On approximation of hypersingular integral operators by bounded ones

Inequalities in real analysis (26D99) Inequalities in approximation (Bernstein, Jackson, Nikol'ski?-type inequalities) (41A17) Linear operator approximation theory (47A58)

Abstract: We solve the Stechkin problem about approximation of generally speaking unbounded hypersingular integral operators by bounded ones. As a part of the proof, we also solve several related and interesting on their own problems. In particular, we obtain sharp Landau-Kolmogorov type inequalities in both additive and multiplicative forms for hypersingular integral operators and prove a sharp Ostrowski type inequality for multivatiate Sobolev classes. We also give some applications of the obtained results, in particular study the modulus of continuity of the hypersingular integral operators, and solve the problem of optimal recovery of the value of a hypersingular integral operator based on the argument known with an error.

Review text

Denote by \(\mathbb{K}\) the family of all open convex symmetric with respect to \(\theta\) sets \(K\subset\mathbb{R}^d\), \(d\geq 2\), with \(\theta\in\int K\). For \(K\in\mathbb{K}\) the norm generated by the set \(K\) is \(|x|_K=\inf\{\lambda>0:x\in\lambda K\}\), the dual norm is \(|z|_{K^{\circ}}=\sup\{(x,z):|x|_K\leq 1\}\). The set of all open convex cones generated by a finite number of points is denoted by \(\mathcal{C}\). For \(C\in\mathcal{C}\), \(p\in [1,\infty]\), the space of functions \(f:C\to\mathbb{R}\) such that \(f\in L_{\infty}(C)\) and all their (distributional) derivatives of the first order belong to \(L_p(C)\) is denoted by \(L_{\infty,p}^1(C)\); \[ W_{\infty,p}^K=\{f\in L_{\infty,p}^1(C):\Vert |\nabla f|_{K^{\circ}}\Vert_{L_p(C)}\leq 1\}. \] For \(p\in(d,\infty]\) \(p'\) is such that \(p^{-1}+(p')^{-1}=1\). The authors solve the Stechkin problem about approximation of an (unbounded) operator \(A:X\to Y\) by bounded operators: to find \[ E_N(A,Q)=\inf\{U(A,S;Q):S\in\mathcal{L}(X,Y),\Vert S\Vert \leq N\}, \] where \[ U(A,S;Q)=\sup_{x\in Q}\Vert Ax-Sx\Vert_Y. \] More precisely, they approximate the hypersingular integral operator \(D_{K,w}:L_{\infty,p}^1(C)\to L_{\infty}(C)\): \[ (D_{K,w}f)(x)= \int_C w(|t|_K)(f(x)-f(x+t))dt, x\in C, \] in \(W_{\infty,p}^K\). More notations to formulate the main result: \begin{itemize} \item[1.] For \(0\leq a<b\leq +\infty\) and \(p\in[1,\infty)\) denote by \(\mathcal{L}_p(a,b)\) the space of all measurable functions with finite norm \[ \Vert w\Vert_{\mathcal{L}_p(a,b)}=\left(\int_a^bt^{d-1}w^p(t)dt\right)^{1/p} \] (with standard extension with essential supremum for \(p=\infty)\). \item[2.] For \(h>0\) and \(p\in[1,\infty]\) denote by \(\mathcal{W}_p(0,h)\) the space of all non-negative functions \(w:(0,h] \to\mathbb{R}\) such that \(w\in\mathcal{L}_1(u,h)\) for all \(u\in (0,h)\) and the function \[ g_{w,h}(u)=\frac{1}{u^{d-1}}\int_u^hw(t)t^{d-1}dt \] belongs to \(\mathcal{L}_p(0,h)\). \item[3.] For \(h>0\) and \(p\in[1,\infty]\) denote by \(\mathcal{W}_p^*(0,h)\) the space of all non-negative functions \(w:(0,h]\to\mathbb{R}\) such that \( w\in\mathcal{L}_1(u,h) \) for all \( u\in (0,h) ,\) for all \(\nu\) close enough to 1 from the left, the function \[ w_{\nu}(u)=\sup_{t\in[\nu u,u]}|w(t)-w(u)| \] belongs to \( \mathcal{W}_p(0,h), \) and \( \lim_{\nu\to 1-0}\Vert g_{w_{\nu},h}\Vert_{\mathcal{L}_p(0,h)}=0. \) \end{itemize} The main result of the article is Theorem 1: Let \( p\in(d,\infty], C\in \mathcal{C}, K \in \mathbb{K} \) be a polytope and \( w\in \mathcal{W}_{p'}(0,1)\cap\mathcal{L}_1(1,\infty), \) or \( K \in \mathbb{ K} \) and \( w\in \mathcal{W}_{p'}^*(0,1)\cap\mathcal{L}_1(1,\infty). \) For all \( N\in (0,\Vert D_{K,w}\Vert) \) (here the norm should be replaced by infinity for unbounded \(D_{K,w} ) \) let \( h_N \) be such that \( 2d \cdot \) meas\((K\cap C)\int_{h_N}^{\infty}w(\rho)\rho^{d-1}d\rho=N.\) Then \[ E_N(D_{K,w},W_{\infty,p}^K)=(d\cdot \mathrm{meas}(K\cap C))^{\frac{1}{p'}}\Vert g_{w,h_N}\Vert_{\mathcal{L}_{p'}(0,n_N)}. \] Moreover, the extremal operator is \[ (D_{K,w,h_N}f)(x)=\int_{C\backslash h_NK}w(|t|_K)(f(x)-f(x+t))dt, x\in C. \] The proof uses new sharp Ostrowski type and Landau-Kolmogorov type inequalities. Several applications are given.

Recommendations

Cites work

Cited in

(7)

This page was built for publication: On approximation of hypersingular integral operators by bounded ones

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2134427)