On some problems studied by R. V. Kadison (Q1412943)

The author gives a survey of progress on problems, introduced by R. V. Kadison and described below, which have turned out to be interrelated. He deals mainly with the following: For a unital C*-algebra \(A\), a bounded unital (algebraic) homomorphism \(u: A \to B(H)\) is called similar to a representation if there is an isomorphism \(S \in B(H)\) such that \(x \mapsto S^{-1}u(x)S\) is a representation of \(A\). The similarity problem is as to whether every bounded homomorphism \(A \to B(H)\) of a concrete C*-algebra on \(H\) is similar to a representation. This can be reduced to the question as to whether bounded homomorphisms are completely bounded [see, for example, \textit{U. Haagerup}, Ann. Math. (2) 118, 215--240 (1983; Zbl 0543.46033)]. The derivation problem for \(A \subset B(H)\) is as to whether every derivation is inner. It is known [\textit{E. Kirchberg}, J. Oper. Theory 36, 59--62 (1996; Zbl 0865.46055)] that the similarity problem is equivalent to the derivation problem. There is also a related formulation, by R. V. Kadison, J. R. Ringrose, et. al., in terms of vanishing of the first Hochschild cohomology group. The distance between two C*-algebras is defined to be the distance between their unit balls. A question, for a bounded operator \(x\) on a Hilbert space \(H\) which commutes with the unit ball of a C*-algebra acting on \(H\), is as to when \(x\) is close to the commutant of the C*-algebra. The author, in [Indiana Univ. Math. J. 26, 891--904 (1977; Zbl 0395.46045)], reformulated closeness so as to be suitable for an abstract C*-algebra \(A\) introducing a property \(D_{k}, k \in \mathbb{N}\), valid for any representation \(\pi\) of \(A\) in a Hilbert space \(H\), to hold if the distance from \(x \in B(H)\) to \(\pi (A)'\) is smaller than \(k \sup \{\|[x,\pi(a)]\|: a \in A, \|a\| \leq 1\}\). For example, nuclear C*-algebras have the property \(D_{1}\). The author states that the distance problem is equivalent to the Kadison problems without stipulating what the `distance problem' is, but he does indicate a proof of the `equivalence'. For a von Neumann algebra \(M\), all derivations are inner if and only if there exists a \(k > 0\) such that \(d(x,M') \leq k \sup \{ \|[x,m]\|: m \in M, \|m\| \leq 1 \}\). One extends this to a C*-algebra by proving the existence of a \(\chi \leq k\) for all such \(k\) and showing every bounded derivation \(\delta: A \to B(H)\) to be inner if and only if \(\|\delta\|_{cb} \leq \chi\|\delta\|\) (see E. Kirchberg, loc.cit.). The author describes the more recent, related problem on the length and the degree of similarity of a C*-algebra. A mapping \(\phi: A \to B(H)\) is said to be \(n\)-multiplicative if there is a factorisation of \(\phi\) such that \(\phi(a_{1}a_{2} \cdots a_{n}) = \phi_{1}(a_{1})\phi_{2}(a_{2})\cdots\phi_{n}(a_{n}), \phi_{i}:A \to B(H)\). The degree of similarity \(d(A)\) can be characterised as the least integer \(n\) such that all \(n\)-multiplicative maps are completely bounded (and hence similar to representations). Also, the length can be defined by factorisation; in order to eliminate having to stipulate the various sized matrices involved one may use the natural embedding of \(M_{n}(A)\) in \(M_{n+1}(A)\) and take \(\mathcal{K}(A)\) to be the completion of \(\bigcup_{k} M_{k}(A)\). The length of \(A\) is said to be \(\leq d\) if for every element \(x \in M_{n}(A)\) there is an \(N(n,x)\) such that \(x\) can be factorised as \(\alpha_{0}D_{1}\alpha_{1} \cdots D_{d}\alpha_{d}\), where the \(D_{i}\) are diagonal matrices in \(M_{n}(A)\) and the \(\alpha_{i}\) are (rectangular) scalar matrices with sizes determined so that the product makes sense; thus \(\alpha_{0} \in M_{n,N}(\mathbb{C}), \alpha_{1} \in M_{N}(\mathbb{C})\), etc. [cf. \textit{D. P. Blecher} and \textit{V. I. Paulsen}, Proc. Am. Math. Soc. 112, 839--850 (1991; Zbl 0767.46037)]. \textit{G. Pisier} [Int. J. Math. 12, 403--414 (2001; Zbl 1112.46309)] showed that the length and the degree of similarity are equal and that the problem as to whether \(A\) has finite length is still another version of the Kadison problem. He has determined the lengths of several types of C*-algebras and conjectured that all lengths can happen, which he has indeed proved for (non-selfadjoint) operator algebras.