\(C^*\)-algebras associated to coverings of \(k\)-graphs (Q936122): Difference between revisions

This is a very interesting paper. The authors take as building blocks the coverings of \(k\)-graphs introduced by \textit{D. Pask, J. Quigg} and \textit{I. Raeburn} [J. Algebra 289, No.~1, 161--191 (2005; Zbl 1067.05031)] and use these to construct new \(C^*\)-algebras: a single covering of \(k\)-graphs gives rise to an embedding of universal algebras, and here a \((k+1)\)-graph is constructed with a \(C^*\)-algebra describing the embedding. Further, they construct an inductive limit of ``covering systems of \(k\)-graphs'' and identify its universal algebra as one corresponding to a \((k+1)\)-graph. This realisation is used to perform \(K\)-theory computations for the \((k+1)\)-graph. One application shows that the Kirchberg algebra \(\mathcal{P}_n\) falls within the framework of these constructions. Another interesting application is an explicit (and new) computation of the \(K\)-theory of certain \(3\)-graph algebras: these algebras cannot be obtained from the rank-\(2\) Bratteli diagram construction of \textit{D. Pask, I. Raeburn, M. Rørdam} and \textit{A. Sims} [J. Funct. Anal. 239, No.~1, 137--178 (2006; Zbl 1112.46042)]; rather they can be characterised as higher-rank analogues of the Bunce-Deddens algebras. A covering of a \(k\)-graph \(\Lambda\) is a surjective \(k\)-graph morphism \(p\) between \(\Lambda\) and a \(k\)-graph \(\Gamma\) such that, for every vertex \(v\in \Gamma^0\), the map \(p\) sends \(\Gamma v\) (consisting of \(\gamma\) in \(\Gamma\) with source \(v\)) one-to-one and onto \(\Lambda p(v)\), and similarly for \(v\Gamma\) (containing elements with range \(v\)) and \(p(v)\Lambda\). A covering system of \(k\)-graphs contains, besides a covering \((\Lambda, \Gamma, p)\), a cocycle \({\mathfrak s}:\Lambda\to S_m\) for \(m\in \mathbb N\setminus\{0\}\); this is a functor from the category \(\Lambda\) to the category \(S_m\) of the permutations on the set \(\{1,2,\dots,m\}\), viewed as a category with a single object. The formal definition of the \((k+1)\)-graph which encodes the data \((\Lambda, \Gamma, p, {\mathfrak s}, S_m)\) is rather involved, see Proposition 2.7. An excellent intuitive description is formulated by the authors as follows: this algebra ``is a \((k+1)\)-graph containing disjoint copies \(\iota(\Lambda)\) and \(j(\Gamma)\) of the \(k\)-graphs in the first \(k\) dimensions. The image \(j(v)\) of a vertex \(v\in \Gamma\) is connected to the image \(\iota(p(v))\) of the vertex it covers in \(\Lambda\) by \(m\) parallel edges \(e(v,1),\dots , e(v,m)\) of degree \(e_{k+1}\). Factorisations of paths in [\dots] (this new \((k+1)\)-graph) involving edges \(e(v,l)\) of degree \(e_{k+1}\) are determined by the unique path-lifting property and the cocycle \({\mathfrak s}\).'' A sequence of covering systems of \(k\)-graphs gives rise to an inductive limit, i.e., a unique \((k+1)\)-graph with appropriate universal properties. The authors present, in fact, a more general construction (requiring a non-negligible portion of technically hard bookkeeping) of a ``matrix of covering systems''. All this (and a lot more) is used in Theorem 3.8 to achieve \(K\)-theory computations of the \(K\)-groups of the algebra in the limit in terms of the \(K\)-groups at the finite levels. Criteria for simplicity and pure infiniteness are given in Section 4. Concrete \(K\)-theory computations are obtained in Section 5: in Section 5.1 for coverings of \(1\)-graphs and in Section 5.2 for coverings of higher-rank graphs. In the last mentioned case, the computations rely on work of \textit{D. Gwion Evans} [New York J. Math. 14, 1--31 (2008; Zbl 1146.46048)], which uses Kasparov's spectral sequence theorem. Section 6 covers a number of examples, among which are the afore-mentioned ``higher-rank'' Bunce-Deddens algebras.