Gelfand and Kolmogorov numbers of Sobolev embeddings of weighted function spaces (Q413472): Difference between revisions

Let \(X\) and \(Y\) be complex Banach spaces and \(\mathcal{L}(X,Y)\) the class of all linear continuous operators \(T:\;X\rightarrow Y\). For \(T\in \mathcal{L}(X,Y)\), the \(n\)-th Kolmogorov number of the operator \(T\) is defined by \[ d_n(T,X,Y):=\inf\{\|Q_N^YT\|:N\subset\subset Y,\;\mathrm{dim}(N)<n\}, \] where \(Q_N^Y\) stands for the natural surjection of \(Y\) onto the quotient space \(Y/N\) and \(A\subset\subset B\) means \(A\) is a closed subspace of a topological vector space \(B\). The \(n\)-th Gelfand number of the operator \(T\) is defined by \[ c_n(T,X,Y):=\inf\{\|TJ_M^X\|:M\subset X,\;\mathrm{codim}(N)<n\}, \] where \(J_M^X\) stands for the natural injection of \(M\) into \(X\). Denote by \(\mathcal{S}'(\mathbb{R}^d)\) the set of all tempered distributions on the Euclidean space \(\mathbb{R}^d\). Let \(s\in\mathbb{R}, p, q\in[1,\infty]\) and \(\omega_{\alpha}(x):=(1+|x|^2)^{\alpha/2}\) for some \(\alpha\in(0,\infty)\) and all \(x\in \mathbb{R}^d\). The weighted Besov and Triebel-Lizorkin spaces, \(B_{p,q}^s(\mathbb{R}^n,\omega_\alpha)\) and \(F_{p,q}^s(\mathbb{R}^n,\omega_\alpha)\) are defined, respectively, as follows: \[ B_{p,q}^s(\mathbb{R}^n,\omega_\alpha) :=\{f\in\mathcal{S}'(\mathbb{R}^n):\|f\|_{B_{p,q}^s(\mathbb{R}^n,\omega_\alpha)} =\|f\omega_\alpha\|_{B_{p,q}^s(\mathbb{R}^n)}<\infty\} \] and \[ F_{p,q}^s(\mathbb{R}^n,\omega_\alpha) :=\{f\in\mathcal{S}'(\mathbb{R}^n):\|f\|_{F_{p,q}^s(\mathbb{R}^n,\omega_\alpha)} =\|f\omega_\alpha\|_{F_{p,q}^s(\mathbb{R}^n)}<\infty\} \] with \(p\neq\infty\) for the \(F\)-space. Let \(A_{p,q}^s(\mathbb{R}^n,\omega_\alpha)\) (resp., \(A_{p,q}^s(\mathbb{R}^n)\)) stand for either \(B_{p,q}^s(\mathbb{R}^n,\omega_\alpha)\) (resp., \(B_{p,q}^s(\mathbb{R}^n)\)) or \(F_{p,q}^s(\mathbb{R}^n,\omega_\alpha)\) (resp., \(F_{p,q}^s(\mathbb{R}^n)\)) with the constraint that, for the \(F\)-spaces \(p<\infty\), holds. In this paper, the authors prove the following: Let \(1\leq p_1,p_2, q_1, q_2\leq\infty\), \(-\infty<s_2<s_1<\infty\) and \(\alpha\in(0,\infty)\). Suppose that \(\delta:=s_1-s_2-d(1/p_1-1/p_2)>0\), \(\theta:=\frac{1/p_1-1/p_2}{1/2-1/p_2}\) and \(1/\widetilde{p}=\mu/d+1/p_1\) with \(\mu:=\min\{\alpha,\delta\}\). Denote by \(d_n\) the \(n\)-th Kolmogorov number of the Sobolev embedding \(A_{p_1,q_1}^{s_1}(\mathbb{R}^d,\omega_\alpha) \hookrightarrow A_{p_2,q_2}^{s_2}(\mathbb{R}^d)\). If \(1\leq p_1\leq p_2\leq\infty\) or \(\widetilde{p}<p_2<p_1\leq\infty\), \(\delta\neq\alpha\), and \(p_2<\infty\) when \(p_1<p_2\), then \(d_n\sim n^{-\lambda}\), where \(\lambda:=\frac{\mu}d\) when \(1\leq p_1\leq p_2\) or \(2<p_1=p_2\leq\infty\), or \(\lambda:=\frac\mu d+\frac1{p_1}-\frac1{p_2}\) when \(\widetilde{p}<p_2<p_1\leq\infty\), or \(\lambda:=\frac\mu d+\frac12-\frac1{p_2}\) when \(1\leq p_1<2<p_2<\infty\) and \(\mu>\frac d{p_2}\), or \(\lambda:=\frac{\mu p_2}{2d}\) when \(1\leq p_1<2<p_2<\infty\) and \(\mu<\frac d{p_2}\), or \(\lambda:=\frac\mu d+\frac1{p_1}-\frac1{p_2}\) when \(2\leq p_1<p_2<\infty\) and \(\mu>\frac d{p_2}\theta\), or \(\lambda:=\frac{\mu p_2}{2d}\) when \(2\leq p_1<p_2<\infty\) and \(\mu<\frac d{p_2}\theta\). Also, similar results corresponding to the \(n\)-th Gelfand number of the Sobolev embedding \(A_{p_1,q_1}^{s_1}(\mathbb{R}^d,\omega_\alpha) \hookrightarrow A_{p_2,q_2}^{s_2}(\mathbb{R}^d)\) are also obtained in the paper.