See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/332153088

Modified Ostrowski Methods For Solving Nonlinear Equation

Article in International Journal of Pure and Applied Mathematics · February 2019

CITATIONS READS

0 445

3 authors:

Parimala Sivakumar Kalyanasundaram Madhu

Pondicherry Engineering College Khalifa University
8 PUBLICATIONS 22 CITATIONS 30 PUBLICATIONS 143 CITATIONS

SEE PROFILE SEE PROFILE

Jayakumar Jayaraman
Pondicherry Engineering College
30 PUBLICATIONS 254 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Development and Analysis of Higher Order Multipoint variants of Newton's method for solving nonlinear equations View project

Mathematical Modelling for epidemics View project

All content following this page was uploaded by Kalyanasundaram Madhu on 02 April 2019.

The user has requested enhancement of the downloaded file.

International Journal of Pure and Applied Mathematics
Volume 120 No. 8 2018, 59-67
ISSN: 1314-3395 (on-line version)
url: http://www.acadpubl.eu/hub/
Special Issue
http://www.acadpubl.eu/hub/

Modified Ostrowski Methods For Solving

Nonlinear Equation
S. Parimala1 , M. Kalyanasundaram2 and J. Jayakumar1
1
Pondicherry Engineering College, Pondicherry, India
2
Saveetha Engineering College, Chennai, India
[email protected], [email protected], [email protected]

Abstract
In this paper, we have obtained modified Ostrowski’s
methods for solving nonlinear equation with convergence
order eight and six. The efficiency indices of the proposed
methods are 1.516 and 1.565, which are better than the effi-
ciency index of Newton’s method (1.414). Also it is observed
from the numerical illustrations, the proposed methods take
less number of iterations than Newton’s method. Few other
methods are compared with the proposed methods, where
the number of iterations for those methods are either same
or more than the presented methods. Some examples are
given to illustrate the performance of the new methods.
AMS Subject Classification: 65H05 · 65D05

Key Words:Non-linear equation, Higher-order method,

Optimal order, Kung-Traub conjecture, Multi-point itera-
tion.

1 Introduction
The numerical solution of a nonlinear equation f (x) = 0 is a funda-
mental task in scientific computation. The most famous approach
is probably Newton’s method (NM):
f (xn )
xn+1 = xn − . (1)
f 0 (xn )

59
International Journal of Pure and Applied Mathematics Special Issue

It uses two function evaluations f and f 0 to achieve second-order

convergence. However, there are many iterative methods available
in the literature, some are specifically developed and analyzed for
solving nonlinear equations that improve classical methods, such as
Newton’s method (NM) and Halley’s iteration method etc. Some
of the iterative methods are given below: Halley’s iteration method
is given by
2f (xn )f 0 (xn )
xn+1 = xn − . (2)
2f 0 (x 2 00
n ) − f (xn )f (xn )

Arithmetic mean Newton’s method (AM) with cubic convergence

(see [15]) is given by
f (xn ) 2f (xn )
yn = xn − , xn+1 = xn − . (3)
f 0 (xn ) f 0 (xn ) + f 0 (yn )

This method is of order three with three function evaluations per

full iteration having efficiency index (EI) = 1.442. The following
method is known as Harmonic mean Newton’s method (HM) with
cubic convergence [4]:
 
f (xn ) 1 1
xn+1 = xn − + . (4)
2 f 0 (xn ) f 0 (yn )

The Newton-Steffensen method (SM) with cubic convergence [13]

is given by
f (xn )2
xn+1 = xn − . (5)
f 0 (x n )[f (xn ) − f (yn )]

Halley’s iteration method has third order convergence with three

function evaluations. Obviously, f 00 is difficult to calculate and
computationally more costly. Therefore, f 00 in Equation (2) is ap-
proximated using the finite difference; still, the convergence order
and total number of function evaluations are maintained [3]. Such
a third order method similar to Equation (2) after approximating
f 00 in Halley’s iteration method is given below:
f (xn ) 2βf (xn )
yn = xn − β , xn+1 = xn − , β 6= 0. (6)
f 0 (xn ) (2β − 1)f 0 (xn ) + f 0 (yn )

The efficiency index (EI) of an iterative method is measured

1
using the formula p d , where p is the local order of convergence
and d is the number of function evaluations per full iteration cycle
[11]. Kung–Traub [7] conjectured that the order of convergence of

60
International Journal of Pure and Applied Mathematics Special Issue

any multi-point without memory method with d function evalua-

tions cannot exceed the bound 2d−1 , the “optimal order”. New-
ton’s method is an optimal second order method, whereas Jarratt’s
method [6] is an example of an optimal fourth order method. Re-
cently, some optimal and non-optimal multi-point iterative methods
have been developed in [1, 2, 8, 9, 12, 14] and the references therein.
In this paper, we present two new three-step iterative methods
for solving nonlinear equations with eighth and sixth order conver-
gence by introducing Newton’s step in Ostrowski’s method [11] and
approximating the derivative to improve the efficiency index by re-
ducing the number of functional evaluations per step. In Section 2
new methods are given. In Section 3, we establish the convergence
order of the new methods. Section 4 confirms the theoretical re-
sults from computational tests and compare the new variants with
classical methods. Finally, Section 5 draws conclusion on our work.

2 Construction of the methods

Let us consider the following fourth order method,
h i
f (xn ) f (xn ) f (xn ) − f (yn )
yn = xn − , zn = xn − , (7)
f 0 (xn ) f 0 (xn ) f (xn ) − 2f (yn )

which is the well known Ostrowski’s method [11].

New Eighth order method (M1): This method is obtained by

introducing one more Newton’s step in the method (7), then we get
f (zn )
xn+1 = zn − . (8)
f 0 (zn )
Thus, we obtain a new three-step iterative method (8), which has
convergence order eight with five function evaluations. The effi-
ciency of the method (8) is found to be EI = 1.516, which shows
that it is higher than Newton’s method (1.414).

New Sixth order method (M2): In order to improve the ef-

ficiency index of the method (8), we modify it by using divided
difference and approximate f 0 (zn ) by a combination of already com-
puted function values as follows:

61
International Journal of Pure and Applied Mathematics Special Issue

f (zn ) − f (yn )
f 0 (zn ) ≈ = f [yn , zn ]. (9)
zn − yn
Using (9) in method (8), we obtain the following method:
f (zn )
xn+1 = zn − . (10)
f [yn , zn ]
This method (10) has EI = 1.565 which is better than Newton’s
method and the method (8). Thus, we have obtained a new three-
step iterative method (10) which has convergence order six with
four function evaluations.

3 Convergence Analysis
Theorem 1. Let a sufficiently smooth function f : D ⊂ R →
R has a simple root x∗ in the open interval D. If x0 is chosen in
a sufficiently small neighborhood of x∗ , then the method (8) is of
local eighth order convergence.

Proof. Let en = xn − α. Using the Taylor series, then we have

f (xn ) = f 0 (α)[en + c2 e2n + c3 e3n + c4 e4n + c5 e5n + c6 e6n + c7 e7n + . . .], (11)
0 0
f (xn ) = f (α)[1 + 2c2 en + 3c3 e2n + 4c4 e3n + 5c5 e4n + 6c6 e5n + . . .], (12)
f (k) (x∗ )
where ck = k!f 0 (x∗ )
, k ≥ 2. Now substituting (11), (12) in (3), then
yn = α + c2 e2n − 2(c22 − c3 )e3n + (4c32 − 7c2 c3 + 3c4 )e4n + (−8c42
+20c22 c3 − 6c23 − 10c2 c4 + 4c5 )e5n + (16c52 − 52c32 c3 + 33c2 c23
(13)
+28c22 c4 − 17c3 c4 − 13c2 c5 + 5c6 )e6n − 2(16c62 − 64c42 c3 − 9c33 + 36c32 c4
+6c24 + 9c22 (7c23 − 2c5 ) + 11c3 c5 + c2 (−46c3 c4 + 8c6 ) − 3c7 )e7n + . . .
Expanding f (yn ) about α and taking into account (13), we have
f (yn ) = f 0 (α)[c2 e2n − 2(c22 − c3 )e3n + (5c32 − 7c2 c3 + 3c4 )e4n − 2(6c42
− 12c22 c3 + 3c23 + 5c2 c4 − 2c5 )e5n + (28c52 − 73c32 c3 + 34c22 c4 − 17c3 c4
(14)
+ c2 (37c23 − 13c5 ) + 5c6 )e6n − 2(32c62 − 103c42 c3 − 9c33 + 52c32 c4 + 6c24
+c22 (80c23 − 22c5 ) + 11c3 c5 + c2 (−52c3 c4 + 8c6 ) − 3c7 )e7n + . . .].
Now, using (11), (12) and (14) in (7), then we have
zn = α + (c32 − c2 c3 )e4n − 2(2c42 − 4c22 c3 + c23 + c2 c4 )e5n + (10c52 − 30c32 c3
+ 12c22 c4 − 7c3 c4 + 3c2 (6c23 − c5 ))e6n − 2(10c62 − 40c42 c3 − 6c33 + 20c32 c4
(15)
+ 3c24 + 8c22 (5c23 − c5 ) + 5c3 c5 + c2 (−26c3 c4 + 2c6 ))e7n + . . . .

62
International Journal of Pure and Applied Mathematics Special Issue

Expanding f (zn ) about α and taking into account (15), we have

f (zn ) = f 0 (α)[(c32 − c2 c3 )e4n − 2(2c42 − 4c22 c3 + c23 + c2 c4 )e5n
(16)
+(10c52 − 30c32 c3 + 18c2 c23 + 12c22 c4 − 7c3 c4 − 3c2 c5 )e6n + . . .],

f 0 (zn ) = f 0 (α)[1 + 2c22 (c22 − c3 )e4n − (4c2 (2c42 − 4c22 c3 + c23 + c2 c4 ))e5n
(17)
+ (2c2 (10c52 − 30c32 c3 + 12c22 c4 − 7c3 c4 + 3c2 (6c23 − c5 )))e6n + . . .].
 
Using (15), (16) and (17) in (8), we get en+1 = c32 − c23 e8n + O(e9n ),
which shows eighth order convergence.

Theorem 2. Let a sufficiently smooth function f : D ⊂ R →

R has a simple root x∗ in the open interval D. If x0 is chosen in
a sufficiently small neighborhood of x∗ , then the method (10) is of
local sixth order convergence.
f (zn )−f (yn )
Proof. Replacing f 0 (zn ) in (8) using the formula f 0 (zn ) ≈ zn −yn
,
we obtain
f (zn ) − f (yn )
= f 0 (α)[(1 + c22 e2n + (−2c32 + 2c2 c3 ))e3n + (c2 (5c32
zn − yn (18)
−7c2 c3 + 3c4 ))e4n − (4c2 )(3c42 − 6c22 c3 + c23 + 3c2 c4 − c5 )e5n + . . .].
 
Using (18) in (10), we get en+1 = c52 − c32 c3 e6n + O(e7n ).

4 Computational examples
The contribution given in Section 3 is supported here through nu-
merical work. The effectiveness of the new methods are compared
to the existing methods such as Newton’s method (NM), Arith-
metic mean Newton method (AM) [15], Harmonic mean Newton
method (HM) [4], Newton-Steffensen method (SM) [13], method
proposed by Chun et al (CH) [2], method proposed by Hu et al
(HF) [5] and method proposed by Noor et al. (NK) [10] . The
methods presented in this paper are denoted as M1 and M2. Nu-
merical results on some test functions are given for the proposed
methods along with some existing methods. Numerical computa-
tions have been carried out in the Matlab software. The compu-
tational order of convergence is given by
ln |(xN − xN −1 )/(xN −1 − xN −2 )|
ρ= .
ln |(xN −1 − xN −2 )/(xN −2 − xN −3 )|

63
International Journal of Pure and Applied Mathematics Special Issue

The test functions and their simple zeros are given below:
3
f1 (x) = sin(2 cos x) − 1 − x2 + esin(x ) , x∗ = −0.7848959876612125352...
2
f2 (x) = xex − sin2 x + 3 cos x + 5, x∗ = −1.2076478271309189270...
∗
f3 (x) = sin(x) + cos(x) + x, x = −0.4566247045676308244...
f4 (x) = (x + 2)ex − 1, x∗ = −0.4428544010023885831...
f5 (x) = x2 + sin ( x5 ) − 14 , x∗ = 0.4099920179891371316...

Table 1: Comparison of Efficiency Indices (EI) and Optimality

Methods p d EI Optimal/Non-optimal
NM 2 2 1.414 Optimal
AM 3 3 1.442 Non-optimal
HM 3 3 1.442 Non-optimal
SM 3 3 1.442 Non-optimal
CH 6 4 1.565 Non-optimal
HF 7 5 1.475 Non-optimal
NK 7 5 1.475 Non-optimal
M1 8 5 1.516 Non-optimal
M2 6 4 1.565 Non-optimal

Table 1 shows the efficiency indices of the new methods with

some known methods. The results of comparisons are given in
Table 2. We observe that if the initial points are very close to
the root, then all the methods take less number of iterations and
produce least error. Also, the present new methods M1 and M2
have better efficiency when compared with other methods. For the
test function f2 (x), HF method diverges, whereas it converges for
the proposed new methods.

5 Conclusions
We have composed known Ostrowski’s method with Newton’s method
obtaining an eighth order method and then replaced the deriva-
tive by a divided difference obtaining a sixth order method with
efficiency index 1.565. It is clear that our new methods require
five function evaluations and four function evaluations per iterative
step respectively to obtain eighth order convergence and sixth order
convergence. The proposed methods M1 and M2 have better effi-
ciency, whereas M2 has the highest efficiency among the methods
discussed. Moreover, the proposed new method M2 requires less
cpu time for convergence when compared with other methods.

64
International Journal of Pure and Applied Mathematics Special Issue

Table 2: Numerical results for test functions

f (x) Methods x0 N ρ error cpu(s) x0 N ρ error cpu(s)
f1 (x) NM −1.2 7 1.99 1.57e-060 0.653 −0.5 8 1.99 6.42e-071 0.518
AM 5 3.00 6.56e-052 0.681 6 2.99 5.53e-147 0.780
HM 5 2.99 1.04e-051 0.647 5 3.00 4.40e-057 0.799
SM 5 2.99 3.12e-072 0.586 6 3.00 7.27e-130 0.710
CH 4 5.99 3.57e-124 0.715 4 5.99 7.65e-67 0.744
HF 4 6.99 1.09e-270 0.691 4 7.00 8.95e-064 0.792
NK 4 7.00 2.25e-297 0.726 4 7.00 8.36e-201 0.732
M1 4 7.99 2.64e-256 0.713 4 8.00 3.85e-254 0.743
M2 4 5.99 2.14e-165 0.628 4 5.99 7.76e-111 0.671
f2 (x) NM −1.7 9 2.00 4.38e-054 0.568 −0.8 9 2.00 2.13e-055 0.585
AM 7 3.00 4.32e-124 0.703 7 3.00 1.03e-086 0.624
HM 6 3.00 1.29e-072 0.577 6 3.00 9.90e-140 0.547
SM 6 2.99 2.79e-051 0.584 7 3.00 3.55e-149 0.617
CH 5 5.99 2.30e-245 0.715 4 5.99 1.34e-189 0.744
HF 4 7.05 1.51e-062 0.520 Div Div Div
NK 4 6.99 1.62e-101 0.584 4 7.00 2.27e-074 0.550
M1 4 8.00 1.28e-164 0.676 4 8.00 1.13e-276 0.637
M2 4 6.00 1.55e-63 0.670 4 6.00 1.65e-95 0.652
f3 (x) NM 0.5 7 1.99 1.08e-055 0.433 −1.5 7 1.99 1.16e-058 0.405
AM 5 2.99 2.75e-066 0.494 6 2.99 9.23e-149 0.517
HM 6 2.99 1.62e-137 0.512 6 2.99 5.06e-143 0.531
SM 5 3.00 1.30e-059 0.559 5 2.99 1.39e-107 0.459
CH 4 5.99 7.81e-78 0.599 4 5.99 2.92e-118 0.544
HF 4 7.00 3.06e-114 0.588 4 7.00 1.94e-253 0.490
NK 4 7.00 1.53e-208 0.673 4 6.99 9.43e-287 0.558
M1 4 7.99 1.95e-283 0.514 4 7.99 1.74e-181 0.550
M2 4 5.99 2.03e-137 0.472 4 5.99 9.29e-162 0.431
f4 (x) NM −0.2 7 1.99 7.24e-052 0.706 −0.9 8 1.99 3.10e-058 0.448
AM 5 2.99 2.67e-061 0.625 6 2.99 1.48e-093 0.490
HM 5 3.00 2.38e-082 0.721 5 2.99 1.83e-076 0.435
SM 5 2.99 1.75e-065 0.650 6 2.99 1.80e-109 0.483
CH 4 5.99 3.90e-156 0.638 4 5.99 1.092e-70 0.618
HF 4 6.99 1.86e-274 0.605 5 7.00 4.10e-275 0.542
NK 4 6.99 3.96e-305 0.728 4 7.00 3.86e-171 0.504
M1 3 7.97 1.31e-58 0.439 4 7.99 1.41e-298 0.616
M2 4 5.99 6.62e-190 0.561 4 5.99 1.31e-117 0.483
f5 (x) NM 0.8 8 1.99 3.21e-072 0.456 0.2 8 1.99 8.25e-076 0.398
AM 6 3.00 1.70e-136 0.505 6 2.99 2.60e-143 0.503
HM 5 2.99 2.35e-094 0.464 5 2.99 1.84e-098 0.441
SM 5 3.00 1.84e-136 0.502 6 3.00 2.82e-143 0.520
CH 4 5.99 3.46e-100 0.665 4 5.99 3.13e-85 0.559
HF 4 6.99 3.55e-159 0.476 4 7.00 2.70e-099 0.465
NK 4 6.99 2.50e-209 0.558 4 7.00 1.26e-209 0.511
M1 4 7.99 1.22e-286 0.502 4 7.99 5.31e-301 0.505
M2 4 5.99 2.37e-121 0.469 4 5.99 2.07e-127 0.462

References
[1] D.K.R. Babajee, K. Madhu, J. Jayaraman, A family of higher order multi-
point iterative methods based on power mean for solving nonlinear equa-
tions, Afr. Mat., 27(5) (2016), 865-876.

65
International Journal of Pure and Applied Mathematics Special Issue

[2] C. Chun, K. Yong-Il, Several new third-order iterative methods for solving
nonlinear equations, Acta Appl. Math., 109 ( 2010), 1053-1063.

[3] J.A. Ezquerro, M.A. Hernandez, A uniparametric halley-type iteration

with free second derivative, Int. J. Pure Appl. Math., 6 (2003), 99-110.

[4] H.H.H. Homeier, On Newton-type methods with cubic convergence, J.

Comp. Appl. Math., 176 (2005), 425-432.

[5] Y. Hu and L. Fang, A seventh-order convergent newton-type method for

solving nonlinear equations, Second International Conference on Compu-
tational Intelligence and Natural Computing, (2010),13-15.

[6] P. Jarratt, Some efficient fourth order multipoint methods for solving
equations, BIT, 9 (1969), 119-124.

[7] H.T. Kung, J.F. Traub, Optimal order of one-point and multipoint iter-
ation. J. Assoc. Comput. Mach., 21(4) (1974), 643-651.

[8] K. Madhu, J. Jayaraman, Higher order methods for nonlinear equations

and their basins of attraction, Mathematics, 4 (2016), (22).

[9] K. Madhu, J. Jayaraman, Two new families of iterative methods for

solving nonlinear equations, Tamsui Oxford Journal of Information and
Mathematical Sciences, 30(1) (2016), 25-38.

[10] M.A. Noor, W.A. Khan, K.I. Noor, E. Al-said, Higher-order iterative
methods free from second derivative for solving nonlinear equations, In-
ternational Journal of the Physical Sciences, 6(8) (2011), 1887-1893.

[11] A M. Ostrowski, Solutions of Equations and System of equations, Aca-

demic Press, New York, (1960).

[12] S. Parimala, K. Madhu, J. Jayaraman, Revisit of Ostrowski’s Method and

Two New Higher Order Methods for Solving Nonlinear equation, Int. J.
Math. And Appl., 6(2-A) 2018, 263-270.

[13] J.R. Sharma, A composite third order newton-steffensen method for solv-
ing nonlinear equations, Appl. Math. Comput, 169 (2005), 242-246.

[14] F. Soleymani, S.K. Khratti, S. Karimi Vanani, Two new classes of optimal
Jarratt-type fourth-order methods, Appl. Math. Lett., 25(5) (2011), 847-
853.

[15] S. Weerakoon, T.G.I. Fernando, A variant of Newton’s method with ac-

celerated third order convergence, Appl. Math. Lett., 13 (2000), 87-93.

66
67
 68

View publication stats