NUMERICAL INTEGRATION

(NUMERICAL QUADRATURE)

Dr. Gabriel Obed Fosu

Department of Mathematics
Kwame Nkrumah University of Science and Technology
Google Scholar: https://scholar.google.com/citations?user=ZJfCMyQAAAAJ&hl=en&oi=ao
ResearchGate ID: https://www.researchgate.net/profile/Gabriel_Fosu2

Dr. Gabby (KNUST-Maths) Numerical Integration 1 / 44

Lecture Outline

1 Trapezium Rule

2 Simpson’s 1/3 Rule

3 Simpson’s 3/8 Rule

4 Romberg Method

5 Integration Rules Based on Non-uniform Mesh Spacing

Dr. Gabby (KNUST-Maths) Numerical Integration 2 / 44

 Introduction

Integration Rules Based on Uniform Mesh Spacing

1 Generally, integration problems deals with finding an approximate value of the integral
Z b
I= w(x) f (x)d x (1)
a
where w(x) > 0 the weight function lies in the open interval (a, b), and I a definite
integral.
2 I reduces to an indefinite integral when the limits of integration are not specified.
b−a
3 When w(x) = 1 and x k ’s are equidistant with x 0 = a, x n = b, h = , where N is the
N
number of subdivision, then the integralZ (1) reduces to
xn
I= f (x)d x (2)
x0

4 This integral (2) defines the area under the curve above the x−axis within the interval
[x 0 , x n ].
Dr. Gabby (KNUST-Maths) Numerical Integration 3 / 44
 Introduction

Integration Rules Based on Uniform Mesh Spacing

The following integration techniques will be employed in finding the value of this
area.
1 Trapezium Method
2 Simpson’s Method
3 Romberg Method
Note
These methods are called Newton–Cotes formulas, or the Newton–Cotes
quadrature rules or simply Newton–Cotes rules.

These techniques are all based on evaluating the integrand at equally spaced
points. They are named after Isaac Newton and Roger Cotes.
Dr. Gabby (KNUST-Maths) Numerical Integration 4 / 44
 Trapezium Rule Simple Trapezium Rule

Simple Trapezium Rule

This is also called the trapezoidal method. Let the curve y = f (x), a ≤ x ≤ b be
approximated by the line joining the points P (a, f (a)) and Q(b, f (b)) on the curve
as illustrated below

Dr. Gabby (KNUST-Maths) Numerical Integration 6 / 44

 Trapezium Rule Simple Trapezium Rule

Simple Trapezium Rule

Then the trapezium method uses the area under the approximate line to obtain
the formula
1 £ ¤
I = h f (a) + f (b) (3)
2
1 h i
= (b − a) f (a) + f (b) ; N =1 (4)
2

Note
The trapezium rule of integration integrates exactly polynomials of degree ≤ 1

Dr. Gabby (KNUST-Maths) Numerical Integration 7 / 44

 Trapezium Rule Simple Trapezium Rule

Example
Approximate the following integrals using the simple trapezium rule
R5
1 2x 2 d x
R3 π/4
2 sin(x)d x
0
Hence determine the absolute error (AE)

Solution R
5
1. Given 3 2x 2 d x , then a = 3 and b = 5. The functional values are
f (a) = f (3) = 2(3)2 = 18
f (b) = f (5) = 2(5)2 = 50

Dr. Gabby (KNUST-Maths) Numerical Integration 8 / 44

 Trapezium Rule Simple Trapezium Rule

Therefore
1 £ ¤
I = (b − a) f (a) + f (b) (5)
2
1
= (5 − 3)[18 + 50] (6)
2
= 68 (7)

For absolute error, the exact solution is required

Exact solution

5 2x 3 ¯¯5
Z ¯
2
2x d x = = 65.33 (8)
3 3 ¯3

AE = |E S − AS| = |65.33 − 68| = 2.66

Dr. Gabby (KNUST-Maths) Numerical Integration 9 / 44
 Trapezium Rule Simple Trapezium Rule

π
R π/4
2. Given 0
sin(x)d x , then a = 0 and b = . The functional values are
4
f (a) = f (0) = sin(0) = 0
f (b) = f (π/4) = sin(π/4) = 0.707. Therefore
1 £ ¤
I = (b − a) f (a) + f (b) (9)
2
1
= (π/4 − 0)[0 + 0.707] (10)
2
π
= (0.707) (11)
8
= 15.9075 (12)
Exact solution
Z π/4
¯π/4
¯
sin(x)d x = − cos(x)¯¯ = − cos(π/4) + cos(0) = 0.293 (13)
0 0

AE = |E S − AS| = |0.293 − 15.9075| = 15.614

Dr. Gabby (KNUST-Maths) Numerical Integration 10 / 44
 Trapezium Rule Composite Trapezium Rule

Composite Trapezium Rule

Here, the idea is to split the interval (a, b) into a sequence of N smaller
b−a
sub-interval with width h = . These yields the composite trapezium formula
N
Z b h
½ · ¸ ¾
f (x)d x = f (x 0 ) + 2 f (x 1 ) + f (x 2 ) + · · · + f (x N 1 ) + f (x N ) (14)
a 2

Note
The nodal points are given by

a = x 0 , x 1 = x 0 + h, x 2 = x 0 + 2h, · · · , x N = x 0 + N h = b (15)

Dr. Gabby (KNUST-Maths) Numerical Integration 11 / 44

 Trapezium Rule Composite Trapezium Rule

Example
Find the approximate solution for the following using the composite trapezium rule
with 4 equal sub-intervals.
R5
1 2x 2 d x
3
R2
2 ln(x) d x
0

Dr. Gabby (KNUST-Maths) Numerical Integration 12 / 44

 Trapezium Rule Composite Trapezium Rule

Solution
b −a 5−3
For 4 sub-intervals =⇒ h = = = 0.5.
N 4
Therefore the x points are

x 0 = 3, x 1 = 3.5, x 2 = 4, x 3 = 4.5, x4 = 5

The functional value are

f (x 0 ) = f (3) = 18 (16)
f (x 1 ) = f (3.5) = 24.5 (17)
f (x 2 ) = f (4) = 32 (18)
f (x 3 ) = f (4.5) = 40.5 (19)
f (x 4 ) = f (5) = 50 (20)

Dr. Gabby (KNUST-Maths) Numerical Integration 13 / 44

 Trapezium Rule Composite Trapezium Rule

Substituting into the formula

Z b hn h i o
f (x)d x = f (x 0 ) + 2 f (x 1 ) + f (x 2 ) + · · · + f (x N −1 ) + f (x N ) (21)
a 2
Z 5
0.5 n h i o
2x 2 d x = f (x 0 ) + 2 f (x 1 ) + f (x 2 ) + f (x 3 ) + f (x 4 ) (22)
3 2
0.5
= {18 + 2[24.5 + 32 + 40.5] + 50} (23)
2
= 65.5 (24)

From above the exact solution is 65.33, therefore the absolute error is

AE = |E S − AS| = |65.33 − 65.5| = 0.17

By comparison, we realize that the composite trapezium is more accurate than

the simple trapezium method.
Dr. Gabby (KNUST-Maths) Numerical Integration 14 / 44
 Simpson’s 1/3 Rule Simple Simpson’s 1/3 Rule

Simple Simpson’s 1/3 Rule

1 This is based on a quadratic curve through equally spaced point rather than a
line as is the case of the simple trapezium rule.
2 The interval (a, b) is subdivided into two equal parts with the step length
h = b−a
2
.
3 We approximate the integral curve by the parabola joining these points.
4 The formula of the Simpson’s 1/3 rule is deduced from interpolation
techniques, and it is given by
Z b h
· µ
a +b
¶ ¸
f (x)d x = f (a) + 4 f + f (b) (25)
a 3 2

5 The Simpson’s rule integrates exactly polynomials of degree ≤ 3.

Dr. Gabby (KNUST-Maths) Numerical Integration 16 / 44
 Simpson’s 1/3 Rule Simple Simpson’s 1/3 Rule

Example
Find the derivative of the following functions using the simple Simpson’s 1/3 rule
R5
1 2x 2 d x
R3 3 1
2 dx
1 1+x
b −a 5−3
For 2 sub-intervals =⇒ h = = = 1.
2 2
Therefore the x points are
[x 0 , x 1 , x 2 ] = [3, 4, 5]
The functional value are f (x 0 ) = f (3) = 18 = f (a) (26)
a +b
µ ¶
f (x 1 ) = f (4) = 32 = f (27)
2
f (x 2 ) = f (5) = 50 = f (b) (28)
Dr. Gabby (KNUST-Maths) Numerical Integration 17 / 44
 Simpson’s 1/3 Rule Simple Simpson’s 1/3 Rule

Substituting into the formula

Z b b−a
· µ
a +b
¶ ¸
f (x)d x = f (a) + 4 f + f (b) (29)
a 6 2
Z 5
5−3
2x 2 d x = [18 + 4(32) + 50] (30)
3 6
1
= (196) (31)
3
= 65.33 (32)

Dr. Gabby (KNUST-Maths) Numerical Integration 18 / 44

 Simpson’s 1/3 Rule Composite Simpson’s 1/3 Rule

Composite Simpson’s 1/3 Rule

1 This is an extension of the simple Simpson’s rule.
2 Here, the given interval [a, b] can be divided into any finite sub-intervals of
equal length say h .
3 For N sub-intervals the formula is given as
Z b hn h i
f (x)d x= f (x 0 ) + 4 f (x 1 ) + f (x 3 ) + f (x 5 ) + · · · + f (x N −1 ) (33)
a 3
h i o
+2 f (x 2 ) + f (x 4 ) + f (x 6 ) + · · · + f (x N −2 ) + f (x N )

4 The composite Simpson’s 1/3 rule is also of order 3, that is the composite
Simpson’s 1/3 rule produces exact results for polynomials of degree ≤ 3.

Dr. Gabby (KNUST-Maths) Numerical Integration 19 / 44

 Simpson’s 1/3 Rule Composite Simpson’s 1/3 Rule

Example
Find the derivative of the following functions using the composite Simpson’s 1/3
rule using 4 sub-intervals
R5
1 2x 2 d x
R3 3 1
2 dx
1 1+x

Dr. Gabby (KNUST-Maths) Numerical Integration 20 / 44

 Simpson’s 1/3 Rule Composite Simpson’s 1/3 Rule

Solution
b −a 5−3
For 4 sub-intervals =⇒ h = = = 0.5.
4 4
Therefore the x points are

x 0 = 3, x 1 = 3.5, x 2 = 4, x 3 = 4.5, x 4 = 5

The functional value are

f (x 0 ) = f (3) = 18 (34)
f (x 1 ) = f (3.5) = 24.5 (35)
f (x 2 ) = f (4) = 32 (36)
f (x 3 ) = f (4.5) = 40.5 (37)
f (x 4 ) = f (5) = 50 (38)

Dr. Gabby (KNUST-Maths) Numerical Integration 21 / 44

 Simpson’s 1/3 Rule Composite Simpson’s 1/3 Rule

Substituting into the formula

Z b hn h i
f (x)d x = f (x 0 ) + 4 f (x 1 ) + f (x 3 ) + f (x 5 ) + · · · + f (x N −1 ) (39)
a 3
h i o
+ 2 f (x 2 ) + f (x 4 ) + f (x 6 ) + · · · + f (x N −2 ) + f (x N )

(40)
Z 5 hn h i h i o
2x 2 d x = f (x 0 ) + 4 f (x 1 ) + f (x 3 ) + 2 f (x 2 ) + f (x 4 )
3 3
0.5 n o
= 18 + 4(24.5 + 40.5) + 2(32) + 50 (41)
3
= 0.167(392) (42)
= 65.474 (43)

Dr. Gabby (KNUST-Maths) Numerical Integration 22 / 44

 Simpson’s 3/8 Rule Simple Simpson’s 3/8 Rule

Simple Simpson’s 3/8 Rule

1 To derive the Simpson’s 1/3 rule, we have approximated f (x) by a quadratic
polynomial.
2 To derive the Simpson’s 3/8 rule, we approximate f (x) by a cubic polynomial.
3 For interpolating by a cubic polynomial, we require four nodal points. Hence,
we subdivide the given interval [a, b] into 3 equal parts so that we obtain four
nodal points.
4 Let h = (b − a)/3, then the nodal points are given by x 0 = a, x 1 = x 0 + h, x 2 =
x 0 + 2h, x 3 = x 0 + 3h.
5 Using the Newton’s forward difference formula, the integral of the cubic
polynomial approximation to f(x) gives the Simpson’s 3/8 rule as
Z b 3h £ ¤
f (x)d x = f (x 0 ) + 3 f (x 1 ) + 3 f (x 2 ) + f (x 3 ) (44)
a 8

Dr. Gabby (KNUST-Maths) Numerical Integration 24 / 44

 Simpson’s 3/8 Rule Composite Simpson’s 3/8 Rule

Composite Simpson’s 3/8 Rule

1 This is an extension of the simple Simpson’s 3/8 rule.
2 In this case, we subdivide [a, b] into a number of subintervals of equal length
such that the number of subintervals is divisible by 3.
3 For example, if we divide [a, b] into 6 parts, then we get 7 nodal points with
Simpson’s 3/8 rule as
Z b Z x3 Z x6
f (x)d x = f (x)d x + f (x)d x
a x0 x3
3h
= [{ f (x 0 ) + 3 f (x 1 ) + 3 f (x 2 ) + f (x 3 )} + { f (x 3 ) + 3 f (x 4 ) + 3 f (x 5 ) + f (x 6 )}]
8
3h
= [ f (x 0 ) + 3 f (x 1 ) + 3 f (x 2 ) + 2 f (x 3 ) + 3 f (x 4 ) + 3 f (x 5 ) + f (x 6 )]
8

Dr. Gabby (KNUST-Maths) Numerical Integration 25 / 44

 Simpson’s 3/8 Rule Composite Simpson’s 3/8 Rule

Example
R2 1
Using the Simpson’s 3/8 rule, evaluate 1
d x with 3 and 6 subintervals.
5 + 3x
Compare with the exact solution.
When n = 3, we have the following step lengths and nodal points.
b −a 2−1 1
h= = = (45)
n 3 3
The nodes are
x 0 = 1, x 1 = x 0 + h = 4/3, x 2 = x 1 + h = 5/3, x 2 = 2.0.
We have the following tables of values
x 1.0 4/3 5/3 2.0
f (x) 0.125 0.11111 0.10000 0.09091
Dr. Gabby (KNUST-Maths) Numerical Integration 26 / 44
 Simpson’s 3/8 Rule Composite Simpson’s 3/8 Rule

n =3
Z b 3h £ ¤
f (x 0 ) + 3 f (x 1 ) + 3 f (x 2 ) + f (x 3 )
f (x)d x = (46)
a 8
Z 2
1 31
dx = [0.125 + 3(0.11111) + 3(0.10000) + 0.09091] (47)
1 5 + 3x 83
= 0.10616 (48)
Exact Solution
1
F (x) = ln(5 + 3x) (49)
3
¯2 1
F (x)¯ = [ln 11 − ln 8] = 0.10615. (50)
¯
1 3
(51)
The magnitude of the error is 0.00001
Dr. Gabby (KNUST-Maths) Numerical Integration 27 / 44
 Simpson’s 3/8 Rule Composite Simpson’s 3/8 Rule

When n=6
b −a 2−1 1
h= = = (52)
6 6 6
We have the following tables of values
x 1.0 7/6 8/6 9/6 10/6 11/6 2.0
f (x) 0.125 0.11765 0.11111 0.10526 0.10000 0.09524 0.09091
Z b 3h
f (x)d x =
[ f (x 0 ) + 3 f (x 1 ) + 3 f (x 2 ) + 2 f (x 3 ) + 3 f (x 4 ) + 3 f (x 5 ) + f (x 6 )] (53)
a 8
Z 2
1 31
dx = [0.125 + 3(0.11765 + 0.11111 + 0.10000 + 0.09524) + 2(0.10526) + 0.09091]
1 5 + 3x 86
(54)
= 0.10615 (55)
Dr. Gabby (KNUST-Maths) Numerical Integration 28 / 44
 Romberg Method

Romberg Method

1 Romberg integration uses the composite trapezium/Simpson’s(1/3) rule

to give preliminary approximations and then applies the Richardson
extrapolation process to improve the approximations.
2 Usually, we start with a coarse step length, then reduce the step lengths and
recompute the value of the integral.
3 The sequence of these values converges to the exact value of the integral.

Definition
The Richardson extrapolation is a sequence acceleration method used to improve
the rate of convergence of a sequence of estimates of some value say A .

Dr. Gabby (KNUST-Maths) Numerical Integration 30 / 44

 Romberg Method Romberg method for the trapezium rule

Romberg method for the trapezium rule

1 The Romberg extrapolation procedure for the composite trapezium rule is

(m) 4m I (m−1) (h/2) − I (m−1) (h)

I (h) = ; m = 1, 2, 3, · · · (56)
4m − 1
2 where I (0) (h) = I (h). Note these initial values are computed from the trapezium
method discussed earlier.
h h
3 This often uses the step lengths h, , ,···
2 22

Dr. Gabby (KNUST-Maths) Numerical Integration 31 / 44

 Romberg Method Romberg method for the trapezium rule

Example
R1 1
Solve 0 d x using Romberg method with two sub-interval. Compute your
1+x
initial values using the composite trapezium method.
Perform two iterations.
Recall that the½composite
· trapezium formula is
h
¸ ¾
Rb
a f (x)d x = f (x 0 ) + 2 f (x 1 ) + f (x 2 ) + · · · + f (x N 1 ) + f (x N ) where h = (b − a)/N
2
The approximations using the trapezium rule to the integral with various values of
the step lengths were obtained as follows:
1st h: N = 2 =⇒ h0 = 1/2, =⇒ I (0) (1/2) = 0.708334 (57)
2nd h: h = h0 /2 = 1/4, =⇒ N = 4 =⇒ I (0) (1/4) = 0.697024 (58)
2 (0)
3rd h: h = h0 /2 = 1/8, =⇒ N = 8 =⇒ I (1/8) = 0.694122. (59)
Dr. Gabby (KNUST-Maths) Numerical Integration 32 / 44
 Romberg Method Romberg method for the trapezium rule

When m = 1, h = 1/2
4m I (m−1) (h/2) − I (m−1) (h)
I (m) (h) = ; m = 1, 2, 3, · · · (60)
4m − 1
4I (0) (1/4) − I (0) (1/2)
I (1) (1/2) = (61)
4−1
4(0.697024) − 0.708334
= (62)
3
= 0.693254 (63)

When m = 2, h = 1/2
42 I (1) (1/4) − I (1) (1/2)
I (2) (1/2) = (64)
42 − 1
But I (1) (1/4) is not known. This should be computed with the same formula with
h = 1/4
Dr. Gabby (KNUST-Maths) Numerical Integration 33 / 44
 Romberg Method Romberg method for the trapezium rule

When m = 1, h = 1/4
4I (0) (1/8) − I (0) (1/4)
I (1) (1/4) = (65)
4−1
4(0.694122) − 0.697024
= (66)
3
= 0.693155 (67)

Thus eq. (64) becomes

When m = 2, h = 1/2
42 I (1) (1/4) − I (1) (1/2)
I (2) (1/2) = (68)
42 − 1
16(0.693155) − 0.693254
= (69)
15
= 0.693148 (70)
Dr. Gabby (KNUST-Maths) Numerical Integration 34 / 44
 Romberg Method Romberg method for the trapezium rule

Exact Solution

F (x) = ln(1 + x) (71)

¯1
F (x)¯ = ln(2) − ln(1) (72)
¯
0
= 0.693147 (73)

Magnitude of the error is 0.000001.

Dr. Gabby (KNUST-Maths) Numerical Integration 35 / 44

 Romberg Method Romberg method for the Simpson’s 1/3 rule

Romberg method for the Simpson’s 1/3 rule

1 The Romberg extrapolation procedure for the composite Simpson’s 1/3 rule is

4m+1 I (m−1) (h/2) − I (m−1) (h)

I (m) (h) = ; m = 1, 2, 3, · · · (74)
4m+1 − 1
2 where I (0) (h) = I (h). Note these initial values are computed from the
Simpson’s 1/3 method discussed earlier.
h h
3 This also uses the step lengths h, , ,···
2 22

Dr. Gabby (KNUST-Maths) Numerical Integration 36 / 44

 Romberg Method Romberg method for the Simpson’s 1/3 rule

Example
R1 1
Solve 0
d x using Romberg method with two sub-interval. Compute your
1+x
initial values using the composite Simpson’s 1/3 method.
Perform two iterations.
The approximations using the Simpson’s 1/3 rule to the integral with various
values of the step lengths were obtained as follows:

1st h: N = 2 =⇒ h0 = 1/2, =⇒ I (0) (1/2) = 0.694444 (75)

2nd h: h = h0 /2 = 1/4, =⇒ N = 4 =⇒ I (0) (1/4) = 0.693254 (76)
3rd h: h = h0 /22 = 1/8, =⇒ N = 8 =⇒ I (0) (1/8) = 0.693155 (77)

Dr. Gabby (KNUST-Maths) Numerical Integration 37 / 44

 Romberg Method Romberg method for the Simpson’s 1/3 rule

When m = 1, h = 1/2
4m+1 I (m−1) (h/2) − I (m−1) (h)
I (m) (h) = ; m = 1, 2, 3, · · · (78)
4m+1 − 1
16I (0) (1/4) − I (0) (1/2)
I (1) (1/2) = (79)
16 − 1
16(0.693254) − 0.694444
= (80)
15
= 0.693175 (81)

When m = 2, h = 1/2
43 I (1) (1/4) − I (1) (1/2)
I (2) (1/2) = (82)
43 − 1
But I (1) (1/4) is not known. This should be computed with the same formula with
h = 1/4
Dr. Gabby (KNUST-Maths) Numerical Integration 38 / 44
 Romberg Method Romberg method for the Simpson’s 1/3 rule

When m = 1, h = 1/4
16I (0) (1/8) − I (0) (1/4)
I (1) (1/4) = (83)
16 − 1
16(0.693155) − 0.693244
= (84)
15
= 0.693148 (85)

Thus eq. (82) becomes

When m = 2, h = 1/2
43 I (1) (1/4) − I (1) (h)
I (2) (1/2) = (86)
43 − 1
64(0.693148) − 0.693175
= (87)
63
= 0.693148 (88)
Dr. Gabby (KNUST-Maths) Numerical Integration 39 / 44
 Romberg Method Romberg method for the Simpson’s 1/3 rule

Exact Solution

F (x) = ln(1 + x) (89)

¯1
F (x)¯ = ln(2) − ln(1) (90)
¯
0
= 0.693147 (91)

Magnitude of the error is 0.000001.

Dr. Gabby (KNUST-Maths) Numerical Integration 40 / 44

 Integration Rules Based on Non-uniform Mesh Spacing

Gaussian Integration (Quadrature)

We have defined the general integration rule as
Z b
I= ω(x) f (x) d (x) (92)
a

We have the following Gaussian integration rules depending on the limits of

integration and on the expression for the weight function w(x).
1 Gauss-Legendre integration rule: [a, b] = [−1, 1] and w(x) = 1.
1
2 Gauss-Chebychev integration rule:[a, b] = [−1, 1] and w(x) = p
1 − x2
−x
3 Gauss-Laguerre integration rule: [a, b] = [0, ∞) and w(x) = e
2
4 Gauss-Hermite integration rule:[a, b] = (−∞, ∞) and w(x) = e −x

Dr. Gabby (KNUST-Maths) Numerical Integration 42 / 44

 Integration Rules Based on Non-uniform Mesh Spacing

Exercise
Evaluate the function Z 6 1
0 1 + x2
using
1 Simple trapezium rule
2 Composite trapezium rule with 6 sub-intervals
3 Simple Simpson’s 1/3 rule
4 Composite Simpson’s 1/3 rule with 8 sub-intervals
5 Romberg method with two sub-interval for three iterations. Compute your
initial values using the Simpson’s 1/3 method.
6 Romberg method with two sub-interval. Compute your initial values using the
Trapezium method. Take ϵ = 1−5
Dr. Gabby (KNUST-Maths) Numerical Integration 43 / 44
 END OF LECTURE
THANK YOU

Dr. Gabby (KNUST-Maths) Numerical Integration 44 / 44