You are on page 1of 38

Katana

one or more roots of an


equation
Finding
0
747
s one
of the most commonly occurring
in scientific and engineering
problem
be given
studies The f f may
n in
explicitly
as a polynomial of degree
or
be defined implicitly
x or flat may
A transcendental y
a transcendental function
root a finite or infinite
have no
may CI
while a polynomial y
no of roots
n reel complex roots
has exactly
isquadratic
cubic biquadratic express
If f
are available
then algebraic formulas
the
the root in terms of
for expressing
On the other hand if f is a
coefficient
f in
polynomial of higher degree or a

algebraic methods are


transcendental f
and recourse must be taken
not available
iteratin
the root by new approximate
to find direct

methods

methods These methods are based


Iterative

If sucessive approximations starting


to
or more initial approximations
with one

iterates
the root we obtain a sequence of
to the
in the limit converges
71k which
one or two
methods determine
root These
time
roots at a

with an example consider solving


0 for a
given an
fate a
I
to
This problem has a practical application
machine divide operation
computers without
a

let a
take an
approximate solution of flat 0

210,76 draw a
tangent
At the point
let x be the
line to the graph y fat
line intersect
at which the tangent
point
x axis To obtain an
equation for X

match the slopes obtained from the

at ie
line and the derivative of f so
tangent

y a

ya

1 0,71 07

i
t

Xo Xi Xo 9 0 1

Ixi a.it
nolz

formula is then obtained


The general iteration
the process with X replacing x
by repeating
ad infinitum
to get
N f IN
Ant Xn 2 9in

more convenient for theoretical purposes


A form
the scaled residue
s obtained by introducing
In 1 a in

Using it
gene pin Itoh IN EN

For the error and


I
1
en
ta an
I I
this
We will analyze the convergence of
method its speed and its dependence on to

First
I 9 sent 1 92in Itn
Tnt
I 1 rn Itn

Innert
Inductively
g ro 30

the erm en to a n a iff


This implies
neo Iff 1201 1 ie
Into as

t CI 911 C I

oeil to it is
In order that sin converge
that slot
necessary
and sufficient
o
f
when 121 is satisfied
For the speed of convergence
enti
ng ng engl
d ent aen

in't
E
e

Rel hint Refn


relative error in Tn Based on
where Rel17in

we
say that en converges to zero quadratically
how will decrease
To illustrate rapidly
Then Rel 15
that Red 1767 0.1 y
suppose
Iterate
A root 9
is called a simple root of

no if 7157 0 f's 40 Then


f
write flat 21 5 g a
we can also

whose him go so and 915770


US

Bisecting lentinuous function


on the
is a
Assume f it also satisfies
and that
internet gib
given
f a f b o

value theorem F SE q s
Using intermediate
0
such that 7157
is the mid point of intend
If my qty
in m or may
either lies q
then the root
to or have the
provided fem to If Fim
required root
re

Repeating the process a no


of times we

the Bisection method


obtain

I
9kt bn an 1 0,112
mkt

he
can muti if flan femur so
faun but
mut ibn if fimin F but

last internet as

we take the mid point of


to the root
an approximation latest
untill the
this process
We repeat the root is as

which contains
internet
length
small as desired say of E
the new internal
of nth
end step Canbut
At the
1 9 1
would be of length

have
we then
Is ee
so
msn.si
iteration required
137 gives the no of
Eg and
E if lb a

achen've an accuracy
to
2 0.001

n 310 x
I
af
Caftan
i As
Ig
ix

15,7167
117 Find a real root of
Eyes
fat It I txt 7

correct to three decimal places


and the last
is cubic
The given y have a
will
Hence fate
term i positive
reel root
negative
Note that 14
71 37
6 71 27 1
71 1
3 and 2
lies between
a root
Therefore

We take
t
2,1 5
lies between
the root
4.875
Since FL E
2 and 2 T

2 2.25
Xz
31
therefore the
1.5781
Since Ff 2.25
in 2 and 2.25
root lies
Susessive approximations
are
given by
2.0625 45 2.0938
Xy
A 2.1016
2.1 94
to
2.1035
Xg 2 IOTT Xg

X 2.1050
X 2.1045
b w is 0 0005
difference to xp
The
that the root is given
Hence we conclude
to three decimal places
2.105 correct
y between
Find the positive root
It the equation
si z é to a

0 and I of
0.05
tolerance of
let fat xe l
SI and fill e i o
7107 1
we have
between o and I
root exists
hence a
N Sin Fan Tolerance

9 0 l
b I e 170

I 0 5 0.1756

2 0.75 0.5878 E 33.377

3 0.625 22 2 7
Y 0.5625 4 11.117

J o 5938 4 5.211
6
6.5781 8 2.711
7 one A W seen
0.5614 G o 69T
9 0.5687 Eg 0.357
10 0.5674
11 0.5669
12 0.5671 4 6 0351

where Ent 4x100


the interval in which the
Find
Exes lies
positive root of the following 9
smallest
tan at tan ha o
at

b I 21 4 0

roots correct to two decimal


Determine the
bisection method
places the
Iteration method based on first degree
n

We know that if y fat is a


first degree
2
is y Fet 4
equation
Then the solution of 767 0 is
easily
computable 0
a 4

Ed
e

far is not a first degree equation


If y

approximate tox by a
hen we
may
and we write
first degree equation
0
y fee 971 91

The soy of a is given by

x E
Go

where a da are arbitrary constant to


determined two appropriate
be by prescribing
conditions on flt and in derivative

Secant method's

affair we
are

determine
two

a
approximations of
and a in
root then

the condition
using

Fk it 964k it 9

a te t 9
fat
where fie f fan and fu
a
flare

we obtain
On solving

1a T
E
i
a N
in we obtain the next
using
to the root given by
approximation aka

xmt

I n
fani g.fi
This is Its in this method
chord

method Geometrically
straight line
we replace
the th f by a

chord passing through


the points
or a
and take
and Nk 17km
ah fie
intersection of the straight
the point of
axis as the next
line with x

the root If the


approximation of
approximation
is such that fk.fi so
then the method is know as

Falsi method Since folk y fkn and


Regula
or method of false position
known before the start of
lock Fi are

iteration the second method


the

requires one function evacuation per step

Secant methed
image

i
No Élan

I
Regala false
method
Use the secant and regula falsi
Efe noon the equation
method to determine the of
He o
Cosa
fee
as
initial approximation
SI Taking the
0 FL 7107 1
40
101 e 2.177979523
X I fix 7117

Secant method
k n fine fifteenth
0.51987117
0.3146653378 0.519871175 0.3176053378
1
0.4467281466
0 20945
0.4467281461 0.203545
2
0.4940173 0 708023
6.5317058606 0 429311 f 01
3

2 3

remaining computer
11 N
The number within the parenthesis denote

exponentiation

0 f x 7107 1
40
101 e 2.177979523
X I fix 7117

0.3146053378
471,747
E i

flask 0.5198717737725

0.4467281466

9,1 7,7627
B 2

0.20354T
flag
secant
Xyz 713 4151 Fan
method
flat flat
regales false
method

fi it
0.4940153
Method
Newton Raphson
in 111 using the
a and a
We determine
conditions
following
Kut 9
Fido
fi a

denotes the differentiation


where a prime value
t 21 Now by substituting
the of
W r
in we
90 and a from
obtain

yn 1fqgt f
method
Newton Raphson
This method is called
also be
The above
iterative formula may
In the
xu in
obtained by passing xp
when the chord passing through
limit se ten
and folks fan becomes the
the point Pik fk
I all
tangent line at the point ok Fk
If
fit
y
Goitcho

inn

to
Newton Raphson's method
Apply
Empty root of the equation
a
determine
cont He
flat
8 when a is the approxima
s t 17Gt e 10
1
and initial approximation
to the root
The iterative genes using the method
1
Newton nephron is given by
of
fi gift
a

k ke they Met fake


06530794 O 460
1.0 0.349
0
0.65307940 0.1217 0.53134337 0.41
I
0 134341051790991 o.gg
2 0.53134337

3 0.51790991 0.152513 0.51775738 g É


0.2195
0.1948 0 51775736
Y 0.51775738 1 10
C 71
i

Pn a t pro
there exists a sequence
If
I independent of
n
constant X
and a
positive
t
fan n Ed lent for all
then X is
value of n
sufficiently large
to x with rate of
said to converge
0 Ph
convergence

an 11 0 137
is
Let Xn be
Def orderfcenergenet
to x let
that conveys
a sequence
constant
en xn se
If 7 a positive
X and a s t
it
im
no

said to converge to a
then xn is
em
war a with asymptotic
of
constant d

It follows that for a sequence


Nof 2 the eons
order
that converges of
the asymptotic relation
satisfies
Item t
discussions
From previous

Ethiopia t let f be a continuous f on

theorem
intend aib and suppose that
the closed
the bisection method
Then
flat fist co

a sequence of approximations
generates
I
Xn which convey to a root By Vt
RE Cais with the property

pin ale
1591 1 9
3
not
Remedy 111 Only existence of a root
uniqueness
which the
with problems for
2 Working that
is known and verifying
rot
analytical satisfied
bound tan all is
em
theoretical human
tool for eliminating
is a powerful
in the development of computer
errors
codes
that an intend fans
3 The requirement that
implies
flat feb
co
i t
be found be used to
method cannot
the bisection
root of even multiplicity
locate
fee x 17
Eg act Cb fla so
for any
715770
also implies
that Bisection method
41
has rate of convergence of
know that
As we
5
EX is
iterative sequence Xn given by
the

thy
Xngty
at em I

atten
anti i
Engen

be defined by
Let 121 a sequence

it
2 end at 2 aft n t
en ta
a 440
Then lima
I t
Fan
4
a Itza I 4 1 a
2
of lim
Since
hey
9
24 EI z
Bisection method
The order of convergena of
error constant
is I with asymptotic

under the assumption linen exist

6 On the plus side only one new function


so the
evaluation is needed per iteration
i.e Hss imputations cost
method inexpensiveis

most importantly
the sequence of approximate
the method is guvranted
generated by
to converge
On the minus side the sequence of
71
the bisection
approximation generated by
method conveys only linearly with r o c
off
theoretical em bound is available
Even though a
As a result
the bound can be overly pessimistic is
which
that an approximation
t is possible
accurate to within the specified convergence
to terminate the iteration
terence will fail
andsecant d
Methodffolkposition
of iterations
It is evident form
the formula
scant method or method of false position
for

t
ni
mfÉ
we need nti function evaluation for

n iteration
t 2
fr x2 we need fo tf
ie
n f 72 2
K n

th t I
In t Fn y z

in approximating
H t compare the error

the sort of

11 fee
13 222 32 1

ii flat tanta a 6
method

using Bisch'm method of Regala Fabri Secant


7
Fom the outset we
might have expected
bisection
faln position to always outperform
secant method
After all
the false position
about the f However
uses more information
that this is
above shown
the example
not the case There is no
general theory
which method will be better
to indicate

for a
given f
which the method
The main advantage
method has over
secant
of false position
is the existence of
the bisection method
error estimate
computable
Secant method
Convergence Analysis

Teasfumt s is a simple root of


in
substitute 71k St Ea
fix o and

ni
i
fi LK I

eat a

Esiliiga
Expanding f ster and 715 4 in Taylor

series about the


point 5 and noting

that 7157 0 we
get
Em

gyjyyyg.gg 1

y
when
festal teh t Ex 715
II
t
F s

Eat I Ek Ek t Off

or
feni CEK.fi
where c
I egg and the higher power
of Ei are neglected

The Relation of the form is called


Error equation Keeping in view of def of
order of converge a we seek a

relation of the form


Ext A Sit
A and are to be determined
when p
we have
From
P
it A Epa

the value of 9kt and Ek in


Substituting
we get
EI cattle't
Ek
comparing the power of

Ptt p A P
Il F
minus find the
Neglecting the sign we

secant method is
rate of convergence f
p
me 1

method of False position

aport method
of approximation
of false position
generated by
P If so what
a root
Pn converge

the order of convergence


is
The squence Pn conveys
let en poi p
and the order
a n yo of
Iff lent to
the
is determined by
convergence
relationship between lent ten
asymptotic
know that Pn satisfies
We

bn
pn
jigger
bn n f's
espn p
Iggy
th f values flan
Now we
approximate
second degree polynomials
and flbn by
fan pl'd
fan I F p an P t
FYI
F bn Dt bn pi
7147
P
FYI
has been taken into
where the fact f p
is then
The term fl bn flan
the aciount

approximately

Y
fin f

Ibn an flip 1731 bn tan Lp

and into 117 factoring the


Substituting
dividing out the tem bn an yields
term bn p
Pu p I bn p bn p F'Int Id Snp

flntfgncbntan
lbn
mft tyypgffg.nl

prptlbn man
2PI
P.fi igpihntan

as the
Letthposition
iteration proceeds one of the iteration
become fixed In fact
will always eventually
will always
position
the method of false the configurations
settle into one of
eventually flip so
I f p o

by i an fixed
ii

time i
an fixed
bn Pn I
i
I
f ipso
i
p f PRO
20
i
f p
an Pat
bn fixed

Eventual configuration for the method of false


position
the enclosing
interval
Because one endpoint of
the other is just pay 141
remains fixed while
becomes

extend
after
where x
m

ant an remains fixed


and f
bn p bn remains fixed
that the sequence
Provided 10121 it then follows
the method of false position
generated by
and the convergence is linear
converges
error constant Al
with asymptotic
in more
one of the configuration
let us consider
detail
I an p
let an is fixed if
an p f in o
and f Pl co
Now an p co

o t 2711Mt and f P fan p f p


Since f p
and

Ift
and the method of false position
tiny Intel
M a similar fashion the
conveyor
can be shown
remaining three configuration
to lead to Incl
we would like to terminate
Zing when lent
the method of false position
tolerances
below a specified convergence
falls
this idea we must have a formula
To implement
lent which involves only quantities
for estimating
be calculated during the course
of
which can

note that
iteration As a starting point

in Pn p

Pn In it Pn i t

Pn Pn tent
From the error evolution equation sent deny
this in u
en Substituting get
is Trey
1in on l
lent
If
1

estimated terms
The valor of A can be wing
from the sequence Pn as follows
Consider the ratio

a i i i
and en Ien ily
relation en Iden
Using the
we find

a i
together constitute
161 and 171
Equation
Ion
a computable estimate for font squinty
condition for the
an appropriate stopping
is to terminate
method of false position
when
the iteration

I ten on ice

where N is obtained from 7

You might also like