2010 Mathematics subject classification:

Front. Appl. Math. Stat.

Frontiers in Applied Mathematics and Statistics

Front. Appl. Math. Stat.

2297-4687

Frontiers Media S.A.

10.3389/fams.2017.00004

Applied Mathematics and Statistics

Original Research

Characterization of Local Besov Spaces via Wavelet Basis Expansions

Derevianko

Nadiia

¹ Myroniuk

Vitalii

¹ Prestin

Jürgen

² ^*

¹Department of Theory of Functions, Institute of Mathematics of NAS of Ukraine Kyiv, Ukraine ²Institut für Mathematik, Universität zu Lübeck Lübeck, Germany

Edited by: Lixin Shen, Syracuse University, USA

Reviewed by: Jian Lu, Shenzhen University, China; Xiaosheng Zhuang, City University of Hong Kong, Hong Kong

*Correspondence: Jürgen Prestin prestin@math.uni-luebeck.de

This article was submitted to Mathematics of Computation and Data Science, a section of the journal Frontiers in Applied Mathematics and Statistics

28 03 2017

2017

21 12 2016 13 03 2017

2017

Derevianko, Myroniuk and Prestin

This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

In this paper we deal with local Besov spaces of periodic functions of one variable. We characterize these spaces in terms of summability conditions on the coefficients in series expansions of their elements with respect to an orthogonal Schauder basis of trigonometric polynomials. We consider a Schauder basis that was constructed by using ideas of a periodic multiresolution analysis and corresponding wavelet spaces. As an interim result we obtain a characterization of local Besov spaces via operators of the orthogonal projection on the corresponding scaling and wavelet spaces. In order to achieve our new results, we substantially use a theorem on the discretization of scaling and wavelet spaces as well as a connection between local and usual classical Besov spaces. The corresponding characterizations are also given for the classical Besov spaces.

local Besov spaces Schauder basis Fourier coefficients periodic multiresolution analysis wavelets scaling functions trigonometric polynomials

2010 Mathematics subject classification: 42A10 42C40.

1. Introduction

One of the crucial problems in the theory of approximation is to describe the smoothness properties of functions by the behavior of the coefficients in their series expansions in terms of given bases or frames. Besov spaces and their generalizations are particularly suitable for such studies. Recent papers describing the smoothness of functions from these spaces by the decay of the coefficient sequences are e.g., Bazarkhanov [1] for Meyer wavelets, Dinh [2] for mixed B-splines, and Hinrichs et al. [3] for Faber-Schauder bases.

In the present paper we consider this problem for local Besov spaces of periodic functions of one variable with respect to some orthogonal trigonometric Schauder basis. Let us first give some motivation of our work. Let

∑n = -∞∞an(f)einx

be the Fourier series of some function f∈L2π2, a_n(f) the Fourier coefficients of f.

In view of the Parseval's equality, it is easy to obtain the following result about the description of the usual classical Besov spaces B2,θα of periodic functions: f∈B2,θα, α > 0, if and only if f∈L2π2 and the norm

(1.1){(|a0(f)|θ+∑n = 1∞[2nα(∑2n−1≤|k|<2n|ak(f)|2)1/2]θ)1/θ,1≤θ<∞,supn∈ℕ(|a0(f)|+2nα(∑2n−1≤|k|<2n|ak(f)|2)1/2),θ=∞,

is finite.

On the other hand, local properties of functions from the Besov spaces can be investigated by expanding them in a series with respect to the Haar basis. To give a short description of these results let for n ∈ ℕ, ν(n) = {k: k = 0, …, 2ⁿ⁻¹−1}, ν(0) = {0} be the sets of indices and {h_{n, k}}, n ∈ ℕ₀, k ∈ ν(n) be the Haar system. This system is an orthogonal Schauder basis in the space L^p([0, 1]), 1 ≤ p < ∞, and for every f ∈ L^p([0, 1])

f(x)=∑n = 0∞∑k∈ν(n)bn,k(f)hn,k(x),

in the sense of the norm of L^p([0, 1]) [4, Chap. 3].

Romanyuk [5] obtained necessary and sufficient conditions on the Fourier-Haar coefficients b_{n, k}(f) at which functions from L^p([0, 1]) belong to the Besov spaces. Namely, let 1 ≤ p, θ < ∞, 0 < α < 1/p, then f∈Bp,θα if and only if f ∈ L^p([0, 1]) and the norm

(1.2)(∑n = 0∞[2n(α-1/p+1/2)(∑k∈ν(n)|bn,k(f)|p)1/p]θ)1/θ

is finite. In this case it is evident that the Fourier-Haar coefficients b_{n, k}(f) describe the local behavior of the function f. Note that V. Romanyuk considered the multivariate case, but since in the present paper we investigate functions of one variable, we formulate his result only in the univariate case.

Our aim in this paper is to combine these two approaches and to describe local smoothness of periodic functions in terms of summability conditions on the Fourier coefficients with respect to an orthogonal Schauder basis of trigonometric polynomials in the space L2πp for all 1 ≤ p ≤ ∞. The local smoothness is understood in the sense of Besov spaces. We call these Besov-type spaces as local Besov spaces (see Subsection 2.1 for a definition).

Note that some results in this direction were obtained by Mhaskar and Prestin [6]. There expansions of functions from the local Besov spaces in series with respect to a system of trigonometric frames were considered and these spaces were described via coefficients of these expansions. However, this system is not a Schauder basis.

Let us sketch the main results of the present paper. Let ψ₀ be a scaling function of a periodic multiresolution analysis (PMRA) generated by de la Vallée Poussin means and ψ_n, n ∈ ℕ, be corresponding wavelets [7]. Let Nn:=3·2n for n ∈ ℕ and N₀: = N₁. By ψ_{n, s} we denote shifts of ψ_n:

(1.3)ψn,s(·):=ψn(·-sπNn), n∈ℕ0, s=0,…,2Nn-1.

We show that for a particular choice of ψ_n the system {ψ_{n, s}} constitutes an orthogonal trigonometric Schauder basis in the space L2πp, 1 ≤ p ≤ ∞, and a function f∈L2πp can be represented by a series (for more detailed information see Subsection 2.2)

(1.4)f(x)=∑n = 0∞∑s = 02Nn-1an,s(f)ψn,s(x), x∈ℝ,

converging in the norm of the space L2πp, where the coefficient functionals a_{n, s}(f) are Fourier coefficients of f with respect to the basis {ψ_{n, s}}:

an,s(f)=〈f,ψn,s〉=12π∫02πf(x)ψn,s(x)dx.

Because ψ_n is even (see Subsection 2.2 for definition) and from (1.3) we conclude that a_{n, s}(f) can be represented in the following way:

an,s(f)=12π∫02πf(x)ψn(sπNn-x)dx=(f*ψn)(sπNn),

where f * g means the convolution

(f*g)(x):=12π∫02πf(y)g(x-y)dy, x∈ℝ.

For more information regarding trigonometric Schauder bases we refer to Lorentz and Saakyan [8], Prestin and Selig [7], Selig [9] and the references cited there.

Let I ⊂ ℝ, |I| < 2π, be some segment and n ∈ ℕ. By κ(I, n) we denote the set of indices s which satisfy the properties s = 0, …, 2N_n − 1 and there exists k ∈ ℤ such that the point (s+2kNn)πNn belongs to the segment I. For 1 ≤ p ≤ ∞, n ∈ ℕ₀, and the segment I we define the following sequence

(1.5)cn(I,p):={Nn1/2-1/p(∑s∈κ(I,n)|an,s(f)|p)1/p, 1≤p<∞,Nn1/2maxs∈κ(I,n)|an,s(f)|, p=∞.

Let 1 ≤ p ≤ ∞, f∈L2πp, x₀ ∈ [0, 2π), 0 < α < 1, 0 < θ ≤ ∞. Then the main result of this paper is written as follows: A function f belongs to the local Besov spaces Bp,θα(x0) if and only if there exists an interval I ⊂ ℝ, |I| < 2π, centered at x₀, such that the norm

{(∑n = 0∞2nθα|cn(I,p)|θ)1/θ, 0<θ<∞, supn∈ℕ02nα|cn(I,p)|, θ=∞,

is finite.

We adopt the following convention regarding constants. The letters C, C_i, i = 1, 2, …, mean positive constants which may depend on parameters fixed for the spaces. Their values are not necessarily the same in different parts of the text. When constants depend on functions f, ζ or some intervals I, we indicate this in brackets.

The present paper is organized as follows: In Subsection 2.1 we define the local Besov spaces and formulate their connection with classical Besov spaces. In Subsection 2.2 we give definitions of the orthogonal trigonometric Schauder basis that we work with and describe expansions of functions from L2πp in a series with respect to this basis. In Section 3 we formulate the main results of this paper. In Section 4 we prove the main auxiliary statements. In Section 5 we give proofs of the main results of the present paper.

2. Preliminaries

As stated in Section 1, our aim in this paper is to describe the local Besov spaces of functions f in terms of summability conditions on the coefficients in a series expansion of f as in (1.4).

Let us first agree about the notation. As usual ℕ is reserved for the natural numbers, by ℕ₀ we denote the natural numbers including 0, by ℤ the set of all integers and by ℝ the set of all real numbers.

Let 1 ≤ p ≤ ∞ and A ⊂ ℝ be a Lebesgue measurable set. By L^p(A) we denote the space of functions f: A → ℝ Lebesgue measurable on A with the finite norm

‖f‖A,p:={(∫A|f(t)|pdt)1/p, 1≤p<∞,ess supt∈A|f(t)|, p=∞.

When A = [0, 2π), we understand by L^p([0, 2π)) the space of functions f defined on the segment [0, 2π) and extended 2π-periodically to the real line with natural modification for the norm ‖f‖[0,2π),p:=(12π∫02π|f(t)|pdt)1/p. For simplicity, by L^∞([0, 2π)) we denote the space of 2π-periodic continuous functions (equipped with ||f||_{[0, 2π), ∞} as its norm). Further, we will write L2πp instead of L^p([0, 2π)) and ||f||_p instead of ||f||_{[0, 2π), p}.

2.1. The local Besov spaces of functions and their connection with classical Besov spaces

Let us introduce the local Besov spaces Bp,θα(x0) of periodic functions. For a function f ∈ L^p(I) where I = [a, b] ⊆ ℝ we define the rth difference operator Δtr by

Δtrf(·)=∑k = 0r(−1)r−k(rk)f(·+kt), r∈ℕ, t∈ ℝ,

and for 0 < δ < (b − a)/r we define the modulus of smoothness by

ωI,r,p(f,δ):=sup0<t≤δ‖Δtrf(·)‖[a,b-rδ],p.

If δ ≥ (b − a)/r, we put

ωI,r,p(f,δ):=inf‖f-P‖I,p,

where the infimum is taken over all algebraic polynomials of degree at most r−1.

It will be convenient for us to use a sequential version of the Besov spaces which we now define. For a sequence a={an}n=0∞ and numbers α, θ > 0 by ||a||_{θ, α} we denote the following norm

‖a‖θ,α:={(∑n = 0∞(2nα|an|)θ)1/θ,0<θ<∞, supn∈ℕ02nα|an|,θ=∞.

The notation a ∈ b_θ,α means that the norm ||a||_θ,α is finite.

Let 1 ≤ p ≤ ∞, 0 < θ ≤ ∞, α > 0 and r = [α]+1. For x₀ ∈ ℝ the local Besov space Bp,θα(x0) is the collection of functions f which satisfy the following properties:

f∈L2πp;

there is a non-degenerate interval I ⊂ ℝ, |I| < 2π, centered at x₀, such that {ωI,r,p(f,2-n)}n=0∞∈bθ,α.

The spaces Bp,θα(x0) were considered in Mhaskar and Prestin [6]. By periodicity we can restrict ourselves to points x₀ ∈ [0, 2π).

In order to prove our main results, we use a connection between the local and the classical Besov spaces of periodic functions. Let us define these Besov spaces.

Let f∈L2πp, 1 ≤ p ≤ ∞. For δ > 0 we define the modulus of smoothness by

ωr,p*(f,δ):=sup0<t≤δ‖Δtrf(·)‖p,

where, in contrast to the modulus of smoothness ω_{I, r, p}(f, δ), the norm is taken over the entire period of f, using the periodicity of f in the case when the translates go outside of [0, 2π). Let 1 ≤ p ≤ ∞, α > 0, 0 < θ ≤ ∞ and r be some integer number greater than α. The classical Besov space Bp,θα* consists of functions f such that f∈L2πp and the sequence {ωr,p*(f,2-n)}n=0∞∈bθ,α for some integer r > α. The space Bp,θα* is independent of the choice of r as long as r > α (see [10, Theorem 10.1, Chapter 2]). One can find more information about Besov spaces in the monographs [10, 11]. The following statement about the connection between the two spaces mentioned above is proved in the paper [6].

Proposition A. Let 1 ≤ p ≤ ∞, f∈L2πp, α > 0, 0 < θ ≤ ∞, x₀ ∈ [0, 2π). Then f∈Bp,θα(x0) if and only if there exists an interval I ⊂ ℝ, |I| < 2π, centered at x₀ such that for every infinitely differentiable on ℝ function ζ supported on I and extended as a 2π-periodic function, the function fζ is in Bp,θα*.

2.2. Expansions in a series

Let us first give some necessary definitions. Let

(2.1)g1(x):={2, -π2+π6≤x≤π2-π6,1-3(2|x|π-1)1+(3(2|x|π-1))2, π2-π6≤|x|≤π2+π6,0, π2+π6≤|x|≤π,

be a coefficient function given on the segment [−π, π] and let as above Nn=3·2n and Mn:=2n, n ∈ ℕ. By φ_n we denote the following function

φn(·):=12Nn∑k=-Nn-MnNn+Mng1(kπ2Nn)cosk·,

which is a modification of the de la Vallée Poussin kernel.

Let as above φ_{n, s} denote shifts of the function φ_n:

φn,s(·):=φn(·-sπNn), n∈ℕ, s = 0,…,2Nn-1.

In Prestin and Selig [7] it is proved that the sequence of spaces {Vn}n=1∞, defined as

Vn:=span{φn,s,s = 0,…,2Nn-1},

provides a PMRA in L2π2, i.e., the system {φ_{n, s}, s = 0, …, 2N_n−1} is a basis in the space V_n; V_n⊂V_{n + 1} for all n ∈ ℕ; it holds that closL2π2(⋃n∈ℕVn) = L2π2. The functions φ_n are called scaling functions of this PMRA and the spaces V_n are called scaling spaces.

The wavelet space W_n which is defined to be the orthogonal complement of V_n with respect to V_{n + 1}, i.e., W_n = V_{n + 1}⊖V_n, is spanned by the translates of the function ψ_n [7]:

Wn=span{ψn,s,s = 0,…,2Nn-1},

where

ψn(·)=1Nn∑k=-Mn2Nn-Mng2(kπNn)cos(Nn+k)·,

and

(2.2)g2(x)={1+3xπ1+(3xπ)2, -π3≤x≤π3,2-3(xπ-1)4+(3(xπ-1))2, π3<x≤2π-π3,

is a coefficient function given on the segment [-π3,2π-π3]. The functions ψ_n are called wavelets. Note that the functions g₁ and g₂ defined on the segments with lengths equal 2π can be extended to ℝ as 2π-periodic functions (and continuous since g₁(−π) = g₁(π) = 0 and g2(-π3)=g1(2π-π3)=0).

For simplicity of notations we denote ψ₀: = φ₁, N₀: = N₁ and for n ∈ ℕ₀, s = 0, …, 2N_n−1,

t2Nn+s(·):=ψn,s(·).

For f∈L2πp, 1 ≤ p ≤ ∞, we define some trigonometric polynomial operators σ_n. Let n ∈ ℕ₀, x ∈ ℝ and

σn(f,x):=∑k = 02Nn-2-1〈f,tk〉tk(x), n≥3,

where 〈f,tk〉=12π∫02πf(x)tk(x)dx are Fourier coefficients of the function f with respect to the basis {tk}k = 0∞. It is convenient for us to put σ_n(f, x) ≡ 0 if n = 0, 1, 2.

We also use the following representation of the operators σ_n:

(2.3)σn(f,x)=12π∫02πf(y)Kn(x,y)dy,

where Kn(x,y)=∑k = 02Nn-2-1tk(x)tk(y).

Using a similar technique as in Prestin and Selig [7], one can prove that the system of polynomials {tk}k = 0∞ is an orthogonal trigonometric Schauder basis in the space L2π∞. In view of Theorem 9 [4, p. 12] and Theorem 6 [4, p. 10] we get that {tk}k = 0∞ is a Schauder basis in the space L2πp, 1 ≤ p < ∞, and

(2.4)‖σn(f,·)‖p≤C‖f(·)‖p,

with some constant C > 0.

By 𝕋_n, n ∈ ℕ, we denote the set of all trigonometric polynomials of the form

T(x)=∑k = 0n(akcoskx+bksinkx), x∈ℝ.

From (2.3) and the representation of the kernel K_n [7, p. 421], it can be derived that

(2.5)σn(T,·)=T(·), T∈T2n-1.

Let

En(f)p:=infT∈Tn‖f-T‖p

be the best approximation of a function f∈L2πp by trigonometric polynomials from 𝕋_n.

A sequence of linear operators Un: L2πp→T2n, n ∈ ℕ₀, is called a sequence of near best approximation (with the constant λ > 0) for L2πp if it satisfies the following condition:

‖f-Un(f)‖p≤λE2n-1(f)p.

For example, in the case p = ∞ the operators of de la Vallée Poussin Vn2n determine a sequence of near best approximation with the constant λ=43+23π [12, Chap. 5, §2].

The following lemma results from the properties (2.4) and (2.5).

Lemma 2.1. {σn(f)}n = 0∞ is a sequence of near best approximation (with some constant) for L2πp, 1 ≤ p ≤ ∞.

Further, for f∈L2πp, 1 ≤ p ≤ ∞, we define operators τ_n, n ∈ ℕ₀, as follows:

τn(f,x)=σn+3(f,x)-σn+2(f,x), x∈ℝ.

From Lemma 2.1 it is easy to see that a function f∈L2πp, 1 ≤ p ≤ ∞, can be represented by the series

(2.6)f(x)=∑n = 0∞τn(f,x), x∈ℝ,

where convergence is understood in the metric of the space L2πp.

Using the definition of the operators τ_n, we can represent them in the following form:

(2.7)τ0(f,x)=∑k = 02N1-1〈f,tk〉tk(x)=∑s = 02N0-1〈f,ψ0,s〉ψ0,s(x),

and for n ∈ ℕ

(2.8)τn(f,x)=∑2Nn≤k<2Nn+1-1〈f,tk〉tk(x)=∑s = 02Nn-1〈f,ψn,s〉ψn,s(x).

From (2.6)–(2.8) we get representation (1.4).

3. Formulations of the main results

In this section we formulate the main results of this paper. Let us first explain the relationship between the local Besov spaces Bp,θα(x0) and the behavior of the operators τ_n near the point x₀. This behavior will be described by the condition that certain norms of the operators belong to a sequential version of the Besov spaces.

Theorem 3.1. Let 1 ≤ p ≤ ∞, f∈L2πp, x₀ ∈ [0, 2π), 0 < α < 1, 0 < θ ≤ ∞. The following statements are equivalent:

f∈Bp,θα(x0);

There exists an interval I ⊂ ℝ, |I| < 2π, centered at x₀ such that the sequence {‖τn(f,·)‖I,p}n=0∞∈bθ,α;

There exists an interval I ⊂ ℝ, |I| < 2π, centered at x₀ such that for every infinitely differentiable on ℝ function ζ supported on I and extended as a 2π-periodic function, the sequence {‖τn(fζ,·)‖p}n=0∞∈bθ,α.

Let further for 1 ≤ p ≤ ∞ and a segment I, {cn(I,p)}n=0∞ be a sequence defined by the formula (1.5). And let ζ be some infinitely differentiable on ℝ function supported on I and extended as a 2π-periodic function. By {dn(I,ζ,p)}n=0∞ we denote the sequence:

dn(I,ζ,p):={Nn1/2-1/p(∑s = 02Nn-1|an,s(fζ)|p)1/p, 1≤p<∞,Nn1/2maxs = 0,…,2Nn-1|an,s(fζ)|, p=∞.

Theorem 3.2. Let 1 ≤ p ≤ ∞, f∈L2πp, x₀ ∈ [0, 2π), 0 < α < 1, 0 < θ ≤ ∞. The following statements are equivalent:

f∈Bp,θα(x0);

There exists an interval I ⊂ ℝ, |I| < 2π, centered at x₀ such that the sequence {cn(I,p)}n = 0∞∈bθ,α;

This theorem is the discrete version of Theorem 3.1 in the sense that the L2πp norm of the operators τ_n is replaced by a corresponding discrete norm (see Theorem 4.1).

For classical Besov spaces Bp,θα*, α > 0, we can obtain a result similar to Theorems 3.1 and 3.2 which is essentially of the same kind as (1.1) and (1.2). To formulate this equivalence we introduce the following sequence

an(p):={Nn1/2-1/p(∑s = 02Nn-1|an,s(f)|p)1/p, 1≤p<∞,Nn1/2maxs = 0,…,2Nn-1|an,s(f)|, p=∞.

Theorem 3.3. Let 1 ≤ p ≤ ∞, f∈L2πp, α > 0, 0 < θ ≤ ∞. The following statements are equivalent:

f∈Bp,θα*;

The sequence {‖τn(f,·)‖p}n = 0∞∈bθ,α;

The sequence {an(p)}n = 0∞∈bθ,α.

4. The main auxiliary statements 4.1. A property of the kernel <italic>K</italic><italic>n</italic>(<italic>x, y</italic>)

In this subsection we present some estimates for the kernel K_n(x, y). Further, by V[f] we mean the total variation of a function f:ℝ → ℝ and by Vab[f] the total variation of f defined on the segment [a, b].

Lemma 4.1. There exists a constant C > 0 such that for every n ∈ ℕ, n ≥ 3:

|Kn(x,y)|≤C2n|x-y|2, 0<|x-y|≤2π.

Proof. Let N = 3·2ⁿ⁻², M = 2ⁿ⁻², n ≥ 3. From Selig [9, pp. 91–93] it is known that

(4.1)Kn(x,y)=QNM(y-x)+cosN(y+x) RM(y-x),

where

QNM(x)=1+2∑k=1N-Mcos kx + ∑k = -M+1M-1(M-k)2M2+k2 cos (N+k)x=∑k = -N+MN-Mcos kx+12∑|k|=N-M+1N+M-1(M-(|k|-N))2M2+(|k|-N)2 cos kx,RM(x)=1+2∑k = 1M-1M2-k2M2+k2 cos kx=∑k = -M+1M-1M2-k2M2+k2 cos kx.

Defining continuous coefficient functions by

g(x)={1-x21+x2, -1≤x≤1,0, else;

fβ(x)={1, |x|≤1-β,(β-(|x|-1))22(β2+(|x|-1)2), 1-β≤|x|≤1+β,0, else,

where β ∈ (0, 1), we can rewrite the polynomials R_M and QNM in the following forms:

RM(x)=∑k∈ℤg(kM)coskx, QNM(x)=∑k∈ℤfMN(kN)coskx.

Let us first estimate |R_M(y − x)|. Using Proposition 2.2 [13], for 0 < |x − y| ≤ 2π we get

(4.2)|RM(y−x)| = |∑k∈ℤg(kM)cosk(y−x)|≤C|x−y|2∑k∈ℤ|Δ2g(kM)|,

where Δ2g(kM)=g(kM)-2g(k+1M)+g(k+2M).

Then, using the mean value theorem, we obtain that there exist points l_k ∈ (k, k+1), k ∈ ℤ, such that

Δ2g(kM)=g(kM)-2g(k + 1M)+g(k+2M)=1M(g′(lk + 1M)-g′(lkM)).

Since the system of the points {lkM,k∈ℤ} is some partition of the real line and g is a function having a first derivative of finite total variation V[g′], it holds that

(4.3)∑k∈ℤ|Δ2g(kM)|=∑k∈ℤ1M|(g′(lk+1M)−g′(lkM))| ≤1MV[g′]=C2nV[g′].

Since g′ = 0 outside of the segment [−1, 1], it holds that V[g′]=V-11[g′].

From the first and second derivatives

g′(x)= -4x(1 + x2)2, g″(x)=4(3x2-1)(1+x2)3,

we see that g″ = 0 at the points -33 and 33 and g′ is monotonously increasing on [-1,-33] and on [33,1] and monotonously decreasing on [-33,33]. Therefore, the extrema are at -33 and 33:

supx∈[−1,1]g′(x)=g′(−33)= −g′(33)=334.

Since g′ (−1) > 0 and g′(1) < 0, the total variation of g′ is

(4.4)V[g′]=V-11[g′]=4supx∈[-1,1]g′(x)=33.

From (4.2), (4.3) and (4.4), we get

(4.5)|RM(y-x)|≤C2n|x-y|2, 0<|x-y|≤2π.

Analogously, we can estimate |QNM(y-x)|. Let us calculate V[fβ′] where

fβ′(x)={-β(β2-(x-1)2)β2+(x - 1)2, 1-β≤x≤1+β,β(β2-(-x-1)2)β2+(-x -1)2, -1-β≤x≤-1+β,0, else.

Since fβ′=0 outside of the segments [−1 − β, −1 + β] and [1 − β, 1 + β] and fβ′ is an odd function, it holds that V[fβ′]=2V-1-β-1+β[fβ′].

From the second derivative

fβ″(x)=2β(1+x)(-3β2+(1+x)2)(β2+(1 + x)2)3, x∈[-1-β,-1+β],

we can see that fβ″=0 at the point − 1 and fβ′ is monotonously increasing on [−1−β, −1] and monotonously decreasing on [−1, −1+β]. This means that

supx∈[-1-β,-1+β]fβ′(x)=fβ′(-1)=1β.

Since fβ′(-1-β)=fβ′(-1+β)=0, it holds that

V-1-β-1+β[fβ′]=2supx∈[-1-β,-1+β]fβ′(x)=2β,

and the total variation of fβ′ is

(4.6)V[fβ′]=4β.

Then, using Proposition 2.2 [13], for 0 < |x − y| ≤ 2π we get

(4.7)|QNM(y−x)| = |∑k∈ℤfMN(kN)cosk(y−x)| ≤C|x−y|2∑k∈ℤ|Δ2fMN(kN)|,

where Δ2fMN(kN)=fMN(kN)-2fMN(k+1N)+fMN(k+2N).

From the mean value theorem it follows that there exist points s_k ∈ (k, k + 1), k ∈ ℤ, such that

(4.8)Δ2fMN(kN)=1N(f′MN(sk+1N)−f′MN(skN)).

Since the system of the points {skN,k∈ℤ} is some partition of the real line and fMN is a function having a first derivative of finite total variation, from (4.6) and (4.8) it can be derived that

(4.9)∑k∈ℤ|Δ2fMN(kN)|≤∑k∈ℤ1N|(f′MN(sk+1N)−f′MN(skN))| ≤1NV[f′MN]=C2n.

In view of (4.7) and (4.9), we have

(4.10)|QNM(y-x)|≤C2n|x-y|2, 0<|x-y|≤2π.

Finally, from (4.1), (4.5) and (4.10) we conclude that

|Kn(x,y)|≤C2n|x-y|2, 0<|x-y|≤2π.

□

4.2. A property of the spaces <italic>V</italic><italic>n</italic> and <italic>W</italic><italic>n</italic>

In this subsection we formulate and prove the main auxiliary statement. Let us make some preparations for this. We use the Minkowski inequality in the following form:

(4.11)(∑l = 1m|∫abfl(x)dx|p)1/p≤∫ab(∑l = 1m|fl(x)|p)1/pdx, 1≤p<∞

(under appropriate conditions on the functions f_l which appear above). One can prove this inequality by using a similar technique as in the proof of the generalized Minkowski inequality [14, pp. 18–19].

Since the coefficient functions g₁ and g₂ defined by the formulas (2.1) and (2.2) on the segments of length 2π have the first derivatives (on corresponding segments) of finite total variation, by similar techniques as in Lemma 4.3 [7] one can prove the following estimates (with some constants C₁ and C₂):

(4.12)‖∑s = 02Nn−1|φn,s(·)|‖∞≤C1Nn,

(4.13)‖∑s = 02Nn−1|ψn,s(·)|‖∞≤C2Nn.

Theorem 4.1. Let n ∈ ℕ and (an,s)s = 02Nn−1 be an arbitrary sequence of real numbers. Then, there exist constants C_i > 0, i = 1, 2, 3, 4, such that for 1 ≤ p < ∞ the following inequalities hold:

(4.14)C12n(12−1p)(∑s = 02Nn−1|an,s|p)1p≤‖∑s = 02Nn−1an,sψn,s(·)‖p≤C22n(12−1p)(∑s = 02Nn−1|an,s|p)1p.

For p = ∞ we have

(4.15)C32n2maxs = 0,…,2Nn−1|an,s|≤‖∑s = 02Nn−1an,sψn,s(·)‖∞ ≤C42n2maxs = 0,…,2Nn−1|an,s|.(4.15)

Proof. For the sake of simplicity we denote

(4.16)τ(x)=∑s = 02Nn−1an,sψn,s(x).

First, we prove the right-hand side of (4.14). For p = 1 we have

(4.17)‖τ(·)‖1=12π∫02π|∑s = 02Nn−1an,sψn,s(x)|dx ≤∑s = 02Nn−1|an,s|12π∫02π|ψn,s(x)|dx=∑s = 02Nn−1|an,s|‖ψn(·−sπNn) ‖1=‖ψn(·)‖1∑s = 02Nn−1|an,s|.

Let us estimate ||ψ_n(·)||₁. For the L2π1-norm of a polynomial T ∈ 𝕋_n it holds that [15, p. 228]

‖T(·)‖1 ≤supx2πm∑s = 0m−1|T(x−2πsm)|, m∈ℕ.

Applying this estimate to ψ_n with m = 2N_n and using the inequality (4.13), we obtain

(4.18)‖ψn(·)‖1≤12Nn‖∑s = 02Nn−1|ψn(·−sπNn) | ‖∞≤C2Nn≤C2−n/2.

From (4.17) and (4.18) we derive

‖τ(·)‖1≤C2−n/2∑s = 02Nn-1|an,s|.

Let 1 < p < ∞. Using the inequality of Hölder for sums

∑k|akbk|≤(∑k|ak|p)1/p(∑k|bk|p′)1/p′, 1p+1p′=1,

and the inequality (4.13), we obtain

|∑s = 02Nn−1an,sψn,s(x)|≤∑s = 02Nn−1|an,s||ψn,s(x)| =∑s = 02Nn−1|an,s||ψn,s(x)|1p+1p′ ≤(∑s = 02Nn−1|an,s|p|ψn,s(x)|)1p ×(∑s = 02Nn−1|ψn,s(x)|)1p′ ≤(∑s = 02Nn−1|an,s|p|ψn,s(x)|)1p ×(supx∈[0,2π)∑s = 02Nn−1|ψn,s(x)|)1p′ ≤C2n2p′(∑s = 02Nn−1|an,s|p|ψn,s(x)|)1p.

Therefore,

|∑s = 02Nn−1an,sψn,s(x)|p≤C2np2p′∑s = 02Nn−1|an,s|p|ψn,s(x)|.

In view of (4.18), we have

‖τ(·)‖pp=12π∫02π|∑|s = 02Nn−1an,sψn,s(x)|pdx ≤C2np2p′∑s = 02Nn−1|an,s|p12π∫02π|ψn,s(x)|dx = C2np2p′‖ψn(·)‖1∑s = 02Nn−1|an,s|p ≤C2np2p′2−n2∑s = 02Nn−1|an,s|p=C2np(12−1p)∑s = 02Nn−1|an,s|p,

and this implies that

‖τ(·)‖p≤C2n(12−1p)(∑s = 02Nn−1|an,s|p)1/p.

Now, we prove the left-hand side of (4.14). Taking the inner product in (4.16) with Tnsψn, we get that for all s = 0, …, 2N_n − 1

(4.19)an,s=〈τ,ψn,s〉=12π∫02πτ(x)ψn(x-sπNn)dx.

Let p = 1. Using (4.19) and the estimation (4.13), we obtain

∑s = 02Nn−1|an,s| =∑s = 02Nn−112π|∫02πτ(x)ψn(x−sπNn)dx|≤12π∫02π|τ(x)|∑s = 02Nn−1|ψn(x−sπNn)|dx≤supx∈[0,2π)∑s = 02Nn−1|ψn(x−sπNn)|×12π∫02π|τ(x)|dx≤C2n/2‖τ(·)‖1.

Finally, C2-n/2∑s = 02Nn-1|an,s|≤‖τ(·)‖1.

Let now 1 < p < ∞. From (4.19) and the inequality (4.11) we derive that

For the L2πp-norm of a polynomial T ∈ T_n we use the following inequality [15, p. 228]:

supx(2πm∑s = 0m-1|T(x-2πsm)|p)1p≤(1+2nπm)‖T(·)‖p, m∈ℕ.

Applying this estimation to τ with m = 2N_n and using the inequality (4.18), we obtain that

(∑s = 02Nn-1|an,s|p)1/p≤C2n(1p-12)‖τ(·)‖p.

Let p = ∞. The right-hand side of the inequality (4.15) follows from obvious inequalities and the inequality (4.13):

‖τ(·)‖∞=supx∈[0,2π)|∑s = 02Nn−1an,sψn,s(x)| ≤supx∈[0,2π)∑s = 02Nn−1|an,s||ψn,s(x)|≤maxs = 0,…,2Nn−1|an,s|supx∈[0,2π]∑s = 02Nn−1|ψn(x−sπNn)|=maxs = 0,…,2Nn−1|an,s|‖∑s = 02Nn−1|ψn(·−sπNn)| ‖∞≤C2Nnmaxs = 0,…,2Nn−1|an,s|≤C2n/2maxs = 0,…,2Nn−1|an,s|.

Let us prove the left-hand side of (4.15). From the inequality (4.19) for s = 0, …, 2N_n − 1 we get

|an,s|≤12π∫02π|τ(x)‖ψn(x-sπ/Nn)|dx≤‖τ‖∞‖ψn‖1≤C2-n/2‖τ‖∞.

Finally, we conclude:

C2n/2maxs = 0,…,2Nn-1|an,s|≤‖τ‖∞.

□

We formulated and proved Theorem 4.1 for a polynomial τ ∈ W_n, but using the same techniques and the inequality (4.12) instead of (4.13) in the corresponding places of the proof, one can prove a similar theorem for a polynomial τ ∈ V_n.

5. Proof of the main results 5.1. Proof of Theorem 3.1

In order to prove Theorem 3.1, we need some known statements from the paper Mhaskar and Prestin [6], Theorem A and Lemma A, and the following Lemma 5.1. Note that in the proofs of this Lemma and Theorem 3.1 we use similar considerations as in the proofs of Lemma 4.2 and Theorem 2.1 in Mhaskar and Prestin [6].

Theorem A. Let 1 ≤ p ≤ ∞, f∈L2πp, α > 0, 0 < θ ≤ ∞, and {U_n} be a sequence of near best approximation (with some constant) for L2πp. The following statements are equivalent:

f∈Bp,θα*;

{E2n(f)p}∈bθ,α;

{||U_n(f) − U_{n − 1}(f)||_p} ∈ b_{θ, α}.

Lemma A. Let 1 ≤ p ≤ ∞, f∈L2πp, α > 0, 0 < θ ≤ ∞, and {U_n} be a sequence of near best approximation (with some constant) for L2πp. If, for some interval I centered at x₀, the sequence {||U_n(f) − U_n−1(f)||_I,p} ∈ b_θ,α, then for every infinitely differentiable on ℝ function ζ supported on I and extended 2π-periodically, the function fζ is in Bp,θα*.

Lemma 5.1. Let I ⊂ ℝ, |I| < 2π, be an interval centered at x₀, J₁ and J be intervals centered at x₀ such that J ⊂ J₁ ⊂ I, ζ be an infinitely differentiable on ℝ function supported on I and extended as a 2π-periodic function such that ζ(x) = 1 for all x ∈ J₁ and let f∈L2π1. Then, for the operator

σn(f,x)=12π∫02πf(y)Kn(x,y)dy, n≥3,

we have

‖σn((1-ζ)f,·)‖J,∞≤C(I,ζ,f)2n.

Proof. Without loss of generality we can assume that J₁ is an interval with length equal to |I|/2 and J is an interval with length equal to |I|/4. For x ∈ J, we have

Since x ∈ J and y ∈ [0, 2π)\J₁, then |I|/8 < |x − y|. Using Lemma 4.1, we get

|σn((1-ζ)f,x)|≤12π∫[0,2π)\J1|f(y)(1-ζ(y))|C2n|x-y|2dy≤C2π(|I|/8)212n∫[0,2π)\J1|f(y)(1-ζ(y))|dy≤C(I,ζ,f)2n.

Therefore,

‖σn((1-ζ)f,·)‖J,∞≤C(I,ζ,f)2n.

□

Proof of Theorem 3.1. Let part (a) hold. In view of Proposition A, it is equivalent to the fact that there exists an interval I centered at x₀ such that for every infinitely differentiable on ℝ function ζ supported on I and extended as a 2π-periodic function, the function fζ is in Bp,θα*. According to Theorem A (with σ_n instead of U_n) it is equivalent to part (c). Thus, parts (a) and (c) are equivalent.

Let part (c) hold and let I be the interval chosen as in that part, let J₁ and J be intervals centered at x₀ such that J ⊂ J₁ ⊂ I and ζ be an infinitely differentiable on ℝ function supported on I and extended as 2π-periodic such that ζ(x) = 1 for all x ∈ J₁.

From the obvious inequalities, we get

‖τn(f,·)‖J,p≤‖τn(fζ,·)‖J,p+‖τn(f(1-ζ),·)‖J,p≤‖τn(fζ,·)‖p+‖τn(f(1-ζ),·)‖J,∞≤‖τn(fζ,·)‖p+ ‖σn(f(1-ζ),·)‖J,∞+ ‖σn-1(f(1-ζ),·)‖J,∞,

and using Lemma 5.1, we obtain

‖τn(f,·)‖J,p≤‖τn(fζ,·)‖p+C(I,ζ,f)2n.

Since α ∈ (0, 1), {2-n}n=1∞∈bθ,α and from the condition of part (c) we know that {‖τn(fζ,·)‖p}n=0∞∈bθ,α. Therefore, {‖τn(f,·)‖J,p}n=0∞∈bθ,α and part (b) is proved.

Let part (b) hold, and I be the interval chosen as in that part and let ζ be an infinitely differentiable on ℝ function supported on I and extended as a 2π-periodic function. In view of Lemma A (applied with σ_n in place of U_n), we get that fϕ∈Bp,θα*. According to Theorem A this means that {‖τn(fζ,·)‖p}n=0∞∈bθ,α. This proves part (c).

□

5.2. Proof of Theorem 3.2

Let us first formulate and prove some auxiliary statements.

Lemma 5.2. There exist constants C₁, C₂ > 0 such that for every n ∈ ℕ and 0 < x ≤ 2π:

(5.1)|φn(x)|≤C123n/2|x|2,

(5.2)|ψn(x)|≤C223n/2|x|2.

Proof. First, we prove estimation (5.1). Note that we can define the scaling functions φ_n using an aperiodic coefficient function

φn(·)=12Nn∑k∈ℤg3(kNn)cos k·,

where

g3(x)={2, |x|≤1-13,1-3(|x|-1)1+9(|x|-1)2, 1-13≤|x|≤1+13,0, else.

Let us calculate V[g3′]. Since g3′=0 outside of the segment [−4/3, 4/3], it holds that V[g3′]=V-4343[g3′]. From the first and second derivatives of g₃

g3′(x)={6-9x(9x2-18x+10)3/2, 23≤x≤43,-6+9x(9x2+18x+10)3/2, -43≤x≤-23,0, else,

g3″(x)={9(18x2-27x+8)(9x2-18x+10)5/2, 23≤x≤43,-9(18x2+27x+8)(9x2+18x+10)5/2, -43≤x≤-23,0, else,

we get that g3″=0 at the points -112(9+17) and -112(9+17) and g3′ is monotonously increasing on the segments [-43,-112(9+17)] and [112(9+17),43] and monotonously decreasing on the segments [-112(9+17),-23] and [23,-112(9+17)]. This means that the extrema are at the points -112(9+17) and 112(9+17):

supx∈[-4/3,4/3]g3′(x)=g3′(-112(9+17))=-g3′(112(9+17))=(3+17)7+1743.

Since g3′(-43)>0 and g3′(43)<0, the total variation of g3′ is

(5.3)V[g3′]=V-4343[g3′]=4supx∈[-4/3,4/3]g3′(x)=(3+17)7+173.

Then, using Proposition 2.2 [13], for x ∈ (0, 2π] we get

(5.4)|φn(x)| =12Nn|∑k∈ℤg3(kNn)coskx| ≤CNn|x|2∑k∈ℤ|Δ2g3(kNn)|,

where Δ2g3(kNn)=g3(kNn)-2g3(k+1Nn)+g3(k+2Nn).

From the mean value theorem we can deduce that there exist points m_k ∈ (k, k + 1) such that

Δ2g3(kNn)=1Nn(g3′(mk+1Nn)-g3′(mkNn)).

Since the system of the points {mkNn,k∈ℤ} is some partition of the real line and g₃ is a function having a first derivative of finite total variation V[g3′], it holds that

(5.5)∑k∈ℤ|Δ2g3(kNn)|=∑k∈ℤ1Nn|(g3′(mk+1Nn)−g3′(mkNn))| ≤1NnV[g3′]=C2n.

Thus, from (5.3)–(5.5) we get

|φn(x)|≤C23n/2|x|2.

The function ψ_n can be estimated in the same way. We have that

ψn(·)=1Nn∑k∈ℤg4(kNn)cosk·,

where

g4(x)={1+3(x-1)1+9(x-1)2, 1-13≤x≤1+13,2-3(x-2)4+9(x-2)2, 1+13<x≤2+23,0, else,

is an aperiodic coefficient function.

From the first and second derivatives g4′ and g4″, we get that the points 16(9+17) and 112(15-17) are extrema of the function g4′. Since g4′(1-13)>0 and g4′(2+23)<0, the total variation of g4′ is

V[g4′]=2g4′(112(15-17))-2g4′(16(9+17))=3(17+3)7+174.

Using the same consideration as above, we have

|ψn(x)|≤C23n/2|x|2V[g4′]≤C23n/2|x|2.

□

The main ingredient to prove the following lemmata is using estimations (5.1) and (5.2). Therefore, we formulate and prove these results for functions ψ_n, n ∈ ℕ, but they are true also for functions φ_n, n ∈ ℕ.

Further, by I′ we denote the complement of the interval I ⊂ [0, 2π) to the segment [0, 2π), i.e., I′: = [0, 2π) \ I. In the case when x₀ = 0 we use corresponding modification: I′: = [−π, π) \ I, where I ⊂ [−π, π).

Lemma 5.3. Let I ⊂ ℝ, |I| < 2π, be an interval centered at x₀, J be an interval centered at x₀ such that J ⊂ I, and ζ be an infinitely differentiable on ℝ function supported on J and extended 2π-periodically, and let f∈L2π1. Then, for 1 ≤ p < ∞

(∑s=02Nn-1|((fζ)*ψn)(sπNn)|p)1p≤(∑s∈κ(I,n)|((fζ)*ψn)(sπNn)|p)1p+C1(I,ζ,f)2-n(32-1p),

and for p = ∞

maxs=0,…,2Nn-1|((fζ)*ψn)(sπNn)|≤maxs∈κ(I,n)|((fζ)*ψn)(sπNn)|+ C2(I,ζ,f)2-3n2.

Proof. Without loss of generality, we can assume that the interval J has a length equal to |I|/2. First, we consider the case 1 ≤ p < ∞. Using the inequality (a + b)^q ≤ a^q + b^q, a, b > 0, 0 < q ≤ 1, with q = 1/p, we have

Let us estimate the second term in this inequality. From the inequality (4.11), for 1 ≤ p < ∞ we have

Since x ∈ J and (s+2kNn)πNn∈I′ for some k ∈ ℤ, we have that |I|/4<|x-(s+2kNn)πNn|. Using Lemma 5.2, we obtain

(∑s∈κ(I′,n)|((fζ)*ψn)(sπNn)|p)1/p≤C23n/2∫J(∑s∈κ(I′,n)|f(x)|p|ζ(x)|p1|x-(s+2kNn)πNn|2p)1/pdx≤C23n/2(|I|/4)2∫J(∑s∈κ(I′,n)|f(x)|p|ζ(x)|p)1/pdx≤C(I)2n(32-1p)∫J|f(x)‖ζ(x)|dx≤C(I,ζ,f)2n(32-1p).

Let p = ∞. From the obvious inequality maxs∈K|as|≤maxs∈K1|as|+maxs∈K2|as|, K = K₁ ∪ K₂, K₁ ∩ K₂ = ∅, where K is a finite set of indices, we get

Let us estimate the second term in the last inequality. Using a similar consideration (with corresponding modification) as in the case 1 ≤ p < ∞, we have

maxs∈κ(I′,n)|((fζ)*ψn)(sπNn)|≤maxs∈κ(I′,n)12π∫J|f(x)‖ζ(x)||ψn(x-sπNn)|dx≤C23n/2maxs∈κ(I′,n)∫J|f(x)‖ζ(x)|dx|x-(s+2kNn)πNn|2≤C(I,ζ,f)23n/2.

□

Lemma 5.4. Let I ⊂ ℝ, |I| < 2π, be an interval centered at x₀, J₁ and J be intervals centered at x₀ such that J ⊂ J₁ ⊂ I, and ζ be an infinitely differentiable on ℝ function supported on I and extended 2π-periodically such that ζ(x) = 1 for all x ∈ J₁, and let f∈L2π1. Then, for 1 ≤ p < ∞, we have

(∑s∈κ(J,n)|(f*ψn)(sπNn)|p)1p≤(∑s∈κ(J,n)|((fζ)*ψn)(sπNn)|p)1p+C1(I,ζ,f)2-n(3/2-1/p),

and for p = ∞

maxs∈κ(J,n)|(f*ψn)(sπNn)|≤maxs∈κ(J,n)|((fζ)*ψn)(sπNn)|+ C2(I,ζ,f)2-3n2.

Proof. Without loss of generality we can assume that |J₁| = |I|/2 and |J| = |I|/4. Applying Minkowski's inequality for sums (with corresponding modification for p = ∞)

(∑k|ak+bk|p)1p≤(∑k|ak|p)1p+(∑k|bk|p)1p,

we have that

Let us estimate the second term in this inequality. From inequality (4.11), for 1 ≤ p < ∞ we have

Since for x ∈ [0, 2π) \ J₁ and (s+2kNn)πNn∈J for some k ∈ ℤ, it holds that |I|/8<|x-(s+2kNn)πNn|, from Lemma 5.2 we get

(∑s∈κ(J,n)|((f(1-ζ))*ψn)(sπNn)|p)1/p≤C23n/2∫[0,2π)\J1(∑s∈κ(J,n)|f(x)|p|1-ζ(x)|p|x-(s+2kNn)πNn|2p)1/pdx≤C23n/2(|I|/8)2∫[0,2π)\J1(∑s∈κ(J,n)|f(x)|p|1-ζ(x)|p)1/pdx≤C(I)2n(32-1p)∫[0,2π)\J1|f(x)‖1-ζ(x)|dx≤C(I,ζ,f)2n(32-1p).

Using a similar consideration (with corresponding modification), for p = ∞ we have

maxs∈κ(J,n)|((f(1-ζ))*ψn)(sπNn)|≤maxs∈κ(J,n)12π∫[0,2π)\J1|f(x)‖1-ζ(x)||ψn(x-sπNn)|dx≤C23n/2maxs∈κ(J,n)∫[0,2π)\J1|f(x)‖1-ζ(x)|dx|x-(s+2kNn)πNn|2≤C(I,ζ,f)23n/2.

□

Proof of Theorem 3.2. The equivalence between (a) and (c) follows from Theorem 4.1 and the equivalence between (a) and (c) of Theorem 3.1.

Let now part (c) hold and I be the interval chosen as in that part. Let J₁ and J be intervals centered at x₀ such that J ⊂ J₁ ⊂ I, and ζ be an infinitely differentiable on ℝ function supported on I and extended as a 2π-periodic function such that ζ(x) = 1 for all x ∈ J₁. Using Lemma 5.4 for n ∈ ℕ and 1 ≤ p ≤ ∞ (with corresponding modification for p = ∞), we get

Since {2-n}n = 0∞∈bθ,α for 0 < α < 1 and from the conditions of part (c), we have that part (b) also holds.

Let part (b) hold. Let I be the interval as in that part and I₁ be an interval centered at x₀ such that I₁ ⊂ I. Let ζ be an infinitely differentiable on ℝ function supported on I₁ and extended 2π-periodically. Then, from Lemma 5.3 for n ∈ ℕ and 1 ≤ p ≤ ∞, we have

Since {2-n}n=0∞∈bθ,α for 0 < α < 1 and from the conditions of part (b) we have that part (c) also holds. □

Note that Theorem 3.1 and Theorem 3.2 were proved for the smoothness parameter α which takes values from the interval (0, 1). The proofs depends on the smoothness properties of the coefficient functions g₁ and g₂ (see (2.1) and (2.2)). One can prove these theorems for other values of the parameter α by taking “smoother” coefficient functions (2.7) and (2.8).

5.3. Proof of theorem 3.3

The equivalence between parts (a) and (b) follows from Lemma 2.1 and the equivalence between parts (a) and (c) of Theorem A. To get equivalence between parts (b) and (c) we use Theorem 4.1 and formulas (2.7) and (2.8). □

Author contributions

All authors listed, have made substantial, direct and intellectual contribution to the work, and approved it for publication.

Conflict of interest statement

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The authors were supported by FP7-People-2011-IRSES Project number 295164 (EUMLS: EU-Ukrainian Mathematicians for Life Sciences) and H2020-MSCA-RISE-2014 Project number 645672 (AMMODIT: Approximation Methods for Molecular Modeling and Diagnosis Tools). The authors would like to thank Viktor Romanyuk (Institute of Mathematics of NAS of Ukraine) for many valuable scientific discussions. Moreover, the authors would like to thank the referees for their helpful remarks and corrections.

References 1. Bazarkhanov

. Wavelet approximation and Fourier widths of classes of periodic functions of several variables. I. Proc Steklov Inst Math. (2010) 269:2–24. translation from: Tr Mat Inst Steklova (2010) 269:8–30. 10.1134/S0081543810020021 2. Dinh

. B-spline quasi-interpolant representations and sampling recovery of functions with mixed smoothness. J Complex. (2011) 27:541–67. 10.1016/j.jco.2011.02.004 3. Hinrichs

Markhasin

Oettershagen

Ullrich

. Optimal quasi-Monte Carlo rules on higher order digital nets for the numerical integration of multivariate periodic functions. Num Math. (2016) 134:163–96. 10.1007/s00211-015-0765-y 4. Kashin

Saakyan

. Orthogonal Series. Translations of Mathematical Monographs, V.75, Providence: American Mathematical Society (1989). 5. Romanyuk

. Multiple Haar basis and m-term approximation for functions from Besov classes. I. Ukr Math J. (2016) 68:625–37. translation from: Ukr Mat Zh. (2016) 68:551–62. 10.1007/s11253-016-1246-x 6. Mhaskar

Prestin

. On local smoothness classes of periodic functions. J Fourier Anal Appl. (2005) 11:353–73. 10.1007/s00041-005-4006-0 7. Prestin

Selig

. On a constructive representation of an orthogonal trigonometric Schauder basis for C_2π. Oper Theory Adv Appl. (2001) 121:402–25. 10.1007/978-3-0348-8276-7_22 8. Lorentz

Saakyan

. Orthonormal trigonometric Schauder bases of optimal degree for C(K). J. Fourier Anal Appl. (1994) 1:103–12. 10.1007/s00041-001-4005-8 9. Selig

. Periodische Wavelet-Packets und Eine Gradoptimale schauderbasis. Shaker Verlag, Aachen: Thesis, Univ. Rostock 1997 (1998). 10. DeVore

Lorentz

. Constructive Approximation. Berlin: Springer Verlag (1993). 11. Nikolskii

. Approximation of Functions of Several Variables and Imbedding Theorems. Berlin: Springer (1975). 12. Dzyadyk

Shevchuk

. Theory of Uniform Approximation of Functions by Polynomials. Berlin: Walter de Gruyter (2008). 13. Mhaskar

Prestin

. On the detection of singularities of a periodic function. Adv Comput Math. (2000) 12:95–131. 10.1023/A:1018921319865 14. Zygmund

. Trigonometric Series. Volume I. Cambridge: Cambridge University Press (1968). 15. Timan

. Theory of Approximation of Functions of a Real Variable. Oxford: Pergamon Press (1963).