Cohomology of a class of Kadison-Singer algebras (Q625783): Difference between revisions

The paper under review is part of a recent burst of activity in the study of non-selfadjoint operator algebras inspired by the work of \textit{L.-M. Ge} and \textit{W. Yuan} [Proc. Natl. Acad. Sci. USA 107, No.~5, 1838--1843 (2010; Zbl 1206.46049); ibid., No.~11, 4840--4844 (2010; Zbl 1206.46050)], who introduced and began to explore a class of algebras they called Kadison-Singer algebras. We begin, inevitably, with some background. Given a (separable, complex) Hilbert space \({\mathcal H}\), a subspace lattice \({\mathcal L}\) is a collection of orthogonal projections onto subspaces of \({\mathcal H}\) such that \({\mathcal L}\) contains \(0\) and the identity projection \(I\) and is closed in the strong operator topology. For a given subspace lattice \({\mathcal L}\), we denote by \(\text{Alg}{\mathcal L}\) the algebra of all bounded linear operators \(A\) on \({\mathcal H}\) for which \(PAP=AP\) for all \(P\in{\mathcal L}\). Similarly, for an algebra \({\mathcal A}\) of bounded linear operators on \({\mathcal H}\), \(\operatorname{Lat}{\mathcal A}\) denotes the subspace lattice containing all projections \(P\) for which \(PAP=AP\) for all \(A\in{\mathcal A}\). A lattice \({\mathcal L}\) is reflexive if \(\text{LatAlg}{\mathcal L}={\mathcal L}\), and an algebra \({\mathcal A}\) is reflexive if \(\operatorname{AlgLat}{\mathcal A}={\mathcal A}\). A subalgebra \({\mathcal A}\) of bounded linear operators on \({\mathcal H}\) is called a Kadison-Singer algebra provided that it is reflexive and also maximal with respect to its diagonal subalgebra \({\mathcal A}\cap {\mathcal A}^*\), in the sense that, if \({\mathcal B}\) is another reflexive algebra such that \({\mathcal A}\subseteq {\mathcal B}\) and \({\mathcal B}\cap {\mathcal B}^*={\mathcal A}\cap {\mathcal A}^*\), then \({\mathcal A}={\mathcal B}\). In particular, it follows that an algebra of the form \(\operatorname{Alg}{\mathcal L}\) is a Kadison-Singer algebra if, and only if, the lattice of projections \({\mathcal L}\) is the minimal reflexive lattice generating the double commutant \({\mathcal L}''\). The specific class of Kadison-Singer algebras studied in this paper is as follows. Begin with a nest \({\mathcal N}\) of projections containing \(0\) and \(I\) and closed in the strong operator topology. Then \({\mathcal N}''\) has a unit separating vector, \(\xi\) with \(\|\xi\|=1\), such that the map \(T\mapsto T\xi\) from \({\mathcal N}''\to {\mathcal H}\) is injective. Let \(P_\xi\) be the orthogonal projection onto the one-dimensional subspace spanned by \(\xi\) and consider the lattice \({\mathcal L}={\mathcal N}\cup\{P_\xi\}\). (The author refers to such a lattice as a one point extension of a nest. It is proven in Theorem 2.4, which builds on a result of \textit{L. Wang} and \textit{W. Yuan} ``On some examples of Kadison-Singer algebras'' (Preprint, per bibl.)], that the algebra \(\operatorname{Alg}{\mathcal L}\) is, in this case, a Kadison-Singer algebra. In parallel to the theory of nest algebras, the author characterizes the rank one elements of \(\operatorname{Alg}{\mathcal L}\) (Corollary 3.3), then uses that to show, in Proposition 3.7, that an operator \(T\) satisfies the condition \(TS-ST\in\operatorname{Alg}{\mathcal L}\) for all \(S\in\operatorname{Alg}{\mathcal L}\) if, and only if \(T\in\operatorname{\mathcal L}\). The author begins the final section of the paper by studying bounded derivations on \(\operatorname{Alg}{\mathcal L}\), where \({\mathcal L}\) once again is a one point extension of a nest. Recall that a derivation \(\delta\) on \(\operatorname{Alg}{\mathcal L}\) (with values in some \(\operatorname{Alg}{\mathcal L}\)-module) is a function that satisfies the equation \(\delta(XY)=\delta(X)Y+X\delta(Y)\) for all \(X\) and \(Y\) in \(\operatorname{Alg}{\mathcal L}\). A derivation is inner if there is an operator \(T\) (in the \(\operatorname{Alg}\)-module) such that \(\delta(A)=TA-AT\) for all \(A\in\operatorname{Alg}{\mathcal L}\). In the language of Hochschild cohomology, if every bounded derivation is inner, the corresponding cohomology group \(H^1\) is 0. (We omit the general definitions of these groups here.) Theorem 4.2 shows that every bounded derivation from \(\operatorname{Alg}{\mathcal L}\), with values in \({\mathcal B(H)}\), is inner. That is, \(H^1(\operatorname{Alg}{\mathcal L},\,{\mathcal B(H)})=0\). It then follows from Proposition 3.7 that every bounded derivation from \(\operatorname{Alg}{\mathcal L}\) into itself is also inner, so that \(H^1(\operatorname{Alg}{\mathcal L},\operatorname{Alg}{\mathcal L})=0\). The paper concludes with Theorem 4.4, which shows that the Hochschild cohomology groups \(H^n(\operatorname{Alg}{\mathcal L},\,{\mathcal B(H)})\) are all 0 for \(n\geq 2\). In proving these results, the author draws on techniques from the work of \textit{E. Christensen} [Math. Ann. 229, 155--161 (1977; Zbl 0356.46057)] and of \textit{E. C. Lance} [Proc. Lond. Math., III. Ser. 43, 334-356 (1981; Zbl 0477.47030)] on the cohomology of nest algebras. As in that earlier work, the rôle of the rank one elements of \(\operatorname{Alg}{\mathcal L}\) is crucial.