International Scholarly Research Notices

International Scholarly Research Notices / 2011 / Article

Research Article | Open Access

Volume 2011 |Article ID 693787 | 12 pages | https://doi.org/10.5402/2011/693787

Eighth-Order Iterative Methods without Derivatives for Solving Nonlinear Equations

Academic Editor: A.-V. Phan
Received29 Mar 2011
Accepted17 May 2011
Published06 Sep 2011

Abstract

A new family of eighth-order derivative-free methods for solving nonlinear equations is presented. It is proved that these methods have the convergence order of eight. These new methods are derivative-free and only use four evaluations of the function per iteration. In fact, we have obtained the optimal order of convergence which supports the Kung and Traub conjecture. Kung and Traub conjectured that the multipoint iteration methods, without memory based on 𝑛 evaluations, could achieve optimal convergence order 2𝑛−1. Thus, we present new derivative-free methods which agree with Kung and Traub conjecture for 𝑛=4. Numerical comparisons are made to demonstrate the performance of the methods presented.

1. Introduction

Consider iterative methods for finding a simple root 𝛼 of the nonlinear equation𝑓(𝑥)=0,(1.1) where𝑓∶𝐷⊂ℝ→ℝ is a scalar function on an open interval 𝐷, and it is sufficiently smooth in a neighbourhood of 𝛼.It is well known that the techniques to solve nonlinear equations have many applications in science and engineering. In this paper, a new family of three-point derivative-free methods of the optimal order eight is constructed by combining optimal two-step fourth-order methods and a modified third step. In order to obtain these new derivative-free methods, we replace derivatives with suitable approximations based on divided difference. In fact, it is well known that the various methods have been used in order to approximate the derivatives by the Newton interpolation, the Hermite interpolation, the Lagrange interpolation, and ration function [1, 2].

The prime motive of this study is to develop a class of very efficient three-step derivative-free methods for solving nonlinear equations. The eighth-order methods presented in this paper are derivative-free and only use four evaluations of the function per iteration. In fact, we have obtained the optimal order of convergence which supports the Kung and Traub conjecture. Kung and Traub conjectured that the multipoint iteration methods, without memory based on 𝑛 evaluations, could achieve optimal convergence order 2𝑛−1.Thus, we present new derivative-free methods which agree with the Kung and Traub conjecture for 𝑛=4.In addition, these new eighth-order derivative-free methods have an equivalent efficiency index to the established Kung and Traub eighth-order derivative-free method presented in [3]. Furthermore, the new eighth-order derivative-free methods have a better efficiency index than the three-step sixth-order derivative-free methods presented recently in [4, 5], and in view of this fact, the new methods are significantly better when compared with the established methods. Consequently, we have found that the new eighth-order derivative-free methods are consistent, stable, and convergent.

This paper is organised as follows. In Section 2 we construct the eighth-order methods that are free from derivatives and prove the important fact that the methods obtained preserve their convergence order. In Section 3 we will briefly state the established Kung and Traub method in order to compare the effectiveness of the new methods. Finally, in Section 4 we demonstrate the performance of each of the methods described.

2. Methods and Convergence Analysis

In this section we will define a new family of eighth-order derivative-free methods. In order to establish the order of convergence of these new methods, we state the three essential definitions.

Definition 2.1. Let 𝑓(𝑥) be a real function with a simple root 𝛼, and let {𝑥𝑛} be a sequence of real numbers that converge towards 𝛼.The order of convergence 𝑚 is given by limğ‘›â†’âˆžğ‘¥ğ‘›+1−𝛼𝑥𝑛−𝛼𝑚=𝜁≠0,(2.1) where 𝜁 is the asymptotic error constant and 𝑚∈ℝ+.

Definition 2.2. Suppose that 𝑥𝑛−1,𝑥𝑛, and 𝑥𝑛+1 are three successive iterations closer to the root 𝛼 of (1.1). Then, the computational order of convergence [6] may be approximated by |||𝑥COC≈ln𝑛+1𝑥−𝛼𝑛−𝛼−1||||||𝑥ln𝑛𝑥−𝛼𝑛−1−𝛼−1|||,(2.2) where 𝑛∈ℕ.

Definition 2.3. Let 𝛽 be the number of function evaluations of the new method. The efficiency of the new method is measured by the concept of efficiency index [7, 8] and defined as 𝜇1/𝛽,(2.3) where 𝜇 is the order of the method.

2.1. The Eighth-Order Derivative-Free Method (RT)

We consider the iteration scheme of the form𝑦𝑛=𝑥𝑛−𝑓𝑥𝑛𝑥𝑓′𝑛,𝑧𝑛=𝑦𝑛−𝑓𝑦𝑛𝑦𝑓′𝑛,𝑥𝑛+1=𝑧𝑛−𝑓𝑧𝑛𝑧𝑓′𝑛.(2.4)

This scheme consists of three steps in which the Newton method is repeated. It is clear that formula (2.4) requires six evaluations per iteration and has an efficiency index of 81/6=1.414, which is the same as the classical Newton method. In fact, scheme (2.4) does not increase the computational efficiency. The purpose of this paper is to establish new derivative-free methods with optimal order; hence, we reduce the number of evaluations to four by using some suitable approximation of the derivatives. To derive higher efficiency index, we consider approximating the derivatives by divided difference method. Therefore, the derivatives in (2.4) are replaced by𝑥𝑓′𝑛𝑤≈𝑓𝑛,𝑥𝑛=𝑓𝑤𝑛𝑥−𝑓𝑛𝑤𝑛−𝑥𝑛=𝑓𝑤𝑛𝑥−𝑓𝑛𝑓𝑥𝑛,𝑦𝑓′𝑛≈𝑓𝑥𝑛,𝑦𝑛𝑓𝑤𝑛,𝑦𝑛𝑓𝑤𝑛,𝑥𝑛,𝑧𝑓′𝑛≈𝑓𝑦𝑛,𝑧𝑛𝑥−𝑓𝑛,𝑦𝑛𝑥+𝑓𝑛,𝑧𝑛.(2.5) Substituting (2.5) into (2.4), we get𝑤𝑛=𝑥𝑛𝑥+𝛽𝑓𝑛,𝑦𝑛=𝑥𝑛−𝑓𝑥𝑛2𝑓𝑤𝑛𝑥−𝑓𝑛,𝑧𝑛=𝑦𝑛−𝑓𝑤𝑛,𝑥𝑛𝑓𝑦𝑛𝑓𝑥𝑛,𝑦𝑛𝑓𝑤𝑛,𝑦𝑛,𝑥𝑛+1=𝑧𝑛−𝑓𝑧𝑛𝑓𝑦𝑛,𝑧𝑛𝑥−𝑓𝑛,𝑦𝑛𝑥+𝑓𝑛,𝑧𝑛.(2.6)

The first step of formula (2.6) is the classical Steffensen second-order method [9], and the second step is the new fourth-order method. Furthermore, we have found that the third step does not produce an optimal order of convergence. Therefore, we have introduced two weight functions in the third step in order to achieve the desired eighth-order derivative-free method. The two weight functions are expressed as𝑓𝑧1−𝑛𝑓𝑤𝑛−1,𝑦1+2𝑓𝑛3𝑓𝑤𝑛2𝑓𝑥𝑛−1.(2.7) Then the iteration scheme (2.4) in its final form is given as𝑤𝑛=𝑥𝑛𝑥+𝛽𝑓𝑛,𝑦𝑛=𝑥𝑛−𝑓𝑥𝑛2𝑓𝑤𝑛𝑥−𝑓𝑛,𝑧𝑛=𝑦𝑛−𝑓𝑤𝑛,𝑥𝑛𝑓𝑦𝑛𝑓𝑥𝑛,𝑦𝑛𝑓𝑤𝑛,𝑦𝑛,𝑥𝑛+1=𝑧𝑛−𝑓𝑧1−𝑛𝑓𝑤𝑛−1𝑦1+2𝑓𝑛3𝑓𝑤𝑛2𝑓𝑥𝑛−1𝑓𝑧𝑛𝑓𝑦𝑛,𝑧𝑛𝑥−𝑓𝑛,𝑦𝑛𝑥+𝑓𝑛,𝑧𝑛,(2.8) where 𝑛∈ℕ, 𝛽∈ℝ+, provided that the denominators in (2.8) are not equal to zero.

Thus the scheme (2.8) defines a new family of multipoint methods with two weight functions. To obtain the solution of (1.1) by the new derivative-free methods, we must set a particular initial approximation𝑥0, ideally close to the simple root. In numerical mathematics it is very useful and essential to know the behaviour of an approximate method. Therefore, we will prove the order of convergence of the new eighth-order method.

Theorem 2.4. Assume that the function𝑓∶𝐷⊂ℝ→ℝ for an open interval D has a simple root 𝛼∈𝐷. Letting 𝑓(𝑥) be sufficiently smooth in the interval D and the initial approximation 𝑥0 is sufficiently close to 𝛼, then the order of convergence of the new derivative-free method defined by (2.8) is eight.

Proof. Let 𝛼 be a simple root of 𝑓(𝑥), that is, 𝑓(𝛼)=0 and ğ‘“î…ž(𝛼)≠0, and the error is expressed as 𝑒=𝑥−𝛼.(2.9)
Using the Taylor expansion, we have 𝑓𝑥𝑛=𝑓(𝛼)+ğ‘“î…ž(𝛼)𝑒𝑛+2−1ğ‘“î…žî…ž(𝛼)𝑒2𝑛+6−1ğ‘“î…žî…žî…ž(𝛼)𝑒3𝑛+24−1𝑓𝑖𝑣(𝛼)𝑒4𝑛+⋯.(2.10)
Taking 𝑓(𝛼)=0 and simplifying, expression (2.10) becomes 𝑓𝑥𝑛=𝑐1𝑒𝑛+𝑐2𝑒2𝑛+𝑐3𝑒3𝑛+𝑐4𝑒4𝑛+⋯,(2.11) where 𝑛∈ℕ and 𝑐𝑘=𝑓(𝑘)(𝛼)(𝑘!)for𝑘=1,2,3,4,….(2.12)
Expanding the Taylor series of 𝑓(𝑤𝑛) and substituting 𝑓(𝑥𝑛) given by (2.11), we have 𝑓𝑤𝑛=𝑐11+𝑐1𝛽𝑒𝑛+3𝛽𝑐1𝑐2+𝛽2𝑐21𝑐2+𝑐2𝑒2𝑛+⋯.(2.13)
Substituting (2.11) and (2.13) in expression (2.8) gives us 𝑦𝑛−𝛼=𝑥𝑛𝑥−𝛼−𝛽𝑓𝑛2𝑓𝑤𝑛𝑥−𝑓𝑛=𝑐2𝑐1𝛽𝑐1𝑒+12𝑛+⋯.(2.14)
The expansion of 𝑓(𝑦𝑛) about 𝛼 is given as 𝑓𝑦𝑛=𝑐1𝑦𝑛−𝛼+𝑐2𝑦𝑛−𝛼2+𝑐3𝑦𝑛−𝛼3+⋯.(2.15)
Simplifying (2.15), we have 𝑓𝑦𝑛=𝑐2𝑐1𝑒𝛽+12𝑛+𝛽𝑐31𝑐3−2𝑐22+3𝛽𝑐21𝑐3+2𝑐1𝑐3−𝛽2𝑐21𝑐22−2𝛽𝑐1𝑐22𝑐1𝑒3𝑛+⋯.(2.16)
The expansion of the particular term used in (2.8) is given as 𝑓𝑤𝑛,𝑥𝑛=𝑓𝑤𝑛𝑥−𝑓𝑛𝑤𝑛−𝑥𝑛=𝑐1+2𝑐2+𝛽𝑐1𝑐2𝑒𝑛+3𝑐3+3𝛽𝑐1𝑐3+𝛽2𝑐21𝑐3+𝛽𝑐22𝑒2𝑛𝑓𝑤+⋯,𝑛,𝑦𝑛=𝑓𝑤𝑛𝑦−𝑓𝑛𝑤𝑛−𝑦𝑛=𝑐1+𝑐2+𝛽𝑐1𝑐2𝑒𝑛+𝛽𝑐21𝑐3+𝑐22+2𝛽𝑐1𝑐22+2𝛽𝑐21𝑐3+𝑐31𝑐3𝑐1𝑒2𝑛𝑓𝑥+⋯,𝑛,𝑦𝑛=𝑓𝑥𝑛𝑦−𝑓𝑛𝑥𝑛−𝑦𝑛=𝑐1+𝑐2𝑒𝑛+𝑐1𝑐3+𝑐22+𝛽𝑐1𝑐22𝑐1𝑒2𝑛𝑓𝑤+⋯,𝑛,𝑥𝑛𝑓𝑥𝑛,𝑦𝑛𝑓𝑤𝑛,𝑦𝑛=1𝑐1+𝛽𝑐21𝑐3−3𝑐22−3𝛽𝑐1𝑐22+𝑐1𝑐3𝑐31𝑒2𝑛+⋯.(2.17)
Substituting appropriate expressions in (2.8), we obtain 𝑧𝑛−𝛼=𝑦𝑛𝑓𝑤−𝛼−𝑛,𝑥𝑛𝑓𝑦𝑛𝑓𝑥𝑛,𝑦𝑛𝑓𝑤𝑛,𝑦𝑛.(2.18)
The Taylor series expansion of 𝑓(𝑧𝑛) about 𝛼 is given as 𝑓𝑧𝑛=𝑐1𝑧𝑛−𝛼+𝑐2𝑧𝑛−𝛼2+𝑐3𝑧𝑛−𝛼3+⋯.(2.19)
Simplifying (2.18), we have 𝑓𝑧𝑛=2𝑐32−𝑐1𝑐2𝑐3+4𝛽𝑐1𝑐32+2𝛽2𝑐21𝑐32−2𝛽𝑐21𝑐2𝑐3−𝑐31𝑐2𝑐3𝑐21𝑒4𝑛+⋯.(2.20)
In order to evaluate the essential terms of (2.8), we expand term by term 𝑓𝑦𝑛,𝑧𝑛=𝑓𝑦𝑛𝑧−𝑓𝑛𝑦𝑛−𝑧𝑛=𝑐1+𝛽𝑐1𝑐22+𝑐22𝑐1𝑒2𝑛𝑓𝑥+⋯,𝑛,𝑧𝑛=𝑓𝑥𝑛𝑧−𝑓𝑛𝑥𝑛−𝑧𝑛=𝑐1+𝑐2𝑒𝑛+𝑐3𝑒2𝑛+⋯.(2.21)
Collecting the above terms, 𝑓𝑦𝜓=𝑛,𝑧𝑛𝑥−𝑓𝑛,𝑦𝑛𝑥+𝑓𝑛,𝑧𝑛=−1=1𝑐1+𝑐2𝑐3+𝛽𝑐1𝑐2𝑐3𝑐31𝑒3𝑛𝑓𝑧+⋯,𝜔=1−𝑛𝑓𝑤𝑛−1=1−𝛽𝑐21𝑐2𝑐3−2𝛽𝑐1𝑐32+𝑐1𝑐2𝑐3−2𝑐32𝑐31𝑒3𝑛𝑦+⋯,𝜉=1+2𝑓𝑛3𝑓𝑤𝑛2𝑓𝑥𝑛−1=1−2𝛽𝑐1𝑐32+2𝑐32𝑐31𝑒3𝑛+⋯,𝜔𝜉=1−𝛽𝑐1𝑐2𝑐3+𝑐2𝑐3𝑐21𝑒3𝑛=1+⋯,𝜓𝜔𝜉𝑐1+−𝛽2𝑐21𝑐42+6𝛽𝑐21𝑐22𝑐3+3𝛽2𝑐31𝑐22𝑐3+𝛽𝑐1𝑐42−2𝛽𝑐31𝑐2𝑐4+2𝑐42−𝑐21𝑐2𝑐4+3𝑐1𝑐22𝑐3−𝛽2𝑐41𝑐2𝑐4𝑐51×𝑒4𝑛+⋯.(2.22)
Substituting appropriate expressions in (2.8), we obtain 𝑒𝑛+1=𝑧𝑛𝑧−𝛼−𝜓𝜔𝜉𝑓𝑛.(2.23)
Simplifying (2.23), we obtain the error equation 𝑒𝑛+1=−𝑐1−730𝛽2𝑐21𝑐22−2𝑐21𝑐42𝑐4+26𝛽𝑐1𝑐72+14𝛽3𝑐31𝑐72−2𝑐21𝑐32𝑐23+9𝛽2𝑐31𝑐52𝑐3−12𝛽2𝑐41𝑐42𝑐4−8𝛽3𝑐51𝑐42𝑐4+3𝛽𝑐21𝑐42𝑐3−8𝛽𝑐31𝑐42𝑐4+9𝛽3𝑐41𝑐52𝑐3−12𝛽2𝑐41𝑐32𝑐23−8𝛽𝑐31𝑐32𝑐23−8𝛽3𝑐31𝑐32𝑐23+8𝑐72+2𝛽4𝑐41𝑐72+𝑐31𝑐22𝑐3𝑐4+𝛽4𝑐71𝑐22𝑐3𝑐4−2𝛽4𝑐61𝑐42𝑐4+4𝛽𝑐41𝑐22𝑐3𝑐4+6𝛽2𝑐51𝑐22𝑐3𝑐4+4𝛽3𝑐51𝑐22𝑐3𝑐4−2𝛽4𝑐61𝑐32𝑐23+3𝛽4𝑐51𝑐52𝑐3𝑒8𝑛.(2.24)
Expression (2.24) establishes the asymptotic error constant for the eighth order of convergence for the new eighth-order derivative-free method defined by (2.8).

2.2. Method 2: Liu 1

The second of three-step eighth-order derivative-free method is constructed by combining the two-step fourth-order method presented by Liu et al. [2], and the third step is developed to achieve the eighth order. As before, we have introduced two weight functions in the third step in order to achieve the desired eighth-order method. In this particular case the two weight functions are expressed as𝑓𝑧1−𝑛𝑓𝑤𝑛−1,𝑦1+2𝑓𝑛3𝑓𝑤𝑛2𝑓𝑥𝑛−1.(2.25) Then the iteration scheme based on Liu et al. method is given as𝑧𝑛=𝑦𝑛−𝑓𝑦𝑛𝑓𝑥𝑛,𝑦𝑛𝑤+𝑓𝑛,𝑦𝑛𝑤−𝑓𝑛,𝑥𝑛𝑥,(2.26)𝑛+1=𝑧𝑛−𝑓𝑧1−𝑛𝑓𝑤𝑛−1𝑓𝑦1+𝑛3𝑓𝑥𝑛𝑓𝑧𝑛2−1𝑓𝑧𝑛𝑓𝑦𝑛,𝑧𝑛𝑥−𝑓𝑛,𝑦𝑛𝑥+𝑓𝑛,𝑧𝑛,(2.27) where 𝑤𝑛,𝑦𝑛 are given in (2.8) and 𝑥0 is the initial approximation provided that the denominators of (2.26)-(2.27) are not equal to zero.

Theorem 2.5. Assume that the function f is sufficiently differentiable and f has a simple root 𝛼∈𝐷. If the initial approximation 𝑥0 is sufficiently close to 𝛼, then the method defined by (2.27) converges to 𝛼with eighth order.

Proof. Using appropriate expressions in the proof of Theorem 2.4 and substituting them into (2.27), we obtain the asymptotic error constant 𝑒𝑛+1=−𝑐1−712𝛽2𝑐21𝑐72−𝑐21𝑐42𝑐4+10𝛽𝑐1𝑐72−𝑐1𝑐52𝑐3+6𝛽3𝑐31𝑐72−2𝑐21𝑐32𝑐23−6𝛽2𝑐41𝑐42𝑐4−4𝛽3𝑐51𝑐42𝑐4−2𝛽𝑐21𝑐52𝑐3−4𝛽𝑐31𝑐42𝑐4+2𝛽3𝑐41𝑐52𝑐3−12𝛽2𝑐41𝑐32𝑐23−8𝛽𝑐31𝑐32𝑐23+3𝑐72+𝛽4𝑐41𝑐72+𝑐31𝑐22𝑐3𝑐4+𝛽4𝑐71𝑐22𝑐3𝑐4−𝛽4𝑐61𝑐42𝑐4+4𝛽𝑐41𝑐22𝑐3𝑐4+6𝛽2𝑐51𝑐22𝑐3𝑐4+4𝛽3𝑐61𝑐22𝑐3𝑐4−2𝛽4𝑐61𝑐32𝑐23+𝛽4𝑐51𝑐52𝑐3−8𝛽3𝑐51𝑐32𝑐23𝑒8𝑛.(2.28)
Expression (2.28) establishes the asymptotic error constant for the eighth order of convergence for the new eighth-order derivative-free method defined by (2.27).

2.3. Method 3: Liu 2

The third of three-step eighth-order derivative-free method is constructed by combining the two-step fourth-order method presented by Liu et al. [2], and the third step is developed to achieve the eighth-order. As before, we have introduced two weight functions in the third step in order to achieve the desired eighth-order method. In this particular case the two weight functions are expressed as𝑓𝑧1−𝑛𝑓𝑤𝑛−1,𝑦(2.29)1+2𝑓𝑛3𝑓𝑤𝑛2𝑓𝑥𝑛−1.(2.30) Then the iteration scheme based on Liu et al. method is given as𝑧𝑛=𝑦𝑛−𝑓𝑦𝑛𝑓𝑥𝑛,𝑦𝑛𝑤−𝑓𝑛,𝑦𝑛𝑤+𝑓𝑛,𝑥𝑛𝑓𝑥𝑛,𝑦𝑛2,𝑥(2.31)𝑛+1=𝑧𝑛−𝑓𝑧1−𝑛𝑓𝑤𝑛−1âŽ¡âŽ¢âŽ¢âŽ£îƒ©ğ‘“î€·ğ‘¦1−𝑛𝑓𝑧𝑛3âŽ¤âŽ¥âŽ¥âŽ¦îƒ¬ğ‘“î€·ğ‘¦1+𝑛3𝑓𝑥𝑛𝑓𝑧𝑛2−1×𝑓𝑧𝑛𝑓𝑦𝑛,𝑧𝑛𝑥−𝑓𝑛,𝑦𝑛𝑥+𝑓𝑛,𝑧𝑛,(2.32) where 𝑤𝑛,𝑦𝑛 are given in (2.8) and 𝑥0 is the initial approximation provided that the denominators of (2.31)-(2.32) are not equal to zero.

Theorem 2.6. Assume that the function f is sufficiently differentiable and f has a simple root 𝛼∈𝐷. If the initial approximation 𝑥0 is sufficiently close to 𝛼, then the method defined by (2.32) converges to 𝛼with eighth order.

Proof. Using appropriate expressions in the proof of Theorem 2.4 and substituting them into (2.32), we obtain the asymptotic error constant 𝑒𝑛+1=−𝑐1−7𝛽4𝑐71𝑐22𝑐3𝑐4−𝛽4𝑐61𝑐42𝑐4−2𝛽4𝑐61𝑐32𝑐23+𝛽4𝑐51𝑐52𝑐3+𝛽4𝑐41𝑐72+4𝛽3𝑐61𝑐22𝑐3𝑐4−5𝛽3𝑐51𝑐42𝑐4+3𝛽3𝑐41𝑐52𝑐3+8𝛽3𝑐31𝑐72+6𝛽2𝑐51𝑐22𝑐3𝑐4−12𝛽2𝑐41𝑐32𝑐23−9𝛽2𝑐41𝑐42𝑐4+3𝛽2𝑐31𝑐52𝑐3+21𝛽2𝑐21𝑐72+4𝛽𝑐41𝑐22𝑐3𝑐4−7𝛽𝑐31𝑐42𝑐4−8𝛽𝑐31𝑐32𝑐23+𝛽𝑐21𝑐52𝑐3+22𝛽𝑐1𝑐72−2𝑐21𝑐32𝑐23−2𝑐21𝑐42𝑐4+8𝑐72+𝑐31𝑐22𝑐3𝑐4−8𝛽3𝑐51𝑐32𝑐23𝑒8𝑛.(2.33)
Expression (2.33) establishes the asymptotic error constant for the eighth order of convergence for the new eighth-order derivative-free method defined by (2.32).

2.4. Method 4: SKK

The fourth of three-step eighth-order derivative-free method is constructed by combining the two-point fourth-order method presented by Khattri and Agarwal [10], and the third point is developed to achieve the eighth order. Here also, we have introduced two weight functions in the third step in order to achieve the desired eighth-order method. In this particular case the two weight functions are expressed as𝑓𝑧1−𝑛𝑓𝑤𝑛−1,𝑦1+3𝑓𝑛3𝑓𝑤𝑛2𝑓𝑥𝑛−1.(2.34) Then the iteration scheme based on the Khattri and Agarwal method is given as𝑤𝑛=𝑥𝑛𝑥−𝛽𝑓𝑛𝑦,(2.35)𝑛=𝑥𝑛−𝑓𝑥𝑛2𝑓𝑥𝑛𝑤−𝑓𝑛𝑧,(2.36)𝑛=ğ‘¦ğ‘›âˆ’îƒ©ğ‘“î€·ğ‘¥ğ‘›î€¸ğ‘“î€·ğ‘¦ğ‘›î€¸ğ‘“î€·ğ‘¥ğ‘›î€¸î€·ğ‘¤âˆ’ğ‘“ğ‘›î€¸îƒªâŽ¡âŽ¢âŽ¢âŽ£ğ‘“î€·ğ‘¦1+𝑛𝑓𝑥𝑛+𝑓𝑦𝑛𝑓𝑥𝑛2+𝑓𝑦𝑛𝑓𝑤𝑛+𝑓𝑦𝑛𝑓𝑤𝑛2âŽ¤âŽ¥âŽ¥âŽ¦ğ‘¥,(2.37)𝑛+1=𝑧𝑛−𝑓𝑤𝑛𝑓𝑤𝑛𝑧−𝑓𝑛𝑦1+3𝑓𝑛3𝑓𝑥𝑛𝑓𝑧𝑛2−1𝑓𝑧𝑛𝑓𝑦𝑛,𝑧𝑛𝑥−𝑓𝑛,𝑦𝑛𝑥+𝑓𝑛,𝑧𝑛,(2.38) where 𝑤𝑛,𝑦𝑛 are given in (2.8) and 𝑥0 is the initial approximation provided that the denominators of (2.35)–(2.37) are not equal to zero.

Theorem 2.7. Assume that the function f is sufficiently differentiable and f has a simple root 𝛼∈𝐷. If the initial approximation 𝑥0 is sufficiently close to 𝛼, then the method defined by (2.38) converges to 𝛼with eighth order.

Proof. Using appropriate expressions in the proof of Theorem 2.4 and substituting them into (2.38), we obtain the asymptotic error constant 𝑒𝑛+1=−𝑐1−78𝛽𝑐31𝑐32𝑐23+𝛽4𝑐51𝑐52𝑐3−2𝛽4𝑐61𝑐32𝑐23+15𝑐72+60𝛽2𝑐21𝑐72−4𝛽𝑐41𝑐22𝑐3𝑐4+6𝛽2𝑐51𝑐22𝑐3𝑐4+𝑐31𝑐22𝑐3𝑐4−4𝛽3𝑐61𝑐22𝑐3𝑐4+15𝛽4𝑐41𝑐72+𝛽4𝑐71𝑐22𝑐3𝑐4−3𝛽4𝑐61𝑐42𝑐4−8𝛽𝑐21𝑐52𝑐3−3𝑐21𝑐42𝑐4+12𝛽𝑐31𝑐42𝑐4−18𝛽2𝑐41𝑐42𝑐4+12𝛽3𝑐51𝑐42𝑐4−2𝑐21𝑐32𝑐23+𝑐1𝑐52𝑐3−48𝛽𝑐1𝑐72−39𝛽3𝑐31𝑐72+𝛽5𝑐61𝑐52𝑐3−3𝛽5𝑐51𝑐72−11𝛽3𝑐41𝑐52𝑐3+8𝛽3𝑐51𝑐32𝑐23+16𝛽𝑐31𝑐52𝑐3−12𝛽2𝑐41𝑐32𝑐23𝑒8𝑛.(2.39)
Expression (2.38) establishes the asymptotic error constant for the eighth order of convergence for the new eighth-order derivative-free method defined by (2.38).

3. The Kung-Traub Eighth-Order Derivative-Free Method

The classical Kung-Traub eighth-order derivative-free method considered is given in [3]. Since this method is well established, we will state the essential expressions used in order to calculate the approximate solution of the given nonlinear equations and thus compare the effectiveness of the new eighth-order derivative-free methods. The Kung-Traub method is given as𝑧𝑛=𝑦𝑛−𝑓𝑥𝑛𝑓𝑤𝑛𝑓𝑦𝑛𝑥−𝑓𝑛1𝑓[]−1𝑤,𝑥𝑓[],𝑥𝑤,𝑦𝑛+1=𝑧𝑛−𝑓𝑤𝑛𝑓𝑥𝑛𝑓𝑦𝑛𝑓𝑧𝑛𝑥−𝑓𝑛×1𝑓𝑧𝑛𝑤−𝑓𝑛1𝑓[]−1𝑦,𝑧𝑓[]−1𝑤,𝑦𝑓𝑦𝑛𝑥−𝑓𝑛1𝑓[]−1𝑤,𝑦𝑓[],𝑤,𝑥(3.1) where 𝑤𝑛,𝑦𝑛 are given in (2.8) and 𝑥0 is the initial approximation provided that the denominators of (3.1) are not equal to zero.

4. Application of the New Derivative-Free Iterative Methods

To demonstrate the performance of the new eighth-order methods, we take ten particular nonlinear equations. We will determine the consistency and stability of results by examining the convergence of the new derivative-free iterative methods. The findings are generalised by illustrating the effectiveness of the eighth-order methods for determining the simple root of a nonlinear equation. Consequently, we will give estimates of the approximate solution produced by the eighth-order methods and list the errors obtained by each of the methods. The numerical computations listed in the tables were performed on an algebraic system called Maple. In fact, the errors displayed are of absolute value, and insignificant approximations by the various methods have been omitted in Tables 1, 2, and 3.


FunctionRoot

𝑓 1 ( 𝑥 ) = ( 𝑥 − 2 ) ( 𝑥 1 0 + 𝑥 + 1 ) e x p ( − 𝑥 − 1 ) , 𝛼 = 2
𝑓 2 ( 𝑥 ) = e x p ( − 𝑥 2 + 𝑥 + 2 ) − c o s ( 𝑥 + 1 ) + 𝑥 3 + 1 , 𝛼 = − 1
𝑓 3 ( 𝑥 ) = s i n ( 𝑥 ) 2 − 𝑥 2 + 1 , 𝛼 = 1 . 4 0 4 4 9 1 6 5 ⋯
𝑓 4 ( 𝑥 ) = 𝑒 − 𝑥 − c o s ( 𝑥 ) , 𝛼 = − 0 . 6 6 6 2 7 3 1 2 6 ⋯
𝑓 5 ( 𝑥 ) = l n ( 𝑥 2 + 𝑥 + 2 ) − 𝑥 + 1 , 𝛼 = 4 . 1 5 2 5 9 0 7 4 ⋯
𝑓 6 ( 𝑥 ) = e x p ( 𝑥 2 + 7 𝑥 − 3 0 ) − 1 , 𝛼 = 1 . 4 0 4 4 9 1 6 5 ⋯
𝑓 7 ( 𝑥 ) = c o s ( 𝑥 ) 2 − 𝑥 5 𝛼 = 1 . 0 8 5 9 8 2 6 8 ⋯
𝑓 8 𝑥 ( 𝑥 ) = s i n ( 𝑥 ) − 2 , 𝛼 = 0
𝑓 9 ( 𝑥 ) = 𝑥 1 0 − 𝑥 3 − 𝑥 + 1 𝛼 = 0 . 5 9 1 4 4 8 0 9 3 ⋯
𝑓 1 0 ( 𝑥 ) = s i n ( 𝑥 ) − 𝑥 + 1 𝛼 = 2 . 6 3 0 6 6 4 1 5 ⋯


FunctionsK-TLiu1Liu2RTSKK

𝑓 1 ( 𝑥 ) , 𝑥 0 = 1 . 9 0 . 1 9 9 𝑒 − 3 7 0 . 1 2 1 𝑒 − 6 6 0 . 1 0 8 𝑒 − 6 6 0 . 1 1 4 𝑒 − 7 5 —
𝑓 2 ( 𝑥 ) , 𝑥 0 = − 1 . 5 0 . 1 6 6 𝑒 − 1 5 8 0 . 8 3 2 𝑒 − 2 0 7 0 . 4 9 0 𝑒 − 2 8 2 0 . 1 2 8 𝑒 − 1 5 6 0 . 1 4 5 𝑒 − 1 0 0
𝑓 3 ( 𝑥 ) , 𝑥 0 = 1 . 6 0 . 1 7 2 𝑒 − 2 5 8 0 . 6 2 9 𝑒 − 4 1 7 0 . 2 0 4 𝑒 − 3 9 3 0 . 4 9 8 𝑒 − 2 9 7 0 . 3 0 8 𝑒 − 2 8 5
𝑓 4 ( 𝑥 ) , 𝑥 0 = 0 . 1 2 5 0 . 1 6 6 𝑒 − 1 3 8 5 0 . 1 0 7 𝑒 − 1 1 3 3 0 . 8 6 8 𝑒 − 8 0 0 0 . 5 6 6 𝑒 − 1 0 6 0 0 . 6 0 9 𝑒 − 2 2 0
𝑓 5 ( 𝑥 ) , 𝑥 0 = 4 . 4 0 . 1 1 5 𝑒 − 9 2 7 0 . 3 5 0 𝑒 − 1 0 1 7 0 . 2 6 8 𝑒 − 1 0 2 2 0 . 4 7 1 𝑒 − 1 0 2 5 0 . 6 3 9 𝑒 − 7 9 1
𝑓 6 ( 𝑥 ) , 𝑥 0 = 3 . 1 0 . 1 1 8 𝑒 − 3 2 0 . 5 8 1 𝑒 − 5 9 0 . 2 3 6 𝑒 − 5 8 0 . 2 2 4 𝑒 − 5 3 0 . 1 9 1 𝑒 − 1 2
𝑓 7 ( 𝑥 ) , 𝑥 0 = 0 . 7 5 0 . 7 5 6 𝑒 − 5 6 8 0 . 1 1 9 𝑒 − 5 7 1 0 . 3 3 3 𝑒 − 6 0 7 0 . 5 2 2 𝑒 − 5 5 7 0 . 3 6 7 𝑒 − 3 3 5
𝑓 8 ( 𝑥 ) , 𝑥 0 = 0 . 5 0 . 2 6 0 𝑒 − 3 1 7 0 . 6 5 1 𝑒 − 3 4 4 0 . 2 0 5 𝑒 − 3 0 1 0 . 2 9 8 𝑒 − 2 8 5 0 . 2 5 3 𝑒 − 6 8 5
𝑓 9 ( 𝑥 ) , 𝑥 0 = 0 . 5 0 . 2 9 3 𝑒 − 4 1 6 0 . 7 1 3 𝑒 − 4 3 3 0 . 3 4 7 𝑒 − 4 4 8 0 . 6 4 4 𝑒 − 4 2 3 0 . 5 3 5 𝑒 − 2 8 3
𝑓 1 0 ( 𝑥 ) , 𝑥 0 = 2 . 5 0 . 6 8 7 𝑒 − 7 3 9 0 . 4 3 9 𝑒 − 6 9 6 0 . 9 8 1 𝑒 − 6 9 8 0 . 1 2 1 𝑒 − 6 9 4 0 . 7 7 0 𝑒 − 6 4 3


FunctionsK-TLiu1Liu2RTSKK

𝑓 1 ( 𝑥 ) , 𝑥 0 = 1 . 9 7.51467.56027.55937.5078—
𝑓 2 ( 𝑥 ) , 𝑥 0 = − 1 . 5 8.00217.97078.00008.00267.8899
𝑓 3 ( 𝑥 ) , 𝑥 0 = 1 . 6 8.00008.00008.00008.00007.9847
𝑓 4 ( 𝑥 ) , 𝑥 0 = 0 . 1 2 5 12.00011.00010.00011.0008.0001
𝑓 5 ( 𝑥 ) , 𝑥 0 = 4 . 4 8.00008.00008.00008.00008.0000
𝑓 6 ( 𝑥 ) , 𝑥 0 = 3 . 1 7.73847.58377.56947.4350—
𝑓 7 ( 𝑥 ) , 𝑥 0 = 0 . 7 5 8.00008.00008.00008.00007.9889
𝑓 8 ( 𝑥 ) , 𝑥 0 = 0 . 5 10.97310.97710.97010.96710.994
𝑓 9 ( 𝑥 ) , 𝑥 0 = 0 . 5 8.00008.00008.00008.00007.9845
𝑓 1 0 ( 𝑥 ) , 𝑥 0 = 2 . 5 8.00008.00008.00008.00008.0000

Remark 4.1. The family of three-step methods requires four function evaluations and has the order of convergence eight. Therefore, this family is of optimal order and supports the Kung-Traub conjecture [3]. To determine the efficiency index of these new derivative-free methods, we will use Definition 2.3. Hence, the efficiency index of the eighth-order derivative-free methods given is 4√8≈1.682.

Remark 4.2. The test functions and their exact root 𝛼 are displayed in Table 1. The differences between the root 𝛼 and the approximation 𝑥𝑛 for test functions with initial approximation 𝑥0 are displayed in Table 2. In fact, 𝑥𝑛 is calculated by using the same total number of function evaluations (TNFEs) for all methods. Here, the TNFE for all the methods is 12. Furthermore, the computational order of convergence (COC) is displayed in Table 3.

5. Remarks and Conclusion

We have demonstrated the performance of a new family of eighth-order derivative-free methods. Convergence analysis proves that the new methods preserve their order of convergence. There are two major advantages of the eighth-order derivative-free methods. Firstly, we do not have to evaluate the derivative of the functions; therefore they are especially efficient where the computational cost of the derivative is expensive, and secondly we have established a higher order of convergence method than the existing derivative-free methods [4, 5]. We have examined the effectiveness of the new derivative-free methods by showing the accuracy of the simple root of a nonlinear equation. The main purpose of demonstrating the new eighth-order derivative-free methods for many different types of nonlinear equations was purely to illustrate the accuracy of the approximate solution, the stability of the convergence, and the consistency of the results and to determine the efficiency of the new iterative method. In addition, it should be noted that like all other iterative methods, the new methods have their own domain of validity and in certain circumstances should not be used.

References

  1. S. D. Conte and C. de Boor, Elementary Numerical Analysis: An Algorithmic Approach, McGraw-Hill, New York, NY, USA, 1981. View at: Zentralblatt MATH
  2. Z. Liu, Q. Zheng, and P. Zhao, “A variant of Steffensen's method of fourth-order convergence and its applications,” Applied Mathematics and Computation, vol. 216, no. 7, pp. 1978–1983, 2010. View at: Publisher Site | Google Scholar | Zentralblatt MATH
  3. H. T. Kung and J. F. Traub, “Optimal order of one-point and multipoint iteration,” Journal of the Association for Computing Machinery, vol. 21, pp. 643–651, 1974. View at: Google Scholar | Zentralblatt MATH
  4. A. Cordero, J. L. Hueso, E. Martinez, and J. R. Torregrosa, “Steffensen type methods for solving nonlinear equations,” Journal of Computational and Applied Mathematics. In press. View at: Publisher Site | Google Scholar
  5. S. K. Khattri and I. K. Argyros, “Sixth order derivative free family of iterative methods,” Applied Mathematics and Computation, vol. 217, no. 12, pp. 5500–5507, 2011. View at: Publisher Site | Google Scholar
  6. S. Weerakoon and T. G. I. Fernando, “A variant of Newton's method with accelerated third-order convergence,” Applied Mathematics Letters, vol. 13, no. 8, pp. 87–93, 2000. View at: Publisher Site | Google Scholar | Zentralblatt MATH | MathSciNet
  7. W. Gautschi, Numerical Analysis: An Introduction, Birkhauser, 1997.
  8. J. F. Traub, Iterative Methods for Solution of Equations, Chelsea Publishing, New York, NY, USA, 1977.
  9. J. F. Steffensen, “Remark on iteration,” Skand Aktuar Tidsr, vol. 16, pp. 64–72, 1933. View at: Google Scholar
  10. S. K. Khattri and R. P. Agarwal, “Derivative-free optimal iterative methods,” Computational Methods in Applied Mathematics, vol. 10, pp. 368–375, 2010. View at: Google Scholar

Copyright © 2011 R. Thukral. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

982 Views | 599 Downloads | 15 Citations
 PDF  Download Citation  Citation
 Download other formatsMore
 Order printed copiesOrder