On an approximation formula

Abstract

We generalize an approximation formula which in some particular cases has been studied by [J.F. Traub 1964] and \ [R.M.Humel and C.S. Secbeck 1949]. Denote by \(I_{x}\) the closed interval determined by the distinct points \(x,x_{0} \in \mathbb{R}\). Consider the nonlinear mapping \(f:I_{x}\rightarrow \mathbb{R}\), which has derivatives up to the order \(2n+1\) on \(I_{x}\), and deonte by \(G\) the set of functions $$G=\big\{g:g(t) =f(x_0) + (t-x_0) \sum \limits_{i=1}^{n} a_i f'(x_0 + b_i(t-x_0) , \ a_i, b_i \in \mathbb{R}, i=1,n, t\in I_x\big\}$$ From the set \(G\) we determine a function \(\bar{g}\) with the properties \(f^{(i)}(x_0) = \bar{g}^{(i)}(x_0)\). We determine the coefficients \(a_{i},b_{i},\ i=1,\ldots,n\) and we also evaluate the remainder \(f(t) -\bar{g}(t)\), \(t\in I_{x}\).

Authors

Ion Păvăloiu
(Tiberiu Popoviciu Institute of Numerical Analysis)

Keywords

PDF

Scanned paper: on the journal website.

Latex version of the paper.

Cite this paper as:

I. Păvăloiu, On an approximation formula, Rev. Anal. Numér. Théor. Approx., 26 (1997) nos. 1-2, pp. 179-184.

About this paper

Print ISSN

1222-9024

Online ISSN

2457-8126

References

[1] C. I. Berezin and N. Jidkov, Metody vychisleny, Fizmatgiz, Moscow (1962).

[2] P. M. Humel and C. L. Seebeck Jr., A generalization of Taylor’s expansion, Amer. Math. Monthly 56 (1949), pp. 243-247.

[3] A. Lupas Calculul valorilor unor functii elementare, Gazeta Matematica (Ser. A) VII, l (1986), pp. 15-26.

[4] J. F. Traub, Iterative Methods for the Solution of Equations, Prentice-Hall, Inc., Englowood Cliffs, N.J., 1964.

Paper (preprint) in HTML form

On an approximation formula

On an approximation formula

Ion Păvăloiu
1991 Mathematics Subject Classification:
65D05, 65D10.

1. Introduction

This Note contains some remarks concerning an approximation formula for functions, which is a generalization of some interpolation formulae given in [3] and [5]. In particular, we shall show that only one of the formulae of this type, mentioned in [5], has a maximal degree of exactness. Some particular cases of such formulae were also mentioned in [5, p.163].

Denote by Ix the closed interval determined by two distinct points x0, x in ℝ. For a (2⁢n+1)-times derivable function f:Ix→ℝ and n∈ℕ, consider the class G of functions given by

(1) G={ g:g⁢(t)=f⁢(x0)+(t−x0)⁢∑i=1nai⁢f′⁢(x0+bi⁢(t−x0)),
ai,bi∈ℝ,i=1,n¯,t∈Ix}.

Consider the following problem: Find a function g¯∈G such that

(2) f(i)⁢(x0)=g¯(i)⁢(x0),i=1,m¯.

In [5] this problem was solved in some particular cases. We shall show that, for m=2⁢n, this problem has a unique solution and we shall give a representation for the remainder.

2. Determination of the approximating function

For m=2⁢n, we are looking for a function g¯ in G verifying conditions (2) and having a maximal degree of approximation.

It is easily seen that conditions (2) lead to the following system, having the real numbers ai,bi,i=1,n¯, as unknowns:

(3) ∑i=1nai⁢bik=1(k+1),k=0,1,…,2⁢n−1.

Consider now a continuous function φ:[0,1]→ℝ and let

(4) ∫01φ⁢(t)⁢𝑑t=∑i=1nai⁢φ⁢(bi)+R⁢[φ]

be a quadrature formula, having {bi}1n as knots and {ai}1n as coefficients. Asking that R⁢[φk]=0 for φk⁢(t)=tk,k=0,2⁢n−1¯, (4) becomes the classical Gauss quadrature formula. On the other hand, the conditions R⁢[φk]=0, for φk⁢(t)=tk,k=0,2⁢n−1¯, lead again to the system (3), implying that bi must be the roots of the Legendre polynomial wn of degree n, i.e., the roots of the equation

(5) wn⁢(i):=n!(2⁢n)!⁢d⁢tn⁢[tn⁢(t−l)n]=0.

The coefficients ai are given by the following formula

(6) ai=(n!)4[(2⁢n)!]2⁢bi⁢(1−bi)⁢[wn′⁢(bi)]2,i=1,n¯,

(see [2, p.261]).

Now, it is clear that the following theorem holds:

Theorem 2.1.

If f:Ix→ℝ is a (2⁢n+1)-times derivable function on Ix, then there exists only one function g¯∈G verifying conditions (2) for m=2⁢n. The parameters {ai}i=1n are given by formula (6), where {bi}i=1n are the roots of equation (5).

3. Determination of the remainder

Consider the approximation formula

(7) f⁢(x)=g¯⁢(x)+r⁢[f],

where g¯∈G is a function verifying (3) and r⁢[f] is the remainder.

In the conditions of Theorem 2.1, it follows that

(8) f′⁢(x0+bi⁢(x−x0))=∑j=12⁢nf(j)⁢(x0)(j−1)!⁢bij−1⁢(x−x0)j−1+ri⁢(x),

where

(9) ri⁢(x)=f(2⁢n+1)⁢(θi)(2⁢n)!⁢bi2⁢n⁢(x−x0)2⁢n,

and θi is a number contained in the open interval determined by x0 and x0+bi⁢(x−x0),1≤i≤n.

From (8) we obtain the equalities

(10) f⁢(x)−f⁢(x0)−(x−x0)⁢f′⁢(x0+bi⁢(x−x0))=
=f(x)−f(x0)−∑j=12⁢nf(j)⁢(x0)(j−1)!bij−1(x−x0)j−ri(x)(x−x0,),i=1,n¯.

Multiplying equalities (10) by ai, taking into account solution (3) and summing up, we obtain

(11) f⁢(x)−g¯⁢(x)=f⁢(x)−∑j=02⁢nf(j)⁢(x0)j!⁢(x−x0)j−∑i=1nai⁢ri⁢(x)⁢(x−x0).

Now, using (9) and Lagrange from of the remainder in the Taylor formula, we get

(12) f⁢(x)−g¯⁢(x)=[f(2⁢n+1)⁢(η)(2⁢n+1)!−∑i=1nai⁢bi2⁢n⁢f(2⁢n+1)⁢(θi)(2⁢n)!]⁢(x−x0)(2⁢n+1),

where η∈Ix.

Setting φ⁢(t)=t2⁢n in (4) and taking into account the form of the remainder term in the Gauss quadrature formula [[2],p.259], we get

∑i=1nai⁢bi2⁢n+[n!]4[(2⁢n)!]2⁢(2⁢n+1)=12⁢n+1,

implying

(13) ∑i=1nai⁢bi2⁢n=[(2⁢n)!]2−[n!]4(2⁢n+1)⁢[(2⁢n)!]2.

Suppose now that the (2⁢n+1)-order derivative of f is bounded on Ix and let

(14) M2⁢n+1=supt∈Ix|f(2⁢n+1)⁢(t)|.

Taking into account relations (12) and (13), one obtains the following delimitation for r⁢[f]

(15) |r⁢[f]|≤M2⁢n+1(2⁢n+1)!⋅2⋅[(2⁢n)!]2+[n!]4[(2⁢n)!]2⁢|x−x0|2⁢n+1.

4. Particular cases

  • a)

    n=1. In this case, b1=12, a1=1 and

    g⁢(x)=f⁢(x0)+(x−x0)⁢f′⁢(x0+12⁢(x−x0)).

    From (15) we get

    |f⁢(x)−g⁢(x)|≤7⁢M324⁢|x−x0|3,

    where M3=supt∈Ix|f′′′⁢(t)|.

  • b)

    n=2. In this case, b1=3−36, b2=3+36,a1=a2=12and

    g⁢(x)=f⁢(x0)+12⁢(x−x0)⁢[f′⁢(x0+3−36⁢(x−x0))+f′⁢(x0+3+36⁢(x−x0))],

    One also obtains the evaluation

    |f⁢(x)−g⁢(x)|≤71⁢M54320⁢|x−x0|5,

    where M5=supt∈Ix|f(5)⁢(t)|.

Remark.

Approximation formula of the type considered in this Note could be useful for the approximate calculation of the values of some functions having rational functions as derivatives.

References

  • [1]
  • [2] C. I. Berezin and N. Jidkov, Metody vychisleny, Fizmatgiz, Moscow (1962).
  • [3] P. M. Humel and C. L. Seebeck Jr., A generalization of Taylor’s expansion, Amer. Math. Monthly 56 (1949), 243–247.
  • [4] A. Lupaş, Calculul valorilor unor funcţii elementare, Gazeta Matematică (Ser. A) VII, 1 (1968), 15–26.
  • [5] J. F. Traub, Iterative Methods for the Solution of Equations, Prentice-Hall, Inc., Englewood Cliffs, N.J., 1964.

Received March 15, 1996

Institutul de Calcul ”Tiberiu Popoviciu”

Str. G. Bilaşcu, nr.37

C.P. 68, O.P. 1, 3400 Cluj-Napoca

România

1997

Related Posts