Abstract

In this note we study the convergence of a generalized Aitken type method for approximating the solutions of nonlinear equations in R. We obtain conditions which assure monotone convergence of the generated sequences. We also obtain a posteriori estimations for the errors.

Author

Keywords

nonlinear equations in R; Aitken method

Cite this paper as:

I. Păvăloiu, E. Cătinaş, On an Aitken type method, Rev. Anal. Numér. Théor. Approx., 36 (2007) no. 2, pp. 173-176. https://doi.org/10.33993/jnaat362-865

PDF

PDF-LaTeX file (on the journal website).

About this paper

Print ISSN

1222-9024

Online ISSN

2457-8126

MR

1222-9024

Online ISSN

2457-8126

Google Scholar citations

[1] M.A. Ostrowski, Solution of Equations and Systems of Equations, Academic Press, New York and London, 1980.
[2] I. Pavaloiu, Solutions Equations by Interpolation, Dacia, Cluj-Napoca, 1981 (in Romanian).
[3] I. Pavaloiu, Bilateral approximation of solution of equations by order three Steffensen-type methods , Studia Univ. “Babes-Bolyai”, Mathematica, LI (2006) no. 3, pp. 105–114.
[4] I. Pavaloiu and N. Pop, Interpolation and Applications, Risoprint, Cluj-Napoca, 2005 (in Romanian).
[5] I. Pavaloiu, Approximation of the roots of equations by Aitken-Steffensen-type monotonic sequences , Calcolo, 32 (1995) nos. 1–2, pp. 69–82.
[6] I. Pavaloiu and E. Catinas, On a Steffensen type method, Proceedings of the Ninth International Symposion on Symbolic and Numeric Algoritms for Scientific Computing (SYNASC 2007), September 26–29, 2007, Timisoara, Romania, IEEE Computer Society, pp. 369–375.
[7] B.A. Turowicz, Sur les derivees d’ordre superieur d’une fonction inverse , Ann. Polon. Math., 8 (1960), pp. 265–269.

Paper (preprint) in HTML form

ON AN AITKEN TYPE METHOD∗

ON AN AITKEN TYPE METHOD∗

Ion Păvăloiu and Emil Cătinaş†
(Date: August 14, 2006.)
Abstract.

In this note we study the convergence of a generalized Aitken type method for approximating the solutions of nonlinear equations in ℝ. We obtain conditions which assure monotone convergence of the generated sequences. We also obtain a posteriori estimations for the errors.

Key words and phrases:
Nonlinear equations, Aitken method.
1991 Mathematics Subject Classification:
65H05.
∗This work was supported by MEdC under grant 2CEEX06-11-96/19.09.2006.
†“T. Popoviciu” Institute of Numerical Analysis, P.O. Box 68-1, Cluj-Napoca, Romania, e-mail: {pavaloiu,ecatinas}@ictp.acad.ro.

1. Introduction

Consider the equation

(1) f⁢(x)=0

where f:[a,b]→ℝ, a,b∈ℝ, a<b.

Consider also the following two equations, both equivalent to (1):

(2) x−gi⁢(x)=0,gi:[a,b]→[a,b],i=1,2.

In order to approximate a root x∗ of (1), we consider the sequence (xn)n≥0, generated by the following relations:

(3) xn+1 =xn−f⁢(xn)[xn,g1⁢(xn);f]−[xn,g1⁢(xn),g2⁢(xn);f]⁢f⁢(xn)⁢f⁢(g1⁢(xn))[xn,g1⁢(xn);f]⁢[xn,g2⁢(xn);f]⁢[g1⁢(xn),g2⁢(xn);f]
x∈[a,b],n=0,1,2,…,

where [x,y;f] denotes the first order divided difference of f on x and y.

Relation (3) suggests an Aitken type method. If we assume that f∈C3⁢[a,b] and f′⁢(x)≠0 for all x∈[a,b], denoting F=f⁢([a,b]), it is known that there exists f−1:F→[a,b] and f−1∈C3⁢([a,b]). Moreover, the following relation holds

(4) (f−1⁢(y))′′′=3⁢(f′′⁢(x))2−f′⁢(x)⁢f′′′⁢(x)(f′⁢(x))5

where y=f⁢(x)and x∈[a,b] (see [2], [3], [5], [8]).

In [4] was studied the convergence of a Steffensen type method, analogous to (3). In order to obtain a control of the error at each iteration step, we have studied in [7] the cases when the generalized Steffensen type method leads to monotone approximations. Under some reasonable conditions of monotonicity and convexity (concavity) on f, in [7] were obtained monotone sequences which yield bilateral approximations to the solution x∗ of (1). A condition such that the Steffensen type method studied in [7] to lead to monotone approximations is that the sign of the expression

(5) Ef⁢(x)=3⁢(f′′⁢(x))2−f′⁢(x)⁢f′′′⁢(x)

to be negative: Ef⁢(x)<0, ∀x∈[a,b].

In this note we shall show that if Ef⁢(x)>0 ∀x∈[a,b], one may construct functions g1 and g2, and on may choose the initial approximations x0∈[a,b] such that, under some reasonable hypotheses on f, the sequence (3) converges monotonically.

In an analogous fashion as in [7], we can easily show that for a given xn∈[a,b], then exists ξn∈]cn,dn[, where [cn,dn] is the smallest interval containing the points xn, g1⁢(xn), g2⁢(xn), x∗, such that

(6) x∗−xn+1=−Ef⁢(ξn)⁢f⁢(xn)⁢f⁢(g1⁢(xn))⁢f⁢(g2⁢(xn))6⁢[f′⁢(ξn)]5,n=0,1,….

2. The convergence of the Aitken method

We consider the following hypotheses on the functions f1,g1, g2, and on the initial approximation x0∈[a,b]:

  • i.

    f∈C3⁢[a,b] and Ef⁢(x)>0, ∀x∈[a,b] where Ef is given by (5);

  • ii.

    f⁢(x0)<0;

  • iii.

    f′⁢(x)>0, ∀x∈[a,b];

  • iv.

    f′′⁢(x)≥0, x∈[a,b];

  • v.

    g1 and g2 are continuous functions, decreasing on [a,b];

  • vi.

    equation (1) has a solution x∗∈]a,b[;

  • vii.

    g1⁢(x0)≤b,g2⁢(x0)≤b;

  • viii.

    equations (1) and (2) are equivalent.

The following result holds:

Theorem 1.

If functions f1,g1,g2 and the initial approximation x0∈[a,b] verify hypotheses i.–viii., then the sequences (xn)n≥0, (g1⁢(xn)), (g2⁢(xn)) generated by (3) have the following properties:

  • j.

    the sequence (xn)n≥0 is increasing;

  • jj.

    the sequences (g1⁢(xn)), (g2⁢(xn))n≥0 are decreasing;

  • jjj.

    limxn=limg1⁢(xn)=limg2⁢(xn)=x∗;

  • jv.

    the following relations hold:

    x∗−xn≤min⁡{g1⁢(xn)−xn,g2⁢(xn)−xn},n=0,1,….
Proof.

Let xn∈[a,b] be an approximation of the solution x∗ such that f⁢(xn)<0, g1⁢(xn)≤b, g2⁢(xn)≤b. From ii. and iii. it follows xn<x∗. By v. we have g1⁢(xn)>x∗, g2⁢(xn)>x∗. Taking into account the fact that f⁢(xn)<0, f⁢(g1⁢(xn))>0 and using hypothesis iii. and iv., from (3) it follows that xn+1>xn. Since f⁢(g2⁢(xn))>0, by iii., i. and (6) we get xn+1<x∗.

By v. and xn<xn+1 it follows g1(xn)>g1(xn+1)) and g2⁢(xn)>g2⁢(xn+1). By v. it also follows that g1⁢(xn+1)>x∗ and g2⁢(xn+1)>x∗. These relations imply j. and jj.

In order to prove jjj., let ℓ=lim xn. For n→∞, from (3) we obtain f⁢(ℓ)=0 and hence ℓ=x∗. Obviously, from the continuity of g1 and g2 it follows that limg1⁢(xn)=g1⁢(x∗)=x∗ and limg2⁢(xn)=g2⁢(x∗)=x∗.

Relations jv. are obvious, and allow us to evaluate the a posteriori error at each iteration step. ∎

3. Construction of functions g1 and g2

The basic conditions on g1 and g2 on v. and vii.

We shall consider a function g:[a,b]→ℝ given by

(7) g⁢(x)=x−λ⁢f⁢(x)

where λ∈ℝ.

If g′⁢(x)≤0, then g is decreasing. This attracts

1−λ⁢f′⁢(x)<0

i.e.,

(8) λ>1f′⁢(x)

But since f′⁢(x)>0 and f′′≥0, we obtain

f′⁢(a)≤f′⁢(x)≤f′⁢(b)

or

(9) 1f′⁢(a)≥1f′⁢(x)≥1f′⁢(b).

By (9), it is obvious that if λ>1f′⁢(a) then (8) is verified for every x∈[a,b], from which g is decreasing.

In order to obtain g⁢(x0)≤b, we need thatx0−λ⁢(x0)≤b, which leads to λ≤x0−bf⁢(x0) which in turn holds if x0 is sufficiently close to x∗.

Let λi, i=1,2, λ1≠λ2 be two numbers such that

1f′⁢(a)≤λi≤x0−bf⁢(x0);

then functions gi, i=1,2 given by

gi⁢(x)=x−λi⁢f⁢(x)

verify hypotheses v. and vii.

4. Numerical example

Consider the equation

f⁢(x)=x−2⁢cos⁡x=0,

with x∈[π6,π2].

Obviously, f′⁢(x)>0, for all x∈[π6,π2], and f′′⁢(x)>0, x∈[π6,π2]. If in (7) we take λ1=0.5, λ2=0.6 and x0=π6, then g1 and g2 verify v. and vii.

We have:

g1⁢(x) =cos⁡x+x2,g2⁢(x)=6⁢cos⁡x+2⁢x5.

Obviously, g1⁢(π6)<π2 and g2⁢(π6)<π2. Function Ef is given by

Ef⁢(x)=4+8⁢cos2⁡x+2⁢sin⁡x>0,∀x∈[π6,π2].

In the table below we have obtained the following results.

Table 1. Numerical results
n xn g1⁢(xn) g2⁢(xn) f⁢(xn)
1 0.5235987755982988 1.127824791583588 1.248669994780646 −1.208452031970579e+0
2 1.027717814817341 1.030632925047758 1.031215947093841 −5.830220460833369e−3
3 1.029866528928396 1.029866529462959 1.029866529569871 −1.069125232788792e−9
4 1.029866529322259 1.029866529322259 1.029866529322259 0

The numerical results confirm that inequalities

0≤x∗−x4<10−15

hold.

References

  • [1]
  • [2] Ostrowski, M.A., Solution of Equations and Systems of Equations, Academic Press, New York and London, 1980.
  • [3] Păvăloiu, I., Solutions Equations by Interpolation, Dacia, Cluj-Napoca, 1981 (in Romanian).
  • [4] Păvăloiu, I., Bilateral approximation of solution of equations by order three Steffensen-Type methods, Studia Univ. “Babeş-Bolyai”, Mathematica, LI (2006) no. 3, pp. 105–114.
  • [5] Păvăloiu, I. and Pop. N., Interpolation and Applications, Risoprint, Cluj-Napoca, 2005 (in Romanian).
  • [6] Păvăloiu, I., Approximation of the roots of equations by Aitken-Steffensen-type monotonic sequences, Calcolo, 32 (1995) nos. 1–2, pp. 69–82.
  • [7] Păvăloiu, I. and Cătinaş, E., On a Steffensen type method, Proceedings of the Ninth International Symposion on Symbolic and Numeric Algoritms for Scientific Computing (SYNASC 2007), September 26–29, 2007, Timişoara, Romania, IEEE Computer Society.
  • [8] Turowicz, B.A., Sur les derivées d’ordre superieur d’une fonction inverse, Ann. Polon. Math., 8 (1960), pp. 265–269.
  • [9]
2007

Related Posts