MATH2033 (2023–2024)

Introduction to Scientific Computation

Numerical Calculus IV

Department of Mathematical Sciences

Numerical Calculus IV 2/30

Composite Numerical Integration

For appropriate functions f and intervals [a, b], we can approximate the
integral
b
∫ f (x) dx
a
by dividing the interval [a, b] into n subintervals [xi−1 , xi ] so that
b n xi
∫ f (x) dx = ∑ ∫ f (x) dx
a i=1 xi−1

and using a numerical integration rule to approximate

xi
∫ f (x) dx
xi−1

for all positive integers i such that i ≤ n.

Numerical Calculus IV 3/30

Composite Trapezoidal Rule

Suppose that n is a positive integer, that a ∈ R, that b ∈ R and that a < b.
b−a
Let h = and let xi = a + ih for all nonnegative integers i such that
n
i ≤ n.
For appropriate functions f , by the trapezoidal rule,
b n xi
∫ f (x) dx = ∑ ∫ f (x) dx
a i=1 xi−1
n
h
≈ ∑ (f (xi−1 ) + f (xi ))
i=1 2
h n
= ∑(f (xi−1 ) + f (xi ))
2 i=1
h
= (f (x0 ) + 2f (x1 ) + 2f (x2 ) + ⋯ + 2f (xn−1 ) + f (xn ))
2
n−1
h
= (f (a) + 2 ∑ f (xi ) + f (b)) .
2 i=1
Numerical Calculus IV 4/30

We have arrived at the composite trapezoidal rule (also called composite

trapezium rule)

b n−1
h
∫ f (x) dx ≈ (f (a) + 2 ∑ f (xi ) + f (b))
a 2 i=1

which uses the values of f at n + 1 evenly spaced points (or uses n

subintervals of equal length or uses n strips of equal width).
Numerical Calculus IV 5/30

Composite Midpoint Rule

Suppose that n is a positive integer, that a ∈ R, that b ∈ R and that a < b
b−a
Let h = and let xi = a + 2ih for all nonnegative integers i such that
2n
i ≤ n.
For appropriate functions f , by the midpoint rule,
b n xi
∫ f (x) dx = ∑ ∫ f (x) dx
a i=1 xi−1
n
≈ ∑ 2hf (xi−1 + h)
i=1
n
= 2h ∑ f (xi−1 + h).
i=1
We have arrived at the composite midpoint rule
b n
∫ f (x) dx ≈ 2h ∑ f (xi−1 + h)
a i=1
which uses the values of f at n evenly spaced points.
Numerical Calculus IV 6/30

Composite Simpson’s Rule

Suppose that n is a positive integer, that a ∈ R, that b ∈ R and that a < b
b−a
Let h = and let xi = a + 2ih for all nonnegative integers i such that
2n
i≤n
Numerical Calculus IV 7/30

For appropriate functions f , by Simpson’s rule,

b n xi
∫ f (x) dx = ∑ ∫ f (x) dx
a i=1 xi−1
n
h
≈ ∑ (f (xi−1 ) + 4f (xi−1 + h) + f (xi ))
i=1 3
h n
= ∑(f (xi−1 ) + 4f (xi−1 + h) + f (xi ))
3 i=1
h
= (f (x0 ) + 4f (x0 + h) + 2f (x1 ) + 4f (x1 + h) + 2f (x2 )
3
+4f (x2 + h) + ⋯ + 2f (xn−1 ) + 4f (xn−1 + h) + f (xn ))
n−1 n−1
h
= (f (a) + 4 ∑ f (xi + h) + 2 ∑ f (xi ) + f (b)) .
3 i=0 i=1
Numerical Calculus IV 8/30

We have arrived at the composite Simpson’s rule

b n−1 n−1
h
∫ f (x) dx ≈ (f (a) + 4 ∑ f (xi + h) + 2 ∑ f (xi ) + f (b))
a 3 i=0 i=1

which uses the values of f at 2n + 1 evenly spaced points.

Numerical Calculus IV 9/30

Problem 1
Let f be given by f (x) = sin(πx).
1
Approximate ∫ f (x) dx using the composite trapezoidal rule with
0
f being evaluated at 5 evenly spaced points.
1
Approximate ∫ f (x) dx using the composite midpoint rule with f
0
being evaluated at 5 evenly spaced points.
1
Approximate ∫ f (x) dx using the composite Simpson’s rule with f
0
being evaluated at 5 evenly spaced points.
Numerical Calculus IV 10/30

Using the composite trapezoidal rule with f being evaluated at 5 evenly

spaced points yields, where h = 0.25,
1
∫ sin(πx) dx
0
h
≈ (f (0) + 2(f (0.25) + f (0.5) + f (0.75)) + f (1))
2
= 0.125(sin(0) + 2(sin(0.25π) + sin(0.5π) + sin(0.75π)) + sin(π))
= 0.60355 . . . .
Using the composite midpoint rule with f being evaluated at 5 evenly
spaced points yields, where h = 0.1,
1
∫ sin(πx) dx
0
≈ 2h(f (0.1) + f (0.3) + f (0.5) + f (0.7) + f (0.9))
= 0.2(sin(0.1π) + sin(0.3π) + sin(0.5π) + sin(0.7π) + sin(0.9π))
= 0.64721 . . . .
Numerical Calculus IV 11/30

Using the composite Simpson’s rule with f being evaluated at 5 evenly

spaced points yields, where h = 0.25,
1
∫ sin(πx) dx
0
h
≈ (f (0) + 4(f (0.25) + f (0.75)) + 2f (0.5) + f (1))
3
1
= (sin(0) + 4(sin(0.25π) + sin(0.75π)) + 2 sin(0.5π) + sin(π))
12
= 0.63807 . . . .

Note that
1 1
1 1 1 2
∫ sin(πx) dx = [− cos(πx)] = − cos(π)+ cos(0) = = 0.63661 . . . .
0 π 0 π π π
Numerical Calculus IV 12/30

The error identity for the composite trapezoidal rule

Theorem
Suppose that a ∈ R, that b ∈ R, that a < b, that f ∈ C 2 [a, b] and that n is
b−a
a positive integer. Let h = and let xi = a + ih for all nonnegative
n
integers i such that i ≤ n. Then there exists ξ ∈ (a, b) for which

b n−1
h b − a 2 ′′
∫ f (x) dx = (f (a) + 2 ∑ f (xi ) + f (b)) − h f (ξ).
a 2 i=1 12
Numerical Calculus IV 13/30

Proof By the error identity for the trapezoidal rule, for all positive
integers i such that i ≤ n, there exists ξi ∈ (xi−1 , xi ) for which
b n xi
∫ f (x) dx = ∑ ∫ f (x) dx
a i=1 xi−1
n
h3
= ∑ ( (f (xi−1 ) + f (xi )) − f ′′ (ξi ))
h
i=1 2 12
h n h3 n ′′
= ∑(f (xi−1 ) + f (xi )) − ∑ f (ξi )
2 i=1 12 i=1
n−1
h nh 2 1 n ′′
= (f (a) + 2 ∑ f (xi ) + f (b)) − h ∑ f (ξi )
2 i=1 12 n i=1
n−1
h b − a 2 1 n ′′
= (f (a) + 2 ∑ f (xi ) + f (b)) − h ∑ f (ξi )
2 i=1 12 n i=1

because nh = b − a.
Numerical Calculus IV 14/30

Now, for all positive integers i such that i ≤ n,

min f ′′ (x) ≤ f ′′ (ξi ) ≤ max f ′′ (x).

x∈[a,b] x∈[a,b]

Hence,
n
n min f ′′ (x) ≤ ∑ f ′′ (ξi ) ≤ n max f ′′ (x)
x∈[a,b] i=1 x∈[a,b]

and
1 n ′′
min f ′′ (x) ≤ ′′
∑ f (ξi ) ≤ max f (x).
x∈[a,b] n i=1 x∈[a,b]
Numerical Calculus IV 15/30

If
1 n ′′ ′′
∑ f (ξi ) = min f (x)
n i=1 x∈[a,b]

then
f ′′ (ξi ) = min f ′′ (x)
x∈[a,b]

for all positive integers i such that i ≤ n and, for all positive integers j
such that j ≤ n,

1 n ′′ 1 n ′′ ′′
∑ f (ξi ) = ∑ f (ξj ) = f (ξj ).
n i=1 n i=1
Numerical Calculus IV 16/30

If
1 n ′′ ′′
∑ f (ξi ) = max f (x)
n i=1 x∈[a,b]

then
f ′′ (ξi ) = max f ′′ (x)
x∈[a,b]

for all positive integers i such that i ≤ n and, for all positive integers j
such that j ≤ n,

1 n ′′ 1 n ′′ ′′
∑ f (ξi ) = ∑ f (ξj ) = f (ξj ).
n i=1 n i=1
Numerical Calculus IV 17/30

If
1 n ′′
min f ′′ (x) ≠ ′′
∑ f (ξi ) ≠ max f (x)
x∈[a,b] n i=1 x∈[a,b]

then
1 n ′′
min f ′′ (x) < ′′
∑ f (ξi ) < max f (x)
x∈[a,b] n i=1 x∈[a,b]

and so, since f ∈ C 2 [a, b], by the Intermediate Value Theorem there
exists c ∈ (min{c1 , c2 }, max{c1 , c2 }) for which

1 n ′′
f ′′ (c) = ∑ f (ξi )
n i=1

where c1 is such that c1 ∈ [a, b] and f ′′ (c1 ) = min f ′′ (x) and c2 is such
x∈[a,b]
that c2 ∈ [a, b] and f ′′ (c2 ) = max f ′′ (x).
x∈[a,b]
Numerical Calculus IV 18/30

Therefore, there exists ξ ∈ (a, b) for which

1 n ′′
f ′′ (ξ) = ∑ f (ξi )
n i=1

and
b n−1
h b − a 2 ′′
∫ f (x) dx = (f (a) + 2 ∑ f (xi ) + f (b)) − h f (ξ).
a 2 i=1 12
Numerical Calculus IV 19/30

The error identity for the composite midpoint rule

Theorem
Suppose that a ∈ R, that b ∈ R, that a < b, that f ∈ C 2 [a, b] and that n is
b−a
a positive integer. Let h = and let xi = a + 2ih for all nonnegative
2n
integers i such that i ≤ n. Then there exists ξ ∈ (a, b) for which
b n
b − a 2 ′′
∫ f (x) dx = 2h ∑ f (xi−1 + h) + h f (ξ).
a i=1 6
Numerical Calculus IV 20/30

The error identity for the composite Simpson’s rule

Theorem
Suppose that a ∈ R, that b ∈ R, that a < b, that f ∈ C 4 [a, b] and that n is
b−a
a positive integer. Let h = and let xi = a + 2ih for all nonnegative
2n
integers i such that i ≤ n. Then there exists ξ ∈ (a, b) for which

b n−1 n−1
h
∫ f (x) dx = (f (a) + 4 ∑ f (xi + h) + 2 ∑ f (xi ) + f (b))
a 3 i=0 i=1
b − a 4 (4)
− h f (ξ).
180
Numerical Calculus IV 21/30

Problem 2
1
Suppose that n is a positive integer. Let h = and let xi = ih for all
n
nonnegative integers i such that i ≤ n. Let f be given by f (x) = sin(πx).
Given that there exists ξ ∈ (0, 1) for which

1 n−1
h2
(f (0) + 2 ∑ f (xi ) + f (1)) − f ′′ (ξ),
h
∫ f (x) dx =
0 2 i=1 12

determine an integer m which is such that taking n ≥ m will guarantee

that
1 n−1
h
∣∫ f (x) dx − (f (0) + 2 ∑ f (xi ) + f (1))∣ < 0.0001.
0 2 i=1
Numerical Calculus IV 22/30

We have that
1 n−1
h
∣∫ f (x) dx − (f (0) + 2 ∑ f (xi ) + f (1))∣
0 2 i=1
1 2 ′′
= ∣− h f (ξ)∣ for some ξ ∈ (0, 1)
12
h2
≤ max ∣f ′′ (x)∣
12 x∈[0,1]
1
= max ∣−π 2 sin(πx)∣
12n2 x∈[0,1]
π2
= .
12n2
Numerical Calculus IV 23/30

Now, since n is positive,

π2 2 π2 π
2
< 0.0001 ⇔ n > ⇔n> √ .
12n 0.0012 0.0012
π
Moreover, √ = 90.6899 . . . and so taking n ≥ m with m = 91
0.0012
guarantees that

1 n−1
h
∣∫ f (x) dx − (f (0) + 2 ∑ f (xi ) + f (1))∣ < 0.0001.
0 2 i=1
Numerical Calculus IV 24/30

Problem 3
1
Suppose that n is a positive integer. Let h = and let xi = 2ih for all
2n
nonnegative integers i such that i ≤ n. Let f be given by f (x) = sin(πx).
Given that there exists ξ ∈ (0, 1) for which
1 n
h2 ′′
∫ f (x) dx = 2h ∑ f (xi−1 + h) + f (ξ),
0 i=1 6

determine an integer m which is such that taking n ≥ m will guarantee

that
1 n
∣∫ f (x) dx − 2h ∑ f (xi−1 + h)∣ < 0.0001.
0 i=1
Numerical Calculus IV 25/30

We have that
1 n
h2 ′′
∣∫ f (x) dx − 2h ∑ f (xi−1 + h)∣ = ∣ f (ξ)∣ for some ξ ∈ (0, 1)
0 i=1 6
h2
≤ max ∣f ′′ (x)∣
6 x∈[0,1]
1
= max ∣−π 2 sin(πx)∣
24n2 x∈[0,1]
π2
= .
24n2
Numerical Calculus IV 26/30

Now, since n is positive,

π2 2 π2 π
2
< 0.0001 ⇔ n > ⇔n> √ .
24n 0.0024 0.0024
π
Moreover, √ = 64.1274 . . . and so taking n ≥ m with m = 65
0.0024
guarantees that
1 n
∣∫ f (x) dx − 2h ∑ f (xi−1 + h)∣ < 0.0001.
0 i=1
Numerical Calculus IV 27/30

Problem 4
1
Suppose that n is a positive integer. Let h = and let xi = 2ih for all
2n
nonnegative integers i such that i ≤ n. Let f be given by f (x) = sin(πx).
Given that there exists ξ ∈ (0, 1) for which

1 n−1 n−1
h
∫ f (x) dx = (f (0) + 4 ∑ f (xi + h) + 2 ∑ f (xi ) + f (1))
0 3 i=0 i=1
h4 (4)
− f (ξ),
180
determine an integer m which is such that taking n ≥ m will guarantee
that
1 n−1 n−1
h
∣∫ f (x) dx − (f (0) + 4 ∑ f (xi + h) + 2 ∑ f (xi ) + f (1))∣ < 0.0001.
0 3 i=0 i=1
Numerical Calculus IV 28/30

We have that
1 n−1 n−1
h
∣∫ f (x) dx − (f (0) + 4 ∑ f (xi + h) + 2 ∑ f (xi ) + f (1))∣
0 3 i=0 i=1
h4 (4)
= ∣− f (ξ)∣ for some ξ ∈ (0, 1)
180
h4
≤ max ∣f (4) (x)∣
180 x∈[0,1]
1
= max ∣π 4 sin(πx)∣
2880n4 x∈[0,1]
π4
= .
2880n4
Numerical Calculus IV 29/30

Now, since n is positive,

π4 4 π4 π
4
< 0.0001 ⇔ n > ⇔n> .
2880n 0.288 0.2881/4
π
Moreover, = 4.2884 . . . and so taking n ≥ m with m = 5
0.2881/4
guarantees that

1 n−1 n−1
h
∣∫ f (x) dx − (f (0) + 4 ∑ f (xi + h) + 2 ∑ f (xi ) + f (1))∣ < 0.0001.
0 3 i=0 i=1
Numerical Calculus IV 30/30

References
Burden, Faires & Burden, Numerical Analysis, 10E
Section 4.4