Ullah, M.

FIVE–POINT THIRTY–TWO OPTIMAL ORDER ITERATIVE METHOD

FOR SOLVING NON-LINEAR EQUATIONS

by

Malik Zaka ULLAH a and Fayyaz AHMAD b*

a Department of Mathematics, Faculty of Science, King Abdulaziz University,
Jeddah, Saudi Arabia
b Department of Mathematics, School of Science, National Textile University,

Faisalabad, Pakistan
Original scientific paper
https://doi.org/10.2298/TSCI21S2401U

A five-point thirty-two convergence order derivative-free iterative method to find

simple roots of non-linear equations is constructed. Six function evaluations are
performed to achieve optimal convergence order26–1 = 32 conjectured by Kung
and Traub [1]. Secant approximation to the derivative is computed around the
initial guess. High order convergence is attained by constructing polynomials of
quotients for functional values.
Key words: non-linear equations, derivative-free, iterative method, simple roots,
optimal order of convergence, Kung-Trub conjecture

Introduction
Non-linear problems are unavoidable in science and technology because the majority
of phenomena in nature have non-linear behavior. Non-linear equations belong to non-linear
problems, and their analytical solution is hard to find in general. Due to this limitation, iterative
methods are good candidates to find the approximation solution of non-linear equations. The
effectiveness of an iterative method for computing the solution of non-linear equations can be
measured by the Ostrowski efficiency index [2]. Optimal efficiency index can be obtained in
Kung and Traub conjecture (K-T) [1]. According to the K-T conjecture, an iterative method to
find a simple root of non-linear equations is optimal if it reaches an order of convergence p =
2n–1, where n is the number of functional evaluations per cycle.
The classical Newton’s method (NM) has quadratic convergence to find the simple
root, and it attains the optimal convergence order in the sense of K-T conjecture. In NM, we
use two functional evaluations and hence obtain optimal convergence order 22–1 = 2. Quadratic
convergence means when we are in the vicinity of the root, the error in the approximation of
root at step k is the square of the error in the approximation of root at step
k + 1. Generally, we observe very fact convergence in this case. But when we deal with a non-
linear equation, it is hard to compute an explicit derivative expression for the derivative then
NM is out of the question. In such a situation, we are looking for derivative-free optimal
convergence order methods. Steffensen’s method (SM) [3] is a derivative-free optimal
convergence method that also offers quadratic convergence. Fourth-order derivative-free
optimal order four iterative is established in [4, 5]. Other Steffensen type optimal order iterative
––––––––––––––
* Corresponding author, e-mail: [email protected]
 Ullah, M. Z

methods for finding simple roots can be found [6-8]. Authors use interpolation polynomial to
enhance the convergence order. Cordero et al. [9] constructed a technique to obtaining
derivative-free methods with optimal convergence order in the sense of K-T conjecture. In the
aforementioned article, the authors provided Steffensen type methods of optimal order four,
eight, and sixteen using Padé approximation. Interested readers can consult articles [9-11] and
references in them.
Thirty-two order optimal method
We constructed a five-point optimal convergence order thirty-two iterative method
for computing simple roots of smooth non-linear equations. Six function evaluations are used
to achieve the optimal convergence order in the sense of K-T conjecture. Let f be a smooth non-
linear function and α be a simple root of f(x) = 0, i.e., f(α) = 0, and f ( )  0. Our proposed
optimal order iterative method is FM32:
 Initialguess = x
 w= xaf
 x
 fw  f x
 df =
 w x
 fx
 y1 = x  df Convergence order = 2, Number of function evaluations = 2

 fy
 y2 = y1  p2 1 Convergence order = 4, Number of function evaluations = 3
 df
FM32 =
 fy
 y3 = y2  p3 2 Convergence order = 8, Number of function evaluations = 4
 df
 f
 y = y  p y3 Convergence order = 16, Number of function evaluations = 5
 4 3 4
df

 fy
 y5 = y4  p5 4 Convergence order = 32, Number of function evaluations = 6
 df

 x = y5

where f x = f ( x), f w = f (w), f y = f ( y1 ), f y = f ( y2 ), f y = f ( y3 ), f y = f ( y4 ),

t0 = f y /f w , t1 = f y /f x , t2 = f y /f w1, t3 = f y /f x , 2 t4 = f y /f y , 3t5 = f y /f w , t64 = f y /f x ,
t7 = f y1 /f y , t8 =1 f y /f y , t92 = f y /f w , 2t10 = f y /f x , 2 t11 1
= f y /f y 3, t12 = f y 3/f y ,
t13 = f y /f y , B =1/(1  adf ) and polynomial of ti ’s are:
3 1 3 2 4 4 4 1 4 2
4 3
p2 = 1  t0  t1

p3 =1  t0  t1  2t2  2t3  t4  t0t1  t03  t02t1  t0t12  t13

p4 =1  (t15t4  t14t4  2t13t42  t13t4  t13t8  t12t42  2t1t43  t13  t12t4  t12t8  t1t42  t12  t1t8  t1 

2t3  4t6 )B  (t15t4  t14t4  2t13t42  t13t4  t13t8  t13 )B2  (t15t4  t14t4  t13t42  t13t4  t13t8  t13 )B3 
Ullah, M. Z

t1  2t3  t4  4t6  2t7  t8  t8t1  t13  t42t1  t43  t4t13  t8t13  2t43t1 –

t42t13  t4t15  B4t4t15  B5t4t15

p5 =1  t17t42t8  t15t42t8  2t13t44t8  2t1t45t8  t1t44t8  3t1t43t82  2t1t13t43  2t1t43t8  t8t42t17 B6  t8t1 

t42t1  t4t13  t8t13  t4t15  2t43t1  t42t13  t8t4t13  t8t42t13  t13t4t15  2t13t8t13  t13  t8  t4 

t8t42t17 B7  8t1t13t4t8  t83t13  2t4t1  2t8t4  t13t4  2t13t8  t1t83  t82t4  t13t4t13  t8t4t15  t13t1 

t13t13  (t17t42t8  t16t42t8  t15t43t8  t15t42t8  2t14t43t8  3t13t44t8  t15t13t4  t15t4t8  t14t42t8  2t13t43t8 

2t13t42t82  t15t4  t14t13t4  t14t4t8  2t13t13t42  2t13t42t8  t13t83  t14t4  t13t13t4  2t13t13t8  2t13t42 

t13t4t8  t13t13  t13t4  t13t8  t13 )B2  (t17t42t8  t16t42t8  t15t43t8  t15t42t8  t14t43t8  2t13t44t8 

t15t13t4  t15t4t8  t14t42t8  t13t43t8  t15t4  t14t13t4  t14t4t8  t13t13t42  t13t42t8  t13t83  t14t4  t13t13t4 

2t13t13t8  t13t42  t13t4t8  t13t13  t13t4  t13t8  t13 )B3  (t17t42t8  t16t42t8  t15t43t8  t15t42t8  t15t13t4 

(t17t42t8  t16t42t8  t15t42t8  t15t13t4  t15t4t8  t15t4 )B5  2t13t4t1  2t13t8t1  t8t43t13  4t8t4t1 

t13t42t13  t15t4t8  t14t42t8  t8t45  t8t43  (t17t42t8  t16t42t8  t15t43t8  t15t42t8  t14t43t8  3t13t44t8 

t15t13t4  t15t4t8  t14t42t8  2t13t43t8  2t13t42t82  t12t44t8  2t1t45t8  t15t4  t14t13t4  t14t4t8  2t13t13t42 

2t13t42t8  t13t83  t12t43t8  t1t44t8  3t1t43t82  t14t4  t13t13t4  2t13t13t8  2t13t42  t13t4t8  t12t13t42 

t12t42t8  t12t83  2t1t13t43  2t1t43t8  t1t42t82  4t1t4t83  t13t13  t13t4  t13t8  t12t13t4  2t12t13t8 

t12t42  t12t4t8  t1t13t42  8t1t13t4t8  2t1t43  t1t42t8  2t1t4t82  t1t83  t13  t12t13  t12t4  t12t8 

2t1t13t4  2t1t13t8  t1t42  4t1t4t8  t12  t1t13  2t1t4  t1t8  t1 )B  t1t42t82  4t1t4t83  t1t13t42 

t1t42t8  2t1t4t82  4t13t8t4  t1  t13t43  2t83t4  t83  t13  t82t43  t43

Theorem 1 Let   I be a simple root of a sufficiently differentiable function
f : I   in an open interval I. If x0 is in the vicinity of α, then the method FM32 has
optimal convergence order 32.
Proof. Let ex be an error in x i.e. ex = x – α, c1 = f ( ), and c j = f ( j ) ( )/[j!f ( )] for
j = 1, 2, 3, 4, 5, 6. The proof of an order of convergence is established by using MAPLE. We
obtain the following error equation:
 Ullah, M. Z

e y = c12a 2  5c1 a  5c16a6  25c15a5  237 c14a 4  985c13a3  1895c12a 2  1683c1 a  561c28 
5

47 c15 a5 1177 c14 a 4 2239c13 a3 4127 c12 a 2

6c3 c16 a 6      1203c1 a  401c1 a  1c26 
2 6 3 3

169c13 a 3 2
4c4 c12 a 2  5c1 a  5c12 a 2  17 / 2c1 a  17 / 2c1 a  12 c25  11c14 a 4 c32  c3 
11

c5 778c12 a 2 2 3c5 1218c1 a 2 5c5 609c32 5c5

  c3   c3    c1 a  12 c2 4 
169 11 389 11 609 11 11
59c1 a 59 3
9c3 c4 c1 a  13 c12 a 2   c2  6c1 a  13 c12 a 2c33  8c33  1/ 48c3 c5 
9 9
1/ 48c42ac1  8c33  1/ 6c3 c5  1/ 6c42c22  3c32c4 c1 a  14 c2  c34c1 a  14 c1 a  1c14a4 

14c13a3  59c12a2  90c1 a  45c24  3c13c3 a3  22c12c3 a2  38c1 c3 a  19c3c22 

c4 c1 a  12 c2  c32c1 a  12 c12a2  5c1 a  5c22  c1 c3 a  c3c2 c112a12  46c111a11 

988c110a10  12724c19a9  105177c18a8  569656c17 a7  2045391c16a6  4906906c15a5 

7867690c14a4  8309000c13a3  5543491c12a2  2117286c1 a  352881c12a2  5c1 a 

241c111a11 5286c110 a10 134157 c19 a 9 213281c18 a8

52 c216  10c3 c112 a12     
5 5 10 2

2760832c17 a 7 9526764c16 a 6 22166539c15 a 5 34785142c14 a 4 14491245c13 a 3

     
5 5 5 5 2

47956127 c12 a 2 9126786c1 a 1521131

   c1 a  1c12 a 2  5c1 a  5c214  10c4 c1 a  12 c12 a 2 
10 5 5

167 c17 a 7 4609c16 a 6 6577 c15 a5 25525c14 a 4 54877 c13 a3

5c1 a  52 c18 a8      
5 10 2 2 2

65281c12 a 2 10069 13 254c111a11 2 5c5

  20138c1 a  c2  40c1 a  12 c112 a12 c32  c3  
2 2 5 2032

8937 c110 a10 2 58c5 137427 c19 a 9 2 1295c5 4183597c18 a8 2 21607 c5

 c3   c3   c3  
8 14895 10 274854 40 4183597

10379137 c17 a 7 2 112279c5 34557833c32 47843c5 6 19562206c15 a 5 2

 c3   a6  c1  c3 
20 20758274 20 5 5

219307 c5 120451369c14 a 4 2 1359007 c5 12387785c13 a3 2 56099c5

  c3   c3   a2 
39124412 20 240902738 2 9910228
Ullah, M. Z

92317 c5 40696989c32 2 7720692c1 a 2 58425c5 1286782c32 11685c5 12

  c1  c3    c2 
4 10 5 10294256 5 8

1211c17 a 7 c 31779c16 a 6 20c6 212635c15 a5

74c18 a8c3 c4  c3 c4  6  c3 c4   c3 c4 
37 2422 74 31779 74

153c6 391695c14 a 4 589c6 1630843c13 a3 1245c6 1907541c12 a 2

  c3 c4   c3 c4   c3 c4 
212635 37 783390 74 1630843 74
1465c6 584540c1 a 45c6 146135c3 c4 225c6
  c3 c4    c1 a  13 c12 a 2  5c1 a  5c211 
1907541 37 58454 37 74
804c19 a9 3 8c3 c5 7 c4 2 74918c18 a8 3 451c3 c5
85c1 a  13 c110 a10 c33  c3    c3  
17 1005 1340 85 37459

303c4 2 730386c17 a 7 3 5180c3 c5 2395c4 2 4159632c16 a 6 3 5307 c3 c5

 2em  c3    c3  
37459 85 365193 243462 85 346636

946c4 2 14573201c15 a 5 3 231696c3 c5 168672 c4 2 32104249 c14 a 4 3 520444 c3 c5

  c3    c3  
86659 85 14573201 14573201 85 32104249

384258c4 2 44447606c13 a3 3 364008c3 c5 38733c4 2 37478714 c12 a 2 3

  c3    c3 
32104249 85 22223803 3174829 85

308577 c3 c5 230964c4 2 3515575c1 a 3 11604c3 c5 8703c4 2 703115c33

   c3    
18739357 18739357 17 703115 703115 17

11604c3 c5 8703c4 2 10 6431c17 a 7 2 5c c 12c4 c5

  c2  205c18 a8c32 c4  c3 c4  3 6  
17 17 205 6431 6431

79214c16 a 6 2 46c3 c6 118c4 c5 498662c15 a 5 2 669c3 c6 894c4 c5

 c3 c4    c3 c4   
205 39607 39607 205 498662 249331

1753771c14 a 4 2 2497 c3 c6 6844c4 c5 3546088c13 a3 2 5185c3 c6 515c4 c5

 c3 c4    c3 c4   
205 1753771 1753771 205 3546088 126646

4083296c12 a 2 2 6045c3 c6 605c4 c5 497360 c1 a 2 37 c3 c6 26c4 c5

 c3 c4    c3 c4   
205 4083296 145832 41 24868 6217

124340c32 c4 185c3 c6 520 c4 c5 4343c3 a 7 c17 3 72 c3 c5

   c1 a  14 c29  106c18 a8c34  c3  
41 41 41 106 4343

92c4 2 63439c16 a 6 4 1536c32 c5 2062c3 c4 2 c c c2 223926c15 a 5 4

  c3    6 4  5  c3 
4343 106 63439 63439 63439 63439 53

6271c32 c5 2921c3 c4 2 c c c2 1698171c14 a 4 4 16926c32 c5 72911c3 c4 2

   6 4  5  c3   
223926 74642 37321 37321 106 566057 1698171
 Ullah, M. Z

56c6 c4 56c52 1794680c13 a 3 4 11073c32 c5 16149c3 c4 2 13c6 c4 13c5 2

   c3     
1698171 1698171 53 358936 358936 358936 358936

2115115c12 a 2 4 66223c32 c5 97342c3 c4 2 16c6 c4 16c52 1300190c1 a 4

 c3      c3 
53 2115115 2115115 423023 423023 53

20472c32 c5 30183c3 c4 2 5c6 c4 5c52 650095c34 25c6 c4 10236 c32 c5

       
650095 650095 130019 130019 106 106 53

30183c3 c4 2 25c52 6458c15 a 5 3 4c c 2

  c1 a  14 c28  267 c1 a  15 c33c4 a 6 c16  c3 c4  6 3 
106 106 267 3229

39c3 c4 c5 9c43 55816c14 a 4 3 25c6 c32 517 c3 c4 c5 67 c43 73052c13 a3 3

   c3 c4     c3 c4 
6458 6458 267 13954 55816 27908 89

37 c6 c32 2371c3 c4 c5 325c43 410678c12 a 2 3 436c6 c32 4723c3 c4 c5 1325c43

    c3 c4    
18263 219156 109578 267 205339 410678 410678

120440c1 a 3 13c6 c32 283c3 c4 c5 10c43 1415c3 c4 c5 260c6 c32 120440c33c4

 c3 c4       
89 6022 24088 3011 267 267 267

400c43 7 421c32 a5 c15 3 23c3 c5 22c4 2 12932c14 a 4 5 1013c33c5

 c2  78c16 a 6 c35  c3    c3  
267 13 842 421 39 25864

2035c32 c4 2 c c c2 3c 2 c 56093c13 a 3 5 5009c33c5 10417 c32 c4 2

  6 4  5 c3  4 5  c3   
25864 12932 12932 25864 39 112186 112186

7 c6 c4 7 c52 21c4 2 c5 109934c12 a 2 5 937 c33c5 26c32 c4 2 8c6 c4 8c5 2

  c3   c3     c3 
56093 56093 112186 39 19988 263 54967 54967

12c4 2 c5 32755c1 a 5 3121c33c5 6611c32 c4 2 c6 c4 c2 3c 2 c

  c3     5 c3  4 5 2em 
54967 13 65510 65510 6551 6551 13102

32755c35 3121c33c5 6611c32 c4 2 5c52 5c6 c4 5c 2 c

     c3  4 5 c1 a  15 c26 
39 78 78 39 39 26

1385c13 a3 3 c c 2 18c3 c4 c5 19c43 13567 c12 a 2 3

166c3 c33c4 a 4 c14  c3 c4  6 3    c3 c4 
83 554 1385 2770 166

34c6 c32 260c3 c4 c5 149c43 10797 c1 a 3 29c6 c32 224c3 c4 c5 130c43

    c3 c4    
13567 13567 13567 83 10797 10797 10797

10797 c33c4 29c6 c32 112c3 c4 c5 65c43 761c33 a3c13 3

    c1 a  16 c25  35c14 a 4 c36  c3 
166 166 83 83 35
Ullah, M. Z

28c3 c5 66c4 2 4286c12 a 2 6 112c5 c34 277c4 2 c33 c6 c4 c2 3c c 2c

   c3     5 c32  3 4 5 
761 761 35 2143 2143 4286 4286 4286

c4 4 1410c1 a 6 196c5 c34 488c4 2 c33 c6 c4 c2 c c 2c c4

  c3     5 c32  3 4 5  4 
4286 7 3525 3525 3525 3525 1175 3525

705c36 28c5 c34 488c4 2 c33 3c c 2 c

    1 / 35c52  1 / 35c6 c4 c32  3 4 5  1 / 35c4 4 c1 a  16 c2 4 
7 5 35 35

43c1 a 3 c c 2 9c c c 3c 3 43c33c4 1 9c c c
50c12 a 2 c33c4  c3 c4  6 3  3 4 5  4   c6 c32  3 4 5 
5 430 430 215 5 50 50

3c43 2 94c1 a 3 2c3 c5 11c4 2 94c33

 c3 c1 a  17 c23  9c12 a 2 c33  c3     4 / 9c3 c5 
25 9 47 94 9

11c4 2 4
 c3 c1 a  17 c2 2  6c36 c4 c1 a  18 c2  c38 c1 a  18 e32  O (e33 )
9

Numerical results
To validate our proposed iterative method FM32, we adopt the following definition of
approximated computational order of convergence (ACOC):
log[| f ( xk  2 )/f ( xk 1 ) |]
ACOC =
log[| f ( xk 1 )/f ( xk ) |]
where xi is the sequence of approximations to root α, which is computed by using FM32. We
choose six non-linear functions to compute the computational order of convergence. Table 1
list the functions and their simple roots. In tab. 2, we have shown that the approximated
computational order of convergence is thirty-two, which supports our theoretically computed
order of convergence. The effect of the parameter a of the accuracy of root approximation is
depicted in figs. 1(a) and 1(b). The higher value of the graph corresponds to higher accuracy.
We conclude that the smaller parameter values a give us better accuracy in approximating the
first-order derivative. The smaller value a improves the accuracy of approximation of the root.

Table 1. Non-linear equations and simple roots

Non-linear functions Simple roots

f1( x) = x cos( x )  x ln(1  x  π)  π

4 2 5 2 2
1.77245

f2 ( x) = x exp( x2 )  sin( x)2  3cos( x) – 0.8077428

f3 ( x) = x3  10 2.154434

f 4 ( x) = sin( x)2  x2  1 1.40449

f5 ( x) = ( x  2)exp( x)  1 – 0.44285

f6 ( x) = ( x  1)3  2 2.25992
 Ullah, M. Z

Table 2. Approximated computational order of convergence, a = 00.01

f1 f2 f3 f4 f5 f6
Initial guess 1.5 –1.2 2.15 1.40 –0.44 2.25
f(x1) 2.36e-1 4.85 6.16e-2 1.11e-2 4.70e-3 4.69e-2
f(x2) 3.82e-25 1.41e-5 5.45e-72 2.65e-62 6.30e-68 1.68e-51
f(x3) 5.16e-849 2.96e-171 8.83e-2314 2.28e-2102 8.65e-2375 4.50e-1733
ACOC 31.9456 29.9191 32.0011 32.0023 31.9991 32.0058
32.0000 31.9999 32.0000 32.0001 32.0000 32.0001

Figure 1. Effect a on the accuracy of root approximation; (a) effect of the parameter a on the accuracy
of root approximation f1(x) = 0 and (b) effect of the parameter a on the accuracy of root approximation
f4(x) = 0

Conclusions
Multi-point numerical iterative methods are numerically stable to compute the simple
roots. We proposed a five-point numerical iterative method to compute the simple roots of non-
linear equations. Our proposed method is numerically stable because it uses the information at
six different points. It has thirty-two order of convergence that is optimal in the sense of K-T
conjecture. Our numerical iterative method does not require the evaluation of the analytical
derivative of the functions. We use the finite difference technique to approximate the derivative.
The discretization of the first-order derivative introduces a parameter a. The small value of the
parameter a can better approximate the first-order derivative. The parameter a can change the
dynamics of the numerical iterative method by changing its path of convergence. To validate
our numerical iterative method, we choose six non-linear equations to find the simple roots. In
all numerical simulations, the value of the parameter a was 0.01. All the initial guesses are
reported in tab. 2. We performed three iterations in all numerical simulations.
Acknowledgment
This project was funded by the Deanship of Scientific Research (DSR) at King
Abdulaziz University, Jeddah, under grant no. G: 351-130-1439. The authors, therefore,
acknowledge with thanks DSR for technical and financial support.
Ullah, M. Z

References
[1] Kung, H. T., Traub, J. F., Optimal Order of One-Point and Multi-Point Iteration, Journal of the Association
for Computing Machinery, 21 (1974), 4, pp. 643-651
[2] Ostrowski, A. M., Solutions of Equations and Systems of Equations, Academic Press, New York, USA,
1966
[3] Ortega, J. M., Rheinboldt, W. C., Iterative Solutions of Non-linear Equations in Several Variables,
Academic Press, New York, USA, 1970
[4] Ren, H., et al., A Class of Two-Step Steffensen Type Methods with Fourth-Order Convergence, Applied
Mathematics and Computation, 209 (2009), 2, pp. 206-210
[5] Feng, X., He, Y., High Order Iterative Methods without Derivatives for Solving Non-linear Equations,
Applied Mathematics and Computation, 186 (2007), 2, pp. 1617-1623
[6] Cordero, A., Torregrosa, J. R., A Class of Steffensen type Methods with Optimal Order of Convergence,
Applied Mathematics and Computation, 217 (2011), 19, pp. 7653-7659
[7] Liu, Z., et al., A Variant of Steffensen’s Method of Fourth-Order Convergence and Its Applications,
Applied Mathematics and Computation, 216 (2010), 7, pp. 1978-1983
[8] Zheng, Q., et al., A Steffensen-Like Method and Its Higher-Order Variants, Applied Mathematics and
Computation, 214 (2009), 1, pp. 10-16
[9] Cordero, A., et al., A New Technique to Obtain Derivative-Free Optimal Iterative Methods for Solving
Non-linear Equations, Journal of Computational and Applied Mathematics, 252 (2013), Nov., pp. 95-102
[10] Soleymani, F., et al., A Class of Three-Step Derivative-Free Root-Solvers with Optimal Convergence
Order, Journal of Applied Mathematics, 2012 (2012), ID 568740
[11] Cordero, A., Torregrosa, J. R., Variants of Newton’s Method Using Fifth-Order Quadrature Formulas,
Applied Mathematics and Computation, 190 (2007), 1, pp. 686-698

Paper submitted: July 15, 2021 © 2021 Society of Thermal Engineers of Serbia.
Paper revised: July 28, 2021 Published by the Vinča Institute of Nuclear Sciences, Belgrade, Serbia.
Paper accepted: August 2, 2021 This is an open access article distributed under the CC BY-NC-ND 4.0 terms and conditions.