Posts by Octavian Agratini

Abstract

In this paper, we define and study a general class of convolution operators based on Landau operators. A property of these new operators is that they reproduce the affine functions, a feature less commonly encountered by integral type operators. Approximation properties in different function spaces are obtained, including quantitative Voronovskaya-type results.

Authors

Octavian Agratini
Babeş-Bolyai University, Cluj-Napoca, Romania
Tiberiu Popoviciu Institute of Numerical Analysis, Romanian Academy

Sorin G. Gal
University of Oradea, Romania
Academy of Romanian Scientists

Keywords

Landau operator; modulus of continuity; weighted space; approximation process; upper estimates; quantitative Voronovskaya-type theorems

PDF

Cite this paper as:

O. Agratini, S.G. Gal,  On Landau-type approximation operators, Mediterranean Journal of Mathematics, 18 (2021) art. no. 64, https://doi.org/10.1007/s00009-021-01712-w

About this paper

Journal

Mediterranean Journal of Mathematics

Publisher Name

Springer

Print ISSN

1660-5446

Online ISSN

1660-5454

Google Scholar Profile

Paper (preprint) in HTML form

On Landau type operators

On Landau type operators

Octavian Agratini Babeş-Bolyai University,
Faculty of Mathematics and Computer Science,
Str. Kogălniceanu, 1, 400084 Cluj-Napoca, Romania
and
Tiberiu Popoviciu Institute of Numerical Analysis,
Romanian Academy,
Str. Fântânele, 57, 400320 Cluj-Napoca, Romania
agratini@math.ubbcluj.ro
 and  Sorin G. Gal University of Oradea,
Department of Mathematics and Computer Science,
Str. Universităţii, 1, 410087 Oradea, Romania
galso@uoradea.ro
Abstract.

In this paper we define and study a general class of convolution operators based on Landau operators. A property of these is that they reproduce the affine functions, a feature less commonly encountered by integral type operators. Approximation properties in different function spaces are obtained.

Keywords and phrases: Landau operator, Korovkin theorem, modulus of smoothness, weighted space, approximation process.

Mathematics Subject Classification: 41A36, 41A25.

1. Introduction

Edmund Landau [2, Eq. (2)] proved

limn01f(ξ)(1(ξx)2)n𝑑ξ201(1u2)n𝑑u=f(x), (1.1)

the convergence being uniform over a compact interval [a,b], where 0<a<b<1. In the first instance, f is supposed given only for x[a,b]. As the author states, the function f can be extended by continuity from [a,b] to [0,1] by usual technique:

f(x)=f(a) for x[0,a)andf(x)=f(b) for x(b,1].

The relation (1.1) was used by Landau to recover the celebrated Weierstrass approximation theorem established in 1885 which says that every continuous function defined on a compact interval can be uniformly approximated as closely as desired by a polynomial function. Examining (1.1) we notice that the author uses a sequence of convolution operators which today bears his name.

A generalization of this class of operators was given by Mamedov and it was described by the following relation [5, Eq. (1)]

(Mnf)(x)=kn1/(2k)Γ(1/(2k))01f(t)(1(tx)2k)n𝑑t, (1.2)

where k is a fixed natural number, x[0,1] and Γ indicates Gamma function. In time, more papers appeared concerning these operators, among the most recent one we mention [1].

In this note we propose another generalization of Landau operators which involves a real parameter α1. Keeping the idea of convolution type operators, this class aim to be associated with functions defined on the whole real axis and the affine functions are fixed points of the operators in question. We obtain evaluations of the approximation error for bounded functions as well as for functions belonging to some weighted spaces.

2. The operators

Let α1 be fixed. For each n, we set

an=01(1y2α)n𝑑yandτp,n=B(n+1,pλ)B(n+1,λ),p, (2.1)

where λ=12α(0,12] and B indicates Beta function. Taking in view these quantities, the following statements hold, their proofs are based on elementary calculus.

The sequence (an)n1 is strictly decreasing and

an=λB(n+1,λ). (2.2)

We have

01yi(1y2α)n𝑑y=λB(n+1,(1+i)λ), (2.3)

and, for p2, (τp,n)n1 is a positive strictly decreasing sequence verifying the relations

τp,n+1=τp,nλ+n+1pλ+n+1<τp,1=λ+1p(1+pλ)<1p. (2.4)
Lemma 2.1.

Let the real sequence (τp,n)n1 be given by (2.1). For any λ>0 and any p=2,3,, we have

limnτp,n=0. (2.5)

Proof. Relation (2.4) implies

τp,n=k=0nλ+kpλ+k.

On the other hand, the following identity

Γ(λ)=limnn!nλλ(λ+1)(λ+n)

takes place, see, e.g., [6, Exercise 1.7]. Consequently, we get

Γ(pλ)Γ(λ)=limnn(p1)λτp,n.

We mention that the existence of the limit of the sequence (τp,n)n1 is based on statement (2.4). Since Γ(pλ)/Γ(λ) belongs to and limnn(p1)λ= for any λ>0 and p=2,3,, we deduce (2.5).

B() denotes the space of all real valued functions defined on and bounded. The space is endowed with the usual sup-norm ,

h=supx|h(x)|.

We consider the operators defined as follows

(Lnf)(x)=1an01f(xy+τ2,n)(1y2α)n𝑑y,x, (2.6)

where f is Lebesgue measurable on the domain.

The operators can be rewritten as follows

(Lnf)(x)=1anx1xf(t+τ2,n)(1(xt)2α)n𝑑t,x. (2.7)

It is observed that for each n, Ln represents a convolution product between the functions fun and vn, where

un(t)=t+τ2,nandvn(t)=(1t2α)n,t.

Also, the operators are linear and positive. If fB(), then the operators are non-expansive, this means

Lnff.

Replacing the sequence (τ2,n)n1 with zero and choosing α=k, we obtain the operators introduced by Mamedov, see (1.2).

3. Results

Set 0={0} and ej, j0, monomials of degree j, e0(x)=1, ej(x)=xj, j.

Theorem 3.1.

The operators Ln, n, defined by (2.6) reproduce affine functions.

Proof. Taking in view that these operators are linear, it is enough to prove Lne0=e0 and Lne1=e1.

Due to the definition of an, it is clear that Lne0=e0. Further, by using (2.3) and (2.2), we get

(Lne1)(x)=1an01y(1y2α)n𝑑y+x+τ2,n=x,

and the proof is completed.

The purpose of introducing the quantity τ2,n in the definition of Ln, n, was precisely to ensure the identity Lne1=e1.

To determine the approximation error, we need to evaluate the second order central moment of our operators.

Lemma 3.2.

For the operators Ln, n, defined by (2.6), the following relation

(Lnφx2)(x)=τ3,nτ2,n2 (3.1)

occurs, where φx(t)=|tx|.

Proof. Clearly,

(Lnφx2)(x)=(Lne2)(x)2x(Lne1)(x)+x2(Lne0)(x). (3.2)

By using (2.3) and (2.2) we can write

(Lne2)(x)=x2+τ3,nτ2,n2. (3.3)

Returning at (3.2) and using Theorem 3.1, we arrive at the desired result. 

Theorem 3.3.

Let the operators Ln, n, be defined by (2.6). For any compact interval K, the following relation

limnLnf=f uniformly on K (3.4)

occurs, provided f is continuous on .

Proof. We use Korovkin theorem [4]. In accordance with Theorem 3.1, the first two Korovkin test functions are fixed points of the operators. Relation (3.3) and Lemma 2.1 involve

limn(Lne2)(x)=x2,x.

We define the lattice homomorphism TK:C()C(K) given by

TK(f)=f|K

for every fC(). Based on above statements, we can write

limnTK(Lnej)=TK(ej),j{0,1,2},

uniformly on K and Korovkin criterion implies (3.4).

We establish the error of approximation with the help of modulus of smoothness defined as follows

ωf(δ)ω(f;δ) =sup{|f(x)f(x′′)|:x,x′′,|xx′′|δ}
=sup0hδsupx|f(x+h)f(x)|,δ0,

where fB().

Theorem 3.4.

Let the operators Ln, n, be defined by (2.6). For any Lebesgue integrable function f belonging to B(), we get

|(Lnf)(x)f(x)|2ω(f;τ3,nτ2,n2),x. (3.5)

Proof. To achieve the statement, we use the following inequality proved by Shisha and Mond [7], that says: if S is a linear positive operator, then one has

|(Sf)(x)f(x)| |f(x)||(Se0)(x)1|
+((Se0)(x)+1δ(Se0)(x)(Sφx2)(x))ω(f;δ), (3.6)

δ>0, for every bounded function f. The proof of (3) is mainly based on the following relations:

|f(x)f(y)|ω(f;|xy|),
ω(f;μδ)(1+μ)ω(f;δ),δ0,μ0.

Applying the inequality (3) for Ln operators, by choosing

δ=τ3,nτ2,n2

and taking into account both Theorem 3.1 and the identity (3.1), we obtain the inequality (3.5).

Further, we analyze the behavior of operators in some weighted spaces. For a given m, we consider the weight

ρm(x)=1+x2m,x,

and the space

Bρm()={f:|f(x)|Mfρm(x),x},

Mf being a positive constant depending only on f. The usual norm of this space is ρm defined by

hρm=supx|h(x)|ρm(x).

The operators Ln, n, are well defined for any Lebesgue integrable function belonging to Bρm() and it is easy to see that

(Lnf)(x)fρm(Lnρm)(x),x, (3.7)

takes place.

Lemma 3.5.

Each operator Ln defined by (2.6) maps Bρm() into Bρm().

Proof. Let n be fixed. In view of (3.7), it is enough to show that

(Lnρm)(x)Mρm(x),x, (3.8)

where M is a constant depending on m. We can write successively

0(Lnρm)(x) =1+1an01j=02m(2mj)(y)2mj(x+τ2,n)j(1y2α)ndy
1+12m+1j=02m(2m+1j)i=0j(ji)12ji|x|i.

We used (2.4) both for p=2m+1j (j=0,2m¯) and for p=2.
The relation we reached allows us to assert that there this a constant M satisfying (3.8).

For weighted functions belonging to the space Bρm(), we give estimates of the error |Lnf(x)f(x)|, n, involving the following weighted modulus of smoothness

Ωm(f;δ)=supx0hδ|f(x+h)f(x)|1+(x+h)2m,δ0. (3.9)

Best of our knowledge, this type of modulus associated to a function defined on +=[0,), appeared for the first time in the papers [3], [8]. Its definition formula ensures that it can be also used for functions defined on which is our case. Based on (3.9), it is obvious that Ωm(f;) is a monotone increasing function. Following the same line as in paper of Yuksel and Ispir [8, Lemma 2 (iv)] we deduce the following property

Ωm(f;λδ)(λ+1)Ωm(f;δ),δ+,λ+. (3.10)
Theorem 3.6.

Let Ln, n, be given by (2.6). For any function fBρm() and Lebesgue integrable, a constant C depending on m exists such that

|(Lnf)(x)f(x)|C(1+x2m)Ωm(f;τ3,nτ2,n2),x, (3.11)

where τp,n is given by (2.1).

Proof. Let x be fixed arbitrarily. Besides the function φx defined in Lemma 3.2, we introduce the function ψx,m given as follows

ψx,m(t)=1+(2|x|+|t|)2m,(x,t)×.

By using the definition of Ωm(f;) and property (3.10), for δ>0 we get

|f(t)f(x)| (1+(|x|+|tx|)2m)(1+1δ|tx|)Ωm(f;δ)
ψx,m(t)(1+1δφx(t))Ωm(f;δ). (3.12)

Since the operators Ln are linear and positive, consequently monotone, relation (3) and Cauchy inequality allow us to write

|(Lnf)(x)f(x)| Ln(|ff(x)|;x)
((Lnψx,m)(x)+1δLnψx,m2(x)(Lnφx2)(x))Ωm(f;δ).

Lemma 3.5 implies the existence of some constants depending on m such that

Lmψx,mM1ρmandLmψx,m2M2ρm.

Taking in view (3.1) and choosing δ=τ3,nτ2,n2, the relation (3.11) follows.

Let denote by Cρm() the subspace of all continuous functions belonging to Bρm() with the property that lim|x||f(x)|/(1+x2m) exists and it is finite. Based on [8, Lemma 2 (ii)], for each fCρm()

limδ0+Ωm(f;δ)=0

takes place. Relation (3.11) corroborated with Lemma 2.1 leads us to the following identity

limnLnffρm=0,fCρm().

References

  • [1] Gal, S., Iancu, I., Quantitative approximation by nonlinear convolution operators of Landau-Choquet type, Carpathian J. Math., 36(2020), accepter paper.
  • [2] Landau, E., Über die Approximation einer stetingen Funktion durch eine ganze rationale Funktion, Rendiconti del Circolo Matematico di Palermo, 25(1908), 337-345.
  • [3] López-Moreno, A.-J., Weighted simultaneous approximation with Baskakov type operators, Acta Math. Hungarica, 104(2004), 143-151.
  • [4] Korovkin, P.P., On convergence of linear positive operators in the space of continuous functions, Dokl. Akad. Nauk SSSR, 90(1953), 961-964.
  • [5] Mamedov, R.G., Approximation of functions by generalized linear Landau operators, (in Russian), Dokl. Akad. Nauk SSSR, 139(1961), No. 1, 28-30.
  • [6] Mocica, Gh., Problems of Special Functions, (in Romanian), Didactical and Pedagogical Press, Bucharest, 1988.
  • [7] Shisa, O., Mond, B., The degree of convergence of linear positive operators, Proc. Nat. Acad. Sci. USA, 60(1968), 1196-1200.
  • [8] Yuksel, I., Ispir, N., Weighted Approximation by a Certain Family of Summation Integral-Type Operators, Computers and Mathematics with Applications, 52(2006), 1463-1470.

[1] Altomare, F., Korovkin-type theorems and approximation by positive linear operators. Surv. Approx. Theory 5, 92–164 (2010), MathSciNet MATH Google Scholar
[2] Binmore, K.G., Mathematical Analysis: A Straightforward Approach. Cambridge University Press, Cambridge (1977), MATH Google Scholar
[3] Chen, Z., Shih, T., A new class of generalized Landau linear positive operator sequence and its properties of approximation. Chin. Q. J. Math. 13(1), 29–43 (1998), MathSciNet MATH Google Scholar
[4] Gal, S.G., Iancu, I., Quantitative approximation by nonlinear convolution operators of Landau–Choquet type. Carpathian J. Math. 37(1) (2021) (to appear)
[5] Gao, J.B., Approximation properties of a kind of generalized discrete Landau operator (Chinese). J. Huazhong Univ. Sci. Technol. 12(5), 1–4 (1984), Google Scholar
[6] Gonska, H.H., Piţul, P., Raşa, I., On Peano’s form of the Taylor remainder, Voronovskaja’s theorem and the commutator of positive linear operators. In: Agratini, O., Blaga, P. (eds.) Numerical Analysis and Approximation Theory. Proc. Int. Conf. Cluj-Napoca, pp. 55–80. Casa Cărţii de Ştiinţă, Cluj-Napoca (2006), Google Scholar
[7] Jackson, D., A proof of Weierstrass’ theorem. Am. Math. Mon. 41(5), 309–312 (1934), MathSciNet Article Google Scholar
[8] Landau, E., Über die Approximation einer stetingen Funktion durch eine ganze rationale Funktion. Rend. Circ. Mat. Palermo 25, 337–345 (1908), Article Google Scholar
[9] López-Moreno, A.-J., Weighted simultaneous approximation with Baskakov type operators. Acta Math. Hung. 104, 143–151 (2004), MathSciNet Article Google Scholar
[10] Mamedov, R.G., Approximation of functions by generalized linear Landau operators (Russian). Dokl. Akad. Nauk SSSR 139(1), 28–30 (1961), MathSciNet Google Scholar
[11] Mamedov, R.G., On the order and on the asymptotic value of the approximation of functions by generalized linear Landau operators (Russian). Akad. Nauk Azerbadzan SSR Trudy Inst. Mat. Meh. 2(10), 49–65 (1963), Google Scholar
[12] Mocica, Gh., Problems of Special Functions (Romanian). Didactical and Pedagogical Press, Bucharest (1988)
[13] Pendina, T.P., Iterations of positive linear operators of exponential type and of Landau polynomials (Russian). In: Geometric Problems of the Theory of Functions and Sets (Russian), pp. 105–111. Kalinin. Gos. Univ., Kalinin (1987)
[14] Shisha, O., Mond, B., The degree of convergence of linear positive operators. Proc. Natl. Acad. Sci. U.S.A. 60, 1196–1200 (1968), MathSciNet Article Google Scholar
[15] Sikkema, P.C., Approximation formulae of Voronovskaya type for certain convolution operators. J. Approx. Theory 26(1), 26–45 (1979), MathSciNet Article Google Scholar
[16] Veselinov, V.M., Certain estimates of the approximation of functions by de la Vallée–Poussin and Landau operators (Russian). An. Univ. Sofia Fac. Math. 66, 153–158 (1974), Google Scholar
[17] Yuksel, I., Ispir, N., Weighted approximation by a certain family of summation integral-type operators. Comput. Math. Appl. 52, 1463–1470 (2006), MathSciNet Article Google Scholar

Related Posts

Shift λ-Invariant Operators

AbstractThe present note is devoted to a generalization of the notion of shift invariant operators that we call it λ-invariant…