M.Sc.

MATHEMATICS

FIRST YEAR -SEMESTER-II

NUMERICAL METHODS

https://mkuniversity.ac.in/dde/

PMATE23

All copy right privileges are reserved

 M.Sc. First YearMathematics

NUMERICAL METHODS

WELCOME

Dear Students

We welcome you as a student of the first year M.Sc., Mathematics

degreeprogramme.

Thispaper deal withthesubjectNUMERICAL METHODS. The

learning material for this paper will be supplemented by contact lectures.
In this book, we are going to learn about Bisection method,Direct methods -
Error analysis for direct methods, Numerical Differentiation , Numerical
integration and Difference equations.

Learning through the Distance Educationmode, as you are all

aware, involves self-learning and self- assessment and in this regard, you
are expected to part in disciplined anddedicatedeffort. Asour part, we
assure of ourguidance and support.
With best regards

DEPARTMENT OF MATHEMATICS-DDE, MKU

 SYLLABUS

NUMERICAL METHODS

Unit I: Introduction- Bisection method - iteration methods based on first degree

equation - iteration methods based on second degree equation -methods for
complex roots - polynomial equations.

Unit II: Introduction - Direct methods - Error analysis for direct methods -
iteration methods - Eigen values and Eigen vectors - Bounds on Eigen values.

Unit III: Introduction - Lagrange and Newton Interpolations - Finite difference

operators - Interpolating polynomials using finite differences - Hermite
interpolations - piecewise and spline interpolation.

Unit IV: Introduction - Numerical Differentiation - Extrapolation methods - partial

Differentiation - Numerical integration - methods based on interpolation -
Composite integration methods - Romberg method.

Unit V: Introduction - Difference equations - Numerical methods - Single step

methods.

Text Book :

M.K.Jain, S.R.K.Iyengar and R.K.Jain; Numerical Methods for Scientific and

Engineering Computation, New Age International Publishers, Fourth Edition, 2013

Reference Book :

C.F.Gerald and P.O.Wheatly; Applied Numerical Analysis, Addison Wesley, Fifth

Edition,1998.
_________________________________________________________________
Lesson Compiled By:
Dr. B. Meera Devi, Assistant Professor,
PG and Research Department of Mathematics,
Sri S. Ramasamy Naidu Memorial College, Sattur.
 Unit – 1 SPACE OF HINTS

1.1: Introduction

Transcendental and Polynomial Equation:

A transcendental number is a real (or) complex number which is not
algebraic.

(i.e.) A transcendental number is not a solution of non-zero polynomial

equation.

A real and complex numbers are transcendental number because

the set of all real complex numbers are uncountable. Alsothe algebraic
numbers are countable.

Transcendental function:
A transcendental function is a function which does not satisfy the
polynomial equation whose coefficients are themselves as a polynomial.
The logarithmic and exponential functions are examples of
transcendental function.

The transcendental function is a term used to describe the

trigonometric functions.

(i.e.)sine, cosine, tangent, cot, sec, cosec.

A function which is transcendental is said to be non-algebraic.

Example:
Examples of algebraic function are rational function and square
root functions.

Definition:
The root of equation is of the form 𝑓 𝑥 = 0 ,

Then 𝑓 𝑥 = 𝑃𝑛 𝑥 = 𝑎0 𝑥 𝑛 + 𝑎1 𝑥 𝑛 −1 + ⋯ + 𝑎𝑛−1 𝑥 𝑛 + 𝑎𝑛 where

𝑎0 ≠ 0 is a polynomial of degree 𝑛 in 𝑥 (or) 𝑓(𝑥) may be a

transcendental function.

1
 Definition:
SPACE FOR HINTS A number 𝜀 is a root of solution of 𝑓 𝑥 = 0 if 𝑓 𝜀 = 0 such a
solution 𝜀 is a root (or) a zero of 𝑓 𝑥 = 0.

Definition:
𝑚
If 𝑓 𝑥 = 𝑥 − 𝜉 𝑔 𝑥 = 0 where 𝑔(𝑥) is bounded and 𝑔(𝜉) ≠ 0
then 𝜉 is called a multiple root of multiplicity 𝑚.

In this case,

𝑓 𝜉 = 𝑓 ′ 𝜉 = ⋯ = 𝑓 𝑚 −1 𝜉 = 0

But 𝑓 𝑚 (𝜉) ≠ 0

For 𝑚 = 1, the number 𝜉 is said to be a simple root.

Direct Methods:
In this method, we find the exact value of the roots in a finite
number of steps.

Example1.1.1:
The root of the linear (or) first degree equation 𝑎0 𝑥 + 𝑎1 = 0
where 𝑎0 ≠ 0

⟹ 𝑎0 𝑥 = −𝑎1
𝑎1
⟹𝑥=− 𝑎0

Similarly, the roots of the quadratic equation (or) second degree

equation is 𝑎0 𝑥 2 + 𝑎1 𝑥 + 𝑎2 = 0 given by,

−𝑎1 ± 𝑎1 2 − 4𝑎0 𝑎2
𝑥=
2𝑎0

Iterative Methods:
These methods are based on the idea of successive approximations, i.e.,
Starting with one (or) more initial approximation to the root then we
obtain asequence of approximations (or) iterates {𝑥𝑘 } which in the limit
converges to the root.This method gives only one root at a time.

2
Definition:
SPACE OF HINTS
A sequence of iterates {𝑥𝑘 } is said to converge to the root 𝜉 if
lim𝑘→∞ 𝑥𝑘 − 𝜉 = 0

(or) lim𝑘 →∞ 𝑥𝑘 = 𝜉. If 𝑥𝑘 , 𝑥𝑘−1 , ⋯ , 𝑥𝑘−𝑚 +1 are 𝑚-approximations to

the root, then amultipoint iteration method is defined as 𝑥𝑘+1 = 𝜑(𝑥𝑘 ,
𝑥𝑘−1 , ⋯ , 𝑥𝑘−𝑚 +1 ) where the function𝜑 is called the multipoint
iteration function.

For, 𝑚 = 1, we get the onepoint iteration method 𝑥𝑘+1 = 𝜑(𝑥𝑘 )

Note:

If 𝜑(𝑥) is continuous in the interval [𝑎, 𝑏] that contains the root 𝜉

and 𝜑′ (𝑥) ≤ 𝑐 < 1, in the interval [𝑎, 𝑏]. Then 𝑥0 ∈ [𝑎, 𝑏] and the
sequence of iterates {𝑥𝑘 } is obtained from 𝑥𝑘+1 = 𝜑(𝑥𝑘 ) converges to
the root of 𝑓 𝑥 = 0.

Initial approximations:
The initial approximation of the root can be obtained by using
graphical method (or) by Intermediate Value theorem.

Theorem1.1.1:Intermediate Value Theorem

If 𝑓(𝑥) is a Continuous function on some interval [𝑎, 𝑏] and
𝑓 𝑎 . 𝑓 𝑏 < 0 then the equation 𝑓 𝑥 = 0 has atleast one real root or
an odd number of real roots in the interval 𝑎, 𝑏 .

Example1.1.2:

Obtain an initial approximation of equation 𝑓 𝑥 = 8𝑥 3 −

12𝑥 2 − 2𝑥 + 3 = 0

Solution:

𝑥 −2 −1 0 1 2 3

𝑓(𝑥) −105 −15 3 −3 15 105

3
 From the above table, we find that the equation 𝑓 𝑥 = 0 has roots in
SPACE FOR HINTS interval −1, 0 , (0, 1) and (1, 2). The exact roots are −0.5 , 0.5 , 1.5.

1.2: Bisection Method

If we know that a root of 𝑓(𝑥) = 0 lies in the interval 𝐼0 = (𝑎0 , 𝑏0 ), we
𝑎 0 +𝑏0
bisect 𝐼0 at the point 𝑚1 = . Denote by 𝐼1 the interval (𝑎0 , 𝑚1 )if
2

𝑓(𝑎0 ) 𝑓(𝑚1 ) < 0 or the interval (𝑚1 , 𝑏0 ) if𝑓(𝑚1 ) 𝑓(𝑏0 ) < 0.

Therefore, the interval 𝐼1 also contains the root.

We bisect the interval 𝐼1 and get a subinlerval 𝐼2 at whose end points

𝑓(𝑥) takes the values of opposite signs and therefore contains the root.
Continuing this procedure, we obtain a sequence of nested sets of
subintervals 𝐼0 ⊃ 𝐼1 ⊃ 𝐼2 … such that each subinterval contains the root.
After repeating the bisection process 𝑞 times, we either find the root or
𝑏0 −𝑎 0
find the interval 𝐼𝑞 of length which contains the root. We take the
2𝑞

midpoint of the last subinterval as the desired approximation to the root.

This root has error not greater than one half of the length of the interval
of which it is the midpoint. Thus, we have
1
𝑚𝑘+1 = 𝛼𝑘 + 𝑏 − 𝑎𝑘 , 𝑘 = 0,1,2, …
2 𝑘

(𝛼𝑘 , 𝑚𝑘+1 ), 𝑖𝑓𝑓 𝛼𝑘 𝑓 𝑚𝑘+1 < 0

Where 𝛼𝑘+1 , 𝑏𝑘+1 =
(𝑚𝑘−1 , 𝑏𝑘 ), 𝑖𝑓𝑓( 𝑚𝑘+1 ) 𝑓 𝑏𝑘 < 0

Example 1.2.1:
Perform five iterations of the bisection method to obtain the
smallest positive root of the equation 𝑓 𝑥 = 𝑥 3 − 5𝑥 + 1 = 0.

Solution:
Given 𝑓 𝑥 = 𝑥 3 − 5𝑥 + 1 = 0

Now, put 𝑥 = 0; 𝑓 0 = 1

𝑥 = 1; 𝑓 1 = 1 − 5 + 1 = −3

4
 ∴ The smallest positive root lies in the interval 0, 1 .
SPACE OF HINTS
Let 𝑎0 = 0 and 𝑏0 = 1
𝑎 0 +𝑏0 0+1
Then 𝑚1 = = = 0.5
2 2

𝑘 𝑎𝑘−1 𝑏𝑘−1 𝑚𝑘 𝑓(𝑎𝑘−1 ) 𝑓(𝑚𝑘 ) 𝑓 𝑚𝑘 𝑓(𝑎𝑘−1 )

𝑎𝑘−1 + 𝑏𝑘−1
=
2
1 0 1 0.5 +𝑣𝑒 −𝑣𝑒 <0
2 0 0.5 1.25 +𝑣𝑒 −𝑣𝑒 <0
3 0 0.25 0.125 +𝑣𝑒 +𝑣𝑒 >0
4 0.125 0.25 0.1875 +𝑣𝑒 +𝑣𝑒 >0
5 0.1875 0.25 0.21875 +𝑣𝑒 −𝑣𝑒 <0
Here the smallest positive root lies in the interval (0.1875,0.21875)

The approximate root of the equation is given by 0.203125.

Example1.2.2:
Perform five iterations of the bisection method to obtain a root
of the equation 𝑓 𝑥 = cos 𝑥 − 𝑥𝑒 𝑥 = 0.

Solution:

Given 𝑓 𝑥 = cos 𝑥 − 𝑥𝑒 𝑥 = 0

Put 𝑥 = 0; 𝑓 0 = 1

𝑥 = 1; 𝑓 1 = −2.1780

∴ the root lies in the interval (0, 1)

𝑎 0 +𝑏0
Take 𝑎0 = 0 𝑏0 = 1𝑚1 = = 0.5
2

𝑘 𝑎𝑘−1 𝑏𝑘−1 𝑎𝑘 + 𝑏𝑘 𝑓(𝑎𝑘−1 ) 𝑓(𝑚𝑘 ) 𝑓 𝑚𝑘 𝑓(𝑎𝑘−1 )

𝑚𝑘 =
2
1 0 1 0.5 +𝑣𝑒 +𝑣𝑒 >0

2 0.5 1 0.75 +𝑣𝑒 −𝑣𝑒 <0

3 0.5 0.75 0.625 +𝑣𝑒 −𝑣𝑒 <0

4 0.5 0.625 0.5625 +𝑣𝑒 −𝑣𝑒 <0

5
 5 0.5 0.5625 0.53125 +𝑣𝑒 −𝑣𝑒 <0
SPACE FOR HINTS
Hence the smallest positive root lies in the interval (0.5, 0.53125). The
approximate root of the equation is given by 0.515625.

Problem:Find the interval in which the smallest positive root of the

following equation lies tan 𝑥 + tanh 𝑥 = 0. Determine the roots correct
to two decimals using the Bisection Method.

Solution:
Given 𝑓 𝑥 = tan 𝑥 + tanh 𝑥 = 0

tan 𝑥 is discontinuous in the interval (1, 2)

∴ the smallest positive root lies in the interval (2, 3)

Put 𝑥 = 0 ; 𝑓 0 = 0

𝑥 = 1 ; 𝑓 1 = 2.319

𝑥 = 2 ; 𝑓 2 = −1.22

𝑥 = 3 ; 𝑓 3 = 0.852

∴ the root lies in the interval (2, 3)

𝑎 0 +𝑏0 2+3
Take 𝑎0 = 2 𝑏0 = 3𝑚1 = = = 2.5
2 2

𝑘 𝑎𝑘−1 𝑏𝑘−1 𝑚𝑘 𝑓(𝑎𝑘−1 ) 𝑓(𝑚𝑘 ) 𝑓 𝑚𝑘 𝑓(𝑎𝑘−1 )

𝑎𝑘 + 𝑏𝑘
=
2

1 2 3 2.5 −𝑣𝑒 +𝑣𝑒 >0

2 2 2.5 2.25 −𝑣𝑒 −𝑣𝑒 >0

3 2.25 2.5 2.375 −𝑣𝑒 +𝑣𝑒 <0

4 2.25 2.375 2.313 −𝑣𝑒 −𝑣𝑒 >0

5 2.313 2.375 2.344 −𝑣𝑒 −𝑣𝑒 >0

6 2.344 2.375 2.36 −𝑣𝑒 −𝑣𝑒 >0

7 2.36 2.375 2.37 −𝑣𝑒 +𝑣𝑒 <0

6
 8 2.36 2.37 2.37 −𝑣𝑒 +𝑣𝑒 <0
SPACE OF HINTS

∴ the root lies in the interval (2.36, 2.37) and the approximate root is
2.36 + 2.37
= 22.37
2
1.3: Iteration method based on first degree equation
If 𝑓 𝑥 = 0 is the first degree equation in 𝑥 then it can be written as

𝑓 𝑥 = 𝑎0 𝑥 + 𝑎1 = 0

⟹ 𝑎0 𝑥 = −𝑎1
−𝑎1
⟹ 𝑥= (1)
𝑎0

Where 𝑎0 ≠ 0, and 𝑎1 are arbitrary parameter to be determined.

Secant and Regula – Falsi Methods:

If 𝑥𝑘−1 and 𝑥𝑘 are two approximate roots then we determine 𝑎0
and 𝑎1 by using thefollowing conditions.

𝑓𝑘−1 = 𝑎0 𝑥𝑘−1 + 𝑎1 (2)

𝑓𝑘 = 𝑎0 𝑥𝑘 + 𝑎1 (3)

Where 𝑓𝑘−1 = 𝑓(𝑥𝑘−1 ) and 𝑓𝑘 = 𝑓(𝑥𝑘 )

(3) - (2) ⟹ 𝑓𝑘 − 𝑓𝑘 −1 = 𝑎0 (𝑥𝑘 − 𝑥𝑘−1 )

𝑓𝑘 − 𝑓𝑘−1
⟹ 𝑎0 = (4)
𝑥𝑘 − 𝑥𝑘−1

Substituting 𝑎0 in (3) we get

𝑓 𝑘 −𝑓 𝑘−1
(3) ⟹ 𝑓𝑘 = 𝑥𝑘 + 𝑎1
𝑥 𝑘 −𝑥 𝑘−1

𝑓𝑘 − 𝑓𝑘−1
⟹ 𝑎1 = 𝑓𝑘 − 𝑥
𝑥𝑘 − 𝑥𝑘−1 𝑘
𝑓𝑘 𝑥𝑘 − 𝑥𝑘−1 − 𝑓𝑘 𝑥𝑘 + 𝑓𝑘−1 𝑥𝑘
=
𝑥𝑘 − 𝑥𝑘−1
𝑓𝑘 𝑥𝑘 − 𝑓𝑘 𝑥𝑘−1 − 𝑓𝑘 𝑥𝑘 + 𝑓𝑘−1 𝑥𝑘
𝑎1 =
𝑥𝑘 − 𝑥𝑘 −1

7
 𝑓𝑘−1 𝑥𝑘 − 𝑓𝑘 𝑥𝑘−1
𝑎1 =
SPACE FOR HINTS 𝑥𝑘 − 𝑥𝑘−1

Substituting 𝑎0 and 𝑎1 value in (1) we get

−(𝑓 𝑘−1 𝑥 𝑘 −𝑓 𝑘 𝑥 𝑘−1 )
𝑥 𝑘 −𝑥 𝑘−1
𝑥= 𝑓 𝑘 −𝑓 𝑘−1
𝑥 𝑘 −𝑥 𝑘−1

−(𝑓𝑘−1 𝑥𝑘 − 𝑓𝑘 𝑥𝑘−1 ) 𝑓𝑘 𝑥𝑘−1 − 𝑓𝑘−1 𝑥𝑘

= =
𝑓𝑘 − 𝑓𝑘−1 𝑓𝑘 − 𝑓𝑘−1
∴The next interation 𝑥𝑘 +1 is given by,
𝑓𝑘 𝑥𝑘 −1 − 𝑓𝑘−1 𝑥𝑘
𝑥𝑘+1 =
𝑓𝑘 − 𝑓𝑘−1

Adding and subtracting 𝑓𝑘 𝑥𝑘 in numerator,

𝑓𝑘 𝑥𝑘 −1 − 𝑓𝑘−1 𝑥𝑘 + 𝑓𝑘 𝑥𝑘 − 𝑓𝑘 𝑥𝑘
𝑥𝑘+1 =
𝑓𝑘 − 𝑓𝑘−1
𝑥𝑘 (𝑓𝑘 − 𝑓𝑘−1 ) 𝑓𝑘 (𝑥𝑘−1 − 𝑥𝑘 )
= +
𝑓𝑘 − 𝑓𝑘−1 𝑓𝑘 − 𝑓𝑘−1
(𝑥𝑘 − 𝑥𝑘−1 )
𝑥𝑘 +1 = 𝑥𝑘 − 𝑓 (5)
𝑓𝑘 − 𝑓𝑘−1 𝑘

This is called secant or Chord Method.

If the next approximation is passing through the points
𝑥𝑘 , 𝑓𝑘 & (𝑥𝑘−1 , 𝑓𝑘−1 ) and take the point of intersection of the straight
line with the 𝑋-axis.

If the approximations are 𝑓𝑘 . 𝑓𝑘−1 < 0 then the above method is known
as Regula-Falsi method.

Example1.3.1:
A real root of the equation 𝑓 𝑥 = 𝑥 3 − 5𝑥 + 1 = 0 lies in the interval
0,1 .Performfour iterations of the secant method and Regula-Falsi
method to obtain this root.

Solution:

8
Secant Method:
SPACE OF HINTS
Given 𝑓 𝑥 = 𝑥 3 − 5𝑥 + 1 = 0

Put 𝑥 = 0 ; 𝑓 0 = 1

𝑥 = 1 ; 𝑓 1 = −3

∴The root lies between (0,1)

(𝑥𝑘 − 𝑥𝑘−1 )
𝑥𝑘+1 = 𝑥𝑘 − 𝑓
𝑓𝑘 − 𝑓𝑘−1 𝑘
𝑥 1 −𝑥 0
Put 𝑘 = 1 ⇒ 𝑥2 = 𝑥1 − 𝑓1
𝑓1 −𝑓0

Where 𝑥0 = 0 , 𝑥1 = 1 , 𝑓0 = 1 , 𝑓1 = −3
1−0
∴ 𝑥2 = 1 − (−3)
−3 − 1
3
=1+ = 1 − 0.75
−4
𝑥2 = 0.25

𝑓2 = 𝑓 𝑥2 = 𝑓 0.25 = −0.2344
𝑥 2 −𝑥 1
Put 𝑘 = 2 ⇒ 𝑥3 = 𝑥2 − 𝑓2
𝑓2 −𝑓1

0.25 − 1
= 0.25 − −0.2344
−0.2344 + 3
0.1758
= 0.25 −
2.7656
= 0.25 − 0.0635

𝑥3 = 0.1864

𝑓3 = 𝑓 𝑥3 = 𝑓 0.1864 = 0.0745
𝑥 3 −𝑥 2
Put 𝑘 = 3 ⇒ 𝑥4 = 𝑥3 − 𝑓3
𝑓3 −𝑓2

0.1864 − 0.25
= 0.1864 − 0.0745
0.0745 + 0.2344
= 0.1864 + 0.0153

= 0.20173

9
 𝑓4 = 𝑓 𝑥4 = 𝑓 0.20173 = −0.00044
SPACE FOR HINTS 𝑥 4 −𝑥 3
Put 𝑘 = 4 ⇒ 𝑥5 = 𝑥4 − 𝑓4
𝑓4 −𝑓3

0.20173 − 0.1864
= 0.20173 − (−0.00044)
−0.00044 − 0.0745
= 0.20173 + 0.00009

= 0.2016

Regula-Falsi Method:

Given the initial values are 𝑥0 = 0, 𝑓0 = 1, 𝑥1 = 1, 𝑓1 = −3;

𝑓 𝑥 = 𝑥 3 − 5𝑥 + 1
The formula for Regula-Falsi method is
(𝑥𝑘 − 𝑥𝑘−1 )
𝑥𝑘+1 = 𝑥𝑘 − 𝑓
𝑓𝑘 − 𝑓𝑘−1 𝑘
𝑥 1 −𝑥 0
Put 𝑘 = 1 ⟹ 𝑥2 = 𝑥1 − 𝑓1
𝑓1 −𝑓0

𝑥2 = 0.25

𝑓2 = 𝑓 𝑥2 = 𝑓 0.25 = −0.2344

Now, 𝑓0 𝑓2 < 0

in the next iteration interval 𝐼 = (𝑥0 , 𝑥2 )

= (0, 0.25)
𝑥 2 −𝑥 1
Put 𝑘 = 2 ⟹ 𝑥3 = 𝑥2 − 𝑓2
𝑓2 −𝑓1

0.25 − 0
= 0.25 − (−0.2344)
−0.2344 − 1
0.0586
= 0.25 + = 0.25 − 0.0474
−1.2344
= 0.2025

𝑓3 = 𝑓 𝑥3 = 𝑓 0.2025 = −0.0042

Now, 𝑓0 𝑓3 < 0

In the next iteration interval 𝐼 = (𝑥0 , 𝑥3 )

10
 = (0,0.2025)
SPACE OF HINTS
𝑥 3 −𝑥 2
Put 𝑘 = 3 ⇒ 𝑥4 = 𝑥3 − 𝑓3
𝑓3 −𝑓2

0.2025
= 0.2025 − (−0.0042)
−0.0042 − 1
= 0.2025 − 0.000846

= 0.2017

𝑓4 = 𝑓 𝑥4 = 𝑓 0.2017 = −0.00029

Now, 𝑓0 𝑓4 < 0

The next iteration the interval 𝐼 = 𝑥0 , 𝑥4

= (0,0.2017)
𝑥 4 −𝑥 3
Put 𝑘 = 4 ⇒ 𝑥5 = 𝑥4 − 𝑓4
𝑓4 −𝑓3

0.2017 − 0
= 0.2017 − (−0.00029)
−0.00029 − 1
= 0.2017 − 0.000058 = 0.2016

By Secant and Regula-Falsi method the approximation root of the

given equation is 𝑥 = 0.2016

Example1.3.2:
Use the Secant and Regula-Falsi methods to determine the root of
the equation 𝑐𝑜𝑠𝑥 − 𝑥𝑒 𝑥 = 0

Solution:

Secant Method:
Given 𝑓 𝑥 = 𝑐𝑜𝑠𝑥 − 𝑥𝑒 𝑥 = 0

Put 𝑥 = 0 ⟹ 𝑓 0 =1

𝑥=1 ⟹ 𝑓 1 = −2.1779

∴The root lies between 0 and 1

Let 𝑥0 = 0 , 𝑓0 = 1 , 𝑥1 = 1 , 𝑓1 = −2.1779

The formula for Secant method is

11
 (𝑥𝑘 − 𝑥𝑘−1 )
𝑥𝑘+1 = 𝑥𝑘 − 𝑓
SPACE FOR HINTS 𝑓𝑘 − 𝑓𝑘−1 𝑘
𝑥 1 −𝑥 0
Put 𝑘 = 1 ⟹ 𝑥2 = 𝑥1 − 𝑓1
𝑓1 −𝑓0

1−0
=1− −2.1779
−2.1779 − 1
= 1 − 0.6853

𝑥2 = 0.3147

𝑓2 = 𝑓 𝑥2 = 0.5198
𝑥 2 −𝑥 1
Put 𝑘 = 2 ⟹ 𝑥3 = 𝑥2 − 𝑓2
𝑓2 −𝑓1

0.3147 − 1
= 0.3147 − (0.5198)
0.5198 + 2.1779
= 0.3147 + 0.1320 = 0.4467

𝑓3 = 𝑓 𝑥3 = 0.2036
𝑥 3 −𝑥 2
Put 𝑘 = 3 ⟹ 𝑥4 = 𝑥3 − 𝑓3
𝑓3 −𝑓2

0.4467 − 0.3147
= 0.4467 − (0.2036)
0.2036 − 0.5198
= 0.4467 + 0.0849

= 0.5316

𝑓4 = 𝑓 𝑥4 = −0.0426
𝑥 4 −𝑥 3
Put 𝑘 = 4 ⇒ 𝑥5 = 𝑥4 − 𝑓4
𝑓4 −𝑓3

0.5316 − 0.4467
= 0.5316 − (−0.0426)
−0.0426 − 0.2036
= 0.5316 − 0.0147

= 0.5169

𝑓5 = 𝑓 𝑥5 = 0.0026
𝑥 5 −𝑥 4
Put 𝑘 = 5 ⟹ 𝑥6 = 𝑥5 − 𝑓5
𝑓5 −𝑓4

12
 0.5169 − 0.5136 SPACE OF HINTS
= 0.5169 − 0.0026
0.0026 + 0.0426
= 0.5169 − 0.000189

= 0.5168 = 0.5167

𝑓6 = 𝑓 𝑥6 = 0.0032
𝑥 6 −𝑥 5
Put 𝑘 = 6 ⟹ 𝑥7 = 𝑥6 − 𝑓6
𝑓6 −𝑓5

0.5177 − 0.5170
= 0.5177 − (0.0017)
0.0017 − 0.0023
= 0.5177 + 0.00007

= 0.5177

𝑓7 = 𝑓 𝑥7 = 0.00017

Regula-Falsi Method:
Given the initial values are

𝑥=0 ⟹ 𝑓 0 =1

𝑥=1 ⟹ 𝑓 1 = −2.1779

The formula for Regula-Falsi method

(𝑥𝑘 − 𝑥𝑘−1 )
𝑥𝑘+1 = 𝑥𝑘 − 𝑓
𝑓𝑘 − 𝑓𝑘−1 𝑘

𝑥2 = 0.3147 and 𝑓 𝑥2 = 0.5198

Now 𝑓1 𝑓2 < 0

∴In the next iteration, interval 𝐼 = (𝑥2 , 𝑥1 )

= (0.3147,1)
𝑥 1 −𝑥 2
Put 𝑘 = 2 ⟹ 𝑥3 = 𝑥1 − 𝑓1
𝑓1 −𝑓2

1 − 0.3147
=1− (−2.1779)
−2.1779 − 0.5198
= 1 − 0.5533

= 0.4476

13
 𝑓3 = 𝑓 𝑥3 = 0.2036
SPACE FOR HINTS
Now 𝑓1 𝑓3 < 0

In the next iteration the interval 𝐼 = (0.4467,1)

𝑥1 − 𝑥3
⟹ 𝑥4 = 𝑥1 − 𝑓
𝑓1 − 𝑓3 1
1 − 0.4476
=1− (−2.1779)
−2.1779 − 0.2036
= 1 − 0.5051

= 0.4940

𝑓4 = 𝑓 𝑥4 = 0.0708

Now 𝑓1 𝑓4 < 0

In the next iteration the interval 𝐼 = 0.4940,1

𝑥1 − 𝑥4
⟹ 𝑥5 = 𝑥1 − 𝑓
𝑓1 − 𝑓4 1
1 − 0.4940
=1− (−2.1779)
−2.1779 − 0.0708
= 1 − 0.4900

= 0.51

𝑓5 = 𝑓 𝑥5 = 0.0234

Now 𝑓1 𝑓3 < 0

In the next iteration the interval 𝐼 = (0.51,1)

𝑥1 − 𝑥5
⟹ 𝑥6 = 𝑥1 − 𝑓
𝑓1 − 𝑓5 1
1 − 0.51
=1− (−2.1779)
−2.1779 − 0.0234
= 1 − 0.4714

= 0.5286

𝑓6 = 𝑓 𝑥6 = −0.0332

14
Now 𝑓5 𝑓6 < 0
SPACE OF HINTS
∴In the next iteration the interval 𝐼 = (0.51,0.5286)
𝑥6 − 𝑥5
⟹ 𝑥7 = 𝑥6 − 𝑓
𝑓6 − 𝑓5 6
0.5286 − 0.51
= 0.5286 − (−0.0332)
−0.0332 − 0.0234
= 0.5286 − 0.0109 = 0.5177

∴ the Secant and Regula-Falsi Method the approximate root of the given
equation is 0.5177

Newton - Raphson Method:

Let 𝑓 𝑥𝑘 = 𝑎0 𝑥𝑘 + 𝑎1 (1)

Then differentiating (1) w.r.to 𝑥,

𝑓 ′ 𝑥𝑘 = 𝑎0 (2)

∴ (1)⇒𝑓 𝑥𝑘 = 𝑓 ′ 𝑥𝑘 𝑥𝑘 + 𝑎1

⟹ 𝑎1 = 𝑓 𝑥𝑘 − 𝑓 ′ 𝑥𝑘 𝑥𝑘 (3)
−𝑎 1
The solution for (1) is 𝑥𝑘+1 = where 𝑎0 ≠ 0 (4)
𝑎0

Substituting (2) and (3) in (4), we get

−(𝑓 𝑥𝑘 − 𝑓 ′ 𝑥𝑘 𝑥𝑘 )
𝑥𝑘+1 =
𝑓 ′ 𝑥𝑘
−𝑓 𝑥𝑘 𝑓 ′ 𝑥𝑘 𝑥𝑘
= +
𝑓 ′ 𝑥𝑘 𝑓 ′ 𝑥𝑘
𝑓 𝑥𝑘
𝑥𝑘+1 = 𝑥𝑘 −
𝑓 ′ 𝑥𝑘

∴ the general formula for Newton – Raphson Method is

𝑓 𝑥𝑘
𝑥𝑘+1 = 𝑥𝑘 − ; where 𝑘 = 0,1,2, ⋯
𝑓 ′ 𝑥𝑘

Example 1.3.3:
Perform four iterations of thee Newton – Raphson method to find the
smallestpositive root of the equation 𝑓 𝑥 = 𝑥 3 − 5𝑥 + 1 = 0

15
 Solution:
SPACE FOR HINTS Given that 𝑓 𝑥 = 𝑥 3 − 5𝑥 + 1 = 0 (1)

𝑓 ′ 𝑥 = 3𝑥 2 − 5

Put 𝑥 = 0 ; 𝑓 0 = 1

𝑥 =1; 𝑓 1 =1−5+1=3

∴ the root lies between in the interval (0,1)

Let the initial approximation 𝑥0 = 0.

The general formula for Newton –Raphson Method is

𝑥𝑘 +1
𝑓 𝑥𝑘
= 𝑥𝑘 − (2)
𝑓 ′ 𝑥𝑘

𝑓 𝑥0
Put 𝑘 = 0 ⟹ 𝑥1 = 𝑥0 −
𝑓 ′ 𝑥0

1 1
=0− = = 0.2
−5 5
𝑓 𝑥1 = 𝑓 0.2 = 0.008

𝑓 ′ 𝑥1 = 𝑓 ′ 0.2 = −4.88
𝑓 𝑥1
Put 𝑘 = 1 ⟹ 𝑥2 = 𝑥1 −
𝑓 ′ 𝑥1

0.008
= 0.2 −
−4.88
= 0.2 + 0.00163

= 0.20164

𝑓 𝑥2 = 0.008

𝑓 ′ 𝑥2 = −4.8780
𝑓 𝑥2
Put 𝑘 = 2 ⟹ 𝑥3 = 𝑥2 −
𝑓 ′ 𝑥2

0.008
= 0.20164 −
−4.8780
= 0.20164

16
𝑓 𝑥3 = 𝑓 0.20164 = −0.0000016
SPACE OF HINTS
′ ′
𝑓 𝑥3 = 𝑓 0.20164 = −4.8780

Put 𝑘 = 2 in (2) we get

𝑓 𝑥3
⟹ 𝑥4 = 𝑥3 −
𝑓 ′ 𝑥3
(−0.0000016)
= 0.20164 −
−4.8780
= 0.20164

𝑓 𝑥4 = 𝑓 0.20164 = −0.0000016

𝑓 ′ 𝑥3 = 𝑓 ′ 0.20164 = -4.8780
In the forth iteration the approximate value of given equation is

𝑥 = 0.20164

Example1.3.4:
Perform four iterations of the Newton Raphson method to obtain the
1
approximation value of (17) 3. Starting with initial approximation
𝑥0 = 2.

Solution:
1
Let the approximation value 𝑥 = (17) 3

⟹ 𝑥 3 = 17

⟹ 𝑥 3 − 17 = 0

∴ 𝑓 𝑥 = 𝑥 3 − 17 = 0

𝑓 ′ 𝑥 = 3𝑥 2

Given 𝑥0 = 2

∴ 𝑓 𝑥0 = 𝑓 2 = −9

𝑓 ′ 𝑥0 = 𝑓 ′ 2 = 12
𝑓 𝑥𝑘
The general formula for Newton-Raphson Method is 𝑥𝑘+1 = 𝑥𝑘 −
𝑓 ′ 𝑥𝑘

17
 𝑓 𝑥0
Put 𝑘 = 0 ⟹ 𝑥1 = 𝑥0 −
𝑓 ′ 𝑥0
SPACE FOR HINTS
−9
=2−
12
= 2.75

𝑓 𝑥1 = 3.7969

𝑓 ′ 𝑥1 = 22.6875
𝑓 𝑥1
Put 𝑘 = 1 ⟹ 𝑥2 = 𝑥1 −
𝑓 ′ 𝑥1

3.7969
= 2.75 −
22.6875
= 2.5826

𝑓 𝑥2 = 0.2255

𝑓 ′ 𝑥2 = 20.0095
𝑓 𝑥2
Put 𝑘 = 2 ⟹ 𝑥3 = 𝑥2 −
𝑓 ′ 𝑥2

0.2255
= 2.5826 −
20.0095
= 2.5713

𝑓 𝑥3 = 0.000365

𝑓 ′ 𝑥3 = 19.8348
𝑓 𝑥3
Put 𝑘 = 3 ⟹ 𝑥4 = 𝑥3 −
𝑓 ′ 𝑥3

0.000365
= 2.5713 −
19.8348
= 2.5713

∴in the fourth iteration, the approximate value is 𝑥 = 2.5713

Example1.3.5:
Apply Newton-Raphson Method to determine a root of the
equation 𝑓 𝑥 = cos 𝑥 − 𝑥𝑒 𝑥 = 0 such that 𝑓 𝑥 ∗ < 10−8 where 𝑥 ∗
is the approximation to the root. Take the initial approximation 𝑥0 = 1

18
 SPACE OF HINTS
Solution:
Let 𝑓 𝑥 = cos 𝑥 − 𝑥𝑒 𝑥

𝑓 ′ 𝑥 = − sin 𝑥 − (𝑒 𝑥 + 𝑥𝑒 𝑥 )

Given 𝑥0 = 1

𝑓 𝑥0 = 𝑓 1 = −2.1780

𝑓 ′ 𝑥0 = −6.2780
𝑓 𝑥𝑘
The general formula for Newton-Raphson Method is 𝑥𝑘+1 = 𝑥𝑘 −
𝑓 ′ 𝑥𝑘

𝑓 𝑥0
Put 𝑘 = 0 ⟹ 𝑥1 = 𝑥0 −
𝑓 ′ 𝑥0

2.1780
=1+
−6.2780
= 0.6531

𝑓 𝑥1 = −0.4607

𝑓 ′ 𝑥1 = −3.7841
𝑓 𝑥1
Put 𝑘 = 1 ⟹ 𝑥2 = 𝑥1 −
𝑓 ′ 𝑥1

0.4607
= 0.6531 +
−3.7841
= 0.5341

𝑓 𝑥2 = −0.04198

𝑓 ′ 𝑥2 =-3.1121
𝑓 𝑥2
Put 𝑘 = 2 ⟹ 𝑥3 = 𝑥2 −
𝑓 ′ 𝑥2

0.04198
= 0.5314 +
−3.1121
= 0.5179

𝑓 𝑥3 = −0.00043

𝑓 ′ 𝑥3 =-3.0429

19
 𝑓 𝑥3
Put 𝑘 = 3 ⟹ 𝑥4 = 𝑥3 −
𝑓 ′ 𝑥3
SPACE FOR HINTS
0.00043
= 0.5179 +
−3.0429
= 0.5178

𝑓 𝑥4 = −0.00013

𝑓 ′ 𝑥4 = −3.0423
𝑓 𝑥4
Put 𝑘 = 4 ⟹ 𝑥5 = 𝑥4 −
𝑓 ′ 𝑥4

0.00013
= 0.5178 +
−3.0423
= 0.5178

∴ The approximation root of the given equation is 0.5178

Example1.3.6:
1
Show that the initial approximation 𝑥0 for finding , where 𝑁 is a
𝑁

positive integer, bythe Newton-Raphson method must satisfy

2
0 < 𝑥0 < , for convergence.
𝑁

Solution:
1
Given 𝑥0 =
𝑁

1 1
(i.e.) 𝑥= ⟹𝑁=
𝑁 𝑥

1 1
⟹𝑁− =0= −𝑁 =0
𝑥 𝑥
1
∴𝑓 𝑥 = −𝑁 (1)
𝑥
1
𝑓′ 𝑥 = − (2)
𝑥2
𝑓 𝑥𝑘
The general formula for Newton-Raphson method is 𝑥𝑘 +1 = 𝑥𝑘 −
𝑓 ′ 𝑥𝑘

1
∴ 𝑓 𝑥𝑘 = −𝑁
𝑥𝑘

20
 𝑓 ′ 𝑥𝑘 = −1/𝑥𝑘 2
SPACE OF HINTS
1
( − 𝑁)
𝑥𝑘
𝑥𝑘 +1 = 𝑥𝑘 − 1
−
𝑥𝑘 2

1−𝑁𝑥 𝑘 )
( )
𝑥𝑘
= 𝑥𝑘 + 1
𝑥𝑘 2

1 𝑁𝑥 𝑘
+ (1 − )
𝑥𝑘 𝑥𝑘
= 1
𝑥𝑘 2

2 − 𝑁𝑥𝑘 𝑥𝑘 2
= ×
𝑥𝑘 1

= 2 − 𝑁𝑥𝑘 𝑥𝑘 (3)
1
Now draw the graph for equations 𝑥 = and𝑥𝑘+1 = (2 − 𝑁𝑥𝑘 )𝑥𝑘
𝑁

1
(i.e.) 𝑦 = 𝑥 = & 𝑦 = 2 − 𝑁𝑥 𝑥
𝑁

𝑦𝑦 = 𝑥

1 1
( , )
𝑁 𝑁

(0,0) 𝑥
2
( , 0)
𝑁

2
From the graph, the initial approximation must satisfy 0 < 𝑥0 < for
𝑁

converges.

21
 2
The initial approximation outside the range 0 < 𝑥0 < , the method
𝑁
SPACE FOR HINTS
diverges.

1.4: Iteration method based on second degreeequation

Let us assume 𝑓(𝑥) be a polynomial of degree two.

(i.e.) 𝑓 𝑥 = 𝑎0 𝑥 2 + 𝑎1 𝑥 + 𝑎2 = 0 𝑎0 ≠ 0 (1)

where 𝑎0 , 𝑎1 , 𝑎2 are arbitrary parameters to be determined by

prescribing three appropriate conditions on 𝑓(𝑥) and / or its derivatives.

Muller Method:
If𝑥𝑘−2 , 𝑥𝑘−1 , &𝑥𝑘−1 and 𝑥𝑘 are three approximations to the root 𝜉
of 𝑓 𝑥 = 0, thento determine 𝑎0 , 𝑎1 , 𝑎2 by using the conditions.

𝑓𝑘−2 = 𝑓 𝑥𝑘−2 = 𝑎0 𝑥 2 𝑘−2 + 𝑎1 𝑥𝑘−2 + 𝑎2

𝑓𝑘−1 = 𝑓 𝑥𝑘−1 = 𝑎0 𝑥 2 𝑘−1 + 𝑎1 𝑥𝑘−1 + 𝑎2 (2)

𝑓𝑘 = 𝑓 𝑥𝑘 = 𝑎0 𝑥 2 𝑘 + 𝑎1 𝑥𝑘 + 𝑎2

Eliminating 𝑎0 , 𝑎1 , & 𝑎2 ,we get

𝑓(𝑥) 𝑥2 𝑥 1
2
𝑓𝑘−2 𝑥 𝑘−2 𝑥𝑘−2 1
𝑓𝑘−1 𝑥 2 𝑘−1 𝑥𝑘−1 1 = 0
𝑓𝑘 𝑥2𝑘 𝑥𝑘 1

𝑥 2 𝑘−2 𝑥𝑘−2 1 𝑥2 𝑥 1
𝑓 𝑥 𝑥 2 𝑘−1 𝑥𝑘−1 1 − 𝑓𝑘−2 𝑥 2 𝑘−1 𝑥𝑘−1 1
𝑥2 𝑘 𝑥𝑘 1 𝑥2𝑘 𝑥𝑘 1
𝑥2 𝑥 1 𝑥2 𝑥 1
2 2
+ 𝑓𝑘−1 𝑥 𝑘−2 𝑥𝑘 −2 1 − 𝑓𝑘 𝑥 𝑘−2 𝑥𝑘−2 1 =0
𝑥2 𝑘 𝑥𝑘 1 𝑥2 𝑘 𝑥𝑘 1

On simplification, we get

22
 𝑥 − 𝑥𝑘 −1 (𝑥 − 𝑥𝑘 )
𝑓 𝑥 = 𝑓 SPACE OF HINTS
𝑥𝑘−2 − 𝑥𝑘 −1 (𝑥𝑘−2 − 𝑥𝑘 ) 𝑘 −2
𝑥 − 𝑥𝑘−2 (𝑥 − 𝑥𝑘 )
+ 𝑓
𝑥𝑘−1 − 𝑥𝑘−2 (𝑥𝑘−1 − 𝑥𝑘 ) 𝑘−1
𝑥 − 𝑥𝑘 −2 (𝑥 − 𝑥𝑘−1 )
+ 𝑓
𝑥𝑘 − 𝑥𝑘 −2 (𝑥𝑘 − 𝑥𝑘−1 ) 𝑘

The above equation can be written as

𝑕 + 𝑕𝑘 𝑕 𝑕 + 𝑕𝑘 + 𝑕𝑘−1 𝑕
𝑓 𝑥 = 𝑓𝑘−2 + 𝑓𝑘−1
(− 𝑕𝑘−1 )(− 𝑕𝑘−1 + 𝑕𝑘 ) 𝑕𝑘−1 (−𝑕𝑘 )
𝑕 + 𝑕𝑘 + 𝑕𝑘−1 (𝑕 + 𝑕𝑘 )
+ 𝑓𝑘
(𝑕𝑘 + 𝑕𝑘−1 )𝑕𝑘
𝑕 + 𝑕𝑘 𝑕 𝑕 + 𝑕𝑘 + 𝑕𝑘−1 𝑕
= 𝑓𝑘−2 − 𝑓𝑘−1
𝑕𝑘−1 𝑕𝑘−1 + 𝑕𝑘 𝑕𝑘−1 𝑕𝑘
𝑕 + 𝑕𝑘 + 𝑕𝑘−1 (𝑕 + 𝑕𝑘 )
+ 𝑓𝑘
(𝑕𝑘 + 𝑕𝑘−1 )𝑕𝑘

Since 𝑓 𝑥 = 0,

𝑕𝑘 ( 𝑕 + 𝑕𝑘 𝑕𝑓𝑘−2 − (𝑕𝑘 + 𝑕𝑘−1 )(𝑕2 + 𝑕𝑕𝑘 + 𝑕𝑕𝑘−1 )𝑓𝑘−1

⟹
𝑕𝑘 𝑕𝑘−1 𝑕𝑘−1 + 𝑕𝑘 + 𝑕2 + 𝑕𝑕𝑘 + 𝑕𝑕𝑘−1 + 𝑕𝑕𝑘 𝑕𝑘 2 + 𝑕𝑘 𝑕𝑘−1 𝑕𝑘−1 𝑓𝑘

=0

⟹ 𝑕2 𝑕𝑘 + 𝑕𝑕𝑘 2 𝑓𝑘−2
− 𝑕2 𝑕𝑘 + 𝑕2 𝑕𝑘−1 + 𝑕𝑕𝑘 2 + 𝑕𝑕𝑘 𝑕𝑘−1 + 𝑕𝑕𝑘 𝑕𝑘−1
+ 𝑕𝑕𝑘−1 2 𝑓𝑘−1

+ 𝑕2 𝑕𝑘−1 + 𝑕𝑕𝑘 𝑕𝑘−1 𝑕𝑘 𝑕𝑘−1 2

+𝑕𝑕𝑘−1 2 + 𝑕𝑕𝑘 𝑕𝑘−1 + 𝑕𝑘 2 𝑕𝑘−1 𝑓𝑘 = 0

⟹ 𝑕2 𝑕𝑘 + 𝑕𝑕𝑘 2 𝑓𝑘−2
− 𝑕2 𝑕𝑘 + 𝑕2 𝑕𝑘−1 + 𝑕𝑕𝑘 2 + 𝑕𝑕𝑘−1 2 + 2 𝑕𝑕𝑘 𝑕𝑘−1 𝑓𝑘−1
+ (𝑕2 𝑕𝑘−1 + 𝑕𝑕𝑘−1 2 + 𝑕𝑘 2 𝑕𝑘−1 + 𝑕𝑘 𝑕𝑘−1 2
+ 2𝑕𝑕𝑘 𝑕𝑘−1 ) 𝑓𝑘 = 0

23
 𝑕 𝑕𝑘 𝑕
Let 𝜆 = , 𝜆𝑘 = ⇒ 𝜆𝜆𝑘 =
𝑕𝑘 𝑕 𝑘−1 𝑕 𝑘−1
SPACE FOR HINTS
Dividing above equation by 𝑕𝑘 2 𝑕𝑘−1 , we get
2
𝑕 𝑕𝑘 𝑕𝑘
+ 𝑓
𝑕𝑘 𝑕𝑘−1 𝑕𝑘−1 𝑘−2

2 2
𝑕 𝑕𝑘 𝑕 𝑕𝑘 𝑕 𝑕𝑘−1 𝑕
− + + + . + 2. 𝑓
𝑕𝑘 𝑕𝑘−1 𝑕𝑘 𝑕𝑘−1 𝑕𝑘 𝑕𝑘 𝑕𝑘 𝑘−1

2
𝑕 𝑕 𝑕𝑘−1 𝑕𝑘−1 𝑕
+ + . +1+ + 2. 𝑓𝑘 = 0
𝑕𝑘 𝑕 𝑘 𝑕𝑘 𝑕𝑘 𝑕𝑘

1
⟹ 𝜆2 𝜆𝑘 + 𝜆. 𝜆𝑘 𝑓𝑘−2 − 𝜆2 . 𝜆𝑘 + 𝜆2 + 𝜆. 𝜆𝑘 + 𝜆. + 2𝜆 𝑓𝑘−1
𝜆𝑘
1 1
+ 𝜆2 + 𝜆. + 1 + + 2𝜆 𝑓𝑘 = 0
𝜆𝑘 𝜆𝑘

Multiplying the above equation by 𝜆𝑘 , we get

𝜆2 𝜆𝑘 2 + 𝜆𝜆𝑘 2 𝑓𝑘−2 − 𝜆2 𝜆𝑘 2 + 𝜆2 𝜆𝑘 + 𝜆𝜆𝑘 2 + 𝜆 + 2𝜆. 𝜆𝑘 𝑓𝑘−1

+ 𝜆2 𝜆𝑘 + 𝜆 + 𝜆𝑘 + 1 + 2. 𝜆. 𝜆𝑘 𝑓𝑘 = 0

⟹ 𝜆2 𝜆𝑘 2 𝑓𝑘−2 − 𝜆𝑘 2 𝑓𝑘−1 − 𝜆𝑘 𝑓𝑘 −1 + 𝜆𝑘 𝑓𝑘

+𝜆 𝜆𝑘 2 𝑓𝑘−2 − 𝜆𝑘 2 𝑓𝑘−1 − 𝑓𝑘−1 − 2𝜆𝑘 𝑓𝑘−1 + 𝑓𝑘 − 2𝜆𝑘 𝑓𝑘

+ 𝜆𝑘 𝑓𝑘 + 𝑓𝑘 = 0

⟹ 𝜆2 . 𝜆𝑘 𝜆𝑘 𝑓𝑘−2 − 𝜆𝑘 𝑓𝑘−1 − 𝑓𝑘 −1 + 𝑓𝑘
+ 𝜆 𝜆𝑘 2 𝑓𝑘−2 − 𝜆𝑘 2 𝑓𝑘 −1 − 𝑓𝑘−1 − 2𝜆𝑘 𝑓𝑘−1 + 𝑓𝑘 − 2𝜆𝑘 𝑓𝑘
+ 𝜆𝑘+1 𝑓𝑘 = 0

⟹ 𝜆2 . 𝜆𝑘 𝜆𝑘 𝑓𝑘−2 − 𝜆𝑘+1 𝑓𝑘−1 + 𝑓𝑘

+ 𝜆 𝜆𝑘 2 𝑓𝑘−2 − 𝜆𝑘+1 2 + 2𝜆𝑘 𝑓𝑘−1 + 1 + 2𝜆𝑘 𝑓𝑘
+ 𝜆𝑘+1 𝑓𝑘 = 0

⟹ 𝜆2 . 𝜆𝑘 𝜆𝑘 𝑓𝑘−2 − 𝜆𝑘−1 𝑓𝑘−1 + 𝑓𝑘

+ 𝜆 𝜆𝑘 2 𝑓𝑘−2 − 𝜆𝑘+1 2 𝑓𝑘−1 + 1 + 2𝜆 𝑓𝑘 + 𝜆𝑘+1 𝑓𝑘 = 0

24
 Take 1 + 𝜆𝑘 = 𝛿𝑘
SPACE OF HINTS
2 2 2
⟹ 𝜆 . 𝜆𝑘 𝜆𝑘 𝑓𝑘−2 − 𝛿𝑘 𝑓𝑘−1 + 𝑓𝑘 + 𝜆 𝜆𝑘 𝑓𝑘−2 − 𝛿𝑘 𝑓𝑘−1
+ 𝛿𝑘 + 𝜆𝑘 𝑓𝑘 + 𝛿𝑘 𝜆𝑘 = 0 (3)

Let 𝑐𝑘 = 𝜆𝑘 𝜆𝑘 𝑓𝑘−2 − 𝛿𝑘 𝑓𝑘−1 + 𝑓𝑘 and 𝑔𝑘 = 𝜆𝑘 2 𝑓𝑘−2 − 𝛿𝑘 2 𝑓𝑘−1 +

(𝜆𝑘 + 𝛿𝑘 )𝑓𝑘

∴ 3 ⟹ 𝜆2 𝑐𝑘 + 𝜆𝑔𝑘 + 𝛿𝑘 𝑓𝑘 = 0

Dividing above equation by 𝜆2 , We get

𝑔𝑘 𝛿𝑘 𝑓𝑘
𝑐𝑘 + + 2 =0
𝜆 𝜆
1
⟹ 1/𝜆2 𝛿𝑘 𝑓𝑘 + 𝑔𝑘 + 𝑐𝑘 = 0
𝜆

1
The above equation is quadratic equation in ,
𝜆

Here 𝑎 = 𝛿𝑘 𝑓𝑘 , 𝑏 = 𝑔𝑘 , 𝑐 = 𝑐𝑘

1 −𝑏 ± 𝑏2 − 4𝑎𝑐 𝑔𝑘 2 − 4𝛿𝑘 𝑓𝑘 𝑐𝑘
= = −𝑔𝑘 ±
𝜆 2𝑎 2𝛿𝑘 𝑓𝑘
−2𝛿𝑘 𝑓𝑘
𝜆=
𝑔𝑘 ± 𝑔𝑘 2 − 𝛿𝑘 𝑓𝑘 𝑐𝑘

𝜆 = 𝜆𝑘+1

The sign in the denominators is chosen such that 𝜆𝑘+1 has the smallest
absolute value. We also know that
𝑕 (𝑥 − 𝑥𝑘 )
𝜆 = 𝜆𝑘+1 = =
𝑕𝑘 (𝑥𝑘 − 𝑥𝑘−1 )

(𝜆𝑘+1 ) 𝑥𝑘 − 𝑥𝑘−1 = 𝑥 − 𝑥𝑘

𝑥 = 𝑥𝑘 + 𝜆𝑘+1 (𝑥𝑘 − 𝑥𝑘−1 )

Replace 𝑥 by 𝑥𝑘+1 , we obtain the Muller Method

(i.e.)𝑥𝑘+1 = 𝑥𝑘 + 𝜆𝑘+1 𝑥𝑘 − 𝑥𝑘−1 ; 𝑘 = 2,3, ⋯

Note:
In Muller method the following conditions are needed

25
 i) 𝑕 = 𝑥 − 𝑥𝑘 ; 𝑕𝑘 = 𝑥𝑘 − 𝑥𝑘−1
SPACE FOR HINTS 𝑕𝑘
ii) 𝜆𝑘 = and 𝛿𝑘 = 1 + 𝜆𝑘
𝑕 𝑘−1

iii) 𝑔𝑘 = 𝜆𝑘 2 𝑓𝑘−2 − 𝛿𝑘 2 𝑓𝑘−1 + (𝜆𝑘 + 𝛿𝑘 )𝑓𝑘 and

𝑐𝑘 = 𝜆𝑘 𝜆𝑘 𝑓𝑘−2 − 𝛿𝑘 𝑓𝑘−1 + 𝑓𝑘
−2𝛿 𝑘 𝑓 𝑘
iv) 𝜆𝑘+1 =
𝑔 𝑘 ± 𝑔 𝑘 2 −4𝛿 𝑘 𝑓 𝑘 𝑐 𝑘

v) 𝑥𝑘+1 = 𝑥𝑘 + 𝜆𝑘+1 𝑥𝑘 − 𝑥𝑘−1 ; 𝑘 = 2, 3, ⋯

Example 1.4.1:
Perform three iterations of the Muller methods to find the smallest
positive rootof the equation 𝑓 𝑥 = 𝑥 3 − 5𝑥 + 1 = 0

Solution:
Given 𝑓 𝑥 = 𝑥 3 − 5𝑥 + 1 = 0

Let 𝑥 = 0 ; 𝑓 𝑥 = 1

𝑥 = 1 ; 𝑓 1 = 1 − 5 + 1 = −3

∴ The smallest positive root of the given equation lies in the interval
(0, 1).

Let the initial approximations 𝑥0 = 0 , 𝑥1 = 0.5 , 𝑥2 = 1

∴ 𝑓0 = 1 , 𝑓1 = −1.375 , 𝑓2 = −3

First iteration:

i) 𝑕𝑘 = 𝑥𝑘 − 𝑥𝑘−1

𝑘 = 1 ⟹ 𝑕1 = 𝑥1 − 𝑥0

= 0.5

𝑘 = 2 ⟹ 𝑕2 = 𝑥2 − 𝑥1

= 1 − 0.5 = 0.5
𝑕𝑘
ii) 𝜆𝑘 =
𝑕 𝑘−1

𝑕2
𝑘 = 2 ⟹ 𝜆2 =
𝑕1

26
 0.5
= =1 SPACE OF HINTS
0.5
𝛿𝑘 = 1 + 𝜆𝑘

Put 𝑘 = 2 ⟹ 𝛿2 = 1 + 𝜆2

=1+1=2

iii) 𝑔2 = 𝜆2 2 𝑓0 − 𝛿2 2 𝑓1 + (𝜆2 + 𝛿2 )𝑓2

= (1)2 1 − 4. −1.375 + (1 + 2)(−3)
= 1 + 5.5 − 9
= −2.5
𝑐𝑘 = 𝜆𝑘 𝜆𝑘 𝑓𝑘−2 − 𝛿𝑘 𝑓𝑘−1 + 𝑓𝑘

Put 𝑘 = 2 ⟹ 𝑐2 = 𝜆2 𝜆2 𝑓0 − 𝛿2 𝑓1 + 𝑓2

= 1 1.1 − 2 −1.375 + −3

= 1 + 2.75 − 3 = 0.75
−2𝛿 𝑘 𝑓 𝑘
iv) 𝜆𝑘+1 =
𝑔 𝑘 ± 𝑔 𝑘 2 −4𝛿 𝑘 𝑓 𝑘 𝑐 𝑘

−2𝛿 2 𝑓2
Put 𝑘 = 2 ⟹ 𝜆3 =
𝑔2 ± 𝑔2 2 −4𝛿 2 𝑓2 𝑐2

−2𝛿2 −3
=
−2.5 − −2.5 2 − 4 × 2 × −3 (0.75)
12
=
−2.5 ± 24.25
12
= = −1.61629
−2.5 − 24.25
v) 𝑥𝑘+1 = 𝑥𝑘 + 𝜆𝑘+1 𝑥𝑘 − 𝑥𝑘−1

Put 𝑘 = 2 ⇒ 𝑥3 = 𝑥2 + 𝜆3 𝑥2 − 𝑥1

= 1 + 1 − 0.5 (−1.61629)

= 0.19185

𝑓3 = 0.04781

27
 Second iteration:
SPACE FOR HINTS i) 𝑕3 = 𝑥3 − 𝑥2

= 0.80815
𝑕3
ii) 𝜆3 =
𝑕2

0.80815
=− = −1.6163
0.5
𝛿3 = 1 + 𝜆3

= 1 − 1.6163 = −0.6163

iii) 𝑔3 = 𝜆3 2 𝑓1 − 𝛿3 2 𝑓2 + (𝜆3 + 𝛿3 )𝑓3

2 2
= −1.6163 −1.375 − −0.6163 −3
+ −1.375 − 0.6163 (0.04781)

= −2.5593

𝑐3 = 𝜆3 𝜆3 𝑓1 − 𝛿3 𝑓2 + 𝑓3

= −1.16163( −1.6163 −1.375 − −0.6163 −3

+ 0.0478

= −0.68098
−2𝛿 3 𝑓3
iv) 𝜆4 =
𝑔3 ± 𝑔3 2 −4𝛿 3 𝑓3 𝑐3

−2 −0.6163 (0.04781)
=
−2.5593 − (−2.5593)2 − 4 −0.6163 −0.68095 (0.04781)

0.058930
=
−2.5593 − 6.46975
0.058930
=
−5.10287
= −0.01154

v) 𝑥4 = 𝑥3 + 𝜆4 𝑥3 − 𝑥2
= 0.19185 + 0.19185 − 1 (−0.01154)
= 0.20118

28
𝑓 𝑥 = 0.002248
SPACE OF HINTS
Third iteration:

i) 𝑕4 = 𝑥4 − 𝑥3
= 0.20118 − 0.19185 = 0.00933
𝑕4
ii) 𝜆4 =
𝑕3

0.00933
= = −0.01154
−0.80815
𝛿4 = 1 + 𝜆4

= 1 − 0.01154 = 0.98846
iii) 𝑔4 = 𝜆4 2 𝑓2 − 𝛿4 2 𝑓3 + (𝜆4 + 𝛿4 )𝑓4
2 2
= −0.01154 −3 − 0.98846 0.04781
+ 0.98846 − 0.01154 0.002242

= −0.04

𝑐4 = 𝜆4 𝜆4 𝑓2 − 𝛿4 𝑓3 + 𝑓4

= −0.01154 −0.01154 −3 − 0.98846 0.04781

+ 0.002242 = 0.00012

−2𝛿4 𝑓4
𝜆5 =
𝑔4 ± 𝑔4 2 − 4𝛿4 𝑓4 𝑐4

−2 × 0.98846(0.002242)
=
−0.04492 − (−0.04492)2 − 4.098864 0.00012 (0.002242)
−0.004432
=
−0.04492 − 0.00201674
−0.004432
= = 0.04933
−0.089828
iv) 𝑥5 = 𝑥4 + 𝜆5 𝑥4 − 𝑥3
= 0.20118 + 0.20118 − 0.19185 0.04933
= 0.201640248

29
 Example1.4.2:
SPACE FOR HINTS Perform five iteration of the Muller Method to find the root of the
equation 𝑓 𝑥 = cos 𝑥 − 𝑥𝑒 𝑥

Solution:
Given that 𝑓 𝑥 = cos 𝑥 − 𝑥𝑒 𝑥

Put 𝑥 = 0 ; 𝑓 0 = 1

𝑥 = 1 ; 𝑓 1 = −2.1780

∴ the root lies between 0 and 1.

First iteration:

Let 𝑥0 = 0 , 𝑥1 = 0.5 , 𝑥2 = 1

𝑓0 = 1 , 𝑓1 = 0.0532 , 𝑓2 = −2.1780

i) 𝑕1 = 𝑥1 − 𝑥0 = 0.5
𝑕2 = 𝑥2 − 𝑥1 = 0.5
𝑕2
ii) 𝜆2 = =1
𝑕1

𝛿2 = 1 + 𝜆2 = 2
iii) 𝑔2 = 𝜆2 2 𝑓0 − 𝛿2 2 𝑓1 + (𝜆2 + 𝛿2 )𝑓2
2 2
= 1 1 − 2 0.0532 + 3 (−2.17801)
= −5.7468
𝑐2 = 𝜆2 𝜆2 𝑓0 − 𝛿2 𝑓1 + 𝑓2
= 1(1.1 − 2 0.0532 − 2.17801)
= −1.2844
−2𝛿 2 𝑓2
iv) 𝜆3 =
𝑔2 ± 𝑔2 2 −4𝛿 2 𝑓2 𝑐2

−2 2 (−2.1780)
=
−5.7468 − (−5.7468)2 − 4 × 2 × −2.1780 (−1.2844)
8.712
=
−5.7468 − 10.64632
8.712
= = −0.96789
−9.0009669

30
 v) 𝑥3 = 𝑥2 + 𝜆3 𝑥2 − 𝑥1
SPACE OF HINTS
= 1 + 0.5 (−0.96789)

= 0.516055

𝑓3 = 0.00503

Second iteration:

i) 𝑕3 = 𝑥3 − 𝑥2

= 0.5161 − 1 = −0.4839
𝑕3
ii) 𝜆3 =
𝑕2

−0.4839
= = −0.967
0.5
𝛿3 = 1 + 𝜆3 = 1 − 0.9678 = 0.0322

𝑔3 = 𝜆3 2 𝑓1 − 𝛿3 2 𝑓2 + (𝜆3 + 𝛿3 )𝑓3
2 2
= −0.9678 0.0532 − 0.0322 −2.1780
+ −0.9678 + 0.0322 0.00503

= 0.04738

𝑐3 = 𝜆3 𝜆3 𝑓1 − 𝛿3 𝑓2 + 𝑓3

= −0.9678( −0.9678 0.0532 − 0.0322 −2.1780

+ 0.00503)

= −0.0229
−2𝛿 3 𝑓3
iii) 𝜆4 =
𝑔3 ± 𝑔3 2 −4𝛿 3 𝑓3 𝑐3

−2 0.0322 (0.00503)
=
0.04738 + 0.04738 2 − 4 0.0322 0.00503 (−0.0229)
−0.000323
=
0.04738 + 0.0022597
−0.000323
= = −0.00340
0.094916
iv) 𝑥4 = 𝑥3 + 𝜆4 𝑥3 − 𝑥2

31
 = 0.5161 + 0.5161 − 1 −0.00340 = 0.51775
SPACE FOR HINTS
𝑓4 = 0.000224

Third iteration:

i) 𝑕4 = 𝑥4 − 𝑥3
= 0.51775 − 0.516055 = 0.00165
𝑕4
ii) 𝜆4 =
𝑕3

0.00165
= = −0.0034
−0.4839
𝛿4 = 1 + 𝜆4
= 1 − 0.0034 = 0.9966

iii) 𝑔4 = 𝜆4 2 𝑓2 − 𝛿4 2 𝑓3 + (𝜆4 + 𝛿4 )𝑓4

2 2
= −0.0034 −2.1780 − 0.9966 0.00503

+ −0.0034 + 0.9966 0.000224

= −0.00479

𝑐4 = 𝜆4 𝜆4 𝑓2 − 𝛿4 𝑓3 + 𝑓4

= −0.0034(−0.0034 −2.1780 − 0.9966 0.00503

+ 0.000224

= −0.0000088
−2𝛿 4 𝑓4
iv) 𝜆5 =
𝑔4 ± 𝑔4 2 −4𝛿 4 𝑓4 𝑐4

−2 0.9966 (0.000224)
=
−0.00479 − (−0.00479)2 − 4 0.9966 0.000224 (−0.0000088)
−0.000446
= = 0.0000465
−9.58000

v) 𝑥5 = 𝑥4 + 𝜆5 𝑥4 − 𝑥3

= 0.51775 + 0.51775 − 0.5161 0.0000465

= 0.51775

𝑓5 = 0.000224

32
Chebyshev Method:
SPACE OF HINTS
Let 𝑓 𝑥 = 𝑎0 𝑥 2 + 𝑎1 𝑥 + 𝑎2 = 0 (1)

Then 𝑓𝑘 = 𝑎0 𝑥𝑘 2 + 𝑎1 𝑥𝑘 + 𝑎2 (2)

𝑓𝑘 ′ = 2𝑎0 𝑥𝑘 + 𝑎1 (3)
𝑓 𝑘 ′′
𝑓𝑘 ′′ = 2𝑎0 ⇒𝑎0 = (4)
2

2𝑓𝑘 ′′ ′
3 ⟹ 𝑓𝑘 = 𝑥 + 𝑎1
2 𝑘
⟹ 𝑎1 = 𝑓𝑘 ′ − 𝑓𝑘 ′′ 𝑥𝑘 (5)

Substituting 𝑎0 , 𝑎1 in (2) we get

𝑓𝑘 ′′ 2
2 ⟹ 𝑓𝑘 = 𝑥𝑘 + 𝑓𝑘 ′ − 𝑓𝑘 ′′ 𝑥𝑘 𝑥𝑘 + 𝑎2
2
𝑓𝑘 ′′ 2
⟹ 𝑎2 = 𝑓𝑘 − 𝑥𝑘 + 𝑓𝑘 ′ − 𝑓𝑘 ′′ 𝑥𝑘 𝑥𝑘
2
𝑓𝑘 ′′ 2
= 𝑓𝑘 − 𝑥𝑘 − 𝑓𝑘 ′ 𝑥𝑘 + 𝑓𝑘 ′′ 𝑥𝑘 2
2

′𝑓𝑘 ′′ 𝑥𝑘 2
= 𝑓𝑘 − 𝑓𝑘 𝑥𝑘 +
2
Substituting 𝑎0 , 𝑎1 , 𝑎2 in (1) we get

1 ⟹𝑓 𝑥 =0
𝑓 𝑘 ′′ 𝑓 𝑘 ′′ 𝑥 𝑘 2
(i.e.) 𝑥 2 + 𝑓𝑘 ′ − 𝑓𝑘 ′′ 𝑥𝑘 𝑥 + 𝑓𝑘 − 𝑓𝑘 ′ 𝑥𝑘 + =0
2 2

𝑓𝑘 ′′ 2
⟹ 𝑥 − 2 𝑥𝑘 𝑥 + 𝑥𝑘 2 + 𝑓𝑘 𝑥 − 𝑥𝑘 + 𝑓𝑘 = 0
2
𝑓𝑘 ′′
⟹ (𝑥 − 𝑥𝑘 )2 + 𝑓𝑘 ′ 𝑥 − 𝑥𝑘 + 𝑓𝑘 = 0
2

′ 𝑓𝑘 ′′
𝑓𝑘 + 𝑓𝑘 𝑥 − 𝑥𝑘 + (𝑥 − 𝑥𝑘 )2 = 0
2
The above equation is the Taylor series expansion of 𝑓(𝑥) and higher
powers are neglected.

33
 𝑥 − 𝑥𝑘 (𝑥 − 𝑥𝑘 )2 ′′
∴ 𝑓𝑘 ′ = −𝑓𝑘 − 𝑓𝑘
SPACE FOR HINTS 1! 2!
−𝑓𝑘 (𝑥 − 𝑥𝑘 )2 ′′
⟹ 𝑥 − 𝑥𝑘 = ′ − 𝑓𝑘
𝑓𝑘 2

−𝑓𝑘 (𝑥 − 𝑥𝑘 )2 𝑓𝑘 ′′
⟹ 𝑥 − 𝑥𝑘 = ′ −
𝑓𝑘 2 𝑓𝑘 ′

Replace 𝑥 by 𝑥𝑘 +1 we get

−𝑓𝑘 𝑥𝑘+1 − 𝑥𝑘 2
𝑓𝑘 ′′
⟹ 𝑥𝑘+1 − 𝑥𝑘 = ′ − (6)
𝑓𝑘 2 𝑓𝑘 ′

By Newton-raphson method,
𝑓𝑘
𝑥𝑘+1 = 𝑥𝑘 −
𝑓𝑘 ′
𝑓𝑘
⟹ 𝑥𝑘+1 − 𝑥𝑘 = − (7)
𝑓𝑘 ′

Substituting (7) in (6) we get

−𝑓 𝑘 2
𝑓𝑘 ( 𝑓 𝑘 ′ ) 𝑓𝑘 ′′
𝑥𝑘+1 − 𝑥𝑘 = − ′ −
𝑓𝑘 2 𝑓𝑘 ′
2
𝑓𝑘 1 −𝑓𝑘 𝑓𝑘 ′′
∴ 𝑥𝑘+1 = 𝑥𝑘 − ′ − ( ′)
𝑓𝑘 2 𝑓𝑘 ′ 𝑓𝑘

Which is called the Chebyshev method.Thismethod requires three

equations for each iterations.

Example1.4.3:
Perform two iterations of the Chebyshev method to find the smallest
positive root of the equation 𝑓 𝑥 = 𝑥 3 − 5𝑥 + 1 = 0

Solution:
Given that 𝑓 𝑥 = 𝑥 3 − 5𝑥 + 1 = 0

Put 𝑥 = 0 ; 𝑓 0 = 1

𝑥 = 1 ; 𝑓 1 = −3

34
∴ The smallest root lies in the interval (0, 1).
SPACE OF HINTS
Take the initial approximation as 𝑥0 = 0.5

𝑓 𝑥 = 𝑥 3 − 5𝑥 + 1 ⟹ 𝑓 0.5 = −1.375 = 𝑓0

𝑓 ′ 𝑥 = 3𝑥 2 − 5 ⟹ 𝑓 ′ 0.5 = −4.25 = 𝑓0 ′

𝑓 ′′ 𝑥 = 6𝑥 ⟹ 𝑓 ′′ 0.5 = 3 = 𝑓0 ′′
The formula for Chebyshev Method is
2
𝑓𝑘 1 −𝑓𝑘 𝑓𝑘 ′′
𝑥𝑘+1 = 𝑥𝑘 − ′ − ( ′)
𝑓𝑘 2 𝑓𝑘 ′ 𝑓𝑘
First iteration:
𝑓0 1 𝑓 𝑓0 ′′
Put 𝑘 = 0 ⇒ 𝑥1 = 𝑥0 − − ( 0′ )2
𝑓0 ′ 2 𝑓0 𝑓0 ′

= 0.5 − 0.323529 − 0.052335(−0.70588)

𝑥1 = 0.213413

𝑓 𝑥1 = −0.05734

𝑓 ′ 𝑥1 = −4.88336

𝑓 ′′ 𝑥1 = 1.28047

Second iteration:
𝑓1 1 𝑓 𝑓1 ′′
Put 𝑘 = 1 ⇒ 𝑥2 = 𝑥1 − − ( 1′ )2
𝑓1 ′ 2 𝑓1 𝑓1 ′

(−0.05734) −0.05734 2 1.28049

= 0.213413 − − 0.5( )
−4.86336 4.86336 −4.86336
= 0.213413 − 0.011790 − 0.02780(−0.26328)

= 0.20164

Example1.4.4:
Perform two iteration of the Chebyshev method to find an
1
approximate value of .Take the initial approximation as 𝑥0 = 0.1
7

Solution:
1 1
Let 𝑥 = ⟹ 7 =
7 𝑥

35
 1
⟹ −7=0
SPACE FOR HINTS 𝑥
1
𝑓 𝑥 = −7=0
𝑥
−1
𝑓′ 𝑥 =
𝑥2
2
𝑓 ′′ 𝑥 =
𝑥3
Given initial approximation 𝑥0 = 0.1
1
𝑓 𝑥 = −7 ⟹ 𝑓 0.1 = 3
𝑥
−1
𝑓′ 𝑥 = ⟹ 𝑓 ′ 0.1 = −100
𝑥2
2
𝑓 ′′ 𝑥 = ⟹ 𝑓 ′′ 0.1 = 2000
𝑥3
The general formula for Chebyshev method is
2
𝑓𝑘 1 −𝑓𝑘 𝑓𝑘 ′′
𝑥𝑘+1 = 𝑥𝑘 − ′ − ( ′)
𝑓𝑘 2 𝑓𝑘 ′ 𝑓𝑘

First iteration:
𝑓0 1 𝑓 𝑓0 ′′
Put 𝑘 = 0 ⇒ 𝑥1 = 𝑥0 − − ( 0′ )2
𝑓0 ′ 2 𝑓0 𝑓0 ′

2
3 1 3 2000
= 0.1 − − ( )
100 2 −100 −100
= 0.1 + 0.03 + 0.00045 × 20 = 0.139

𝑓 0.139 = 0.19424

𝑓 ′ 0.139 = −51.75716

𝑓 ′′ 0.139 = 749.70727
𝑓1 1 𝑓 𝑓1 ′′
Second iteration: Put 𝑘 = 1 ⇒ 𝑥2 = 𝑥1 − − ( 1′ )2
𝑓1 ′ 2 𝑓1 𝑓1 ′

= 0.139 + 0.0037529 − 7.04212(9.615935)

= 0.14285

36
Example1.4.5:
SPACE OF HINTS
Using ChebyshevMethod find the root of the equation

𝑓 𝑥 = cos 𝑥 − 𝑥𝑒 𝑥 = 0

Solution:

Given 𝑓 𝑥 = cos 𝑥 − 𝑥𝑒 𝑥 = 0

𝑓 ′ 𝑥 = − sin 𝑥 − (𝑥 + 1) 𝑒 𝑥

𝑓 ′′ 𝑥 = − cos 𝑥 − (𝑥 + 2)𝑒 𝑥

Put 𝑥 = 0 ⟹ 𝑓 0 = 1

𝑥 = 1 ⟹ 𝑓 1 = −2.1780

∴The root lies between the interval 0, 1 .

Take the initial approximation 𝑥0 = 0.1

𝑓 𝑥 = cos 𝑥 − 𝑥𝑒 𝑥 = 0

𝑓 ′ 𝑥 = − sin 𝑥 − 𝑥𝑒 𝑥 − 1 𝑒 𝑥

𝑓 ′′ 𝑥 = − cos 𝑥 − 𝑥𝑒 𝑥 + 2𝑒 𝑥

𝑓 1 = −2.177980 𝑓 ′ 1 = −6.278035 𝑓 ′′ 1 = −8.695148

The formula for Chebyshev Method is

2
𝑓𝑘 1 −𝑓𝑘 𝑓𝑘 ′′
𝑥𝑘+1 = 𝑥𝑘 − ′ − ( ′)
𝑓𝑘 2 𝑓𝑘 ′ 𝑓𝑘

First iteration:
𝑓0 1 𝑓 𝑓0 ′′
Put 𝑘 = 0 ⇒ 𝑥1 = 𝑥0 − − ( 0′ )2
𝑓0 ′ 2 𝑓0 𝑓0 ′

= 1 − 0.346920 − 0.060176(1.385011)

= 0.569735

𝑓 𝑥1 = −0.165132

𝑓 ′ 𝑥1 = −3.314384

𝑓 ′′ 𝑥1 = −5.384818

37
 Second iteration:
SPACE FOR HINTS 𝑓1 1 𝑓 𝑓1 ′′
Put 𝑘 = 1 ⇒ 𝑥2 = 𝑥1 − − ( 1′ )2
𝑓1 ′ 2 𝑓1 𝑓1 ′

= 0.569735 − 0.049822 − 0.00201622

= 0.517896

Problem:

The equation 𝑓 𝑥 = 𝑥 4 − 𝑥 − 10 = 0 has a root in the interval

(1,2). Take the initial approximation as 𝑥0 = 1.5. Performthree iteration
of the Chebyshev method.

Solution:
Given that 𝑓 𝑥 = 𝑥 4 − 𝑥 − 10 = 0

Given that the root lies in the interval 1,2 .

Take the initial approximation as 𝑥0 = 1.5

𝑓 𝑥 = 𝑥 4 − 𝑥 − 10 ⟹ 𝑓 1.5 = −6.4375

𝑓 ′ 𝑥 = 4𝑥 3 − 1 ⟹ 𝑓 ′ 1.5 = 12.5

𝑓 ′′ 𝑥 = 12𝑥 2 ⟹ 𝑓 ′′ 1.5 = 27

The formula for Chebyshev Method is

2
𝑓𝑘 1 −𝑓𝑘 𝑓𝑘 ′′
𝑥𝑘+1 = 𝑥𝑘 − ′ − ( ′)
𝑓𝑘 2 𝑓𝑘 ′ 𝑓𝑘
First iteration:
𝑓0 1 𝑓 𝑓0 ′′
Put 𝑘 = 0 ⇒ 𝑥1 = 𝑥0 − ′ − ( 0′ )2
𝑓0 2 𝑓0 𝑓0 ′

= 1.5 + 0.515 − (0.1326)(2.16)

𝑥1 = 1.7286

𝑓 𝑥1 = −2.8001

𝑓 ′ 𝑥1 = 19.6606

𝑓 ′′ 𝑥1 = 35.8567

38
Second iteration:
SPACE OF HINTS
𝑓1 1 𝑓 𝑓 ′′
Put 𝑘 = 1 ⇒ 𝑥2 = 𝑥1 − − ( 1′ )2 1 ′
𝑓1 ′ 2 𝑓1 𝑓1

= 1.7286 + 0.1424 − (0.0101)(1.8238)

= 1.8526

𝑓 𝑥2 = −0.0731

𝑓 ′ 𝑥2 = 24.4334

𝑓 ′′ 𝑥2 = 41.1855

Third iteration:
𝑓2 1 𝑓 𝑓 ′′
Put 𝑘 = 2 ⟹ 𝑥3 = 𝑥2 − − ( 2′ )2 ( 2 ′ )
𝑓2 ′ 2 𝑓2 𝑓2

= 1.8526 + 0.00299 − (0.000004)(1.6856)

= 1.8556

Multipoint iteration Method:

We Know that 𝑓𝑘 = 𝑎0 𝑥𝑘 2 + 𝑎1 𝑥𝑘 + 𝑎2

𝑓𝑘 ′ = 2𝑎0 𝑥𝑘 + 𝑎1

𝑓𝑘 ′′ = 2𝑎0

𝑓𝑘 ′ 𝑓𝑘 ′′ 2
𝑓𝑘 + 𝑥 − 𝑥𝑘 + 𝑥 − 𝑥𝑘 =0 (1)
1! 2

𝑓𝑘 ′ 𝑓𝑘 ′′
⟹ 𝑓𝑘 𝑥 − 𝑥𝑘 + 𝑥 − 𝑥𝑘 =0
1! 2

𝑓𝑘 ′ 𝑓𝑘 ′′
⟹ 𝑥 − 𝑥𝑘 + 𝑥 − 𝑥𝑘 = −𝑓𝑘
1! 2

−𝑓𝑘
⟹ 𝑥 − 𝑥𝑘 = ′
𝑓𝑘 𝑓 𝑘 ′′
+ 𝑥 − 𝑥𝑘
1! 2!

Now replace 𝑥 by 𝑥𝑘+1 ,

−𝑓𝑘
𝑥𝑘+1 − 𝑥𝑘 =
𝑓𝑘 ′ 𝑓 𝑘 ′′ 𝑥 𝑘+1 −𝑥 𝑘
+
1! 2!

39
 −𝑓𝑘
= 1 (2)
SPACE FOR HINTS 𝑓 ′ (𝑥𝑘 + 𝑥𝑘+1 − 𝑥𝑘 )
2

By Newton-Raphson Method,
𝑓𝑘
𝑥𝑘+1 = 𝑥𝑘 −
𝑓𝑘 ′
𝑓𝑘
⟹ 𝑥𝑘+1 − 𝑥𝑘 = − (3)
𝑓𝑘 ′

Substituting (3)in (2) on R.H.S, we get

−𝑓𝑘
𝑥𝑘+1 − 𝑥𝑘 = 1 𝑓𝑘
𝑘 = 0,1,2, ⋯
𝑓 ′ 𝑥𝑘 + ′
2 𝑓𝑘

𝑓𝑘
⟹ 𝑥𝑘+1 = 𝑥𝑘 − 1 𝑓𝑘
𝑓 ′ 𝑥𝑘 +
2 𝑓𝑘 ′

Which is called multipoint iteration method.

For computation purpose, this method havethe two-stage method

1 𝑓𝑘
(1) 𝑥 ∗ 𝑘+1 = 𝑥𝑘 − 𝐼
2 𝑓𝑘 ′
𝑓𝑘
(2) 𝑥𝑘+1 = 𝑥𝑘 − ′ ∗
𝐼𝐼
𝑓 (𝑥 𝑘+1 )

𝑓𝑘
(i.e.) 𝑥𝑘+1 = 𝑥𝑘 − , 𝑘 = 0,1,2, ⋯
𝑓 ′ ∗ 𝑥 𝑘+1

Example1.4.6:
Perform three iterations of the multipoint iteration method to
findthe smallest positive root of the equation 𝑓 𝑥 = 𝑥 3 − 5𝑥 + 1 = 0

Solution:
The smallest positive root of the equation is find as follows

Put 𝑥 = 0 ⟹ 𝑓 0 = 1

𝑥 = 1 ⟹ 𝑓 1 = −3

∴The smallest +veroot lies in the interval (0,1).

Take the initial approximation 𝑥0 = 0.5

40
General formula for multipoint iteration method is
SPACE OF HINTS
1 𝑓𝑘
𝑥 ∗ 𝑘+1 = 𝑥𝑘 −
2 𝑓𝑘 ′
𝑓𝑘
𝑥𝑘+1 = 𝑥𝑘 −
𝑓 ′ ∗ (𝑥𝑘+1 )

𝑓 𝑥 = 𝑥 3 − 5𝑥 + 1

𝑓 ′ 𝑥 = 3𝑥 2 − 5

First iteration:
1 𝑓0
Take 𝑘 = 0 ⇒ 𝑥1 ∗ = 𝑥0 −
2 𝑓0 ′

1 −1.375
= 0.5 −
2 −4.25
= 0.5 − 0.161764 = 0.33824

𝑓 ′ 𝑥1 ∗ = 𝑓 ′ (0.33824)

= −4.65678
𝑓0
𝑥1 = 𝑥0 −
𝑓1 ′ ∗
−1.375
= 0.5 − = 0.2
−4.65678
𝑓 𝑥1 = −0.01507

𝑓 ′ 𝑥1 = −4.87426

Second iteration:
1 𝑓1
𝑘 = 1 ⇒ 𝑥2 ∗ = 𝑥1 −
2 𝑓1 ′
1 −0.01507
= 0.20473 −
2 −4.87426
= 0.20473 − 0.0015458 = 0.20318

𝑓 ′ 𝑥2 ∗ = 𝑓 ′ (0.20318)

= −4.87615

41
 𝑓1
𝑥2 = 𝑥1 −
SPACE FOR HINTS 𝑓2 ′ ∗
0.01507
= 0.20473 +
−4.87615
= 0.20164

𝑓 𝑥2 = −0.000001581

𝑓 ′ 𝑥2 = −4.87802

Third iteration:
1 𝑓2
𝑘 = 2 ⇒ 𝑥3 ∗ = 𝑥2 −
2 𝑓2 ′
1 0.000001581
= 0.20164 − (− )
2 −4.87802
= 0.20164 − 0.000000162

= 0.20164

𝑓 ′ 𝑥3 ∗ = −4.87802
𝑓2
𝑥3 = 𝑥2 −
𝑓3 ′ ∗

= 0.20164

Example1.4.7:
Perform three iterations of the multipoint iteration method to find the
root of the equation 𝑓 𝑥 = cos 𝑥 − 𝑥𝑒 𝑥 = 0

Solution:
Put 𝑥 = 0 ⟹ 𝑓 0 = 1

𝑥 = 1 ⟹ 𝑓 1 = −2.1780

∴The root lies between the interval (0,1)

Take the initial approximation 𝑥0 = 0.5

General formula for multipoint iteration method is

42
 1 𝑓𝑘
𝑥 ∗ 𝑘+1 = 𝑥𝑘 − SPACE OF HINTS
2 𝑓𝑘 ′
𝑓𝑘
𝑥𝑘+1 = 𝑥𝑘 − ′ ∗
𝑓 (𝑥𝑘+1 )

𝑓 𝑥 = cos 𝑥 − 𝑥𝑒 𝑥

𝑓 ′ 𝑥 = − sin 𝑥 − 𝑥 𝑒 𝑥 − 𝑒 𝑥

= − sin 𝑥 − (𝑥 + 1) 𝑒 𝑥

First iteration:
1 𝑓0
Take 𝑘 = 0 ⇒ 𝑥1 ∗ = 𝑥0 −
2 𝑓0 ′

1 0.05322
= 0.5 −
2 −2.9525
= 0.5 + 0.009013 = 0.509013

𝑓 ′ 𝑥1 ∗ = 𝑓 ′ 0.509013

= −2.997783
𝑓0
𝑥1 = 𝑥0 −
𝑓1 ′ ∗
0.05322
= 0.5 −
−2.997783
= 0.51775

𝑓 𝑥1 = 0.000022

𝑓 ′ 𝑥1 = −3.04209

Second iteration:
1 𝑓1
𝑘 = 1 ⟹ 𝑥2 ∗ = 𝑥1 −
2 𝑓1 ′
1 0.00022
= 0.51775 − ( )
2 −3.04209
= 0.51775 + 0.0000072 = 0.51775

𝑓 ′ 𝑥2 ∗ = 𝑓 ′ (0.51775)

= −3.04209

43
 𝑓1
𝑥2 = 𝑥1 −
SPACE FOR HINTS 𝑓2 ′ ∗
0.000022
= 0.51775 −
−3.04209
= 0.51775 + 0.0000072 = 0.51775

∴The root of the equation 𝑥 = 0.51775.

1.5: Methods for Complex roots

The root of an equation 𝑓 𝑥 = 0 in which 𝑍 is complex variable.

Substituting 𝑍 = 𝑥 + 𝑖𝑦 in the above equation we get

𝑓 𝑥 + 𝑖𝑦 = 𝑢 𝑥, 𝑦 + 𝑖𝑣 𝑥, 𝑦 = 0

Where 𝑢(𝑥, 𝑦)and 𝑣(𝑥, 𝑦)are the real and imaginary parts of
𝑓(𝑥)respectively.

Comparing real and imaginary parts,we get 𝑢 𝑥, 𝑦 = 0 & 𝑣 𝑥, 𝑦 = 0

The formula for Complex method is

𝑥𝑘+1 𝑥𝑘 −1 𝑢𝑘
𝑦𝑘+1 + 𝑦𝑘 − 𝐽𝑘 𝑣𝑘
𝑢𝑥 𝑢𝑦 𝑘
Now,𝐽𝑘 = 𝑣 𝑘 𝑣𝑦 𝑘
𝑥𝑘

1 𝑢𝑥 𝑘 −𝑢𝑦 𝑘
𝐽𝑘 −1 =
𝐽𝑘 −𝑣𝑥 𝑘 𝑣𝑦 𝑘

Example1.5.1:
Obtain the Complex root of the equation 𝑓 𝑧 = 𝑍 3 + 1 = 0.Correct to
eight decimal places use the initial approximation to a root as 𝑥0 , 𝑦0 =
1+𝑖 3
0.25, 0.25 comparewith the exact values of the roots
2

Solution:

Given𝑓 𝑧 = 𝑍 3 + 1 = 0

Substituting 𝑧 = 𝑥 + 𝑖𝑦 we get

𝑓 𝑥 + 𝑖𝑦 = (𝑥 + 𝑖𝑦)3 + 1

44
 𝑢 𝑥, 𝑦 + 𝑖𝑣 𝑥, 𝑦 = 𝑥 3 + 3𝑥 2 𝑖𝑦 + 3𝑥(𝑖𝑦)2 + (𝑖𝑦)3 + 1
SPACE OF HINTS
3 2 2 3
= 𝑥 + 3𝑥 𝑖𝑦 − 3𝑥𝑦 − 𝑖𝑦 + 1

= 𝑥 3 − 3𝑥𝑦 2 + 1 + 𝑖(3𝑥 2 𝑦 − 𝑦 3 )

Equating real and imaginary parts on both sides, 𝑢 = 𝑥 3 + 3𝑥𝑦 2 +

1;𝑣 = 3𝑥 2 𝑦 − 𝑦 3
𝑢𝑘
To find 𝑣 :
𝑘

𝑢𝑘 𝑥 3 − 3𝑥𝑘 𝑦𝑘 2 + 1
(i.e.) 𝑣 = 𝑘
𝑘 3𝑥𝑘 2 𝑦𝑘 − 𝑦𝑘 3
𝑢𝑥 𝑢𝑦 𝑘 3𝑥𝑘 2 − 3𝑦𝑘 2 −6𝑥𝑘 𝑦𝑘
𝐽𝑘 = 𝑣 𝑘 𝑣𝑦 𝑘 =
𝑥𝑘 6𝑥𝑘 𝑦𝑘 3𝑥𝑘 2 − 3𝑦𝑘 2

3𝑥𝑘 2 − 3𝑦𝑘 2 −6𝑥𝑘 𝑦𝑘

𝐽𝑘 =
6𝑥𝑘 𝑦𝑘 3𝑥𝑘 2 − 3𝑦𝑘 2

= (3𝑥𝑘 2 − 3𝑦𝑘 2 )2 + 36𝑥𝑘 2 𝑦𝑘 2

= 9𝑥𝑘 4 + 9𝑦𝑘 4 − 18𝑥𝑘 2 𝑦𝑘 2 + 36𝑥𝑘 2 𝑦𝑘 2

= 9𝑥𝑘 4 + 9𝑦𝑘 4 + 18𝑥𝑘 2 𝑦𝑘 2 = 9 𝑥𝑘 4 + 𝑦𝑘 4 + 2𝑥𝑘 2 𝑦𝑘 2

𝐽𝑘 = 9(𝑥𝑘 2 + 𝑦𝑘 2 )2

−1 1 3𝑥𝑘 2 − 3𝑦𝑘 2 6𝑥𝑘 𝑦𝑘

𝐽𝑘 =
𝐽𝑘 −6𝑥𝑘 𝑦𝑘 3𝑥𝑘 2 − 3𝑦𝑘 2

−1 1 3𝑥𝑘 2 − 3𝑦𝑘 2 6𝑥𝑘 𝑦𝑘

∴ 𝐽𝑘 =
9 𝑥𝑘 2 + 𝑦𝑘 2 2 −6𝑥𝑘 𝑦𝑘 3𝑥𝑘 2 − 3𝑦𝑘 2

∴The general formula for Methods for Complex root is,

𝑥𝑘+1 𝑥𝑘 −1 𝑢𝑘
𝑦𝑘+1 = 𝑦𝑘 − 𝐽𝑘 𝑣𝑘

First iteration:
𝑥1 𝑥0 𝑢0
Put 𝑘 = 0 ⇒ 𝑦 = 𝑦 − 𝐽0 −1 𝑣
1 0 0

Given initial approximation is 𝑥0 , 𝑦0 = (0.25,0.25)

𝐽𝑘 = 9(𝑥𝑘 2 + 𝑦𝑘 2 )2 =0.140625

45
 1 3𝑥𝑘 2 − 3𝑦𝑘 2 6𝑥𝑘 𝑦𝑘
𝐽𝑘 −1 =
SPACE FOR HINTS 𝐽𝑘 −6𝑥𝑘 𝑦𝑘 3𝑥𝑘 2 − 3𝑦𝑘 2
0 2.667
𝐽𝑘 −1 =
−2.667 0
𝑢𝑘 𝑥𝑘 3 − 3𝑥𝑘 𝑦𝑘 2 + 1
𝑣𝑘 =
3𝑥𝑘 2 − 𝑦𝑘 3
𝑢𝑘 0.96875
𝑣𝑘 = 0.03125
𝑥𝑘+1 𝑥𝑘 −1 𝑢𝑘
𝑦𝑘+1 = 𝑦𝑘 − 𝐽𝑘 𝑣𝑘
0.1667
=
2.8334
Second iteration:

𝑥1 , 𝑦1 = (0.1667,2.8334)

Put 𝑘 = 1, 𝐽𝑘 = 9(𝑥1 2 + 𝑦1 2 )2 = 9(0.16672 + 2.83342 )2

= 584.0842
1
𝐽1 −1 =
584.0842
3 0.1667 2 − 3 2.8334 2
6 0.1667 2.8334
−6 0.1667 2.8334 3 0.1667 2 − 3 2.8334 2

1 −0.04109 0.00485
=
584.0842 −0.00485 −0.04109
−0.04109 0.00485
=
−0.00485 −0.04109
𝑢1 𝑥1 3 − 3𝑥1 𝑦1 2 + 1
𝑣1 = 3𝑥1 2 𝑦1 − 𝑦1 3
−3.01025
=
−22.5108
𝑥2 𝑥1 𝑢1
𝑦2 = 𝑦 − 𝐽1 −1 𝑣
1 1

0.1667 −0.04109 0.00485 −3.01025

= −
2.8334 −0.00485 −0.04109 −22.5108
𝑥2 0.1522
𝑦2 = 1.8938

46
Third iteration:
SPACE OF HINTS
𝑥2 , 𝑦2 = (0.1522, 1.8938)
𝑥3 𝑥2 𝑢2
Put 𝑘 = 2, 𝑦 = 𝑦 − 𝐽2 −1 𝑣
3 2 2

1 3𝑥2 2 − 3𝑦2 2 6𝑥2 𝑦2

𝐽2 −1 =
𝐽2 −6𝑥2 𝑦2 3𝑥2 2 − 3𝑦2 2

𝐽2 = 9((0.1522)2 + (1.8931)2 )2

= 117.2655
1 −10.6899 1.7293
𝐽2 −1 = [ ]
117.2655 −1.7293 −10.6899
−0.0912 0.01475
=
−0.01475 −0.0912
𝑢2 𝑥2 3 − 3𝑥2 𝑦2 2 + 1
𝑣2 =
3𝑥2 2 𝑦2 − 𝑦2 3
0.63395
=
−6.6605
𝑥3 0.1522 −0.0912 0.01475
𝑦3 = 1.8938 − −0.01475 −0.0912
0.19269
=
1.2772
Fourth iteration: 𝑥3 , 𝑦3 = (0.19269, 1.2772)
𝑥4 𝑥3 𝑢3
Put 𝑘 = 3, 𝑦 = 𝑦 − 𝐽3 −1 𝑣
4 3 3

𝐽𝑘 = 9(0.19269)2 + (1.2772)2 )2

= 25.05377
1 −4.78272 1.476257
𝐽3 −1 =
25.05377 1.476257 −4.78272
−0.19089 0.058923
=
0.058923 −0.19089
𝑢3 0.06438
𝑣3 = −1.94144
𝑥4 0.31932
𝑦4 =
0.91041

47
 Fifth iteration:
SPACE FOR HINTS 𝑥4 , 𝑦4 = (0.31932,0.91041)
𝑥5 𝑥4 −1 𝑢4
Put 𝑘 = 4, 𝑦5 = 𝑦4 − 𝐽4 𝑣4

𝐽4 = 7.79769
1 −2.18067 1.74422
𝐽4 −1 =
7.79769 −1.74422 −2.18067
𝑢4 𝑥4 3 − 3𝑥4 𝑦4 2 + 1
𝑣4 =
3𝑥4 2 𝑦4 − 𝑦4 3
0.23854
=
0.47612
𝑥5 0.31932 −0.17322
𝑦5 = −0.91041 − 0.07978
0.49253
=
0.83063
Sixth iteration:
𝑥6 𝑥5 −1 𝑢5
Put 𝑘 = 5, 𝑦6 = 𝑦5 − 𝐽5 𝑣5
1 −1.342081 2.45466
𝐽5 −1 =
7.82654 −2.45466 −1.342081
−0.17148 0.31363
=
−0.31363 −0.17148
𝑢5 𝑥5 3 − 3𝑥5 𝑦5 2 + 1
𝑣5 =
3𝑥5 2 𝑦5 − 𝑦5 3
0.100023
=
0.03141
𝑥6 0.49253 −0.0073001 0.499833
𝑦6 = 0.83063 − 0.020337 = 0.86738

Seventh iteration:

𝑥6 , 𝑦6 = (0.499833,0.86738)
𝑥7 𝑥6 −1 𝑢6
Put 𝑘 = 6, 𝑦7 = 𝑦6 − 𝐽6 𝑣6
1 −1.50755 2.601270
𝐽6 −1 =
9.03948 −2.601270 −1.50755

48
 0.16677 0.287767
= SPACE OF HINTS
−0.287767 0.16677
𝑥7 0.49983 0.000545 + 0.000713 0.498572
𝑦7 = 0.86738 − 0.000939 + 0.000412 = 0.866028

1.6: Polynomial Equation:

The root of the polynomial equation of degree 𝑛 is given by

𝑃𝑛 𝑥 = 𝑎0 𝑥 𝑛 + 𝑎1 𝑥 𝑛−1 + ⋯ + 𝑎𝑛−1 𝑥 + 𝑎𝑛 = 0 where 𝑎0 ≠ 0 and

𝑎0 , 𝑎1 , ⋯ , 𝑎𝑛 are real Numbers.

Definition:
When in a polynomial (terms being written in order) a +sign follows a
+sign or –sign follows a –sign continuous or a permanence of signs is
said to occur.But if +sign follows a –sign or a –sign follows a +sign,
then change of sign is said to occur.

Example:

Consider the polynomial equation 𝑃5 𝑥 = 8𝑥 5 + 12𝑥 4 − 10𝑥 3 +

17𝑥 2 − 18𝑥 + 5 = 0

The above equation has four sign changes and one continuous of signs.

Desearte’s Rule of Signs:

The number ofpositive real roots of 𝑃𝑛 𝑥 = 0 cannot exceed the

number of sign changes in𝑃𝑛 𝑥 and the number of negative real root of
𝑃𝑛 𝑥 = 0 cannot exceed the number of signchanges in 𝑃𝑛 −𝑥

Example 1.6.1:
Consider the Polynomial equation

𝑃5 𝑥 = 8𝑥 5 + 12𝑥 4 − 10𝑥 3 + 17𝑥 2 − 18𝑥 + 5 = 0

The number of sign changes in 𝑃5 𝑥 = 4

𝑃5 −𝑥 = −8𝑥 5 + 12𝑥 4 + 10𝑥 3 + 17𝑥 2 + 18𝑥 + 5 = 0

The number of sign changes in 𝑃5 −𝑥 = 1

∴The number of positive real root of 𝑃5 𝑥 = 0 cannot exceed 4.

49
 Similarlythe number of sign changes of 𝑃5 −𝑥 = 1
SPACE FOR HINTS Hence the number of negative real root of 𝑃5 𝑥 = 0 cannot exceed 1.

Sturm’s function (or) Sturm Sequence:

Let 𝑓(𝑥) be a given polynomial of degree 𝑛 and let 𝑓1 (𝑥)
represent its first order

derivate. Denote by 𝑓2 (𝑥) the reminder of 𝑓(𝑥) divided by 𝑓1 (𝑥) taken

with the reverse sign and 𝑓3 (𝑥) for the reminder of 𝑓1 (𝑥) divided by
𝑓2 (𝑥) with the reverse sign and so on until a constantis arrived.
𝑓 𝑥
(i.e.) 𝑓2 𝑥 = − 𝑟𝑒𝑚𝑖𝑛𝑑𝑒𝑟 𝑜𝑓
𝑓1 𝑥

𝑓1 𝑥
𝑓3 𝑥 = − 𝑟𝑒𝑚𝑖𝑛𝑑𝑒𝑟 𝑜𝑓
𝑓2 𝑥

∴we obtain a sequence of function 𝑓 𝑥 , 𝑓1 𝑥 , 𝑓2 (𝑥) is called the

Sturm function or Sturm sequence.

Theorem1.6.2: (Sturm theorem)

The number of rootof the equation 𝑓 𝑥 = 0 on [𝑎, 𝑏] equals the
difference between the number of changes of sign in the Sturm sequence
at 𝑥 = 𝑎 and 𝑥 = 𝑏 provided that 𝑓(𝑎) ≠ 0, 𝑓(𝑏) ≠ 0.

Example1.6.3:
Find the number of root of complex roots of the polynomial equations

(i) 𝑃3 𝑥 = 𝑥 3 − 5𝑥 + 1 = 0
(ii) 𝑃4 𝑥 = 4𝑥 4 + 2𝑥 2 − 1 = 0 using Sturm method.

Solution:

(i) The Given equation is 𝑓 𝑥 = 𝑥 3 − 5𝑥 + 1 = 0 = 0

𝑓 ′ 𝑥 = 3𝑥 2 − 5 𝑥 3
𝑓(𝑥)
Let 𝑓1 𝑥 = 𝑓 ′ 𝑥 = 3𝑥 2 − 5 3𝑥 2 − 5 𝑥 3 − 5𝑥 + 1 =
𝑓1 (𝑥)

𝑥 3 −5𝑥+1 5𝑥
𝑥3 −
3𝑥 2 −5 3

50
 −10𝑥
+1 SPACE OF HINTS
3
𝑓 𝑥
𝑓2 𝑥 = − 𝑟𝑒𝑚𝑖𝑛𝑑𝑒𝑟 𝑜𝑓
𝑓1 𝑥
10𝑥 10𝑥
=− − +1 = −1
3 3
10𝑥 − 3
= =0
3
∴ 𝑓2 𝑥 = 10𝑥 − 3

𝑓1 𝑥 3𝑥 9
𝑓3 𝑥 = − 𝑟𝑒𝑚𝑖𝑛𝑑𝑒𝑟 𝑜𝑓 +
𝑓2 𝑥 10 100
−473 473
=− = =0 10𝑥 − 3 3𝑥 2 − 5
100 100

9𝑥
𝑓3 𝑥 = 473 3𝑥 2 −
10

9𝑥
Let 𝑉(𝑎) denote the number of sign changes in the Sturms −5
10

9𝑥 27
equation at 𝑥 = 𝑎. −
10 100

27 473
The following table, is the construction of sign changes −5=
100 100

in 𝑉(𝑎).

𝑥=𝑎 𝑓(𝑥) 𝑓1 (𝑥) 𝑓2 (𝑥) 𝑓3 (𝑥) 𝑉(𝑥)

−∞ − + − + 3
−3 − + − + 3
−2 + + − + 2
−1 + − − + 2
0 + − − + 2
1 − − + + 1
2 − + + + 1
3 + + + + 0
∞ + + + + 0
∴We find that from the above table, the polynomial has 3 real roots
lying the interval (−3, −2), 0,1 , & (2,3) .

51
 ∴The given polynomial has no complex root.
SPACE FOR HINTS 𝑥

4𝑥 3 + 𝑥 4𝑥 4 + 2𝑥 2 − 1

𝑓 ′ 𝑥 = 16 + 4𝑥 = 𝑓1 𝑥 4𝑥 4 + 𝑥 2

⟹ 𝑓1 𝑥 = 4𝑥 3 + 4𝑥 2 − 1
𝑓 𝑥
Now, 𝑓2 𝑥 = − 𝑟𝑒𝑚𝑖𝑛𝑑𝑒𝑟 𝑜𝑓
𝑓1 𝑥

= − 𝑥 2 − 1 = −𝑥 2 + 1
𝑓1 𝑥
𝑓3 𝑥 = − 𝑟𝑒𝑚𝑖𝑛𝑑𝑒𝑟 𝑜𝑓 − 4𝑥
𝑓2 𝑥

= − 5𝑥 = −5𝑥 = 0 − 𝑥 2 + 1 4𝑥 3 + 𝑥

⇒ −𝑥 = 0 4𝑥 3 − 4𝑥

𝑓2 𝑥
𝑓4 𝑥 = − 𝑟𝑒𝑚𝑖𝑛𝑑𝑒𝑟 𝑜𝑓 5𝑥
𝑓3 𝑥

= − 1 = −1

Now let V(a) denote the number of changes in the Sturm‟s sequence at
𝑥 = 𝑎.

𝑥=𝑎 𝑓(𝑥) 𝑓1 (𝑥) 𝑓2 (𝑥) 𝑓3 (𝑥) 𝑓4 (𝑥) 𝑉(𝑎)

−∞ + − − + − 3
−2 + − − + − 3
−1 + − 0 + − 3
0 − 0 + 0 − 2
1 + + 0 − − 1
2 + + − − − 1
∞ + + − − − 1
From the above table, we find that the given polynomial has two real
roots lies in intervals (−1,0) and (0,1).

The polynomial has no complex root.

Example1.6.4:

52
 Find the number of real roots and complex roots of the
SPACE OF HINTS
polynomial 𝑃4 𝑥 = 𝑥 4 − 4𝑥 3 + 3𝑥 2 + 4𝑥 − 4 using Sturm sequence.
Solution:Given equation is 𝑓 𝑥 = 𝑥 4 − 4𝑥 3 + 3𝑥 2 + 4𝑥 − 4
𝑓 ′ 𝑥 = 4𝑥 3 − 12𝑥 2 + 6𝑥 + 4
⇒ 𝑓1 𝑥 = 2𝑥 3 − 6𝑥 2 + 3𝑥 + 2
𝑓 𝑥
Now, 𝑓2 𝑥 = − 𝑟𝑒𝑚𝑖𝑛𝑑𝑒𝑟 𝑜𝑓
𝑓1 𝑥

3𝑥 2 9𝑥 𝑥 1
=− − + −3 +
2 2 2 2

2𝑥 3 − 6𝑥 2 + 3𝑥 + 2 𝑥 4 − 4𝑥 3 + 3𝑥 2 + 4𝑥 − 4
3𝑥 2
= 3𝑥 2 − 9𝑥 + 6 𝑥 4 − 3𝑥 3 + +𝑥
2

2 3
3𝑥 2
= 𝑥 − 3𝑥 + 2 −𝑥 + + 3𝑥 − 4
2
𝑓1 𝑥 3𝑥
𝑓3 𝑥 = − 𝑟𝑒𝑚𝑖𝑛𝑑𝑒𝑟 𝑜𝑓 𝑥 3 − 3𝑥 2 + +1
𝑓2 𝑥 2

3𝑥 2 9𝑥
= − −𝑥 + 2 = 𝑥 − 2 − + −3
2 2
2𝑥

𝑥 2 − 3𝑥 + 2 2 𝑥 3 − 6𝑥 2 + 3𝑥 + 2

𝑓2 𝑥
𝑓4 𝑥 = − 𝑟𝑒𝑚𝑖𝑛𝑑𝑒𝑟 𝑜𝑓 2𝑥 3 − 6𝑥 2 + 4𝑥
𝑓3 𝑥

=0 − 𝑥 + 2𝑥 − 1

𝑥 − 2 𝑥 2 − 3𝑥 + 2

𝑥 2 − 2𝑥

−𝑥 + 2

−𝑥 + 2
0

When we divide 𝑓2 (𝑥) by 𝑓3 (𝑥) we get zero as the reminder.

53
 ∴ 𝑓3 𝑥 is the last element of the Sturm sequence hence 𝑥 = 2 is the
SPACE FOR HINTS double root.

Let 𝑉(𝑎) denote the number of sign changes in the Sturm‟s sequence
𝑥 = 𝑎 of the polynomial. We divide each element of the Sturm sequence
by (𝑥 − 2) and obtain the new sequences asfollows.

∗
𝑥 4 − 4𝑥 3 + 3𝑥 2 + 4𝑥 − 4
𝑓 𝑥 =
(𝑥 − 2)

𝑓 ∗ 𝑥 = 𝑥 3 − 2𝑥 2 − 𝑥 + 2

∗ 2𝑥 3 − 6𝑥 2 + 3𝑥 + 2
𝑓1 (𝑥) =
(𝑥 − 2)

= 2𝑥 2 − 2𝑥 − 1

𝑥 2 − 3𝑥 + 2
𝑓2 ∗ 𝑥 =
𝑥−2
=𝑥−1
𝑥−2
𝑓3 ∗ 𝑥 =
𝑥−2
=1

𝑥=𝑎 𝑓∗ 𝑓1 ∗ (𝑥) 𝑓2 ∗ (𝑥) 𝑓3 ∗ (𝑥) 𝑉(𝑥)

−∞ − + − + 3

−2 − + − + 3

−1.5 − + + + 2

−1 0 + − + 2

0 + − − + 2

1.5 − + + + 1

2 0 + + + 0

2.5 + + + + 0

∞ + + + + 0

54
From the above table, the polynomial has three real roots in the interval
SPACE OF HINTS
−1.5, 0 , (0, 1.5) and (1.5,2.5) and also 𝑥 = 2 lies in the interval
(1.5, 2.5)

∴The given polynomial has four real roots lies in the interval
−1.5,0 , (0,1.5) and two roots liesin the interval (1.5, 2.5).

Birge-Vieta Method:

Steps for Birge-Vieta Method:

𝑏𝑛
Step1: The formula for the Method is 𝑃𝑘+1 = 𝑃𝑘 −
𝑐𝑛 −1

Step2:Calculate the co. efficient of 𝑏𝑛 &𝑐𝑛−1 from thefollowing

synthetic division.

𝑝 𝑎0 𝑎1 𝑎2 ⋯ 𝑎𝑛

0 𝑝𝑏0 𝑝𝑏1 ⋯ 𝑝𝑏𝑛−1

𝑏0 𝑏1 𝑏2 ⋯ 𝑏𝑛−1 𝑏𝑛

𝑝𝑐0 𝑝𝑐1 ⋯ 𝑝𝑐𝑛 −2

𝑐0 𝑐1 ⋯ 𝑐𝑛−1

Step3:
The deflated polynomial is 𝑄𝑛−1 𝑥 = 𝑏0 𝑥 𝑛−1 + 𝑏1 𝑥 𝑛−2 + ⋯ + 𝑏𝑛−1

Example1.6.5:
Use synthetic division and perfect two iterations of the Birge-Vieta
Method to find the smallest positive root of the polynomial

𝑃3 𝑥 = 2𝑥 3 − 5𝑥 + 1 use the initial approximation 𝑃0 = 0.5 also

obtain the deflated polynomial.

Solution:
𝑏𝑛
The formula for Birge-Vieta Method is 𝑃𝑘+1 = 𝑃𝑘 − ,
𝑐𝑛 −1

𝑘 = 0,1,2, ⋯

55
 First iteration:
SPACE FOR HINTS The given equation is polynomial of order 3.
𝑏3
𝑘 = 0, 𝑃1 = 𝑃0 −
𝑐2

Given𝑃3 𝑥 = 2𝑥 3 − 5𝑥 + 1 = 0 &𝑃0 = 0.5

0.5 2 0 −5 1

0 1 0.5 − 2.25

2 1 − 4.5 − 1.25

0 1 1

2 2 − 3.5
−1.25 1.25
𝑃1 = 0.5 − = 0.5 −
−3.5 3.5
= 0.1429
𝑏3
Second iteration:𝑘 = 1, 𝑃2 = 𝑃1 −
𝑐2

0.1429 2 0 −5 1

0 0.2858 0.0408 − 0.7087

2 0.2858 − 4.9592 0.2913

0 0.2858 0.08168

2 0.5716 − 4.87752
0.29134
𝑃2 = 0.1429 −
−4.87752
= 0.2026
𝑏3
Third iteration:𝑘 = 2, 𝑃3 = 𝑃2 −
𝑐2

0.2026 2 0 −5 1

0 0.4052 0.0820 − 0.9964

2 0.4052 − 4.9179 0.0036

56
Now 𝑏3 = 𝑃3 𝑃2 = 0.0036, this is givesthe error in satisfying the
SPACE OF HINTS
given equation 𝑃3 𝑥 = 0.

∴The deflected polynomial after two iteration is given by

𝑄2 𝑥 = 2𝑥 2 + 0.4052𝑥 − 4.9179.

Example1.6.6:
Use synthetic division and perform two iteration of the Birge-Vieta
Method to find the smallest positive root of the equation

𝑥 4 − 3𝑥 3 + 3𝑥 2 − 3𝑥 + 2 = 0 use the initial approximation 𝑃0 = 0.5

Solution:The formula for Birge-Vieta Method is 𝑃𝑘+1 = 𝑃𝑘 −

𝑏𝑛
𝑘 = 0, 1, 2, ⋯
𝑐𝑛 −1

First iteration: The given equation is apolynomialoforder 4.

∴𝑛=4
𝑏4
𝑘 = 0, 𝑃1 = 𝑃0 −
𝑐3

0.5 1 −3 3 −3 2

0 0.5 − 1.25 0.875 − 1.0625

1 − 2.5 1.75 − 2.125 0.9375

0 0.5 −1 0.375

1 −2 0.75 − 1.75
0.9375 0.9375
𝑃1 = 0.5 − = 0.5 + = 1.0357
−1.75 1.75
𝑏4
Second iteration:𝑘 = 1, 𝑃2 = 𝑃1 −
𝑐3

1.0357 1 −3 3 −3 2

0 1.0357 − 2.0344 1.00007 − 2.0713

1 − 1.9643 0.9656 − 1.9999 − 0.0713

1.0357 − 0.9617 0.00404

1 − 0.9286 0.0039 − 1.9959

57
 0.0713
𝑃2 = 1.0357 − −
SPACE FOR HINTS −1.9959
= 0.9999

Third iteration:
𝑏4
𝑘 = 2, 𝑃3 = 𝑃2 −
𝑐3

0.9999 1 −3 3 −3 2

0 0.9999 − 1.9999 1 − 1.9998

1 − 2.0001 1.0001 − 2 0.0002

Now 𝑏4 = 𝑃4 𝑃2 = 0.0002

∴The deflated polynomial 𝑄3 𝑥 = 𝑏0 𝑥 3 + 𝑏1 𝑥 2 + 𝑏2 𝑥 + 𝑏3

𝑄3 𝑥 = 𝑥 3 − 2.0001𝑥 2 + 1.0001𝑥 − 2

and the exact root of the equation is0.999~1.

Bairstow Method:

Steps for Bairstow Method:

1) The input values are 𝑝0 and 𝑞0 are 𝑝1 = 𝑝0 + ∆𝑝 , 𝑞1 = 𝑞0 + ∆𝑞

𝑏𝑛 𝑐𝑛 −3−𝑏𝑛 −1 𝑐𝑛 −2
Where ∆𝑝 = −
𝑐𝑛 −2 2 −𝑐𝑛 −3 𝑐𝑛 −1 −𝑏𝑛 −1

𝑏𝑛−1 𝑐𝑛−1 − 𝑏𝑛−1 − 𝑏𝑛 𝑐𝑛−2

∆𝑞 = −
𝑐𝑛−2 2 − 𝑐𝑛−3 𝑐𝑛−1 − 𝑏𝑛−1

For computing 𝑏𝑘 ′ 𝑠 &𝑐𝑘 ′ we use the following synthetic division

−𝑝0 𝑎0 𝑎1 𝑎2 ⋯ 𝑎𝑛−1 𝑎𝑛

−𝑏0 𝑝0 − 𝑏1 𝑝0 ⋯ − 𝑏𝑛−2 𝑝0 − 𝑏𝑛−1 𝑝0

−𝑞0 − 𝑏0 𝑞0 ⋯ − 𝑏𝑛−3 𝑞0 − 𝑏𝑛−2 𝑞0

−𝑝0 𝑏0 𝑏1 𝑏2 ⋯ 𝑏𝑛−1 𝑏𝑛

−𝑝0 𝑐0 − 𝑝0 𝑐1 ⋯ − 𝑝0 𝑐𝑛−2

−𝑞0 − 𝑞0 𝑐0 ⋯ − 𝑞0 𝑐𝑛−3

𝑐0 𝑐1 𝑐2 𝑐𝑛−1

58
The deflated polynomial is given by 𝑄𝑛−2 𝑥 = 𝑏0 𝑥 𝑛−2 + 𝑏1 𝑥 𝑛−4 +
SPACE OF HINTS
⋯ + 𝑏𝑛−2 .

Example1.6.7:
Perform two iterations of the Bairstow Method to extract a
quadratic factor𝑥 2 + 𝑝𝑥 + 𝑞 from the polynomial 𝑃3 𝑥 = 𝑥 3 + 𝑥 2 −
𝑥 + 2 = 0 use the initial approximation 𝑝0 = −0.9 , 𝑞0 = 0.9.

Solution:
The initial values are 𝑝0 = −0.9 , 𝑞0 = 0.9

∴ −𝑝0 = 0.9 &−𝑞0 = −0.9

0.9 1 1 −1 2

0.9 1.71 − 0.171

−0.9 −0.9 − 1.71

0.9 1 1.9 − 0.190.119 = 𝑏3

−0.90.9 2.52

−0.9
1 2.8 1.43 = 𝑐2
𝑏𝑛 𝑐𝑛 −3−𝑏𝑛 −1 𝑐𝑛 −2
The general formula ∆𝑝 = −
𝑐𝑛 −2 2 −𝑐𝑛 −3 𝑐𝑛 −1 −𝑏𝑛 −1

𝑏3 𝑐0 −𝑏2 𝑐1
Put 𝑛 = 3, ∆𝑝 = −
𝑐1 2 −𝑐0 𝑐2 −𝑏2

0.651
∆𝑝 = − = −0.104
6.22
𝑏𝑛 −1 𝑐𝑛 −1 −𝑏𝑛 −1 −𝑏𝑛 𝑐𝑛 −2
The general formula ∆𝑞 = −
𝑐𝑛 −2 2 −𝑐𝑛 −3 𝑐𝑛 −1 −𝑏𝑛 −1

𝑏2 𝑐2 −𝑏2 −𝑏3 𝑐1
Put 𝑛 = 3, ∆𝑞 = −
𝑐1 2 −𝑐0 𝑐2 −𝑏2

0.641
=− − = 0.1031
6.22
𝑝1 = 𝑝0 + ∆𝑝 = −0.9 − 0.1047 = −1.0047

𝑞1 = 𝑞0 + ∆𝑞 = 0.9 + 0.1031 = 1.0031

59
 ∴ −𝑝1 = 1.0047& − 𝑞1 = −1.0031
SPACE FOR HINTS
Second iteration:
1.0047 1 1 −1 2
1.0047 2.0141 0.011
−1.0031 − 1.0031 − 2.0109
1 2.0047 0.011 0.0002 = 𝑏2
1.0047 3.0235
−1.0031
1 3.0094 2.0314 = 𝑐2

𝑏𝑛 𝑐𝑛 − 3 − 𝑏𝑛−1 𝑐𝑛−2
∆𝑝 = −
𝑐𝑛 −2 2 − 𝑐𝑛−3 𝑐𝑛−1 − 𝑏𝑛−1
𝑏3 𝑐0 − 𝑏2 𝑐1
=−
𝑐1 2 − 𝑐0 𝑐2 − 𝑏2
0.03290
=− = 0.0047
7.0361
𝑏𝑛−1 𝑐𝑛−1 − 𝑏𝑛−1 − 𝑏𝑛 𝑐𝑛−2 𝑏2 𝑐2 − 𝑏2 − 𝑏3 𝑐1
∆𝑞 = − = −
𝑐𝑛−2 2 − 𝑐𝑛−3 𝑐𝑛−1 − 𝑏𝑛−1 𝑐1 2 − 𝑐0 𝑐2 − 𝑏2
0.0216
=− = −0.0031
7.0361
𝑝2 = 𝑝1 + ∆𝑝 = −1.0047 + 0.0047 = −1

𝑞2 = 𝑞1 + ∆𝑞 = 1.0031 − 0.0031 = 1

Hence the extracted values of 𝑝 and 𝑞 are 𝑝 = −1 and 𝑞 = 1.

The extracted quadratic factor is 𝑥 2 + 𝑝𝑥 + 𝑞 = 𝑥 2 − 𝑥 + 1.

Example1.6.8:
Perform one iteration of the Bairstow method to exact a quadratic
factor 𝑥 2 + 𝑝𝑥 + 𝑞from polynomial 𝑥 4 + 𝑥 3 + 2𝑥 2 + 𝑥 + 1 = 0 use
initial approximation 𝑝0 = 0.5, 𝑞0 = .5

Solution:The initial values are 𝑝0 = 0.5 , 𝑞0 = 0.5

60
 ∴ −𝑝0 = −0.5 &−𝑞0 = −0.5
SPACE OF HINTS
First iteration:
Given polynomial 𝑥 4 + 𝑥 3 + 2𝑥 2 + 𝑥 + 1 = 0

−0.5 1 1 2 1 1

−0.5 − 0.25 − 0.625 − 0.0625

−0.5 − 0.5 − 0.25 − 0.625

1 0.5 1.25 0.125 0.3125 = 𝑏4

−0.5 0 − 0.375

−0.5 0

1 0 0.75 − 0.25 = 𝑐3

𝑏𝑛 𝑐𝑛 − 3 − 𝑏𝑛−1 𝑐𝑛−2
∆𝑝 = −
𝑐𝑛 −2 2 − 𝑐𝑛−3 𝑐𝑛−1 − 𝑏𝑛−1
𝑏4 𝑐1 − 𝑏3 𝑐2
=−
𝑐2 2 − 𝑐1 𝑐3 − 𝑏3
0.09375
=− − = 0.1667
0.5625
𝑏𝑛−1 𝑐𝑛−1 − 𝑏𝑛−1 − 𝑏𝑛 𝑐𝑛−2
∆𝑞 = −
𝑐𝑛−2 2 − 𝑐𝑛−3 𝑐𝑛−1 − 𝑏𝑛−1

𝑏3 𝑐3 − 𝑏3 − 𝑏4 𝑐2
=−
𝑐2 2 − 𝑐1 𝑐3 − 𝑏3

0.28125
=− − = 0.5
0.5625
𝑝1 = 𝑝0 + ∆𝑝 = 0.5 + 0.1667 = 0.6667 ≅ 1

𝑞1 = 𝑞0 + ∆𝑞 = 0.5 + 0.5 = 1

∴The exacted quadratic factor is 𝑥 3 + 𝑝𝑥 + 𝑞 = 𝑥 2 + 𝑥 + 1.

61
 Graff’s root squaring Method:This method is used to find the root of a
SPACE FOR HINTS polynomial with real coefficients.The roots may be real and distinct, real
and equal(or) complex.

Real and Distinct root:

Consider the polynomial equation of degree 𝑛
𝐴1 𝑥 = 𝑎0 𝑥 𝑛 + 𝑎1 𝑥 𝑛 −1 + ⋯ + 𝑎𝑛 −1 𝑥 + 𝑎𝑛 = 0 where 𝑎0 ≠ 0 (1)
Separating the even and odd functions of 𝑥 in (1) and squaring we get
𝑛 −1
(𝑎0 𝑥 𝑛 + 𝑎2 𝑥 𝑛−2 + ⋯ )2 = (𝑎 1 𝑥 + 𝑎3 𝑥 𝑛 −3 + ⋯ )2 (2)
Now, simplifying the equation,we get
𝑎0 2 𝑥 2𝑛 − 𝑎1 2 − 2𝑎0 𝑎2 𝑥 2𝑛 −2 + 𝑎2 2 + 2𝑎0 𝑎4 − 2𝑎1 𝑎3 𝑥 2𝑛 −4
− ⋯ − (−1)𝑛 𝑎𝑛 2 = 0
⟹ 𝑎0 2 (𝑥 𝑛 )2 − 𝑎1 2 − 2𝑎0 𝑎2 (𝑥 𝑛−1 )2
+ 𝑎2 2 + 2𝑎0 𝑎4 − 2𝑎1 𝑎3 (𝑥 𝑛−2 )2
− ⋯ − (−1)𝑛 𝑎𝑛 2 = 0 (3)
Put 𝑥 2 = −𝑧 and 𝑎0 2 = 𝑏0 , 𝑎1 2 − 2𝑎0 𝑎2 = 𝑏1 &𝑎2 2 + 2𝑎0 𝑎4 −
2𝑎1 𝑎3 = 𝑏2 and so on
⟹ 𝑏0 (−𝑧)𝑛 + 𝑏1 (−𝑧)𝑛−1 + 𝑏2 (−𝑧)𝑛−2 − ⋯ − (−1)𝑛 𝑏𝑛 = 0 (4)
∴All the 𝑏𝑘 ‟s are known in terms of 𝑎𝑘 ‟s.The root of the equation (4) are
−𝜉1 2 − 𝜉2 2 ⋯ − 𝜉𝑛 2
where 𝜉1 , 𝜉2 , … , 𝜉𝑛 are the roots of givenequation.
The coefficients𝑏𝑘 ‟s can beobtain by the following table:

𝑎0 𝑎1 𝑎2 𝑎3 𝑎4 ⋯ 𝑎𝑛

62
 SPACE OF HINTS
2 2 2 2 2
𝑎0 𝑎1 𝑎2 𝑎3 ⋯ 𝑎𝑛
−2𝑎0 𝑎2 − 2𝑎1 𝑎3 − 2𝑎2 𝑎4 ⋯

2𝑎0 𝑎4 + 2𝑎1 𝑎3 + ⋯
⋮

𝑏0 𝑏1 𝑏2 𝑏3 ⋯ 𝑏𝑛

The (𝐾 + 1)𝑡𝑕 column in the above table can be obtained as follows:

The terms attended in sign starting with a positive sign. The first term is
the square of the (𝐾 + 1)𝑡𝑕 co. efficient 𝑎𝑘 . The second term is true the
product of the 𝑛 nearest neighboring coefficient 𝑎𝑘−1 and 𝑎𝑘+1 . The
third is twice the product of the next neighboring coefficients𝑎𝑘−2 and
𝑎𝑘+2 . The procedure is contained until there are no available coefficients
to form the cross product. This procedure can be repeated 𝑚 times and
we obtain the equation.

𝐵0 𝑥 𝑛 + 𝐵1 𝑥 𝑛 −1 + 𝐵2 𝑥 𝑛−2 + ⋯ + 𝐵𝑛−1 𝑥 + 𝐵𝑛 = 0 (5)

Whose roots 𝑅1 , 𝑅2 , ⋯ , 𝑅𝑛 are the 2𝑚 th power of the roots of a equation

(1) with opposite signs.
𝑚
𝑅𝑖 = −𝜉 2 𝑖 = 1,2,3, ⋯ , 𝑛

If we assume the 𝜉1 > 𝜉2 > ⋯ > 𝜉𝑛 .

Then 𝑅1 ≫ 𝑅2 ≫ ⋯ ≫ 𝑅𝑛 .

∴The roots of equation (1) differ in magnitude then the 2𝑚 th power of

roots are widely separated

for large values of 𝑚, we get

𝐵1
− = 𝑅𝑖 ≈ 𝑅1
𝐵0

63
 𝐵2
= 𝑅𝑖 𝑅𝑗 ≈ 𝑅1 𝑅2
SPACE FOR HINTS 𝐵0
𝐵3
− = 𝑅𝑖 𝑅𝑗 𝑅𝑘 ≈ 𝑅1 𝑅2 𝑅3
𝐵0
(−1)𝑛 𝐵𝑛
≈ 𝑅1 𝑅2 ⋯ 𝑅𝑛
𝐵0

∴From the above equation, we get

−𝐵𝑖
𝑅𝑖 = , 𝑖 = 1,2, ⋯ , 𝑛
𝐵𝑖−1
𝐵𝑖 2𝑚
⟹ 𝑅𝑖 = = 𝜉𝑖 (or)
𝐵𝑖−1

Taking 𝑙𝑜𝑔 on both sides, we get

2𝑚
𝐵𝑖
𝑙𝑜𝑔 𝜉𝑖 = 𝑙𝑜𝑔
𝐵𝑖−1

2𝑚 𝑙𝑜𝑔 𝜉𝑖 = 𝑙𝑜𝑔 𝐵𝑖 − 𝑙𝑜𝑔 𝐵𝑖−1

𝑙𝑜𝑔 𝜉𝑖 = 2−𝑚 𝑙𝑜𝑔 𝐵𝑖 − 𝑙𝑜𝑔 𝐵𝑖−1 𝑖 = 1, 2, 3, ⋯ , 𝑚

This determines the absolute values of the root and substitution in the
original equation (1) willgives the sign of the roots.

The squaring process is stopped when another squaring process

produces new coefficients that are almost the squares of the
corresponding coefficients 𝑏𝑘 ‟s

(i.e.) When the crass product terms become negligible in comparison to

square terms.

Real and Equal roots:

After few squaring, if the magnitude of the coefficients 𝑏𝑘 ‟s about

half the square

of the magnitude of the corresponding coefficients in the previous

equation then its indicates that 𝜀𝑘 is a double root.

We can find this double root by using the following procedure.

64
 −𝐵𝑘 −𝐵𝑘+1
𝑅𝑘 = and 𝑅𝑘+1 =
𝐵𝑘−1 𝐵𝑘 SPACE OF HINTS

𝐵𝑘+1
𝑅𝑘 𝑅𝑘+1 ~𝑅𝑘 2 =
𝐵𝑘−1

2 2(2𝑚 )
𝐵𝑘+1
𝑅𝑘 = 𝜉𝑘 =
𝐵𝑘−1
This gives the magnitude of the double root substituting in the given
equation can find it‟s sign.This double root can also be found directly.

Complex root:

If 𝜀𝑘 and 𝜀𝑘+1 from a complex pair. If 𝜉𝑘 and 𝜉𝑘 +1 = 𝛽𝑘 exp ±𝜑𝑘 is

the complex Pair then the coefficients fluctuate in magnitude and sign
by an amount 2𝛽𝑘 𝑚 cos(𝑚 𝜑𝑘 )

For 𝑚 sufficiently large, 𝛽𝑘 can be determined from the relation

(2𝑚 ) 𝐵𝑘+1
𝛽𝑘 2 = .
𝐵𝑘−1

Example 1.6.9: Find all the roots of the polynomial

𝑥 3 − 6𝑥 2 + 11𝑥 − 6 = 0 using the Graffe‟s rootSquaring method.

Solution:
Given polynomial equation is 𝑥 3 − 6𝑥 2 + 11𝑥 − 6 = 0
𝑚 2𝑚 𝑎0 𝑎1 𝑎2 𝑎3

0 1 1 −6 11 −6

1 36 121 36
−22 −72
1 2 1 14 49 36

1 196 2401 1296

−98 −1008
2 4 1 98 1393 1296

9604 1940449 1679616

65
 −2786 −254016
SPACE FOR HINTS
3 8 1 6818 1686433 1679616

46485124 2.84406 × 1012 2.82111 × 1012

−3372866 − 2.29032 × 1010
4 16 1 43112258 2.821157 × 1012 2.8211 × 1012

Successive approximations to the roots are found by the formula

(2𝑚 )−1
2𝑚
𝐵𝑖 𝐵𝑖
𝑅𝑖 = 𝜉𝑖 = ⇒ 𝜉𝑖 =
𝐵𝑖−1 𝐵𝑖−1

𝑚 2𝑚 𝛼1 𝛼2 𝛼3

1 2 3.7414 1.8708 0.8571

2 4 3.1463 1.9417 0.9821

3 8 3.0144 1.9914 0.9995

4 16 3.0003 1.9998 1.0000

𝛼1 ≅ 3𝛼2 ≅ 2𝛼3 ≅ 1

∴The exact roots of the given equation is3,2,1.

Example 1.6.10: Find all the roots of the polynomial

𝑥 3 − 4𝑥 2 + 5𝑥 − 2 = 0 using the Graffe‟s rootSquaring method.

Solution: Given polynomial equation is 𝑥 3 − 4𝑥 2 + 5𝑥 − 2 = 0

Coefficients in the root squaring by Graff‟s Method.

𝑚 2𝑚 𝑎0 𝑎1 𝑎2 𝑎3

0 1 1 −4 5 −2

1 16 25 4
−10 −16
1 2 1 6 9 4

66
 1 36 81 16
SPACE OF HINTS
−18 −48
2 4 1 18 33 16

1 324 1089 256

−66 −576
3 8 1 258 513 256

1 66564 263169 65636

−1026 −132096
4 16 1 65538 131073 65536

1 0.4295 × 1010 1.718 × 1010 0.4295 × 1010

−262146 −0.859 × 1010
5 32 1 0.4295 × 1010 0.859 × 1010 0.4295 × 1010

We notice that above table,the magnitude of the 𝐵2 is half the square of

the magnitude of thecorresponding co. efficient privies equation.

This phenomenon indicates that 𝜉2 is a double root.

∴ The magnitude of the roots are

2.2𝑛
𝐵𝑘+1
𝜉𝑘 =
𝐵𝑘−1

2.32 64 𝐵3 0.4295×10 10
Put𝑘 = 2 , 𝜉2 = 𝜉2 = =
𝐵1 0.4295×10 10

1
𝜉2 = 164 = 1;

∴ 𝜉3 is also 1.

Now theanother root

2𝑚
𝐵𝑖 𝐵1
𝜉1 = =
𝐵𝑖−1 𝐵0

32
0.4295 × 1010
𝜉1 =
1
1
𝜉1 = (0.4295 × 1010 )32 = 2

∴The exact root of the given equation is 2, 1, 1.

Example 1.6.11:

67
 Find all the roots of the polynomial 𝑥 4 − 𝑥 3 + 3𝑥 2 + 𝑥 − 4 = 0
SPACE FOR HINTS using the Graffe‟s root squaring Method.

Solution:
The Coefficients in the successive root squaring are given in the
table.

𝑚 2𝑚 𝑎0 𝑎1 𝑎2 𝑎3 𝑎4

0 1 1 −1 3 1 −4

1 1 9 1 16
−6 2 24
−8
1 2 1 −5 3 25 16

1 25 9 625 256
−6 250 −96
32
2 4 1 19 291 529 256

1 361 84681 279841 65636

−582 −20102 −148992
512
3 8 1 −221 65091 130849 65536

1 48841 0.4237 × 1010 1.712 × 1010 0.4295 × 1010

−130182 0.0057 × 1010 −0.853 × 1010
0.00001 × 1010
4 16 1 −81341 0.4295 × 1010 0.859 × 1010 0.4295 × 1010

Here 𝐵1 values are alternatively in sign.

∴we have a pair of complex roots which can be obtained by using

𝐵0 , 𝐵1 , &𝐵2 , 𝑚 = 4 again we get the negative value of 𝐵1 .

68
∴For calculation purpose we use 𝑚 = 3 values for real roots.
SPACE OF HINTS
Real roots:

2𝑚
𝐵𝑖
𝜉𝑖 =
𝐵𝑖−1

8
𝐵3
𝑖 = 3 ⇒ 𝜉3 =
𝐵2

8
130849
𝜉3 = =2
65091
𝜉3 = 1.0912

Also𝑝4 𝜉3 = +𝑣𝑒 value.

∴ 𝜉3 = 1.0912 is a +𝑣𝑒 root

Now another real root, put 𝑖 = 4

8 𝐵4
𝜉4 =
𝐵3

65536 1
𝜉4 = ( )8 = 0.9172
130849

𝑝4 𝜉4 = −𝑣𝑒

∴ 𝜉4 = −0.9172 is a – 𝑣𝑒 root.

To obtain a pair of complex roots 𝑝 ± 𝑖𝑞, we have 𝑝2 + 𝑞2 = 𝛽2 2 (1)

2.2𝑚
𝐵2 65091
𝛽2 = =( )
𝐵0 1
16
𝛽2 = (65091)
1
𝛽2 = (65091)16 = 1.9991
−𝑎 1
Also, we know that sum of the roots 𝜉3 + 𝜉4 + 2𝑝 =
𝑎0

−(−1)
1.0912 − 0.9172 + 2𝑝 = =1
1
0.174 + 2𝑝 = 1

2𝑝 = 1 − 0.174

⇒ 𝑝 = 0.413

69
 2
(1)⇒ 0.413 + 𝑞2 = 1.9991 2

SPACE FOR HINTS ⟹ 0.1705 + 𝑞2 = 3.9964

𝑞2 = 3.8258

⟹ 𝑞 = 1.9560

Hence the pair of complex roots 𝑝 ± 𝑖𝑞 is given by

0.413 + 𝑖 1.9560 & 0.413 − 𝑖(1.9560)

Here the roots of the given polynomial are 1.0912, −0.9172 and
0.413 ± 𝑖(1.9560).

Unit – 2

2.1: Introduction

1) A real matrix 𝐴is said to be non-singular, if 𝐴 ≠ 0.

2) A real matrix 𝐴 is said to be Symmetric if𝐴 = 𝐴.
3) A real matrix 𝐴 is said to be Skew Symmetricif𝐴 = −𝐴𝑇 .
4) A real matrix 𝐴 is said to be orthogonal if𝐴−1 = 𝐴𝑇 .
5) A real matrix 𝐴 is said to be Null if 𝑎𝑖𝑗 = 0 ∀ 𝑖, 𝑗 = 1 𝑡𝑜𝑛.
𝑎𝑖𝑗 = 0 , 𝑖 ≠ 𝑗
6) A real matrix 𝐴 is said to be Diagonal Matrixif .
𝑎𝑖𝑗 ≠ 0 , 𝑖 = 𝑗
𝑎𝑖𝑗 = 0 , 𝑖 ≠ 𝑗
7) A real matrix 𝐴 is said to be Unit Matrix, if .
𝑎𝑖𝑗 = 1 , 𝑖 = 𝑗
8) A real matrix 𝐴 is said to be Lower Triangular Matrix, if

𝑎𝑖𝑗 = 0 , 𝑗 < 𝑖.

9) A real matrix 𝐴 is said to be Upper Triangular Matrix, if

𝑎𝑖𝑗 = 0 , 𝑖 < 𝑗.

10)A real matrix 𝐴 is said to be Band Matrix, if 𝑎𝑖𝑗 = 0 for𝑗 > 𝑖 + 𝑝

and𝑖 > 𝑗 + 𝑞 with band width𝑝 + 𝑞 = 1.

11) A real matrix 𝐴 is said to be Tridiagonal, if𝑎𝑖𝑗 = 0 for 𝑖 − 𝑗 > 1.

70
 12) A real matrix𝐴 is said to be Diagonally Dominant, if
SPACE OF HINTS
𝑛

𝑎𝑖𝑗 ≥ 𝑎𝑖𝑗 , 𝑖 = 1 𝑡𝑜 𝑛.
𝑗 =1,𝑖≠𝑗

13) A real matrix 𝐴 is said to be Strictly Diagonally Dominant, if the

strict inequality is satisfied for all 𝑖.

14)Hermitian matrix:- A complex matrix𝐴is said to Hermitian

denoted by 𝑨∗ (or)𝑨𝑯 if𝑨 = 𝑨 𝑻 where𝑨is the complex

conjugate of𝑨.

15)Permutation matrix:If it has exactly one 1 in each row and

column and allother entries 0.

16)A complex matrix𝐴is said tounitary if𝐀−𝟏 = 𝐀 𝐓 .

17)A complex matrix𝐴is said to be normal if𝑨𝑨∗ = 𝑨∗ 𝑨.

18)A matrix 𝐴 is said to be a Positive Definite Matrix if 𝑋 ∗ 𝐴𝑋 > 0

for any vector 𝑿 ≠ 𝟎 &𝑿∗ = 𝑿𝑻 . Further 𝑿∗ 𝑨𝑿 = 𝟎iff𝑿 = 𝟎.

19)If 𝑨 is Hermitian, strictly diagonal dominant matrix with positive

real diagonalentries, then 𝑨 is positive definite.

Note:-

i. If 𝐴 is non-singular and positive definite then 𝑩 = 𝑨∗ 𝑨is

Hermitian and positive definite.
ii. The eigen values of a positive definite matrix are all positive.
iii. All the leading minors of 𝐴 are positive.

Note:-

1) A real matrix 𝐵 is said to have property 𝐴 iff there exist a

permutation matrix 𝑷 such that

71
 𝑨𝟏𝟏 𝑨𝟏𝟐
𝑷𝑩𝑷𝑻 =
𝑨𝟐𝟏 𝑨𝟐𝟐
SPACE FOR HINTS
where𝑨𝟏𝟏 and𝑨𝟐𝟐 are diagonal matrices.
2) The minor𝑀𝑖𝑗 of𝑎𝑖𝑗 is the determinant of the 𝑛 − 1 × 𝑛 −
1 submatrix obtained by deleting the𝑖𝑡𝑕 row and𝑗𝑡𝑕 column in𝑨.
The cofactor is defined by𝐴𝑖𝑗 = (−1)𝑖+𝑗 𝑀𝑖𝑗
3) The transpose of the matrix of the cofactors of the elements of 𝐴
is called the adjoint matrix and is denoted by 𝒂𝒅𝒋 𝑨 .
𝟏
4) The inverse of a matrix 𝐴 is defined by 𝑨−𝟏 = 𝒂𝒅𝒋 𝑨 .
𝑨

Remarks:

The inverse of a symmetric matrix is symmetric. The inverse of a

lower triangular matrix is lower triangular. The inverse of a upper
triangular matrix is upper triangular.

Example: 2.1.1

12 4 −1
Show that the matrix 4 7 1 is positive definite.
−1 1 6

Solution:

12 4 −1 𝑥1
Let 𝐴 = 4 7 𝑥
1 and let 𝑋 = 2
−1 1 6 𝑥3

𝑋∗ = 𝑋 𝑇
= 𝑥1 𝑥2 𝑥2

12 4 −1 𝑥1
∗
𝑋 𝐴𝑋 = 𝑥1 𝑥2 𝑥2 4 7 1 𝑥2
−1 1 6 𝑥3

72
 𝑥1
= 12𝑥1 + 4𝑥2 − 𝑥3 4𝑥1 + 7𝑥2 + 𝑥3 −𝑥1 + 𝑥2 + 6𝑥3 𝑥2 SPACE OF HINTS
𝑥3

= 12𝑥1 𝑥1 + 4𝑥2 𝑥1 − 𝑥3 𝑥1 + 4𝑥1 𝑥2 + 7𝑥2 𝑥2 + 𝑥3 𝑥2 − 𝑥1 𝑥3 + 𝑥2 𝑥3

+ 6𝑥3 𝑥3

2 2 2
= 12 𝑥1 + 7 𝑥2 + 6 𝑥3 + 4(𝑥2 𝑥1 + 𝑥1 𝑥2 ) − (𝑥3 𝑥1 + 𝑥1 𝑥3 )
+ (𝑥3 𝑥2 + 𝑥2 𝑥3 )

Let 𝑥1 = 𝑝1 + 𝑖𝑞1 , 𝑥2 = 𝑝2 + 𝑖𝑞2 𝑎𝑛𝑑 𝑥3 = 𝑝3 + 𝑖𝑞3

2
⟹ 𝑥1 = 𝑝1 2 + 𝑞1 2 ; 𝑥2 2
= 𝑝2 2 + 𝑞2 2 ; 𝑥3 2
= 𝑝3 2 + 𝑞3 2

Now, 𝑥1 𝑥2 = (𝑝1 − 𝑖𝑞1 )(𝑝2 + 𝑖𝑞2 )

= 𝑝1 𝑝2 + 𝑖𝑝1 𝑞2 − 𝑖𝑞1 𝑝2 + 𝑞1 𝑞2

= 𝑝1 𝑝2 + 𝑞1 𝑞2 + 𝑖(𝑝1 𝑞2 − 𝑞1 𝑝2 )

𝑥2 𝑥1 = 𝑝1 + 𝑖𝑞1 (𝑝2 − 𝑖𝑞2 )

= 𝑝1 𝑝2 − 𝑖𝑝1 𝑞2 + 𝑖𝑞1 𝑝2 + 𝑞1 𝑞2

= 𝑝1 𝑝2 + 𝑞1 𝑞2 + 𝑖(𝑞1 𝑝2 − 𝑝1 𝑞2 )

𝑥1 𝑥3 = 𝑝1 + 𝑖𝑞1 (𝑝3 − 𝑖𝑞3 )

= 𝑝1 𝑝3 + 𝑞1 𝑞3 + 𝑖(𝑞1 𝑝3 − 𝑝1 𝑞3 )

𝑥2 𝑥3 = 𝑝2 𝑝3 + 𝑞2 𝑞3 + 𝑖(𝑝2 𝑞3 − 𝑞2 𝑝3 )

𝑋 ∗ 𝐴𝑋 = 12 𝑝1 2 + 𝑞1 2 + 8 𝑝1 𝑝2 + 𝑞1 𝑞2 − 2 𝑝1 𝑝3 + 𝑞1 𝑞3
+ 2 𝑝2 𝑞3 + 𝑞2 𝑝3 + 7 𝑝2 2 + 𝑞2 2 + 6(𝑝3 2 + 𝑞3 2 )

= 4𝑝1 2 + 4𝑝2 2 + 8𝑝1 𝑝2 + 4𝑞1 2 + 4𝑞2 2 + 8𝑞1 𝑞2

+ 𝑝1 2 + 𝑝3 2 − 2𝑝1 𝑝3 + 𝑞1 2 + 𝑞3 2 − 2𝑞1 𝑞3
+ 𝑝2 2 + 𝑝3 2 + 2𝑝2 𝑝3 + (𝑞2 2 + 𝑞3 2 + 2𝑞2 𝑞3 )

73
 = 4(𝑝1 + 𝑝2 )2 + 4(𝑞1 + 𝑞2 )2 + (𝑝1 − 𝑝3 )2 + (𝑞1 − 𝑞3 )2
SPACE FOR HINTS + (𝑝2 + 𝑝3 )2 + (𝑞2 + 𝑞3 )2

>0

∴ 𝑋 ∗ 𝐴𝑋 > 0

Hence the matrix 𝐴 is positive definite.

Note:

12 4 −1
𝐴= 4 7 1
−1 1 6

Now, the leading minor of the matrix A is

12 = 12 is positive

12 4
= 68 is positive
4 7

12 4 −1
𝐴 = 4 7 1 = 12 42 − 1 − 4 24 + 1 − 1 4 + 7
−1 1 6

= 492 − 100 − 11 = 381 = +𝑣𝑒

Here all these values are positive.

∴ 𝐴is positive definite.

Note:

Here 𝐴 is Hermitian and strictly diagonally dominant matrix.

Because𝐴 = 𝐴 𝑇 and all diagonal entries are ≥ 1 (greater than or equal
to one).

∴The given matrix 𝐴 is positive definite.

Example: 2.1.2

74
 2 −1 0 −1
−1 2 −1 0 SPACE OF HINTS
Show that the matrix 𝐴 = posseses
0 −1 2 0
0 0 −1 2
property 𝐴.

Solution:

0 0 0 1
0 1 0 0
Consider the permutation matrix 𝑃 =
0 0 1 0
1 0 0 0

0 0 0 1 2 −1 0 −1 0 0 0 1
0 1 0 0 −1 2 −1 0 0 1 0 0
𝑃𝐴𝑃𝑇 =
0 0 1 0 0 −1 2 0 0 0 1 0
1 0 0 0 0 0 −1 2 1 0 0 0

0 0 −1 2 0 0 0 1
−1 2 −1 0 0 1 0 0
=
0 −1 2 0 0 0 1 0
2 −1 0 −1 1 0 0 0

2 0 −1 −1
0 2 −1 −1
=
0 −1 2 0
−1 −1 0 2

In the above matrix, the submatrix 𝐴11 and 𝐴22 are diagonal
matrices and hence the matrix 𝐴 as property 𝐴.

Example: 2.1.3

Find the inverse of the following matrix

1 0 0 0 0 0 0
𝑥 1 0 0 0 0 0
𝑥2 𝑥 1 0 0 0 0
𝑥3 𝑥2 𝑥 1 0 0 0
⋮ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮
𝑛−1 𝑛 −2
𝑥 𝑥 𝑥 𝑛−3 ⋯ 𝑥2 𝑥 1

Solution:

75
 We know that, the inverse of the lower triangular matrix is also a
SPACE FOR HINTS triangular matrix.

Let the inverse of the matrix 𝐴be𝐴−1 .

Using the identity 𝐴𝐴−1 = 𝐼.

𝑙11 0 0 0 0
𝑙21 𝑙22 0 0 0
Now, consider 𝐴−1 = 𝑙31 𝑙32 𝑙33 0 0
⋮ ⋮ ⋮ ⋮ ⋮
𝑙𝑛1 𝑙𝑛2 𝑙𝑛3 ⋯ 𝑙𝑛𝑛

Now, 𝐴𝐴−1 = 𝐼

1 0 0 0 0 0 0 𝑙11 0 0 0 0
𝑥 1 0 0 0 0 0 𝑙21 𝑙22 0 0 0
𝑥2 𝑥 1 0 0 0 0
⟹ 𝑙31 𝑙32 𝑙33 0 0
𝑥3 𝑥2 𝑥 1 0 0 0
⋮ ⋮ ⋮ ⋮ ⋮
⋮ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮
𝑙𝑛1 𝑙𝑛2 𝑙𝑛3 ⋯ 𝑙𝑛𝑛
𝑥 𝑛−1 𝑥 𝑛 −2 𝑥 𝑛 −3 ⋯ 𝑥2 𝑥 1
1 0 0 0 ⋯ 0
0 1 0 0 ⋯ 0
0 0 1 0 ⋯ 0
=
0 0 0 1 ⋯ 0
⋮ ⋮ ⋮ ⋮ ⋱ ⋮
0 0 0 0 ⋯ 1

𝑙11 = 1, also𝑙22 = 𝑙33 = ⋯ = 𝑙𝑛𝑛 = 1

𝑥𝑙11 + 𝑙21 = 0 ⟹ 𝑥 + 𝑙21 = 0 ⟹ 𝑙21 = −𝑥

𝑥 2 𝑙11 + 𝑥𝑙21 + 𝑙31 = 0

⟹ 𝑥 2 − 𝑥 2 + 𝑙31 = 0

⟹ 𝑙31 = 0

𝑥𝑙22 + 𝑙32 = 0 ⟹ 𝑙32 = −𝑥

1, 𝑖 = 𝑗
∴ 𝑙𝑖𝑗 = −𝑥, 𝑖 = 𝑗 + 𝑖
0, 𝑜𝑡𝑕𝑒𝑟𝑒𝑤𝑖𝑠𝑒

76
Hence the matrix 𝐴−1 is
SPACE OF HINTS

𝑙11 0 0 0 0 1 0 0 ⋯ 0 0
𝑙21 𝑙22 0 0 0 −𝑥 1 0 ⋯ 0 0
0 −𝑥 1 ⋯ 0 0
𝐴−1 = 𝑙31 𝑙32 𝑙33 0 0 =
0 0 −𝑥 ⋱ 0 0
⋮ ⋮ ⋮ ⋮ ⋮
⋮ ⋮ ⋮ ⋱ 1 ⋮
𝑙𝑛1 𝑙𝑛2 𝑙𝑛3 ⋯ 𝑙𝑛𝑛
0 0 0 0 −𝑥 1

Linear system of equations

The given system of equation can be written as

𝑨𝑿 = 𝒃 (𝐼)

the matrix 𝐴 𝑏 is called the “augmented matrix” where𝑏is an column

matrix and 𝐴 is an𝑛 × 𝑛 square matrix.

If all𝑏𝑖 are zero then the system of equation is said to be “Homogenous”

and if at least one𝑏𝑖 is not zero, then it is said to be “in-homogenous”.

The in-homogenous system has a unique solution iff 𝐴 ≠ 0

𝑎11 𝑎12 ⋯ 𝑎14

𝑎21 𝑎22 ⋯ 𝑎24
𝑖𝑒) det 𝐴 = ⋮ ⋮ ⋱ ⋮ ≠0
𝑎𝑛1 𝑎𝑛2 ⋯ 𝑎𝑛𝑛

The solution to the above system can be written as 𝑋 = 𝐴−1 𝑏.

The homogenous system𝑏𝑖 = 0, 𝑖 = 1 𝑡𝑜 𝑛possess only a trivial solution

𝑥1 = 𝑥2 = ⋯ = 0.

If det 𝐴 ≠ 0

Now, consider this system in which the parameter 𝜆occurs and

determine values of𝜆, called “eigen values” for which the system has a
non-trivial solution. Such a solution is called an eigenvector and the
entire system is called an eigenvalue problem or the characteristic value
problem.

77
 In 𝐼 𝑏 = 0, then the above system can be written
SPACE FOR HINTS
𝑨𝒙 = 𝝀𝒙 (1)

⟹ 𝑨𝒙 − 𝝀𝒙 = 𝟎

⟹ 𝑨−𝝀 𝒙=𝟎 (2)

In order that (2) has a non-trivial solution 𝒙 ≠ 𝟎, the determinant of the

matrix 𝑨 − 𝝀𝑰 = 𝟎

𝑖𝑒) 𝐝𝐞𝐭 𝑨 − 𝝀𝑰 = 𝟎 (3)

(3) is called the “characteristic equation”. Then the 𝑛roots 𝜆1 , 𝜆2 , … 𝜆𝑛

are called eigenvalues of 𝐴 which are distinct or repeated, real or
complex. The largest eigen value in modulus is called the “special
radius” of 𝐴. Corresponding to each eigen value𝜆𝑖 , there exist eigen
vector𝑥𝑖 which is a non-trivial solution of 𝐴 − 𝜆𝑖 𝐼 𝑥𝑖 = 0.

In the𝑛 eigen values 𝜆𝑖 , 𝑖 = 1 𝑡𝑜 𝑛 of 𝐴 are different and 𝑥𝑖 are

the corresponding eigen vector then any vector𝒚, in the eigenspace can
be expressed a linear combination of the vectors 𝑥𝑖 , 𝑦 = 𝑐1 𝑥1 + ⋯ +
𝑐𝑛 𝑥𝑛 . The method to determine eigenvalues and eigenvectors of the
system may be classified into two types.

i. Direct method: These methods produce the exact solution after a

finite number of steps (disregarding the round off errors)
ii. Iterative method: These methods give a sequence of
approximate solutions, which converges when the number of
steps tend to infinity.

2.2: Direct method

The system of equations𝑨 𝑿 = 𝒃

It can be directly solved in thefollowing cases:

78
(i). A = D.
SPACE OF HINTS

The equations (I) become

𝑎11 𝑥1 = 𝑏1

𝑎22 𝑥2 = 𝑏2

𝑎𝑛𝑛 𝑥𝑛 = 𝑏𝑛

The solution is given by

𝑏𝑖
𝑥𝑖 = , 𝑖 = 1 𝐼 𝑛, 𝑎𝑖𝑖 ≠ 0 (4)
𝑎𝑖𝑖

(ii). A = L

The equations may be written as

𝑎11 𝑥1 = 𝑏1

𝑎21 𝑥1 + 𝑎22 𝑥2 = 𝑏2

𝑎31 𝑥1 + 𝑎32 𝑥2 + 𝑎33 𝑥3 = 𝑏3 (5)

𝑎𝑛1 𝑥1 + 𝑎𝑛2 𝑥2 + ⋯ + 𝑎𝑛𝑛 𝑥𝑛 = 𝑏𝑛

Solving the first equation andthen successively solving thesecond, third

and so on, weobtain

𝑏1
𝑥1 =
𝑎11

(𝑏2 − 𝑎21 𝑥1 )
𝑥2 =
𝑎22

79
 (𝑏3 − 𝑎31 𝑥1 − 𝑎32 𝑥2 )
𝑥3 =
SPACE FOR HINTS 𝑎33

𝑛−1
𝑏𝑛 − 𝑗 =1 𝑎𝑛𝑗 𝑥𝑗
𝑥𝑛 =
𝑎𝑛𝑛

where 𝑎𝑖𝑖 ≠ 0, 𝑖 = 1(𝐼)𝑛.

Since the unknowns are solvedby forward substitution, this method is

called the forwardsubstitution method.

(iii). 𝑨 = 𝑼

The system of equations (I), becomes

𝑎11 𝑥1 + 𝑎12 𝑥2 + ⋯ + ⋯+ 𝑎1𝑛 𝑥𝑛 = 𝑏1

𝑎22 𝑥2 + ⋯ + ⋯+ 𝑎2𝑛 𝑥𝑛 = 𝑏2

𝑎𝑛−1,𝑛−1 𝑥𝑛−1 + 𝑎𝑛 −1,𝑛 𝑥𝑛 = 𝑏𝑛−1

𝑎𝑛𝑛 𝑥𝑛 = 𝑏𝑛

Solving for the unknowns in theorder𝑥𝑛 , 𝑥𝑛 −1 , … 𝑥1 ,we get

𝑏𝑛
𝑥𝑛 =
𝑎𝑛𝑛

(𝑏𝑛−1 − 𝑎𝑛 −1,𝑛 𝑥𝑛 )
𝑥𝑛−1 =
𝑎𝑛−1, 𝑛−1

𝑛
𝑏1 − 𝑗 =2 𝑎𝑖𝑗 𝑥𝑗
𝑥1 =
𝑎11

The unknowns are solved by backsubstitution and this method is called

80
the back substitution method.Thus, the equations (I) are exactlysolvable,
SPACE OF HINTS
if the matrix A in (3) canbe transformed into any one of theforms D, L
or U.

Cramer Rule

Step 1:Cramer‟s Rule

The given system of equation can be written as 𝑨𝑿 = 𝒃.

Step 2:

𝐵𝑖
The value of𝑥𝑖 = , 𝑖 = 1 𝑡𝑜 𝑛, where 𝐵𝑖 is the determinant of
𝐴

the matrix 𝐵𝑖 , obtained by replacing the𝑖𝑡𝑕 column of 𝐴 and 𝐵.This

process is known as Cramer‟s rule.

Example: 2.2.1

Solve the equation𝑥1 + 2𝑥2 − 𝑥3 = 2, 3𝑥1 + 6𝑥2 + 𝑥3 = 1, 3𝑥1 +

3𝑥2 + 2𝑥3 = 3

i. By using Cramer‟s rule

ii. By determining the inverse ofthe co-efficient matrix.

Solution:

i. By using Cramer‟s rule

The given system of equation can be written as 𝑨𝑿 = 𝒃 where

1 2 −1 𝑥1 2
𝐴= 3 6 1 , 𝑥
𝑋 = 2 𝑎𝑛𝑑 𝑏 = 1
3 3 2 𝑥3 3

𝐴 = 1 12 − 3 − 2 6 − 3 − 1 9 − 18

=9−6+9

= 12

81
 2 2 −1
𝐵1 = 1 6 1
SPACE FOR HINTS
3 3 2

= 2 12 − 3 − 2 2 − 3 − 1 3 − 18

= 18 + 2 + 15

= 35

1 2 −1
𝐵2 = 3 1 1
3 3 2

=1 2−3 −2 6−3 −1 9−3

= −1 − 6 − 6

= −13

1 2 2
𝐵3 = 3 6 1
3 3 3

= 1 18 − 3 − 2 9 − 3 + 2(9 − 18)

= 15 − 12 − 18

= −15

𝐵1 35 𝐵2 −13 𝐵3 −15
∴ 𝑥1 = = ; 𝑥2 = = ; 𝑥3 = =
𝐴 12 𝐴 12 𝐴 12

35
12
𝑥1
−13
∴ 𝑥1 , 𝑥2 , 𝑥3 𝑜𝑟 𝑥2 = 12
𝑥3
−15
12

ii. The inverse of the co-efficient matrix 𝑋 = 𝐴−1 𝑏 where

82
 1
𝐴−1 = 𝑎𝑑𝑗𝐴 SPACE OF HINTS
𝐴
Here 𝐴 = 12.

6 1 3 1 3 6
𝐴11 = = 9 ; 𝐴12 = − = −3 ; 𝐴13 = = −9
3 2 3 2 3 3

2 −1 1 −1 1 2
𝐴21 = − = −7 ; 𝐴22 = = 5 ; 𝐴23 = −
3 2 3 2 3 3
=3

2 −1 1 −1 1 2
𝐴31 = = 8 ; 𝐴32 = − = −4 ; 𝐴33 = =0
6 1 3 1 3 6
𝑇 𝑇
𝐴11 𝐴12 𝐴13 9 −3 −9
𝑎𝑑𝑗𝐴 = 𝐴21 𝐴22 𝐴23 = −7 5 3
𝐴31 𝐴32 𝐴33 8 −4 0

9 −7 8
∴ 𝑎𝑑𝑗𝐴 = −3 5 −4
−9 3 0

1 1 9 −7 8
𝐴−1 = 𝑎𝑑𝑗 𝐴 = −3 5 −4
𝐴 12
−9 3 0

𝑋 = 𝐴−1 𝑏

1 9 −7 8 2
= −3 5 −4 1
12
−9 3 0 3

1 18 − 7 + 24
= −6 + 5 − 12
12
−18 + 3

1 35
= −13
12
−15

83
 35
12
SPACE FOR HINTS
−13
∴𝑋= 12
−15
12

GaussElimenation Method:

Here the unknowns are eliminated by combining equations such

that the 𝑛 equations in 𝑛 unknowns are reduced to an equivalent upper
triangular system which is then solved by back substitution method.

Consider the 3 × 3 system

𝑎11 𝑥1 + 𝑎12 𝑥2 + 𝑎13 𝑥3 = 𝑏1

𝑎21 𝑥1 + 𝑎22 𝑥2 + 𝑎23 𝑥3 = 𝑏2 (1)
𝑎31 𝑥1 + 𝑎32 𝑥2 + 𝑎33 𝑥3 = 𝑏3

In the first stage of elimination, multiply the first row (1) by

𝑎21 𝑎31
𝑎11 and 𝑎11 respectively and subtract from the second and third

row weget

𝑎22 (2) 𝑥2 + 𝑎23 (2) 𝑥3 = 𝑏2 (2)

(2)
𝑎32 (2) 𝑥2 + 𝑎33 (2) 𝑥3 = 𝑏3 (2)

where

𝑎21 𝑎21
𝑎22 (2) = 𝑎22 − 𝑎12 ; 𝑎23 (2) = 𝑎23 − 𝑎
𝑎11 𝑎11 13

𝑎31 𝑎31
𝑎32 (2) = 𝑎32 − 𝑎12 ; 𝑎33 (2) = 𝑎33 − 𝑎
𝑎11 𝑎11 13

𝑎21 𝑎31
𝑏2 (2) = 𝑏2 − 𝑏1 ; 𝑏3 (2) = 𝑏3 − 𝑏
𝑎11 𝑎11 1

84
In the second stage of elimination, multiply the first row in (2)
SPACE OF HINTS
𝑎 (2)
by 32 and subtract from the second row in (2), we get
𝑎22 (2)

𝑎33 (2) 𝑥3 = 𝑏3 (3) (3)

𝑎 32 (2)
where𝑎33 (2) = 𝑎33 (2) − 𝑎23 (2)
𝑎 22 (2)

Collecting the first equation from

(3) (2) 𝑎32 (2) (3)

𝑏3 = 𝑏3 𝑏
𝑎22 (2) 2

Each stage ie) from (1), (2),(3). We obtain the system

𝑎11 (1) 𝑥1 + 𝑎12 (1) 𝑥2 + 𝑎13 (1) 𝑥3 = 𝑏1 (1)

𝑎22 (2) 𝑥2 + 𝑎23 (2) 𝑥3 = 𝑏2 (2)

𝑎33 (3) 𝑥3 = 𝑏3 (3)

where 𝑎𝑖𝑗 (1) = 𝑎𝑖𝑗 , 𝑏𝑖 (1) = 𝑏𝑖 , 𝑖, 𝑗 = 1, 2, 3.

The system is an upper triangular system and can be solved suing

the back substitution method.

Example: 2.2.2

Solve the equations 10𝑥1 − 𝑥2 + 2𝑥3 = 4, 𝑥1 + 10𝑥2 − 𝑥3 = 3,

2𝑥1 + 3𝑥2 + 20𝑥3 = 7 using Gauss elimination method.

Solution:

The given system of equation can be written as 𝑨𝑿 = 𝑩

10 −1 2 𝑥1 4
𝑖𝑒) 1 10 𝑥
−1 2 = 3
2 3 20 𝑥3 7

85
 The augmented matrix is
SPACE FOR HINTS
10 −1 2 4 𝑅
10 −1 2 4 101 −6 26 𝑅2 → 𝑅2 − 1
0 10
1 10 −1 3 ≅ 10 5 10
𝑅
2 3 20 7 16 2018 31 𝑅3 → 𝑅3 − 1
0 5
5 101 5

10 −1 2 4
101 −6 26
0 10 16
≅ 10 5 10 𝑅3 → 𝑅3 − 𝑅2
16 2018 543 101 5
0
5 101 101

From the above matrix, the system of equation can be written as

10𝑥1 − 𝑥2 + 2𝑥3 = 4 (1)

101 −6 26
𝑥2 𝑥3 = (2)
10 5 10

2018 543
𝑥3 = (3)
101 101

543 101
⟹ 𝑥3 = ×
101 2018

⟹ 𝑥3 = 0.2691

101 26 6
2 ⟹ 𝑥2 = + (0.2691)
10 10 5

10
𝑥2 = 2.9229 ×
101

𝑥2 = 0.2894

(1) ⟹ 10𝑥1 = 4 + 𝑥2 − 2𝑥3

= 4 + 0.2894 − (2 × 0.2691)

𝑥1 = 0.37512

86
Now the solution of the system of equation is 𝑥1 = 0.3751,
SPACE OF HINTS

𝑥2 = 0.2894 𝑎𝑛𝑑𝑥3 = 0.2691.

Example:2.2.3

Solve the equations 𝑥1 + 𝑥2 + 𝑥3 = 6, 3𝑥1 + 3𝑥2 + 4𝑥3 = 20 𝑎𝑛𝑑

2𝑥1 + 𝑥2 + 3𝑥3 = 13 using the Gauss elimination method.

Solution:

The given system of equation can be written as 𝑨𝑿 = 𝑩

1 1 1 𝑥1 6
𝑖𝑒) 3 3 𝑥
4 2 = 20
2 1 3 𝑥3 13

The augmented matrix is

1 1 1 6 1 1 1 6 𝑅 → 𝑅 − 3𝑅
2 2 1
3 3 4 20 ≅ 0 0 1 2 𝑅 → 𝑅 − 2𝑅
2 1 3 13 0 −1 11 3 3 1

1 1 16
≅ 0 0 1 2 𝑅3 ↔ 𝑅2
0 −1 11

From the above matrix, the system of equation can be written as

𝑥1 + 𝑥2 + 𝑥3 = 6 (1)

−𝑥2 + 𝑥3 = 1 (2)

𝑥3 =
2 (3) ∴The solution
to the given system of equation is

𝑥1 = 3, 𝑥2 = 1 𝑎𝑛𝑑 𝑥3 = 2.

Example: 2.2.4

87
 Solve the equations𝑥1 + 𝑥2 + 𝑥3 = 6, 3𝑥1 + (3 + 𝜖)𝑥2 + 4𝑥3 =
SPACE FOR HINTS 20𝑎𝑛𝑑 2𝑥1 + 𝑥2 + 3𝑥3 = 13using Gauss elimination method where is
small such that 1 ± 𝜖 2 ≈ 1

Solution:

The given system of equation can be written as 𝑨𝑿 = 𝑩

1 1 1 𝑥1 6
𝑖𝑒) 3 3+𝜖 4 𝑥2 = 20
2 1 3 𝑥3 13

The augmented matrix is

1 1 1 6 1 1 1 6 𝑅 → 𝑅 − 3𝑅
2 2 1
3 3+𝜖 4 20 ≅ 0 𝜖 1 2 𝑅 → 𝑅 − 2𝑅
3 3 1
2 1 3 13 0 −1 11

1 1 16
≅ 0 −1 1 1 𝑅2 ↔ 𝑅3
0 𝜖 12

1 1 1 6
≅ 0 −1 1 1 𝑅3 → 𝑅3 + 𝜖𝑅2
0 0 1+𝜖 2+𝜖

From the above matrix, the system of equation can be written as

𝑥1 + 𝑥2 + 𝑥3 = 6 (1)

−𝑥2 + 𝑥3 = 1 (2)

1 + 𝜖 𝑥3 = 2 + 𝜖 (3)

2+𝜖
𝑥3 =
1+𝜖

2+𝜖
2 ⟹ −𝑥2 + =1
1+𝜖

2+𝜖
⟹ 𝑥2 = −1 +
1+𝜖

88
 1
𝑥2 = SPACE OF HINTS
1+𝜖

1 2+𝜖
3 ⟹ 𝑥1 + + =6
1+𝜖 1+𝜖

3+𝜖
⟹ 𝑥1 + =6
1+𝜖

3+𝜖
⟹ 𝑥1 = 6 −
1+𝜖

3 + 5𝜖
𝑥1 =
1+𝜖

∴The solution to the given system of equation is

3 + 5𝜖 1 2+𝜖
𝑥1 = , 𝑥2 = 𝑎𝑛𝑑𝑥3 =
1+𝜖 1+𝜖 1+𝜖

Example: 2.2.5

2 1 1 −2 𝑥1 −10
4 0 2 1 𝑥2 8
Solve the system of equation
3 2 2 0 𝑥3 = 7
1 3 2 −1 𝑥4 −5

Solution:

The augmented matrix is,

2 1 1 −2
2 1 1 −2 −10 0 −2 0 5 −10 𝑅2 → 𝑅2 − 2𝑅1
1 1 3
4 0 2 1 8 0 2 2 3 28 𝑅3 → 𝑅3 − 2 𝑅1
≅
3 2 2 0 7 22
1
1 3 2 −1 −5 0 5 3 0 0 𝑅4 → 𝑅4 − 𝑅
2 2 2 1

89
 2 1 1 −2
0 −2 0 5 −10 1
SPACE FOR HINTS 1 17 𝑅3 → 𝑅3 − 𝑅2
0 0 2 4 28 4
≅
29 5
3 25 35 𝑅4 → 𝑅4 + 𝑅2
0 0 2 4 4

2 1 1 −2
0 −2 0 5 −10
0 0 1 17
≅ 2 4 28 𝑅 → 𝑅 − 3𝑅
4 4 3
29
0 0 0 −26 −52
4

−26
𝑥 = −52 1
4 4

4
⟹ 𝑥4 = −52 ×
−26

𝑥4 = 8

1 17
𝑥3 + 𝑥 = 29 (2)
2 4 4

1 17
⟹ 𝑥3 = 29 − 𝑥
2 4 4

𝑥3 = −10

−2𝑥2 + 5𝑥4 = 28 (3)

⟹ −2𝑥2 = 28 − 5 × 8

𝑥2 = 6

2𝑥1 + 𝑥2 + 𝑥3 − 2𝑥4 = −10

2𝑥1 = −10 − 𝑥2 − 𝑥3 + 2𝑥4

= −10 − 6 + 10 + 2(8)

2𝑥1 = 10 ⟹ 𝑥1 = 5

90
∴The solution is 𝑥1 = 5, 𝑥2 = 6, 𝑥3 = −10 𝑎𝑛𝑑 𝑥4 = 8.
SPACE OF HINTS

Gauss-Jordon Elimination method

1. Here the coefficient matrix is reduced to a diagonal matrix rather

than a triangular matrix.
2. At all steps of gauss elimination method, the eliminates is alone
not only in the equation below but the equation above the pivots,
producing the solution without using the back substitution
method on the completion of the Gauss-Jordon method, the given
equation becomes,

1 0 ⋯ 0 𝑥1 𝑑1
0 1 ⋯ 0 𝑥2 𝑑2
⋮ ⋮ ⋱ ⋮ ⋮ = ⋮
0 0 ⋯ 1 𝑥𝑛 𝑑𝑛

The solution is given by 𝑥𝑖 = 𝑑𝑖 , 𝑖 = 1,2, … , 𝑛.

𝐺𝑎𝑢𝑠𝑠𝐽𝑜𝑟𝑑𝑜𝑛
Hence the gauss Jordon method gives 𝐴 | 𝑏 𝐼|𝑑

For inverse of 𝐴, we start with a augmented matrix of 𝐴 with the identity

matrix 𝐼 of the same order

𝐺𝑎𝑢𝑠𝑠𝐽𝑜𝑟𝑑𝑜𝑛
𝑖𝑒) 𝐴 | 𝐼 𝐼 | 𝐴−1

Example:2.2.6

Find the inverse of the co-efficient matrix of the system

1 1 1 𝑥1 1
4 3 𝑥
−1 2 = 6 .By the Gauss Jordon method with partial and
3 5 3 𝑥3 4
hence solve the system.

Solution:

Using the augmented matrix

91
 1 1 1 0 1 0
𝐴 𝐼 = 4 3 −1 1 0 0
SPACE FOR HINTS
3 5 3 0 0 1

4 3 −1 0 1 0
~ 1 1 1 1 0 0 𝑅1
3 5 3 0 0 1
1 3 4
−1
40
1
4 0 𝑅1
↔ 𝑅2 ~ 𝑅1 →
1 1 1 1 0 0 4
3 5 3 0 0 1

1 3 −1 1
4 40 4 0

1 5 −1 𝑅2 → 𝑅2 − 𝑅1
~ 0 4 4 1 4 0
𝑅3 → 𝑅3 − 3𝑅1
11 15 −3
0 4 4 0 4 1

1 3 −1 1
4 40 4 0

~ 0 1 5 4 −1 0 𝑅2 → 4𝑅2
11 15 −3
0 4 4 0 4 1

1 0 −16 3
4 −3 1 0 𝑅1 → 𝑅1 − 𝑅2
~ 4
4 −1 0 11
0 1 5 −11 2 1 𝑅3 → 𝑅3 − 𝑅
0 0 −10 4 2

−3 1 0
1 0 −4 4 −1 0 𝑅3
~ 0 1 5 11 −2 −1 𝑅3 →
10 10 10 −10
0 0 1

7 1 −2
5 5 5
1 0 0 −3 1 𝑅1 → 𝑅1 + 4𝑅3
~ 0 2 0 2
1 0
𝑅2 → 𝑅2 − 5𝑅3
0 0 1
11 −2 −1
10 10 10

92
 ∴ 𝐴|𝐼 = 𝐼|𝐴−1
SPACE OF HINTS

7 1 −2
5 5 5
−3 0 1
Hence 𝐴−1 = 2 2
11 −2 −1
10 10 10

We know that, 𝐴𝑋 = 𝑏 ⟹ 𝑋 = 𝐴−1 𝑏

7 1 −2
5 5 5
𝑥1 1
−3 0 1
𝑥
𝑖𝑒) 2 = 2 2 6
𝑥3 4
11 −2 −1
10 10 10

7 6 2 5
5+ 5− 5 5
𝑥1
−3 4 1
⟹ 𝑥2 = 2+0+ 2 = 2
𝑥3
11 12 4 −5
10 − 10 − 10 10

1
𝑥1 1
2
𝑥2 =
𝑥3 −1
2

Problem:1

Solve the following system of equation Gauss-Jordon

2 2 1 𝑥1 1
4 2 3 𝑥2 = 2
1 1 1 𝑥3 3

Solution:Using the augmented matrix

93
 2 2 11 0 0
𝐴𝐼 = 4 2 30 1 0
SPACE FOR HINTS
1 1 10 0 1

2 2 11 0 0
~ 1 1 10 0 1 𝑅2 ↔ 𝑅3
4 2 30 1 0

1 1 1 1 0 0
2 2 𝑅1
~ 𝑅1 →
1 1 1 0 0 1 𝑅2
4 2 3 0 1 0

1 1 1 1 0 0
2 2
𝑅2 → 𝑅2 − 𝑅1
~ 0 0 1 −1 0 1 𝑅3 → 𝑅3 − 4𝑅1
2 2
0 −2 1 −2 1 0

1 1 1 1 0 0
2 2
~ 0 −2 1 −2 1 0 𝑅2 ↔ 𝑅3
0 0 1 −1 0 1
2 2

1 1 1 1 0 0
2 2

0 1 −1 1 −1 −𝑅2
~ 2 2 0 𝑅2 →
2
0 0 1 −1 0 1
2 2

−1 2 1
2 0
1 0 1
0 1 −1
2 1 −1
~ 2 0 𝑅1 → 𝑅1 − 𝑅2
0 0 1
2 −1 0 1
2

94
 −1 2 1
2 0
1 0 1 SPACE OF HINTS
0 1 −1
2 1 −1
~ 2 0 𝑅3 → 2𝑅3
0 0 1
2 −1 0 1
2

1 1 −2
2 2
1 0 1 𝑅1 → 𝑅1 − 𝑅3
~ 0 1 01 −1 1 𝑅3
2 2 𝑅2 → 𝑅2 +
0 0 1 2
−1 0 2

𝐴 𝐼 = [𝐼|𝐴−1 ]

1 1 −2
2 2
Hence 𝐴−1 = 1 −1 1
2 2
−1 0 2

We know that, 𝐴𝑋 = 𝑏 ⟹ 𝑋 = 𝐴−1 𝑏

1 1 −2
2 2
𝑥1 1
𝑖𝑒) 𝑥2 = 1 −1 1 2
𝑥3 2 2 3
−1 0 2

1
+1−6
2
= 1
−1+3
2
−1 + 0 + 6

−9
2
𝑥1
𝑥2 = 5
𝑥3 2
5

Problem:2

95
 1 2 1
Find the inverse of the matrix by Gauss-Jordon 2 3 −1
SPACE FOR HINTS
2 −1 3
Solution:
Using augmented matrix
1 2 1 1 0 0
𝐴𝐼 = 2 3 −1 0 1 0
2 −1 3 0 0 1

1 2 1 1 0 0
2 1 −1 1 0 𝑅 → 𝑅2
~ 3 30 3 2
3
2 −1 3 0 0 1

1 2 1 1 0 0
−1 2
~ 0 −1 −2 3 1 0 𝑅2 → 𝑅2 − 𝑅
3 3 3 1
𝑅3 → 𝑅3 − 2𝑅1
0 −5 1 −2 0 1

1 2 1 1 0 0
~ 0 1 3 2 −1 0 𝑅2 → −3𝑅2
0 −5 1 −2 0 1

1 0 −5 −3 2 0 𝑅 → 𝑅 − 2𝑅
1 1 2
~ 0 1 3 2 −1 0
𝑅 → 𝑅3 + 5𝑅2
0 0 16 8 −5 1 3

−3 2 0
1 0 −5 2 −1 0 𝑅3
~ 0 1 3 8 −5 1 𝑅3 →
16 16 16 16
0 0 1

−8 7 5
16 16 16
1 0 0 8 −1 −3 𝑅1 → 𝑅1 + 5
~ 0 1 0 16 16 16
𝑅2 → 𝑅2 − 3
0 0 1
8 −5 1
16 16 16

𝐴 𝐼 = [𝐼|𝐴−1 ]

96
 −8 7 5
16 16 16 SPACE OF HINTS

8 −1 −3
Hence 𝐴−1 = 16 16 16
8 −5 1
16 16 16

Triangularization method:

This method is also known as thedecomposition method or

thefactorization method. In this method,the coefficient matrix A of the
systemof equations (I) is decomposed orfactorized into the product of a
lowertriangular matrix 𝑳 and an uppertriangular matrix 𝑼. We write
thematrix A as𝑨 = 𝑳 𝑼

where
𝑙11 0 0 … 0
𝑙21 𝑙22 0 … 0
𝑙31 𝑙32 𝑙33 … 0
⋮ ⋮ ⋮
𝑙𝑛1 𝑙𝑛2 𝑙𝑛3 … 𝑙𝑛𝑛

Example: 2.2.7

Consider the equation 𝑥1 + 𝑥2 + 𝑥3 = 1, 4𝑥1 + 3𝑥2 − 𝑥3 = 6 𝑎𝑛𝑑

3𝑥1 + 2𝑥2 + 3𝑥3 = 4 use the decomposition method to solve the

system.

Solution:

The given system of equation can be written as 𝑨𝑿 = 𝒃

1 1 1 𝑥1 1
𝑖𝑒) 4 3 −1 𝑥2 = 6
3 5 3 𝑥3 4

In Triangularization method 𝐴 = 𝐿𝑈 and let 𝑙𝑖𝑖 = 1

97
 1 1 1 1 0 0 𝑢11 𝑢12 𝑢13
4 3 𝑙
−1 = 21 1 0 0 𝑢22 𝑢23
SPACE FOR HINTS
3 5 3 𝑙31 𝑙32 1 0 0 𝑢33

1 1 1 𝑢11 𝑢12 𝑢13

4 3 −1 = 𝑢11 𝑙21 𝑢12 𝑙21 + 𝑢22 𝑢13 𝑙21 + 𝑢23
3 5 3 𝑙31 𝑢11 𝑙31 𝑢12 +𝑙32 𝑢22 𝑢13 𝑙31 +𝑙32 𝑢23 + 𝑢33

𝑢11 = 1 𝑢12 = 1 𝑢13 = 1

𝑢11 𝑙21 = 4 ⟹ 𝑙21 = 4

𝑢12 𝑙21 + 𝑢22 = 3 ⟹ 4 + 𝑢22 = 3 ⟹ 𝑢22 = −1

𝑢13 𝑙21 + 𝑢23 = −1 ⟹ 4 + 𝑢23 = −1 ⟹ 𝑢23 = −5

𝑙31 𝑢11 = 3 ⟹ 𝑙31 = 3

𝑙31 𝑢12 +𝑙32 𝑢22 = 5 ⟹ 3 − 𝑙32 = 5 ⟹ 𝑙32 = −2

𝑢13 𝑙31 +𝑙32 𝑢23 + 𝑢33 = 3 ⟹ 3 + −2 × −5 + 𝑢33 = 3

⟹ 𝑢33 = −10

1 0 0 1 0 0
∴ 𝐿 = 𝑙21 1 0 = 4 1 0
𝑙31 𝑙32 1 3 −2 1

𝑢11 𝑢12 𝑢13 1 1 1

𝑈= 0 𝑢22 𝑢21 = 0 −1 −5
0 0 𝑢33 0 0 −10

We know that, 𝐴𝑋 = 𝑏

⟹ 𝐿𝑈𝑋 = 𝑏 [∵ 𝐴 = 𝐿𝑈] 1

Here 𝑈𝑋 = 𝑍 (2)

∴ 1 ⟹ 𝐿𝑍 = 𝑏

98
 1 0 0 𝑧1 1
4 1 0 𝑧2 = 6 SPACE OF HINTS
3 −2 1 𝑧3 4

𝑧1 = 1

4𝑧1 + 𝑧2 = 6 ⟹ 4 + 𝑧2 = 6 ⟹ 𝑧2 = 2

3𝑧1 − 2𝑧2 + 𝑧3 = 4 ⟹ 𝑧3 = 5

2 ⟹ 𝑈𝑋 = 𝑍

1 1 1 𝑥1 1
𝑖𝑒) 0 −1 −5 𝑥2 = 2
0 0 𝑥
−10 3 5

−1
∴ −10𝑥3 = 5 ⟹ 𝑥3 =
2

5 1
−𝑥2 − 5𝑥3 = 2 ⟹ 𝑥2 = −2 + ⟹ 𝑥2 =
2 2

𝑥1 + 𝑥2 + 𝑥3 = 1 ⟹ 𝑥1 = 1 − 𝑥2 − 𝑥3

1 1
⟹1− + ⟹ 𝑥1 = 1
2 2

3 2 1
Example: 2.2.8 Find the inverse of the matrix𝐴 = 2 3 2 using
1 2 2
𝐿𝑈 decomposition method. Take 𝑢11 = 𝑢22 = 𝑢33 = 1.

Solution:

The given system of equation can be written as 𝐴𝑋 = 𝑏.

In Triangularization method 𝐴 = 𝐿𝑈and 𝑢𝑖𝑖 = 1.

3 2 1 𝑙11 0 0 1 𝑢12 𝑢13

𝑖𝑒) 2 3 2 = 𝑙21 𝑙22 0 0 1 𝑢21
1 2 2 𝑙31 𝑙32 𝑙33 0 0 1

99
 3 2 1 𝑙11 𝑢12 𝑙11 𝑢13 𝑙11
SPACE FOR HINTS
⟹ 2 3 2 = 𝑙21 𝑙12 𝑢21 + 𝑙22 𝑢13 𝑙21 + 𝑙22 𝑢23
1 2 2 𝑙31 𝑙31 𝑢12 +𝑙32 𝑢13 𝑙31 +𝑙32 𝑢23 + 𝑙33

𝑙11 = 3 𝑙21 = 2 𝑙31 = 1

2
𝑢12 𝑙11 = 2 ⟹ 𝑢12 =
3

1
𝑢13 𝑙11 = 1 ⟹ 𝑢13 =
3

4 5
𝑙12 𝑢21 + 𝑙22 = 3 ⟹ 𝑙22 = 3 − ⟹ 𝑙22 =
3 3

1 5
𝑢13 𝑙21 + 𝑙22 𝑢23 = 2 ⟹ 2 + 𝑢23 = 2
3 3

2 3
⟹ 𝑢23 = 2 − ×
3 5

4
⟹ 𝑢23 =
5

2 4
𝑙31 𝑢12 +𝑙32 = 2 ⟹ 𝑙32 = 2 − ⟹ 𝑙32 =
3 3

1 4 4
𝑢13 𝑙31 +𝑙32 𝑢23 + 𝑙33 = 2 ⟹ + × + 𝑙33 = 2
3 3 5

1 16 3
⟹ 𝑙33 = 2 − − ⟹ 𝑙33 =
3 15 5

3 0 0
𝑙11 0 0 2 5 0
3
∴ 𝐿 = 𝑙21 1 0 =
𝑙31 𝑙32 𝑙33 1 4 3
3 5

100
 1 2 1
3 3 SPACE OF HINTS
1 𝑢12 𝑢13
𝑈= 0 1 𝑢21 = 0 1 4
0 0 1 5
0 0 1

To find𝑨−𝟏 :𝑖𝑒) 𝐴−1 = (𝐿𝑈)−1 = 𝑈 −1 𝐿−1

1 2 1
3 3
𝑈 = 0 1 4 =1 1−0 =1
5
0 0 1

3 0 0
2 5 0
3 15
𝐿 = =3 −0 =3
4 3 15
1 3 5

1
𝑈 −1 = 𝑎𝑑𝑗 𝑈
𝑈

𝑇
1 0 0 1 −2 1
−2 3 5
3 1 0
𝑎𝑑𝑗 𝑈 = = 0 1 −4
1 −4 5
5 5 1
0 0 1

1 −2 1
3 5
𝑈 −1 = 0 1 −4
5
0 0 1

1
𝐿−1 = 𝑎𝑑𝑗 𝐿
𝐿

101
 𝑇
1 −6 1
5
SPACE FOR HINTS 1 0 0
9 −12 −6 9
𝑎𝑑𝑗 𝐿 = 0 5 3 = 5 5 0

0 0 15 1 −4 5
3

1 0 0
3
1 0 0
1 −6 9 0 = −2 5 3 0
𝐿−1 = 5 5 5
3
1 −4 5 1 −4 5
3 3 3

𝐴−1 = 𝑈 −1 𝐿−1

1 0 0
1 −2 1 3
3 5
−2 3 0
= 0 1 −4 5 5
5
1 −4 5
0 0 1 3 3 3

1 4 1 6 4 1
3+ 15 + 15 0 − 15 − 15 0 + 0 + 3

= 0 − 2 5 − 4 15 0 + 3 5 + 16 15 0 + 0 − 4 3

0+0+1 3 0+0−4 3 0+0+5 3

2 −2 1
3 3 3
−2 5 −4
= 3 3 3
1 −4 5
3 3 3

102
 1 2 −2 1
𝐴−1 = −2 5 −4 SPACE OF HINTS
3
1 −4 5

Example: 2.2.9

Show that the 𝐿𝑈 decomposition to solve the system of equation

1 1 −1 𝑥1 2
2 2 5 𝑥2 = −3 exact solution is 𝑥1 = 1, 𝑥2 = 0, 𝑥3 = −1.
3 2 −3 𝑥3 6

Solution:

Let 𝐴 = 𝐿𝑈 and take 𝑙𝑖𝑖 = 1

1 1 −1 1 0 0 𝑢11 𝑢12 𝑢13

2 2 𝑙
5 = 21 1 0 0 𝑢22 𝑢21
3 2 −3 𝑙31 𝑙32 1 0 0 𝑢33

𝑢11 𝑢12 𝑢13

= 11 𝑙21
𝑢 𝑢12 𝑙21 + 𝑢22 𝑢13 𝑙21 + 𝑢23
𝑙31 𝑢11 𝑙31 𝑢12 +𝑙32 𝑢22 𝑢13 𝑙31 +𝑙32 𝑢23 + 𝑢33

𝑢11 = 1 𝑢12 = 1 𝑢13 = −1

𝑢11 𝑙21 = 2 ⟹ 𝑙21 = 2

𝑢12 𝑙21 + 𝑢22 = 2 ⟹ 2 × 1 + 𝑢22 = 2 ⟹ 𝑢22 = 0

𝑢13 𝑙21 + 𝑢23 = 5 ⟹ 2 × (−1) + 𝑢23 = 5 ⟹ 𝑢23 = 7

𝑙31 𝑢11 = 3 ⟹ 𝑙31 = 3

𝑙31 𝑢12 +𝑙32 𝑢22 = 2 ⟹ 3 × 1 + 𝑙32 × 0 = 2 ⟹ 3 = 2 it is not

possible.

𝑢13 𝑙31 +𝑙32 𝑢23 + 𝑢33 = −3

∴ In the above process 𝑢22 = 0 ⟹ 𝐿𝑈 decomposition method fails as

the pivot element 𝑢22 = 0.

103
 Now, take 𝑢𝑖𝑖 = 1, 𝐴 = 𝐿𝑈
SPACE FOR HINTS
1 1 −1 𝑙11 0 0 1 𝑢12 𝑢13
2 2 5 = 𝑙21 𝑙22 0 0 1 𝑢21
3 2 −3 𝑙31 𝑙32 𝑙33 0 0 1

𝑙11 𝑢12 𝑙11 𝑢13 𝑙11

= 𝑙21 𝑙21 𝑢21 + 𝑙22 𝑢13 𝑙21 + 𝑙22 𝑢23
𝑙31 𝑙31 𝑢12 +𝑙32 𝑢13 𝑙31 +𝑙32 𝑢23 + 𝑙33

𝑙11 = 1 𝑙21 = 2 𝑙31 = 3

𝑢12 𝑙11 = 1 ⟹ 𝑢12 = 1

𝑢13 𝑙11 = −1 ⟹ 𝑢13 = −1

𝑙21 𝑢21 + 𝑙22 = 2 ⟹ 2 × 1 + 𝑙22 = 2 ⟹ 𝑙22 = 0

𝑢13 𝑙21 + 𝑙22 𝑢23 = 5 ⟹ 2 × −1 + 0 × 𝑢23 = 5

⟹ −2 = 5 it is not possible.

∴ 𝐿𝑈 decomposition method fails as the pivot 𝑙22 = 0. Interchanging the

first and 3rd row of the system of equation.

3 2 −3 𝑥1 6
2 2 5 𝑥2 = −3
1 1 𝑥
−1 3 2

Now let 𝐴 = 𝐿𝑈 and let 𝑙𝑖𝑖 = 1

3 2 −3 1 0 0 𝑢11 𝑢12 𝑢13

2 2 5 = 𝑙21 1 0 0 𝑢22 𝑢21
1 1 −1 𝑙31 𝑙32 1 0 0 𝑢33

𝑢11 𝑢12 𝑢13

= 𝑢11 𝑙21 𝑢12 𝑙21 + 𝑢22 𝑢13 𝑙21 + 𝑢23
𝑙31 𝑢11 𝑙31 𝑢12 +𝑙32 𝑢22 𝑢13 𝑙31 +𝑙32 𝑢23 + 𝑢33

𝑢11 = 3 𝑢12 = 2 𝑢13 = −3

104
 2
𝑢11 𝑙21 = 2 ⟹ 𝑙21 = SPACE OF HINTS
3

2
𝑢12 𝑙21 + 𝑢22 = 2 ⟹ × 2 + 𝑢22 = 2
3

4 2
⟹ 𝑢22 = 2 − ⟹ 𝑢22 =
3 3

2
𝑢13 𝑙21 + 𝑢23 = 5 ⟹ × −3 + 𝑢23 = 5 ⟹ 𝑢23 = 7
3

1
𝑙31 𝑢11 = 1 ⟹ 𝑙31 =
3

2 2 1
𝑙31 𝑢12 +𝑙32 𝑢22 = 1 ⟹ × 2 + 𝑙32 × = 1 ⟹ 𝑙32 =
3 3 2

𝑢13 𝑙31 +𝑙32 𝑢23 + 𝑢33 = −1

1 1 −7
⟹ × −3 + × 7 + 𝑢33 = −1 ⟹ 𝑢33 =
3 2 2

1 0 0
2 1 0
3
∴𝐿=
1 1
3 2 1

3 2 −3
0 2 7
3
𝑈=
0 0 −7
2

We know that, 𝐴𝑋 = 𝑏

⟹ 𝐿𝑈𝑋 = 𝑏 (1)

Here 𝑈𝑋 = 𝑍 2

105
 1 ⟹ 𝐿𝑍 = 𝑏 (3)
SPACE FOR HINTS
1 0 0
2 1 0 𝑧1
3 6
𝑧2 = −3
1
3
1
2 1 𝑧3 2

𝑧1 = 6

2
𝑧 + 𝑧2 = −3 ⟹ 𝑧2 = −7
3 1

1 1 7
𝑧1 + 𝑧2 + 𝑧3 = 2 ⟹ 𝑧3 =
3 2 2

6
𝑧1
−7
∴ 𝑧2 = 7
𝑧3 2

2 ⟹ 𝑈𝑋 = 𝑍

3 2 −3
2 6
0 3 7 𝑥1
−7
⟹ 𝑥2 = 7
0 0 −7 𝑥3 2
2

−7 7
𝑥3 = ⟹ 𝑥3 = −1
2 2

2
𝑥 + 7𝑥3 = −7 ⟹ 𝑥2 = 0
3 2

3𝑥1 + 2𝑥2 − 3𝑥3 = 6 ⟹ 𝑥1 = 1

𝑥1 1
𝑥
∴ 2 = 0
𝑥3 −1

Example: 2.2.10

106
Solve the system of equations 𝐴𝑋 = 𝑏,
SPACE OF HINTS

2 1 1 −2 −10
4 0 2 1 8
where 𝐴 = 𝑏= .
3 2 2 0 7
1 3 2 −1 −5

Using the 𝐿𝑈decomposition method. Take all the diagonal elements of L

as I. Also find𝐴−1 .

Solution:

Let 𝐴 = 𝐿𝑈 and given 𝑙𝑖𝑖 = 1.

2 1 1 −2 1 0 0 0 𝑢11 𝑢12 𝑢13 𝑢14

4 0 2 1 𝑙 1 0 0 0 𝑢22 𝑢23 𝑢24
= 21 0 0 𝑢33 𝑢34
3 2 2 0 𝑙31 𝑙32 1 0
1 3 2 −1 𝑙41 𝑙42 𝑙43 1 0 0 0 𝑢44

𝑢11 𝑢12 𝑢13 𝑢14

𝑢11 𝑙21 𝑢12 𝑙21 + 𝑢22 𝑢13 𝑙21 + 𝑢23 𝑢14 𝑙21 + 𝑢24
= 𝑢 𝑙 𝑢12 𝑙31 + 𝑢22 𝑙32 𝑢13 𝑙31 + 𝑢23 𝑙32 + 𝑢33 𝑢14 𝑙31 + 𝑢34 𝑙32 + 𝑢34
11 31
𝑢11 𝑙41 𝑢12 𝑙41 + 𝑢22 𝑙42 𝑢13 𝑙41 + 𝑢23 𝑙42 + 𝑙43 𝑢33 𝑢14 𝑙41 + 𝑢24 𝑙42 + 𝑢34 𝑙43 + 𝑢44

𝑢11 = 2 𝑢12 = 1 𝑢13 = 1 𝑢14 = −2

𝑢11 𝑙21 = 4 ⟹ 𝑙21 = 2

𝑢12 𝑙21 + 𝑢22 = 0 ⟹ 𝑢22 = −2

𝑢13 𝑙21 + 𝑢23 = 2 ⟹ 𝑢23 = 0

𝑢14 𝑙21 + 𝑢24 = 1 ⟹ 𝑢24 = 5

𝑢11 𝑙31 = 3 ⟹ 𝑙31 = 3 2

−1
𝑢12 𝑙31 + 𝑢22 𝑙32 = 2 ⟹ 𝑙32 =
4

1
𝑢13 𝑙31 + 𝑢23 𝑙32 + 𝑢33 = 2 ⟹ 𝑢33 =
2

107
 17
𝑢14 𝑙31 + 𝑢34 𝑙32 + 𝑢34 = 0 ⟹ 𝑢34 =
SPACE FOR HINTS 4

1
𝑢11 𝑙41 = 4 ⟹ 𝑙41 =
2

−5
𝑢12 𝑙41 + 𝑢22 𝑙42 = 3 ⟹ 𝑙42 =
4

𝑢13 𝑙41 + 𝑢23 𝑙42 + 𝑙43 𝑢33 = 2 ⟹ 𝑙43 = 3

−13
𝑢14 𝑙41 + 𝑢24 𝑙42 + 𝑢34 𝑙43 + 𝑢44 = −1 ⟹ 𝑢44 =
2

1 0 0 0
2 1 0 0
3 −1 1 0
∴ 𝐿= 2 4
1 −5 3 1
2 4

2 1 1 −2
0 −2 0 5
0 0 1 17
𝑈= 2 4

0 0 0 −13
4

We know that, 𝐴𝑋 = 𝑏

⟹ 𝐿𝑈𝑋 = 𝑏 (1)

Here 𝑈𝑋 = 𝑍 2

1 ⟹ 𝐿𝑍 = 𝑏 (3)

108
 1 0 0 0
2 1 0 0 𝑧1 −10
SPACE OF HINTS
3 −1 0 𝑧2
2 4 1 8
𝑧3 = 7
1
2
−5
4 3 1 𝑧4 −5

𝑧1 = −10

2𝑧1 + 𝑧2 = 8 ⟹ 𝑧2 = 28

3 1
𝑧1 − 𝑧2 + 𝑧3 = 7 ⟹ 𝑧3 = 29
2 4

1 5
𝑧1 − 𝑧2 + 3𝑧3 + 𝑧4 = −5 ⟹ 𝑧4 = −52
2 4

𝑧1 −10
𝑧2 28
∴ 𝑧 =
3 29
𝑧4 −52

2 1 1 −2
0 −2 0 5 𝑥1 −10
0 0 1 17 𝑥2 28
2 ⟹ 2 4
𝑥3 = 29
0 0 0 −13 𝑥4 −52
4

−13
𝑥 = −52 ⟹ 𝑥4 = 8
4 4

1 17
𝑥3 + 𝑥 = 29 ⟹ 𝑥3 = −10
2 4 4

−2𝑥2 + 5𝑥4 = 28 ⟹ 𝑥2 = 6

2𝑥1 + 𝑥2 + 𝑥3 − 2𝑥4 = −10 ⟹ 𝑥1 = 5

109
 𝑥1 5
𝑥2 6
SPACE FOR HINTS ∴ 𝑥 =
3 −10
𝑥4 8

To find 𝑨−𝟏 :𝑖𝑒) 𝐴−1 = (𝐿𝑈)−1 = 𝑈 −1 𝐿−1

1
𝐿 = 1 𝑎𝑛𝑑 𝑈 =
13

1
𝑈 −1 = 𝑎𝑑𝑗 𝑈
𝑈

13 0 0 0
2
13 −13 0 0
𝑎𝑑𝑗 𝑈 = 4 2
−13 0 26 0
−8 −5 17 −2

13 13 −13 −8
2 4
1 −13
𝑈 −1 = 0 2 0 −5
13
0 0 26 0
0 0 17 −2

1
𝐿−1 = 𝑎𝑑𝑗 𝑈
𝐿

𝑇
1 −2 −2 3
1
1 0 1 14 2
𝐿−1 =
1
0 0 1 −3
0 0 0 1

1 0 0 0
−2 1 0 0
−2 1 1 0
𝐿−1 = 4

3 1 −3 1
2

110
 𝐴−1 = 𝑈 −1 𝐿−1
SPACE OF HINTS

13 13 −13 −8 1 0 0 0
2 4 −2 1 0 0
1 −13 −2 1 1 0
= 0 0 −5 4
13 2
0 0 26 0 3 1 −3 1
2
0 0 17 −2

2 −4 11 −8
1 −2 −9 15 −5
𝐴−1 =
13 −1 15 −25 17
−6 −1 6 −2

Cholesky method:

This method is also known as square – root method. The given system
of equation is of the form 𝐴𝑋 = 𝑏. (1)

Where the coefficient matrix 𝐴 is symmetric and positive definite, then

the matrix 𝐴 can be decomposed as,

𝐴 = 𝐿𝐿𝑇 (2)

(or)𝐴 = 𝑈𝑈 𝑇 (3)

where 𝑙𝑖𝑗 = 0, 𝑖 < 𝑗 𝑎𝑛𝑑 𝐿 = 𝑙𝑖𝑗

Similarly 𝑈 = 𝑢𝑖𝑗 , 𝑢𝑖𝑗 = 0, 𝑖 > 𝑗.

∴ 1 can be written as

𝐿𝐿𝑇 𝑋 = 𝑏 𝑜𝑟 𝑈𝑈 𝑇 𝑋 = 𝑏 (4)

Let 𝐿−1 𝑋 = 𝑧 (5)

4 ⟹ 𝐿𝑍 = 𝑏 (6)

111
 From (6), we find the unknown value 𝑧, is obtained from by forward
SPACE FOR HINTS substitution and that determine the value 𝑥𝑖 , 𝑖 = 1,2, … , 𝑛 by backward
substitution. The inverse of 𝐴 can be obtained from

2 ⟹ 𝐴−1 = 𝐿𝐿𝑇 −1

∴ 𝐴−1 = 𝐿𝑇 −1 −1
𝐿

Example: 2.2.11

1 2 3 𝑥1 5
Solve the system of equations 2 8 22 𝑥2 = 6 using
3 22 82 𝑥3 −10
cholesky method.

Solution:

1 2 3
Given 𝐴 = 2 8 22
3 22 82

The matrix 𝐴 is symmetric and positive definite.

1 2
Since 1 = 1, =8−4=4
2 8

1 2 3
2 8 22 = 1 × 656 − 484 − 2 × 164 − 66 + 3 × (44 − 24)
3 22 82

= 36

Here all the values are positive

By cholesky method 𝐴 = 𝐿𝐿−1

1 2 3 𝑙11 0 0 𝑙11 𝑙12 𝑙13

𝑖𝑒) 2 8 22 = 𝑙21 𝑙22 0 0 𝑙22 𝑙23
3 22 82 𝑙31 𝑙32 𝑙33 0 0 𝑙33

112
 𝑙11 2 𝑙11 𝑙21 𝑙11 𝑙31
SPACE OF HINTS
= 𝑙21 𝑙11 𝑙21 2 + 𝑙22 2 𝑙21 𝑙31 + 𝑙22 𝑙32
𝑙31 𝑙11 𝑙21 𝑙31 + 𝑙22 𝑙32 𝑙31 2 + 𝑙32 2 + 𝑙33 2

𝑙11 2 = 1 ⟹ 𝑙11 = 1 𝑙11 𝑙21 = 2 ⟹ 𝑙21 = 2

𝑙11 𝑙31 = 3 ⟹ 𝑙31 = 3

𝑙21 2 + 𝑙22 2 = 8 ⟹ 4 + 𝑙22 2 = 8 ⟹ 𝑙22 2 = 4 ⟹ 𝑙22 = 2

𝑙21 𝑙31 + 𝑙22 𝑙32 = 22 ⟹ 𝑙32 = 8

𝑙31 2 + 𝑙32 2 + 𝑙33 2 = 82 ⟹ 𝑙33 2 = 9 ⟹ 𝑙33 = 3

1 0 0
∴𝐿= 2 2 0
8 8 3

Now,𝐴𝑋 = 𝑏 ⟹ 𝐿𝐿−1 𝑋 = 𝑏 (1)

Here, 𝐿 −1 𝑋 = 𝑧 2

∴ 1 ⟹ 𝐿𝑍 = 𝑏 (3)

1 0 0 𝑧1 5
𝑖𝑒) 2 2 0 𝑧2 = 6
8 8 3 𝑧3 −10

𝑧1 = 5 2𝑧1 + 2𝑧2 = 6 ⟹ 𝑧2 = −2

3𝑧1 + 8𝑧2 + 3𝑧3 = −10 ⟹ 𝑧3 = −3

1 2 3 𝑥1 5
∴ 2 ⟹ 0 2 𝑥
8 2 = −2
0 0 3 𝑥3 −3

3𝑥3 = −3 ⟹ 𝑥3 = −1

2𝑥2 + 8𝑥3 = −2 ⟹ 𝑥2 = 3

113
 𝑥1 + 2𝑥2 + 3𝑥3 = 5 ⟹ 𝑥1 = 2
SPACE FOR HINTS
𝑥1 2
∴ The solution of given system of equation is 𝑥2 = 3 .
𝑥3 −1

Example: 2.2.12

4 −1 0 0 𝑥1 1
−1 4 −1 0 𝑥2 0
Solve the equation
0 −1 4 −1 𝑥3 = 0 using the cholesky
0 0 −1 4 𝑥4 0
method. Determine 𝐴−1 .

Solution:

4 −1 0 0
−1 4 −1 0
Given 𝐴 =
0 −1 4 −1
0 0 −1 4

The matrix A is symmetric and positive definite.

4 −1
Since 4 = 4, = 16 − 1 = 15,
−1 4

4 −1 0
−1 4 −1 = 4 × 16 − 1 + 1 × −4 = 56
0 −1 4

4 −1 0 0
4 −1 0 −1 −1 0
−1 4 −1 0
= 4 −1 4 −1 + 0 4 −1
0 −1 4 −1
0 −1 4 0 −1 4
0 0 −1 4

= 4 × 56 + 16 − 1 + 1 × 0

= 224 − 15 = 209

(or)𝐴 is Hermitian and strictly diagonally dominant. Because

𝐴 = 𝐴 𝑇 and all diagonal entries are ≥ 1.

∴ The given matrix 𝐴 is positive definite. Let 𝐴 = 𝐿𝐿𝑇

114
 𝑙11 0 0 0 𝑙11 𝑙21 𝑙31 𝑙41
𝑙 𝑙22 0 0 0 𝑙22 𝑙23 𝑙24 SPACE OF HINTS
𝐴 = 21
𝑙31 𝑙32 𝑙33 0 0 0 𝑙33 𝑙34
𝑙41 𝑙42 𝑙43 𝑙44 0 0 0 𝑙44

𝑙11 2 𝑙11 𝑙21 𝑙11 𝑙31 𝑙11 𝑙41

𝑙 𝑙 𝑙21 2 + 𝑙22 2 𝑙21 𝑙31 + 𝑙22 𝑙32 𝑙21 𝑙41 + 𝑙22 𝑙42
= 21 11
𝑙31 𝑙11 𝑙21 𝑙31 + 𝑙22 𝑙32 𝑙31 2 + 𝑙32 2 + 𝑙33 2 𝑙31 𝑙41 + 𝑙32 𝑙42 + 𝑙33 𝑙43
𝑙41 𝑙11 𝑙41 𝑙21 + 𝑙42 𝑙22 𝑙41 𝑙31 + 𝑙42 𝑙32 + 𝑙43 𝑙32 𝑙41 2 + 𝑙42 2 + 𝑙43 2 + 𝑙44 2

𝑙11 2 = 4 ⟹ 𝑙11 = 2

−1
𝑙11 𝑙21 = −1 ⟹ 𝑙21 =
2

𝑙31 𝑙11 = 0 ⟹ 𝑙31 = 0

𝑙41 𝑙11 = 0 ⟹ 𝑙41 = 0

15 15
𝑙21 2 + 𝑙22 2 = 4 ⟹ 𝑙22 2 = ⟹ 𝑙22 =
4 4

4
𝑙21 𝑙31 + 𝑙22 𝑙32 = −1 ⟹ 𝑙32 = −
15

𝑙21 𝑙41 + 𝑙22 𝑙42 = 0 ⟹ 𝑙42 = 0

56 56
𝑙31 2 + 𝑙32 2 + 𝑙33 2 = 4 ⟹ 𝑙33 2 = ⟹ 𝑙33 =
15 15

15
𝑙31 𝑙41 + 𝑙32 𝑙42 + 𝑙33 𝑙43 = −1 ⟹ 𝑙43 = −
56

115
 209 209
SPACE FOR HINTS 𝑙41 2 + 𝑙42 2 + 𝑙43 2 + 𝑙44 2 = 4 ⟹ 𝑙44 2 = ⟹ 𝑙44 =
56 56

2 0 0 0
−1 15 0 0
2 4

∴𝐿= 0 − 4 15 56 0
15

0 0 − 15 56 209
56

𝐴𝑋 = 𝑏 ⟹ 𝐿𝐿−1 𝑋 = 𝑏 (1)

Here, 𝐿 −1 𝑋 = 𝑧 2

∴ 1 ⟹ 𝐿𝑍 = 𝑏 (3)

2 0 0 0
−1 15 0 0
2 4
𝑧1 1
𝑧2 0
− 4 15 56
0 15 0 𝑧3 = 0
𝑧4 0

0 0 − 15 56 209
56

1
2𝑧1 = 1 ⟹ 𝑧1 =
2

−1 15 1
𝑧1 + 𝑧2 = 0 ⟹ 𝑧2 =
2 4 60

4 56 1
− 𝑧2 + 𝑧3 = 0 ⟹ 𝑧3 =
15 15 840

116
 15 209 1 SPACE OF HINTS
− 𝑧3 + 𝑧4 = 0 ⟹ 𝑧4 =
56 56 11704

(2) ⟹ 𝐿 −1 𝑋 = 𝑧

2 −1 0 0
2
1
2
0 15 4
4 − 15 0
𝑥1 1
𝑥2 60
56 15 𝑥3 = 1
0 0 15 − 56 𝑥4 840
1
209 11704
0 0 0 56

209 1 1
𝑥4 = ⟹ 𝑥4 =
56 11704 209

56 15 1 56 1 1 4
𝑥3 − 𝑥4 = ⟹ 𝑥3 = ⟹ 𝑥3 =
15 56 840 15 840 209 209

15 4 1 15
𝑥2 − 𝑥3 = ⟹ 𝑥2 =
4 15 60 209

1 1 56
2𝑥1 − 𝑥2 = ⟹ 𝑥1 =
2 2 209

56
209

𝑥1 15
209
𝑥2
∴ 𝑥 =
3 4
𝑥4 209
1
209

117
 To find 𝑳−𝟏
SPACE FOR HINTS
2 0 0 0
−1 15 0 0
2 4

𝐿= 0 − 4 15 56 0
15

0 0 − 15 56 209
56

209 209 209 1

4 60 840 56

0 2 209 15 2 209 210 2 1 14

𝑎𝑑𝑗 𝐿 =

0 0 15 15
56 209 56

0 0 0 56

𝐿 =1
209

𝐿−1

209 0 0 0
4

209 2 209 15 0 0
60
1
=
209 209
840 2 209 210 15
56 209 0

1 2 1 14 15 56
56 56

118
 0.5 0 0 0
0.1291 0.5164 0 0 SPACE OF HINTS
𝐿−1 =
0.0345 0.1380 0.5175 0
0.00921 0.0369 0.1387 0.5176

0.5 0.1291 0.0345 0.00921

0 0.5164 0.1380 0.0369
𝐿−1 𝑇
=
0 0 0.5175 0.1387
0 0 0 0.5176

𝐴−1 = 𝐿−1 𝑇 𝐿−1

0.2679 0.0718 0.0191 0.0048

0.0718 0.2871 0.0766 0.0192
𝐴−1 =
0.0191 0.0765 0.2810 0.0718
0.0048 0.0192 0.0718 0.2679

Partition method

Partition method is used to find the inverse of a matrix 𝐴.Let the

𝐵 𝐶
co-efficient matrix 𝐴 be partition as 𝐴 = where 𝐵 is an𝑟 ×
𝐸 𝐷
𝑟 matrix,

𝐶 is an 𝑟 × 𝑠 matrix, 𝐸 is an 𝑠 × 𝑟 matrix, 𝐷 is an 𝑠 × 𝑠 matrix.

Here 𝑟 and 𝑠 are positive integers with 𝑟 = 𝑠 = 𝑛.Further, we assume

𝑋 𝑌
that 𝐴−1 is partitioned as 𝐴−1 = where 𝑋, 𝑌, 𝑍 𝐴 𝑎𝑛𝑑 𝑉 are
𝑍 𝑉
matrices of the same order as 𝐵, 𝐶, 𝐸, 𝐷 respectively. Then we have

𝐵 𝐶 𝑋 𝑌 𝐼 0
𝐴𝐴−1 = = 1
𝐸 𝐷 𝑍 𝑉 0 𝐼2

Where𝐼1 and𝐼2 are identity matrices of order 𝑟 and 𝑠 respectively. We

have

𝐵𝑋 + 𝐶𝑍 = 𝐼1 (1)

𝐵𝑌 + 𝐶𝑉 = 0 (2)

𝐸𝑋 + 𝐷𝑍 = 0 (3)

119
 𝐸𝑌 + 𝐷𝑉
SPACE FOR HINTS = 𝐼2 (4) 2 ⟹ 𝐵𝑌
= −𝐶𝑉

−𝐶𝑉
⟹𝑌= = −𝐵−1 𝐶𝑉
𝐵

⟹ 𝑌 = −𝐵−1 𝐶𝑉

4 ⟹ 𝐸𝑌 + 𝐷𝑉 = 𝐼2

⟹ 𝐸 −𝐵−1 𝐶𝑉 + 𝐷𝑉 = 𝐼2

⟹ 𝐷 − 𝐸𝐵−1 𝐶 𝑉 = 𝐼2

𝐼2
⟹𝑉=
𝐷 − 𝐸𝐵−1 𝐶

⟹ 𝑉 = 𝐷 − 𝐸𝐵−1 𝐶 −1

1 ⟹ 𝐵𝑋 + 𝐶𝑍 = 𝐼1

⟹ 𝑋 = 𝐵−1 (𝐼1 − 𝐶𝑍)

⟹ 𝑋 = 𝐵−1 − 𝐵−1 𝐶𝑍

(3) ⟹ 𝐸𝑋 + 𝐷𝑍 = 0

⟹ 𝐸 𝐵−1 − 𝐵−1 𝐶𝑍 + 𝐷𝑍 = 0

⟹ 𝐸𝐵 −1 − 𝐸𝐵 −1 𝐶𝑍 + 𝐷𝑍 = 0

⟹ 𝐷 − 𝐸𝐵 −1 𝐶 𝑍 = −𝐸𝐵−1

1
= 𝑍
𝑉

= −𝐸𝐵−1

∴ 𝑍 = −𝐸𝐵−1 𝐷 − 𝐸𝐵 −1 𝐶 −1

120
⟹ 𝑍 = −𝑉𝐸𝐵−1
SPACE OF HINTS

Note :Steps for partition method

𝐵 𝐶 𝑋 𝑌
1. Partition the given matrix 𝐴 as 𝐴 = and 𝐴−1 =
𝐸 𝐷 𝑍 𝑉
2. Find 𝐵−1
3. Find 𝑉, 𝑉 = 𝐷 − 𝐸𝐵−1 𝐶 −1

4. Find 𝑌, 𝑌 = −𝐵−1 𝐶𝑉
5. Find 𝑍, 𝑍 = −𝑉𝐸𝐵−1
6. Find 𝑋, 𝑋 = 𝐵−1 − 𝐵−1 𝐶𝑍

Example: 2.2.13

Determine the inverse of the matrix using the partition

1 1 1
method 4 3 −1 . Hence find the solution of the system of equations
3 5 3

𝑥1 + 𝑥2 + 𝑥3 = 1, 4𝑥1 + 3𝑥2 − 𝑥3 = 6,

3𝑥1 + 5𝑥2 + 3𝑥3 = 4.

Solution:

The given system of equation can be written as𝑨𝑿 = 𝒃

1 1 1 𝑥1 1
Given 𝐴 = 4 3 −1 ; 𝑋 = 𝑥2 ; 𝑏 = 6
3 5 3 𝑥3 4

1 1 1
𝑋 𝑌
A can be partitioned as 𝐴 = 4 3 −1 and𝐴−1 =
𝑍 𝑉
5 5 3

1 1 1
From 𝐴, 𝐵 = , 𝑐= , 𝐷 = 3, 𝐸 = 3 5
4 3 −1

To find 𝑩−𝟏

1
𝐵−1 = 𝑎𝑑𝑗 𝐵
𝐵

121
 3 −1
=−
−4 1
SPACE FOR HINTS

−3 1
𝐵−1 =
4 −1

To find 𝑽

𝑉 = 𝐷 − 𝐸𝐵−1 𝐶 −1

−3 1 1
𝐷 − 𝐸𝐵−1 𝐶 = 3 − 3 5
4 −1 −1

−4
=3− 3 5
5

= 3 − [−12 + 25] = 3 − 13 = 10

𝑉 = 𝐷 − 𝐸𝐵−1 𝐶 −1

−1
= −10

−1
𝑉=
10

To find 𝒀

𝑌 = −𝐵−1 𝐶𝑉

−3 1 1 −1
𝑌=−
4 −1 −1 10

1 4
=
10 5

−2
5
𝑌=
1
2

To find 𝒁

𝑍 = −𝐸𝐵−1 𝑉

122
 −1 −3 1
𝑍=− 3 5 SPACE OF HINTS
10 4 −1

−1
= 11 −2
10

11 −1
𝑍= 10 5

To find 𝑿

𝑋 = 𝐵 −1 − 𝐵−1 𝐶𝑍

−3 1 −3 1 1 11 10 −1
= − 5
4 −1 4 −1 −1

−44 4
10 5
−3 1
= −
4 −1 55 −5
10 5

14 1
10 5
=
−15 0
10

1.4 0.2
𝑋=
−1.5 0

1.4 0.2 −0.4

𝑋 𝑌
∴ 𝐴−1 = = −1.5 0 0.5
𝑍 𝑉
1.1 −0.2 −0.1

We know that, 𝐴𝑋 = 𝑏 ⟹ 𝑋 = 𝐴−1 𝑏

1.4 0.2 −0.4 1

∴ 𝑋 = −1.5 0 0.5 6
1.1 −0.2 −0.1 4

1.4 + 1.2 − 1.6

= −1.5 + 0 + 2.0
1.1 − 1.2 − 0.4

123
 1
𝑋= 0.5
SPACE FOR HINTS
−0.5

∴ 𝑥1 = 1, 𝑥2 = 0.5, 𝑥3 = −0.5.

Example: 2.2.14

2 1 1 −2
4 0 2 1
Find the inverse of the matrix 𝐴 = using partition
3 2 2 0
1 3 2 −1
method. Hence, solve the system of equations 𝐴𝑋 = 𝑏, where

𝑏 = −10 8 7 −5 𝑇 .

Solution:

The matrix A can be partitioned as

2 1 1 −2
4 0 2 1 𝑋 𝑌
𝐴= and𝐴−1 =
3 2 2 0 𝑍 𝑉
1 3 2 −1

2 1 1 −2 2 0 3 2
Here 𝐵 = , 𝐶= , 𝐷= , 𝐸=
4 0 2 1 2 −1 1 3

To find 𝑩−𝟏

1
𝐵−1 = 𝑎𝑑𝑗 𝐵
𝐵

1 0 −1
=−
4 −4 2

0 1
4
−1
𝐵 =
1 −1
2

124
To find 𝑽 ∶
SPACE OF HINTS

V = D − EB −1 C −1

0 1
4
2 0 3 2 1 −2
𝐷 − 𝐸𝐵−1 𝐶 = −
2 −1 1 3 1 −1 2 1
2

1 1
2 4
2 0 3 2
= −
2 −1 1 3 0 −5
2

3 −17
2 4
2 0
= −
2 −1 1 −29
2 4

1 17
2 4
−1
𝐷 − 𝐸𝐵 𝐶 =
1 25
2 4

−1
1 17
2 4
𝑉=
1 25
2 4

−1.9231 1.3077
𝑉=
0.4615 −0.1538

To find 𝒀

𝑌 = −𝐵−1 𝐶𝑉

0 1
4
1 −2 −1.9231 1.3077
𝑌=−
1 −1 2 1 0.4615 −0.1538
2

125
 0 1
4
SPACE FOR HINTS −2.8461 1.6153
=−
1 −1 −3.3847 2.4616
2

−0.8462 0.6153
=−
−1.1537 0.3845

0.8462 −0.6153
𝑌=
1.1537 −0.3845

To find 𝒁

𝑍 = −𝑉𝐸𝐵 −1

0 1
4
3 2 −1.9231 1.3077
𝑍=−
1 3 1 −1 0.4615 −0.1538
2

2 −1
4
−1.9231 1.3077
=−
3 −5 0.4615 −0.1538
4

0.0769 −1.1537
=−
0.4616 0.0769

−0.0769 1.1537
𝑍=
−0.4616 −0.0769

To find 𝑿

𝑋 = 𝐵−1 − 𝐵−1 𝐶𝑍

0 1 0 1
4 4
1 −2 −0.0769 1.1537
𝑋= −
1 −1 1 −1 2 1 −0.4616 −0.0769
2 2

126
 0 1 1 1
4 2 4 SPACE OF HINTS
−0.0769 1.1537
= −
1 −1 0 −15 −0.4616 −0.0769
2 2

0 1
4
−0.0539 0.5577
= −
1 −1 1.154 0.1923
2

0.1539 −0.3077
=
−0.154 −0.6923

𝑋 𝑌
We know that,𝐴−1 =
𝑍 𝑉

0.1539 −0.3077 0.8462 −0.6154

−0.154 −.6923 1.1537 −0.3845
𝐴−1 =
−0.0769 1.1537 −1.931 1.3077
−0.4616 −0.0769 0.4615 −0.1538

We know that, 𝐴𝑋 = 𝑏 ⟹ 𝑋 = 𝐴−1 𝑏

0.1539 −0.3077 0.8462 −0.6154 −10

−0.154 −.6923 1.1537 −0.3845 8
𝑋=
−0.0769 1.1537 −1.931 1.3077 7
−0.4616 −0.0769 0.4615 −0.1538 −5

5
6
𝑋=
−10
8

∴ 𝑥1 = 5, 𝑥2 = 6, 𝑥3 = −10, 𝑥4 = 8.

2.3:Error Analysis for Direct Methods

The methods discussed in the previous section involve only a finite

number of arithmetic operations. The number of divisions and
multiplications involved in solving the system of equation is usually
called the operational count for that method. In the following, we obtain
the operational count for Gauss elimination procedure.

127
SPACE FOR HINTS

Operational count for Gauss elimination

Number of divisions:
First step [first equation (division by first pivot)]: n
Second step [second equation (division by second pivot)]: n – 1
.........................................
𝑛𝑡𝑕 step [nth equation (divisionby nth pivot)]: 1
𝑛
Total number of divisions= 𝛴𝑛 = (𝑛 + 1).
2

Number of multiplications:
First step: second equation : n

third equation : n

.........................

nth equation: n

Total multiplications in the first step is 𝑛(𝑛 − 1) .

Hence, total multiplications in the forward elimination

𝑛
= 𝛴(𝑛 − 1)𝑛 = 𝛴 (𝑛2 − 𝑛) = (𝑛 + 1)(𝑛 − 1).
3

For back substitution, we have the number of multiplications as

(n – 1)th equation : 1

(n – 2)th equation : 2

............................

first equation : n –1

128
Total multiplications in the back substitution
SPACE OF HINTS
𝑛
= (𝑛 − 1) = (𝑛 − 1).
2
𝑛 𝑛
Total multiplications= 𝑛 + 1 𝑛 − 1 + (𝑛 − 1)
3 2

𝑛
= 𝑛 − 1 (2𝑛 + 5)
6

Operational count = Total number of divisions and multiplications

𝑛 𝑛
= 𝑛 + 1 + 𝑛 − 1 (2𝑛 + 5)
2 6
𝑛 2
= (𝑛 + 3𝑛 − 1)
3

𝑛3
For large 𝑛, operational count = .
3

𝑛
Total additions and subtractions = 𝑛 − 1 2𝑛 + 5 .
6

We find that for large 𝑛, the

𝑛3
(i). Gauss – Jordan elimination method requires operations.
2

𝑛3
(ii). 𝑳𝑼 decomposition methodrequires operations (sameas in Gauss
3

eliminationmethod).

𝑛3
(iii). Cholesky method requires operations.
6

If all calculations are performed exactly, then only we can hope to find
the exact solution to the system. Usually, during computation, it will be
necessary to round or chop the numbers. This will introduce round-off
errors in the computation. Because ofthis, the methods used will produce
results which will differ considerably from the exact solution. The exact
solution x and the corresponding approximate solution 𝑥 will
satisfyrespectively the equations

129
 𝑨𝑿 = 𝒃
SPACE FOR HINTS
𝑨 + 𝜹𝑨 𝒙 = 𝒃 + 𝜹𝒃 (1)

where δA and 𝜹𝒃 are the changes in 𝑨 and𝒃 respectively, due to round-

offerror.

From (1) we obtain

−1
𝑥 − 𝑥 = 𝐴 + 𝛿𝐴 𝑏 + 𝛿𝑏 − 𝐴−1 𝑏
−1
= 𝐴 + 𝛿𝐴 − 𝐴−1 𝑏 + (𝐴 + 𝛿𝐴)−1 𝛿𝑏 (2)

which may be called the error equation. In order to estimate the error
vector 𝜖 = 𝑥 − 𝑥, we recall theconcept of a norm of a vector x and
amatrix𝐴.

Vector Norm:
The non-negative quantity || 𝑥 || isa measure of the size or length of
avector satisfying:

(i). || 𝒙 || > 0, 𝑓𝑜𝑟 𝒙 ≠ 𝟎 𝑎𝑛𝑑 ||𝟎|| = 0,

(ii). || 𝑐𝒙 || = |𝑐| || 𝒙 ||, for anarbitrary complex number 𝑐,

(iii). ||𝒙 + 𝒚|| ≤ || 𝒙 || + || 𝒚 ||.

The most commonly used norms are

i. Absolute norm (𝑙1 norm)

𝒙 = 𝑥𝑖 (1)
𝑖=1

ii. Euclidean norm

𝑛 1
2
∗ 1 2
𝑥 2 = (𝑥 𝑥) 2 = 𝑥𝑖 (2)
𝑖=1

iii. Maximum norm(𝑙∞ norm)

130
 𝑥 ∞ = max 𝑥𝑖 (3)
1≤𝑖≤𝑛 SPACE OF HINTS

Matrix Norm:
The matrix norm, || 𝑨 ||, is a non-negative number which satisfies the
properties:

i. ||𝑨|| > 0, 𝑖𝑓 𝑨 ≠ 𝟎 𝑎𝑛𝑑 ||𝟎|| = 0,

ii. || 𝑐𝑨|| = |𝑐| || 𝑨 ||, for anarbitrary complex number 𝑐,

iii. || 𝑨 + 𝑩 || ≤ || 𝑨 || + || 𝑩 ||,

iv. || 𝑨𝑩 || ≤ || 𝑨 || ||𝑩||.

The most commonly used norms are:

(i). Frobenius or Euclidean norm

1
𝑛 2 2

𝐹(𝐴) = 𝑎𝑖𝑗
𝑖,𝑗 =1

(ii). Maximum norm

A = A ∞

= max𝑖 𝑘 𝑎𝑖𝑘 (maximumabsolute row sum).

||𝐴|| = ||𝐴||1

= max𝑘 𝑖 𝑎𝑖𝑘 (maximumabsolute column sum).

(iii). Hilbert norm or Spectral norm

||𝐴||2 = 𝜆, 𝑤𝑕𝑒𝑟𝑒 𝜆 = 𝜌(𝐴∗ 𝐴) .

If A is Hermitian or real and symmetric, then

𝜆 = 𝜌(𝐴2 ) = 𝜌2 (𝐴)

131
 So that ||𝐴||2 = 𝜌(𝐴).
SPACE FOR HINTS
The matrix norm must be consistentwith the vector norm that we
areusing for any vector x and matrix A, i.e.

𝐴𝑥 ≤ 𝐴 𝑥 (1)

It may be verified that the norm

𝐴 = max 𝑎𝑖𝑗 (2)

𝑖
𝑗 =1

is consistent with maximum norm ||𝑥 ||.

Error Estimate:
−1
𝑥−𝑥 ≤ 𝐴 + 𝛿𝐴 − 𝐴−1 𝑏 + (𝐴 + 𝛿𝐴)−1 𝛿𝑏

Bu (𝐴 + 𝛿𝐴)−1 = (𝐴 + 𝛿𝐴)−1 − 𝐴−1 + 𝐴−1

≤ 𝐴−1 𝐼 − (𝐼 + 𝐴−1 𝛿𝐴) −1

= 𝐴−1 (𝐼 + 𝐴−1 𝛿𝐴)−1 (𝐼 + 𝐴−1 𝛿𝐴 − 𝐼)

= 𝐴−1 (𝐼 + 𝐴−1 𝛿𝐴)−1 𝐴−1 𝛿𝐴

≤ 𝐴−1 𝐴−1 𝛿𝐴 / (𝐼 − 𝐴−1 𝛿𝐴 )

where 𝐴−1 𝛿𝐴 is assumed to beless than 1. Hence, we obtain

𝐴−1
𝑥−𝑥 ≤ 𝐴 𝐴−1 𝛿𝐴 𝑏 + 𝛿𝑏 (1)
𝐴−1 𝛿𝐴

Since 𝐴−1 𝛿𝐴 ≤ 𝐴−1 𝛿𝐴 , the equation(1) may be written as

𝐴−1
𝑥−𝑥 ≤ 𝒙 𝜹𝑨 + 𝛿𝑏
(1 − 𝐴−1 𝛿𝐴 )

or

132
 𝑥−𝑥 𝐴−1 𝒙 𝛿𝐴 𝛿𝑏
≤ 𝐴 [ + ] SPACE OF HINTS
𝒙 (1 − 𝐴−1 𝛿𝐴 ) 𝐴 𝒙 𝐴 𝒙

or

𝑥−𝑥 𝐾(𝐴) 𝛿𝑏 𝛿𝐴
≤ + (2)
𝒙 (1 − 𝐴−1 𝛿𝐴 ) 𝑏 𝐴

where the quantity 𝐾(𝐴) = ||𝐴−1 || is called the condition number of

the matrix A and is denoted by cond (𝑨). The left-hand side in
(2)givesthe overall relative error in x. The first term inside the brackets
on the right hand is the overall relative error in b and the second term is
the relative error in A.

If there is no error in b then ||𝛿𝑏|| = 0 and (2) becomes

𝑥−𝑥 𝐾(𝐴) 𝛿𝐴
≤ .
𝒙 (1 − 𝐴−1 𝛿𝐴 ) 𝐴

If there is no error in A then 𝛿𝐴 = 0and (2) gives

𝒙−𝒙 𝛿𝒃
≤𝐾 𝑨 [
𝒙 𝒃

If 𝐾(𝑨) is small, then small changesin A or b produce only small

changesin x. If 𝐾(𝑨) is large, then smallrelative changes in A or b will
produce large relative changes in x, and the system of equations𝑨𝑿 =
𝒃is said to be ill conditioned. If 𝐾(𝑨) isnear unity, then the
system𝑨𝑿 = 𝒃 is wellconditioned

If || . || is the spectral norm, then

−1
𝜆
𝐾 𝑨 = 𝐴 2 𝐴 2 =
𝜇

133
 where λ and μ are the largest and smallest eigenvalues in modulus
𝜆∗
SPACE FOR HINTS of𝑨∗ 𝑨. If 𝑨 is Hermitian or real and symmetric, we have 𝐾(𝑨) =
𝜇∗

where 𝜆∗ and 𝜇∗ are the largest andthe smallest eigenvalues in 𝑨

Example: 2.3.1

Determine the Euclidean and the maximum absolute row sum norms of
1 7 −4
the matrix 𝐴 = 4 −4 9
12 −1 3

Solution:

1
3 2 2
Euclidean norm = 𝐹(𝐴) = 𝑖,𝑗 =1 𝑎𝑖𝑗

∴ [𝐹 𝐴 ]2 = 12 + 72 + (−4)2 + 42 + (−4)2 + 92 + 122 + (−1)2

+ 32

= 1 + 49 + 16 + 16 + 16 + 81 + 144 + 1 + 9 = 333

Hence 𝐹 𝐴 = 333 = 18.25

3
Maximum absolute row sum norms = max𝑖 𝑘=1 𝑎𝑖𝑘

= max 12, 17, 16 = 17

Example: 2.3.2

2.1 1.8 𝑥1 2.1

Find the condition number of the system
6.2 5.3 𝑥2 = 6.2

Solution:

Using the spectral norm, we have

42.85 36.64
𝐴∗ 𝐴 = 𝐴𝑇 𝐴 =
36.64 31.33

134
The eigenvalues of 𝐴∗ 𝐴 are the solutions of
SPACE OF HINTS
42.85 − 𝜆 36.64
=0
36.64 31.33 − 𝜆

⟹ 𝜆2 − 74.18𝜆 + 0.0009 = 0

The solutions of this equation are 𝜆1 = 74.17999 and

𝜆2 = 0.00001213

The condition number is

𝜆1
𝐾 𝑨 = = 24.72.73
𝜆2

Hence, this system of equation is highly ill conditioned and is very

sensitive to round-off errors.

Example: 2.3.3

Determine the condition number of the matrix

1 7 −4
𝐴= 4 −4 9
12 −1 3

Solution:

Using the maximum absolute row sum norm, we have

31 −44 17
8 8 8
−31 44 −17 −44 56 −20
−1 1
𝐴 =− 44 −56 20 = 8 8 8
8
−17 20 −7
17 −20 7
8 8 8

A ∞ = 𝑚𝑎𝑥𝑖𝑚𝑢𝑚 𝑎𝑏𝑠𝑜𝑙𝑢𝑡𝑒 𝑟𝑜𝑤 𝑠𝑢𝑚 𝑛𝑜𝑟𝑚 𝑓𝑜𝑟 𝐴

= 𝑚𝑎𝑥{14, 29, 50} = 50

135
 𝐴−1 ∞ = 𝑚𝑎𝑥𝑖𝑚𝑢𝑚 𝑎𝑏𝑠𝑜𝑙𝑢𝑡𝑒 𝑟𝑜𝑤 𝑠𝑢𝑚 𝑛𝑜𝑟𝑚 𝑓𝑜𝑟 𝐴−1
SPACE FOR HINTS
31 44 17 44 56 20 17 20 7
= 𝑚𝑎𝑥 + + , + + , + +
8 8 8 8 8 8 8 8 8

92 44
= 𝑚𝑎𝑥 , 15, = 15
8 8

Therefore 𝐾 𝐴 = A ∞ = 𝐴−1 ∞ = 50 × 15 = 750.

2.4 Iteration methods:

Jacobi Iteration method:

The system of equation can be written as

𝑎11 𝑥1 + 𝑎12 𝑥2 + ⋯ + 𝑎1𝑛 𝑥𝑛 = 𝑏1
𝑎21 𝑥1 + 𝑎22 𝑥2 + ⋯ + 𝑎2𝑛 𝑥𝑛 = 𝑏2
⋮
𝑎𝑛1 𝑥1 + 𝑎𝑛2 𝑥2 + ⋯ + 𝑎𝑛𝑛 𝑥𝑛 = 𝑏𝑛

The jacobi iteration method (or) Gauss-Jacobi iteration method can be

written as,

𝑘
(𝑘+1)
𝑏1 − (𝑎12 𝑥2 + 𝑎13 𝑥3 𝑘 + ⋯ + 𝑎1𝑛 𝑥𝑛 𝑘 )
𝑥1 =
𝑎11
(𝑘+1)
𝑏2 − (𝑎21 𝑥2 + 𝑎23 𝑥3 𝑘 + ⋯ + 𝑎2𝑛 𝑥𝑛 𝑘 )
𝑘
𝑥2 = (2)
𝑎22
⋮
𝑏𝑛 − (𝑎𝑛1 𝑥1 + 𝑎𝑛2 𝑥2 𝑘 + ⋯ + 𝑎𝑛𝑛 −1 𝑥𝑛−1 𝑘 )
𝑘
𝑥𝑛 (𝑘+1) =
𝑎𝑛𝑛

We replace the complete vector 𝑋 (𝑘) in the RHS of 2 at the end

of each iteration, this method is also called the method of simultaneous
displacement.

The matrix form the above method is,

𝑋 (𝑘+1) = −𝐷 −1 𝐿 + 𝑈 𝑋 𝑘
+ 𝐷 −1 𝑏 3

𝑋 (𝑘+1) = 𝐻𝑋 (𝑘) + 𝐶, 𝑘 = 0,1,2, …

136
Where 𝐻 = −𝐷 −1 𝐿 + 𝑈 , 𝐶 = −𝐷 −1 𝑏 and 𝐿 and 𝑈 are respectively
SPACE OF HINTS
lower and upper triangular matrices with zero diagonal entries and 𝐷 is
the diagonal matrix such that

𝐴=𝐿+𝐷+𝑈

∴ 3 can be written as,

𝑋 (𝑘+1) = 𝑋 𝑘
−𝑋 𝑘
− 𝐷 −1 𝐿 + 𝑈 𝑋 𝑘
+ 𝐷 −1 𝑏

𝑘
=𝑋 − 𝐷 −1 𝐴𝑋 𝑘
+ 𝐷 −1 𝑏

𝑖𝑒) 𝑋 (𝑘+1) − 𝑋 𝑘
= −𝐷 −1 𝐴𝑋 𝑘
+ 𝐷 −1 𝑏

= 𝐷 −1 (𝑏 − 𝐴𝑋 𝑘
)

𝑘
𝑉
= 𝐷 −1 𝑟 𝑘
(4)

𝑘
where 𝑉 = 𝑋 𝑘+1 − 𝑋 𝑘
, 𝑟 𝑘
= 𝑏 − 𝐴𝑋 𝑘
and𝑉 𝑘
is the error in the
𝑘
approximation and 𝑟 is the residual vector.

𝑘 𝑘 𝑘+1 𝑘 𝑘
Now, 4 ⟹ 𝐷𝑉 =𝑟 and 𝑋 =𝑋 +𝑉
theseequation is the Jacobi iteration method in an error format.

Example: 2.4.1

Solve the system of equation 4𝑥1 + 𝑥2 + 𝑥3 = 1,

𝑥1 + 5𝑥2 + 2𝑥3 = −6, 𝑥1 + 2𝑥2 + 3𝑥3 = −4 using the Jacobi iteration

method given in equation 3 and its error format equation in 4 . Take
0
the initial approximation as 𝑥 = 0.5 −0.5 −0.5 𝑇 and perform
three iterations in each case. The exact solution is𝑥1 = 1,

𝑥2 = −1, 𝑥3 = −1.

Solution:

Given equation of the form 𝑨𝑿 = 𝒃.

137
 4 1 1 𝑥1 2
Where 𝐴 = 1 5 2 𝑋 = 𝑥2 𝑏 = −6
SPACE FOR HINTS 𝑥3
1 2 3 −4

(i). In Jacobi method, consider 𝐴 = 𝐿 + 𝐷 + 𝑈.

0 0 0 4 0 0 0 1 1
Where 𝐿 = 1 0 0 𝐷= 0 5 0 𝑈= 0 0 2
1 2 0 0 0 3 0 0 0

The formula for Jacobi iteration method is,

𝑋 (𝑘+1) = 𝐻𝑋 (𝑘) + 𝐶 where 𝐻 = −𝐷 −1 𝐿 + 𝑈 , 𝐶 = −𝐷 −1 𝑏

1 0 0
4

0 1 0
Now, clearly, 𝐷 −1 = 5

0 0 1
3

1 0 0 0 1 1
4 4 4
1 0 1 1 1 2
𝐻=− 0 5 0 1 0 2 =− 5 0 5
1 2 0
0 0 1 1 2 0
3 3 3

0 −1 4 −1 4

1 −2 5
𝐻= − 5 0

−1 3 −2 3 0

𝐶 = −𝐷 −1 𝑏

138
 1 0 0 1
4 2 SPACE OF HINTS

1 2 −6
= 0 5 0 5
−6 =
−4
0 0 13 −4
3

∴ 𝑋 𝑘+1 = 𝐻𝑋 𝑘
+𝐶

0 −1 4 −1 4 1
2
1 −2 5 𝑋 −6
𝑋 (𝑘+1) = − 5 0 𝑘
+ 5 ; 𝑘 = 0,1,2, …

−1 3 −2 3 0 −4
3

Put 𝑘 = 0,

0 −1 4 −1 4 1
2
1 −2 5 𝑋 −6
𝑋 (1) = − 5 0 0
+ 5

−1 3 −2 3 0 −4
3

Given 𝑥 (0) = [0.5 −0.5 −0.5]𝑇

0 −1 4 −1 4 1
2
1 0.5 −6
= − 5 0 −2 5 5
−0.5 +
−0.5
−1 3 −2 3 0 −4
3

0.25 0.5
= 0.1 + −1.2
−0.1667 −1.333

0.75
𝑋 (1) = −1.1
−1.1663

139
 Put 𝑘 = 0,
SPACE FOR HINTS
0 −1 4 −1 4 1
2
1
𝑋 (2) = − 5 0 − 2 5 𝑋 1 + −6 5

−1 3 −2 3 0 −4
3

0 −1 4 −1 4 1
2
1 0.75 −6
= − 5 0 −2 5 5
−1.1 +
−1.1663
−1 3 −2 3 0 −4
3

1.067
(2)
𝑋 = −0.8835
−0.85

Put 𝑘 = 0,

0 −1 4 −1 4 1
2
1
𝑋 (3) = − 5 0 − 2 5 𝑋 2 + −6 5

−1 3 −2 3 0 −4
3

0 −1 4 −1 4 1
2
1 1.067 −6
= − 5 0 −2 5 5
−0.8835 +
−0.85
−1 3 −2 3 0 −4
3

0.9334
= −1.0734
−1.0997

140
 1
𝑋 (3) ≅ −1 SPACE OF HINTS
−1

𝑥1 = 1, 𝑥2 = −1, 𝑥3 = −1

(ii). Jacobi iteration method in an error formula is,

𝑋 (𝑘+1) = 𝑋 (𝑘) + 𝑉 (𝑘)

𝑘 𝑘 𝑘
where 𝑟 = 𝑏 − 𝐴𝑋 𝑎𝑛𝑑 𝑉 = 𝐷 −1 𝑟 𝑘

Put 𝑘 = 0,

0 0
𝑟 = 𝑏 − 𝐴𝑋

2 4 1 1 0.5
= −6 − 1 5 2 −0.5
−4 1 2 3 −0.5

1
0
𝑟 = −3
−2

0
𝑉 = 𝐷 −1 𝑟 0

1 0 0
4
1 1
= 0 5 0
−3
−2
0 0 1
3

0.25
0
𝑉 = −0.6
−0.6667

𝑋 (1) = 𝑋 0 + 𝑉 0

0.5 0.25
= −0.5 −0.6
−0.5 −0.6667

141
 0.75
𝑋 (1) = −1.1
SPACE FOR HINTS
−1.1667

Put 𝑘 = 1,

1 1
𝑟 = 𝑏 − 𝐴𝑋

2 4 1 1 0.75
= −6 − 1 5 2 −1.1
−4 1 2 3 −1.1667

2 0.7333
= −6 − −7.0834
−4 −4.95

1.2667
1
𝑟 = 1.0834
0.95

𝑉 (1) = 𝐷 −1 𝑟 1

1 0 0
4
1 1.2667
= 0 5 0
1.0834
0.95
0 0 1
3

0.3167
(1)
𝑉 = 0.2167
0.3167

𝑋 (2) = 𝑋 (1) + 𝑉 (1)

0.75 0.3167
= −1.1 + 0.2167
−1.1667 0.3167

1.0667
(2)
𝑋 = −0.8833
−0.85

Put 𝑘 = 2,

142
 2 2
𝑟 = 𝑏 − 𝐴𝑋
SPACE OF HINTS

2 4 1 1 1.0667
= −6 − 1 5 2 −0.8833
−4 1 2 3 −0.85

−0.5335
2
𝑟 = −0.9502
−0.7501

𝑉 (2) = 𝐷 −1 𝑟 2

1 0 0
4
1 −0.5335
= 0 5 0
−0.9502
−0.7501
0 0 1
3

−0.1334
𝑉 2() = −0.19
−0.25

𝑋 (3) = 𝑋 (2) + 𝑉 (2)

1.0667 −0.1334 0.9333

= −0.8833 + −0.19 = −1.0733
−0.85 −0.25 −1.1

1
𝑋 (1) ≅ −1
−1

Gauss-Seidel Iteration Method

The given system of equation can be written as

𝑎11 𝑥1 + 𝑎12 𝑥2 + ⋯ + 𝑎1𝑛 𝑥𝑛 = 𝑏1

𝑎21 𝑥1 + 𝑎22 𝑥2 + ⋯ + 𝑎2𝑛 𝑥𝑛 = 𝑏2
⋮
𝑎𝑛1 𝑥1 + 𝑎𝑛2 𝑥2 + ⋯ + 𝑎𝑛𝑛 𝑥𝑛 = 𝑏𝑛

The above equation can be written as,

143
 𝑎11 𝑥1 (𝑘+1) = − 𝑎12 𝑥2 (𝑘) + 𝑎13 𝑥3 (𝑘) + ⋯ + 𝑎1𝑛 𝑥𝑛 (𝑘) + 𝑏1
SPACE FOR HINTS
𝑎21 𝑥1 (𝑘+1) + 𝑎22 𝑥2 (𝑘+1) = − 𝑎23 𝑥3 (𝑘) + ⋯ + 𝑎2𝑛 𝑥𝑛 (𝑘) + 𝑏2

𝑎𝑛1 𝑥1 (𝑘+1) + 𝑎𝑛2 𝑥2 (𝑘+1) + ⋯ + 𝑎𝑛𝑛 𝑥𝑛 (𝑘+1) = 𝑏𝑛

∴ 𝐷 + 𝐿 𝑋 (𝑘+1) = −𝑈𝑋 (𝑘) + 𝑏 (2)

−𝑈𝑋 (𝑘) + 𝑏
⟹ 𝑋 (𝑘+1) =
𝐷+𝐿

⟹ 𝑋 (𝑘+1) = 𝐷 + 𝐿 −1
(𝑏 − 𝑈𝑋 (𝑘) )

⟹ 𝑋 (𝑘+1) = 𝐷 + 𝐿 −1
𝑏− 𝐷+𝐿 −1
𝑈𝑋 (𝑘)

𝑘+1 𝑘
⟹𝑋 = 𝐻𝑋 + 𝑐 ; 𝑘 = 0,1,2, … (3)

−1 −1
where 𝐻 = − 𝐷 + 𝐿 𝑈, 𝑐 = 𝐷+𝐿 𝑏

Gauss-Seidel method is also called the method of successive

displacement

3 can be written as,

𝑋 (𝑘+1) = 𝑋 (𝑘) − 𝑋 (𝑘) + 𝐻𝑋 (𝑘) + 𝑐

= 𝑋 (𝑘) − 𝑋 (𝑘) − 𝐷 + 𝐿 −1
𝑈𝑋 (𝑘) + 𝐷 + 𝐿 −1
𝑏

= 𝑋 (𝑘) − (𝐼 + 𝐷 + 𝐿 −1
𝑈)𝑋 (𝑘) + 𝐷 + 𝐿 −1
𝑏

= 𝑋 (𝑘) − 𝐷 + 𝐿 −1
[ 𝐷 + 𝐿 + 𝑈]𝑋 (𝑘) + 𝐷 + 𝐿 −1
𝑏

= 𝑋 (𝑘) − 𝐷 + 𝐿 −1
(𝐴𝑋 (𝑘) ) + 𝐷 + 𝐿 −1
𝑏

𝑋 (𝑘+1) = 𝑋 (𝑘) − 𝐷 + 𝐿 −1
𝑏 − 𝐴𝑋 (𝑘) (4)

𝑋 (𝑘+1) − 𝑋 (𝑘) = 𝐷 + 𝐿 −1
𝑏 − 𝐴𝑋 (𝑘) (5)

∴ 𝑉 (𝑘) = 𝐷 + 𝐿 −1 (𝑘)
𝑟

144
Where𝑉 (𝑘) = 𝑋 (𝑘+1) − 𝑋 (𝑘) , 𝑟 (𝑘) = 𝑏 − 𝐴𝑋 (𝑘)
SPACE OF HINTS
(𝑘+1) (𝑘) (𝑘)
From 5 ⟹ 𝑋 −𝑋 𝐷 + 𝐿 = 𝑏 − 𝐴𝑋

𝑉 (𝑘) 𝐷 + 𝐿 = 𝑟 (𝑘)

∴ 𝑋 (𝑘+1) = 𝑋 (𝑘) + 𝑉 (𝑘)

The above equation describes the Gauss-Seidel method is an error

format.

Example: 2.4.2

Solve the system of equations 42 − 𝑥2 = 7, −𝑥1 + 2𝑥2 − 𝑥3 = 1, −𝑥2 +

2𝑥3 = 1 using the Gauss-Seidel method given in equations and its error
format given in equations. Take the initial approximation as 𝑋 (0) = 0
and perform three iterations.

Solution:

The formula for Gauss-Seidel method is

𝑋 𝑘+1 = 𝐻𝑋 𝑘 + 𝐶

−1 −1
Where 𝐻 = − 𝐷 + 𝐿 𝑈, 𝑐 = 𝐷+𝐿 𝑏.

2 −1 0
𝐴 = −1 2 −1
0 −1 2

0 0 0 2 0 0
From matrix 𝐴, 𝐿 = −1 0 0 , 𝐷= 0 2 0,
0 −1 0 0 0 2

0 −1 0
𝑈= 0 0 −1
0 0 0

2 0 0
𝐷 + 𝐿 = −1 2 0
0 −1 2

145
 −1
1
𝐷+𝐿 = 𝑎𝑑𝑗 (𝐷 + 𝐿)
SPACE FOR HINTS 𝐷+𝐿

1 0 0
2

−1
1 1 0
⟹ 𝐷+𝐿 = 4 2
1 1 1
8 4 2

−1
To find H & C:𝐻 = − 𝐷 + 𝐿 𝑈

1 0 0
2
1 1 0 −1 0
=− 4 2 0
0 0 −1
0 0 0
1 1 1
8 4 2

0 1 0
2
1 1
𝐻= 0 4 2

0 1 1
8 4

−1
𝑐 = 𝐷+𝐿 𝑏

1 0 0
2
1 1 7
= 4 2 0
1
1
1 1 1
8 4 2

146
 7
2 SPACE OF HINTS

9
𝑐= 4
13
8

∴ 𝑋 (𝑘+1) = 𝐻𝑋 (𝑘) + 𝑐

0 1 0 7
2 2
1 1 9
⟹ 𝑋 𝑘+1 = 0 4 2 𝑋𝑘 + 4

0 1 1 13
8 4 8

1stIteration:𝑘 = 0

0 1 0 7 0 1 0 7
2 2 2 2
1 1 9 1 1 0 9
𝑋 1
= 0 4 2 𝑋 + 0 4 = 0 4 2 0 + 4
0
0 1 1 13 0 1 1 13
8 4 8 8 4 8

7
2
9
𝑋 (1) = 4
13
8

2nd Iteration:𝑘 = 1

0 1 0 7
2 2
1 1 9
𝑋 (2) = 0 4 2 𝑋 (1) + 4

0 1 1 13
8 4 8

147
 0 1 0 7 7
2 2 2
SPACE FOR HINTS
1 1 9 9
= 0 4 2 4 + 4

0 1 1 13 13
8 4 8 8

1.125 3.5
= 2.1875 + 2.25
0.6875 1.625

4.625
(2)
𝑋 = 3.625
2.3125

3rdIteration: 𝑘 = 2

0 1 0 7
2 2
1 1 9
𝑋 (3) = 0 4 2 𝑋 (2) + 4

0 1 1 13
8 4 8

0 1 0 7
2 2
1 1 4.625 9
= 0 4 2 3.625 + 4
2.3125
0 1 1 13
8 4 8

5.3125
𝑋 (3) = 4.3125
2.6563

Using the Gauss-Seidal formula

𝑟 (𝑘) = 𝑏 − 𝐴𝑋 (𝑘)

𝑉 (𝑘) = 𝐷 + 𝐿 −1 (𝑘)
𝑟

𝑋 (𝑘+1) = 𝑋 (𝑘) + 𝑉 (𝑘)

148
1st Iteration:𝑘 = 0
SPACE OF HINTS

𝑟 (0) = 𝑏 − 𝐴𝑋 (0)

7 2 −1 0 0 7
(0)
= 1 − −1 2 −1 0 𝑟 = 1
1 0 −1 2 0 1

𝑉 (0) = 𝐷 + 𝐿 −1 (0)
𝑟

1 0 0
2
1 1 7
= 4 2 0
1
1
1 1 1
8 4 2

3.5
(0)
𝑉 = 2.25
1.625

𝑋 (1) = 𝑋 (0) + 𝑉 (0)

0 3.5
= 0 + 2.25
0 1.625

3.5
(1)
𝑋 = 2.25
1.625

2ndIteration: 𝑘 = 1

𝑟 (1) = 𝑏 − 𝐴𝑋 (1)

7 2 −1 0 3.5
= 1 − −1 2 −1 2.25
1 0 −1 2 1.625

149
 7 4.75
= 1 − −0.625
SPACE FOR HINTS
1 1

2.25
(1)
𝑟 = 1.625
0

𝑉 (1) = 𝐷 + 𝐿 −1 (1)
𝑟

1 0 0
2
1 1 2.25
= 4 2 0
1.625
0
1 1 1
8 4 2

1.125
(1)
𝑉 = 1.375
0.6875

𝑋 (2) = 𝑋 (1) + 𝑉 (1)

3.5 1.125
= 2.25 + 1.375
1.625 0.6875

4.625
(2)
𝑋 = 3.625
2.3125

3rdIteration:𝑘 = 2

𝑟 (2) = 𝑏 − 𝐴𝑋 (2)

7 2 −1 0 4.625
= 1 − −1 2 −1 3.625
1 0 −1 2 2.3125

7 5.625
= 1 − 2.625
1 1

150
 1.375
𝑟 (2) = 0.6875 SPACE OF HINTS
0

𝑉 (2) = 𝐷 + 𝐿 −1 (2)
𝑟

1 0 0
2
1 1 1.375
= 4 2 0
0.6875
0
1 1 1
8 4 2

0.6875
𝑉 (2) = 0.6875
0.3438

𝑋 (3) = 𝑋 (2) + 𝑉 (2)

4.625 0.6875
= 3.625 + 0.6875
2.3125 0.3438

5.3125
𝑋 (3) = 4.3125
2.6563

Successive Over Relaxation (SOR) Method

This method is the generalization of the Gauss-Seidel method. This

method is used for when the coefficient matrix of the system of equation
is symmetric and has „𝑃𝑟𝑜𝑝𝑒𝑟𝑡𝑦 𝐴‟. The auxiliary vector 𝑥 can be
written as,

𝑋 (𝑘+1) = −𝐷 −1 𝐿𝑋 (𝑘+1) − 𝐷 −1 𝑈𝑋 (𝑘) + 𝐷 −1 𝑏 (1)

The final solution of the system is written as,

𝑋 (𝑘+1) = 𝑋 (𝑘) + 𝑤 𝑋 (𝑘+1) − 𝑋 (𝑘)

𝑋 (𝑘+1) = 1 − 𝑤 𝑋 (𝑘) + 𝑤𝑋 (𝑘+1) (2)

151
 (𝑘+1)
𝑋 (𝑘+1) − 1 − 𝑤 𝑋 (𝑘)
⟹𝑋 =
SPACE FOR HINTS 𝑤

Substituting (2) in (1), we get

𝑋 (𝑘+1) = [ 1 − 𝑤 𝑋 (𝑘) + 𝑤𝑋 (𝑘+1) ](−𝐷 −1 𝐿) − 𝐷 −1 𝑈𝑋 (𝑘) + 𝐷 −1 𝑏

𝑋 (𝑘+1) − 1 − 𝑤 𝑋 (𝑘)
= −𝐷 −1 𝐿𝑋 (𝑘+1) − 𝐷 −1 𝑈𝑋 (𝑘) + 𝐷 −1 𝑏
𝑤

After simplification,

𝑋 (𝑘+1) = 𝐷 + 𝑤𝐿 −1
𝐷 1 − 𝑤 − 𝑤𝑈 𝑋 (𝑘) + 𝑤(𝐷 + 𝑤𝐿)−1 𝑏 (3)

𝑋 (𝑘+1) = 𝐻𝑋 (𝑘) + 𝑐 (4)

−1
where 𝐻 = 𝐷 + 𝑤𝐿 𝐷 1 − 𝑤 − 𝑤𝑈 , 𝑐 = 𝑤(𝐷 + 𝑤𝐿)−1 𝑏

Add and sub 𝑋 (𝑘) in (3) of RHS.

𝑋 (𝑘+1) = 𝑋 (𝑘) − [𝑋 (𝑘) − 𝐷 + 𝑤𝐿 −1

𝐷 1 − 𝑤 − 𝑤𝑈 𝑋 (𝑘) ]
+ 𝑤(𝐷 + 𝑤𝐿)−1 𝑏

= 𝑋 (𝑘) − 𝐷 + 𝑤𝐿 −1
𝐷 + 𝑤𝐿 − 𝐷 1 − 𝑤 + 𝑤𝑈 𝑋 (𝑘)
+ 𝑤(𝐷 + 𝑤𝐿)−1 𝑏

= 𝑋 (𝑘) − 𝐷 + 𝑤𝐿 −1
𝐷 + 𝑤𝐿 − 𝐷 + 𝐷𝑤 + 𝑤𝑈 𝑋 (𝑘)
+ 𝑤(𝐷 + 𝑤𝐿)−1 𝑏

= 𝑋 (𝑘) − 𝐷 + 𝑤𝐿 −1
𝐿 + 𝐷 + 𝑈 𝑤𝑋 (𝑘) + 𝑤(𝐷 + 𝑤𝐿)−1 𝑏

= 𝑋 (𝑘) − 𝐷 + 𝑤𝐿 −1
𝐴𝑤𝑋 (𝑘) + 𝑤(𝐷 + 𝑤𝐿)−1 𝑏

= 𝑋 (𝑘) + 𝐷 + 𝑤𝐿 −1
𝑤[𝑏 − 𝐴𝑋 (𝑘) ]

= 𝑋 (𝑘) + 𝐷 + 𝑤𝐿 −1
𝑤𝑟 𝑘
(5)

𝑘
where 𝑟 = [𝑏 − 𝐴𝑋 (𝑘) ] is residual and 𝑉 (𝑘) = 𝐷 + 𝑤𝐿 −1
𝑤𝑟 𝑘

(5) ⟹ 𝑋 (𝑘+1) = 𝑋 (𝑘) − 𝑉 (𝑘)

152
The above equation is the SOR method in its error format.
SPACE OF HINTS

Note:

1. When 𝑤 = 1, the above method reduces to the Gauss-Seidel method.

The quantity𝑤 is called the relaxation parameter and 𝑋 𝑘+1 is weighted
mean of 𝑋 𝑘+1 and 𝑋 𝑘 .

2. If 𝑤 > 1 then the method is called over relaxation method and

if𝑤 < 1, the method under relaxation method. Since the weights are
non-negative and for 0 ≤ 𝑤 ≤ 1.

Theorem: 2.4.1

Let A be a square Matrix. Then lim𝑚 →∞ 𝐴𝑚 = 0if 𝐴 < 1 (or) iff

𝜌 𝐴 < 1.

Proof:

If 𝐴 < 1 , then we have

𝐴𝑚 = 𝐴𝑚 −1 . 𝐴

≤ 𝐴𝑚 −1 𝐴

≤ 𝐴𝑚 −2 . 𝐴 𝐴

≤ 𝐴𝑚 −2 𝐴 . 𝐴

≤ 𝐴𝑚 −2 𝐴 2

𝐴𝑚 ≤ 𝐴 𝑚

Hence lim𝑚 →∞ 𝐴𝑚 ≤ lim𝑚 →∞ 𝐴 𝑚

∴ lim 𝐴𝑚 = 0 𝑎𝑛𝑑 𝐴 = 0 𝑖𝑓𝑓 𝐴 = 0

𝑚 →∞

Assume that all the eigenvalues of 𝐴 are distinct.

153
 Then there exist a similarity transformation 𝑆 such that, 𝑨 =
SPACE FOR HINTS 𝑺−𝟏 𝑫𝑺where 𝑫 is the diagonal matrix, having eigenvalues of 𝑨 on the
diagonal.

𝑖𝑒) 𝐴 = 𝑆 −1 𝐷𝑆

𝐴2 = 𝐴. 𝐴 = (𝑆 −1 𝐷𝑆)(𝑆 −1 𝐷𝑆)

𝐴2 = 𝑆 −1 𝐷 2 𝑆

Similarly, 𝐴3 = 𝑆 −1 𝐷 3 𝑆

𝜆1 𝑚 0 ⋯ 0
𝑚
In general, 𝐴𝑚 = 𝑆 −1 𝐷 𝑚 𝑆 where 𝐷 𝑚 = 0 𝜆2 ⋯ 0
⋮ ⋮ ⋱ ⋮
0 0 ⋯ 𝜆𝑛 𝑚

∴ lim 𝐴𝑚 = 0 𝑖𝑓𝑓 ∴ lim 𝐷 𝑚 = 0

𝑚 →∞ 𝑚 →∞

𝑖𝑒) 𝑖𝑓𝑓 𝜆𝑖 < 1, ∀𝑖

𝑖𝑒) 𝑖𝑓𝑓 𝜌 𝐴 < 1.

Theorem: 2.4.2

The infinite series 𝐼 + 𝐴 + 𝐴2 + ⋯ converges ifflim𝑚 →∞ 𝐴𝑚 = 0. The

series converges to(𝐼 − 𝐴)−1 .

Proof:

Let the infinite series 𝐼 + 𝐴 + 𝐴2 + ⋯ be converges.

Let 𝑆𝑛 = 𝐼 + 𝐴 + 𝐴2 + ⋯ + 𝐴𝑚 + ⋯

We know that 𝐼 + 𝑥 + 𝑥 2 + ⋯ = (𝐼 − 𝑥)−1

∴ 𝑆𝑛 = (𝐼 − 𝐴)−1

Now, lim𝑚 →∞ 𝑆𝑛 = (𝐼 − 𝐴)−1 = 𝐿 (𝑠𝑎𝑦)

154
 𝐴𝑚 = 𝑆𝑚 − 𝑆𝑚 −1
SPACE OF HINTS

Applying lim𝑚 →∞ on both sides,

lim 𝐴𝑚 = lim 𝑆𝑚 − lim 𝑆𝑚 −1 = 𝐿 − 𝐿 = 0

𝑚 →∞ 𝑚 →∞ 𝑚 →∞

∴ lim 𝐴𝑚 = 0
𝑚 →∞

Suppose thatlim𝑚 →∞ 𝐴𝑚 = 0.

To prove the infinite series 𝐼 + 𝐴 + 𝐴2 + ⋯ converges

Now, by theorem 2.4.1,lim𝑚 →∞ 𝐴𝑚 = 0 ⇔ 𝜌 𝐴 < 1.

−1
Hence 𝐼 − 𝐴 ≠ 0 and hence 𝐼 − 𝐴 exists.

Theorem: 2.4.3

𝑘+1 𝑘
The iteration method of the form 𝑋 = 𝐻𝑋 + 𝐶, 𝑘=
0,1,2, … for the solution of𝑨𝑿 = 𝒃converges to the exact solution for
any initial vector if 𝐻 < 1 (or) if𝜌 𝐻 < 1.

Proof:

Given 𝑋 (𝑘+1) = 𝐻𝑋 (𝑘) + 𝑐

Without loss of generality, we take the initial method 𝑋 (0) and put
𝑘 = 0 in (1) we get

𝑋 (1) = 𝐻𝑋 (0) + 𝑐

𝑋 (1) = 𝑐

Put 𝑘 = 1, 𝑋 (2) = 𝐻𝑋 1 + 𝑐

= 𝐻𝑐 + 𝑐

𝑋 (2) = (𝐻 + 𝐼)𝑐

Put 𝑘 = 2, 𝑋 (3) = 𝐻𝑋 2 + 𝑐

155
 𝑋 (3) = (𝐼 + 𝐻 + 𝐻 2 + ⋯ )𝑐
SPACE FOR HINTS
Applying lim𝑚 →∞ on both sides,

lim 𝑋 (𝑘+1) = lim (𝐼 + 𝐻 + 𝐻 2 + ⋯ + 𝐻 𝑘 + ⋯ )𝑐

𝑘→∞ 𝑘→∞

= (𝐼 − 𝐻)−1 (𝑏𝑦 𝑡𝑕𝑒𝑜𝑟𝑒𝑚: 2.4.2)

If 𝐻 < 1 (or) if 𝜌 𝐻 < 1

In the case of the Jacobi method, we have

−1
𝐼−𝐻 𝑐 = [𝐼 − (−𝐷 −1 (𝐿 + 𝑈))]−1 𝐷 −1 𝑏

{∵ 𝐻 = −𝐷 −1 𝐿 + 𝑈 , 𝑐 = 𝐷 −1 𝑏}

= 𝐷𝐷 −1 + 𝐷 −1 (𝐿 + 𝑈)) −1
𝐷 −1 𝑏

= 𝐷 −1 (𝐷 + 𝐿 + 𝑈)) −1
𝐷 −1 𝑏

−1
= (𝐷 + 𝐿 + 𝑈)) 𝐷𝐷 −1 𝑏

= 𝐴−1 𝐷𝐷 −1 𝑏 [ since 𝐴 = 𝐷 + 𝐿 + 𝑈]

= 𝐴−1 𝑏

−1
𝐼−𝐻 𝑐=𝑋

−1
𝐼−𝐻 𝑐 satisfies the system of vectors X.

Similarly, the result is can be proved for the gauss-Seidel and

𝑆𝑂𝑅 methods.

Theorem: 2.4.4

A necessary and sufficient condition for convergence of an iterative

method of the form 𝑋 (𝑘+1) = 𝐻𝑋 (𝑘) + 𝑐, 𝑘 = 0,1,2, … is that eigen
vales of the iteration matrix satisfy, 𝜆𝑖 (𝑯) < 1, 𝑖 = 1 𝑡𝑜 𝑛.

156
Proof:
SPACE OF HINTS

We prove the result when the iteration matrix A has n independent

eigenvectors.

𝑖𝑒) 𝑋1 , 𝑋2 , … 𝑋𝑛 with eigenvalues 𝜆1 , 𝜆2 , … 𝜆𝑛 respectively. The

error vector 𝜖 (0) can be written as 𝜖 (0) = 𝑐1 𝑋1 + 𝑐2 𝑋2 + ⋯ + 𝑐𝑛 𝑋𝑛

Using 𝜖 (𝑘) = 𝐻 𝑘 𝜖 (0) , 𝑘 = 0,1,2, … we get

𝑘
𝜖 = 𝐻 𝑘 𝑐1 𝑋1 + 𝑐2 𝑋2 + ⋯ + 𝑐𝑛 𝑋𝑛

= 𝑐1 𝐻 𝑘 𝑋1 + 𝑐2 𝐻 𝑘 𝑋2 + ⋯ + 𝑐𝑛 𝐻 𝑘 𝑋𝑛 (1)

Since 𝐴𝑋 = 𝜆𝑋 ⟹ 𝐻𝑋 = 𝜆𝑋

𝑖𝑒) 𝐻 𝑘 𝑋 = 𝜆𝑖 𝑘 𝑋 𝑖 = 1 𝑡𝑜 𝑛

∴ 𝜖 𝑘
= 𝑐1 𝜆1 𝑘 𝑋1 + 𝑐2 𝜆2 𝑘 𝑋2 + ⋯ + 𝑐𝑛 𝜆𝑛 𝑘 𝑋𝑛

Necessary part:

Let us assume that the iteration method 𝑋 𝑘+1 = 𝐻𝑋 𝑘 + 𝐶,

𝑘 = 0,1,2, … is converges.

Hence lim𝑚 →∞ 𝜖 (𝑘) = 0for any arbitrary initial vector and for any
arbitrary error vector𝜖 (0) .

Then by (1), the magnitudes, of the eigenvalue 𝜆𝑖 < 1, 𝑖 = 1 𝑡𝑜 𝑛.

Hence 𝐻 satisfies 𝜆𝑖 (𝐻) < 1

Sufficient part:

For 𝜆𝑖 < 1, 𝑖 = 1 𝑡𝑜 𝑛

Then by (1),lim𝑘 →∞ 𝜖 (𝑘) = 𝑛

𝑖=1 𝐶𝑖 lim𝑘→∞ 𝜆𝑖 (𝑘)

157
 ∴ lim 𝜖 (𝑘) = 0 ∴ lim 𝜆𝑖 (𝑘) = 0
𝑘→∞ 𝑘→∞
SPACE FOR HINTS

Hence the above iteration method isconverges.

Definition:

The rate of convergence of an iteration method is given by 𝑉 =

− log10 𝜌 𝑯 or also as where 𝜌 𝑯 is the spectral radius of 𝑯.

Theorem: 2.4.5

If 𝑨 is strictly diagonally dominant matrix, then the Jacobi

iteration scheme converges for any initial starting vector.

Proof:

The Jacobi iteration scheme is given by

𝑋 (𝑘+1) = 𝐻𝑋 (𝑘) + 𝑐 (1)

where 𝐻 = −𝐷 −1 𝐿 + 𝑈 , 𝑐 = 𝐷 −1 𝑏

𝑋 (𝑘+1) = −𝐷 −1 𝐿 + 𝑈 𝑋 (𝑘) + 𝐷 −1 𝑏

= −𝐷 −1 𝐴 − 𝐷 𝑋 (𝑘) + 𝐷 −1 𝑏 {𝑠𝑖𝑛𝑐𝑒 𝐷 + 𝑈 + 𝐿
= 𝐴}

= 𝐼−𝐷 −1 𝐴 𝑋 (𝑘) + 𝐷 −1 𝑏

By theorem 2.4.4, the iteration scheme will be convergent if

𝐼−𝐷 −1 𝐴 < 1 (2)

Using absolute row sum norm,

𝑛
1
𝑎𝑖𝑗 < 1, ∀𝑖
𝑎𝑖𝑖
𝑗 =1
𝑖≠𝑗

158
 1 1
⟹ < 𝑛 , ∀𝑖 SPACE OF HINTS
𝑎𝑖𝑖 𝑗 =1 𝑎𝑖𝑗
𝑖≠𝑗

⟹ 𝑎𝑖𝑖 > 𝑎𝑖𝑗 , ∀𝑖

𝑗 =1
𝑖≠𝑗

which is true, since the matrix 𝑨 is strictly diagonally dominant.

Theorem: 2.4.6

If A is a strictly diagonally dominant matrix, then the Gauss-Seidel

iteration scheme converges for any initial starting vector.

Proof:

Gauss-Seidel iteration scheme is given by

𝑋 (𝑘+1) = 𝐻𝑋 (𝑘) + 𝑐 (1)

where 𝐻 = −(𝐷 + 𝐿)−1 𝑈, 𝑐 = (𝐷 + 𝐿)−1 𝑏

𝑋 (𝑘+1) = −(𝐷 + 𝐿)−1 𝑈𝑋 (𝑘) + (𝐷 + 𝐿)−1 𝑏

−1
=− 𝐷+𝐿 𝐴 − 𝐷 + 𝐿 𝑋 (𝑘) + 𝐷 + 𝐿 −1
𝑏

{𝑠𝑖𝑛𝑐𝑒 𝐷 + 𝑈 + 𝐿 = 𝐴 ⟹ 𝑈 = 𝐴 − (𝐿 + 𝐷)}

−1
= 𝐼− 𝐷 + 𝐿 𝐴 𝑋 (𝑘) + 𝐷 + 𝐿 −1
𝑏

The iteration scheme will be convergent if

−1
𝜌 𝐼− 𝐷 + 𝐿 𝐴 <1

−1
Let 𝜆 be the eigen value of 𝐼− 𝐷 + 𝐿 𝐴.

−1
Then 𝐼− 𝐷 + 𝐿 𝐴 𝑋 = 𝜆𝑋 {∵ 𝐴𝑋 = 𝜆𝑋}

Multiply by 𝐷 + 𝐿

159
 ⟹ 𝐷 + 𝐿 𝑋 − 𝐴𝑋 = 𝜆 𝐷 + 𝐿 𝑋
SPACE FOR HINTS
⟹ 𝐷+𝐿 −𝐴 𝑋 =𝜆 𝐷+𝐿 𝑋

⟹ −𝑈𝑋 = 𝜆 𝐷 + 𝐿 𝑋 {𝑠𝑖𝑛𝑐𝑒 𝐷 + 𝑈 + 𝐿 = 𝐴

⟹ −𝑈 = (𝐷 + 𝐿 − 𝐴)}

𝑛 𝑖

⟹− 𝑎𝑖𝑗 𝑋𝑗 = 𝜆 𝑎𝑖𝑗 𝑋𝑗
𝑗 =𝑖+1 𝑗 =𝑖+1

𝑛 𝑖−1

⟹− 𝑎𝑖𝑗 𝑋𝑗 = 𝜆 𝑎𝑖𝑗 𝑋𝑗 + 𝜆𝑎𝑖𝑖 𝑋𝑖

𝑗 =𝑖+1 𝑗 =𝑖+1

𝑛 𝑖−1

⟹ 𝜆𝑎𝑖𝑖 𝑋𝑖 = − 𝑎𝑖𝑗 𝑋𝑗 − 𝜆 𝑎𝑖𝑗 𝑋𝑗

𝑗 =𝑖+1 𝑗 =𝑖+1

𝑛 𝑖−1

⟹ 𝜆𝑎𝑖𝑖 𝑋𝑖 ≤ 𝑎𝑖𝑗 𝑋𝑗 + 𝜆 𝑎𝑖𝑗 𝑋𝑗 (2)

𝑗 =𝑖+1 𝑗 =𝑖+1

Since 𝑋 is an eigen vector, 𝑋 ≠ 0.

Without loss of generality, we assume that 𝑥 ∞ = 1.

Choose an index 𝑖 such that 𝑥𝑖 = 1 and 𝑥𝑗 ≤ 1 ∀ 𝑗 ≠ 𝑖.

𝑛 𝑖−1

2 ⟹ 𝜆 𝑎𝑖𝑖 . 1 ≤ 𝑎𝑖𝑗 + 𝜆 𝑎𝑖𝑗

𝑗 =𝑖+1 𝑗 =1

𝑖−1 𝑛

𝜆 𝑎𝑖𝑖 − 𝑎𝑖𝑗 ≤ 𝑎𝑖𝑗

𝑗 =1 𝑗 =𝑖+1

𝑛
𝑗 =𝑖+1 𝑎𝑖𝑗
∴ 𝜆 ≤ (3)
𝑖−1
𝑎𝑖𝑖 − 𝑗 =1 𝑎𝑖𝑗

160
Since A is strictly diagonally dominant
SPACE OF HINTS
𝑖−1

𝑖𝑒) 𝑎𝑖𝑖 > 𝑎𝑖𝑗

𝑗 =1

𝑖−1 𝑛

⟹ 𝑎𝑖𝑖 − 𝑎𝑖𝑗 > 𝑎𝑖𝑗

𝑗 =1 𝑗 =𝑖+1

𝑖−1
𝑎𝑖𝑖 − 𝑗 =1 𝑎𝑖𝑗
⟹ 𝑛
>1
𝑗 =𝑖+1 𝑎𝑖𝑗

𝑛
𝑗 =𝑖+1 𝑎𝑖𝑗
⟹ <1
𝑎𝑖𝑖 − 𝑖−1 𝑗 =1 𝑎𝑖𝑗

−1
From 3 ⟹ 𝜆 < 1 since 𝜌 𝐼− 𝐷 + 𝐿 𝐴 <1

∴ The gauss-Seidal scheme converges for any initial starting vector.

Example:2.4.3

Consider the system of equations

1 −𝑎 𝑥1 𝑏1
= where 𝑎 is a real constant.
−𝑎 1 𝑥2 𝑏2

(i). For which values of 𝑎, the Jacobi and Gauss-Seidal methods

converge.

(ii). For find the value of 𝑤 which minimizes the special radius of the
SOR iteration matrix.

Solution:

(i). The two iteration methods are Gauss Jacobi and Gauss-Seidel.

The formula for Gauss-Jacobi method is

𝑋 (𝑘+1) = 𝐻𝑋 (𝑘) + 𝑐 where 𝐻 = −𝐷 −1 𝐿 + 𝑈 , 𝑐 = 𝐷 −1 𝑏 (1)

161
 𝑋 (𝑘+1) = −𝐷 −1 𝐿 + 𝑈 𝑋 (𝑘) + 𝐷 −1 𝑏
SPACE FOR HINTS
1 −𝑎 0 0 0 −𝑎 1 0
Given 𝐴 = , 𝐿= , 𝑈= , 𝐷=
−𝑎 1 −𝑎 0 0 0 0 1

1 0 0 −𝑎
𝐷 −1 = , 𝐿+𝑈 =
0 1 −𝑎 0

1 0 0 −𝑎
Now, 𝐻=−
0 1 −𝑎 0

0 −𝑎
=−
−𝑎 0

0 𝑎
𝐻=
𝑎 0

1 0 𝑏1 𝑏
𝑐= = 1
0 1 𝑏2 𝑏2

0 𝑎 (𝑘) 𝑏
(1) ⟹ 𝑋 (𝑘+1) = 𝑋 + 1
𝑎 0 𝑏2

∴ The eigen value of Jacobi iteration matrix is given by, 𝐻 − 𝜆𝐼 = 0

0 𝑎 𝜆 0
⟹ − =0
𝑎 0 0 𝜆

−𝜆 𝑎
⟹ =0
𝑎 −𝜆

⟹ 𝜆2 − 𝑎2 = 0

⟹ 𝜆2 = 𝑎2

⟹ 𝜆 = ±𝑎

∴𝜆= 𝑎 <1

∴ the condition for the convergesof the Jacobi iteration method is 𝑎 <
1.

162
Hence the spectral radius of Jacobi iteration matrix becomes𝜌 𝐻𝑗 =
SPACE OF HINTS
𝜆= 𝑎.

The formula for Gauss-Seidel method is,

𝑋 𝑘+1 = 𝐻𝑋 𝑘 + 𝑐 where 𝐻 = −(𝐷 + 𝐿)−1 𝑈, 𝑐 = (𝐷 + 𝐿)−1 𝑏 (2)

1 0 −1 1 0
𝐷+𝐿 = ⟹ 𝐷+𝐿 =
−𝑎 1 𝑎 1

1 0 0 −𝑎
𝐻=−
𝑎 1 0 0

0 −𝑎
=−
0 −𝑎2

0 𝑎
𝐻=
0 𝑎2

1 0 𝑏1
𝑐=
𝑎 1 𝑏2

0 𝑎 𝑘 1 0 𝑏1
2 ⟹ 𝑋 𝑘+1 = 2 𝑋 +
0 𝑎 𝑎 1 𝑏2

The eigenvalue of the Gauss-Seidel iteration matrix 𝐻 is given by,

𝐻 − 𝜆𝐼 = 0

−𝜆 𝑎
𝑖𝑒) 2 =0
0 𝑎 −𝜆

⟹ −𝜆 𝑎2 − 𝜆 = 0

⟹ −𝜆 𝑎2 − 𝜆 = 0

⟹ 𝑎2 − 𝜆 = 0 𝑜𝑟 𝜆 = 0

⟹ 𝑎2 = 𝜆 𝑜𝑟 𝜆 = 0

∴ The spectral radius of Gauss-Seidel iteration matrix becomes

𝜌 𝐻𝐺𝑆 = 𝑎2

163
 For convergence 𝑎2 < 1 ⟹ 𝑎 < 1.
SPACE FOR HINTS
(ii). The optimal relaxation factor for SOR method is given by

2 1 − 1 − 𝜇2
𝑊𝑜𝑝𝑡 = 𝑕𝑒𝑟𝑒 𝜇 = 𝑎
𝜇2

2 1 − 1 − 𝑎2
=
𝑎2

2 1 − 1 − (0.5)2
= (𝑓𝑜𝑟 𝑎 = 0.5)
(0.5)2

𝑊𝑜𝑝𝑡 = 1.0718

Example: 2.4.4

The system of equations𝑨𝑿 = 𝒃is to be solved iteratively by

1 𝑘 2
𝑋𝑛+1 = 𝑀𝑋𝑛 + 𝑏, suppose 𝐴 = , 𝑘≠ , 𝑘 𝑖𝑠 𝑟𝑒𝑎𝑙.
2𝑘 1 2

(a). Find a necessary and sufficient condition on 𝑘 for convergence of

the Jacobi method.

(b). For 𝑘 = 0.25 determine the optimal relaxation factor 𝑤, if the

system is to be solved with relaxation method.

Solution:

(a). The formula for Jacobi iteration method is

𝑋 (𝑛+1) = 𝐻𝑋 (𝑛) + 𝑐 where 𝐻 = −𝐷 −1 𝐿 + 𝑈 , 𝑐 = 𝐷 −1 𝑏 (1)

1 𝑘
Given 𝐴 =
2𝑘 1

We know that, 𝐴 = 𝐿 + 𝑈 + 𝐷

0 0 0 𝑘 1 0
⟹𝐿= , 𝑈= , 𝐷=
2𝑘 0 0 0 0 1

164
 1 1 1 0
𝐷 −1 = 𝑎𝑑𝑗 𝐷 = SPACE OF HINTS
𝐷 1 0 1

1 0
𝐷 −1 =
0 1

0 𝑘
𝐿+𝑈 =
2𝑘 0

1 0 0 𝑘 0 𝑘
∴𝐻=− =−
0 1 2𝑘 0 2𝑘 0

0 𝑘 𝑛 1 0 𝑏1
1 ⟹ 𝑋 𝑛+1 = − 𝑋 +
2𝑘 0 0 1 𝑏2

The eigenvalue of the Jacobi method for matrix H is 𝐻 − 𝜆𝐼 = 0

−𝜆 −𝑘
𝑖𝑒) =0
−2𝑘 −𝜆

⟹ 𝜆2 − 2𝑘 2 = 0

⟹ 𝜆2 = 2𝑘 2

⟹𝜆 =± 2𝑘

∴ 𝜌 𝐻𝑗 = 2 𝑘

1
For convergence 𝜌 𝐻𝑗 < 1 ⟹ 2 𝑘 < 1 ⟹ 𝑘 <
2

(b). The optimal relaxation factor is

2 1 − 1 − 𝜇2 2
𝑊𝑜𝑝𝑡 = = 1 − 1 − 2𝑘 2
𝜇2 2𝑘 2

𝑊𝑜𝑝𝑡 = 1.03337 for 𝑘 = 0.25,

2.5: Eigen values and eigen vectors

Consider the eigenvalue problem

165
 𝑨𝑿 = 𝝀 𝑿 (1)
SPACE FOR HINTS
The eigenvalues of a matrix 𝐴 are given by the roots of the
characteristic equation

𝑑𝑒𝑡 𝐴 − 𝜆𝐼 = 0 (2)

which when simplified gives the polynomial equation

𝑝(𝜆) = −1 𝑛 𝜆𝑛 + 𝑎1 𝜆𝑛−1 +. . . +𝑎𝑛 = 0 (3)

The coefficients of the above polynomial can be determined with the

help of the Faddeev-Leverrier method, we have

𝐵1 = 𝐴 𝑎𝑛𝑑 𝑎1 = 𝑡𝑟 𝐵1

1
𝐵2 = 𝐴 𝐵1 − 𝑎1 𝐼 𝑎𝑛𝑑 𝑎2 = 𝑡𝑟 𝐵2
2

1
𝐵3 = 𝐴 𝐵2 − 𝑎2 𝐼 𝑎𝑛𝑑 𝑎3 = 𝑡𝑟 𝐵3
3

1
𝐵𝑘 = 𝐴 𝐵𝑘−1 − 𝑎𝑘−1 𝐼 𝑎𝑛𝑑 𝑎𝑘 = 𝑡𝑟 𝐵𝑘
𝑘

1
𝐵𝑛 = 𝐴 𝐵𝑛−1 − 𝑎𝑛−1 𝐼 𝑎𝑛𝑑 𝑎𝑛 = 𝑡𝑟 𝐵𝑛
𝑛

where 𝑡𝑟 𝐴 = 𝑎11 + 𝑎22 + . . . + 𝑎𝑛𝑛 .

A nonzero vector 𝑥𝑖 such that 𝑨 𝒙𝒊 = 𝝀𝒊 𝒙𝒊 (4)

is called the eigenvector or characteristic vector corresponding to 𝜆𝑖 .

Multiplying (4) by an arbitrary constant 𝑐 and putting 𝑦𝑖 = 𝑐 𝑥𝑖 , we get

𝐴𝑦𝑖 = 𝜆𝑖 𝑦𝑖 (5)

166
which shows that an eigenvector is determined only to within an
SPACE OF HINTS
arbitrary multiplicative constant.Premultiplying (1) 𝑚 times by A we
obtain

𝐴𝑚 𝑥 = 𝜆𝑚 𝑥 (6)

which shows that 𝜆𝑚 is an eigenvalue of 𝐴𝑚 and x is the corresponding

eigenvector.

Substituting (6) into (3) we get

𝑝(𝐴) = 0

which gives the result that a square matrix A satisfies its own
characteristic equation. This result is known as the Cayley-Hamilton
theorem. We also find

𝑑𝑒𝑡 𝐴𝑇 − 𝜆𝐼 = 𝑑𝑒𝑡 𝐴 − 𝜆𝐼 = 0. (7)

Thus, 𝐴 and 𝐴𝑇 have the same eigenvalues. For distinct eigenvalues, if

𝑢1 , 𝑢2 , . . . . 𝑢𝑛 are the eigenvectors of A and 𝑣1 , 𝑣2 , . . . . 𝑣𝑛 are the
eigenvectors of 𝐴𝑇 then we have

𝐴𝑢𝑖 = 𝜆𝑖 𝑢𝑖 (8)

𝐴𝑇 𝑣𝑗 = 𝜆𝑗 𝑣𝑗 (9)

Taking transpose of (8) and (9) and post-multiplying by 𝑢𝑖 , we get

𝑣𝑗 𝑇 𝐴 𝑢𝑖 = 𝜆𝑖 𝑣𝑗 𝑇 𝑢𝑖

𝑣𝑗 𝑇 𝐴 𝑢𝑖 = 𝜆𝑗 𝑣𝑗 𝑇 𝑢𝑖

Subtracting these two equations, we get

(𝜆𝑖 − 𝜆𝑗 )𝑣𝑗 𝑇 𝑢𝑖 = 0.

𝐼𝑓 𝑖 ≠ 𝑗, 𝑡𝑕𝑒𝑛 𝜆𝑖 ≠ 𝜆𝑗 𝑎𝑛𝑑 𝑤𝑒 𝑕𝑎𝑣𝑒

167
 𝑣𝑗 𝑇 𝑢𝑖 = 0.
SPACE FOR HINTS
If 𝑖 = 𝑗, 𝑣𝑗 𝑇 𝑢𝑖 ≠ 0 and since the length of eigenvectors is arbitrarywe
normalize them such that

𝑣𝑗 𝑇 𝑢𝑖 = 1

Thus, we have

0, 𝑖 ≠ 𝑗
𝑣𝑗 𝑇 𝑢𝑖 =
1 𝑖=𝑗

Let A and B be two square matrices of same order. If a non-singular

matrix 𝑆 can be determined such that

𝐵 = 𝑆 −1 𝐴 𝑆 (10)

then the matrices A and B are said to be similar and those equation is

called a similarity transformation.

The matrix 𝑆 is called the similarity matrix. We may write

𝐴 = 𝑆 𝐵 𝑆 −1 (11)

If 𝜆𝑖 is an eigenvalue of A and 𝑢𝑖 is the corresponding eigenvector then

𝐴𝑢𝑖 = 𝜆𝑗 𝑢𝑖

or 𝑆 −1 𝐴𝑢𝑖 = 𝜆 𝑆 −1 𝑢𝑖 (12)

Substituting 𝑢𝑖 = 𝑆𝑣𝑖 in (12) and using (10) we get

𝐵𝑣𝑖 = 𝜆𝑖 𝑣𝑖 (13)

Thus, 𝑆 −1 𝐴𝑆 has the same eigenvalues as A and its eigenvectors 𝑣𝑖 are

obtained from the relation 𝑣𝑖 = 𝑆 −1 𝑢𝑖

168
A similarity transformation, where S is the matrix of eigenvectors,
SPACE OF HINTS
reduces a matrix A to its diagonal form. The eigenvalues of A are
located on the leading diagonal, of this diagonal matrix.

If the eigenvectors of A are linearly independent, then 𝑆 −1 exists and the

matrix A is said to be diagonalizable.

Suppose that the matrix A has eigenvalues 𝜆𝑖 with eigenvectors 𝑢𝑖 and

that A has an inverse 𝐴−1 . Then 𝐴𝑢𝑖 = 𝜆𝑖 𝑢𝑖

Which may be written as

1
𝐴−1 = 𝑢𝑖 = 𝑢
𝜆𝑖 𝑖

The inverse matrix 𝐴−1 has the same eigenvectors as A but has the
1
eigenvalues .
𝜆𝑖

Example: 2.5.1 For the matrix

1 2 −2 3 2 2
(i). 𝐴 = 1 1 1 (ii). 𝐴 = 2 5 2
1 3 −1 2 2 3

(a). Find all the eigenvalues and corresponding eigenvectors.

(b). Verify that 𝑺−𝟏 𝑨𝑺is a diagonal matrix, where 𝑺 is the matrix of
eigenvectors.

Solution:

1 2 −2
(i). Given 𝐴 = 1 1 1
1 3 −1

The given matrix can be written as 𝐴𝑋 = 𝜆𝑋

𝑖𝑒) 𝐴 − 𝜆𝐼 𝑋 = 0 (1)

169
 1 2 −2 𝜆 0 0
1 1 1 − 0 𝜆 0 =0
SPACE FOR HINTS
1 3 −1 0 0 𝜆

1−𝜆 2 −2
⟹ 1 1−𝜆 1 =0
1 3 −1 − 𝜆

⟹ 1−𝜆 1−𝜆 − 1+𝜆 −3 −2 − 1+𝜆 −1

+2 3−1+𝜆 =0

⟹ 1 − 𝜆 −1 + 𝜆 − 𝜆 − 𝜆2 − 3 − 2 −1 − 𝜆 − 1 + 2 2 + 𝜆 = 0

⟹ 1 − 𝜆 −4 − 𝜆2 − 2 −2 − 𝜆 + 2 𝜆 + 2 = 0

⟹ 1 − 𝜆 𝜆2 − 4 = 0

⟹ 1 − 𝜆 = 0 𝑜𝑟 𝜆2 − 4 = 0

⟹𝜆=1 𝑜𝑟 𝜆 = ±2

∴ The eigen values are 𝜆 = 1, 2, − 2.

Case: 1when 𝜆 = 1

1 ⟹ 𝐴−𝐼 𝑋 =0

1−1 2 −2
𝑖𝑒) 1 1−1 1 𝑋=0
1 3 −1 − 1

0 2 −2 𝑥1
1 0 1 𝑥2 = 0
1 3 −2 𝑥3

0𝑥1 + 2𝑥2 − 2𝑥3 = 0

𝑥1 + 0𝑥2 + 𝑥3 = 0

𝑥1 + 3𝑥2 − 2𝑥3 = 0

𝑥1 𝑥2 𝑥3
⟹ = = =𝑘
2 − 0 −2 − 0 0 − 2

170
 𝑥1 𝑥2 𝑥3
⟹ = = =𝑘 SPACE OF HINTS
2 −2 −2
𝑥1 2𝑘
⟹ 𝑥2 = −2𝑘
𝑥3 −2𝑘
𝑥1 1
1
Put 𝑘 = − ⟹ 𝑥2 = −1
2
𝑥3 −1

1
∴ The eigen vectors corresponding the eigenvalue 𝜆 = 1 is 𝑋 = −1
−1

Case: 2when 𝜆 = 2

1 ⟹ 𝐴 − 2𝐼 𝑋 = 0

1−2 2 −2
1 1−2 1 𝑋=0
1 3 −1 − 2

−1 2 −2 𝑥1
1 −1 1 𝑥2 = 0
1 3 −3 𝑥3

⟹ −𝑥1 + 2𝑥2 + 2𝑥3 = 0

⟹ 𝑥1 − 𝑥2 + 𝑥3 = 0

⟹ 𝑥1 + 3𝑥2 + 3𝑥3 = 0

𝑥1 𝑥2 𝑥3
⟹ = = = 𝑘 (𝑠𝑎𝑦)
2 − 2 −2 + 1 1 − 2

𝑥1 𝑥2 𝑥3
⟹ = = =𝑘
0 −1 −1
𝑥1 0
𝑥2 = −𝑘
𝑥3 −𝑘
𝑥1 0
𝑥
Put 𝑘 = −1, ⟹ 2 = 1
𝑥3 1

171
 Case: 3when 𝜆 = −2
SPACE FOR HINTS
1 ⟹ 𝐴−𝐼 𝑋 =0

3 2 −2
1 3 1 𝑋=0
1 3 1

3 2 −2 𝑥1
1 3 1 𝑥2 = 0
1 3 1 𝑥3

3𝑥1 + 2𝑥2 − 2𝑥3 = 0

𝑥1 + 3𝑥2 + 𝑥3 = 0

𝑥1 + 3𝑥2 + 𝑥3 = 0

𝑥1 𝑥2 𝑥3
⟹ = = = 𝑘(𝑠𝑎𝑦)
8 −5 7

𝑥1 𝑘
𝑥2 = −5𝑘
𝑥3 7𝑘
𝑥1 8
𝑥
Put 𝑘 = 1, ⟹ 2 = −5
𝑥3 7

−1 0 8
Now, 𝑆 = 1 1 −5 ⟹ 𝑆 = −1 7 + 5 + 8 1 − 1 = −12
1 1 7

−1
1 −12 8 8 −12 12 0
𝑆 = 12 15 −3 , 𝑎𝑑𝑗 𝑆 = 8 15 −1
12
0 −1 1 8 −3 1

−1
1 −12 8 8 1 2 −2 −1 0 8
𝑆 𝐴𝑆 = 12 15 −3 1 1 1 1 1 −5
12
0 −1 1 1 3 −1 1 1 7

1 12 0 0
= 0 24 0
12
0 0 −24

172
 1 0 0
𝑆 −1 𝐴𝑆 = 0 2 0 SPACE OF HINTS
0 0 −2

3 2 2
(i). Given 𝐴 = 2 5 2
2 2 3

The characteristic equation is 𝐴𝑋 = 𝜆𝑋

𝑖𝑒) 𝐴 − 𝜆𝐼 𝑋 = 0

3−𝜆 2 2
𝐴 − 𝜆𝐼 = 2 5−𝜆 2
2 2 3−𝜆

3−𝜆 2 2
𝐴 − 𝜆𝐼 = 0 ⟹ 2 5−𝜆 2 =0
2 2 3−𝜆

⟹ 3−𝜆 5 − 𝜆 3 − 𝜆 − 4 − 2 6 − 2𝜆 − 4 + 2 4 − 10 + 2𝜆
=0

⟹ 3 − 𝜆 15 − 8𝜆 + 𝜆2 − 4 − 4 + 4𝜆 + 4𝜆 − 12 = 0

⟹ 33 − 24𝜆 + 3𝜆2 − 11𝜆 + 8𝜆2 − 𝜆3 − 4 + 8𝜆 − 12 = 0

⟹ −𝜆3 + 11𝜆2 − 27𝜆 + 17 = 0

⟹ 𝜆3 − 11𝜆2 + 27𝜆 − 17 = 0

𝜆 = 1, 𝜆2 − 10𝜆 + 17 = 0 1 1 − 11 27 − 17

10± 100−4(17)
𝜆= 0 1 − 10 17
2

10± 32
= 1 − 10 17 0
2

𝜆 =5±2 2

The eigenvalues are 𝜆 = 1, 5 + 2 2, 5 − 2 2.

173
 Case: 1When 𝜆 = 1
SPACE FOR HINTS
2 2 2
2 4 2 𝑋=0
2 2 2
𝑥1 𝑥2 𝑥3
⟹ = = =𝑘
4−8 4−4 8−4
𝑥1 𝑥2 𝑥3
⟹ = = =𝑘
−4 0 4
𝑥1 −4
𝑥2 = 0
𝑥3 4
𝑥1 −1
𝑥
Put 𝑘 = 4 ⟹ 2 = 0
𝑥3 1

Case: 2 when 𝜆 = 5 + 2 2

−2 − 2 2 2 2
2 −2 2 2 𝑋=0
2 2 −2 − 2 2

𝑥1 𝑥2 𝑥3
⟹ = = =𝑘
4+4 2 4+4+4 2 4 2+8−4

𝑥1 𝑥2 𝑥3
⟹ = = =𝑘
4+4 2 8+4 2 4 2+4

𝑥1 4+4 2
𝑥2 = 8 + 4 2
𝑥3
4+4 2

𝑥1 1+ 2 1
⟹ 𝑥2 = 2 + 2 = 2
𝑥3 1
1+ 2

Case: 3 when 𝜆 = 5 − 2 2

174
 −2 + 2 2 2 2
SPACE OF HINTS
2 2 2 2 𝑋=0
2 2 −2 + 2 2

𝑥1 𝑥2 𝑥3
⟹ = = =𝑘
4−4 2 4+4−4 2 −2 2 + 8 − 4

𝑥1 𝑥2 𝑥3
⟹ = = =𝑘
4−4 2 8−4 2 4−2 2

𝑥1 4−4 2
𝑥2 = 8 − 4 2
𝑥3
4−4 2

𝑥1 1− 2 1
𝑥
⟹ 2 = 2− 2 = − 2
𝑥3 1
1− 2

1 1 1
𝑆= 0 2 2
−2 2 2 2

𝑆 =1 2 + 2 − − 2 + 1( 2)

2 2 2 2
𝑎𝑑𝑗 𝑆 = 0 2 −2
−2 2 2 2

2 2 0 −2 2 2 0 −2
−1
1 1
𝑆 = 2 2 2 = 1 2 1
4 2 4
2 −2 2 1 2 1

2 0 −2 3 2 2 1 1 1
−1
1
𝑆 𝐴𝑆 = 1 2 1 2 5 2 0 2 2
4
1 2 1 2 2 3 −2 2 2 2

1 0 0
−1
𝑆 𝐴𝑆 = 0 5+2 2 0
0 0 5− 2

175
 2.6: Bounds on eigenvalues
SPACE FOR HINTS
Theorem:2.6.1State and prove Gerschgorin theorem

The largest eigenvalue is modulus of a square matrix 𝑨 cannot exceed

the largest sum of the modulus of the moduli of the elements along any
row (or) any column.

Proof:

Let 𝜆𝑖 be an eigenvalue of 𝑨and 𝑿𝒊 be the corresponding eigenvector.

Then 𝐴𝑋𝑖 = 𝜆𝑖 𝑋𝑖 (1)

The system of homogeneous equation can be written as,

𝑎11 𝑥𝑖,1 + 𝑎12 𝑥𝑖,2 + ⋯ + 𝑎1𝑛 𝑥𝑖,𝑛 = 𝜆𝑖 𝑥𝑖,1

𝑎21 𝑥𝑖,1 + 𝑎22 𝑥𝑖,2 + ⋯ + 𝑎2𝑛 𝑥𝑖,𝑛 = 𝜆𝑖 𝑥𝑖,2
⋮
𝑎𝑛1 𝑥𝑖,1 + 𝑎𝑛2 𝑥𝑖,2 + ⋯ + 𝑎𝑛𝑛 𝑥𝑖,𝑛 = 𝜆𝑖 𝑥𝑖,𝑛

Let 𝑥𝑖,𝑘 = max𝑟 𝑥𝑖,𝑟

Now, select the 𝑘 𝑡𝑕 equation and divide it by 𝑥𝑖,𝑘

𝑎𝑘1 𝑥𝑖,1 + 𝑎𝑘2 𝑥𝑖,2 + ⋯ + 𝑎𝑘𝑛 𝑥𝑖,𝑛 = 𝜆𝑖 𝑥𝑖,𝑘

𝑥𝑖,1 𝑥𝑖,2 𝑥𝑖,𝑛

⟹ 𝜆𝑖 = 𝑎𝑘1 + 𝑎𝑘2 + ⋯ + 𝑎𝑘𝑛
𝑥𝑖,𝑘 𝑥𝑖,𝑘 𝑥𝑖,𝑘

𝑥 𝑖,1 𝑥 𝑖,2 𝑥 𝑖,𝑛

and 𝜆𝑖 = 𝑎𝑘1 + 𝑎𝑘2 + ⋯ + 𝑎𝑘𝑛 (2)
𝑥 𝑖,𝑘 𝑥 𝑖,𝑘 𝑥 𝑖,𝑘

𝑥 𝑖,𝑗
Since ≤ 1, 𝑗 = 1 𝑡𝑜 𝑛.
𝑥 𝑖,𝑘

2 ⟹ 𝜆𝑖 ≤ 𝑎𝑘1 + 𝑎𝑘2 + ⋯ + 𝑎𝑘𝑛 (3)

Since 𝑘 is arbitrary,

176
 𝑛
SPACE OF HINTS
3 ⟹ 𝜆 ≤ max 𝑎𝑖𝑗
𝑖
𝑗 =1

Since the eigenvalue of 𝑨𝑻 are the same as those of 𝑨.

∴ Applying the above procedure to𝑨𝑻 , we see that the theorem is true for
columns.

𝜆 ≤ max 𝑎𝑖𝑗
𝑗
𝑖=1

Theorem:2.6.2State and prove Brauer theorem

Let 𝑃𝑘 be the sum of the moduliof the elements along the 𝑘 𝑡𝑕 row
excluding the diagonal element 𝑎𝑘𝑘 . Then every eigenvalue of 𝑨 lies
inside or on the boundary of atleast one of the circles 𝜆 − 𝑎𝑘𝑘 = 𝑃𝑘 ,
𝑘 = 1 𝑡𝑜 𝑛.

Proof:

Let be 𝜆𝑖 an eigenvalue of 𝐴 and 𝑋𝑖 be the corresponding eigen vector.

Then 𝐴𝑋𝑖 = 𝜆𝑖 𝑋𝑖 (1)

Then system of homogenous equation can be written as,

𝑎11 𝑥𝑖,1 + 𝑎12 𝑥𝑖,2 + ⋯ + 𝑎1𝑛 𝑥𝑖,𝑛 = 𝜆𝑖 𝑥𝑖,1

𝑎21 𝑥𝑖,1 + 𝑎22 𝑥𝑖,2 + ⋯ + 𝑎2𝑛 𝑥𝑖,𝑛 = 𝜆𝑖 𝑥𝑖,2
⋮
𝑎𝑛1 𝑥𝑖,1 + 𝑎𝑛2 𝑥𝑖,2 + ⋯ + 𝑎𝑛𝑛 𝑥𝑖,𝑛 = 𝜆𝑖 𝑥𝑖,𝑛

Let 𝑥𝑖,𝑘 = max𝑟 𝑥𝑖,𝑟

Now, select the 𝑘 𝑡𝑕 equation

𝑎𝑘1 𝑥𝑖,1 + 𝑎𝑘2 𝑥𝑖,2 + ⋯ + 𝑎𝑘𝑛 𝑥𝑖,𝑛 = 𝜆𝑖 𝑥𝑖,𝑘

177
 𝑥𝑖,1 𝑥𝑖,2 𝑥𝑖,𝑛
⟹ 𝜆𝑖 = 𝑎𝑘1 + 𝑎𝑘2 + ⋯ + 𝑎𝑘𝑘 (1) + 𝑎𝑘𝑛
SPACE FOR HINTS 𝑥𝑖,𝑘 𝑥𝑖,𝑘 𝑥𝑖,𝑘

𝑥𝑖,1 𝑥𝑖,2 𝑥𝑖,𝑛

⟹ 𝜆𝑖 − 𝑎𝑘𝑘 = 𝑎𝑘1 + 𝑎𝑘2 + ⋯ + 𝑎𝑘𝑛
𝑥𝑖,𝑘 𝑥𝑖,𝑘 𝑥𝑖,𝑘

𝑥𝑖,1 𝑥𝑖,2 𝑥𝑖,𝑛

⟹ 𝜆𝑖 − 𝑎𝑘𝑘 = 𝑎𝑘1 + 𝑎𝑘2 + ⋯ + 𝑎𝑘𝑛 (2)
𝑥𝑖,𝑘 𝑥𝑖,𝑘 𝑥𝑖,𝑘

𝑥 𝑖,𝑗
Since ≤ 1, 𝑗 = 1 𝑡𝑜 𝑛.
𝑥 𝑖,𝑘

2 ⟹ 𝜆𝑖 − 𝑎𝑘𝑘 ≤ 𝑎𝑘1 + 𝑎𝑘2 + ⋯ + 𝑎𝑘𝑛

≤ 𝑎𝑘𝑖 = 𝑃𝑘
𝑖=1
𝑖≠𝑘

The eigenvalue of 𝑨 lie inside (or) on the union of the above circles.

Since 𝑨 and 𝑨𝑻 have the same eigenvalues, hence all the eigenvalues lie
in the union of the 𝑛 circles.

Applying the above procedure for the column, we get

𝜆𝑖 − 𝑎𝑘𝑘 ≤ 𝑎𝑗𝑘 , 𝑘 = 1 𝑡𝑜 𝑛
𝑖=1
𝑗 ≠𝑘

Note:

The bounds obtained here are all independent. Hence all the
eigenvalues of A must lie in the intersection of these bounds. These
circles are called the Gerschgorin bounds. If any of the Gerschgorin
circle is isolated, then it contains exactly one eigenvalue.

178
Example:2.6.1
SPACE OF HINTS

1 2 −1
Estimate the eigenvalues of the matrix 𝐴 = 1 1 1 using the
1 3 −1
Gerschgorin bounds. The eigen values lie in the regions

(i). 𝜆 ≤ 5, 𝜆 ≤6

(ii). The union of the circles

𝜆 − 1 ≤ 3, 𝜆 − 1 ≤ 2, 𝜆 + 1 ≤ 4 and

(iii). The union of the circles

𝜆 − 1 ≤ 2, 𝜆 − 1 ≤ 5, 𝜆+1 ≤2

Solution:

1 2 −1
(i). Given 𝐴 = 1 1 1
1 3 −1

The eigenvalues lie in the region is given by, 𝜆 ≤ max (𝑟𝑜𝑤 𝑠𝑢𝑚)

𝑖𝑒) 𝜆 ≤ max 4, 3, 5 ⟹ 𝜆 ≤5

Similarly, 𝜆 ≤ max (𝑐𝑜𝑙𝑢𝑚𝑛 𝑠𝑢𝑚)

𝑖𝑒) 𝜆 ≤ max(3, 6, 3)

⟹ 𝜆 ≤6

∴ Eigen values lie in the region 𝜆 ≤ 5, 𝜆 ≤6

To find the union of circles:

𝑛 𝑛

𝜆𝑖 − 𝑎𝑘𝑘 ≤ 𝑎𝑘𝑖 𝑎𝑛𝑑 𝜆𝑖 − 𝑎𝑘𝑘 ≤ 𝑎𝑗𝑘

𝑖=1 𝑖=1
𝑖≠𝑘 𝑗 ≠𝑘

179
 Row Column
SPACE FOR HINTS
𝑖𝑒) 𝜆 − 1 ≤ 3, 𝜆−1 ≤2

𝜆 − 1 ≤ 2, 𝜆−1 ≤5

𝜆 + 1 ≤ 4, 𝜆+1 ≤2

The first union of circle gives 𝜆 − 1 ≤ 3, 𝜆 + 1 ≤ 4 and

thesecond union of circle gives 𝜆 − 1 ≤ 5, 𝜆 + 1 ≤ 2.

In the intersection of these circles gives the required region and is

shown in the following figure.

The required region is 𝜆 − 1 ≤ 3, 𝜆 + 1 ≤ 4 , 𝜆 − 1 ≤ 5.

Now, 𝜆 − 1 ≤ 3 ⟹ −3 ≤ 𝜆 − 1 ≤ 3 ⟹ −2 ≤ 𝜆 ≤ 4

𝜆 + 1 ≤ 4 ⟹ −4 ≤ 𝜆 + 1 ≤ 4 ⟹ −3 ≤ 𝜆 ≤ 5

𝜆 − 1 ≤ 5 ⟹ −5 ≤ 𝜆 − 1 ≤ 5 ⟹ −4 ≤ 𝜆 ≤ 6

𝜆 + 1 ≤ 2 ⟹ −2 ≤ 𝜆 + 1 ≤ 2 ⟹ −3 ≤ 𝜆 ≤ 1

𝜆𝑖𝑚𝑔

𝜆𝑟𝑒𝑎𝑙

-5 -4 -3 -2 -1 0 1 2 3 4 5 6

180
 SPACE OF HINTS

Hence the intersection of region is −2 ≤ 𝜆 ≤ 1 and it is given in the

above figure.

∴ −2 ≤ 𝜆 ≤ 1 is an Gerschgorin bound.

Example:2.6.2

Find the interval which contains the eigenvalue of the symmetric matrix
3 2 2
𝐴= 2 5 2
2 2 3

Solution: The eigen values lie in the region

(i). 𝜆 ≤ max 𝑟𝑜𝑤 𝑠𝑢𝑚 = max 7, 9, 7 = 9

⟹ 𝜆 ≤9

𝜆 ≤ max 𝑐𝑜𝑙𝑢𝑚𝑛 𝑠𝑢𝑚 = max 7, 9, 7 = 9

⟹ 𝜆 ≤9

To find the union of the circles

𝜆𝑖 − 𝑎𝑘𝑘 ≤ 𝑎𝑘𝑖
𝑖=1
𝑖≠𝑘
𝑛

𝜆𝑖 − 𝑎𝑘𝑘 ≤ 𝑎𝑗𝑘
𝑖=1
𝑗 ≠𝑘

Row Column

𝑖𝑒) 𝜆 − 3 ≤ 4, 𝜆−3 ≤4

𝜆 − 5 ≤ 4, 𝜆−5 ≤4

𝜆 − 3 ≤ 4, 𝜆−3 ≤4

181
 ∴ The union of the circles gives the following interval
SPACE FOR HINTS
−4 ≤ 𝜆 − 3 ≤ 4 −4≤𝜆−5≤4 −4≤𝜆−3≤4

⟹ −1 ≤ 𝜆 ≤ 7 ⟹1≤𝜆≤9 ⟹ −1 ≤ 𝜆 ≤ 7

Hence the intersection of the region is −1 ≤ 𝜆 ≤ 9

Example:2.6.3

2 2
Find 𝑨10 when 𝑨 =
2 −1

Solution: Reduce the given matrix 𝑨 to its diagonal form by a

similarity transformation. Alsothe given matrix 𝑨 is symmetric which
implies there exist a orthogonal matrix 𝑺 which reduces to𝑫.

∴ 𝑆 −1 𝐴𝑆 = 𝐷 𝑜𝑟 𝐴 = 𝑆𝐷𝑆 𝑇

𝐴𝑚 has an eigen value 𝜆𝑚 . Hence 𝐴10 = 𝑆𝐷10 𝑆 𝑇

cos 𝜃 − sin 𝜃
Let 𝑆 =
sin 𝜃 cos 𝜃

cos 𝜃 sin 𝜃 2 2 cos 𝜃 − sin 𝜃

Then 𝑆 −1 𝐴𝑆 =
− sin 𝜃 cos 𝜃 2 −1 sin 𝜃 cos 𝜃

2 cos 𝜃 + 2 sin 𝜃 2 cos 𝜃 − sin 𝜃 cos 𝜃 − sin 𝜃

=
−2 sin 𝜃 + 2 cos 𝜃 −2 sin 𝜃 − cos 𝜃 sin 𝜃 cos 𝜃

= 2 cos 2 𝜃 + 2 sin 𝜃 cos 𝜃 + 2 cos 𝜃 sin 𝜃 − sin2 𝜃 −2 cos 𝜃 sin 𝜃 − 2 sin2 𝜃 + 2 cos 2 𝜃 − sin 𝜃 cos 𝜃
−2 sin 𝜃 cos 𝜃 + 2 cos 2 𝜃 − 2 sin2 𝜃 − cos 𝜃 sin 𝜃 2 sin2 𝜃 − 2 sin 𝜃 cos 𝜃 − 2 cos 𝜃 sin 𝜃 − cos 2 𝜃

2 cos2 𝜃 + 4 sin 𝜃 cos 𝜃 − sin2 𝜃 −3 sin 𝜃 cos 𝜃 + 2 cos2 𝜃 − sin2 𝜃

=
−3 sin 𝜃 cos 𝜃 + 2 cos 2 𝜃 − sin2 𝜃 2 sin2 𝜃 − 4 cos 𝜃 sin 𝜃 − cos 2 𝜃

2 cos2 𝜃 + 2 sin 2𝜃 − sin2 𝜃 −3

2 sin 2𝜃 + 2 cos 2𝜃
=
−3 2 sin2 𝜃 − 2 sin 2𝜃 − cos2 𝜃
2 sin 2𝜃 + 2 cos 2𝜃

We know that, 𝑆 −1 𝐴𝑆 = 𝐷

182
 ∴ −3 2 sin 2𝜃 + 2 cos 2𝜃 = 0
SPACE OF HINTS

3
2 sin 2𝜃 = 2 cos 2𝜃

𝑠𝑖𝑛𝑒 2𝜃 2.2
=
𝑐𝑜𝑠 2𝜃 3

4
⟹ tan 2𝜃 =
3

⟹ 2𝜃 = tan−1 4 3

⟹ 𝜃 = 0.4637

2.99987 0 3 0
𝑆 −1 𝐴𝑆 = =
0 −1.999 0 −2

2 1
Let us consider the orthogonal matrix 𝑆 = . Here 𝑆 𝑇 = 𝑆 −1
−1 2

To find𝑨𝟏𝟎 = 𝑺𝑫𝟏𝟎 𝑺𝑻 :

1 1 2 1
𝑆 −1 = 𝑎𝑑𝑗 𝑆 =
𝑆 5 −1 2

2 1
5 5
−1 2 1
𝑆𝑆 =
−1 2 −1 2
5 5

4 1 2 2
5+ 5 5− 5
=
2 2 1 4
5− 5 5+ 5

1 0
𝑆𝑆 −1 =
0 1

2 1 2 1 4+1 2−2
𝑆𝑆 𝑇 = =
−1 2 −1 2 2−2 1+4

183
 1 0
𝑆𝑆 𝑇 = 5
0 1
SPACE FOR HINTS

1 2 1
∴ 𝑆 −1 = 𝑆 𝑇 =
5 −1 2

𝐴10 = 𝑆𝐷10 𝑆 𝑇

=
2 −1 310 0 1 2 1
1 2 0 −210 5 −1 2

1 2 −1 310 × 2 310
=
5 1 2 −210 210 × 2

1 4 × 310 + 210 2 × 310 − 211

𝐴10 =
5 310 × 2 − 211 310 + 212

184
 SPACE OF HINTS

Unit-3
3.1: Introduction
Consider the problem of approximating a given function by a class of
simpler functions, mainly polynomials. There are two main uses of
interpolation or interpolating polynomials. The first use is in
reconstructing the function 𝑓(𝑥) when it is not explicitly and only the
values of 𝑓(𝑥) and /or its certain order derivatives at a set of points,
called nodes, tabular points or arguments are known.

The second use is to replace the function 𝑓(𝑥) by an

interpolating polynomial 𝑃(𝑥) so that many common operations such as
determination of roots, differentiation and integration etc. which are
intended for function 𝑓(𝑥) may be performed using 𝑃 𝑥 . In
approximation, we measure the derivation of the given function 𝑓(𝑥)
from the approximating polynomial P(𝑥) for all values of 𝑥 over a given
interval 𝑎, 𝑏 .

Definition:3.1.1

A Polynomial 𝑃(𝑥) is called an interpolating polynomial if the values

of 𝑃(𝑥) and /or its certain order derivatives consider with those of 𝑓(𝑥)
and /or its same order derivatives at one or more tabular points.

Taylor Series:
If the polynomial P(𝑥) is written as the Taylor‟s expansion, for the
function 𝑓(𝑥) about a point 𝑥0 , 𝑥0 ∈ [𝑎, 𝑏] in the form

185
 1
𝑃 𝑥 = 𝑓 𝑥0 + 𝑥 − 𝑥0 𝑓 ′ 𝑥0 + (𝑥 − 𝑥0 )2 𝑓 ′′ 𝑥0 + ⋯
SPACE FOR HINTS 2!
1
+ 𝑥
𝑛!
− 𝑥0 𝑛 𝑓 𝑛
𝑥0 (1)

then 𝑃(𝑥) may be regarded as an indepolating polynomial of degree 𝑛 ,

satisfying the conditions,
𝑘 𝑘
𝑃 𝑥0 = 𝑓 𝑥0 , 𝑘 = 0,1,2, ⋯ 𝑛 (2)
𝑥−𝑥 0 𝑛 +1 𝑛+1
The term 𝑅𝑛 = 𝑓 𝜉 , 𝑥0 < 𝜉 < 𝑥 (3)
𝑛+1 !

Which has been neglected in the Taylor series and is called the
reminder (or) truncation error.
𝑛+1
Also,𝑀𝑛+1 = max𝑎≤𝑥≤𝑏 𝑓 (𝑥)

Hence, the Taylor series becomes

1 𝑛+1 𝑛+1
𝑥 − 𝑥0 𝑓 (𝑥) ≤ 𝜖 𝑤𝑕𝑒𝑟𝑒 𝜖 > 0
𝑛+1 !
1 𝑛+1
⟹ 𝑥 − 𝑥0 𝑀𝑛+1 ≤ 𝜖 𝑤𝑕𝑒𝑟𝑒 𝜖 > 0
𝑛+1 !
where 𝜖 is the errorterm.

Example 3.1.1:

Obtain the Taylor Series approximation about 𝑥 = 1 up to second

1
degree terms for the function 𝑓 𝑥 = . Find a bound on the error if
1+𝑥 2

this approximation is to be used in [1, 1.4].

Solution:
1
Given 𝑓 𝑥 =
1+𝑥 2

1 + 𝑥 2 0 − 1. (2𝑥)
𝑓′ 𝑥 =
(1 + 𝑥 2 )2
−2𝑥
=
(1 + 𝑥 2 )2

186
 1 + 𝑥2 2
−2 − −2𝑥 2 1 + 𝑥 2 2𝑥
𝑓 ′′ 𝑥 = SPACE OF HINTS
(1 + 𝑥 2 )4

1 + 𝑥 2 [ −2 1 + 𝑥 2 + 8𝑥 2 ]
=
(1 + 𝑥 2 )4

′′′
1 + 𝑥 2 3 2 6𝑥 − 2 3𝑥 2 − 1 3(1 + 𝑥 2 )2 2𝑥
𝑓 𝑥 =
(1 + 𝑥 2 )6

2 2
12𝑥 1 + 𝑥 2 − 12𝑥 3𝑥 2 − 1
= 1+𝑥
1 + 𝑥2 6

12𝑥 + 12𝑥 3 − 36𝑥 3 + 12𝑥

=
(1 + 𝑥 2 )4

24𝑥 − 24𝑥 3
=
(1 + 𝑥 2 )4

24𝑥(1 − 𝑥 2 )
=
(1 + 𝑥 2 )4

To find at 𝑥 = 1,
1
𝑓 1 =
2
1
𝑓′ 1 = −
2
2(3 − 1) 1
𝑓 ′′ 1 = =
23 2
24 0
𝑓 ′′′ 1 = =0
24
Hence, the Taylor series approximation is given by
2
′
𝑥 − 𝑥0
𝑓 𝑥 = 𝑓 𝑥0 + 𝑥 − 𝑥0 𝑓 𝑥0 + 𝑓 ′′ 𝑥0 + ⋯
2!
2
𝑥−1 ′
𝑓 𝑥 =𝑓 1 + 𝑥−1 𝑓 1 + 𝑓 ′′ (1)
2!
2
1 1 𝑥−1 1
= + 𝑥−1 − +
2 2 2! 2
2
1 1 𝑥−1
𝑓 𝑥 = − 𝑥−1 +
2 2 4

187
 The error bound is given by,
SPACE FOR HINTS (𝑥−1)3
𝑅2 ≤ 𝑀3 where 𝑀3 = max1≤𝑥≤1.4 𝑓 ′′′ (𝑥)
3!

Now,𝑀3 = max 𝑓 ′′′ (𝑥)

1≤𝑥≤1.4

24 1.4 [1 − 1.4 2 ]
=
[1 + 1.4 2 ]4
24 1.4 (0.96)
=
(2.96)4
32.256
=
76.7656
= 0.42019

(𝑥 − 1)3
∴ 𝑅3 ≤ 𝑀3
3!
3
1.4 − 1
= 0.42019
3!
= 0.00448

∴The maximum absolute error occurs at 𝑥 = 1.4 and is given by

0.00448.

Example 3.1.2:
Obtain polynomial approximation 𝑃(𝑥) to 𝑓 𝑥 = 𝑒 −𝑥 using the
Taylors expansion about 𝑥0 = 0 and determine

i) 𝑥 when the error in 𝑃(𝑥) obtained from the first four terms
only is to be less then 10−6 after rounding.
ii) The number of terms in the approximation to find results
correct to 10−10 for 0 ≤ 𝑥 ≤ 1.
Solution:
Given 𝑓 𝑥 = 𝑒 −𝑥
𝑓 ′ 𝑥 = −𝑒 −𝑥
𝑓 ′′ 𝑥 = (−1)2 𝑒 −𝑥
𝑟
Hence integral 𝑓 𝑥 = (−1)𝑟 𝑒 −𝑥

188
 𝑟
∴At 𝑥0 = 0 , 𝑓 0 = (−1)𝑟 𝑒 −0 = (−1)𝑟 , 𝑟 = 0,1,2, ⋯
SPACE OF HINTS
′ 𝑥−𝑥 0 2 ′′
∴The polynomial 𝑃 𝑥 = 𝑓 𝑥0 + 𝑥 − 𝑥0 𝑓 𝑥0 + 𝑓 𝑥0 +
2!

⋯
𝑥2 𝑥3
= 1 + 𝑥 −1 + 1 + −1 + ⋯
2! 3!
𝑥2 𝑥3
𝑃 𝑥 =1−𝑥+ − +⋯
2 6
1 𝑛+1
The error term is 𝑥 − 𝑥0 𝑀𝑛+1 ≤ 𝜖
𝑛+1 !

1
𝑥 − 0 4 𝑀4 ≤ 𝜖
4!
1 4
𝑥 𝑀4 ≤ 𝜖
24
Let us consider 𝜖 = 5 × 10−7
𝑥 4 𝑀4 ≤ 5 × 10−7 × 24 (1)
Where 𝑀4 = max0≤𝑥≤1 𝑓 4 (𝑥)

= max 𝑒 −𝑥 = 1
0≤𝑥≤1

∴ 𝑀4 = 1

1 ⟹ 𝑥 4 . 1 ≤ 5 × 10−7 × 24 = 120 × 10−7

𝑥 ≤ 0.05886
Now, the number of terms in the approximation is given by,
1
< 5 × 10−11
𝑛+1 !
1
< 5 × 10−11
𝑛 + 1 𝑛!
1
< 𝑛 + 1 𝑛!
5 × 10−11
⟹ 0.2 × 1011 < 𝑛 + 1 !

⟹ 20000000000 < 𝑛 + 1 !

∴ 𝑛 ≥ 14.

3.2: Lagrange and Newton interpolations

189
 Lagrange‟s interpolation formula is,
SPACE FOR HINTS 𝑥 − 𝑥1 𝑥 − 𝑥0
𝑃 𝑥 = 𝑓 𝑥0 + 𝑓(𝑥1 )
𝑥0 − 𝑥1 𝑥1 − 𝑥0
𝑥 𝑥
𝑃 𝑥 = 𝐼0 𝑓 𝑥0 + 𝐼1 𝑓 𝑥1 (1)
𝑥 𝑥−𝑥 1 𝑥 𝑥−𝑥 0
where 𝑙0 = , 𝑙1 =
𝑥 0 −𝑥 1 𝑥 1 −𝑥 0

The functions 𝑙0 (𝑥) and 𝑙1 (𝑥) are called the Lagrange‟s fundamental
polynomials andit can be verified that satisfying conditions

𝑙0 𝑥 + 𝑙1 𝑥 = 1

𝑙0 𝑥0 = 1, 𝑙0 𝑥1 = 0

𝑙1 (𝑥0 ) = 0, 𝑙1 (𝑥1 ) = 1
1 𝑖𝑓 𝑖 = 𝑗
(i.e.)𝑙𝑖 (𝑥𝑗 ) = 𝛿𝑖𝑗 =
0 𝑖𝑓 𝑖 ≠ 𝑗

The degree of the polynomials 𝑙0 (𝑥) and 𝑙1 (𝑥) is one.

∴ (1) is the linear Lagrange‟s interpolating polynomial.

Iterated linear interpolation:

1
1 can be written as 𝑃 𝑋 = 𝑥1 − 𝑥 𝑓 𝑥0 − 𝑥0 − 𝑥 𝑓 𝑥1
𝑥 1 −𝑥 0

1 𝑓0 (𝑥0 ) 𝑥0 − 𝑥
=
𝑥1 − 𝑥0 𝑓1 (𝑥1 ) 𝑥1 − 𝑥
1 𝐼0 (𝑥) 𝑥0 − 𝑥
𝑃 𝑋 = (2)
𝑥1 − 𝑥0 𝐼1 (𝑥) 𝑥1 − 𝑥

Where 𝐼0 𝑥 = 𝑓(𝑥0 ) and 𝐼1 𝑥 = 𝑓 𝑥1 = 𝐼0,1 (𝑥)

∴ (2) is called Aitkin’s or iterated linear interpolating polynomial.
Newton’s Divided Difference Interpolation:
Newton‟s divided difference formula is
𝑓 𝑥1 − 𝑓 𝑥0
𝑃 𝑥 = 𝑓 𝑥0 + 𝑥 − 𝑥0
𝑥1 − 𝑥0

= 𝑓 𝑥0 + 𝑥 − 𝑥0 𝑓 𝑥0 , 𝑥1
𝑓 𝑥 1 −𝑓 𝑥 0
where 𝑓 𝑥0 , 𝑥1 =
𝑥 1 −𝑥 0

190
Here the ratio 𝑓 𝑥0 , 𝑥1 is called the first divided difference of 𝑓(𝑥)
SPACE OF HINTS
relative to 𝑥0 and 𝑥1 .

Example 3.2.1:
Given 𝑓 2 = 4 , 𝑓 2.5 = 5.5 find the linear interpolating
polynomial using

i) Lagrange interpolation
ii) Aitkin‟s interpolation
iii) Newton‟s divided difference interpolation, hence find an
approximation value of 𝑓(2.2).
Solution:
i) Given 𝑥0 = 2 𝑥1 = 2.5
𝑓 𝑥0 = 4 𝑎𝑛𝑑 𝑓 𝑥1 = 5.5
The formula for Lagrange‟s interpolation is
𝑥 − 𝑥1 𝑥 − 𝑥0
𝑃 𝑥 = 𝑓 𝑥0 + 𝑓(𝑥1 )
𝑥0 − 𝑥1 𝑥1 − 𝑥0
𝑥 − 2.5 𝑥 − 2 (5.5)
= 4 +
2 − 2.5 0.5
= −8 𝑥 − 2.5 + 11(𝑥 − 2)
= −8𝑥 + 20 + 11𝑥 − 22
𝑃 𝑥 = 3𝑥 − 2
Now 𝑃 𝑥 = 𝑓 𝑥 = 3𝑥 − 2
𝑓 2.2 = 3 2.2 − 2 = 4.6.
ii) It can iterated linear interpolating polynomial is
1
𝑃 𝑋 = 𝑥 − 𝑥 𝑓 𝑥0 − 𝑥0 − 𝑥 𝑓 𝑥1
𝑥1 − 𝑥0 1
1 𝐼0 (𝑥0 ) 𝑥0 − 𝑥
=
𝑥1 − 𝑥0 𝐼1 (𝑥1 ) 𝑥1 − 𝑥
1 𝑓(𝑥0 ) 𝑥0 − 𝑥
=
2.5 − 2 𝑓(𝑥1 ) 𝑥1 − 𝑥

191
 1 4 2−𝑥
=
SPACE FOR HINTS 0.5 5.5 2.5 − 𝑥
= 3𝑥 − 2
∴ 𝑃 𝑥 = 3𝑥 − 2
𝑃 𝑥 =𝑓 𝑥
𝑓 2.2 = 3 2.2 − 2 = 4.6
iii) Newton‟s divided difference interpolating polynomial is
𝑓 𝑥1 − 𝑓 𝑥0
𝑃 𝑥 = 𝑓 𝑥0 + 𝑥 − 𝑥0
𝑥1 − 𝑥0
5.5 − 4
=4+ 𝑥−2
2.5 − 2
1.5
=4+ 𝑥−2
0.5
= 4 + 3 𝑥 − 2 = 4 + 3𝑥 − 6
𝑃 𝑥 = 3𝑥 − 2
𝑝 𝑥 = 𝑓(𝑥)
𝑓 2.2 = 3 2.2 = 3 2.2 − 2 = 4.6.
Truncation error bounds:
The polynomial 𝑃 𝑥 coincides with the function 𝑓 𝑥 at 𝑥0 and 𝑥1 and
it derivatives at all other points, in the interval 𝑥0 , 𝑥1 . This derivation
is called truncation error.

The truncation error is given by,

𝐸1 𝑓; 𝑥 = 𝑓 𝑥 − 𝑃(𝑥)

The truncation error in linear interpolation is given by,

1
𝐸1 𝑓; 𝑥 = 𝑥 − 𝑥0 𝑥 − 𝑥1 𝑓 ′′ (𝜉)
2
where 𝑓 ′′ (𝑥) ≤ 𝑀2 , 𝑥 ∈ 𝑥0 , 𝑥1

Then step size is given by,

𝑕2
max 𝑓 ′′ (𝑥) ≤ 𝜖 𝑤𝑕𝑒𝑟𝑒 𝑕 = 𝑥1 − 𝑥0
8 𝑎≤𝑥≤𝑏
Example 3.2.2:

192
 Using the data sin 0.1 = 0.09983 and sin 0.2 = 0.19867, find an
SPACE OF HINTS
(0.15) by Lagrange‟s interpolation. Obtain a
approximate value of sin⁡
bound on the truncation error.

Solution:
Given sin 0.1 = 0.09983

sin 0.2 = 0.19867

(i.e.) 𝑥0 = 0.1 , 𝑥1 = 0.2

𝑓 𝑥0 = 0.09983

𝑓 𝑥1 = 0.1986
The formula for Lagrange‟s interpolation is,
𝑥 − 𝑥1 𝑥 − 𝑥0
𝑓 𝑥 = 𝑓 𝑥0 + 𝑓(𝑥1 )
𝑥0 − 𝑥1 𝑥1 − 𝑥0
(𝑥 − 0.20) 𝑥 − 0.1
= 0.09983 + (0.19867)
0.1 − 0.2 0.2 − 0.1
= −0.9983 𝑥 − 0.2 + 1.9867(𝑥 − 0.1)

= −0.9983𝑥 + 0.19966 + 1.9867𝑥 − 0.19867

𝑓 𝑥 = 0.9884𝑥 + 0.00099

Put 𝑥 = 0.15 𝑓 0.15 = 0.14826 + 0.00099 = 0.14925

1
The truncation error is 𝐸1 𝑓; 𝑥 = 𝑥 − 𝑥0 𝑥 − 𝑥1 𝑓 ′′ (𝜉)
2

Given 𝑓 𝑥 = sin 𝑥

𝑓 ′ 𝑥 = cos 𝑥

𝑓 ′′ 𝑥 = − sin 𝑥

∴ 𝑓 ′′ 𝜉 = − sin 𝜉, where 𝜉 ∈ 0.1, 0.2

Now, max0.1≤𝜖≤0.2 − sin 𝜉 = max − sin 0,1 − sin⁡

(0.2)

= 𝑚𝑎𝑥 0.09983 , 019867

= 0.19867

193
 1
∴ 𝐸1 𝑓, 𝑥 = 0.15 − 0.1 0.15 − 0.2 (0.19867)
SPACE FOR HINTS 2
1
= 0.05 −0.05 (0.19867)
2
⇒ 𝐸1 𝑓, 𝑥 ≤ −0.000248 = 0.00025

Hence the maximum absolute error is 0.00025.

Example 3.2.3:
Determine the step size 𝑕 that can be used in the tabulation of
𝑓 𝑥 = sin 𝑥 in theinterval [1, 3] so that the linear interpolation will be
correct to four decimal places after rounding.

Solution:
The formula for step size 𝑕 is given by,

𝑕2
max 𝑓 ′′ (𝑥) ≤ 𝜖
8 𝑎≤𝑥≤𝑏
Now, max 𝑓 ′′ (𝑥) = max − sin 𝑥
𝑎≤𝑥≤𝑏 1≤𝑥≤3

= max − sin 1 − sin 2 − sin 3

= max −0.8415 − 0.90929 − 0.1411

= 0.9095 ≅ 1

𝑕2
. 1 ≤ 5 × 10−5
8
𝑕2 ≤ 40 × 10−5 ⇒ 𝑕2 ≤ 0.02.

Example 3.2.4:
The function 𝑓 𝑥 = sin 𝑥 is defined on the interval [1, 3]

i) Obtain the Lagrange linear interpolating polynomial in this

interval and find the bound on the truncation error. Obtain the
approximation value of 𝑓(1.5) and 𝑓(2.5)
ii) Divide the interval [1, 3] into two sub intervals [1, 2] and
2, 3 . Obtain the Lagrange linear interpolating polynomial in
each subinterval and find the bound on the truncation error.

194
 Hence find the approximate value of 𝑓(1.5) and 𝑓 2.5 .
SPACE OF HINTS
Compare with the exact values.

Solution:
i) Given 𝑓 𝑥 = sin 𝑥
The interval is [1, 3] , i.e.)𝑥0 = 1 , 𝑥1 = 3
∴ 𝑓 𝑥0 = 0.8415 𝑓 𝑥1 = 0.14112
The formula for Lagrange‟s linear interpolation is,
𝑥 − 𝑥1 𝑥 − 𝑥0
𝑓 𝑥 = 𝑓 𝑥0 + 𝑓(𝑥1 )
𝑥0 − 𝑥1 𝑥1 − 𝑥0
𝑥−3 𝑥−1
= 0.8415 + (0.14112)
1−3 3−1
𝑥−3 𝑥−1
= 0.8415 + (0.14112)
−2 2
= −0.4208𝑥 + 1.2623 + 0.0706𝑥 − 0.0706
𝑓 𝑥 = −0.3502𝑥 + 1.1917
∴ 𝑓 1.5 = 0.6664 𝑓 2.5 = 0.3162
Lagrange‟s truncation error bound is given by,
1
𝐸1 𝑓, 𝑥 = 𝑥 − 𝑥0 𝑥 − 𝑥1 𝑓 ′′ (𝜉)
2
1
𝐸1 𝑓, 𝑥 ≤ (𝑥 − 𝑥0 ) 𝑥 − 𝑥1 𝑓 ′′ (𝜉)
2
Now,𝑓 𝑥 = sin 𝑥
𝑓 ′ 𝑥 = cos 𝑥
𝑓 ′′ 𝑥 = − sin 𝑥
𝑓 ′′ (𝜖) = − sin 𝜖
max − sin 𝜖 = max 0.8415 , 0.9093 ,0.1411
1≤𝑥≤3

= 0.9093
Now, the truncation error bound at 𝑥 = (1.5) is
1
𝐸1 𝑓, 𝑥 ≤ 1.5 − 1 1.5 − 3 (0.9093)
2

195
 1
≤ 0.5 −1.5 (0.9093)
SPACE FOR HINTS 2
≤ 0.34098
The truncation error bounded at 𝑥 = 2.5 is
1
𝐸1 𝑓, 𝑥 ≤ 2.5 − 1 2.5 − 3 (0.9093)
2
≤ 0.34098
ii) Divide the given interval into two intervals [1, 2] and [2, 3]
Case i: In the interval [1, 2]
𝑥0 = 1, 𝑥1 = 2
𝑓 𝑥 = sin 𝑥
𝑓 1 = 0.8415
𝑓 2 = 0.9093
The formula for Lagrange‟s linear interpolation is
𝑥 − 𝑥1 𝑥 − 𝑥0
𝑓 𝑥 = 𝑓 𝑥0 + 𝑓(𝑥1 )
𝑥0 − 𝑥1 𝑥1 − 𝑥0
𝑥−2 𝑥−1
= (0.8415) + (0.9093)
1−2 2−1
= − 𝑥 − 2 0.8415 + 𝑥 − 1 0.9093

= −0.8415𝑥 + 1.683 + 0.9093

= 0.0678𝑥 + 0.7737

∴ 𝑓 1.5 = 0.8754

The truncation error bound is,

1
𝐸1 𝑓, 𝑥 ≤ (𝑥 − 𝑥0 ) 𝑥 − 𝑥1 𝑓 ′′ (𝜉)
2

Where 𝑓 ′′ (𝜉) = − sin 𝜉

max 𝑓 ′′ (𝜉) = max −0.8415 , −0.9093 , −0.9975

1≤𝑥≤2

= 0.9975
1
∴ 𝐸1 [ 𝑓, 𝑥] ≤ 1.5 − 1 (1.5 − 2) 0.9975
2

196
 = 0.1247
SPACE OF HINTS
Case ii: In the interval [2, 3]

𝑥0 = 2 , 𝑥1 = 3

𝑓 𝑥 = sin 𝑥

𝑓 2 = 0.9093

𝑓 3 = 0.1411

The formula for Lagrange‟s linear interpolation is,

𝑥 − 𝑥1 𝑥 − 𝑥0
𝑓 𝑥 = 𝑓 𝑥0 + 𝑓(𝑥1 )
𝑥0 − 𝑥1 𝑥1 − 𝑥0
𝑥−3 𝑥−2
= 0.9093 + (0.1411)
2−3 3−2
= −0.9093𝑥 + 2.7279 + 0.1411𝑥 − 0.2822

𝑓 𝑥 = −0.7682𝑥 + 2.4457

𝑓 2.5 = 0.5252

The truncation error bound is,

1
𝐸1 𝑓, 𝑥 ≤ (𝑥 − 𝑥0 ) 𝑥 − 𝑥1 𝑓 ′′ (𝜉)
2

Where 𝑓 ′′ 𝜉 = − sin 𝜉

max 𝑓 ′′ (𝜉) = max −0.9093 , −0.5985 , 0.1411

2≤𝑥≤3

= 0.9093
1
𝐸1 𝑓, 𝑥 ≤ 2.5 − 2 (2.5 − 3) 0.9093
2
= 0.1137

Exact Value:Given 𝑓 𝑥 = sin 𝑥

𝑓 1.5 = sin 1.5 = 0.9975

𝑓 2.5 = 0.5985

In case ii) 𝑓 1.5 = 0.8754 and 𝑓 2.5 = 0.5252 is nearer to the exact
value.

197
SPACE FOR HINTS

Quadratic interpolation:
The quadratic interpolation polynomial in Lagrange‟s Method is,
𝑥 − 𝑥1 𝑥 − 𝑥2 𝑥 − 𝑥0 (𝑥 − 𝑥2 )
𝑓 𝑥 = 𝑓 𝑥0 + 𝑓 𝑥1
𝑥0 − 𝑥1 𝑥0 − 𝑥2 𝑥1 − 𝑥0 𝑥1 − 𝑥2
𝑥 − 𝑥0 𝑥 − 𝑥1
+ 𝑓(𝑥2 )
𝑥2 − 𝑥0 𝑥2 − 𝑥0

= 𝑙0 𝑥 𝑓 𝑥0 + 𝑙1 𝑥 𝑓 𝑥1 + 𝑙2 𝑥 𝑓(𝑥2 )
The truncation error in the Lagrange‟s quadratic interpolation is
𝑥 − 𝑥0 𝑥 − 𝑥1 𝑥 − 𝑥2 ′′′
𝐸2 𝑓, 𝑥 = 𝑓 (𝜉)
6
where 𝑓 ′′′ 𝜉 = max𝑥 0 ≤𝑥≤𝑥 2 𝑓 ′′′ (𝑥)

Higher order interpolation:

The Lagrange fundamental polynomial of degree 𝑛 based on 𝑛 + 1
distinct points𝑎 ≤ 𝑥0 < 𝑥1 < 𝑥2 < ⋯ < 𝑥𝑛 ≤ 𝑏 and can be written in
the form
𝑥 − 𝑥0 𝑥 − 𝑥1 ⋯ 𝑥 − 𝑥𝑖−1 𝑥 − 𝑥𝑖+1 ⋯ 𝑥 − 𝑥𝑛
𝑙𝑖 (𝑥) = ,
𝑥𝑖 − 𝑥0 𝑥𝑖 − 𝑥1 ⋯ 𝑥𝑖 − 𝑥𝑖−1 𝑥𝑖 − 𝑥𝑖+1 ⋯ 𝑥𝑖 − 𝑥𝑛

𝑖 = 0,1,2, ⋯ 𝑛
𝑛

∴𝑓 𝑥 = 𝑙𝑖 𝑥 𝑓(𝑥𝑖 )
𝑖=1

The above equation is the Lagrange‟s interpolating polynomial of

degree 𝑛.

The truncation error in Lagrange‟s interpolation in given by,

1
𝐸𝑛 𝑓, 𝑥 = 𝑥 − 𝑥0 𝑥 − 𝑥1 ⋯ 𝑥 − 𝑥𝑛 𝑓 (𝑛+1) (𝜉)
𝑛+1 !

198
 𝑛+1
where 𝑓 𝜉 = max𝑥 0 ≤𝑥≤𝑥 2 𝑓 (𝑛+1) (𝑥) = 𝑀𝑛+1
SPACE OF HINTS
Example 3.2.5:Given that 𝑓 0 = 1 , 𝑓 1 = 3 , 𝑓 3 = 55 find the
unique polynomial of degree 2 or less which fits the given data.

Find the bound on the error.

Solution:
Given 𝑥0 = 0 , 𝑥1 = 1 , 𝑥2 = 3

𝑓 𝑥0 = 1 𝑓 𝑥1 = 3 𝑓 𝑥2 = 55

The formula for Lagrange‟s quadratic interpolation is

𝑥 − 𝑥1 𝑥 − 𝑥2 𝑥 − 𝑥0 (𝑥 − 𝑥2 )
𝑓 𝑥 = 𝑓 𝑥0 + 𝑓 𝑥1
𝑥0 − 𝑥1 𝑥0 − 𝑥2 𝑥1 − 𝑥0 𝑥1 − 𝑥2
𝑥 − 𝑥0 𝑥 − 𝑥1
+ 𝑓(𝑥2 )
𝑥2 − 𝑥0 𝑥2 − 𝑥0

𝑥−1 𝑥−3 𝑥−0 𝑥−3

= 1 + 3
0−1 0−3 1−0 1−3
𝑥−0 𝑥−1
+ (55)
3−0 3−1

𝑥 − 1 (𝑥 − 3) 𝑥 𝑥 − 3 𝑥 𝑥−1
= + 3 + 55
3 −2 6
𝑥 2 − 4𝑥 + 3 3𝑥 2 − 9𝑥 55𝑥 2 − 55𝑥
= + +
3 −2 6
1
= 2𝑥 2 − 8𝑥 + 6 − 9𝑥 2 + 27𝑥 + 55𝑥 2 − 55𝑥
6
1
= 48𝑥 2 − 36𝑥 + 6 = 8𝑥 2 − 6𝑥 + 1
6
The quadratic error formula is,
𝑥−𝑥 0 𝑥−𝑥 1 𝑥−𝑥 2
𝐸2 𝑓, 𝑥 ≤ 𝑓 ′′′ (𝜉) where 𝑀3 = max0≤𝑥≤3 𝑓 ′′′ (𝜉)
6

1
= 𝑀 [ max 𝑥 𝑥 − 1 𝑥 − 3 ]
6 3 0≤𝑥≤3

199
 1
= 2.1126 𝑀3 (since the maximum of 𝑥 𝑥 − 1 𝑥 − 3 occurs at
6
SPACE FOR HINTS
𝑥 = 2.2152

= 0.3521 𝑀3

Example 3.2.6
The following values of the function 𝑓 𝑥 = sin 𝑥 + cos 𝑥, are given

𝑥 10° 20° 30°

𝑓(𝑥) 1.1585 1.2817 1.3660

Construct a quadratic interpolating polynomial that fits the data. Hence

𝜋
find 𝑓 . Compare with the exact value.
2

Solution:

Given 𝑥0 = 10° , 𝑥1 = 20° , 𝑥2 = 30°

𝑓 𝑥0 = 1.1585 𝑓 𝑥1 = 1.2817 𝑓 𝑥2 = 1.3660

𝜋
The finite value of 𝑥 = is a radian.
12

∴We convert the given data into radian.

𝜋 𝜋
(i.e.)𝑥0 = 10 × = = 0.1745 𝑓 𝑥0 = 1.1585
180 18

𝜋 𝜋
𝑥1 = 20 × = = 0.3491; 𝑓 𝑥1 = 1.2817
180 9
𝜋 𝜋
𝑥2 = 30 × = = 0.5236 ; 𝑓 𝑥2 = 1.3660.
180 6

∴The Lagrange‟s quadratic interpolating polynomial is

𝑥 − 𝑥1 𝑥 − 𝑥2 𝑥 − 𝑥0 (𝑥 − 𝑥2 )
𝑓 𝑥 = 𝑓 𝑥0 + 𝑓 𝑥1
𝑥0 − 𝑥1 𝑥0 − 𝑥2 𝑥1 − 𝑥0 𝑥1 − 𝑥2
𝑥 − 𝑥0 𝑥 − 𝑥1
+ 𝑓(𝑥2 )
𝑥2 − 𝑥0 𝑥2 − 𝑥0

200
 𝑥 − 0.3491 𝑥 − 0.5236
= 1.5885 SPACE OF HINTS
0.1745 − 0.3491 0.1745 − 0.5236
𝑥 − 0.1745 𝑥 − 0.5236
+ 1.2817
0.3491 − 0.1745)(0.3491 − 0.5236
𝑥 − 0.1745 𝑥 − 0.3491
+ (1.3660)
0.5236 − 0.1745 0.5236 − 0.1745

𝑥 − 0.3491 𝑥 − 0.5236
= 1.5885
0.06096
𝑥 − 0.1745 𝑥 − 0.5236
+ 1.2817
−0.03048
𝑥 − 0.1745 𝑥 − 0.3491
+ (1.3660)
0.06093

= 19.0043 𝑥 2 − 0.8727 𝑥 + 0.1828

− 42.0505 𝑥 2 − 0.6981 𝑥 + 0.0914
+ 22.4192(𝑥 2 − 0.5236 𝑥 + 0.0609)

= 𝑥 2 19.0043 − 42.0505 + 22.4192

+ 𝑥 −16.5850 + 29.3555 − 11.7387
+ 3.4739 − 3.8434 + 1.3653

= 𝑥 2 −0.627 + 1.0390𝑥 + 0.9955

𝜋
𝑥= = 0.2618
12
𝜋
𝑓 = 1.2245
12
Given 𝑓 𝑥 = sin 𝑥 + cos 𝑥
𝜋 𝜋 𝜋
∴𝑓 = sin + 𝑐𝑜𝑠 = sin 0.2618 + cos 0.2618
12 12 12
= 1.2247
𝜋
Hence the exact value 𝑓 = 1.2247 and approximate value is
12
𝜋
𝑓 = 1.2245.
12

Iterated interpolation:

201
 The iterated interpolation table is,
SPACE FOR HINTS

𝑥0 − 𝑥 = 𝐼0 (𝑥)

𝑥0 𝐼0,1 (𝑥)

𝑥1 − 𝑥 = 𝐼1 𝑥 = 𝑥 𝐼0,1,2 (𝑥)

𝑥1 𝐼0,2 (𝑥)
𝑥2 − 𝑥1 = 𝐼2 (𝑥)

𝑥2

⋮ 𝐼0,1,𝑛−1 (𝑥)

⋮ ⋮ 𝐼0, 𝑛−1 (𝑥) ⋯ 𝐼0,1,2,⋯,𝑛 (𝑥)

𝑥𝑛−1 − 𝑥 = 𝐼𝑛−1 (𝑥) 𝐼0,1,𝑛 (𝑥)

𝑥𝑛−1 𝐼0,𝑛 (𝑥)
𝑥𝑛 − 𝑥 = 𝐼𝑛 (𝑥)

𝑥𝑛

1 𝐼0,1,⋯,𝑛−1 (𝑥) 𝑥𝑛−1 − 𝑥

where𝐼0,1,2,⋯,𝑛 𝑥 =
𝑥 𝑛 −𝑥 𝑛 −1 𝐼0,1,⋯,𝑛 (𝑥) 𝑥𝑛 − 𝑥

Newton’s Divided Difference Interpolation:

Newton‟s divided difference interpolating polynomial is given by

𝑃𝑛 𝑥 = 𝑓 𝑥0
+ 𝑥 − 𝑥0 𝑓 𝑥0 , 𝑥1 ⋯ 𝑥 − 𝑥0 𝑥 − 𝑥1 ⋯ 𝑥
− 𝑥𝑛 −1 𝑓 𝑥0 , 𝑥1 , ⋯ , 𝑥𝑛

Example 3.2.7:
Find the unique polynomial of degree 2 or less such that 𝑓 0 = 1,
𝑓 1 = 3 , 𝑓 3 = 55, using i) The iterated interpolation

202
ii) The Newton divided difference interpolation.
SPACE OF HINTS
Solution:
Given 𝑥0 = 0 , 𝑥1 = 1 , 𝑥2 = 3

𝑓 𝑥0 = 1 , 𝑓 𝑥1 = 3 , 𝑓 𝑥2 = 55
i) The formula for iterated interpolation is
1 𝑓(𝑥0 ) 𝑥0 − 𝑥
𝐼0,1 𝑥 =
𝑥1 − 𝑥0 𝑓(𝑥1 ) 𝑥1 − 𝑥
1 1 0−𝑥
=
1−0 3 1−𝑥
= 1 − 𝑥 + 3𝑥 = 1 + 2𝑥
1 𝑓(𝑥0 ) 𝑥0 − 𝑥
𝐼0,2 𝑥 =
𝑥2 − 𝑥0 𝑓(𝑥2 ) 𝑥2 − 𝑥
1 1 0−𝑥
=
3 55 3 − 𝑥
1 1
= 3 − 𝑥 + 55𝑥 = (3 + 54𝑥)
3 3
= 1 + 18𝑥
1 𝐼0,1 𝑥 𝑥1 − 𝑥
𝐼0,1,2 𝑥 =
𝑥2 − 𝑥1 𝐼0,2 𝑥 𝑥2 − 𝑥
1 1 + 2𝑥 1−𝑥
=
3 − 1 + 18𝑥
1 3−𝑥
1
= 1 + 2𝑥 3 − 𝑥 − 1 − 𝑥 (1 + 18𝑥)
2

1
= (3 − 𝑥 + 6𝑥 − 2𝑥 2 − 1 − 18𝑥 + 𝑥 + 18𝑥 2 )
2
1
= (16𝑥 2 − 12𝑥 + 2)
2
= 8𝑥 2 − 6𝑥 + 1
The formula for divided difference method is,

𝑃 𝑥 = 𝑓 𝑥0 + 𝑥 − 𝑥0 𝑓 𝑥0 , 𝑥1
+ 𝑥 − 𝑥0 𝑥 − 𝑥1 𝑓(𝑥0 , 𝑥1 , 𝑥2 )

and divided difference table is,

203
SPACE FOR HINTS

𝑥 𝑓(𝑥) First order divided Second order divide

Difference Difference

0 1
3−1
=2
1−0

1 3 26 − 2
55 − 3 =8
=8 3−0
3−1

3 55

Now 𝑃 𝑥 = 1 + 𝑥 − 0 2 + 𝑥 − 0 𝑥 − 1 8
= 1 + 2𝑥 + 8𝑥 2 − 8𝑥
= 8𝑥 2 − 6𝑥 + 1
Example 3.2.8:
1
Calculate the 𝑛𝑡𝑕 divided difference of , based on the points
𝑥

𝑥0 , 𝑥1 , ⋯ 𝑥𝑛 .

Solution:
1
Given 𝑓 𝑥 = based on the points 𝑥0 , 𝑥1 , ⋯ 𝑥𝑛 .
𝑥

The 1𝑠𝑡 order divided difference is

𝑓 𝑥1 − 𝑓(𝑥0 )
𝑓 𝑥0 , 𝑥1 =
𝑥1 − 𝑥0
1 1
− 𝑥0 − 𝑥1 1
𝑥1 𝑥0
= = ×
𝑥1 − 𝑥0 𝑥0 𝑥1 𝑥1 − 𝑥0
−(𝑥1 − 𝑥0 ) 1
= ×
𝑥0 𝑥1 𝑥1 − 𝑥0

204
 −1
𝑓 𝑥0 , 𝑥1 = SPACE OF HINTS
𝑥0 𝑥1
−1
Similarly, 𝑓 𝑥1 , 𝑥2 =
𝑥1 𝑥2

Now, the 2𝑛𝑑 order divided difference is,

𝑓 𝑥1 , 𝑥2 − 𝑓(𝑥0 , 𝑥1 )
𝑓 𝑥0 , 𝑥1 , 𝑥2 =
𝑥2 − 𝑥0
1 1 1 1
− + −
𝑥1 𝑥2 𝑥0 𝑥1 𝑥0𝑥1 𝑥1𝑥2
= =
𝑥2 − 𝑥0 𝑥2 − 𝑥0
𝑥2 − 𝑥0 1
= ×
𝑥0 𝑥1 𝑥2 𝑥2 − 𝑥0
(−1)2
𝑓 𝑥0 , 𝑥1 , 𝑥2 = and so on
𝑥0 𝑥1 𝑥2

The above result is true for 𝑥0 , 𝑥1 , 𝑥2 , ⋯ , 𝑥𝑘 .

(−1)𝑘
(i.e.) 𝑓 𝑥0 , 𝑥1 , 𝑥2 , ⋯ , 𝑥𝑘 =
𝑥 0 ,𝑥 1 ,𝑥 2 ,⋯,𝑥 𝑘

Now,
𝑓 𝑥0 , 𝑥1 , 𝑥2 , ⋯ , 𝑥𝑘+1 − 𝑓 𝑥0 , 𝑥1 , 𝑥2 , ⋯ , 𝑥𝑘
𝑓 𝑥0 , 𝑥1 , 𝑥2 , ⋯ , 𝑥𝑘+1 =
𝑥𝑘+1 − 𝑥0

(−1)𝑘+1 𝑥𝑘+1 − 𝑥0
=
𝑥0 , 𝑥1 , 𝑥2 , ⋯ , 𝑥𝑘+1 𝑥𝑘+1 − 𝑥0

(−1)𝑘+1
𝑓 𝑥0 , 𝑥1 , 𝑥2 , ⋯ , 𝑥𝑘+1 =
𝑥0 , 𝑥1 , 𝑥2 , ⋯ , 𝑥𝑘 +1

∴The above result has for 𝑘 + 1 terms.

Hence, by induction we obtain
(−1)𝑛
Then 𝑓 𝑥0 , 𝑥1 , 𝑥2 , ⋯ , 𝑥𝑛 = .
𝑥 0 , 𝑥 1 , 𝑥 2 ,⋯,𝑥 𝑛

Example 3.2.9:
Construct the divided difference table for the data
𝑥 0.5 1.5 3.0 5.0 6.5 8.0

𝑓(𝑥) 1.625 5.875 31.0 131.0 282.125 521.0

205
 Hence find the interpolating polynomial and an approximation to the
SPACE FOR HINTS value of 𝑓(7).
Solution:
Newton‟s divided difference table is,

𝑥 𝑓(𝑥) First order 2nd order divided 3rd order 4th order
divided difference divided divided
difference difference difference

0.5 1.625
5.875 − 1.625
1.5 − 0.5
1.5 5.875 = 4.25 16.75 − 4.25
3 − 0.5
=5 9.5 − 5
31 − 5.875 3 − 0.5
3.0 31.0 3 − 1.5 =1 1−1
= 16.75 =0
6.5 − 0.5
50 − 16.75
5 − 1.5 14.5 − 9.5
131 − 31 = 9.5 6.5 − 1.5
5.0 131.0 5−3 =1 1−1
= 50 100.75 − 50 =0
8 − 1.5
6.5 − 3
282.12 − 131 = 14.5
6.5 282.125 6.5 − 5 19.5 − 14.5
= 100.75 8−3
159.25 − 100.75 = 1
521 − 282.125 8−5
8.0 521.0 8 − 6.5 = 19.5
= 159.25

The formula for divided difference method is,

𝑃 𝑥 = 𝑓 𝑥0 + 𝑥 − 𝑥0 𝑓 𝑥0 , 𝑥1 + 𝑥 − 𝑥0 𝑥 − 𝑥1 𝑓 𝑥0 , 𝑥1 , 𝑥2

+ 𝑥 − 𝑥0 𝑥 − 𝑥1 𝑥 − 𝑥2 𝑓 𝑥0 , 𝑥1 , 𝑥2 , 𝑥3 + ⋯

= 1.625 + 𝑥 − 0.5 4.25 + 𝑥 − 0.5 𝑥 − 1.5 5

+ 𝑥 − 0.5 𝑥 − 1.5 (𝑥 − 3)

206
 = 1.625 + 4.25𝑥 − 2.125 + 5𝑥 2 − 10𝑥 + 3.75 + 𝑥 3 − 2𝑥 2
SPACE OF HINTS
2
+ 0.75𝑥 − 3𝑥 + 6𝑥 − 2.25

= 𝑥 3 + 𝑥 2 5 − 5 + 𝑥 4.25 − 10 + 0.75 + 6 + (1.625

− 2.125 + 3.75 − 2.25)

𝑃 𝑥 = 𝑥3 + 𝑥 + 1 ∴ 𝑓 𝑥 = 𝑥3 + 𝑥 + 1

Hence, 𝑓 7 = 351.

3.3: Finite difference Operator

Let the tabular points 𝑥0 , 𝑥1 , 𝑥2 , ⋯ , 𝑥𝑛 be equally spaced, then

𝑥𝑖 = 𝑥0 + 𝑖𝑕 , 𝑖 = 0,1,2, ⋯ 𝑛. Now, we define the following operators.

i) Shift operator: 𝐸𝑓 𝑥𝑖 = 𝑓(𝑥𝑖 + 𝑕)
ii) Forward difference operator: ∆𝑓 𝑥𝑖 = 𝑓 𝑥𝑖 + 𝑕 − 𝑓(𝑥𝑖 )
iii) Backward difference operator: ∆𝑓 𝑥𝑖 = 𝑓 𝑥𝑖 + 𝑕 − 𝑓(𝑥𝑖 )
𝑕 𝑕
iv) Centre difference operator: 𝛿𝑓 𝑥𝑖 = 𝑓 𝑥𝑖 + − 𝑓 𝑥𝑖 −
2 2
1 𝑕 𝑕
v) Average operator : 𝜇𝑓 𝑥𝑖 = 𝑓 𝑥𝑖 + + 𝑓 𝑥𝑖 −
2 2 2

Relationship between the above Operators:

𝐸 ∆ ∇ 𝛿
𝐸 𝐸 ∆+1 (1 − ∇)−1 1
1 + 𝛿2
2
1
+𝛿 1 + 𝛿2
∆ 𝐸−1 ∆ (1 − ∇)−1 − 1 4

1 2 1
∇ 1 − 𝐸 −1 1 − (1 + ∆)−1 ∇ 𝛿 +𝛿 1 + 𝛿2
2 4

1 1 1 1
𝛿 𝐸2 − 𝐸− 2 ∆(1 + ∆)−2 ∇(1 − ∇)− 2 1 1
− 𝛿2 + 𝛿 1 + 𝛿2
2 4
1 1
1 1 1 1+ ∆ 1− ∇
(𝐸 + 𝐸 −
) 2 2 1 1
𝜇 2 2 1 1
− 𝛿2 + 𝛿 1 + 𝛿2
2 (1 + ∆)2 (1 − ∇)− 2 2 4

207
 1
SPACE FOR HINTS 1 + 𝛿2
4

1
∆𝑓𝑖 = ∇𝑓𝑖+1 = 𝛿𝑓𝑖 +
2
∆= 𝐸 − 1 ⇒ 𝐸 = ∆ + 1
∇= 1 − 𝐸 −1 ⇒ 𝐸 −1 = 1 − ∇
1 −1
𝐸= = 1−∇
1−∇
1 1
𝛿 = 𝐸 2 − 𝐸 −2
1 1 1
𝜇= 𝐸 2 + 𝐸 −2
2
Example 3.3.1: Show that
1
i) 𝛿 = ∇(1 − ∇)− 2
𝛿2 1
ii) 𝜇 = (1 + )2
4

Solution:
1 1
i) Consider ∇(1 − ∇)−2 = 1 − 𝐸 −1 (1 − (1 − 𝐸 −1 ))−2
1
= 1 − 𝐸 −1 (1 − 1 − 𝐸 −1 )−2
1
= 1 − 𝐸 −1 (𝐸 2 )
1 1
= 𝐸 2 − 𝐸 −2 = 𝛿
1
∴ ∇(1 − ∇)−2 = 𝛿
1 1
ii) 𝛿 = 𝐸 2 − 𝐸 −2
1 1
𝛿 2 = (𝐸 2 − 𝐸 −2 )2 = 𝐸 + 𝐸 −1 − 2
𝛿2 𝐸+𝐸 −1 −2
Now =
4 4

𝛿2 𝐸 + 𝐸 −1 − 2 4 + 𝐸 + 𝐸 −1 − 2
1+ =1+ =
4 4 4
𝐸 + 𝐸 −1 + 2
=
4

208
 1
1 1 1 1
2 2 − 2 SPACE OF HINTS
𝛿 𝐸 + 𝐸 −1 + 2 2 (𝐸 + 𝐸 )
2 2 2
⟹ 1+ = =
4 4 22
1 1
𝐸 2 + 𝐸 −2
=
2
1
𝛿2 2 1 1 1
∴ 1+ = 𝐸 2 + 𝐸 −2 = 𝜇.
4 2
3.4: Interpolating Polynomials Using Finite Difference:
Gregory Newton Forward Difference Interpolation:
𝑥 − 𝑥0 𝑥 − 𝑥0 𝑥 − 𝑥1 2
𝑃 𝑥 = 𝑓 𝑥0 + ∆𝑓 𝑥0 + ∆ 𝑓 𝑥0
1! 𝑕 2! 𝑕2
𝑥 − 𝑥0 𝑥 − 𝑥1 𝑥 − 𝑥2 ⋯ 𝑥 − 𝑥𝑛−1 𝑛
+ ∆ 𝑓(𝑥0 )
𝑛! 𝑕𝑛
This is called Gregory-Newton forward difference interpolating
polynomial.
𝑥−𝑥 0
Put 𝑢 =
𝑕

𝑢 𝑢 𝑢−1 2
𝑃 𝑥0 + 𝑕𝑢 = 𝑓 𝑥0 + ∆𝑓 𝑥0 + ∆ 𝑓 𝑥0 + ⋯
1! 2!
𝑢 𝑢 − 1 ⋯ (𝑢 − 𝑛 − 1) 𝑛
+ ∆ 𝑓 𝑥0
𝑛!
𝑢 𝑢 −1 ⋯ 𝑢−𝑛
An error term is 𝐸𝑛 𝑓; 𝑥 = 𝑕𝑛+1 𝑓 𝑛+1
(𝜉)
𝑛+1 !

Gregory-Newton Backward Difference Interpolation:

𝑢 𝑢 𝑢+1 2
𝑃 𝑥𝑛 + 𝑕𝑢 = 𝑓 𝑥𝑛 + ∇𝑓 𝑥𝑛 + ∇ 𝑓 𝑥𝑛 + ⋯
1! 2!
𝑢 𝑢 + 1 ⋯ (𝑢 + 𝑛 − 1) 𝑛
+ ∇ 𝑓 𝑥𝑛
𝑛!
𝑢 𝑢+1 ⋯ 𝑢+𝑛
The error term is, 𝐸𝑛 𝑓; 𝑥 = 𝑕𝑛+1 𝑓 𝑛+1
(𝜉)
𝑛+1 !

Stirling Interpolating Interpolation:

Stirling interpolating polynomial is given by,

209
 𝑢 𝑢2
𝑃 𝑥 = 𝑓 𝑥0 + 𝛿𝑓1 + 𝛿𝑓−1 + 𝑓0 𝛿 2
SPACE FOR HINTS 2 2 2 2!
𝑢 𝑢 2 − 12 1 2
+ 𝛿 𝑓1 + 𝛿 3 𝑓−1
3! 2 2 2

𝑢 𝑢 2 − 12 ⋯ 𝑢 2 − 𝑝 − 1 2
1 2𝑝−1
+⋯ 𝛿 𝑓1 + 𝛿 2𝑝−1 𝑓−1
2𝑝 − 1 ! 2 2 2

𝑢 2 𝑢 2 − 12 𝑢 2 − 22 ⋯ 𝑢 2 − 𝑝 − 1 2
𝛿 2𝑝 𝑓0
2𝑝 !
𝑥−𝑥 0
where 𝑢 =
𝑕

Bessel’s Formula:

𝑦0 + 𝑦1 1 𝑢 𝑢−1 ∆2 𝑦 − 1 + ∆2 𝑦0
𝑃 𝑥 = + 𝑢 − ∆𝑦0 +
2 2 2! 2
1
𝑢− 𝑢(𝑢 − 1)
2
+ ∆3 𝑦 − 1
3!
𝑢+1 𝑢 𝑢−1 𝑢−2 ∆4 𝑦 − 2 + ∆4 𝑦 − 1
+
4! 2

𝑥−𝑥 0
Where 𝑢 = , Here 𝑥0 = 𝑚𝑖𝑑 𝑣𝑎𝑙𝑢𝑒.
𝑕

Example 3.4.1:
For the following data calculate the difference and obtain the
forward and backward difference polynomials. Interpolate 𝑥 = 0.25 and
𝑥 = 0.35.
𝑥 0.1 0.2 0.3 0.4 0.5

𝑓(𝑥) 1.40 1.56 1.76 2.00 2.28

Solution:
The difference table is
𝑥 𝑓(𝑥) ∆ ∆2 ∆3 ∆4

210
 0.1 1.40
SPACE OF HINTS
0.16

0.2 1.56 0.04

0.20 0

0.3 1.76 0.04 0

0.24 0

0.4 2.00 0.04

0.28

0.5 2.28

Forward difference polynomial is,

𝑥 − 𝑥0 𝑥 − 𝑥0 (𝑥 − 𝑥1 ) 2
𝑃 𝑥 = 𝑓 𝑥0 + ∆𝑓 𝑥0 + ∆ 𝑓(𝑥0 ) + ⋯
1! 𝑕 2! 𝑕2
0.16 𝑥 − 0.1 (𝑥 − 0.2) 0.04
= 1.40 + 𝑥 − 0.1 + +0+0
0.1 2 0.01
= 1.4 + 1.6𝑥 − 0.16 + (𝑥 2 − 0.3𝑥 + 0.02)2

= 1.4 + 1.6𝑥 − 0.16 + 2𝑥 2 − 0.6𝑥 + 0.04

𝑃 𝑥 = 2𝑥 2 + 𝑥 + 1.28

Put 𝑥 = 0.25 we get

𝑃 0.25 = 1.655

Backward difference polynomial is,

𝑥 − 𝑥0 𝑥 − 𝑥0 (𝑥 − 𝑥1 ) 2
𝑃 𝑥 = 𝑓 𝑥0 + ∇𝑓 𝑥0 + ∇ 𝑓(𝑥0 ) + ⋯
1! 𝑕 2! 𝑕2
𝑥 − 0.5 0.28 𝑥 − 0.5 𝑥 − 0.4 0.04
= 2.28 + +
0.1 2 0.01
= 2𝑥 2 + 𝑥 + 1.28

Put 𝑥 = 0.35 we get

𝑃 0.35 = 1.875
𝑢 𝑢+1
Now, 𝑃 𝑥 = 𝑓 𝑥𝑛 + 𝑢∇𝑓 𝑥𝑛 + ∇2 𝑓 𝑥𝑛 + ⋯
2!

211
 𝑥−𝑥 𝑛 0.35−0.5
where 𝑢 = = = −1.5
𝑕 0.1
SPACE FOR HINTS
𝑢 = −1.5
−1.5 (−0.5)
𝑃 0.35 = 2.28 + −1.5 0.28 + 0.04
2
= 2.28 − 0.42 + 0.015
= 1.875.
Example 3.4.2:
Determine the step size 𝑕 that can be used in the tabulation of a function
𝑓(𝑥),𝑎 ≤ 𝑥 ≤ 𝑏, at equally spaced nodal points so that the truncation
error of the quadratic interpolation is less than 𝜖.

Solution:

Let 𝑥𝑖−1 , 𝑥𝑖 , 𝑥𝑖+1 denote there consecutive equispaced points with

stepsize 𝑕.

The truncation error of the quadratic Lagrange interpolation is bounded

by,
𝑀3
𝐸2 (𝑓, 𝑥) ≤ max 𝑥 − 𝑥𝑖−1 𝑥 − 𝑥𝑖 (𝑥 − 𝑥𝑖+1 )
6
where 𝑥𝑖−1 ≤ 𝑥 ≤ 𝑥𝑖+1

and 𝑀3 = max𝑎≤𝑥≤𝑏 𝑓 ′′′ (𝑥)

(𝑥−𝑥 𝑖 )
Let 𝑡 = then 𝑥 − 𝑥𝑖 − 1 = 𝑥 − 𝑥𝑖 + 𝑕
𝑕

= 𝑡𝑕 + 𝑕
= 𝑡+1 𝑕
𝑥 − 𝑥𝑖+1 = 𝑥 − 𝑥𝑖 − 𝑕
= 𝑡𝑕 − 𝑕 = 𝑡 − 1 𝑕
Now 𝑥 − 𝑥𝑖−1 𝑥 − 𝑥𝑖 𝑥 − 𝑥𝑖+1 = 𝑡 + 1 𝑕 𝑡𝑕 𝑡 − 1 𝑕

= 𝑡 + 1 𝑡(𝑡 − 1)𝑕3

= 𝑡(𝑡 2 − 1)𝑕3

= 𝑔(𝑡) (say)

Now put 𝑔′ 𝑡 = 0 we get 𝑔 𝑡 = (𝑡 3 − 𝑡)𝑕3 𝑔′′ 𝑡 = 6𝑡

212
 1
⟹ 3𝑡 3 − 1 𝑕3 = 0 Put 𝑡 = − we attains a minimum value. SPACE OF HINTS
3

3𝑡 2 − 1 = 0 ⇒ 3𝑡 2 = 1
1
𝑡2 =
3

1
𝑡=±
3

For these values of 𝑡, we obtain

max 𝑥 − 𝑥𝑖−1 𝑥 − 𝑥𝑖 (𝑥 − 𝑥𝑖+1 ) = 𝑕3 𝑡(𝑡 2 − 1)
−1 1
= 𝑕3 −1
3 3
−1 −2
= 𝑕3
3 3
2
∴ 𝑚𝑎𝑥 𝑥 − 𝑥𝑖−1 𝑥 − 𝑥𝑖 (𝑥 − 𝑥𝑖+1 ) = 𝑕3
3 3
Hence the truncation error of the quadratic interpolation is
𝑀3 3 2
𝐸2 (𝑓, 𝑥) ≤ 𝑕
6 3 3
𝑕3 𝑀3
≤
9 3
Given the truncation error of the quadratic interpolation is less than 𝜖.
𝑕 3 𝑀3
(i.e.) <𝜖
9 3

9𝜖 3
𝑕3 <
𝑀3
1
3
9𝜖 3
𝑕< .
𝑀3
Example 3.4.3:
Determine the step size 𝑕 that can be used in the tabulation of a
function 𝑓(𝑥) , 𝑎 ≤ 𝑥 ≤ 𝑏 at equally spaced nodal points so the
truncation error of the interpolation is less than 𝜖.

Solution:

213
 Let 𝑥0 , 𝑥1 , 𝑥2 , 𝑥3 denote four consecutive equispaced points with
SPACE FOR HINTS stepsize 𝑕.

The truncation error of the cubic Lagrange interpolation is denoted by

𝑀4
𝐸3 (𝑓, 𝑥) ≤ max 𝑥 − 𝑥0 𝑥 − 𝑥1 (𝑥 − 𝑥2 )(𝑥 − 𝑥3 ) where
24

𝑥0 ≤ 𝑥 ≤ 𝑥3
and 𝑀4 = max𝑥 0 ≤𝑥≤𝑥 3 𝑓 (4) (𝑥)
1 (𝑥 1 +𝑥 2 )
𝑡 = (𝑥 − (or)
𝑕 2
𝑥1 + 𝑥2
⟹ 𝑡𝑕 = 𝑥 −
2
𝑥1 + 𝑥2
⟹ 𝑥 = 𝑡𝑕 +
2
𝑥1 + 𝑥2
𝑥 − 𝑥0 = 𝑡𝑕 + − 𝑥0
2
2𝑥0 + 3𝑕
= 𝑡𝑕 + − 𝑥0
2
3𝑕 3
= 𝑡𝑕 + = 𝑡+ 𝑕
2 2
𝑥1 + 𝑥2
𝑥 − 𝑥1 = 𝑡𝑕 + − 𝑥1
2
2𝑥1 + 𝑕
= 𝑡𝑕 + − 𝑥1
2
𝑕 1
= 𝑡𝑕 + = 𝑡 + 𝑕
2 2
𝑥1 + 𝑥2
𝑥 − 𝑥2 = 𝑡𝑕 + − 𝑥2
2
𝑕 1
= 𝑡𝑕 − = 𝑡 − 𝑕
2 2
𝑥1 + 𝑥2
𝑥 − 𝑥3 = 𝑡𝑕 + − 𝑥3
2
3𝑕 3
= 𝑡𝑕 − = 𝑡− 𝑕
2 2
Now
3 1 1
𝑥 − 𝑥0 𝑥 − 𝑥1 𝑥 − 𝑥2 𝑥 − 𝑥3 = 𝑡 + 𝑕 𝑡+ 𝑕 𝑡− 𝑕
2 2 2

214
 9 1 4
= 𝑡2 − 𝑡2 − 𝑕 SPACE OF HINTS
4 4
= 𝑔(𝑡)𝑕4

Put 𝑔′ 𝑡 = 0 we get
9 1
𝑡2 − + 𝑡2 − 2𝑡 = 0 (or)
4 4
5
4𝑡 𝑡 2 − = 0 (or)
4
5
𝑡 = 0 (or) 𝑡 2 =
4

5
𝑡=±
4

𝑔′′ 𝑡 = 12𝑡 2 − 5
max 𝑥 − 𝑥0 𝑥 − 𝑥1 𝑥 − 𝑥2 (𝑥 − 𝑥3 ) = 𝑕4 max 𝑔(𝑡)
𝑥 0 ≤𝑥≤𝑥 3 5 5
− <𝑡<
2 2

𝑕4
Hence 𝐸3 (𝑓; 𝑥) ≤
24
𝑕4
Choose 𝑕 ∋ : 𝑀4 < 𝜖
24
1
24 𝜖 4
𝑕<
𝑀4
3.5: Hermite Interpolation:
The formula for Hermite interpolation is
𝑛 𝑛

𝑃 𝑥 = 𝐴𝑖 𝑥 𝑓 𝑥𝑖 + 𝐵𝑖 𝑥 𝑓 ′ (𝑥𝑖 )
𝑖=0 𝑖=0

where 𝐴𝑖 (𝑥) and 𝐵𝑖 (𝑥) are polynomial of degree ≤ 2𝑛 + 1 and

𝐴𝑖 𝑥 = 1 − 2 𝑥 − 𝑥𝑖 𝑙𝑖 ′ 𝑥𝑖 𝑙𝑖 2 𝑥

𝐵𝑖 𝑥 = 𝑥 − 𝑥𝑖 𝑙𝑖 2 𝑥

Example 3.5.1:
Determine the parameters in the formula
3 2
𝑃 𝑥 = 𝑎0 𝑥 − 𝑎 + 𝑎1 𝑥 − 𝑎 + 𝑎2 𝑥 − 𝑎 + 𝑎3 such that
P 𝑎 = 𝑓 𝑎 , 𝑃′ 𝑎 = 𝑓 ′ 𝑎 , 𝑃 𝑏 = 𝑓 𝑏 , 𝑃′ 𝑏 = 𝑓 ′ (𝑏) .
Solution:

215
 3 2
Given 𝑃 𝑥 = 𝑎0 𝑥 − 𝑎 + 𝑎1 𝑥 − 𝑎 + 𝑎2 𝑥 − 𝑎 + 𝑎3 (1)
SPACE FOR HINTS Put 𝑥 = 𝑎 in (1), we get
𝑃 𝑎 = 𝑓 𝑎 = 𝑎3
∴ 𝑎3 = 𝑓(𝑎)
Differentiating equation (1) with respect to𝑥 , we get
𝑃′ 𝑥 = 𝑎0 3(𝑥 − 𝑎)2 + 2𝑎1 𝑥 − 𝑎 + 𝑎2 (2)
Put 𝑥 = 𝑎 in above equation (2), we get
𝑃′ 𝑎 = 𝑎2 = 𝑓 ′ (𝑎)
∴ 𝑎2 = 𝑓 ′ (𝑎)
Put 𝑥 = 𝑏 in (1), we get

𝑓 𝑏 = 𝑃 𝑏 = 𝑎0 (𝑏 − 𝑎)3 + 𝑎1 (𝑏 − 𝑎)2 + 𝑎2 𝑏 − 𝑎 + 𝑎3 (3)

𝑓 𝑏 = 𝑎0 (𝑏 − 𝑎)3 + 𝑎1 (𝑏 − 𝑎)2 + 𝑓 ′ 𝑎 𝑏 − 𝑎 + 𝑓(𝑎)

𝑓 𝑏 − 𝑓(𝑎)
= 𝑎0 (𝑏 − 𝑎)2 + 𝑎1 𝑏 − 𝑎 + 𝑓 ′ 𝑎 (4)
𝑏−𝑎
Differentiating (3) with respect to 𝑏 ,we get
𝑓 ′ 𝑏 = 𝑎0 3(𝑏 − 𝑎)2 + 2𝑎1 𝑏 − 𝑎 + 𝑎2
= 3𝑎0 (𝑏 − 𝑎)2 + 2𝑎1 𝑏 − 𝑎 + 𝑓 ′ (𝑎)

𝑓 ′ 𝑏 − 𝑓 ′ 𝑎 = 3𝑎0 𝑏 − 𝑎 2
+ 2𝑎1 𝑏 − 𝑎 (5)

Multiply 4 × 3 − 5
𝑓 𝑏 −𝑓 𝑎
⟹3 − 𝑓′ 𝑏 − 𝑓′ 𝑎
𝑏−𝑎
= 3𝑎1 𝑏 − 𝑎 − 2𝑎1 𝑏 − 𝑎 + 3𝑓 ′ (𝑎)

= 𝑎1 𝑏 − 𝑎 + 3𝑓 ′ (𝑎)

𝑓 𝑏 −𝑓 𝑎
3 − 𝑓′ 𝑏 − 𝑓′ 𝑎 − 3𝑓 ′ 𝑎 = 𝑎1 𝑏 − 𝑎
𝑏−𝑎

𝑓 𝑏 −𝑓 𝑎 𝑓 ′ 𝑏 + 2𝑓 ′ 𝑎
⟹ 𝑎1 = 3 −
(𝑏 − 𝑎)2 𝑏−𝑎

Substituting 𝑎1 value in (5) we get

216
 𝑓 ′ 𝑏 − 𝑓1 𝑎
SPACE OF HINTS

2
𝑓 𝑏 −𝑓 𝑎
= 3𝑎0 𝑏 − 𝑎 +6 𝑏−𝑎
𝑏−𝑎 2

𝑓 ′ 𝑏 + 2𝑓 ′ 𝑎
− 2 × (𝑏
𝑏−𝑎
− 𝑎)

2
𝑓 𝑏 −𝑓 𝑎
= 3𝑎0 𝑏 − 𝑎 +6 − 2 𝑓′ 𝑏 + 𝑓′ 𝑎
(𝑏 − 𝑎)

2
𝑓 𝑏 −𝑓 𝑎
= 3𝑎0 𝑏 − 𝑎 +6 − 2𝑓 ′ 𝑏 − 2𝑓 ′ 𝑎
(𝑏 − 𝑎)

𝑓 𝑏 −𝑓 𝑎
𝑓 ′ 𝑏 − 𝑓 1 𝑎 + 2𝑓 ′ 𝑏 + 4𝑓 ′ 𝑎 = 3𝑎0 𝑏 − 𝑎 2
+6
(𝑏 − 𝑎)

𝑓 𝑏 −𝑓 𝑎
3𝑓 ′ 𝑏 + 3𝑓 ′ 𝑎 = 3𝑎0 𝑏 − 𝑎 2
+6
(𝑏 − 𝑎)

2
𝑓′ 𝑏 + 𝑓′ 𝑎 𝑓 𝑏 −𝑓 𝑎
𝐷𝑖𝑣𝑖𝑑𝑒𝑑 𝑏𝑦 3 𝑏 − 𝑎 ⟹ 2
= 𝑎0 + 2
𝑏−𝑎 (𝑏 − 𝑎)3

𝑓′ 𝑏 + 𝑓′ 𝑎 𝑓 𝑏 −𝑓 𝑎
∴ 𝑎0 = − 2
𝑏−𝑎 2 (𝑏 − 𝑎)3

𝑓 𝑎 −𝑓 𝑏 𝑓′ 𝑎 + 𝑓′ 𝑏
𝑜𝑟 𝑎0 = 2 +
(𝑏 − 𝑎)3 𝑏−𝑎 2

Example 3.5.2:
Given the following values 𝑓 𝑥 and 𝑓 ′ (𝑥)

𝑥 𝑓(𝑥) 𝑓 ′ (𝑥)

−1 1 −5

0 1 1

1 3 7

217
 Estimate the value of 𝑓(−0.5) and 𝑓(0.5) using the Hermite
SPACE FOR HINTS 33 97
interpolation. The exact values are 𝑓 −0.5 = and 𝑓 0.5 = .
64 64

Solution:
Here 𝑖 = 0 , 1 , 2 and 𝑛 = 2
Given 𝑥0 = −1 , 𝑥1 = 0 , 𝑥2 = 1
𝑓 𝑥0 = 1 , 𝑓 𝑥1 = 1 , 𝑓 𝑥2 = 3
𝑓 ′ 𝑥0 = 5 , 𝑓 ′ 𝑥1 = 1 , 𝑓 ′ 𝑥2 = 7
The formula for Hermite interpolation is,
2 2

𝑃 𝑥 = 𝐴𝑖 𝑥 𝑓 𝑥𝑖 + 𝐵𝑖 𝑥 𝑓 ′ (𝑥𝑖 )
𝑖=0 𝑖=0

Where 𝐴𝑖 𝑥 = 1 − 2 𝑥 − 𝑥𝑖 𝑙𝑖 𝑥𝑖 𝑙𝑖 2 (𝑥) and

′

𝐵𝑖 𝑥 = 𝑥 − 𝑥𝑖 𝑙𝑖 2 (𝑥)
1
𝐴0 𝑥 = 3𝑥 5 − 2𝑥 4 − 5𝑥 3 + 4𝑥 2
4
Put 𝑖 = 1, 𝐴1 𝑥 = 1 − 2 𝑥 − 𝑥1 𝑙1 ′ 𝑥1 𝑙1 2 (𝑥)
𝑥 − 𝑥0 (𝑥 − 𝑥2 )
𝑙1 𝑥 =
𝑥1 − 𝑥0 (𝑥1 − 𝑥2 )
𝑥 + 1 (𝑥 − 1) 𝑥 2 − 1
= =
0 + 1 (0 − 1) −1
= − 𝑥2 − 1 = 1 − 𝑥2

𝑙1 2 𝑥 = (1 − 𝑥 2 )2 = 1 + 𝑥 4 − 2𝑥 2

𝑙1 ′ 𝑥 = −2𝑥

𝑥1 = 0 ⇒ 𝑙1 ′ 1 = 0
𝐴1 𝑥 = 1 − 2 𝑥 − 0 0 1 + 𝑥 4 − 2𝑥 2
= 1 + 𝑥 4 − 2𝑥 2 = 𝑥 4 − 2𝑥 2 + 1
Put 𝑖 = 2, 𝐴2 𝑥 = 1 − 2 𝑥 − 𝑥2 𝑙2 ′ 𝑥2 𝑙2 2 (𝑥)
𝑥 − 𝑥0 (𝑥 − 𝑥1 )
𝑙2 𝑥 =
𝑥2 − 𝑥0 (𝑥2 − 𝑥1 )
𝑥+1 𝑥−0 𝑥2 + 𝑥
= =
1+1 1−0 2

218
 1
𝑙2 ′ 𝑥 = (2𝑥 + 1) SPACE OF HINTS
2
1 3
𝑙2 ′ 𝑥 = 3 =
2 2
2
2 𝑥2 + 𝑥 1 4
𝑙2 𝑥 = = (𝑥 + 𝑥 2 + 2𝑥 3 )
2 4
2 𝑥−1 3 1 4
𝐴2 𝑥 = 1 − (𝑥 + 𝑥 2 + 2𝑥 3 )
2 4
1
= 1 − 3𝑥 + 3 (𝑥 4 + 𝑥 2 + 2𝑥 3 )
4
1
= 4𝑥 4 + 4𝑥 2 + 8𝑥 3 − 3𝑥 5 − 3𝑥 3 − 6𝑥 4
4
1
= (−3𝑥 5 − 2𝑥 4 + 5𝑥 3 + 4𝑥 2 )
4
𝑖 = 0, 𝐵0 𝑥 = 𝑥 − 𝑥0 𝑙0 2 (𝑥)

𝑥 4 + 𝑥 2 − 2𝑥 3 1
= 𝑥+1 = 𝑥 5 + 𝑥 3 − 2𝑥 4 + 𝑥 4 + 𝑥 2 − 2𝑥 3
4 4

1 5
= (𝑥 − 𝑥 4 − 𝑥 3 + 𝑥 2 )
4
𝑖 = 1, 𝐵1 𝑥 = (𝑥 − 𝑥1 )𝑙1 2 𝑥
= 𝑥 − 0 (1 + 𝑥 4 − 2𝑥 2 )
= 𝑥 + 𝑥 5 − 2𝑥 3
= 𝑥 5 − 2𝑥 3 + 1
𝑖 = 2, 𝐵2 𝑥 = (𝑥 − 𝑥0 )𝑙2 2 𝑥
𝑥 4 + 𝑥 2 + 2𝑥 3
= (𝑥 − 1)
4
1 5
= 𝑥 + 𝑥 3 + 2𝑥 4 − 𝑥 4 − 𝑥 2 − 2𝑥 3
4
1
= 𝑥5 + 𝑥4 − 𝑥3 − 𝑥2
4
2 2

∴𝑃 𝑥 = 𝐴𝑖 𝑥 𝑓 𝑥𝑖 + 𝐵𝑖 𝑥 𝑓 ′ (𝑥𝑖 )
𝑖=0 𝑖=0

219
 = 𝐴0 𝑥 𝑓 𝑥0 + 𝐴1 𝑥 𝑓 𝑥1 + 𝐴2 𝑥 𝑓 𝑥2 + 𝐵0 𝑥 𝑓 ′ 𝑥0
SPACE FOR HINTS + 𝐵1 𝑥 𝑓 ′ 𝑥1 + 𝐵2 𝑥 𝑓 ′ (𝑥2 )
1
= 3𝑥 5 − 2𝑥 4 − 5𝑥 3 + 4𝑥 2 1 + 𝑥 4 − 2𝑥 2 + 1 1
4
1
+ −3𝑥 5 − 2𝑥 4 + 5𝑥 3 + 4𝑥 2 3
4
1
+ 𝑥 5 − 𝑥 4 − 𝑥 3 + 𝑥 2 −1
4
+ 𝑥 5 − 2𝑥 3 + 𝑥 1
1 5
+ 𝑥 + 𝑥 4 − 𝑥 3 − 𝑥 2 (7)
4
= 1/4 3𝑥 5 − 2𝑥 4 − 5𝑥 3 + 4𝑥 2 + 4𝑥 4 − 8𝑥 2 + 4 − 9𝑥 5
− 6𝑥 4 + 15𝑥 3 + 12𝑥 2 − 5𝑥 5 + 5𝑥 4 + 5𝑥 3
− 5𝑥 2 + 4𝑥 5 − 8𝑥 3 + 4𝑥 + 7𝑥 5 + 7𝑥 4 − 7𝑥 3
− 7𝑥 2
1
𝑓 𝑥 = 8𝑥 4 − 4𝑥 2 + 4𝑥 + 4 = 2𝑥 4 − 𝑥 2 + 𝑥 + 1
4
97
𝑓 0.5 = 1.375 exact value is
64
33
𝑓 −0.5 = 0.375 exact value is .
64

3.6: Piecewise and Spline Interpolation:

Piecewise linear interpolation:
For 𝑥 ∈ 𝑥𝑖−1 , 𝑥𝑖 ,
𝑥−𝑥 𝑖 𝑥−𝑥 𝑖−1
then 𝑃𝑖,1 𝑥 = 𝑓 𝑥𝑖−1 + 𝑓 𝑥𝑖 , 𝑖 = 1,2,3, ⋯ , 𝑛
𝑥 𝑖−1 −𝑥 𝑖 𝑥 𝑖 −𝑥 𝑖−1
𝑥−𝑥 𝑖+1 𝑥−𝑥 𝑖
For 𝑥 ∈ 𝑥𝑖 , 𝑥𝑖+1 , then 𝑃𝑖+1,1 𝑥 = 𝑓 𝑥𝑖 + 𝑓 𝑥𝑖+1 ,
𝑥 𝑖 −𝑥 𝑖+1 𝑥 𝑖+1 −𝑥 𝑖

𝑖 = 1,2,3, ⋯ , 𝑛
Then the interpolating polynomial is,
𝑛

𝑃 𝑥 = 𝑝𝑖,1 (𝑥)
𝑖=0

which agree with 𝑓(𝑥) at 𝑥𝑖 , 𝑖 = 0,1,2, ⋯ , 𝑛 is linear in each

220
subinterval [𝑥𝑖−1 , 𝑥𝑖 ] can be written as
SPACE OF HINTS
𝑁

𝑃1 𝑥 = 𝑁𝑖 𝑥 𝑓(𝑥𝑖 )
𝑖=0

where 𝑁𝑖 𝑥 is called a shape function and is denoted by

0, 𝑥 ≤ 𝑥𝑖−1
𝑥−𝑥 𝑖−1
, 𝑥𝑖−1 ≤ 𝑥 ≤ 𝑥𝑖
𝑥 𝑖 −𝑥 𝑖−1
𝑥 𝑖+1 −𝑥
𝑁𝑖 𝑥 = , 𝑥𝑖 ≤ 𝑥 ≤ 𝑥𝑖+1
𝑥 𝑖+1 −𝑥 𝑖

0 𝑥 ≥ 𝑥𝑖+1

The error in the piecewise is given by

1
𝑓 𝑥 − 𝑃𝑖,1 𝑥 = 𝑥 − 𝑥𝑖−1 𝑥 − 𝑥𝑖 𝑓 ′′ 𝜖 , 𝑥𝑖−1 < 𝜖𝑖 < 𝑥𝑖
2!
Example 3.6.1:
Obtain the piecewise linear interpolating polynomials for the function
𝑓(𝑥) defined by the data. Hence estimate the value of 𝑓 3 & 𝑓(7).
𝑥 1 2 4 8

𝑓(𝑥) 3 7 21 73

Solution:
In the interval [1, 2], we have
𝑥−2 𝑥−1
𝑃1 𝑥 = 𝑃1,1 𝑥 = ×3+ 𝑥
1−2 2−1
= − 𝑥 − 2 3 + 7(𝑥 − 1)
= −3𝑥 + 6 + 7𝑥 − 7 = 4𝑥 − 1

In the interval [2, 4], we have

𝑥−4 𝑥−2
𝑃1 𝑥 = 𝑃2,1 𝑥 = 7 + (21)
2−4 4−2
7 21
=− 𝑥−4 + (𝑥 − 2)
2 2

221
 7𝑥 28 21 42
=− + + 𝑥−
SPACE FOR HINTS 2 2 2 2
14 14
= 𝑥− = 7𝑥 − 7
2 2
In the interval [4, 8], we have
𝑥−8 𝑥−4
𝑃1 𝑥 = 𝑃3,1 𝑥 = 21 + (73)
4−8 8−4
21 73
=− 𝑥−8 + (𝑥 − 4)
4 4
21𝑥 73
=− + 42 + 𝑥 − 73
4 4
= 13𝑥 − 31
∴ Piecewise linear interpolating polynomials are given by,
4𝑥 − 1 1≤𝑥≤2
𝑃1 𝑥 = 7𝑥 − 7 2𝑥 ≤ 𝑥 ≤ 4
13𝑥 − 31 4 ≤ 𝑥 ≤ 8
Using the polynomial in the interval [2, 4]
𝑓 3 = 7 3 − 7 = 21 − 7 = 14
𝑓 7 = 13 7 − 31 = 91 − 31 = 60
∴ 𝑓 3 = 14 & 𝑓 7 = 60

Piecewise Quadratic Interpolation:

For 𝑥 ∈ (𝑥𝑖−1 , 𝑥𝑖+1 ) then the quadratic interpolating polynomial is,
𝑥 − 𝑥𝑖 𝑥 − 𝑥𝑖+1
𝑃𝑖,2 𝑥 = 𝑓 𝑥𝑖−1
𝑥𝑖−1 − 𝑥𝑖 𝑥𝑖−1 − 𝑥𝑖+1
𝑥 − 𝑥𝑖−1 𝑥 − 𝑥𝑖+1
+ 𝑓 𝑥𝑖
𝑥𝑖 − 𝑥𝑖−1 𝑥𝑖 − 𝑥𝑖+1
𝑥 − 𝑥𝑖−1 𝑥 − 𝑥𝑖
+ 𝑓(𝑥𝑖+1 )
𝑥𝑖+1 − 𝑥𝑖−1 𝑥𝑖+1 − 𝑥𝑖
The error in the piecewise quadratic interpolation is given by
1
𝑓 𝑥 − 𝑃𝑖,2 𝑥 = 𝑥 − 𝑥𝑖−1 𝑥 − 𝑥𝑖 𝑥 − 𝑥𝑖+1 𝑓 ′′′ 𝜉𝑖 ,
3!

222
 𝑥𝑖−1 < 𝜉𝑖 < 𝑥𝑖+1 .
SPACE OF HINTS
Example 3.6.2:
Obtain the piecewise quadratic interpolating polynomials for the
function 𝑓 𝑥 defined by the data
𝑥 −3 −2 −1 1 3 6 7

𝑓(𝑥) 369 222 171 165 207 990 1779

Hence find an approximation value of 𝑓(−2.5) and 𝑓(6.5).

Solution:
Consider the groups of nodal points −3, −2, −1 , −1, 1, 3 , 3, 6, 7 .
On each of these subintervals, we write the quadratic interpolating
polynomials.
Consider the interval [−3, −1]
𝑃1,2 𝑥
𝑥+2 𝑥+1
= 369
−3 + 2 −3 + 1
𝑥+3 𝑥+1
+ 222
−2 + 3 −2 + 1
𝑥+3 𝑥+4
+ (171)
−1 + 3 −1 + 2
369 222
= 𝑥 2 + 3𝑥 + 2 − 𝑥 2 + 4𝑥 + 3
2 −1
𝑥 2 + 5𝑥 + 6
+ 171
2
369𝑥 2 + 1107𝑥 + 738 − 444𝑥 2 − 1776𝑥 − 1332 + 171𝑥 2 + 855𝑥 + 1026
=
2
96𝑥 2 + 186𝑥 + 432
=
2
= 48𝑥 2 + 93𝑥 + 216

Now consider the interval [−1, 3 ]

223
 𝑥−1 𝑥−3 𝑥+1 𝑥−3
𝑃2,2 𝑥 = 171 + 165
SPACE FOR HINTS −1 − 1 1 − 3 1+1 1−3
𝑥+1 𝑥−1
+ (207)
3+1 3−1
171 2 165 2 207 2
= 𝑥 − 4𝑥 + 3 − 𝑥 − 2𝑥 − 3 + (𝑥 − 1)
8 4 8
1
= 171𝑥 2 − 684𝑥 + 513 − 330𝑥 2 + 660𝑥 + 990 + 207𝑥 2
8
− 207
1
= (48𝑥 2 − 24𝑥 + 1296)
8
= 6𝑥 2 − 3𝑥 + 162
Consider the interval [3, 7]
𝑥−6 𝑥−7 𝑥−3 𝑥−7
𝑃3,2 𝑥 = 207 + 996
3−6 3−7 6−3 6−7
𝑥−3 𝑥−6
+ 1779
7−3 7−6
207 𝑥 2 − 10𝑥 + 21
= 𝑥 2 − 13𝑥 + 149 − 990
12 −3
1779
+ 𝑥 2 − 9𝑥 + 18
4
= 132𝑥 2 − 927𝑥 + 1880

48𝑥 2 + 93𝑥 + 216 − 3 ≤ 𝑥 ≤ −1

∴ 𝑃2 𝑥 = 6𝑥 2 − 3𝑥 + 162 − 1 ≤ 𝑥 ≤ 3
132𝑥 2 − 927𝑥 + 1800 3 ≤ 𝑥 ≤ 7
∴ 𝑓 −2.5 = 48(−2.5)2 + 93 −2.5 + 216 = 283.5

𝑓 6.5 = 132(6.5)2 − 927 6.5 + 1800 = 1351.5.

Piecewise Cubic Interpolation:

Example 3.6.3:
The function 𝑓 𝑥 defined by the data given below
𝑥 −3 −2 −1 1 3 6 7

224
 𝑓(𝑥) 369 222 171 165 207 990 1779
SPACE OF HINTS

Obtain the approximation values of 𝑓(−2.5) and 𝑓 6.5 , using the

Piecewise Cubic Interpolation.
Solution:
Consider the groups of nodal points as −3, −2, −1, 1 and 1, 3,
6, 7 .
First we find the Newton‟s divided difference table for the set of points
−3, −2, −1, 1
𝑥 𝑓(𝑥) First divided difference Second divided Third divided
difference difference

−3 369
222 − 369
= −147
222 −2 + 3 −51 + 147
−2 = 48
171 − 222 −1 + 3
= −51 16 − 48
−1 + 2 −3 + 51 = −8
1+3
−1 171 165 − 171 = 16
= −3 1+2
1+1
1 165

Therefore, on −3, 1 , we have

𝑃1,3 = 𝑓 𝑥0 + 𝑥 − 𝑥0 𝑓 𝑥0 , 𝑥1 + 𝑥 − 𝑥0 𝑥 − 𝑥1 𝑓 𝑥0 , 𝑥1 , 𝑥2
+ 𝑥 − 𝑥0 𝑥 − 𝑥1 𝑥 − 𝑥2 𝑓(𝑥0 , 𝑥1 , 𝑥2 , 𝑥3 )
= 369 + 𝑥 + 3 −147 + 𝑥 + 3 𝑥 + 2 48
+ 𝑥 + 3 𝑥 + 2 𝑥 + 1 −8
= 369 − 147𝑥 − 441 + 48𝑥 2 + 240 + 288 − 8𝑥 3 − 48𝑥 2
− 88𝑥 − 481
= −8𝑥 3 + 5𝑥 + 168
Now 𝑓 −2.5 = −8(−2.5)3 + 5 −2.5 + 168 = 280.5
On 1, 7 , we have the following difference table.
𝑥 𝑓(𝑥) First divided Second divided Third divided
difference difference

225
 difference
SPACE FOR HINTS
1 165
207 − 165
= 21
207 3−1 261 − 21
3 = 48
990 − 207 6 − 1
= 261 132 − 48
6−3 789 − 261 = 14
= 132 7−1
6 990 1779 − 990
= 789 7 − 3
7−6
7 1779

Hence,
𝑃2,3 = 𝑓 𝑥0 + 𝑥 − 𝑥0 𝑓 𝑥, 𝑥1 + 𝑥 − 𝑥0 𝑥 − 𝑥1 𝑓 𝑥0 , 𝑥1 , 𝑥2
+ 𝑥 − 𝑥0 𝑥 − 𝑥1 𝑥 − 𝑥2 𝑓(𝑥0 , 𝑥1 , 𝑥2 , 𝑥3
= 165 + 𝑥 − 1 21 + 𝑥 − 1 𝑥 − 3 48 + 𝑥 − 1
𝑥 − 3 𝑥 − 6 (14)

= 165 + 21𝑥 − 21 + 48𝑥 2 − 192𝑥 + 144 + 14𝑥 3 − 140𝑥 2

+ 378 − 252

= 14𝑥 3 − 92𝑥 2 + 207𝑥 + 36

𝑓 6.5 = 14(6.5)3 − 92(6.5)2 + 207 6.5 + 36

= 1339.25
Piecewise Cubic Interpolation using Hermite Type data:
Formula for Piecewise cubic Hermite interpolating polynomial is,
𝑃𝑖,3 (𝑥) = 𝐴𝑖−1 𝑥 𝑓𝑖−1 𝑥 + 𝐴𝑖 𝑥 𝑓𝑖 𝑥 + 𝐵𝑖−1 𝑥 𝑓𝑖−1 ′ 𝑥
+ 𝐵𝑖 𝑥 𝑓𝑖 ′ (𝑥)
(𝑥−𝑥 𝑖 )2 2 𝑥 𝑖−1 −𝑥
where 𝐴𝑖−1 𝑥 = 1+
𝑥 𝑖−1 −𝑥 𝑖 2 𝑥 𝑖−1 −𝑥 𝑖

(𝑥 − 𝑥𝑖−1 )2 2 𝑥 − 𝑥𝑖
𝐴𝑖 𝑥 = 1 +
𝑥𝑖−1 − 𝑥𝑖 2 𝑥𝑖−1 − 𝑥𝑖
𝑥 − 𝑥𝑖−1 (𝑥 − 𝑥𝑖 )2
𝐵𝑖−1 𝑥 =
𝑥𝑖−1 − 𝑥𝑖 2

226
 𝑥 − 𝑥𝑖 (𝑥 − 𝑥𝑖−1 )2
𝐵𝑖 𝑥 = SPACE OF HINTS
𝑥𝑖−1 − 𝑥𝑖 2
Example 3.6.4:
Using the following values of 𝑓(𝑥) and 𝑓 ′ (𝑥)
𝑥 𝑓(𝑥) 𝑓 ′ (𝑥)

−1 1 −5

0 1 1

1 3 7

Estimate the value of 𝑓(−0.5) and 𝑓(0.5) using preceuise cubic

Hermite interpolation.
Solution:
Here 𝑥𝑖−1 = −1 , 𝑥𝑖 = 0 , 𝑥𝑖+1 = 1

𝑓 𝑥𝑖−1 = 1, 𝑓 𝑥𝑖 = 1, 𝑓 𝑥𝑖+1 = 3

𝑓 ′ 𝑥𝑖−1 = −5, 𝑓 ′ 𝑥𝑖 = 1, 𝑓 ′ 𝑥𝑖+1 = 7

Piecewise Cubic Hermite interpolating polynomial becomes, in the

interval (−1, 0)
𝑃𝑖,3 𝑥 = 𝐴𝑖−1 𝑥 𝑓𝑖−1 𝑥 + 𝐴𝑖 𝑥 𝑓𝑖 𝑥 + 𝐵𝑖−1 𝑥 𝑓𝑖−1 ′ 𝑥
+ 𝐵𝑖 𝑥 𝑓𝑖 ′ (𝑥)
(𝑥−𝑥 𝑖 )2 2 𝑥 𝑖−1 −𝑥
Where 𝐴𝑖−1 𝑥 = 1+
𝑥 𝑖−1 −𝑥 𝑖 2 𝑥 𝑖−1 −𝑥 𝑖

𝑥2 2 𝑥+1
= 1+ = 𝑥 2 (1 + 2𝑥 + 2)
1 1
= 2𝑥 3 + 3𝑥 2

(𝑥 − 𝑥𝑖−1 )2 2 𝑥 − 𝑥𝑖
𝐴𝑖 𝑥 = 2
1+
𝑥𝑖−1 − 𝑥𝑖 𝑥𝑖−1 − 𝑥𝑖
2
𝑥+1 2𝑥
= 1+
1 −1
= 𝑥 2 + 2𝑥 + 1 1 − 2𝑥

= 𝑥 2 + 2𝑥 + 1 − 2𝑥 3 − 4𝑥 2 − 2𝑥

227
 = −2𝑥 3 − 3𝑥 2 + 1
SPACE FOR HINTS
𝑥 − 𝑥𝑖−1 (𝑥 − 𝑥𝑖 )2
𝐵𝑖−1 𝑥 =
𝑥𝑖−1 − 𝑥𝑖 2

𝑥 + 1 𝑥2
= = 𝑥3 + 𝑥2
1
𝑥 − 𝑥𝑖 (𝑥 − 𝑥𝑖−1 )2
𝐵𝑖 𝑥 =
𝑥𝑖−1 − 𝑥𝑖 2
2
𝑥 𝑥+1
= = 𝑥(𝑥 2 + 2𝑥 + 1)
1
= 𝑥 3 + 2𝑥 2 + 𝑥

∴ 𝑃3 𝑥 = 𝐴𝑖−1 𝑥 𝑓𝑖−1 𝑥 + 𝐴𝑖 𝑥 𝑓𝑖 𝑥 + 𝐵𝑖−1 𝑥 𝑓𝑖−1 ′ 𝑥

+ 𝐵𝑖 𝑥 𝑓𝑖 ′ 𝑥 + ⋯

= 2𝑥 3 + 3𝑥 2 1 + −2𝑥 3 − 3𝑥 2 + 1 + 𝑥 + 𝑥 2 −5
+ 𝑥 3 + 2𝑥 2 + 𝑥 (1)

= 2𝑥 3 + 3𝑥 2 − 2𝑥 3 − 3𝑥 2 + 1 − 5𝑥 3 − 5𝑥 2 + 𝑥 3 + 3𝑥 2 + 𝑥

= −4𝑥 3 − 3𝑥 2 + 𝑥 + 1

𝑓 −0.5 = −4(−0.5)3 − 3(−0.5)2 + −0.5 + 1 = 0.25

Now consider the interval [0, 1]

Here 𝑥𝑖−1 = 0 , 𝑥𝑖 = 1 ⇒ 𝑓𝑖−1 = 1 , 𝑓𝑖 = 3 , 𝑓𝑖−1 ′ = 1 , 𝑓𝑖 ′ = 7

𝑃𝑖,3 𝑥 = 𝐴𝑖−1 𝑥 𝑓𝑖−1 𝑥 + 𝐴𝑖 𝑥 𝑓𝑖 𝑥 + 𝐵𝑖−1 𝑥 𝑓𝑖−1 ′ 𝑥

+ 𝐵𝑖 𝑥 𝑓𝑖 ′ (𝑥)
2
𝑥−1 2 0−𝑥
𝐴𝑖−1 𝑥 = 2
1+
0−1 0−1

2
2 −𝑥
= 𝑥−1 1+
−1
2
= 𝑥−1 + (1 + 2𝑥)

= 𝑥 2 − 2𝑥 + 1 + 2𝑥 3 − 4𝑥 2 + 12𝑥

= 2𝑥 3 − 3𝑥 2 + 1

228
 2
𝑥−0 2 0−𝑥
𝐴𝑖 𝑥 = 2
1+ SPACE OF HINTS
0−1 0−1

𝑥2
= 1−2 𝑥−1
1
= 𝑥 2 (1 − 2𝑥 + 2)

= 𝑥 2 − 2𝑥 3 + 2𝑥 2 = −2𝑥 3 + 3𝑥 2

(𝑥 − 0)(𝑥 − 1)2
𝐵𝑖−1 𝑥 = 2
= 𝑥(𝑥 − 1)2 = 𝑥(𝑥 2 − 2𝑥 + 1)
(0 − 1)

= 𝑥 3 − 2𝑥 2 + 𝑥

(𝑥 − 1)(𝑥 − 0)2
𝐵𝑖 𝑥 = 2
= 𝑥 − 1 𝑥2 = 𝑥3 − 𝑥2
(0 − 1)

𝑃3 𝑥 = 2𝑥 3 − 3𝑥 2 + 1 1 + −2𝑥 3 + 3𝑥 2 3 + 𝑥 3 − 2𝑥 2 + 𝑥 1
+ 𝑥 3 − 𝑥 2 (7)

= 2𝑥 3 − 3𝑥 2 + 1 − 6𝑥 3 + 9𝑥 2 + 𝑥 3 − 2𝑥 2 + 𝑥 + 7𝑥 3 − 7𝑥 2

𝑃3 𝑥 = 4𝑥 3 − 3𝑥 2 + 𝑥 + 1 = 𝑓(𝑥)

𝑓 0.5 = 1.25

Example 3.6.5:

𝑆3 (𝑥) is the piecewise cubic Hermite interpolating approximation of

𝑓 𝑥 = sin 𝑥 cos 𝑥in the abscissas 0, 1, 1.5, 2, 3. Estimate the error
max0≤𝑥≤3 𝑓 𝑥 − 𝑆3 (𝑥)

Solution:
Given 𝑓 𝑥 = sin 𝑥 cos 𝑥
sin 2𝑥
𝑓 𝑥 =
2
1
𝑓′ 𝑥 = . cos 2𝑥 . 2 = cos 2𝑥
2
𝑓 ′′ 𝑥 = − sin 2𝑥 . 2 = −2 sin 2𝑥

𝑓 ′′′ 𝑥 = −4 cos 2𝑥

𝑓 ′′′′ 𝑥 = 8 sin 2𝑥

229
 The error term 𝐸𝑖 is given by,
SPACE FOR HINTS 1
𝐸𝑖 = (𝑥 − 𝑥𝑖−1 )2 (𝑥 − 𝑥𝑖 )2 + 𝜖 ; 𝑥𝑖−1 < 𝜖 < 𝑥𝑖
4!
Here 𝑀𝑖 = max𝑥 𝑖−1 ≤𝑥≤𝑥 𝑖 𝑓 𝑖𝑣 𝑥 = max 8 sin 2𝑥
𝑥 𝑖−1 ≤𝑥≤𝑥 𝑖

Now 𝑀1 = max 𝑓 𝑖𝑣 𝑥 = max 8 sin 2𝑥

0≤𝑥≤1 𝑥 𝑖−1 ≤𝑥≤𝑥 𝑖

𝑀2 = max 8 sin 2𝑥 = 7.9799 ≅ 8

1≤𝑥≤1.5

𝑀3 = max 8 sin 2𝑥 = 6.0544

1.5≤𝑥≤2

𝑀4 = max 8 sin 2𝑥 = 7.9693 ≅ 8

2≤𝑥≤3

Also, the maximum value of 𝑥 − 𝑥𝑖−1 (𝑥 − 𝑥𝑖 ) is obtained at

𝑥𝑖 + 𝑥𝑖−1
𝑥=
2
1
∴ max (𝑥 − 𝑥𝑖−1 )2 (𝑥 − 𝑥𝑖 )2
𝑥 𝑖−1 ≤𝑥≤𝑥 𝑖 4!
2 2
1 𝑥𝑖−1 + 𝑥𝑖 𝑥𝑖−1 + 𝑥𝑖
= − 𝑥𝑖−1 − 𝑥𝑖
4! 2 2

2 2
1 𝑥𝑖 + 𝑥𝑖−1 𝑥𝑖−1 + 𝑥𝑖
=
4! 2 2
1
= (𝑥 − 𝑥𝑖−1 )2 (𝑥𝑖 − 𝑥𝑖−1 )2
24 × 16 𝑖
1
= (𝑥 − 𝑥𝑖−1 )4
384 𝑖

1
∴ at 0,1 , 𝐸1 = (1 − 0)4 𝑀1
384

1
= 8 = 0.0208
384
1
at 1,1.5 , 𝐸2 = (1.5 − 1)4 𝑀2
384

1
= (0.5)4 × 7.2744
384

230
 = 0.0012
SPACE OF HINTS
1
at 1.5,2 , 𝐸3 = (2 − 1.5)4 𝑀3
384

1
= (0.5)4 6.0544 = 0.00099
384
1
at 2,3 , 𝐸4 = (3 − 2)4 𝑀4
384

1
= 8 = 0.0208
384
The maximum error term attained at the interval 0, 1 & (2, 3)

∴ max 𝑓 𝑥 − 𝑆3 𝑥 = max( 𝐸1 , 𝐸2 , 𝐸3 , 𝐸4
0≤𝑥≤3

= 0.0208

and the error 𝐸 is 0.0208.

Spline interpolation

Definition 3.6.1:

A spline function of degree 𝑛 with nodes 𝑥𝑖 , 𝑖 = 0, 1, … , 𝑛 is a function

𝐹(𝑥) satisfying the properties

i) 𝐹 𝑥𝑖 = 𝑓 𝑥𝑖 , 𝑖 = 0, 1, … , 𝑛.
ii) On each subintervals 𝑥𝑖−1 , 𝑥𝑖 , 1 ≤ 𝑖 ≤ 𝑛, 𝐹(𝑥) is a polynomial
of degree 𝑛.
iii) 𝐹(𝑥) and its first (𝑛 − 1) derivatives are continuous
interpolation.

Quadratic Spline interpolation:

The formula for quadratic spline interpolation is,

𝐹 𝑥𝑖 = 𝑃𝑖 𝑥 = 𝑎𝑖 𝑥 2 + 𝑏𝑖 𝑥 + 𝑐𝑖 , 𝑖 = 1,2,3, ⋯ , 𝑛 1

𝑥𝑖−1 ≤ 𝑥 ≤ 𝑥𝑖

𝑏1 𝑥0 + 𝑐1 = 𝑓0

𝑏1 𝑥1 + 𝑐1 = 𝑓1 (2)

𝑎2 𝑥1 2 + 𝑏2 𝑥1 + 𝑐2 = 𝑓1

231
 𝑎2 𝑥2 2 + 𝑏2 𝑥2 + 𝑐2 = 𝑓2 (3)
SPACE FOR HINTS 2𝑎2 𝑥1 + 𝑏2 = 2𝑎1 𝑥1 + 𝑏1

𝑎3 𝑥2 2 + 𝑏3 𝑥2 + 𝑐3 = 𝑓2

𝑎3 𝑥3 2 + 𝑏3 𝑥3 + 𝑐3 = 𝑓3 (4)

2𝑎3 𝑥2 + 𝑏3 = 2𝑎2 𝑥2 + 𝑏2

Example 3.6.1:
Given the data

𝑥 0 1 2 3

𝑓(𝑥) 1 2 33 244

fit quadratic splines with 𝑀 0 = 𝑓 ′′ 0 = 0. Hence find an estimate of

𝑓(2.5).

Solution:
The spline approximation is

𝑃1 𝑥 = 𝑎1 𝑥 2 + 𝑏1 𝑥 + 𝑐1 , 0 ≤ 𝑥 ≤ 1 (1)

𝑃2 𝑥 = 𝑎2 𝑥 2 + 𝑏2 𝑥 + 𝑐2 , 1 ≤ 𝑥 ≤ 2 (2)

𝑃3 𝑥 = 𝑎3 𝑥 2 + 𝑏3 𝑥 + 𝑐3 , 2 ≤ 𝑥 ≤ 3 (3)

Since 𝑀 0 = 𝑓 ′′ 0 = 0

1 ⟹ 𝑃1 ′ 𝑥 = 2𝑎1 𝑥 + 𝑏1

𝑃1 ′′ 𝑥 = 2𝑎1

Now, 𝑥0 = 0 , 𝑥1 = 1 , 𝑥2 = 2 , 𝑥3 = 3

𝑓 𝑥0 = 1 , 𝑓 𝑥1 = 2 , 𝑓 𝑥2 = 33 , 𝑓 𝑥3 = 244

1 ⟹ 𝑏1 𝑥 + 𝑐1 = 𝑓0

⟹ 𝑏1 0 + 𝑐1 = 1

⟹ 𝑐1 = 1 (4)

1 ⟹ 𝑏1 𝑥 + 𝑐1 = 𝑓1

⟹ 𝑏1 + 1 = 2

232
 ⟹ 𝑏1 = 1 (5)
SPACE OF HINTS
2
(2) ⟹ 𝑎2 𝑥1 + 𝑏2 𝑥1 + 𝑐2 = 𝑓1

⟹ 𝑎2 + 𝑏2 + 𝑐2 = 2 (6)

⟹ 𝑎2 𝑥2 2 + 𝑏2 𝑥2 + 𝑐2 = 𝑓2

⟹ 4𝑎2 + 2𝑏2 + 𝑐2 = 33 7

⟹ 2𝑎2 𝑥1 + 𝑏2 = 2𝑎1 𝑥1 + 𝑏1

⟹ 2𝑎2 + 𝑏2 = 1 8

8 × 2 ⟹ 4𝑎2 + 2𝑏2 = 2

(4) ⟹ 2 + 𝑐2 = 33
∴ 𝑐2 = 31

Solving equation (6) and (7) we get

𝑎2 + 𝑏2 + 𝑐2 = 2

4𝑎2 + 2𝑏2 + 𝑐2 = 33

-3𝑎2 − 𝑏2 = −31

3𝑎2 + 𝑏2 = 31

2𝑎2 + 𝑏2 = 1

𝑎2 = 30

∴ 𝑎2 = 30

(8) ⟹ 2𝑎2 + 𝑏2 = 1

∴ 𝑏2 = −59

3 ⟹ 𝑎3 𝑥2 2 + 𝑏3 𝑥2 + 𝑐3 = 𝑓2

⟹ 4𝑎3 + 2𝑏3 + 𝑐3 = 33 (9)

𝑎3 𝑥3 2 + 𝑏3 𝑥3 + 𝑐3 = 𝑓3

⟹ 9𝑎3 + 3𝑏3 + 𝑐3 = 244 (10)

2𝑎3 𝑥2 + 𝑏3 = 2𝑎2 𝑥2 + 𝑏2

4𝑎3 + 𝑏3 = 2 × 30 × 2 − 59 = 61 ⟹ 4𝑎3 + 𝑏3 = 61 (11)

233
 10) − (9 ⇒ 5𝑎3 + 𝑏3 = 211 (12)
SPACE FOR HINTS
11 ⟹ 4𝑎3 + 𝑏3 = 61

𝑎3 = 150 ⟹ 𝑎3 = 150

11 ⟹ 4 150 + 𝑏3 = 61 ⇒ 𝑏3 = 61 − 600 = −539

𝑏3 = −539

9 ⟹ 9 150 − 3 539 + 𝑐3 = 244

⟹ 1350 − 1617 + 𝑐3 = 244

𝑐3 = 511

Hence 𝑎1 = 0 , 𝑏1 = 1 , 𝑐1 = 1

𝑎2 = 30 , 𝑏2 = −59 , 𝑐2 = 31

𝑎3 = 150 , 𝑏3 = −539 , 𝑐3 = 511

∴ 𝑃1 𝑥 = 𝑥 + 1 , 0 ≤ 𝑥 ≤ 1

𝑃2 𝑥 = 30𝑥 2 + −59 𝑥 + 31 = 30𝑥 2 − 59𝑥 + 31 , 1 ≤ 𝑥 ≤ 2

𝑃3 𝑥 = 150𝑥 2 − 539𝑥 + 511 , 2 ≤ 𝑥 ≤ 3

𝑓 2.5 = 150(2.5)2 − 539 2.5 + 511 = 101.

Cubic spline interpolation:

1
The formula for cubic spline interpolation is, 𝐹 𝑥 = (𝑥𝑖 −
6𝑕
1
𝑥)3 𝑀𝑖−1 + (𝑥 − 𝑥𝑖−1 )3 𝑀𝑖 + 𝑥𝑖 − 𝑥
𝑕

𝑕2 1 𝑕2
𝑓𝑖−1 − 𝑀𝑖−1 + 𝑥 − 𝑥𝑖−1 𝑓𝑖 − 𝑀
6 𝑕 6 𝑖
6
and 𝑀𝑖−1 + 4𝑀𝑖 + 𝑀𝑖+1 = (𝑓𝑖+1 − 2𝑓𝑖 + 𝑓𝑖−1 )
𝑕2

Example 3.6.2:
Obtain the cubic spline approximation for the function defined by the
data

𝑥 0 1 2 3

𝑓(𝑥) 1 2 33 244

234
with 𝑀 0 = 0, 𝑀 3 = 0. Hence find an estimate of 𝑓(2.5)
SPACE OF HINTS
Solution:
Here 𝑕 = 1 and we know that,
6
𝑚𝑖−1 + 4𝑚𝑖 + 𝑀𝑖+1 = 𝑓𝑖+1 − 2𝑓𝑖 + 𝑓𝑖−1 where 𝑖 = 1, 2
𝑕2

For 𝑖 = 1 , 𝑀0 + 4𝑀1 + 𝑀2 = 6(𝑓2 − 2𝑓1 + 𝑓0 )

Given 𝑀 0 = 0 and 𝑀 3 = 0 ⟹ 𝑀0 = 0 & 𝑀3 = 0

∴ 4𝑀1 + 𝑀2 = 6 33 − 4 + 1 = 180

∴ 4𝑀1 + 𝑀2 = 180 (1)

Now, put 𝑖 = 2, 𝑀1 + 4𝑀2 + 𝑀3 = 6(𝑓3 − 2𝑓2 + 𝑓1 )

∴ 𝑀1 + 4𝑀2 = 6 244 − 66 + 2 = 1080

∴ 𝑀1 + 4𝑀2 = 1080 2

1 × 4 ⟹ 16𝑀1 + 4𝑀2 = 720

𝑀1 + 4𝑀2 = 1080

15𝑀1 = −360

⟹ 𝑀1 = −24

−24 + 4𝑀2 = 1080

⟹ 𝑀2 = 276

The cubic splines in the corresponding intervals are obtained as follows

1
𝐹 𝑥 = 𝑥𝑖 − 𝑥 3 𝑀𝑖−1 + 𝑥 − 𝑥𝑖−1 3 𝑀𝑖
6𝑕
1 𝑕2
+ 𝑥𝑖 − 𝑥 𝑓𝑖−1 − 𝑀𝑖−1
𝑕 6
1 𝑕2
+ 𝑥 − 𝑥𝑖−1 𝑓𝑖 − 𝑀𝑖
𝑕 6

Consider the interval 0, 1 𝑖 = 1, 𝑕 = 1

235
 1 1
𝐹 𝑥 = (𝑥 − 𝑥)3 𝑀0 + (𝑥 − 𝑥0 )3 𝑀1 + 𝑥1 − 𝑥 𝑓0 − 𝑀0
SPACE FOR HINTS 6 1 6
1
+ 𝑥 − 𝑥0 (𝑓1 − 𝑀1 )
6
1 24
= ( 1 − 𝑥 30 + 𝑥 − 0 3
24 + 1 − 𝑥 1 − 0 + 𝑥 − 0 2+
6 6
1
= −24𝑥 3 + 1 − 𝑥 + 6𝑥
6
= −4𝑥 3 + 5𝑥 + 1

Consider the interval [1, 2], 𝑖 = 2, 𝑕 = 1

1 1
𝐹 𝑥 = (𝑥 − 𝑥)3 𝑀1 + (𝑥 − 𝑥1 )3 𝑀2 + 𝑥2 − 𝑥 𝑓1 − 𝑀1
6 2 6
1
+ 𝑥 − 𝑥1 (𝑓2 − 𝑀2 )
6

1 1
= ( 2 − 𝑥 3 (−24) + 𝑥 − 1 3
276 + 2 − 𝑥 2− −24
6 6

276
+ 𝑥−1 33 +
6
= −4 8 − 𝑥 3 − 12𝑥 + 6𝑥 2 + 𝑥 3 − 1 − 3𝑥 2 + 3𝑥 + 46
+ 2 − 𝑥 6 + (𝑥 − 1)(−13)

= 𝑥 3 4 + 46 + 𝑥 2 −24 − 138 + 𝑥 48 + 138 − 6 − 13

− 32 − 46 + 12 + 13

= 50𝑥 3 − 162𝑥 2 + 167𝑥 − 53

Consider the interval 2, 3 , 𝑖 = 3, 𝑕 = 1

1 276
𝐹(𝑥) = ( 3 − 𝑥 3 (276) + 𝑥 − 3 3
0 + 3−𝑥 33 −
6 6
+ 𝑥 − 2 244 − 0
1
= 27 − 27𝑥 + 9𝑥 2 − 𝑥 3 46 + 3 − 𝑥 −13 + 𝑥 − 2 (244 − 0)
6
= 𝑥 3 −46 + 𝑥 2 414 + 𝑥 −1242 + 13 + 244 + 1242 − 39 − 488

236
= −46𝑥 3 + 414𝑥 2 − 985𝑥 + 715
SPACE OF HINTS
3
𝐹 𝑥 = −4𝑥 + 5𝑥 + 1 , 0 ≤ 𝑥 ≤ 1

𝐹 𝑥 = 50𝑥 3 − 16𝑥 2 + 167𝑥 − 53 , 1 ≤ 𝑥 ≤ 2

𝐹 𝑥 = −46𝑥 3 + 414𝑥 2 − 985𝑥 + 715 ,2 ≤ 𝑥 ≤ 3

∴ 𝑓 2.5 = −46(2.5)3 + 414(2.5)2 − 985 2.5 + 715

= 121.25

Unit – 4

Differentiation and Integration

4.1:Numerical Differentiation
Numerical differentiation methods are obtained using one of the
following three techniques:

(i). Methods based on interpolation.

237
 (ii). Methods based on finite difference operators.
SPACE FOR HINTS (iii). Methods based on undetermined coefficients.

Methods based on interpolation

The values of 𝑓(𝑥) at a set of points 𝑥0 , 𝑥1 , 𝑥2 , … , 𝑥𝑛 , the general
approach for deriving the numerical differentiation methods is to first
obtain the interpolating polynomial 𝑃𝑛 (𝑥).
𝑟
Then differentiate this polynomial 𝑟 times (𝑛 ≥ 𝑟) to get 𝑃𝑛 𝑥 .
𝑟 𝑟
The value of 𝑃𝑛 𝑥𝑘 gives the approximate value of 𝑓 (𝑥) at the
nodal point 𝑥𝑘 .

𝑃𝑛 𝑥 𝑎𝑛𝑑 𝑓 𝑥 have the same values at the nodal points, yet the
derivatives may differ considerably at these points.

The situation may be worse at a non-nodal point. The approximation

may further weaken as the order of derivative increases. The quantity
𝑟 𝑟 𝑟
𝐸 𝑥 =𝑓 𝑥 − 𝑃𝑛 𝑥 (1) This is
called the error of approximation in the 𝑟 𝑡𝑕 order derivative at any point
𝑥.

Non-uniform nodal points

Linear interpolation:
The formula for linear interpolation is,
𝑓1 − 𝑓0
𝑃1 ′ 𝑥 = 𝑥 ∈ 𝑥0 , 𝑥1
𝑥1 − 𝑥0
𝑥 0 −𝑥 1
The truncation error is𝐸1 ′ 𝑥0 = 𝑓 ′′ 𝜉 , 𝑥 < 𝜉 < 𝑥1 .
2

Quadratic interpolation:
The formula for quadratic interpolation is,

238
 2𝑥 − 𝑥1 − 𝑥2 2𝑥 − 𝑥 − 𝑥2
𝑃2 ′ 𝑥 = 𝑓0 + 𝑓 SPACE OF HINTS
(𝑥0 − 𝑥1 )(𝑥0 − 𝑥2 ) (𝑥1 − 𝑥0 )(𝑥1 − 𝑥2 ) 1
2𝑥 − 𝑥0 − 𝑥1
+ 𝑓
(𝑥2 − 𝑥0 )(𝑥2 − 𝑥1 ) 2

and the error term is,

1
𝐸2 ′ 𝑥0 = 𝑥 − 𝑥1 𝑥0 − 𝑥2 𝑓 ′′′ 𝜉 , 𝑥0 < 𝜉 < 𝑥2
6 0
Similarly
𝑓0 𝑓1
𝑃2 ′′ 𝑥 = 2 +
𝑥0 − 𝑥1 𝑥0 − 𝑥2 𝑥1 − 𝑥0 𝑥1 − 𝑥2
𝑓3
+ , 𝑥 ∈ 𝑥0 , 𝑥1
𝑥2 − 𝑥0 𝑥2 − 𝑥1
The truncation error is,
1
𝐸2 ′′ 𝑥0 = 2𝑥0 − 𝑥1 − 𝑥2 𝑓 ′′′ 𝜉
3
1
+ 𝑥 − 𝑥1 𝑥0 − 𝑥2 𝑓 𝐼𝑣 𝜂1 + 𝑓 𝐼𝑣 𝜂2
24 0
where 𝜂1 , 𝜂2 ∈ 𝑥0 , 𝑥1

Uniform nodal points

Liner interpolation
𝑓1 − 𝑓0 𝑓1 − 𝑓0
𝑓 ′ 𝑥0 ≅ 𝑃1 ′ 𝑥 = , 𝐸1 ′ 𝑥0 = 𝑓 ′ 𝑥0 −
𝑕 𝑕
𝑓1 − 𝑓0 𝑓1 − 𝑓0
𝑓 ′ 𝑥1 ≅ 𝑃1 ′ 𝑥 = , 𝐸1 ′ 𝑥1 = 𝑓 ′ 𝑥1 −
𝑕 𝑕
Quadratic interpolation
1
𝑓 ′ 𝑥0 = −3𝑓0 + 4𝑓1 − 𝑓2
2𝑕
1
𝑓 ′ 𝑥1 = 𝑓 − 𝑓0
2𝑕 2
1
𝑓 ′ 𝑥2 = 𝑓 − 4𝑓1 + 3𝑓2
2𝑕 0
And the truncation error is,

239
 −𝑕2 ′′′
𝐸2 ′ 𝑥0 = 𝑓 𝜉 , 𝑥0 < 𝜉 < 𝑥2
SPACE FOR HINTS 3

′ −𝑕2 ′′′
𝐸2 𝑥1 = 𝑓 𝜂1 , 𝑥0 < 𝜂1 < 𝑥2
6
−𝑕2 ′′′
𝐸2 ′ 𝑥2 = 𝑓 𝜂2 , 𝑥0 < 𝜂2 < 𝑥2
3
Similarly,
1
𝑓 ′′ 𝑥0 ≅ 𝑃2 ′′ 𝑥0 = 𝑓 − 2𝑓1 + 𝑓2
𝑕2 0
1
𝑓 ′′ 𝑥1 ≅ 𝑃2 ′′ 𝑥1 = 𝑓 − 2𝑓1 + 𝑓2
𝑕2 0
1
𝑓 ′′ 𝑥2 ≅ 𝑃2 ′′ 𝑥2 = 𝑓 − 2𝑓1 + 𝑓2
𝑕2 0
The truncation error is given by,
1
𝐸2 ′′ 𝑥0 = 𝑓 ′′ 𝑥0 − 𝑓 𝑥0 − 2𝑓 𝑥2 + 𝑓 𝑥2
𝑕2
= −𝑕𝑓 ′′′ 𝜉1 , 𝑥0 < 𝜉1 < 𝑥2

′′ −𝑕2 𝑖𝑣
𝐸2 𝑥1 = 𝑓 𝜉2 , 𝑥0 < 𝜉2 < 𝑥2
12
𝐸2 ′′ 𝑥2 = 𝑕𝑓 ′′′ 𝜉3 , 𝑥0 < 𝜉3 < 𝑥2

Example: 4.1.1
Given the following values of 𝑓 𝑥 = ln 𝑥find the approximate value of
𝑓 ′ 2.0 using linear and quadratic interpolation of 𝑓 ′′ 2.0 and using
quadratic interpolation. Find an upper bound on the error.

I 0 1 2

𝑥𝑖 2.0 2.2 2.6

𝑓𝑖 0.69315 0.78846 0.95551

Solution:
(i). Given 𝑥0 = 2.0 , 𝑥1 = 2.2 , 𝑥2 = 2.6

𝑓0 = 0.69315 , 𝑓1 = 0.78846 , 𝑓2 = 0.95551

240
The formula for linear interpolation is (non-uniform)
SPACE OF HINTS
𝑓1 − 𝑓0
𝑓 ′ 𝑥0 =
𝑥1 − 𝑥0
0.78846 − 0.69315
= = 0.4765
2.2 − 2.0
Quadratic interpolation
2𝑥0 − 𝑥1 − 𝑥2 𝑥0 − 𝑥2
𝑓 ′ 𝑥0 = 𝑓0 + 𝑓
(𝑥0 − 𝑥1 )(𝑥0 − 𝑥2 ) (𝑥1 − 𝑥0 )(𝑥1 − 𝑥2 ) 1
𝑥0 − 𝑥1
+ 𝑓
(𝑥2 − 𝑥0 )(𝑥2 − 𝑥1 ) 2

𝑓 ′ 2.0
4 − 2.2 − 2.6
= (0.69315)
2.0 − 2.2 2.0 − 2.6
2 − 2.6
+ (0.78846)
2.2 − 2 2.2 − 2.6
2 − 2.2
+ (0.95551)
2.6 − 2 2.6 − 2.2

= 0.04621 + 0.05913 − 0.00796 = 0.4961

To find 𝒇′′ 𝒙𝟎

𝑓 ′′ 𝑥0
𝑓0 𝑓1
=2 +
(𝑥0 − 𝑥1 )(𝑥0 − 𝑥2 ) (𝑥1 − 𝑥0 )(𝑥1 − 𝑥2 )
𝑓3
+
(𝑥2 − 𝑥0 )(𝑥2 − 𝑥1 )

𝑓 ′′ 2.0
0.69315 0.78846
=2 +
(2 − 2.2)(2 − 2.6) (2.2 − 2)(2.2 − 2.6)
0.95551
+
(2.6 − 2)(2.6 − 2.2)

𝑓 ′′ 2.0 = −0.19642

241
 𝑥0 − 𝑥1 ′′
𝐸1 ′ 𝑥0 = 𝑓 𝜉 , 𝑥0 < 𝜉 < 𝑥1
SPACE FOR HINTS 2
1
𝐸2 ′ 𝑥0 = 𝑥0 − 𝑥1 𝑥0 − 𝑥2 𝑓 ′′′ 𝜉 , 𝑥0 < 𝜉 < 𝑥2
6
1
𝐸2 ′′ 𝑥0 = 2𝑥0 − 𝑥1 − 𝑥2 𝑓 ′′′ 𝜉
3
1
+ 𝑥0 − 𝑥1 𝑥0 − 𝑥2 𝑓 𝐼𝑣 𝜂1 + 𝑓 𝐼𝑣 𝜂2
24

𝑥0 < 𝜉, 𝜂1 , 𝜂2 < 𝑥2

Given 𝑓 𝑥 = ln 𝑥
1 1 2 −6
𝑓′ 𝑥 = , 𝑓 ′′ 𝑥 = − , 𝑓 ′′′ 𝑥 = , 𝑓 𝑖𝑣 𝑥 =
𝑥 𝑥2 𝑥3 𝑥4
1
Now, 𝑀2 = max𝑥 0 <𝜉 <𝑥 1 𝑓 ′′ (𝜖) = max2<𝜉 <2.2 − = 0.25
𝑥2

2
𝑀3 = max 𝑓 ′′′ (𝜖) = max = 0.25
𝑥 0 <𝜉 <𝑥 2 2<𝜉 <2.6 𝑥3
6
𝑀4 = max 𝑓 𝑖𝑣 (𝜖) = max − = 0.375
𝑥 0 <𝜉 , 𝜂 1 ,𝜂 2 <𝑥 2 2<𝜉 ,𝜂 1 ,𝜂 2 <2.6 𝑥4
2 − 2.2
𝐸1 ′ 2.0 = 0.25 = 0.025
2
1
𝐸2 ′ 2.0 = 2 − 2.2 2 − 2.6 0.25 = 0.005
6
𝐸2 ′′ 2.0
1
= 4 − 2.2 − 2.6 0.25
3
1
+ 2 − 2.2 2 − 2.6 0.375 + 0.375
24

𝐸2 ′′ 2.0 = 0.0704

Example: 4.1.2

The following data for the given function 𝑓 𝑥 = 𝑥 4 is given

𝑥 0.4 0.6 0.8

𝑓(𝑥) 0.0256 0.1296 0.4096

242
find 𝑓 ′ 0.8 and 𝑓 ′′ 0.8 using quadratic interpolation. Compare the
SPACE OF HINTS
exact solution obtain the bound on the truncation errors.

Solution:
Here intervals are equal.

∴ Using uniform-nodal quadratic interpolation method. We have,

1
𝑓 ′ 𝑥2 = 𝑓 − 4𝑓1 + 3𝑓2
2𝑕 0
Here 𝑕 = 0.2,
1
𝑓 ′ 0.8 = 0.0256 − 4(0.1296) + 3(0.4096) = 1.84
2 × 0.2
1
Similarly, 𝑓 ′′ 𝑥2 = 𝑓0 − 2𝑓1 + 𝑓2
𝑕2

1
= 0.0256 − 2(0.1296) + 0.4096 = 4.4
0.04
To find the exact solution:
Given 𝑓 𝑥 = 𝑥 4

𝑓 ′ 𝑥 = 4𝑥 3 , 𝑓 ′′ 𝑥 = 12𝑥 2 , 𝑓 ′′′ 𝑥 = 24𝑥

𝑓 ′ 0.8 = 4 × (0.8)3 = 2.048

𝑓 ′′ 0.8 = 12 × (0.8)2 = 7.68

To find the error term:

′ −𝑕2 ′′′
𝐸2 𝑥2 = 𝑓 𝜂2
3
𝐸2 ′′ 𝑥2 = 𝑕𝑓 ′′′ 𝜖3

𝑀3 = max 𝑓 ′′′ (𝑥) = max 24𝑥 = 19.2

0.4<𝑥<0.8 0.4<𝑥<0.8

2
′ 0.2
𝐸2 𝑥2 ≤ 19.2 = 0.256
3

𝐸2 ′′ 𝑥2 ≤ 0.2 19.2 = 3.84

243
 Example: 4.1.3
SPACE FOR HINTS a) Derive the formulas for the first derivative of 𝑦 = 𝑓 𝑥 of
𝑂 𝑕2 using
i. Forward difference approximation
ii. Backward difference approximation
iii. Central difference approximation
b) When 𝑓 𝑥 = sin 𝑥estimate 𝑓 ′ 𝜋 4 with 𝑕 = 𝜋 4using the
above formulas. Obtain the bounds on the truncation error and
compare with the exact solution.

Solution:
(a). (i) Newton‟s forward difference formula is given by,
𝑢(𝑢 − 1) 2 (𝑥 − 𝑥0 )
𝑓 𝑥 = 𝑓0 + 𝑢∆𝑓0 + ∆ 𝑓0 𝑤𝑕𝑒𝑟𝑒 𝑢 =
2 𝑕
𝜕𝑓 𝜕𝑢
𝑓′ 𝑥 = .
𝜕𝑢 𝜕𝑥
𝜕𝑢 1 𝜕𝑓 (2𝑢 − 1) 2
= , = ∆𝑓0 + ∆ 𝑓0
𝜕𝑥 𝑕 𝜕𝑢 2
1 (2𝑢 − 1) 2
∴ 𝑓′ 𝑥 = ∆𝑓0 + ∆ 𝑓0
𝑕 2

′ ′
𝑕2
𝐸 𝑥0 = 𝐸 𝑢=0 ≤ 𝑀3 , 𝑤𝑕𝑒𝑟𝑒 𝑀3 = max 𝑓 ′′′ 𝜉 .
3 𝑥 0 <𝜖<𝑥 2

ii) Newton‟s backward difference formula is given by,

𝑢(𝑢 + 1) 2 (𝑥 − 𝑥2 )
𝑓 𝑥 = 𝑓2 + 𝑢∇𝑓2 + ∆ 𝑓2 , 𝑤𝑕𝑒𝑟𝑒 𝑢 =
2 𝑕
𝜕𝑓 𝜕𝑢
𝑓′ 𝑥 = .
𝜕𝑢 𝜕𝑥
𝜕𝑢 1 𝜕𝑓 (2𝑢 + 1) 2
= , = ∇𝑓2 + ∆ 𝑓2
𝜕𝑥 𝑕 𝜕𝑢 2
1 (2𝑢 + 1) 2
∴ 𝑓′ 𝑥 = ∇𝑓2 + ∇ 𝑓2
𝑕 2

244
 𝑕2
𝐸 ′ 𝑥2 = 𝐸′ 𝑢 = 0 ≤ 𝑀, 𝑤𝑕𝑒𝑟𝑒 𝑀3 = max 𝑓 ′′′ 𝜉 SPACE OF HINTS
3 3 𝑥 0 <𝜖<𝑥 2

iii) Central difference formula is,

𝑢 (𝑥 − 𝑥2 )
𝑓 𝑥 = 𝑓0 + 𝛿𝑓1 + 𝛿𝑓−1 𝑤𝑕𝑒𝑟𝑒 𝑢 =
2 2 2 𝑕
Here 𝑥0 is the middle value
𝜕𝑓 𝜕𝑢 𝜕𝑢 1
𝑓′ 𝑥 = . , =
𝜕𝑢 𝜕𝑥 𝜕𝑥 𝑕
𝜕𝑓 1
= 2 𝛿𝑓1 + 𝛿𝑓−1
𝜕𝑢 2 2

= 1 2 (𝑓1 − 𝑓0 ) + (𝑓0 − 𝑓−1 )

= 1 2 𝑓1 − 𝑓−1

1 1
∴ 𝑓 ′ 𝑥 = 1 2 𝑓1 − 𝑓−1 = 𝑓 − 𝑓−1
𝑕 2𝑕 1

𝑕2
The error term is, 𝐸 ′ 𝑥 = 𝑀3
6

(b). 𝑓 𝑥 = sin 𝑥

Forward difference table

𝑥 𝑓(𝑥) ∆ ∆2

𝜋 0.7071
4
0.1589
𝜋 0.8660 −0.059
3
0.0999
0.9659
5𝜋
12

1 (2𝑢 − 1) 2
𝑓′ 𝑥 = ∆𝑓0 + ∆ 𝑓0
𝑕 2

245
 12 0.059
𝑓′ 𝜋 4 = 0.1589 +
SPACE FOR HINTS 𝜋 2
𝑓 ′ 𝜋 4 = 0.71963

Now,

′
𝑕2
𝐸 𝑥 ≤ 𝑀3 , 𝑤𝑕𝑒𝑟𝑒 𝑀3 = max 𝑓 ′′′ 𝜉
3 𝜋
4 <𝜖<
5𝜋
12

∴ 𝑀3 = max − cos 𝑥
𝜋 5𝜋
4 <𝜉< 12

𝑀3 = cos 𝜋 4 = 0.70711

𝜋 2

𝐸′ 𝑥 ≤ 12 × 0.70711 = 0.0162
3
Backward difference table

𝑥 𝑓 𝑥 ∇ ∇2
𝜋 0.2588
12
0.2412
𝜋 0.5 -0.0341
6
0.2071

𝜋 0.7071
4
1 2𝑢 + 1 2
𝑓′ 𝑥 = ∇𝑓2 + ∇ 𝑓2
𝑕 2

(𝑥 − 𝑥0 ) 𝑥2 − 𝑥2
𝑢= = =0
𝑕 𝑕
12 1
𝑓′ 𝜋 4 = 0.2071 + −0.0341
𝜋 2
= 0.72593

𝑓 ′ 𝜋 4 ≅ 0.7260

𝑕2
𝐸′ 𝑥 ≤ 𝑀 , 𝑤𝑕𝑒𝑟𝑒 𝑀3 = 𝜋 max𝜋 − cos 𝑥
3 3 12 <𝜉 < 4

246
 𝜋 2

= 12 0.96593 ; 𝑀3 = cos 𝜋 12 = 0.96593

SPACE OF HINTS
3
𝐸′ 𝑥 = 0.022

Central difference table

𝑥 𝑓(𝑥) 𝛿

𝜋 6 0.5

0.20711

𝜋 4 0.7071

0.1589

𝜋 3 0.8660
1
𝑓′ 𝜋 4 = 𝑓 − 𝑓−1
2𝑕 1
1
= 0.8660 − 0.5 = 0.6991
2 × 𝜋 12

′
𝑕2
𝐸 𝑥 ≤ 𝑀3 , 𝑤𝑕𝑒𝑟𝑒 𝑀3 = 𝜋 max𝜋 − cos 𝑥 = 0.86603
3 6<𝜉 < 3

𝜋 2

= 12 0.86603
3
𝐸′ 𝑥 = 0.00989 ≅ 0.0099

The exact solution is, 𝑓 𝑥 = sin 𝑥 , 𝑓 ′ 𝑥 = cos 𝑥

𝑓 ′ 𝜋 4 = cos 𝜋 4 = 0.70711

Methods based on undetermined coefficient

Example: 4.1.4
A differentiation rule of the form

𝑕𝑓 ′ (𝑥2 ) = 𝛼0 𝑓(𝑥0 ) + 𝛼1 𝑓(𝑥1 ) + 𝛼2 𝑓(𝑥3 ) + 𝛼3 𝑓(𝑥4 )

where 𝑥𝑗 = 𝑥0 + 𝑗𝑕, 𝑗 = 0, 1, 2, 3, 4is given.

247
 Determine the values of 𝛼0 , 𝛼1 , 𝛼2 𝑎𝑛𝑑 𝛼4 so that the rule is exact for a
SPACE FOR HINTS polynomial of degree 4.

i. Find the error term

ii. Obtain the expression for the round-off error in calculating
𝑓 ′ 𝑥2 .
iii. Calculate 𝑓 ′ 𝑥2 using five place values of 𝑓 𝑥 = sin 𝑥with

𝑕 = 0.1 compare the result with the exact value cos(0.3)

Solution:

𝑕𝑓 ′ 𝑥2 = 𝛼0 𝑓 𝑥0 + 𝛼1 𝑓 𝑥1 + 𝛼2 𝑓 𝑥3 + 𝛼3 𝑓 𝑥4 (1)

Given 𝑥𝑗 = 𝑥0 + 𝑗𝑕, 𝑗 = 0, 1, 2, 3, 4

Consider the intervals 𝑥0 , 𝑥1 , 𝑥2 , 𝑥3 , 𝑥4

∴ We have 𝑥0 = 𝑥2 − 2𝑕 , 𝑥1 = 𝑥1 − 𝑕

𝑥3 = 𝑥2 + 𝑕 , 𝑥4 = 𝑥1 + 2𝑕

∴ 1 ⟹ 𝑕𝑓 ′ 𝑥2 = 𝛼0 𝑓 𝑥2 − 2𝑕 + 𝛼1 𝑓 𝑥1 − 𝑕 + 𝛼2 𝑓 𝑥2 + 𝑕

+𝛼3 𝑓 𝑥1 + 2𝑕

By Tayler series expansion, we have

𝑓 ′ 𝑥0 𝑓 ′′ 𝑥0 2
𝑓 𝑥 = 𝑓 𝑥0 + 𝑥 − 𝑥0 + 𝑥 − 𝑥0 +⋯
1! 2!
𝑕𝑒𝑟𝑒 𝑥0 = 𝑥2 ,

𝑥 = 𝑥2 − 2𝑕

′
𝑓 ′ 𝑥2 𝑓 ′′ 𝑥2 2
∴ 𝑕𝑓 𝑥2 = 𝛼0 𝑓 𝑥2 + −2𝑕 + −2𝑕 +⋯
1! 2!
𝑓 ′′ 𝑥2
+ 𝛼1 𝑓 𝑥2 + 𝑓 ′ 𝑥2 −𝑕 + −𝑕 2
+⋯
2!

′
𝑓 ′′ 𝑥2 2
+𝛼2 𝑓 𝑥2 + 𝑓 𝑥2 𝑕 + 𝑕 +⋯
2!
𝑓 ′′ 𝑥2
+𝛼3 𝑓 𝑥2 + 𝑓 ′ 𝑥2 2𝑕 + 2𝑕 2
+⋯
2!

248
 = 𝛼0 + 𝛼1 + 𝛼2 + 𝛼3 𝑓 𝑥2
SPACE OF HINTS
′
+ 𝑓 𝑥2 −2𝑕𝛼0 − 𝑕𝛼1 + 𝑕𝛼2 + 2𝑕𝛼3
𝑓 ′′ 𝑥0
+ 4𝑕2 𝛼0 + 𝑕2 𝛼1 + 𝑕2 𝛼2 + 4𝑕2 𝛼3 + ⋯
2

𝑕𝑓 ′ 𝑥2 = 𝛼0 + 𝛼1 + 𝛼2 + 𝛼3 𝑓 𝑥2
+ 𝑕𝑓 ′ 𝑥2 −2𝛼0 − 𝛼1 + 𝛼2 + 2𝛼3
𝑓 ′′ 𝑥0 2
+ 𝑕 4𝛼0 + 𝛼1 + 𝛼2 + 4𝛼3 + ⋯
2

Comparing constants, we get 𝛼0 + 𝛼1 + 𝛼2 + 𝛼3 = 0

Comparing the coefficient of 𝑕𝑓 ′ 𝑥2 on both sides, we get

−2𝛼0 − 𝛼1 + 𝛼2 + 2𝛼3 = 0
𝑓 ′′ 𝑥 0
Comparing the coefficient of 𝑕2 on both sides, we get
2

4𝛼0 + 𝛼1 + 𝛼2 + 4𝛼3 = 0
𝑓 ′′′ 𝑥 0
Comparing the coefficient of 𝑕3 on both sides, we get
3!

−8𝛼0 − 𝛼1 + 𝛼2 + 8𝛼3 = 0
𝐵𝑖
Using Cramer‟s rule, we find 𝑥𝑖 = , 𝑖 = 0,1,2,3.
𝐴𝑖

The given system of equation can be written as, 𝐴𝑋 = 𝐵

1 1 1 1 𝛼0 0
−2 −1 1 2 𝛼1 1
𝐴= , 𝑋= , 𝐵=
4 1 1 4 𝛼2 0
−8 −1 1 8 𝛼3 0
−1 1 2 −2 1 2 −2 1 2
𝐴 =1 1 1 4 − 4 1 4 + 4 1 4
−1 1 8 −8 1 8 −8 1 8
−2 −1 1
− 4 1 1
−8 −1 1

= −12 + 48 + 48 − 12

𝐴 = 72

249
 0 1 1 1
1 1 1
1 −1 1 2
SPACE FOR HINTS Now, 𝐵1 = =0−1 1 1 4
0 1 1 4
−1 1 8
0 −1 1 8
= −4 − 12 − 2 = 6
1 0 1 1
−2 1 1 2
𝐵2 =
4 0 1 4
−8 0 1 8
1 1 2 −2 1 2 −2 1 1
=1 0 1 4 −0+ 4 0 4 +1 4 0 1
0 1 8 −8 0 8 −8 0 1

= 1 8 − 4 + −2 0 + 1 32 + 32 + 2 0 + 2 6 − 0
+1 4−8 +0

𝐵2 = −48

𝐵3 = 48

𝐵4 = −6
𝐵1 6 1 𝐵2 −48 −2
⟹ 𝛼0 = = = 𝛼1 = = =
𝐴 72 12 𝐴 72 3
𝐵3 48 2 𝐵4 −6 −1
𝛼2 = = = 𝛼3 = = =
𝐴 72 3 𝐴 72 12
1 −2 2 1
∴ 𝑕𝑓 ′ 𝑥2 = 𝑓 𝑥0 + 𝑓 𝑥1 + 𝑓 𝑥2 − 𝑓(𝑥3 )
12 3 3 12
1 1 −2 2 1
∴ 𝑓 ′ 𝑥2 = 𝑓 𝑥0 + 𝑓 𝑥1 + 𝑓 𝑥2 − 𝑓(𝑥3 )
𝑕 12 3 3 12
i. The error term becomes
5 5
𝛼0 −2𝑕 𝛼1 −𝑕 𝛼2 𝑕5 𝛼3 2𝑕 5
𝐸= + + + 𝑓5 𝜉 ,
5! 5! 5! 5!
𝑥2 − 2𝑕 ≤ 𝑥 ≤ 𝑥2 + 2𝑕
𝑕5
= [−32𝛼0 − 𝛼1 + 𝛼2 + 32𝛼3 ]𝑓 5 𝜉
120
ii. To find round of error in 𝑓 ′ (𝑥2 )

Let 𝜉0 , 𝜉1 , 𝜉2 , 𝜉3 be the round of errors in

250
𝑓 𝑥2 − 2𝑕 , 𝑓 𝑥2 − 𝑕 , 𝑓 𝑥2 + 𝑕 , 𝑓 𝑥2 + 2𝑕 respectively.
SPACE OF HINTS
1 1 2 2 1
Then 𝑅𝐸 ≤ 𝜖 − 𝜖1 + 𝜖2 − 𝜖
𝑕 12 0 3 3 12 3

𝜉 = max 𝜉0 , 𝜉1 , 𝜉2 , 𝜉3
1 1 2 2 1
𝑅𝐸 ≤ 𝜉− 𝜉+ 𝜉− 𝜉
𝑕 12 3 3 12
1 1 2 2 1
≤ 𝜉 + 𝜉 + 𝜉 + 𝜉
𝑕 12 3 3 12
1 2 4
≤ 𝜉 + 𝜉
𝑕 12 3
1 2 + 16 𝜉
≤
𝑕 12
1 3
≤ 𝜉
𝑕 2
1 1
iii. 𝑓 ′ 𝑥2 = 𝑓 𝑥0 − 8𝑓 𝑥1 + 8𝑓 𝑥3 − 𝑓(𝑥4 )
𝑕 12

Let 𝑥0 = 0.1, 𝑥1 = 0.2, 𝑥2 = 0.3, 𝑥3 = 0.4, 𝑥4 = 0.5

1
𝑓 ′ 0.3 = 𝑓 𝑥0 − 8𝑓 𝑥1 + 8𝑓 𝑥3 − 𝑓(𝑥4 )
12

1
= sin 0.1 − 8 sin 0.2 + 8 sin 0.4 − sin 0.5
12 × 0.1
1
= 0.09983 − 1.5894 + 3.11535 − 0.4794
1.2
= 0.95532

Exact value is 𝑓 𝑥 = sin 𝑥 , 𝑓 ′ 𝑥 = cos 𝑥 = cos 0.3 = 0.95534

4.2 Extrapolation methods

To obtain differentiation methods of higher order, we require a
large number of tabular points and thus a large number of function
evaluations.Consequently there is a possibility that the roundoff errors
may increase so much that the numerical results may become
meaningless.

251
 Generally it is possible to obtain highly accurate results by
SPACE FOR HINTS combining the computed values obtained by using a certain method with
different step sizes.

Let 𝑔 𝑕 denotes the approximate value of 𝑔, obtained by using a

method of order 𝑝, with step length 𝑕 and 𝑔(𝑞𝑕) denote the value of 𝑔
obtained by using the same method of order 𝑝, with step length 𝑞𝑕, we
have𝑔 𝑕 = 𝑔 + 𝑐𝑕𝑝 + 𝑂(𝑕𝑝+1 )

𝑔 𝑞𝑕 = 𝑔 + 𝑐𝑞𝑝 𝑕𝑝 + 𝑂(𝑕𝑝+1 )

Eliminating 𝑐 from the above equation, we get

𝑞𝑝 𝑔 𝑕 − 𝑔(𝑞𝑕)
𝑔= + 𝑂(𝑕𝑝+1 )
𝑞𝑝 − 1
Hence, we obtain

1
𝑞𝑝 𝑔 𝑕 − 𝑔 𝑞𝑕
𝑔 𝑕 = 𝑝
= 𝑔 + 𝑂(𝑕𝑝+1 )
𝑞 −1

Which is of order 𝑝 + 1. This technique of combining two computed

values obtained by using the same method with two different step sizes,
to obtain a higher order method is called the Extrapolation method or
Richardson’s extrapolation.

Richardson’s extrapolation
The formula for Richardson‟s extrapolation is,

4𝑚 𝑔 𝑚 −1 𝑕 𝑚 −1
2 −𝑔 𝑕
𝑚
𝑔 𝑕 = , 𝑚 = 1,2, … 𝑤𝑕𝑒𝑟𝑒 𝑔(0) 𝑕
4𝑚 −1
= 𝑔(𝑕)

Extrapolation table:

𝑜𝑟𝑑𝑒𝑟 𝑆𝑒𝑐𝑜𝑛𝑑 𝐹𝑜𝑢𝑟𝑡𝑕 𝑆𝑖𝑥𝑡𝑕 𝐸𝑖𝑔𝑡𝑕

252
 𝑕 𝑔 𝑕
SPACE OF HINTS
′
𝑔 (𝑕)
𝑕 𝑔(𝑕 2) 𝑔(2) (𝑕)
2
𝑔(1) (𝑕 2) 𝑔(3) (𝑕)

𝑔(2) (𝑕 2)
𝑕 𝑔(𝑕 22 )
22
𝑔(1) (𝑕 22 )

𝑕 𝑔(𝑕 23 )
23

Example: 4.2.1
The following table of values is given

𝑥 -1 1 2 3 4 5 7

𝑓(𝑥) 1 1 16 81 256 625 2401

𝑓 𝑥 2 −𝑓(𝑥 0 )
Using the formula 𝑓 ′ 𝑥1 = and the Richardson extrapolation.
2𝑕

Find 𝑓 ′ 3 .

Solution:
𝑓 𝑥 2 −𝑓(𝑥 0 )
Let 𝑓 ′ 𝑥1 =
2𝑕

Using Tayler series expansion,

𝑓 ′ 𝑥0 𝑓 ′′ 𝑥0 2
𝑓 𝑥 = 𝑓 𝑥0 + 𝑥 − 𝑥0 + 𝑥 − 𝑥0 +⋯ (1)
1! 2!
Put 𝑥 = 𝑥0 , 𝑥0 = 𝑥1 𝑖𝑛 1 ,

∴ 1 𝑏𝑒𝑐𝑜𝑚𝑒𝑠
𝑓 ′ 𝑥1 𝑓 ′′ 𝑥1 2
𝑓 𝑥0 = 𝑓 𝑥1 + 𝑥0 − 𝑥1 + 𝑥0 − 𝑥1 +⋯
1! 2!

′
𝑓 ′′ 𝑥1 2
= 𝑓 𝑥1 + 𝑓 𝑥1 (−𝑕) + −𝑕 +⋯
2

253
 𝑕2 ′′
𝑓 𝑥0 = 𝑓 𝑥1 − 𝑕𝑓 ′ 𝑥1 + 𝑓 𝑥1 + ⋯ (2)
SPACE FOR HINTS 2
Put 𝑥 = 𝑥2 , 𝑥0 = 𝑥1 𝑖𝑛 1 ,
𝑓 ′ 𝑥1 𝑓 ′′ 𝑥1 2
𝑓 𝑥2 = 𝑓 𝑥1 + 𝑥2 − 𝑥1 + 𝑥2 − 𝑥1 +⋯
1! 2!

′
𝑓 ′′ 𝑥1 2
= 𝑓 𝑥1 + 𝑓 𝑥1 𝑕 + 𝑕 +⋯
2

′
𝑕2 ′′
𝑓 𝑥2 = 𝑓 𝑥1 + 𝑕𝑓 𝑥1 + 𝑓 𝑥1 + ⋯ (3)
2
3 − 2 ⟹ 𝑓 𝑥2 − 𝑓 𝑥0
2𝑕3 ′′′ 2𝑕5 5
= 2𝑕𝑓 ′ 𝑥1 + 𝑓 𝑥1 + 𝑓 𝑥1 + ⋯
3! 5!

′
𝑕2 ′′′ 𝑕4 5
= 2𝑕 𝑓 𝑥1 + 𝑓 𝑥1 + 𝑓 𝑥1 + ⋯
3! 5!

𝑓 𝑥2 − 𝑓 𝑥0 𝑕2 𝑕4 5
⟹ = 𝑓 ′ 𝑥1 + 𝑓 ′′′ 𝑥1 + 𝑓 𝑥1 + ⋯
2𝑕 6 120
Extrapolation table becomes,

𝑕\𝑜𝑟𝑑𝑒𝑟 𝑆𝑒𝑐𝑜𝑛𝑑 𝑓𝑜𝑢𝑟𝑡𝑕 𝑠𝑖𝑥𝑡𝑕

𝑕=4 𝑔 𝑕 = 𝑓(4)

𝑔′ (𝑕)
𝑕 𝑔(2) (𝑕)
2=2 𝑔 𝑕 2
𝑔(1) (𝑕 2)
= 𝑔(2)
𝑕
22 = 1

𝑔 𝑕 22

= 𝑔(1)

Taking 𝑕 = 4, the above table becomes, we know that

254
 𝑓 𝑥2 − 𝑓 𝑥0
𝑔 𝑥 = SPACE OF HINTS
2𝑕
Also

𝑥0 = −1, 𝑥1 = 3, 𝑥3 = 7, 𝑓 𝑥0 = 1, 𝑓 𝑥1 = 81, 𝑓 𝑥2 = 2401

2401 − 1
𝑕 = 4, 𝑔 4 = = 300
8
Take 𝑕 = 2, 𝑥0 = 1, 𝑥1 = 3, 𝑥2 = 5, 𝑓 𝑥0 = 1,

𝑓 𝑥1 = 81, 𝑓 𝑥2 = 625
625 − 1
𝑕 = 2, 𝑔 2 = = 156
4
Take 𝑕 = 1, 𝑥0 = 2, 𝑥1 = 3, 𝑥2 = 4, 𝑓 𝑥0 = 16,

𝑓 𝑥1 = 81, 𝑓 𝑥2 = 256
256 − 1
𝑕 = 1, 𝑔 1 = = 120
2
Richardson‟s extrapolation formula for higher derivatives is,

4𝑚 𝑔 𝑚 −1 𝑕 𝑚 −1
2 −𝑔 𝑕
𝑚
𝑔 𝑕 = ,𝑚
4𝑚 − 1
= 1,2, … 4

𝑤𝑕𝑒𝑟𝑒 𝑔0 𝑕 = 𝑔 𝑕

Put 𝑚 = 1in (4), we get,

4𝑔 0 𝑕 0
4𝑔 𝑕 2 − 𝑔 𝑕
2 −𝑔 𝑕
𝑔′ 𝑕 = =
4−1 3
4𝑔 2 − 𝑔(4) 4 × 156 − 300
𝑕 = 4, 𝑔′ 4 = = = 108
3 3
4𝑔 1 − 𝑔(2) 4 × 120 − 156
𝑕 = 2, 𝑔′ 2 = = = 108
3 3
Put 𝑚 = 2 in (4), we get

255
 ′′ 2
42 𝑔 1
2 −𝑔 1
4 16 108 − 108
𝑔 4 =𝑔 4 = = = 108
SPACE FOR HINTS 42 − 1 15
The extrapolation table becomes

𝑕\𝑜𝑟𝑑𝑒𝑟 𝑠𝑒𝑐𝑜𝑛𝑑 𝑓𝑜𝑢𝑟𝑡𝑕 𝑠𝑖𝑥𝑡𝑕

4 300

108

2 156 108

108
1 120

From the given data 𝑓 𝑥 = 𝑥 4

∴ 𝑓 ′ 𝑥 = 4𝑥 3

𝑓 ′ 3 = 4 × 33 = 108

∴ The exact value is obtained at the second column

Example: 4.2.2
−𝑦 𝑥 + 2𝑕 + 4𝑦 𝑥 + 𝑕 − 3𝑦(𝑥)
Define𝑆 𝑕 =
2𝑕
a) Show that 𝑦 ′ 𝑥 − 𝑆 𝑕 = 𝑐1 𝑕2 + 𝑐2 𝑕3 + 𝑐1 𝑕4 + ⋯ and
state𝑐1 .
b) Calculate 𝑦 ′ (0.398) as accurately as possible using the table
below and with the aid of the approximation 𝑆 𝑕 . Give an error
estimate (the values in the table are correctly rounded).

𝑥 0.398 0.399 0.400 0.401 0.402

𝑓(𝑥) 0.408591 0.409671 0.410752 0.411834 0.412915

Solution:
Expanding 𝑦(𝑥 + 𝑕) and 𝑦(𝑥 + 2𝑕) in Taylor series about the
point 𝑥,

We obtain from the given formula

256
 𝑕2 ′′′ 𝑕3 7 2 𝑣
𝑆 𝑕 = 𝑦′ 𝑥 − 𝑦 𝑥 − 𝑦 𝑖𝑣 𝑥 − 𝑕 𝑦 𝑥 −⋯ SPACE OF HINTS
3 4 603
= 𝑦 ′ 𝑥 − 𝑐1 𝑕2 − 𝑐2 𝑕3 − 𝑐1 𝑕4 + ⋯

⟹ 𝑦 ′ 𝑥 − 𝑆 𝑕 = 𝑐1 𝑕2 − 𝑐2 𝑕3 − 𝑐1 𝑕4 + ⋯
𝑦 ′′′ 𝑥
where 𝑐1 = 3

(b). Taking 𝑕 = 0.001, we obtain, using the given formula and the data
−𝑦 0.400 + 4𝑦 0.399 − 𝑦(0.398)
𝑦 ′ 0.398 ≈
2(0.001)

= 1.0795

The error in the formula is given by

𝑕2
Error = 𝑦 ′′′ 𝑥0
3

1 3
≈ ∆ 𝑦0
3𝑕
1
= (𝑦 − 3𝑦2 + 3𝑦1 − 𝑦0 )
3𝑕 3
1
= 𝑦 0.401 − 3𝑦 0.400 + 3𝑦 0.399 − 𝑦 0.398 =0
3(0.001)
Taking the next term in the error formula, we have
𝑕3
Error = 𝑦 𝑖𝑣 𝑥0
4

1 4
≈ ∆ 𝑦0
4𝑕
1
= (𝑦 − 4𝑦3 + 6𝑦2 −4𝑦1 − 𝑦0 )
4𝑕 4
1
= 𝑦 0.402 − 4𝑦 0.401 + 6𝑦 0.400 − 4𝑦 0.399
0.004
+ 𝑦(0.398)

= −0.0005.

4.3: Partial differentiation

257
 Let the values of the function 𝑓 𝑥, 𝑦 be given at a set of points
SPACE FOR HINTS 𝑥𝑖 , 𝑦𝑗 in the plane with spacing 𝑕 and 𝑘 in 𝑥 and 𝑦 directions
respectively, we have

𝑥𝑖 = 𝑥0 + 𝑖𝑕, 𝑦𝑗 = 𝑦0 + 𝑗𝑘, 𝑖, 𝑗 = 1, 2, …

Then the derivative is given by,

𝑓 𝑖+1,𝑗 −𝑓 𝑖,𝑗
+ 𝑂(𝑕)
𝑕
𝜕𝑓 𝑓 𝑖,𝑗 −𝑓 𝑖−1,𝑗
= + 𝑂(𝑕) where 𝑓𝑖,𝑗 = 𝑓 𝑥𝑖 , 𝑦𝑗
𝜕𝑥 𝑥𝑖 , 𝑦 𝑗 𝑕
𝑓 𝑖+1,𝑗 −𝑓 𝑖−1,𝑗
+ 𝑂(𝑕2 )
2𝑕

Similarly we can write,

𝑓𝑖,𝑗 +1 − 𝑓𝑖,𝑗
+ 𝑂(𝑘)
𝑘
𝜕𝑓 𝑓𝑖,𝑗 − 𝑓𝑖,𝑗 −1
= + 𝑂(𝑘)
𝜕𝑥 𝑥𝑖 , 𝑦 𝑗 𝑘
𝑓𝑖,𝑗 +1 − 𝑓𝑖,𝑗 −1
+ 𝑂(𝑘 2 )
2𝑘
𝜕2 𝑓 1
= 𝑓 − 2𝑓𝑖,𝑗 + 𝑓𝑖+1,𝑗 + 𝑂(𝑕2 )
𝜕𝑥 2
𝑥𝑖 , 𝑦 𝑗
𝑕2 𝑖−1,𝑗

𝜕2 𝑓 1
= 𝑓 − 2𝑓𝑖,𝑗 + 𝑓𝑖,𝑗 +1 + 𝑂(𝑘 2 )
𝜕𝑦 2 𝑥𝑖 , 𝑦 𝑗
𝑘 2 𝑖,𝑗 −1

Example: 4.3.1
Find the Jacobian matrix for the system of equations

𝑓1 𝑥, 𝑦 = 𝑥 2 + 𝑦 2 − 𝑥 = 0, 𝑓2 𝑥, 𝑦 = 𝑥 2 − 𝑦 2 − 𝑦 = 0

At the point 1,1 using the methods

𝜕𝑓 𝑓 𝑖+1,𝑗 −𝑓 𝑖−1,𝑗 𝜕𝑓 𝑓 𝑖,𝑗 +1 −𝑓 𝑖,𝑗 −1
= = with 𝑕 = 𝑘 = 1.
𝜕𝑥 𝑥 𝑖 ,𝑦 𝑗 2𝑕 𝜕𝑦 𝑥 𝑖 ,𝑦 𝑗 2𝑘

Solution:

258
 The Jacobian matrix is given by,
SPACE OF HINTS
𝜕𝑓1 𝜕𝑓1
𝜕𝑥 𝜕𝑦
𝐽=
𝜕𝑓2 𝜕𝑓2
𝜕𝑥 𝜕𝑦
𝜕𝑓 𝑓 𝑖+1,𝑗 −𝑓 𝑖−1,𝑗
Given = and
𝜕𝑥 𝑥 𝑖 ,𝑦 𝑗 2𝑕

𝜕𝑓 𝑓𝑖,𝑗 +1 − 𝑓𝑖,𝑗 −1
=
𝜕𝑦 𝑥 𝑖 ,𝑦 𝑗
2𝑘

Now 𝑕 = 1,
𝜕𝑓1 𝑓1 (1 + 𝑕, 1) − 𝑓1 (1 − 𝑕, 1) 𝑓1 (2,1) − 𝑓1 (0, 1)
= =
𝜕𝑥 1,1 2𝑕 2
3−1
= =1
2
𝜕𝑓2 𝑓2 (2,1) − 𝑓2 (0.1) (4 − 1 − 1) + 2
= = =2
𝜕𝑥 (1,1) 2 2

Put 𝑘 = 1,
𝜕𝑓1 𝑓1 (1,2) − 𝑓1 (1,2) 4 − 0
= = =2
𝜕𝑦 (1,1)
2 2

𝜕𝑓2 𝑓2 (1,2) − 𝑓1 (1,0)

= = −3
𝜕𝑦 (1,1)
2

1 2
∴ The Jacobian matrix 𝐽 =
2 −3

4.4: Numerical Integration

Definition:
An integration method of the form
𝑛
𝑏
𝐼= 𝑤 𝑥 𝑓 𝑥 𝑑𝑥 ≅ 𝜆𝑘 𝑓𝑘 𝑤𝑕𝑒𝑟𝑒 𝑥𝑘 , 𝑘 = 0,1, … , 𝑛
𝑎 𝑘=0

259
 are called the abscissas or nodes distributed within the limits of
SPACE FOR HINTS integration 𝑎, 𝑏 and 𝜆𝑘 , 𝑘 = 0,1, … , 𝑛 are called the weights of the
integration rule or the quadratic formula. The above integration is said to
be of order 𝑝 , if it produces exact results 𝑅𝑛 = 0 for all polynomials
of degree less than or equal to 𝑝.

Methods Based on Interpolation:

Newton-Cotes Methods:
The newton-Cotes formula is
𝑛
𝑏
𝑤 𝑥 𝑓 𝑥 𝑑𝑥 = 𝜆𝑘 𝑓𝑘
𝑎 𝑘=0

𝑏−𝑎
where 𝑥𝑘 are equispaced with 𝑥0 = 𝑎, 𝑥𝑛 = 𝑏 with spacing 𝑕 = .
𝑛

Here𝑤 𝑥 = 1 and the weights 𝜆𝑘 ‟s are called Cotes numbers.

Trapezoidal rule
The formula for Trapezoidal rule is,
𝑏
𝑏−𝑎
𝐼= 𝑓 𝑥 𝑑𝑥 = 𝑓 𝑎 + 𝑓(𝑏)
𝑎 2

The truncation error is,

3
− 𝑏−𝑎
𝑅1 = 𝑓 ′′ 𝜂 , 𝜂 ∈ 0,1
12
𝑕3 ′′
(𝑖𝑒. )𝑅1 = − 𝑓 𝜂
12

Simpson’s rule:
The formula for Simpson‟s rule is,
𝑏
𝑏−𝑎 𝑎+𝑏
𝐼= 𝑓 𝑥 𝑑𝑥 = 𝑓 𝑎 + 4𝑓 + 𝑓(𝑏)
𝑎 6 2

The truncation error is,

260
 𝑕4 − 𝑏 − 𝑎 5
−𝑕5 𝑖𝑣
𝑅2 = = 𝑓 𝑖𝑣 𝜂 = 𝑓 𝜂 SPACE OF HINTS
3! 2880 90
𝒕𝒉
Simpson’s 𝟑 𝟖 rule:

𝑡𝑕
The formula for Simpson‟s 3 8 rule is,

𝑥3
3𝑕
𝑓 𝑥 𝑑𝑥 = 𝑓 𝑥0 + 3𝑓 𝑥1 + 3𝑓 𝑥2 + 𝑓(𝑥3 )
𝑥0 8

he truncation error is,

5
− 𝑏−𝑎 −𝑕5 𝑖𝑣
𝑅2 = 𝑓 𝑖𝑣 𝜂 = 𝑓 𝜂
2880 90
Example: 4.4.1
1 𝑑𝑥
Find the approximate value of 𝐼 = 0 1+𝑥
using

i. Trapezoidal rule
ii. Simpson‟s rule.

Obtain a bound for the errors. The exact value of 𝐼 = ln 2 = 0.693147 .

correct to six decimal places.

Solution:
1 𝑑𝑥
Given 𝐼 = 0 1+𝑥

1 −1 2
Here 𝑓 𝑥 = , 𝑓′ 𝑥 = , 𝑓 ′′ 𝑥 =
1+𝑥 1+𝑥 2 1+𝑥 3

i. The formula for trapezoidal rule is,

𝑏
𝑏−𝑎
𝐼= 𝑓 𝑥 𝑑𝑥 = 𝑓 𝑎 + 𝑓(𝑏)
𝑎 2

Here 𝑎 = 0, 𝑏 = 1

∴ 𝑓 𝑎 = 1, 𝑓 𝑏 =1 2

1−0 1
Hence 𝐼 = 1+ = 0.75
2 2

261
 − 𝑏−𝑎 3
The truncation error 𝑅1 = 𝑓 ′′ 𝜂
12
SPACE FOR HINTS
Now,
2
𝑀2 = max 𝑓 ′′ 𝜂 = max 3
=2
𝑎≤𝑥≤𝑏 0≤𝑥≤1 1+𝑥
3
− 1−0 2 1
∴ 𝑅1 = (2) = ≤ = 0.1667
12 12 6

ii. Simpson‟s rule

The formula for Simpson‟s rule is, 𝑓 0.5 = 2 3

𝑏−𝑎 𝑎+𝑏
𝐼= 𝑓 𝑎 + 4𝑓 + 𝑓(𝑏)
6 2
1−0
= 1 + 2.6667 + 0.5
6
𝐼 = 0.69444
The error term is,
5
− 𝑏−𝑎
𝑅2 = 𝑓 𝑖𝑣 𝜂
2880
24
𝑀4 = max 𝑓 𝑖𝑣 𝜂 = max 5
= 24
0≤𝑥≤1 0≤𝑥≤1 1+𝑥

1 5
𝑅2 = . 24 = 0.00833
2880
Exact value
1 𝑑𝑥
Given 𝑓(𝑥) = 0 1+𝑥

1
𝑑𝑥 1
𝐼= = log 1 + 𝑥 0
0 1+𝑥

𝐼 = 0.69315

Open Type Integration Rule

i. Mid-point rule:

262
 𝑛 = 2 , 𝑥0 = 𝑎, 𝑥0 + 𝑕, 𝑥0 + 2𝑕 = 𝑏
SPACE OF HINTS
𝑏 3
𝑕 ′′′
∴ 𝑓 𝑥 𝑑𝑥 = 2𝑕𝑓 𝑥0 + 𝑕 + 𝑓 (𝜉1 )
𝑎 3

ii. Two-point rule:

𝑛 = 3 , 𝑥0 = 𝑎, 𝑥0 + 𝑕, 𝑥0 + 2𝑕, 𝑥0 + 3𝑕 = 𝑏
𝑏
3𝑕 3
∴ 𝑓 𝑥 𝑑𝑥 = 𝑓 𝑥0 + 𝑕 + 𝑓(𝑥0 + 2𝑕) + 𝑕3 𝑓 ′′ (𝜉1 )
𝑎 2 4
iii. Three-point rule:

𝑛 = 3 , 𝑥0 = 𝑎, 𝑥0 + 𝑕, 𝑥0 + 2𝑕, 𝑥0 + 3𝑕 = 𝑏
𝑏
4𝑕
∴ 𝑓 𝑥 𝑑𝑥 = 2𝑓 𝑥0 + 𝑕 − 𝑓 𝑥0 + 2𝑕 + 2𝑓 𝑥0 + 3𝑕
𝑎 3
14 5 𝑖𝑣
+ 𝑕 𝑓 (𝜉3 )
45
where 𝑎 < 𝜉1 , 𝜉2 , 𝜉3 < 𝑏

Example: 4.4.2
1 sin 𝑥
Find the approximate value of 𝐼 = 0 𝑥
𝑑𝑥 using

i. Mid-point rule
ii. Two-point open type rule
Solution:
1 sin 𝑥
Given 𝐼 = 0 𝑥
𝑑𝑥
sin 𝑥 𝑏−𝑎 1
Here 𝑓 𝑥 = , 𝑎 = 0, 𝑏 = 1, 𝑕 = = 𝑥0 = 0,
𝑥 2 2

1
𝑥1 = , 𝑥2 = 1
2
1 sin 1 2
Now 𝑓 𝑥0 + 𝑕 = 𝑓 𝑥1 = 𝑓 = 1 = 0.9589
2 2

i. Mid-point rule is,

263
 𝑏
𝑕3 ′
𝑓 𝑥 𝑑𝑥 = 2𝑕𝑓 𝑥0 + 𝑕 + 𝑓 (𝜖1 )
SPACE FOR HINTS 𝑎 3

=2 1 2 0.9589 + 𝑒𝑟𝑟𝑜𝑟

= 0.9589
ii. Two-point rule is,
𝑏
3𝑕 3
𝑓 𝑥 𝑑𝑥 = 𝑓 𝑥0 + 𝑕 + 𝑓(𝑥0 + 2𝑕) + 𝑕3 𝑓 ′′ (𝜖2 )
𝑎 2 4
𝑏−𝑎 1 1 2
Here 𝑕 = = , 𝑛 = 3, 𝑥0 = 0, 𝑥1 = , 𝑥2 = , 𝑥3 = 1
𝑛 3 3 3

1
3𝑕
𝐼= 𝑓 𝑥 𝑑𝑥 = 𝑓 1 3 +𝑓 2 3
0 8

1
= 0.9816 + 0.92756
2

𝐼 = 0.9546

Example: 4.4.3
Find the reminder of the Simpson three-eighth rule
𝑥3
3𝑕
𝑓 𝑥 𝑑𝑥 = 𝑓 𝑥0 + 3𝑓 𝑥1 + 3𝑓 𝑥2 + 𝑓(𝑥3 )
𝑥0 8

For equally spaced points𝑥𝑖 = 𝑥0 + 𝑖𝑕, 𝑖 = 1,2,3. Use this rule to

1 𝑑𝑥
approximate the value of the original 𝐼 = 0 1+𝑥
. Also, find a bound on

the error.

Solution:

We can show that the rule is exact for 𝑓 𝑥 = 𝑥 0 , 𝑥 1 , 𝑥 2 , 𝑥 3 . Now

consider

𝑓 𝑥 = 𝑥 0 = 1. We get,
𝑥3
3𝑕
𝑑𝑥 = 𝑥3 − 𝑥0 = × 1 + 3 + 3 + 1 = 3𝑕
𝑥0 8

264
 𝐶
The error term is given by, 𝐸 = 𝑓 𝑖𝑣 (𝜂)
4! SPACE OF HINTS

The error constant is given by,

𝑥3
3𝑕
𝐶= 𝑥 4 𝑑𝑥 − 𝑥 4 + 3(𝑥1 )4 + 3(𝑥2 )4 + 𝑥3 4
𝑥0 8 0
𝑥3
𝑥5 3𝑕
= − 𝑥 4 + 3(𝑥1 )4 + 3(𝑥2 )4 + 𝑥3 4
5 𝑥0
8 0

𝑥3 5 𝑥0 5 3𝑕
= − − 𝑥 4 + 3(𝑥0 + 𝑕)4 + 3(𝑥0 + 2𝑕)4 + (𝑥0 + 3𝑕)4
5 5 8 0
(𝑥0 + 3𝑕)5 𝑥0 5
= −
5 5
3𝑕
− 𝑥0 4 + 3 𝑥0 4 + 4𝑥0 3 𝑕 + 4𝐶2 𝑥0 2 𝑕2 + 4𝐶3 𝑥0 𝑕3 + 𝑕4
8
+3 𝑥0 4 + 4𝑥0 3 (2𝑕) + 4𝐶2 𝑥0 2 (2𝑕)2 + 4𝐶3 𝑥0 (2𝑕)3 + (2𝑕)4

+ 𝑥0 4 + 4𝑥0 3 (3𝑕) + 4𝐶2 𝑥0 2 (3𝑕)2 + 4𝐶3 𝑥0 (3𝑕)3 + (3𝑕)4 ]

1 5
= 𝑥 + 5𝐶1 𝑥0 4 3𝑕 + 5𝐶2 𝑥0 3 3𝑕 2
+ 5𝐶3 𝑥0 2 3𝑕 3
5 0
4 5
𝑥0 5
+ 5𝐶4 𝑥0 3𝑕 + 5𝐶5 3𝑕 −
5
3𝑕
− 𝑥 4 + 3 𝑥0 4 + 4𝑥0 3 𝑕 + 4𝐶2 𝑥0 2 𝑕2 + 4𝐶3 𝑥0 𝑕3 + 𝑕4
8 0
+3 𝑥0 4 + 4𝑥0 3 (2𝑕) + 4𝐶2 𝑥0 2 (2𝑕)2 + 4𝐶3 𝑥0 (2𝑕)3 + (2𝑕)4

+ 𝑥0 4 + 4𝑥0 3 (3𝑕) + 4𝐶2 𝑥0 2 (3𝑕)2 + 4𝐶3 𝑥0 (3𝑕)3 + (3𝑕)4

265
 5 3𝑕 9𝑕 9𝑕 3𝑕
= 𝑥0 4 (3𝑕) − − − −
SPACE FOR HINTS 5 8 8 8 8
3
10 2
36𝑕2 72𝑕2 36𝑕2
+ 𝑥0 9𝑕 − − −
5 8 8 8
10(27𝑕3 ) 54𝑕3 216𝑕3 162𝑕3
+ 𝑥0 2 − − −
5 8 8 8
36𝑕4 288𝑕4 324𝑕4
4
243𝑕5 9𝑕5
+ 𝑥0 81𝑕 − − − + −
8 8 8 5 8
144𝑕5 243𝑕5
− −
8 8

−36
= 𝑥0 4 0 + 𝑥0 3 0 + 𝑥0 2 0 + 𝑥0 0 + 𝑕5
40
−9 5
𝐶= 𝑕
10
𝐶
Hence the error term is 𝐸 = 𝑓 𝑖𝑣 𝜂
4!

9 𝑕5 𝑖𝑣
=− 𝑓 𝜂
10 24

−3𝑕5 𝑖𝑣
𝐸= 𝑓 𝜂
80
Given 𝑥𝑖 = 𝑥0 + 𝑖𝑕, 𝑖 = 1,2,3.
𝑏−𝑎 1−0 1
∴ By two point rule 𝑛 = 3 , 𝑕= = =
𝑛 3 3

∴ 𝑥0 = 0, 𝑥1 = 𝑥0 + 𝑕 = 1 3 , 𝑥2 = 𝑥0 + 2𝑕 = 2 3 , 𝑥3

= 𝑥0 + 3𝑕 = 𝑏 = 1

3𝑕
∴𝐼= 𝑓 𝑥0 + 3𝑓 𝑥1 + 3𝑓 𝑥2 + 𝑓(𝑥3 )
8

3 13
1
= 1+3 3 4 +3 3 5 +
8 2

266
 1 9 9 1
= 1+ + + SPACE OF HINTS
8 4 5 2
𝐼 = 0.69375

4.5: Composite Integration Methods

Trapezoidal rule:
The formula for Trapezoidal rule is,
𝑕
𝐼= 𝑓 + 2 𝑓1 + 𝑓2 + ⋯ + 𝑓𝑁−1 + 𝑓𝑁
2 0
Simpson’s rule: (No of interval is even)
The formula for Simpson‟s rule is,
𝑕
𝐼= 𝑓 + 4 𝑓1 + 𝑓3 + ⋯ + 𝑓2𝑁−1 + 2 𝑓2 + 𝑓4 + ⋯ + 𝑓2𝑁−2 + 𝑓2𝑁
3 0
Gauss-Legendre three point formula:
1
1
𝑓 𝑥 𝑑𝑥 = 5𝑓 − 3 5 + 8𝑓 0 + 5𝑓 3
5
−1 9

Example: 4.5.1
1 𝑑𝑥
Evaluate the integral 𝐼 = 0 1+𝑥
using

i. Composite trapezoidal rule

ii. Composite Simpson‟s rule

With 2, 4 𝑎𝑛𝑑 8 equal subintervals.

Solution:
1 𝑑𝑥 1
i. When 𝑁 = 2, given 𝐼 = 0 1+𝑥
; 𝑓 𝑥 =
1+𝑥
𝑏−𝑎 1−0 1
we have 𝑕 = = = and three nodes 0, 1 2 and 1.
𝑁 2 2

𝑥 0 1 1
2

𝑓 𝑥 1 2 1
3 2

Trapezoidal rule:

267
 1
2
SPACE FOR HINTS 𝑕 17
𝐼 = 𝑓0 + 2𝑓1 + 𝑓2 = 1+2 2 3 +1 2 =
2 2 24

= 0.70833

Simpson’s rule:
1
2
𝑕
𝐼 = 𝑓0 + 4𝑓1 + 𝑓2 = 1 + 4 2 3 + 1 2 = 0.69444
3 3

ii. Four subintervals

𝑏−𝑎 1−0 1
When 𝑁 = 4, = =
𝑁 4 4

𝑥 0 1 1 3 1
4 2 4

𝑓 𝑥 1 4 2 4 1
5 3 7 2

Trapezoidal rule:
𝑕
𝐼= 𝑓 + 2 𝑓1 + 𝑓2 + 𝑓3 + 𝑓4
2 0
1
4
= 1+2 4 5+2 3+4 7 +1 2
2

𝐼 = 0.697024

Simpson’s rule:
𝑕
𝐼= 𝑓 + 4 𝑓1 + 𝑓3 + 2𝑓2 + 𝑓4
3 0
1
4
4 4 2 1
= 1+4 + +2 +
3 5 7 3 2
1
𝐼= 8.3190 = 0.69325
12

268
iii. 8 subintervals
SPACE OF HINTS
𝑥 0 1 1 3 1 5 3 7 1
8 4 8 2 8 4 8

𝑓(𝑥) 1 8 4 8 2 8 4 8 1
9 5 11 3 13 7 15 2

Trapezoidal rule:
𝑕
𝐼= 𝑓 + 2 𝑓1 + 𝑓2 + ⋯ + 𝑓7 + 𝑓8
2 0
1
8
= 1 + 2 8 9 + 4 5 + 8 11 + 2 3 + 8 13 + 4 7 + 8 15
2

+1 2

𝐼 = 0.69412

Simpson’s rule:
𝑕
𝐼= 𝑓 + 4 𝑓1 + 𝑓3 + 𝑓5 + 𝑓7 + 2(𝑓2 + 𝑓4 + 𝑓6 ) + 𝑓8
3 0
1
8
8 8 8 8 4 2 4 1
= 1+4 + + + +2 + + +
3 7 11 13 15 5 3 7 2
𝐼 = 0.693155

The exact solution is,

1
𝑑𝑥 1
𝐼= = log 𝑒 (1 + 𝑥) 0
0 1+𝑥

= ln 2 − ln 0

𝐼 = 0.693147

Example: 4.5.2

269
 1 𝑑𝑥
Evaluate the integral 𝐼 = 0 1+𝑥
by subdividing the interval
SPACE FOR HINTS
0,1 into two equal parts and then applying the Gauss-Legendre three
point formula
1
1
𝑓 𝑥 𝑑𝑥 = 5𝑓 − 3 5 + 8𝑓 0 + 5𝑓 3
5
−1 9

Solution:
The formula for Gauss-Legendre three-point method is,
1
1
𝑓 𝑥 𝑑𝑥 = 5𝑓 − 3 5 + 8𝑓 0 + 5𝑓 3
5
−1 9

Given
1 1 1
𝑑𝑥 2 𝑑𝑥 𝑑𝑥
𝐼= = + = 𝐼1 + 𝐼2
1+𝑥 0 1+𝑥 1 1+𝑥
0 2

For 𝐼1 , 𝐼2 the transformation 𝑡 + 1 = 4𝑥 and 𝑦 + 3 = 4𝑥 in the first and

second integrals respectively.

Find the value of 𝑡 and 𝑦.

𝑡+1
Consider, 𝑡 + 1 = 4𝑥 ⟹ 𝑥 =
4

𝑑𝑡
𝑑𝑥 =
4
𝑦+3
Similarly, in 𝐼2 , 𝑦 + 3 = 4𝑥 ⟹ 𝑥 =
4 𝑥 1 2 1
𝑑𝑡 𝑦 −1 1
𝑑𝑥 =
4
𝑥 0 1 2
1 1 1
𝑑𝑥 𝑑𝑡 𝑑𝑡 𝑡 −1 1
∴ 𝐼1 = = 𝑡+1
=
0 1+𝑥 −1 4 1+ −1 𝑡 + 5
4

1
∴𝑓 𝑡 =
𝑡+5
Applying Legendre formula

270
 1
𝑓 − 35 = = 0.23666 SPACE OF HINTS
3
− +5
5

1
𝑓 0 = = 0.2
0+5

3 1
𝑓 5 = = 0.17317
3
+5
5

1
∴ 𝐼1 = 5 0.23666 + 8 0.2 + 5(0.17317) = 0.40546
9
Similarly,
1 1 1
𝑑𝑥 𝑑𝑡 𝑑𝑡
∴ 𝐼2 = = 𝑦+3
=
1 1+𝑥 −1 4 1+ −1 𝑦 + 7
2 4

1
∴𝑓 𝑦 =
𝑦+7

Applying Legendre formula

1
𝑓 − 35 = = 0.16064
3
− +7
5

1
𝑓 0 = = 0.14286
0+7

3 1
𝑓 5 = = 0.128617
3
+7
5

1
∴ 𝐼2 = 5 0.16064 + 8 0.14286 + 5(0.128617) = 0.287682
9
∴ 𝐼 = 𝐼1 + 𝐼2 = 0.40546 + 0.287682 = 0.693146
1 𝑑𝑥 1
The exact solution is, 𝐼 = 0 1+𝑥
= log 𝑒 (1 + 𝑥) 0

= ln 2 − ln 0

𝐼 = 0.693147

4.6: Romberg Integration

271
 Trapezoidal rule with Romeberg Integration:
SPACE FOR HINTS
4𝑚 𝐼𝑇 𝑚 −1 𝑕 𝑚 −1
2 − 𝐼𝑇 𝑕
𝑚
𝐼𝑇 𝑕 = , 𝑚 = 1,2, …
4𝑚 − 1
Simpson’s rule with Romberg integration:

4𝑚 +1 𝐼𝑆 𝑚 −1 𝑕 𝑚 −1
2 − 𝐼𝑆 𝑕
𝑚
𝐼𝑆 𝑕 = , 𝑚 = 1,2, …
4𝑚 +1 − 1
Example: 4.6.1
1 𝑑𝑥
Find the approximate value of the integral 𝐼 = 0 1+𝑥
using

i. Composite trapezoidal rule with 2, 3, 5, 9 nodes and Romberg

integration
ii. Composite Simpson‟s rule with 3, 5, 9 nodes and Romberg
integration.

Obtain the number of function evaluations required to get an accuracy of

10−6 when the integral is evaluated directly by using the trapezoidal and
Simpson‟s rules.

Solution:
1 𝑑𝑥
Given 𝐼 = 0 1+𝑥

1
Here 𝑓 𝑥 =
1+𝑥

Number of nodes = 9

∴𝑁 =9−1=8
𝑏−𝑎 1−0 1
∴𝑕= = =
𝑁 8 8
𝑥 0 1 2 3 4 5 6 7 1
8 8 8 8 8 8 8

𝑓(𝑥) 1 8 4 8 2 8 4 8 1
9 5 11 3 13 7 15 2

272
Consider number of nodes = 2
SPACE OF HINTS
𝑥 0 1

𝑓(𝑥) 1 1 2
1
2
𝑕 1
𝐼𝑇 = 𝑓0 + 𝑓1 = 1 + = 0.75
2 2 2

Number of nodes = 3, 𝑕 = 1 2

𝑥 0 4 8 1

𝑓(𝑥) 1 8 12 1 2
𝑕 1 8 1
𝐼𝑇 = 𝑓0 + 2𝑓1 + 𝑓2 = 1 + 2 + = 0.70833
2 2 12 2
𝑏−𝑎 1
Number of nodes = 5, 𝑁 = 4, 𝑕 = =
𝑁 4

𝑥 0 2 8 4 8 6 1
8

𝑓(𝑥) 1 8 10 8 12 8 1 2
14

𝑕
𝐼𝑇 = 𝑓 + 2(𝑓1 + 𝑓2 + 𝑓3 ) + 𝑓4
2 0
1 8 8 8 1
= 1+2 + + + = 0.697024
8 10 12 14 2
𝑏−𝑎 1
Number of nodes = 9, 𝑁 = 8, 𝑕 = =
𝑁 8

𝑕
𝐼𝑇 = 𝑓 + 2 𝑓1 + 𝑓2 + ⋯ + 𝑓7 + 𝑓8
2 0
1 8 8 8 8 8 8 8 1
= 1+2 + + + + + + +
8 9 10 11 12 13 14 15 2
𝐼𝑇 = 0.694122
Using Trapezoidal rule with Romberg integration,

273
 4𝑚 𝐼𝑇 𝑚 −1 𝑕 𝑚 −1
SPACE FOR HINTS 2 − 𝐼𝑇 𝑕
𝑚
𝐼𝑇 𝑕 = , 𝑚 = 1,2, …
4𝑚 − 1
Put 𝑚 = 1 𝑎𝑛𝑑 𝑕 = 1 (Fourth order derivatives)

4𝑚 𝐼𝑇 0 1 0
2 − 𝐼𝑇 𝑕
4 0.70833 − 0.75
𝐼𝑇 ′ 𝑕 = =
4−1 3
𝐼𝑇 ′ 1 = 0.69444

𝑕 = 1 2 we get,

0 1 0 1
4𝐼𝑇 4 − 𝐼𝑇 2
′ 1 4 0.697024 − 0.70833
𝐼𝑇 = =
2 4−1 3
= 0.693254

𝑕 = 1 4 we get,

0 1 0 1
4𝐼𝑇 8 − 𝐼𝑇 4
′ 1 4 0.694122 − 0.697024
𝐼𝑇 = =
4 4−1 3
= 0.693155

Put 𝑚 = 2 𝑎𝑛𝑑 𝑕 = 1 (Sixth order derivatives)

42 𝐼𝑇 1 1 1
2 − 𝐼𝑇 1
16 0.693254 − 0.69444
𝐼𝑇 (2) 𝑕 = =
42 − 1 15
= 0.693175

𝑕 = 1 2 we get,

1 1 1 1
16𝐼𝑇 4 − 𝐼𝑇 2
(2) 1 16 0.693155 − 0.693254
𝐼𝑇 = =
2 15 15

274
 = 0.693148
SPACE OF HINTS
Put 𝑚 = 3 𝑎𝑛𝑑 𝑕 = 1 (Eigth order derivatives)

43 𝐼𝑇 2 1 2
2 − 𝐼𝑇 1
64 0.693148 − 0.693175
𝐼𝑇 (3) 1 = =
43 −1 15
= 0.693148

𝑕\𝑜𝑟𝑑𝑒𝑟 𝑆𝑒𝑐𝑜𝑛𝑑 𝐹𝑜𝑢𝑟𝑡𝑕 𝑆𝑖𝑥𝑡𝑕 𝐸𝑖𝑔𝑡𝑕

1 0.75
0.69444
1 0.70833 0.693175
2
0.693254 0.693148

0.697024 0.693148
1
4 0.693155

0.694122

1
8

1 𝑑𝑥 1
Exact solution is, 𝐼 = 0 1+𝑥
= log 𝑒 (1 + 𝑥) 0

= ln 2 − ln 0

𝐼 = 0.693147
𝑏−𝑎
Simpson’s rule:Number of nodes = 3 𝑁 = 2 𝑕= =1 2
𝑁

𝑥 0 1 2 1

𝑓(𝑥) 1 2 3 1 2
1
1 2
𝑕 2 1
𝑓𝑑𝑥 = 𝑓0 + 4𝑓1 + 𝑓2 = 1+4 +
0 3 3 3 2

275
 = 0.69444
SPACE FOR HINTS
Number of nodes = 5, 𝑁 = 4 𝑕=1 4

𝑥 0 1 4 1 2 3 1
4

𝑓(𝑥) 1 4 5 2 3 4 1 2
7

1
1 4
4 4 2 1
𝑓𝑑𝑥 = 1+4 + +2 + = 0.693254
0 3 5 7 3 2
𝑏−𝑎 1
Number of nodes = 9 𝑁 = 8, 𝑕 = =
𝑁 8

𝑥 0 1 2 3 4 5 6 7 1
8 8 8 8 8 8 8

𝑓(𝑥) 1 8 4 8 2 8 4 8 1
9 5 11 3 13 7 15 2

1
1 8
8 8 8 8 4 2 4 1
𝑓𝑑𝑥 = 1+4 + + + +2 + + +
0 3 9 11 13 15 5 3 7 2

= 0.693155
Using Simpson‟s rule with Romberg integration,

4𝑚 +1 𝐼𝑆 𝑚 −1 1 𝑚 −1
2 − 𝐼𝑆 𝑕
𝑚
𝐼𝑆 𝑕 = , 𝑚 = 1,2, …
4𝑚 +1 − 1
Put 𝑚 = 1, 𝑕 = 1 2 we get,

0 1 0 1
42 𝐼𝑆 4 − 𝐼𝑆 2
(1) 1
𝐼𝑆 =
2 4−1

276
 16 0.693254 − 0.69444
= = 0.0.693175 SPACE OF HINTS
15
𝑕 = 1 4 we get,

0 1 0 1
42 𝐼𝑆 8 − 𝐼𝑆 4
(1) 1
𝐼𝑆 =
4 42 − 1
16 0.693155 − 0.693254
=
15
= 0.693148

Put 𝑚 = 2, 𝑕 = 1 2 we get,

1 1 1 1
43 𝐼𝑆 4 − 𝐼𝑆 2
(2) 1
𝐼𝑆 =
2 43 − 1
64 0.693148 − 0.693175
=
63
= 0.693148

𝑕\𝑜𝑟𝑑𝑒𝑟 𝑓𝑜𝑢𝑟𝑡𝑕 𝑠𝑖𝑥𝑡𝑕 Eighth

1 0.69444
2
0.693175

0.693254 0.693148
1
4 0.693148

0.693155

1
8

The error term in trapezoidal rule is,

277
 𝑕2 ′′
𝑅1 = 𝑓 𝜖 , 0<𝜖<1
SPACE FOR HINTS 12
1
Given 𝑓 𝑥 = , 0≤𝑥≤1
1+𝑥

−1 2
𝑓′ 𝑥 = , 𝑓 ′′ 𝑥 =
1+𝑥 2 1+𝑥 3

Put 𝑥 = 0, 𝑓 ′′ 𝜖 = 2

𝑥 = 1, 𝑓 ′′ 1 = 1 4

∴ 1 4 ≤ 𝑓 ′′ 𝜖 ≤ 2

1 𝑕2 𝑕2 ′′
𝑕2
4 12 ≤ 12 𝑓 𝜖 ≤ 2 12

𝑕2 𝑕2 ′′ 𝑕2
≤ 𝑓 𝜖 ≤
48 12 6
Given error accuracy is 10−6

𝑕2
∴ ≤ 10−6
48
𝑕2 ≤ 10−6 . (48)
1
𝑕 ≤ 0.000048 2

𝑕 ≅ 0.007

the error term in Simpson‟s rule is,

2 −6 24
𝑓 ′′ 𝑥 = 3
, 𝑓 ′′′ 𝑥 = 4
, 𝑓 𝑖𝑣 𝑥 = 5
1+𝑥 1+𝑥 1+𝑥

0 ≤ 𝑥 ≤ 1 ⟹ 𝑥 = 0 𝑓 𝑖𝑣 0 = 24

𝑥 = 1 𝑓 𝑖𝑣 1 = 3 4

𝑕4 𝑖𝑣
𝑅2 = 𝑓 𝜂
180

∴ 3 4 ≤ 𝑓 𝑖𝑣 𝜂 ≤ 24

278
 𝑕4 3 𝑕4 𝑖𝑣
𝑕4
4 ≤ 𝑓 𝜂 ≤ 24 SPACE OF HINTS
180 180 180

𝑕4 𝑕4 𝑖𝑣
2𝑕4
≤ 𝑓 𝜂 ≤
240 180 15

Given error accuracy is10−6

𝑕4
∴ ≤ 10−6
240
𝑕4 ≤ 10−6 . (240)
1
𝑕 ≤ 10−6 × 240 4

𝑕 ≅ 0.1245

Unit - 5
Ordinary differential equations: Initial value problems
5.1: Introduction
An ordinary differential equation is a relation between a function,
its derivatives, and the variable upon which they depend. The most
general form of an ordinary differential equation is given by

𝜑 𝑡, 𝑦, 𝑦 ′ , 𝑦 ′′ , … , 𝑦 𝑚
=0 (1)

279
 where 𝑚 represents the highest order derivative, and 𝑦 and its
SPACE FOR HINTS derivatives are functions of 𝑡. The order of the differential equation is
the order of its highest derivative and its degree is the degree of the
derivate of the highest order after the equation has been rationalized in
derivatives.

If no product of the dependent variable 𝑦(𝑡) with itself or any one

of its derivatives occur, then the equation is said to be linear, otherwise
it is nonlinear.

A linear differential equation of order 𝑚 can be expressed in the form

𝑚
𝑝
𝜑𝑝 𝑡 𝑦 𝑡 = 𝑟(𝑡) (2)
𝑝=0

in which 𝜑𝑝 𝑡 are functions. If the general nonlinear differential

equation (1) of order 𝑚 can be written as

𝑦 (𝑚 ) = 𝐹(𝑡, 𝑦, 𝑦 ′ , 𝑦 ′′ , … , 𝑦 𝑚 −1
(3)

Then the equation (3) is called a canonical representation of the

differential equation (1). In such a form, the highest order derivative is
expressed in terms of the lower order derivatives and the independent
variable.

Initial value problem:

A general solution of an ordinary differential equation such as (1) is a
relation between 𝑦, 𝑡 and 𝑚 arbitrary constants, which satisfies the
equation, but which contains no derivatives. The solution may be an
implicit relation of the form

𝑤 𝑡, 𝑦, 𝑐1 , 𝑐2 , … , 𝑐𝑚 = 0 (4)

or an explicit function of 𝑡 of the form

𝑦 = 𝑤 𝑡, 𝑐1 , 𝑐2 , … , 𝑐𝑚 (5)

The 𝑚 arbitrary constants 𝑐1 , 𝑐2 , … , 𝑐𝑚 can be determined by

prescribing 𝑚 conditions of the form

280
 𝑣
𝑦 𝑡0 = 𝜂𝑣 , 𝑣 = 0, 1, 2, … , 𝑚 − 1 (6) SPACE OF HINTS

At one point 𝑡 = 𝑡0 , which are called initial conditions. The point 𝑡0 is

called the initial point. The differential equation (1) together with the
initial conditions (6) is called an 𝑚𝑡𝑕 order initial value problem.

If the 𝑚 conditions are prescribed at more than one point to determine

the 𝑚 arbitrary constants in the general solution (4), these are called
boundary conditions.
The differential equation (1) together with the boundary conditions is
known as the boundary value problem.

Reduction of Higher Order Equations to the System of First Order

Differential Equations:
The solution of first order initial value problem is
𝑑𝑢
= 𝑓 𝑡, 𝑢 , 𝑢 𝑡0 = 𝜂0
𝑑𝑡
𝑢′ 𝑡 = 𝑓 𝑡, 𝑢 , 𝑢 𝑡0 = 𝜂0

Example: 5.1.1
Convert the following second order initial value problem into a
system of first order initial problems 𝑡𝑦 ′′ − 𝑦 ′ + 4𝑡 3 𝑦 = 0

𝑦 1 = 1, 𝑦 ′ 1 = 2.

Solution:
Given 𝑡𝑦 ′′ − 𝑦 ′ + 4𝑡 3 𝑦 = 0

⟹ 𝑡𝑦 ′′ = 𝑦 ′ − 4𝑡 3 𝑦
1 ′
∴ 𝑦 ′′ = 𝑦 − 4𝑡 3 𝑦 (1)
𝑡
Let 𝑣1 = 𝑦 𝑡 , 𝑣2 = 𝑦 ′ (𝑡)

Now 𝑣1 ′ = 𝑦 ′ 𝑡 , 𝑣2 ′ = 𝑦 ′′ (𝑡)

𝑖𝑒) 𝑣1 ′ = 𝑦 ′ 𝑡 = 𝑣2

The system of equation becomes,

281
 ∴ 𝑣1 = 𝑦 𝑡 = 𝑦 1 = 1 (𝑡 = 1)
SPACE FOR HINTS
Similarly, 𝑣1 ′ = 𝑦 ′ 𝑡 = 𝑦 ′ 1 = 2 (𝑡 = 1)

Substitute the values in (1),

1 ′
𝑦 ′′ = 𝑦 − 4𝑡 3 𝑦
𝑡
1
𝑣2 ′ = 𝑣 − 4𝑡 3 𝑣1 𝑎𝑛𝑑 𝑣1 1 = 1, 𝑣1 ′ (1) = 2
𝑡 2
Taking 𝑣1 ′ = 𝑓 𝑡, 𝑣 𝑎𝑛𝑑 𝑣1 1 = 1 = 𝑑(𝑠𝑎𝑦)

Hence the second order derivative can be changed into the first order
initial value problem, where 𝑣 = 𝑣1 , 𝑣2 𝑇 , 𝑓 = 𝑓1 , 𝑓2 𝑇 ,

𝑇
1
𝑑 = 𝑑1 , 𝑑2 𝑎𝑛𝑑 𝑓1 = 𝑣2 𝑎𝑛𝑑 𝑓2 = 𝑣2 − 4𝑡 3 𝑣1
𝑡
where 𝑑1 = 1, 𝑑2 = 2.

Example: 5.1.2
Convert the following system of two simultaneous third-order equations
into a system of six simultaneous first-order equations.

𝑦1 ′′′ = 𝑓1 𝑡, 𝑦1 , 𝑦2 , 𝑦1 ′ , 𝑦2 ′ , 𝑦1 ′′ , 𝑦2 ′′

𝑦2 ′′′ = 𝑓2 𝑡, 𝑦1 , 𝑦2 , 𝑦1 ′ , 𝑦2 ′ , 𝑦1 ′′ , 𝑦2 ′′

𝑦1 𝑡0 = 𝑎0 , 𝑦1 ′ 𝑡0 = 𝑎1 , 𝑦 ′′ 𝑡0 = 𝑎2 ,

𝑦2 = 𝑏0 , 𝑦2 ′ 𝑡0 = 𝑏1 , 𝑦 ′′ 𝑡0 = 𝑏2

Solution:
𝑦1 ′′′ = 𝑓1 𝑡, 𝑦1 , 𝑦2 , 𝑦1 ′ , 𝑦2 ′ , 𝑦1 ′′ , 𝑦2 ′′
(1)
𝑦2 ′′′ = 𝑓2 𝑡, 𝑦1 , 𝑦2 , 𝑦1 ′ , 𝑦2 ′ , 𝑦1 ′′ , 𝑦2 ′′

Let
𝑢1 𝑡 = 𝑦1 𝑡 , 𝑢3 𝑡 = 𝑦1 ′ 𝑡 , 𝑢5 𝑡 = 𝑦1 ′′ 𝑡
(2)
𝑢2 𝑡 = 𝑦2 𝑡 , 𝑢4 𝑡 = 𝑦2 ′ 𝑡 , 𝑢6 𝑡 = 𝑦2 ′′ 𝑡

282
 𝑢1 ′ 𝑡 = 𝑦1 ′ 𝑡 , 𝑢3 ′ 𝑡 = 𝑦1 ′′ 𝑡 , 𝑢5 ′ 𝑡 = 𝑦1 ′′′ 𝑡
⟹ ′ (3) SPACE OF HINTS
𝑢2 𝑡 = 𝑦2 ′ 𝑡 , 𝑢4 ′ 𝑡 = 𝑦2 ′′ 𝑡 , 𝑢6 ′ 𝑡 = 𝑦2 ′′′ 𝑡

From the above set of equations (2) & (3), we have

∴ 𝑢1 ′ 𝑡 = 𝑢3 𝑡 𝑢3 ′ 𝑡 = 𝑢5 𝑡

𝑢2 ′ 𝑡 = 𝑢4 𝑡 𝑢4 ′ 𝑡 = 𝑢6 𝑡

Given 𝑦1 𝑡0 = 𝑎0 , 𝑦2 (𝑡0 ) = 𝑏0

𝑦1 ′ 𝑡0 = 𝑎1 , 𝑦2 ′ 𝑡0 = 𝑏1

𝑦 ′′ 𝑡0 = 𝑎2 , 𝑦 ′′ 𝑡0 = 𝑏2

2 ⟹ 𝑢1 𝑡0 = 𝑎0 , 𝑢2 𝑡0 = 𝑎1 , 𝑢3 𝑡0 = 𝑎2 , 𝑢4 𝑡0 = 𝑏0 ,

𝑢5 𝑡0 = 𝑏1 , 𝑢6 𝑡0 = 𝑏2

Similarly, 𝑢′ 𝑡0 = 𝑦1 ′ 𝑡0 = 𝑓(𝑡, 𝑢) 𝑢 𝑡0 = 𝑢0

Substitute thisvalues in (1), we get

𝑦1 ′′′ = 𝑢5 ′ 𝑡 = 𝑓1 𝑡, 𝑢1 , 𝑢2 , 𝑢3 , 𝑢4 , 𝑢5 , 𝑢6 , 𝑢5 𝑡0 = 𝑎2

𝑦2 ′′′ = 𝑢6 ′ 𝑡 = 𝑓2 𝑡, 𝑢1 , 𝑢2 , 𝑢3 , 𝑢4 , 𝑢5 , 𝑢6 , 𝑢6 𝑡0 = 𝑏2

Existence and uniqueness

Theorem: 5.1.1
We assume that 𝑓 𝑡, 𝑢 satisfies the following conditions:

i. 𝑓 𝑡, 𝑢 is a real function.
ii. 𝑓 𝑡, 𝑢 is defined and continuous in the strip 𝑡 ∈ 𝑡0 , 𝑏 ,

𝑢 ∈ −∞, ∞

iii. There exists a constant 𝐿 such that for any 𝑡 ∈ 𝑡0 , 𝑏 and any 𝑣1
and 𝑣2 𝑓 𝑡, 𝑣1 − 𝑓 𝑡, 𝑣2 ≤ 𝐿 𝑣1 − 𝑣2 where 𝐿 is called the
Lipschitz constant.
𝑑𝑢
Then for any 𝑢0 the initial value problem = 𝑓 𝑡, 𝑢 and
𝑑𝑡

𝑢 𝑡0 = 𝜂0 has a unique solution 𝑢 𝑡 for 𝑡 ∈ 𝑡0 , 𝑏 .

Systems of Linear First Order Differential Equations with Constant

Coefficients:

283
 A first order linear system with constant coefficients is of the form
SPACE FOR HINTS 𝑢′ = 𝐴𝑢 + 𝑏 𝑡 (1)

where 𝐴 = 𝑎𝑖𝑗 , 𝑖, 𝑗 = 1, 2, … , 𝑚 is an 𝑚 × 𝑚 constant matrix and 𝑏(𝑡)

is an 𝑚-dimensional vector.

The general solution of (1) may be written as

𝑢 𝑡 = 𝑘 𝑖 𝑒 𝜆 𝑖 𝑡 𝑢𝑖 + 𝜙 𝑡 (2)
𝑖=1

where the matrix 𝑨 has distinct eigenvalues 𝜆𝑖 and the corresponding

eigenvectors 𝑢𝑖 .

The function 𝜙 𝑡 is the particular solution of (1). The matrix equation

(1) may be decoupled by applying a similarity transformation.

Let 𝑈 = [𝑢1 , 𝑢2 , … , 𝑢𝑚 ] be the matrix of eigenvectors. Set 𝑢 = 𝑈𝑧. Then

𝑢′ = 𝑈𝑧 ′ = 𝐴 𝑈 𝑧 + 𝑏 𝑡 (3)

Premultiplying (3) by 𝑈 −1 , we get

𝑈 −1 𝑈 𝑧 ′ = 𝑈 −1 𝐴𝑈𝑧 + 𝑈 −1 𝑏(𝑡)

or 𝑧 ′ = 𝐷𝑧 + 𝑕 𝑡 (4)

where 𝑕 𝑡 = 𝑈 −1 𝑏(𝑡) and 𝐷 = 𝑈 −1 𝐴𝑈 is the diagonal matrix with

diagonal elements 𝜆1 , 𝜆2 , … , 𝜆𝑚 .

(4) gives a degenerate system for the solution of 𝑧1 , 𝑧2 , … , 𝑧𝑚 .

The solution (4) is given by

𝑧𝑖 𝑡 = 𝑒 𝜆 𝑖 𝑡 𝑒 −𝜆 𝑖 𝑡 𝑕𝑖 𝑡 𝑑𝑡 + 𝑐𝑖 , 𝑖 = 1, 2, … , 𝑚.

The solution to the linear 1st order differential equation is

𝒖 𝒕 = 𝒄𝟏 𝒆𝝀𝟏 𝒕 𝒗(𝟏) 𝒕 + 𝒄𝟐 𝒆𝝀𝟐 𝒕 𝒗(𝟐) (𝒕).

Theorem: 5.1.2 (stability)

284
The solutions to the system 𝑢′ = 𝐴𝑢 + 𝑏 𝑡 with 𝑏 𝑡 = 0 are said to be
SPACE OF HINTS
stable.

𝑖𝑒) 𝑢𝑖 → 0 𝑎𝑠 . 𝑡 → ∞ iff all the eigen values of the matrix 𝑨 have

negative real parts.

Example: 5.1.3
𝑑𝑢
Find the solution of the system of equations = 𝐴𝑢 where
𝑑𝑡

𝑢 = 𝑢1 𝑢2 𝑇 and 𝐴 = −3 4
.
−2 3
Solution:
The solution to the system of equations can be obtained by finding
−3 4
the eigenvalues and eigenvectors of the given matrix 𝐴 = .
−2 3
The characteristic equation of 𝐴 is 𝐴 − 𝜆𝐼 = 0.
−3 − 𝜆 4
=0
−2 3−𝜆
⟹ −3 − 𝜆 3 − 𝜆 + 8 = 0

⟹ −9 + 3𝜆 − 3𝜆 + 𝜆2 + 8 = 0

⟹ 𝜆2 − 1 = 0 ⟹ 𝜆 = ±1

The eigenvalues are 𝜆 = 1, −1

To find the eigenvectors corresponding eigenvalues 𝐴 − 𝜆 𝑋 = 0

−4 4 𝑢1
For 𝜆 = 1, 𝑢2 = 0
−2 2
−4𝑢1 + 4𝑢2 = 0

−2𝑢1 + 2𝑢2 = 0

⟹ 𝑢1 = 𝑢2

𝑢2 = 1 ⟹ 𝑢1 = 1
𝑢1 1
∴ The eigenvector corresponding the eigenvalues 𝜆 = 1 is 𝑢 =
2 1
1
𝑣 (1) =
1

285
 −2 4 𝑢1
For 𝜆 = −1, 𝑢2 = 0
−2 4
SPACE FOR HINTS
−2𝑢1 + 4𝑢2 = 0

−2𝑢1 + 4𝑢2 = 0

⟹ 𝑢1 = 2𝑢2

𝑢2 = 1 ⟹ 𝑢1 = 2
𝑢1 2
∴ The eigen vector corresponding the eigenvalues 𝜆 = −1 is 𝑢 =
2 1
2
𝑣 (2) =
1
Hence the solution to the system of equation is

𝑢 𝑡 = 𝑐1 𝑒 𝜆 1 𝑡 𝑉 1 𝑡 + 𝑐2 𝑒 𝜆 2 𝑡 𝑉 2 (𝑡)
1 2
= 𝑐1 𝑒 𝑡 + 𝑐2 𝑒 −𝑡
1 1
𝑐1 𝑒 𝑡 + 2𝑐2 𝑒 −𝑡
𝑢 𝑡 =
𝑐1 𝑒 𝑡 + 𝑐2 𝑒 −𝑡

Hence the solution to the system of equation is,

𝑢1 𝑡 = 𝑐1 𝑒 𝑡 + 2𝑐2 𝑒 −𝑡 , 𝑢2 (𝑡) = 𝑐1 𝑒 𝑡 + 𝑐2 𝑒 −𝑡

Example: 5.1.4
Find the solution of the initial value problem
𝑑𝑢 𝑇 −2 1 𝑢1
= 𝐴𝑢, 𝑢 0 = 1, 0 𝑤𝑕𝑒𝑟𝑒 𝐴 = 𝑎𝑛𝑑 𝑢 = 𝑢 .Is
𝑑𝑡 1 −20 2

the system asymptotically stable?

Solution:
The solution to the system of linear equation can be obtained by finding
the eigenvalues and eigenvectors of the given matrix 𝐴.

The characteristic equation 𝐴 is 𝐴 − 𝜆𝐼 = 0.

286
 −2 − 𝜆 1
=0 SPACE OF HINTS
1 −20 − 𝜆
⟹ −2 − 𝜆 −20 − 𝜆 − 1 = 0

⟹ 40 + 2𝜆 + 20𝜆 + 𝜆2 − 1 = 0

⟹ 𝜆2 + 22𝜆 + 39 = 0

−22 ± (22)2 − 4 × 1 × 39
𝜆= = −11 ± 82
2
∴ The eigen values are 𝜆1 = −11 + 82 , 𝜆2 = −11 − 82.

To find the eigenvector corresponding eigenvalues, 𝐴 − 𝜆 𝑢 = 0

−2 + 11 − 82 1 𝑢1
For 𝜆1 = −11 + 82 , 𝑢 =0
1 −20 + 11 − 82 2

9 − 82 𝑢1 + 𝑢2 = 0

𝑢1 − 9 − 82 𝑢2 = 0

⟹ 9 − 82 𝑢1 = −𝑢2 ⟹ 82 − 9 𝑢1 = 𝑢2

Put 𝑢1 = 1 ⟹ 𝑢2 = 82 − 9

∴ The eigen vector corresponding the eigen value, 𝜆1 = −11 + 82is

𝑢1 1
𝑢2 = = 𝑣 (1)
82 − 9

−2 + 11 + 82 1 𝑢1
For 𝜆 = −11 − 82 , =0
1 −20 + 11 + 82 𝑢2

9 + 82 𝑢1 + 𝑢2 = 0

𝑢1 + −9 + 82 𝑢2 = 0

⟹ 𝑢1 = − −9 + 82 𝑢2

Put 𝑢2 = 1 ⟹ 𝑢1 = 9 − 82

∴ The eigenvector corresponding the eigenvalue, 𝜆1 = −11 − 82is

𝑢1 9 − 82 = 𝑣 (2)
𝑢2 = 1

287
 The general solution to the given system of equation is given by
SPACE FOR HINTS
𝑢 𝑡 = 𝑐1 𝑒 𝜆 1 𝑡 𝑣 1 𝑡 + 𝑐2 𝑒 𝜆 2 𝑡 𝑣 2 (𝑡)

1
= 𝑐1 𝑒 (−11+ 82)𝑡
82 − 9

+ 𝑐2 𝑒 (−11− 82)𝑡 9 − 82
1

𝑐1 𝑒 (−11+ 82)𝑡
+ (9 − 82)𝑐2 𝑒 (−11− 82)𝑡
𝑢 𝑡 =
( 82 − 9)𝑐1 𝑒 (−11+ 82)𝑡
+ 𝑐2 𝑒 (−11− 82)𝑡

Substitute the initial condition 𝑢 0 = 0,1 𝑇 , we get,

𝑐1 + (9 − 82)𝑐2
𝑢 0 =
( 82 − 9)𝑐1 + 𝑐2

1 𝑐 + (9 − 82)𝑐2
𝑖𝑒) = 1
0 ( 82 − 9)𝑐1 + 𝑐2

𝑐1 + 9 − 82 𝑐2 = 1 (1)

82 − 9 𝑐1 + 𝑐2 = 0 (2)

1 ⟹ 𝑐1 = 1 − 9 − 82 𝑐2

(2) ⟹ (1 − 9 − 82 𝑐2 ) 82 − 9 + 𝑐2 = 0
2
⟹ 82 − 9 + 𝑐2 82 − 9 =0
2
⟹ 𝑐2 82 − 9 + 1 = 9 − 82

9 − 82
𝑐2 = 2 = −0.05522
82 − 9 +1

1 ⟹ 𝑐1 = 1 − 9 − 82 −0.05522 = 0.99694

Hence the solution to the system is

288
 1
𝑢1 𝑡 = 0.99694𝑒 (−11+ 82)𝑡
SPACE OF HINTS
82 − 9

− 0.05522𝑒 −(11+ 82)𝑡 9 − 82

1

Example: 5.1.5
Obtain a general solution of the system of equations
𝑑𝑢1 𝑑𝑢2
= −5𝑢1 + 2𝑢2 + 𝑡, = 2𝑢1 − 2𝑢2 + 𝑒 −𝑡
𝑑𝑡 𝑑𝑡
Solution:
𝑑𝑢
The given system of equation can be written as, = 𝐴𝑢 + 𝑏
𝑑𝑡

where
−5 2 𝑢1 𝑡
𝐴= , 𝑢= 𝑢 , 𝑏= .
2 −2 2 𝑒 −𝑡
The solution to the system of linear equation can be obtained by finding
the eigenvalues and eigenvectors of the given matrix 𝐴.

The characteristic equation of 𝐴 is 𝐴 − 𝜆𝐼 = 0

−5 − 𝜆 2
=0
2 −2 − 𝜆
⟹ −5 − 𝜆 −2 − 𝜆 = 0

⟹ 10 + 7𝜆 + 6 = 0

⟹ 𝜆+1 𝜆+6 =0

⟹ 𝜆 = −1, −6
To find the eigenvectors,
−4 2 𝑢1
For 𝜆 = −1, =0
2 −1 𝑢2
−4𝑢1 + 2𝑢2 = 0, 2𝑢1 − 𝑢2 = 0

⟹ −2𝑢1 + 𝑢2 = 0, 2𝑢1 = 𝑢2

2𝑢1 = 𝑢2

Put 𝑢1 = 1, 𝑢2 = 2

289
 𝑢1 1
∴ 𝑢 = = 𝑣 (1)
SPACE FOR HINTS 2 2
1 2 𝑢1
For 𝜆 = −6, =0
2 4 𝑢2
𝑢1 + 2𝑢2 = 0, 2𝑢1 + 4𝑢2 = 0

𝑢1 + 2𝑢2 = 0

𝑢1 = 2𝑢2

Put 𝑢2 = −1, 𝑢1 = 2
𝑢1 2
∴ 𝑢 = = 𝑣 (2)
2 −1
The formula for this method is, 𝑢′ = 𝑢𝑧 ′ 𝑎𝑛𝑑 𝑢𝑧 ′ = 𝐴 ∪ 𝑧 + 𝑏(𝑡)

Premultiplying𝑢−1 in the above equation, we get

𝑧 ′ = 𝑢−1 𝐴 𝑈 𝑧 + 𝑢−1 𝑏(𝑡)

𝑧 ′ = 𝐷𝑧 + 𝑕(𝑡)

where 𝐷 = 𝑢−1 𝐴 𝑈 and 𝑕 𝑡 = 𝑢−1 𝑏(𝑡)

To find 𝒖:

𝑈 =The eigen vector matrix

2 1
𝑈= , 𝑈 = −1 − 4 = 5
−1 2
To find 𝒉(𝒕):

𝑕 𝑡 = 𝑈 −1 𝑏(𝑡)
1 2 −1 𝑡 1 2𝑡 − 𝑒 −𝑡
= =
5 1 2 𝑒 −𝑡 5 𝑡 + 2𝑒 −𝑡
To find 𝑫:
1 2 −1 −5 2 2 1
𝐷 = 𝑈 −1 𝐴𝑈 =
5 1 2 2 −2 −1 2
−6 0
𝐷=
0 −1

290
 ∴ 𝑧 −1 = 𝐷𝑧 + 𝑕(𝑡)
SPACE OF HINTS

−6 0 𝑧1 1 2𝑡 − 𝑒 −𝑡
= +
0 −1 𝑧2 5 𝑡 + 2𝑒 −𝑡

𝑧1 ′ = −6𝑧1 + 1 5 2𝑡 − 𝑒 −𝑡

𝑧2 ′ = −𝑧2 + 1 5 𝑡 + 2𝑒 −𝑡

The solution to the 1st order differential equations are

𝑧𝑖 (𝑡) = 𝑒 𝜆 𝑖 𝑡 𝑒 𝜆 𝑖 𝑡 𝑕𝑖 (𝑡)𝑑𝑡 + 𝑐𝑖

𝑧1 (𝑡) = 𝑒 𝜆 1 𝑡 𝑒 𝜆 1 𝑡 𝑕1 (𝑡)𝑑𝑡 + 𝑐1

𝑧1 𝑡 = 𝑐1 𝑒 𝜆 1 𝑡 + 𝑒 𝜆 1 𝑡 𝑒 −𝜆 1 𝑡 𝑕1 (𝑡)𝑑𝑡

𝑧1 𝑡 = 𝑒 𝜆 1 𝑡 𝑒 𝜆 1 𝑡 𝑕2 (𝑡)𝑑𝑡 + 𝑐1

𝑒 −6𝑡
= 𝑒 −6𝑡 (−2𝑡 + 𝑒 −𝑡 )𝑑𝑡 + 𝑐1
5

𝑒 −6𝑡 𝑢 = −2𝑡 ; 𝑑𝑣 = 𝑒 6𝑡 𝑑𝑡
= 𝑐1 𝑒 −6𝑡 + −2𝑡𝑒 −6𝑡 + 𝑒 −5𝑡 𝑑𝑡 𝑒 6𝑡
5 𝑑𝑢 = −2𝑑𝑡; 𝑣 =
6
𝑒 −6𝑡 𝑒 6𝑡
= 𝑐1 𝑒 −6𝑡 + −2𝑡𝑒 −6𝑡 − −2𝑑𝑡 + 𝑒 5𝑡 𝑑𝑡
5 6

−6𝑡
𝑒 −6𝑡 −𝑡𝑒 6𝑡 1 𝑒 6𝑡 𝑒 5𝑡
= 𝑐1 𝑒 + + . +
5 3 3 6 5

−6𝑡
−𝑡 1 𝑒 −𝑡
𝑧1 𝑡 = 𝑐1 𝑒 + + +
5 90 25

−6𝑡
18𝑒 −𝑡 − 30𝑡 + 5
= 𝑐1 𝑒 +
450

Similarly,

291
 𝑧2 𝑡 = 𝑐2 𝑒 −𝑡 + 𝑒 −𝑡 𝑒 𝑡 1 5 𝑡 + 2𝑒 −𝑡 𝑑𝑡 𝜆2 = −1
SPACE FOR HINTS

𝑒 −𝑡 𝑢 = 𝑡; 𝑑𝑣 = 𝑒 𝑡 𝑑𝑡
= 𝑐2 𝑒 −𝑡 + 𝑡𝑒 𝑡 𝑑𝑡 + 2𝑑𝑡
5 𝑑𝑢 = 𝑑𝑡 ; 𝑣 = 𝑒𝑡
𝑒 −𝑡
= 𝑐2 𝑒 −𝑡 + 𝑡𝑒 𝑡 − 𝑒 𝑡 𝑑𝑡 + 2𝑑𝑡
5
1
𝑧2 𝑡 = 𝑐2 𝑒 −𝑡 + 𝑡 − 1 + 2𝑡𝑒 𝑡
5
The general solution is 𝑢 = 𝑈𝑍
𝑢1 2 1 𝑧1
𝑖𝑒) 𝑢2 = −1 2 𝑧2
𝑢1 2𝑧1 + 𝑧2
𝑢2 =
−𝑧1 + 2𝑧2

𝑢1 = 2𝑧1 + 𝑧2

−6𝑡
18𝑒 −𝑡 − 30𝑡 + 5 1
= 2 𝑐1 𝑒 + + 𝑐2 𝑒 −𝑡 + 𝑡 − 1 + 2𝑡𝑒 𝑡 −
450 5
1
𝑢1 = 2𝑐1 𝑒 −6𝑡 + 𝑐2 𝑒 −𝑡 + 180𝑡𝑒 −𝑡 − 36𝑒 −𝑡 + 150𝑡 − 100
450
𝑢2 = 2𝑧2 − 𝑧1

−𝑡
1 18𝑒 −𝑡 − 30𝑡 + 5
= 2 𝑐2 𝑒 + 𝑡 − 1 + 2𝑡𝑒 𝑡 − 𝑐1 𝑒 −6𝑡
+
5 450
1
𝑢2 = 2𝑐2 𝑒 −𝑡 − 𝑐1 𝑒 −6𝑡 + 360𝑡𝑒 −𝑡 + 18𝑒 −𝑡 + 150𝑡 − 175
450
5.2 Difference Equations

Let 𝑥𝑛 , 𝑥𝑛 +1 , … , 𝑥𝑛 +𝑘 be a set of 𝑘 + 1 equispaced tabular points

with spacing 𝑕 and 𝑢𝑛 , 𝑢𝑛+1 , … , 𝑢𝑛+𝑘 be the corresponding values of a
function 𝑢(𝑥)

At these points, ie., 𝑥𝑛+1 = 𝑥𝑛 + 𝑖𝑕, 𝑢𝑛 +1 = 𝑢 𝑥𝑛 +𝑖 , 𝑖 = 0, 1, … , 𝑘, for

some integer 𝑘. A relationship between 𝑢𝑛 and the differences
∆𝑢𝑛 , ∆2 𝑢𝑛, … , ∆𝑘 𝑢𝑛 is called a difference equation and hence it can be
regarded as a relation among 𝑢𝑛 , 𝑢𝑛+1 , … , 𝑢𝑛+𝑘 . The order of a

292
difference equation is the number of intervals separating the largest and
SPACE OF HINTS
the smallest argument of the dependent variable.

A difference equation of order 𝑘, in its most general form can be

written as

𝐹 𝑢𝑛 , 𝑢𝑛+1 , … , 𝑢𝑛+𝑘 = 0 (1) If the

function 𝐹 is linear in 𝑢𝑛 , 𝑢𝑛+1 , … , 𝑢𝑛+𝑘 , then the difference equation
is called linear.

A linear difference equation of order 𝑘 can be written as

𝑎0 𝑢𝑛+𝑘 + 𝑎1 𝑢𝑛+𝑘−1 + ⋯ , 𝑎𝑘 𝑢𝑛 = 𝑔𝑛 , 𝑎0 ≠ 0 (2)

If the coefficients 𝑎0 , 𝑎1 , … , 𝑎𝑘 in (2) are constants, then the equation (2)

defines a linear difference equation of order 𝑘 with constant coefficients.

The solution of thee difference equation (2) consists of a solution

𝑢𝑛 (𝐻) , called complementary solution of the homogeneous part

𝑎0 𝑢𝑛 +𝑘 + 𝑎1 𝑢𝑛+𝑘−1 + ⋯ , 𝑎𝑘 𝑢𝑛 = 0 (3)

And a particular solution 𝑢𝑛 (𝑝) of the inhomogeneous part.

The general solution of (2) is written as

𝑢𝑛 = 𝑢𝑛 (𝐻) + 𝑢𝑛 𝑝
(4)

For solving (3), we assume a solution of the form 𝑢𝑛 = 𝐴𝜉 𝑛 , where

𝐴 ≠ 0 is a constant.

Substituting in (3), we get

𝐴 𝑎0 𝜉 𝑘 + 𝑎1 𝜉 𝑘−1 + ⋯ + 𝑎0 𝜉 𝑛 = 0

or 𝑎0 𝜉 𝑘 + 𝑎1 𝜉 𝑘−1 + ⋯ + 𝑎0 = 0 (5)

which is a polynomial of degree 𝑘. The equation (5) is called

characteristic equation of the difference equation (3).

Example: 5.2.1
Find the general solution of the difference equation

i. ∆2 𝑢𝑛 − 3∆𝑢𝑛 + 2𝑢𝑛 = 0

293
 ii. ∆2 𝑢𝑛 + ∆𝑢𝑛 + 1 4 𝑢𝑛 = 0
SPACE FOR HINTS

iii. ∆2 𝑢𝑛 − 2∆𝑢𝑛 + 2𝑢𝑛 = 0

iv. ∆2 𝑢𝑛+1 − 1 3 ∆2 𝑢𝑛 = 0

Is the solution bound? We substitute for the forward differences in the

differential equations.

Solution:
We know that, the forward difference table for 3 values.

∆ ∆2

𝑢𝑛

𝑢𝑛+1 − 𝑢𝑛

𝑢𝑛 +1 𝑢𝑛+2 − 𝑢𝑛+1 − 𝑢𝑛 +1 − 𝑢𝑛 = 𝑢𝑛+2 − 2𝑢𝑛 +1 − 𝑢𝑛

𝑢𝑛+2 − 𝑢𝑛 +1

𝑢𝑛+22

i. Given ∆2 𝑢𝑛 − 3∆𝑢𝑛 + 2𝑢𝑛 = 0

𝑢𝑛 +2 − 2𝑢𝑛+1 + 𝑢𝑛 − 3 𝑢𝑛+1 − 𝑢𝑛 + 2𝑢𝑛 = 0

𝑢𝑛+2 − 5𝑢𝑛+1 + 6𝑢𝑛 = 0 (1)

Now, substituting 𝑢𝑛 = 𝐴𝜉 𝑕 , where 𝐴 ≠ 0, a constant. Then the

characteristic equation is,

1 ⟹ 𝐴𝜉 𝑛+2 − 5𝐴𝜉 𝑛+1 + 6𝐴𝜉 𝑛 = 0

𝐴𝜉 𝑛 𝜉 2 − 5𝜉 + 6 = 0

⟹ 𝜉 2 − 5𝜉 + 6 = 0

𝜉 = 2, 3

∴ Roots are 2 and 3. Here the roots are real and distinct.

∴ The general solution of the difference equation is,

𝑢𝑛 = 𝑐1 2𝑛 + 𝑐2 3𝑛

294
Since 𝑢𝑛 → ∞ 𝑎𝑠 𝑛 → ∞,
SPACE OF HINTS
Hence the solution is unbounded.

ii. Given ∆2 𝑢𝑛 + ∆𝑢𝑛 + 1 4 𝑢𝑛 = 0

𝑢𝑛+2 − 2𝑢𝑛+1 + 𝑢𝑛 + 𝑢𝑛+1 − 𝑢𝑛 + 1 4 𝑢𝑛 = 0

𝑢𝑛+2 − 5𝑢𝑛+1 + 6𝑢𝑛 = 0 (2)

Substituting 𝑢𝑛 = 𝐴𝜉 𝑕 , we get,

(2) ⟹ 𝐴𝜉 𝑛+2 − 𝐴𝜉 𝑛 +1 + 1 4 𝐴𝜉 𝑛 = 0

1
𝐴𝜉 𝑛 𝜉 2 − 𝜉 + =0
4
1
⟹ 𝜉2 − 𝜉 + =0
4
1 1
𝜉= ,
2 2
1 1
Roots are and . Here the roots are real and equal.
2 2

∴ The general solution of the difference equation is

𝑛
1
𝑢𝑛 = 𝑐1 + 𝑛𝑐2 .
2
Since 𝑢𝑛 → 0 𝑎𝑠 𝑛 → ∞, hence the solution is bounded.

iii. Given ∆2 𝑢𝑛 − 2∆𝑢𝑛 + 2𝑢𝑛 = 0

𝑢𝑛 +2 − 2𝑢𝑛+1 + 𝑢𝑛 − 2 𝑢𝑛+1 − 𝑢𝑛 + 2𝑢𝑛 = 0

𝑢𝑛+2 − 5𝑢𝑛+1 + 6𝑢𝑛 = 0 3

Substituting 𝑢𝑛 = 𝐴𝜉 𝑕 , we get,

(2) ⟹ 𝐴𝜉 𝑛 +2 − 4𝐴𝜉 𝑛 +1 + 5𝐴𝜉 𝑛 = 0

𝐴𝜉 𝑛 𝜉 2 − 4𝜉 + 5 = 0

⟹ 𝜉 2 − 4𝜉 + 5 = 0

295
 4 ± 16 − 20
𝜉= =2±𝑖
SPACE FOR HINTS 2
Here the roots are imaginary

∴ The solution 𝑢𝑛 = 𝑟 𝑛 𝑒 𝑖𝑛𝜃 , 𝑟 = 𝛼 2 + 𝛽2 , where

𝛼 = coefficient of real part
𝛽 = coefficient of imaginary part

∴ 𝑟 = 4+1= 5
𝛽
tan 𝜃 = 𝛼

𝜃 = tan−1 1 2 = 0.4636

𝑛 𝑛
∴ 𝑢𝑛 = 5 𝑒 𝑖𝑛𝜃 = 5 (𝑐1 cos 𝑛𝜃 + 𝑐2 cos 𝑛 𝜃
𝑛
= 5 cos n 0.4636 + 𝑖 sin 𝑛 0.4636

as 𝑛 → ∞ , 𝑢𝑛 → ∞

∴ The solution is unbounded.

iv. Given ∆2 𝑢𝑛+1 − 1 3 ∆2 𝑢𝑛 = 0

∆2 𝑢𝑛 = 𝑢𝑛+2 − 2𝑢𝑛+1 + 𝑢𝑛
𝑠𝑖𝑚𝑖𝑙𝑎𝑟𝑙𝑦
∆2 𝑢𝑛+1 = 𝑢𝑛+3 − 2𝑢𝑛+2 + 𝑢𝑛+1

𝑢𝑛+3 − 2𝑢𝑛+2 + 𝑢𝑛+1 − 1 3 𝑢𝑛+3 − 2𝑢𝑛+2 + 𝑢𝑛+1 = 0

1 2 1
𝑢𝑛+3 − 𝑢𝑛+2 2 + + 𝑢𝑛+1 1 + − 𝑢𝑛 = 0
3 3 3

𝑢𝑛+3 − 7 3 𝑢𝑛+2 + 5 3 𝑢𝑛+1 − 1 3 𝑢𝑛 = 0

⟹ 3𝑢𝑛+3 − 7𝑢𝑛+2 + 5𝑢𝑛+1 − 𝑢𝑛 = 0

Substituting 𝑢𝑛 = 𝐴𝜉 𝑕 , (∴ 𝐴 ≠ 0)

⟹ 3𝐴𝜉 𝑛 +3 − 7𝐴𝜉 𝑛 +2 + 5𝐴𝜉 𝑛+1 − 𝐴𝜉 𝑛 = 0

296
 ⟹ 𝐴𝜉 𝑛 3𝜉 3 − 7𝜉 2 + 5𝜉 − 1 = 0
SPACE OF HINTS
3 2
⟹ 3𝜉 − 7𝜉 + 5𝜉 − 1 = 0

⟹ 𝜉 =1, 3𝜉 2 − 4𝜉 + 1 = 0

𝜉 = 1, 1, 1 3

Here one root is distinct and also two roots are real and equal.
𝑛
𝑛
𝑢𝑛 = 𝑐1 + 𝑛𝑐2 (1) + 𝑐3 1 as 𝑛 → ∞ , 𝑢𝑛 → ∞
3

∴ The solution is unbounded.

Example: 5.2.2

Solve the difference equation ∆2 𝑦𝑗 + 3∆𝑦𝑗 − 4𝑦𝑗 = 𝑗 2 with the initial

conditions 𝑦0 = 0, 𝑦2 = 2.

Solution:
Substituting the forward difference formula in the given equation,
we have

∆2 𝑦𝑗 = 𝑦𝑗 +2 − 2𝑦𝑗 +1 + 𝑦𝑗

∆𝑦𝑗 = 𝑦𝑗 +1 − 𝑦𝑗

∴ The given equation becomes

𝑦𝑗 +2 − 2𝑦𝑗 +1 + 𝑦𝑗 + 3 𝑦𝑗 +1 − 𝑦𝑗 − 4𝑦𝑗 = 𝑗 2

⟹ 𝑦𝑗 +2 + 𝑦𝑗 +1 − 6𝑦𝑗 = 𝑗 2 (1)

The general solution of the difference equation is of the form

𝑦𝑗 = 𝑦 (𝐻) + 𝑦 (𝑃)

where𝑦 (𝐻) is the complementary solution and 𝑦 (𝑃) is the particular

solution.

Substitute 𝑦𝑗 = 𝐴𝜉𝑗 ,

297
 ∴ The auxiliary equation is 𝑦𝑗 +2 + 𝑦𝑗 +1 − 6𝑦𝑗 = 0
SPACE FOR HINTS
𝐴𝜉𝑗 +2 + 𝐴𝜉𝑗 +1 − 6𝐴𝜉𝑗 = 0

⟹ 𝐴𝜉𝑗 𝜉 2 + 𝜉1 − 6 = 0

⟹ 𝜉 2 + 𝜉1 − 6 = 0

𝜉 = 2, −3

∴ The roots are real and distinct.

∴ The complementary solution is 𝑦 (𝐻) = 𝑐1 −3 𝑗

+ 𝑐2 2 𝑗

To find the particular solution:

We assume that the particular solution in the form

𝑝
𝑦 = 𝑎𝑗 2 + 𝑏𝑗 + 𝑐 and substitute into the equation (1). We get,

𝑎(𝑗 + 2)2 + 𝑏(𝑗 + 1) + 𝑐 + 𝑎(𝑗 + 1)2 + 𝑏(𝑗 + 1) + 𝑐

− 6 𝑎𝑗 2 + 𝑏𝑗 + 𝑐 = 𝑗 2

⟹ −4𝑎𝑗 2 + 𝑗 6𝑎 − 4𝑏 + 5𝑎 + 3𝑏 − 4𝑐 = 𝑗 2

Comparing 𝑗 2 coefficient on both sides, we get

−4𝑎 = 1 ⟹ 𝑎 = − 1 4

Comparing 𝑗 coefficient on both sides,

6𝑎 6 −1 −3
6𝑎 − 4𝑏 = 0 ⟹ 𝑏 = = × ⟹𝑏=
4 4 4 8
Comparing constant on both sides,

5𝑎 + 3𝑏 − 4𝑐 = 0 ⟹ 5 −1 4 + 3 −3 8 − 4𝑐 = 0

1 −5 9 −19
⟹𝑐= − ⟹𝑐=
4 4 8 32

𝑝
−1 2 −3 −19
∴ 𝑦 = 𝑎𝑗 2 + 𝑏𝑗 + 𝑐 ⟹ 𝑦 𝑝
= 𝑗 + 𝑗+ .
4 8 32
Hence the general solution becomes, 𝑦𝑗 = 𝑦 (𝐻) + 𝑦 (𝑃)

298
 𝑗 𝑗
−1 2 −3 −19
𝑦𝑗 = 𝑐1 −3 + 𝑐2 2 + 𝑗 + 𝑗+ (2) SPACE OF HINTS
4 8 32
Given the initial conditions are 𝑦0 = 0, 𝑦2 = 2

Put 𝑗 = 0 𝑖𝑛 2
−19
⟹ 𝑦0 = 𝑐1 + 𝑐2 + = 0 ⟹ 𝑐1 + 𝑐2
32
−19
= (3)
32

𝑗 = 2 𝑖𝑛 2

2 2
−1 2 −3 −19
⟹ 𝑦2 = 𝑐1 −3 + 𝑐2 2 + 2 + .2 +
4 8 32
75 139
2 = 9𝑐1 + 4𝑐2 − ⟹ 9𝑐1 + 4𝑐2 = (4)
32 32
Solving (3) & (4), we get
76
3 × 4 ⟹ 𝑐1 + 𝑐2 =
32
139
4 ⟹ 9𝑐1 + 4𝑐2 =
32
−63 −63
−5𝑐1 = ⟹ 𝑐1 =
32 160
−63 76 32
3 ⟹ + 𝑐2 = ⟹ 𝑐2 =
160 32 160

𝑗 𝑗
−1 2 −3 −19
∴ 𝑦𝑗 = 𝑐1 −3 + 𝑐2 2 + 𝑗 + 𝑗+
4 8 32
−63 𝑗
32 𝑗
−1 2 −3 −19
= −3 + 2 + 𝑗 + 𝑗+
160 160 4 8 32
1 𝑗 𝑗
𝑦𝑗 = 63 −3 + 32 2 − 40𝑗 2 − 60𝑗 − 95 .
160
Example: 5.2.3
Find the range for 𝛼, so that the roots of the characteristic equation
of the difference equations

i. 1 − 5𝛼 𝑦𝑛 +2 − 1 + 8𝛼 𝑦𝑛 +1 + 𝛼𝑦𝑛 = 0

299
 ii. 1 − 9𝛼 𝑦𝑛 +3 − 1 + 19𝛼 𝑦𝑛 +2 + 5𝛼𝑦𝑛 +1 − 𝛼𝑦𝑛 = 0
SPACE FOR HINTS Are less than one in magnitude.

Solution:
i. The characteristic equation of the difference equation isobtain by
putting 𝑦𝑛 = 𝐴𝜉 𝑕

∴The given equation is 1 − 5𝛼 𝐴𝜉 𝑛+2 − 1 + 8𝛼 𝐴𝜉 𝑛 +1 + 𝛼𝜉 𝑛 = 0

𝐴𝜉 𝑛 1 − 5𝛼 𝐴𝜉 2 − 1 + 8𝛼 𝜉1 + 𝛼 = 0

⟹ 1 − 5𝛼 𝜉 𝑛 +2 − 1 + 8𝛼 𝜉1 + 𝛼 = 0 (1)

Since the coefficient involve a parameter 𝛼, we use Routh-Hurwitz

criterion and setting
1+𝑧
𝜉= 𝑖𝑛 (1),
1−𝑧

2
1+𝑧 1+𝑧
1 − 5𝛼 − 1 + 8𝛼 +𝛼 =0
1−𝑧 1−𝑧
2 2
1 − 5𝛼 𝑧 + 1 − 1 + 8𝛼 1 + 𝑧 1 − 𝑧 + 𝛼 1 − 𝑧 =0

1 − 5𝛼 (1 + 𝑧 2 + 2𝑧) − 1 + 8𝛼 1 − 𝑧 2 + 𝛼 1 + 𝑧 2 − 2𝑧 = 0

1 + 𝑧 2 + 2𝑧 − 5𝛼 − 5𝛼𝑧 2 − 10𝛼𝑧 − 1 + 𝑧 2 − 8𝛼 + 8𝛼𝑧 2 + 𝛼 + 𝛼𝑧 2

−2𝛼𝑧 = 0

𝑧 2 2 + 4𝛼 + 𝑧 2 − 12𝛼 − 12𝛼 = 0

The Routh-Hurwitz criterion is satisfied if

2 + 4𝛼 > 0, 2 − 12𝛼 > 0, −12𝛼 > 0.

The second and third inequalities are satisfied for all 𝛼 < 0.

From the first inequality 2 + 4𝛼 > 0

⟹ 𝛼 = −1 2 ∴ ∀ 𝛼 = −1 2 , 0

For 𝛼 = 0, 1 ⟹ 𝜉 2 − 𝜉 = 0

𝜉 𝜉 − 1 = 0 ⟹ 𝜉 = 0, 1

This indicate that 𝜉 = 1.

300
 ii. The characteristic equation of the difference equation is obtained
SPACE OF HINTS
by putting 𝑦𝑛 = 𝐴𝜉 𝑕

∴ Given equation becomes,

1 − 9𝛼 𝐴𝜉 𝑛+3 − 1 + 19𝛼 𝐴𝜉 𝑛+2 + 5𝛼𝐴𝜉 𝑛 +1 − 𝛼𝐴𝜉 𝑛 = 0

⟹ 𝐴𝜉 𝑛 1 − 9𝛼 𝜉 3 − 1 + 19𝛼 𝜉 2 + 5𝛼𝜉 − 𝛼 = 0

⟹ 1 − 9𝛼 𝜉 3 − 1 + 19𝛼 𝜉 2 + 5𝛼𝜉 − 𝛼 = 0 (2)

Since the coefficient involve a parameter 𝛼, we use the Routh-Hurwitz

1+𝑧
criterion and setting 𝜉 = 𝑖𝑛 (2), we get
1−𝑧

3 2
1+𝑧 1+𝑧 1+𝑧
1 − 9𝛼 − 1 + 19𝛼 + 5𝛼 −𝛼 =0
1−𝑧 1−𝑧 1−𝑧
3 2
⟹ 1 − 9𝛼 1 + 𝑧 − 1 + 19𝛼 1 + 𝑧 1−𝑧
2 3
+ 5𝛼 1 + 𝑧 1 − 𝑧 −𝛼 1−𝑧 =0

⟹ 1 − 9𝛼 1 + 𝑧 3 + 3𝑧 2 + 3𝑧 − 1 + 19𝛼 1 + 𝑧 2 + 2𝑧 1 − 𝑧
+ 5𝛼 1 + 𝑧 (1 + 𝑧 2 − 2𝑧) − 𝛼(1 − 𝑧 3 + 3𝑧 2 − 3𝑧) = 0

⟹ 𝑧 3 2 + 16𝛼 + 𝑧 2 4 − 16𝛼 + 2 − 48𝛼 − 24𝛼 = 0

By applying Routh-Hurwitz criterion if,

𝑎0 = 2 + 16𝛼 > 0 , 𝑎1 = 4 − 16𝛼 > 0 , 𝑎2 = 2 − 48𝛼 > 0 ,

𝑎3 = −24𝛼 > 0

𝑎1 𝑎2 − 𝑎0 𝑎3 = 4 − 16𝛼 2 − 48𝛼 − (2 + 16𝛼)(−24𝛼)

= 8 1 − 4𝛼 1 − 24𝛼 + 48 1 + 8𝛼 𝛼

= 8 1 − 28𝛼 + 96𝛼 2 + 6𝛼 + 48𝛼 2

= 8 144𝛼 2 − 22𝛼 + 1 > 0

From 𝑎0 , 2 + 16𝛼 = 0 ⟹ 𝛼 = −1 8

∴ 𝛼 lies between −1 8 , 0 .

5.3: Numerical Methods

301
 Euler Method
SPACE FOR HINTS
𝑢𝑗 +1 = 𝑢𝑗 + 𝑕𝑓𝑗 , 𝑗 = 0,1, … , 𝑁 − 1 𝑤𝑕𝑒𝑟𝑒 𝑓𝑗 = 𝑓 𝑡𝑗 , 𝑢𝑗
𝑕2
The truncation error formula is 𝑇 ≤ 𝑀2 where 𝑀2 = max 𝑡 0 ,𝑏 𝑢′′ 𝜉
2

and

𝑇 = max 𝑇𝑗 +1
𝑡 0 ,𝑏

Example: 5.3.1
Show that in Euler method the bound of the truncation error,
when applied to the test equation 𝑢′ = 𝜆𝑢 , 𝑢 𝑎 = 𝐵 , 𝜆 > 0 can be
written as,
𝑕𝑀
𝑢 𝑡𝑗 − 𝑢(𝑡𝑗 , 𝑢) ≤ exp 𝜆 𝑡𝑖 − 𝑎 −1
2𝜆
where 𝑀 = max 𝑢′′ (𝑡) . Generalize the result when applied to the
problem 𝑢′ = 𝑓(𝑡, 𝑢) , 𝑢 𝑎 = 𝐵 .

Solution:
𝑕𝑀
To prove: 𝑢 𝑡𝑗 − 𝑢(𝑡𝑗 , 𝑢) ≤ exp 𝜆 𝑡𝑖 − 𝑎 −1
2𝜆

Given 𝑢′ = 𝜆𝑢 , 𝑢 𝑎 = 𝐵 , 𝜆 > 0

The Euler formula is, 𝑢𝑗 +1 = 𝑢𝑗 + 𝑕𝑓𝑗 , 𝑕𝑒𝑟𝑒 𝑓𝑗 = 𝑓 𝑡𝑗 , 𝑢𝑗 = 𝜆𝑢𝑗

𝑢𝑗 +1 = 𝑢𝑗 + 𝜆𝑕𝑢𝑗

𝑢𝑗 +1
= 𝑢𝑗 1
+ 𝜆𝑕 (1)

The truncation error formula is, adding the truncation error 𝑇𝑗 +1 to

equation (1), we get

𝑢 𝑡𝑗 +1 = 𝐸 𝜆𝑕 𝑢 𝑡𝑗 + 𝑇𝑗 +1

𝑢 𝑡𝑗 +1 = 1 + 𝜆𝑕 𝑢 𝑡𝑗 + 𝑇𝑗 +1 (2)

where 𝑇𝑗 +1 is the truncation error given by,

302
 𝑕2 𝑢′′ (𝜉) 𝑕2 𝑀
𝑇𝑗 +1 = ≤ SPACE OF HINTS
2 2
2 − 1 ⟹ 𝑢 𝑡𝑗 +1 − 𝑢𝑗 +1 = 1 + 𝜆𝑕 𝑢 𝑡𝑗 − 𝑢𝑗 + 𝑇𝑗 +1 (3)

𝜖𝑗 +1 = 1 + 𝜆𝑕 𝜖𝑗 + 𝑇𝑗 +1 , 𝑤𝑕𝑒𝑟𝑒 𝜖𝑗 = 𝑢 𝑡𝑗 − 𝑢𝑗 .

Hence 𝜖𝑗 +1 = 1 + 𝜆𝑕 𝜖𝑗 + 𝑇𝑗 +1 (4)

Let 𝐴 = 1 + 𝜆𝑕,

𝑕2 ′′ 𝑕2 𝑀
𝑇𝑗 +1 = 𝑢 (𝜉) ≤ (5)
2 2
where 𝑀 = max 𝑡 0 ,𝑏 𝑢′′ 𝑡
𝑎≤𝑡≤𝑏

max 𝜖𝑗 = 𝐸𝑗 , max 𝜖𝑗 +1 = 𝐸𝑗 +1 & max 𝑇𝑗 +1 = 𝑇.

𝑕2 𝑀
4 ⟹ 𝐸𝑗 +1 ≤ 𝐴𝐸𝑗 + 𝑇 𝑤𝑕𝑒𝑟𝑒 𝑇 =
2
Setting 𝑗 = 0,1,2, … we get

𝐸1 ≤ 𝐴𝐸0 + 𝑇

𝐸2 ≤ 𝐴𝐸1 + 𝑇 = 𝐴 𝐴𝐸0 + 𝑇 + 𝑇
= 𝐴2 𝐸0 + (1 + 𝐴)𝑇

Similarly, 𝐸3 ≤ 𝐴𝐸2 + 𝑇 = 𝐴3 𝐸0 + (1 + 𝐴 + 𝐴2 )𝑇 and so on.

In general, 𝐸𝑗 ≤ 𝐴𝑗 𝐸0 + (1 + 𝐴 + ⋯ + 𝐴𝑗 −1 )𝑇

𝐴𝑗 −1
𝑗
𝐸𝑗 ≤ 𝐴 𝐸0 + 𝑇 𝑤𝑕𝑒𝑟𝑒 𝐴 ≠ 1 6
𝐴−1

Let 𝐸0 = 0, 𝑖𝑒) there is no initial error.

2
𝑛𝑥 𝑛𝑥
∴ exp 𝑛𝑥 = 1 + + +⋯
1! 2!
𝑗
1 + 𝜆𝑕 < exp 𝜆𝑕𝑗 < exp 𝜆 𝑡𝑗 − 𝑎 , 𝜆 > 0.

Since 𝜆 > 0 , we get

𝑕2 𝑀
6 ⟹ 𝐸𝑗 ≤ exp 𝜆 𝑡𝑗 − 𝑎 −1
2𝜆𝑕

303
 𝑕𝑀
≤ exp 𝜆 𝑡𝑗 − 𝑎 −1
SPACE FOR HINTS 2𝜆
Since max 𝜖𝑗 = 𝐸𝑗 , we get

𝜖𝑗 ≤ 𝑢 𝑡𝑗 − 𝑢 𝑡𝑗 , 𝑕

𝑕𝑀
≤ exp 𝜆 𝑡𝑗 − 𝑎 −1
2𝜆
𝜕(𝜆𝑢 )
since 𝜆 = ,the result can be generalized as
𝜕𝑢

The result can be generalized as,

𝑕𝑀 𝜕𝑓
𝑢 𝑡𝑗 − 𝑢 𝑡𝑗 , 𝑕 ≤ exp 𝜆 𝑡𝑗 − 𝑎 − 1 𝑤𝑕𝑒𝑟𝑒 ≤ 𝐿.
2𝐿 𝜕𝑦

Example: 5.3.2

The system 𝑦 ′ = 𝑧 , 𝑧 ′ = −𝑏𝑦 − 𝑎𝑧 where 0 < 𝑎 < 2 𝑏 , 𝑏 > 0 is to

be integrated by Euler method with known initial values. What is the
largest step length 𝑕 for which all solutions of the corresponding
difference equation all bounded?

Solution:
Given 𝑦 ′ = 𝑧 , 𝑧 ′ = −𝑏𝑦 − 𝑎𝑧

The Euler formula is, 𝑢𝑗 +1 = 𝑢𝑗 + 𝑕𝑓𝑗 , 𝑕𝑒𝑟𝑒 𝑓𝑗 = 𝑓 𝑡𝑗 , 𝑢𝑗

𝑦𝑗 +1 = 𝑦𝑗 + 𝑕𝑧𝑗

𝑧𝑗 +1 = 𝑧𝑗 + 𝑕 −𝑏𝑦𝑗 − 𝑎𝑧𝑗 = −𝑏𝑕𝑦𝑗 − 𝑎𝑕𝑧𝑗 + 𝑧𝑗

𝑧𝑗 +1 = −𝑏𝑕𝑦𝑗 + 1 − 𝑎𝑕 𝑧𝑗

which may be written as,

𝑦𝑗 +1 1−𝜖 𝑕 𝑦𝑗
𝑧𝑗 +1 = −𝑏𝑕 1 − 𝑎𝑕 − 𝜖 𝑧𝑗
𝑢𝑗 +1 = 𝐴𝑢𝑗

1 𝑕 𝑦𝑗
where 𝐴 = , 𝑢𝑗 = 𝑧
−𝑏𝑕 1 − 𝑎𝑕 𝑗

The characteristic equation of 𝐴 is 𝐴 − 𝜉𝐼 = 0

304
 1−𝜉 𝑕
⟹ =0 SPACE OF HINTS
−𝑏𝑕 1 − 𝑎𝑕 − 𝜉

⟹ 1 − 𝜉 1 − 𝑎𝑕 − 𝜉 − −𝑏𝑕2 = 0

⟹ 1 − 𝑎𝑕 − 2𝜉 − 𝑎𝑕𝜉 + 𝜉 2 + 𝑏𝑕2 = 0

⟹ 𝜉 2 − 2 − 𝑎𝑕 𝜉 + 1 − 𝑎𝑕 + 𝑏𝑕2 = 0 ∗
1+𝑧
Putting 𝜉 = in ∗ , we get,
1−𝑧

2
1+𝑧 1+𝑧
− 2 − 𝑎𝑕 + 1 − 𝑎𝑕 + 𝑏𝑕2 = 0
1−𝑧 1−𝑧
1 + 𝑧 2 + 2𝑧 − 2 − 𝑎𝑕 1 − 𝑧 2 + 1 + 𝑧 2 − 2𝑧 − 𝑎𝑕 − 𝑎𝑕2
+ 2𝑎𝑕𝑧 + 𝑏𝑕2 + 𝑏𝑕2 𝑧 2 − 2𝑏𝑕2 𝑧 = 0

4 − 2𝑎𝑕 + 𝑏𝑕2 𝑧 2 + 2𝑕 𝑎 − 𝑏𝑕 𝑧 + 𝑏𝑕2 = 0

The Routh-Hurwitz Criterion, is satisfied.

𝑖𝑒) 𝑎0 𝑧 2 + 𝑎1 𝑧 + 𝑎2 = 0

𝑎0 = 4 − 2𝑎𝑕 + 𝑏𝑕2 > 0 1

𝑎1 = 2𝑕 𝑎 − 𝑏𝑕 > 0 2

𝑎2 = 𝑏𝑕2
>0 3

1 ⟹ 4 + 𝑏𝑕2 − 4𝑕 𝑏 + 4𝑕 𝑏 − 2𝑎𝑕 > 0

2
⟹ 2−𝑕 𝑏 + 2𝑕 2 𝑏 − 𝑎 > 0

2 ⟹ 𝑎 − 𝑏𝑕 > 0 ⟹ 𝑎 > 𝑏𝑕 ⟹ 𝑕 < 𝑎 𝑏

3 ⟹ 𝑏𝑕2 > 0𝑖𝑒) 𝑏 > 0

The third inequality satisfied.

Hence we obtain 𝑕 < 𝑎 𝑏

For 𝑕 < 𝑎 𝑏 𝑜𝑟 𝑎 = 𝑏𝑕 .

The characteristic equation becomes,

∗ ⟹ 𝜉 2 − 2 − 𝑎𝑕 𝜉 + 1 − 𝑏𝑕2 + 𝑏𝑕2 = 0

305
 ⟹ 𝜉 2 − 2 − 𝑎𝑕 𝜉 + 1 = 0
SPACE FOR HINTS
2 − 𝑏𝑕2 ± 2 − 𝑏𝑕2 2 −4
𝜉=
2
∴ 2 − 𝑏𝑕2 2
−4<0

∴ The roots are complex pail and 𝜉 = 1.

∴ The largest step length is 𝑕 ≤ 𝑎 𝑏

Backward Euler method

The formula for backward Euler method is, 𝑢𝑗 +1 = 𝑢𝑗 + 𝑕𝑓𝑗 +1

The truncation error at 𝑡 = 𝑡𝑗 +1 is given by

−𝑕2 ′′
𝑇𝑗 +1 = 𝑢 𝜉 , 𝑡𝑗 < 𝜉 < 𝑡𝑗 +1
2
Example:5.3.3

Solve the initial value problem 𝑢′ = −2𝑡𝑢2 , 𝑢 0 = 1 with 𝑕 = 0.2 on

the interval 0, 0.4 using the backward Euler method.

Solution:

Given 𝑢′ = −2𝑡𝑢2 , 𝑢 0 = 1 , 𝑕 = 0.2

𝑖𝑒) 𝑓 = −2𝑡𝑢2
The formula for backward Euler method is,

𝑢𝑗 +1 = 𝑢𝑗 + 𝑕𝑓𝑗 +1 1

Here 𝑓𝑗 +1 = −2𝑡𝑗 +1 𝑢𝑗 +1 2

1 ⟹ 𝑢𝑗 +1 = 𝑢𝑗 − 2𝑕 𝑡𝑗 +1 𝑢𝑗 +1 2

This is an implicit equation in 𝑢𝑗 +1

We generally use Newton Rapson method, given 𝑕 = 0.2

𝑓 𝑢𝑗 +1 = 𝑢𝑗 +1 − 𝑢𝑗 + 2𝑕 𝑡𝑗 +1 𝑢𝑗 +1 2

𝐹 𝑢𝑗 +1 = 𝑢𝑗 +1 − 𝑢𝑗 + 0.4 𝑡𝑗 +1 𝑢𝑗 +1 2

𝐹 ′ 𝑢𝑗 +1 = 1 + 4𝑕 𝑡𝑗 +1 . 𝑢𝑗 +1

306
 𝐹 ′ 𝑢𝑗 +1 = 1 + 0.8 𝑡𝑗 +1 . 𝑢𝑗 +1 SPACE OF HINTS

The formula,
𝑠
𝑠+1 𝑠
𝐹(𝑢𝑗 +1 )
𝑢𝑗 +1 = 𝑢𝑗 +1 −
𝐹 ′ 𝑢𝑗 +1 𝑠

0
Taking 𝑢𝑗 +1 = 𝑢𝑗
0
𝑗 = 0 , 𝑡1 = 0.2 , 𝑢1 = 𝑢0 = 1 (𝑔𝑖𝑣𝑒𝑛 )
0
𝑠 = 0 ⟹ 𝐹 𝑢1 = 𝑢1 − 𝑢0 + 0.4 𝑡1 𝑢1 2 = 1 − 1 + 0.4 0.2 12

= 0.08

𝐹 ′ 𝑢1 0
= 1 + 0.8 0.2 1 = 1.16

1 0
𝐹 𝑢1 0 0.08
𝑢1 = 𝑢1 − ′ =1− = 0.9310
𝐹 𝑢1 0 1.16
𝑠 𝑠 𝑠
𝑗 𝑠 𝑢𝑗 +1 𝐹 𝑢𝑗 +1 𝐹 ′ 𝑢𝑗 +1 𝑠
𝑠+1 𝑠
𝐹(𝑢𝑗 +1 )
𝑢𝑗 +1 = 𝑢𝑗 +1 −
𝐹 ′ 𝑢𝑗 +1 𝑠

0 0 1 0.08 1.16 0.9310

0 1 0.9310 0.00034 1.14896 0.9307

0 2 0.9307 0.0690 1.14891 0.8706

0 3 0.8706 0.00054 1.13930 0.8701

∴ 𝑢 0.2 = 0.9307 = 𝑢1

For 𝑡 = 0.4 , 𝑢1 = 0.9307 = 𝑢2 0 , 𝑢0 = 1

𝑠 𝑠 𝑠+1
𝑗 𝑠 𝑢𝑗 +1 𝐹 𝑢𝑗 +1 𝐹 ′ 𝑢𝑗 +1 𝑠
𝑢𝑗 +1
𝑠
𝑠
𝐹(𝑢𝑗 +1 )
= 𝑢𝑗 +1 −
𝐹 ′ 𝑢𝑗 +1 𝑠

0 0 0.9307 0.13859 1.297824 0.82391

0 1 0.8239 0.10861 1.263648 0.82249

1 0 0.9307 0.13859 1.297824 0.82391

1 1 0.8239 0.0018 1.263648 0.82249

307
 1 2 0.82249 0.00003 1.263196 0.82247
SPACE FOR HINTS
𝑢 0.4 = 0.82247

Midpoint method
The equation may be written as, 𝑢𝑗 +1 = 𝑢𝑗 −1 + 2𝑕 𝑓𝑗 ,

𝑗 = 0,1, … , 𝑁 − 1. This is called midpoint method.

Truncation error,

𝑕3 ′′′
𝑇𝑗 +1 = 𝑢 𝜉 , 𝑡𝑗 −1 < 𝜉 < 𝑡𝑗 +1
3
𝑕3
𝑇𝑗 +1 ≤ 𝑀3 𝑤𝑕𝑒𝑟𝑒 𝑀3 = max 𝑢′′′ 𝑡
3 𝑡∈ 𝑡 0 ,𝑏

Example: 5.3.4
Solve the initial value problem 𝑢′ = −2𝑡𝑢2 , 𝑢 0 = 1using the
mid-point method with 𝑕 = 0.2 over the interval 0,1 . Use the Tayler
series method of second order to compute 𝑢 0.2 . Determine the
percentage relative error at 𝑡 = 1.

Solution:

Given 𝑢′ = −2𝑡𝑢2 , 𝑓𝑗 = −2𝑡𝑗 𝑢𝑗 2

The midpoint method 𝑢𝑗 +1 = 𝑢𝑗 −1 + 2𝑕 𝑓𝑗 , 𝑗 = 0,1, …

𝑢𝑗 +1 = 𝑢𝑗 −1 − 4𝑕𝑡𝑗 𝑢𝑗 2

𝑕 = 0.2, 𝑢𝑗 +1 = 𝑢𝑗 −1 − 0.8 𝑕𝑡𝑗 𝑢𝑗 2 , 𝑗 = 0,1,2,3,4. 1

Calculating 𝑢1 from the Tayler series method of 2nd order. We have,

𝑕2 ′′ ′
𝑢1 = 𝑢 𝑕 = 𝑢 0 + 𝑕𝑢 0 + 𝑢 0 2
2
Where 𝑢 0 = 1, 𝑢′ 𝑡 = −2𝑡𝑢2 , 𝑢′′ = −2𝑢2

𝑢′ 0 = 0 , 𝑢′′ 0 = −2

Substituting values in (2),

0.04
𝑢1 = 𝑢 0.2 = 1 + 0 + −2 = 0.96
2

308
Now, we obtain from (1),𝑢2 , 𝑢3 , 𝑢4 , 𝑢5 ,we have
SPACE OF HINTS
Put 𝑗 = 1, 𝑢0 = 1, 𝑢1 = 0.96 , 𝑡1 = 0.2

𝑢 0.4 = 𝑢2 = 𝑢0 − 0.8 𝑡1 𝑢1 2
2
= 1 − 0.8 0.2 0.96

𝑢 0.4 = 𝑢2 = 0.852544

Put 𝑗 = 2, 𝑢1 = 0.96, 𝑢2 = 0.852544 , 𝑡2 = 0.4

𝑢 0.6 = 𝑢3 = 𝑢1 − 0.8 𝑡2 𝑢2 2
2
= 0.96 − 0.8 0.4 0.852544

𝑢 0.6 = 𝑢3 = 0.7274139

Put 𝑗 = 3, 𝑢2 = 0.852544, 𝑢3 = 0.7274139 , 𝑡3 = 0.6

𝑢 0.8 = 𝑢4 = 𝑢2 − 0.8 𝑡3 𝑢3 2
2
= 0.852544 − 0.8 0.6 0.7274139

𝑢 0.8 = 𝑢4 = 0.5985611

Put 𝑗 = 4, 𝑢3 = 0.7274139, 𝑢4 = 0.5985611 , 𝑡4 = 0.8

𝑢 1.0 = 𝑢5 = 𝑢3 − 0.8 𝑡4 𝑢4 2
2
= 0.7274139 − 0.8 0.8 0.5985611

𝑢 1.0 = 𝑢5 = 0.4981176

The exact solution is 𝑢 𝑡 = 1 1 + 𝑡 2

1
∴ 𝑢 1 = = 0.5
1+1
The percentage error is defined as
𝑢 − 𝑢∗ 𝑢 𝑖𝑠 𝑡𝑕𝑒 𝑒𝑥𝑎𝑐𝑡 𝑠𝑜𝑙𝑢𝑡𝑖𝑜𝑛
× 100 𝑤𝑕𝑒𝑟𝑒 ∗
𝑢 𝑢 𝑖𝑠 𝑡𝑕𝑒 𝑎𝑝𝑝𝑟𝑜𝑥𝑖𝑚𝑎𝑡𝑒 𝑠𝑜𝑙𝑢𝑡𝑖𝑜𝑛

0.5 − 0.4981176
∴ 𝑝. 𝑒 = × 100 = 0.38 %
0.5
Example: 5.3.5
Find singlestep methods for the differential equation

309
 𝑦 ′ = 𝑓 𝑡, 𝑦 which produce exact results for
SPACE FOR HINTS i. 𝑦 𝑡 = 𝑎 + 𝑏𝑒 −𝑡
ii. 𝑦 𝑡 = 𝑎 + 𝑏 cos 𝑡 + 𝑐 sin 𝑡

Solution:
i. For eliminating the arbitrary parameter 𝑎 and 𝑏 we need 3
relations 𝑦 𝑡 in 𝑦 ′ 𝑡 and to eliminate 𝑎 and 𝑏.

Now, assume a relation between 𝑦 𝑡𝑗 +1 , 𝑦 𝑡𝑗 &𝑦 ′ 𝑡𝑗 to obtain a

singlestep method.

We have,

𝑦 𝑡𝑗 +1 = 𝑎 + 𝑏𝑒 −𝑡 𝑗 +1 ⟹ 𝑦𝑗 +1 − 𝑎 − 𝑏𝑒 −𝑡 𝑗 +1 = 0

𝑦 𝑡𝑗 = 𝑎 + 𝑏𝑒 −𝑡 𝑗 ⟹ 𝑦𝑗 − 𝑎 − 𝑏𝑒 −𝑡 𝑗 = 0

𝑦 ′ 𝑡𝑗 = −𝑏𝑒 −𝑡 𝑗 ⟹ 𝑦𝑗 ′ + 𝑏𝑒 −𝑡 𝑗 = 0

𝑦𝑗 +1 −1 −𝑒 −𝑡 𝑗 +1
𝑜𝑟 𝑦𝑗 −1 −𝑒 −𝑡 𝑗 = 0
𝑦𝑗 ′ 0 −𝑒 −𝑡 𝑗

𝑦𝑗 +1 −𝑒 −𝑡 𝑗 + 1 𝑦𝑗 𝑒 −𝑡 𝑗 + 𝑦𝑗 ′ 𝑒 −𝑡 𝑗 − 𝑒 −𝑡 𝑗 +1 𝑦𝑗 ′ = 0

⟹ − 𝑦𝑗 +1 𝑒 −𝑡 𝑗 + 𝑦𝑗 𝑒 −𝑡 𝑗 + 𝑦𝑗 ′ 𝑒 −𝑡 𝑗 − 𝑦𝑗 ′ 𝑒 −𝑡 𝑗 +1 = 0

𝑒 −𝑡 𝑗 +1
⟹ − 𝑦𝑗 +1 + 𝑦𝑗 + 𝑦𝑗 − 𝑦𝑗 ′ ′
= 0 (since dividev by𝑒 −𝑡 𝑗 )
𝑒 −𝑡 𝑗
⟹ − 𝑦𝑗 +1 + 𝑦𝑗 + 𝑦𝑗 ′ 1 − 𝑒 −𝑕 = 0

⟹ 𝑦𝑗 + 𝑦𝑗 ′ 1 − 𝑒 −𝑕 = 𝑦𝑗 +1

ii. Given 𝑦 𝑡 = 𝑎 + 𝑏 cos 𝑡 + 𝑐 sin 𝑡

We have three arbitrary parameters, 𝑎, 𝑏 & 𝑐

We require four relations to eliminate the parameter

𝑦 𝑡𝑗 +1 = 𝑎 + 𝑏 cos(𝑡𝑗 +1 ) + 𝑐 sin(𝑡𝑗 +1 ) 1

𝑦 ′ 𝑡𝑗 +1 = −𝑏 sin(𝑡𝑗 +1 ) + 𝑐 cos 𝑡𝑗 +1 2

310
 𝑦 𝑡𝑗 = 𝑎 + 𝑏 cos 𝑡𝑗 +1 + 𝑐 sin 𝑡𝑗 +1 3 SPACE OF HINTS

𝑦 ′ 𝑡𝑗 = −𝑏 sin(𝑡𝑗 ) + 𝑐 cos 𝑡𝑗 4

Eliminating ‘𝑎’ from 1 & 3 ,

𝑦 𝑡𝑗 +1 − 𝑦 𝑡𝑗 = 𝑏 cos(𝑡𝑗 +1 ) − cos(𝑡𝑗 ) + 𝑐 sin(𝑡𝑗 +1 ) − sin(𝑡𝑗 )

𝑦 𝑡𝑗 +1 − 𝑦 𝑡𝑗 − 𝑏 cos 𝑡𝑗 +1 − cos 𝑡𝑗 − 𝑐 sin 𝑡𝑗 +1 − sin 𝑡𝑗 =0

2 ⟹ 𝑦𝑗 +1 ′ + 𝑏 sin(𝑡𝑗 +1 ) − 𝑐 cos 𝑡𝑗 +1 = 0

4 ⟹ 𝑦𝑗 ′ + 𝑏 sin(𝑡𝑗 ) − 𝑐 cos 𝑡𝑗 = 0

𝑦𝑗 +1 − 𝑦𝑗 − cos 𝑡𝑗 +1 − cos 𝑡𝑗 − sin 𝑡𝑗 +1 − sin 𝑡𝑗

𝑦 ′ 𝑡𝑗 +1 sin(𝑡𝑗 +1 ) − cos(𝑡𝑗 +1 ) =0
𝑦 ′ 𝑡𝑗 sin(𝑡𝑗 ) − cos(𝑡𝑗 )

⟹ 𝑦𝑗 +1 − 𝑦𝑗 − sin 𝑡𝑗 +1 cos 𝑡𝑗 + cos 𝑡𝑗 +1 sin 𝑡𝑗

+ cos 𝑡𝑗 +1 − cos 𝑡𝑗 − cos 𝑡𝑗 𝑦𝑗 +1 ′ − cos 𝑡𝑗 +1 𝑦𝑗 ′
− sin 𝑡𝑗 +1 − sin 𝑡𝑗 𝑦𝑗 +1 ′ sin 𝑡𝑗 − 𝑦𝑗 ′ sin 𝑡𝑗 +1 =0

⟹ 𝑦𝑗 +1 − 𝑦𝑗 sin 𝑡𝑗 − 𝑡𝑗 +1 − cos 𝑡𝑗 cos 𝑡𝑗 +1 𝑦𝑗 +1 ′

+ 𝑦𝑗 ′ cos2 𝑡𝑗 +1 + 𝑦𝑗 +1 ′ cos2 𝑡𝑗 − 𝑦𝑗 ′ cos 𝑡𝑗 cos 𝑡𝑗 +1
− 𝑦𝑗 +1 ′ sin 𝑡𝑗 +1 sin 𝑡𝑗 + 𝑦𝑗 ′ sin2 𝑡𝑗 +1 + 𝑦𝑗 +1 ′ sin2 𝑡𝑗
− 𝑦𝑗 ′ sin 𝑡𝑗 sin 𝑡𝑗 +1 = 0

⟹ 𝑦𝑗 +1 − 𝑦𝑗 sin 𝑡𝑗 − 𝑡𝑗 +1
− 𝑦𝑗 +1 ′ cos 𝑡𝑗 cos 𝑡𝑗 +1 − cos2 𝑡𝑗 +1
+ sin 𝑡𝑗 sin 𝑡𝑗 +1 − sin2 𝑡𝑗
+ 𝑦𝑗 ′ cos2 𝑡𝑗 +1 − cos 𝑡𝑗 cos 𝑡𝑗 +1 + sin2 𝑡𝑗 +1
− sin 𝑡𝑗 sin 𝑡𝑗 +1 =0

⟹ 𝑦𝑗 +1 − 𝑦𝑗 sin 𝑡𝑗 − 𝑡𝑗 +1 − 𝑦𝑗 +1 ′ cos 𝑡𝑗 − 𝑡𝑗 +1 − 1
+ 𝑦𝑗 ′ 1 − cos 𝑡𝑗 − 𝑡𝑗 +1 =0

311
 ⟹ 𝑦𝑗 +1 − 𝑦𝑗 sin −𝑕 − 𝑦𝑗 +1 ′ cos(−𝑕) − 1 + 𝑦𝑗 ′ 1 − cos(−𝑕)
SPACE FOR HINTS
=0

⟹ − sin 𝑕 𝑦𝑗 +1 − 𝑦𝑗 − 𝑦𝑗 +1 ′ cos 𝑕 − 1 + 𝑦𝑗 ′ 1 − cos 𝑕 = 0

⟹ − sin 𝑕 𝑦𝑗 +1 − 𝑦𝑗 = − 𝑦𝑗 +1 ′ + 𝑦𝑗 ′ 1 − cos 𝑕

1 − cos 𝑕
⟹ 𝑦𝑗 +1 − 𝑦𝑗 = 𝑦𝑗 +1 ′ + 𝑦𝑗 ′
sin 𝑕
(since divided by − sin 𝑕 )
1 − cos 𝑕
⟹ 𝑦𝑗 +1 = 𝑦𝑗 + 𝑦𝑗 +1 ′ + 𝑦𝑗 ′
sin 𝑕
5.4: Singlestep Methods
A general singlestep method can be written as

𝑢𝑗 +1 = 𝑢𝑗 + 𝑕𝜙(𝑡𝑗 +1 , 𝑡𝑗 , 𝑢𝑗 +1 , 𝑢𝑗 , 𝑕)

where 𝜙 is a function of the arguments 𝑡𝑗 , 𝑡𝑗 +1 , 𝑢𝑗 , 𝑢𝑗 +1 , 𝑕 and also

depends on 𝑓

Taylor Series Method:

The Taylor series method of order 𝑝 is given by

𝑢𝑗 +1 = 𝑢𝑗 + 𝑕𝜙(𝑡𝑗 , 𝑢𝑗 , 𝑕) where 𝑗 = 0, 1, … , 𝑁 − 1

The truncation error of this method is given by

1
𝑇𝑗 +1 = 𝑕𝑝+1 𝑢𝑝+1 𝑡𝑗 + 𝜃𝑕 .
𝑝+1 !

Euler-Cauchy Method:
The formula for Euler-Cauchy method is given by
𝑕
𝑢𝑗 +1 = 𝑢𝑗 + [𝐾1 + 𝐾2 ] where 𝐾1 = 𝑕𝑓𝑗 and 𝐾2 = 𝑕𝑓(𝑡𝑗 +1 , 𝑢𝑗 + 𝐾1 )
2

Modified Euler-Cauchy Method

The formula for Modified Euler-Cauchy method is given by

𝑢𝑗 +1 = 𝑢𝑗 + 𝐾2

312
 𝑕 1
where 𝐾1 = 𝑕𝑓(𝑡𝑗 , 𝑢𝑗 ) and 𝐾2 = 𝑕𝑓(𝑡𝑗 + , 𝑢𝑗 + 𝐾1 )
2 2 SPACE OF HINTS

Heun Method:
1
𝑢𝑗 +1 = 𝑢𝑗 + [𝐾1 + 𝐾2 ]
2

where 𝐾1 = 𝑕𝑓(𝑡𝑗 , 𝑢𝑗 ) and 𝐾2 = 𝑕𝑓(𝑡𝑗 + 𝑕, 𝑢𝑗 + 𝐾1 )

which reduces to the trapezoidal rule when 𝑓(𝑡, 𝑢) is independent of 𝑢.

This method is called as the Heun method.

Example: 5.4.1

Given the initial value problem𝑢′ = 𝑡 2 + 𝑢2 , 𝑢 0 = 0.

Determine the first three non-zero terms in the Tayler series for 𝑢 𝑡 and
hence obtain the value for 𝑢 1 . Also determine 𝑡 when the error in 𝑢 𝑡
obtain from the first two non-zero terms is to be less than10−6 after
rounding.

Solution:

Given 𝑢′ = 𝑡 2 + 𝑢2 , 𝑢 0 = 0

⟹ 𝑢′ 0 = 0

𝑢′′ 𝑡 = 2𝑡 + 2𝑢𝑢′ , 𝑢′′ 0 = 0

𝑢′′′ 𝑡 = 2 + 2(𝑢𝑢′′ + 𝑢′ 𝑢′ ) , 𝑢′′′ 0 = 2

𝑢4 𝑡 = 2 𝑢𝑢 ′′′ + 𝑢′′ 𝑢′ + 2 𝑢𝑢′′ + 𝑢′ 𝑢′′ = 2𝑢𝑢′′′ + 6𝑢 ′ 𝑢′′ ,

𝑢4 0 = 0

𝑢5 𝑡 = 2 𝑢𝑢4 + 𝑢3 𝑢′ + 6 𝑢′ 𝑢3 + 𝑢′′ 𝑢′′

= 2𝑢𝑢4 + 8𝑢′ 𝑢3 + 6𝑢′′ 𝑢′′ ,

𝑢5 0 = 0

𝑢6 𝑡 = 2 𝑢𝑢5 + 𝑢4 𝑢′ + 8 𝑢′′ 𝑢3 + 𝑢′ 𝑢4
+ 6 +𝑢3 𝑢′′ + 𝑢′′ 𝑢3

= 2𝑢𝑢 5 + 10𝑢4 𝑢′ + 20𝑢′′ 𝑢′′′ , 𝑢6 0 = 0

313
 𝑢7 𝑡 = 2 𝑢𝑢6 + 𝑢5 𝑢′ + 10 𝑢4 𝑢′′ + 𝑢′ 𝑢5
SPACE FOR HINTS
+ 20 𝑢′′ 𝑢4 + 𝑢3 𝑢3

= 2𝑢𝑢6 + 12𝑢5 𝑢′ + 30𝑢′′ 𝑢′′ + 20 𝑢 3 2 ,

𝑢7 0 = 80

𝑢8 𝑡 = 2 𝑢𝑢7 + 𝑢6 𝑢′ + 12 𝑢5 𝑢′′ + 𝑢′ 𝑢6
+ 30 𝑢′′ 𝑢5 + 𝑢4 𝑢3

= 2𝑢𝑢7 + 14𝑢6 𝑢′ + 42𝑢′′ 𝑢5 + 70𝑢4 𝑢3 , 𝑢8 0 = 0

𝑢9 𝑡 = 2 𝑢𝑢8 + 𝑢7 𝑢′ + 14 𝑢6 𝑢′′ + 𝑢′ 𝑢7
+ 42 𝑢′′ 𝑢6 + 𝑢5 𝑢3 + 70 𝑢5 𝑢 3 + 𝑢4 𝑢4

= 2𝑢𝑢8 + 16𝑢7 𝑢′ + 56𝑢′′ 𝑢6 + 112𝑢 5 𝑢3 + 70 𝑢4 2 ,

𝑢9 0 = 0

𝑢10 𝑡 = 2 𝑢𝑢9 + 𝑢8 𝑢′ + 16 𝑢7 𝑢′′ + 𝑢′ 𝑢 8

+ 56 𝑢′′ 𝑢7 + 𝑢6 𝑢3 + 112 𝑢6 𝑢3 + 𝑢5 𝑢4
+ 70 2𝑢5 𝑢4

= 2𝑢𝑢9 + 18𝑢8 𝑢′ + 72𝑢′′ 𝑢7 + 168𝑢 6 𝑢3 + 252𝑢5 𝑢4 ,

𝑢10 0 = 0

𝑢11 𝑡 = 2 𝑢𝑢10 + 𝑢9 𝑢′ + 18 𝑢8 𝑢′′ + 𝑢′ 𝑢 9

+ 72 𝑢′′ 𝑢8 + 𝑢7 𝑢3 + 168 𝑢6 𝑢4 + 𝑢7 𝑢3
+ 252 𝑢4 𝑢6 + 𝑢5 𝑢5

= 2𝑢𝑢10 + 20𝑢9 𝑢′ + 90𝑢′′ 𝑢8 + 240𝑢7 𝑢3 + 420𝑢6 𝑢4

+ 252 𝑢5 2 , 𝑢11 0 = 38400

∴ The Tayler series for 𝑢 𝑡 becomes,

2
𝑡−0 ′ 𝑡−0
𝑢 𝑡 =𝑢 0 + 𝑢 0 + 𝑢′′ 0 + ⋯
1! 2!
𝑡 3 ′′′ 𝑡7 7 𝑡 11 11
𝑢 𝑡 = 𝑢 0 + 𝑢 0 + 𝑢 0 +⋯
3! 7! 11!

314
 𝑡3 𝑡7 𝑡 11
= 2 + 80 + 38400 + ⋯ SPACE OF HINTS
2.3 5040 39916800
𝑡3 𝑡7 2𝑡 11
𝑢 𝑡 = + + +⋯
3 63 2079
1 1 2
𝑢 1 = + + +⋯
3 63 2079
If only the first two terms are used them the value (or) 𝑡 is obtained
2𝑡 11
from, ≤𝜖
2079

Let us consider 𝜖 = 0.5 × 10−7

2𝑡 ′′
⟹ ≤ 0.5 × 10−7
2079
⟹ 2𝑡 ′′ ≤ 1.0395 × 10−4

⟹ 𝑡 ′′ ≤ 5.1995 × 10−5

𝑡 ≤ 0.4079

Example: 5.4.2
Find the three term Tayler series solution for the third order
initial value problem𝑤 ′′′ + 𝑤𝑤 ′′ = 0, 𝑤 0 = 0 , 𝑤 ′ 0 =
0 , 𝑤 ′′ 0 = 1 find the bound on the error for 𝑡 ∈ 0, 0.2

Solution:
Given 𝑤 ′′′ + 𝑤𝑤 ′′ = 0, 𝑤 0 = 0 , 𝑤 ′ 0 = 0 , 𝑤 ′′ 0 = 1

𝑤 ′′′ + 𝑤𝑤 ′′ = 0

𝑤 ′′′ 0 = −𝑤𝑤 ′′ = 0

𝑤 4 𝑡 = − 𝑤𝑤 ′′ + 𝑤 ′ 𝑤 ′′ , 𝑤4 0 = 0

𝑤 5 𝑡 = − 𝑤𝑤 4 + 𝑤 ′ 𝑤 ′′′ + 𝑤 ′ 𝑤 ′′′ + 𝑤 ′′ 𝑤 ′′ ,
𝑤 5 0 = −1

𝑤6 𝑡
= − 𝑤𝑤 5 + 𝑤 ′ 𝑤 4 + 𝑤 ′′ 𝑤 ′′′ + 𝑤 4 𝑤 ′
+ 𝑤 ′′ 𝑤 ′′′ + 𝑤 ′′ 𝑤 ′′′ + 𝑤 ′′ 𝑤 ′′′

315
 = − 𝑤𝑤 5 + 3𝑤 ′ 𝑤 4 + 4𝑤 ′′ 𝑤 ′′′ , 𝑤6 0 = 0
SPACE FOR HINTS
𝑤7 𝑡
= − 𝑤𝑤 6 + 𝑤 ′ 𝑤 6 + 3𝑤 5 𝑤 ′ + 3𝑤 4 𝑤 ′′
+ 4𝑤 ′′ 𝑤 4 + 4𝑤 ′′′ 𝑤 ′′′

= − 4𝑤𝑤 5 + 𝑤𝑤 6 + 7𝑤 4 𝑤 ′′ + 4𝑤 ′′′ 𝑤 ′′′ ,

𝑤7 0 = 0

𝑤8 𝑡
= − 𝑤𝑤 7 + 𝑤 ′ 𝑤 6 + 4𝑤 5 𝑤 ′′ + 4𝑤 ′ 𝑤 6
+ 7𝑤 ′′ 𝑤 5 + 7𝑤 ′′′ 𝑤 4 + 4𝑤 ′′′ 𝑤 4
+ 4𝑤 4 𝑤 ′′′

= − 𝑤𝑤 7 + 5𝑤 ′ 𝑤 6 + 11𝑤 ′′ 𝑤 5 + 7𝑤 ′′′ 𝑤 4
+ 8𝑤 ′′′ 𝑤 4 , 𝑤 8 0 = 11

𝑤9 𝑡
= − 𝑤𝑤 8 + 𝑤 ′ 𝑤 7 + 5𝑤 ′ 𝑤 5 + 5𝑤 6 𝑤 ′′
+ 11𝑤 ′′ 𝑤 6 + 11𝑤 ′′′ 𝑤 5 + 7𝑤 4 𝑤 4
+ 7𝑤 ′′′ 𝑤 5 + 8𝑤 ′′′ 𝑤 5 + 8𝑤 4 𝑤 4

= − 𝑤𝑤 8 + 6𝑤 ′ 𝑤 7 + 16𝑤 ′′ 𝑤 6
+ 18𝑤 ′′′ 𝑤 5 + 7𝑤 4 𝑤 4 + 8𝑤 ′′′ 𝑤 5
+ 8𝑤 4 𝑤 4 , 𝑤9 0 = 0

𝑤 10 𝑡
= − 𝑤𝑤 9 + 𝑤 ′ 𝑤 8 + 6𝑤 ′′ 𝑤 7 + 6𝑤 ′ 𝑤 8
+ 16𝑤 ′′′ 𝑤 6 + 16𝑤 ′′′ 𝑤 6 + 18𝑤 ′′′ 𝑤 6
+ 18𝑤 4 𝑤 5 + 7𝑤 5 𝑤 4 + 7𝑤 4 𝑤 5 + 8𝑤 ′′′ 𝑤 3
+ 8𝑤 5 𝑤 4 + 8𝑤 5 𝑤 4 + 8𝑤 5 𝑤 4

316
 SPACE OF HINTS
9 ′ 8 ′′ 7
= − 𝑤𝑤 + 7𝑤 𝑤 + 22𝑤 𝑤
+ 34𝑤 ′′′ 𝑤 6 + 56𝑤 4 𝑤 5
+ 8𝑤 ′′′ 𝑤 6 , 𝑤 10 0 = 0

𝑤 11 𝑡
= − 𝑤𝑤 10 + 𝑤 ′ 𝑤 9 + 7𝑤 ′′ 𝑤 8 + 7𝑤 ′ 𝑤 9
+ 22𝑤 ′′′ 𝑤 7 + 22𝑤 ′′ 𝑤 8 + 35𝑤 ′′′ 𝑤 7
+ 35𝑤 4 𝑤 6 + 56𝑤 5 𝑤 5 + 56𝑤 4 𝑤 6
+ 8𝑤 4 𝑤 6 + 8𝑤 ′′′ 𝑤 7

= − 𝑤𝑤 10 + 8𝑤 ′ 𝑤 9 + 29𝑤 ′′ 𝑤 8
+ 65𝑤 ′′′ 𝑤 7 + 91𝑤 4 𝑤 6
+ 56𝑤 5 𝑤 5 , 𝑤 11 0 = −375

Tayler series:

𝑡 ′ 𝑡 2 ′′ 𝑡 3 ′′′
𝑤 𝑡 = 𝑤 𝑡 + 𝑤 𝑡 + 𝑤 𝑡 +⋯
1! 2! 3!
𝑡 2 ′′ 𝑡5 𝑡8 𝑡 11 11
= 𝑤 𝑡 + 𝑤5 𝑡 + 𝑤8 𝑡 + 𝑤 𝑡
2! 5! 8! 11!
𝑡2 𝑡5 𝑡8 𝑡 11
= 1 + −1 + 11 + −375
2! 5! 8! 11!
The error term is,

11
𝑡 11
𝐸 ≤ max 𝑤 𝑡
0≤𝑡≤0.2 11!
11
−375 0.2
= 1.924001924 × 10−13
11!

𝐸 ≤ 1.924001924 × 10−13

Example: 5.4.3

317
 Given the initial value problem𝑢′ = −2𝑡𝑢2 , 𝑢 0 = 1 estimate
SPACE FOR HINTS 𝑢 0.4 using

i. Modified Euler Cauchy method

ii. Heun method with 𝑕 = 0.2

Compare the results with the exact solution 𝑢 𝑡 = 1 1 + 𝑡 2

Solution:
i. The formula for modified Euler Cauchy method is

𝑢𝑗 +1 = 𝑢𝑗 + 𝑘2 𝑤𝑕𝑒𝑟𝑒 𝐾1 = 𝑕𝑓 𝑡𝑗 , 𝑢𝑗 ; 𝐾2

𝐾1
= 𝑕𝑓 𝑡𝑗 + 𝑕 2 , 𝑢𝑗 + 2

Here 𝑓 𝑡𝑗 , 𝑢𝑗 = 𝑢′ = −2𝑡𝑗 𝑢𝑗 2 also given 𝑢 0 = 1

𝑕 = 0.2 , 𝑘1 = −2 −2𝑡𝑗 𝑢𝑗 2

𝑗 = 0 ⟹ 𝑡0 = 0 , 𝑢0 = 1

𝐾1 = 0.2 −2 0 1 =0

𝐾2 = 0.2 𝑓 𝑡𝑗 + 0.2 2 , 𝑢𝑗 + 0 2

= 0.2 𝑓 0 + 0.2 2 , 1 + 0

= 0.2 𝑓 0.1 , 1
2
𝐾2 = −2 −2 0.1 1 = −0.04

∴ 𝑢𝑗 +1 = 𝑢𝑗 + 𝑘2

𝑗 = 0 , ⟹ 𝑢1 = 𝑢0 + 𝑘2 = 1 − 0.04 = 0.96

𝑢 0.2 = 𝑢1 = 0.96

For 𝑗 = 1, 𝐾1 = 𝑕𝑓 𝑡1 , 𝑢1

𝑡1 = 0.2 , 𝑢1 = 0.96

318
 𝐾1 = −2 −2𝑡𝑗 𝑢𝑗 2 = 0.2 −2 0.2 0.96 2
SPACE OF HINTS

𝐾1 = −0.073728

𝐾2 = 0.2 𝑓 𝑡𝑗 + 0.2 2 , 𝑢𝑗 + −0.073728 2

𝐾2 = 0.2 𝑓 0.3 , 0.93136

2
𝑓 𝑡𝑗 , 𝑢𝑗 = −2 0.3 0.923136 = −0.102262

𝑢𝑗 +1 = 𝑢𝑗 + 𝐾2

𝑗 = 1 ⟹ 𝑢2 = 𝑢1 + 𝐾2 = 0.96 + −0.102262

𝑢2 = 𝑢 0.4 = 0.857738

ii. The formula for Heun method is,

𝑢𝑗 +1 = 𝑢𝑗 + 1 2 𝐾1 + 𝐾2 𝑤𝑕𝑒𝑟𝑒 𝐾1 = 𝑕𝑓 𝑡𝑗 , 𝑢𝑗 ; 𝐾2

= 𝑕𝑓 𝑡𝑗 + 𝑕 , 𝑢𝑗 + 𝐾1

Here 𝑓 𝑡𝑗 , 𝑢𝑗 = 𝑢 ′ = −2𝑡𝑗 𝑢𝑗 2 , 𝑢 0 = 1

𝑡0 = 0 , 𝑢0 = 1 , 𝑕 = 0.2

𝐾1 = 0.2 −2 0 1 =0

𝐾2 = 𝑕𝑓 𝑡𝑗 + 𝑕 , 𝑢𝑗 + 𝑘1

= 0.2 𝑓 0 + 0.2 , 1 = 0.2 𝑓 0.2 , 1

= 0.2 −2 0.2 1

𝐾2 = −0.08

𝑢𝑗 +1 = 𝑢𝑗 + 1 2 𝐾1 + 𝐾2

𝑢1 = 𝑢0 + 1 2 𝐾1 + 𝐾2

𝑢 0.2 = 𝑢1 = 0.96

319
 For 𝑗 = 1 , 𝑡1 = 0.2 , 𝑢1 = 0.96
SPACE FOR HINTS 𝐾1 = −0.073728

𝐾2 = 𝑕𝑓 𝑡𝑗 + 𝑕 , 𝑢𝑗 + 𝑘1

= 0.2 𝑓 0.2 + 0.2 , 0.96 + −0.073728

= 0.2 𝑓 0.4 , 0.886272

2
= 0.2 −2 0.4 0.886272

𝐾2 = −0.125676

𝑢𝑗 +1 = 𝑢𝑗 + 1 2 𝐾1 + 𝐾2

𝑗=1, 𝑢1 = 𝑢 0.4

= 0.96 + 1 2 −0.073728 − 0.125676

𝑢1 = 0.860298

𝑢′ = −2𝑡𝑢2
1
𝑢 𝑡 =
1 + 𝑡2
1
𝑢 0.2 = = 0.961538
1 + 0.22
1
𝑢 0.4 = = 0.862069
1 + 0.42
The absolute errors in the numerical solutions are,

Modified Euler-Cauchy method, 𝜖 0.2 = 0.001538 ,

𝜖 0.4 = 0.004331

Heun method, 𝜖 0.2 = 0.001538 , 𝜖 0.4 = 0.001771

Example: 5.4.4

Solve the initial value problem 𝑢′ = −2𝑡𝑢2 , 𝑢 0 = 1with 𝑕 = 0.2 on

the interval 0 , 0.4 use the fourth order classical Runge-kutta method.
Compare with the exact solution.

320
Solution:
SPACE OF HINTS
The formula for fourth order method is,

𝑢𝑗 +1 = 𝑢𝑗 + 1 6 𝐾1 + 2 𝐾2 + 𝐾3 + 𝐾4 where 𝐾1 = 𝑕𝑓 𝑡𝑗 , 𝑢𝑗

𝐾1
𝐾2 = 𝑕𝑓 𝑡𝑗 + 𝑕 2 , 𝑢𝑗 + 2

𝐾2
𝐾3 = 𝑕𝑓 𝑡𝑗 + 𝑕 2 , 𝑢𝑗 + 2

𝐾4 = 𝑕𝑓 𝑡𝑗 + 𝑕 , 𝑢𝑗 + 𝐾3

For 𝑗 = 0 , 𝑡0 = 0 , 𝑢0 = 1
2
𝐾1 = 𝑕𝑓 𝑡𝑗 , 𝑢𝑗 = 0.2 −2 0 0.1 =0

⟹ 𝐾1 = 0

𝐾1
𝐾2 = 𝑕𝑓 𝑡𝑗 + 𝑕 2 , 𝑢𝑗 + 2

0.2 0
= 𝑕𝑓 0 + , 1+
2 2
= 0.2 𝑓 0.1 , 1
2
= 0.2 −2 0.1 1

𝐾2 = −0.04

𝐾2
𝐾3 = 𝑕𝑓 𝑡𝑗 + 𝑕 2 , 𝑢𝑗 + 2

= 𝑕𝑓 0 + 0.2 2 , 1 + 0 2

= 0.2𝑓 0.1 , 0.98

𝐾3 = −0.038416

𝐾4 = 𝑕𝑓 𝑡𝑗 + 𝑕 , 𝑢𝑗 + 𝐾3

= 0.2 𝑓 0.2 , 1 − 0.038416

321
 = 0.2 𝑓 0.2 , 0.961584
SPACE FOR HINTS 2
= 0.2 −2 0.2 0.961584

𝐾4 = −0.073972

∴ 𝑢𝑗 +1 = 𝑢𝑗 + 1 6 𝐾1 + 2 𝐾2 + 𝐾3 + 𝐾4

𝑢1 = 𝑢0 + 1 6 𝐾1 + 2 𝐾2 + 𝐾3 + 𝐾4

𝑢 0.2 ≈ 𝑢1 = 1 + 1 6 0 + 2 −0.04 − 0.038416 − 0.073972

𝑢 0.2 = 0.961533

For 𝑗 = 1 , 𝑡1 = 0.2 , 𝑢1 = 0.961533

𝐾1 = 𝑕𝑓 𝑡𝑗 , 𝑢𝑗 = −0.073964

𝐾1
𝐾2 = 𝑕𝑓 𝑡𝑗 + 𝑕 2 , 𝑢𝑗 + 2

0.2 −0.073964
= 𝑕𝑓 0.2 + , 0.961533 +
2 2
= 0.2 𝑓 0.3 , 0.924551
2
= 0.2 −2 0.3 0.924551

𝐾2 = −0.1025753462

𝐾2
𝐾3 = 𝑕𝑓 𝑡𝑗 + 𝑕 2 , 𝑢𝑗 + 2

= 𝑕𝑓 0.2 + 0.2 2 , 0.961533 + −0.1025753462 2

= 0.2𝑓 0.3 , 0.910245

𝐾3 = −0.0994255

𝐾4 = 𝑕𝑓 𝑡𝑗 + 𝑕 , 𝑢𝑗 + 𝑘3

= 0.2 𝑓 0.2 + 0.2 , 1 − 0.0994255

322
 = 0.2 𝑓 0.4, 0.9005745
SPACE OF HINTS
2
= 0.2 −2 0.4 0.9005745

𝐾4 = −0.1297655088

∴ 𝑢𝑗 +1 = 𝑢𝑗 + 1 6 𝐾1 + 2 𝐾2 + 𝐾3 + 𝐾4

𝑢2 = 𝑢1 + 1 6 𝐾1 + 2 𝐾2 + 𝐾3 + 𝐾4

𝑢 0.2 = 𝑢2
= 0.961533
+ 1 6 −0.073964

+ 2 −0.102575 − 0.0994255
− 0.1297955088

𝑢 0.2 = 0.8602395819

*******

323