Dilla University

College off Natural and Computational Science

Department of Mathematics

Module on
on Numerical Analysis I

Prepare by: Asnake Muluye (Med/Msc

/Msc)
Shimsue Nuru (Msc)
Wondwosen Gebeyaw (Msc)
Edited By:
Gezahegn Anberber (Msc)
June 2007 EC

1
 Module Introduction

This module consists of six chapters. The first chapter deals with one of the central concept in numerical
analysis which is error analysis. In this unit we will look briefly at many terms and concepts related to
numerical errors and discuss the sources, types and measures of errors focusing on how to measure and
control them. The second chapter deals with computing the approximate solution of different non-linear
algebraic equations, which are either complicated or impossible to solve by using analytical methods,
using different numerical methods focusing on how to minimize and control the approximation errors and
also dealing with the computer algorithms of each methods . Like that of the previous chapter, the third
chapter deals with how to compute the approximate solution of a variety of systems of both linear and
non-linear equations using different numerical methods, by considering the two broad categories which
are Direct Methods and Indirect(or Iterative) Methods, by explaining the merits and demerits of each
methods in both cases .

The last three chapters focuses on one of the most important concept in numerical analysis, which is the
approximation of complicated functions by using simpler functions such as polynomials, trigonometric,
rational functions and so on, that are easy for mathematical operations; in which the different
mathematical operations intended for the complicated functions can be done by using the simpler
approximating function; focusing on one of the most widely used techniques of approximation called
polynomial interpolation and consider its application. Briefly, chapter four discuss about difference
operators which are important in developing the interpolating polynomials whereas, chapter five focuses
on the development of the different polynomial interpolation techniques. Finally, chapter six explains the
applications of polynomial interpolation, by giving special focuses on the two basic uses of polynomial
interpolation, which are numerical integration and differentiation.

2
Module Objectives:

At the completion of this module students will be able to:

 Understand clearly the need for Numerical Analysis in mathematics.

 Solve a variety of problems using numerical methods whenever, analytical methods are
difficult or impossible to apply.
 Understand the importance of numerical analysis in solving problems which arises in
science, engineering, and social science fields.
 Understand that approximation is the central concept in Numerical Analysis.
 Identify the sources, types and measures of approximation errors and able to bound and
minimize them.
 Correctly identify the importance of computers on ;implementing numerical methods.
 Compute approximate solutions of non-linear algebraic equations.
 Demonstrate the application of numerical methods.
 Develop and implement different algorithms for the numerical methods which are used to
solve real life problems.
 Define and solve system of both linear and non-linear equations numerically.
 Understand the basic principles on approximation of functions.
 Identify different numerical methods for construction of interpolating polynomials.
 Apply polynomial interpolation to compute missing functional values of a tabulated function.
 Demonstrate the applications of polynomial interpolation especially on numerical
integration and differentiation, together with solving exercises.

3
 Table of content
CHAPTER ONE ............................................................................................................................. 6
1.1. Introduction ................................................................................................................................... 6
1.2. Errors............................................................................................................................................. 7
1.2.1. Sources of Errors................................................................................................................... 8
1.2.2. Measuring Errors................................................................................................................... 8
1.2.3. Classification of Errors ....................................................................................................... 16
1.3. Computer Representation of Numbers ........................................................................................ 19
1.3.1. Fixed Point Representation ................................................................................................. 19
1.3.2. Floating Point Representation ............................................................................................. 22
1.4. Propagation of Errors .................................................................................................................. 27
1.5. Stability of Algorithms and Conditioning numbers .................................................................... 30
1.5.1. Stability of Algorithms........................................................................................................ 31
1.5.2. Conditioning or Condition of a Problem............................................................................. 32
CHAPTER TWO .......................................................................................................................... 36
2. SOLUTION OF ALGEBRAIC AND TRANSCENDENTAL EQUATIONS ..................... 36
2.1. Introduction ................................................................................................................................. 36
2.2. Preliminaries: .............................................................................................................................. 37
2.2.5. The secant Method: ................................................................................................................... 56
CHAPTER THREE ...................................................................................................................... 69
3.2. Exact Method .............................................................................................................................. 70
3.2.1. Gaussian Elimination .......................................................................................................... 71
3.2.2. Gaussian Elimination Method ............................................................................................. 75
3.2.3. The backward or forward substitution method formula ...................................................... 80
3.2.4. Gaussian Elimination with partial pivoting......................................................................... 81
3.2.5. Gauss-Jordan Method ......................................................................................................... 82
3.2.6. Matrix Inversion Using Jordan Elimination........................................................................ 86
3.2.7. LU Matrix Decomposition .................................................................................................. 87
3.3. Indirect/Iterative/ methods of solving systems of linear equation ........................................ 91
3.3.1. Gauss-Seidel Method .......................................................................................................... 91

4
 UNITE FOUR ....................................................................................................................... 100
4.1. INTRODUCTION................................................................................................................... 100
4.2. FORWARD DIFFERENCE OPERATOR ........................................................................... 100
4.2.1. The Operator E .................................................................................................................. 103
4.2.2. Relation between the Operator E and ∆ .................................................................................. 104
4.3. Backward Differences ............................................................................................................. 106
4.4. Central Deference ................................................................................................................... 109
UNIT FIVE ................................................................................................................................ 111
5. INTERPOLATION ............................................................................................................. 111
5.1. Introduction .................................................................................................................................. 111
5.2. Interpolation with Evenly Spaced Points ................................................................................... 112
5.2.3 Gauss interpolating polynomial ............................................................................................... 121
5.3 Interpolation with Unevenly Spaced Points ................................................................................ 131
5.3.2 Newton’s Divided Difference Interpolation............................................................................. 138
CHAPTER SIX ......................................................................................................................... 146
6.1 Introduction ................................................................................................................................... 147
6.2 Numerical Differentiation ............................................................................................................ 147
6.2.1 Formulae for derivatives .......................................................................................................... 148
6.2.2 Maxima and Minima of a tabulated function ........................................................................... 155
6.3 Numerical Integration .................................................................................................................. 158
6.3.1 Newton-Cotes quadrature formula ........................................................................................... 159
6.3.2 Errors in quadrature formulae .................................................................................................. 167
6.3.3 Romberg's Method ................................................................................................................... 169
6.3.4 Euler-Maclaurin formula.......................................................................................................... 172
6.4 Method of undetermined coefficients .......................................................................................... 175
6.4.1 Differentiation formulae .......................................................................................................... 175
6.4.2 Integration formulae................................................................................................................. 176
6.5 Numerical Double Integration ..................................................................................................... 178

5
 CHAPTER ONE

1. BASIC CONCEPTS IN ERROR ESTIMATION

1.1.Introduction

Numerical Analysis is the area of mathematics that creates, analyzes, and implements algorithms
for solving numerically the problems of continuous mathematics. Such problems originate
generally from real world applications of algebra, geometry, and calculus , and they involve
variables which varies continuously; these problems occurs throughout the natural sciences,
social sciences, engineering, medicine, and business.
During the past half century, the growth in power and availability of digital computers has led to
an increasing use of realistic mathematical models in science and engineering, and numerical
analysis of increasing sophistication has been needed to solve these more detailed mathematical
models of the real world problems.
Thus to solve any real life problem by using numerical methods the following three steps are
mostly considered:
 Converting the real life (or physical) problem in to a mathematical model.
 Apply an appropriate numerical method which can solve the mathematical model and
develop an algorithm for the method.
 Finally, implement the algorithm on a computational tools (most commonly on
computers) to compute the required result.
More often all the above steps are exposed to an error due to different assumptions and
limitations. So most numerical methods give answers that are only approximations to the desired
true solution, and it's important to understand and to be able, if possible, to estimate or bound the
resulting error. Therefore, the study of errors is the central concern of numerical analysis.
This chapter examines the various sources and types of errors that may occur in a problem. The
representation of numbers in computers is examined, along with the error in computer arithmetic.
General results on the propagation of errors in a calculation are also considered. Finally, the
concept of stability and conditioning of problems on numerical methods are introduced and
illustrated.

6
After reading this chapter, students will be able to:
 define and explain about numerical analysis
 understand a lot of things about errors in numerical methods
 identify, measures and control errors in numerical calculations
 differentiate the different sources and types of errors
 represent numbers on computers and calculate the maximum error bounds which
corresponds to the representation
 understand and control the propagation of errors in numerical calculations
 define and understand the algorithms of numerical methods
 identify the stability of algorithms
 identify well-conditioned and ill-conditioned problems by calculating their condition
numbers.
1.2. Errors

After completing this sections lesson, students are expected to:

 identify the different sources and types of errors
 find the true and relative true error,
 calculate the true and relative approximate error,
 relate the absolute relative approximate error to the number of significant digits
One of the most important aspect of numerical analysis is the study of errors; because, errors
occur at any stage of the process of solving problems numerically. By the error we simply mean
that the difference between the true ( or exact ) value and the approximate value. Therefore;

Error= True value -Approximate value

Whenever, we solve any problem using numerical analysis, errors will arise during the
calculations. To be able to deal with the issue of errors, we need to

 identify where the error is coming from, followed by

 quantifying the error, and lastly
 Minimize the error as per our needs.

7
 1.2.1. Sources of Errors

Errors in a solution of a problem are due to the following sources:

i. To solve any physical problem, mathematical models are formulated to describe
them, but these models do not represent the problems exactly because different
assumptions and inexact initial data's are used in the development of the models,
as a result errors are induced.
ii. The numerical methods used to solve the mathematical models are often not
exact, because this methods replaces an infinite process in to finite, as a
consequence errors arises.
iii. Computation of the final result is done using computational tools like tables,
calculators, computers, and etc (but mostly by implementing the algorithms of the
numerical methods on different computer packages ) but these tools have limited
space; because of this only limited part of the results are considered, thus errors
are induced.

1.2.2. Measuring Errors

Suppose is the exact value and is the approximate value of , thus the error occurred on
such approximation can be measured using one of the following ways depending on the accuracy
required.
a) Tue Error
True error denoted by Et is the difference between the true value (also called the exact value)

and the approximate value .

True Error  True value – Approximate value
= −

8
 Example 1.1

The derivative of a function f (x ) at a particular value of x can be approximately calculated

by
f ( x  h)  f ( x )
f ( x) 
h
0 .5 x
For f ( x )  7e and h  0.3 , find
a) the approximate value of f ( 2)
b) the true value of f ( 2)
c) the true error for part (a)
f ( x  h)  f ( x )
Solution: a) f ( x) 
h
For x  2 and h  0.3 ,
f (2  0.3)  f (2)
f (2) 
0.3

f (2.3)  f (2)

0.3

7e 0.5( 2.3)  7e 0.5( 2)


0.3

22.107  19.028

0.3
 10.265

b) The exact value of f ( 2) can be calculated by using our knowledge of differential

calculus.

f ( x )  7 e 0 .5 x
f ' ( x)  7  0.5  e0.5 x
 3 . 5e 0.5 x
So the true value of f ' ( 2) is
f ' (2)  3.5e 0.5( 2)
 9.5140
c) True error is calculated as
Et = True value – Approximate value
 9.5140  10.265
 0.75061

9
 b) Absolute True Error
Absolute true error which is denoted by is the absolute value of the true error, that means:
 –
=| |=| − |
Since when we talk about error the main focus lies on the magnitude of the error than the sign,
absolute errors are used in place of actual errors.
In Example 1.1 above the absolute error becomes
= | | =  0.75061 = 0.75061

The magnitude of true error does not show how bad the error is. An absolute true error of
E A  0.75061 may seem to be small, but if the function given in the Example 1.1 were

f ( x)  7  10 6 e 0.5 x , the absolute true error in calculating f ( 2) with h  0.3, would be

E A  0.75061 10 6. This value of absolute true error is smaller, even when the two problems
are similar in that they use the same value of the function argument, x  2 and the step size,
h  0.3 . This brings us to the definition of relative true error.
c) Relative True Error
Relative true error is denoted by R and is defined as the ratio between the true error and the

True Error
true value. 
True Value
−
= =

Example 1.2
For the problem in Example 1.1 above, find the relative true error at x  2 .
Solution
From Example 1.1,
Et = True value – Approximate value
 9.5140  10.265
 0.75061
Relative true error is calculated as
True Error
t 
True Value

10
  0.75061

9.5140
 0.078895
Relative true errors are also presented as percentages. For this example,
t  0.0758895  100%
 7.58895%
Absolute relative true errors may also need to be calculated. In such cases,
t | 0.075888 |

= 0.0758895
= 7.58895%
d) Approximate Error
In the previous section, we discussed how to calculate true errors. Such errors are calculated
only if true values are known. An example where this would be useful is when one is checking if
a program is in working order and you know some examples where the true error is known. But
mostly we will not have the luxury of knowing true values as why would you want to find the
approximate values if you know the true values. So when we are solving a problem numerically,
we will only have access to approximate values. We need to know how to quantify error for such
cases.
Approximate error is denoted by Ea and is defined as the difference between the present
approximation and previous approximation.
Approximate Error  Present Approximation – Previous Approximation
Example1. 3
For the problem in Example 1.1 above, find the following
a) f ( 2 ) using h  0.3
b) f ( 2 ) using h  0.15
c) approximate error for the value of f ( 2) for part (b)
Solution
a) The approximate expression for the derivative of a function is
f ( x  h)  f ( x )
f ' ( x)  .
h
For x  2 and h  0.3 ,

11
 f (2  0.3)  f (2)
f ' (2) 
0.3
f (2.3)  f (2)

0.3

7e 0.5( 2.3)  7e 0.5( 2)


0.3
22.107  19.028

0.3
 10.265
b) Repeat the procedure of part (a) with h  0 .15,
f ( x  h)  f ( x )
f ( x) 
h
For x  2 and h  0.15 ,
f (2  0.15)  f (2)
f ' (2) 
0.15
f (2.15)  f (2)

0.15
0.5 ( 2.15 )
7e  7e 0.5( 2 )

0.15
20.50  19.028

0.15
 9.8799
c) So the approximate error, Ea is
E a  Present Approximation – Previous Approximation
 9.8799  10.265
 0.38474
The magnitude of approximate error does not show how bad the error is . An approximate error
of E a  0.38300 may seem to be small; but for f ( x)  7  10 6 e 0.5 x , the approximate error in

calculating f ' ( 2) with h  0.15 would be Ea  0.38474  106 . This value of approximate error
is smaller, even when the two problems are similar in that they use the same value of the
function argument, x  2 , and h  0.15 and h  0.3 . This brings us to the definition of relative
approximate error.

e) Relative Approximate Error

Relative approximate error is denoted by a and is defined as the ratio between the approximate
error and the present approximation.

12
 Approximate Error
Relative Approximate Error 
Present Approximation
Example 1.4
For the problem in Example 1.1 above, find the relative approximate error in calculating f ( 2)
using values from h  0.3 and h  0.15 ?
Solution
From Example1.3, the approximate value of f ( 2)  10.263 using h  0.3 and f ' ( 2)  9.8800
using h  0.15 .
E a  Present Approximation – Previous Approximation
 9.8799  10.265
 0.38474
The relative approximate error is calculated as
Approximate Error
a 
Present Approximation
 0.38474

9.8799
 0.038942
Relative approximate errors are also presented as percentages. For this example,
a  0.038942  100%
=  3.8942%
Absolute relative approximate errors may also need to be calculated. In this example
a  | 0.038942 |

 0.038942 or 3.8942%
f) The Limiting Errors
In a numerical method that uses iterative methods, a user can calculate relative approximate
error a at the end of each iteration. The user may pre-specify a minimum acceptable tolerance

called the pre-specified tolerance, s . If the absolute relative approximate error a is less than

or equal to the pre-specified tolerance s , that is, |a | s , then the acceptable error has been
reached and no more iterations would be required.

13
Alternatively, one may pre-specify how many significant digits they would like to be correct in
their answer. In that case, if one wants at least m significant digits to be correct in the answer,
then you would need to have the absolute relative approximate error, |a | 0.5  10 2 m %.

Example1.5
If one chooses 6 terms of the Maclaurin series for e x to calculate e 0.7 , how many significant
digits can you trust in the solution? Find your answer without knowing or using the exact
answer.
2
Solution: e x  1  x  x  .......... .......
2!
Using 6 terms, we get the current approximation as
0.7 2 0.7 3 0.7 4 0.7 5
e 0.7  1  0.7    
2! 3! 4! 5!
 2.0136
Using 5 terms, we get the previous approximation as
0.7 2 0.7 3 0.7 4
e 0.7  1  0.7   
2! 3! 4!
 2.0122
The percentage absolute relative approximate error is
2.0136  2.0122
a   100
2.0136
 0.069527%
Since a  0.5  10 22 % , at least 2 significant digits are correct in the answer of

e 0.7  2 .0136
Q: But what do you mean by significant digits?
A: Significant digits are important in showing the truth one has in a reported number. For
example, if someone asked me what the population of my country is, I would respond, “The
population of Ethiopia is around 80 million”. But if someone was going to give me a 100 birr for
every citizen of the country, I would have to get an exact count. That count would have been
80,079,587 in year 2007 G.C. So you can see that in my statement that the population is 80
million, that there is only one significant digit, that is, 8, and in the statement that the population
is 80,079,587, there are eight significant digits. So, how do we differentiate the number of digits

14
correct in 80,000,000 and 80,079,587? Well for that, one may use scientific notation. For our
data we show
80,000,000  8  10 7
80,079,587  8.0079587  10 7
to signify the correct number of significant digits
Example1.6
Give some examples of showing the number of significant digits.
Solution
a) 0.0459 has three significant digits
b) 4.590 has four significant digits
c) 4008 has four significant digits
d) 4008.0 has five significant digits
e) 1.079  10 3 has four significant digits
f) 1.0790  10 3 has five significant digits
g) 1.07900  10 3 has six significant digits
Activity 1.1:
1. Let = 23.457609 be the value of the number and its approximate value is = 23.458,
then find
a) The true error
b) the absolute true error
c) the relative true error
d) both the percentage absolute and relative errors

′
2. For the question in Example 1.1 above compute the approximate value of (2) for
ℎ = 0.05 & ℎ = 0.001 and then find
a) the approximate absolute error
b) the approximate relative error
c) and compare this errors with their corresponding errors in example 1.3 and 1.4
3. Let = 20 and = 1000 are the exact measurements and if = 15 and
= 1005 are their respective approximations, then compare the significance of the two
errors by using

15
 a) absolute true error
b) absolute relative error
c) percentage absolute error and percentage relative error

1.2.3. Classification of Errors

The errors induced by the different sources mentioned above are broadly classified as the
following three types of errors:
1) Inherent Errors: are errors which occur in the development of the mathematical model
for a given physical problem. These types of errors are mostly unavoidable, and they are
caused due to:
 The approximate value of the initial data.
 The different assumption taken on the model.
 The limitations of the computing aids
Even if such errors are beyond the control of the numerical analyst, they can be
minimized by selecting:
 A better initial data.
 A better mathematical model that represent the problem.
 Computing aids of higher precisions.
2) Truncation ( or numerical ) Errors
Truncation error is defined as the error caused by truncating (or cutting) an infinite
mathematical procedure in to a finite one. For example, the Maclaurin series for e x is given as
x2 x3
ex  1 x    ....................
2! 3!
This series has an infinite number of terms but when using this series to calculate e x , only a
finite number of terms can be used. For example, if one uses three terms to calculate e x , then
x2
ex  1 x  .
2!
the truncation error for such an approximation is
 x2 
Truncation error = e x  1  x  ,
 2! 

16
 x3 x4
   .......................
3! 4!
But, how can truncation error be controlled in this example? We can use the concept of relative
approximate error to see how many terms need to be considered. Assume that one is calculating
e 1.2 using the Maclaurin series, then

1.2 1.2 2 1.2 3

e  1  1.2    ...................
2! 3!

Let us assume one wants the absolute relative approximate error to be less than 1%. In Table 1,
we show the value of e 1.2 , approximate error and absolute relative approximate error as a
function of the number of terms, n .

n e 1.2 Ea a %
1 1 − −
2 2.2 1.2 54.546
3 2.92 0.72 24.658
4 3.208 0.288 8.9776
5 3.2944 0.0864 2.6226
6 3.3151 0.020736 0.62550

Using 6 terms of the series yields a a < 1%.

3) Rounding Error
1
A computer can only represent a number approximately. For example, a number like may be
3
represented as 0.333333 on a PC. Then the round off error in this case is:
1
 0.333333  0.00000033 .
3
Then there are other numbers that cannot be represented exactly. For example,  and 2 are
numbers that need to be approximated in computer calculations.
Q: What problems can be created by round off errors?

17
A: Twenty-eight Americans were killed on February 25, 1991. An Iraqi Scud hit the Army
barracks in Dhahran, Saudi Arabia. The patriot defense system had failed to track and intercept
the Scud. What was the cause for this failure?
The Patriot defense system consists of an electronic detection device called the range gate. It
calculates the area in the air space where it should look for a Scud. To find out where it should
aim next, it calculates the velocity of the Scud and the last time the radar detected the Scud.
Time is saved in a register that has 24 bits length. Since the internal clock of the system is
measured for every one-tenth of a second, 1/10 is expressed in a 24 bit-register as
0.00011001100110011001100. However, this is not an exact representation. In fact, it would
need infinite numbers of bits to represent 1/10 exactly. So, the error in the representation in
decimal format is

Figure 1.1 Patriot missile (Courtesy of the US Armed Forces,

http://www.redstone.army.mil/history/archives/patriot/patriot.html)

1
 (0  2 1  0  2 2  0  23  1 2  4  ...  1 2 22  0  2  23  0  2 24 )
10
 9.537 108

The battery was on for 100 consecutive hours, hence causing an inaccuracy of
s 3600s
 9.537  10 8  100 hr 
0.1s 1hr
 0.3433s

18
The shift calculated in the range gate due to 0.3433s was calculated as 687m . For the Patriot
missile defense system, the target is considered out of range if the shift was going to more than
137 m .
1.3.Computer Representation of Numbers
After completing this sections lesson, students are expected to:
 identify how to represent numbers on a computer using different techniques
 represent numbers both using fixed point and floating point representation
 understand about the round-off errors
 measure the maximum error bounds in number representation
A computer is in general built to handle pieces of information of a fixed size called a word. The
number of digits in a word (usually binary) is called the Word-Length of the computer. Typically
word-lengths are 16, 32, 48, or 64 bits. A real or integer number is usually stored in a word.
Integers can be exactly represented, provided that the word-length suffices to store all the digits
in its representation. But it is not easy to store real numbers, even the smallest rational numbers
such as 1/3 = 0.3333. .. can't be represented on a word-length.
Since there is a fixed space (word-length) of memory in the digital computer, a given number in
a certain base must be represented in a finite space in the memory of the computer. Thus, all
digits of a given number may not be represented in the memory. There are two conventional
ways for the representation of data in the word-length.
1.3.1. Fixed Point Representation
In the first generation of computers calculations were made on a Fixed Point Number system,
that is, real numbers were represented with a fixed number of t binary digits. If the word-length
of the computer is s+1 bits (including the sign bit), then only numbers in the interval =
[−2 ,2 ] are permitted.
Suppose the number to be represented has n digits. In the fixed point representation system
the n digits are subdivided in to & where is reserved for the integral part and is
reserved for the fractional part.

19
 Integral part Decimal point Fractional part
= +

Consider an old time cash register that would ring any purchase between 0 and 999.99 units of
money. Note that there are five (not six) working spaces in the cash register (the decimal
number is shown just for clarification).

Q: How will the smallest number 0 be represented?

A: The number 0 will be represented as

0 0 0 . 0 0

Q: How will the largest number 999.99 be represented?

A: The number 999.99 will be represented as

9 9 9 . 9 9

Q: Now look at any typical number between 0 and 999.99, such as 256.78. How would it be
represented?
A: The number 256.78 will be represented as
2 5 6 . 7 8
Q: What is the smallest change between consecutive numbers?
A: It is 0.01, like between the numbers 256.78 and 256.79.
Q: What amount would one pay for an item, if it costs 256.789?
A: The amount one would pay would be rounded off to 256.79 or chopped to 256.78. In either
case, the maximum error in the payment would be less than 0.01.
Q: What magnitude of relative errors would occur in a transaction?
A: Relative error for representing small numbers is going to be high, while for large numbers the
relative error is going to be small.
For example, for 256.786, rounding it off to 256.79 accounts for a round-off error of
256.786  256.79  0.004 . The relative error in this case is
 0.004
t   100
256.786
 0.001558% .

20
For another number, 3.546, rounding it off to 3.55 accounts for the same round-off error of
3.546  3.55  0.004 . The relative error in this case is
 0.004
t   100
3.546

 0.11280%
Example 1.7
Let = 13042, then represent using a fixed point representation if;
i) =3 & =2
Solution:

130 42

Then = 130.42
So, the computer understands the number as 130.42
ii) =3=
Solution:

130 420

Then = 130.420
So, the computer understands the number as 130.420
iii) =1 & =4

Solution:

1 30 42

Then = 1.3042
So, the computer understands the number as 1.3042
iv) =2=

Solution:

21
 13 04

Then = 13.04
So, the computer understands the number as 13.04
As we can observe from the above example this way of representing numbers has different
limitations specially its exposed to large absolute or relative errors because the numbers must be
rounded to be represented.
1.3.2. Floating Point Representation
One of the most common way for the representation of numbers on a digital computers is the
Floating Point, in which the position of the decimal ( binary ) is not fixed at the outset; rather its
position with respect to the first digit is indicated for each number separately.
On a digital computers any n-digit floating point number in base , can be written in general
form as:
= ±(. … )×
Where are digits or bits with values from 0 to ( − 1) .

and the whole are called Mantissa.

= the Base.
= the integral exponent.
Common choices of the base are
 Binary ( Base Two ) which are most commonly used on digital computers.
 Decimal ( Base Ten ) which are used in our day to day activities.
 Hexadecimal ( Base Sixteen ) which are an alternative for base two on some
machines.
Since base ten is familiar to us let study the floating point representation of numbers by using
this base, thus any real number can be approximately represented by the floating point
representation in base ten as:
= ±(0. … ) × 10 Where 0 ≤ ≤9 & ≠0
Thus the standard floating point representation slot of n-bits is given by:
t-bits 1-bit r-bits

22
 … ±

mantissa sign exponent = + +1

Let us go back to the example where we have five spaces available for a number. Let us also
limit ourselves to positive numbers with positive exponents for this example. If we use the same
five spaces, then let us use four for the mantissa and the last one for the exponent. So the
smallest number that can be represented is 1 but the largest number would be 9.999  10 9 . By
using the floating-point representation, what we lose in accuracy, we gain in the range of
numbers that can be represented. For our example, the maximum number represented changed
from 999.99 to 9.999  10 9 .

What is the error in representing numbers in the scientific format? Take the previous example of
256.78. It would be represented as 2.568  10 2 and in the five spaces as

2 5 6 8 2

Another example, the number 576329.78 would be represented as 5.763  10 5 and in five spaces
as
5 7 6 3 5
So, how much error is caused by such representation. In representing 256.78, the round off error
created is 256.78  256.8  0.02 , and the relative error is
 0.02
t   100  0.0077888% ,
256.78
In representing 576329.78 , the round off error created is 576329 .78  5.763  10 5  29 .78 , and
the relative error is
29.78
t   100  0.0051672% .
576329.78
What you are seeing now is that although the errors are large for large numbers, but the relative
errors are of the same order for both large and small numbers.
Example 1.8
Represent the following numbers by their standard floating point representation on a five digit
computational tool that means when = 5.

23
 i) = 13456
ii) = 134.5634
iii) = 1345.267
iv) = 1345.236
and compute their absolute true error of approximation?
Solution:
i) = 0.13456 × 10
13456 + 05
and also = 0.13456 × 10 thus no error is induced
ii) but here = 0.13456 × 10
13456 + 03

and also = 0.1345634 × 10 thus there is an error, that is

= | − | = |0.1345634 × 10 − 0.13456 × 10 |

= |0.1345634 − 0.13456| × 10
= 0.0000034 × 10
= 0.0034
Note:
1) A number cannot be represented exactly if it contains more than t-bits in the mantissa.
2) In the case of binary ( base two ) representation, there are two most commonly used
standard notations:
i) 32 bits of word length, of which
24 bits are reserved for the mantissa
7 >> >> >> >> >> exponent
1 bit is used by the sign.
ii) 64 bits of word length, of which
52 bits are reserved for the mantissa
11 >> >> >> >> >> exponent
1 bit is used by the sign.

24
In a floating point representation system with t digits for the mantissa a number with mantissa
greater than t digits cannot be represented exactly. Thus, such a number most somehow be
rounded-off to t-digits. There are two ways of reducing the number of digits of a given number,
which are discussed as follows.
i) By Chopping
In a t-digit computation, if all the digits of the mantissa to the right of the digit are just
dropped off , then we say that the number is approximated by chopping.
Let = ± × 10 where = 0. … … .
Then in a t-digit computer the mantissa is chopped as:
= 0. … where all the digits starting from to the right are dropped.
Thus, = × 10 = 0. … × 10
 Error=| − | = 0.000 … 0 … × 10
= 0. … × 10 × 10
= . … × 10 × 10 × 10
≤ 9.999 … × 10 since 0 ≤ ≤9
≤ 10 × 10 = 10
Therefore, the maximum absolute error bound by chopping is
Error=| − | ≤ 10
Example 1.9
Let = 4 and = 14.28625. Using chopping find the maximum error committed in a 4 digit
computation?
Solution
= 14.28625 = 0.1428625 × 10 , on a 4 digit computation x is approximated by
= 0.1428 × 10 using chopping
Then, the maximum absolute error bound becomes
Error=| − | = |0.1428625 − 0.1428| × 10
= 0.0000625 × 10 = 0.625 × 10 × 10
= 0.625 × 10 = 6.25 × 10 < 9.9 × 10
< 10 × 10 = 10
Therefore, the maximum error committed is less than 10 = 0.01 .
Or simply since = 4, = 2 , we have

25
 Error=| − | ≤ 10 = 10 = 10 = 0.01 .
ii) Rounding / Rounding off/
Let x = ±0. … … × 10
In this case if
 ≥ 5 , we add one to , i.e. we round up.
 < 5 , we merely chopped off all the terms after .
Thus,
(0. … + 10 ) × 10 , ≥5
=
0. … × 10 , <5
For Case Ι , i.e., ≥5:
The Absolute Error becomes,
=| − |=
= |0. …′ + 1 − 0. … | × 10
= 0.000 … 0 ∝ ∝ … ∝ × 10
But ≥5 implies ∝ <5
= | − | =∝ .∝ … ∝ × 10 × 10 × 10
≤ 4.999 … × 10
< 5 × 10 × 10 = 0.5 × 10
Therefore, the maximum absolute error bound by rounding up is

E=| − | ≤ × 10

For Case ΙΙ , i.e., < 5 , Show that the maximum error bound is the same as the
previous case , i.e.
E=| − | ≤ × 10

Thus, the round off error in a t digit computation is:

E=| − | ≤ × 10

Example 1.10
Let = 4 and = 14.28625. Using Rounding find the maximum error committed in a 4 digit
computation?
Solution

26
 = 14.28625 = 0.1428625 × 10 , on a 4 digit computation x is approximated by
= 0.1429 × 10 using Rounding
Then, the maximum absolute error bound becomes
Error=| − | = |0.1428625 − 0.1429| × 10
= 0.0000375 × 10 = 0.375 × 10 × 10
= 0.375 × 10 = 3.75 × 10 < 4.99 × 10
< 5 × 10 = × (10 )

Therefore, the maximum error committed is less than × (10 ) = 0.005 .

Or simply since = 4, = 2 , we have

Error=| − | ≤ × (10 ) = × (10 ) = × (10 ) = 0.005 .

Activity 1.2
1. Compute the absolute and relative errors committed when the numbers in Example 1.8
( ) are represented on a five digit computational tool?
2. Derive a formula for the maximum relative error bound when a number = ± × 10
where = 0. … … . is represented on a t-digit computational tool ,using both
chopping and rounding?
3. Find the maximum absolute error committed in a six digit computation when x=4/3 is
approximated Using chopping?
4. Find the maximum error committed in a six digit computation approximated by Using
Rounding for :
a) x=4/9
b) x=2/3

1.4. Propagation of Errors

After completing this sections lesson, students are expected to:
 define and understand propagation of errors in numerical methods
 identify the different sources of propagated errors
 measure and control the propagated errors

27
If a calculation is made with numbers that are not exact, then the calculation itself will have an
error. How do the errors in each individual number propagate through the calculations. Let’s
look at the concept via some examples.
Example 1.11
Find the bounds for the propagation error in adding two numbers. For example if one is
calculating X  Y where
X  1.5  0.05 ,
Y  3.4  0.04 .
Solution
By looking at the numbers, the maximum possible value of X and Y are
X  1.55 and Y  3.44
Hence
X  Y  1.55  3.44  4.99
is the maximum value of X  Y .
The minimum possible value of X and Y are
X  1.45 and Y  3.36 .
Hence
X  Y  1.45  3.36
 4.81
is the minimum value of X  Y .
Hence
4.81  X  Y  4.99.
One can find similar intervals of the bound for the other arithmetic operations of
X  Y , X * Y , and X / Y . What if the evaluations we are making are function evaluations instead?
How do we find the value of the propagation error in such cases.
If f is a function of several variables X 1 , X 2 , X 3 ,......., X n 1 , X n , then the maximum possible

value of the error in f is

f f f f
f  X 1  X 2  .......  X n 1  X n
X 1 X 2 X n 1 X n

Example 1.12

28
The strain in an axial member of a square cross-section is given by
F

h 2E
where
F =axial force in the member, N
h = length or width of the cross-section, m
E =Young’s modulus, Pa
Given
F  72  0.9 N
h  4  0.1 mm
E  70  1.5 GPa
Find the maximum possible error in the measured strain.
Solution
72

(4  10 ) (70  10 9 )
3 2

 64 .286  10 6
 64 .286 

  
  F  h  E
F h E
 1
 2
F h E
 2F
 3
h h E
 F
 2 2
E h E
1 2F F
E  2
F  3 h  2 2 E
h E h E h E

1 2  72
 3 2 9
 0.9   0.0001
(4  10 ) (70  10 ) (4  10 ) (70  10 9 )
3 3

72
 3 2 9 2
 1.5  10 9
(4  10 ) (70  10 )

29
  8.0357  10 7  3.2143  10 6  1.3776  10 6
 5.3955  10 6
 5.3955 
Hence
 (64 .286   5.3955  )
implying that the axial strain,  is between 58 .8905  and 69 .6815 
Example 1.13
Subtraction of numbers that are nearly equal can create unwanted inaccuracies. Using the
formula for error propagation, show that this is true.
Solution
Let
z  x y
Then
z z
z  x  y
x y

 (1)x  (1)y

 x  y

So the absolute relative change is

z x  y

z x y

As x and y become close to each other, the denominator becomes small and hence create large
relative errors.
For example if
x  2  0.001
y  2.003  0.001

z 0.001  0.001

z | 2  2.003 |
= 0.6667
= 66.67%
1.5. Stability of Algorithms and Conditioning numbers
30
After completing this sections lesson, students are expected to:
 identify whether an algorithm is stable or unstable
 identify whether a given problem is well or ill conditioned
 calculate the condition numbers of different problems
Since we must live with errors in our numerical computations, the next question natural is
regarding appraisal of a given computed solution: In view of the fact that the problem and the
numerical algorithm both yields error, can we trust the numerical solution of a nearby problem
(or the same problem with slightly different data) to differ by only a little from our computed
solution? A negative answer could make our computed solution meaningless.
This question can be complicated to answer in general, and it leads to notations such as problem
sensitivity and algorithm stability. If the problem is too sensitive, or ill-conditioned, meaning
that even a small perturbation in the data produces a large difference in the result.

1.5.1. Stability of Algorithms

Definition: An algorithm is an unambiguous and precise description of operations executed on

input data in a finite number of steps, transforming it to the desired output.
Definition: A propagated error is an error that occurs during the implementation of the steps of
an algorithm. Suppose a given algorithm has an initial error :
i) The propagated error generated will have a linear growth if the final error induced after
applying the algorithm is a constant multiple of the initial error, i.e.,
=
ii) The growth is exponential if
= , where n is the number of steps in the algorithm.
Linear growth is usually avoidable and when & are small the results are generally
acceptable, as a consequence, an algorithm that exhibits linear growth is called a stable
algorithm.
Exponential growth of error should be avoided, since the term becomes large even for
relatively small values of n, this leads to unacceptable result regardless of . An algorithm
which exhibits such an error is called unstable algorithm.

31
 Figure 1.2 graphs of stable and unstable algorithms.

1.5.2. Conditioning or Condition of a Problem

The words condition and conditioning are used informally to indicate how sensitive the solution
of a problem may be to small relative change in the input data. The condition of a numerical
problem is a qualitative or quantitative statement about how it is to solve, irrespective of the
algorithm used to solve it.
As a qualitative example, consider the solution of two simultaneous linear equations. The
problem may be described graphically by the pair of straight lines representing each equation:
the solution is then the point of intersection of the lines.

well Conditioned ill-conditioned.

Figure 1.2 graph of solution of two simultaneous linear equations
The left-hand problem is easier to solve than the right hand one, irrespective of the graphical
algorithm used. For example, a better (or worse) algorithm is to use a sharper (or blunter) pencil:
but in any case it should be possible to measure the coordinates of the solution more exactly in
the left-hand case than the right.

32
Quantitatively, the condition number K of a problem is a measure of the sensitivity of the
problem to a small perturbation or change. If this number is large, it indicates the problem is ill-
conditioned problem; in contrast, if the number is modest , the problem is recognized as a well-
conditioned problem.
For example, we can consider the problem of evaluating a differentiable function ( ) . Let be
a point close to . In this case K is a function of defined as the relative change in ( )
caused by a unit relative change in . That is
|[ ( ) ( )]/ ( )|
( ) = lim → |( )/ |
| ( ) ( )|
= lim → | |
( )
. ,( )
=
( )

Example 1.14
Suppose ( ) = √ . We get
,
. ( ) .[1/2 ] 1
( )= = =2
( ) √

So K is a constant which implies that taking square roots is equally well conditioned for all
non-negative, and that the relative error is reduced by half in the process.
Example 1.15
Suppose ( )= . In this case we get

,
. ( ) .[1/(1− )2
( )= = =
( ) 1/(1− ) 1−

So ( ) can get arbitrarily large for values of close to 1 and can be used to measure the
relative error in ( ) for such values, e.g. if = 1.000001 then the relative error will increase
by a factor of about 10 .
Review Exercise

1. Round-off the following numbers to four decimal places.

a) 0.235082 b) 0.0022218 c) 4.50089 d) 2.36425 e) 1.3456
2. The following numbers are correct to the last digits, find the sum.
i) 2.56, 4.5627, 1.253, 1.0534

33
 ii) 1.3526, 2.00462, 1.532, 28.201, 31.0012
3. Find the relative error in computation of − for = 12.05 and = 8.02 having
absolute errors ∆ = 0.005 and ∆ = 0.001.
4. Find the relative error in computation of + for = 11.75 and = 6.56 having
absolute errors ∆ = 0.001 and ∆ = 0.003 respectively.
5. If =4 − 5 , find the percentage error in at = 1 , if the error is = 0.04.

6. If be represented approximately by 0.8333, find both relative error and percentage

error.
7. If ( ) = 4 cos − 6 , find the relative percentage error in ( ) for = 0 , if the
error is = 0.005.
8. Determine the number of correct digits in the number given its relative error .
i) = 386.4, = 0.3.
ii) = 86.34, = 0.1.
iii) = 0.4785, = 0.2 × 10 .
9. Evaluate = √5.01 − √5 correct to three significant digits.
10. If is approximated to 0.6667. Find

i) absolute error
ii) relative error and
iii) percentage error

11. If = and error in , , be 0.001, 0.002, 0.003, compute the relative error

in . Where = = = 1.
12. If the true value of a number is 2.546282 and 2.5463 is its approximate value; find the
absolute error, relative error and the percentage error in the number.
13. If = 10.00 ± 0.05, = 0.0356 ± 0.002, = 15300 ± 100, and = 62000 ± 500
then find the maximum value of the absolute error in
i) + + +
ii) +5 −
iii)
14. If (0.31 + 2.73)/( + 0.35)

34
 where the coefficients are rounded off find the relative and absolute error in when
= 0.5 ± 0.1.
15. If = 5.43 and = 3.82 , where denote the length and breadth of a
rectangular plate, measured accurate up to 1 , find the error in computing the area.

35
 CHAPTER TWO
2. SOLUTION OF ALGEBRAIC AND TRANSCENDENTAL
EQUATIONS
2.1. Introduction:
Consider the equation
f ( x)  0 (2.1)
which may be given explicitly as a polynomial of degree n in x or f (x) may be defined
implicitly as a transcendental function. An equation which contains polynomials, exponential
functions, logarithmic functions, trigonometric functions etc. is called a transcendental equation.
Finding one or more roots of Eq. (2.1) is one of the more commonly occurring problems of
applied mathematics. Since there is no general formula for the solution of polynomial equations,
no general formula will exist for the solution of an arbitrary nonlinear equation of the form Eq.
(2.1) where f is a continuous real-valued function. How can we then decide whether or not such
an equation possesses a solution in the set of real numbers, and how can we find a solution? The
present chapter is devoted to the study of these questions. Our goal is to develop simple
numerical methods for the approximate solution of the Eq. (2.1) where f is a real-valued
function, defined and continuous on a bounded and closed interval of the real line.
Unit Objectives:
1. At the end of the unit students will be able to:
 use the bisection method to finding roots of a nonlinear equation,
 enumerate the advantages and disadvantages of the bisection method
 use the Regula-Falsi method to finding roots of a nonlinear equation,
 enumerate the advantages and disadvantages of the Regula-Falsi method
 use the Newton-Raphson method to solve a nonlinear equation,
 discuss the advantages and drawbacks of the Newton-Raphson method.
 use the secant method to solve a nonlinear equation.
 discuss the advantages and drawbacks of the secant method
 use the Iteration method to numerically solve a nonlinear equation.
 discuss the advantages and drawbacks of the Iteration method
 discuss about the convergence of iteration methods

36
 2.1. Preliminaries:

Definition 2.1: A number  is the solution of f ( x )  0 if f ( )  0. Such a solution  is a root

or a zero of f ( x )  0.

Geometrically the root of the equation Eq. (2.1) is the value of x at which the graph of
y  f (x) intersects the x  axis.
Definition 2.2: (Simple root) A number  is a simple root of f ( x )  0 if f ( )  0. and

f ' ( )  0. Then, we can write f (x) as f ( x )  ( x   ) g ( x ), g ( )  0.

For example, since ( x  1) is a factor of f ( x)  x 3  x  2  0, we can write

f ( x)  ( x  1)( x 2  x  2)  ( x  1) g ( x), g (1)  0.

Alternately, we find f 1  0 , f '  x   3 x 2  1, , f ' 1  4  0. Hence, x  1 is a simple root of

f ( x)  x 3  x  2  0.
Definition 2.3: (Multiple root) A number  is a multiple root, of multiplicity m, of f ( x)  0 , if

f ( )  0, f ' ( )  0,..., f ( m1) ( )  0, and f m    0 . Then, we can write f (x) as

f  x   ( x   ) m g ( x ), g ( )  0.

For example, consider the equation f ( x)  x 3  3x 2  4  0. We find

f (2) = 8 – 12 + 4 = 0, f '  x   3x 2  6 x, f ' (2)  12 – 12 = 0, f ' '  x   6 x  6, f ' ' 2   6  0.

Hence, x  2 is a multiple root of multiplicity 2 (double root) of f ( x)  x 3  3x 2  4  0.

We can write f ( x)  ( x  2) 2 ( x  1)  ( x  2) 2 g ( x), g (2)  3  0.

In this chapter, we shall be considering the case of simple roots only.
There are two types of methods that can be used to find the roots of Eq. (2.1).
(i) Direct Method: These methods give the exact value of the roots (in the absence of round off
errors) in a finite number of steps. These methods determine all the roots at the same time.

For example, a direct method gives the root of a linear or first degree equation
a 0 x  a1  0 (2.2)
a1
x= (2.3)
a0

37
Similarly, the roots of the quadratic equation
a 0 x 2  a1 x  a 2  0 (2.4)

 a1  a12  4a 0 a 2
are given by x = (2.5)
2a 0
(ii) Iterative methods: These methods are based on the idea of successive approximations, i.e
starting with one or more initial approximations to the root ,we obtain a sequence of
approximations or iterates {xk } which in the limit as k  , converge to the exact root  . An

iterative method for finding a root of the equation f ( x)  0 can be obtained as

xk 1   ( xk ), k  0,1, 2,... (2.6)

This method uses one initial approximation to the root x0 . The sequence of approximations
is given by
x1   ( x0 ), x2   ( x1 ), x3   ( x2 ), ... (2.7)

The function  is called an iteration function and x0 is called an initial approximation.

If a method uses two initial approximations x0 , x1, to the root, then we can write the

method as
xk 1   ( xk 1, xk ), k  1, 2,3,... (2.8)

Convergence of iterative methods:

The sequence of iterates, xk , is said to converge to the exact root  , if

lim xk   , or lim xk    0 (2.9)

k  k 

The error of approximation at the kth iterate is defined as  k  xk   .

Then, we can write (2.9) as

lim xk    lim  k  0 (2.10)

k  k 

Remark: Given one or two initial approximations to the root, we require a suitable iteration
function  for a given function f (x ) , such that the sequence of iterates, xk , converge to the

exact root  . Further, we also require a suitable criterion to terminate the iteration

38
Criterion to terminate iteration procedure:
Since, we cannot perform infinite number of iterations, we need a criterion to stop the iterations.
we can select a tolerance   0 and generate x1 , x 2 , x3 ,..., x k , x k 1 , until one of the following
conditions is met:
(i) The equation f ( x )  0 is satisfied to a given accuracy or f ( xk ) is bounded by an error

tolerance  .
f ( xk )   . (2.11)

(ii) The magnitude of the difference between two successive iterates is smaller than a given

accuracy or an error bound  .

xk 1  xk   . Or (2.12)

x k 1  xk
(iii) Relative error   . (2.13)
xk 1

Generally, we use the second criterion. For example, if we require two decimal place accuracy,

then we iterate until xk 1  xk  0.005. If we require three decimal place accuracy, then we

iterate until xk 1  xk  0.0005. But inequality (Eq. 2.13) is the best stopping criterion to apply

because it comes closest to testing relative error.

As we have seen earlier, we require a suitable iteration function and suitable initial
approximation(s) to start the iteration procedure. In the next section, we give a method to find
initial approximation.
Initial Approximations:
Initial approximations to the root are often known from the physical consideration of the
problem. Otherwise, graphical methods are generally used to obtain initial approximations to the
root (See section 2.2.1).
Another commonly used method to obtain initial approximation to the root is based on
intermediate value theorem, which stases;

39
Theorem 2.1: If f (x ) is continuous on some interval [a, b] and f ( a ) f (b )  0, then the equation
f ( x )  0 has at least one real root or an odd number of real roots in the interval (a, b).

This result is very simple to use. We set up a table of values of f (x ) for various values of x
Studying the changes in signs in the values of f (x ) , we determine the intervals in which the
roots lie. For example, if f (1) and f ( 2) are of opposite signs, then there is a root in the interval
(1,2).
Example 2.1: Determine an interval of length one unit in which the negative real root, which is
smallest in magnitude lies for the equation f ( x)  9 x 3  18x 2  37 x  70  0.

Solution: Let f ( x)  9 x 3  18x 2  37 x  70  0. Since, the smallest negative real root in magnitude
is required, we form a table of values for x  0,
x -5 -4 -3 -2 -1 0

f (x ) -560 -210 -40 4 -24 -70

table 2.1 values of f (x )

Since , f (  2) f (  1)  0, the negative root of smallest magnitude lies in the interval (-2, -1).
Activity 2.1:
1.Define a (i) root, (ii) simple root and (iii) multiple root of an algebraic equation f ( x )  0.
2. State the intermediate value theorem.
3. How can we find an initial approximation to the root of f ( x )  0 ?
4. Show that the following equations have at least one solution in the given intervals.
i) x cos x  2 x 2  3x  1  0, [0.2, 0.3] and [1.2, 1.3]

ii) ( x  2) 2  ln x  0, [1, 2] and [e, 4]

5. Find intervals containing solutions to the following equations.
i) x  3  x  0 ii) 4 x 2  e x  0 iii) x 3  2 x 2  4 x  2  0
6. What are the criteria used to terminate an iterative procedure?

2.3. Solution of Non-linear Equations:

2.3.1. Graphical Methods:

40
One alternative to obtain an approximate solution is to plot the function and determine where it
cross the x-axis. If the equation f ( x)  0 can be conveniently written in the form f1 ( x)  f 2 ( x),

then the point of intersection the graph of y  f1 ( x ) and y  f 2 ( x) gives the root of f ( x)  0.

Example 2.2: Determine the approximate solution of f ( x)  e  x cos x  x  0.

Solution: Let f1 ( x)  e  x cos x and f 2 ( x )   x Such that f ( x )  f1 ( x)  f 2 ( x )

Draw the graphs of f1 ( x ) and  f 2 ( x ) on the same coordinate axis.

1  f 2 ( x)  x
0.5
f1 ( x)  e  x cos x
(0,0) 0.5 1
Figure 2.1
0.5 is approximate solution of f ( x )  0

Remark: Graphical techniques are of limited practical value because they are not precise.
However, graphical methods can be utilized to obtain rough estimates of the roots. These
estimates can be employed as starting guesses for numerical methods which will be discussed in
the next sections.
2.2.2 Bisection method:

This method is based on the repeated application of The Intermediate Value Theorem.
Suppose f is a continuous function defined on the interval [a , b ], with f (a) and f (b)
of opposite sign. The Intermediate Value Theorem implies that a number m exists in ( a, b)
with f (m)  0. Although the procedure will work when there is more than one root in the
interval (a, b ), we assume for simplicity that the root in this interval is unique. The method
calls for a repeated halving (or bisecting) of subintervals of [a, b] and, at each step, locating
the half containing m .
a1  b1
To begin, set a1  a and b1  b and let m1 be the midpoint of [a , b ]; that is, m1 
2
• If f (m1 )  0, then m  m1 , and we are done.

41
 • If f (m1 )  0, then f (m1 ) has the same sign as either f (a1 ) or f (b1 ).

• If f (m1 ) and f (a1 ) have the same signs, then m (m1 , b1 ). Set a2  m1 and b2  b1.

• If f (m1 ) and f (a1 ) have opposite signs, then m (a1 , m1 ). Set a2  a1 and b2  m1.

Then reapply the process to the interval [a2 , b2 ]. After repeating the bisection process a number
of times, we either find the root or find subinterval which contains the root. We take the
midpoint of the last subinterval as an approximation to the root.
The method is shown graphically in the Fig. 2.2
y (b, f (b))

b m2 m3 m1 a x

(a, f ( a ))
Figure 2.2
Graphical representation of bisection method.

Example 2.3: Perform five iterations of the bisection method to obtain the smallest positive root

of the equation f ( x)  x 3  5x  1  0.

Solution: Since, f (0)  0 and f (1)  0, the smallest positive root lies in the interval (0,1).

a0  b0 0  1
Taking a0  0 , b0  1 , we get m1    0.5 , f (m1 )  f (0.5)  1.375  0,
2 2
and f (a0 ) f (m1 )  0,

Thus, the root lies in the interval (0,0.5).

a1  b1 0  0.5
Taking a1  0 , b1  0.5 , we get m2    0.25 ,
2 2
f (m2 )  f (0.25)  0.234375  0,
and f (a1 ) f (m2 )  0,

42
 Thus, the root lies in the interval (0,0.25).
The sequence of intervals given in the table 2.2
n a n 1 bn 1 mn f (mn ) f (an1 )

1 0 1 0.5 <0
2 0 0.5 0.25 <0
3 0 0.25 0.125 >0
4 0.125 0.25 0.1875 >0
5 0.1875 0.25 0.21875 <0
Hence, the root lies in (0.1875,0.21875).The approximate root is taken as the midpoint of this
interval ,that is 0.203125.
Example 2.4:
Find the interval in which the smallest positive root of the equation f ( x)  x 3  x  4  0 lies.
Determine the root correct to two decimal places using the bisection method.
Solution:
For f ( x)  x 3  x  4, we find f (0)  4, f (1)  4, f (2)  2.
Therefore, the root lies in the interval (1,2). The sequence of intervals given in the table 2.3
n an bn mn f ( mn ) f ( a n )

1 1 2 1.5 >0
2 1.5 2 1.75 >0
3 1.75 2 1.875 <0
4 1.75 1.875 1.8125 >0
5 1.75 1.8125 1.78125 >0
6 1.78125 1.8125 1.796875 <0
7 1.78125 1.796875 1.7890625 >0
8 1.7890625 1.796875 1.792969 >0
9 1.792969 1.796875 1.794922 >0
10 1.794922 1.796875 1.795898 > 0.
After 10 iterations, we find that the root lies in the interval (1.795898, 1.796875).Therefore,
the approximate root is m = 1.796387. The root correct to two decimal places is 1.80.

43
Example 2.5: Show that f ( x)  x 3  4 x 2  10  0. has a root in [1,2], and use the Bisection
method to determine an approximation to the root that is accurate to at least within 10−4.
Solution: Because f (1)  5, f (2)  14 the Intermediate Value Theorem ensures that this
continuous function has a root in [1,2].
For the first iteration of the Bisection method we use the fact that at the midpoint of
[1,2] we have f (1.5)  2.375  0. This indicates that we should select the interval [1,1.5]
for our second iteration. Then again take the mid point of [1,1.5] , we find that
f (1.25)  1.796875  0. so our new interval becomes [1.25,1.5], whose midpoint is 1.375.
Continuing in this manner gives the values in Table 2.4.
n an bn mn f ( mn )

1 1.0 2.0 1.5 2.375

2 1.0 1.5 1.25 -1.79687
3 1.25 1.5 1.375 0.16211
4 1.25 1.375 1.3125 −0.84839
5 1.3125 1.375 1.34375 −0.35098
6 1.34375 1.375 1.359375 −0.09641
7 1.359375 1.375 1.3671875 0.03236
8 1.359375 1.3671875 1.36328125 −0.03215
9 1.36328125 1.3671875 1.365234375 0.000072
10 1.36328125 1.365234375 1.364257813 −0.01605
11 1.364257813 1.365234375 1.364746094 −0.00799
12 1.364746094 1.365234375 1.364990235 −0.00396
13 1.364990235 1.365234375 1.365112305 −0.00194
After 13 iterations, m13  1.365112305 approximates the root m with an error

m  m13  b14  a14  |1.365234375 − 1.365112305| = 0.000122070.

m  m13 b14  a14
Since a14  m , we have < ≤ 9.0 × 10−5,
m a14
So, the approximation is correct to at least within 10−4. The correct value of m to nine decimal
places is m  1.365230013. Note that m 9 is closer to m than is the final approximation m13 .

44
You might suspect this is true because f (m9 )  f (m13 ) , but we cannot be sure of this

unless the true answer is known.

The Bisection method, though conceptually clear, has significant drawbacks. It is relatively

slow to converge (that is, n may become quite large before m  mn is sufficiently small), and a

good intermediate approximation might be inadvertently discarded. However, the method has the
important property that it always converges to a solution.
Theorem 2.2: Let f be continuous on [a, b] and f (a ) f (b)  0. Then the Bisection method

generates a sequence mn n1 approximating the root  with the property

ba
mn    , when n  1 .
2n
Proof: For each n  1, we have
ba
bn  an  and  ∈ (an , bn ). (2.14)
2 n1
an  bn
Since mn  for all n  1,
2
bn  an b  a
it follows that mn     n .
2 2
Example 2.6: Determine approximately how many iterations are necessary to solve
f ( x)  x 3  4 x 2  10  0 with accuracy of ε =10-5 for a  1 and b  2 .
ba
Solution: This requires finding an integer n that will satisfies mn    
2n
2 1
i.e. mn    n
 2 n  10 5
2
To determine n we use logarithms to the base 10
n 5
2
Since 2  n  10  5  log10 < log10
10  5

2
  n log10  5

5
 n> 2
 16.6 .
log10
It would appear to require 17 iterations to obtain an approximate accuracy to 10-5.

45
REMARK: If an error tolerance ε is prescribed, then the approximate number of the iterations
[log(b  a)  log ]
required may be determined from the relation n  .
log 2
Activity 2.3:
1. Determine the number of iterations necessary to solve f ( x)  x 3  4 x 2  10  0 with
accuracy 10−3 using a  1 and b  2 .

2. Perform five iterations of the bisection method to obtain the smallest positive root of the
following equations
i) x 5  4 x  2  0 ii ) cos x  3x  1 iii) x 3  2 x 2  1  0 iv) 5 x 3  20 x 2  3  0

3. Find the root of the equation sin x  1  x 3 , which lies in the interval (  2, 1), correct to
three decimal places.
4. Use the Bisection method to find solutions, accurate to within 10−5 for the following equations
i) 3x  e x  0 for 1  x  2 ii ) 2 x  3 cos x  e x  0 for 0  x  1

ALGORITHM: The Bisection Method:

INPUT: endpoints a, b; tolerance TOL; maximum number of iterations N0.
OUTPUT: approximate solution p or message of failure.
Step 1 Set i = 1;
FA = f (a).
Step 2 While i ≤ N0 do Steps 3–6.
Step 3 Set p = a + (b − a)/2; (Compute pi.)
FP = f ( p).
Step 4 If FP = 0 or (b − a)/2 < TOL then
OUTPUT (p); (Procedure completed successfully.) STOP.
Step 5 Set i = i + 1.
Step 6 If FA ・ FP > 0 then set a = p; (Compute ai, bi.)
FA = FP
else set b = p. (FA is unchanged.)
Step7 OUTPUT (‘Method failed after N0 iterations, N0 =’, N0);
(The procedure was unsuccessful.)
STOP.

46
2.2.3 Method of False Position:

The method is also called linear interpolation method or chord method or regula-falsi method.
At the start of all iterations of the method, we require the interval in which the root lies. Let the
root of the equation f ( x)  0, lie in the interval ( xk 1 , xk ), that is, f k 1 f k  0, where

f ( xk 1 )  f k 1 , and f ( xk )  f k . Then , P( xk 1 , f k 1 ) and Q ( xk , f k ) are points on the curve

f ( x)  0. Draw a straight line joining the points P and Q (see Fig. 2.3). The line PQ is taken as

an approximation of the curve in the interval [ xk 1 , xk ]. The equation of the line PQ is given by

y  fk x  xk

f k 1  f k xk 1  xk
The point of intersection of this line PQ with the x  axis is taken as the next approximation
to the root. Setting y  0, and solving for x, we get

xk 1  xk x  x k 1
x  xk  ( ) fk  x k  ( k ) fk
f k 1  f k f k  f k 1
The next approximation to the root is taken as
x k  x k 1
x k 1  x k  ( ) fk
f k  f k 1
Simplifying, we can also write the approximation as

xk 1 f k  xk f k 1
xk 1 
f k  f k 1 k  1, 2,3,... (2.15)

Therefore, starting with the initial interval ( x0 , x1 ), in which the root lies, we compute

x0 f1  x1 f 0
x2 
f1  f 0

Now, if f ( x0 ) f ( x2 )  0, then the root lies in the interval ( x0 , x2 ). Otherwise, the root lies in

the interval ( x2 , x1 ). The iteration is continued using the interval in which the root lies, until
the required accuracy criterion satisfied.

47
 The method is shown graphically by

y P
x3
x0 x2 X

x1
Q
Figure 2.3
Method of false position
Remark : i) At the start of each iteration, the required root lies in an interval, whose length is
decreasing. Hence, the method always converges.
ii) The method of false position has a disadvantage. If the root lies initially in the
interval ( x0 , x1 ), then one of the end points is fixed for all iterations.
Example 2.7: Locate the intervals which contain the smallest positive real roots of the equation
x 3  3x  1  0. Obtain these roots correct to three decimal places, using the
method of false position.
Solution: We form the following table of values for the function f ( x ).
x 0 1 2 3

f (x ) 1 -1 3 19

Table 2.5
There is one positive real root in the interval (0,1). and another in the interval (1,2).
There is no real root for x  2 as f ( x )  0, for all x  2.
we find the root in (0,1). We have
x 0  0, x1  1, f 0  f ( x 0 )  f (0 )  1, f 1  f ( x1 )  f (1)   1 .

x0 f1  x1 f 0 0 1
x2    0.5 , f ( x 2 )  f ( 0 .5 )   0 .375 .
f1  f 0 1 1

Since, f (0) f (0.5)  0, the root lies in the interval (0,0.5).

48
 x0 f 2  x 2 f 0 0  0.5(1)
x3    0.36364 , f ( x3 )  f (0 .36364 )  0 .04283 .
f2  f0  0.375  1

Since f (0) f (0.36364)  0, the root lies in the interval (0,0.36364 ).

x0 f 3  x3 f 0 0  0.36364(1)
x4    0.34870 , f ( x 4 )  f ( 0 .34870 )   0 .00370 .
f3  f0  0.04283  1

Since f (0) f (0.34870)  0, the root lies in the interval (0,0.34870 ).

x0 f 4  x4 f 0 0  0.3487(1) , f ( x 5 )  f (0.34741 )  0 .00030 .

x5    0.34741
f 4  f0  0.00370  1

Since f (0) f (0.34741)  0, the root lies in the interval (0,0.34741).

x 0 f 5  x 5 f 0 0  0 .34741 (1)
x6   0 .347306
f5  f0  0 .0003  1

Now , x6  x5  | 0.347306 – 0.34741 | ≈ 0.0001 < 0.0005.

The root has been computed correct to three decimal places. The required root can be
taken as x  x 6  0.347306. We may also give the result as 0.347, even though x6 is more

accurate. Note that the left end point x  0 is fixed for all iterations.
Example 2.8: Find the root correct to two decimal places of the equation, cos x  xe x ,
using the method of false position.
Solution: Define f ( x)  cos x  xe x  0. There is no negative root for the equation. We have
f ( 0 )  1, f (1)   2 .17798 .
Since, f (0) f (1)  0, the root lies in the interval (0,1).

x0 f1  x1 f 0 0  1(1)
x2    0.31467 , f ( x 2 )  f ( 0 .31467 )  0 .51986 .
f1  f 0  2.177985  1

Since, f (0.31467) f (1)  0, the root lies in the interval (0.31467 ,1).

49
 x2 f1  x1 f 2 0.31467(2.17798)  1(0.51986)
x3    0.44673 , f ( x )  f ( 0.44673 )  0.20354 .
f1  f 2  2.17798  0.51986 3

Since, f (0.44673) f (1)  0, the root lies in the interval (0.44673,1).

x3 f1  x1 f 3 0.44673(2.17798)  91)  (1)0.20354

x4    0.49402 , f ( x 4 )  f ( 0 .49402 )  0 .07079 .
f1  f 3  2.17798  0.20354

Since, f (0.49402) f (1)  0, the root lies in the interval (0.49402 ,1).

x4 f1  x1 f 4 0.49402(2.17798)  (1)0.07079 , f ( x5 )  f (0 .50995 )  0 .02360 .

x5    0.50995
f1  f 4  2.17798  0.07079
Since, f (0.50995) f (1)  0, the root lies in the interval (0.50995 ,1).

x 5 f1  x1 f 5 0 .50995 (  2 .17798 )  (1) 0 .02360

x6    0 .51520
f1  f 5  2 .17798  0 .02360 , f ( x 6 )  f (0.51520 )  0 .00776 .
Since, f (0.51520) f (1)  0, the root lies in the interval (0.51520 ,1).

x 6 f1  x1 f 6 0 .51520 (  2 .17798 )  (1) 0 .00776

x7    0 .51692
f1  f 6  2 .17798  0 .00776

Now , x7  x6  | 0.51692 – 0.51520 | ≈ 0.00172 < 0.005.

The root has been computed correct to two decimal places. The required root can be
taken as x  x 7  0.51692.

Note that the right end point x  2 is fixed for all iterations.
Activity 2.4:
In the following problems, find the root as specified using the regula-falsi method.
1. Find the positive root of x 3  2 x  5. (Do only four iterations).
2. Find an approximate root of − 1.2 = 0 , correct to three decimal places.
3. Solve the equation , x tan x  1, starting with a  2 .5, and b  3, correct to three decimal places.

4. Find the smallest positive root of , x  e  x  0, correct to three decimal places.

5. Find the smallest positive root of x 4  x  10  0, correct to three decimal places.

50
ALGORITHM: The False Position Method:
INPUT: initial approximations p0, p1; tolerance TOL; maximum number of iterations N0.
OUTPUT: approximate solution p or message of failure.
Step 1 Set i = 2;
q0 = f ( p0);
q1 = f ( p1).
Step 2 While i ≤ N0 do Steps 3–7.
Step 3 Set p = p1 − q1( p1 − p0)/(q1 − q0). (Compute pi.)
Step 4 If | p − p1| < TOL then
OUTPUT (p); (The procedure was successful.)
STOP.
Step 5 Set i = i + 1;
q = f ( p).
Step 6 If q ・ q1 < 0 then set p0 = p1;
q0 = q1.
Step 7 Set p1 = p;
q1 = q.
Step 8 OUTPUT (‘Method failed after N0 iterations, N0 =’, N0);
(The procedure unsuccessful.) STOP.
2.2.4. Newton-Raphson Method:
This method is also called Newton’s method.
Let x0 be an initial approximation to the root of f ( x )  0. Then , P( x0 , f 0 ), where f 0  f ( x0 ), is a

point on the curve. Draw the tangent to the curve at P, (see Fig.2.4). We approximate the curve
in the neighborhood of the root by the tangent to the curve at the point P. The point of
intersection of the tangent with the x  axis is taken as the next approximation to the root. The
process is repeated until the required accuracy is obtained. The equation of the tangent to the
curve y  f (x) at the point , P( x0 , f 0 ), is given by

y  f ( x0 )  ( x  x0 ) f ' ( x0 )
where f ' ( x0 ) is the slope of the tangent to the curve at P.

51
Setting y  0 and solving for x, we get
f ( x0 )
x  x0  , f ' ( x0 )  0
f ' ( x0 )
The next approximation to the root is given by
f ( x0 )
x1  x0  , f ' ( x0 )  0
f ' ( x0 )
We repeat the procedure. The iteration method is defined as
f ( xk )
x k 1 = x k  , f ' ( xk )  0 k  0,1, 2,... (2.16)
f ( xk )

Which is the Newton-Raphson formula.

Alternate derivation of the method

Let xk be an approximation to the root of the equation f ( x )  0. Let x be an increment in

x such that xk  x is the exact root, that is f ( xk  x)  0.

Expanding in Taylor’s series about the point xk . we get

(x) 2
f ( xk )  xf ' ( xk )  f ' ' ( xk )  ...  0.
2!
Neglecting the second and higher powers of  x , we obtain
f ( xk )
f ( xk )  xf ' ( xk )  0, or x   .
f ' ( xk )
Hence, we obtain the iteration method
f ( xk )
xk 1  xk  x  xk  . f ' ( xk )  0
f ' ( xk )
which is same as the method derived earlier.
Geometrically, the method consists in replacing the part of the curve between the point
( x0 , f ( x0 )), and the x  axis by means of the tangent to the curve at the point.
The method is shown graphically in the Fig.2.4
y
P ( x0 , f 0 )

52
 x1 x 0 x

Figure 2.4
Newton-Raphson Method
Remark:1. Convergence of the Newton’s method depends on the initial approximation to the
root. If the approximation is far away from the exact root, the method diverges.
However, if a root lies in a small interval ( a , b ) and x 0  ( a , b ), then the method
converges.
2. The computational cost of the method is one evaluation of the function f (x ) and one
evaluation of the derivative f ' ( x ), for each iteration.
Example 2.9: Perform four iterations of the Newton’s method to find the smallest positive root
of the equation f ( x)  x 3  5 x  1  0.
Solution: We have f (0 )  1, f (1)   3 .
Since , f (0) f (1)  0, the smallest positive root lies in the interval (0,1).

Let f ( x)  x 3  5 x  1  0 and f ' ( x)  3 x 2  5

Now applying the Newton’s method, we obtain
xk3  5 xk  1 2 xk3  1
xk 1  xk   2 k  0,1,2,...
3xk2  5 3xk  5

Let x0  0.5 . We have the following results.

2 x03  1 2(0.5)3  1
x1    0.176471,
3 x02  5 3(0.5) 2  5
2 x13  1 2(0.176471) 3  1
x2  2   0.201568,
3x1  5 3(0.176471) 2  5

2 x23  1 2(0.201568) 3  1
x3    0.201640,
3x22  5 3(0.201568) 2  5

2 x33  1 2(0.201640) 3  1
x4    0.201640
3x32  5 3(0.201640) 2  5
Therefore, the root correct to six decimal places is x  0.201640.

53
Example2.10: Derive the Newton’s method for finding the qth root of a positive number N ,
1 1
q
N , where N  0, q  0. Hence, compute 17 , correct to four decimal places,
3

assuming the initial approximation as x0  2.

1

Solution: Let x  N , or x q  N . Define f ( x)  x q  N . Then , f ' ( x )  qx q 1 .

q

Newton’s method gives the iteration

xkq  N qxkq  xkq  N (q  1) xkq  N
xk 1  xk   
qxkq1 qxkq1 qxkq 1
1
For computing 17 3 , we have q  3 and N  17. Hence, the method becomes

2 xk3  17
xk 1  ,
3 xk2
With x0  2 , we obtain the following results.

2 x03  17 2(2) 3  17 k  0,1,2,...

x1    2.75,
3 x02 3(2) 2

2 x13  17 2(2.75) 3  17
x2    2.582645,
3 x12 3( 2.75) 2

2 x23  17 2( 2.582645) 3  17
x3    2.571332,
3x22 3(2.582645) 2

2 x43  17 2(2.571332) 3  17
x4    2.571282,
3 x42 3( 2.571332) 2

Now , x4  x3  | 2.571282 – 2.571332 | = 0.00005.

We may take x  2.571282 as the required root correct to four decimal places.

54
Activity2.5:

1. Given the following equations :

i) x 4  x 2  80  0 ii) 2 xe 2 x  sin x iii) cos x  x 2  x  0

determine the initial approximations for finding the smallest positive root. Use these to

find the root correct to three decimal places .Use Newton-Raphson method.

2. Using Newton-Raphson method solve = 12.34 with x0  10 root correct to four

decimal places.

3. Use Newton’s method to find solutions accurate to within 10−4 for the following problems.

i) x 3  2 x 2  5  0 , [1,4] iii) x 3  3 x 2  1  0 , [3,2]

ii) x  cos x  0 , [0, 2 ] iv) x  0.8  0.2 sin x  0 , [0, 2 ]

ALGORITHM : The Newton Method

Problem: To find a solution to f (x) = 0 given an initial approximation p0:
INPUT initial approximation p0; tolerance TOL; maximum number of iterations N0.
OUTPUT approximate solution p or message of failure.
Step 1 Set i = 1.
Step 2 While i ≤ N0 do Steps 3–6.
Step 3 Set p = p0 − f ( p0)/f1( p0). (Compute pi.)
Step 4 If | p − p0| < TOL then
OUTPUT (p); (The procedure was successful.)
STOP.
Step 5 Set i = i + 1.
Step 6 Set p0 = p. (Update p0.)
Step 7 OUTPUT (‘The method failed after N0 iterations, N0 =’, N0);
(The procedure was unsuccessful.)
STOP.

55
2.2.5. The secant Method:
We have seen that the Newton-Raphson Method requires the evaluation of derivatives of the
functions and this is not always possible, particularly in the case of functions arising in practical
problems. In the secant method ,the derivative at xk is approximated by the formula

f ( xk )  f ( xk 1 )
f ( x k )  (2.17)
x k  x k 1
Hence Newton-Raphson Method formula (2.16) becomes
f ( x k )( x k  xk 1 )
x k 1  x k  , f ( x k )  f ( xk 1 )  0 , k  1, 2,3,... (2.18)
f ( x k )  f ( x k 1 )

Which is the secant method formula .

Remark: This method requires two initial guesses, but unlike the bisection method, the two
initial guesses do not need to bracket the root of the equation. The secant method is an
open method and may or may not converge. However, when secant method converges, it
will typically converge faster than the bisection method. However, since the derivative is
approximated as given by Eq. (2.17), it typically converges slower than the Newton-
Raphson method.
Example 2.11: A root of the equation f ( x)  x 3  5 x  1  0 lies in the interval (0,1). Perform
four iterations of the secant method to obtain this root.

Solution: We have x0  0, x1  1, f (0)  1, and f (1)  3,

Now applying the secant method, we obtain
f ( x1 )( x1  x0 )  3(1  0)
x2  x1   1  0.25, f ( x 2 )  0.234375
f ( x1 )  f ( x0 )  3 1

f ( x 2 )( x2  x1 )  0.234375(0.25  1)
x3  x 2   0.25   0.186441, f ( x3 )  0.074276
f ( x 2 )  f ( x1 )  0.234375  3
f ( x3 )( x3  x2 ) 0.074276(0.186441  0.25)
x 4  x3   0.186441   0.201736, f ( x 4 )  0.000470
f ( x3 )  f ( x 2 ) 0.074276  0.234375

f ( x 4 )( x 4  x3 )  0.000470(0.201736  0.186441)
x5  x 4   0.201736   0.201640.
f ( x 4 )  f ( x3 )  0.000470  0.074276

56
Example 2.12: Given f ( x)  x 4  x  10  0 .Determine the initial approximations for finding the
smallest positive root. Use these to find the root correct to three decimal places
using the secant method.
Solution: We have f ( x)  x 4  x  10  0,
we find that f (0)  10, and f (1)  10, f (2)  4.
Hence, the smallest positive root lies in the interval (1,2).
The Secant method gives the iteration scheme

f ( xi )( xi  xi 1 )
xi 1  xi  i  1,2,3,...
f ( xi )  f ( xi 1 )

With x0  1, x1  2, we obtain the sequence of iterates

x2  1, 1.7143,

x3  1, 1.8385,

x4  1, 1.8578,

x5  1, 1.8556,

x6  1, 1.8556.
The root correct to three decimal places is 1.856.

Activity2.6:

1.Use secant method to obtain the smallest positive root , correct to three decimal places, of the
following questions
i) x 3  3 x 2  3  0

ii ) x 3  x 2  x  7  0

iii) x  e  x  0
2. Use the secant method to find solutions, accurate to within 10−5 for the following problems.
i) x 2  4 x  4  ln x  0 for 1  x  2
ii ) x  1  2 sin x  0 for 0  x  0.5

57
ALGORITHM: The Secant Method

Problem: To find a solution to f (x) = 0 given initial approximations p0 and p1:

INPUT initial approximations p0, p1; tolerance TOL; maximum number of iterations N0.
OUTPUT approximate solution p or message of failure.
Step 1 Set i = 2;
q0 = f ( p0);
q1 = f ( p1).
Step 2 While i ≤ N0 do Steps 3–6.
Step 3 Set p = p1 − q1( p1 − p0)/(q1 − q0). (Compute pi.)
Step 4 If | p − p1| < TOL then
OUTPUT (p); (The procedure was successful.)
STOP.
Step 5 Set i = i + 1.
Step 6 Set p0 = p1; (Update p0, q0, p1, q1.)
q0 = q 1 ;
p1 = p;
q1 = f ( p).
Step 7 OUTPUT (‘The method failed after N0 iterations, N0 =’, N0);
(The procedure was unsuccessful.)
STOP.
2.2.6. Iteration Method:

The method is also called method of successive approximations or fixed point iteration method.

The first step in this method is to rewrite the given equation f ( x)  0 in an equivalent form as
x   (x) (2.19)
There are many ways of rewriting f ( x)  0 in this form.

For example, f ( x)  x 3  5 x  1  0, can be rewritten in the following forms.

1
x3 1 5x 1
x , x  (5 x  1) 3 , x  , etc. (2.20)
5 x
Now, finding a root of f ( x)  0 is same as finding a number  such that    ( ), that is,

58
a fixed point of  (x). A fixed point of a function  is a point  such that    ( ).
Using Eq. (2.19), the iteration method is written as
xk 1   ( xk ), k  0,1, 2,... (2.21)

The function  (x) is called the iteration function.

Starting with the initial approximation x0 , we compute the next approximations as

x1   ( x0 ), x2   ( x1 ), x3   ( x2 ), … (2.22)
The stopping criterion is same as used earlier.
Since, there are many ways of writing f ( x)  0 as x   (x), it is important to know whether all
or at least one of these iteration methods converges.
Convergence of an iteration method xk 1   ( xk ), k  0,1, 2,... depends on the choice of the

iteration function  (x), and a suitable initial approximation x0 , to the root.

Consider again, the iteration methods given in Eq.(2.20), for finding a root of the equation
f ( x)  x 3  5 x  1  0. The positive root lies in the interval (0,1).

xk3  1
(i ) xk 1  , k  0,1, 2,... (2.23)
5
With x0  1, we get the sequence of approximations as

x1  0.4, x2  0.2128, x3  0.20193, x4  0.20165, x5  0.20164.

The method converges and x  x5  0.20164 is taken as the required approximation to

the root.
1
(ii ) x k 1  (5 x k  1) 3 , k  0,1, 2,... (2.24)

With x0  1, we get the sequence of approximations as

x1  1.5874, x2  1.9072, x3  2.0437, x4  2.0968,...

which does not converge to the root in (0,1).

5 xk  1
(iii) x k 1  k  0,1, 2,... (2.25)
xk

With x0  1, we get the sequence of approximations as

x1  2.0, x2  2.1213, x3  2.1280, x4  2.1284,...

59
which does not converge to the root in (0,1).
Now, we derive the condition that the iteration function  (x) should satisfy in order that
the method converges.
Condition of convergence

The iteration method for finding a root of f ( x)  0 is written as

xk 1   ( xk ), k  0,1, 2,... (2.26)

Let  be the exact root. That is,

   ( ). (2.27)
We define the error of approximation at the kth iterate as  k  xk   , k  0,1,2,...
Subtracting (2.27) from (2.26), we obtain
xk 1     ( xk )   ( )

 ( xk   ) ' (t k ) (using the mean value theorem) (2.28)

or  k 1   ' (t k ) k , xk  t k   .

Setting k  k  1, we get  k   ' (t k 1 ) k 1 , xk 1  t k 1   .

Hence,  k 1   ' (t k ) ' (t k 1 ) k 1.

Using (2.28) recursively, we get
 k 1   ' (t k ) ' (t k 1 )... ' (t0 ) 0 .
The initial error  0 is known and is a constant. We have

 k 1  ' (tk ) ' (tk 1 ) ...' (t0 )  0 .

Let ' (tk )  c, k  0,1, 2,...

Then,  k 1  ck 1  0 . (2.29)

For convergence, we require that  k 1  0 as k  . This result is possible, if and only

if c  1. Therefore, the iteration method (2.26) converges, if and only if

' (xk )  c  1, k  0,1, 2,...

or

| φ′(x) |≤ c < 1, for all x in the interval (a, b). (2.30)

60
We can test this condition using x0 , the initial approximation, before the computations
are done.
Let us now check whether the methods (2.23), (2.24), (2.25) converge to a root in (0, 1) of
the equation f ( x)  x 3  5 x  1  0.

x3 1 3x 2 3x 2
(i) We have  ( x)  ,  ' ( x)  ,  ' ( x)   1,
5 5 5
for all x in 0 < x <1. Hence, the method converges to a root in (0, 1).
1
5
(ii) We have  ( x)  (5 x  1) 3 ,  ' ( x)  2
, .
3
3(5 x  1)

Now ' ( x)  1, when x is close to 1 and ' ( x)  1, in the other part of the interval.

Convergence is not guaranteed.

5x  1 1
(iii) We have  ( x )  ,  ' ( x)  3 1
,
x 2 2
2 x (5 x  1)

Again ' (x)  1, when x is close to 1 and ' (x)  1, in the other part of the interval.
Convergence is not guaranteed.
Remark: Sometimes, it may not be possible to find a suitable iteration function  (x) by
manipulating the given function f (x) . Then, we may use the following procedure.
Write f ( x)  0 as x  x  f ( x)   ( x) , where  is a constant to be determined.

Let x0 be an initial approximation contained in the interval in which the root lies.
For convergence, we require
' ( x0 )  1  f ' ( x)  1. (2.31)

Simplifying, we find the interval in which  lies. We choose a value for  from
this interval and compute the approximations. A judicious choice of a value in
this interval may give faster convergence.
Example2.13: Find the smallest positive root of the equation x 3  x  10  0 , using the general
iteration method.
Solution: We have f ( x)  x 3  x  10 , f(0) = – 10, f(1) = – 10,
f(2) = 8 – 2 – 10 = – 4, f(3) = 27 – 3 – 10 = 14.

61
Since, f(2) f(3) < 0, the smallest positive root lies in the interval (2 , 3).
1
Write x 3  x  10 , and x  ( x  10) 3   ( x) . We define the iteration method as
1
1
xk 1  ( xk  10) . We obtain  ' ( x) 
3
2 ,
3( x  10) 3

We find ' ( x)  1, for all x in the interval (2, 3). Hence, the iteration converges.

Let x0 = 2.5. We obtain the following results.

x1 = (12.5)1/3 = 2.3208,
x2 = (12.3208)1/3 = 2.3097,

x3 = (12.3097)1/3 = 2.3090,

x4 = (12.3090)1/3 = 2.3089.

Since, | x4 – x3 | = |2.3089 – 2.3090 | = 0.0001,

we take the required root as x ≈ 2.3089.

Example2.14: Find the smallest negative root in magnitude of the equation 3x 4  x 3  12 x  4  0
, using the method of successive approximations.
Solution: We have
f ( x)  3x 4  x 3  12x  4  0 , f(0) = 4, f(– 1) = 3 – 1 – 12 + 4 = – 6.
Since, f(– 1) f(0) < 0, the smallest negative root in magnitude lies in the interval (– 1, 0).
Write the given equation as
4
x(3x 3  x 2  12)  4  0 , and x     ( x).
(3 x  x 2  12)
3

The iteration method is written as

4
xk 1   3 2
(3xk  xk  12)

4(9 x 2  2 x)
We obtain  ' ( x) 
(3x 3  x 2  12) 2

We find ' ( x)  1, for all x in the interval (– 1, 0). Hence, the iteration converges.

Let x0 = – 0.25. We obtain the following results.

62
 x1 = − =−
( . ) ( . )
= −0.33290,

x2 = − =− ( . ) ( . )
= −0.33333,

x3 = − =−
( . ) ( . )
= −0.33333.

The required approximation to the root is x ≈ – 0.33333.

Example 2.15: The equation x 2  ax  b  0 , has two real roots α and β. Show that the iteration
method
 (axk  b)
(a) xk 1  , is convergent near x = α, if | α | > | β |,
xk
b
(b) xk 1  , is convergent near x = α, if | α | < | β |.
xk  a
Solution: We have α + β = – a, α β = b.
 (ax  b)  b  
(a)  ( x)  ,  '' ( x)   2 For convergence to α ,
x x2 x
 ( )
we have  '' ( )  <1 , if | α | > | β |.
2
b b 
(b)  (x)  ,  '' ( x )  2
 For convergence to α,
xa ( x  a) (x     )2

( )
we have  '' ( )  <1 , if | α | < | β |.
2

Example 2.16: The equation f ( x)  3x 3  4 x 2  4 x  1  0 has a root in the interval (– 1, 0).

Determine an iteration function  (x) , such that the sequence of iterations
obtained from
xk 1   ( xk ) ,

x0  0.5 , k = 0, 1,..., converges to the root.

Solution: We illustrate the method given in 2.31. We write the given equation as
x  x   (3x 3  4 x 2  4 x  1)   ( x)
where α is a constant to be determined such that
 ' ( x)  1  f ' ( x)

63
 = 1   (9 x 2  8 x  4) < 1

for all x ∈ (– 1, 0). This condition is also to be satisfied at the initial approximation.

Setting x0  0.5 , we get  ' ( x0 )  1  f ' ( x0 ) = | 1 + |< 1

or −1 < 1 + <1 or - < α < 0,

Hence,  takes negative values. The interval for α depends on the initial approximation x0 .

Let us choose the value   0.5 . We obtain the iteration method as

3 2
xk 1  xk  o.5(3xk  4xk  4 xk  1)
3 2
 o.5(3xk  4xk  2xk  1) =  ( xk )
Starting with x0  0.5 , we obtain the following results.
3 2
x1   ( x0 )  o.5(3x0  4x0  2x0  1)
= – 0.5 [3(– 0.5)3 + 4(– 0.5)2 + 2(– 0.5) + 1] = – 0.3125.
3 2
x2   ( x1 )  o.5(3x1  4 x1  2 x1  1)
= – 0.5[3(– 0.3125)3 + 4(– 0.3125)2 + 2(– 0.3125) + 1] = – 0.337036.
3 2
x3   ( x2 )  o.5(3x2  4 x2  2 x2  1)
= – 0.5[3 (– 0.337036)3 + 4(– 0.337036)2 + 2(– 0.337036) + 1] = – 0.332723.
3 2
x4   ( x3 )  o.5(3x3  4x3  2x3  1)
= – 0.5[3(– 0.332723)3 + 4(– 0.332723)2 + 2(– 0.332723) + 1] = – 0.333435.
3 2
x5   ( x4 )  o.5(3x4  4 x4  2 x4  1)
= – 0.5[3(– 0.333435)3 + 4(– 0.333435)2 + 2(– 0.333435) + 1] = – 0.333316.
Since | x5  x4 | = | – 0.333316 + 0.333435 | = 0.000119 < 0.0005, the result is correct to
three decimal places.
We can take the approximation as x  x5  – 0.333316. The exact root is x  – 1/3.

We can verify that  ' ( x j )  1, for all j.

64
Activity2.7

1.Define the fixed point iteration method to obtain a root of f ( x)  0. When does the method
Converge?
2 .In the following problems, find the smallest positive root as specified using fixed point
iteration method.
i) x 2  5 x  1  0 , correct to four decimal places.

ii ) x 5  64x  30  0 , correct to four decimal places.

iii) x  e  x , correct to two decimal places.

3. Find the smallest negative root in magnitude of 3x 3  x  1  0 , correct to four decimal

places. Use fixed point iteration method.

2.3 Convergence of the Iteration Methods:

We now study the rate at which the iteration methods converge to the exact root, if the initial
approximation is sufficiently close to the desired root.
Define the error of approximation at the kth iterate as  k  xk   , k  0,1,2,...
Definition: An iterative method is said to be of order p or has the rate of convergence p , if p is
the largest positive real number for which there exists a finite constant c  0 , such that
p
 k 1  c  k . (2.32)

The constant c , which is independent of k , is called the asymptotic error constant and it
depends on the derivatives of f (x ) at x   .
Let us now obtain the orders of the methods that were derived earlier.
Method of false position:
We have noted earlier that if the root lies initially in the interval ( x0 , x1 ) , then one of the end

points is fixed for all iterations. If the left end point x0 is fixed and the right end point moves
towards the required root, the method behaves like

x0 f k  x k f 0
xk 1 
fk  f0

Substituting xk   k   , xk 1   k 1   , x0   0   , we expand each term in Taylor’s

series and simplify using the fact that f ( )  0 . We obtain the error equation as

65
 f ' ' ( )
 k 1  c 0 k , where c 
f ' ( )
Since  0 is finite and fixed, the error equation becomes

 k 1  c   k , where c   c 0 . (2.33)

Hence, the method of false position has order 1 or has linear rate of convergence.
Method of successive approximations or fixed point iteration method
We have xk 1   ( xk ), and    ( ),

Subtracting, we get xk 1     ( xk )   ( )   (  xk   )   ( )

 [ ( )  ( xk   ) ' ( )  ...]   ( )

or  k 1   k  ' ( )  0( k2 ).

Therefore,  k 1  c  k , xk  t k   , and c  ' ( ) . (2.34)

Hence, the fixed point iteration method has order 1 or has linear rate of convergence.

Newton-Raphson method:
f ( xk )
The method is given by xk 1  xk  , f ' ( xk )  0
f ' ( xk )

Substituting xk   k   , xk 1   k 1   , we obtain

f ( k   )
 k 1     k   
f ' ( k   )

Expand the terms in Taylor’s series. Using the fact that f ( )  0 , and canceling f '' ( ) , we
1
[ k f ' ( )   k2 f '' ( )  ...]
obtain  k 1   k  2
f ( )   k f '' ( )
'

f '' ( ) 2 f '' ( )
 k 1   k  [ k  '  k  ...][1  '  k  ...]1
2 f ( ) f ( )
f '' ( ) 2 f '' ( )
 k 1   k  [ k  '  k  ...][1  '  k  ...]
2 f ( ) f ( )

f '' ( ) 2 f '' ( ) 2
 k 1   k  [ k   k  ...]   k  ...
2 f ' ( ) f ' ( )

Neglecting the terms containing  k3 and higher powers of  k , we get

66
 f ' ' ( )
 k 1  c k2 , where c  ,
f ' ( )
2
and  k 1  c  k . (2.35)

Therefore, Newton’s method is of order 2 or has quadratic rate of convergence.

Remark: What is the importance of defining the order or rate of convergence of a method?
Suppose that we are using Newton’s method for computing a root of f ( x )  0 . Let us
assume that at a particular stage of iteration, the error in magnitude in computing the root
is 10–1 =0.1. We observe from (2.35) , that in the next iteration, the error behaves like
2 –2
c (0.1) = c (10 ).That is, we may possibly get an accuracy of two decimal places.
Because of the quadratic convergence of the method, we may possibly get an accuracy of
four decimal places in the next iteration. However, it also depends on the value of c .
From this discussion, we conclude that both fixed point iteration and regula-falsi methods
converge slowly as they have only linear rate of convergence. Further, Newton’s method
converges at least twice as fast as the fixed point iteration and regula-falsi methods.
Review Exercises
1. Find all intervals in which the roots of the following equations lies
i) x 4  2 x 3  4 x 2  4 x  4  0 ii ) x 3  7 x 2  14 x  6  0
2. Use the Bisection method to find solutions, accurate to within 10−5 for the following problems.
i) x 2  4 x  4  ln x  0 for 1  x  2 and 2  x  4
ii) x  1  2 sinx  0 for 0  x  0.5 and 0.5  x  1
3. a. Sketch the graphs of y  x and y  2 sin x.
b. Use the Bisection method to find an approximation to within 10−5 to the first positive value
of x with x  2 sin x. .
4. Find the iterative methods based on the Newton-Raphson method for finding N , 1 / N, N1 / 3,
where N is a positive real number. Apply the methods to N = 22 to obtain the results
correct to two decimal places.
5. The fourth-degree polynomial f ( x)  230x 4  18x 3  9 x 2  221x  9 has two real zeros, one in
[−1, 0] and the other in [0, 1]. Attempt to approximate these zeros to within 10−6 using
the (i). Method of False Position (ii). Secant method (iii). Newton’s method.

67
Use the endpoints of each interval as the initial approximations in (i) and (ii) and the midpoints
as the initial approximation in (iii).
6. Given the following equations :
i) x 4  x  10  0 ,

ii) cos x  x 2  x  0

iii) e x  2  x  2 cos x  6  0
determine the initial approximations for finding the smallest positive root. Use these to
find the root correct to three decimal places with the following methods:
(i) Secant method, (ii) Regula-Falsi method (iii )Newton-Raphson method.
7. Use a fixed-point iteration method to determine a solution accurate to within 10−2 for
x 4  3x 2  3  0 on [1, 2]. Use x0  1 .

8. (a) Show that the equation log ex  x 2  1 has exactly two real roots, α1 = 0.45 and α2 = 1.

(b) Determine for which initial approximation x0 , the iteration

xn1  1  log exn

converges to α1 or α2

9.The equation x 2  ax  b  0 has two real roots α and β. Show that the iteration method
 ( xk2  b)
xk 1  is convergent near x = α if 2| α | < | α + β |.
a
10. What is the rate of convergence of the following methods:
(i) The Bisection method (ii) Method of false position, (iii) Newton-Raphson method,
(iv) Secant method, (v) Fixed point iteration method?

68
 CHAPTER THREE

3. SYSTEMS OF EQUATIONS

3.1. INTRODUCATION

Systems of equations are used to represent physical problems that involve the
interaction of various properties. The variables in the system represent the
properties being studied, and the equations describe the interaction between the
variables.
The system is easiest to study when the equations are all linear. Often the number
of equations is the same as the number of variables, for only in this case is it likely
that a unique solution will exist.
Although not all physical problems can be reasonably represented using a linear
system with the same number of equations as unknowns, the solutions to many
problems either have this form or can be approximated by such a system. In fact,
this is quite often the only approach that can give quantitative information about a
physical problem.
System of linear equations occur in solving problems in a wide verity of displaces,
including mathematics and statistics ,the physical, biological, and social sciences,
as well as engineering and business. There are two different approaches for finding
the numerical solution of a system of equations, namely

69
Unit Objectives:
At the end of the unit students will be able to:
 use the Exact method to finding roots of system of linear equation,
 use the Gaussian Elimination method to finding solution of system of linear equation,
 use the gaussian elimination method to solve system of linear equation,
 use the the backward or forward substitution method formula to solve a nonlinear equation.
 use the gaussian elimination with partial pivoting method to numerically solve a nonlinear
equation.
 discuss the advantages and drawbacks of LU Matrix Decomposition method
1.2.Exact Method

In this chapter we consider exact methods for approximating the solution of a system of n linear
equations in n unknowns. An exact method is one that gives the exact solution to the system, if it
is assumed that all calculations can be performed without round-off error effects. This
assumption is idealized. We will need to consider quite carefully the role of finite-digit
arithmetic error in the approximation to the solution to the system and how to arrange the
calculations to minimize its effect.

Consider the following general system of n linear equations with n variables.

+ + + ⋯ =
⋮ ⋱ ⋮
+ + + ⋯ =

, , ⋯
⋮ ⋱ ⋮ ⋮ = ⋮
, , ⋯

Which is the form of = . Where

, , ⋯
= ⋮ ⋱ ⋮ , = ⋮ = ⋮
, , ⋯

NB: A linear system of n equations in n variables, with coefficient matrix A and constant
vector ≠ 0, has a unique solution iff the determinant of A ≠ 0.

70
 1.2.1. Gaussian Elimination

If you have studied linear algebra or matrix theory, you probably have been introduced
to Gaussian elimination, the most elementary method for systematically determining the solution
of a system of linear equations. Variables are eliminated from the equations until one equation
involves only one variable, a second equation involves only that variable and one other, a third
has only these two and one additional, and so on. The solution is found by solving for the
variable in the single equation, using this to reduce the second equation to one that now contains
a single variable, and so on, until values for all the variables are found.

Three operations are permitted on a system of equations ( )

Operation on system of Equation

1. Equation can be multiplied by any nonzero constant λ,

with the resulting equation used in place of . This
operation is denoted (λ ) → ( ).
2. Equations and can be transposed in order. This
operation is denoted ( ) ↔ ( ).
3. Equations and can be transposed in order. This
operation is denoted ( ) ↔ ( ).

By a sequence of the operations just given, a linear system can be transformed to a more easily
solved linear system with the same solutions. The sequence of operations is illustrated in the
next example.

EXAMPLE 1 The four equations

1: + + 3 = 4,

2∶ 2 + − + = 1,

71
 3∶ 3 − − + 2 = −3,

4∶ − + 2 + 3 − = 4,

Will be solved for , , , and . First use equation to eliminate the unknown from
, , and by performing ( −2 )→( ),

( −3 )→( ), and

( + )→( ).

The resulting system is

1: + + 3 = 4,

2∶ − − − 5 = −7,

3∶ − 4 − − 7 = −15,

4∶ 3 + 3 + 2 = 8,

Where, for simplicity, the new equations are again labeled , , , and .

In the new system, is used to eliminate from and by

( −4 )→( ) and ( +3 )→( ), resulting in

E1: + + 3 = 4,

E2: − − − 5 = −7,

E3: 3 + 13 = 13,

E4: − 13 = −13.

The system of equations is now in triangular (or reduced) form and can be solved for the
unknowns by a backward-substitution process. Noting that E4 implies

= 1, we can solve for :

72
 1 1
= (13 − 13 ) = (13 − 13) = 0
3 3

Continuing, gives

= − (−7 + 5 + ) = −(−7 + 5 + 0) = 2,

and gives

=4−3 − = 4 − 3 − 2 = −1.

The solution is, therefore, = −1, = 2, = 0, and = 1. It is easy to verify that these
values solve the original system of equations.

When performing the calculations of Example 1, we did not need to write out the full equations
at each step or to carry the variables , , , and through the calculations, since they
always remained in the same column. The only variation from system to system occurred in the
coefficients of the unknowns and in the values on the right side of the equations. For this reason,
a linear system is often replaced by a matrix, a rectangular array of elements in which not only is
the value of an element important, but also its position in the array. The matrix contains all the
information about the system that is necessary to determine its solution in a compact form. The
notation for an n ×m (n by m) matrix will be a capital letter, such as A, for the matrix and
lowercase letters with double subscripts, such as aij, to refer to the entry at the intersection of the
ith row and jth column; that is

⋯
A= = ⋮ ⋱ ⋮
⋯

EXAMPLE 2 The matrix

2 −1 7
A=
3 1 0

is a 2 ×3 matrix with a11 = 2, a12 = −1, a13 = 7, a21 = 3, a22 = 1, and a23 = 0.

The 1 ×n matrix A = [a11 a12 · · ·a1n] is called an n-dimensional row vector, and an n × 1
matrix

73
 A= ..
.

is called an n-dimensional column vector. Usually the unnecessary subscript is omitted for
vectors and a boldface lowercase letter is used for notation. So

X= ..
.

A system of n linear equations in the n unknowns x1, x2 , . . . , xn has the form

+ +··· + =

+ +··· + =

⋮ ⋱ ⋮ = ⋮

+ + ⋯+ =

An n × (n+1) matrix can be used to represent this linear system by first constructing

⋯
= = ⋮ ⋱ ⋮ and = ..
.
⋯

and then combining these matrices to form the augmented matrix:

⋯
⋮ ⋱ ⋮ ⋮ .
⋯

Repeating the operations involved in Example 1 with the matrix notation results in first
considering the augmented matrix:
+ + 3 = 4,

2 + − + = 1,

3 − − + 2 = −3,

74
 − + 2 + 3 − = 4,

1 1 0 3 4
2 1 −1 1 1
3 −1 −1 2 −3
−1 2 3 − 1 4

Performing the operations as described in that example produces the matrices

1 1 0 3 1 1 0 3
0 −1 −1 −5 0 −1 −1 −5
and
0 −4 −1 −7 0 0 3 13
0 3 3 2 0 0 0 − 13

1.2.2. Gaussian Elimination Method

The latter matrix can now be transformed into its corresponding linear system and solutions for
, , , and obtained. The procedure involved in this process is called Gaussian
Elimination with Backward Substitution.
The general Gaussian elimination procedure applied to the linear system
: + +··· + =
: + +··· + =

⋮ ⋱ ⋮ = ⋮

: + + ⋯+ =

is handled in a similar manner. First form the augmented matrix A:

⋯
= ⋮ ⋱ ⋮
⋯

Suppose that a11 ≠ 0. To convert the entries in the first column, below a11,

to zero, we perform the operations ( − ) → ( ) for each k = 2,3 . . . , n

for an appropriate multiplier mk1. We first designate the diagonal element in the column, a11 as
the pivot element. The multiplier for the kth row is defined by = / .
Performing the operations ( − ) → ( ) for each

75
k = 2,3, . . . , n eliminates (that is, change to zero) the coefficient of in each of these rows:

…
…
E2 − m21E1 → E2 , E3 − m31E1 → E3, …,
⋮ ⋱ ⋮ ⋮
…
En − mn1E1 → En
…
0 …
⋮ ⋱ ⋮ ⋮
0 …

Although the entries in rows 2, 3, . . . , n are expected to change, for ease of notation, we again
denote the entry in the ith row and the jth column by aij.

If the pivot element a22 ≠ 0, we form the multipliers mk2 = ak2/a22 and perform the operations
( − )→ for each k = 3. . . n obtaining

…
0 …
E3 − m32E2 → E3… En − mn2E2 → En
⋮ ⋱ ⋮ ⋮
0 …

…
0 …
⋮ ⋱ ⋮ ⋮
0 0 0 …
Since the new linear system is triangular, backward substitution can be performed. Solving the
nth equation for xn gives

Solving the (n − 1)st equation for xn−1 and using the known value for xn yields

and Continuing this process

,

Note:
Consider the system =
Let us denote the original system by

76
 ⋯
⋮ ⋱ ⋮ =
⋮ ⋮
⋯

Step 1: Assume ≠ 0. Then define the row multipliers:

Then eliminate from the last (n-1) equations by subtracting the multiple of the first
eqn from the ith eqn.
( ) ( )
As a result, the first row of and are left unchanged & the remaining rows are
changed.
And so, we get a new system, denoted by:
( ) ( )
=
.

⋯ ⎡ ⎤
0 ⋮ ⋱ ⋮ ⋮ =⎢ ⎥
0 ⋯ ⎢⋮⎥
⎣ ⎦
Where the new coefficients are given by
= −
= − i, j= 2, 3, 4, . . ., n
( )
Step 2 If ≠ 0, we can eliminate from the last (n-2) eqns by generating the multiplier:

= , i = 3, 4, 5,…

Thus we get a new system denoted by

( ) ( )
=
The coefficients are obtained by using:
= −
= − i, j= 3, 4, . . ., n
We continue to eliminate the unknowns, going onto columns, 3, 4, & so on & this is
expressed generally as:
Step K

77
 ( ) ( ) ( )
Let 1 ≤ ≤ − 1. assume that = eliminated at successive stage, & has
the form
…
⎡ ⎤
⎢ 0 … ⎥
⎢ ⋮ ⎥
⎢⋮ 0 0 0 … ⎥
⎢ ⎥
⎢ ⋮ ⎥
⎣ 0 0 0 … ⎦
Assume that multipliers

= , = + 1, + 2, …

And use those to remove from + 1 through n.

= −
= − i, j = k+1, k+2, k+ 3, ... n
The earlier rows 1 through k are left undistributed and zeros are introduced into column k
below the diagonal element.
By continuing this manner, after n-1 steps we obtain

⋯ ⎡ ⎤
0 ⋮ ⋱ ⋮ ⋮ =⎢ ⎥
0 0 0 0 ⋯ ⎢⋮⎥
⎣ ⎦
Then, using backward substitution formula
∑
= = , − 1, − 2, … , 1

We solve for x’s

Example 3: Solve the linear system using Gaussian elimination
+2 + = 0
2 +2 +3 = 3
− −3 = 2
Solution
The augmented matrix
1 2 1 ⋮ 0
[ ⋮ ]= 2 2 3 ⋮ 3
−1 − 3 0 ⋮ 2
Step 1 ≠ , and then generate the multipliers

78
 = = = 2 and

= = = −1

= − , = − , = 2,3
= −2 = 2 − 2(2) = −2
= − −1 = 3 − 2(1) = 1
= − (−1) = 0 + (1)(1) = 1
= − , = 2,3
= − = 3 − 2(0) = 3
= − = 2 − (−1)(0) = 2
1 2 1 ⋮ 0
⟹ 0 −2 1 ⋮ 3
0 −1 1 ⋮ 2
Step 2 ≠ , and then generate the multipliers

= = =

1 2 1 ⋮ 0
⟹ 0 −2 1 ⋮ 3 ⟹ = = 1, = −1, =1
0 0 ⋮

79
 Quick Exercise for student

1. Solve the linear system using Gaussian elimination

x + 2x + x = 3
2x + 3x + 3x = 10
3x − x + 2x = 13
2. Solve the linear system using Gaussian elimination
2.11 − 4.21 + 0.92 = 2.01
4.01 + 10.2 − 1.12 = −3.09
1.09 + 0.987 + 0.832 = 4.21

1.2.3. The backward or forward substitution method formula

System AX=B, where the matrix A is upper triangular has the form
+ +··· + =

+ +··· + =

⋮ ⋱ ⋮ = ⋮

+ ⋯+ =

⋮ ⋱ ⋮ = ⋮

=
≠0
= 1, 2, … ℎ ℎ ℎ
, , ,… , ,

80
from

−
=
, ,

⋮
The backward substitution formula (BSF)

1.2.4. Gaussian Elimination with partial pivoting

Hey, not bad. Now you’re making progress! The general rule to follow is: at each
elimination stage, arrange the rows of the augmented matrix so that the new pivot elt is
larger in absolute value than elt beneath it in its column.
Example
Solve the following system using Gaussian elimination with partial pivoting.
+ −2 = 3
4 −2 + = 5
3 − +3 = 8
Solution
Consider the augmented matrix
[1] 1 −2 ⋮ 3
4 −2 1 ⋮ 5
3 −1 3 ⋮ 8
→ Interchanging the 2nd and the 1st rows because ≤ to get a multiplier of
small magnitude
4 −2 1 ⋮ 5
1 1 −2 ⋮ 3
1 8 1 ⋮ 8
→ Eliminate from the 2nd and 3erd rows to get
Apply the row operations = − (0.5) = −
(0.75) for 1 ≤ ≤ 4 to the above augmented matrix. And hence the modified
augment matrix is given by

81
 4 −2 1 ⋮ 5
0 1.5 − 2.25 ⋮ 1.75
0 0.5 2.25 ⋮ 4.25

Apply the row operations

= − (0.3333) 1 ≤ ≤ 4 to the above augmented matrix .
Hence we have
4 −2 1 ⋮ 5
0 1.5 − 2.25 ⋮ 1.75
0 0 2.9999 ⋮ 3.6667
Thus, by back-substitution = 1.2223, = 3.0001 = 2.4445. Hence the required
numerical solution is given by = (1.222, 3, 2.445) .
Group Work activity
Solve the following system using Gaussian elimination with partial pivoting

+ − =
+ − =
− + =
1.2.5. Gauss-Jordan Method

The Gauss-Jordan method consists of transforming the linear system AX = B into

equivalent system
IX = B , where I is the identity matrix of order n, so that X = B
Description of the method
Consider
⋯ ,
… ,
⋮ = ⋮
⋮
… ,

Step1
Assume that ≠0
=1 , . .

= , = 1, 2, 3, … + 1

Now, make the non-diagonal elements of the first column zero by applying
= − , = 1, 2, 3, … = 1, 2, 3, … +1

82
 Then the new system is:
⋯ ,
0 … ,
⋮ = ⋮
⋮
0 … ,

Step 2
Assume that ≠0
=1 , . .

= , = 2, 3, … + 1

Now, make the non-diagonal elts of the 2nd column zero by applying
= − , where
= 1, 2, 3, …
= 1, 2, 3, … +1
Continue the process until the system takes the form:
IX = B
0 0 0 ,
0 0 …0 ,
= ⋮
⋮ ⋮
0 0 0… ,

This implies that

= , , = , , = , , … = ,

Example 4: Solve the following system using Jordan Method

3 + +6 = 2
2 + +3 = 7
+ + = 4
Solution
3 1 6 ⋮ 2
2 1 3 ⋮ 7
1 1 1 ⋮ 4
→ = , = 1, 2, 3, 4

83
 1 1 2 ⋮ 23
3
2 1 3 ⋮ 7
1 1 1 ⋮ 4
→ → −2 , → − ,

1 1 ⋮ 23
3 2
0 1 −3 ⋮ 17
0 2 −1 ⋮ 10
3 3
→ → −1 3 , → −2 3 ,

≠ 0, and so make = 1, by → = , = 2, 3, 4 … + 1

1 0 3 ⋮ −5
0 1 − 3 ⋮ 17
0 0 1 ⋮ −8
→ Make all the non- diagonal entries of column zero
→ → −3 , → +3 ,
1 0 0 ⋮ 19
0 1 0 ⋮ −7
0 0 1 ⋮ −8
Therefore = 19, = −7, and = −8

Example 2: Use the Gauss-Jordan elimination method to solve the linear system

 1 2 3  x1   3 
    
  3 1 5  x 2     2 
 2 4  1  x    1 
  3   

First form the augmented matrix M = [A , B]

 1 2 3 3 
 
M    3 1 5  2
 2 4 1 1
 

84
Then perform Gauss-Jordan elimination.

 1 2 3 3 
 
M    3 1 5  2
 2 4 1 1
 
1 2 3 3 
 
 0 7 14 7 
0 0  7  7
 
1 0 1 1 
 
0 1 2 1 
0 0  7  7
 

1 0 0 2
 
0 1 0 1
0 0 1 1
 

 x1   2 
   
Hence, the solution is  X    x2     1
 x  1 
 3  

Exercise ( )

1. Solve the following system using Jordan Method

+2 + = 3
2 +3 +3 = 10
3 − +2 = 13

2. Solve the following system using Jordan Method

−2 −6 = 14
9 +4 + = −17
+6 = 4

85
 Remark: Gauss- Jordan method is more expensive than Gaussian elimination for
solving system of linear equation.
It involves approximately 50% more arithmetic operation than Gaussian elimination does
1.2.6. Matrix Inversion Using Jordan Elimination

Let A be an matrix such that ≠ .

To find the inverse of A use the ff transformation

( : ) to ( : ) where B is the inverse of A.

Example: Find the inverse of A by using Jordan Elimination

1 −1 −2
= 2 −3 −5
−1 3 5

Solution:

1 −1 −2 ⋮ 1 0 0
2 −3 −5 ⋮ 0 1 0
−1 3 5 ⋮ 0 0 1

→ =1

→ −2

→ +

1 −1 −2 ⋮ 1 0 0
⟹ 0 − 1 − 1 ⋮ −2 1 0
0 2 3 ⋮ 1 0 1

→ = −1 so make = 1 by

→−

86
 1 −1 −2 ⋮ 1 0 0
⟹ 0 1 1 ⋮ 2 −1 0
0 2 3 ⋮ 1 0 1

→ +

→ −2

1 0 −1 ⋮ 3 −1 0
⟹ 0 1 1 ⋮ 2 −1 0
0 0 1 ⋮ −3 2 1

→ =1

→ +

→ −

1 0 0 ⋮ 0 1 1
⟹ 0 1 0 ⋮ 5 −3 −1
0 0 1 ⋮ −3 2 1

0 1 1
Therefore = 5 −3 −1
−3 2 1

1.2.7. LU Matrix Decomposition

Procedure:

Let = be a given system of linear equation

Step 1: write = as

LUX=B

⇔ L(UX)=B

Step 2: set Y=UX so that LY=B

Step 3: solve the equation LY=B for Y.

Step 4: solve Y=UX for X. then X is the solution of the system.

87
Remark: make sure A is non-Singular matrix.

Example Solve the following system using LU decomposition.

− +2 + = 0
8 +6 = 10
−2 + 5 = −11
Solution:

−1 2 1
= 0 8 6
−2 0 5

→Check:

Det A = −1(40 − 0) − (0 + 12) + 1(0 + 16) = −48 ≠ 0 so A is non-singular.

Now factorize A as

A=LU

−1 2 1 0 0 1
= 0 8 6 = 0 0 1
−2 0 5 0 0 1

−1 2 1
= 0 8 6 = + +
−2 0 5 + +

⟹ = −1, = 0, = −2

= 2, ⟹ = −2

Similarly, = −1, = 8, =3 2 , = −4, =6

−1 2 1 −1 0 0 1
= 0 8 6 = 0 0 1
−2 0 5 0 0 1

88
 −1 2 1 −1 0 0 1 −2 −1
⟹ = = 0 8 6 = 0 8 0 0 1 32
−2 0 5 −2 −4 6 0 0 1

Now = ⇔ =

Set = , where Y is a column vector

⟹ =

−1 0 0 0
0 8 0 = 10
−2 −4 6 −11

0
⇔ 8 = 10
−2 −4 +6 −11

⟹ = 0, 8 = 10 ⟹ =5 4 & = −1

0
= = 5
4
−1

Finally solve for X the equation UX = Y

1 −2 −1 0
⟹ 0 1 3 = 5
2 4
0 0 1 −1

⟹ = −1, =2 =3

Hence

3
= 2
−1

89
 Exercise

Solve the following by LU decomposition

+ + = 0
1. 2 − + = 6
3 +2 −4 = −4

2 +6 +4 = 5
2. 6 + 19 + 12 = 6
2 + 8 + 14 = 7

2 + +3 = −1
3. 4 + +7 = 5
−6 − 22 − 12 = −2

1. Tri-Diagonal Matrix

Definition: Let A = ( ) be a square matrix of order n such that = 0 whenever| − | > 1,

then A is called a tri diagonal matrix. That is

∝ 0 0 … 0
⎛ … 0⎞
=⎜0 ∝ … ⋮ ⎟
⋮ ⋱
⎝0 0 0 0 … ⎠

Exercise: Solve using LU decomposition method

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

Hint: A=LU for tri diagonal matrix has the form:

90
 2 −1 0 0 0 0 0 1 0 0
–1 2 −1 0 0 0 0 1 0
=
0 −1 2 −1 0 0 0 0 1
0 0 −1 2 0 0 0 0 0

1.3.Indirect/Iterative/ methods of solving systems of linear equation

The two common iterative methods are:

i. The Gauss-Jacobi Method &

ii. The Gauss-Seidel

1.3.1. Gauss-Seidel Method

Iteration is a popular technique finding roots of equations. Generalization of fixed point iteration
can be applied to systems of linear equations to produce accurate results. The Gauss-Seidel
method is the most common iterative method and is attributed to Philipp Ludwig von Seidel
(1821-1896).
Consider that the n×n square matrix A is split into three parts, the main diagonal D, below
diagonal L and above diagonal U. We have A = D - L - U.

 a1,1 a1,2 a1,3 a1,nt a1,n1 a1,n 

 
 a2,1 a2,2 a2,3 a2,n2 a2,n1 a2,n 
a a3,2 a3,3 a3,n2 a3,n1 a3,n 
3,1
 
 . 
      
A 
 . 
 
 an2,1 an2,2 an2,3 an2,n2 an2,n1 a n2,n 
a an1,2 an1,3 an1,n2 an1,n1 an1,n 
n1,1
 
a a a  a a a 
 n,1 n, 2 n,3 n,n2 n,n1 n,n 

91
 =

D=

U=

92
L=

The solution to the linear system AX=B can be obtained starting with P0, and using iteration
scheme
= +

where
=( − )

and

=( − )

A sufficient condition for the method to be applicable is that A is strictly diagonally dominant
dominant.

Example1 Use Gauss-Seidel

Seidel iteration to solve the linear system

Try 10 iterations.

93
The system can be expressed as

Using 10 iterations we have:

94
Hence,

For the purpose of hand calculation let’s see 3 set of linear equations containing 3 unknowns.

a11x1  a12x2  a13x3  b1

a21x1  a22x2  a23x3  b2
a31x1  a32x2 a33x3  b3

x
If the diagonal elements are all nonzero, the first equation can be solved for 1, the second for

x2 and the third for x 3 to yield:

(a) = =

(b) =

(c) =

Steps to be followed

i. Using the initial guess x2 = x3 = 0.0 solve for x1 from (a)

ii. Using the values of x1 from step i and x3 = 0.0 solve for x 2 from (b)

iii. Using the value of x1 from step i and that of x 2 from step ii solve for x3 from (c)

95
iv. Using the value of x 2 from step ii and that of x3 from step iii solve for x1 from (a)

v. Using the value of x1 from step iv and that of x3 from step iii solve for x 2 from (b)

vi. Using the value of x1 from step iv and that of x 2 from step v solve for x3 from(c)

vii. Repeat the process until the required accuracy is achieved.

Example2 Use the Gauss-Seidel method to obtain the solution of the following system of linear
equations.

5x1  x2  x3  4
x1  3x2  x3  2
 x1  x2  4 x3  3

4  x2  x3
Solving for: x1 from eq1 x1 = 5

2  x1  x 3
x2 from eq2 x2 =
3

3  x1  x2
x3 from eq3 x3 =
4
Executing the above steps repetitively we will have the following result.

96
 x1 x2 x3

0.8 0.4 0.85

0.71 0.146667 0.890833

0.651167 0.152667 0.874625

0.655608 0.156589 0.874755

0.656367 0.156293 0.875019

0.656255 0.156242 0.875003

0.656248 0.15625 0.875

0.65625 0.15625 0.875

0.65625 0.15625 0.875

As we can see the values start to repeat after the 8th iteration hence we can stop the calculation
and take the final values as the solution of the linear system of equations.

Hence, x1 = 0.65625

x2 = 0.15625

x3 = 0.875

97
3. Review Exersise

1. Find the inverse of the following matrices if possible

 1 3 7 5
 1 2 3   1 0 1
1 3    1 2 1 3  
(a )   (b)   3 1 2  (c )  (d )  2 1 2 
 8 4   9 5 6 2 0 1 4 1 2 1
     
 1 1 1 3 


1 1 0 
 
2. Find the values of x for which the matrix A= 1 0  1 is invertible. In that case give A-1.
1 2 x 
 

3. Find the determinants of the matrices in 1above.

 1  2 0  1 2 0
   
4. Given that  2 1 0 A
   2 1 0   5I 3 , what is det(A)?
 0 0 1  0 0 1
   

5. Find the determinant and inverse of the following matrix

1 0 0 0
  1 1 1  x  m x x
0 1 0 0    
(a)  (b)  a b c (c) x xm x 
0 0 1 0 a2

a
  b2 c 2   x
 x x 
 b c d 

6. Solve each of the following system of linear equations by using:

i) Gaussian elimination method

ii) Cramer’s rule, whenever possible

2x  5 y  1
(a )
3x  2 y  4

98
 2x  y  6z  6 3x  4 y  7 z  0
3 x  2 y  z  1
(b) c ) 3 x  2 y  2 z  2 d ) y  2z  3
 x  2 y  9z  9
x  y  2z  4 x  3 y  z  5

x  y  z  w  4
2 y  z  3w  4
e)
2 x  y  z  2w  5
x yw4

7. Find the LU decomposition of the matrix

 25 5 1
A   64 8 1
144 12 1

8. Use the LU decomposition method to solve the following simultaneous linear equations.

 25 5 1  a1  106.8 
 64 8 1 a   177.2 
   2  
144 12 1  a3  279.2

9. A man refused to tell anyone his age, but he likes to drop hints about it. He then remarks that

twice his mother’s age add up to 140 and also that his age plus his father’s age add up to 105

Furthermore, he says that the sum of his age and his mother’s age is 30 more than his father’s

age. Calculate the man’s age or show that his hints contradict one a

99
 UNITE FOUR

2. FINITE DIFFERENCES
1.1. INTRODUCTION

Numerical Analysis is a branch of mathematics which leads to approximate solution by repeated

application of four basic operations of Algebra. The knowledge of finite differences is essential
for the study of Numerical Analysis. In this section we introduce few basic operators.
Unit Objectives:
At the end of the unit students will be able to:
 Define forward difference operator
 use forward difference operator, for interpolation
 identify relation between the operator E and ∆
 use the the backward or forward substitution method formula to solve a nonlinear equation.
 use the gaussian elimination with partial pivoting method to numerically solve a nonlinear
equation.
 discuss the advantages and drawbacks of LU Matrix Decomposition method
1.2.FORWARD DIFFERENCE OPERATOR

Let = ( ) be any function given by the values , , ,…, , which it takes for the
equidistant values , , ,…, , of the independent variable x, then − , −
, − ,…, − , are called the first differences of the function y.
They are denoted by ∆ , ∆ , … etc.
We have ∆ = −
∆ = −
...
∆ = −
The symbol ∆ is called the difference operator. The differences of the first differences denoted
by ∆ ,∆ ,..., ∆ are called second differences, where
∆ = ∆(∆ ) = ∆( − )= ∆ −∆
=( − )−( − )
= − 2 +

100
 ∆ = ∆(∆ ) = − 2 +
∆ is called the second difference operator.
∆ = ∆ − ∆ = − 3 + 3 −
...
∆ = ∆ − ∆
1.2.1. Difference Table
It is a convenient method for displaying the successive differences of a function. The following
table is an example to show how the differences are formed

X Y ∆ ∆ ∆ ∆ ∆ ∆

∆ ∆

∆ ∆

∆ ∆ ∆

∆ ∆ ∆

∆ ∆ ∆

∆ ∆

∆ ∆

101
The above table is called a diagonal difference table. The first term in the table is . It is called
the leading term.

The differences, ∆ , ∆ , ∆ ,., are called the leading differences. The differences ∆ with
a fixed subscript are called forward differences. In forming such a difference table care must be
taken to maintain correct sign.

Example . Construct a forward difference table for the following data

X 0 10 20 30

Y 0 0.174 0.347 0.518

Solution

X y ∆ ∆ ∆

0 0

0.174

10 0.174 −.001

0.173 −.001

20 0.347 −.002

. 171

30 0.518

102
Group work Exercise

1. Construct a difference table for y = f(x) = x3 + 2x + 1 for x = 1, 2, 3, 4, 5

2. By constructing a difference table and taking the second order differences as constant find
the sixth term of the series 8, 12, 19, 29, 42.

1.2.2. The Operator E

Let y = f (x) be function of x and x, x + h, x + 2h, x + 3h, , etc., be the consecutive values of x,
then the operator E is defined as

Ef(x) = f (x + h),

E is called shift operator. It is also called displacement operator.

Note: E is only a symbol but not an algebraic sum.

E2f (x) means the operator E is applied twice on f (x), i.e.,

E2f (x) = E[Ef(x)]

= Ef(x + h)

= f (x + 2h)

Similarly En f (x) = f (x + nh) and

E–n f (x) = f (x – nh).

The operator E has the following properties:

1. E(f1 (x) + f2 (x) + … + fn (x)) = Ef 1(x) + Ef2 (x) + … + Efn (x)

2. E(cf(x)) = cEf(x) (where c is constant)

103
 3. E m (En f (x)) = En (Em f (x)) = En+m f (x) where m, n are positive integers

4. If n is positive integer En[E-n f (x)] = f (x)

Alternative notation: If y0 , y1 , y2 , …, yn , …, etc., are consecutive values of the function y = f

(x) corresponding to equally spaced values x0 , x1 , x2 , …, xn , etc., of x then in alternative
notation

Ey0 = y 1

Ey1 =y2

And in general Eny0 =yn

Theorem: If n is a positive integer then yn = y0 + nc1 ∆ y0 + nc2∆2 y0 + …+ ∆n y0 .

Proof :From the definition

y1 = Ey0 = (1+ ∆)y0 = y0+ ∆y0

y2 = E2y0 = (1 + ∆)2y0 =( 1 + 2c1∆ + ∆2 )y0

= y0 + 2c1∆y0 + ∆2 y0

Similarly we get yn = Eny0 = (1+ ∆n) y0

=( 1+ nc1∆ + ... + ∆n )y0

= y0 + nc1∆y0 + ... + ∆ny0,

Relation between the Operator E and ∆

From the definition of ∆, we know that

∆f x= f(x+h) – f(x),

where h is the interval of differencing. Using the operator E we can write

∆f(x) = Ef(x) − f( x)

104
 ⇒ ∆f(x) = (E – 1)f(x) .

The above relation can be expressed as an identity

∆=E−1

i.e., E = 1+ ∆ .

Example : Show that

E∆ ≡ ∆E

Proof E∆f(x) = E(f( x + h) − f (x))

= Ef(x + h) – Ef(x)

= f(x + 2h) – f(x + h)

= ∆(f(x +h)

= ∆Ef(x)

∴ E∆ ≡ ∆E.

∆f ( x )

∆
Example: Evaluate .

Solution:

Let h be the interval of differencing

∆
= (∆ )

(E − 1 )

=( − 2 + 1)

=( −2+ )x3
105
 = −2 +

= (x + h)3 – 2x3 + (x – h)3

= 6xh.

Example: Given u0 = 3, u1 = 12, u2 = 81, u3 = 200, u4 = 100, and u5 = 8, find ∆5 u0 .

Solution ∆5u0 = (E–1)5 u0

= (E5 – 5E4 + 10E3 – 10E2 + 5E –1)u0

=u5 – 5u4 + 10u3 – 10u2 + 5u –u0

= 8 –500 + 2000 – 810 + 60 –3

= 755.

Group Work Exercise:

Find the first term of the series whose second and subsequent terms are

8, 3, 0, –1, 0, …

1.3.Backward Differences

Let y = f(x) be a function given by the values y0 , y1, … yn which it takes for the equally spaced

Values x0 , x1 , …, xn of the independent variable x. Then y1– y0 , y2– y1, …, yn – yn-1 are called
the first backward differences of y = f (x). They are denoted by ∇y0 , ∇y1 ,..., ∇yn , respectively.
Thus we have

y1− y0 = ∇y1

y2 − y1 = ∇y2

yn − yn − 1 = ∇yn ,

where ∇ is called the backward difference operator.

106
x y

Note: In the above table the differences ∇ny with a fixed subscript i, lie along the diagonal
upward sloping.

Alternative notation: Let the function y = f(x) be given at equal spaces of the independent
variable x at x = a, a + h, a + 2h, … then we define

∇f(x) = F(a) – F(a – h)

Where ∇ is called the backward difference operator, h is called the interval of differencing.

In general we can define

∇f(x) = f(x) – f(x- h)

We observe that ∇f(x + h) = f (x + h) – f(x) = ∆f(x)

∇f(x + 2h) = f (x + 2h) – f(x + h) = ∆f(x + h)

∇f(x + nh) = f (x + nh) – f(x + (n-1) h) = ∆f(x + h)

=∇f(x + (n – 1) h)

Similarly we get

107
 ∇2f(x + 2h) = ∇( ∇f(x + 2h)

= ∇ (∆f(x +h)

= ∆ (∆ f(x)

= ∆ ( ( ))

….

f(x + nh) = ∆ ( ( ))

Relation between E and ∇ :

∇f(x) = f( x) – f(x - h) = f (x ) − E −1 f (x )

⇒ ∇ = 1 − E −1

or ∇= .

Example : Prove the following

(a) (1 + ∆)(1 − ∆)= 1

(b) ∆∇ = ∆ − ∇

(c) ∇ = E −1∆.

Solution

(a) (1+ ∆) (1− ∇) f(x) = E( f( x))

= Ef (x – h) = f(x) = 1.f (x)

(b) ∇∆ f(x) = (E - 1) (1- E −1) f x

= (E – 1)[ f (x) − f(x – h)]

= Ef( x) – f( x) – Ef( x – h) + f (x – h)

= f(x + h) – f(x) – f( x) + f (x – h)

108
 = [(E – 1) – (1− E −1)] f x

= (∆ − ∇) f x

∴ ∆∇f(x) = (∆ − ∇) f(x)

∴ ∆∇ = ∆ − ∇.

(c) ∇ f(x) = (1 - E −1) f(x) = f (x) – f(x – h)

and E −1 ∆f (x) = E −1 [f (x + h) − f( x)

= f (x) − f (x –h) ∇

∴ ∇ = E −1 ∆

1.4.Central Deference

The central deference operator is defined by the relation

− = , − = ,... − =

Similarly, higher-order central differences can be defined. With the values of x and y as in the
preceding two tables a central difference table can be formed.

x y

109
It is clear from the three tables that in definite numerical case the same number occurred in the
same positions whether we use forward, backward or central differences.

Thus we obtain

∆ =∇ =

∆ =∇ =

Exercise

1. Find the forward difference table corresponding to the data points (1, 3), (2, 5), (3, 7), and
(4, 10).
2. Find the forward table corresponding to the data points (1, 3), (2, 5), (3, 8), and (4, 10).
3. Find the backward difference table corresponding to the data points (1, 3), (3, 5), (5, 7),
and (7, 10).
4. Find the backward difference table corresponding to the data points (1, 3), (3, 5), (5, 8),
and (7, 10).

110
 UNIT FIVE

5. INTERPOLATION
5.1. Introduction

In this chapter, we discuss the problem of approximating a given function by polynomials.

There are two main uses of these approximating polynomials. The first use is to reconstruct the
function f (x ) when it is not given explicitly and only values of f (x ) are given at a set of distinct
points called nodes or tabular points. The second use is to perform the required operations which
were intended for f (x ) , like determination of roots, differentiation and integration etc. can be
carried out using the approximating polynomial p (x ) . The approximating polynomial p (x ) can
be used to predict the value of f (x ) at a non tabular point. The deviation of p (x ) from f (x ) ,
that is f (x ) – p (x ) , is called the error of approximation.
Let f (x ) be a continuous function defined on some interval [a,b], and be prescribed at
n  1 distinct tabular points x0 , x , x 2 , . . . , x0 such that a  x0  x1  x 2  ...  xn  b . The distinct
tabular points x0 , x1 , x 2 , . . . , xn may be non-equispaced or equispaced, that is xi  xi 1  h , for

i  1,2,3,..., n. The problem of polynomial approximation is to find a polynomial p n (x ) , of degree

≤ n, which fits the given data exactly, that is,
p n ( xi ) = f n ( xi ) , i  0,1,2,..., n. (5.1)

The polynomial p n (x ) is called the interpolating polynomial. The conditions given in

(5.1) are called the interpolating conditions. The interpolation polynomial fitting a given data is
unique. We may express it in various forms but are otherwise the same polynomial. For
example, f ( x )  x 2  2 x  1 can be written as x 2  2 x  1 =  2  ( x  1)  ( x  1)( x  2).

111
Unit Objectives:
At the end of the unit students will be able to:
 Derive Newton’s method of interpolation,
 Solve problems using Newton’s method of interpolation
 Derive Lagrangian method of interpolation,
 Solve problems using Lagrangian method of interpolation
 Derive Newton’s divided difference method of interpolation
 Apply Newton’s divided difference method of interpolation

5.2. Interpolation with Evenly Spaced Points

5.2.1 Newton’s Forward Difference Interpolation Formula

Given the set of (n  1) values, viz., ( x0 , y 0 ), ( x1 , y1 ), ( x 2 , y 2 ), . . . , ( x n , y n ) of x and y , it is

required to find y n (x) , a polynomial of nth degree such that y and y n (x) agree at the tabulated

points. Let the values of x be equidistant, that is,

xi  xi 1  h , for i  1,2,3,..., n.

Therefore , x1  x0  h , x 2  x0  2 h , etc.

xi  x0  ih , for i  1,2,3,..., n.

Since y n (x) is a polynomial of nth degree, it may be written as

y n ( x )  a0  a1 ( x  x0 )  a 2 ( x  x0 )( x  x1 )  a3 ( x  x0 )( x  x1 )( x  x 2 )  ...

 a n ( x  x0 )( x  x1 )( x  x2 )...( x  xn 1 ). ( 5.2)

The (n  1) unknowns a0 , a1 , a 2 , . . . , an can be found as follows

Put x  x0 in ( 5.2) we obtain ,

y n ( x0 )  y 0  a 0 i.e a0  y0 (since the other terms in ( 5.2) vanish)

112
 again put x  x1 in ( 5.2) we obtain ,

y n ( x1 )  y1  a 0  a1 ( x1  x0 )  y 0  a1 ( x1  x0 ) (since the other terms in ( 5.2) vanish)

y1  y 0  a1 ( x1  x0 ) then solve for a1 we obtain ,

y1  y0 y0 y0
a1   i.e a1 
x1  x0 h h

Similarly put x  xi , for i  2,3,..., n  1 , in ( 5.2) we obtain

2 y0 3 y0 n y0
a2  a  a 
2!h 2 , 3!h 3 , . . .,
3 n
n! h n

now substitute a0 , a1 , a 2 , . . . , an in ( 5.2)

y 2 y0 3 y0 n y0
by a0  y0 , a1  0 , a2  a  a  we obtain ,
2!h 2 , 3!h 3 , . . .,
3 n
h n! h n

y0 2 y0 3 y0
yn ( x)  y0  ( x  x0 )  ( x  x0 )( x  x1 )  ( x  x0 )( x  x1 )( x  x2 ) 
h 2! h 2 3! h3

n y0
...  ( x  x0 )( x  x1 )( x  x2 )...( x  xn1 ). ( 5.3)
n! h n

Put x  x0  ph in ( 5.3), we get

p( p  1) 2 p ( p  1)( p  2) 3
y n ( x)  y0  py0   y0   y0  ...
2! 3!

p( p  1)( p  2)...( p  n  1) n
  y0 (5.4)
n!

This is Newton’s forward difference interpolation formula and useful for interpolating near the
beginning of a set of tabular values.
5.2.2 Newton’s Backward Difference Interpolation Formula

113
Given the set of (n  1) values, viz., ( x0 , y 0 ), ( x1 , y1 ), ( x 2 , y 2 ), . . . , ( x n , y n ) of x and y , it is

required to find y n (x) , a polynomial of nth degree such that y and y n (x) agree at the tabulated

points. Let the values of x be equidistant, that is,

xi  xi 1  h , for i  1,2,3,..., n.

Therefore , x1  x0  h , x 2  x0  2 h , etc.

xi  x0  ih , for i  1,2,3,..., n.

Instead of assuming y n (x) as in ( 5.2), if we choose it in the form

y n ( x )  a 0  a1 ( x  x n )  a 2 ( x  xn )( x  x n 1 )  a3 ( x  x n )( x  xn 1 )( x  x n  2 )  ...

 a n ( x  x n )( x  xn 1 )( x  x n  2 )...( x  x1 ). ( 5.5)

then the (n  1) unknowns a0 , a1 , a 2 , . . . , an can be found as follows

Put x  x n in ( 5.5) we obtain ,

y n ( xn )  y n  a0 i.e a0  y n (since the other terms in ( 5.5) vanish)

again put x  xn 1 in ( 5.5) we obtain ,

y n ( xn1 )  y n 1  a0  a1 ( xn1  xn )  y n  a1 ( xn 1  xn )

y n 1  y n  a1 ( x n 1  x n ) then solve for a1 we obtain ,

yn  yn1 yn y h
a1   i.e a1 
xn  xn1 h h

Similarly put x  xn i , for i  2,3,..., n  1 , in ( 5.5) we obtain

 2 yn 3 yn  n yn
a2  a  a 
2! h 2 , 3!h3 , . . .,
3 n
n!h n

now substitute a0 , a1 , a 2 , . . . , an in ( 5.5)

114
 y h  2 yn 3 yn  n yn
by a0  y n a1  a2  a  a  we obtain ,
, h 2! h 2 ,
3
3!h 3 ,. . .
n
n!h n

yn  2 yn  3 yn
y n ( x)  y n  ( x  xn )  ( x  x n )( x  x n 1 )  ( x  xn )( x  xn1 )( x  xn2 ) 
h 2!h 2 3!h 3

 n yn
...  ( x  xn )( x  xn1 )( x  xn2 )...( x  x1 ). (5.6)
n!h n

Put x  x n  ph in (5.6), we get

p( p  1) 2 p( p  1)( p  2) 3
yn ( x)  y n  py n   yn   yn  ...
2! 3!

p( p  1)( p  2)...( p  n  1) n
  yn . (5.7)
n!

This is Newton’s backward difference interpolation formula and useful for interpolating near the
end of the tabular values.

Example 5.1: Using Newton’s forward difference interpolation formula, find the form of the
function y (x ) from the following table
x 0 1 2 3

f (x ) 1 2 1 10

Solution: We have the following forward difference table for the data.
The forward differences can be written in a tabular form as in Table 5.1.

115
 x f (x )  2 3

0 1

1 2 -2

-1 12

2 1 10

3 10

Table 5.1. forward differences.

Since n  3 ,the cubic Newton’s forward difference interpolation polynomial becomes:
p( p  1) 2 p( p  1)( p  2) 3
y3 ( x)  y0  py0   y0   y0 .
2! 3!
x  x0 x  0
where p  x
h 1
x( x  1) x( x  1)( x  2)
y3 ( x)  1  x(1)  (2)  (12).
2! 3!
y3 ( x)  1  x  ( x 2  x)  2( x)(x  1)(x  2).

y3 ( x)  1  x  ( x 2  x)  2( x)(x 2  3x  2).

y3 ( x)  2 x3  7 x 2  6x  1.

Example 5.2: Find the interpolating polynomial corresponding to the data (1,5),(2,9),(3,14),and
(4,21). Using Newton’s backward difference interpolation polynomial.
Solution: We have the following backward difference table for the data.
The backward differences can be written in a tabular form as in Table 5.2.

116
 x f (x )  2 3

1 5

2 9 1

5 1

3 14 2

4 21

Table 5.2. backward differences.

Since n  3 ,the cubic Newton’s backward difference interpolation polynomial becomes:
p( p  1) 2 p( p  1)( p  2) 3
y 3 ( x )  y n  p y n   yn   yn .
2! 3!
x  x3 x  4
where p   x  4.
h 1
( x  4)( x  4  1) ( x  4)( x  4  1)( x  4  2)
y3 ( x)  21  ( x  4)(7)  (2)  (1).
2 6
( x  4)( x  3) ( x  4)( x  3)( x  2)
y3 ( x)  21  ( x  4)(7)  ( 2)  (1).
2 6
x 3 x 2 26x
y3 ( x)     1.
6 2 6
Example 5.3: The table below gives the value of tan x for 0.10  x  0.30 :
x 0.10 0.15 0.20 0.25 0.30

f (x ) 0.1003 0.1511 0.2027 0.2553 0.3093

Find:
i) tan 0.12 ii ) tan 0.26

117
Solution: We have the following forward difference table for the data.
The forward differences can be written in a tabular form as in Table 5.3.

x f (x )  2 3 4

0.10 0.1003

0.0508

0.15 0.1511 0.0008

0.0516 0.0002

0.20 0.2027 0.0010 0.0002

0.0526 0.0004

0.25 0.2553 0.0014

0.0540

0.30 0.3093

Table 5.3 forward differences.

i) To find tan 0.12 we use Newton’s forward difference interpolation polynomial.

x  x0 0.12  0.10
we have x  0.12 , h  xi 1  xi  0.05 and p    0.4.
h 0.05
Hence formula (5.4) gives
p( p  1) 2 p( p  1)( p  2) 3 p( p  1)( p  2)( p  3) 4
y 4 ( x)  y 0  py0   y0   y0   y0 .
2! 3! 4!

118
 0.4(0.4  1) 0.4(0.4  1)(0.4  2)
y 4 (0.12)  tan 0.12  0.1003  0.4(0.0508)  (0.0008)  (0.0002) 
2 6
0.4(0.4  1)(0.4  2)(0.4  3)
(0.0002).
24
=0.1205.

ii) To find tan 0.26 we use Newton’s back ward difference interpolation polynomial.
x  x4 0.26  0.30
we have x  0.26 , h  xi 1  xi  0.05 and p    0.8.
h 0.05
Hence formula (5.7) gives
p ( p  1) 2 p ( p  1)( p  2) 3 p( p  1)( p  2)( p  3) 4
y 4 ( x)  y 4  py 4   y4   y4   y4 .
2! 3! 4!
 0.8(0.8  1)
y 4 (0.26)  tan 0.26  0.3093  0.8(0.0540)  (0.0014) 
2
 0.8(0.8  1)(0.8  2)  0.8(0.8  1)(0.8  2)(0.8  3)
(0.0004)  (0.0002).
6 24

=0.2662
Example 5.4: Using Newton’s forward difference formula , find the sum
Sn  13  23  33  ...  n3 .
Solution:

We have Sn1  13  23  33  ...  n3  (n 1)3 .

Hence Sn1  Sn  (n  1)3 .

or Sn  (n  1)3 .

It follows that 2 Sn  Sn1  Sn  (n  2)3  (n  1)3  3n2  9n  7.

3 Sn  3(n 1) 2  9(n 1)  7  (3n 2  9n  7)  6n  12.

4 Sn  6(n  1) 12  (6n 12)  6.

Since 5 Sn  6 Sn  ...  0, Sn is a fourth degree polynomial in n.

Further, S1  1, S1  8, 2 S1  19, 3 S 1  18, 4 S1  6.

Hence formula (5.4) gives

119
 (n  1)(n  2) (n  1)(n  2)(n  3)
S n  1  (n  1)(8)  (19)  (18) 
2 6

(n  1)(n  2)(n  3)(n  4)

 (6).
24
1 1 1
 n4  n3  n 2
4 2 4
2
 n(n  1) 
 .
 2 

Activity 5.1:

1. Find f (x ) as a polynomial in x for the following data by Newton’s forward difference

formula

x 3 4 5 6 7 8 9

f (x ) 13 21 31 43 57 73 91

Hence, interpolate at x = 3.5 .

2. Find f (12) by Newton’s backward difference formula, for the data

x 4 5 7 10 11 13

f (x ) 48 100 294 900 1210 2028

3. Given

x 0.20 0.22 0.24 0.26 0.28 0.3

f (x ) 1.6596 1.6698 1.6804 1.6912 1.7024 1.7139

Using Newton’s difference interpolation formula, find f (0.23) and f (0.29) .

120
5.2.3 Gauss interpolating polynomial

Newton- Gregory forward and backward difference interpolation polynomials are suitable for
interpolating near the beginning and end of the data points given respectively. For interpolating
near the middle, we are going to present the method called Gauss interpolating polynomial. Such
formulae involve difference near the horizontal line through x  xi . Let ( xi , f ( xi )) for 0  i  n

be n  1 data points such that x i 1  x i  h. Let x i be any arbitrary data such that n f ( xi ) exist,

and let x  x i  ph .

Gauss’ forward formula

We consider the following deference table in which the central ordinate is taken for convenience
as y0 corresponding to x  x 0 .
The difference used in this formula lie on the line shown in the Table 5.1 The formula is,
therefore , of the form
y p ( x )  y0  G1y0  G2 2 y 1  G3 3 y 1  G4 4 y  2  ... (5.8)

where G1 , G 2 , . . . have to be determined. The y p on the left side can be expressed in terms of

y 0 , y 0 and higher order difference of y 0 , as follows:

Table 5.4. Gauss’ forward formula

121
 x f (x )  2 3 4 5 6

x 3 y 3

y 3

x2 y2 2 y3

y  2 3 y 3

x 1 y 1 2 y  2 4 y3

y 1 3 y  2 5 y 3

x0 y0 2 y 1 4 y  2 6 y3

y 0 3 y 1 5 y  2

x1 y1 2 y0 4 y 1

 y1 3 y0

x2 y2 2 y1

y 2

x3 y3

Clearly, y0  E p y0

122
  (1  ) p y0

p( p  1) 2 p( p  1)( p  2) 3
 y0  py0   y0   y0  ...
2! 3!

Similarly, the right side of (5.8) can be expressed in terms of y 0 , y 0 and higher order

differences.
We have 2 y 1  2 E 1 y0

 2 (1  ) 1 y0

 2 (1    2  3  ...) y0

 2 ( y0  y0  2 y 0  3 y0  ...)

 2 y0  3 y 0  4 y 0  5 y0  ...

3 4 5 6
3 y 1   y0   y0   y0   y0  ...

4 y 2  4 E 2 y0

 4 (1  ) 2 y0

 4 ( y0  2y0  32 y0  43 y0  ...)

 4 y0  25 y0  36 y 0  47 y0  ...

Hence (5.8) gives the identity

p ( p  1) 2 p( p  1)( p  2) 3 p( p  1)( p  2)...( p  3) 4

y0  py0   y0   y0   y0  ...
2! 3! 4!

 y 0  G1y 0  G2 (2 y0  3 y 0  4 y0  5 y 0  ...)  G3 (3 y0  4 y0  5 y0  6 y0  ...) 

G4 (4 y0  25 y0  36 y0  47 y0  ...)  (5.9)

Equating the coefficients of y 0 , 2 y0 , 3 y0 , etc., on both sides of (5.9) we obtain

123
 G1  p, 
p ( p  1) 
G2  , 
2! 
( p  1) p ( p  1)  (5.10)
G3  , 
3! 
( p  1) p( p  1)( p  2) 
G4  ,
4! 

now substitute G1 , G 2 , G3 , . . . , G n in ( 5.8) we get

This is Gauss’ forward formula

Gauss’ backward formula

This formula uses the difference which lie on the line shown in Table 5.5

Table 5.5. Gauss’ backward formula

x f (x )  2 3 4 5 6

x 3 y 3

y 3

x2 y2 2 y3

y  2 3 y 3

x 1 y 1 2 y  2 4 y3

p ( p y
1)1 2 ( p  1) p ( p  1) 3 y3 2 ( p  1) p( p  1)(5 yp3 2) 4
y p ( x)  y0  py0   y 1   y 1   y  2  ...(5.11)
2! 3! 4!
x0 y0 2 y 1 4 y  2 6 y3

124
 y 0 3 y 1 5 y  2

x1 y1 2 y0 4 y 1

 y1 3 y0

x2 y2 2 y1

y 2

x3 y3

Gauss’ backward formula can be therefore be assumed to be of the form

(5.12)
y p ( x )  y 0  G '1y 1  G ' 2 2 y 1  G ' 3 3 y  2  G ' 4 4 y  2  ...

where G '1 , G ' 2 , . . . , have to be determined. Following the same procedure as in Gauss’ forward
formula, we obtain

125
 G '1  p, 

' p ( p  1) 
G2  ,
2! 

' ( p  1) p ( p  1)  (5.13)
G3  , 
3! 
' ( p  2)( p  1) p ( p  1) 
G4  ,...
4! 

now substitute G '1 , G ' 2 , G '3 , . . . , G ' n in ( 5.12) we get

p ( p  1) 2 ( p  1) p ( p  1) 3 ( p  2)( p  1) p ( p  1) 4
y p ( x)  y0  py 1   y 1   y2   y  2  ...(5.14)
2! 3! 4!

This is Gauss’ backward formula

Example 5.5: For the following table ,find the values of e1.17 using the Gauss’ forward
formula

x 1.00 1.05 1.10 1.15 1.20 1.25 1.30

ex 2.7183 2.8577 3.oo42 3.1582 3.3201 3.4903 3.6693

Solution:
Clearly, h  x i 1  x i  0.05 and x  1.17 is near to x3  1.15 . Hence

x  x3 1.17  1.15
p   0.4. Again the corresponding difference table is given by
h 0.05

x ex  2 3 4 5 6

1.00 2.7183

126
 0.1394

1.05 2.8577 0.0071

0.1465 0.0004

1.10 3.oo42 0.0076 0

0.1540 0.0004 0

1.15 3.1582 0.0079 0 0.0001

0.1619 0.0004 0.0001

1.20 3.3201 0.0083 0.0001

0.1702 0.0005

1.25 3.4903 0.0088

0.1790

1.30 3.6693

Table 5.6. Gauss’ forward difference

The corresponding Gauss’ forward interpolation polynomial is given by

p( p  1) 2 ( p  1) p ( p  1) 3 ( p  1) p ( p  1)( p  2) 4
y6 ( x)  y0  py0   y 1   y 1   y 2  ...
2! 3! 4!

0.4(0.4  1) (0.4  1)0.4(0.4  1) (0.4  1)0.4(0.4  1)(0.4  2)

e1.17  3.1582  (o.4)(0.1619)  (0.0079)  (0.0004)  (0)  ...
2! 3! 4!

=3.2221

127
Example 5.6: For the following table ,find the values of e1.17 using the Gauss’ backward
formula

x 1.00 1.05 1.10 1.15 1.20 1.25 1.30

ex 2.7183 2.8577 3.oo42 3.1582 3.3201 3.4903 3.6693

Solution: Clearly, h  x i 1  x i  0.05 and x  1.17 is near to x3  1.15 . Hence

x  x3 1.17  1.15
p   0.4. Again the corresponding difference table is given by
h 0.05
x ex  2 3 4 5 6

1.00 2.7183

0.1394

1.05 2.8577 0.0071

0.1465 0.0004

1.10 3.oo42 0.0076 0

0.1540 0.0004 0

1.15 3.1582 0.0079 0 0.0001

0.1619 0.0004 0.0001

1.20 3.3201 0.0083 0.0001

0.1702 0.0005

1.25 3.4903 0.0088

0.1790

128
 1.30 3.6693

Table 5.7. Gauss’ backward difference

The corresponding Gauss’ backward interpolation polynomial is given by

p ( p  1) 2 ( p  1) p ( p  1) 3 ( p  2)( p  1) p ( p  1) 4
y6 ( x)  y0  py 1   y 1   y 2   y  2  ...
2! 3! 4!

0.4(0.4  1) (0.4  1)0.4(0.4  1) (0.4  2)0.4(0.4  1)(0.4  1)

e1.17  3.1582  (o.4)(0.1540)  (0.0079)  (0.0004)  (0)  ...
2! 3! 4!
=3.2221

129
Activity 5.2:

1. For the following table ,find the values of f (5.5) using the Gauss’ forward

formula

x 3 4 5 6 7 8 9

f (x ) 13 21 31 43 57 73 91

2. For the following table ,find the values of f (8.5) using the Gauss’ backward

formula

x 4 5 7 10 11 13

f (x ) 48 100 294 900 1210 2028

3. Given

x 0.20 0.22 0.24 0.26 0.28 0.3

f (x ) 1.6596 1.6698 1.6804 1.6912 1.7024 1.7139

Using Gauss’ difference interpolation formula, find f (0.25) .

130
5.3 Interpolation with Unevenly Spaced Points

5.3.1 Lagrange Interpolation

Let the data

x x0 x1 x2 … xn

f (x ) f ( x0 ) f ( x1 ) f ( x2 ) … f ( xn )

be given at distinct unevenly spaced points or non-uniform points x 0 , x1 , x 2 , . . . , x n . This data

may also be given at evenly spaced points. For this data, we can fit a unique polynomial of
degree ≤ n. Since the interpolating Polynomial must use all the ordinates f ( x 0 ), f ( x1 ),... , f ( x n ).
it can be written as a linear combination of these ordinates. That is,
we can write the polynomial as
pn (x ) =  0 ( x ) f ( x0 ) +  1 ( x ) f ( x1 ) + ... +  n ( x ) f ( x n )

=  0 ( x ) f 0 +  1 ( x ) f 1 + ... +  n (x ) f n
n
pn ( x)    i ( x) f i . (5.15)
i 0

where f ( x i ) = f i and  i (x ) ), i  0,1,2,..., n. are polynomials of degree n. This polynomial fits the
data given in (5.1) exactly.
At x = x 0 , we get

f ( x 0 )  p n ( x 0 ) =  0 ( x 0 ) f ( x0 ) +  1 ( x 0 ) f ( x1 ) + ... +  n ( x 0 ) f ( x n )

This equation is satisfied only when  0 ( x 0 )  1 and  i ( x 0 )  0 , i  0 .

At a general point x = xi , we get

f ( x i )  p n ( xi ) =  0 ( x i ) f ( x0 ) +  1 ( x i ) f ( x1 ) + ... +  i ( x i ) f ( x i ) + ...  n ( x i ) f ( x n )

This equation is satisfied only when  i ( x i )  1 and  j ( xi )  0 , i  j .

131
Therefore,  i (x ) , which are polynomials of degree n, satisfy the conditions
0 ≠
 i (x j ) = (5.16)
1 =
Since,  i (x ) = 0 at x  x0 , x1 , x 2 , . . . , x i 1 , x i 1 , . . . , x n , we know that

( x  x 0 ) , ( x  x1 ) , ( x  x 2 ) , . . . , ( x  x i 1 ) , ( x  xi 1 ) , . . . , ( x  x n ).

are factors of  i ( x ). The product of these factors is a polynomial of degree n.

Therefore, we can write
 i (x )  c ( x  x 0 ) ( x  x1 ) ( x  x 2 )... ( x  x i 1 ) ( x  x i 1 )... ( x  x n ).

where c is a constant.
Now, since  i ( xi ) = 1, we get

 i ( xi ) = 1 = c ( xi  x0 ) ( x i  x1 ) ( xi  x 2 )... ( x i  x i 1 ) ( xi  xi 1 )... ( xi  x n ).

1
Hence, C
( xi  x0 )( xi  x1 )...( xi  xi 1 )( xi  xi 1 )...( xi  xn )
( x  x0 )( x  x1 )...( x  xi 1 )( x  xi 1 )...( x  xn )
Therefore,  i ( x) 
( xi  x0 )( xi  x1 )...( xi  xi 1 )( xi  xi 1 )...( xi  xn ) (5.17)

Note that the denominator on the right hand side of  i (x ) is obtained by setting x = x i in

the numerator. The polynomial given in (5.15) where  i (x ) are defined by (5.17) , i.e,
n
pn ( x)    i ( x) f i (5.18)
i 0

is called the Lagrange interpolating polynomial and  i (x ) are called the Lagrange fundamental
polynomials.
We can write the Lagrange fundamental polynomials  i (x ) in a simple notation.

Denote w(x)  ( x  x 0 ) ( x  x1 ) ( x  x 2 )... ( x  x n ).

which is the product of all factors. Differentiating w(x) with respect to x and substituting x = x i
we get
w' ( xi )  ( x i  x 0 ) ( x i  x1 ) ( xi  x 2 )... ( x i  x i 1 ) ( xi  xi 1 )... ( xi  x n ).

since all other terms vanish. Therefore, we can also write  i (x ) as

132
 w( x )
 i ( x) 
( x  xi ) w' ( xi )
So that (5.18) becomes
n
w( x)
pn ( x)   fi . (5.19)
i 0 ( x  xi ) w' ( xi )

Let us derive the linear and quadratic interpolating polynomials.

LINEAR INTERPOLATION
For n = 1, we have the data

x f (x )

x0 f ( x0 )

x1 f ( x1 )

The Lagrange fundamental polynomials are given by

( x  x1 ) ( x  x0 )
 0 ( x)  ,  1 ( x) 
( x0  x1 ) ( x1  x0 )
The Lagrange linear interpolation polynomial is given by
1
p1 ( x )    i ( x ) f i
i 0

p1 ( x ) =  0 ( x ) f ( x0 ) +  1 ( x ) f ( x1 ) .

( x  x1 ) ( x  x0 )
= f ( x0 ) + f ( x1 ) .
( x0  x1 ) ( x1  x0 )

QUADRATIC INTERPOLATION

133
 For n = 2, we have the data
x f (x )

x0 f ( x0 )

x1 f ( x1 )

x2 f ( x2 )

The Lagrange fundamental polynomials are given by

( x  x1 )( x  x2 ) ( x  xo )( x  x 2 ) ( x  xo )( x  x1 )
 0 ( x)  ,  1 ( x)  ,  2 ( x) 
( xo  x1 )( xo  x2 ) ( x1  xo )( x1  x 2 ) ( x 2  xo )( x 2  x1 )

The Lagrange quadratic interpolation polynomial is given by

2
p2 ( x)    i ( x) f i
i 0

p2 ( x ) =  0 ( x ) f ( x 0 ) +  1 ( x ) f ( x1 ) +  2 ( x ) f ( x 2 ) .

( x  x1 )( x  x2 ) ( x  xo )( x  x 2 ) ( x  xo )( x  x1 )
= f ( x0 ) + f ( x1 ) + f ( x 2 ).
( xo  x1 )( xo  x2 ) ( x1  xo )( x1  x 2 ) ( x2  xo )( x2  x1 )

Example 5. 7: Determine the linear Lagrange interpolating polynomial that passes through the
points (2,4) and (5, 1).
Solution: In this case we have
( x  x1 ) ( x  5) 1 ( x  x0 ) ( x  2) 1
 0 ( x)    ( x  5) ,  1 ( x)    ( x  2)
( x0  x1 ) (2  5)  3 ( x1  x0 ) (5  2) 3
The Lagrange linear interpolation polynomial is given by
p1 ( x ) =  0 ( x ) f ( x0 ) +  1 ( x ) f ( x1 ) .

1 1
= ( x  5) (4) + ( x  2) (1)
3 3
= x6

134
Example 5. 8: 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.
Solution: We have x 0  0 , f 0  1 , x1  1 , f1  3 , x2  3 , f 2  55 . The Lagrange fundamental
polynomials are given by
( x  x1 )( x  x 2 ) ( x  1)( x  3) 1 2
 0 ( x)    0 ( x)   ( x  4 x  3).
( xo  x1 )( xo  x2 ) (1)(3) 3

( x  xo )( x  x 2 ) ( x  0)( x  3) 1
 1 ( x)    1 ( x)   (3x  x 2 ).
( x1  xo )( x1  x2 ) (1)(2) 2

( x  xo )( x  x1 ) ( x  0)( x  1) 1 2
 2 ( x)    2 ( x)   ( x  x).
( x 2  xo )( x 2  x1 ) (3)(2) 6
Hence, the Lagrange quadratic polynomial is given by
P2(x) =  0 ( x ) f (x0) +  1 ( x ) f (x1) +  2 ( x ) f (x2)

1 1 1
 ( x 2  4 x  3)(1)  (3 x  x 2 )(3)  ( x 2  x)(55).
3 2 6
= 8x2 – 6x + 1.
Example 5.9: Using Lagrange interpolation formula, find the form of the function y(x) from the
following table
x 0 1 3 4

f (x ) -12 0 12 24

Solution: Since y  0 when x  1 , it follow that x  1 is a factor. Let y  ( x  1) R( x)

y
Then R( x)  .We now tabulate the value of x and R(x).
( x  1)

135
 x 0 3 4

R (x ) 12 6 8

Applying Lagrange interpolation formula to the above table, we find

( x  3)( x  4) ( x  0)( x  4) ( x  0)( x  3)
R( x)  (12)  (6)  (8)
(3)(4) (3)(1) (4)(1)
 ( x  3) ( x  4)  2 x ( x  4)  2 x ( x  3)

 x 2  5 x  12.
Hence the required polynomial approximate to y (x) is given by
y (x)  ( x  1) ( x 2  5 x  12 )

 x 3  6 x 2  17 x  12.
Remark: For a given data, it is possible to construct the Lagrange interpolation polynomial.
However, it is very difficult and time consuming to collect and simplify the coefficients
of x i , i  0,1,2,..., n. Now, assume that we have determined the Lagrange interpolation
polynomial of degree n based on the data values ( xi , f ( x i )) , i  0,1,2,..., n at the (n + 1)

distinct points. Suppose that to this given data, a new value ( x n 1 , f ( x n 1 )) at the distinct

point x n 1 is added at the end of the table. If we require the Lagrange interpolating
polynomial for this new data, then we need to compute all the Lagrange fundamental
polynomials again. The nth degree polynomial obtained earlier is of no use. This is the
disadvantage of the Lagrange interpolation. However, Lagrange interpolation is a
fundamental result and is used in proving many theoretical results of interpolation.

Activity 5.3:

1. Write the property satisfied by Lagrange fundamental polynomials  i ( xi ) .

136
2. What is the disadvantage of Lagrange interpolation?

3.Given the table of values

x 150 152 154 156

f ( x)  x 12.247 12.329 12.410 12.490

evaluate √155 using Lagrange’s interpolation formula.

4.The following values of the function f ( x)  sin x  cos x , are given

x 100 200 300

f (x ) 1.1585 1.2817 1.3660

Construct the quadratic Lagrange interpolating polynomial that fits the data. Hence, find


f ( 12 ) . Compare with the exact value.

5. Construct the Lagrange interpolating polynomials for the following functions

i) f ( x)  sin(ln x) , x 0 = 2.0, x1 = 2.4, x2 = 2.6,

ii ) f ( x)  sin x  cos x , x 0 = 0, x1 = 0.25, x2 = 0.5, x 3 = 1.0

137
5.3.2 Newton’s Divided Difference Interpolation

Divided Differences
Let the data values, ( xi , f ( x i )) , i  0,1,2,..., n, be given. We define the divided differences as
follows. First divided difference:
Consider any two consecutive data values ( x i , f ( x i )), and ( x i 1 , f ( xi 1 )). Then,
we define the first divided difference as
f ( xi 1 )  f ( xi )
f [ xi , xi 1 ]  i  0,1,2,..., n  1.
xi 1  xi
f ( x1 )  f ( x0 ) f ( x2 )  f ( x1 )
Therefore, f [ x0 , x1 ]  , f [ x1 , x2 ]  etc.
x1  x0 x2  x1
f ( xi ) f ( xi 1 )
Note that f [ x i , xi 1 ]  f [ x i 1 , xi ]   .
xi  xi 1 xi 1  xi
We say that the divided differences are symmetrical about their arguments.
Second divided difference Consider any three consecutive data values ( x i , f ( x i )), ( x i 1 , f ( xi 1 )),

( x i  2 , f ( x i  2 )). Then, we define the second divided difference as

f [ xi 1 , xi  2 ]  f [ xi , xi1 ]
f [ xi , xi 1 , xi 2 ]  i  0,1,2,..., n  2.
xi  2  xi
f [ x1 , x2 ]  f [ x0 , x1 ] f [ x2 , x3 ]  f [ x1 , x2 ]
Therefore, f [ x0 , x1 , x2 ]  , f [ x1 , x2 , x3 ]  etc.
x 2  x0 x3  x1
The nth divided difference using all the data values in the table, is defined as

f [ x1 , x 2 ,..., x n ]  f [ x0 , x1 ,..., x n 1 ]
f [ x0 , x1 ,..., x n ]  (5.20)
x n  x0

The divided differences can be written in a tabular form as in Table 5.8.

x f (x ) First d.d Second d.d Third d.d

x0 f ( x0 )

138
 f [ x 0 , x1 ]

x1 f ( x1 ) f [ x 0 , x1 , x 2 ]

f [ x1 , x 2 ] f [ x 0 , x1 , x 2 , x3 ]

x2 f ( x2 ) f [ x1 , x 2 , x3 ]

f [ x 2 , x3 ]

x3 f ( x3 )

Table 5.8. Divided differences (d.d).

Example 5.10: Obtain the divided difference table for the data
x -1 0 2 3

f (x ) -8 3 1 12

139
Solution: We have the following divided difference table for the data.
The divided differences can be written in a tabular form as in Table 5.9.
x f (x ) First d.d Second d.d Third d.d

-1 -8

=11

0 3 =-4

= −1 =2
Table 5.9
Divided 2 1 =4 differences
(d.d).
=11

3 12

Activity5.4 : Obtain the divided difference table for the data

x -4 -1 0 2 5

f (x ) 245 23 4 6 335

We mentioned earlier that the interpolating polynomial representing a given data values is
unique, but the polynomial can be represented in various forms.
We write the interpolating polynomial as
f (x )  p n (x )

 c 0  ( x  x 0 )c1  ( x  x 0 )( x  x1 ) c 2  ...  ( x  x 0 )( x  x1 )... ( x  x n 1 ) c n . (5.21)

The polynomial fits the data p n ( x i )  f ( x i )  f i .

Setting p n ( x 0 )  f 0 , we obtain

p n ( x 0 )  f 0  c0 since all the remaining terms vanish.

140
Setting p n ( x1 )  f 1 , we obtain

f1  c 0 f  f0
f 1  c 0  ( x  x 0 )c1 , or c1   1  f [ x 0 , x1 ] .
x1  x0 x1  x0
Setting p n ( x 2 )  f 2 , we obtain f 2   c 0  ( x  x 0 )c1  ( x  x 0 )( x  x1 ) c 2 ,

f 2  f 0  ( x2  x0 ) f [ x0 , x1 ]
or c2 
( x2  x0 )( x2  x1 )

1 f  f0
 [ f 2  f 0  ( x2  x0 )( 1 )]
( x2  x0 )( x2  x1 ) x1  x0
f0 f1 f2
  
( x0  x1 )( x0  x2 ) ( x1  x0 )( x1  x2 ) ( x2  x0 )( x2  x1 )
= f [ x 0 , x1 , x 2 ].

By induction, we can prove that c n  f [ x 0 , x1 ,..., x n ].

Hence, we can write the interpolating polynomial as
p n (x )  f ( x 0 )  ( x  x 0 ) f [ x 0 , x1 ]  ( x  x 0 )( x  x1 ) f [ x 0 , x1 , x 2 ]  ...

 ( x  x 0 )( x  x1 )... ( x  x n 1 ) f [ x 0 , x1 ,..., x n ]. (5.22)

This polynomial is called the Newton’s divided difference interpolating polynomial.

For an example of a second order polynomial, given ( x 0 , y 0 ), ( x1 , y1 ), and ( x 2 , y 2 ).

p 2 ( x )  f ( x 0 )  ( x  x 0 ) f [ x 0 , x1 ]  ( x  x 0 )( x  x1 ) f [ x 0 , x1 , x 2 ].

Example 5.11: Find f (x ) as a polynomial in x for the following data by Newton’s divided
difference formula
x -4 -1 0 2 5

f (x ) 1245 33 5 9 1335

Solution: First we form the divided difference table for the data.

141
x f (x ) First d.d Second d.d Third d.d Fourth d.d

-4 1245

-404

-1 33 94

-28 -14

0 5 10 3

2 13

2 9 88

442

5 1335

The Newton’s divided difference formula gives

f (x )  f ( x 0 )  ( x  x 0 ) f [ x 0 , x1 ]  ( x  x 0 )( x  x1 ) f [ x 0 , x1 , x 2 ]

 ( x  x 0 )( x  x1 )( x  x 2 ) f [ x 0 , x1 , x 2 , x 3 ]

 ( x  x 0 )( x  x1 )( x  x 2 )( x  x 3 ) f [ x 0 , x1 , x 2 , x 3 , x 4 ]

= 1245 + ( x + 4)(- 404) + ( x + 4)( x + 1)(94) + ( x + 4)( x + 1) (x) (-14)

+ ( x + 4)( x + 1) (x) ( x -2)(3)

= 1245 – 404 x – 1616 + ( x 2 + 5 x + 4)(94) + ( x 3 + 5 x 2 + 4 x )(-14)

+ ( x 4 + 3 x 3 – 6 x 2 – 8 x )(3)
= 3 x 4 – 5 x 3 + 6 x 2 – 14 x + 5.
Example 5.12: Find f (x ) as a polynomial in x for the following data by Newton’s divided
difference formula
x –2 –1 0 1 3 4

f (x ) 9 16 17 18 44 81

142
Hence, interpolate at x = 0.5 and x = 3.1.
Solution: We form the divided difference table for the given data.
x f (x ) First d.d Second d.d Third d.d Fourth d.d

–2 9

-1 16 -3

1 1

0 17 0 0

1 1

1 18 4 0

13 1

3 44 8

37

4 81

Since, the fourth order differences are zeros, the data represents a third degree polynomial.
Newton’s divided difference formula gives the polynomial as
f (x )  f ( x 0 )  ( x  x 0 ) f [ x 0 , x1 ]  ( x  x 0 )( x  x1 ) f [ x 0 , x1 , x 2 ]

 ( x  x 0 )( x  x1 )( x  x 2 ) f [ x 0 , x1 , x 2 , x 3 ]

= 9 + ( x + 2)(7) + ( x + 2)( x + 1)(– 3) + ( x + 2)( x + 1) (x) (1)

= 9 + 7 x + 14 – 3 x 2 – 9 x – 6 + x 3 + 3 x 2 + 2 x = x 3 + 17.
Hence, f (0.5)  (0.5)3 + 17 = 17.125.
f (3.1)  (3.1)3 + 17 = 47.791.

143
Remark: Newton’s divided difference interpolating polynomial possesses the permanence
property. Suppose that we add a new data value ( x n 1 , f ( x n 1 )) at the distinct point x n 1 , at
the end of the given table of values. This new data of values can be represented by a
(n + 1)th degree polynomial. Now, the (n + 1)th column of the divided difference table
contains the (n + 1)th divided difference. Therefore, we require to add the term
( x  x 0 )( x  x1 )... ( x  x n 1 )( x  x n ) f [ x 0 , x1 ,..., x n , x n 1 ].

to the previously obtained nth degree interpolating polynomial given in (5.22).

Activity5.5:

1. Find f (x ) as a polynomial in x for the following data by Newton’s divided difference

formula

x 1 3 4 5 7 10

f (x ) 3 31 69 131 351 1011

Hence, interpolate at x = 3.5 and x = 8.0.

2. Does the Newton’s divided difference interpolating polynomial have the permanence

property?

3. Using divided difference formula, find f (3) given

f (1)  = – 26, f (2)  12, f (4)  256, and f (6)  844.

4. Using Newton’s divided difference interpolation, find y (10 ) given that

y (5)  12, y (6)  13, y (9)  14, y (11)  16.

5. Find f (8) by Newton’s divided difference formula, for the data

144
 x 4 5 7 10 11 13

f (x ) 48 100 294 900 1210 2028

5.4 Review Exercises

1. Does the Lagrange interpolating polynomial have the permanence property?
2 .For the data ( xi , f i ), i  0,1,2,..., n. construct the Lagrange fundamental polynomials  i ( x j )

using the information that they satisfy the conditions  i ( x j )  0, for i  j and  i ( x j )  1. for i  j.

3. Given

145
 x 0.20 0.22 0.24 0.26 0.28 0.3

f (x ) 1.6596 1.6698 1.6804 1.6912 1.7024 1.7139

Using Gauss’ difference interpolation formula, find f (0.262)

4. Construct the Lagrange interpolating polynomials for the following functions
i) f ( x )  e 2 x cos 3 x x 0  0, x1  0.3, x2  0.6,

ii ) f ( x)  ln x x 0  1, x1  1.1, x2  1.3, x 3  1 . 4,

5.Use appropriate Lagrange interpolating polynomials of degrees one, two, and three to
approximate each of the following:
i) f (8.4) if f (8.1)  16.94410, f (8.3)  17.56492, f (8.6)  18.50515, f (8.7)  18.82091
1
ii ) f ( ) if f (0.75)  −0.07181250, f (0.5)  −0.02475000, f (0.25)  0.33493750,
3
6.Use the Lagrange and the Newton-divided difference formulas to calculate f (3) from
the following table :.
x 0 1 2 4 5 6

f (x ) 1 14 15 5 6 19

CHAPTER SIX

6 Applications of Polynomial Interpolation

146
6.1 Introduction

After reading this chapter, students will be able to:

 understand the different applications of polynomial interpolation.
 derive and use different numerical differentiation methods.
 compute the maxima and minima of a tabulated functions.
 understand the need for numerical integration.
 derive and use different quadrature formulae to compute definite integrals.
 measure and calculate the maximum error bounds in quadrature formulae.
 derive and use the Romberg's method and Euler-Maclaurin formulae.
 apply method of undetermined coefficient to derive numerical formulae for both
differentiation and integration.
 derive and use numerical double integration formulae.
6.2 Numerical Differentiation

After completing this sections lesson, students are expected to:

 understand the need for numerical differentiation.
 calculate derivatives using forward difference formula.
 calculate derivatives using backward difference formula.
 calculate derivatives using central difference formula.
 compute the maxima and minima of a tabulated functions.

Numerical differentiation is the process of calculating the values of the derivative of a function at

some assigned values of from the given set of values ( , ). To compute , we first replace

the exact relation = ( ) by the best interpolating polynomial = ∅( ) and then differentiate
the later as many times as we desire. the choice of the interpolation formula to be used, will

depend on the assigned value of at which is desired, that means:

1. If the values of are equi-spaced and is required

i. near the beginning of the table, we employ Newton's forward formula.

ii. near the end of the table, we use Newton's backward formula.

147
 iii. near the middle of the table, is calculated by means of Stirling's or

Bessel's formula.
2. If the values are not equi-spaced, we use Newton's divided difference formula to
represent the function.

Hence corresponding to each of the interpolation formulae we can derive the formula for finding
the derivative.

Obs. while using these formulae, it must be observed that the table of values defines the function
at these points only and does not completely define the function and the function may not be
differentiable at all. As such, the process of numerical differentiation should be used if the
tabulated values are such that the differences of some order are constants. Otherwise, errors are
bound to creep in which go on increasing as derivatives of higher order are found. This is due to
the fact that the difference between ( ) and the approximating polynomial ( ) may be small
at the data points but ( ) − ′( ) may be large.

6.2.1 Formulae for derivatives

Consider the function = ( ) which is tabulated for the equally spaced data points , that is,
= + ℎ, for = 0, 1, 2, . . . , , with their corresponding functional values = ( ) , for
= 0,1,2, … , . Depending on these data points we can derive different numerical differentiation
formulae as follows.

1. Derivatives using forward difference formula

Newton's forward interpolation formula (p. ) is

( ) ( )( )
= + Δ + ∆ + ∆ +. ..
! !

Differentiating both sides with respect to , we have

=Δ + !
∆ + !
∆ +. ..

Since = , therefore =

148
Now = . = [Δ + !
∆ + !
∆ +. . . ]

. . . . . . . . . . . . . . . . . . . . . . . . . . . (1)

= , =0

= [Δ − ∆ + ∆ − ∆ + ∆ − ∆ +. . . ]

. . . . . . . . . . . . . . . . . . . . . . . . . . (2)

again differentiating (1) w.r.t , we get

1 1 1
= = . = . = . .
ℎ ℎ ℎ

1 1 2 6 −6 12 − 36 + 22 1
= . . = ∆ + ∆ + ∆ +. . .
ℎ ℎ 2! 3! 4! ℎ

Putting = 0, we obtain

= ∆ −∆ + ∆ − ∆ + ∆ ...

. . . . . . . . . . . . . . . . . . . . . . . . . . (3)

Similarly,

= ∆ − ∆ +. . .

. . . . . . . . . . . . . . . . . . . . . . . . . . (4)

2. Derivatives using Backward difference formula

Newton's backward interpolation formula (p. ) is

( + 1) ( + 1)( + 2)
= + ∇ + ∇ + ∇ +. ..
2! 3!
149
Differentiating both sides with respect to , we have

=∇ + !
∇ + !
∇ +. ..

Since = , therefore =

Now = . = [∇ + ∇ + ∇ +. . . ]
! !

. . . . . . . . . . . . . . . . . . . . . . . . . . . (5)

= , =0

= [∇ = ∇ + ∇ + ∇ + ∇ + ∇ +. . . ]

. . . . . . . . . . . . . . . . . . . . . . . . . . (6)

again differentiating (5) w.r.t , we get

1 1 1
= = . = . = . .
ℎ ℎ ℎ

1 1 2 6 +6 6 + 18 + 11 1
= . . = ∇ + ∇ + ∇ +. . .
ℎ ℎ 2! 3! 12 ℎ

Putting = 0, we obtain

= ∇ +∇ + ∇ + ∇ + ∇ +. . .

. . . . . . . . . . . . . . . . . . . . . . . . . . (7)

Similarly,

= ∇ + ∇ +. . .

. . . . . . . . . . . . . . . . . . . . . . . . . . (8)

3. Derivatives using central difference formulae

150
 Stirling's formula (p. ) is
p ∆y + ∆y p p(p − 1 ) ∆ y +∆ y
=y + + ∆ y +
1! 2 2! 3! 2
p (p − 1 )
+
∆ y +⋯
4!
Differentiating both sides with respect to p, we get
∆y + ∆y 2p 3p − 1 ∆ y +∆ y 4p − 2p
= + ∆ y + + ∆ y +⋯
2 2! 3! 2 4!

Since = , therefore =

∆ ∆ ∆ ∆
Now = . = + p∆ y + + ∆ y +⋯

= , = 0. Hence putting = 0, we get

∆ ∆ ∆ ∆ ∆ ∆
= − + +⋯

. . . . . . . . . . . . . . . . . . . . . . . . . . (9)
Similarly

= ∆ y − ∆ y + ∆ y −. . .

. . . . . . . . . . . . . . . . . . . . . . . . . . (10)

Obs: we can similarly use any other interpolation formula for computing derivatives.

Example 6.1: Given that

: 1.0 1.1 1.2 1.3 1.4 1.5 1.6

: 7.898 8.403 8.781 9.129 9.451 9.750 10.031

Then find and at ( ) = 1.1 ( ) = 1.6

Solution:

(a) The difference table is

151
X Y ∆ ∆ ∆ ∆ ∆ ∆

1.0 7.989

0.414

1.1 8.403 -0.036

0.378 0.006

1.2 8.781 -0.030 -0.002

0.348 0.004 0.002

1.3 9.129 -0.026 0.000 -0.003

0.322 0.004 -0.001

1.4 9.451 -0.023 -0.001

0.299 0.005

1.5 9.750 -0.018

0.281

1.6 10.031

We have

= [Δ − ∆ + ∆ − ∆ + ∆ − ∆ +. . . ] ... (i) and

= ∆ −∆ + ∆ − ∆ + ∆ ... ... (ii)

Here ℎ = 0.1, = 1.1, Δ = 0.378, ∆ = −0.003 etc. Substituting these values in

equations (i) and (ii), we get

= .
0.378 − (−0.003) + (0.004) − (0) + (−0.001) − (−0.003)
.

152
 = 3.946

=( . )
−0.03 − (0.04) + (0) − (−0.001) + (−0.003)
.

= −3.545

(b) We use the above difference table and the backward difference operator ∇ instead of Δ.

= [∇ + ∇ + ∇ + ∇ + ∇ + ∇ +. . . ] ... (i) and

= ∇ +∇ + ∇ + ∇ + ∇ +. . . ... (ii)

Here ℎ = 0.1, = 1.6, ∇ = 0.281, ∇ = −0.018, etc. putting these values in (i) and (ii)
we get

= .
0.281 + (−0.018) + (0.005) + (−0.001) + (−0.001) + (−0.003)
.

= 2.727

= −0.018 + 0.005 + (−0.001) + (−0.001) + (−0.003)

. ( . )

= −1.703

Example 6.2: A slider in a machine moves along a fixed straight rod. its distance cm along the
rod is given below for a various of the time seconds. Find the velocity of the slider and its
acceleration when = 0.3 second.

= 0 0.1 0.2 0.3 0.4 0.5 0.6

= 30.13 31.62 32.87 33.64 33.95 33.81 33.24

solution:

the difference table is:

∆ ∆ ∆ ∆ ∆ ∆

153
0 30.13

1.49

0.1 31.62 -0.24

1.25 -0.24

0.2 32.87 -0.48 0.26

0.77 0.02 -0.27

0.3 33.64 -0.46 -0.01 0.29

0.31 0.01 0.02

0.4 33.95 -0.45 0.01

-0.14 0.02

0.5 33.81 -0.43

-0.57

0.6 33.24

As the derivatives are required near the middle of the table, we use Stirling's formulae:

∆ ∆ ∆ ∆ ∆ ∆
= − + +⋯ ...(1)

= ∆ x − ∆ x + ∆ x . . . ... (2)

Here ℎ = 0.1, = 0.3, ∆x = 0.31, ∆x = 0.77, ∆ x = −0.46, etc.

putting these values in (1) and (2), we get

. . . . . .
= .
− + −⋯
.

154
 = 5.33

= .
−0.46 − (−0.01) + (0.29). . .
.

= −45.6

Hence the required velocity is 5.33 cm/sec and acceleration is -45.6 cm/sec2.

6.2.2 Maxima and Minima of a tabulated function

Forward interpolation formula is:

( ) ( )( )
= + Δ + !
∆ + !
∆ +. ..

differentiating it with respect to , we get

=Δ + !
∆ + !
∆ +. .. (1)

For maxima or minima, / = 0. Hence equating the right hand side of (1) to zero and
retaining only up to third differences, we obtain

Δ + !
∆ + !
∆ +. . . = 0

i.e. ∆ + (∆ −∆ ) + Δ − ∆ + ∆ =0

Substituting the values of Δ , ∆ and ∆ from the difference table, we solve this quadratic
for . Then the corresponding values of are given by = + ℎ at which is maximum or
minimum.

Example 6.3: From the table below for what values of , is minimum? Also find this value of
.

= 3 4 5 6 7 8

= 0.205 0.240 0.259 0.262 0.250 0.224

solution:

155
The difference table is:

∆ ∆ ∆ ∆ ∆

3 0.205

0.035

4 0.240 -0.016

0.019 0.000

5 0.259 -0.016 0.001

0.003 0.001 -0.001

6 0.262 -0.015 0.000

-0.012 0.001

7 0.250 -0.014

-0.026

8 0.224

Taking = 3 , we have = 0.205 , ∆ = 0.035 , ∆ = −0.016 and ∆ =0

therefore, Newton's forward difference formula gives

( − 1)
= 0.205 + (0.035) + (−0.016) ...( )
2!

Differentiating it w.r.t. p, we have

156
 2 −1
= 0.035 + (−0.016)
2!

For to be minimum , / =0

Thus, 0.035 − 0.008(2 − 1) = 0

which gives = 2. .6875

therefore, = + ℎ = 3 + 2.6875 × 1 = 5.6875

Hence is minimum when = 5.6875 .

Putting = 2. .6875 in ( ) the minimum value of becomes

2.6875 × (2.6875 − 1)
= 0.205 + 2.6875 × (0.035) + (−0.016)
2

= 0.2628

Activity 6.1

1. Find (0) and (0) from the following table:

: 0 1 2 3 4 5

: 4 8 15 7 6 2

157
 2. Find the first and second derivatives of ( ) at = 1.5 if
: 1.5 2.0 2.5 3.0 3.5 4.0

( ): 3.375 7.000 13.625 24.000 38.875 59.000

3. Given the following table of values of and

: 1.00 1.05 1.10 1.15 1.20 1.25 1.30

: 1.0000 1.0247 1.0488 1.0723 1.0954 1.1180 1.1401

Find and at i) = 1.00. ii) = 1.25

4. Given sin 0° = 0.0000, sin 10° = 0.1736, sin 20° = 0.3420, sin 30° = 0.5000, and
sin 40° = 0.6428, then
i) find the value of sin 23°

ii) find the numerical value of at = 10° for = sin ,

iii) find the numerical value of at = 20° for = sin .

5. The population of a certain town (as obtained from census data) is shown in the following
table:

: 1951 1961 1971 1981 1991

: 19.96 39.65 58.81 77.21 94.61

( in thousands)
Estimate the population in the years 1966 and 1993. And also find the rate of growth of
population in 1981.

6.3 Numerical Integration

After completing this sections lesson, students are expected to:

158
  understand the need for numerical integration techniques.
 derive the multiple-segment quadrature formulae of integration.
 use the multiple-segment quadrature formulae of integration to solve problems.
 calculate the truncation errors in each quadrature formulae.
 derive and use the Romberg's method and Euler-Maclaurin formulae.

The process of evaluating a definite integral from a set of tabulated values of the
integrand f(x) is called numerical Integration . This process when applied to a function
of a single variable is known as quadrature.
The problem of numerical integration, like that of numerical differentiation, is solved by
representing f(x) by an interpolation formula and then integrating it between the given
limits. In this way, we can derive quadrature formula for approximate integration of
function defined by a set of numerical values only.

6.3.1 Newton-Cotes quadrature formula

b
Let I   f  x dx
a

where = ( ) takes the values , y , y , … . , y for = ,x ,x ,….,x

Let us divide the interval ( , ) in to sub-interval of width ℎ so that =a,
=x +h, = x + 2h , .... , = x + nh = b . Then

159
 = ( )

Putting = + ℎ, =ℎ on the above integral gives

= ℎ∫ ( + ℎ)
( ) ( )( ) ( )( )( )
= ℎ∫ [ + ∆ + !
∆ + !
∆ + !
∆
( )( )( )( ) ( )( )( )( )( )
+ !
∆ + !
∆ + ⋯]

By Newton's forward interpolation formula , so integrating this term by term,

gives us
( ) ( )
=∫ ( ) = ℎ[ + ∆ + ∆ + ∆ +
∆ ∆
+ − + −3 !
+ −2 + − + 12 !
+
∆
− + 17 − + − 60 !
+ ⋯] . . . . (1)

This is known as Newton-Cotes quadrature formula. From this general formula ,

we deduce the following important quadrature rules by taking = 1,2,3, . . ..
1. Trapezoidal Rule
Putting = 1 in equ (1) above and taking the curve through ( , ) and
( , ) as a straight line i.e. a polynomial of first degree so that
differences of order higher than first becomes zero, we get

1 ℎ
( ) =ℎ + ∆ = ( + )
2 2
Similarly
1 ℎ
( ) =ℎ + ∆ = ( + )
2 2
.............................................................................
1 ℎ
( ) =ℎ + ∆ = ( + )
( ) 2 2
adding these n integrals, we obtain

160
 ∫ ( ) = [( + ) + 2( + + ⋯+ )] . . . . (2)

This is known as the Trapezoidal rule.

Obs. The area of each strip (trapezium) is found separately. Then the area
under the curve and the ordinates and + ℎ is approximately equal
to the sum of the areas of the trapeziums.

2. Simpson's one-third (1/3) rule

Putting = 2 in equ (1) above and taking the curve through ( , )
,( , ) and ( , ) as a parabola i.e. a polynomial of second degree so
that differences of order higher than second vanish , we get
1 ℎ
( ) = 2ℎ +∆ + ∆ = ( +4 + )
6 3
Similarly
ℎ
( ) = ( +4 + )
3
..........................................................................

∫ ( )
( ) = ( +4 + ), being even.

adding all these integrals, we have when n is even

∫ ( ) = [( + ) + 4( + + ⋯+ )+

+2( + +⋯+ )] . . . . (3)

This is known as the Simpson's one-third rule or simply Simpson's rule
and it is most commonly used.
Obs. While applying (3), the given interval must be divided in to even
number of equal sub-intervals, since we find the area of two strips at a
time.
3. Simpson's three-eighth (3/8) rule
Putting = 3 in equ (1) above and taking the curve through ( , )
= 0,1,2,3 as a polynomial of third order so that differences of order
higher than third vanish, we get

161
 3 3 1
( ) = 3ℎ + ∆ + ∆ + ∆
2 2 8
3ℎ
= ( +3 +3 + )
8
similarly
3ℎ
( ) = ( +3 +3 + )
8
Adding all such expressions from to + ℎ , where n is a multiple of
3, we obtain
3ℎ
( ) = [( + ) + 3( + + + +⋯+ )
8
+2( + + ⋯+ )] . . . . . . . (4)

Which is known as Simpson's three-eighth rule.

Obs. While applying (4), the number of sub-intervals should be taken as a
multiple of 3.

4. Boole's rule
Putting = 4 in equ (1) above and neglecting all differences above the
fourth order, we get
5 2 7
( ) = 4ℎ + 2∆ + ∆ + ∆ + ∆
3 3 90
2ℎ
= (7 + 32 + 12 + 32 +7 )
45
similarly
2ℎ
( ) = (7 + 32 + 12 + 32 +7 )
45
and so on.
Adding all these integrals from to + ℎ , where n is a multiple of 4, we
obtain

162
 2ℎ
( ) = (7 + 32 + 12 + 32 + 14 +
45
+32 + 12 + 32 + 14 …) . . . . . (5)
This is known as Boole's rule.
Obs. While applying (5), the number of sub-intervals should be taken as a
multiple of 4.

5. Weddle's rule
Putting = 6 in equ (1) above and neglecting all differences above the sixth
order, we obtain
9 123 11 1 41
( ) = 6ℎ + 3∆ + ∆ + 4∆ + ∆ + ∆ + . ∆
2 60 20 6 140

If we replace ∆ by ∆ , the error made will be negligible. Therefore,

3ℎ
( ) = ( +5 + +6 + +5 + )
10
Similarly

∫ ( ) = ( +5 + +6 + +5 + )

and so on.
Adding all these integrals from to + ℎ , where n is a multiple of 6, we
obtain
3ℎ
( ) = ( +5 + +6 + +
10
+5 +2 +5 + + ⋯) . . . . . (6)
This is known as Weddle's rule.
Obs. While applying (6), the number of sub-intervals should be taken as a
multiple of 6.
Weddle's rule is generally more accurate than any of the others. Of the two
Simpson's rules, the 1/3 rule is better.

163
Example 6.4:
Evaluate ∫ by using

i) Trapezoidal rule
ii) Simpson's 1/3 rule
iii) Simpson's 3/8 rule
iv) Weddle's rule and compare the results with its actual value.
Solution:
Divide the interval (0,6) into six equal parts each of width h=1. The values of
( )= are given below:

0 1 2 3 4 5 6

( ) 1 0.5 0.2 0.1 0.0588 0.0385 0.027

=

i) Trapezoidal rule
ℎ
= [( + ) + 2( + + + + )]
1+ 2

= [(1 + 0.027) + 2(0.5 + 0.2 + 0.1 + 0.0588 + 0.0385)]

= 1.4108.
ii) Simpson's 1/3 rule
ℎ
= [( + ) + 4( + + ) + 2( + )]
1+ 3

= [(1 + 0.027) + 4(0.5 + 0.1 + 0.0385) + 2(0.2 + 0.0588)]

= 1.3662.
iii) Simpson's 3/8 rule
3ℎ
= [( + ) + 3( + + + ) + 2( )]
1+ 8

= [(1 + 0.027) + 3(0.5 + 0.2 + 0.0588 + 0.0385) + 2(0.1)]

164
 = 1.3571.
iv) Weddle's rule
3ℎ
= ( +5 + +6 + +5 + )
1+ 10
= 0.3[1 + 5(0.5) + 0.2 + 6(0.1) + 0.0588 + 5(0.0385) + 0.027
= 1.3735 .
Also

∫ = [tan ] = tan 6 = 1.4056

This shows that the value of the integral found by Weddle's rule is the
nearest to the actual value followed by its value given by Simpson's 1/3
rule.

Example 6.5:
The velocity ( / ) of a moped which starts from rest, is given at
fixed intervals of time ( ) as follows:
2 4 6 8 10 12 14 16 18 20

10 18 25 29 32 20 11 5 2 0

Estimate approximately the distance covered in 20 minutes.

Solution:
If ( ) be the distance covered in ( ), then

= , therefore

| | =∫ = [ + 4. + 2. ] , by Simpson's 1/3 rule

Here ℎ = 2, =0, = 10, = 18, = 25, etc

= + = 0 + 0 = 0.
= + + + + = 10 + 25 + 32 + 11 + 2 = 80
= + + + = 18 + 29 + 20 + 5 = 72
Hence the required distance
| | = (0 + 4 × 80 + 2 × 72) = 309.33 .

165
Activity 6.2
1. Use trapezoidal rule to evaluate ∫ considering five sub-
intervals.

2. Evaluate ∫ using

i) Trapezoidal rule taking ℎ = 1/4.

ii) Simpson's 1/3 rule taking ℎ = 1/4.
iii) Simpson's 3/8 rule taking ℎ = 1/6.
iv) Weddle's rule taking ℎ = 1/6.
Hence compute an approximate value of in each case.

3. Evaluate ∫ taking 7 ordinates by applying Simpson's 3/8 rule.

And deduce the value of log 2.

/
4. Integrate numerically ∫ √cos .
5. Given that
: 4.0 4.2 4.4 4.6 4.8 5.0 5.2

log : 1.3863 1.4351 1.4816 1.5261 1.5686 1.6094 1.6487

.
Evaluate ∫ log using
i) Trapezoidal rule
ii) Simpson's 1/3 rule
iii) Simpson's 3/8 rule
iv) Weddle's rule.
6. The following table gives the velocity of a particle at time :
( ): 0 2 4 6 8 10 12

( / ): 4 6 16 34 60 94 136

Find the distance moved by the particle in 12 seconds and also the

166
 acceleration at = 2 .

6.3.2 Errors in quadrature formulae

The error in the quadrature formulae is given by

=∫ −∫ ( )
where ( ) is the polynomial representing the function = ( ), in the interval
[ , ].
1) Error in Trapezoidal rule
Expanding = ( ) around = by Taylor's series, we get
( )
= +( − ) + +⋯ . . . (1)
!

Therefore,
( − )
= +( − ) + +⋯
2!

= ℎ+ !
+ !
+⋯ . . . (2)

Also = area of the first trapezium in the interval [ , ]

= ℎ( + ) . . . (3)

Putting = +ℎ and = in (1), we get

= +ℎ + + +⋯
! !

Substituting this value of in (3), we get

1 ℎ ℎ
= ℎ + +ℎ + + ′ +⋯
2 2! 3!

=ℎ + + +⋯ . . . (4)
. !

Therefore, Error in the interval [ , ]

=∫ − [(2) − (4)]

= !
− . !
ℎ +⋯ =− +⋯

that is, principal part of the error in [ , ]=−

167
 Similarly principal part of the error in [ , ]=− and so on.

Hence the total error =− [ + + ⋯+ ]

Assuming that ( ) is the largest of the n quantities , , . .., ,

we obtain
( )
<− ( )=− ( ) . . . (5)

since ℎ = −
Hence the error in the trapezoidal rule is of order ℎ .

2) Error in Simpson's 1/3 rule

Expanding = ( ) around = by Taylor's series, we get (1).
Therefore, over the first doubt strip, we get
( − )
= +( − ) + +⋯
2!

= 2ℎ + !
+ !
+ !
+ !
+⋯ (6)

Also = area over the first doubt strip by Simpson's 1/3 rule

= ℎ( +4 + ) . . . (7)

Putting = +ℎ and = in (1), we get

ℎ ℎ
= +ℎ + + +⋯
2! 3!
Again putting = + 2ℎ and = in (1), we have
4ℎ 8ℎ
= + 2ℎ + + +⋯
2! 3!
Substituting these values of and in (7), we get

ℎ ℎ ℎ
= [ + 4( +ℎ + + +⋯)+
3 2! 3!
+( + 2ℎ + !
+ !
+ ⋯ )]

= 2ℎ + 2ℎ + + + + ⋯ . . . (8)

168
 Therefore, Error in the interval [ , ]
=∫ − [(6) − (8)]

= − ℎ +⋯=−

that is, principal part of the error in [ , ]=−

Similarly principal part of the error in [ , ]=− and so on.

Hence the total error =− + + ⋯+ ( )

Assuming that ( ) is the largest of , , . .., ( ) , we

obtain
( )
<− ( )=− ( ) . . . (9)

since 2 ℎ = −
Hence the error in Simpson's 1/3 rule is of order ℎ .
3) Error in Simpson's 3/8 rule
Proceeding as above, here the principal part of the error in the interval
[ , ]

=− . . . (10)

4) Error in Boole's rule

In this case, the principal part of the error in the interval [ , ]

=− . . . (11)

5) Error in Weddle's rule

In this case, the principal part of the error in the interval [ , ]

=− . . . (12)

6.3.3 Romberg's Method

In section 6.2.1 above, we have derived approximate quadrature formulae with

the help of finite differences method. Romberg's method provides a simple
modification to these quadrature formulae for finding their better approximations.

169
As an illustration, let us improve upon the value of the integral

=∫ ( ) ,
by Trapezoidal rule. If , be the values of with sub-intervals of width ℎ , ℎ
and , be their corresponding errors respectively, then
( ) ( )
=− ( ), =− ( )

Since ( ) is also the largest value of ( ), we can reasonably assume that

( ) and ( ) are very nearly equal. Therefore,

= or = . . . (1)

Now since = + = + , therefore

− = − . . . (2)
From (1) and (2), we have

= ( − )

Hence = + = + ( − ) , that is;

= . . . (3)

which is a better approximation of .

To evaluate systematically, we take

ℎ =ℎ and ℎ = ℎ

so that (3) gives

( / )
= ( / )
= , that is;

= [4 (ℎ/2) − (ℎ)] . . . (4)

Now we use the trapezoidal rule several times successively having ℎ and apply
(4) to each pair of values as per the following scheme:
(ℎ)

(ℎ, ℎ/2)

(ℎ/2) (ℎ, ℎ/2, ℎ/4)

170
 (ℎ/2, ℎ/4) (ℎ, ℎ/2, ℎ/4, ℎ/8)

(ℎ/4) (ℎ/2, ℎ/4, ℎ/8)

(ℎ/4, ℎ/8)

(ℎ/8)

The computation is continued till successive values are close to each other. This
method is called Richardson's deferred approach to the limit and its systematic
refinement is called Romberg's method.
Example 6.6:
Use Romberg's method to compute ∫ correct to 4 decimal places.

Solution:
Let's take ℎ = 0.5, 0.25, and 0.125 successively and evaluate the given integral
using Trapezoidal rule.
i) When ℎ = 0.5, the values of = (1 + ) are
: 0 0.5 1.0

: 1 0.8 0.5

.
therefore, = [1 + 2 × 0.8 + 0.5] = 0.775

ii) When ℎ = 0.25, the values of = (1 + ) are

: 0 0.25 0.5 0.75 1.0

: 1 0.9412 0.8 0.64 0.5

.
therefore, = [1 + 2(0.9412 + 0.8 + 0.64) + 0.5] = 0.7828

iii) When ℎ = 0.125, we find that = 0.7848

Thus we have
(ℎ) = 0.7750, (ℎ/2) = 0.7828, (ℎ/4) = 0.7848
Now using (4) above, we obtain

171
 (ℎ, ℎ/2) = [4 (ℎ/2) − (ℎ)] = [3.1312 − 0.775] = 0.7854

(ℎ/2, ℎ/4) = [4 (ℎ/4) − (ℎ/2)] = [3.1392 − 0.7828] = 0.7855

and (ℎ, ℎ/2, ℎ/4) = [4 (ℎ/2, ℎ/4) − (ℎ, ℎ/2)] = [3.142 − 0.7854] = 0.7855

Therefore, the table of these values is

0.7750

0.7854

0.7828 0.7855

0.7855

0.7848

Hence the value of the integral = 0.7855.

6.3.4 Euler-Maclaurin formula

Taking ∆ ( ) = ( ), we define the inverse operator ∆ as

( )=∆ ( ) . . . (1)
Now ( )− ( )=∆ ( )= ( )
( )− ( )=∆ ( )= ( )
..........................................................
( )− ( )=∆ ( )= ( )
Adding all these, we get
( )− ( )=∑ ( ) . . . (2)

172
where , ,…, are the ( + 1) equi-spaced values of with difference ℎ.
From (1) above, we have
( )=∆ ( ) = ( − 1) ( )=( − 1) ( ) since =

= 1+ℎ + + +⋯ −1 ( )
! !

= (ℎ ) 1+ + +⋯ ( )
! !

= 1− + − +⋯ ( )

= ∫ ( ) − ( )+ ( )− ( )+⋯ . . . (3)

Putting = and = in (3) and then subtracting, we get

( )− ( )= ∫ ( ) − [ ( ) − ( )] + [ ( )− ( )]

− [ ( )− ( )] + ⋯ . . . (4)

1 1 ℎ
( )= ( ) − [ ( ) − ( )] + [ ( )− ( )]
ℎ 2 12

− [ ( )− ( )] + ⋯

that is;

1 1 ℎ
( ) = ( )+ [ ( ) − ( )] − [ ( )− ( )]
ℎ 2 12

+ [ ( )− ( )] + ⋯

= [ ( ) + 2 ( ) + 2 ( ) + ⋯+ 2 ( )+ ( )]

− [ ( )− ( )] + [ ( )− ( )] + ⋯

Hence ∫ ( ) = [ +2 +2 +⋯+2 + ]

− ( − )+ ( − )+⋯ . . . (5)

which is called the Euler-Maclaurin formula.

Obs. The first term on the right-hand side of (5) represents the approximate value
of the integral obtained from trapezoidal rule and the other terms denotes the

173
successive corrections to this value. This formula is often used to find the sum of
a series of the form
( )+ ( + ℎ) + ⋯ + ( + ℎ) .
Example 6.7:
Using Euler-Maclaurin formula, find the value of log 2 from ∫ ?

Solution:
Taking = , = 0, = 10, ℎ = 0.1, we have

= −( )
and = −( )

Then Euler-Maclaurin formula gives

.
∫ = + .
+ .
+ .
+ .
+ .
+ .
+ .
+
( . ) ( . )
+ .
+ .
+ − ( )
−( )
+ ( )
−( )

= 0.693773 − 0.000625 + 0.000001 = 0.693149

Also ∫ = |log(1 + )| = log 2

Hence log 2 = 0.693149 approximately.

Example 6.8:
Apply Euler-Maclaurin formula to evaluate
+ + + ⋯+

Solution:
Taking = , = 51, = 24, ℎ = 2, we have

=− and =−

Then Euler-Maclaurin formula gives

2 1 2 2 2 1
= + + +⋯+ +
2 51 53 55 97 99
( ) ( )
− − + −

Therefore,
1 1 1 1 1 1 1 1 1 1 1
+ + + ⋯+ = + + + −
51 53 55 99 2 2 51 99 3 51 99

174
 − − +⋯

= + 0.000243 + 0.0000022 − ⋯

= 0.00499 approximately.

6.4 Method of undetermined coefficients

After completing this sections lesson, students are expected to:

 identify the importance of methods of undetermined coefficients.
 derive different numerical integrations formulae using methods of undetermined
coefficients.
 derive different numerical integration formulae using the methods of
undetermined coefficients.
The method of undetermined coefficient is based on imposing certain conditions on a
pre-assigned formula involving certain unknown coefficients and then using these
conditions for evaluating these unknown coefficients. Assuming the formula to be exact
for the polynomials 1, , ,…, respectively and taking for ( ), we shall
determine the unknown coefficients to derive the formulae.

6.4.1 Differentiation formulae

We first derive 2-term formula by assuming

= + . . . (1)
where the unknown constants , are determined by making (1) exact for
( ) = 1 and respectively.
So, putting ( ) = 1, successively in (1), we get
0= + and 1 = + ( + ℎ)
whence = 1/ℎ and = −1/ℎ

Hence = ( − ) . . . (2)

The 3-term formula can be derived by taking

= + + . . . (3)

175
where the unknown constants , , are determined by making (3) exact for
( ) = 1, and respectively. Therefore,
0= + +
1= ( − ℎ) + + ( + ℎ)
2 = ( − ℎ) + + ( + ℎ)
To solve these equations, we shift the origin to that is, = 0. As such,
being slope of the tangent to the curve = ( ) at = remains unaltered.
Thus the equations reduce to
+ + =0
− + = 1/ℎ
and + =0

giving =− , = 0, = 1/2ℎ

Hence = ( − ) . . . (4)

Similarly for second order derivative, taking

′ = + +
and making it exact for ( ) = 1, and and putting = 0, we get

′ = ( −2 + ) . . . (5)

6.4.2 Integration formulae

The 2-term formula is derived by assuming

∫ = + . . . (6)

where the unknown constants , are determined by making (6) exact for
( ) = 1 and respectively.
So, putting ( ) = 1, successively in (6), we get

+ =∫ 1. =ℎ

+ ( + ℎ) = ∫ . = [( + ℎ) − ]

To solve these, we shift the origin to and take = 0 . Therefore, the above
equations reduce to

176
 + = ℎ and = ℎ,

whence = 1/ℎ and = 1/ℎ

Hence ∫ = ( + ) . . . (7)

Which is Trapezoidal rule.

The 3-term formula is derived by assuming

∫ = + + . . . (8)

where the unknown constants , , are determined by making (8) exact for
( ) = 1, and respectively.
So, putting ( ) = 1, , successively in (8), we obtain

+ + =∫ 1. = 2ℎ

( − ℎ) + + ( + ℎ) = ∫ .

= [( + ℎ) − ( − ℎ) ]

( − ℎ) + + ( + ℎ) = ∫ .

= [( + ℎ) − ( − ℎ) ]

To solve these equations, we shift the origin to and take = 0 . Therefore,

the above equations reduce to
+ + = 2ℎ
− + =0
and + = ℎ

Solving these, we get = ℎ= , = ℎ

Hence ∫ = ( +4 + ) . . . (9)

Which is Simpson's rule

Activity 6.3
1. Obtain an estimate of the number of sub-intervals that should be chosen so as

to guarantee that the error committed in evaluating ∫ by trapezoidal rule

177
 is less than 0.001.

2. Use Romberg's method to compute ∫ . Hence evaluate log 2 correct

to four decimal places.

.
3. Apply Romberg's method to Evaluate ∫ log , given that

: 4.0 4.2 4.4 4.6 4.8 5.0 5.2

log : 1.3863 1.4351 1.4816 1.5261 1.5686 1.6094 1.6487

/
4. Using Euler-Maclaurin formula, find the value of ∫ sin correct to five
decimal places.
5. Apply Euler-Maclaurin formula, to evaluate
i) + + +⋯+

ii) ( )
+( )
+( )
+ ⋯+ ( )

6. Assuming that ∫ ( ) = ℎ( + )+ℎ ( + ) , derive

the quadrature formula, using the method of undetermined coefficients.

6.5 Numerical Double Integration

After completing this sections lesson, students are expected to:

 understand the need for numerical double integration techniques.
 derive and use trapezoidal formula for calculating double integrals.
 derive and use Simpson's formula for calculating double integrals.
The double integral

= ( , )

178
is evaluated numerically by two successive integrations in and directions considering
one variable at a time. Repeated application of trapezoidal rule ( or Simpson's rule )
yields formulae for evaluating .
1) Trapezoidal rule
Dividing the interval ( , ) into equal sub-intervals each of length ℎ and
the interval ( , ) into equal sub-intervals each of length , we have:
= + ℎ, = , =
= + , = , = .
Using trapezoidal rule in both directions, we get
ℎ
= [ ( , )+ ( , ) + 2{ ( , ) + ( , ) + ⋯ + ( , )}]
2

= [( + ) + 2( + + ⋯+ )

+( + )+2 + + ⋯+ ,

+2 ∑ ( + )+2 + + ⋯+ , ]
where = , .

2) Simpson's rule
Dividing the interval ( , ) into 2 equal sub-intervals each of length ℎ
and the interval ( , ) into 2 equal sub-intervals each of length . Then
applying Simpson's rule in both directions, we get
ℎ
( , ) = [ ( , )+4 ( , )+ ( , )]
3

= [ , +4 , + , +4 , +4 , + ,

+ , +4 , + , ]
Adding all such intervals, we obtain the value of .
Example 6.9:

179
 Using trapezoidal rule evaluate

=∫ ∫ , taking four sub-intervals.

Solution:
Taking ℎ = = 0.25 so that = = 4, we obtain

= [ ( , ) + ( , ) + 2( ( , . ) + ( , . ) + ( , . ))

+ ( , ) + ( , ) + 2( ( , . ) + ( , . ) + ( , . ))

+2{ ( . , ) + ( . , ) + 2( ( . , . ) + ( . , . ) + ( . , . ))

+ ( . , ) + ( . , ) + 2( ( . , . ) + ( . , . ) + ( . , . ))

+ ( . , ) + ( . , ) + 2( ( . , . ) + ( . , . ) + ( . , . ) )}]

= 0.3407.
Example 6.10:
Apply Simpson's rule to evaluate the integral
. .
=∫ ∫ , taking two sub-intervals.

Solution:
Taking ℎ = 0.2 = 0.3 so that = = 2, we get

= [ (4,2) + 4 (4,2.3) + (4,2.6)

+4{ (4.2,2) + 4 (4.2,2.3) + (4.2,2.6)}

+ (4.4,2) + 4 (4.4,2.3) + (4.4,2.6)]
.
= [0.6559 + 4(0.6246) + 0.5962]
.
= × 3.7505 = 0.025.

Activity 6.4
1. Evaluate ∫ ∫ using trapezoidal rule (by taking ℎ = = 0.5).
2. Using Trapezoidal and Simpson's rules, evaluate

∫ ∫ , by taking two sub-intervals.

3. Using Trapezoidal rules, evaluate

180
 ∫ ∫ , by taking two sub-intervals.

. .
4. Evaluate ∫ ∫ , using Simpson's rule.

Review Exercise
1. Find the first and second derivative of the function tabulated below, at the point = 1.1:
: 1.0 1.2 1.4 1.6 1.8 2.0

181
 ( ): 0 0.128 0.544 1.296 2.432 4.00

2. For the following values of and , find the first derivative at = 4.

: 1 2 4 8 10

: 0 1 5 21 27

3. Find the value of cos(1.74) using the values given in the table below:
: 1.70 1.74 1.78 1.82 1.86

sin : 0.9916 0.9857 0.9781 0.9691 0.9584

4. Find at = 1 from the following table by constructing a central difference table:

: 0.7 0.8 0.9 1.0 1.1 1.2 1.3

: 0.644218 0.717356 0.783327 0.841471 0.891207 0.932039 0.963558

5. If = ( ) and denotes ( + ℎ), prove that , if power of ℎ above ℎ be neglected.

= ( − )− ( − )+ ( − )

6. Using the following data , find for which is minimum and find this value of .
: 0.60 0.65 0.70 0.75

: 0.6221 0.6155 0.6138 0.6170

7.

182
Bibliography

Brice Carnahan, H. L. (1969). Applied Numerical Methods. New York: John Wiely and Sons Inc.
E.Atkinson, K. (1978). An Introductory to Numerical Analysis. Canada: John Willey and Sons
Inc.
J.Stoer, R. B. (1991). Introduction to Numerica Analysis. New York: Springer-velag New York
Inc.
Jain, M. (2007). Numerical methods for engineering and Computation . New Dehil: New age
Intrnational.

183
Lee W. Johnson, R. R. (1977). Numerical Analysis. Philippines: Addison-Wesley Publishing
Campany Inc.
P. Kandasamy. 2008. Numerical Method. Ram Nagar, New Delhi, 110055.

Purna Chandra. 2008. Numerical analysis. New Delhi, 110001.

R.L. Burden, J. F. (2010). Numerical Analysis.
Roslonice, S. (2008). Fundamental of Numerical Methods for Eletrical Engineering. Berlin
Heidelberg: Springer Verlag.
Sastry, S. (2002). Introduct Methods of Numerical Analysis

184