Carleson potentials and the reproducing kernel thesis for embedding theorems (Q938809)

Let \(\mu\) be a non-negative measure in the unit disc \(\mathbf{D}\). The famous Carleson Embedding Theorem states that the following assertions are equivalent: {\parindent8mm \begin{itemize}\item[(i)] The Hardy space \(H^2(\mathbf{D})\) is continuously embedded in \(L^2(\mu)\). \item[(ii)] \(C(\mu)^2 : = \sup_{z \in \mathbf{D}}\| k_z\|^2_{L^2(\mu)} < \infty, \) where \(\displaystyle{k_z(\zeta) = \frac{(1 - |z|^2)^{1/2}}{1 - \zeta \overline{z}}}\) is the reproducing kernel for the Hardy space \(H^2(\mathbf{D})\). \item[(iii)] \(\mu \) is a Carleson measure on \(\mathbf{D}\). \end{itemize}} The ``reproducing kernel thesis'' is the idea that it is sufficient to check the boundedness of an operator only on the reproducing kernel. Thus, the Carleson embedding theorem in an example of this. In the paper, the authors present a new simple proof of the implication \(\text{(ii)} \Rightarrow \text{(i)}\) which is different from others previously known. In fact, they prove (Theorem 0.2) that if \(\sup_{\lambda \in supp \mu} \int_{\mathbf{D}}|k_{\lambda}(z)|^2d\mu(z) = : A < \infty\) then \(\int_{\mathbf{D}}|f(z)|^2 d\mu(z) \leq 2eA\|f\|^2_{H^2(\mathbf{D})}\). This fact was known before. So, for instance, N. Nikolski gave a proof of this result, the constant being 32 [\textit{N. Nikolski} ``Treatise on the Shift Operators,'' Springer Verlag, Berlin, (1986; Zbl 0587.47036)] and later the constant being 16 [\textit{N. K. Nikolski}, Operators, functions, and systems: an easy reading. Volume I: Hardy, Hankel, and Toeplitz. Transl. from the French by Andreas Hartmann. Mathematical Surveys and Monographs. 92. (Providence), RI: American Mathematical Society (AMS). (2002; Zbl 1007.47001)]. The proof is ``conformally invariant'' and using Green's formula, it is shown a lemma, which the authors attribute to Uchiyama, from which a corollary is deduced stating that if \(\varphi\) is subharmonic and bounded in \(\mathbf{D}\), then \(d\nu(z) = \frac{1}{2\pi}\Delta\varphi(z) \log\frac{1}{|z|}dA(z)\) is a Carleson measure on \(\mathbf{D}\). In fact, it was B. Berndtson who firstly showed that essentially all Carleson measures are of the form \(\Delta\varphi \log\left(1/|z|\right)dA(z)\). It turns out that, with the invariant Laplacian instead of the usual Laplacian, the Uchiyama lemma can be extended to the unit ball \(\mathbf{B}_n\) of \(\mathbf{C}^n\) (Lemma 2.1) and then, with the standard modification for high dimensions, this version of the Carleson Embedding Theorem can be extended to \(H^2(\mathbf{B}_n)\). The authors also point out that their method to state Theorem 0.2 can be applied to obtain a better value of the constant of interpolation \(C : = 2e\delta^{-1}(1 + 2\log \delta^{-1})\), which was proved by different methods by \textit{A. Nicolau, J. Ortega-Cerdà} and \textit{K. Seip}, [Pac. J. Math. 213, No. 2, 389--398 (2004; Zbl 1053.30027)].