Curvature invariant on noncommutative polyballs (Q2344266)

The paper under review is part of a long and steady sequence of papers by the author, where he develops aspects of multivariable operator theory (model theory, von~Neumann inequalities, numerical invariants, functional calculus and noncommutative function theory; see, for example, [Indiana Univ. Math. J. 55, No. 2, 389--442 (2006; Zbl 1104.47013); Mem. Am. Math. Soc. 964, 124 p. (2010; Zbl 1194.47001); ``Berezin transforms on noncommutative polydomains'', Trans. Am. Math. Soc., in press (2015; \url{doi:10.1090/tran/6466 })]). Let \(\mathcal{H}\) be an infinite dimensional and separable Hilbert space. Given a tuple \(\mathbf{n} = (n_1, \dots, n_k)\), the \textit{noncommutative polyball} \(\mathbf{B_n}(\mathcal{H})\) is the set of all \(k\)-tuples \(T = (T_1, \dots, T_k)\) where each \(T_i\) is in itself an \(n_i\)-tuple \((T_{i1}, \dots, T_{in_i})\) such that \(\sum_{p=1}^{n_i} T_{ip}T_{ip}^* \leq I\); moreover, it is assumed that \(T_{ip}\) commutes with \(T_{jq}\) for all \(i \neq j\). This elaborate setup is aimed at encompassing as many examples as possible: when \(k=1\), we are dealing with the set of all (noncommuting) row contractions, and when \(n_1 = \dots = n_k = 1\), we are dealing with the set of all \(k\)-tuples of commuting contractions. In a previous paper, the author showed that there is a concrete universal model \(S\) consisting of commuting tuples of (noncommuting) row isometries, and this is quite analogous to the classical Sz.-Nagy-Foias model that allows to view every pure contraction as the co-restriction of the unilateral shift to a co-invariant subspace. The universal model plays an important role in this paper (as it does in large parts of the theory). The central goal of this paper is to introduce, study, and exploit a numerical invariant for tuples in polyballs, called the \textit{curvature invariant} (it should mentioned that the word ``curvature'' should not lead one to hope for far-reaching connections with differential geometry -- as far as I know, there is not much more than inspiration and suggestive analogies). Curvature invariants for tuples of operators were defined previously by several authors: first \textit{W. Arveson} [J. Reine Angew. Math. 522, 173--236 (2000; Zbl 0951.47008)] for the commutative case, then \textit{D. W. Kribs} [Integral Equations Oper. Theory 41, No. 4, 426--454 (2001; Zbl 0994.47007)] and the author [Adv. Math. 158, No. 2, 264--309 (2001; Zbl 1002.46029)] for the noncommutative (free) case, and later \textit{P. S. Muhly} and \textit{B. Solel} [Proc. Lond. Math. Soc. (3) 87, No. 3, 748--778 (2003; Zbl 1057.46053)] in the greater generality of CP maps, and others (see the introduction of the paper). For \(T \in \mathbf{B_n}(\mathcal{H})\), denote by \(\Phi_{T_i}\) the CP map given by \(\Phi_{T_i} (a) = \sum_j T_{ij}a T_{ij}^*\). Define also \(\mathbf{\Delta_T} = (\operatorname{id} - \Phi_{T_1}) \circ \dots \circ (\operatorname{id} - \Phi_{T_k})\). For \(T \in \mathbf{B_n}(\mathcal{H})\), the author defines its curvature as the limit of a certain asymptotic formula. He proves existence of the given limit, and shows that it can be computed in several different ways. One of the formulas obtained (which the reader of the review may take as definition) is \[ \operatorname{curv}(T) = \lim_{(q_1, \dots, q_k) \in \mathbb{Z}_+^k} = \frac{1}{n_1^{q_1} \dots n_k^{q_k}} \operatorname{trace} \left[\Phi_{T_1}^{q_1} \circ \dots \circ \Phi_{T_k}^{q_k} (\mathbf{\Delta_T} (I)) \right]. \] In the third section of the paper, the author shows that the curvature can take any value in the interval \([0,\infty)\). The thorough study of the curvature, related invariants and their properties goes on, and one cannot hope to cover them all in this review (the reader may want to look at the introduction of the paper for more details). Here I will just mention a few results to give a flavour of the type of results one can get from the study of this invariant. After obtaining several formulas for the curvature invariant, the author turns to applying it. The first application of the curvature invariant is to show that it detects the universal model: Theorem. A tuple \(T \in \mathbf{B_n}(\mathcal{H})\) is unitarily equivalent to the universal model \(S\) for \(\mathbf{B_n}(\mathcal{H})\) if and only if \(\operatorname{curv}(T) = \operatorname{rank}\mathbf{\Delta_T}(I)\). The author also proves that the curvature invariant is a complete invariant for finite rank Beurling type invariant subspaces of the universal model \(S \otimes I_{\mathcal{H}}\). However, the curvature invariant is not a complete invariant in general: for every \(t \in (0,1)\), the author exhibits an uncountable family of tuples in the polyball which have curvature \(t\) but are not unitarily equivalent. The author shows that the curvature invariant enjoys some stability and continuity properties. For example, compressing a tuple to a finite co-dimensional co-invariant subspace preserves the curvature. In the fifth and last section of this paper, the author turns to \textit{commutative} polyballs (the definition is as for the noncommutative polyballs, with the difference that now all operators are assumed to commute). He defines a curvature invariant in this setting, too. The author obtains some analogous results with somewhat different methods. For example, in the commutative case it is also true that the curvature detects the universal model (for the commutative polyball). The study in the paper under review has been continued in a subsequent paper [the author, ``Euler characteristic on noncommutative polyballs'', J. Reine Angew. Math., in press (2014; \url{doi:10.1515/crelle-2014-0103})], where the author (following Arveson) introduces a noncommutative Euler characteristic, and proves an analogue of the Gauss-Bonnet-Chern theorem: for certain graded tuples, the curvature equals the Euler characteristic.