Lecture Note for Numerical Analysis (14) Gauss Quadrature Formula

1. Concept of Gauss Quadrature

 Gauss quadrature implements a strategy of positioning any two points on a curve to define a straight line that
would balance the positive and negative errors.
 Hence the area evaluated under this straight line provides an improved estimate of the integral.

2. Standard Form of Integration for Gauss Quadrature Application

 Gauss quadrature can be standardized by using the integration interval of
x  [1, 1]
to calculate
xf
I  f ( x ) dx, x  [ x0 , x f ]
x0

 Affine Transformation

2 x  x0 x  x0 x  x0
t x f , t  [ 1, 1]  x  f t f
x f  x0 x f  x0 2 2
Using Affine transformation we can redefine above integration as

2 x  x0
dt  dx  dx  f dt
x f  x0 2
xf x f  x0 1  x f  x0 x f  x0  x f  x0 1
I  f ( x)dx  1  2
f  t   dt   g (t )dt
x0 2 2  2 1

1
3. Derivation of Gauss Quadrature Formula
 If we use the standard form, the integration can be estimated by integrating the following form

1
I   f ( x)dx
1

 General form of n-point Gauss Quadrature Formula

1
I   f ( x)dx
1

 w0 f  x0   w1 f  x1   w2 f  x2   w3 f  x3   w4 f  x4     wn 1 f  xn 1 
where
xj, j  0,1,2, , n  1 : Gauss Quadrature points

wj , j  0,1,2, , n  1 : Gauss Quadrature weights

x j ' s, w j ' s should be determined to balance the positive and negative errors

1
(3-1) Two-point Gauss quadrature formula to exactly integrate I  
1
f ( x)dx with f ( x )  1, x, x 2 , x 3

I  w0 f ( x0 )  w1 f ( x1 )
1
f ( x)  1  w0 f ( x0 )  w1 f ( x1 )  w0  w1   1dx  2
1 w0  w1  2
1
f ( x)  x  w0 f ( x0 )  w1 f ( x1 )  w0 x0  w1 x1   xdx  0 w0 x0   w1 x1
1

2
 2 2 2
f ( x)  x 2  w0 f ( x0 )  w1 f ( x1 )  w0 x0  w1x1   x 2 dx 
2 2 1 w0 x0  w1x1 
1 3 3
3 3
1 w0 x0   w1 x1
f ( x)  x  w0 f ( x0 )  w1 f ( x1 )  w0 x0  w x   x dx  0
3 3 3 3
1 1 1

2
 From the 2nd and 4th equations,
2 2
x0  x1  x0   x1
From the 2nd equation,

w0 x0  w1 x1  w0 x0  w1 x0  x0 ( w0  w1 )  0  w0  w1 since x0  0
From the 1st equation,

2 w0  2  w0  w1  1
From the 3rd equation
1
x0    0.5773503
2 2 2 2 2 3
w0 x0  w1 x1   2 x0  
3 3 1
x1   0.5773503
3
Therefore the integration formula becomes
I  w0 f ( x0 )  w1 f ( x1 )
 1   1 
 f   f 
 3  3

(3-2) Performance of two-point Gauss quadrature formula

(a) 2nd order polynomial function: f ( x )  a  bx  cx 2
1
- Exact integration of I   f ( x)dx
1

1 2
I   (a  bx  cx 2 )dx  2a  c
1 3
- Gauss Quadrature formula results in the same value

 1   1 
I  w0 f ( x0 )  w1 f ( x1 )  f    f 
 3  3
 1 1  1 1 2
 a b  c   a  b  c   2a  c
 3 3  3 3 3
(b) 3rd order polynomial function: f ( x )  a  bx  cx 2  dx 3
1 2
- Exact integration: I   (a  bx  cx 2  dx3 )dx  2a  c
1 3
- Gauss Quadrature formula results in the same value

 1   1 
I  w0 f ( x0 )  w1 f ( x1 )  f    f 
 3  3
 1 1 1   1 1 1  2
 a b c d   a  b c d   2a  c
 3 3 3 3  3 3 3 3 3

3
 (c) 4th order polynomial function: f ( x )  a  bx  cx 2  dx 3  ex 4
1 2 2
- Exact integration: I   (a  bx  cx 2  dx3  ex 4 )dx  2a  c  e
1 3 5
- Gauss Quadrature formula generates some error

 1   1 
I  w0 f ( x0 )  w1 f ( x1 )  f    f 
 3  3
 1 1 1 1  1 1 1 1 2 2
 a b c d  e   a  b c d  e   2a  c  e
 3 3 3 3 9  3 3 3 3 9 3 9

In general, the n-point Gauss Quadrature formula gives the exact integration up to (2n-1)th order

polynomials

4
4. Application of Gauss Quadrature Formula

 Given
- Number of Gauss Quadrature Point: n
- Data for Gauss Quadrature Point
(a) Arguments :  j, j  1,2,  , n ( j  ( 1,1) )

(b) Weights : wj , j  1,2,  , n

- Function and the integration interval

xf
I  f ( x ) dx, x  [ x0 , x f ]
x0

 Transform using the affine transformation to get the standard form

2 x f  x0
d  dx  dx  d
x f  x0 2
xf x f  x0
 x f  x0 x  x0 
1
I  f 
f ( x)dx  
 f  d
x0 2 1
 2 2 
x  x0 1 x  x0 ~ ~ 1
 f
2 1
g ( )d  f
2
I  I   g ( )d
1

 Integration Formula
n
~ 1
I   g ( )d   w j g ( j )
1
j 1

 x  x0  ~  x f  x0  n  x  x0  n  x  x0 x  x0 
I   f  I    w j g ( j )   f  w j f  f j  f 
 2   2  j 1  2  j 1  2 2 

 Weights and function arguments x in Gauss Quadrature for the different number of quadrature points

Points n Weighting Factors wj Function Arguments j

2 w1  1.000000000  1  0.577350269
w2  1.000000000  2  0.577350269

3 w1  0.555555556  1  0.774596669
w2  0.888888889  2  0.000000000
w3  0.555555556  3  0.774596669

4 w1  0.347854845  1  0.861136312
w2  0.652145155  2  0.339981044
w3  0.652145155  3  0.339981044
w4  0.347854845  4  0.861136312

5
 5 w1  0.236926885  1  0.906179846
w2  0.478628670  2  0.538469310
w3  0.568888889  3  0.000000000
w4  0.478628670  4  0.538469310
w5  0.236926885  5  0.906179846

6 w1  0.171324492  1  0.932469514
w2  0.360761573  2  0.661209386
w3  0.467913935  3  0.238619186
w4  0.467913935  4  0.238619186
w5  0.360761573  5  0.661209386
w6  0.171324492  6  0.932469514

Example 1]
 Given for n=3

Points n Weighting Factors wj Function Arguments j

w1  0.555555556
 1  0.774596669
w2  0.888888889
 2  0.000000000
3
w3  0.555555556
 3  0.774596669

5
I   f ( x)dx, x  [0, 5]
0  Exact Integral
5
 
I   3x 2  2 x dx  x3  x 2
0
  5

0
 53  52  150.0
f ( x)  3x 2  2 x

 Solution
x f  x0 x f  x0
I  2.5 w j f 2.5 j  2.5 
n
 2.5,  2.5
j 1 2 2
 2.5w1 f 2.5 1  2.5  w2 f 2.5 2  2.5  w3 f 2.5 3  2.5
x1  2.5 1  2.5  2.5  0.774596669  2.5  0.56350833
x2  2.5 2  2.5  2.5  0.0  2.5  2.5
x3  2.5 3  2.5  2.5  0.774596669  2.5  4.436491672

f1  f (0.56350833)  3x 2  2 x  x  0.56350833
 2.079641560

f2  f (2.5)  3x  2 x 2
 23.75
x  2.5

f3  f (4.436491672)  3 x  2 x  2
 67.920358425
x  4.436491672

I  2.5(0.555555556 f1  0.888888889 f 2  0.555555556 f 3 )  150.0

6
5. Pseudo code for the Gauss Quadrature

Program main
% Input
n =3;
xmin = 0.0;
xmax = 5.0;
dx_plus = 0.5*(xmax+xmin);
dx_minus = 0.5*(xmax-xmin);
% Guass quadrature nodes and weights
call gauss_node(n,tau,w);
% Integration using Guass quadrature formula
gauss_integral = 0.0;
do j=1, n
x = dx_minus*tau(j)+dx_plus;
call f(x,y);
gauss_integral = gauss_integral + w(j)*y;
end do;
gauss_integral = dx_minus*guass_integral;
end program main

function gauss_node(n,tau,w)
if n=2, then
w1  1.000000000 , w2  1.000000000
1  0.577350269,  2  0.577350269
else if n=3, then
w1  0.555555556, w2  0.888888889, w3  0.555555556
1  0.774596669,  2  0.000000000,  3  0.774596669
else if n=4, then
w1  0.347854845, w2  0.652145155, w3  0.652145155 , w4  0.347854845
1  0.861136312,  2  0.339981044,  3  0.339981044 ,  4  0.861136312
else
print*,‘node number n exceeds the maximum allowed node number= 4.’
print*, ‘Please add the node information (Gauss quadrature points and weights’
stop
end if
end function gauss_node

function f(x,y);% Function to be integrated, which should be specified by the user

y=3.0*x*x*x+2.0*x;
end function f

7
6. Advanced Topics on Gauss Quadrature
(6-1) Theorem (Gauss quadrature) Let { j } Nj 0 be the set of the zeros of the polynomial p N 1 ( )
orthogonal with respect to a weighting function () . Then there exists a unique set of quadrature weights
defined by
1 N
(   k )
w j    j()()d where  j()  
1
k 0 ( j   k )
k j

and the exact weighted integration of the polynomial function , the degree of which is less than or equal to
2N+1, can be obtained with
N
p()()d   p( j )wj ,
1
1
j 0
p()  P2 N 1

The quadrature weights are all positive and expressed as

2
k pN 
wj  N 1 , j  0,1,, N
k N pN ( j ) pN 1 ( j )
Where kN and k N 1 are the leading coefficients p N ( ) and p N 1 ( ) , respectively. And
1
 p N ()  ( )p N ()d  is the weighted function norm of p N () .
2
pN 

1

(6-2) Othogonal Polynomial example: Legendre polynomials orthogonal w.r.t () 1

 Important Characteristics of the Legendre polynomials

■ Legendre polynomials: Eigen functions of the sigular Sturm-Lioville problem

d  2 dLN ( ) 
(1   )   N ( N  1) LN ( )  0 with LN ( 1 )  1
d  d 

■ Leading coefficient of N-th order Legendre polynomial

( 2 N )!
kN  in L N ( )  k N  N
 k N 1 N 1    k1 1  k 0
2 N ( N !) 2

■ Orthogonality with unit weighting function

1 2
(Lj , Lk )   Lj ( ) ( ) Lk ( )d   jk
1 2 j 1

■ Three term recurrence formula

( N  1) LN 1 ( )  (2 N  1)LN ( )  NLN 1 ( ), with L0 ( )  1, L1 ( )  

■ Recurrence relation for derivatives

( 2 N  1) LN ( )  LN 1 ( )  LN 1 ( ), N 1
N ( N  1)
(1   2 ) LN ( )  LN 1 ( )  LN 1 ( ) 
2N  1

8
  Derivation of Legendre polynomials using the recurrence relation

L0 ( )  1
L1 ( )  
3 1 1
L2 ( )  L1 ( )  L0 ( )  (3 2  1)
2 2 2
5 2 5 2 1
L3 ( )  L2 ( )  L1 ( )  (3 3   )    (5 3  3 )
3 3 6 3 2
7 3 7 3 1
L4 ( )  L3 ( )  L2 ( )  (5 4  3 2 )  (3 2  1)  (35 4  30 2  3)
4 4 8 8 8
9 4 9 2 1
L5 ( )  L4 ( )  L3 ( )  (35 5  30 3  3 )  (5 3  3 )  (63 5  70 3  15 )
5 5 40 5 8
11 5 11 5
L6 ( )  L5 ( )  L4 ( )  (63 6  70 4  15 2 )  (35 4  30 2  3)
6 6 48 48
1
 (231 6  315 4  105 2  5)
16

( N  1) LN 1 ( )  ( 2 N  1)LN ( )  NLN 1 ( )
(6-3) Gauss Quadrature nodes { j } Nj 0 and weights {w j } Nj 0

 Nodes are zeros of LN 1( )

N  0  L1 ( )    0 0  0
1
N  1  L2 ( )  (3 2  1)  0   0  1 / 3 ,  1  1 / 3
2
1
N  2  L3 ( )  (5 3  3 )  0   0   3 / 5 ,  1  0,  0  3 / 5
2

N  j  L j 1 ( )  0 , the roots of which are generally calculated using the numerical root finding method.
2
k LN 
 Weights: w j  N 1 , j  0,1,, N
k N LN ( j ) LN 1 ( j )
Using the following relations
k N 1 (2 N  2)! 2 N ( N !)!2 (2 N  2)(2 N  1) (2 N  1)
 N 1  
k N 1 2 ( N  1)!2 (2 N )! 2( N  1) 2 ( N  1)
1 2
LN   ( LN , LN ) 1   LN ( ) ( ) LN ( )d 
2
1 2N  1
(1   2j )
(1   ) L N 1 ( )  ( N  1) L N ( )  L N 1 ( )   LN ( j ) 
2
LN 1 ( j )
( N  1)

The quadrature weights can be written as (also, weights are generally calculated using numerical methods)
( 2 N  1) 2 ( N  1)
wj  , j  0,1,  , N
( N  1) ( 2 N  1) (1   j ) LN 1 ( j ) LN 1 ( j )
2

2
 2
j 
(1   ) LN 1 ( j ) 
2

9
Gauss Quadrature Nodes and Weights
N  0   0  0, L1( )  1  w0  2
N  1   0  1 / 3 ,  1  1 / 3 , L2 ( )  3  w0  w1  2 /1  1 / 33  1
1
N  2   0   3 / 5 ,  1  0,  2  3 / 5 , L3 ( )  (15 2  3)
2
 1 
2
 5
w0  w2  2 / 1  3 / 5 9  3 
 2   9

  3   8
2

w1  2 /     
  2   9

15  225  105 1
N  3  (35 4  30 2  3)  0   0 ,1, 2 ,3    0.86113631,  0.33998104 , L4 ( )  (105 3  60 )
35 8
w0  w2  0.34785484
w1  w3  0 .65214515


N No Node weight N No Node weight

1 0.000000000000E+00 .200000000000E+01 1 -.577350269190E+00 .100000000000E+01

2 0.577350269190E+00 .100000000000E+01
1 2
1 -.774596669241E+00 0.555555555556E+00 1 -.861136311594E+00 .347854845137E+00
2 0.000000000000E+00 0.888888888889E+00 2 -.339981043585E+00 .652145154863E+00
3 0.774596669241E+00 0.555555555556E+00 4 3 0.339981043585E+00 .652145154863E+00
3
4 0.861136311594E+00 .347854845137E+00
1 -.906179845939E+00 0.236926885056E+00 1 -.932469514203E+00 0.171324492379E+00
2 -.538469310106E+00 0.478628670499E+00 2 -.661209386466E+00 0.360761573048E+00
6
5 3 0.000000000000E+00 0.568888888889E+00 3 -.238619186083E+00 0.467913934573E+00
4 0.538469310106E+00 0.478628670499E+00 4 0.238619186083E+00 0.467913934573E+00
5 0.906179845939E+00 0.236926885056E+00 5 0.661209386466E+00 0.360761573048E+00
6 0.932469514203E+00 0.171324492379E+00
1 -.949107912343E+00 0.129484966169E+00 1 -.960289856498E+00 0.101228536290E+00
2 -.741531185599E+00 0.279705391489E+00 2 -.796666477414E+00 0.222381034453E+00
3 -.405845151377E+00 0.381830050505E+00 8 3 -.525532409916E+00 0.313706645878E+00
7
4 0.000000000000E+00 0.417959183673E+00 4 -.183434642496E+00 0.362683783378E+00
5 0.405845151377E+00 0.381830050505E+00 5 0.183434642496E+00 0.362683783378E+00
6 0.741531185599E+00 0.279705391489E+00 6 0.525532409916E+00 0.313706645878E+00
7 0.949107912343E+00 0.129484966169E+00 7 0.796666477414E+00 0.222381034453E+00
8 0.960289856498E+00 0.101228536290E+00
1 -.968160239508E+00 0.812743883616E-01 1 -.973906528517E+00 0.666713443087E-01
2 -.836031107327E+00 0.180648160695E+00 2 -.865063366689E+00 0.149451349151E+00
3 -.613371432701E+00 0.260610696403E+00 10 3 -.679409568299E+00 0.219086362516E+00
9
4 -.324253423404E+00 0.312347077040E+00 4 -.433395394129E+00 0.269266719310E+00
5 0.000000000000E+00 0.330239355001E+00 5 -.148874338982E+00 0.295524224715E+00
6 0.324253423404E+00 0.312347077040E+00 6 0.148874338982E+00 0.295524224715E+00
7 0.613371432701E+00 0.260610696403E+00 7 0.433395394129E+00 0.269266719310E+00
8 0.836031107327E+00 0.180648160695E+00 8 0.679409568299E+00 0.219086362516E+00
9 0.968160239508E+00 0.812743883616E-01 9 0.865063366689E+00 0.149451349151E+00
10 0.973906528517E+00 0.666713443087E-01
1 -.978228658146E+00 0.556685671162E-01 1 -.981560634247E+00 0.471753363865E-01
2 -.887062599768E+00 0.125580369465E+00 2 -.904117256370E+00 0.106939325995E+00
3 -.730152005574E+00 0.186290210928E+00 12 3 -.769902674194E+00 0.160078328543E+00
11
4 -.519096129207E+00 0.233193764592E+00 4 -.587317954287E+00 0.203167426723E+00
5 -.269543155952E+00 0.262804544510E+00 5 -.367831498998E+00 0.233492536538E+00
6 0.000000000000E+00 0.272925086778E+00 6 -.125233408511E+00 0.249147045813E+00
7 0.269543155952E+00 0.262804544510E+00 7 0.125233408511E+00 0.249147045813E+00
8 0.519096129207E+00 0.233193764592E+00 8 0.367831498998E+00 0.233492536538E+00
9 0.730152005574E+00 0.186290210928E+00 9 0.587317954287E+00 0.203167426723E+00
10 0.887062599768E+00 0.125580369465E+00 10 0.769902674194E+00 0.160078328543E+00
11 0.978228658146E+00 0.556685671162E-01 11 0.904117256370E+00 0.106939325995E+00
12 0.981560634247E+00 0.471753363865E-01

10