Clark theory in the Drury-Arveson space (Q2444478)

Let \(H^2_d\) be the Drury-Arveson space (a.k.a. \textit{the \(d\)-shift space} or \textit{symmetric Fock space}), and let \(b\) be a contractive multiplier on \(H^2_d\). The paper under review studies several problems in the \textit{de Branges-Rovnyak type space} \(\mathcal{H}(b)\), which is the reproducing kernel Hilbert space on the unit ball \(\mathbb{B}_d\) with kernel \[ k^b(z,w) = \frac{1-b(z) \bar b(w)}{1 - \langle z, w\rangle} . \] \textit{D. N. Clark} studied these spaces in the case \(d=1\) (see [J. Anal. Math. 25, 169--191 (1972; Zbl 0252.47010)]), focusing on rank one perturbations of the (restriction of) the backward shift to these spaces (in fact, Clark considered the case where \(b\) is inner). The author of the paper under review seeks to ``exend the basic elements of Clark's theory'' to the \(d>1\) setting. It turns out that there are several obstructions to obtaining straightforward generalizations. The author first has to find (and justify) what the correct generalizations are, and then he proves them, somewhat surprisingly, by drawing heavily on techniques from noncommutative multivariable operator theory. The first step is a several variables substitute to the formula \[ \frac{1+b(z)}{1-b(z)} = \int_{\mathbb{T}} \frac{1+ z \bar \zeta}{1 - z \bar \zeta} d \mu(\zeta) + i \operatorname{Im} \frac{1+b(0)}{1-b(0)} . \] The substitute for the above formula is obtained by lifting attention from Drury-Arveson space and its multipliers to the full Fock space and the noncommutative disc algebra \(\mathcal{A}\). Instead of the measure \(\mu\) one has now a positive functional \(\mu\) on a certain operator system that lies in \(\mathcal{A} + \mathcal{A}^*\). This linear functional allows to construct a Hilbert space \(P^2(\mu)\) analogous to the one studied in the classical setting. The movement from the commutative to the noncommutative setting, and back, is what powers also the rest of this paper, and what makes it of particular interest. The next result is that the noncommutative Fantappié transform, which is the relevant analogue of the classical Fantappié transform in the multivariable setting, is a unitary from \(P^2(\mu)\) onto \(\mathcal{H}(b)\). At this point, the author introduces the notion of a \textit{quasi-extreme multiplier} \(b\). In the one variable case, \(b\) is quasi-extreme if and only if it is an extreme point in the unit ball of \(H^\infty\) (unfortunately, there is no such characterization in the multivariable setting), and it is in this case where Clark obtained his strongest results. Accordingly, for the second half of the paper the author concentrates mostly on the case where \(b\) is a quasi-extreme multiplier, and the main results are obtained in this case. The next goal of the author is to study the multidimensional analogue of rank one perturbations of the backward shift (restricted to \(\mathcal{H}(b)\)). The correct interpretation of ``restriction of the backward shift'' in the multivariable case turns out to be ``solution of the Gleason problem''. The Gleason problem in the spaces \(\mathcal{H}(b)\) was solved by \textit{J. A. Ball} and \textit{V. Bolotnikov} [CRM Proceedings and Lecture Notes 51, 1--39 (2010; Zbl 1210.47041)]. In the paper under review, it is shown that if \(b\) is ``quasi-extreme'', then there is always a unique, contractive, ``extremal'' solution \(X = (X_1, \dots, X_d)\) to the Gleason problem in \(\mathcal{H}(b)\). Finally, the author uses the results obtained to study certain rank one perturbation of the solution \(X\) to the Gleason problem. He finds a unitary equivalent model for these operators on the space \(P^2(\mu)\), and partially determines their spectrum. He finds that a point \(\zeta \in \partial \mathbb{B}_d\) is a joint eigenvalue for \(X\) if and only if \(b\) has finite angular derivative at \(\zeta\), and finds the eigenvector (by the way, there is an unfortunate typo that reccurs a couple of times in the last section: \(\mathbb{B}^d\) is written instead of \(\partial \mathbb{B}^d\)). Moreover, the essential spectrum of \(X\) is shown to contain the entire unit sphere \(\partial \mathbb{B}_d\).