A Carathéodory theorem for the bidisk via Hilbert space methods (Q662396): Difference between revisions

This paper presents an interesting new approach to Julia-Wolff-Carathéodory theorems for bounded functions in the bidisk complementing well the results obtained by the reviewer [J. Anal. Math. 74, 275--306 (1998; Zbl 0912.32005)]. Let \(\varphi:\mathbb{D}^2\to\mathbb{D}\) be a bounded holomorphic function on the unit bidisk \(\mathbb{D}^2\) and take \(\tau\in\partial\mathbb{D}^2\). Then the authors prove that: (i) if \(\varphi\) admits a (suitably defined) angular gradient \(\nabla\varphi(\tau)\in\mathbb{C}^2\) at~\(\tau\) (in the terminology of the paper, \(\tau\) is a \(C\)-point for \(\varphi\)), then \(\nabla\varphi(\lambda)\to\nabla\varphi(\tau)\) as \(\lambda\to\tau\) nontangentially in \(\mathbb{D}^2\); (ii) if the liminf of \(\big(1-|\varphi(\lambda)|\big)\big/\big(1-\|\lambda\|\big)\) as \(\lambda\to\tau\) is finite (in the terminology of the paper, \(\tau\) is a \(B\)-point for \(\varphi\)), then the directional derivative \(D^{}_{-\delta}\varphi(\tau)\) of \(\varphi\) at~\(\tau\) in the direction~\(-\delta\) exists for all appropriate directions \(\delta\in\mathbb{C}^2\), and it depends holomorphically on~\(\delta\). It should be remarked that a \(C\)-point is a \(B\)-point always, whereas, contrarily to what happens in one variable, the converse in general is not true; a counterexample is given in the paper. The proofs, following ideas due to \textit{D. Sarason} [Sub-Hardy Hilbert spaces in the unit disk. New York: Wiley (1993; Zbl 1253.30002)] relies upon an abstract way of realizing the function \(\varphi\) via a holomorphic map \(u\) from \(\mathbb{D}^2\) to a Hilbert space~\(\mathcal{M}\). Thus most of the paper is devoted to describing how to go from properties of the function \(\varphi\) to properties of the model \((\mathcal{M},u)\) and, more to the point, conversely. At present this approach works in the bidisk only; it would be interesting to understand whether it could be applied in other domains, too.