Applied Mathematics and Computation 215 (2009) 12881292

Contents lists available at ScienceDirect

Applied Mathematics and Computation

journal homepage: www.elsevier.com/locate/amc

A new fourth-order iterative method for nding multiple roots

of nonlinear equations
Li Shengguo a,*, Liao Xiangke b, Cheng Lizhi a,1
a
b

School of Science, National University of Defense Technology, Changsha 410073, China

School of Computer, National University of Defense Technology, Changsha 410073, China

a r t i c l e

i n f o

a b s t r a c t
In this paper, we present a new fourth-order method for nding multiple roots of nonlinear
equations. It requires one evaluation of the function and two of its rst derivative per iteration. Finally, some numerical examples are given to show the performance of the presented method compared with some known third-order methods.
2009 Elsevier Inc. All rights reserved.

Keywords:
Jarratt method
Multiple roots
Nonlinear equations
Iterative methods
Root nding

1. Introduction
Finding the roots of nonlinear equations is very important in numerical analysis and has many applications in engineering
and other applied sciences. In this paper, we consider iterative methods to nd a multiple root a of multiplicity m, i.e.,
f j a 0; j 0; 1; . . . ; m  1 and f m a0, of a nonlinear equation f x 0.
The modied Newtons method for multiple roots is quadratically convergent and it is written as [1]

xn1 xn  m

f xn
;
f 0 xn

which requires the knowledge of the multiplicity m.

In recent years, some modications of the Newton method for multiple roots have been proposed and analyzed, which
also require the knowledge of the multiplicity m. Most of these methods are of third-order convergence. For example, see
Traub [2], Hansen and Patrick [3], Victory and Neta [4], Dong [5,6], Osada [7], Neta [8], Chun and Neta [9], Chun et al.
[10], etc.
The third-order Chebyshevs method for nding multiple roots [2,8] is given by

xn1 xn 

m3  m
m2 f xn 2 f 0 xn
;
un 
2
2
f 0 xn 3

where

un

f xn
:
f 0 xn

* Corresponding author.
E-mail addresses: lsg_998@yahoo.com.cn (L. Shengguo), clzcheng@vip.sina.com (C. Lizhi).
1
This work is partially supported by National Natural Science Foundation of China (60573027).
0096-3003/$ - see front matter 2009 Elsevier Inc. All rights reserved.
doi:10.1016/j.amc.2009.06.065

L. Shengguo et al. / Applied Mathematics and Computation 215 (2009) 12881292

1289

The cubically convergent Halleys method which is a special case of the Hansen and Patricks method [3], is written as

xn1 xn 

f xn
m1 0
f xn
2m

00

 f x2fn f0 xnxn

The third-order Osada method [7], is written as

1
1
f 0 xn
xn1 xn  mm 1un m  12 00
:
2
2
f xn

Dong [5] has developed two third-order methods requiring two evaluations of the function and one of its rst derivative

8
p
< yn xn  mun ;

1m
f yn
: xn1 yn  m 1  p1
;
f 0 xn
m
8
< yn xn  un ;
: xn1 yn 

f yn

m1

m1
m

f xn f yn

un :

In [11], Neta and Johnson have proposed a fourth-order method requiring one-function and three-derivative evaluation
per iteration. This method is based on the Jarratt method [12] given by the iteration function

xn1 xn 

f xn
;
a1 f 0 xn a2 f 0 yn a3 f 0 gn

where

8
y xn  aun ;
>
< n
v n ff0xynn ;
>
:
gn xn  bun  cv n :

Neta and Johnson [11] give a table of values for the parameters a; b; c; a1 ; a2 ; a3 for several values of m. But, they do not give a
closed formula for general case.
Neta [13] has developed another fourth-order method requiring one-function and three-derivative evaluation per iteration. This method is based on Murakamis method [14] given by

xn1 xn  a1 un  a2 v n  a3 w3  wxn ;
where un is dened by (3),

v n ; yn

10

and gn are given by (9) and

f xn
;
f 0 gn
f xn
wxn
:
b1 f 0 xn b2 f 0 yn
w3 xn

11

A table of values for the parameters a; b; c; a1 ; a2 ; a3 ; b1 ; b2 for several values of m is also given by Neta [13].
In this paper, we propose a new fourth-order method for multiple roots, which requires one-function and two-derivative
evaluation per iteration, while the method proposed in [11] or [13] requires one-function and three-derivative evaluation
per iteration. So, the method presented here requires less function evaluations than these methods in [11,13]. Moreover,
the presented method here has a closed formula. The presented method is obtained by investigating the following iteration
functions

yn xn  hun ;
0

12

cf yn
xn1 xn  bff 0 xxnndf
0 y un ;
n

where h; b; c and d are parameters to be determined. When h 23 ; b  12 ; c  32 and d 3, (12) reduces to the fourth-order Jarratt method [15], which is dened by

8
< yn xn  23 un ;

: xn1 xn  1  32

f 0 yn f 0 xn
3f 0 yn f 0 xn

un :

13

By specially choosing the parameters, we get a new fourth-order method based on (12) in Section 2. Analysis of convergence
shows that the new method has fourth-order convergence. Finally, we use some numerical examples to compare the new
fourth-order method with some known third-order methods. From the results, we can see that the fourth-order method converges faster than these third-order methods and requires less function evaluations.

1290

L. Shengguo et al. / Applied Mathematics and Computation 215 (2009) 12881292

2. The fourth-order method

For simplicity, we dene

Aj

f mj a
;
f m a

j 1; 2; . . . ;

mh
:
m

14

Let a 2 R be a multiple root of multiplicity m of a sufciently smooth function f x; en xn  a and ^en yn  a, where yn is
dened in (12). Using the Taylor expansion of f xn and f yn about a, we have

f xn


f m a m 
en 1 C 1 en C 2 e2n C 3 e3n C 4 e4n Oe5n ;
m!

15

f yn


f m a m 
^en 1 C 1 ^en C 2 ^e2n C 3 ^e3n C 4 ^e4n O^e5n ;
m!

16

f 0 xn


f m a m1 
1 D1 en D2 e2n D3 e3n D4 e4n Oe5n ;
e
m  1! n

17

f 0 yn


f m a m1 
^e
1 D1 ^en D2 ^e2n D3 ^e3n D4 ^e4n O^e5n ;
m  1! n

18

un

en
1 C 1  D1 en C 2  D2 D21  C 1 D1 e2n C 3  D3 D1  C 1 D2 D2  C 2 C 1 D1  D21 D1 e3n
m
Oe4n ;
m!
A ; Dj
mj! j

19

m1!
A
mj1! j

where C j

and un is dened by (3).

Now, substituting ^en en  hun , (17)(19) into the error equation

en1 en 

bf 0 xn cf 0 yn
un ;
f 0 xn df 0 yn

20

and using a Computer Algebra System such as Maple, we can get

en1 K 1 en K 2 e2n K 3 e3n K 4 e4n Oe5n ;

21

where

K1 1 
K2

m  hb mclm
;
mm  h mdlm

"
m  h2 dh2 mh2  3mh m2 lm
m2 m 1h  m  mdlm 2

22

b

m 1h2  mh  m2 1 dlm
m2 m 1h  m  mdlm 2

clm A1 :

23

Before we list K 3 , we choose b and c to annihilate the coefcients K 1 and K 2

mm 1h2  mh  m2 1 dlm
;
hmh h  2m

mm  h2
m 1h2 m2  3mh

md:
hmh h  2m
hmh h  2mlm

24

In this case, we have

K 3 um; hA21 wm; hA2 ;

25

where

um; h

2m 12 h3  2mm 1m 5h2 m2 m 2m 7h  2m3 m 3

2m2 m 12 h  mh  m  mdlm mh h  2m

wm; h

2m 12 h2  mm2 9m 6h 2m2 3 m
2m2 m 12 mh h  2mh  m  mdlm

m 2h2  mm 4h 2m2
;
m2 m 1m 2mh h  2m

and A1 and A2 are dened in (14). Let um; h 0 and wm; h 0, we can get

dlm
26

27

L. Shengguo et al. / Applied Mathematics and Computation 215 (2009) 12881292

2m
;
m2

m

m
mh
m
d

:
m
m2

1291

28

Substituting h and d into b and c, we can get

m2
;
2
1 mm  2
c  m :
2
m
b

29

m2

Substituting (28) and (29) into (12), we can get a fourth-order method

8
2m
>
>
un ;
> yn xn 
>
m2
>
<
 m
2
1
m
mm  2 m2
f 0 yn  m2 f 0 xn
2
>
>
>
un ;
xn1 xn 
 m
>
>
m
:
f 0 xn  m2
f 0 yn

30

which satises the error equation

en1 K 4 e4n Oe5n ;

31

where

K4

m3 2m2 2m  2
3m4 m 13

A31 

A1 A2
mm 12 m 2

mA3

32

m 1m 3m 23

and Ai ; i 1; 2; 3 are dened in (14). This Eq. (31) indicates that the presented method (30) is of fourth-order convergence.
When m 1, the presented method (30) reduces to the Jarratt method for simply roots (13).
Per iteration the presented method requires one evaluation of the function, two of its rst derivative. We consider the
1
denition of efciency index [16] as pw , where p is the order of the method and w is the number of function evaluations
per iteration required by the method. If we assume that
p all the evaluations have the samepcost
as function one, we have that
1:587, which is better than 2 1:414 of Newtons method,
the
presented method has the efciency index 3 4p
p

3
3 1:442 of third-order methods ((2), (4)(7)) and 4 4 1:414 of the fourth-order methods proposed in [11] and [12].
3. Numerical example
In this section, we employ the presented method (30) (JM) to solve some nonlinear equations and compare it with the
Newton method (1) (NM) and some known third-order methods such as the Chebyshevs method (2) (CM), Halleys method
(4) (HM), Osadas method (5) (OM) and Dongs method (7) (DM).
All computations were done using the MAPLE using 1024 digit oating point arithmetics (Digits = 1024). We use the following functions, which have also been considered in [9,10], respectively.
m

x

1.3652300134140968457608068290

f2 x sin x  x 1

1.4044916482153412260350868178

f3 x x2  ex  3x 25

0.2575302854398607604553673049

f4 x cos x  x3

0.7390851332151606416553120876

f5 x xex  sin2 x 3 cos x 54

1.2076478271309189270094167584

3.0

2.8424389537844470678165859402

f x
3

f1 x x 4x  10
2

f6 x ex

7x30

 14

f7 x ex x  202

Table 1 shows the number of iterations required such that jf xj < 1060 and the values of jf xj after required iterations in
parentheses. From the results presented in Table 1, we can see that, for the functions we tested, the method proposed in this
paper converges faster than the third-order methods compared. As regards the number of function evaluations, the presented fourth-order method and the third-order methods compared require three function evaluations, while the modied
Newton method requires two function evaluations per iteration. From the data displayed in Table 2, it is seen that also in
terms of function evaluations, the fourth-order method is competitive to the third-order methods and the modied Newton
method.

1292

L. Shengguo et al. / Applied Mathematics and Computation 215 (2009) 12881292

Table 1
The number of iterations.
f x

x0

NM

CM

HM

OM

DM

JM

f1

0.3
3.0

54(5.36e88)
7(9.76e116)

9(1.35e134)
5(2.09e179)

58(3.16e73)
4(1.07e81)

27(5.68e92)
5(3.15e146)

121(3.92e165)
4(1.27e84)

6(2.12e185)
4(2.45e201)

f2

2.7
3.5

7(1.62e90)
7(9.20e76)

5(5.91e108)
5(2.14e114)

5(1.10e108)
5(2.22e137)

5(1.09e88)
5(1.66e92)

5(5.32e154)
5(8.24e125)

4(2.01e134)
4(1.25e91)

f3

1.8
3.0

5(1.15e120)
6(6.10e124)

3(8.40e77)
4(9.22e72)

4(4.12e148)
4(6.02e139)

3(7.98e131)
4(3.31e77)

4(3.79e166)
4(9.19e133)

3(3.48e105)
3(1.22e67)

f4

1.7
2.2

5(8.38e96)
5(4.17e81)

4(2.93e125)
4(1.55e76)

4(5.36e131)
4(5.54e90)

4(7.15e121)
4(2.43e63)

4(1.24e148)
4(3.06e108)

3(4.60e132)
3(4.95e82)

f5

3.5
2.5

17(1.78e74)
11(1.14e105)

11(5.58e63)
7(7.23e90)

9(4.66e121)
6(6.96e148)

12(5.55e73)
8(1.12e163)

9(1.61e88)
6(1.55e123)

9(2.08e163)
6(6.81e228)

f6

3.2
4.2

8(7.39e106)
22(4.37e74)

5(1.15e84)
15(3.68e122)

4(1.13e62)
12(5.76e149)

5(7.46e64)
16(2.07e104)

5(3.86e179)
12(2.49e114)

4(5.01e101)
11(2.97e68)

f7

3.0
5.5

5(2.16e70)
9(9.78e110)

4(1.36e173)
6(5.42e133)

3(4.33e72)
5(6.19e99)

4(1.99e144)
6(1.92e69)

3(7.88e69)
5(2.79e85)

3(2.19e146)
4(1.31e61)

Table 2
The number of function evaluations.
f x

x0

NM

CM

HM

OM

DM

JM

f1

0.3
3.0

108
14

27
15

174
12

81
15

363
12

18
12

f2

2.7
3.5

14
14

15
15

15
15

15
15

15
15

12
12

f3

1.8
3.0

10
12

9
12

12
12

9
12

12
12

9
9

f4

1.7
2.2

10
10

12
12

12
12

12
12

12
12

9
9

f5

3.5
2.5

34
22

33
21

27
18

36
24

27
18

27
18

f6

3.2
4.2

16
44

15
45

12
36

15
48

15
36

12
33

f7

3.0
5.5

10
18

12
18

9
15

12
18

9
15

9
12

Acknowledgements
The authors would like to thank the referees for their helpful comments and suggestions, which greatly improved the
original manuscript.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]

E. Schrder, ber unendlich viele Algorithmen zur Ausung der Gleichungen, Math. Ann. 2 (1870) 317365.
J.F. Traub, Iterative Methods for the Solution of Equations, Chelsea Publishing Company, New York, 1977.
E. Hansen, M. Patrick, A family of root nding methods, Numer. Math. 27 (1977) 257269.
H.D. Victory, B. Neta, A higher order method for multiple zeros of nonlinear functions, Int. J. Comput. Math. 12 (1983) 329335.
C. Dong, A basic theorem of constructing an iterative formula of the higher order for computing multiple roots of an equation, Math. Numer. Sinica 11
(1982) 445450.
C. Dong, A family of multipoint iterative functions for nding multiple roots of equations, Int. J. Comput. Math. 21 (1987) 363367.
N. Osada, An optimal multiple root-nding method of order three, J. Comput. Appl. Math. 51 (1994) 131133.
B. Neta, New third order nonlinear solvers for multiple roots, Appl. Math. Comput. 202 (2008) 162170.
C. Chun, B. Neta, A third-order modication of Newtons method for multiple roots, Appl. Math. Comput. (2009), doi:10.1016/i,amc.1009.01. 087.
C. Chun, H.J. Bae, B. Neta, New families of nonlinear third-order solvers for nding multiple roots, Comput. Math. Appl. 57 (2009) 15741582.
B. Neta, A.N. Johnson, High order nonlinear solver for multiple roots, Comput. Math. Appl. 55 (2008) 20122017.
P. Jarratt, Multipoint iterative methods for solving certain equations, Comput. J. 8 (1966) 398400.
B. Neta, Extension of Murakamis high order nonlinear solver to multiple roots, Int. J. Comput. Math., doi:10.1080/00207160802272263.
T. Murakami, Some fth order multipoint iterative formulae for solving equations, J. Inform. Process. 1 (1978) 138139.
I.K. Argyros, D. Chen, Q. Qian, The Jarratt method in Banach space setting, J. Comput. Appl. Math. 51 (1994) 103106.
W. Gautschi, Numerical Analysis: An introduction, Birkhuser, 1997.