An efficient Newton-type method with fifth-order convergence for solving nonlinear equations

An efficient Newton-type method with fifth-order convergence for solving nonlinear equations

Liang Fang^{I, II}; Li Sun^II; Guoping He^{II, III}

^IDepartment of Mathematics and System Science, Taishan University, 271021, Tai'an, China

^IIDepartment of Mathematics, Shanghai Jiaotong University, 200240, Shanghai, China

^IIICollege of Information Science and Engineering, Shandong University of Science and Technology, 266510, Qingdao, China, E-mail: [e-mail]

ABSTRACT

In this paper, we present an efficient Newton-like method with fifth-order convergence for nonlinear equations. The algorithm is free from second derivative and it requires four evaluations of the given function and its first derivative at each iteration. As a consequence, its efficiency index is equal to ⁴√5 which is better than that of Newton's method √2. Several examples demonstrate that the presented algorithm is more efficient and performs better than Newton's method.

Mathematical subject classification: 41A25, 65H05, 65D99.

Key words: Newton's method, iterative method, nonlinear equations, order of convergence.

1 Introduction

We consider iterative methods to find a simple root α of a nonlinear equation f(x) = 0, where f : D ⊆ → is a scalar function and it is sufficiently smooth in a neighborhood of α, and D is an open interval.

It is well known that classical Newton's method (named NM for simplicity)

is one of basic and important iterative methods for solving nonlinear equations and it converges quadratically [1] in a neighborhood of α.

Recently, some fifth-order iterative methods have been proposed and analyzed for solving nonlinear equations. These methods improve some classical methods such as Newton's method, Cauchy's method and Halley's method [2-6]. It is said that the improved methods are efficient and can compete with Newton's method. For more details about them see [2-6] and the references therein. It should be noted that most of these fifth-order methods depend on the second derivative which strictly reduces their practical applications.

Motivated by the recent activities in this direction, we present and analyze a new fifth-order iterative method in this paper.

The paper is organized as follows. In section 2, we propose the new method and analyze its convergence. Some numerical results are given in section 3. Finally, we draw some conclusions in section 4.

2 The new method and its convergence analysis

Consider the iteration scheme (FLM)

where y_n = x_n.

For the method defined by (2), we have the following result.

Theorem 2.1. Assume that the function f : D ⊆ → has a single root α Є D, where D is an open interval. Assume furthermore that f(x) is a sufficiently differentiable function in the neighborhood of α, then the order of convergence of the method FLM defined by (2) is five, and it satisfies the error equation

Proof. Let α be the simple root of f(x), c_k = f^(k)(α)/(k!f'(α)), k = 2,3,... and e_n = x_n - α. Consider the iteration function F(i) defined by

where

By some computations using Maple we can obtain

On the other hand, from the Taylor expansion of F(x_n) around α, we have

Therefore, we obtain e_n+1 = , which means that the method defined by (2) is fifth-order convergent.

□

Now, we consider efficiency index defined as , where p is the order of the method and ω is the number of function evaluations required by the method at each iteration. It's easy to see that our method FLM has the efficiency index which is better than that of Newton's method .

3 Numerical results

In this section, we employ the new method FLM to solve some nonlinear equations and compare it with NM, (PPM) [7] defined by

and two-step method (TSM)

where y_n = x_n.

All experiments were performed on a personal computer (IBM R40e) with Intel(R) Pentium(R) 4 CPU 2.00 GHz and 512 MB memory. The operating system was Windows XP (SP2) and the implementations were done in MATLAB 7.0.1 with double precision. We use the following stopping criteria for computer programs: (i)|f(x_n)| < 1.E - 14, and (ii) |x_n+1 - x_n| < 1.E - 14.

Displayed in Table 1 are the order of convergence of different methods. The number of iterations (IT) and function evaluations (NFE) required such that the stopping criterion is satisfied are listed in Table 2. Table 3 gives the comparison of number of operations needed for obtaining solutions.

We used the following test functions and display the approximate zeros x* found up to the 14^th decimal place.

f₁(x) = x³ + 4x² - 10, x* = 1.36523001341410.

f₂(x) = x³ -2x² + x - 1, x* = 1.75487766624670.

f₃(x) = - sin²(x) + 3cos(x) + 5, x* = -1.20764782713092.

From Table 1 and 3, we can see that, the algorithm FLM increase three order than classical Newton's method, but, only at the cost of small number of operations. We know, generally speaking, the increasing of order of convergence is more or less built on the cost of number of operations.

The computational results show that the present method requires less IT and NFE than NM, PPM and TSM as far as the numerical results are concerned. Therefore, the new method is of practical interest.

4 Conclusions

We present a new modification of Newton-type method with fifth-order convergence for solving nonlinear equations. At each iteration, it requires four evaluations of the function and its first derivative. Analysis of efficiency shows that the new algorithm is more efficient and it performs better than classical Newton's method and some other methods.

Acknowledgements. The authors would like to thank the anonymous referees for there constructive suggestions and criticisms that led to an improved presentation.

Received: 12/XII/07. Accepted: 06/VI/08.

#747/07.

• [1] A.M. Ostrowski, Solution of Equations in Euclidean and Banach Space Academic Press, New York (1973).
• [2] M.A. Noor and K.I. Noor, Fifth-order iterative methods for solving nonlinear equations Appl. Math. Comput., 188 (2007), 406-410.
• [3] J. Kou and Y. Li, The improvements of Chebyshev-Halley methods with fifth-order convergence Appl. Math. Comput., 188 (2007), 143-147.
• [4] M. Grau and M. Noguera, A Variant of Cauchy's Method with Accelerated Fifth-Order Convergence Applied Mathematics Letters, 17 (2004), 509-517.
• [5] J. Kou, Y. Li and X. Wang, Some modifications of Newton's method with fifth-order convergence Appl. Math. Comput., 209 (2007), 146-152.
• [6] J. Kou, The improvements of modified Newton's method. Appl. Math. Comput., 189 (2007), 602-609.
• [7] F.A. Potra and V. Pták, Nondiscrete induction and iterative processes. Research Notes in Mathematics, vol. 103, Pitman, Boston (1984).

Publication Dates

• Publication in this collection
  29 Oct 2008
• Date of issue
  2008

History

• Accepted
  06 June 2008
• Received
  12 Dec 2007