Interpolating sequences for pairs of spaces (Q6108683)

Let \(\mathcal{H}\) be a reproducing kernel Hilbert space on a nonempty set \(X\); for instance, \(X\) might be the open unit disk in the complex plane, and \(\mathcal{H}\) might be the Hardy space, the Bergman space, and so on. Let \(M(\mathcal{H})\) denote the multiplier algebra of \(\mathcal{H}\), which consists of functions \(\phi:X\to \mathbb{C}\) for which \(\phi\cdot \mathcal{H}\subset \mathcal{H}\). A sequence \(\{\lambda_i\}\subset X\) is called an \textit{interpolating sequence for} \(M(\mathcal{H})\) if, for each sequence \(\{w_i\}\subset \ell^\infty\), there is a multiplier \(\phi\) such that \(\phi(\lambda_i)=w_i\) for all \(i\). In the same way, \(\{\lambda_i\}\subset X\) is called an \textit{interpolating sequence for} \(\mathcal{H}\) if, for each sequence \(\{w_i\}\subset \ell^2\), there is a function \(f\in \mathcal{H}\) such that \[\frac{f(\lambda_i)}{\Vert k_{\lambda_i}\Vert}=w_i,\] where \(k_{\lambda_i}\) is the reproducing kernel corresponding to the point \(\lambda_i\). It is known that every \(M(\mathcal{H})\)-interpolating sequence is an \(\mathcal{H}\)-interpolating sequence, however, these two notions coincide in spaces with \textit{complete Pick kernels} [\textit{J.~Agler} and \textit{J.~E. McCarthy}, J. Funct. Anal. 175, No.~1, 111--124 (2000; Zbl 0957.47013)]. Recall that a reproducing kernel \(s\) is called an \textit{irreducible complete Pick kernel} if \[s_w(z)=\frac{f(z)\overline{f(w)}}{1-\langle b(z), b(w)\rangle_{\mathcal{K}}},\] where \(f:X\to\mathbb{C}\) is non-vanishing, and \(b\) is a function from \(X\) into the open unit ball of some Hilbert space \(\mathcal{K}\). The kernel \(s\) is called a \textit{normalized complete Pick kernel} if, in addition, \(f\) is identically \(1\), and there is a point \(w_0\in X\) such that \(b(w_0)=0\). The present paper concerns the concept of interpolating sequences for multipliers between functional Hilbert spaces on a set \(X\). Let \(k\) and \(l\) be two non-zero reproducing kernels on a set \(X\), and let \(\mathcal{H}_k\) and \(\mathcal{H}_l\) be the corresponding reproducing kernel Hilbert spaces, respectively. We write \(\phi\in M(\mathcal{H}_k, \mathcal{H}_l)\) if \(\phi\cdot \mathcal{H}_k\subset \mathcal{H}_l\); that is, \(\phi\) is a multiplier from \(\mathcal{H}_k\) to \(\mathcal{H}_l\). A sequence \(\{\lambda_i\}\subset X\) is called an \textit{interpolating sequence for} \(M(\mathcal{H}_k, \mathcal{H}_l)\) if, for each sequence \(\{w_i\}\subset \ell^\infty\), there is a multiplier \(\phi\in M(\mathcal{H}_k, \mathcal{H}_l)\) such that \[\phi(\lambda_i)=w_i\frac{\Vert l_{\lambda_i}\Vert }{\Vert k_{\lambda_i}\Vert} \text{ for all } i\ge 1.\] In [\textit{A.~Aleman} et al., Int. Math. Res. Not. 2019, No.~12, 3832--3854 (2019; Zbl 1443.46016)], the authors investigated interpolating sequences for pairs of Hilbert spaces. It was proved, among other things, that multiplier interpolating sequences, or \(M(\mathcal{H}_k,\mathcal{H}_l)\)-interpolating sequences, enjoy the following properties: \begin{itemize} \item They satisfy the Carleson measure condition for \(\mathcal{H}_k\): a sequence \(\{\lambda_i\}\) satisfies the \emph{Carleson measure condition for} \(\mathcal{H}_k\) if there exists \(C>0\) such that \[\int_X |f|^2 \, d\mu=\sum_{i=1}^\infty \frac{|f(\lambda_i)|^2}{\Vert k_{\lambda_i}\Vert^2}\le C\Vert f\Vert^2_{\mathcal{H}_k}\text{ for all } f\in \mathcal{H}_k.\] \item They are weakly separated by \(l\): a sequence \(\{\lambda_i\}\) is said to be \textit{weakly separated by} \(l\) if there exists \(\epsilon>0\) such that, for all \(i\neq j\), \(d_l(\lambda_i, \lambda_j)\ge \epsilon\), where \[d_l(z,w)=\sqrt{1-\frac{|\langle l_z,l_w\rangle|^2}{\Vert l_z\Vert^2\Vert l_w\Vert^2}},\quad z,w\in X.\] \end{itemize} The authors of [\textit{A. Aleman} et al., loc. cit.] then raised the question whether the converse is also true; more precisely, they asked: {Question}. Let \(s\) be a normalized complete Pick kernel on \(X\) and let \(l=gs\), where \(g\) is a kernel on \(X\). Is it true that a sequence \(\{\lambda_i\}\subset X\) is \(M(\mathcal{H}_s, \mathcal{H}_l)\)-interpolating if and only if it satisfies the Carleson measure condition for \(\mathcal{H}_s\) and is weakly separated by \(l\)? The author shows here that the above conditions are not sufficient in general, and a stronger notion of separation is needed. He defines the notion of \(n\)-weak separation in the following way (see Definition~1.3): for \(n\ge 2\), \(\{\lambda_i\}\) is said to be \(n\)-\textit{weakly separated by} a kernel \(k\) if there is \(\epsilon>0\) such that, for every \(n\)-point subset \(\{a_1,\dots ,a_n\}\subset \{\lambda_i\}\), we have \[\operatorname{dist}\left( \frac{k_{a_1}}{\Vert k_{a_1}\Vert}, \operatorname{span}\left\{\frac{k_{a_2}}{\Vert k_{a_2}\Vert}, \dots, \frac{k_{a_n}}{\Vert k_{a_n}\Vert} \right\} \right)\ge \epsilon.\] Note that if \(n=2\), the two concepts of separation coincide. Under these conditions, the main result of the paper under review reads as follows (see Theorem~1.4): {Theorem}. Let \(s\) be a normalized complete Pick kernel on \(X\) and let \(l=gs\), where \(g\) is a positive semi-definite kernel on \(X\). Then a sequence \(\{\lambda_i\}\subset X\) is \(M(\mathcal{H}_s, \mathcal{H}_l)\)-interpolating if and only if it satisfies the Carleson measure condition for \(\mathcal{H}_s\) and is \(n\)-weakly separated by \(l\), for all~$n\ge2$.