Comparison radius and mean topological dimension: Rokhlin property, comparison of open sets, and subhomogeneous C*-algebras (Q2164786)

The article under review studies the structural interplay between discrete topological dynamical systems and their associated transformation group C*-algebras. Given a discrete amenable group \(\Gamma\) and a free minimal action \(\Gamma\curvearrowright X\) on a compact metrizable space via homeomorphisms, one constructs the crossed product C*-algebra \(C(X)\rtimes\Gamma\) in a way that the original dynamics is encoded in the multiplicative structure of the C*-algebra. Although this construction yields plenty of interesting examples, the study of crossed product C*-algebras is notoriously difficult, which calls for methods and results that allow one to transfer structural results back and forth between the original dynamics and the algebra. The results in the present article make progress on two important questions: \begin{itemize} \item[(1)] If the action \(\Gamma\curvearrowright X\) is `sufficiently nice', is the C*-algebra \(C(X)\rtimes\Gamma\) in the classifiable class captured by Elliott's program? As we know by now [\textit{J.-L. Tu}, \(K\)-Theory 17, No. 3, 215--264 (1999; Zbl 0939.19001); \textit{J. Castillejos} et al., Invent. Math. 224, No. 1, 245--290 (2021; Zbl 1467.46055); \textit{A. Tikuisis} et al., Ann. Math. (2) 185, No. 1, 229--284 (2017; Zbl 1367.46044)], one can equivalently ask: is \(C(X)\rtimes\Gamma\) Jiang-Su stable? \item[(2)] What does the abstract C*-algebraic isomorphism class of \(C(X)\rtimes\Gamma\) remember about the structure of the action \(\Gamma\curvearrowright X\)? \end{itemize} A folklore conjecture by Phillips and Toms (originally stated for \(\Gamma=\mathbb Z\)) asserts that the so-called radius of comparison of the C*-algebra \(C(X)\rtimes\Gamma\) coincides with half of the mean dimension for the action \(\Gamma\curvearrowright X\) [\textit{E. Lindenstrauss} and \textit{B. Weiss}, Isr. J. Math. 115, 1--24 (2000; Zbl 0978.54026)]. A positive solution for this ambitious conjecture would give satisfactory answers to both of the above questions, since in a manner of speaking, the radius of comparison of a C*-algebra can be seen as a numerical value measuring how `far away' it is from being classifiable. The author then makes progress on this conjecture for a certain class of well-behaved actions \(\Gamma\curvearrowright X\). In this context he proves that the comparison radius of \(C(X)\rtimes\Gamma\) is bounded above by half of the mean dimension of \(\Gamma\curvearrowright X\). The method of proof is largely based on highly involved groupoid techniques, which are outside the scope of this review. To make more precise what `well-behaved' means here, one asks whether \begin{itemize} \item[(a)] \(\Gamma\curvearrowright X\) satisfies a version of a Rokhlin property that is weaker than almost finiteness in the sense of \textit{H. Matui} [Proc. Lond. Math. Soc. (3) 104, No. 1, 27--56 (2012; Zbl 1325.19001)] and \textit{D. Kerr} [J. Eur. Math. Soc. (JEMS) 22, No. 11, 3697--3745 (2020; Zbl 1465.37010)]; \item[(b)] the inclusion \(C(X)\subset C(X)\rtimes\Gamma\) satisfies a version of strict comparison for positive elements. In order to avoid excessive technical details, let us consider a stronger version of what is assumed in the article: If \(f,g\in C(X)\) are nonnegative functions with open supports \(U\) and \(V\) such that \(\mu(U)<\mu(V)\) holds for all \(\Gamma\)-invariant Borel probability measures \(\mu\) on \(X\), then \(f\) is Cuntz-subequivalent to \(g\) inside the C*-algebra \(C(X)\rtimes\Gamma\). \end{itemize} A priori, these two conditions have the drawback of being difficult to verify on the nose, but the article also shows that they are automatic in some important cases. Firstly, it is shown that the conditions are redundant for \(\Gamma=\mathbb{Z}\). This alone is impressive because together with the aforementioned main result, the author recovers and substantially generalizes his earlier breakthrough work [\textit{G. A. Elliott} and \textit{Z. Niu}, Duke Math. J. 166, No. 18, 3569--3594 (2017; Zbl 1410.46046)]. Secondly, the two conditions are also shown to be redundant if one assumes that \(\Gamma\) has local subexponential growth and \(X\) admits a free factor system on the Cantor set. This part of the article relies on the results in [\textit{D. Kerr} and \textit{G. Szabó}, Commun. Math. Phys. 374, No. 1, 1--31 (2020; Zbl 1446.46040)] and a result of \textit{T. Downarowicz} and \textit{G.-H. Zhang} [``The comparison property of amenable groups'', Preprint (2020), \url{arXiv:1712.05129}]. It is noteworthy that in a follow-up article [\textit{Z. Niu}, Trans. Am. Math. Soc. 374, No. 10, 7525--7551 (2021; Zbl 1485.46065)], the author extends the results of the present article further. There it is shown that the two aforementioned conditions are further redundant for \(\Gamma=\mathbb{Z}^d\), which leads to a further generalization of the breakthrough result in [\textit{G. A. Elliott} and \textit{Z. Niu}, Duke Math. J. 166, No. 18, 3569--3594 (2017; Zbl 1410.46046)].