using the Composite Trapezoidal rule with n  8.

(b) Using the Composite Trapezoidal rule, determine the value of n and h required to
2
approximate 
0
e 2 x sin 3 x dx to within 104.

Solutions:

(a) Clearly, f  C 2   ,  .

  ( )   
With a   , b   and n  8, we have that h   and x j    j  
8 4 4
 3   3 
 x1     , x2      , x3      , x4      0,
4 4 2 2 4 4
5  3  7 3
x5     , x6     and x7    
4 4 2 2 4 4
2
  
 
   
1
1 1
 f (a )  f ( )  
2 2
e e
 2  2
2
 3 
 
 4 
  
 
 3 
  
9
1 1
f ( x1 )  f    
2 32
e e
 4   2  2
2
  
 
 2 
  
 
 
  
1
1 1
f ( x2 )  f     
2 8
e e
 2   2  2
2
  
 
 4 
  
 
 
  
1
1 1
f ( x3 )  f     
2 32
e e
 4   2  2
1
f ( x4 )  f (0) 
 2

5
 2
 
 
 4 
 
 
 
  
1
1 1
f ( x5 )  f    
2 32
e e
 4   2  2
2
 
 
 2 
 
 
 
  
1
1 1
f ( x6 )  f    
2 8
e e
2  2  2
2
 3 
 
 4  2
    
   

 3
 

  
9

1
1 1 1 1
f ( x7 )  f    and f (b)  f ( )  
2 32 2 2
e e e e .
 4   2  2  2  2

 h 81 


f ( x) dx   f ( )  2 f ( x j )  f ( ) 
2 j 1 

  f ( )  2 f ( x1 )  2 f ( x2 )  2 f ( x3 )  2 f ( x4 )  2 f ( x5 )  2 f ( x6 )  2 f ( x7 )  f ( ) 
8

  1   1 8   1   1 
1 9 1 1
   
1
 e  2   2 e   2   2 
2 32 32
 e e
8   2   2    2    2    2 
 1 32   1 8   1 32  2 
1 1 9 1
   
1
2 e   2 e   2 e  e 
  2    2    2   2 

  2 
1 9 1 1
  
2
 .     1
32 8 32

 e 2e 2e 2e
8  2  
 0.680163689
(b) Here, a  0, b  2 and f ( x)  e2 x sin 3x

 f ( x)  2e2 x sin 3x  3e2 x cos 3x

f ( x)  2 2e2 x sin 3x  3e2 x cos3x   3 2e2 x cos3x  3e2 x sin 3x   12e2 x cos3x  5e2 x sin 3x.

The error function for the Composite Trapezoidal rule is

ba 2
E( f )   h f (  )
12

ba 2  2 
  h f (  )  104   h 2  Max 12e 2 x cos 3 x  5e 2 x sin 3 x   10 4.
12  12  0,2

Since Max 12e2 x cos 3 x  5e2 x sin 3 x   705.3601029, we have that

0,2

6 104
h2 
705.3601029
6
  h  8.506293417 107  0.0009222956911.
ba
Since h  , we have that
n
2 2
 0.0009222956911 n  2168.501945
n 0.0009222956911
 n  2168.

Theorem 6.1.3

ba
Let f  C 4 [a, b], n be even, h  and x j  a  jh for each j  0,1, 2,..., n. There exists a
n
  (a, b) for which the Composite Simpson’s rule for n subintervals can be written with its
error term as

h  b  a 4 (4)
n1 n

f ( x) dx   f (a)  2 f ( x2 j )  4 f ( x2 j 1 )  f (b)  
b 2 2


a 3  j 1 j 1  180
h f (  ).

Example 6.1.3

(a) A car laps a race track in 84 seconds. The speed of the car at each 6  second interval

is determined by using a radar gun and is given from the beginning of the lap, in
feet/second, by the entries in the following table:

Time 0 6 12 18 24 30 36 42 48 54 60 66 72 78 84
Speed (ft/s) 124 134 148 156 147 133 121 109 99 85 78 89 104 116 123

How long is the track?

(b) Determine the value of n and h required to approximate

2  1 

0   dx
 x4

to within 10 5 and compute the approximation using Composite Simpson’s rule.

7
Solutions:

(a) Here h  6, x0  0, x1  6, x2  12,..., x13  78, x14  84.

Since n  14, we can use Composite Simpson’s rule:

h 
1
14 14

f ( x) dx   f ( x0 )  2 f ( x2 j )  4 f ( x2 j 1 )  f ( x14 ) 
84 2 2

 Length of the track, l  

0 3  j 1 j 1 

h
  f ( x0 )  2 f ( x2 )  2 f ( x4 )  2 f ( x6 )  2 f ( x8 )  2 f ( x10 )  2 f ( x12 )  4 f ( x1 )  4 f ( x3 )  4 f ( x5 ) 
3
4 f ( x7 )  4 f ( x9 )  4 f ( x11 )  4 f ( x13 )  f ( x14 ) 
h
 124  2(148)  2(147)  2(121)  2(99)  2(78)  2(104)  4(134)  4(156)  4(133) 
3
4(109)  4(85)  4(89)  4(116)  (123)
 4232 ft.

b  a 4 (4) 1 1 2
(b) E( f )   h f (  ), where f ( x)   f ( x)   , f ( x)  ,
180 x4 ( x  4) 2
( x  4)3

6 24
f ( x)   and f (4) ( x)  .
( x  4) 4
( x  4)5

24
 f (4) (  ) 
(   4)5

 2 4  (4)
 E( f )   h  f (  )  105
 180 
h4  24  5
 Max    10
90 0,2  (   4)5 
h4  4  5
  5   10
15  4 
 h 4  105 (3840)  0.0384
 h  0.442672767.
ba
h  0.442672767
n
2
n  4.518
0.442672767
 n  4.
Using Composite Simpson’s rule, we take n  6 so that

8
  1  h
 dx   f ( x0 )  2 f ( x2 )  2 f ( x4 )  4 f ( x1 )  4 f ( x3 )  4 f ( x5 )  f ( x6 )]
2

0 
 x4 3
2 
1  1   1   1   1   1  1 
 6   2 2   2 4   4 1   4   4 5  
3  4  3  4   3  4   3  4   1  4   3  4  2  4 
 0.405466374.
2  1 

2
The exact value is   dx  ln x  4 0  ln 6  ln 4 0.405465108
0
 x4

giving the error | ln 32  0.405466374 | 0.00000126589184  105.

THE END!