NUMERICAL DIFFERENTIATION AND INTEGRATION

Lecture series on “ Numerical Techniques and Programming in

MATLAB”

VBS Purvanchal University, Jaunpur

July 22-28, 2016

by

Dr. Sarvesh Kumar,

IIST Trivandrum

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 1
 Numerical Differentiation

Basic problems

• Derive a formula that approximates the derivative of a function in

terms of linear combination of function values ( Function may be

known )

• Approximate the value of a derivative of a function defined by discrete

data.

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 2
 Solution Approaches ..

• Use Taylor Series Expansion.

• Pass a polynomial through the given data and differentiate the

interpolating polynomial.

Applications

To solve Ordinary and Partial Differential Equations.

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 3
 First Derivative..

Let f : [a, b] −→ R, then the derivative of f is another function say

0
f : [a, b] −→ R and defined by

f (c + h) − f (c)
f 0(c) = lim , ∀c ∈ (a, b).
h→0 h

Geometrically Speaking f 0(c) is the slope of tangent to the curve f (x)

at x = c.

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 4
 Taylor Series

Derivative of a function at x = x0

Suppose f has two continuous derivatives. Then, by Taylor’s Theorem

h2 00 0
f (x0 + h) = f (x0) + hf (x0) + f (θ)
2

where θ ∈ (x0, x0 + h). Now,

f (x0 + h) − f (x0)
f 0(x0) ≈
h

Called Forward Formula

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 5
 Error

Error = |true value − approximate value|

h ”
|ED (f )| ≤ max |f (θ)|
θ∈[a,b] 2

Example

Example1. Using Taylor series find the derivative of f (x) = x2 at x=1,

with h = .1. Also compute the error.

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 6
 Other formulae

Backward Formula

f (x0) − f (x0 − h)
f 0(x0) ≈
h

Central Formula

f (x0 + h) − f (x0 − h)
f 0(x0) ≈
2h

Similarly, we can drive (Second Derivative )

00 f (x0 + h) − 2f (x0) + f (x0 − h)

f (x0) ≈
h2

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 7
 Derivative for discrete data using interpolating

polynomial

x x0 x1 x2 ... xn

f (x) f (x0) f (x1) f (x3) ... f (xn)

Assumption: x0, x1 · · · xn are equispaced i.e., xi − xi−1 = h. Where

the explicit nature of the function f is not known.

Remark 1: We can used Newton’s Forward or Backward formula

depends the location of the point

Remark 2: If data is not equispaced then Lagrange interpolating

polynomial can be used.

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 8
Using Newton’s forward difference formula

s(s − 1) 2 s(s − 1)(s − 2) 3

f (x) ≈ Pn(x) = f (x0)+s∆f (x0)+ ∆ f (x0)+ ∆ f (x0)
2! 3!
s(s − 1)(s − 2) · · · (s − n + 1) n
··· ∆ f (x0)
n!

where

x = x0 + sh

We use pn(x) to calculate the derivatives of f.

That is f 0(x) ' p0n(x) for all x ∈ [x0, xn].

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 9
For a given x,

f 0(x) ' p0n(x)

dpn ds
=
ds dx
1 dpn
=
h ds

Similarly,
2
00 d pn
f (x) '
dx2
d dpn
( )
dx dx
d dpn ds
= ( )
dx ds dx
1 d dpn
= ( )
h dx ds
Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 10
 1 d2 p n 1
= ( 2 )
h ds h
1 d2 p n
= 2 2
h ds

Thus in general,
k
(k) 1 d pn
f (x) = k k
h ds

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 11
 Example 1

Using Taylor series expansion (forward formula) and Newton forward

divided difference, compute first and second derivative at x = 2 for the

following tabulated function

x 1 2 3 4

f (x) 2 5 7 10

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 12
 Solution

x f (x) ∆ ∆2 ∆3

1 2

2 5 -1

2 2

3 7 1

4 10

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 13
Here h = 1 Using Taylor series

f (2 + h) − f (2) f (3) − f (2)

f 0(2) = = =2
h 1

00 f (2 + h) − 2f (2) + f (2 − h) f (3) − 2f (2) + f (1)

f (2) = 2
= = −1
h 1

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 14
Using Newton forward divided difference formula

s(s − 1) 2 s(s − 1)(s − 2) 3

f (x) ≈ Pn(x) = f (x0)+s∆f (x0)+ ∆ f (x0)+ ∆ f (x0)
2! 3!
2
 
1 dpn 1 2s − 1 2 3s − 6s + 2 3
f 0(x) ≈ = ∆f (x0) + ∆ f (x0) + ∆ f (x0)
h ds h 2! 3!
Here x = 2, x0 = 1 , s = 1 and h = 1

0 1 1
f (2) = 3 − − = 13/6
2 3

00 1 d2 p n 1  2 3

f (x) ≈ 2 2 = 2 ∆ f (x0) + (s − 1)∆ f (x0)
h ds h
f 00(2) = −1

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 15
 Example 2

Calculate f (4)(0.15)

x 0.1 0.2 0.3 0.4 0.5 0.6

f (X) 0.425 0.475 0.400 0.450 0.525 0.575

Solution:

Newton’s forward difference formula:

2 3 2
1 s − s 2 s − 3s + 2s 3
p5(x) = f (x0) + s4 f (x0) + 4 f (x0) + 4 f (x0)+
2 6

s4 − 6s3 + 11s2 − 6s 4 s5 − 10s4 + 35s3 − 50s2 + 24s 5

+ 4 f (x0) + 4 f (x0)
24 120
Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 16
Differentiating this 4-times we get,

d4f dp45 1 4 1 5
' = [4 f (x 0 ) + (5s − 10)4 f (x0)
dx4 dx4 h4 5

1 4
= 4 [4 f (x0) + (s − 2)45f (x0)]
h

1 2
= [−035 + (0.5 − 2)(0.4)] = −95.00 × 10
(0.1)4

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 17
 x f (x) 41 f 42 f 43 f 44 f 45 f

0.1 0.425

0.050

0.2 0.475 -0.125

-0.075 0.25

0.3 0.400 0.125 -0.35

‘ 0.050 -0.100 0.4

0.4 0.450 0.025 0.05

0.075 -0.05

0.5 0.525 -0.025

0.050

0.6 0.575
Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 18
 Numerical Integration

If f : [a, b] −→ R is differentiable then, we obtain a new function f 0 :

[a, b] −→ R, called the derivative of f . Likewise, if f : [a, b] −→ R is

integrable, then we obtain a new function F : [a, b] −→ R defined by

Z x
F (x) = f (t)dt ∀x ∈ [a, b].
a

Z b
Observation: If f is nonnegative function, then f (x)dx is represent
a
the area under the curve f (x).

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 19
 Antiderivative

Antiderivative: Let F : [a, b] −→ R be such that f = F 0, then F is called

an antiderivative of f .

Recall

Fundamental Theorem of Calculus: Let f : [a, b] −→ R is integrable

and has an antiderivative F , then

Z b
f (x)dx = F (b) − F (a).
a

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 20
 Basic Problems

• Difficult to find an antiderivative of the function (for example f (x) =

2
e−x )

• Function is given in the tabular form.

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 21
 Newton-Cotes Methods/Formulae

The derivation of Newton-Cotes formula is based on Polynomial

Interpolation.

x x0 x1 x2 ... xn

f (x) f (x0) f (x1) f (x3) ... f (xn)

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 22
 The idea is:

Rb
Replace f by pn(x) and evaluate a
pn(x)dx

That is,
Z b Z b n
Z bX
f (x)dx ' pn(x)dx = li(x)f (xi)dx
a a a i=0
n
X Z b
= f (xi) li(x)dx
i=0 a

Xn
= Aif (xi)
i=0
Rb
Where Ai = l (x)dx
a i
called weights.

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 23
 Types of Newton-Cotes Formulae

• Trapezoidal Rule (Two pint formula)

• Simpson’s 1/3 Rule (Three Point formula)

• Simpson’s 3/8 Rule (Four point formula)

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 24
 Trapezoidal Rule

Since it is two point formula, it uses the first order interpolation

polynomial P1(x).

Z b Z x1
f (x) ≈ P1(x)dx
a x0

P1(x) = f (x0) + s∆f (x0)

x − x0
s=
h
Now, dx = h ds at x = x0, s = 0 and at x = x1, s = 1.

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 25
Hence,

Z b Z 1
h
f (x)dx ≈ (f (x0) + s∆f (x0))hds = [f (x0) + f (x1)]
a 0 2

OR
b
b−a
Z
f (x)dx ≈ [f (a) + f (b)]
a 2
Error
T (b − a)3 00
E =− f (ξ),
12
where a < ξ < b

Remark: x0 = a and x1 = b.

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 26
 1
Basic Simpson’s 3 Rule

Rb
To evaluate a
f (x)dx.

• f will be replaced by a polynomial of degree 2 which interpolates f at

a+b a+b
a, 2 and b. . Here, x0 = a, x1 = 2 , x2 =b

b
b−a
Z
b+a
f (x) dx = [f (a) + 4f ( ) + f (b).]
a 6 2

Error
s−h5f (4)(ξ)
E =
90
for some ξ ∈ (a, b).

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 27
 3
Basic Simpson’s 8 Rule

• f is replaced by p3(x) which interpolates f at x0 = a, x1 = a +

b−a
h, x2 = a + 2h, x3 = a + 3h = b. where h = 3 . Thus we get:

Z b
3h
f (x)dx ' [f0 + 3f1 + 3f2 + f3]
a 8

−3h5 (4)
s
Error: E = f (ξ), where a < ξ < b.
80

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 28
 Example

1
R2 4
R2
Using Trapezoidal and Simpson 3 rules find 0
x dx and 0
sinxdx and

find the upper bound for the error.

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 29
 Composite Rules

Note that if the integral [a, b] is large, then the error in the Trapezoidal

rule will be large.

Idea

Error can be reduced by dividing the interval [a, b] into equal subinterval

and apply quadrature rules in each subinterval.

Composite Trapezoidal Rule

b−a
h= , xi = x0 + ih
n
Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 30
Composite Rule

Z b Z xn n Z
X xi
f (x)dx = f (x)dx = f (x)dx
a x0 i=1 xi−1

Now apply Trapezoidal rule on each [xi−1, xi], we have

Z b
h
f (x)dx = [f (x0) + 2 ∗ (f (x1) + f (x2) · · · f (xn−1)) + f (xn)]
a 2

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 31
 Error in composite Trapezoidal rule

CT h2 00
E = −(b − a) f (ξ), ξ ∈ [a, b]
12

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 32
 1
The Composite Simpson’s 3 Rule

• [a, b] will be will be divided into 2n equal subintervals and we apply

1
basic Simpson’s 3 rule on each of the n intervals [x2i−2, x2i] for i =

1, 2, 3, · · · , n.

b−a
Thus here h = 2n .

Then
Z b Z b=x2n
f (x)dx = f (x)dx
a a=x0
Z x2 Z x4 Z x2i Z x2n
= f (x)dx + f (x)dx + · · · + f (x)dx + · · · + f (x)dx
x0 x2 x2i−2 x2n−2

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 33
 h h
= [f (x0) + 4f (x1) + f (x2)] + [f (x2) + 4f (x3) + f (x4)]+
3 3
h
+ · · · + [f (x2n−2) + 4f (x2n−1) + f (x2n)]
3
h
= {f (x0) + 4 × [f (x1) + f (x3) + f (x5) + · · · + f (x2n−1)]+
3
+2 × [f (x2) + f (x4) + f (x6) + · · · + f (x2n−2)] + f (x2n)}

CS h4 (4)
E = −(b − a) f (ξ),
180
where ξ ∈ [a, b]

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 34
 Example

R1 1
Evaluate the integral −1
x2 exp(−x)dx by composite Simpson’s 3 rule

with spacing h = 0.25

1
Solution: According to composite Simpson’s 3 rule:

Z 1
h
x2 exp(−x)dx = [f (x0) + 4f (x1) + 2f (x2) + 4f (x3) + 2f (x4)+
−1 3

+4f (x5) + 2f (x6) + 4f (x7) + f (x8)]

Here f (x0) = f (−1) = 2.7183

f (x1) = f (−0.75) = 1.1908

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 35
f (x2) = f (−0.5) = 0.4122

f (x3) = f (−0.25) = 0.0803

f (x4) = f (0) = 0

f (x5) = f (0.25) = 0.0487

f (x6) = f (0.50) = 0.1516

f (x7) = f (0.75) = 0.2657

f (x8) = f (1) = 0.3679

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 36
Substituting these values in the above formula we get:

Z 1
x2 exp(−x)dx ' 0.87965
−1

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 37
 Example

Find the minimum no. of subintervals, used in composite Trapezoidal

Z 1
−x4
and Simpson’s 1/3 rule in order to find the integral e dx such that
0
the error can not exceed by .00001.

Sol. For the composite Trapezoidal rule, we have

00
13 max0<ξ<1 |f (ξ)|
≤ .00001
12n2trap
For the composite Simpson 1/3 rule, we have

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 38
 14 max0<ξ<1 |f (4)(ξ)|
≤ .00001
180n4simp
Now,

00
max |f (ξ)| ≤ 3.5, max |f (4)(ξ)| ≤ 95
0<ξ<1 0<ξ<1

(Please verify )

Hence

ntrap = 171, nsimp = 16

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 39
 3
Composite Simpson’s 8 rule

b−a 3
• [a, b] is divided into 3n equal subintervals. (h = 3n . and we apply 8

rule on each of the n intervals [x3i−3, x3i] for i = 1, 2, 3, · · · , n.)

Hence,

Z b Z x3 Z x6 Z x3n =b
f (x)dx ' f (x)dx + f (x)dx + · · · + f (x)dx
a x0 =a x3 x3n3

3h 3h
= [f0 + 3f1 + 3f2 + f3] + [f3 + 3f4 + 3f5 + f6]+
8 8
3h
+··· + [f3n−3 + 3f3n−2 + 3f3n−1 + f3n]
8
Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 40
 3h
= [f0 + 3f1 + 3f2 + 2f3 + 3f4 + 3f5 + 2f6 + 3f7 + · · · + 3f3n−1 + f3n]
8
Remember:

• f with suffices of multiple 3 are multiplied by 2.

• Others by 3, except the end points.

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 41
 Example

3
Use composite simpson’s 8 rule, find the velocity after 18 seconds, if a

rocket has acceleration as given in the table:

t= 0 2 4 6 8 10 12 14 16 18

a= 40 60 70 75 80 83 85 87 88 88

f0 f1 f2 f3 f4 f5 f6 f7 f8 f9

3h
Sol: Velocity v = 8 [f0 +3f1 +3f2 +2f3 +3f4 +3f5 +2f6 +3f7 +3f8 +f9 ]=
3
4 [40+3×60+3×70+2×75+3×80+383+2×83+2×85+3×87+3×88+88]

= 1389 units.

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 42
 Method of Undetermined Parameters

The Newton - Cotes integration rules are all of the form

I(f ) ' A0f (x0) + A1f (x1) + A2f (x2) + · · · + Anf (xn)

Also, note that the weights Ai’s do not depend on the given function.

Hence, if the error is of the form

E(I) = Const × f (r+1)(η).

Then the rule must be exact for all polynomials of degree ≤ r

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 43
 Therefore

If we wish to construct a rule of the form

I(f ) ' A0f (x0) + A1f (x1) + a2f (x2) + · · · + Anf (xn)

(n-fixed) which is exact for polynomials of degree as high as possible,

i.e., we want

E(I) = Const × f (r+1)(η),

with r as large as possible.

This way of constructing integration rules is called the ” Method of

Undetermined Parameters”.

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 44
 Example

Suppose we want to derive an integration formula of the form:

Rb
a
f (x)dx = A0 f (a) + A1 f (b) + α f 00(ξ).

We assume that:The rule is exact for the polynomials 1, x, x2.

Now, taking f (x) = 1, we get b − a = A0 + A1

b2 −a2
Taking f (x) = x we get 2 = A0 a + A1 b

b−a
Solving the above two equations we get, A0 = A1 = 2 .
Z b
b−a
Thus, f (x)dx = [f (a) + f (b)] + α f 00(ξ)
a 2

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 45
Now if we take f (x) = x2, we get:

b3 − a3 b−a 2
=( )(a + b2) + 2! α
3 2

(b−a)3
=⇒ α = − 12

Thus
b
b−a (b − a)3 00
Z
f (x)dx = [f (a) + f (b)] − f (ξ)
a 2 12

We see that: This is exactly the trapezoidal rule. Similarly, Simpson’s

1 3
3 and 8 rules can be derived.

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 46
 Thus in the Method of Undetermined Parameters

• We aim directly for a formula of a preselected type.

Working Method:

• We impose certain conditions on a formula of desired form and use

these conditions to determine the values of the unknown coefficients in

the formula.

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 47
 The Error term in the Simpson’s 38 -rule, using Method

of Undetermined Parameters

Start with:

Z x3
3h
f (x)dx = [f0 + 3f1 + 3f2 + f3] + α f (4)(ξ)
x0 8

for some suitable ξ ∈ (x0, x3).

Takeing f (x) = x4 in the above integration rule we get:

x53 − x50 3h 4
= [x0 + 3x41 + 3x42 + x43] + α 4!
5 8
Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 48
 x53 − x50 3h 4
4!α = − [x0] + 3(x0 + h)4 + 3(x0 + 2h)4 + (x0 + 3h)4
5 8

(x0 + 3h)5 − x50 3h 4

= − [x0] + 3(x0 + h)4 + 3(x0 + 2h)4 + (x0 + 3h)4
5 8

Without loss of generality, we can take: x0 = 0.

243 5 3h5
We have: 4!α = 5 h − 8 [0 + 3 + 3 × 16 + 81] Thus

9 5
4!α = − h
10

That is,

3 5
α=− h
80
Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 49
Therefore the error in the Simpson’s rule is =

3 5 (4)
− h f (ξ)
80

for some suitable ξ ∈ (a, b).

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 50
 Recall

The Newton - Cotes integration rules are all of the form

I(f ) ' A0f (x0) + A1f (x1) + A2f (x2) + · · · + Anf (xn)

Also, note that the weights Ai’s do not depend on the given function.

Hence, if the error is of the form E(I) = Const × f (r+1)(η). Then the

rule must be exact for all polynomials of degree ≤ r.

Remark: In these quadrature the points xi are fixed.

Ques: Can we improve the accuracy by choosing some suitable xi

Ans: Using Gaussian Quadrature rule one can improve the accuracy.

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 51
 Example

Find x0, x1, A0, A1 and α so that the following rule is exact for all

polynomials of degree≤ 3.

Z 1
f (x)dx = A0f (x0) + A1f (x1) + αf (4)(ξ)
−1

(There are 4 unknowns and hence we have chosen the 4-th derivative

in the error term.)

Taking f (x) = 1, x, x2, x3 we get:

A0 + A1 = 2

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 52
A0 x 0 + A1 x 1 = 0

2
A0x20 + A1x21 = 3

A0x30 + A1x31 = 0

On solving these equations we get:

A0 = A1 = 1 x0 = − √13 and x1 = √1 .
3

R1
Thus the integration rule is: −1
f (x)dx = f (− √13 ) + f ( √13 ) + αf (4)(ξ).

Now if we take f (x) = x4 we get

2 2
= + α4!
5 9
Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 53
 1 8 1
=⇒ α = ( )=
4! 45 135

Thus the expected integration rule is:

Z 1
1 1 1 (4)
f (x)dx = f (− √ ) + f ( √ ) + f (ξ).
−1 3 3 135

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 54
 In general

Giving a positive integer n, we wish to determine 2n + 2 numbers

x0, x1, · · · xn and A0, A1, · · · , An so that the sum

I(f ) ' A0f (x0) + A1f (x1) + A2f (x2) + · · · + Anf (xn),

Z b
provides the exact value of f (x)dx for f (x) = 1, x, x2, · · · x2n+1.
a
Or What we want is that the quadrature rule is exact for all polynomials

of degree ≤ 2n + 1.

Remark: Here we have to solve system of nonlinear equations, which is

some time is not an easy job.

Lecture series on “Numerical techniques and MATLAB”, July 22-28, 2016 Dr. Sarvesh Kumar, IIST Trivandrum. 55