\(K\)-theory for semigroup \(C^*\)-algebras and partial crossed products (Q2113473)

In this important article, the author computes the \(K\)-theory for a large class of inverse semigroup \(C^*\)-algebras, in a very explicit way, as the direct sum of the \(K\)-theory groups of certain subgroups. The assumption requires that these subgroups satisfy the Baum-Connes conjecture, allowing further simplification. The inverse semigroup \(C^*\)-algebras in question are first replaced by partial crossed products of certain partial group actions, so that the latter are also computed quite explicitly. If the ambient group satisfies the ``strong'' Baum-Connes conjecture, then the computation also covers the KK-equivalence class of the crossed product. The author describes many special cases where his formula applies. First, it covers any countable subsemigroup with the independence property of a group that satisfies the Baum-Connes conjecture. Secondly, it covers countable right LCM monoids that embed into a group that satisfies the Baum-Connes conjecture. As more concrete classes of examples, the author treats Artin monoids, Baumslag-Solitar monoids, one-relator monoids, \(ax+b\)-type semigroups attached to congruence monoids, and certain tiling \(C^*\)-algebras. In each case, the computation is quite explicit. The article uses two key ideas. The first is to replace a partial crossed product for a partial group action by a global crossed product for the Morita enveloping action. This allows to apply the Baum-Connes conjecture, which is only formulated for global actions. The second idea is to replace this global action by an equivariantly KK-equivalent one that is more ``discrete''. For the discrete version, the \(K\)-theory then decomposes rather naturally as a direct sum. Standard techniques to compute the \(K\)-theory of crossed products using the Baum-Connes conjecture give spectral sequences, but these do not provide enough information even for actions of groups as simple as \(\mathbb{Z}^k\) for moderately large~\(k\). The discretization trick used here bypasses this difficulty.