No label defined (Q4356035)

A dimension-drop \(C^*\)-algebra is of form \(I_n= \{f\in M_n(C[0, 1]): f(0), f(1)\in \mathbb{C}\}\), i.e. it is an interval algebra with dimension drops at both ends. The authors simplify proofs for the limit algebras by taking `good' quotients and reducing to dimension-drop algebras. They then use the facts that \(K_0(I_n)= \mathbb{Z}\) and \(K_1(I_n)= \mathbb{Z}/n\mathbb{Z}\). The \(KK(I_m, I_n)\) are shown to be representable as triplets of generators and more generally the \(KK\)-class of a homomorphism in the inductive system is a direct sum of these. Let \(T(B)\) denote the tracial state space of a suitable \(C^*\)-algebra \(B\). The canonical embedding of \(C[0,1]\) as the centre of \(I_n\) induces an isomorphism of \(T(I_n)\) onto \(T(C[0,1])\). The \(C^*\)-algebra \(A\) is taken to be the inductive limit of the system \((A_m, \varphi_{m,n})\) where the \(A_m\) are finite direct sums of algebras like \(M_q (I_r)\). Also \(A\) must be assumed to be simple. [The first-named author has informed the reviewer that the inductive limit algebras must be assumed simple. Indeed the invariant \(K_0+K_1\), is not sufficient in the non-simple case.] The authors use a continuous affine mapping \(\psi_T\) to lift mappings \(A\to B\) to mappings \(T(A)\to T(B)\). Isomorphisms \(\psi_0\) and \(\psi_1\) of the scaled ordered groups \(K_0\) and \(K_1\) lift to commutative diagrams in \(K_0\) and \(K_1\), respectively, preserving the compatability of \(\psi_0\) and \(\psi_1\), with respect to the pairing between the equivalence classes of projections and the trace; \(\psi_T\) lifts to an approximately commutative diagram. The connection between trace and projection comes from duality of \(T(A)\) and \(K_0(A)\), viz. \(\langle \varphi(g),\tau\rangle= \langle g,\varphi_T\tau\rangle\) where \(\tau\in T(A)\), \(g\in K_0(A)\). The main theorem is that a complete invariant is given by \(\psi_0\) and \(\psi_1\), homomorphisms of the scaled ordered groups \(K_0\) and \(K_1\), respectively, and the continuous affine mapping \(\psi_T\). In the proof for \(C^*\)-algebras of real rank or when \(K_1 (A)\) is trivial the mapping \(\psi_T\) is unnecessary as an invariant. To prove the theorem, first it is shown in Section 4 in a long-winded proof (aren't they all?) that, except when the inductive limit is a finite-dimensional AF-algebra, the inductive systems of mappings may be assumed injective. The uniqueness proof involves \(KK\)-theory and so also the existence proof must take \(KK\)-theory into account; the authors prove manoueverability of the \(KK\)-theory in a limited space; for example, in the simplest case for any \(\theta= (q,s,t)\in KK(I_m, I_n)\)) with \(q\geq 2m\), they can approximate \(\varphi_T\) by an average of affine maps close to \(\psi_T\) (cf. Thomsen 1994). The \(KK\)-classes arise as follows: A unital mapping from \(A_m\) to \(B_m\) is called of standard form if, essentially, restricted to dimension drop algebras the mapping is, up to unitary equivalence, diagonalisable with entries continuous `eigenvalue' mappings from \([0,1]\) to the spectrum of \(A\). Approximate equivalence of maps \(\varphi_0\) and \(\varphi_1\) from \(A\) to \(B\), both of standard form is, to start with, defined by sequences of unitaries in \(B\); it is shown that unital \(*\)-mappings of standard form are approximately equivalent if the eigenvalue mappings are the same; the authors use the \(KK\)-class of \(\varphi_0\), say, to obtain a unitary in \(B\) giving an approximate equivalence. Approximately equivalent maps have the same \(KK\)-class; conversely, for \(\varphi_0\) and \(\varphi_1\) in standard form, if \([\varphi_0]= [\varphi_1]\in KK(A, B)\) and the induced affine space maps satisfy the conditions in Elliot [1993] then (via passing to subsequences and relabelling) there is approximate unitary equivalence of \(\varphi_0\) and \(\varphi_1\) average of the affine maps close to \(\varphi_T\) (average of a finite family as in Thomsen 1994). Hence uniqueness. For the entire collection see [Zbl 0855.00023].