Life Science Journal 2013;10(2s) http://www.lifesciencesite.

com

Some Families of Higher Order Three-Step Iterative Techniques

1 1 2 2
Nazir Ahmad Mir, Saher Akmal Khan, Naila Rafiq, Nusrut Yasmin.
1
Department of Basic Sciences, Riphah International University, Islamabad, Pakistan,
2
Centre for Advanced Studies in Pure and Applied Mathematics, Bahauddin Zakariya University, Multan, Pakistan.
nazirahmad.mir@gmail.com

Abstract: We suggest and analyze four new families of higher order three-step iterative techniques for solving non-
linear equations. Each of these families has seventh order convergence. Per iteration, these families of methods
require three function and one of its derivative evaluations. Thus, each of these families has 1.627 computational
efficiency. Several examples are given to illustrate the performance and efficiency of these methods. Comparison
with other similar methods is also given
[Nazir Ahmad Mir, Saher Akmal Khan, Naila Rafiq, Nusrut Yasmin. Some Families of Higher Order Three-Step
Iterative Techniques. Life Sci J 2013;10(2s):43-53] (ISSN:1097-8135). http://www.lifesciencesite.com. 8

Keywords: Iterative methods; Three-step methods; Predictor-Corrector type methods; Convergence order.

Introduction f (y n )
Solving non-linear equation is one of the most zn = yn  '
, (3)
important problems in numerical analysis. Newton's f (y n ) f (y n )
method is the most significant iterative method for
solving non-linear equation. where  is a real number and y n is defined by (2).
Consider the approximation, by Chun [5]:
f (x ) = 0 (1) ' (1)
f (x n )[f (x n )  f ( y n )]
f ' (y n ) = (4)
A non-linear equation cannot be solved in general f (x n )  f ( y n )
analytically. Therefore, a number of approximate
Using (4) in (3) , we obtain a new family of
techniques have been developed to compute the zeros
of non-linear equations. These techniques have been Iterative Methods,
developed using error analysis, Adomian zn = yn 
Decomposition methods, Quadrature formulas, etc.
f ( y n )(f (x n )  f ( y n )
In recent years, mathematicians have developed ,
'
many two-step and three-step iterative methods, see f (x n )(f (x n )  f ( y n ))   f ( y n )(f (x n )  f ( y n )
[1-8]. (5)
In this paper, four new families of higher order three-
step methods are developed using variants of
where   R and y n is defined by (2).
Changbum Chun [5] and Li. Tai fang [7] techniques
by applying binomial approximation on them and Second family of methods is obtained by applying
using the strategy of Weihong Bi et al [6]. binomial expansion upto first order on (5) , namely
Convergence analysis of these families of methods is
discussed. Each of these families is of seventh order
f ( y n )[f (x n )  f ( y n )]
convergence. Per iteration,. these families of methods zn = yn 
require three function and one of its derivative f ' (x n )[f (x n )  f ( y n )]
evaluations. Thus, each of these families has 1.627 (6)
  f ( y n )[f (x n )  f ( y n )] 
computational efficiency. 1  ' ,
1 Construction of families of three-step iterative  f (x n )[f (x n )  f ( y n )] 
methods
Consider Newton's Method:
where   R and y n is given by (2).
f (x n )
yn = xn  (2) Similarly, third family of methods is obtained using
f ' (x n ) second approximation on (5), that is
A family of iterative method by Li-Fang et.al [7] is
given by:

http://www.lifesciencesite.com 43 lifesciencej@gmail.com
 Life Science Journal 2013;10(2s) http://www.lifesciencesite.com

f ( y n )[f (x n )  f ( y n )] f (z n )  f ( y n ) 1 ''
zn = yn  f ' (z n )   f ( y n )(z n  y n )
f ' (x n )[f (x n )  f ( y n )] zn  yn 2
  f ( y n )[f (x n )  f ( y n )]  1
1  f ' (x )[f (x )  f ( y )]  (7) = f [z n , y n ]  f '' ( y n )(z n  y n )
n n n
2
 , (12)
   f ( y )[f (x )  f ( y )] 2  In order to avoid the computation of the second
  ' n n n
  ''
  f (x n )[f (x n )  f ( y n )]   derivative, we approximate f ( y n ) as follows:

where   R and y n is given by (2). f '' ( y n )  2f [z n , x x , x n ]

Fourth family of methods is given by using family of 2(f [z n , x x ]  f ' (x n )) (13)
C.Chun and Y.Ham [8], and applying first binomial = ,
approximation on it, that is zn xn
z n = y n  [2f (x n )  (2  1)f ( y n )] where z n and x n are sufficiently close to y n
[(2f (x n )  (2   5)f ( y n )] f ( y n ) (8) when n is sufficiently big integer.
Substituting (13) into (12) and replacing
4f 2 (x n ) f ' (x n )
Further, considering the strategy of Weihong et.al f ' (z n ) with approximation in (12) , we propose
[6], we propose three- step family of methods the following three-step family of methods from(9):
combining equations (2) and (5) : f (x n )
yn = xn  ,
f ' (x n )
f (x n )
yn = xn  zn = yn 
f ' (x n )
f ( y n )(f (x n )  f ( y n ))
zn = yn  '
,
f (x n )(f (x n )  f ( y n ))   f ( y n )(f (x n )  f ( y n ))
f ( y n )( f (x n )  f ( y n ))
' (14)
f (x n )(f (x n )  f ( y n ))   f ( y n )(f (x n )  f ( y n )) and
(9)
and x n 1 = z n  G ( n )
f (z )
x n 1 = z n  G ( n ) ' n , f (z n )
f (z n ) ,
f [z n , y n ]  f [z n , x n , x n ]( z n  y n )
f (z n )
where n = and G (t ) represents a real- f (z n )
f (x n ) where n = and G (t ) represents a
valued function.
f (x n )
' real-valued function.
Using the Taylor expansion, f ( z n ) and f ( z n )
Similarly from (2), (6) and the third equation of
can be approximated as:
(14), we have second family of three-step methods:
f (z n ) = f (z n  y n  y n )  f ( y n )  f (x n )
1 (10) yn = xn  ,
f ( y n )(z n  y n )  f '' ( y n )(z n  y n ) 2 ,
' f ' (x n )
2 f ( y )[f (x n )  f ( y n )]
'' zn = yn  ' n
f ' (z n ) ; f ' ( y n )  f ( y n )( z n  y n ), (11) f (x n )[f (x n )  f ( y n )]
this implies (15)
  f ( y n )[f (x n )  f ( y n )] 
1  ' 
 f (x n )[f (x n )  f ( y n )] 

and

http://www.lifesciencesite.com 44 lifesciencej@gmail.com
 Life Science Journal 2013;10(2s) http://www.lifesciencesite.com

x n 1 = z n  G ( n ) f (z n )
where n = and G (t ) represents a real-
f (z n ) f (x n )
, valued function.
f [z n , y n ]  f [z n , x n , x n ]( z n  y n )
We now discuss convergence for families of three
f (z n ) step methods (14  17).
where  n = and G (t ) represents a real-
f (x n ) 2 Convergence Analysis
valued function. Theorem 1
Also from (2), (7) and third equation of (14) , third Assume that f : D  R  R . is a scalar function
family of methods is given by: on some open interval D . Suppose x D ,
f (x n ) f ' (x  )  0. If the initial point x 0 is sufficiently
yn = xn  ,
f ' (x n ) close to x  and G is any function with
f ( y )[f (x n )  f ( y n )] G (0) = 1 , G ' (0) = 1 , G '' (0) <  , then the
zn = yn  ' n
f (x n )[f (x n )  f ( y n )]
family of three-step methods (14) is of convergence
  f ( y n )[f (x n )  f ( y n )]  order seven for any   R .
1  f ' (x )[f (x )  f ( y )]   (16) Proof.
n n n
 
  f ( y )[f (x )  f ( y )] 2  Let e n = x n  x  , en = y n  x  and
 ' n n n
  (k )
f (x ) 

 f ( x n )[ f ( x n )  f ( y n )]   d n = z n  x  . Denote c k = ,
k !f ' (x  )
and k = 2, 3,.... Using the Taylor expansion and taking
x n 1 = z n  G ( n ) into account f ( x ) = 0 , we have.
f (z n ) f (x n )  f ' ( x  )[e n  c 2e n2  c 3e n3  c 4e n4  c 5e n5  O (e n6 )]
, (18)
f [z n , y n ]  f [z n , x n , x n ]( z n  y n )
f ' ( x n )  f ' ( x  )[1  2c 2e n  3c 3e n2  4c 4e n3  5c 5e n4  O (e n5 )]
(19)
f (z n )
where n = and G (t ) represents
f (x n ) From((18) and (19), we get:
a real-valued function. f (x n )
Fourth family is obtained by combining (2), (8) and '
= e n  c 2e n2  [2c 22  2c 3 ]e n3
f (x n )
third equation of (14) :
[7c 2c 3  4c 23  3c 4 ]e n4 (20)
f (x )
yn = xn  ' n , [8c 24  20c 22c 3  6c 32  10c 2c 4  4c 5 ]e n5
f (x n )
zn = yn  O (e n6 )
[2f (x n )  (2  1)f ( y n )][(2f (x n )  (2  5)f ( y n )] (17) Thus, from (2) , we have:
4f 2 (x n ) e n: = c 2e n2  [2c 3  2c 22 ]e n3  [4c 23  7c 2c 3  3c 4 ]e n4 
f (y n ) [ 8c 24  20c 22c 3  6c 32  10c 2c 4  4c 5 ]e n5  O (e n6 )
f ' (x n ) (21)
and Again expanding f ( y n ) about

x , we have:
x n 1 = z n  G ( n )
f ( y n ) = f ' (x  )[e n:  c 2e n: 2  O (e n: 3 )]
f (z n )
,
f [z n , y n ]  f [z n , x n , x n ]( z n  y n )

http://www.lifesciencesite.com 45 lifesciencej@gmail.com
 Life Science Journal 2013;10(2s) http://www.lifesciencesite.com

f ( y n ) = f ' (x  )[c 2e n2  (2c 3  2c 22 )e n3  f (z n )  f (x n ) f (z n )  f (x n )

f [z n , x n ] = =
(3c 4  5c 23  7c 2c 3 )e n4  zn xn d n  en
(22)
(12c 24  24c 22c 3  10c 2c 4  6c 32  4c 5 )e n5  f ' (x  )
O (e n6 )]. [d n  c 2d n2  c 3d 23  O (d n4 )  (e n  c 2e n2  c 3e n3 
From (18), (19) and (22) , we c 4e n4  O (e n5 ))]
have
d n  en
f ( y n )(f ( x n )  f ( y n ))
f' ( x n )[f ( x n )  f ( y n )]   f ( y n )(f (x n )  f ( y n )] This implies,
2 2 3
= c e  (2c 3  2c )e  (3c 4  6c 2c 3   c  c )e
2 n 2 n
2
2
3
2
4
n f [z n , x n ] = f ' (x  )[1 c2 (en  d n )  c3 (en2  d nen  d n2 )
 (4c 2c 3  10c 24  4c 23  4c 5  4c 32  8c 2c 4 )e n5 c4 (en3  en2d n  end n2  d n3 ) O (d n4 )]
O (e n6 ) (27)
Now,
(23)
From (5), we have: f [z n , x n ]  f ' (x n )
f [z n , x n , x n ] =
f (y n )[f (x n )  f (y n )] zn xn
dn =en:  '
f (x n )[f (x n ) f (y n )] f (y n )[f (x n )  f (y n )] f [z n , x n ]  f ' (x n )
=
d n  en
Using (21) and (23) in the above result, we obtain:
d n = (3c 23  c 2c3  c 22 )e n4  (20c 22c3  4c 2c3 18c 24 Then by (19) and (27), we obtain:
3 2 5 6
4c  2c  2c 2c 4 )e  O (e )
2 3 n n f [ z n , x n , x n ] = f' ( x  )[c 2  c 3 (e n  (e n  d n ))
(24)
c 4 (e n2  (d n  e n )e n  (d 2n  e n2  d n e n )  O (e n3 )]
Expanding f ( z n ) about x  , we have:
f ( z n ) = f ' ( x  )[d n  c 2d n2  c 3d n3  c 4d n4  O (d n5 )] Using (24), we have:
(25) f [ z n , x n , x n ] = f ' ( x  )[c 2  2c 3e n  3c 4e n2  O (e n3 )]
Thus by first equation of (22) and (25), we have: (28)
f (z n )  f (y n ) '  Equations (24) further gives:
f [z n , y n ] = = f (x )
zn yn f [ z n , x n , x n ]( z n  y n ) = f ' ( x  )
[dn c2dn2 c3d23 O(dn4 )  (en: c2en: 2 c3en: 3 O(en: 4 ))] [c 2  2c 3e n  3c 4e n2 ](d n  e n )
dn en: Using (24), we obtain:
f [ z n , x n , x n ]( z n  y n ) = f ' ( x  )
This implies
f [z n , y n ] = f ' (x  )[1  c 2d n  c 2e n:  c 3e n: 2  O (e n6 )] [ c 22e n2  ( 4c 2c 3  2c 23 )e n3
( 4c 32   c 23  6c 2c 4  c 24  10c 22c 3 )e n4  O (e n5 )]
Using(21) and (24), we have: (29)
Equation (26) and (29) , imply:
f [ z n , y n ] = f' ( x  )[1  c 22e n2  2c 2 (c 3  c 22 )e n3 f [z n , y n ]  f [z n , x n , x n ](z n  y n ) = f ' ( x  )
 (7c 24  7c 22c 3  3c 2c 4   c 23 )e n4  O (e n5 )] [1  2c 2c 3e n3  (6c 24  3c 22c 3  3c 2c 4  2 c 23  4c 32 )e n4  O (e n5 )]
(26) (30)
Moreover, by (18) and (25), we have: Now from (18) and (25), we have:
f (z n ) d n
n = =  c 2d n  O (e n5 ) (31)
f (x n ) e n

http://www.lifesciencesite.com 46 lifesciencej@gmail.com
 Life Science Journal 2013;10(2s) http://www.lifesciencesite.com

Using the Taylor expansion and considering '

With G (0) = 1 , error equation is given by
G '' (0) <  , we get e n 1 = (c 22c 32  9c 26   2c 24  6 c 25 )e n7 (37)
' 2
G (  n ) = G (0)  G (0) n  O (  ) n
We have the same error equation for family of
Using (31) , we have:
methods (15) and (16) with seventh order
d  convergence.
G ( n ) = G (0)  G ' (0)  n  c 2d n   O ( n2 )
3 THE CONCRETE ITERATIVE METHODS
 en 
The three-step family of methods (14) suggests
(32)
From (24), we have some new methods which are stated as under:
3.1 METHOD 1 (Abbreviated as MK1)
dn For the given function G , defined by
= (3c 23  c 2c 3  c 22 )e n3 
en t
G (t ) = 1  , (38)
(20c 22c 3  4c 2c 3  18c 24  4c 23  2c 32  2c 2c 4 )e n4 1  t
O (e n5 ) where   R , it can easily be seen that the
(33) function G (t ) of (38) satisfies the conditions of
Thus, the error equation for family of methods (14) Theorem 1. Hence, we get a new two-parameter
is given by: seventh-order family of methods from (14):
e n 1 = x n 1  x  = z n  G ( n ) f (x n )
yn = xn 
f (z n ) f ' (x n )
x 
f [ z n , y n ]  f [ z n , x n , x n ](z n  y n )
zn = yn 
f (z n )
= d n  G ( n ) f ( y n )(f ( x n )  f ( y n ))
f [ z n , y n ]  f [ z n , x n , x n ](z n  y n ) '
f (x n )(f ( x n )  f ( y n ))   f ( y n )(f ( x n )  f ( y n ))
(39)
= dn 
dn
[G (0)  G ' (0)(  c 2d n )  O ( n2 )][d n  c 2d n2  O (d n3 )] f (x n )  (   1)f (z n )
en x n 1 = z n 
1  2c 2c 3e n3  (6c 24  3c 22c 3  3c 2c 4  2c 23  4c 32 )e n4  O (e n5 ) f (x n )   f (z n )
(34) f (z n )
d f [z n , y n ]  f [z n , x n , x n ](z n  y n )
= dn [G(0)dn G (0) n dn G' (0)c2dn2
'

en 3.2 METHOD 2 (Abbreviated as MK2)

d For the function G , defined by
G(0)c2dn2 G' (0)c2 n dn2 G' (0)c22dn3 ...]
en 1
3 4 2 3 2 4 5
G (t ) = 1
(40)
[1 2c c e  (6c  3c c 3c2c4  2c  4c )e O(e )]
2 3 n 2 2 3 2 3 n n 
(1   t )
On simplification, we have: where   0 , it can easily be seen that the function
' 3 2 2
e n 1 = [1  G (0)]d n  [ G (0)(3c  c 2c 3   c ) 2 2 G (t ) of (40) satisfies the conditions of Theorem
1. Hence, we get a new two-parameter seventh-order
2c 2c 3G (0)(3c 23  c 2c 3   c 22 )]e n7  O (e n8 ) family of methods from(14):
(35) f (x n )
Thus, the convergence order of the family (14) is yn = xn 
f ' (x n )
seventh order if G (0) = 1 and the error equation is
given by
f ( y n )(f (x n )  f ( y n ))
en1 =[G' (0)(3c23 c2c3 c22 )2 2c2c3 (3c23 c2c3 c22)]en7 zn = yn 
f ' (x n )(f ( x n )  f ( y n ))   f ( y n )(f (x n )  f ( y n ))
O(en8 ) (41)
(36)

http://www.lifesciencesite.com 47 lifesciencej@gmail.com
 Life Science Journal 2013;10(2s) http://www.lifesciencesite.com

1 1
 f (x n )    f (x n ) 
x n 1 = z n    x n 1 = z n   
 f (x n )   f (z n )   f (x n )   f (z n ) 
f (z n ) f (z n )
f [z n , y n ]  f [z n , x n , x n ](z n  y n ) f [z n , y n ]  f [z n , x n , x n ](z n  y n )
3.3 METHOD 3 (Abbreviated as MK3) 3.5 METHOD 5 (Abbreviated as MK5)
For the function G , defined by (38) For the function G , defined by (38)
t
t G (t ) = 1 
G (t ) = 1  , 1  t
1  t where   R , it can easily be seen that the
where   R , it can easily be seen that the function G (t ) of (38) satisfies the conditions of
function G (t ) of (38) satisfies the conditions of Theorem 1. Hence, we get a new two-parameter
Theorem 1. Hence, we get a new two-parameter seventh-order family of methods from(16):
seventh-order family of methods from (15): f (x n )
f (x ) yn = xn 
yn = xn  ' n f ' (x n )
f (x n ) f ( y )[f (x n )  f ( y n )]
f ( y )[f (x n )  f ( y n )] zn = yn  ' n
zn = yn  ' n f (x n )[f (x n )  f ( y n )]
f (x n )[f (x n )  f ( y n )]
(42)   f ( y n )[f (x n )  f ( y n )] 
  f ( y n )[f (x n )  f ( y n )]  1  f ' (x )[f (x )  f ( y )]  (44)
1  '   n n n

 f (x n )[f (x n )  f ( y n )]     f ( y )[f (x )  f ( y )] 2 
  ' n n n
 
f (x n )  (   1)f (z n )   f (x n )[f (x n )  f ( y n )]  
x n 1 = z n 
f (x n )   f (z n )
f (x n )  (   1)f (z n )
f (z n ) x n 1 = z n 
f (x n )   f (z n )
f [z n , y n ]  f [z n , x n , x n ](z n  y n )
3.4 METHOD 4 (Abbreviated as MK4) f (z n )
For the function G , defined by (40) f [z n , y n ]  f [z n , x n , x n ](z n  y n )
1 3.6 METHOD 6 (Abbreviated as MK6)
G (t ) = 1
, For the function G , defined by (40)

(1   t ) 1
G (t ) = 1
,
where   0 , it can easily be seen that the function 
G (t ) of (40) satisfies the conditions of Theorem (1   t )
1. Hence, we get a new two-parameter seventh-order where   0 , it can easily be seen that the function
family of methods from(15): G (t ) of (40) satisfies the conditions of Theorem
f (x ) 1. Hence, we get a new two-parameter seventh-order
yn = xn  ' n family of methods from(16):
f (x n )
f (x n )
f ( y )[f (x n )  f ( y n )] yn = xn 
zn = yn  ' n f ' (x n )
f (x n )[f (x n )  f ( y n )]
(43)
  f ( y n )[f (x n )  f ( y n )] 
1  ' 
 f (x n )[f (x n )  f ( y n )] 

http://www.lifesciencesite.com 48 lifesciencej@gmail.com
 Life Science Journal 2013;10(2s) http://www.lifesciencesite.com

f ( y n )[f (x n )  f ( y n )] (47)
zn = yn  Now, from equation(17), we get:
f ' (x n )[f (x n )  f ( y n )]
  f ( y n )[f (x n )  f ( y n )]  [2f (xn ) (2 1)f (y n )][(2f (xn ) (2 5)f (yn )] f (yn )
dn =en 
1  f ' (x )[f (x )  f ( y )]  (45) 4f 2(xn ) f ' (x n )
n n n
  (48)
   f ( y )[f (x )  f ( y )] 2  5
  ' n n n
  Using equation(46) and (48) and for  = , we
  f (x n )[f (x n )  f ( y n )]   2
1
have:
 f (x n )   d n = (5c 23  c2c3 )e n4  (32c 22c3  36c24  2c2c4  2c32 )e n5
x n 1 = z n   
 f (x n )   f (z n )  O (e n6 )
f (z n ) (49)

f [z n , y n ]  f [z n , x n , x n ](z n  y n )
By Taylor Series about x , we have:
f ( z n ) = f ' ( x  )[d n  c 2d n2  c 3d n3  c 4d n4  O (d n5 )]
Theorem Now,
Assume that f : D  R  R is a scalar function f (zn )f (yn )

f [zn , yn ]= =
on some open interval D . Suppose x  D , z n yn
f ' (x  )  0. If the initial point x 0 is sufficiently [dn c2dn2 c3d23 O(dn4)(en: c2en: 2 ce :3 :4
3 n O(en ))]
f ' (x )
close to x  and G is any function with dn en:
G (0) = 1 , G ' (0) = 1 , G '' (0) <  , then the
Using (21) and (49), we get:
family of three-step methods (17) is of convergence
f [ z n , y n ] = f ' ( x  )[1  c 2e n2  (2c 2c 3  2c 23 )e n3
5
order seven for = . ( 7c 22c 3  3c 2c 4  9c 24 )e n4  O (e n5 )]
2
(50)
, en = y n  x
 
Proof. Let e n = x n  x and Now, Using (18) and (49), we have:
 f ( k ) (x  ) f [z n , x n ] = f ' (x  )[1  c 2 (en  d n )  c3 (en2  d ne n  d n2 )
dn = z n  x . Denote ck = ,
k !f ' (x  ) c 4 (e n3  en2d n  e nd n2  d n3 )  O (d n4 )]
k = 2,3,... Using the Taylor expansion and taking (51)
into account f ( x ) = 0 , we have: Also by definition,
From (19)and (22), we have: f [z n , x n ]  f ' (x n )
f [z n , x n , x n ] =
f (y n ) zn xn
'
= c2en2  (2c3  4c22 )en3  (13c23 14c2c3  3c4 )en4 
f (x n ) f [z n , x n ]  f ' (x n )
=
(38c24  64c22c3 12c32  20c2c4  4c5 )en5 O (en6 ) d n  en
(46)
Thus, using (19), (22) and (46), we Using (19) and (51), we get:
get
2f (x n )  (2  1)f ( y n )][(2f (x n )  (2  5)f ( y n )] f ( y n )
f [z n , x n , x n ] = f ' (x  )[c2  c3 (en  (en  dn ))
4f 2 (x n ) f ' (x n )
= c 2e n2  (2c3  (2  7)c 22 )en3  (26c 2c3  ( 2 17 c4 (en2  (dn en )en  (d 2n en2  d nen ) O (en3 )]
141 3 Using (49), we obtain:
 )c 2  (8  26)c 2c3  3c4 )e n4
4
331 2 283 4 f [z n , x n , x n ] = f ' (x  )[c2  2c3en 3c4en2 O(en3 )] (52)
(4c5  (6 2  94  )c2 c3  (10 2  96  )c2
2 2 This implies ,
(18  8 )c32  (12  38)c 2c 4 )en5  O (en6 )

http://www.lifesciencesite.com 49 lifesciencej@gmail.com
 Life Science Journal 2013;10(2s) http://www.lifesciencesite.com

f [ z n , x n , x n ](z n  y n ) = e n 1 = [10G ' (0)c 24c 3  (G ' (0)  2)c 22c 32  25G ' (0)c 26 10c 24c 3 ]e n7
O (e n8 ).
f ' ( x  )[c 2  2c 3e n  3c 4e n2 ](d n  e n )
(58)
'
Using (49), we have: With G (0) = 1 , error equation is given by
'  2 2
f [zn ,xn ,xn ](zn yn )=f (x )[c e (4c2c3 2c )e
2 n
3 3
2 n
e n 1 = (c 22c 32  25c 26 )e n7  O (e n8 ) (59)
2 4
(10c c c 6c2c4 4c )e Oe 2 4
( )] 5 4 The CONCRETE FAMILY OF THREE-STEP
2 3 2 3 n n ITERATIVE METHODS
(53)
The three-step family of methods (17) suggest some
Equations (50) and (53) , imply,
new methods which are stated as under:
f [z n , y n ] f [z n , x n , x n ](z n  y n ) = 4.1 Method 7 (Abbreviated as MK7)
f ' (x )[1 2c2c3en3 (4c32 3c2c4 10c24 3c22c3)en4 For the function G , defined by
t
O(en5 )] G (t ) = 1  , (60)
(54)
1  t
Using the Taylor expansion and considering where   R , it can be easily seen that the
''
G (0) <  , we get: function G (t ) of (60) satisfies the conditions of
Theorem 2. Hence, we get a new one-parameter
G (  n ) = G (0)  G ' (0)  n  O (  n2 ) seventh order family of methods from(17):
Now from (18) and (49), we have: f (x n )
yn = xn 
f (z n ) d n f ' (x n )
n = =  c 2d n  O (e n5 )
f (x n ) e n
Using this result, we have: [2f (x n )  f ( y n )] f ( y n )
zn = yn  (61)
d  2f ( x n ) f ' (x n )
G ( n ) = G (0)  G ' (0)  n  c 2d n   O ( n2 )
 en 
f (x n )  (   1)f (z n )
From (49) , we have x n 1 = z n 
f (x n )   f (z n )
dn
=(5c23 c2c3)en3 (32c22c3 36c24 2c32 2c2c4)en4 O(en5) f (z n )
en
(55) f [z n , y n ]  f [z n , x n , x n ](z n  y n )
Thus, the error equation for family of methods (17)
is given by 4.2 Method 8 ( Abbreviated as MK8)
e n 1 = d n  For the function G , defined by

d 1
[G (0) G' (0)( n  c 2d n ) O (n2 )][d n  c2d n2 O (d n3 )] G (t ) = 1
(62)
en 
(1   t )
1  2c2c3en3  (10c 24  3c22c3  3c2c4  4c32 )e n4  O (en5 )
where   0 , it can easily be seen that the function
(56)
G (t ) of (61) satisfies the conditions of Theorem
dn
= [1  G (0)]d n  [G ' (0)  2G (0)c 2c 3e n3 ]d n 2. Hence, we get a new one-parameter seventh-order
en family of methods
[G (0)  G ' (0)]c 2d n2 f (x n )
yn = xn 
[10c 24  3c 22c 3  3c 2c 4  4c 32 )]G (0)e n4d n  O (e n8 ) f ' (x n )
(57) [f (x n )  2f ( y n )] f ( y n )
zn = yn  (63)
Thus, the convergence order of family of methods f (x n ) f ' (x n )
(17) is seventh-order if G (0) = 1 , and the error
equation is given by

http://www.lifesciencesite.com 50 lifesciencej@gmail.com
 Life Science Journal 2013;10(2s) http://www.lifesciencesite.com

1 The execution is stopped at the third iteration for the

 f (x n )  sake of comparison of the methods.
x n 1 = z n    We use the following functions from [2]:
 f (x n )   f (z n ) 
f 1 (x ) = x3 + 4x 2  15 ,
f (z n )
x  = 1.631980805566063518,
f [z n , y n ]  f [z n , x n , x n ](z n  y n )
1
f 2 ( x ) = sin( x )  x ,
Thus, we have constructed four families of three-step 2

methods MK 1 to MK 8. Per iteration, each of x = 1.895494267033980947,
these families of methods require three evaluations of
the function and one evaluation of its first derivative. f 3 ( x ) = e  x  cos( x ), x  = 1.746139530408012418.
We consider the definition of efficiency index as f 4 ( x ) = sin 2 ( x )  x 2  1 ,
1
w
p , where p is the order of the method and w is the x  = 1.4.4491648215341226
number of function evaluations per iteration required f 5 (x ) = cos( x )  x ,
by the method. We observed that each family of
4
x  = 0.7390851332151606417,
methods has the efficiency index 7 ; 1.627. 2
5 Numerical Examples f 6 ( x ) = xe x  sin 2( x )  3cos( x )  5,
Now the Methods MK 1 to MK 8 are x  = 1.207647827130919,
implemented to solve some non-linear equations and 2
compared with Kou’s method of seventh order [2]. f 7 (x ) = 10xe  x  1, x  = 1.67963061042845.
Table 1 shows the difference of two consecutive
iterates, function value, total number of function
evaluatons (TNFE) and total number of iterations.

Table 1. Comparison of various iterative methods

G7 MK 1 MK 2 MK 3 MK 4
1 1  =1
 = , = 2  = , = 2
4 4
f 1, x o = 2
TNFE 12 12 12 12 12
ITERATIONS 3 3 3 3 3
x n 1  x n .36e  20 . 10e  37 . 11e  37 .11e  37 .10e  37
f (x n ) .56e  102 .95e  266 .15e  269 .13e  265 .95e  266
f 2, x o = 2
TNFE 12 12 12 12 12
ITERATIONS 3 3 3 3 3
x n 1  x n . 12e  28 . 83e  56 .84e  56 .83e  56 .83e  56
f (x n ) .52e  145 .3e  349 .3e  349 .3e  349 .3e  349
f 3, x o = 2
TNFE 12 12 12 12 12
ITERATIONS 3 3 3 3 3
x n 1  x n . 169e  25 . 87e  52 . 8e  52 .89e  52 .87e  52
f (x n ) .1e  130 .3e  349 . 41e  287 .3e  349 .3e  349

http://www.lifesciencesite.com 51 lifesciencej@gmail.com
 Life Science Journal 2013;10(2s) http://www.lifesciencesite.com

f 4 , x o = 1.6
TNFE 12 12 12 12 12
ITERATIONS 3 3 3 3 3
x n 1  x n .26e  20 . 27e  38 . 3e  38 .271e  38 .27e  38
f (x n ) .21e  102 . 41e  269 . 82e  269 .45e  269 .41e  269
f 5,x o = 1
TNFE 12 12 12 12 12
ITERATIONS 3 3 3 3 3
x n 1  x n .15e  27 . 26e  59 . 25e  59 .256e  59 .26e  59
f (x n ) .15e  140 0 0 0 0
f 6 , x o = 1
TNFE 12 12 12 12 12
ITERATIONS 4 4 4 4 4
x n 1  x n .98e  17 .19e  13 .15e  14 .2e  13 .2e  13
f (x n ) .21e  83 .23e  92 .34e  100 .24e  92 .23e  92
f 7 , x o = 1.8
TNFE 12 12 12 12 12
ITERATIONS 3 3 3 3 3
x n 1  x n .27e  20 .35e  33 .29e  33 .35e  33 .5e  33
f (x n ) .57e  102 .14e  232 .33e  233 .15e  232 .14e  232

Table 2. Comparison of various iterative methods

G7 MK 5 MK 6 MK 7 MK 8
f 1, x o = 2  =1  =1
TNFE 12 12 12 12 12
ITERATIONS 3 3 3 3 3
x n 1  x n .36e  20 .105e  37 .11e  37 .23e  33 .25e  33
f (x n ) .56e  102 .96e  266 .15e  265 .97e  235 .19e  234
f 2, x o = 2
TNFE 12 12 12 12 12
ITERATIONS 3 3 3 3 3
x n 1  x n .12e  28 .83e  56 .16e  46 .23e  51 .24e  51
f (x n ) .52e  145 .3e  349 .58e  327 .3e  349 .3e  349
f 3, x o = 2
TNFE 12 12 12 12 12
ITERATIONS 3 3 3 3 3
x n 1  x n .17e  25 .87e  52 .79e  52 .282e  50 .269e  50

http://www.lifesciencesite.com 52 lifesciencej@gmail.com
 Life Science Journal 2013;10(2s) http://www.lifesciencesite.com

f (x n ) .102e  130 .3e  349 .3e  297 .3e  349 .3e  349
f 4 , x o = 1.6
TNFE 12 12 12 12 12
ITERATIONS 3 3 3 3 3
x n 1  x n .265e  20 .268e  38 .27e  38 .1e  34 .11e  34
f (x n ) .21e  102 .414e  269 .82e  269 .15e  243 .31e  243
f 5,x o = 1
TNFE 12 12 12 12 12
ITERATIONS 3 3 3 3 3
x n 1  x n .15e  27 .26e  59 .42e  59 .17e  56 .18e  56
f (x n ) .15e 0140 0 .5e  329 0 0
f 6 , x o = 1
TNFE 12 12 12 12 12
ITERATIONS 3 3 3 3 3
x n 1  x n .98e  17 .2e  13 .15e  14 .16e  6 .9e  9
f (x n ) .21e  83 .23e  92 .34e  100 .18e  43 0.26e  59
f 7 , x o = 1.8
TNFE 12 12 12 12 12
ITERATIONS 3 3 3 3 3
x n 1  x n .27e  20 .35e  33 .28e  33 .57e  29 .45e  29
f (x n ) .57e  102 .14e  232 .33e  233 .1e  202 .2e  203

6 Conclusion 3. M.A. Noor, New classes of iterative methods for

In this work, we presented eight three-step families of
seventh order convergent methods. We observe that
these iterative methods are comparable with seventh
order method (G 7 ) of Kou, et.al [2] as cited in the
Table 1 and Table 2, and in almost all the cases,
these families of methods as compared to Kou et.al
method of seventh order [2] give better results in
terms of absolute error and function value.
Corresponding Author:
Nazir Ahmad Mir
Department of Basic Sciences,
Riphah International University
Islamabad, Pakistan
E-mail: nazirahmad.mir@gmail.com.
References
1. C. Chun, Y. Li, X, Wang, Some sixth-order variants of
Ostrowski root-finding methods, Appl. Math. Comput
193(2007) 389-394.
2. J. Kou, Y. Li, X.Wang, Some variants of Ostrowski's
method with seventh-order convergence, J. Comput.
Appl. Math.190 (2007)111-115.
nonlinear equations, Appl. Math. Comput. (2007),
doi:10.1016/j.amc.2007.02.2008.
4. M.A. Noor, New classes of iterative methods for non-
linear equations, Appl. Math. Comput. (2007),
doi:10,1016/j.amc.2007.02.036.
5. C. Chun, A simply constructed third-order
modifications of Newton's method, J. Comput. Appl.
Math. (2007), doi: 10.1016/j.cam.2007.07.004.
6. Weihong Bi, Hongmin Ren, Qingbiao Wu, Three-step
iterative methods with eighth-order convergence for
soloving non-lnear equations, J. comput. Appl. Math
(2008)
7. Li Tai-fang ^{∗}, Li De-sheng, Xu Zhao-di, Fang Ya-
ling new iterative methods for non-linear equaitons,
Appl. Math. Comput 197 (2008) 755-759.
8. C. Chun, Y. Ham, Some fourth-order modifications of
Newton's Method., Appl. Math. Comput. (2007), doi:
10.1016/j.amc.2007.08.003.
1/2/2013

http://www.lifesciencesite.com 53 lifesciencej@gmail.com