Lecture -6

Gaussian quadrature formula:

This formula is applicable only when the function f  x  is explicitly known in terms of its

argument and the limits of integration are -1 to +1. According to this formula the integral is
expressed as a weighted sum of n- terms in the form
1 n

 f ( x) dx  i 1 wi f ( xi ) ...(c)
1

where , ,…, are unknown coefficients known as weights and −1 < < <⋯<
< 1 are also unknown values of x called weighting factors. This is known as n-terms formula
in which there are 2n unknowns. These 2n unknowns will be evaluated by requiring that
“formula (C) be true for all polynomials of degree (2n-1) { i.e. one less to the number of
unknowns} or less. Then, it must certainly be true for simplest cases i.e. for
, , ,……. ”. The formulae based upon this principle are known as Gaussian
quadrature formulae. The satisfaction of (C) for ( ) = , ,……. will lead to a set of
2n linear equations in ′s. Their solution will give the values of ′s and ′s. Substituting back
these values in (C), one gets n-terms Gaussian quadrature formula.

Example: Derivation of 2-terms formula:

For 2-terms formula (C) becomes

∫ ( ) = ( )+ ( ) ...(D)
Since (D) has four unknowns, as per principle of the method, it must be true for all polynomials
of degree 3 or less. Then, it must certainly be true for the simplest cases ( ) = , , ,
( ) = x0 ; ∫ x0 ≡2= .1 + .1 ...(1)

( ) = x1 ; ∫ x1 =0= + ...(2)

( )= ; ∫ ≡ = + ...(3)

( )= ; ∫ =0= + ...(4)
From equations (2) and (4) i.e. (2)  − (4), we get ( − ) = 0  either
 =0 = 0 or =  .
The only valid solution is =− , other reduces our formula (D) into a single term or invalid.

We then find = = 1 and − = = .

√

Thus, two-terms formula becomes

1
∫ ( ) = (− ) f( ).
√ 3
Similarly, the three-terms formula is given by

∫ ( ) =  x1  +  x2  +  x3 
Where = = , = , − x1  x3  3 / 5, =0

and thus ∫ ( ) = − + (0) + ( )}

In fact, the values of x in two-terms or three-terms formula are the zeros of Legendre
polynomials ( ), ( ), respectively.
Similarly, the higher terms formula can be obtained.
If the limits of integration are a to b instead of -1 to +1, then the liner transformation
( ) ( )
= =( )

1
 b  a  t  b  a  
will change the limits as follows: ∫ ( ) = ( ) 
1
f
 2
 dt

EXAMPLES
Ex.1: Calculate the value of the integral
. 1 3
∫ log e x by (a) Trapezoidal rule (b) Simpson’s rd rule (c) Simpson’s th rule
3 8
taking h = 0.2 retaining the numerical values to 7 decimals. Compare the results with exact
values.
Solution:
∶ 4.0 4.2 4.4 4.6 4.8 5.0 5.2

log e x : 1.3862944 1.4350845 1.4816045 1.5260563 1.5686159 1.6094379 1.6486586

(a) By Trapezoidal rule, we have

 .
∫ . log e x = [ . + 2{ . + . + . + . + . }+ . ]

=1.8276551
1
(b) By Simpson’s rd rule, we have
3
.
∫ . log e x = [ . + 4{ . + . + . } + 2{ . + . }+ . ]

= 1.8278472
3
(c) By Simpson's th rule, we have
8
3
log e x dx  h  y4.0  3 y4.2  y4.4  y4.8  y5.0   2 y4.6  y5.2 
5.2
4.0 8
= 1.8278470.
.
Actual value of ∫. = log e x − 1 5.2
4.0
= 1.8278475

Hence the errors due to different formulae are

(a) 0.0001924 (b) 0.0000003 (c) 0.0000005
1
We observe that Simpson’s rd rule is more accurate
3
1
Ex.2: Calculate by Simpson’s rd rule an approximate value of the integral
3
3

x
4
dx by taking seven equidistant ordinates. Compare it with the exact value and the value
3

obtained by using the Trapezoidal rule.

Solution:
x -3 -2 -1 0 1 2 3
y= 81 16 1 0 1 16 81

By Simpson’s rule, we get

3
h
3
 x 4 dx 
3
+ 4{ + + } + 2{ + } + y3 = 98
 3
3  x5 
 x dx     97.2
4
The exact value of
3  5  3
However, by Trapezoidal rule
3
h

3
x 4 dx 
2
+ 2{ + + + + }  y3 = 115.

In this case we observe that the Trapezoidal rule does not give an accurate result.

Example 3: Use Romberg’s method to compute I = ∫ correct to three decimal places

with Trapezoidal rule.

Solution:
We take h = 0.5,0.25 and 0.125 successively and using T.R., we get
I(0.5) = 0.7084, (0.25) = 0.6970 and I (0.125) = 0.6941
 I more correct  Iless correct 
Now, using the formula: Iimproved = I more correct +  
 2n  1 

I(0.5,0.25) = 0.6970 + (0.6970 − 0.7084) = 0.6932

I(0.25,0.125) = 0.6941 + (0.6941 − 0.6970) = 0.6931

Finally,
I(0.5,0.25,0.125) = 0.6931 + (0.6931 − 0.6932)

= 0.6931
In tabular form
I Error ( ) ( ) ( )
0.5 0.7084
0.6932
0.25 0.6970 0.6931
0.6931
0.125 0.6931

Example 4: Evaluate the integral

 /2
I sin x dx , using two-terms Gaussian formula.
0

Solution: First we must change the limits of integration from (0,  /2) to (-1, 1).
Let =
 / 2  t   / 2 =

( + 1), so =

dt .
2 4
 
∴ I = ∫ sin( ( + 1))

  
= 1.0. 1− + 1.0  1+
√ √

= 0.99847