On \(n\)-widths of classes of holomorphic functions with reproducing kernels (Q1331503)

For a subset \(A\) of a Banach space \(X\), the Kolmogorov and Gel'fand \(n\)-widths are defined by \(d_ n (A,X) = \inf_{X_ n} \sup_{x \in A} \inf_{y \in X_ n}\) \(\| x - y \|\) and \(d^ n(A,X) = \inf_{Y_ n} \sup_{x \in A \cap Y_ n} \| x \|\), respectively, where \(X_ n\) runs over all \(n\)-dimensional subspaces and \(Y_ n\) over all closed subspaces of codimension \(n\) of \(X\). The inear \(n\)-width is defined by \(\delta_ n (A,X) = \inf_{T_ n} \sup_{x \in A} \| x - T_ nx \|\), where \(T_ n\) varies over all linear operators on \(X\) such that \(\dim T_ n (X) \leq n\). In a previous paper [J. Approximation Theory 67, No. 3, 347-356 (1991; Zbl 0746.41026)] the authors investigated the \(n\)-width of the unit ball of the Hardy space \(H_ p\) in the \(L_ q\) metric. This time they consider the case of Hilbert spaces of holomorphic functions with reproducing kernel in the space \(C(E)\). The first result obtained refers to the Hardy space \(H^ \alpha_ 2\), \(\alpha>-1\), of all functions \(f\), holomorphic in the unit disc \(\Delta \subseteq \mathbb{C}\) and such that \(\| f \|^ 2_ \alpha = \pi^{-1} \int_ \Delta | f(z) |^ 2 (1 - | z |^ 2)^ \alpha dm (z)\) is finite, where \(dm\) denotes the Lebesgue area measure on \(\Delta\). If \(\Gamma_ r\) is the circle of radius \(r\), \(0<r<1\), with center 0, then the \(n\)-widths of the unit ball \(BH^ \alpha_ 2 \) of \(H^ \alpha_ 2\) are given by \[ \delta_ n \bigl( BH^ \alpha_ 2, C(\Gamma_ r) \bigr) = d^ n \bigl( B H^ \alpha_ 2, C(\Gamma) \bigr) = \Bigl[(1- r^ 2)^{-\alpha - 2} - \sum^{n-1}_{k=0} b^ 2_ k \Bigr]^{1/2}, \] where \((b_ n)^ \infty_{n=0}\) is the decreasing rearrangement of the sequence \[ a_ k = \bigl[ \Gamma (\alpha + k + 2)/ \bigl( k! \Gamma (\alpha + 1) \bigr) \bigr]^{1/2} \cdot r^ k. \] The case of finite subsets of a domain \(\Omega \subseteq \mathbb{C}^ n\) is considered in Section 3. For instance, if \(X\) is a Hilbert space of holomorphic functions with a reproducing kernel \(K(z,w)\) (satisfying some suplementary conditions), then for every compact subset \(E\) of \(\Omega\) there are orthonormal functions \(f_ 1, \dots, f_ n\) in \(X\) such that \[ \sup_{z \in E} \bigl[ K(z,z) - \sum^ n_{j=1} \bigl | f_ j (z) \bigr |^ 2 \bigr]= \delta^ 2_ n, \] and the equality holds at least on \(n+1\) points in \(E\). In Section 4 is considered the case when the compact set \(E\) lies in the interval \((-1,1)\) and the restriction of the kernel \(K\) to \((-1,1) \times (-1,1)\) is strictly totally positive. In this case the real and complex \(n\)-widths agree and there exists always an optimal subspace. The final section, Section 5, is dealing with the ``\(n\)-skeleton'' property in \(H_ 2\), i.e. the existence of a subset \(A=\{x_ 1,\dots,x_{n+1}\}\) of the compact set \(E \subset (-1,1)\) such that \(d^ n (BH_ 2, C(E)) = d^ n (BH_ 2, C(A))\). It is shown that \(\{- r,r\}\) is a 1-skeleton and \(\{-r,0,r\}\) is a 2-skeleton for the set \([- r,r] \subset (-1,1)\). The paper contains many other deep and difficult results which cannot be presented in detail in a review of reasonable length.